summaryrefslogtreecommitdiffstats
diff options
context:
space:
mode:
authorMeghana M <mmadhusu@redhat.com>2015-06-29 06:16:47 +0530
committerKaleb KEITHLEY <kkeithle@redhat.com>2015-06-29 09:42:56 -0700
commit64f36a04d07fc03aeda2ca7921f874ae0db19fa7 (patch)
treede489fc412e7058d26284852cb23ce019fd8a08b
parentb7b2b14c1f1ea1d232c5a5d5901c00bd187aa841 (diff)
NFS-Ganesha : Exporting volume fails
Due to a recent fix, the dbus-send.sh looked for a .export_added file even before it was created. This resulted in the ganesha.enable option failing consistently. Fixing it. Change-Id: I26a68578551b6e38e49a9997e6f6f983fd668971 BUG: 1236561 Signed-off-by: Meghana M <mmadhusu@redhat.com> Reviewed-on: http://review.gluster.org/11456 Reviewed-by: soumya k <skoduri@redhat.com> Reviewed-by: Kaleb KEITHLEY <kkeithle@redhat.com> Reviewed-by: Niels de Vos <ndevos@redhat.com> Tested-by: Gluster Build System <jenkins@build.gluster.com>
-rwxr-xr-xextras/ganesha/scripts/dbus-send.sh2
-rwxr-xr-xextras/hook-scripts/start/post/S31ganesha-start.sh2
2 files changed, 2 insertions, 2 deletions
diff --git a/extras/ganesha/scripts/dbus-send.sh b/extras/ganesha/scripts/dbus-send.sh
index b96bba90e45..d838cabf7b2 100755
--- a/extras/ganesha/scripts/dbus-send.sh
+++ b/extras/ganesha/scripts/dbus-send.sh
@@ -22,7 +22,7 @@ function check_cmd_status()
function dynamic_export_add()
{
count=`ls -l $GANESHA_DIR/exports/*.conf | wc -l`
- if [ "$count" = "0" ] ;
+ if [ "$count" = "1" ] ;
then
EXPORT_ID=2
else
diff --git a/extras/hook-scripts/start/post/S31ganesha-start.sh b/extras/hook-scripts/start/post/S31ganesha-start.sh
index dc3eaeaaa9f..f1ca72b38d2 100755
--- a/extras/hook-scripts/start/post/S31ganesha-start.sh
+++ b/extras/hook-scripts/start/post/S31ganesha-start.sh
@@ -61,7 +61,7 @@ function write_conf()
function export_add()
{
count=`ls -l $GANESHA_DIR/exports/*.conf | wc -l`
- if [ "$count" = "0" ] ;
+ if [ "$count" = "1" ] ;
then
EXPORT_ID=2
else
dth: 0.4%;'/> -rw-r--r--xlators/cluster/afr/src/afr-inode-read.c1272
-rw-r--r--xlators/cluster/afr/src/afr-inode-read.h39
-rw-r--r--xlators/cluster/afr/src/afr-inode-write.c1850
-rw-r--r--xlators/cluster/afr/src/afr-inode-write.h66
-rw-r--r--xlators/cluster/afr/src/afr-lk-common.c1249
-rw-r--r--xlators/cluster/afr/src/afr-mem-types.h35
-rw-r--r--xlators/cluster/afr/src/afr-open.c233
-rw-r--r--xlators/cluster/afr/src/afr-self-heal-algorithm.c167
-rw-r--r--xlators/cluster/afr/src/afr-self-heal-algorithm.h22
-rw-r--r--xlators/cluster/afr/src/afr-self-heal-common.c1296
-rw-r--r--xlators/cluster/afr/src/afr-self-heal-common.h80
-rw-r--r--xlators/cluster/afr/src/afr-self-heal-data.c1107
-rw-r--r--xlators/cluster/afr/src/afr-self-heal-entry.c813
-rw-r--r--xlators/cluster/afr/src/afr-self-heal-metadata.c458
-rw-r--r--xlators/cluster/afr/src/afr-self-heal.h33
-rw-r--r--xlators/cluster/afr/src/afr-self-heald.c1761
-rw-r--r--xlators/cluster/afr/src/afr-self-heald.h66
-rw-r--r--xlators/cluster/afr/src/afr-transaction.c1109
-rw-r--r--xlators/cluster/afr/src/afr-transaction.h48
-rw-r--r--xlators/cluster/afr/src/afr.c322
-rw-r--r--xlators/cluster/afr/src/afr.h503
-rw-r--r--xlators/cluster/afr/src/pump.c335
-rw-r--r--xlators/cluster/afr/src/pump.h23
-rw-r--r--xlators/cluster/dht/src/Makefile.am13
-rw-r--r--xlators/cluster/dht/src/dht-common.c1750
-rw-r--r--xlators/cluster/dht/src/dht-common.h349
-rw-r--r--xlators/cluster/dht/src/dht-diskusage.c225
-rw-r--r--xlators/cluster/dht/src/dht-hashfn.c96
-rw-r--r--xlators/cluster/dht/src/dht-helper.c496
-rw-r--r--xlators/cluster/dht/src/dht-inode-read.c279
-rw-r--r--xlators/cluster/dht/src/dht-inode-write.c535
-rw-r--r--xlators/cluster/dht/src/dht-layout.c216
-rw-r--r--xlators/cluster/dht/src/dht-linkfile.c182
-rw-r--r--xlators/cluster/dht/src/dht-mem-types.h23
-rw-r--r--xlators/cluster/dht/src/dht-rebalance.c1208
-rw-r--r--xlators/cluster/dht/src/dht-rename.c185
-rw-r--r--xlators/cluster/dht/src/dht-selfheal.c557
-rw-r--r--xlators/cluster/dht/src/dht-shared.c758
-rw-r--r--xlators/cluster/dht/src/dht.c497
-rw-r--r--xlators/cluster/dht/src/nufa.c395
-rw-r--r--xlators/cluster/dht/src/switch.c310
-rw-r--r--xlators/cluster/ha/src/Makefile.am7
-rw-r--r--xlators/cluster/ha/src/ha-helpers.c24
-rw-r--r--xlators/cluster/ha/src/ha-mem-types.h21
-rw-r--r--xlators/cluster/ha/src/ha.c32
-rw-r--r--xlators/cluster/ha/src/ha.h24
-rw-r--r--xlators/cluster/map/src/Makefile.am7
-rw-r--r--xlators/cluster/map/src/map-helper.c24
-rw-r--r--xlators/cluster/map/src/map-mem-types.h21
-rw-r--r--xlators/cluster/map/src/map.c27
-rw-r--r--xlators/cluster/map/src/map.h22
-rw-r--r--xlators/cluster/stripe/src/Makefile.am11
-rw-r--r--xlators/cluster/stripe/src/stripe-helpers.c675
-rw-r--r--xlators/cluster/stripe/src/stripe-mem-types.h28
-rw-r--r--xlators/cluster/stripe/src/stripe.c3101
-rw-r--r--xlators/cluster/stripe/src/stripe.h163
-rw-r--r--xlators/cluster/unify/src/Makefile.am16
-rw-r--r--xlators/cluster/unify/src/unify-mem-types.h41
-rw-r--r--xlators/cluster/unify/src/unify-self-heal.c1239
-rw-r--r--xlators/cluster/unify/src/unify.c4589
-rw-r--r--xlators/cluster/unify/src/unify.h146
-rw-r--r--xlators/debug/error-gen/src/Makefile.am7
-rw-r--r--xlators/debug/error-gen/src/error-gen-mem-types.h20
-rw-r--r--xlators/debug/error-gen/src/error-gen.c991
-rw-r--r--xlators/debug/error-gen/src/error-gen.h36
-rw-r--r--xlators/debug/io-stats/src/Makefile.am7
-rw-r--r--xlators/debug/io-stats/src/io-stats-mem-types.h21
-rw-r--r--xlators/debug/io-stats/src/io-stats.c829
-rw-r--r--xlators/debug/trace/src/Makefile.am8
-rw-r--r--xlators/debug/trace/src/trace-mem-types.h21
-rw-r--r--xlators/debug/trace/src/trace.c3075
-rw-r--r--xlators/debug/trace/src/trace.h98
-rw-r--r--xlators/encryption/Makefile.am2
-rw-r--r--xlators/encryption/crypt/Makefile.am (renamed from xlators/cluster/unify/Makefile.am)0
-rw-r--r--xlators/encryption/crypt/src/Makefile.am24
-rw-r--r--xlators/encryption/crypt/src/atom.c962
-rw-r--r--xlators/encryption/crypt/src/crypt-common.h141
-rw-r--r--xlators/encryption/crypt/src/crypt-mem-types.h43
-rw-r--r--xlators/encryption/crypt/src/crypt.c4498
-rw-r--r--xlators/encryption/crypt/src/crypt.h899
-rw-r--r--xlators/encryption/crypt/src/data.c769
-rw-r--r--xlators/encryption/crypt/src/keys.c302
-rw-r--r--xlators/encryption/crypt/src/metadata.c605
-rw-r--r--xlators/encryption/crypt/src/metadata.h74
-rw-r--r--xlators/encryption/rot-13/src/Makefile.am7
-rw-r--r--xlators/encryption/rot-13/src/rot-13.c69
-rw-r--r--xlators/encryption/rot-13/src/rot-13.h20
-rw-r--r--xlators/features/Makefile.am3
-rw-r--r--xlators/features/changelog/Makefile.am3
-rw-r--r--xlators/features/changelog/lib/Makefile.am (renamed from xlators/storage/bdb/Makefile.am)2
-rw-r--r--xlators/features/changelog/lib/examples/c/get-changes.c87
-rw-r--r--xlators/features/changelog/lib/examples/python/changes.py32
-rw-r--r--xlators/features/changelog/lib/examples/python/libgfchangelog.py64
-rw-r--r--xlators/features/changelog/lib/src/Makefile.am37
-rw-r--r--xlators/features/changelog/lib/src/changelog.h31
-rw-r--r--xlators/features/changelog/lib/src/gf-changelog-helpers.c180
-rw-r--r--xlators/features/changelog/lib/src/gf-changelog-helpers.h97
-rw-r--r--xlators/features/changelog/lib/src/gf-changelog-process.c571
-rw-r--r--xlators/features/changelog/lib/src/gf-changelog.c515
-rw-r--r--xlators/features/changelog/src/Makefile.am19
-rw-r--r--xlators/features/changelog/src/changelog-encoders.c176
-rw-r--r--xlators/features/changelog/src/changelog-encoders.h46
-rw-r--r--xlators/features/changelog/src/changelog-helpers.c693
-rw-r--r--xlators/features/changelog/src/changelog-helpers.h395
-rw-r--r--xlators/features/changelog/src/changelog-mem-types.h29
-rw-r--r--xlators/features/changelog/src/changelog-misc.h101
-rw-r--r--xlators/features/changelog/src/changelog-notifier.c314
-rw-r--r--xlators/features/changelog/src/changelog-notifier.h19
-rw-r--r--xlators/features/changelog/src/changelog-rt.c72
-rw-r--r--xlators/features/changelog/src/changelog-rt.h33
-rw-r--r--xlators/features/changelog/src/changelog.c1477
-rw-r--r--xlators/features/compress/Makefile.am (renamed from xlators/protocol/legacy/lib/Makefile.am)2
-rw-r--r--xlators/features/compress/src/Makefile.am17
-rw-r--r--xlators/features/compress/src/cdc-helper.c547
-rw-r--r--xlators/features/compress/src/cdc-mem-types.h22
-rw-r--r--xlators/features/compress/src/cdc.c342
-rw-r--r--xlators/features/compress/src/cdc.h107
-rw-r--r--xlators/features/filter/src/Makefile.am7
-rw-r--r--xlators/features/filter/src/filter-mem-types.h20
-rw-r--r--xlators/features/filter/src/filter.c24
-rw-r--r--xlators/features/gfid-access/Makefile.am (renamed from xlators/performance/stat-prefetch/Makefile.am)0
-rw-r--r--xlators/features/gfid-access/src/Makefile.am15
-rw-r--r--xlators/features/gfid-access/src/gfid-access-mem-types.h23
-rw-r--r--xlators/features/gfid-access/src/gfid-access.c1172
-rw-r--r--xlators/features/gfid-access/src/gfid-access.h128
-rw-r--r--xlators/features/glupy/Makefile.am (renamed from xlators/protocol/legacy/server/Makefile.am)2
-rw-r--r--xlators/features/glupy/doc/README.md44
-rw-r--r--xlators/features/glupy/doc/TESTING9
-rw-r--r--xlators/features/glupy/doc/test.vol10
-rw-r--r--xlators/features/glupy/src/Makefile.am20
-rw-r--r--xlators/features/glupy/src/debug-trace.py774
-rw-r--r--xlators/features/glupy/src/glupy.c2470
-rw-r--r--xlators/features/glupy/src/glupy.h69
-rw-r--r--xlators/features/glupy/src/gluster.py841
-rw-r--r--xlators/features/glupy/src/helloworld.py19
-rw-r--r--xlators/features/glupy/src/negative.py92
-rw-r--r--xlators/features/index/Makefile.am3
-rw-r--r--xlators/features/index/src/Makefile.am17
-rw-r--r--xlators/features/index/src/index-mem-types.h22
-rw-r--r--xlators/features/index/src/index.c1489
-rw-r--r--xlators/features/index/src/index.h73
-rw-r--r--xlators/features/locks/src/Makefile.am17
-rw-r--r--xlators/features/locks/src/clear.c424
-rw-r--r--xlators/features/locks/src/clear.h76
-rw-r--r--xlators/features/locks/src/common.c293
-rw-r--r--xlators/features/locks/src/common.h93
-rw-r--r--xlators/features/locks/src/entrylk.c207
-rw-r--r--xlators/features/locks/src/inodelk.c347
-rw-r--r--xlators/features/locks/src/locks-mem-types.h21
-rw-r--r--xlators/features/locks/src/locks.h65
-rw-r--r--xlators/features/locks/src/posix.c1227
-rw-r--r--xlators/features/locks/src/reservelk.c31
-rw-r--r--xlators/features/locks/tests/unit-test.c22
-rw-r--r--xlators/features/mac-compat/src/Makefile.am7
-rw-r--r--xlators/features/mac-compat/src/mac-compat.c52
-rw-r--r--xlators/features/marker/Makefile.am2
-rw-r--r--xlators/features/marker/src/Makefile.am8
-rw-r--r--xlators/features/marker/src/marker-common.c37
-rw-r--r--xlators/features/marker/src/marker-common.h27
-rw-r--r--xlators/features/marker/src/marker-mem-types.h24
-rw-r--r--xlators/features/marker/src/marker-quota-helper.c52
-rw-r--r--xlators/features/marker/src/marker-quota-helper.h27
-rw-r--r--xlators/features/marker/src/marker-quota.c261
-rw-r--r--xlators/features/marker/src/marker-quota.h50
-rw-r--r--xlators/features/marker/src/marker.c729
-rw-r--r--xlators/features/marker/src/marker.h49
-rw-r--r--xlators/features/marker/utils/Makefile.am3
-rw-r--r--xlators/features/marker/utils/src/Makefile.am22
-rw-r--r--xlators/features/marker/utils/src/gsyncd.c346
-rw-r--r--xlators/features/marker/utils/src/procdiggy.c124
-rw-r--r--xlators/features/marker/utils/src/procdiggy.h26
-rw-r--r--xlators/features/marker/utils/syncdaemon/Makefile.am6
-rw-r--r--xlators/features/marker/utils/syncdaemon/README.md81
-rw-r--r--xlators/features/marker/utils/syncdaemon/__codecheck.py46
-rw-r--r--xlators/features/marker/utils/syncdaemon/__init__.py0
-rw-r--r--xlators/features/marker/utils/syncdaemon/configinterface.py224
-rw-r--r--xlators/features/marker/utils/syncdaemon/gconf.py20
-rw-r--r--xlators/features/marker/utils/syncdaemon/gsyncd.py369
-rw-r--r--xlators/features/marker/utils/syncdaemon/libcxattr.py72
-rw-r--r--xlators/features/marker/utils/syncdaemon/master.py518
-rw-r--r--xlators/features/marker/utils/syncdaemon/monitor.py123
-rw-r--r--xlators/features/marker/utils/syncdaemon/repce.py225
-rw-r--r--xlators/features/marker/utils/syncdaemon/resource.py837
-rw-r--r--xlators/features/marker/utils/syncdaemon/syncdutils.py269
-rw-r--r--xlators/features/path-convertor/src/Makefile.am7
-rw-r--r--xlators/features/path-convertor/src/path-mem-types.h20
-rw-r--r--xlators/features/path-convertor/src/path.c29
-rw-r--r--xlators/features/protect/Makefile.am (renamed from xlators/protocol/legacy/client/Makefile.am)0
-rw-r--r--xlators/features/protect/src/Makefile.am21
-rw-r--r--xlators/features/protect/src/prot_client.c215
-rw-r--r--xlators/features/protect/src/prot_dht.c168
-rw-r--r--xlators/features/protect/src/prot_server.c51
-rw-r--r--xlators/features/qemu-block/Makefile.am1
-rw-r--r--xlators/features/qemu-block/src/Makefile.am155
-rw-r--r--xlators/features/qemu-block/src/bdrv-xlator.c397
-rw-r--r--xlators/features/qemu-block/src/bh-syncop.c48
-rw-r--r--xlators/features/qemu-block/src/clock-timer.c60
-rw-r--r--xlators/features/qemu-block/src/coroutine-synctask.c116
-rw-r--r--xlators/features/qemu-block/src/monitor-logging.c50
-rw-r--r--xlators/features/qemu-block/src/qb-coroutines.c662
-rw-r--r--xlators/features/qemu-block/src/qb-coroutines.h30
-rw-r--r--xlators/features/qemu-block/src/qemu-block-memory-types.h25
-rw-r--r--xlators/features/qemu-block/src/qemu-block.c1140
-rw-r--r--xlators/features/qemu-block/src/qemu-block.h109
-rw-r--r--xlators/features/quiesce/src/Makefile.am7
-rw-r--r--xlators/features/quiesce/src/quiesce-mem-types.h20
-rw-r--r--xlators/features/quiesce/src/quiesce.c650
-rw-r--r--xlators/features/quiesce/src/quiesce.h23
-rw-r--r--xlators/features/quota/src/Makefile.am7
-rw-r--r--xlators/features/quota/src/quota-mem-types.h23
-rw-r--r--xlators/features/quota/src/quota.c876
-rw-r--r--xlators/features/quota/src/quota.h52
-rw-r--r--xlators/features/read-only/src/Makefile.am9
-rw-r--r--xlators/features/read-only/src/read-only-common.c133
-rw-r--r--xlators/features/read-only/src/read-only-common.h76
-rw-r--r--xlators/features/read-only/src/read-only.c22
-rw-r--r--xlators/features/read-only/src/worm.c36
-rw-r--r--xlators/features/trash/src/Makefile.am7
-rw-r--r--xlators/features/trash/src/trash-mem-types.h23
-rw-r--r--xlators/features/trash/src/trash.c109
-rw-r--r--xlators/features/trash/src/trash.h22
-rw-r--r--xlators/lib/src/libxlator.c479
-rw-r--r--xlators/lib/src/libxlator.h104
-rw-r--r--xlators/meta/src/Makefile.am5
-rw-r--r--xlators/meta/src/meta-mem-types.h20
-rw-r--r--xlators/meta/src/meta.c20
-rw-r--r--xlators/meta/src/meta.h20
-rw-r--r--xlators/meta/src/misc.c20
-rw-r--r--xlators/meta/src/misc.h20
-rw-r--r--xlators/meta/src/tree.c22
-rw-r--r--xlators/meta/src/tree.h20
-rw-r--r--xlators/meta/src/view.c20
-rw-r--r--xlators/meta/src/view.h20
-rw-r--r--xlators/mgmt/glusterd/src/Makefile.am55
-rw-r--r--xlators/mgmt/glusterd/src/glusterd-brick-ops.c1084
-rw-r--r--xlators/mgmt/glusterd/src/glusterd-geo-rep.c3451
-rw-r--r--xlators/mgmt/glusterd/src/glusterd-handler.c2511
-rw-r--r--xlators/mgmt/glusterd/src/glusterd-handshake.c1115
-rw-r--r--xlators/mgmt/glusterd/src/glusterd-hooks.c531
-rw-r--r--xlators/mgmt/glusterd/src/glusterd-hooks.h89
-rw-r--r--xlators/mgmt/glusterd/src/glusterd-locks.c637
-rw-r--r--xlators/mgmt/glusterd/src/glusterd-locks.h51
-rw-r--r--xlators/mgmt/glusterd/src/glusterd-log-ops.c118
-rw-r--r--xlators/mgmt/glusterd/src/glusterd-mem-types.h32
-rw-r--r--xlators/mgmt/glusterd/src/glusterd-mgmt-handler.c924
-rw-r--r--xlators/mgmt/glusterd/src/glusterd-mgmt.c1893
-rw-r--r--xlators/mgmt/glusterd/src/glusterd-mgmt.h45
-rw-r--r--xlators/mgmt/glusterd/src/glusterd-mountbroker.c94
-rw-r--r--xlators/mgmt/glusterd/src/glusterd-mountbroker.h24
-rw-r--r--xlators/mgmt/glusterd/src/glusterd-op-sm.c3876
-rw-r--r--xlators/mgmt/glusterd/src/glusterd-op-sm.h107
-rw-r--r--xlators/mgmt/glusterd/src/glusterd-pmap.c140
-rw-r--r--xlators/mgmt/glusterd/src/glusterd-pmap.h26
-rw-r--r--xlators/mgmt/glusterd/src/glusterd-quota.c120
-rw-r--r--xlators/mgmt/glusterd/src/glusterd-rebalance.c1052
-rw-r--r--xlators/mgmt/glusterd/src/glusterd-replace-brick.c1447
-rw-r--r--xlators/mgmt/glusterd/src/glusterd-rpc-ops.c1193
-rw-r--r--xlators/mgmt/glusterd/src/glusterd-sm.c148
-rw-r--r--xlators/mgmt/glusterd/src/glusterd-sm.h60
-rw-r--r--xlators/mgmt/glusterd/src/glusterd-snapshot.c5590
-rw-r--r--xlators/mgmt/glusterd/src/glusterd-store.c2798
-rw-r--r--xlators/mgmt/glusterd/src/glusterd-store.h158
-rw-r--r--xlators/mgmt/glusterd/src/glusterd-syncop.c1639
-rw-r--r--xlators/mgmt/glusterd/src/glusterd-syncop.h71
-rw-r--r--xlators/mgmt/glusterd/src/glusterd-utils.c6057
-rw-r--r--xlators/mgmt/glusterd/src/glusterd-utils.h364
-rw-r--r--xlators/mgmt/glusterd/src/glusterd-volgen.c1593
-rw-r--r--xlators/mgmt/glusterd/src/glusterd-volgen.h122
-rw-r--r--xlators/mgmt/glusterd/src/glusterd-volume-ops.c1561
-rw-r--r--xlators/mgmt/glusterd/src/glusterd-volume-set.c1452
-rw-r--r--xlators/mgmt/glusterd/src/glusterd.c820
-rw-r--r--xlators/mgmt/glusterd/src/glusterd.h636
-rw-r--r--xlators/mount/fuse/src/Makefile.am22
-rw-r--r--xlators/mount/fuse/src/fuse-bridge.c3091
-rw-r--r--xlators/mount/fuse/src/fuse-bridge.h313
-rw-r--r--xlators/mount/fuse/src/fuse-helpers.c350
-rw-r--r--xlators/mount/fuse/src/fuse-mem-types.h22
-rw-r--r--xlators/mount/fuse/src/fuse-resolve.c701
-rwxr-xr-xxlators/mount/fuse/utils/mount.glusterfs.in399
-rwxr-xr-xxlators/mount/fuse/utils/mount_glusterfs.in3
-rw-r--r--xlators/nfs/server/src/Makefile.am23
-rw-r--r--xlators/nfs/server/src/acl3.c708
-rw-r--r--xlators/nfs/server/src/acl3.h31
-rw-r--r--xlators/nfs/server/src/mount3.c1142
-rw-r--r--xlators/nfs/server/src/mount3.h39
-rw-r--r--xlators/nfs/server/src/mount3udp_svc.c189
-rw-r--r--xlators/nfs/server/src/nfs-common.c92
-rw-r--r--xlators/nfs/server/src/nfs-common.h22
-rw-r--r--xlators/nfs/server/src/nfs-fops.c437
-rw-r--r--xlators/nfs/server/src/nfs-fops.h41
-rw-r--r--xlators/nfs/server/src/nfs-generics.c43
-rw-r--r--xlators/nfs/server/src/nfs-generics.h30
-rw-r--r--xlators/nfs/server/src/nfs-inodes.c73
-rw-r--r--xlators/nfs/server/src/nfs-inodes.h19
-rw-r--r--xlators/nfs/server/src/nfs-mem-types.h30
-rw-r--r--xlators/nfs/server/src/nfs.c935
-rw-r--r--xlators/nfs/server/src/nfs.h43
-rw-r--r--xlators/nfs/server/src/nfs3-fh.c183
-rw-r--r--xlators/nfs/server/src/nfs3-fh.h55
-rw-r--r--xlators/nfs/server/src/nfs3-helpers.c459
-rw-r--r--xlators/nfs/server/src/nfs3-helpers.h23
-rw-r--r--xlators/nfs/server/src/nfs3.c681
-rw-r--r--xlators/nfs/server/src/nfs3.h122
-rw-r--r--xlators/nfs/server/src/nlm4.c2525
-rw-r--r--xlators/nfs/server/src/nlm4.h77
-rw-r--r--xlators/nfs/server/src/nlmcbk_svc.c117
-rw-r--r--xlators/performance/Makefile.am2
-rw-r--r--xlators/performance/io-cache/src/Makefile.am8
-rw-r--r--xlators/performance/io-cache/src/io-cache.c463
-rw-r--r--xlators/performance/io-cache/src/io-cache.h30
-rw-r--r--xlators/performance/io-cache/src/ioc-inode.c22
-rw-r--r--xlators/performance/io-cache/src/ioc-mem-types.h20
-rw-r--r--xlators/performance/io-cache/src/page.c74
-rw-r--r--xlators/performance/io-threads/src/Makefile.am7
-rw-r--r--xlators/performance/io-threads/src/io-threads.c1020
-rw-r--r--xlators/performance/io-threads/src/io-threads.h33
-rw-r--r--xlators/performance/io-threads/src/iot-mem-types.h21
-rw-r--r--xlators/performance/md-cache/Makefile.am1
-rw-r--r--xlators/performance/md-cache/src/Makefile.am25
-rw-r--r--xlators/performance/md-cache/src/md-cache-mem-types.h24
-rw-r--r--xlators/performance/md-cache/src/md-cache.c2303
-rw-r--r--xlators/performance/open-behind/Makefile.am1
-rw-r--r--xlators/performance/open-behind/src/Makefile.am15
-rw-r--r--xlators/performance/open-behind/src/open-behind-mem-types.h21
-rw-r--r--xlators/performance/open-behind/src/open-behind.c1001
-rw-r--r--xlators/performance/quick-read/src/Makefile.am7
-rw-r--r--xlators/performance/quick-read/src/quick-read-mem-types.h22
-rw-r--r--xlators/performance/quick-read/src/quick-read.c3536
-rw-r--r--xlators/performance/quick-read/src/quick-read.h61
-rw-r--r--xlators/performance/read-ahead/src/Makefile.am7
-rw-r--r--xlators/performance/read-ahead/src/page.c66
-rw-r--r--xlators/performance/read-ahead/src/read-ahead-mem-types.h20
-rw-r--r--xlators/performance/read-ahead/src/read-ahead.c421
-rw-r--r--xlators/performance/read-ahead/src/read-ahead.h23
-rw-r--r--xlators/performance/readdir-ahead/Makefile.am3
-rw-r--r--xlators/performance/readdir-ahead/src/Makefile.am15
-rw-r--r--xlators/performance/readdir-ahead/src/readdir-ahead-mem-types.h24
-rw-r--r--xlators/performance/readdir-ahead/src/readdir-ahead.c560
-rw-r--r--xlators/performance/readdir-ahead/src/readdir-ahead.h46
-rw-r--r--xlators/performance/stat-prefetch/src/Makefile.am14
-rw-r--r--xlators/performance/stat-prefetch/src/stat-prefetch-mem-types.h36
-rw-r--r--xlators/performance/stat-prefetch/src/stat-prefetch.c4263
-rw-r--r--xlators/performance/stat-prefetch/src/stat-prefetch.h106
-rw-r--r--xlators/performance/symlink-cache/src/Makefile.am7
-rw-r--r--xlators/performance/symlink-cache/src/symlink-cache.c56
-rw-r--r--xlators/performance/write-behind/src/Makefile.am7
-rw-r--r--xlators/performance/write-behind/src/write-behind-mem-types.h23
-rw-r--r--xlators/performance/write-behind/src/write-behind.c3615
-rw-r--r--xlators/playground/Makefile.am2
-rw-r--r--xlators/playground/template/Makefile.am2
-rw-r--r--xlators/playground/template/src/Makefile.am16
-rw-r--r--xlators/playground/template/src/template.c49
-rw-r--r--xlators/playground/template/src/template.h24
-rw-r--r--xlators/protocol/auth/addr/src/Makefile.am11
-rw-r--r--xlators/protocol/auth/addr/src/addr.c36
-rw-r--r--xlators/protocol/auth/login/src/Makefile.am7
-rw-r--r--xlators/protocol/auth/login/src/login.c22
-rw-r--r--xlators/protocol/client/src/Makefile.am13
-rw-r--r--xlators/protocol/client/src/client-callback.c25
-rw-r--r--xlators/protocol/client/src/client-handshake.c995
-rw-r--r--xlators/protocol/client/src/client-helpers.c124
-rw-r--r--xlators/protocol/client/src/client-lk.c380
-rw-r--r--xlators/protocol/client/src/client-mem-types.h23
-rw-r--r--xlators/protocol/client/src/client-rpc-fops.c (renamed from xlators/protocol/client/src/client3_1-fops.c)3430
-rw-r--r--xlators/protocol/client/src/client.c809
-rw-r--r--xlators/protocol/client/src/client.h134
-rw-r--r--xlators/protocol/legacy/Makefile.am3
-rw-r--r--xlators/protocol/legacy/client/src/Makefile.am21
-rw-r--r--xlators/protocol/legacy/client/src/client-mem-types.h43
-rw-r--r--xlators/protocol/legacy/client/src/client-protocol.c6683
-rw-r--r--xlators/protocol/legacy/client/src/client-protocol.h178
-rw-r--r--xlators/protocol/legacy/client/src/saved-frames.c196
-rw-r--r--xlators/protocol/legacy/client/src/saved-frames.h79
-rw-r--r--xlators/protocol/legacy/lib/src/Makefile.am14
-rw-r--r--xlators/protocol/legacy/lib/src/protocol.c108
-rw-r--r--xlators/protocol/legacy/lib/src/protocol.h1118
-rw-r--r--xlators/protocol/legacy/lib/src/transport.c422
-rw-r--r--xlators/protocol/legacy/lib/src/transport.h106
-rw-r--r--xlators/protocol/legacy/server/src/Makefile.am27
-rw-r--r--xlators/protocol/legacy/server/src/authenticate.c249
-rw-r--r--xlators/protocol/legacy/server/src/authenticate.h60
-rw-r--r--xlators/protocol/legacy/server/src/server-helpers.c622
-rw-r--r--xlators/protocol/legacy/server/src/server-helpers.h48
-rw-r--r--xlators/protocol/legacy/server/src/server-mem-types.h39
-rw-r--r--xlators/protocol/legacy/server/src/server-protocol.c6587
-rw-r--r--xlators/protocol/legacy/server/src/server-protocol.h191
-rw-r--r--xlators/protocol/legacy/server/src/server-resolve.c658
-rw-r--r--xlators/protocol/legacy/transport/Makefile.am3
-rw-r--r--xlators/protocol/legacy/transport/ib-verbs/Makefile.am1
-rw-r--r--xlators/protocol/legacy/transport/ib-verbs/src/Makefile.am19
-rw-r--r--xlators/protocol/legacy/transport/ib-verbs/src/ib-verbs-mem-types.h39
-rw-r--r--xlators/protocol/legacy/transport/ib-verbs/src/ib-verbs.c2625
-rw-r--r--xlators/protocol/legacy/transport/ib-verbs/src/ib-verbs.h220
-rw-r--r--xlators/protocol/legacy/transport/ib-verbs/src/name.c712
-rw-r--r--xlators/protocol/legacy/transport/ib-verbs/src/name.h47
-rw-r--r--xlators/protocol/legacy/transport/socket/Makefile.am1
-rw-r--r--xlators/protocol/legacy/transport/socket/src/Makefile.am19
-rw-r--r--xlators/protocol/legacy/transport/socket/src/name.c740
-rw-r--r--xlators/protocol/legacy/transport/socket/src/name.h44
-rw-r--r--xlators/protocol/legacy/transport/socket/src/socket-mem-types.h36
-rw-r--r--xlators/protocol/legacy/transport/socket/src/socket.c1625
-rw-r--r--xlators/protocol/legacy/transport/socket/src/socket.h129
-rw-r--r--xlators/protocol/server/src/Makefile.am20
-rw-r--r--xlators/protocol/server/src/authenticate.c109
-rw-r--r--xlators/protocol/server/src/authenticate.h19
-rw-r--r--xlators/protocol/server/src/server-handshake.c208
-rw-r--r--xlators/protocol/server/src/server-helpers.c1296
-rw-r--r--xlators/protocol/server/src/server-helpers.h68
-rw-r--r--xlators/protocol/server/src/server-mem-types.h20
-rw-r--r--xlators/protocol/server/src/server-resolve.c78
-rw-r--r--xlators/protocol/server/src/server-rpc-fops.c6179
-rw-r--r--xlators/protocol/server/src/server.c764
-rw-r--r--xlators/protocol/server/src/server.h124
-rw-r--r--xlators/protocol/server/src/server3_1-fops.c5235
-rw-r--r--xlators/storage/Makefile.am6
-rw-r--r--xlators/storage/bd/Makefile.am3
-rw-r--r--xlators/storage/bd/src/Makefile.am20
-rw-r--r--xlators/storage/bd/src/bd-aio.c527
-rw-r--r--xlators/storage/bd/src/bd-aio.h41
-rw-r--r--xlators/storage/bd/src/bd-helper.c783
-rw-r--r--xlators/storage/bd/src/bd.c2404
-rw-r--r--xlators/storage/bd/src/bd.h178
-rw-r--r--xlators/storage/bdb/src/Makefile.am18
-rw-r--r--xlators/storage/bdb/src/bctx.c341
-rw-r--r--xlators/storage/bdb/src/bdb-ll.c1464
-rw-r--r--xlators/storage/bdb/src/bdb-mem-types.h42
-rw-r--r--xlators/storage/bdb/src/bdb.c3603
-rw-r--r--xlators/storage/bdb/src/bdb.h530
-rw-r--r--xlators/storage/posix/src/Makefile.am17
-rw-r--r--xlators/storage/posix/src/posix-aio.c569
-rw-r--r--xlators/storage/posix/src/posix-aio.h39
-rw-r--r--xlators/storage/posix/src/posix-handle.c173
-rw-r--r--xlators/storage/posix/src/posix-handle.h49
-rw-r--r--xlators/storage/posix/src/posix-helpers.c610
-rw-r--r--xlators/storage/posix/src/posix-mem-types.h21
-rw-r--r--xlators/storage/posix/src/posix.c2136
-rw-r--r--xlators/storage/posix/src/posix.h99
-rw-r--r--xlators/system/posix-acl/src/Makefile.am10
-rw-r--r--xlators/system/posix-acl/src/posix-acl-xattr.c28
-rw-r--r--xlators/system/posix-acl/src/posix-acl-xattr.h44
-rw-r--r--xlators/system/posix-acl/src/posix-acl.c418
-rw-r--r--xlators/system/posix-acl/src/posix-acl.h71
455 files changed, 125134 insertions, 82969 deletions
diff --git a/xlators/Makefile.am b/xlators/Makefile.am
index b1643d26c..f60fa85ce 100644
--- a/xlators/Makefile.am
+++ b/xlators/Makefile.am
@@ -1,3 +1,4 @@
-SUBDIRS = cluster storage protocol performance debug features encryption mount nfs mgmt system
+SUBDIRS = cluster storage protocol performance debug features encryption mount nfs mgmt system \
+ playground
CLEANFILES =
diff --git a/xlators/bindings/python/src/Makefile.am b/xlators/bindings/python/src/Makefile.am
index c0b9141c6..90370d861 100644
--- a/xlators/bindings/python/src/Makefile.am
+++ b/xlators/bindings/python/src/Makefile.am
@@ -9,7 +9,7 @@ pythondir = $(xlatordir)/python
python_so_SOURCES = python.c
-AM_CFLAGS = -fPIC -D_FILE_OFFSET_BITS=64 -D_GNU_SOURCE -Wall \
+AM_CFLAGS = -fPIC $(GF_CPPFLAGS) -Wall \
-I$(top_srcdir)/libglusterfs/src -shared -nostartfiles \
$(PYTHON_CPPLAGS) -DGLUSTER_PYTHON_PATH=\"$(pythondir)\"
diff --git a/xlators/bindings/python/src/gluster.py b/xlators/bindings/python/src/gluster.py
index ee0eb1310..337c983ec 100644
--- a/xlators/bindings/python/src/gluster.py
+++ b/xlators/bindings/python/src/gluster.py
@@ -1,19 +1,12 @@
-# Copyright (c) 2007 Chris AtLee <chris@atlee.ca>
-# This file is part of GlusterFS.
-#
-# GlusterFS is free software; you can redistribute it and/or modify
-# it under the terms of the GNU General Public License as published
-# by the Free Software Foundation; either version 3 of the License,
-# or (at your option) any later version.
-#
-# GlusterFS is distributed in the hope that it will be useful, but
-# WITHOUT ANY WARRANTY; without even the implied warranty of
-# MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU
-# General Public License for more details.
+
+# Copyright (c) 2007-2012 Red Hat, Inc. <http://www.redhat.com>
+# This file is part of GlusterFS.
#
-# You should have received a copy of the GNU General Public License
-# along with this program. If not, see
-# <http://www.gnu.org/licenses/>.
+# This file is licensed to you under your choice of the GNU Lesser
+# General Public License, version 3 or any later version (LGPLv3 or
+# later), or the GNU General Public License, version 2 (GPLv2), in all
+# cases as published by the Free Software Foundation.
+
from ctypes import *
from glustertypes import *
from glusterstack import *
diff --git a/xlators/bindings/python/src/glusterstack.py b/xlators/bindings/python/src/glusterstack.py
index ba24c8165..0c071ae98 100644
--- a/xlators/bindings/python/src/glusterstack.py
+++ b/xlators/bindings/python/src/glusterstack.py
@@ -1,19 +1,12 @@
-# Copyright (c) 2007 Chris AtLee <chris@atlee.ca>
-# This file is part of GlusterFS.
-#
-# GlusterFS is free software; you can redistribute it and/or modify
-# it under the terms of the GNU General Public License as published
-# by the Free Software Foundation; either version 3 of the License,
-# or (at your option) any later version.
-#
-# GlusterFS is distributed in the hope that it will be useful, but
-# WITHOUT ANY WARRANTY; without even the implied warranty of
-# MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU
-# General Public License for more details.
+
+# Copyright (c) 2007-2012 Red Hat, Inc. <http://www.redhat.com>
+# This file is part of GlusterFS.
#
-# You should have received a copy of the GNU General Public License
-# along with this program. If not, see
-# <http://www.gnu.org/licenses/>.
+# This file is licensed to you under your choice of the GNU Lesser
+# General Public License, version 3 or any later version (LGPLv3 or
+# later), or the GNU General Public License, version 2 (GPLv2), in all
+# cases as published by the Free Software Foundation.
+
from ctypes import *
from glustertypes import *
diff --git a/xlators/bindings/python/src/glustertypes.py b/xlators/bindings/python/src/glustertypes.py
index e9069d07c..98437d22e 100644
--- a/xlators/bindings/python/src/glustertypes.py
+++ b/xlators/bindings/python/src/glustertypes.py
@@ -1,19 +1,12 @@
-# Copyright (c) 2007 Chris AtLee <chris@atlee.ca>
-# This file is part of GlusterFS.
-#
-# GlusterFS is free software; you can redistribute it and/or modify
-# it under the terms of the GNU General Public License as published
-# by the Free Software Foundation; either version 3 of the License,
-# or (at your option) any later version.
-#
-# GlusterFS is distributed in the hope that it will be useful, but
-# WITHOUT ANY WARRANTY; without even the implied warranty of
-# MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU
-# General Public License for more details.
+
+# Copyright (c) 2007-2012 Red Hat, Inc. <http://www.redhat.com>
+# This file is part of GlusterFS.
#
-# You should have received a copy of the GNU General Public License
-# along with this program. If not, see
-# <http://www.gnu.org/licenses/>.
+# This file is licensed to you under your choice of the GNU Lesser
+# General Public License, version 3 or any later version (LGPLv3 or
+# later), or the GNU General Public License, version 2 (GPLv2), in all
+# cases as published by the Free Software Foundation.
+
from ctypes import *
import collections
diff --git a/xlators/bindings/python/src/python.c b/xlators/bindings/python/src/python.c
index 3310a2115..9b96790de 100644
--- a/xlators/bindings/python/src/python.c
+++ b/xlators/bindings/python/src/python.c
@@ -1,22 +1,12 @@
/*
- Copyright (c) 2007-2010 Chris AtLee <chris@atlee.ca>
+ Copyright (c) 2008-2012 Red Hat, Inc. <http://www.redhat.com>
This file is part of GlusterFS.
- GlusterFS is free software; you can redistribute it and/or modify
- it under the terms of the GNU General Public License as published
- by the Free Software Foundation; either version 3 of the License,
- or (at your option) any later version.
-
- GlusterFS is distributed in the hope that it will be useful, but
- WITHOUT ANY WARRANTY; without even the implied warranty of
- MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU
- General Public License for more details.
-
- You should have received a copy of the GNU General Public License
- along with this program. If not, see
- <http://www.gnu.org/licenses/>.
+ This file is licensed to you under your choice of the GNU Lesser
+ General Public License, version 3 or any later version (LGPLv3 or
+ later), or the GNU General Public License, version 2 (GPLv2), in all
+ cases as published by the Free Software Foundation.
*/
-
#include <Python.h>
#ifndef _CONFIG_H
@@ -45,7 +35,7 @@ python_writev (call_frame_t *frame,
xlator_t *this,
fd_t *fd,
struct iovec *vector,
- int32_t count,
+ int32_t count,
off_t offset)
{
python_private_t *priv = (python_private_t *)this->private;
@@ -148,7 +138,7 @@ init (xlator_t *this)
Py_InitializeEx(0);
if (!this->children) {
- gf_log ("python", GF_LOG_ERROR,
+ gf_log ("python", GF_LOG_ERROR,
"FATAL: python should have exactly one child");
return -1;
}
@@ -166,7 +156,7 @@ init (xlator_t *this)
}
priv->pInterp = Py_NewInterpreter();
-
+
// Adjust python's path
PyObject *syspath = PySys_GetObject("path");
PyObject *path = PyString_FromString(GLUSTER_PYTHON_PATH);
@@ -188,7 +178,7 @@ init (xlator_t *this)
priv->pVectorType = PyObject_GetAttrString(priv->pGlusterModule, "iovec");
gf_log("python", GF_LOG_DEBUG, "Loading script...%s", priv->scriptname);
-
+
priv->pScriptModule = AnonModule_FromFile(priv->scriptname);
if (!priv->pScriptModule || PyErr_Occurred())
{
@@ -217,7 +207,7 @@ init (xlator_t *this)
return 0;
}
-void
+void
fini (xlator_t *this)
{
python_private_t *priv = (python_private_t*)(this->private);
diff --git a/xlators/bindings/python/src/testxlator.py b/xlators/bindings/python/src/testxlator.py
index 507455c85..59a991dca 100644
--- a/xlators/bindings/python/src/testxlator.py
+++ b/xlators/bindings/python/src/testxlator.py
@@ -1,19 +1,12 @@
-# Copyright (c) 2007 Chris AtLee <chris@atlee.ca>
-# This file is part of GlusterFS.
-#
-# GlusterFS is free software; you can redistribute it and/or modify
-# it under the terms of the GNU General Public License as published
-# by the Free Software Foundation; either version 3 of the License,
-# or (at your option) any later version.
-#
-# GlusterFS is distributed in the hope that it will be useful, but
-# WITHOUT ANY WARRANTY; without even the implied warranty of
-# MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU
-# General Public License for more details.
-#
-# You should have received a copy of the GNU General Public License
-# along with this program. If not, see
-# <http://www.gnu.org/licenses/>.
+"""
+ Copyright (c) 2007-2012 Red Hat, Inc. <http://www.redhat.com>
+ This file is part of GlusterFS.
+
+ This file is licensed to you under your choice of the GNU Lesser
+ General Public License, version 3 or any later version (LGPLv3 or
+ later), or the GNU General Public License, version 2 (GPLv2), in all
+ cases as published by the Free Software Foundation.
+"""
"""
This is a test translator written in python.
diff --git a/xlators/cluster/afr/src/Makefile.am b/xlators/cluster/afr/src/Makefile.am
index 16ed25af1..35d18a6c0 100644
--- a/xlators/cluster/afr/src/Makefile.am
+++ b/xlators/cluster/afr/src/Makefile.am
@@ -1,21 +1,31 @@
xlator_LTLIBRARIES = afr.la pump.la
xlatordir = $(libdir)/glusterfs/$(PACKAGE_VERSION)/xlator/cluster
-afr_common_source = afr-dir-read.c afr-dir-write.c afr-inode-read.c afr-inode-write.c afr-open.c afr-transaction.c afr-self-heal-data.c afr-self-heal-common.c afr-self-heal-metadata.c afr-self-heal-entry.c afr-self-heal-algorithm.c afr-lk-common.c afr-self-heald.c $(top_builddir)/xlators/lib/src/libxlator.c
+afr_common_source = afr-dir-read.c afr-dir-write.c afr-inode-read.c \
+ afr-inode-write.c afr-open.c afr-transaction.c afr-self-heal-data.c \
+ afr-self-heal-common.c afr-self-heal-metadata.c afr-self-heal-entry.c \
+ afr-self-heal-algorithm.c afr-lk-common.c afr-self-heald.c \
+ $(top_builddir)/xlators/lib/src/libxlator.c
-afr_la_LDFLAGS = -module -avoidversion
+afr_la_LDFLAGS = -module -avoid-version
afr_la_SOURCES = $(afr_common_source) afr.c
afr_la_LIBADD = $(top_builddir)/libglusterfs/src/libglusterfs.la
-pump_la_LDFLAGS = -module -avoidversion
+pump_la_LDFLAGS = -module -avoid-version
pump_la_SOURCES = $(afr_common_source) pump.c
pump_la_LIBADD = $(top_builddir)/libglusterfs/src/libglusterfs.la
-noinst_HEADERS = afr.h afr-transaction.h afr-inode-write.h afr-inode-read.h afr-dir-read.h afr-dir-write.h afr-self-heal.h afr-self-heal-common.h afr-self-heal-algorithm.h pump.h afr-mem-types.h afr-common.c afr-self-heald.h $(top_builddir)/xlators/lib/src/libxlator.h $(top_builddir)/glusterfsd/src/glusterfsd.h
+noinst_HEADERS = afr.h afr-transaction.h afr-inode-write.h afr-inode-read.h \
+ afr-dir-read.h afr-dir-write.h afr-self-heal.h afr-self-heal-common.h \
+ afr-self-heal-algorithm.h pump.h afr-mem-types.h afr-common.c \
+ afr-self-heald.h $(top_builddir)/xlators/lib/src/libxlator.h \
+ $(top_builddir)/glusterfsd/src/glusterfsd.h
-AM_CFLAGS = -fPIC -D_FILE_OFFSET_BITS=64 -D_GNU_SOURCE -Wall -D$(GF_HOST_OS) \
- -I$(top_srcdir)/libglusterfs/src -I$(top_srcdir)/contrib/md5 -shared -nostartfiles $(GF_CFLAGS) \
- -I$(top_srcdir)/xlators/lib/src
+AM_CPPFLAGS = $(GF_CPPFLAGS) \
+ -I$(top_srcdir)/libglusterfs/src -I$(top_srcdir)/xlators/lib/src \
+ -I$(top_srcdir)/rpc/rpc-lib/src
+
+AM_CFLAGS = -Wall $(GF_CFLAGS)
CLEANFILES =
diff --git a/xlators/cluster/afr/src/afr-common.c b/xlators/cluster/afr/src/afr-common.c
index c9a8b5955..af01f2ef2 100644
--- a/xlators/cluster/afr/src/afr-common.c
+++ b/xlators/cluster/afr/src/afr-common.c
@@ -1,20 +1,11 @@
/*
- Copyright (c) 2007-2011 Gluster, Inc. <http://www.gluster.com>
+ Copyright (c) 2008-2012 Red Hat, Inc. <http://www.redhat.com>
This file is part of GlusterFS.
- GlusterFS is free software; you can redistribute it and/or modify
- it under the terms of the GNU General Public License as published
- by the Free Software Foundation; either version 3 of the License,
- or (at your option) any later version.
-
- GlusterFS is distributed in the hope that it will be useful, but
- WITHOUT ANY WARRANTY; without even the implied warranty of
- MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU
- General Public License for more details.
-
- You should have received a copy of the GNU General Public License
- along with this program. If not, see
- <http://www.gnu.org/licenses/>.
+ This file is licensed to you under your choice of the GNU Lesser
+ General Public License, version 3 or any later version (LGPLv3 or
+ later), or the GNU General Public License, version 2 (GPLv2), in all
+ cases as published by the Free Software Foundation.
*/
#include <libgen.h>
@@ -58,10 +49,9 @@
#include "afr-self-heald.h"
#include "pump.h"
-#define AFR_ICTX_OPENDIR_DONE_MASK 0x0000000200000000ULL
-#define AFR_ICTX_SPLIT_BRAIN_MASK 0x0000000100000000ULL
+#define AFR_ICTX_OPENDIR_DONE_MASK 0x0000000100000000ULL
#define AFR_ICTX_READ_CHILD_MASK 0x00000000FFFFFFFFULL
-
+#define AFR_STATISTICS_HISTORY_SIZE 50
int
afr_lookup_done_success_action (call_frame_t *frame, xlator_t *this,
gf_boolean_t fail_conflict);
@@ -92,6 +82,11 @@ afr_xattr_req_prepare (xlator_t *this, dict_t *xattr_req, const char *path)
path, priv->pending_key[i]);
/* 3 = data+metadata+entry */
}
+ ret = dict_set_int32 (xattr_req, GF_GFIDLESS_LOOKUP, 1);
+ if (ret) {
+ gf_log (this->name, GF_LOG_DEBUG, "%s: failed to set gfidless "
+ "lookup", path);
+ }
}
int
@@ -123,6 +118,13 @@ afr_lookup_xattr_req_prepare (afr_local_t *local, xlator_t *this,
loc->path, GLUSTERFS_ENTRYLK_COUNT);
}
+ ret = dict_set_uint32 (local->xattr_req, GLUSTERFS_PARENT_ENTRYLK, 0);
+ if (ret < 0) {
+ gf_log (this->name, GF_LOG_WARNING,
+ "%s: Unable to set dict value for %s",
+ loc->path, GLUSTERFS_PARENT_ENTRYLK);
+ }
+
ret = dict_get_ptr (local->xattr_req, "gfid-req", gfid_req);
if (ret) {
gf_log (this->name, GF_LOG_DEBUG,
@@ -200,60 +202,86 @@ out:
return ret;
}
-afr_inode_ctx_t*
-afr_inode_ctx_get_from_addr (uint64_t addr, int32_t child_count)
+void
+afr_inode_ctx_destroy (afr_inode_ctx_t *ctx)
{
- int ret = -1;
- afr_inode_ctx_t *ctx = NULL;
- size_t size = 0;
+ if (!ctx)
+ return;
+ GF_FREE (ctx->fresh_children);
+ GF_FREE (ctx);
+}
- GF_ASSERT (child_count > 0);
+afr_inode_ctx_t*
+__afr_inode_ctx_get (inode_t *inode, xlator_t *this)
+{
+ int ret = 0;
+ uint64_t ctx_addr = 0;
+ afr_inode_ctx_t *ctx = NULL;
+ afr_private_t *priv = NULL;
- if (!addr) {
- ctx = GF_CALLOC (1, sizeof (*ctx),
- gf_afr_mt_inode_ctx_t);
- if (!ctx)
- goto out;
- size = sizeof (*ctx->fresh_children);
- ctx->fresh_children = GF_CALLOC (child_count, size,
- gf_afr_mt_int32_t);
- if (!ctx->fresh_children)
- goto out;
- } else {
- ctx = (afr_inode_ctx_t*) (long) addr;
+ priv = this->private;
+ ret = __inode_ctx_get (inode, this, &ctx_addr);
+ if (ret < 0)
+ ctx_addr = 0;
+ if (ctx_addr != 0) {
+ ctx = (afr_inode_ctx_t*) (long) ctx_addr;
+ goto out;
}
- ret = 0;
+ ctx = GF_CALLOC (1, sizeof (*ctx),
+ gf_afr_mt_inode_ctx_t);
+ if (!ctx)
+ goto fail;
+ ctx->fresh_children = GF_CALLOC (priv->child_count,
+ sizeof (*ctx->fresh_children),
+ gf_afr_mt_int32_t);
+ if (!ctx->fresh_children)
+ goto fail;
+ ret = __inode_ctx_put (inode, this, (uint64_t)ctx);
+ if (ret) {
+ gf_log_callingfn (this->name, GF_LOG_ERROR, "failed to "
+ "set the inode ctx (%s)",
+ uuid_utoa (inode->gfid));
+ goto fail;
+ }
+
out:
- if (ret && ctx) {
- if (ctx->fresh_children)
- GF_FREE (ctx->fresh_children);
- GF_FREE (ctx);
- ctx = NULL;
+ return ctx;
+
+fail:
+ afr_inode_ctx_destroy (ctx);
+ return NULL;
+}
+
+afr_inode_ctx_t*
+afr_inode_ctx_get (inode_t *inode, xlator_t *this)
+{
+ afr_inode_ctx_t *ctx = NULL;
+
+ LOCK (&inode->lock);
+ {
+ ctx = __afr_inode_ctx_get (inode, this);
}
+ UNLOCK (&inode->lock);
return ctx;
}
void
-afr_inode_get_ctx (xlator_t *this, inode_t *inode, afr_inode_params_t *params)
+afr_inode_get_ctx_params (xlator_t *this, inode_t *inode,
+ afr_inode_params_t *params)
{
GF_ASSERT (inode);
GF_ASSERT (params);
- int ret = 0;
afr_inode_ctx_t *ctx = NULL;
afr_private_t *priv = NULL;
int i = 0;
- uint64_t ctx_addr = 0;
int32_t read_child = -1;
int32_t *fresh_children = NULL;
priv = this->private;
LOCK (&inode->lock);
{
- ret = __inode_ctx_get (inode, this, &ctx_addr);
- if (ret < 0)
- goto unlock;
- ctx = afr_inode_ctx_get_from_addr (ctx_addr, priv->child_count);
+ ctx = __afr_inode_ctx_get (inode, this);
if (!ctx)
goto unlock;
switch (params->op) {
@@ -272,12 +300,6 @@ afr_inode_get_ctx (xlator_t *this, inode_t *inode, afr_inode_params_t *params)
if (ctx->masks & AFR_ICTX_OPENDIR_DONE_MASK)
params->u.value = _gf_true;
break;
- case AFR_INODE_GET_SPLIT_BRAIN:
- params->u.value = _gf_false;
- if (ctx->masks & AFR_ICTX_SPLIT_BRAIN_MASK)
- params->u.value = _gf_true;
- ;
- break;
default:
GF_ASSERT (0);
break;
@@ -290,11 +312,16 @@ unlock:
gf_boolean_t
afr_is_split_brain (xlator_t *this, inode_t *inode)
{
- afr_inode_params_t params = {0};
+ afr_inode_ctx_t *ctx = NULL;
+ gf_boolean_t spb = _gf_false;
- params.op = AFR_INODE_GET_SPLIT_BRAIN;
- afr_inode_get_ctx (this, inode, &params);
- return params.u.value;
+ ctx = afr_inode_ctx_get (inode, this);
+ if (!ctx)
+ goto out;
+ if ((ctx->mdata_spb == SPB) || (ctx->data_spb == SPB))
+ spb = _gf_true;
+out:
+ return spb;
}
gf_boolean_t
@@ -303,11 +330,10 @@ afr_is_opendir_done (xlator_t *this, inode_t *inode)
afr_inode_params_t params = {0};
params.op = AFR_INODE_GET_OPENDIR_DONE;
- afr_inode_get_ctx (this, inode, &params);
+ afr_inode_get_ctx_params (this, inode, &params);
return params.u.value;
}
-
int32_t
afr_inode_get_read_ctx (xlator_t *this, inode_t *inode, int32_t *fresh_children)
{
@@ -315,7 +341,7 @@ afr_inode_get_read_ctx (xlator_t *this, inode_t *inode, int32_t *fresh_children)
params.op = AFR_INODE_GET_READ_CTX;
params.u.read_ctx.children = fresh_children;
- afr_inode_get_ctx (this, inode, &params);
+ afr_inode_get_ctx_params (this, inode, &params);
return params.u.read_ctx.read_child;
}
@@ -377,31 +403,14 @@ afr_inode_ctx_set_opendir_done (afr_inode_ctx_t *ctx)
}
void
-afr_inode_ctx_set_splitbrain (afr_inode_ctx_t *ctx, gf_boolean_t set)
-{
- uint64_t remaining_mask = 0;
- uint64_t mask = 0;
-
- if (set) {
- remaining_mask = (~AFR_ICTX_SPLIT_BRAIN_MASK & ctx->masks);
- mask = (0xFFFFFFFFFFFFFFFFULL & AFR_ICTX_SPLIT_BRAIN_MASK);
- ctx->masks = remaining_mask | mask;
- } else {
- ctx->masks = (~AFR_ICTX_SPLIT_BRAIN_MASK & ctx->masks);
- }
-}
-
-void
-afr_inode_set_ctx (xlator_t *this, inode_t *inode, afr_inode_params_t *params)
+afr_inode_set_ctx_params (xlator_t *this, inode_t *inode,
+ afr_inode_params_t *params)
{
GF_ASSERT (inode);
GF_ASSERT (params);
- int ret = 0;
afr_inode_ctx_t *ctx = NULL;
afr_private_t *priv = NULL;
- uint64_t ctx_addr = 0;
- gf_boolean_t set = _gf_false;
int32_t read_child = -1;
int32_t *fresh_children = NULL;
int32_t *stale_children = NULL;
@@ -409,10 +418,7 @@ afr_inode_set_ctx (xlator_t *this, inode_t *inode, afr_inode_params_t *params)
priv = this->private;
LOCK (&inode->lock);
{
- ret = __inode_ctx_get (inode, this, &ctx_addr);
- if (ret < 0)
- ctx_addr = 0;
- ctx = afr_inode_ctx_get_from_addr (ctx_addr, priv->child_count);
+ ctx = __afr_inode_ctx_get (inode, this);
if (!ctx)
goto unlock;
switch (params->op) {
@@ -432,33 +438,26 @@ afr_inode_set_ctx (xlator_t *this, inode_t *inode, afr_inode_params_t *params)
case AFR_INODE_SET_OPENDIR_DONE:
afr_inode_ctx_set_opendir_done (ctx);
break;
- case AFR_INODE_SET_SPLIT_BRAIN:
- set = params->u.value;
- afr_inode_ctx_set_splitbrain (ctx, set);
- break;
default:
GF_ASSERT (0);
break;
}
- ret = __inode_ctx_put (inode, this, (uint64_t)ctx);
- if (ret) {
- gf_log_callingfn (this->name, GF_LOG_ERROR, "failed to "
- "set the inode ctx (%s)",
- uuid_utoa (inode->gfid));
- }
}
unlock:
UNLOCK (&inode->lock);
}
void
-afr_set_split_brain (xlator_t *this, inode_t *inode, gf_boolean_t set)
+afr_set_split_brain (xlator_t *this, inode_t *inode, afr_spb_state_t mdata_spb,
+ afr_spb_state_t data_spb)
{
- afr_inode_params_t params = {0};
+ afr_inode_ctx_t *ctx = NULL;
- params.op = AFR_INODE_SET_SPLIT_BRAIN;
- params.u.value = set;
- afr_inode_set_ctx (this, inode, &params);
+ ctx = afr_inode_ctx_get (inode, this);
+ if (mdata_spb != DONT_KNOW)
+ ctx->mdata_spb = mdata_spb;
+ if (data_spb != DONT_KNOW)
+ ctx->data_spb = data_spb;
}
void
@@ -467,7 +466,7 @@ afr_set_opendir_done (xlator_t *this, inode_t *inode)
afr_inode_params_t params = {0};
params.op = AFR_INODE_SET_OPENDIR_DONE;
- afr_inode_set_ctx (this, inode, &params);
+ afr_inode_set_ctx_params (this, inode, &params);
}
void
@@ -486,7 +485,7 @@ afr_inode_set_read_ctx (xlator_t *this, inode_t *inode, int32_t read_child,
params.op = AFR_INODE_SET_READ_CTX;
params.u.read_ctx.read_child = read_child;
params.u.read_ctx.children = fresh_children;
- afr_inode_set_ctx (this, inode, &params);
+ afr_inode_set_ctx_params (this, inode, &params);
}
void
@@ -499,7 +498,7 @@ afr_inode_rm_stale_children (xlator_t *this, inode_t *inode,
params.op = AFR_INODE_RM_STALE_CHILDREN;
params.u.read_ctx.children = stale_children;
- afr_inode_set_ctx (this, inode, &params);
+ afr_inode_set_ctx_params (this, inode, &params);
}
gf_boolean_t
@@ -543,6 +542,10 @@ afr_is_read_child (int32_t *success_children, int32_t *sources,
gf_boolean_t success_child = _gf_false;
gf_boolean_t source = _gf_false;
+ if (child < 0) {
+ return _gf_false;
+ }
+
GF_ASSERT (success_children);
GF_ASSERT (child_count > 0);
@@ -559,29 +562,69 @@ out:
return (success_child && source);
}
+int32_t
+afr_hash_child (int32_t *success_children, int32_t child_count,
+ unsigned int hmode, uuid_t gfid)
+{
+ uuid_t gfid_copy = {0,};
+ pid_t pid;
+
+ if (!hmode) {
+ return -1;
+ }
+
+ if (gfid) {
+ uuid_copy(gfid_copy,gfid);
+ }
+ if (hmode > 1) {
+ /*
+ * Why getpid? Because it's one of the cheapest calls
+ * available - faster than gethostname etc. - and returns a
+ * constant-length value that's sure to be shorter than a UUID.
+ * It's still very unlikely to be the same across clients, so
+ * it still provides good mixing. We're not trying for
+ * perfection here. All we need is a low probability that
+ * multiple clients won't converge on the same subvolume.
+ */
+ pid = getpid();
+ memcpy (gfid_copy, &pid, sizeof(pid));
+ }
+
+ return SuperFastHash((char *)gfid_copy,
+ sizeof(gfid_copy)) % child_count;
+}
+
/* If sources is NULL the xattrs are assumed to be of source for all
* success_children.
*/
int
-afr_select_read_child_from_policy (int32_t *success_children, int32_t child_count,
- int32_t prev_read_child,
- int32_t config_read_child, int32_t *sources)
+afr_select_read_child_from_policy (int32_t *success_children,
+ int32_t child_count, int32_t prev_read_child,
+ int32_t config_read_child, int32_t *sources,
+ unsigned int hmode, uuid_t gfid)
{
int32_t read_child = -1;
int i = 0;
GF_ASSERT (success_children);
- read_child = prev_read_child;
+ read_child = config_read_child;
if (afr_is_read_child (success_children, sources, child_count,
read_child))
goto out;
- read_child = config_read_child;
+ read_child = prev_read_child;
if (afr_is_read_child (success_children, sources, child_count,
read_child))
goto out;
+ read_child = afr_hash_child (success_children, child_count,
+ hmode, gfid);
+ if (afr_is_read_child (success_children, sources, child_count,
+ read_child)) {
+ goto out;
+ }
+
for (i = 0; i < child_count; i++) {
read_child = success_children[i];
if (read_child < 0)
@@ -601,7 +644,7 @@ out:
void
afr_set_read_ctx_from_policy (xlator_t *this, inode_t *inode,
int32_t *fresh_children, int32_t prev_read_child,
- int32_t config_read_child)
+ int32_t config_read_child, uuid_t gfid)
{
int read_child = -1;
afr_private_t *priv = NULL;
@@ -611,7 +654,8 @@ afr_set_read_ctx_from_policy (xlator_t *this, inode_t *inode,
priv->child_count,
prev_read_child,
config_read_child,
- NULL);
+ NULL,
+ priv->hash_mode, gfid);
if (read_child >= 0)
afr_inode_set_read_ctx (this, inode, read_child,
fresh_children);
@@ -667,8 +711,11 @@ afr_get_call_child (xlator_t *this, unsigned char *child_up, int32_t read_child,
GF_ASSERT (call_child);
GF_ASSERT (last_index);
GF_ASSERT (fresh_children);
- GF_ASSERT (read_child >= 0);
+ if (read_child < 0) {
+ ret = -EIO;
+ goto out;
+ }
priv = this->private;
*call_child = -1;
*last_index = -1;
@@ -717,81 +764,66 @@ out:
}
void
+afr_xattr_array_destroy (dict_t **xattr, unsigned int child_count)
+{
+ afr_reset_xattr (xattr, child_count);
+ GF_FREE (xattr);
+}
+
+void
afr_local_sh_cleanup (afr_local_t *local, xlator_t *this)
{
afr_self_heal_t *sh = NULL;
afr_private_t *priv = NULL;
- int i = 0;
-
sh = &local->self_heal;
priv = this->private;
- if (sh->buf)
- GF_FREE (sh->buf);
+ if (sh->data_sh_info && strcmp (sh->data_sh_info, ""))
+ GF_FREE (sh->data_sh_info);
+
+ if (sh->metadata_sh_info && strcmp (sh->metadata_sh_info, ""))
+ GF_FREE (sh->metadata_sh_info);
+
+ GF_FREE (sh->buf);
- if (sh->parentbufs)
- GF_FREE (sh->parentbufs);
+ GF_FREE (sh->parentbufs);
if (sh->inode)
inode_unref (sh->inode);
- if (sh->xattr) {
- afr_reset_xattr (sh->xattr, priv->child_count);
- GF_FREE (sh->xattr);
- }
-
- if (sh->child_errno)
- GF_FREE (sh->child_errno);
+ afr_xattr_array_destroy (sh->xattr, priv->child_count);
- if (sh->pending_matrix) {
- for (i = 0; i < priv->child_count; i++) {
- GF_FREE (sh->pending_matrix[i]);
- }
- GF_FREE (sh->pending_matrix);
- }
+ GF_FREE (sh->child_errno);
- if (sh->delta_matrix) {
- for (i = 0; i < priv->child_count; i++) {
- GF_FREE (sh->delta_matrix[i]);
- }
- GF_FREE (sh->delta_matrix);
- }
+ afr_matrix_cleanup (sh->pending_matrix, priv->child_count);
+ afr_matrix_cleanup (sh->delta_matrix, priv->child_count);
- if (sh->sources)
- GF_FREE (sh->sources);
+ GF_FREE (sh->sources);
- if (sh->success)
- GF_FREE (sh->success);
+ GF_FREE (sh->success);
- if (sh->locked_nodes)
- GF_FREE (sh->locked_nodes);
+ GF_FREE (sh->locked_nodes);
if (sh->healing_fd) {
fd_unref (sh->healing_fd);
sh->healing_fd = NULL;
}
- if (sh->linkname)
- GF_FREE ((char *)sh->linkname);
+ GF_FREE ((char *)sh->linkname);
- if (sh->success_children)
- GF_FREE (sh->success_children);
+ GF_FREE (sh->success_children);
- if (sh->fresh_children)
- GF_FREE (sh->fresh_children);
+ GF_FREE (sh->fresh_children);
- if (sh->fresh_parent_dirs)
- GF_FREE (sh->fresh_parent_dirs);
+ GF_FREE (sh->fresh_parent_dirs);
loc_wipe (&sh->parent_loc);
loc_wipe (&sh->lookup_loc);
- if (sh->checksum)
- GF_FREE (sh->checksum);
+ GF_FREE (sh->checksum);
- if (sh->write_needed)
- GF_FREE (sh->write_needed);
+ GF_FREE (sh->write_needed);
if (sh->healing_fd)
fd_unref (sh->healing_fd);
}
@@ -800,34 +832,26 @@ afr_local_sh_cleanup (afr_local_t *local, xlator_t *this)
void
afr_local_transaction_cleanup (afr_local_t *local, xlator_t *this)
{
- int i = 0;
- afr_private_t * priv = NULL;
+ afr_private_t *priv = NULL;
+ int i = 0;
priv = this->private;
- for (i = 0; i < priv->child_count; i++) {
- if (local->pending && local->pending[i])
- GF_FREE (local->pending[i]);
- }
-
- GF_FREE (local->pending);
-
- if (local->internal_lock.locked_nodes)
- GF_FREE (local->internal_lock.locked_nodes);
+ afr_matrix_cleanup (local->pending, priv->child_count);
+ afr_matrix_cleanup (local->transaction.txn_changelog,
+ priv->child_count);
- if (local->internal_lock.inode_locked_nodes)
- GF_FREE (local->internal_lock.inode_locked_nodes);
+ GF_FREE (local->internal_lock.locked_nodes);
- if (local->internal_lock.entry_locked_nodes)
- GF_FREE (local->internal_lock.entry_locked_nodes);
+ for (i = 0; local->internal_lock.inodelk[i].domain; i++) {
+ GF_FREE (local->internal_lock.inodelk[i].locked_nodes);
+ }
- if (local->internal_lock.lower_locked_nodes)
- GF_FREE (local->internal_lock.lower_locked_nodes);
+ GF_FREE (local->internal_lock.lower_locked_nodes);
+ afr_entry_lockee_cleanup (&local->internal_lock);
GF_FREE (local->transaction.pre_op);
- GF_FREE (local->transaction.child_errno);
- GF_FREE (local->child_errno);
GF_FREE (local->transaction.eager_lock);
GF_FREE (local->transaction.basename);
@@ -835,6 +859,8 @@ afr_local_transaction_cleanup (afr_local_t *local, xlator_t *this)
loc_wipe (&local->transaction.parent_loc);
loc_wipe (&local->transaction.new_parent_loc);
+
+ GF_FREE (local->transaction.postop_piggybacked);
}
@@ -861,14 +887,16 @@ afr_local_cleanup (afr_local_t *local, xlator_t *this)
if (local->xattr_req)
dict_unref (local->xattr_req);
- if (local->child_up)
- GF_FREE (local->child_up);
+ if (local->dict)
+ dict_unref (local->dict);
+
+ GF_FREE(local->replies);
- if (local->fresh_children)
- GF_FREE (local->fresh_children);
+ GF_FREE (local->child_up);
- if (local->fd_open_on)
- GF_FREE (local->fd_open_on);
+ GF_FREE (local->child_errno);
+
+ GF_FREE (local->fresh_children);
{ /* lookup */
if (local->cont.lookup.xattrs) {
@@ -886,27 +914,23 @@ afr_local_cleanup (afr_local_t *local, xlator_t *this)
inode_unref (local->cont.lookup.inode);
}
- if (local->cont.lookup.postparents)
- GF_FREE (local->cont.lookup.postparents);
+ GF_FREE (local->cont.lookup.postparents);
- if (local->cont.lookup.bufs)
- GF_FREE (local->cont.lookup.bufs);
+ GF_FREE (local->cont.lookup.bufs);
- if (local->cont.lookup.success_children)
- GF_FREE (local->cont.lookup.success_children);
+ GF_FREE (local->cont.lookup.success_children);
- if (local->cont.lookup.sources)
- GF_FREE (local->cont.lookup.sources);
+ GF_FREE (local->cont.lookup.sources);
+ afr_matrix_cleanup (local->cont.lookup.pending_matrix,
+ priv->child_count);
}
{ /* getxattr */
- if (local->cont.getxattr.name)
- GF_FREE (local->cont.getxattr.name);
+ GF_FREE (local->cont.getxattr.name);
}
{ /* lk */
- if (local->cont.lk.locked_nodes)
- GF_FREE (local->cont.lk.locked_nodes);
+ GF_FREE (local->cont.lk.locked_nodes);
}
{ /* create */
@@ -948,20 +972,32 @@ afr_local_cleanup (afr_local_t *local, xlator_t *this)
{ /* removexattr */
GF_FREE (local->cont.removexattr.name);
}
-
+ { /* xattrop */
+ if (local->cont.xattrop.xattr)
+ dict_unref (local->cont.xattrop.xattr);
+ }
+ { /* fxattrop */
+ if (local->cont.fxattrop.xattr)
+ dict_unref (local->cont.fxattrop.xattr);
+ }
{ /* symlink */
GF_FREE (local->cont.symlink.linkpath);
}
{ /* opendir */
- if (local->cont.opendir.checksum)
- GF_FREE (local->cont.opendir.checksum);
+ GF_FREE (local->cont.opendir.checksum);
}
{ /* readdirp */
if (local->cont.readdir.dict)
dict_unref (local->cont.readdir.dict);
}
+
+ if (local->xdata_req)
+ dict_unref (local->xdata_req);
+
+ if (local->xdata_rsp)
+ dict_unref (local->xdata_rsp);
}
@@ -1044,34 +1080,144 @@ afr_update_loc_gfids (loc_t *loc, struct iatt *buf, struct iatt *postparent)
uuid_copy (loc->pargfid, postparent->ia_gfid);
}
+/*
+ * Quota size xattrs are not maintained by afr. There is a
+ * possibility that they differ even when both the directory changelog xattrs
+ * suggest everything is fine. So if there is at least one 'source' check among
+ * the sources which has the maximum quota size. Otherwise check among all the
+ * available ones for maximum quota size. This way if there is a source and
+ * stale copies it always votes for the 'source'.
+ * */
+
+static void
+afr_handle_quota_size (afr_local_t *local, xlator_t *this,
+ dict_t *rsp_dict)
+{
+ int32_t *sources = NULL;
+ dict_t *xattr = NULL;
+ data_t *max_data = NULL;
+ int64_t max_quota_size = -1;
+ data_t *data = NULL;
+ int64_t *size = NULL;
+ int64_t quota_size = -1;
+ afr_private_t *priv = NULL;
+ int i = 0;
+ int ret = -1;
+ gf_boolean_t source_present = _gf_false;
+
+ priv = this->private;
+ sources = local->cont.lookup.sources;
+
+ if (rsp_dict == NULL) {
+ gf_log_callingfn (this->name, GF_LOG_ERROR, "%s: Invalid "
+ "response dictionary", local->loc.path);
+ return;
+ }
+
+ for (i = 0; i < priv->child_count; i++) {
+ if (sources[i]) {
+ source_present = _gf_true;
+ break;
+ }
+ }
+
+ for (i = 0; i < priv->child_count; i++) {
+ /*
+ * If there is at least one source lets check
+ * for maximum quota sizes among sources, otherwise take the
+ * maximum of the ones present to be on the safer side.
+ */
+ if (source_present && !sources[i])
+ continue;
+
+ xattr = local->cont.lookup.xattrs[i];
+ if (!xattr)
+ continue;
+
+ data = dict_get (xattr, QUOTA_SIZE_KEY);
+ if (!data)
+ continue;
+
+ size = (int64_t*)data->data;
+ quota_size = ntoh64(*size);
+ gf_log (this->name, GF_LOG_DEBUG, "%s: %d, size: %"PRId64,
+ local->loc.path, i, quota_size);
+ if (quota_size > max_quota_size) {
+ if (max_data)
+ data_unref (max_data);
+
+ max_quota_size = quota_size;
+ max_data = data_ref (data);
+ }
+ }
+
+ if (max_data) {
+ ret = dict_set (rsp_dict, QUOTA_SIZE_KEY, max_data);
+ if (ret) {
+ gf_log (this->name, GF_LOG_ERROR, "%s: Failed to set "
+ "quota size", local->loc.path);
+ }
+
+ data_unref (max_data);
+ }
+}
+
int
afr_lookup_build_response_params (afr_local_t *local, xlator_t *this)
{
- int32_t read_child = -1;
struct iatt *buf = NULL;
struct iatt *postparent = NULL;
dict_t **xattr = NULL;
+ int32_t *success_children = NULL;
+ int32_t *sources = NULL;
+ afr_private_t *priv = NULL;
+ int32_t read_child = -1;
int ret = 0;
+ int i = 0;
GF_ASSERT (local);
buf = &local->cont.lookup.buf;
postparent = &local->cont.lookup.postparent;
xattr = &local->cont.lookup.xattr;
+ priv = this->private;
read_child = afr_inode_get_read_ctx (this, local->cont.lookup.inode,
- NULL);
+ local->fresh_children);
+ if (read_child < 0) {
+ ret = -1;
+ goto out;
+ }
+ success_children = local->cont.lookup.success_children;
+ sources = local->cont.lookup.sources;
+ memset (sources, 0, sizeof (*sources) * priv->child_count);
+ afr_children_intersection_get (local->fresh_children, success_children,
+ sources, priv->child_count);
+ if (!sources[read_child]) {
+ read_child = -1;
+ for (i = 0; i < priv->child_count; i++) {
+ if (sources[i]) {
+ read_child = i;
+ break;
+ }
+ }
+ }
if (read_child < 0) {
ret = -1;
goto out;
}
+
gf_log (this->name, GF_LOG_DEBUG, "Building lookup response from %d",
read_child);
if (!*xattr)
*xattr = dict_ref (local->cont.lookup.xattrs[read_child]);
+
*buf = local->cont.lookup.bufs[read_child];
*postparent = local->cont.lookup.postparents[read_child];
+ if (dict_get (local->xattr_req, QUOTA_SIZE_KEY))
+ afr_handle_quota_size (local, this, *xattr);
+
if (IA_INVAL == local->cont.lookup.inode->ia_type) {
/* fix for RT #602 */
local->cont.lookup.inode->ia_type = buf->ia_type;
@@ -1087,6 +1233,7 @@ afr_lookup_update_lk_counts (afr_local_t *local, xlator_t *this,
uint32_t inodelk_count = 0;
uint32_t entrylk_count = 0;
int ret = -1;
+ uint32_t parent_entrylk = 0;
GF_ASSERT (local);
GF_ASSERT (this);
@@ -1102,43 +1249,103 @@ afr_lookup_update_lk_counts (afr_local_t *local, xlator_t *this,
&entrylk_count);
if (ret == 0)
local->entrylk_count += entrylk_count;
+ ret = dict_get_uint32 (xattr, GLUSTERFS_PARENT_ENTRYLK,
+ &parent_entrylk);
+ if (!ret)
+ local->cont.lookup.parent_entrylk += parent_entrylk;
}
+/*
+ * It's important to maintain a commutative property on do_*_self_heal and
+ * found*; once set, they must not be cleared by a subsequent iteration or
+ * call, so that they represent a logical OR of all iterations and calls
+ * regardless of child/key order. That allows the caller to call us multiple
+ * times without having to use a separate variable as a "reduce" accumulator.
+ */
static void
afr_lookup_set_self_heal_params_by_xattr (afr_local_t *local, xlator_t *this,
dict_t *xattr)
{
+ afr_private_t *priv = NULL;
+ int i = 0;
+ int ret = -1;
+ void *pending_raw = NULL;
+ int32_t *pending = NULL;
+
GF_ASSERT (local);
GF_ASSERT (this);
GF_ASSERT (xattr);
- if (afr_sh_has_metadata_pending (xattr, this)) {
- local->self_heal.do_metadata_self_heal = _gf_true;
- gf_log(this->name, GF_LOG_DEBUG,
- "metadata self-heal is pending for %s.",
- local->loc.path);
- }
+ priv = this->private;
- if (afr_sh_has_entry_pending (xattr, this)) {
- local->self_heal.do_entry_self_heal = _gf_true;
- gf_log(this->name, GF_LOG_DEBUG,
- "entry self-heal is pending for %s.", local->loc.path);
- }
+ for (i = 0; i < priv->child_count; i++) {
+ ret = dict_get_ptr (xattr, priv->pending_key[i],
+ &pending_raw);
+ if (ret != 0) {
+ continue;
+ }
+ pending = pending_raw;
- if (afr_sh_has_data_pending (xattr, this)) {
- local->self_heal.do_data_self_heal = _gf_true;
- gf_log(this->name, GF_LOG_DEBUG,
- "data self-heal is pending for %s.", local->loc.path);
+ if (pending[AFR_METADATA_TRANSACTION]) {
+ gf_log(this->name, GF_LOG_DEBUG,
+ "metadata self-heal is pending for %s.",
+ local->loc.path);
+ local->self_heal.do_metadata_self_heal = _gf_true;
+ }
+
+ if (pending[AFR_ENTRY_TRANSACTION]) {
+ gf_log(this->name, GF_LOG_DEBUG,
+ "entry self-heal is pending for %s.",
+ local->loc.path);
+ local->self_heal.do_entry_self_heal = _gf_true;
+ }
+
+ if (pending[AFR_DATA_TRANSACTION]) {
+ gf_log(this->name, GF_LOG_DEBUG,
+ "data self-heal is pending for %s.",
+ local->loc.path);
+ local->self_heal.do_data_self_heal = _gf_true;
+ }
}
}
+void
+afr_lookup_check_set_metadata_split_brain (afr_local_t *local, xlator_t *this)
+{
+ int32_t *sources = NULL;
+ afr_private_t *priv = NULL;
+ int32_t subvol_status = 0;
+ int32_t *success_children = NULL;
+ dict_t **xattrs = NULL;
+ struct iatt *bufs = NULL;
+ int32_t **pending_matrix = NULL;
+
+ priv = this->private;
+
+ sources = GF_CALLOC (priv->child_count, sizeof (*sources),
+ gf_afr_mt_int32_t);
+ if (NULL == sources)
+ goto out;
+ success_children = local->cont.lookup.success_children;
+ xattrs = local->cont.lookup.xattrs;
+ bufs = local->cont.lookup.bufs;
+ pending_matrix = local->cont.lookup.pending_matrix;
+ afr_build_sources (this, xattrs, bufs, pending_matrix,
+ sources, success_children, AFR_METADATA_TRANSACTION,
+ &subvol_status, _gf_false);
+ if (subvol_status & SPLIT_BRAIN)
+ local->cont.lookup.possible_spb = _gf_true;
+out:
+ GF_FREE (sources);
+}
+
static void
afr_detect_self_heal_by_iatt (afr_local_t *local, xlator_t *this,
struct iatt *buf, struct iatt *lookup_buf)
{
if (PERMISSION_DIFFERS (buf, lookup_buf)) {
/* mismatching permissions */
- gf_log (this->name, GF_LOG_INFO,
+ gf_log (this->name, GF_LOG_DEBUG,
"permissions differ for %s ", local->loc.path);
local->self_heal.do_metadata_self_heal = _gf_true;
}
@@ -1146,27 +1353,45 @@ afr_detect_self_heal_by_iatt (afr_local_t *local, xlator_t *this,
if (OWNERSHIP_DIFFERS (buf, lookup_buf)) {
/* mismatching permissions */
local->self_heal.do_metadata_self_heal = _gf_true;
- gf_log (this->name, GF_LOG_INFO,
+ gf_log (this->name, GF_LOG_DEBUG,
"ownership differs for %s ", local->loc.path);
}
if (SIZE_DIFFERS (buf, lookup_buf)
&& IA_ISREG (buf->ia_type)) {
- gf_log (this->name, GF_LOG_INFO,
+ gf_log (this->name, GF_LOG_DEBUG,
"size differs for %s ", local->loc.path);
local->self_heal.do_data_self_heal = _gf_true;
}
if (uuid_compare (buf->ia_gfid, lookup_buf->ia_gfid)) {
/* mismatching gfid */
- gf_log (this->name, GF_LOG_WARNING,
+ gf_log (this->name, GF_LOG_DEBUG,
"%s: gfid different on subvolume", local->loc.path);
}
}
static void
-afr_detect_self_heal_by_lookup_status (afr_local_t *local, xlator_t *this,
- gf_boolean_t split_brain)
+afr_detect_self_heal_by_split_brain_status (afr_local_t *local, xlator_t *this)
+{
+ gf_boolean_t split_brain = _gf_false;
+ afr_self_heal_t *sh = NULL;
+
+ sh = &local->self_heal;
+
+ split_brain = afr_is_split_brain (this, local->cont.lookup.inode);
+ split_brain = split_brain || local->cont.lookup.possible_spb;
+ if ((local->success_count > 0) && split_brain &&
+ IA_ISREG (local->cont.lookup.inode->ia_type)) {
+ sh->force_confirm_spb = _gf_true;
+ gf_log (this->name, GF_LOG_DEBUG,
+ "split brain detected during lookup of %s.",
+ local->loc.path);
+ }
+}
+
+static void
+afr_detect_self_heal_by_lookup_status (afr_local_t *local, xlator_t *this)
{
GF_ASSERT (local);
GF_ASSERT (this);
@@ -1177,24 +1402,11 @@ afr_detect_self_heal_by_lookup_status (afr_local_t *local, xlator_t *this,
local->self_heal.do_entry_self_heal = _gf_true;
local->self_heal.do_gfid_self_heal = _gf_true;
local->self_heal.do_missing_entry_self_heal = _gf_true;
- gf_log(this->name, GF_LOG_INFO,
+ gf_log(this->name, GF_LOG_DEBUG,
"entries are missing in lookup of %s.",
local->loc.path);
- //If all self-heals are needed no need to check for other rules
- goto out;
- }
-
- if ((local->success_count > 0) && split_brain &&
- IA_ISREG (local->cont.lookup.inode->ia_type)) {
- local->self_heal.do_data_self_heal = _gf_true;
- local->self_heal.do_gfid_self_heal = _gf_true;
- local->self_heal.do_missing_entry_self_heal = _gf_true;
- gf_log (this->name, GF_LOG_WARNING,
- "split brain detected during lookup of %s.",
- local->loc.path);
}
-out:
return;
}
@@ -1204,6 +1416,8 @@ afr_can_self_heal_proceed (afr_self_heal_t *sh, afr_private_t *priv)
GF_ASSERT (sh);
GF_ASSERT (priv);
+ if (sh->force_confirm_spb)
+ return _gf_true;
return (sh->do_gfid_self_heal
|| sh->do_missing_entry_self_heal
|| (afr_data_self_heal_enabled (priv->data_self_heal) &&
@@ -1237,6 +1451,7 @@ afr_lookup_select_read_child (afr_local_t *local, xlator_t *this,
dict_t **xattrs = NULL;
int32_t *success_children = NULL;
afr_transaction_type type = AFR_METADATA_TRANSACTION;
+ uuid_t *gfid = NULL;
GF_ASSERT (local);
GF_ASSERT (this);
@@ -1250,8 +1465,9 @@ afr_lookup_select_read_child (afr_local_t *local, xlator_t *this,
ia_type = local->cont.lookup.bufs[success_children[0]].ia_type;
type = afr_transaction_type_get (ia_type);
xattrs = local->cont.lookup.xattrs;
+ gfid = &local->cont.lookup.buf.ia_gfid;
source = afr_lookup_select_read_child_by_txn_type (this, local, xattrs,
- type);
+ type, *gfid);
if (source < 0) {
gf_log (this->name, GF_LOG_DEBUG, "failed to select source "
"for %s", local->loc.path);
@@ -1279,7 +1495,8 @@ afr_launch_self_heal (call_frame_t *frame, xlator_t *this, inode_t *inode,
void (*gfid_sh_success_cbk) (call_frame_t *sh_frame,
xlator_t *this),
int (*unwind) (call_frame_t *frame, xlator_t *this,
- int32_t op_ret, int32_t op_errno))
+ int32_t op_ret, int32_t op_errno,
+ int32_t sh_failed))
{
afr_local_t *local = NULL;
char sh_type_str[256] = {0,};
@@ -1302,7 +1519,7 @@ afr_launch_self_heal (call_frame_t *frame, xlator_t *this, inode_t *inode,
if (background)
bg = "background";
- gf_log (this->name, GF_LOG_INFO,
+ gf_log (this->name, GF_LOG_DEBUG,
"%s %s self-heal triggered. path: %s, reason: %s", bg,
sh_type_str, local->loc.path, reason);
@@ -1373,7 +1590,7 @@ afr_conflicting_iattrs (struct iatt *bufs, int32_t *success_children,
child2 = &bufs[success_children[i-1]];
if (FILETYPE_DIFFERS (child1, child2)) {
- gf_log (xlator_name, GF_LOG_WARNING, "%s: filetype "
+ gf_log (xlator_name, GF_LOG_DEBUG, "%s: filetype "
"differs on subvolumes (%d, %d)", path,
success_children[i-1], success_children[i]);
conflicting = _gf_true;
@@ -1382,7 +1599,7 @@ afr_conflicting_iattrs (struct iatt *bufs, int32_t *success_children,
if (!gfid || uuid_is_null (child1->ia_gfid))
continue;
if (uuid_compare (*gfid, child1->ia_gfid)) {
- gf_log (xlator_name, GF_LOG_WARNING, "%s: gfid differs"
+ gf_log (xlator_name, GF_LOG_DEBUG, "%s: gfid differs"
" on subvolume %d", path, success_children[i]);
conflicting = _gf_true;
goto out;
@@ -1465,13 +1682,11 @@ afr_lookup_set_self_heal_params (afr_local_t *local, xlator_t *this)
int32_t child1 = -1;
int32_t child2 = -1;
afr_self_heal_t *sh = NULL;
- gf_boolean_t split_brain = _gf_false;
priv = this->private;
sh = &local->self_heal;
- split_brain = afr_is_split_brain (this, local->cont.lookup.inode);
- afr_detect_self_heal_by_lookup_status (local, this, split_brain);
+ afr_detect_self_heal_by_lookup_status (local, this);
if (afr_lookup_gfid_missing_count (local, this))
local->self_heal.do_gfid_self_heal = _gf_true;
@@ -1498,23 +1713,28 @@ afr_lookup_set_self_heal_params (afr_local_t *local, xlator_t *this)
afr_lookup_set_self_heal_params_by_xattr (local, this,
xattr[child1]);
}
- if (afr_open_only_data_self_heal (priv->data_self_heal)
- && !split_brain)
+ if (afr_open_only_data_self_heal (priv->data_self_heal))
sh->do_data_self_heal = _gf_false;
+ if (sh->do_metadata_self_heal)
+ afr_lookup_check_set_metadata_split_brain (local, this);
+ afr_detect_self_heal_by_split_brain_status (local, this);
}
int
afr_self_heal_lookup_unwind (call_frame_t *frame, xlator_t *this,
- int32_t op_ret, int32_t op_errno)
+ int32_t op_ret, int32_t op_errno,
+ int32_t sh_failed)
{
afr_local_t *local = NULL;
+ int ret = -1;
+ dict_t *xattr = NULL;
local = frame->local;
if (op_ret == -1) {
local->op_ret = -1;
- if (afr_error_more_important (local->op_errno, op_errno))
- local->op_errno = op_errno;
+ local->op_errno = afr_most_important_error(local->op_errno,
+ op_errno, _gf_true);
goto out;
} else {
@@ -1522,6 +1742,23 @@ afr_self_heal_lookup_unwind (call_frame_t *frame, xlator_t *this,
}
afr_lookup_done_success_action (frame, this, _gf_true);
+ xattr = local->cont.lookup.xattr;
+ if (xattr) {
+ ret = dict_set_int32 (xattr, "sh-failed", sh_failed);
+ if (ret)
+ gf_log (this->name, GF_LOG_ERROR, "%s: Failed to set "
+ "sh-failed to %d", local->loc.path, sh_failed);
+
+ if (local->self_heal.actual_sh_started == _gf_true &&
+ sh_failed == 0) {
+ ret = dict_set_int32 (xattr, "actual-sh-done", 1);
+ if (ret)
+ gf_log(this->name, GF_LOG_ERROR, "%s: Failed to"
+ " set actual-sh-done to %d",
+ local->loc.path,
+ local->self_heal.actual_sh_started);
+ }
+ }
out:
AFR_STACK_UNWIND (lookup, frame, local->op_ret, local->op_errno,
local->cont.lookup.inode, &local->cont.lookup.buf,
@@ -1595,7 +1832,8 @@ afr_lookup_perform_self_heal (call_frame_t *frame, xlator_t *this,
afr_lookup_set_self_heal_params (local, this);
if (afr_can_self_heal_proceed (&local->self_heal, priv)) {
- if (afr_is_transaction_running (local))
+ if (afr_is_transaction_running (local) &&
+ (!local->allow_sh_for_running_transaction))
goto out;
reason = "lookup detected pending operations";
@@ -1656,26 +1894,23 @@ afr_lookup_done_success_action (call_frame_t *frame, xlator_t *this,
int32_t read_child = -1;
int32_t ret = -1;
afr_local_t *local = NULL;
- afr_private_t *priv = NULL;
+ gf_boolean_t fresh_lookup = _gf_false;
local = frame->local;
- priv = this->private;
+ fresh_lookup = local->cont.lookup.fresh_lookup;
if (local->loc.parent == NULL)
fail_conflict = _gf_true;
- if (afr_conflicting_iattrs (local->cont.lookup.bufs,
- local->cont.lookup.success_children,
- priv->child_count, local->loc.path,
- this->name)) {
+ if (afr_lookup_conflicting_entries (local, this)) {
if (fail_conflict == _gf_false)
ret = 0;
goto out;
}
- if (!afr_is_transaction_running (local)) {
- ret = afr_lookup_select_read_child (local, this, &read_child);
- if (ret)
+ ret = afr_lookup_select_read_child (local, this, &read_child);
+ if (!afr_is_transaction_running (local) || fresh_lookup) {
+ if (read_child < 0)
goto out;
ret = afr_lookup_set_read_ctx (local, this, read_child);
@@ -1686,11 +1921,9 @@ afr_lookup_done_success_action (call_frame_t *frame, xlator_t *this,
ret = afr_lookup_build_response_params (local, this);
if (ret)
goto out;
- if (afr_is_fresh_lookup (&local->loc, this)) {
- afr_update_loc_gfids (&local->loc,
- &local->cont.lookup.buf,
- &local->cont.lookup.postparent);
- }
+ afr_update_loc_gfids (&local->loc,
+ &local->cont.lookup.buf,
+ &local->cont.lookup.postparent);
ret = 0;
out:
@@ -1701,6 +1934,135 @@ out:
return ret;
}
+int
+afr_lookup_get_latest_subvol (afr_local_t *local, xlator_t *this)
+{
+ afr_private_t *priv = NULL;
+ int32_t *success_children = NULL;
+ struct iatt *bufs = NULL;
+ int i = 0;
+ int child = 0;
+ int lsubvol = -1;
+
+ priv = this->private;
+ success_children = local->cont.lookup.success_children;
+ bufs = local->cont.lookup.bufs;
+ for (i = 0; i < priv->child_count; i++) {
+ child = success_children[i];
+ if (child == -1)
+ break;
+ if (uuid_is_null (bufs[child].ia_gfid))
+ continue;
+ if (lsubvol < 0) {
+ lsubvol = child;
+ } else if (bufs[lsubvol].ia_ctime < bufs[child].ia_ctime) {
+ lsubvol = child;
+ } else if ((bufs[lsubvol].ia_ctime == bufs[child].ia_ctime) &&
+ (bufs[lsubvol].ia_ctime_nsec < bufs[child].ia_ctime_nsec)) {
+ lsubvol = child;
+ }
+ }
+ return lsubvol;
+}
+
+void
+afr_lookup_mark_other_entries_stale (afr_local_t *local, xlator_t *this,
+ int subvol)
+{
+ afr_private_t *priv = NULL;
+ int32_t *success_children = NULL;
+ struct iatt *bufs = NULL;
+ int i = 0;
+ int child = 0;
+
+ priv = this->private;
+ success_children = local->cont.lookup.success_children;
+ bufs = local->cont.lookup.bufs;
+ memcpy (local->fresh_children, success_children,
+ sizeof (*success_children) * priv->child_count);
+ for (i = 0; i < priv->child_count; i++) {
+ child = local->fresh_children[i];
+ if (child == -1)
+ break;
+ if (child == subvol)
+ continue;
+ if (uuid_is_null (bufs[child].ia_gfid) &&
+ (bufs[child].ia_type == bufs[subvol].ia_type))
+ continue;
+ afr_children_rm_child (success_children, child,
+ priv->child_count);
+ local->success_count--;
+ }
+ afr_reset_children (local->fresh_children, priv->child_count);
+}
+
+void
+afr_succeed_lookup_on_latest_iatt (afr_local_t *local, xlator_t *this)
+{
+ int lsubvol = 0;
+
+ if (!afr_lookup_conflicting_entries (local, this))
+ goto out;
+
+ lsubvol = afr_lookup_get_latest_subvol (local, this);
+ if (lsubvol < 0)
+ goto out;
+ afr_lookup_mark_other_entries_stale (local, this, lsubvol);
+out:
+ return;
+}
+
+gf_boolean_t
+afr_is_entry_possibly_under_creation (afr_local_t *local, xlator_t *this)
+{
+ /*
+ * We need to perform this test in lookup done and treat on going
+ * create/DELETE as ENOENT.
+ * Reason:
+ Multiple clients A, B and C are attempting 'mkdir -p /mnt/a/b/c'
+
+ 1 Client A is in the middle of mkdir(/a). It has acquired lock.
+ It has performed mkdir(/a) on one subvol, and second one is still
+ in progress
+ 2 Client B performs a lookup, sees directory /a on one,
+ ENOENT on the other, succeeds lookup.
+ 3 Client B performs lookup on /a/b on both subvols, both return ENOENT
+ (one subvol because /a/b does not exist, another because /a
+ itself does not exist)
+ 4 Client B proceeds to mkdir /a/b. It obtains entrylk on inode=/a with
+ basename=b on one subvol, but fails on other subvol as /a is yet to
+ be created by Client A.
+ 5 Client A finishes mkdir of /a on other subvol
+ 6 Client C also attempts to create /a/b, lookup returns ENOENT on
+ both subvols.
+ 7 Client C tries to obtain entrylk on on inode=/a with basename=b,
+ obtains on one subvol (where B had failed), and waits for B to unlock
+ on other subvol.
+ 8 Client B finishes mkdir() on one subvol with GFID-1 and completes
+ transaction and unlocks
+ 9 Client C gets the lock on the second subvol, At this stage second
+ subvol already has /a/b created from Client B, but Client C does not
+ check that in the middle of mkdir transaction
+ 10 Client C attempts mkdir /a/b on both subvols. It succeeds on
+ ONLY ONE (where Client B could not get lock because of
+ missing parent /a dir) with GFID-2, and gets EEXIST from ONE subvol.
+ This way we have /a/b in GFID mismatch. One subvol got GFID-1 because
+ Client B performed transaction on only one subvol (because entrylk()
+ could not be obtained on second subvol because of missing parent dir --
+ caused by premature/speculative succeeding of lookup() on /a when locks
+ are detected). Other subvol gets GFID-2 from Client C because while
+ it was waiting for entrylk() on both subvols, Client B was in the
+ middle of creating mkdir() on only one subvol, and Client C does not
+ "expect" this when it is between lock() and pre-op()/op() phase of the
+ transaction.
+ */
+ if (local->cont.lookup.parent_entrylk && local->enoent_count)
+ return _gf_true;
+
+ return _gf_false;
+}
+
+
static void
afr_lookup_done (call_frame_t *frame, xlator_t *this)
{
@@ -1717,8 +2079,18 @@ afr_lookup_done (call_frame_t *frame, xlator_t *this)
priv = this->private;
local = frame->local;
+ if (afr_is_entry_possibly_under_creation (local, this)) {
+ local->op_ret = -1;
+ local->op_errno = ENOENT;
+ goto unwind;
+ }
+
if (local->op_ret < 0)
goto unwind;
+
+ if (local->cont.lookup.parent_entrylk && local->success_count > 1)
+ afr_succeed_lookup_on_latest_iatt (local, this);
+
gfid_miss_count = afr_lookup_gfid_missing_count (local, this);
up_children_count = afr_up_children_count (local->child_up,
priv->child_count);
@@ -1770,25 +2142,20 @@ afr_lookup_done (call_frame_t *frame, xlator_t *this)
* others in that they must be given higher priority while
* returning to the user.
*
- * The hierarchy is ESTALE > ENOENT > others
- *
+ * The hierarchy is ESTALE > EIO > ENOENT > others
*/
-
-gf_boolean_t
-afr_error_more_important (int32_t old_errno, int32_t new_errno)
+int32_t
+afr_most_important_error(int32_t old_errno, int32_t new_errno,
+ gf_boolean_t eio)
{
- gf_boolean_t ret = _gf_true;
-
- /* Nothing should ever overwrite ESTALE */
- if (old_errno == ESTALE)
- ret = _gf_false;
-
- /* Nothing should overwrite ENOENT, except ESTALE/EIO*/
- else if ((old_errno == ENOENT) && (new_errno != ESTALE)
- && (new_errno != EIO))
- ret = _gf_false;
-
- return ret;
+ if (old_errno == ESTALE || new_errno == ESTALE)
+ return ESTALE;
+ if (eio && (old_errno == EIO || new_errno == EIO))
+ return EIO;
+ if (old_errno == ENOENT || new_errno == ENOENT)
+ return ENOENT;
+
+ return new_errno;
}
int32_t
@@ -1807,8 +2174,9 @@ afr_resultant_errno_get (int32_t *children,
} else {
child = i;
}
- if (afr_error_more_important (op_errno, child_errno[child]))
- op_errno = child_errno[child];
+ op_errno = afr_most_important_error(op_errno,
+ child_errno[child],
+ _gf_false);
}
return op_errno;
}
@@ -1820,8 +2188,8 @@ afr_lookup_handle_error (afr_local_t *local, int32_t op_ret, int32_t op_errno)
if (op_errno == ENOENT)
local->enoent_count++;
- if (afr_error_more_important (local->op_errno, op_errno))
- local->op_errno = op_errno;
+ local->op_errno = afr_most_important_error(local->op_errno, op_errno,
+ _gf_false);
if (local->op_errno == ESTALE) {
local->op_ret = -1;
@@ -1868,12 +2236,79 @@ afr_lookup_handle_first_success (afr_local_t *local, xlator_t *this,
afr_set_root_inode_on_first_lookup (local, this, inode);
}
+static int32_t
+afr_discovery_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
+ int32_t op_ret, int32_t op_errno, dict_t *dict,
+ dict_t *xdata)
+{
+ int ret = 0;
+ char *pathinfo = NULL;
+ gf_boolean_t is_local = _gf_false;
+ afr_private_t *priv = NULL;
+ int32_t child_index = -1;
+
+ if (op_ret != 0) {
+ goto out;
+ }
+
+ ret = dict_get_str (dict, GF_XATTR_PATHINFO_KEY, &pathinfo);
+ if (ret != 0) {
+ goto out;
+ }
+
+ ret = afr_local_pathinfo (pathinfo, &is_local);
+ if (ret) {
+ goto out;
+ }
+
+ priv = this->private;
+ /*
+ * Note that one local subvolume will override another here. The only
+ * way to avoid that would be to retain extra information about whether
+ * the previous read_child is local, and it's just not worth it. Even
+ * the slowest local subvolume is far preferable to a remote one.
+ */
+ if (is_local) {
+ child_index = (int32_t)(long)cookie;
+ gf_log (this->name, GF_LOG_INFO,
+ "selecting local read_child %s",
+ priv->children[child_index]->name);
+ priv->read_child = child_index;
+ }
+
+out:
+ STACK_DESTROY(frame->root);
+ return 0;
+}
+
+static void
+afr_attempt_local_discovery (xlator_t *this, int32_t child_index)
+{
+ call_frame_t *newframe = NULL;
+ loc_t tmploc = {0,};
+ afr_private_t *priv = this->private;
+
+ newframe = create_frame(this,this->ctx->pool);
+ if (!newframe) {
+ return;
+ }
+
+ tmploc.gfid[sizeof(tmploc.gfid)-1] = 1;
+ STACK_WIND_COOKIE (newframe, afr_discovery_cbk,
+ (void *)(long)child_index,
+ priv->children[child_index],
+ priv->children[child_index]->fops->getxattr,
+ &tmploc, GF_XATTR_PATHINFO_KEY, NULL);
+}
+
static void
afr_lookup_handle_success (afr_local_t *local, xlator_t *this, int32_t child_index,
int32_t op_ret, int32_t op_errno, inode_t *inode,
struct iatt *buf, dict_t *xattr,
struct iatt *postparent)
{
+ afr_private_t *priv = this->private;
+
if (local->success_count == 0) {
if (local->op_errno != ESTALE) {
local->op_ret = op_ret;
@@ -1886,6 +2321,11 @@ afr_lookup_handle_success (afr_local_t *local, xlator_t *this, int32_t child_ind
afr_lookup_cache_args (local, child_index, xattr,
buf, postparent);
+
+ if (local->do_discovery && (priv->read_child == (-1))) {
+ afr_attempt_local_discovery(this,child_index);
+ }
+
local->cont.lookup.success_children[local->success_count] = child_index;
local->success_count++;
}
@@ -1933,6 +2373,7 @@ afr_lookup_cont_init (afr_local_t *local, unsigned int child_count)
struct iatt *iatts = NULL;
int32_t *success_children = NULL;
int32_t *sources = NULL;
+ int32_t **pending_matrix = NULL;
GF_ASSERT (local);
local->cont.lookup.xattrs = GF_CALLOC (child_count,
@@ -1965,6 +2406,11 @@ afr_lookup_cont_init (afr_local_t *local, unsigned int child_count)
goto out;
local->cont.lookup.sources = sources;
+ pending_matrix = afr_matrix_create (child_count, child_count);
+ if (NULL == pending_matrix)
+ goto out;
+ local->cont.lookup.pending_matrix = pending_matrix;
+
ret = 0;
out:
return ret;
@@ -1982,37 +2428,51 @@ afr_lookup (call_frame_t *frame, xlator_t *this,
int call_count = 0;
uint64_t ctx = 0;
int32_t op_errno = 0;
-
+ int allow_sh = 0;
priv = this->private;
- ALLOC_OR_GOTO (local, afr_local_t, out);
+ AFR_LOCAL_ALLOC_OR_GOTO (local, out);
local->op_ret = -1;
frame->local = local;
local->fop = GF_FOP_LOOKUP;
- if (!strcmp (loc->path, "/" GF_REPLICATE_TRASH_DIR)) {
- op_errno = ENOENT;
+ loc_copy (&local->loc, loc);
+ ret = loc_path (&local->loc, NULL);
+ if (ret < 0) {
+ op_errno = EINVAL;
goto out;
}
- loc_copy (&local->loc, loc);
+ if (local->loc.path &&
+ (strcmp (local->loc.path, "/" GF_REPLICATE_TRASH_DIR) == 0)) {
+ op_errno = EPERM;
+ ret = -1;
+ goto out;
+ }
- ret = inode_ctx_get (loc->inode, this, &ctx);
+ ret = inode_ctx_get (local->loc.inode, this, &ctx);
if (ret == 0) {
/* lookup is a revalidate */
local->read_child_index = afr_inode_get_read_ctx (this,
- loc->inode,
- NULL);
+ local->loc.inode,
+ NULL);
} else {
LOCK (&priv->read_child_lock);
{
- local->read_child_index = (++priv->read_child_rr)
- % (priv->child_count);
+ if (priv->hash_mode) {
+ local->read_child_index = -1;
+ }
+ else {
+ local->read_child_index =
+ (++priv->read_child_rr) %
+ (priv->child_count);
+ }
}
UNLOCK (&priv->read_child_lock);
+ local->cont.lookup.fresh_lookup = _gf_true;
}
local->child_up = memdup (priv->child_up,
@@ -2040,24 +2500,33 @@ afr_lookup (call_frame_t *frame, xlator_t *this,
/* By default assume ENOTCONN. On success it will be set to 0. */
local->op_errno = ENOTCONN;
- local->call_count = afr_up_children_count (local->child_up,
- priv->child_count);
- ret = afr_lookup_xattr_req_prepare (local, this, xattr_req, loc,
+ ret = dict_get_int32 (xattr_req, "allow-sh-for-running-transaction",
+ &allow_sh);
+ dict_del (xattr_req, "allow-sh-for-running-transaction");
+ local->allow_sh_for_running_transaction = allow_sh;
+
+ ret = afr_lookup_xattr_req_prepare (local, this, xattr_req, &local->loc,
&gfid_req);
if (ret) {
local->op_errno = -ret;
goto out;
}
afr_lookup_save_gfid (local->cont.lookup.gfid_req, gfid_req,
- loc);
+ &local->loc);
local->fop = GF_FOP_LOOKUP;
+ if (priv->choose_local && !priv->did_discovery) {
+ if (gfid_req && __is_root_gfid(gfid_req)) {
+ local->do_discovery = _gf_true;
+ priv->did_discovery = _gf_true;
+ }
+ }
for (i = 0; i < priv->child_count; i++) {
if (local->child_up[i]) {
STACK_WIND_COOKIE (frame, afr_lookup_cbk,
(void *) (long) i,
priv->children[i],
priv->children[i]->fops->lookup,
- loc, local->xattr_req);
+ &local->loc, local->xattr_req);
if (!--call_count)
break;
}
@@ -2151,8 +2620,11 @@ __afr_fd_ctx_set (xlator_t *this, fd_t *fd)
goto out;
}
- INIT_LIST_HEAD (&fd_ctx->paused_calls);
+ pthread_mutex_init (&fd_ctx->delay_lock, NULL);
INIT_LIST_HEAD (&fd_ctx->entries);
+ fd_ctx->call_child = -1;
+
+ INIT_LIST_HEAD (&fd_ctx->eager_locked);
ret = __fd_ctx_set (fd, this, (uint64_t)(long) fd_ctx);
if (ret)
@@ -2180,134 +2652,70 @@ afr_fd_ctx_set (xlator_t *this, fd_t *fd)
/* {{{ flush */
int
-afr_flush_unwind (call_frame_t *frame, xlator_t *this)
+afr_flush_cbk(call_frame_t *frame, void *cookie, xlator_t *this,
+ int32_t op_ret, int32_t op_errno, dict_t *xdata)
{
afr_local_t * local = NULL;
- call_frame_t *main_frame = NULL;
-
- local = frame->local;
-
- LOCK (&frame->lock);
- {
- if (local->transaction.main_frame)
- main_frame = local->transaction.main_frame;
- local->transaction.main_frame = NULL;
- }
- UNLOCK (&frame->lock);
-
- if (main_frame) {
- AFR_STACK_UNWIND (flush, main_frame,
- local->op_ret, local->op_errno);
- }
-
- return 0;
-}
-
-
-int
-afr_flush_wind_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
- int32_t op_ret, int32_t op_errno)
-{
- afr_local_t * local = NULL;
- afr_private_t * priv = NULL;
int call_count = -1;
- int child_index = (long) cookie;
- int need_unwind = 0;
local = frame->local;
- priv = this->private;
LOCK (&frame->lock);
{
- if (afr_fop_failed (op_ret, op_errno))
- afr_transaction_fop_failed (frame, this, child_index);
-
if (op_ret != -1) {
if (local->success_count == 0) {
local->op_ret = op_ret;
}
local->success_count++;
-
- if (local->success_count == priv->wait_count) {
- need_unwind = 1;
- }
}
local->op_errno = op_errno;
}
UNLOCK (&frame->lock);
- if (need_unwind)
- afr_flush_unwind (frame, this);
+ call_count = afr_frame_return (frame);
- call_count = afr_frame_return (frame);
-
- if (call_count == 0) {
- local->transaction.resume (frame, this);
- }
+ if (call_count == 0)
+ AFR_STACK_UNWIND(flush, frame, local->op_ret,
+ local->op_errno, NULL);
return 0;
}
-
-int
-afr_flush_wind (call_frame_t *frame, xlator_t *this)
+static int
+afr_flush_wrapper (call_frame_t *frame, xlator_t *this, fd_t *fd, dict_t *xdata)
{
- afr_local_t *local = NULL;
- afr_private_t *priv = NULL;
- int i = 0;
- int call_count = -1;
+ int i = 0;
+ afr_local_t *local = NULL;
+ afr_private_t *priv = NULL;
+ int call_count = -1;
- local = frame->local;
priv = this->private;
-
- call_count = afr_up_children_count (local->child_up, priv->child_count);
-
- if (call_count == 0) {
- local->transaction.resume (frame, this);
- return 0;
- }
-
- local->call_count = call_count;
+ local = frame->local;
+ call_count = local->call_count;
for (i = 0; i < priv->child_count; i++) {
if (local->child_up[i]) {
- STACK_WIND_COOKIE (frame, afr_flush_wind_cbk,
+ STACK_WIND_COOKIE (frame, afr_flush_cbk,
(void *) (long) i,
priv->children[i],
priv->children[i]->fops->flush,
- local->fd);
-
+ local->fd, NULL);
if (!--call_count)
break;
+
}
}
return 0;
}
-
-int
-afr_flush_done (call_frame_t *frame, xlator_t *this)
-{
- afr_local_t *local = NULL;
-
- local = frame->local;
-
- local->transaction.unwind (frame, this);
-
- AFR_STACK_DESTROY (frame);
-
- return 0;
-}
-
-
int
-afr_flush (call_frame_t *frame, xlator_t *this, fd_t *fd)
+afr_flush (call_frame_t *frame, xlator_t *this, fd_t *fd, dict_t *xdata)
{
afr_private_t *priv = NULL;
afr_local_t *local = NULL;
- call_frame_t *transaction_frame = NULL;
+ call_stub_t *stub = NULL;
int ret = -1;
int op_errno = 0;
@@ -2317,47 +2725,27 @@ afr_flush (call_frame_t *frame, xlator_t *this, fd_t *fd)
priv = this->private;
- transaction_frame = copy_frame (frame);
- if (!transaction_frame) {
- op_errno = ENOMEM;
- goto out;
- }
-
- ALLOC_OR_GOTO (transaction_frame->local, afr_local_t, out);
- local = transaction_frame->local;
-
- ret = afr_local_init (local, priv, &op_errno);
- if (ret < 0)
- goto out;
-
- local->op = GF_FOP_FLUSH;
-
- local->transaction.fop = afr_flush_wind;
- local->transaction.done = afr_flush_done;
- local->transaction.unwind = afr_flush_unwind;
+ AFR_LOCAL_ALLOC_OR_GOTO (frame->local, out);
+ local = frame->local;
- local->fd = fd_ref (fd);
+ ret = afr_local_init(local, priv, &op_errno);
+ if (ret < 0)
+ goto out;
- local->transaction.main_frame = frame;
- local->transaction.start = 0;
- local->transaction.len = 0;
-
- ret = afr_open_fd_fix (transaction_frame, this, _gf_false);
- if (ret) {
- op_errno = -ret;
+ local->fd = fd_ref(fd);
+ stub = fop_flush_stub (frame, afr_flush_wrapper, fd, xdata);
+ if (!stub) {
+ ret = -1;
+ op_errno = ENOMEM;
goto out;
}
- afr_transaction (transaction_frame, this, AFR_DATA_TRANSACTION);
+ afr_delayed_changelog_wake_resume (this, fd, stub);
+ ret = 0;
- ret = 0;
out:
- if (ret < 0) {
- if (transaction_frame)
- AFR_STACK_DESTROY (transaction_frame);
-
- AFR_STACK_UNWIND (flush, frame, -1, op_errno);
- }
+ if (ret < 0)
+ AFR_STACK_UNWIND(flush, frame, -1, op_errno, NULL);
return 0;
}
@@ -2371,8 +2759,6 @@ afr_cleanup_fd_ctx (xlator_t *this, fd_t *fd)
uint64_t ctx = 0;
afr_fd_ctx_t *fd_ctx = NULL;
int ret = 0;
- afr_fd_paused_call_t *paused_call = NULL;
- afr_fd_paused_call_t *tmp = NULL;
ret = fd_ctx_get (fd, this, &ctx);
if (ret < 0)
@@ -2381,28 +2767,18 @@ afr_cleanup_fd_ctx (xlator_t *this, fd_t *fd)
fd_ctx = (afr_fd_ctx_t *)(long) ctx;
if (fd_ctx) {
- if (fd_ctx->pre_op_done)
- GF_FREE (fd_ctx->pre_op_done);
+ GF_FREE (fd_ctx->pre_op_done);
- if (fd_ctx->opened_on)
- GF_FREE (fd_ctx->opened_on);
+ GF_FREE (fd_ctx->opened_on);
- if (fd_ctx->locked_on)
- GF_FREE (fd_ctx->locked_on);
+ GF_FREE (fd_ctx->locked_on);
- if (fd_ctx->pre_op_piggyback)
- GF_FREE (fd_ctx->pre_op_piggyback);
- list_for_each_entry_safe (paused_call, tmp, &fd_ctx->paused_calls,
- call_list) {
- list_del_init (&paused_call->call_list);
- GF_FREE (paused_call);
- }
+ GF_FREE (fd_ctx->pre_op_piggyback);
+ GF_FREE (fd_ctx->lock_piggyback);
- if (fd_ctx->lock_piggyback)
- GF_FREE (fd_ctx->lock_piggyback);
+ GF_FREE (fd_ctx->lock_acquired);
- if (fd_ctx->lock_acquired)
- GF_FREE (fd_ctx->lock_acquired);
+ pthread_mutex_destroy (&fd_ctx->delay_lock);
GF_FREE (fd_ctx);
}
@@ -2440,14 +2816,25 @@ afr_release (xlator_t *this, fd_t *fd)
/* {{{ fsync */
int
+afr_fsync_unwind_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
+ int32_t op_ret, int32_t op_errno, struct iatt *prebuf,
+ struct iatt *postbuf, dict_t *xdata)
+{
+ AFR_STACK_UNWIND (fsync, frame, op_ret, op_errno, prebuf, postbuf,
+ xdata);
+ return 0;
+}
+
+int
afr_fsync_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
int32_t op_ret, int32_t op_errno, struct iatt *prebuf,
- struct iatt *postbuf)
+ struct iatt *postbuf, dict_t *xdata)
{
afr_local_t *local = NULL;
int call_count = -1;
int child_index = (long) cookie;
int read_child = 0;
+ call_stub_t *stub = NULL;
local = frame->local;
@@ -2463,13 +2850,13 @@ afr_fsync_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
local->op_ret = 0;
if (local->success_count == 0) {
- local->cont.fsync.prebuf = *prebuf;
- local->cont.fsync.postbuf = *postbuf;
+ local->cont.inode_wfop.prebuf = *prebuf;
+ local->cont.inode_wfop.postbuf = *postbuf;
}
if (child_index == read_child) {
- local->cont.fsync.prebuf = *prebuf;
- local->cont.fsync.postbuf = *postbuf;
+ local->cont.inode_wfop.prebuf = *prebuf;
+ local->cont.inode_wfop.postbuf = *postbuf;
}
local->success_count++;
@@ -2482,9 +2869,32 @@ afr_fsync_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
call_count = afr_frame_return (frame);
if (call_count == 0) {
- AFR_STACK_UNWIND (fsync, frame, local->op_ret, local->op_errno,
- &local->cont.fsync.prebuf,
- &local->cont.fsync.postbuf);
+ /* Make a stub out of the frame, and register it
+ with the waking up post-op. When the call-stub resumes,
+ we are guaranteed that there was no post-op pending
+ (i.e changelogs were unset in the server). This is an
+ essential "guarantee", that fsync() returns only after
+ completely finishing EVERYTHING, including the delayed
+ post-op. This guarantee is expected by FUSE graph switching
+ for example.
+ */
+ stub = fop_fsync_cbk_stub (frame, afr_fsync_unwind_cbk,
+ local->op_ret, local->op_errno,
+ &local->cont.inode_wfop.prebuf,
+ &local->cont.inode_wfop.postbuf,
+ xdata);
+ if (!stub) {
+ AFR_STACK_UNWIND (fsync, frame, -1, ENOMEM, 0, 0, 0);
+ return 0;
+ }
+
+ /* If no new unstable writes happened between the
+ time we cleared the unstable write witness flag in afr_fsync
+ and now, calling afr_delayed_changelog_wake_up() should
+ wake up and skip over the fsync phase and go straight to
+ afr_changelog_post_op_now()
+ */
+ afr_delayed_changelog_wake_resume (this, local->fd, stub);
}
return 0;
@@ -2493,7 +2903,7 @@ afr_fsync_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
int
afr_fsync (call_frame_t *frame, xlator_t *this, fd_t *fd,
- int32_t datasync)
+ int32_t datasync, dict_t *xdata)
{
afr_private_t *priv = NULL;
afr_local_t *local = NULL;
@@ -2508,7 +2918,7 @@ afr_fsync (call_frame_t *frame, xlator_t *this, fd_t *fd,
priv = this->private;
- ALLOC_OR_GOTO (frame->local, afr_local_t, out);
+ AFR_LOCAL_ALLOC_OR_GOTO (frame->local, out);
local = frame->local;
ret = afr_local_init (local, priv, &op_errno);
@@ -2519,13 +2929,17 @@ afr_fsync (call_frame_t *frame, xlator_t *this, fd_t *fd,
local->fd = fd_ref (fd);
+ if (afr_fd_has_witnessed_unstable_write (this, fd)) {
+ /* don't care. we only wanted to CLEAR the bit */
+ }
+
for (i = 0; i < priv->child_count; i++) {
if (local->child_up[i]) {
STACK_WIND_COOKIE (frame, afr_fsync_cbk,
(void *) (long) i,
priv->children[i],
priv->children[i]->fops->fsync,
- fd, datasync);
+ fd, datasync, xdata);
if (!--call_count)
break;
}
@@ -2534,7 +2948,7 @@ afr_fsync (call_frame_t *frame, xlator_t *this, fd_t *fd,
ret = 0;
out:
if (ret < 0)
- AFR_STACK_UNWIND (fsync, frame, -1, op_errno, NULL, NULL);
+ AFR_STACK_UNWIND (fsync, frame, -1, op_errno, NULL, NULL, NULL);
return 0;
}
@@ -2544,7 +2958,8 @@ out:
int32_t
afr_fsyncdir_cbk (call_frame_t *frame, void *cookie,
- xlator_t *this, int32_t op_ret, int32_t op_errno)
+ xlator_t *this, int32_t op_ret, int32_t op_errno,
+ dict_t *xdata)
{
afr_local_t *local = NULL;
int call_count = -1;
@@ -2564,7 +2979,7 @@ afr_fsyncdir_cbk (call_frame_t *frame, void *cookie,
if (call_count == 0)
AFR_STACK_UNWIND (fsyncdir, frame, local->op_ret,
- local->op_errno);
+ local->op_errno, xdata);
return 0;
}
@@ -2572,7 +2987,7 @@ afr_fsyncdir_cbk (call_frame_t *frame, void *cookie,
int32_t
afr_fsyncdir (call_frame_t *frame, xlator_t *this, fd_t *fd,
- int32_t datasync)
+ int32_t datasync, dict_t *xdata)
{
afr_private_t *priv = NULL;
afr_local_t *local = NULL;
@@ -2587,7 +3002,7 @@ afr_fsyncdir (call_frame_t *frame, xlator_t *this, fd_t *fd,
priv = this->private;
- ALLOC_OR_GOTO (frame->local, afr_local_t, out);
+ AFR_LOCAL_ALLOC_OR_GOTO (frame->local, out);
local = frame->local;
ret = afr_local_init (local, priv, &op_errno);
@@ -2601,7 +3016,7 @@ afr_fsyncdir (call_frame_t *frame, xlator_t *this, fd_t *fd,
STACK_WIND (frame, afr_fsyncdir_cbk,
priv->children[i],
priv->children[i]->fops->fsyncdir,
- fd, datasync);
+ fd, datasync, xdata);
if (!--call_count)
break;
}
@@ -2610,7 +3025,7 @@ afr_fsyncdir (call_frame_t *frame, xlator_t *this, fd_t *fd,
ret = 0;
out:
if (ret < 0)
- AFR_STACK_UNWIND (fsyncdir, frame, -1, op_errno);
+ AFR_STACK_UNWIND (fsyncdir, frame, -1, op_errno, NULL);
return 0;
}
@@ -2621,7 +3036,7 @@ out:
int32_t
afr_xattrop_cbk (call_frame_t *frame, void *cookie,
xlator_t *this, int32_t op_ret, int32_t op_errno,
- dict_t *xattr)
+ dict_t *xattr, dict_t *xdata)
{
afr_local_t *local = NULL;
int call_count = -1;
@@ -2630,8 +3045,11 @@ afr_xattrop_cbk (call_frame_t *frame, void *cookie,
LOCK (&frame->lock);
{
- if (op_ret == 0)
+ if (op_ret == 0) {
+ if (!local->cont.xattrop.xattr)
+ local->cont.xattrop.xattr = dict_ref (xattr);
local->op_ret = 0;
+ }
local->op_errno = op_errno;
}
@@ -2641,7 +3059,7 @@ afr_xattrop_cbk (call_frame_t *frame, void *cookie,
if (call_count == 0)
AFR_STACK_UNWIND (xattrop, frame, local->op_ret, local->op_errno,
- xattr);
+ local->cont.xattrop.xattr, xdata);
return 0;
}
@@ -2649,7 +3067,7 @@ afr_xattrop_cbk (call_frame_t *frame, void *cookie,
int32_t
afr_xattrop (call_frame_t *frame, xlator_t *this, loc_t *loc,
- gf_xattrop_flags_t optype, dict_t *xattr)
+ gf_xattrop_flags_t optype, dict_t *xattr, dict_t *xdata)
{
afr_private_t *priv = NULL;
afr_local_t *local = NULL;
@@ -2664,7 +3082,7 @@ afr_xattrop (call_frame_t *frame, xlator_t *this, loc_t *loc,
priv = this->private;
- ALLOC_OR_GOTO (frame->local, afr_local_t, out);
+ AFR_LOCAL_ALLOC_OR_GOTO (frame->local, out);
local = frame->local;
ret = afr_local_init (local, priv, &op_errno);
@@ -2678,7 +3096,7 @@ afr_xattrop (call_frame_t *frame, xlator_t *this, loc_t *loc,
STACK_WIND (frame, afr_xattrop_cbk,
priv->children[i],
priv->children[i]->fops->xattrop,
- loc, optype, xattr);
+ loc, optype, xattr, xdata);
if (!--call_count)
break;
}
@@ -2687,7 +3105,7 @@ afr_xattrop (call_frame_t *frame, xlator_t *this, loc_t *loc,
ret = 0;
out:
if (ret < 0)
- AFR_STACK_UNWIND (xattrop, frame, -1, op_errno, NULL);
+ AFR_STACK_UNWIND (xattrop, frame, -1, op_errno, NULL, NULL);
return 0;
}
@@ -2698,7 +3116,7 @@ out:
int32_t
afr_fxattrop_cbk (call_frame_t *frame, void *cookie,
xlator_t *this, int32_t op_ret, int32_t op_errno,
- dict_t *xattr)
+ dict_t *xattr, dict_t *xdata)
{
afr_local_t *local = NULL;
@@ -2708,8 +3126,12 @@ afr_fxattrop_cbk (call_frame_t *frame, void *cookie,
LOCK (&frame->lock);
{
- if (op_ret == 0)
+ if (op_ret == 0) {
+ if (!local->cont.fxattrop.xattr)
+ local->cont.fxattrop.xattr = dict_ref (xattr);
+
local->op_ret = 0;
+ }
local->op_errno = op_errno;
}
@@ -2719,7 +3141,7 @@ afr_fxattrop_cbk (call_frame_t *frame, void *cookie,
if (call_count == 0)
AFR_STACK_UNWIND (fxattrop, frame, local->op_ret, local->op_errno,
- xattr);
+ local->cont.fxattrop.xattr, xdata);
return 0;
}
@@ -2727,7 +3149,7 @@ afr_fxattrop_cbk (call_frame_t *frame, void *cookie,
int32_t
afr_fxattrop (call_frame_t *frame, xlator_t *this, fd_t *fd,
- gf_xattrop_flags_t optype, dict_t *xattr)
+ gf_xattrop_flags_t optype, dict_t *xattr, dict_t *xdata)
{
afr_private_t *priv = NULL;
afr_local_t *local = NULL;
@@ -2742,7 +3164,7 @@ afr_fxattrop (call_frame_t *frame, xlator_t *this, fd_t *fd,
priv = this->private;
- ALLOC_OR_GOTO (frame->local, afr_local_t, out);
+ AFR_LOCAL_ALLOC_OR_GOTO (frame->local, out);
local = frame->local;
ret = afr_local_init (local, priv, &op_errno);
@@ -2756,7 +3178,7 @@ afr_fxattrop (call_frame_t *frame, xlator_t *this, fd_t *fd,
STACK_WIND (frame, afr_fxattrop_cbk,
priv->children[i],
priv->children[i]->fops->fxattrop,
- fd, optype, xattr);
+ fd, optype, xattr, xdata);
if (!--call_count)
break;
}
@@ -2765,7 +3187,7 @@ afr_fxattrop (call_frame_t *frame, xlator_t *this, fd_t *fd,
ret = 0;
out:
if (ret < 0)
- AFR_STACK_UNWIND (fxattrop, frame, -1, op_errno, NULL);
+ AFR_STACK_UNWIND (fxattrop, frame, -1, op_errno, NULL, NULL);
return 0;
}
@@ -2774,7 +3196,7 @@ out:
int32_t
afr_inodelk_cbk (call_frame_t *frame, void *cookie,
- xlator_t *this, int32_t op_ret, int32_t op_errno)
+ xlator_t *this, int32_t op_ret, int32_t op_errno, dict_t *xdata)
{
afr_local_t *local = NULL;
@@ -2795,7 +3217,7 @@ afr_inodelk_cbk (call_frame_t *frame, void *cookie,
if (call_count == 0)
AFR_STACK_UNWIND (inodelk, frame, local->op_ret,
- local->op_errno);
+ local->op_errno, xdata);
return 0;
}
@@ -2803,7 +3225,8 @@ afr_inodelk_cbk (call_frame_t *frame, void *cookie,
int32_t
afr_inodelk (call_frame_t *frame, xlator_t *this,
- const char *volume, loc_t *loc, int32_t cmd, struct gf_flock *flock)
+ const char *volume, loc_t *loc, int32_t cmd,
+ struct gf_flock *flock, dict_t *xdata)
{
afr_private_t *priv = NULL;
afr_local_t *local = NULL;
@@ -2818,7 +3241,7 @@ afr_inodelk (call_frame_t *frame, xlator_t *this,
priv = this->private;
- ALLOC_OR_GOTO (frame->local, afr_local_t, out);
+ AFR_LOCAL_ALLOC_OR_GOTO (frame->local, out);
local = frame->local;
ret = afr_local_init (local, priv, &op_errno);
@@ -2832,7 +3255,7 @@ afr_inodelk (call_frame_t *frame, xlator_t *this,
STACK_WIND (frame, afr_inodelk_cbk,
priv->children[i],
priv->children[i]->fops->inodelk,
- volume, loc, cmd, flock);
+ volume, loc, cmd, flock, xdata);
if (!--call_count)
break;
@@ -2842,14 +3265,15 @@ afr_inodelk (call_frame_t *frame, xlator_t *this,
ret = 0;
out:
if (ret < 0)
- AFR_STACK_UNWIND (inodelk, frame, -1, op_errno);
+ AFR_STACK_UNWIND (inodelk, frame, -1, op_errno, NULL);
return 0;
}
int32_t
afr_finodelk_cbk (call_frame_t *frame, void *cookie,
- xlator_t *this, int32_t op_ret, int32_t op_errno)
+ xlator_t *this, int32_t op_ret, int32_t op_errno,
+ dict_t *xdata)
{
afr_local_t *local = NULL;
@@ -2870,7 +3294,7 @@ afr_finodelk_cbk (call_frame_t *frame, void *cookie,
if (call_count == 0)
AFR_STACK_UNWIND (finodelk, frame, local->op_ret,
- local->op_errno);
+ local->op_errno, xdata);
return 0;
}
@@ -2878,7 +3302,8 @@ afr_finodelk_cbk (call_frame_t *frame, void *cookie,
int32_t
afr_finodelk (call_frame_t *frame, xlator_t *this,
- const char *volume, fd_t *fd, int32_t cmd, struct gf_flock *flock)
+ const char *volume, fd_t *fd, int32_t cmd, struct gf_flock *flock,
+ dict_t *xdata)
{
afr_private_t *priv = NULL;
afr_local_t *local = NULL;
@@ -2893,7 +3318,7 @@ afr_finodelk (call_frame_t *frame, xlator_t *this,
priv = this->private;
- ALLOC_OR_GOTO (frame->local, afr_local_t, out);
+ AFR_LOCAL_ALLOC_OR_GOTO (frame->local, out);
local = frame->local;
ret = afr_local_init (local, priv, &op_errno);
@@ -2907,7 +3332,7 @@ afr_finodelk (call_frame_t *frame, xlator_t *this,
STACK_WIND (frame, afr_finodelk_cbk,
priv->children[i],
priv->children[i]->fops->finodelk,
- volume, fd, cmd, flock);
+ volume, fd, cmd, flock, xdata);
if (!--call_count)
break;
@@ -2917,15 +3342,14 @@ afr_finodelk (call_frame_t *frame, xlator_t *this,
ret = 0;
out:
if (ret < 0)
- AFR_STACK_UNWIND (finodelk, frame, -1, op_errno);
+ AFR_STACK_UNWIND (finodelk, frame, -1, op_errno, NULL);
return 0;
}
int32_t
-afr_entrylk_cbk (call_frame_t *frame, void *cookie,
- xlator_t *this, int32_t op_ret, int32_t op_errno)
-
+afr_entrylk_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
+ int32_t op_ret, int32_t op_errno, dict_t *xdata)
{
afr_local_t *local = NULL;
int call_count = -1;
@@ -2945,7 +3369,7 @@ afr_entrylk_cbk (call_frame_t *frame, void *cookie,
if (call_count == 0)
AFR_STACK_UNWIND (entrylk, frame, local->op_ret,
- local->op_errno);
+ local->op_errno, xdata);
return 0;
}
@@ -2954,7 +3378,8 @@ afr_entrylk_cbk (call_frame_t *frame, void *cookie,
int32_t
afr_entrylk (call_frame_t *frame, xlator_t *this,
const char *volume, loc_t *loc,
- const char *basename, entrylk_cmd cmd, entrylk_type type)
+ const char *basename, entrylk_cmd cmd, entrylk_type type,
+ dict_t *xdata)
{
afr_private_t *priv = NULL;
afr_local_t *local = NULL;
@@ -2969,7 +3394,7 @@ afr_entrylk (call_frame_t *frame, xlator_t *this,
priv = this->private;
- ALLOC_OR_GOTO (frame->local, afr_local_t, out);
+ AFR_LOCAL_ALLOC_OR_GOTO (frame->local, out);
local = frame->local;
ret = afr_local_init (local, priv, &op_errno);
@@ -2983,7 +3408,7 @@ afr_entrylk (call_frame_t *frame, xlator_t *this,
STACK_WIND (frame, afr_entrylk_cbk,
priv->children[i],
priv->children[i]->fops->entrylk,
- volume, loc, basename, cmd, type);
+ volume, loc, basename, cmd, type, xdata);
if (!--call_count)
break;
@@ -2993,7 +3418,7 @@ afr_entrylk (call_frame_t *frame, xlator_t *this,
ret = 0;
out:
if (ret < 0)
- AFR_STACK_UNWIND (entrylk, frame, -1, op_errno);
+ AFR_STACK_UNWIND (entrylk, frame, -1, op_errno, NULL);
return 0;
}
@@ -3001,7 +3426,7 @@ out:
int32_t
afr_fentrylk_cbk (call_frame_t *frame, void *cookie,
- xlator_t *this, int32_t op_ret, int32_t op_errno)
+ xlator_t *this, int32_t op_ret, int32_t op_errno, dict_t *xdata)
{
afr_local_t *local = NULL;
@@ -3022,7 +3447,7 @@ afr_fentrylk_cbk (call_frame_t *frame, void *cookie,
if (call_count == 0)
AFR_STACK_UNWIND (fentrylk, frame, local->op_ret,
- local->op_errno);
+ local->op_errno, xdata);
return 0;
}
@@ -3031,7 +3456,8 @@ afr_fentrylk_cbk (call_frame_t *frame, void *cookie,
int32_t
afr_fentrylk (call_frame_t *frame, xlator_t *this,
const char *volume, fd_t *fd,
- const char *basename, entrylk_cmd cmd, entrylk_type type)
+ const char *basename, entrylk_cmd cmd,
+ entrylk_type type, dict_t *xdata)
{
afr_private_t *priv = NULL;
afr_local_t *local = NULL;
@@ -3046,7 +3472,7 @@ afr_fentrylk (call_frame_t *frame, xlator_t *this,
priv = this->private;
- ALLOC_OR_GOTO (frame->local, afr_local_t, out);
+ AFR_LOCAL_ALLOC_OR_GOTO (frame->local, out);
local = frame->local;
ret = afr_local_init (local, priv, &op_errno);
@@ -3060,7 +3486,7 @@ afr_fentrylk (call_frame_t *frame, xlator_t *this,
STACK_WIND (frame, afr_fentrylk_cbk,
priv->children[i],
priv->children[i]->fops->fentrylk,
- volume, fd, basename, cmd, type);
+ volume, fd, basename, cmd, type, xdata);
if (!--call_count)
break;
@@ -3070,14 +3496,14 @@ afr_fentrylk (call_frame_t *frame, xlator_t *this,
ret = 0;
out:
if (ret < 0)
- AFR_STACK_UNWIND (fentrylk, frame, -1, op_errno);
+ AFR_STACK_UNWIND (fentrylk, frame, -1, op_errno, NULL);
return 0;
}
int32_t
afr_statfs_cbk (call_frame_t *frame, void *cookie,
xlator_t *this, int32_t op_ret, int32_t op_errno,
- struct statvfs *statvfs)
+ struct statvfs *statvfs, dict_t *xdata)
{
afr_local_t *local = NULL;
int call_count = 0;
@@ -3108,7 +3534,7 @@ afr_statfs_cbk (call_frame_t *frame, void *cookie,
if (call_count == 0)
AFR_STACK_UNWIND (statfs, frame, local->op_ret, local->op_errno,
- &local->cont.statfs.buf);
+ &local->cont.statfs.buf, xdata);
return 0;
}
@@ -3116,7 +3542,7 @@ afr_statfs_cbk (call_frame_t *frame, void *cookie,
int32_t
afr_statfs (call_frame_t *frame, xlator_t *this,
- loc_t *loc)
+ loc_t *loc, dict_t *xdata)
{
afr_private_t * priv = NULL;
int child_count = 0;
@@ -3133,7 +3559,7 @@ afr_statfs (call_frame_t *frame, xlator_t *this,
priv = this->private;
child_count = priv->child_count;
- ALLOC_OR_GOTO (frame->local, afr_local_t, out);
+ AFR_LOCAL_ALLOC_OR_GOTO (frame->local, out);
local = frame->local;
ret = afr_local_init (local, priv, &op_errno);
@@ -3147,7 +3573,7 @@ afr_statfs (call_frame_t *frame, xlator_t *this,
STACK_WIND (frame, afr_statfs_cbk,
priv->children[i],
priv->children[i]->fops->statfs,
- loc);
+ loc, xdata);
if (!--call_count)
break;
}
@@ -3156,14 +3582,15 @@ afr_statfs (call_frame_t *frame, xlator_t *this,
ret = 0;
out:
if (ret < 0)
- AFR_STACK_UNWIND (statfs, frame, -1, op_errno, NULL);
+ AFR_STACK_UNWIND (statfs, frame, -1, op_errno, NULL, NULL);
return 0;
}
int32_t
afr_lk_unlock_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
- int32_t op_ret, int32_t op_errno, struct gf_flock *lock)
+ int32_t op_ret, int32_t op_errno, struct gf_flock *lock,
+ dict_t *xdata)
{
afr_local_t * local = NULL;
int call_count = -1;
@@ -3173,7 +3600,7 @@ afr_lk_unlock_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
if (call_count == 0)
AFR_STACK_UNWIND (lk, frame, local->op_ret, local->op_errno,
- lock);
+ lock, xdata);
return 0;
}
@@ -3195,7 +3622,7 @@ afr_lk_unlock (call_frame_t *frame, xlator_t *this)
if (call_count == 0) {
AFR_STACK_UNWIND (lk, frame, local->op_ret, local->op_errno,
- &local->cont.lk.ret_flock);
+ &local->cont.lk.ret_flock, NULL);
return 0;
}
@@ -3209,7 +3636,7 @@ afr_lk_unlock (call_frame_t *frame, xlator_t *this)
priv->children[i],
priv->children[i]->fops->lk,
local->fd, F_SETLK,
- &local->cont.lk.user_flock);
+ &local->cont.lk.user_flock, NULL);
if (!--call_count)
break;
@@ -3222,7 +3649,7 @@ afr_lk_unlock (call_frame_t *frame, xlator_t *this)
int32_t
afr_lk_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
- int32_t op_ret, int32_t op_errno, struct gf_flock *lock)
+ int32_t op_ret, int32_t op_errno, struct gf_flock *lock, dict_t *xdata)
{
afr_local_t *local = NULL;
afr_private_t *priv = NULL;
@@ -3257,12 +3684,12 @@ afr_lk_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
priv->children[child_index],
priv->children[child_index]->fops->lk,
local->fd, local->cont.lk.cmd,
- &local->cont.lk.user_flock);
+ &local->cont.lk.user_flock, xdata);
} else if (local->op_ret == -1) {
/* all nodes have gone down */
AFR_STACK_UNWIND (lk, frame, -1, ENOTCONN,
- &local->cont.lk.ret_flock);
+ &local->cont.lk.ret_flock, NULL);
} else {
/* locking has succeeded on all nodes that are up */
@@ -3280,7 +3707,7 @@ afr_lk_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
*/
AFR_STACK_UNWIND (lk, frame, local->op_ret, local->op_errno,
- &local->cont.lk.ret_flock);
+ &local->cont.lk.ret_flock, NULL);
}
return 0;
@@ -3289,7 +3716,7 @@ afr_lk_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
int
afr_lk (call_frame_t *frame, xlator_t *this,
- fd_t *fd, int32_t cmd, struct gf_flock *flock)
+ fd_t *fd, int32_t cmd, struct gf_flock *flock, dict_t *xdata)
{
afr_private_t *priv = NULL;
afr_local_t *local = NULL;
@@ -3303,7 +3730,7 @@ afr_lk (call_frame_t *frame, xlator_t *this,
priv = this->private;
- ALLOC_OR_GOTO (frame->local, afr_local_t, out);
+ AFR_LOCAL_ALLOC_OR_GOTO (frame->local, out);
local = frame->local;
ret = afr_local_init (local, priv, &op_errno);
@@ -3327,12 +3754,12 @@ afr_lk (call_frame_t *frame, xlator_t *this,
STACK_WIND_COOKIE (frame, afr_lk_cbk, (void *) (long) 0,
priv->children[i],
priv->children[i]->fops->lk,
- fd, cmd, flock);
+ fd, cmd, flock, xdata);
ret = 0;
out:
if (ret < 0)
- AFR_STACK_UNWIND (lk, frame, -1, op_errno, NULL);
+ AFR_STACK_UNWIND (lk, frame, -1, op_errno, NULL, NULL);
return 0;
}
@@ -3348,8 +3775,7 @@ afr_forget (xlator_t *this, inode_t *inode)
goto out;
ctx = (afr_inode_ctx_t *)(long)ctx_addr;
- if (ctx->fresh_children)
- GF_FREE (ctx->fresh_children);
+ GF_FREE (ctx->fresh_children);
GF_FREE (ctx);
out:
return 0;
@@ -3416,7 +3842,7 @@ find_child_index (xlator_t *this, xlator_t *child)
int32_t
afr_notify (xlator_t *this, int32_t event,
- void *data, ...)
+ void *data, void *data2)
{
afr_private_t *priv = NULL;
int i = -1;
@@ -3429,12 +3855,22 @@ afr_notify (xlator_t *this, int32_t event,
int ret = -1;
int call_psh = 0;
int up_child = AFR_ALL_CHILDREN;
+ dict_t *input = NULL;
+ dict_t *output = NULL;
priv = this->private;
if (!priv)
return 0;
+ /*
+ * We need to reset this in case children come up in "staggered"
+ * fashion, so that we discover a late-arriving local subvolume. Note
+ * that we could end up issuing N lookups to the first subvolume, and
+ * O(N^2) overall, but N is small for AFR so it shouldn't be an issue.
+ */
+ priv->did_discovery = _gf_false;
+
had_heard_from_all = 1;
for (i = 0; i < priv->child_count; i++) {
if (!priv->last_event[i]) {
@@ -3462,9 +3898,19 @@ afr_notify (xlator_t *this, int32_t event,
case GF_EVENT_CHILD_UP:
LOCK (&priv->lock);
{
+ /*
+ * This only really counts if the child was never up
+ * (value = -1) or had been down (value = 0). See
+ * comment at GF_EVENT_CHILD_DOWN for a more detailed
+ * explanation.
+ */
+ if (priv->child_up[idx] != 1) {
+ priv->up_count++;
+ }
priv->child_up[idx] = 1;
- priv->up_count++;
+ call_psh = 1;
+ up_child = idx;
for (i = 0; i < priv->child_count; i++)
if (priv->child_up[i] == 1)
up_children++;
@@ -3474,12 +3920,6 @@ afr_notify (xlator_t *this, int32_t event,
"going online.", ((xlator_t *)data)->name);
} else {
event = GF_EVENT_CHILD_MODIFIED;
- gf_log (this->name, GF_LOG_INFO, "subvol %d came up, "
- "start crawl", idx);
- if (had_heard_from_all) {
- call_psh = 1;
- up_child = idx;
- }
}
priv->last_event[idx] = event;
@@ -3491,8 +3931,22 @@ afr_notify (xlator_t *this, int32_t event,
case GF_EVENT_CHILD_DOWN:
LOCK (&priv->lock);
{
+ /*
+ * If a brick is down when we start, we'll get a
+ * CHILD_DOWN to indicate its initial state. There
+ * was never a CHILD_UP in this case, so if we
+ * increment "down_count" the difference between than
+ * and "up_count" will no longer be the number of
+ * children that are currently up. This has serious
+ * implications e.g. for quorum enforcement, so we
+ * don't increment these values unless the event
+ * represents an actual state transition between "up"
+ * (value = 1) and anything else.
+ */
+ if (priv->child_up[idx] == 1) {
+ priv->down_count++;
+ }
priv->child_up[idx] = 0;
- priv->down_count++;
for (i = 0; i < priv->child_count; i++)
if (priv->child_up[i] == 0)
@@ -3520,10 +3974,11 @@ afr_notify (xlator_t *this, int32_t event,
break;
- case GF_EVENT_TRIGGER_HEAL:
- gf_log (this->name, GF_LOG_INFO, "Self-heal was triggered"
- " manually. Start crawling");
- call_psh = 1;
+ case GF_EVENT_TRANSLATOR_OP:
+ input = data;
+ output = data2;
+ ret = afr_xl_op (this, input, output);
+ goto out;
break;
default:
@@ -3568,18 +4023,13 @@ afr_notify (xlator_t *this, int32_t event,
}
}
UNLOCK (&priv->lock);
- if (up_children > 1) {
- gf_log (this->name, GF_LOG_INFO, "All subvolumes came "
- "up, start crawl");
- call_psh = 1;
- }
}
ret = 0;
if (propagate)
ret = default_notify (this, event, data);
- if (call_psh)
- afr_proactive_self_heal (this, up_child);
+ if (call_psh && priv->shd.iamshd)
+ afr_proactive_self_heal ((void*) (long) up_child);
out:
return ret;
@@ -3611,8 +4061,8 @@ afr_local_init (afr_local_t *local, afr_private_t *priv, int32_t *op_errno)
local->op_ret = -1;
local->op_errno = EUCLEAN;
- local->child_up = GF_CALLOC (sizeof (*local->child_up),
- priv->child_count,
+ local->child_up = GF_CALLOC (priv->child_count,
+ sizeof (*local->child_up),
gf_afr_mt_char);
if (!local->child_up) {
if (op_errno)
@@ -3630,6 +4080,27 @@ afr_local_init (afr_local_t *local, afr_private_t *priv, int32_t *op_errno)
*op_errno = ENOTCONN;
goto out;
}
+
+ local->child_errno = GF_CALLOC (priv->child_count,
+ sizeof (*local->child_errno),
+ gf_afr_mt_int32_t);
+ if (!local->child_errno) {
+ if (op_errno)
+ *op_errno = ENOMEM;
+ goto out;
+ }
+
+ local->transaction.postop_piggybacked = GF_CALLOC (priv->child_count,
+ sizeof (int),
+ gf_afr_mt_int32_t);
+ if (!local->transaction.postop_piggybacked) {
+ if (op_errno)
+ *op_errno = ENOMEM;
+ goto out;
+ }
+
+ local->append_write = _gf_false;
+
ret = 0;
out:
return ret;
@@ -3641,16 +4112,6 @@ afr_internal_lock_init (afr_internal_lock_t *lk, size_t child_count,
{
int ret = -ENOMEM;
- lk->inode_locked_nodes = GF_CALLOC (sizeof (*lk->inode_locked_nodes),
- child_count, gf_afr_mt_char);
- if (NULL == lk->inode_locked_nodes)
- goto out;
-
- lk->entry_locked_nodes = GF_CALLOC (sizeof (*lk->entry_locked_nodes),
- child_count, gf_afr_mt_char);
- if (NULL == lk->entry_locked_nodes)
- goto out;
-
lk->locked_nodes = GF_CALLOC (sizeof (*lk->locked_nodes),
child_count, gf_afr_mt_char);
if (NULL == lk->locked_nodes)
@@ -3670,10 +4131,62 @@ out:
return ret;
}
+void
+afr_matrix_cleanup (int32_t **matrix, unsigned int m)
+{
+ int i = 0;
+
+ if (!matrix)
+ goto out;
+ for (i = 0; i < m; i++) {
+ GF_FREE (matrix[i]);
+ }
+
+ GF_FREE (matrix);
+out:
+ return;
+}
+
+int32_t**
+afr_matrix_create (unsigned int m, unsigned int n)
+{
+ int32_t **matrix = NULL;
+ int i = 0;
+
+ matrix = GF_CALLOC (sizeof (*matrix), m, gf_afr_mt_int32_t);
+ if (!matrix)
+ goto out;
+
+ for (i = 0; i < m; i++) {
+ matrix[i] = GF_CALLOC (sizeof (*matrix[i]), n,
+ gf_afr_mt_int32_t);
+ if (!matrix[i])
+ goto out;
+ }
+ return matrix;
+out:
+ afr_matrix_cleanup (matrix, m);
+ return NULL;
+}
+
+int
+afr_inodelk_init (afr_inodelk_t *lk, char *dom, size_t child_count)
+{
+ int ret = -ENOMEM;
+
+ lk->domain = dom;
+ lk->locked_nodes = GF_CALLOC (sizeof (*lk->locked_nodes),
+ child_count, gf_afr_mt_char);
+ if (NULL == lk->locked_nodes)
+ goto out;
+ ret = 0;
+out:
+ return ret;
+}
+
int
afr_transaction_local_init (afr_local_t *local, xlator_t *this)
{
- int i = 0;
int child_up_count = 0;
int ret = -ENOMEM;
afr_private_t *priv = NULL;
@@ -3684,6 +4197,14 @@ afr_transaction_local_init (afr_local_t *local, xlator_t *this)
if (ret < 0)
goto out;
+ if ((local->transaction.type == AFR_DATA_TRANSACTION) ||
+ (local->transaction.type == AFR_METADATA_TRANSACTION)) {
+ ret = afr_inodelk_init (&local->internal_lock.inodelk[0],
+ this->name, priv->child_count);
+ if (ret < 0)
+ goto out;
+ }
+
ret = -ENOMEM;
child_up_count = afr_up_children_count (local->child_up,
priv->child_count);
@@ -3693,12 +4214,6 @@ afr_transaction_local_init (afr_local_t *local, xlator_t *this)
local->first_up_child = afr_first_up_child (local->child_up,
priv->child_count);
- local->child_errno = GF_CALLOC (sizeof (*local->child_errno),
- priv->child_count,
- gf_afr_mt_int32_t);
- if (!local->child_errno)
- goto out;
-
local->transaction.eager_lock =
GF_CALLOC (sizeof (*local->transaction.eager_lock),
priv->child_count,
@@ -3707,44 +4222,27 @@ afr_transaction_local_init (afr_local_t *local, xlator_t *this)
if (!local->transaction.eager_lock)
goto out;
- local->pending = GF_CALLOC (sizeof (*local->pending),
- priv->child_count,
- gf_afr_mt_int32_t);
-
- if (!local->pending)
- goto out;
-
local->fresh_children = afr_children_create (priv->child_count);
if (!local->fresh_children)
goto out;
- if (local->fd) {
- local->fd_open_on = GF_CALLOC (sizeof (*local->fd_open_on),
- priv->child_count,
- gf_afr_mt_char);
- if (!local->fd_open_on)
- goto out;
- }
-
local->transaction.pre_op = GF_CALLOC (sizeof (*local->transaction.pre_op),
priv->child_count,
gf_afr_mt_char);
if (!local->transaction.pre_op)
goto out;
- for (i = 0; i < priv->child_count; i++) {
- local->pending[i] = GF_CALLOC (sizeof (*local->pending[i]),
- 3, /* data + metadata + entry */
- gf_afr_mt_int32_t);
- if (!local->pending[i])
- goto out;
- }
+ local->pending = afr_matrix_create (priv->child_count,
+ AFR_NUM_CHANGE_LOGS);
+ if (!local->pending)
+ goto out;
- local->transaction.child_errno =
- GF_CALLOC (sizeof (*local->transaction.child_errno),
- priv->child_count,
- gf_afr_mt_int32_t);
- local->transaction.erase_pending = 1;
+ local->transaction.txn_changelog = afr_matrix_create (priv->child_count,
+ AFR_NUM_CHANGE_LOGS);
+ if (!local->transaction.txn_changelog)
+ goto out;
+
+ INIT_LIST_HEAD (&local->transaction.eager_locked);
ret = 0;
out:
@@ -3840,7 +4338,7 @@ afr_set_low_priority (call_frame_t *frame)
int
afr_child_fd_ctx_set (xlator_t *this, fd_t *fd, int32_t child,
- int flags, int32_t wbflags)
+ int flags)
{
int ret = 0;
uint64_t ctx = 0;
@@ -3865,7 +4363,6 @@ afr_child_fd_ctx_set (xlator_t *this, fd_t *fd, int32_t child,
fd_ctx->opened_on[child] = AFR_FD_OPENED;
if (!IA_ISDIR (fd->inode->ia_type)) {
fd_ctx->flags = flags;
- fd_ctx->wbflags = wbflags;
}
ret = 0;
out:
@@ -3918,6 +4415,32 @@ afr_priv_destroy (afr_private_t *priv)
goto out;
inode_unref (priv->root_inode);
GF_FREE (priv->shd.pos);
+ GF_FREE (priv->shd.pending);
+ GF_FREE (priv->shd.inprogress);
+// for (i = 0; i < priv->child_count; i++)
+// if (priv->shd.timer && priv->shd.timer[i])
+// gf_timer_call_cancel (this->ctx, priv->shd.timer[i]);
+ GF_FREE (priv->shd.timer);
+
+ if (priv->shd.healed)
+ eh_destroy (priv->shd.healed);
+
+ if (priv->shd.heal_failed)
+ eh_destroy (priv->shd.heal_failed);
+
+ if (priv->shd.split_brain)
+ eh_destroy (priv->shd.split_brain);
+
+ for (i = 0; i < priv->child_count; i++)
+ {
+ if (priv->shd.statistics[i])
+ eh_destroy (priv->shd.statistics[i]);
+ }
+
+ GF_FREE (priv->shd.statistics);
+
+ GF_FREE (priv->shd.crawl_events);
+
GF_FREE (priv->last_event);
if (priv->pending_key) {
for (i = 0; i < priv->child_count; i++)
@@ -3944,3 +4467,125 @@ xlator_subvolume_count (xlator_t *this)
i++;
return i;
}
+
+inline gf_boolean_t
+afr_is_errno_set (int *child_errno, int child)
+{
+ return child_errno[child];
+}
+
+inline gf_boolean_t
+afr_is_errno_unset (int *child_errno, int child)
+{
+ return !afr_is_errno_set (child_errno, child);
+}
+
+void
+afr_prepare_new_entry_pending_matrix (int32_t **pending,
+ gf_boolean_t (*is_pending) (int *, int),
+ int *ctx, struct iatt *buf,
+ unsigned int child_count)
+{
+ int midx = 0;
+ int idx = 0;
+ int i = 0;
+
+ midx = afr_index_for_transaction_type (AFR_METADATA_TRANSACTION);
+ if (IA_ISDIR (buf->ia_type))
+ idx = afr_index_for_transaction_type (AFR_ENTRY_TRANSACTION);
+ else if (IA_ISREG (buf->ia_type))
+ idx = afr_index_for_transaction_type (AFR_DATA_TRANSACTION);
+ else
+ idx = -1;
+ for (i = 0; i < child_count; i++) {
+ if (is_pending (ctx, i)) {
+ pending[i][midx] = hton32 (1);
+ if (idx == -1)
+ continue;
+ pending[i][idx] = hton32 (1);
+ }
+ }
+}
+
+gf_boolean_t
+afr_is_fd_fixable (fd_t *fd)
+{
+ if (!fd || !fd->inode)
+ return _gf_false;
+ else if (fd_is_anonymous (fd))
+ return _gf_false;
+ else if (uuid_is_null (fd->inode->gfid))
+ return _gf_false;
+
+ return _gf_true;
+}
+
+void
+afr_handle_open_fd_count (call_frame_t *frame, xlator_t *this)
+{
+ afr_local_t *local = NULL;
+ inode_t *inode = NULL;
+ afr_inode_ctx_t *ctx = NULL;
+
+ local = frame->local;
+
+ if (local->fd)
+ inode = local->fd->inode;
+ else
+ inode = local->loc.inode;
+
+ if (!inode)
+ return;
+
+ LOCK (&inode->lock);
+ {
+ ctx = __afr_inode_ctx_get (inode, this);
+ ctx->open_fd_count = local->open_fd_count;
+ }
+ UNLOCK (&inode->lock);
+}
+
+int
+afr_initialise_statistics (xlator_t *this)
+{
+ afr_private_t *priv = NULL;
+ int ret = -1;
+ int i = 0;
+ int child_count = 0;
+ eh_t *stats_per_brick = NULL;
+ shd_crawl_event_t ***shd_crawl_events = NULL;
+ priv = this->private;
+
+ priv->shd.statistics = GF_CALLOC (sizeof(eh_t *), priv->child_count,
+ gf_common_mt_eh_t);
+ if (!priv->shd.statistics) {
+ ret = -1;
+ goto out;
+ }
+ child_count = priv->child_count;
+ for (i=0; i < child_count ; i++) {
+ stats_per_brick = eh_new (AFR_STATISTICS_HISTORY_SIZE,
+ _gf_false,
+ _destroy_crawl_event_data);
+ if (!stats_per_brick) {
+ ret = -1;
+ goto out;
+ }
+ priv->shd.statistics[i] = stats_per_brick;
+
+ }
+
+ shd_crawl_events = (shd_crawl_event_t***)(&priv->shd.crawl_events);
+ *shd_crawl_events = GF_CALLOC (sizeof(shd_crawl_event_t*),
+ priv->child_count,
+ gf_afr_mt_shd_crawl_event_t);
+
+ if (!priv->shd.crawl_events) {
+ ret = -1;
+ goto out;
+ }
+ ret = 0;
+out:
+ return ret;
+
+}
diff --git a/xlators/cluster/afr/src/afr-dir-read.c b/xlators/cluster/afr/src/afr-dir-read.c
index 389515e3c..689dd84e6 100644
--- a/xlators/cluster/afr/src/afr-dir-read.c
+++ b/xlators/cluster/afr/src/afr-dir-read.c
@@ -1,20 +1,11 @@
/*
- Copyright (c) 2007-2011 Gluster, Inc. <http://www.gluster.com>
+ Copyright (c) 2008-2012 Red Hat, Inc. <http://www.redhat.com>
This file is part of GlusterFS.
- GlusterFS is free software; you can redistribute it and/or modify
- it under the terms of the GNU General Public License as published
- by the Free Software Foundation; either version 3 of the License,
- or (at your option) any later version.
-
- GlusterFS is distributed in the hope that it will be useful, but
- WITHOUT ANY WARRANTY; without even the implied warranty of
- MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU
- General Public License for more details.
-
- You should have received a copy of the GNU General Public License
- along with this program. If not, see
- <http://www.gnu.org/licenses/>.
+ This file is licensed to you under your choice of the GNU Lesser
+ General Public License, version 3 or any later version (LGPLv3 or
+ later), or the GNU General Public License, version 2 (GPLv2), in all
+ cases as published by the Free Software Foundation.
*/
@@ -51,7 +42,7 @@
int
afr_examine_dir_sh_unwind (call_frame_t *frame, xlator_t *this, int32_t op_ret,
- int32_t op_errno)
+ int32_t op_errno, int32_t sh_failed)
{
afr_local_t *local = NULL;
@@ -60,7 +51,7 @@ afr_examine_dir_sh_unwind (call_frame_t *frame, xlator_t *this, int32_t op_ret,
afr_set_opendir_done (this, local->fd->inode);
AFR_STACK_UNWIND (opendir, frame, local->op_ret,
- local->op_errno, local->fd);
+ local->op_errno, local->fd, NULL);
return 0;
}
@@ -99,7 +90,7 @@ __checksums_differ (uint32_t *checksum, int child_count,
int32_t
afr_examine_dir_readdir_cbk (call_frame_t *frame, void *cookie,
xlator_t *this, int32_t op_ret, int32_t op_errno,
- gf_dirent_t *entries)
+ gf_dirent_t *entries, dict_t *xdata)
{
afr_private_t * priv = NULL;
afr_local_t * local = NULL;
@@ -137,7 +128,7 @@ afr_examine_dir_readdir_cbk (call_frame_t *frame, void *cookie,
}
list_for_each_entry_safe (entry, tmp, &entries->list, list) {
- entry_cksum = gf_rsync_weak_checksum (entry->d_name,
+ entry_cksum = gf_rsync_weak_checksum ((unsigned char *)entry->d_name,
strlen (entry->d_name));
local->cont.opendir.checksum[child_index] ^= entry_cksum;
}
@@ -152,7 +143,7 @@ afr_examine_dir_readdir_cbk (call_frame_t *frame, void *cookie,
(void *) (long) child_index,
priv->children[child_index],
priv->children[child_index]->fops->readdir,
- local->fd, 131072, last_offset);
+ local->fd, 131072, last_offset, NULL);
return 0;
@@ -175,7 +166,7 @@ out:
afr_set_opendir_done (this, inode);
AFR_STACK_UNWIND (opendir, frame, local->op_ret,
- local->op_errno, local->fd);
+ local->op_errno, local->fd, NULL);
}
}
@@ -208,7 +199,7 @@ afr_examine_dir (call_frame_t *frame, xlator_t *this)
(void *) (long) i,
priv->children[i],
priv->children[i]->fops->readdir,
- local->fd, 131072, 0);
+ local->fd, 131072, 0, NULL);
if (!--call_count)
break;
@@ -222,7 +213,7 @@ afr_examine_dir (call_frame_t *frame, xlator_t *this)
int32_t
afr_opendir_cbk (call_frame_t *frame, void *cookie,
xlator_t *this, int32_t op_ret, int32_t op_errno,
- fd_t *fd)
+ fd_t *fd, dict_t *xdata)
{
afr_private_t *priv = NULL;
afr_local_t *local = NULL;
@@ -242,8 +233,7 @@ afr_opendir_cbk (call_frame_t *frame, void *cookie,
{
if (op_ret >= 0) {
local->op_ret = op_ret;
- ret = afr_child_fd_ctx_set (this, fd, child_index,
- 0, 0);
+ ret = afr_child_fd_ctx_set (this, fd, child_index, 0);
if (ret) {
local->op_ret = -1;
local->op_errno = -ret;
@@ -263,7 +253,7 @@ unlock:
goto out;
if (!afr_is_opendir_done (this, local->fd->inode) &&
- up_children_count > 1) {
+ up_children_count > 1 && priv->entry_self_heal) {
/*
* This is the first opendir on this inode. We need
@@ -291,7 +281,7 @@ unlock:
out:
AFR_STACK_UNWIND (opendir, frame, local->op_ret,
- local->op_errno, local->fd);
+ local->op_errno, local->fd, NULL);
return 0;
}
@@ -317,7 +307,7 @@ afr_opendir (call_frame_t *frame, xlator_t *this,
child_count = priv->child_count;
- ALLOC_OR_GOTO (frame->local, afr_local_t, out);
+ AFR_LOCAL_ALLOC_OR_GOTO (frame->local, out);
local = frame->local;
ret = afr_local_init (local, priv, &op_errno);
@@ -336,7 +326,7 @@ afr_opendir (call_frame_t *frame, xlator_t *this,
(void*) (long) i,
priv->children[i],
priv->children[i]->fops->opendir,
- loc, fd);
+ loc, fd, NULL);
if (!--call_count)
break;
@@ -346,7 +336,7 @@ afr_opendir (call_frame_t *frame, xlator_t *this,
ret = 0;
out:
if (ret < 0)
- AFR_STACK_UNWIND (opendir, frame, -1, op_errno, fd);
+ AFR_STACK_UNWIND (opendir, frame, -1, op_errno, fd, NULL);
return 0;
}
@@ -368,85 +358,6 @@ struct entry_name {
struct list_head list;
};
-
-static gf_boolean_t
-remembered_name (const char *name, struct list_head *entries)
-{
- struct entry_name *e = NULL;
- gf_boolean_t ret = _gf_false;
-
- list_for_each_entry (e, entries, list) {
- if (!strcmp (name, e->name)) {
- ret = _gf_true;
- goto out;
- }
- }
-
-out:
- return ret;
-}
-
-
-static void
-afr_remember_entries (gf_dirent_t *entries, fd_t *fd)
-{
- struct entry_name *n = NULL;
- gf_dirent_t *entry = NULL;
- int ret = 0;
- uint64_t ctx = 0;
- afr_fd_ctx_t *fd_ctx = NULL;
-
- ret = fd_ctx_get (fd, THIS, &ctx);
- if (ret < 0) {
- gf_log (THIS->name, GF_LOG_INFO,
- "could not get fd ctx for fd=%p", fd);
- return;
- }
-
- fd_ctx = (afr_fd_ctx_t *)(long) ctx;
-
- list_for_each_entry (entry, &entries->list, list) {
- n = GF_CALLOC (1, sizeof (*n), gf_afr_mt_entry_name);
- n->name = gf_strdup (entry->d_name);
- INIT_LIST_HEAD (&n->list);
-
- list_add (&n->list, &fd_ctx->entries);
- }
-}
-
-
-static off_t
-afr_filter_entries (gf_dirent_t *entries, fd_t *fd)
-{
- gf_dirent_t *entry = NULL;
- gf_dirent_t *tmp = NULL;
- int ret = 0;
- uint64_t ctx = 0;
- afr_fd_ctx_t *fd_ctx = NULL;
- off_t offset = 0;
-
- ret = fd_ctx_get (fd, THIS, &ctx);
- if (ret < 0) {
- gf_log (THIS->name, GF_LOG_INFO,
- "could not get fd ctx for fd=%p", fd);
- return -1;
- }
-
- fd_ctx = (afr_fd_ctx_t *)(long) ctx;
-
- list_for_each_entry_safe (entry, tmp, &entries->list, list) {
- offset = entry->d_off;
-
- if (remembered_name (entry->d_name, &fd_ctx->entries)) {
- list_del (&entry->list);
- GF_FREE (entry);
- }
- }
-
- return offset;
-}
-
-
static void
afr_forget_entries (fd_t *fd)
{
@@ -472,159 +383,55 @@ afr_forget_entries (fd_t *fd)
}
}
-
-int32_t
-afr_readdir_cbk (call_frame_t *frame, void *cookie,
- xlator_t *this, int32_t op_ret, int32_t op_errno,
- gf_dirent_t *entries)
+static void
+afr_readdir_filter_trash_dir (gf_dirent_t *entries, fd_t *fd)
{
- afr_local_t * local = NULL;
gf_dirent_t * entry = NULL;
gf_dirent_t * tmp = NULL;
- local = frame->local;
-
- if (op_ret == -1)
- goto out;
-
list_for_each_entry_safe (entry, tmp, &entries->list, list) {
- if ((local->fd->inode == local->fd->inode->table->root)
- && !strcmp (entry->d_name, GF_REPLICATE_TRASH_DIR)) {
+ if (__is_root_gfid (fd->inode->gfid) &&
+ !strcmp (entry->d_name, GF_REPLICATE_TRASH_DIR)) {
list_del_init (&entry->list);
GF_FREE (entry);
}
}
-
-out:
- AFR_STACK_UNWIND (readdir, frame, op_ret, op_errno, entries);
-
- return 0;
}
-
int32_t
-afr_readdirp_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
- int32_t op_ret, int32_t op_errno, gf_dirent_t *entries)
+afr_readdir_cbk (call_frame_t *frame, void *cookie,
+ xlator_t *this, int32_t op_ret, int32_t op_errno,
+ gf_dirent_t *entries, dict_t *xdata)
{
- afr_private_t * priv = NULL;
- afr_local_t * local = NULL;
- xlator_t ** children = NULL;
- int32_t next_call_child = -1;
- int ret = 0;
- gf_dirent_t * entry = NULL;
- gf_dirent_t * tmp = NULL;
- int32_t *last_index = NULL;
- int32_t read_child = -1;
- int32_t *fresh_children = NULL;
- uint64_t ctx = 0;
- afr_fd_ctx_t *fd_ctx = NULL;
- off_t offset = 0;
- int32_t call_child = -1;
+ afr_local_t *local = NULL;
- priv = this->private;
- children = priv->children;
+ if (op_ret == -1)
+ goto out;
local = frame->local;
+ afr_readdir_filter_trash_dir (entries, local->fd);
- read_child = (long) cookie;
- last_index = &local->cont.readdir.last_index;
- fresh_children = local->fresh_children;
-
- /* the value of the last_index changes if afr_next_call_child is
- * called. So to find the call_child of this callback use last_index
- * before the next_call_child call.
- */
- if (*last_index == -1)
- call_child = read_child;
- else
- call_child = fresh_children[*last_index];
-
- if (priv->strict_readdir) {
- ret = fd_ctx_get (local->fd, this, &ctx);
- if (ret < 0) {
- gf_log (this->name, GF_LOG_INFO,
- "could not get fd ctx for fd=%p", local->fd);
- op_ret = -1;
- op_errno = -ret;
- goto out;
- }
-
- fd_ctx = (afr_fd_ctx_t *)(long) ctx;
-
- if (op_ret == -1) {
- next_call_child = afr_next_call_child (fresh_children,
- local->child_up,
- priv->child_count,
- last_index,
- read_child);
- if (next_call_child < 0)
- goto out;
- gf_log (this->name, GF_LOG_TRACE,
- "starting readdir afresh on child %d, offset %"PRId64,
- next_call_child, (uint64_t) 0);
-
- fd_ctx->failed_over = _gf_true;
-
- STACK_WIND_COOKIE (frame, afr_readdirp_cbk,
- (void *) (long) read_child,
- children[next_call_child],
- children[next_call_child]->fops->readdirp,
- local->fd,
- local->cont.readdir.size, 0,
- local->cont.readdir.dict);
- return 0;
- }
- }
-
- if (op_ret != -1) {
- list_for_each_entry_safe (entry, tmp, &entries->list, list) {
- if ((local->fd->inode == local->fd->inode->table->root)
- && !strcmp (entry->d_name, GF_REPLICATE_TRASH_DIR)) {
- list_del_init (&entry->list);
- GF_FREE (entry);
- }
- }
- }
-
- if (priv->strict_readdir) {
- if (fd_ctx->failed_over) {
- if (list_empty (&entries->list)) {
- gf_log (this->name, GF_LOG_DEBUG,
- "no entries found");
- goto out;
- }
-
- offset = afr_filter_entries (entries, local->fd);
+out:
+ AFR_STACK_UNWIND (readdir, frame, op_ret, op_errno, entries, NULL);
+ return 0;
+}
- afr_remember_entries (entries, local->fd);
- if (list_empty (&entries->list)) {
- /* All the entries we got were duplicate. We
- shouldn't send an empty list now, because
- that will make the application stop reading. So
- try to get more entries */
+int32_t
+afr_readdirp_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
+ int32_t op_ret, int32_t op_errno, gf_dirent_t *entries,
+ dict_t *xdata)
+{
+ afr_local_t *local = NULL;
- gf_log (this->name, GF_LOG_TRACE,
- "trying to fetch non-duplicate entries "
- "from offset %"PRId64", child %s",
- offset, children[call_child]->name);
+ if (op_ret == -1)
+ goto out;
- STACK_WIND_COOKIE (frame, afr_readdirp_cbk,
- (void *) (long) read_child,
- children[call_child],
- children[call_child]->fops->readdirp,
- local->fd, local->cont.readdir.size, offset,
- local->cont.readdir.dict);
- return 0;
- }
- } else {
- afr_remember_entries (entries, local->fd);
- }
- }
+ local = frame->local;
+ afr_readdir_filter_trash_dir (entries, local->fd);
out:
- AFR_STACK_UNWIND (readdirp, frame, op_ret, op_errno, entries);
-
+ AFR_STACK_UNWIND (readdirp, frame, op_ret, op_errno, entries, NULL);
return 0;
}
@@ -632,15 +439,14 @@ int32_t
afr_do_readdir (call_frame_t *frame, xlator_t *this,
fd_t *fd, size_t size, off_t offset, int whichop, dict_t *dict)
{
- afr_private_t * priv = NULL;
- xlator_t ** children = NULL;
- int call_child = 0;
- afr_local_t *local = NULL;
- uint64_t ctx = 0;
- afr_fd_ctx_t *fd_ctx = NULL;
- int ret = -1;
- int32_t op_errno = 0;
- uint64_t read_child = 0;
+ afr_private_t *priv = NULL;
+ xlator_t **children = NULL;
+ int call_child = 0;
+ afr_local_t *local = NULL;
+ afr_fd_ctx_t *fd_ctx = NULL;
+ int ret = -1;
+ int32_t op_errno = 0;
+ uint64_t read_child = 0;
VALIDATE_OR_GOTO (frame, out);
VALIDATE_OR_GOTO (this, out);
@@ -649,7 +455,7 @@ afr_do_readdir (call_frame_t *frame, xlator_t *this,
priv = this->private;
children = priv->children;
- ALLOC_OR_GOTO (frame->local, afr_local_t, out);
+ AFR_LOCAL_ALLOC_OR_GOTO (frame->local, out);
local = frame->local;
ret = afr_local_init (local, priv, &op_errno);
@@ -665,48 +471,38 @@ afr_do_readdir (call_frame_t *frame, xlator_t *this,
read_child = afr_inode_get_read_ctx (this, fd->inode,
local->fresh_children);
ret = afr_get_call_child (this, local->child_up, read_child,
- local->fresh_children,
- &call_child,
- &local->cont.readdir.last_index);
+ local->fresh_children,
+ &call_child,
+ &local->cont.readdir.last_index);
if (ret < 0) {
op_errno = -ret;
goto out;
}
+ fd_ctx = afr_fd_ctx_get (fd, this);
+ if (!fd_ctx) {
+ op_errno = EBADF;
+ goto out;
+ }
+
+ if ((offset == 0) || (fd_ctx->call_child == -1)) {
+ fd_ctx->call_child = call_child;
+ } else if ((priv->readdir_failover == _gf_false) &&
+ (call_child != fd_ctx->call_child)) {
+ op_errno = EBADF;
+ goto out;
+ }
+
local->fd = fd_ref (fd);
local->cont.readdir.size = size;
local->cont.readdir.dict = (dict)? dict_ref (dict) : NULL;
- if (priv->strict_readdir) {
- ret = fd_ctx_get (fd, this, &ctx);
- if (ret < 0) {
- gf_log (this->name, GF_LOG_INFO,
- "could not get fd ctx for fd=%p", fd);
- op_errno = -ret;
- goto out;
- }
-
- fd_ctx = (afr_fd_ctx_t *)(long) ctx;
-
- if (fd_ctx->last_tried != call_child) {
- gf_log (this->name, GF_LOG_TRACE,
- "first up child has changed from %d to %d, "
- "restarting readdir from offset 0",
- fd_ctx->last_tried, call_child);
-
- fd_ctx->failed_over = _gf_true;
- offset = 0;
- }
-
- fd_ctx->last_tried = call_child;
- }
-
if (whichop == GF_FOP_READDIR)
STACK_WIND_COOKIE (frame, afr_readdir_cbk,
(void *) (long) call_child,
children[call_child],
children[call_child]->fops->readdir, fd,
- size, offset);
+ size, offset, dict);
else
STACK_WIND_COOKIE (frame, afr_readdirp_cbk,
(void *) (long) call_child,
@@ -714,19 +510,18 @@ afr_do_readdir (call_frame_t *frame, xlator_t *this,
children[call_child]->fops->readdirp, fd,
size, offset, dict);
- ret = 0;
+ return 0;
out:
- if (ret < 0)
- AFR_STACK_UNWIND (readdir, frame, -1, op_errno, NULL);
+ AFR_STACK_UNWIND (readdir, frame, -1, op_errno, NULL, NULL);
return 0;
}
int32_t
afr_readdir (call_frame_t *frame, xlator_t *this, fd_t *fd, size_t size,
- off_t offset)
+ off_t offset, dict_t *xdata)
{
- afr_do_readdir (frame, this, fd, size, offset, GF_FOP_READDIR, NULL);
+ afr_do_readdir (frame, this, fd, size, offset, GF_FOP_READDIR, xdata);
return 0;
}
diff --git a/xlators/cluster/afr/src/afr-dir-read.h b/xlators/cluster/afr/src/afr-dir-read.h
index 7e50a1c8c..09456d159 100644
--- a/xlators/cluster/afr/src/afr-dir-read.h
+++ b/xlators/cluster/afr/src/afr-dir-read.h
@@ -1,20 +1,11 @@
/*
- Copyright (c) 2007-2011 Gluster, Inc. <http://www.gluster.com>
- This file is part of GlusterFS.
-
- GlusterFS is free software; you can redistribute it and/or modify
- it under the terms of the GNU General Public License as published
- by the Free Software Foundation; either version 3 of the License,
- or (at your option) any later version.
-
- GlusterFS is distributed in the hope that it will be useful, but
- WITHOUT ANY WARRANTY; without even the implied warranty of
- MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU
- General Public License for more details.
-
- You should have received a copy of the GNU General Public License
- along with this program. If not, see
- <http://www.gnu.org/licenses/>.
+ Copyright (c) 2008-2012 Red Hat, Inc. <http://www.redhat.com>
+ This file is part of GlusterFS.
+
+ This file is licensed to you under your choice of the GNU Lesser
+ General Public License, version 3 or any later version (LGPLv3 or
+ later), or the GNU General Public License, version 2 (GPLv2), in all
+ cases as published by the Free Software Foundation.
*/
#ifndef __DIR_READ_H__
@@ -23,14 +14,14 @@
int32_t
afr_opendir (call_frame_t *frame, xlator_t *this,
- loc_t *loc, fd_t *fd);
+ loc_t *loc, fd_t *fd, dict_t *xdata);
int32_t
afr_releasedir (xlator_t *this, fd_t *fd);
int32_t
afr_readdir (call_frame_t *frame, xlator_t *this,
- fd_t *fd, size_t size, off_t offset);
+ fd_t *fd, size_t size, off_t offset, dict_t *xdata);
int32_t
@@ -39,7 +30,7 @@ afr_readdirp (call_frame_t *frame, xlator_t *this,
int32_t
afr_checksum (call_frame_t *frame, xlator_t *this,
- loc_t *loc, int32_t flags);
+ loc_t *loc, int32_t flags, dict_t *xdata);
#endif /* __DIR_READ_H__ */
diff --git a/xlators/cluster/afr/src/afr-dir-write.c b/xlators/cluster/afr/src/afr-dir-write.c
index 91aa2a9e7..1943b719b 100644
--- a/xlators/cluster/afr/src/afr-dir-write.c
+++ b/xlators/cluster/afr/src/afr-dir-write.c
@@ -1,20 +1,11 @@
/*
- Copyright (c) 2007-2011 Gluster, Inc. <http://www.gluster.com>
+ Copyright (c) 2008-2012 Red Hat, Inc. <http://www.redhat.com>
This file is part of GlusterFS.
- GlusterFS is free software; you can redistribute it and/or modify
- it under the terms of the GNU General Public License as published
- by the Free Software Foundation; either version 3 of the License,
- or (at your option) any later version.
-
- GlusterFS is distributed in the hope that it will be useful, but
- WITHOUT ANY WARRANTY; without even the implied warranty of
- MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU
- General Public License for more details.
-
- You should have received a copy of the GNU General Public License
- along with this program. If not, see
- <http://www.gnu.org/licenses/>.
+ This file is licensed to you under your choice of the GNU Lesser
+ General Public License, version 3 or any later version (LGPLv3 or
+ later), or the GNU General Public License, version 2 (GPLv2), in all
+ cases as published by the Free Software Foundation.
*/
@@ -47,31 +38,222 @@
#include "afr.h"
#include "afr-transaction.h"
-void
-afr_build_parent_loc (loc_t *parent, loc_t *child)
+int
+afr_build_parent_loc (loc_t *parent, loc_t *child, int32_t *op_errno)
{
- char *tmp = NULL;
+ int ret = -1;
+ char *child_path = NULL;
if (!child->parent) {
- //this should never be called with root as the child
- GF_ASSERT (0);
- loc_copy (parent, child);
- return;
+ if (op_errno)
+ *op_errno = EINVAL;
+ goto out;
}
- tmp = gf_strdup (child->path);
- parent->path = gf_strdup (dirname (tmp));
- GF_FREE (tmp);
+ child_path = gf_strdup (child->path);
+ if (!child_path) {
+ if (op_errno)
+ *op_errno = ENOMEM;
+ goto out;
+ }
+ parent->path = gf_strdup( dirname (child_path) );
+ if (!parent->path) {
+ if (op_errno)
+ *op_errno = ENOMEM;
+ goto out;
+ }
+ parent->inode = inode_ref (child->parent);
+ uuid_copy (parent->gfid, child->pargfid);
- parent->name = strrchr (parent->path, '/');
- if (parent->name)
- parent->name++;
+ ret = 0;
+out:
+ GF_FREE(child_path);
- parent->inode = inode_ref (child->parent);
- parent->parent = inode_parent (parent->inode, 0, NULL);
+ return ret;
+}
- if (!uuid_is_null (child->pargfid))
- uuid_copy (parent->gfid, child->pargfid);
+void
+__dir_entry_fop_common_cbk (call_frame_t *frame, int child_index,
+ xlator_t *this, int32_t op_ret,
+ int32_t op_errno, inode_t *inode,
+ struct iatt *buf, struct iatt *preparent,
+ struct iatt *postparent, struct iatt *prenewparent,
+ struct iatt *postnewparent)
+{
+ afr_local_t *local = NULL;
+
+ local = frame->local;
+
+ if (afr_fop_failed (op_ret, op_errno))
+ afr_transaction_fop_failed (frame, this, child_index);
+
+ if (op_ret > -1) {
+ local->op_ret = op_ret;
+
+ if ((local->success_count == 0) ||
+ (child_index == local->read_child_index)) {
+ local->cont.dir_fop.preparent = *preparent;
+ local->cont.dir_fop.postparent = *postparent;
+ if (buf)
+ local->cont.dir_fop.buf = *buf;
+ if (prenewparent)
+ local->cont.dir_fop.prenewparent = *prenewparent;
+ if (postnewparent)
+ local->cont.dir_fop.postnewparent = *postnewparent;
+ }
+
+ local->cont.dir_fop.inode = inode;
+
+ local->fresh_children[local->success_count] = child_index;
+ local->success_count++;
+ local->child_errno[child_index] = 0;
+ } else {
+ local->child_errno[child_index] = op_errno;
+ }
+
+ local->op_errno = op_errno;
+}
+
+int
+afr_mark_new_entry_changelog_cbk (call_frame_t *frame, void *cookie,
+ xlator_t *this,
+ int32_t op_ret, int32_t op_errno,
+ dict_t *xattr, dict_t *xdata)
+{
+ int call_count = 0;
+
+ call_count = afr_frame_return (frame);
+ if (call_count == 0) {
+ AFR_STACK_DESTROY (frame);
+ }
+ return 0;
+}
+
+void
+afr_mark_new_entry_changelog (call_frame_t *frame, xlator_t *this)
+{
+ call_frame_t *new_frame = NULL;
+ afr_local_t *local = NULL;
+ afr_local_t *new_local = NULL;
+ afr_private_t *priv = NULL;
+ dict_t **xattr = NULL;
+ int32_t **changelog = NULL;
+ int i = 0;
+ GF_UNUSED int op_errno = 0;
+
+ local = frame->local;
+ priv = this->private;
+
+ new_frame = copy_frame (frame);
+ if (!new_frame) {
+ goto out;
+ }
+
+ AFR_LOCAL_ALLOC_OR_GOTO (new_frame->local, out);
+ new_local = new_frame->local;
+ changelog = afr_matrix_create (priv->child_count, AFR_NUM_CHANGE_LOGS);
+ if (!changelog)
+ goto out;
+
+ xattr = GF_CALLOC (priv->child_count, sizeof (*xattr),
+ gf_afr_mt_dict_t);
+ if (!xattr)
+ goto out;
+ for (i = 0; i < priv->child_count; i++) {
+ if (local->child_errno[i])
+ continue;
+ xattr[i] = dict_new ();
+ if (!xattr[i])
+ goto out;
+ }
+
+ afr_prepare_new_entry_pending_matrix (changelog,
+ afr_is_errno_set,
+ local->child_errno,
+ &local->cont.dir_fop.buf,
+ priv->child_count);
+
+ new_local->pending = changelog;
+ uuid_copy (new_local->loc.gfid, local->cont.dir_fop.buf.ia_gfid);
+ new_local->loc.inode = inode_ref (local->cont.dir_fop.inode);
+ new_local->call_count = local->success_count;
+
+ for (i = 0; i < priv->child_count; i++) {
+ if (local->child_errno[i])
+ continue;
+
+ afr_set_pending_dict (priv, xattr[i], changelog, i, LOCAL_LAST);
+ STACK_WIND_COOKIE (new_frame, afr_mark_new_entry_changelog_cbk,
+ (void *) (long) i, priv->children[i],
+ priv->children[i]->fops->xattrop,
+ &new_local->loc, GF_XATTROP_ADD_ARRAY,
+ xattr[i], NULL);
+ }
+ new_frame = NULL;
+out:
+ if (new_frame)
+ AFR_STACK_DESTROY (new_frame);
+ afr_xattr_array_destroy (xattr, priv->child_count);
+ return;
+}
+
+gf_boolean_t
+afr_is_new_entry_changelog_needed (glusterfs_fop_t fop)
+{
+ glusterfs_fop_t fops[] = {GF_FOP_CREATE, GF_FOP_MKNOD, GF_FOP_NULL};
+ int i = 0;
+
+ for (i = 0; fops[i] != GF_FOP_NULL; i++) {
+ if (fop == fops[i])
+ return _gf_true;
+ }
+ return _gf_false;
+}
+
+void
+afr_dir_fop_mark_entry_pending_changelog (call_frame_t *frame, xlator_t *this)
+{
+ afr_local_t *local = NULL;
+ afr_private_t *priv = NULL;
+
+ local = frame->local;
+ priv = this->private;
+
+ if (local->op_ret < 0)
+ goto out;
+
+ if (local->success_count == priv->child_count)
+ goto out;
+
+ if (!afr_is_new_entry_changelog_needed (local->op))
+ goto out;
+
+ afr_mark_new_entry_changelog (frame, this);
+
+out:
+ return;
+}
+
+void
+afr_dir_fop_done (call_frame_t *frame, xlator_t *this)
+{
+ afr_local_t *local = NULL;
+ afr_private_t *priv = NULL;
+
+ local = frame->local;
+ priv = this->private;
+
+ if (local->cont.dir_fop.inode == NULL)
+ goto done;
+ afr_set_read_ctx_from_policy (this, local->cont.dir_fop.inode,
+ local->fresh_children,
+ local->read_child_index,
+ priv->read_child,
+ local->cont.dir_fop.buf.ia_gfid);
+done:
+ local->transaction.unwind (frame, this);
+ afr_dir_fop_mark_entry_pending_changelog (frame, this);
+ local->transaction.resume (frame, this);
}
/* {{{ create */
@@ -81,7 +263,6 @@ afr_create_unwind (call_frame_t *frame, xlator_t *this)
{
call_frame_t *main_frame = NULL;
afr_local_t *local = NULL;
- struct iatt *unwind_buf = NULL;
local = frame->local;
@@ -95,18 +276,14 @@ afr_create_unwind (call_frame_t *frame, xlator_t *this)
UNLOCK (&frame->lock);
if (main_frame) {
- if (local->cont.create.read_child_buf.ia_ino) {
- unwind_buf = &local->cont.create.read_child_buf;
- } else {
- unwind_buf = &local->cont.create.buf;
- }
-
AFR_STACK_UNWIND (create, main_frame,
local->op_ret, local->op_errno,
local->cont.create.fd,
- local->cont.create.inode,
- unwind_buf, &local->cont.create.preparent,
- &local->cont.create.postparent);
+ local->cont.dir_fop.inode,
+ &local->cont.dir_fop.buf,
+ &local->cont.dir_fop.preparent,
+ &local->cont.dir_fop.postparent,
+ local->xdata_rsp);
}
return 0;
@@ -117,32 +294,24 @@ int
afr_create_wind_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
int32_t op_ret, int32_t op_errno,
fd_t *fd, inode_t *inode, struct iatt *buf,
- struct iatt *preparent, struct iatt *postparent)
+ struct iatt *preparent, struct iatt *postparent,
+ dict_t *xdata)
{
afr_local_t *local = NULL;
- afr_private_t *priv = NULL;
uint64_t ctx = 0;
afr_fd_ctx_t *fd_ctx = NULL;
int ret = 0;
int call_count = -1;
int child_index = -1;
- int32_t *fresh_children = NULL;
local = frame->local;
- priv = this->private;
child_index = (long) cookie;
LOCK (&frame->lock);
{
- if (afr_fop_failed (op_ret, op_errno))
- afr_transaction_fop_failed (frame, this, child_index);
-
- if (op_ret != -1) {
- local->op_ret = op_ret;
-
+ if (op_ret > -1) {
ret = afr_fd_ctx_set (this, fd);
-
if (ret < 0) {
gf_log (this->name, GF_LOG_ERROR,
"could not set ctx on fd=%p", fd);
@@ -153,7 +322,6 @@ afr_create_wind_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
}
ret = fd_ctx_get (fd, this, &ctx);
-
if (ret < 0) {
gf_log (this->name, GF_LOG_ERROR,
"could not get fd ctx for fd=%p", fd);
@@ -167,23 +335,14 @@ afr_create_wind_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
fd_ctx->opened_on[child_index] = AFR_FD_OPENED;
fd_ctx->flags = local->cont.create.flags;
- if (local->success_count == 0)
- local->cont.create.buf = *buf;
-
- if (child_index == local->read_child_index) {
- local->cont.create.read_child_buf = *buf;
- local->cont.create.preparent = *preparent;
- local->cont.create.postparent = *postparent;
- }
-
- local->cont.create.inode = inode;
-
- fresh_children = local->fresh_children;
- fresh_children[local->success_count] = child_index;
- local->success_count++;
+ if (local->success_count == 0) {
+ if (xdata)
+ local->xdata_rsp = dict_ref(xdata);
+ }
}
-
- local->op_errno = op_errno;
+ __dir_entry_fop_common_cbk (frame, child_index, this,
+ op_ret, op_errno, inode, buf,
+ preparent, postparent, NULL, NULL);
}
unlock:
@@ -191,15 +350,8 @@ unlock:
call_count = afr_frame_return (frame);
- if (call_count == 0) {
- afr_set_read_ctx_from_policy (this, inode,
- local->fresh_children,
- local->read_child_index,
- priv->read_child);
- local->transaction.unwind (frame, this);
-
- local->transaction.resume (frame, this);
- }
+ if (call_count == 0)
+ afr_dir_fop_done (frame, this);
return 0;
}
@@ -235,8 +387,9 @@ afr_create_wind (call_frame_t *frame, xlator_t *this)
&local->loc,
local->cont.create.flags,
local->cont.create.mode,
+ local->umask,
local->cont.create.fd,
- local->cont.create.params);
+ local->xdata_req);
if (!--call_count)
break;
}
@@ -264,13 +417,14 @@ afr_create_done (call_frame_t *frame, xlator_t *this)
int
afr_create (call_frame_t *frame, xlator_t *this,
loc_t *loc, int32_t flags, mode_t mode,
- fd_t *fd, dict_t *params)
+ mode_t umask, fd_t *fd, dict_t *params)
{
- afr_private_t *priv = NULL;
- afr_local_t *local = NULL;
- call_frame_t *transaction_frame = NULL;
- int ret = -1;
- int op_errno = 0;
+ afr_private_t *priv = NULL;
+ afr_local_t *local = NULL;
+ afr_internal_lock_t *int_lock = NULL;
+ call_frame_t *transaction_frame = NULL;
+ int ret = -1;
+ int op_errno = 0;
VALIDATE_OR_GOTO (frame, out);
VALIDATE_OR_GOTO (this, out);
@@ -286,7 +440,7 @@ afr_create (call_frame_t *frame, xlator_t *this,
goto out;
}
- ALLOC_OR_GOTO (transaction_frame->local, afr_local_t, out);
+ AFR_LOCAL_ALLOC_OR_GOTO (transaction_frame->local, out);
local = transaction_frame->local;
ret = afr_local_init (local, priv, &op_errno);
@@ -302,22 +456,41 @@ afr_create (call_frame_t *frame, xlator_t *this,
}
UNLOCK (&priv->read_child_lock);
+ local->op = GF_FOP_CREATE;
local->cont.create.flags = flags;
local->cont.create.mode = mode;
local->cont.create.fd = fd_ref (fd);
+ local->umask = umask;
if (params)
- local->cont.create.params = dict_ref (params);
+ local->xdata_req = dict_ref (params);
local->transaction.fop = afr_create_wind;
local->transaction.done = afr_create_done;
local->transaction.unwind = afr_create_unwind;
- afr_build_parent_loc (&local->transaction.parent_loc, loc);
+ ret = afr_build_parent_loc (&local->transaction.parent_loc, loc,
+ &op_errno);
+ if (ret)
+ goto out;
local->transaction.main_frame = frame;
local->transaction.basename = AFR_BASENAME (loc->path);
+ int_lock = &local->internal_lock;
+
+ int_lock->lockee_count = 0;
+ ret = afr_init_entry_lockee (&int_lock->lockee[0], local,
+ &local->transaction.parent_loc,
+ local->transaction.basename,
+ priv->child_count);
+ if (ret)
+ goto out;
- afr_transaction (transaction_frame, this, AFR_ENTRY_TRANSACTION);
+ int_lock->lockee_count++;
+ ret = afr_transaction (transaction_frame, this, AFR_ENTRY_TRANSACTION);
+ if (ret < 0) {
+ op_errno = -ret;
+ goto out;
+ }
ret = 0;
out:
@@ -325,7 +498,7 @@ out:
if (transaction_frame)
AFR_STACK_DESTROY (transaction_frame);
AFR_STACK_UNWIND (create, frame, -1, op_errno,
- NULL, NULL, NULL, NULL, NULL);
+ NULL, NULL, NULL, NULL, NULL, NULL);
}
return 0;
@@ -340,7 +513,6 @@ afr_mknod_unwind (call_frame_t *frame, xlator_t *this)
{
call_frame_t *main_frame = NULL;
afr_local_t *local = NULL;
- struct iatt *unwind_buf = NULL;
local = frame->local;
@@ -354,17 +526,13 @@ afr_mknod_unwind (call_frame_t *frame, xlator_t *this)
UNLOCK (&frame->lock);
if (main_frame) {
- if (local->cont.mknod.read_child_buf.ia_ino) {
- unwind_buf = &local->cont.mknod.read_child_buf;
- } else {
- unwind_buf = &local->cont.mknod.buf;
- }
-
AFR_STACK_UNWIND (mknod, main_frame,
local->op_ret, local->op_errno,
- local->cont.mknod.inode,
- unwind_buf, &local->cont.mknod.preparent,
- &local->cont.mknod.postparent);
+ local->cont.dir_fop.inode,
+ &local->cont.dir_fop.buf,
+ &local->cont.dir_fop.preparent,
+ &local->cont.dir_fop.postparent,
+ NULL);
}
return 0;
@@ -375,58 +543,25 @@ int
afr_mknod_wind_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
int32_t op_ret, int32_t op_errno, inode_t *inode,
struct iatt *buf, struct iatt *preparent,
- struct iatt *postparent)
+ struct iatt *postparent, dict_t *xdata)
{
- afr_local_t *local = NULL;
- afr_private_t *priv = NULL;
int call_count = -1;
int child_index = -1;
- int32_t *fresh_children = NULL;
-
- local = frame->local;
- priv = this->private;
child_index = (long) cookie;
LOCK (&frame->lock);
{
- if (afr_fop_failed (op_ret, op_errno))
- afr_transaction_fop_failed (frame, this, child_index);
-
- if (op_ret != -1) {
- local->op_ret = op_ret;
-
- if (local->success_count == 0)
- local->cont.mknod.buf = *buf;
-
- if (child_index == local->read_child_index) {
- local->cont.mknod.read_child_buf = *buf;
- local->cont.mknod.preparent = *preparent;
- local->cont.mknod.postparent = *postparent;
- }
-
- local->cont.mknod.inode = inode;
-
- fresh_children = local->fresh_children;
- fresh_children[local->success_count] = child_index;
- local->success_count++;
- }
-
- local->op_errno = op_errno;
+ __dir_entry_fop_common_cbk (frame, child_index, this,
+ op_ret, op_errno, inode, buf,
+ preparent, postparent, NULL, NULL);
}
UNLOCK (&frame->lock);
call_count = afr_frame_return (frame);
- if (call_count == 0) {
- afr_set_read_ctx_from_policy (this, inode,
- local->fresh_children,
- local->read_child_index,
- priv->read_child);
- local->transaction.unwind (frame, this);
-
- local->transaction.resume (frame, this);
- }
+ if (call_count == 0)
+ afr_dir_fop_done (frame, this);
return 0;
}
@@ -460,7 +595,8 @@ afr_mknod_wind (call_frame_t *frame, xlator_t *this)
priv->children[i]->fops->mknod,
&local->loc, local->cont.mknod.mode,
local->cont.mknod.dev,
- local->cont.mknod.params);
+ local->umask,
+ local->xdata_req);
if (!--call_count)
break;
}
@@ -485,14 +621,15 @@ afr_mknod_done (call_frame_t *frame, xlator_t *this)
int
-afr_mknod (call_frame_t *frame, xlator_t *this,
- loc_t *loc, mode_t mode, dev_t dev, dict_t *params)
+afr_mknod (call_frame_t *frame, xlator_t *this, loc_t *loc, mode_t mode,
+ dev_t dev, mode_t umask, dict_t *params)
{
- afr_private_t * priv = NULL;
- afr_local_t * local = NULL;
- call_frame_t * transaction_frame = NULL;
- int ret = -1;
- int op_errno = 0;
+ afr_private_t *priv = NULL;
+ afr_local_t *local = NULL;
+ afr_internal_lock_t *int_lock = NULL;
+ call_frame_t *transaction_frame = NULL;
+ int ret = -1;
+ int op_errno = 0;
VALIDATE_OR_GOTO (frame, out);
VALIDATE_OR_GOTO (this, out);
@@ -508,7 +645,7 @@ afr_mknod (call_frame_t *frame, xlator_t *this,
goto out;
}
- ALLOC_OR_GOTO (transaction_frame->local, afr_local_t, out);
+ AFR_LOCAL_ALLOC_OR_GOTO (transaction_frame->local, out);
local = transaction_frame->local;
ret = afr_local_init (local, priv, &op_errno);
@@ -524,21 +661,40 @@ afr_mknod (call_frame_t *frame, xlator_t *this,
}
UNLOCK (&priv->read_child_lock);
+ local->op = GF_FOP_MKNOD;
local->cont.mknod.mode = mode;
local->cont.mknod.dev = dev;
+ local->umask = umask;
if (params)
- local->cont.mknod.params = dict_ref (params);
+ local->xdata_req = dict_ref (params);
local->transaction.fop = afr_mknod_wind;
local->transaction.done = afr_mknod_done;
local->transaction.unwind = afr_mknod_unwind;
- afr_build_parent_loc (&local->transaction.parent_loc, loc);
+ ret = afr_build_parent_loc (&local->transaction.parent_loc, loc,
+ &op_errno);
+ if (ret)
+ goto out;
local->transaction.main_frame = frame;
local->transaction.basename = AFR_BASENAME (loc->path);
+ int_lock = &local->internal_lock;
+
+ int_lock->lockee_count = 0;
+ ret = afr_init_entry_lockee (&int_lock->lockee[0], local,
+ &local->transaction.parent_loc,
+ local->transaction.basename,
+ priv->child_count);
+ if (ret)
+ goto out;
- afr_transaction (transaction_frame, this, AFR_ENTRY_TRANSACTION);
+ int_lock->lockee_count++;
+ ret = afr_transaction (transaction_frame, this, AFR_ENTRY_TRANSACTION);
+ if (ret < 0) {
+ op_errno = -ret;
+ goto out;
+ }
ret = 0;
out:
@@ -546,7 +702,7 @@ out:
if (transaction_frame)
AFR_STACK_DESTROY (transaction_frame);
AFR_STACK_UNWIND (mknod, frame, -1, op_errno,
- NULL, NULL, NULL, NULL);
+ NULL, NULL, NULL, NULL, NULL);
}
return 0;
@@ -562,7 +718,6 @@ afr_mkdir_unwind (call_frame_t *frame, xlator_t *this)
{
call_frame_t *main_frame = NULL;
afr_local_t *local = NULL;
- struct iatt *unwind_buf = NULL;
local = frame->local;
@@ -576,17 +731,13 @@ afr_mkdir_unwind (call_frame_t *frame, xlator_t *this)
UNLOCK (&frame->lock);
if (main_frame) {
- if (local->cont.mkdir.read_child_buf.ia_ino) {
- unwind_buf = &local->cont.mkdir.read_child_buf;
- } else {
- unwind_buf = &local->cont.mkdir.buf;
- }
-
AFR_STACK_UNWIND (mkdir, main_frame,
local->op_ret, local->op_errno,
- local->cont.mkdir.inode,
- unwind_buf, &local->cont.mkdir.preparent,
- &local->cont.mkdir.postparent);
+ local->cont.dir_fop.inode,
+ &local->cont.dir_fop.buf,
+ &local->cont.dir_fop.preparent,
+ &local->cont.dir_fop.postparent,
+ NULL);
}
return 0;
@@ -597,58 +748,25 @@ int
afr_mkdir_wind_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
int32_t op_ret, int32_t op_errno, inode_t *inode,
struct iatt *buf, struct iatt *preparent,
- struct iatt *postparent)
+ struct iatt *postparent, dict_t *xdata)
{
- afr_local_t *local = NULL;
- afr_private_t *priv = NULL;
int call_count = -1;
int child_index = -1;
- int32_t *fresh_children = NULL;
-
- local = frame->local;
- priv = this->private;
child_index = (long) cookie;
LOCK (&frame->lock);
{
- if (afr_fop_failed (op_ret, op_errno))
- afr_transaction_fop_failed (frame, this, child_index);
-
- if (op_ret != -1) {
- local->op_ret = op_ret;
-
- if (local->success_count == 0)
- local->cont.mkdir.buf = *buf;
-
- if (child_index == local->read_child_index) {
- local->cont.mkdir.read_child_buf = *buf;
- local->cont.mkdir.preparent = *preparent;
- local->cont.mkdir.postparent = *postparent;
- }
-
- local->cont.mkdir.inode = inode;
-
- fresh_children = local->fresh_children;
- fresh_children[local->success_count] = child_index;
- local->success_count++;
- }
-
- local->op_errno = op_errno;
+ __dir_entry_fop_common_cbk (frame, child_index, this,
+ op_ret, op_errno, inode, buf,
+ preparent, postparent, NULL, NULL);
}
UNLOCK (&frame->lock);
call_count = afr_frame_return (frame);
- if (call_count == 0) {
- afr_set_read_ctx_from_policy (this, inode,
- local->fresh_children,
- local->read_child_index,
- priv->read_child);
- local->transaction.unwind (frame, this);
-
- local->transaction.resume (frame, this);
- }
+ if (call_count == 0)
+ afr_dir_fop_done (frame, this);
return 0;
}
@@ -682,7 +800,8 @@ afr_mkdir_wind (call_frame_t *frame, xlator_t *this)
priv->children[i],
priv->children[i]->fops->mkdir,
&local->loc, local->cont.mkdir.mode,
- local->cont.mkdir.params);
+ local->umask,
+ local->xdata_req);
if (!--call_count)
break;
}
@@ -706,16 +825,16 @@ afr_mkdir_done (call_frame_t *frame, xlator_t *this)
return 0;
}
-
int
afr_mkdir (call_frame_t *frame, xlator_t *this,
- loc_t *loc, mode_t mode, dict_t *params)
+ loc_t *loc, mode_t mode, mode_t umask, dict_t *params)
{
- afr_private_t * priv = NULL;
- afr_local_t * local = NULL;
- call_frame_t * transaction_frame = NULL;
- int ret = -1;
- int op_errno = 0;
+ afr_private_t *priv = NULL;
+ afr_local_t *local = NULL;
+ afr_internal_lock_t *int_lock = NULL;
+ call_frame_t *transaction_frame = NULL;
+ int ret = -1;
+ int op_errno = 0;
VALIDATE_OR_GOTO (frame, out);
VALIDATE_OR_GOTO (this, out);
@@ -731,7 +850,7 @@ afr_mkdir (call_frame_t *frame, xlator_t *this,
goto out;
}
- ALLOC_OR_GOTO (transaction_frame->local, afr_local_t, out);
+ AFR_LOCAL_ALLOC_OR_GOTO (transaction_frame->local, out);
local = transaction_frame->local;
ret = afr_local_init (local, priv, &op_errno);
@@ -748,19 +867,38 @@ afr_mkdir (call_frame_t *frame, xlator_t *this,
UNLOCK (&priv->read_child_lock);
local->cont.mkdir.mode = mode;
+ local->umask = umask;
if (params)
- local->cont.mkdir.params = dict_ref (params);
+ local->xdata_req = dict_ref (params);
+ local->op = GF_FOP_MKDIR;
local->transaction.fop = afr_mkdir_wind;
local->transaction.done = afr_mkdir_done;
local->transaction.unwind = afr_mkdir_unwind;
- afr_build_parent_loc (&local->transaction.parent_loc, loc);
+ ret = afr_build_parent_loc (&local->transaction.parent_loc, loc,
+ &op_errno);
+ if (ret)
+ goto out;
local->transaction.main_frame = frame;
local->transaction.basename = AFR_BASENAME (loc->path);
+ int_lock = &local->internal_lock;
+
+ int_lock->lockee_count = 0;
+ ret = afr_init_entry_lockee (&int_lock->lockee[0], local,
+ &local->transaction.parent_loc,
+ local->transaction.basename,
+ priv->child_count);
+ if (ret)
+ goto out;
- afr_transaction (transaction_frame, this, AFR_ENTRY_TRANSACTION);
+ int_lock->lockee_count++;
+ ret = afr_transaction (transaction_frame, this, AFR_ENTRY_TRANSACTION);
+ if (ret < 0) {
+ op_errno = -ret;
+ goto out;
+ }
ret = 0;
out:
@@ -769,7 +907,7 @@ out:
AFR_STACK_DESTROY (transaction_frame);
AFR_STACK_UNWIND (mkdir, frame, -1, op_errno,
- NULL, NULL, NULL, NULL);
+ NULL, NULL, NULL, NULL, NULL);
}
return 0;
@@ -785,7 +923,6 @@ afr_link_unwind (call_frame_t *frame, xlator_t *this)
{
call_frame_t *main_frame = NULL;
afr_local_t *local = NULL;
- struct iatt *unwind_buf = NULL;
local = frame->local;
@@ -799,17 +936,13 @@ afr_link_unwind (call_frame_t *frame, xlator_t *this)
UNLOCK (&frame->lock);
if (main_frame) {
- if (local->cont.link.read_child_buf.ia_ino) {
- unwind_buf = &local->cont.link.read_child_buf;
- } else {
- unwind_buf = &local->cont.link.buf;
- }
-
AFR_STACK_UNWIND (link, main_frame,
local->op_ret, local->op_errno,
- local->cont.link.inode,
- unwind_buf, &local->cont.link.preparent,
- &local->cont.link.postparent);
+ local->cont.dir_fop.inode,
+ &local->cont.dir_fop.buf,
+ &local->cont.dir_fop.preparent,
+ &local->cont.dir_fop.postparent,
+ NULL);
}
return 0;
@@ -820,59 +953,25 @@ int
afr_link_wind_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
int32_t op_ret, int32_t op_errno, inode_t *inode,
struct iatt *buf, struct iatt *preparent,
- struct iatt *postparent)
+ struct iatt *postparent, dict_t *xdata)
{
- afr_local_t *local = NULL;
- afr_private_t *priv = NULL;
int call_count = -1;
int child_index = -1;
- int32_t *fresh_children = NULL;
-
- local = frame->local;
- priv = this->private;
child_index = (long) cookie;
LOCK (&frame->lock);
{
- if (afr_fop_failed (op_ret, op_errno))
- afr_transaction_fop_failed (frame, this, child_index);
-
- if (op_ret != -1) {
- local->op_ret = op_ret;
-
- if (local->success_count == 0) {
- local->cont.link.buf = *buf;
- }
-
- if (child_index == local->read_child_index) {
- local->cont.link.read_child_buf = *buf;
- local->cont.link.preparent = *preparent;
- local->cont.link.postparent = *postparent;
- }
-
- local->cont.link.inode = inode;
-
- fresh_children = local->fresh_children;
- fresh_children[local->success_count] = child_index;
- local->success_count++;
- }
-
- local->op_errno = op_errno;
+ __dir_entry_fop_common_cbk (frame, child_index, this,
+ op_ret, op_errno, inode, buf,
+ preparent, postparent, NULL, NULL);
}
UNLOCK (&frame->lock);
call_count = afr_frame_return (frame);
- if (call_count == 0) {
- afr_set_read_ctx_from_policy (this, inode,
- local->fresh_children,
- local->read_child_index,
- priv->read_child);
- local->transaction.unwind (frame, this);
-
- local->transaction.resume (frame, this);
- }
+ if (call_count == 0)
+ afr_dir_fop_done (frame, this);
return 0;
}
@@ -901,11 +1000,12 @@ afr_link_wind (call_frame_t *frame, xlator_t *this)
for (i = 0; i < priv->child_count; i++) {
if (local->transaction.pre_op[i]) {
- STACK_WIND_COOKIE (frame, afr_link_wind_cbk, (void *) (long) i,
+ STACK_WIND_COOKIE (frame, afr_link_wind_cbk,
+ (void *) (long) i,
priv->children[i],
priv->children[i]->fops->link,
&local->loc,
- &local->newloc);
+ &local->newloc, local->xdata_req);
if (!--call_count)
break;
@@ -931,13 +1031,14 @@ afr_link_done (call_frame_t *frame, xlator_t *this)
int
afr_link (call_frame_t *frame, xlator_t *this,
- loc_t *oldloc, loc_t *newloc)
+ loc_t *oldloc, loc_t *newloc, dict_t *xdata)
{
- afr_private_t * priv = NULL;
- afr_local_t * local = NULL;
- call_frame_t * transaction_frame = NULL;
- int ret = -1;
- int op_errno = 0;
+ afr_private_t *priv = NULL;
+ afr_local_t *local = NULL;
+ afr_internal_lock_t *int_lock = NULL;
+ call_frame_t *transaction_frame = NULL;
+ int ret = -1;
+ int op_errno = 0;
VALIDATE_OR_GOTO (frame, out);
VALIDATE_OR_GOTO (this, out);
@@ -953,7 +1054,7 @@ afr_link (call_frame_t *frame, xlator_t *this,
goto out;
}
- ALLOC_OR_GOTO (transaction_frame->local, afr_local_t, out);
+ AFR_LOCAL_ALLOC_OR_GOTO (transaction_frame->local, out);
local = transaction_frame->local;
ret = afr_local_init (local, priv, &op_errno);
@@ -962,6 +1063,8 @@ afr_link (call_frame_t *frame, xlator_t *this,
loc_copy (&local->loc, oldloc);
loc_copy (&local->newloc, newloc);
+ if (xdata)
+ local->xdata_req = dict_ref (xdata);
LOCK (&priv->read_child_lock);
{
@@ -970,24 +1073,41 @@ afr_link (call_frame_t *frame, xlator_t *this,
}
UNLOCK (&priv->read_child_lock);
+ local->op = GF_FOP_LINK;
local->transaction.fop = afr_link_wind;
local->transaction.done = afr_link_done;
local->transaction.unwind = afr_link_unwind;
- afr_build_parent_loc (&local->transaction.parent_loc, newloc);
+ ret = afr_build_parent_loc (&local->transaction.parent_loc, newloc,
+ &op_errno);
+ if (ret)
+ goto out;
local->transaction.main_frame = frame;
local->transaction.basename = AFR_BASENAME (newloc->path);
+ int_lock = &local->internal_lock;
+
+ int_lock->lockee_count = 0;
+ ret = afr_init_entry_lockee (&int_lock->lockee[0], local,
+ &local->transaction.parent_loc,
+ local->transaction.basename,
+ priv->child_count);
+ if (ret)
+ goto out;
- afr_transaction (transaction_frame, this, AFR_ENTRY_TRANSACTION);
-
+ int_lock->lockee_count++;
+ ret = afr_transaction (transaction_frame, this, AFR_ENTRY_TRANSACTION);
+ if (ret < 0) {
+ op_errno = -ret;
+ goto out;
+ }
ret = 0;
out:
if (ret < 0) {
if (transaction_frame)
AFR_STACK_DESTROY (transaction_frame);
AFR_STACK_UNWIND (link, frame, -1, op_errno,
- NULL, NULL, NULL, NULL);
+ NULL, NULL, NULL, NULL, NULL);
}
return 0;
@@ -1003,7 +1123,6 @@ afr_symlink_unwind (call_frame_t *frame, xlator_t *this)
{
call_frame_t *main_frame = NULL;
afr_local_t *local = NULL;
- struct iatt *unwind_buf = NULL;
local = frame->local;
@@ -1017,17 +1136,13 @@ afr_symlink_unwind (call_frame_t *frame, xlator_t *this)
UNLOCK (&frame->lock);
if (main_frame) {
- if (local->cont.symlink.read_child_buf.ia_ino) {
- unwind_buf = &local->cont.symlink.read_child_buf;
- } else {
- unwind_buf = &local->cont.symlink.buf;
- }
-
AFR_STACK_UNWIND (symlink, main_frame,
local->op_ret, local->op_errno,
- local->cont.symlink.inode,
- unwind_buf, &local->cont.symlink.preparent,
- &local->cont.symlink.postparent);
+ local->cont.dir_fop.inode,
+ &local->cont.dir_fop.buf,
+ &local->cont.dir_fop.preparent,
+ &local->cont.dir_fop.postparent,
+ NULL);
}
return 0;
@@ -1038,58 +1153,25 @@ int
afr_symlink_wind_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
int32_t op_ret, int32_t op_errno, inode_t *inode,
struct iatt *buf, struct iatt *preparent,
- struct iatt *postparent)
+ struct iatt *postparent, dict_t *xdata)
{
- afr_local_t *local = NULL;
- afr_private_t *priv = NULL;
int call_count = -1;
int child_index = -1;
- int32_t *fresh_children = NULL;
-
- local = frame->local;
- priv = this->private;
child_index = (long) cookie;
LOCK (&frame->lock);
{
- if (afr_fop_failed (op_ret, op_errno))
- afr_transaction_fop_failed (frame, this, child_index);
-
- if (op_ret != -1) {
- local->op_ret = op_ret;
-
- if (local->success_count == 0)
- local->cont.symlink.buf = *buf;
-
- if (child_index == local->read_child_index) {
- local->cont.symlink.read_child_buf = *buf;
- local->cont.symlink.preparent = *preparent;
- local->cont.symlink.postparent = *postparent;
- }
-
- local->cont.symlink.inode = inode;
-
- fresh_children = local->fresh_children;
- fresh_children[local->success_count] = child_index;
- local->success_count++;
- }
-
- local->op_errno = op_errno;
+ __dir_entry_fop_common_cbk (frame, child_index, this,
+ op_ret, op_errno, inode, buf,
+ preparent, postparent, NULL, NULL);
}
UNLOCK (&frame->lock);
call_count = afr_frame_return (frame);
- if (call_count == 0) {
- afr_set_read_ctx_from_policy (this, inode,
- local->fresh_children,
- local->read_child_index,
- priv->read_child);
- local->transaction.unwind (frame, this);
-
- local->transaction.resume (frame, this);
- }
+ if (call_count == 0)
+ afr_dir_fop_done (frame, this);
return 0;
}
@@ -1124,7 +1206,8 @@ afr_symlink_wind (call_frame_t *frame, xlator_t *this)
priv->children[i]->fops->symlink,
local->cont.symlink.linkpath,
&local->loc,
- local->cont.symlink.params);
+ local->umask,
+ local->xdata_req);
if (!--call_count)
break;
@@ -1151,13 +1234,14 @@ afr_symlink_done (call_frame_t *frame, xlator_t *this)
int
afr_symlink (call_frame_t *frame, xlator_t *this,
- const char *linkpath, loc_t *loc, dict_t *params)
+ const char *linkpath, loc_t *loc, mode_t umask, dict_t *params)
{
- afr_private_t * priv = NULL;
- afr_local_t * local = NULL;
- call_frame_t * transaction_frame = NULL;
- int ret = -1;
- int op_errno = 0;
+ afr_private_t *priv = NULL;
+ afr_local_t *local = NULL;
+ afr_internal_lock_t *int_lock = NULL;
+ call_frame_t *transaction_frame = NULL;
+ int ret = -1;
+ int op_errno = 0;
VALIDATE_OR_GOTO (frame, out);
VALIDATE_OR_GOTO (this, out);
@@ -1173,7 +1257,7 @@ afr_symlink (call_frame_t *frame, xlator_t *this,
goto out;
}
- ALLOC_OR_GOTO (transaction_frame->local, afr_local_t, out);
+ AFR_LOCAL_ALLOC_OR_GOTO (transaction_frame->local, out);
local = transaction_frame->local;
ret = afr_local_init (local, priv, &op_errno);
@@ -1190,19 +1274,38 @@ afr_symlink (call_frame_t *frame, xlator_t *this,
UNLOCK (&priv->read_child_lock);
local->cont.symlink.linkpath = gf_strdup (linkpath);
+ local->umask = umask;
if (params)
- local->cont.symlink.params = dict_ref (params);
+ local->xdata_req = dict_ref (params);
+ local->op = GF_FOP_SYMLINK;
local->transaction.fop = afr_symlink_wind;
local->transaction.done = afr_symlink_done;
local->transaction.unwind = afr_symlink_unwind;
- afr_build_parent_loc (&local->transaction.parent_loc, loc);
+ ret = afr_build_parent_loc (&local->transaction.parent_loc, loc,
+ &op_errno);
+ if (ret)
+ goto out;
local->transaction.main_frame = frame;
local->transaction.basename = AFR_BASENAME (loc->path);
+ int_lock = &local->internal_lock;
+
+ int_lock->lockee_count = 0;
+ ret = afr_init_entry_lockee (&int_lock->lockee[0], local,
+ &local->transaction.parent_loc,
+ local->transaction.basename,
+ priv->child_count);
+ if (ret)
+ goto out;
- afr_transaction (transaction_frame, this, AFR_ENTRY_TRANSACTION);
+ int_lock->lockee_count++;
+ ret = afr_transaction (transaction_frame, this, AFR_ENTRY_TRANSACTION);
+ if (ret < 0) {
+ op_errno = -ret;
+ goto out;
+ }
ret = 0;
out:
@@ -1210,7 +1313,7 @@ out:
if (transaction_frame)
AFR_STACK_DESTROY (transaction_frame);
AFR_STACK_UNWIND (symlink, frame, -1, op_errno,
- NULL, NULL, NULL, NULL);
+ NULL, NULL, NULL, NULL, NULL);
}
return 0;
@@ -1225,7 +1328,6 @@ afr_rename_unwind (call_frame_t *frame, xlator_t *this)
{
call_frame_t *main_frame = NULL;
afr_local_t *local = NULL;
- struct iatt *unwind_buf = NULL;
local = frame->local;
@@ -1239,19 +1341,14 @@ afr_rename_unwind (call_frame_t *frame, xlator_t *this)
UNLOCK (&frame->lock);
if (main_frame) {
- if (local->cont.rename.read_child_buf.ia_ino) {
- unwind_buf = &local->cont.rename.read_child_buf;
- } else {
- unwind_buf = &local->cont.rename.buf;
- }
-
AFR_STACK_UNWIND (rename, main_frame,
local->op_ret, local->op_errno,
- unwind_buf,
- &local->cont.rename.preoldparent,
- &local->cont.rename.postoldparent,
- &local->cont.rename.prenewparent,
- &local->cont.rename.postnewparent);
+ &local->cont.dir_fop.buf,
+ &local->cont.dir_fop.preparent,
+ &local->cont.dir_fop.postparent,
+ &local->cont.dir_fop.prenewparent,
+ &local->cont.dir_fop.postnewparent,
+ NULL);
}
return 0;
@@ -1262,7 +1359,8 @@ int
afr_rename_wind_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
int32_t op_ret, int32_t op_errno, struct iatt *buf,
struct iatt *preoldparent, struct iatt *postoldparent,
- struct iatt *prenewparent, struct iatt *postnewparent)
+ struct iatt *prenewparent, struct iatt *postnewparent,
+ dict_t *xdata)
{
afr_local_t * local = NULL;
int call_count = -1;
@@ -1276,38 +1374,22 @@ afr_rename_wind_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
{
if (afr_fop_failed (op_ret, op_errno) && op_errno != ENOTEMPTY)
afr_transaction_fop_failed (frame, this, child_index);
+ local->op_errno = op_errno;
+ local->child_errno[child_index] = op_errno;
- if (op_ret != -1) {
- if (local->success_count == 0) {
- local->op_ret = op_ret;
-
- if (buf) {
- local->cont.rename.buf = *buf;
- }
-
- local->success_count++;
- }
-
- if (child_index == local->read_child_index) {
- local->cont.rename.read_child_buf = *buf;
+ if (op_ret > -1)
+ __dir_entry_fop_common_cbk (frame, child_index, this,
+ op_ret, op_errno, NULL, buf,
+ preoldparent, postoldparent,
+ prenewparent, postnewparent);
- local->cont.rename.preoldparent = *preoldparent;
- local->cont.rename.postoldparent = *postoldparent;
- local->cont.rename.prenewparent = *prenewparent;
- local->cont.rename.postnewparent = *postnewparent;
- }
- }
-
- local->op_errno = op_errno;
}
UNLOCK (&frame->lock);
call_count = afr_frame_return (frame);
- if (call_count == 0) {
- local->transaction.unwind (frame, this);
- local->transaction.resume (frame, this);
- }
+ if (call_count == 0)
+ afr_dir_fop_done (frame, this);
return 0;
}
@@ -1341,7 +1423,7 @@ afr_rename_wind (call_frame_t *frame, xlator_t *this)
priv->children[i],
priv->children[i]->fops->rename,
&local->loc,
- &local->newloc);
+ &local->newloc, NULL);
if (!--call_count)
break;
}
@@ -1366,13 +1448,15 @@ afr_rename_done (call_frame_t *frame, xlator_t *this)
int
afr_rename (call_frame_t *frame, xlator_t *this,
- loc_t *oldloc, loc_t *newloc)
+ loc_t *oldloc, loc_t *newloc, dict_t *xdata)
{
- afr_private_t * priv = NULL;
- afr_local_t * local = NULL;
- call_frame_t * transaction_frame = NULL;
- int ret = -1;
- int op_errno = 0;
+ afr_private_t *priv = NULL;
+ afr_local_t *local = NULL;
+ afr_internal_lock_t *int_lock = NULL;
+ call_frame_t *transaction_frame = NULL;
+ int ret = -1;
+ int op_errno = 0;
+ int nlockee = 0;
VALIDATE_OR_GOTO (frame, out);
VALIDATE_OR_GOTO (this, out);
@@ -1388,7 +1472,7 @@ afr_rename (call_frame_t *frame, xlator_t *this,
goto out;
}
- ALLOC_OR_GOTO (transaction_frame->local, afr_local_t, out);
+ AFR_LOCAL_ALLOC_OR_GOTO (transaction_frame->local, out);
local = transaction_frame->local;
ret = afr_local_init (local, priv, &op_errno);
@@ -1400,18 +1484,61 @@ afr_rename (call_frame_t *frame, xlator_t *this,
local->read_child_index = afr_inode_get_read_ctx (this, oldloc->inode, NULL);
+ local->op = GF_FOP_RENAME;
local->transaction.fop = afr_rename_wind;
local->transaction.done = afr_rename_done;
local->transaction.unwind = afr_rename_unwind;
- afr_build_parent_loc (&local->transaction.parent_loc, oldloc);
- afr_build_parent_loc (&local->transaction.new_parent_loc, newloc);
+ ret = afr_build_parent_loc (&local->transaction.parent_loc, oldloc,
+ &op_errno);
+ if (ret)
+ goto out;
+ ret = afr_build_parent_loc (&local->transaction.new_parent_loc, newloc,
+ &op_errno);
+ if (ret)
+ goto out;
local->transaction.main_frame = frame;
local->transaction.basename = AFR_BASENAME (oldloc->path);
local->transaction.new_basename = AFR_BASENAME (newloc->path);
+ int_lock = &local->internal_lock;
+
+ int_lock->lockee_count = nlockee = 0;
+ ret = afr_init_entry_lockee (&int_lock->lockee[nlockee], local,
+ &local->transaction.new_parent_loc,
+ local->transaction.new_basename,
+ priv->child_count);
+ if (ret)
+ goto out;
+
+ nlockee++;
+ ret = afr_init_entry_lockee (&int_lock->lockee[nlockee], local,
+ &local->transaction.parent_loc,
+ local->transaction.basename,
+ priv->child_count);
+ if (ret)
+ goto out;
+
+ nlockee++;
+ if (local->newloc.inode && IA_ISDIR (local->newloc.inode->ia_type)) {
+ ret = afr_init_entry_lockee (&int_lock->lockee[nlockee], local,
+ &local->newloc,
+ NULL,
+ priv->child_count);
+ if (ret)
+ goto out;
+
+ nlockee++;
+ }
+ qsort (int_lock->lockee, nlockee, sizeof (*int_lock->lockee),
+ afr_entry_lockee_cmp);
+ int_lock->lockee_count = nlockee;
- afr_transaction (transaction_frame, this, AFR_ENTRY_RENAME_TRANSACTION);
+ ret = afr_transaction (transaction_frame, this, AFR_ENTRY_RENAME_TRANSACTION);
+ if (ret < 0) {
+ op_errno = -ret;
+ goto out;
+ }
ret = 0;
out:
@@ -1420,7 +1547,7 @@ out:
AFR_STACK_DESTROY (transaction_frame);
AFR_STACK_UNWIND (rename, frame, -1, op_errno,
- NULL, NULL, NULL, NULL, NULL);
+ NULL, NULL, NULL, NULL, NULL, NULL);
}
return 0;
@@ -1450,8 +1577,9 @@ afr_unlink_unwind (call_frame_t *frame, xlator_t *this)
if (main_frame) {
AFR_STACK_UNWIND (unlink, main_frame,
local->op_ret, local->op_errno,
- &local->cont.unlink.preparent,
- &local->cont.unlink.postparent);
+ &local->cont.dir_fop.preparent,
+ &local->cont.dir_fop.postparent,
+ NULL);
}
return 0;
@@ -1461,7 +1589,7 @@ afr_unlink_unwind (call_frame_t *frame, xlator_t *this)
int
afr_unlink_wind_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
int32_t op_ret, int32_t op_errno, struct iatt *preparent,
- struct iatt *postparent)
+ struct iatt *postparent, dict_t *xdata)
{
afr_local_t * local = NULL;
int call_count = -1;
@@ -1474,36 +1602,15 @@ afr_unlink_wind_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
if (child_index == local->read_child_index) {
local->read_child_returned = _gf_true;
}
-
- if (afr_fop_failed (op_ret, op_errno))
- afr_transaction_fop_failed (frame, this, child_index);
-
- if (op_ret != -1) {
- if (local->success_count == 0) {
- local->op_ret = op_ret;
- local->cont.unlink.preparent = *preparent;
- local->cont.unlink.postparent = *postparent;
- }
-
- if (child_index == local->read_child_index) {
- local->cont.unlink.preparent = *preparent;
- local->cont.unlink.postparent = *postparent;
- }
-
- local->success_count++;
- }
-
- local->op_errno = op_errno;
+ __dir_entry_fop_common_cbk (frame, child_index, this,
+ op_ret, op_errno, NULL, NULL,
+ preparent, postparent, NULL, NULL);
}
UNLOCK (&frame->lock);
call_count = afr_frame_return (frame);
-
- if (call_count == 0) {
- local->transaction.unwind (frame, this);
-
- local->transaction.resume (frame, this);
- }
+ if (call_count == 0)
+ afr_dir_fop_done (frame, this);
return 0;
}
@@ -1536,7 +1643,8 @@ afr_unlink_wind (call_frame_t *frame, xlator_t *this)
(void *) (long) i,
priv->children[i],
priv->children[i]->fops->unlink,
- &local->loc);
+ &local->loc, local->xflag,
+ local->xdata_req);
if (!--call_count)
break;
@@ -1562,13 +1670,14 @@ afr_unlink_done (call_frame_t *frame, xlator_t *this)
int32_t
afr_unlink (call_frame_t *frame, xlator_t *this,
- loc_t *loc)
+ loc_t *loc, int xflag, dict_t *xdata)
{
- afr_private_t * priv = NULL;
- afr_local_t * local = NULL;
- call_frame_t * transaction_frame = NULL;
- int ret = -1;
- int op_errno = 0;
+ afr_private_t *priv = NULL;
+ afr_local_t *local = NULL;
+ afr_internal_lock_t *int_lock = NULL;
+ call_frame_t *transaction_frame = NULL;
+ int ret = -1;
+ int op_errno = 0;
VALIDATE_OR_GOTO (frame, out);
VALIDATE_OR_GOTO (this, out);
@@ -1584,7 +1693,7 @@ afr_unlink (call_frame_t *frame, xlator_t *this,
goto out;
}
- ALLOC_OR_GOTO (transaction_frame->local, afr_local_t, out);
+ AFR_LOCAL_ALLOC_OR_GOTO (transaction_frame->local, out);
local = transaction_frame->local;
ret = afr_local_init (local, priv, &op_errno);
@@ -1592,17 +1701,38 @@ afr_unlink (call_frame_t *frame, xlator_t *this,
goto out;
loc_copy (&local->loc, loc);
+ local->xflag = xflag;
+ if (xdata)
+ local->xdata_req = dict_ref (xdata);
+ local->op = GF_FOP_UNLINK;
local->transaction.fop = afr_unlink_wind;
local->transaction.done = afr_unlink_done;
local->transaction.unwind = afr_unlink_unwind;
- afr_build_parent_loc (&local->transaction.parent_loc, loc);
+ ret = afr_build_parent_loc (&local->transaction.parent_loc, loc,
+ &op_errno);
+ if (ret)
+ goto out;
local->transaction.main_frame = frame;
local->transaction.basename = AFR_BASENAME (loc->path);
+ int_lock = &local->internal_lock;
+
+ int_lock->lockee_count = 0;
+ ret = afr_init_entry_lockee (&int_lock->lockee[0], local,
+ &local->transaction.parent_loc,
+ local->transaction.basename,
+ priv->child_count);
+ if (ret)
+ goto out;
- afr_transaction (transaction_frame, this, AFR_ENTRY_TRANSACTION);
+ int_lock->lockee_count++;
+ ret = afr_transaction (transaction_frame, this, AFR_ENTRY_TRANSACTION);
+ if (ret < 0) {
+ op_errno = -ret;
+ goto out;
+ }
ret = 0;
out:
@@ -1610,7 +1740,7 @@ out:
if (transaction_frame)
AFR_STACK_DESTROY (transaction_frame);
AFR_STACK_UNWIND (unlink, frame, -1, op_errno,
- NULL, NULL);
+ NULL, NULL, NULL);
}
return 0;
@@ -1642,8 +1772,9 @@ afr_rmdir_unwind (call_frame_t *frame, xlator_t *this)
if (main_frame) {
AFR_STACK_UNWIND (rmdir, main_frame,
local->op_ret, local->op_errno,
- &local->cont.rmdir.preparent,
- &local->cont.rmdir.postparent);
+ &local->cont.dir_fop.preparent,
+ &local->cont.dir_fop.postparent,
+ NULL);
}
return 0;
@@ -1653,7 +1784,7 @@ afr_rmdir_unwind (call_frame_t *frame, xlator_t *this)
int
afr_rmdir_wind_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
int32_t op_ret, int32_t op_errno, struct iatt *preparent,
- struct iatt *postparent)
+ struct iatt *postparent, dict_t *xdata)
{
afr_local_t * local = NULL;
int call_count = -1;
@@ -1667,36 +1798,22 @@ afr_rmdir_wind_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
if (child_index == read_child) {
local->read_child_returned = _gf_true;
}
-
if (afr_fop_failed (op_ret, op_errno) && (op_errno != ENOTEMPTY))
afr_transaction_fop_failed (frame, this, child_index);
-
- if (op_ret != -1) {
- if (local->success_count == 0) {
- local->op_ret = op_ret;
- local->cont.rmdir.preparent = *preparent;
- local->cont.rmdir.postparent = *postparent;
-
- }
-
- if (child_index == read_child) {
- local->cont.rmdir.preparent = *preparent;
- local->cont.rmdir.postparent = *postparent;
- }
-
- local->success_count++;
- }
-
local->op_errno = op_errno;
+ local->child_errno[child_index] = op_errno;
+ if (op_ret > -1)
+ __dir_entry_fop_common_cbk (frame, child_index, this,
+ op_ret, op_errno, NULL, NULL,
+ preparent, postparent, NULL,
+ NULL);
+
}
UNLOCK (&frame->lock);
call_count = afr_frame_return (frame);
-
- if (call_count == 0) {
- local->transaction.unwind (frame, this);
- local->transaction.resume (frame, this);
- }
+ if (call_count == 0)
+ afr_dir_fop_done (frame, this);
return 0;
}
@@ -1729,7 +1846,8 @@ afr_rmdir_wind (call_frame_t *frame, xlator_t *this)
(void *) (long) i,
priv->children[i],
priv->children[i]->fops->rmdir,
- &local->loc, local->cont.rmdir.flags);
+ &local->loc, local->cont.rmdir.flags,
+ NULL);
if (!--call_count)
break;
@@ -1755,13 +1873,15 @@ afr_rmdir_done (call_frame_t *frame, xlator_t *this)
int
afr_rmdir (call_frame_t *frame, xlator_t *this,
- loc_t *loc, int flags)
+ loc_t *loc, int flags, dict_t *xdata)
{
- afr_private_t * priv = NULL;
- afr_local_t * local = NULL;
- call_frame_t * transaction_frame = NULL;
- int ret = -1;
- int op_errno = 0;
+ afr_private_t *priv = NULL;
+ afr_local_t *local = NULL;
+ afr_internal_lock_t *int_lock = NULL;
+ call_frame_t *transaction_frame = NULL;
+ int ret = -1;
+ int op_errno = 0;
+ int nlockee = 0;
VALIDATE_OR_GOTO (frame, out);
VALIDATE_OR_GOTO (this, out);
@@ -1777,7 +1897,7 @@ afr_rmdir (call_frame_t *frame, xlator_t *this,
goto out;
}
- ALLOC_OR_GOTO (transaction_frame->local, afr_local_t, out);
+ AFR_LOCAL_ALLOC_OR_GOTO (transaction_frame->local, out);
local = transaction_frame->local;
ret = afr_local_init (local, priv, &op_errno);
@@ -1787,23 +1907,53 @@ afr_rmdir (call_frame_t *frame, xlator_t *this,
local->cont.rmdir.flags = flags;
loc_copy (&local->loc, loc);
+ local->op = GF_FOP_RMDIR;
local->transaction.fop = afr_rmdir_wind;
local->transaction.done = afr_rmdir_done;
local->transaction.unwind = afr_rmdir_unwind;
- afr_build_parent_loc (&local->transaction.parent_loc, loc);
+ ret = afr_build_parent_loc (&local->transaction.parent_loc, loc,
+ &op_errno);
+ if (ret)
+ goto out;
local->transaction.main_frame = frame;
local->transaction.basename = AFR_BASENAME (loc->path);
+ int_lock = &local->internal_lock;
+
+ int_lock->lockee_count = nlockee = 0;
+ ret = afr_init_entry_lockee (&int_lock->lockee[nlockee], local,
+ &local->transaction.parent_loc,
+ local->transaction.basename,
+ priv->child_count);
+ if (ret)
+ goto out;
- afr_transaction (transaction_frame, this, AFR_ENTRY_TRANSACTION);
+ nlockee++;
+ ret = afr_init_entry_lockee (&int_lock->lockee[nlockee], local,
+ &local->loc,
+ NULL,
+ priv->child_count);
+ if (ret)
+ goto out;
+
+ nlockee++;
+ qsort (int_lock->lockee, nlockee, sizeof (*int_lock->lockee),
+ afr_entry_lockee_cmp);
+ int_lock->lockee_count = nlockee;
+
+ ret = afr_transaction (transaction_frame, this, AFR_ENTRY_TRANSACTION);
+ if (ret < 0) {
+ op_errno = -ret;
+ goto out;
+ }
ret = 0;
out:
if (ret < 0) {
if (transaction_frame)
AFR_STACK_DESTROY (transaction_frame);
- AFR_STACK_UNWIND (rmdir, frame, -1, op_errno, NULL, NULL);
+ AFR_STACK_UNWIND (rmdir, frame, -1, op_errno, NULL, NULL, NULL);
}
return 0;
diff --git a/xlators/cluster/afr/src/afr-dir-write.h b/xlators/cluster/afr/src/afr-dir-write.h
index 0290c6350..02f0a3682 100644
--- a/xlators/cluster/afr/src/afr-dir-write.h
+++ b/xlators/cluster/afr/src/afr-dir-write.h
@@ -1,20 +1,11 @@
/*
- Copyright (c) 2007-2011 Gluster, Inc. <http://www.gluster.com>
- This file is part of GlusterFS.
-
- GlusterFS is free software; you can redistribute it and/or modify
- it under the terms of the GNU General Public License as published
- by the Free Software Foundation; either version 3 of the License,
- or (at your option) any later version.
-
- GlusterFS is distributed in the hope that it will be useful, but
- WITHOUT ANY WARRANTY; without even the implied warranty of
- MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU
- General Public License for more details.
-
- You should have received a copy of the GNU General Public License
- along with this program. If not, see
- <http://www.gnu.org/licenses/>.
+ Copyright (c) 2008-2012 Red Hat, Inc. <http://www.redhat.com>
+ This file is part of GlusterFS.
+
+ This file is licensed to you under your choice of the GNU Lesser
+ General Public License, version 3 or any later version (LGPLv3 or
+ later), or the GNU General Public License, version 2 (GPLv2), in all
+ cases as published by the Free Software Foundation.
*/
#ifndef __DIR_WRITE_H__
@@ -23,38 +14,34 @@
int32_t
afr_create (call_frame_t *frame, xlator_t *this,
loc_t *loc, int32_t flags, mode_t mode,
- fd_t *fd, dict_t *params);
+ mode_t umask, fd_t *fd, dict_t *xdata);
int32_t
afr_mknod (call_frame_t *frame, xlator_t *this,
- loc_t *loc, mode_t mode, dev_t dev, dict_t *params);
+ loc_t *loc, mode_t mode, dev_t dev, mode_t umask, dict_t *xdata);
int32_t
afr_mkdir (call_frame_t *frame, xlator_t *this,
- loc_t *loc, mode_t mode, dict_t *params);
+ loc_t *loc, mode_t mode, mode_t umask, dict_t *xdata);
int32_t
afr_unlink (call_frame_t *frame, xlator_t *this,
- loc_t *loc);
+ loc_t *loc, int xflag, dict_t *xdata);
int32_t
afr_rmdir (call_frame_t *frame, xlator_t *this,
- loc_t *loc, int flags);
+ loc_t *loc, int flags, dict_t *xdata);
int32_t
afr_link (call_frame_t *frame, xlator_t *this,
- loc_t *oldloc, loc_t *newloc);
+ loc_t *oldloc, loc_t *newloc, dict_t *xdata);
int32_t
afr_rename (call_frame_t *frame, xlator_t *this,
- loc_t *oldloc, loc_t *newloc);
+ loc_t *oldloc, loc_t *newloc, dict_t *xdata);
int
afr_symlink (call_frame_t *frame, xlator_t *this,
- const char *linkpath, loc_t *oldloc, dict_t *params);
-
-int32_t
-afr_setdents (call_frame_t *frame, xlator_t *this,
- fd_t *fd, int32_t flags, dir_entry_t *entries, int32_t count);
+ const char *linkpath, loc_t *oldloc, mode_t umask, dict_t *params);
#endif /* __DIR_WRITE_H__ */
diff --git a/xlators/cluster/afr/src/afr-inode-read.c b/xlators/cluster/afr/src/afr-inode-read.c
index 162c7eed4..e06e3b2f2 100644
--- a/xlators/cluster/afr/src/afr-inode-read.c
+++ b/xlators/cluster/afr/src/afr-inode-read.c
@@ -1,20 +1,11 @@
/*
- Copyright (c) 2007-2011 Gluster, Inc. <http://www.gluster.com>
+ Copyright (c) 2008-2012 Red Hat, Inc. <http://www.redhat.com>
This file is part of GlusterFS.
- GlusterFS is free software; you can redistribute it and/or modify
- it under the terms of the GNU General Public License as published
- by the Free Software Foundation; either version 3 of the License,
- or (at your option) any later version.
-
- GlusterFS is distributed in the hope that it will be useful, but
- WITHOUT ANY WARRANTY; without even the implied warranty of
- MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU
- General Public License for more details.
-
- You should have received a copy of the GNU General Public License
- along with this program. If not, see
- <http://www.gnu.org/licenses/>.
+ This file is licensed to you under your choice of the GNU Lesser
+ General Public License, version 3 or any later version (LGPLv3 or
+ later), or the GNU General Public License, version 2 (GPLv2), in all
+ cases as published by the Free Software Foundation.
*/
@@ -58,7 +49,7 @@
int32_t
afr_access_cbk (call_frame_t *frame, void *cookie,
- xlator_t *this, int32_t op_ret, int32_t op_errno)
+ xlator_t *this, int32_t op_ret, int32_t op_errno, dict_t *xdata)
{
afr_private_t * priv = NULL;
afr_local_t * local = NULL;
@@ -92,12 +83,13 @@ afr_access_cbk (call_frame_t *frame, void *cookie,
(void *) (long) read_child,
children[next_call_child],
children[next_call_child]->fops->access,
- &local->loc, local->cont.access.mask);
+ &local->loc, local->cont.access.mask,
+ NULL);
}
out:
if (unwind) {
- AFR_STACK_UNWIND (access, frame, op_ret, op_errno);
+ AFR_STACK_UNWIND (access, frame, op_ret, op_errno, xdata);
}
return 0;
@@ -105,7 +97,8 @@ out:
int32_t
-afr_access (call_frame_t *frame, xlator_t *this, loc_t *loc, int32_t mask)
+afr_access (call_frame_t *frame, xlator_t *this, loc_t *loc, int32_t mask,
+ dict_t *xdata)
{
afr_private_t *priv = NULL;
xlator_t **children = NULL;
@@ -124,7 +117,9 @@ afr_access (call_frame_t *frame, xlator_t *this, loc_t *loc, int32_t mask)
children = priv->children;
- ALLOC_OR_GOTO (frame->local, afr_local_t, out);
+ AFR_SBRAIN_CHECK_LOC (loc, out);
+
+ AFR_LOCAL_ALLOC_OR_GOTO (frame->local, out);
local = frame->local;
ret = afr_local_init (local, priv, &op_errno);
@@ -156,12 +151,12 @@ afr_access (call_frame_t *frame, xlator_t *this, loc_t *loc, int32_t mask)
(void *) (long) call_child,
children[call_child],
children[call_child]->fops->access,
- loc, mask);
+ loc, mask, xdata);
ret = 0;
out:
if (ret < 0)
- AFR_STACK_UNWIND (access, frame, -1, op_errno);
+ AFR_STACK_UNWIND (access, frame, -1, op_errno, NULL);
return 0;
}
@@ -173,7 +168,7 @@ out:
int32_t
afr_stat_cbk (call_frame_t *frame, void *cookie,
xlator_t *this, int32_t op_ret, int32_t op_errno,
- struct iatt *buf)
+ struct iatt *buf, dict_t *xdata)
{
afr_private_t * priv = NULL;
afr_local_t * local = NULL;
@@ -207,12 +202,12 @@ afr_stat_cbk (call_frame_t *frame, void *cookie,
(void *) (long) read_child,
children[next_call_child],
children[next_call_child]->fops->stat,
- &local->loc);
+ &local->loc, NULL);
}
out:
if (unwind) {
- AFR_STACK_UNWIND (stat, frame, op_ret, op_errno, buf);
+ AFR_STACK_UNWIND (stat, frame, op_ret, op_errno, buf, xdata);
}
return 0;
@@ -220,7 +215,7 @@ out:
int32_t
-afr_stat (call_frame_t *frame, xlator_t *this, loc_t *loc)
+afr_stat (call_frame_t *frame, xlator_t *this, loc_t *loc, dict_t *xdata)
{
afr_private_t *priv = NULL;
afr_local_t *local = NULL;
@@ -239,7 +234,9 @@ afr_stat (call_frame_t *frame, xlator_t *this, loc_t *loc)
children = priv->children;
- ALLOC_OR_GOTO (frame->local, afr_local_t, out);
+ AFR_SBRAIN_CHECK_LOC (loc, out);
+
+ AFR_LOCAL_ALLOC_OR_GOTO (frame->local, out);
local = frame->local;
ret = afr_local_init (local, priv, &op_errno);
@@ -267,12 +264,12 @@ afr_stat (call_frame_t *frame, xlator_t *this, loc_t *loc)
STACK_WIND_COOKIE (frame, afr_stat_cbk, (void *) (long) call_child,
children[call_child],
children[call_child]->fops->stat,
- loc);
+ loc, xdata);
ret = 0;
out:
if (ret < 0)
- AFR_STACK_UNWIND (stat, frame, -1, op_errno, NULL);
+ AFR_STACK_UNWIND (stat, frame, -1, op_errno, NULL, NULL);
return 0;
}
@@ -284,7 +281,8 @@ out:
int32_t
afr_fstat_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
- int32_t op_ret, int32_t op_errno, struct iatt *buf)
+ int32_t op_ret, int32_t op_errno, struct iatt *buf,
+ dict_t *xdata)
{
afr_private_t *priv = NULL;
afr_local_t *local = NULL;
@@ -318,12 +316,12 @@ afr_fstat_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
(void *) (long) read_child,
children[next_call_child],
children[next_call_child]->fops->fstat,
- local->fd);
+ local->fd, NULL);
}
out:
if (unwind) {
- AFR_STACK_UNWIND (fstat, frame, op_ret, op_errno, buf);
+ AFR_STACK_UNWIND (fstat, frame, op_ret, op_errno, buf, xdata);
}
return 0;
@@ -332,7 +330,7 @@ out:
int32_t
afr_fstat (call_frame_t *frame, xlator_t *this,
- fd_t *fd)
+ fd_t *fd, dict_t *xdata)
{
afr_private_t *priv = NULL;
afr_local_t *local = NULL;
@@ -354,7 +352,9 @@ afr_fstat (call_frame_t *frame, xlator_t *this,
VALIDATE_OR_GOTO (fd->inode, out);
- ALLOC_OR_GOTO (frame->local, afr_local_t, out);
+ AFR_SBRAIN_CHECK_FD (fd, out);
+
+ AFR_LOCAL_ALLOC_OR_GOTO (frame->local, out);
local = frame->local;
ret = afr_local_init (local, priv, &op_errno);
@@ -383,20 +383,17 @@ afr_fstat (call_frame_t *frame, xlator_t *this,
local->fd = fd_ref (fd);
- ret = afr_open_fd_fix (frame, this, _gf_false);
- if (ret) {
- op_errno = -ret;
- goto out;
- }
+ afr_open_fd_fix (fd, this);
+
STACK_WIND_COOKIE (frame, afr_fstat_cbk, (void *) (long) call_child,
children[call_child],
children[call_child]->fops->fstat,
- fd);
+ fd, xdata);
ret = 0;
out:
if (ret < 0)
- AFR_STACK_UNWIND (fstat, frame, -1, op_errno, NULL);
+ AFR_STACK_UNWIND (fstat, frame, -1, op_errno, NULL, NULL);
return 0;
}
@@ -408,7 +405,7 @@ out:
int32_t
afr_readlink_cbk (call_frame_t *frame, void *cookie,
xlator_t *this, int32_t op_ret, int32_t op_errno,
- const char *buf, struct iatt *sbuf)
+ const char *buf, struct iatt *sbuf, dict_t *xdata)
{
afr_private_t * priv = NULL;
afr_local_t * local = NULL;
@@ -442,12 +439,13 @@ afr_readlink_cbk (call_frame_t *frame, void *cookie,
children[next_call_child],
children[next_call_child]->fops->readlink,
&local->loc,
- local->cont.readlink.size);
+ local->cont.readlink.size, NULL);
}
out:
if (unwind) {
- AFR_STACK_UNWIND (readlink, frame, op_ret, op_errno, buf, sbuf);
+ AFR_STACK_UNWIND (readlink, frame, op_ret, op_errno, buf, sbuf,
+ xdata);
}
return 0;
@@ -456,7 +454,7 @@ out:
int32_t
afr_readlink (call_frame_t *frame, xlator_t *this,
- loc_t *loc, size_t size)
+ loc_t *loc, size_t size, dict_t *xdata)
{
afr_private_t *priv = NULL;
xlator_t **children = NULL;
@@ -475,7 +473,9 @@ afr_readlink (call_frame_t *frame, xlator_t *this,
children = priv->children;
- ALLOC_OR_GOTO (frame->local, afr_local_t, out);
+ AFR_SBRAIN_CHECK_LOC (loc, out);
+
+ AFR_LOCAL_ALLOC_OR_GOTO (frame->local, out);
local = frame->local;
ret = afr_local_init (local, priv, &op_errno);
@@ -506,12 +506,12 @@ afr_readlink (call_frame_t *frame, xlator_t *this,
(void *) (long) call_child,
children[call_child],
children[call_child]->fops->readlink,
- loc, size);
+ loc, size, xdata);
ret = 0;
out:
if (ret < 0)
- AFR_STACK_UNWIND (readlink, frame, -1, op_errno, NULL, NULL);
+ AFR_STACK_UNWIND (readlink, frame, -1, op_errno, NULL, NULL, NULL);
return 0;
}
@@ -526,7 +526,7 @@ struct _xattr_key {
};
-void
+int
__gather_xattr_keys (dict_t *dict, char *key, data_t *value,
void *data)
{
@@ -538,13 +538,14 @@ __gather_xattr_keys (dict_t *dict, char *key, data_t *value,
xkey = GF_CALLOC (1, sizeof (*xkey), gf_afr_mt_xattr_key);
if (!xkey)
- return;
+ return -1;
xkey->key = key;
INIT_LIST_HEAD (&xkey->list);
list_add_tail (&xkey->list, list);
}
+ return 0;
}
@@ -574,7 +575,7 @@ __filter_xattrs (dict_t *dict)
int32_t
afr_getxattr_cbk (call_frame_t *frame, void *cookie,
xlator_t *this, int32_t op_ret, int32_t op_errno,
- dict_t *dict)
+ dict_t *dict, dict_t *xdata)
{
afr_private_t * priv = NULL;
afr_local_t * local = NULL;
@@ -608,7 +609,8 @@ afr_getxattr_cbk (call_frame_t *frame, void *cookie,
children[next_call_child],
children[next_call_child]->fops->getxattr,
&local->loc,
- local->cont.getxattr.name);
+ local->cont.getxattr.name,
+ NULL);
}
out:
@@ -616,39 +618,620 @@ out:
if (op_ret >= 0 && dict)
__filter_xattrs (dict);
- AFR_STACK_UNWIND (getxattr, frame, op_ret, op_errno, dict);
+ AFR_STACK_UNWIND (getxattr, frame, op_ret, op_errno, dict, xdata);
}
return 0;
}
int32_t
-afr_getxattr_unwind (call_frame_t *frame,
- int op_ret, int op_errno, dict_t *dict)
+afr_getxattr_unwind (call_frame_t *frame, int op_ret, int op_errno,
+ dict_t *dict, dict_t *xdata)
+
+{
+ AFR_STACK_UNWIND (getxattr, frame, op_ret, op_errno, dict, xdata);
+ return 0;
+}
+
+int32_t
+afr_fgetxattr_clrlk_cbk (call_frame_t *frame, void *cookie,
+ xlator_t *this, int32_t op_ret, int32_t op_errno,
+ dict_t *dict, dict_t *xdata)
+{
+ afr_local_t *local = NULL;
+ afr_private_t *priv = NULL;
+ xlator_t **children = NULL;
+ dict_t *xattr = NULL;
+ char *tmp_report = NULL;
+ char lk_summary[1024] = {0,};
+ int serz_len = 0;
+ int32_t callcnt = 0;
+ long int cky = 0;
+ int ret = 0;
+
+ priv = this->private;
+ children = priv->children;
+
+ local = frame->local;
+ cky = (long) cookie;
+
+ LOCK (&frame->lock);
+ {
+ callcnt = --local->call_count;
+ if (op_ret == -1)
+ local->child_errno[cky] = op_errno;
+
+ if (!local->dict)
+ local->dict = dict_new ();
+ if (local->dict) {
+ ret = dict_get_str (dict, local->cont.getxattr.name,
+ &tmp_report);
+ if (ret)
+ goto unlock;
+ ret = dict_set_dynstr (local->dict,
+ children[cky]->name,
+ gf_strdup (tmp_report));
+ if (ret)
+ goto unlock;
+ }
+ }
+unlock:
+ UNLOCK (&frame->lock);
+
+ if (!callcnt) {
+ xattr = dict_new ();
+ if (!xattr) {
+ op_ret = -1;
+ op_errno = ENOMEM;
+ goto unwind;
+ }
+ ret = dict_serialize_value_with_delim (local->dict,
+ lk_summary,
+ &serz_len, '\n');
+ if (ret) {
+ op_ret = -1;
+ op_errno = ENOMEM;
+ gf_log (this->name, GF_LOG_ERROR,
+ "Error serializing dictionary");
+ goto unwind;
+ }
+ if (serz_len == -1)
+ snprintf (lk_summary, sizeof (lk_summary),
+ "No locks cleared.");
+ ret = dict_set_dynstr (xattr, local->cont.getxattr.name,
+ gf_strdup (lk_summary));
+ if (ret) {
+ op_ret = -1;
+ op_errno = ENOMEM;
+ gf_log (this->name, GF_LOG_ERROR,
+ "Error setting dictionary");
+ goto unwind;
+ }
+
+ unwind:
+ // Updating child_errno with more recent 'events'
+ local->child_errno[cky] = op_errno;
+ op_errno = afr_resultant_errno_get (NULL, local->child_errno,
+ priv->child_count);
+ AFR_STACK_UNWIND (fgetxattr, frame, op_ret, op_errno, xattr,
+ xdata);
+
+ if (xattr)
+ dict_unref (xattr);
+ }
+
+ return ret;
+}
+
+int32_t
+afr_getxattr_clrlk_cbk (call_frame_t *frame, void *cookie,
+ xlator_t *this, int32_t op_ret, int32_t op_errno,
+ dict_t *dict, dict_t *xdata)
+{
+ afr_local_t *local = NULL;
+ afr_private_t *priv = NULL;
+ xlator_t **children = NULL;
+ dict_t *xattr = NULL;
+ char *tmp_report = NULL;
+ char lk_summary[1024] = {0,};
+ int serz_len = 0;
+ int32_t callcnt = 0;
+ long int cky = 0;
+ int ret = 0;
+
+ priv = this->private;
+ children = priv->children;
+
+ local = frame->local;
+ cky = (long) cookie;
+
+ LOCK (&frame->lock);
+ {
+ callcnt = --local->call_count;
+ if (op_ret == -1)
+ local->child_errno[cky] = op_errno;
+
+ if (!local->dict)
+ local->dict = dict_new ();
+ if (local->dict) {
+ ret = dict_get_str (dict, local->cont.getxattr.name,
+ &tmp_report);
+ if (ret)
+ goto unlock;
+ ret = dict_set_dynstr (local->dict,
+ children[cky]->name,
+ gf_strdup (tmp_report));
+ if (ret)
+ goto unlock;
+ }
+ }
+unlock:
+ UNLOCK (&frame->lock);
+
+ if (!callcnt) {
+ xattr = dict_new ();
+ if (!xattr) {
+ op_ret = -1;
+ op_errno = ENOMEM;
+ goto unwind;
+ }
+ ret = dict_serialize_value_with_delim (local->dict,
+ lk_summary,
+ &serz_len, '\n');
+ if (ret) {
+ op_ret = -1;
+ op_errno = ENOMEM;
+ gf_log (this->name, GF_LOG_ERROR,
+ "Error serializing dictionary");
+ goto unwind;
+ }
+ if (serz_len == -1)
+ snprintf (lk_summary, sizeof (lk_summary),
+ "No locks cleared.");
+ ret = dict_set_dynstr (xattr, local->cont.getxattr.name,
+ gf_strdup (lk_summary));
+ if (ret) {
+ op_ret = -1;
+ op_errno = ENOMEM;
+ gf_log (this->name, GF_LOG_ERROR,
+ "Error setting dictionary");
+ goto unwind;
+ }
+
+ unwind:
+ // Updating child_errno with more recent 'events'
+ local->child_errno[cky] = op_errno;
+ op_errno = afr_resultant_errno_get (NULL, local->child_errno,
+ priv->child_count);
+ AFR_STACK_UNWIND (getxattr, frame, op_ret, op_errno, xattr, xdata);
+
+ if (xattr)
+ dict_unref (xattr);
+ }
+
+ return ret;
+}
+
+/**
+ * node-uuid cbk uses next child querying mechanism
+ */
+int32_t
+afr_getxattr_node_uuid_cbk (call_frame_t *frame, void *cookie,
+ xlator_t *this, int32_t op_ret, int32_t op_errno,
+ dict_t *dict, dict_t *xdata)
+{
+ afr_private_t *priv = NULL;
+ afr_local_t *local = NULL;
+ xlator_t **children = NULL;
+ int unwind = 1;
+ int curr_call_child = 0;
+
+ priv = this->private;
+ children = priv->children;
+
+ local = frame->local;
+
+ if (op_ret == -1) { /** query the _next_ child */
+
+ /**
+ * _current_ becomes _next_
+ * If done with all childs and yet no success; give up !
+ */
+ curr_call_child = (int) ((long)cookie);
+ if (++curr_call_child == priv->child_count)
+ goto unwind;
+
+ gf_log (this->name, GF_LOG_WARNING,
+ "op_ret (-1): Re-querying afr-child (%d/%d)",
+ curr_call_child, priv->child_count);
+
+ unwind = 0;
+ STACK_WIND_COOKIE (frame, afr_getxattr_node_uuid_cbk,
+ (void *) (long) curr_call_child,
+ children[curr_call_child],
+ children[curr_call_child]->fops->getxattr,
+ &local->loc,
+ local->cont.getxattr.name,
+ NULL);
+ }
+
+ unwind:
+ if (unwind)
+ AFR_STACK_UNWIND (getxattr, frame, op_ret, op_errno, dict,
+ NULL);
+
+ return 0;
+}
+
+int32_t
+afr_getxattr_lockinfo_cbk (call_frame_t *frame, void *cookie,
+ xlator_t *this, int32_t op_ret, int32_t op_errno,
+ dict_t *dict, dict_t *xdata)
+{
+ int call_cnt = 0, len = 0;
+ char *lockinfo_buf = NULL;
+ dict_t *lockinfo = NULL, *newdict = NULL;
+ afr_local_t *local = NULL;
+
+ LOCK (&frame->lock);
+ {
+ local = frame->local;
+
+ call_cnt = --local->call_count;
+
+ if ((op_ret < 0) || (!dict && !xdata)) {
+ goto unlock;
+ }
+
+ if (xdata) {
+ if (!local->xdata_rsp) {
+ local->xdata_rsp = dict_new ();
+ if (!local->xdata_rsp) {
+ local->op_ret = -1;
+ local->op_errno = ENOMEM;
+ goto unlock;
+ }
+ }
+ }
+
+ if (!dict) {
+ goto unlock;
+ }
+
+ op_ret = dict_get_ptr_and_len (dict, GF_XATTR_LOCKINFO_KEY,
+ (void **)&lockinfo_buf, &len);
+
+ if (!lockinfo_buf) {
+ goto unlock;
+ }
+
+ if (!local->dict) {
+ local->dict = dict_new ();
+ if (!local->dict) {
+ local->op_ret = -1;
+ local->op_errno = ENOMEM;
+ goto unlock;
+ }
+ }
+ }
+unlock:
+ UNLOCK (&frame->lock);
+
+ if (lockinfo_buf != NULL) {
+ lockinfo = dict_new ();
+ if (lockinfo == NULL) {
+ local->op_ret = -1;
+ local->op_errno = ENOMEM;
+ } else {
+ op_ret = dict_unserialize (lockinfo_buf, len,
+ &lockinfo);
+
+ if (lockinfo && local->dict) {
+ dict_copy (lockinfo, local->dict);
+ }
+ }
+ }
+
+ if (xdata && local->xdata_rsp) {
+ dict_copy (xdata, local->xdata_rsp);
+ }
+
+ if (!call_cnt) {
+ newdict = dict_new ();
+ if (!newdict) {
+ local->op_ret = -1;
+ local->op_errno = ENOMEM;
+ goto unwind;
+ }
+
+ len = dict_serialized_length (local->dict);
+ if (len == 0) {
+ goto unwind;
+ }
+
+ lockinfo_buf = GF_CALLOC (1, len, gf_common_mt_char);
+ if (!lockinfo_buf) {
+ local->op_ret = -1;
+ local->op_errno = ENOMEM;
+ goto unwind;
+ }
+
+ op_ret = dict_serialize (local->dict, lockinfo_buf);
+ if (op_ret < 0) {
+ local->op_ret = -1;
+ local->op_errno = -op_ret;
+ }
+
+ op_ret = dict_set_dynptr (newdict, GF_XATTR_LOCKINFO_KEY,
+ (void *)lockinfo_buf, len);
+ if (op_ret < 0) {
+ local->op_ret = -1;
+ local->op_errno = -op_ret;
+ goto unwind;
+ }
+
+ unwind:
+ AFR_STACK_UNWIND (getxattr, frame, op_ret,
+ op_errno, newdict,
+ local->xdata_rsp);
+ }
+
+ dict_unref (lockinfo);
+ return 0;
+}
+
+int32_t
+afr_fgetxattr_lockinfo_cbk (call_frame_t *frame, void *cookie,
+ xlator_t *this, int32_t op_ret, int32_t op_errno,
+ dict_t *dict, dict_t *xdata)
{
- AFR_STACK_UNWIND (getxattr, frame, op_ret, op_errno, dict);
+ int call_cnt = 0, len = 0;
+ char *lockinfo_buf = NULL;
+ dict_t *lockinfo = NULL, *newdict = NULL;
+ afr_local_t *local = NULL;
+
+ LOCK (&frame->lock);
+ {
+ local = frame->local;
+
+ call_cnt = --local->call_count;
+
+ if ((op_ret < 0) || (!dict && !xdata)) {
+ goto unlock;
+ }
+
+ if (xdata) {
+ if (!local->xdata_rsp) {
+ local->xdata_rsp = dict_new ();
+ if (!local->xdata_rsp) {
+ local->op_ret = -1;
+ local->op_errno = ENOMEM;
+ goto unlock;
+ }
+ }
+ }
+
+ if (!dict) {
+ goto unlock;
+ }
+
+ op_ret = dict_get_ptr_and_len (dict, GF_XATTR_LOCKINFO_KEY,
+ (void **)&lockinfo_buf, &len);
+
+ if (!lockinfo_buf) {
+ goto unlock;
+ }
+
+ if (!local->dict) {
+ local->dict = dict_new ();
+ if (!local->dict) {
+ local->op_ret = -1;
+ local->op_errno = ENOMEM;
+ goto unlock;
+ }
+ }
+ }
+unlock:
+ UNLOCK (&frame->lock);
+
+ if (lockinfo_buf != NULL) {
+ lockinfo = dict_new ();
+ if (lockinfo == NULL) {
+ local->op_ret = -1;
+ local->op_errno = ENOMEM;
+ } else {
+ op_ret = dict_unserialize (lockinfo_buf, len,
+ &lockinfo);
+
+ if (lockinfo && local->dict) {
+ dict_copy (lockinfo, local->dict);
+ }
+ }
+ }
+
+ if (xdata && local->xdata_rsp) {
+ dict_copy (xdata, local->xdata_rsp);
+ }
+
+ if (!call_cnt) {
+ newdict = dict_new ();
+ if (!newdict) {
+ local->op_ret = -1;
+ local->op_errno = ENOMEM;
+ goto unwind;
+ }
+
+ len = dict_serialized_length (local->dict);
+ if (len <= 0) {
+ goto unwind;
+ }
+
+ lockinfo_buf = GF_CALLOC (1, len, gf_common_mt_char);
+ if (!lockinfo_buf) {
+ local->op_ret = -1;
+ local->op_errno = ENOMEM;
+ goto unwind;
+ }
+
+ op_ret = dict_serialize (local->dict, lockinfo_buf);
+ if (op_ret < 0) {
+ local->op_ret = -1;
+ local->op_errno = -op_ret;
+ }
+
+ op_ret = dict_set_dynptr (newdict, GF_XATTR_LOCKINFO_KEY,
+ (void *)lockinfo_buf, len);
+ if (op_ret < 0) {
+ local->op_ret = -1;
+ local->op_errno = -op_ret;
+ goto unwind;
+ }
+
+ unwind:
+ AFR_STACK_UNWIND (fgetxattr, frame, op_ret,
+ op_errno, newdict,
+ local->xdata_rsp);
+ }
+
+ dict_unref (lockinfo);
+
return 0;
}
int32_t
+afr_fgetxattr_pathinfo_cbk (call_frame_t *frame, void *cookie,
+ xlator_t *this, int32_t op_ret, int32_t op_errno,
+ dict_t *dict, dict_t *xdata)
+{
+ afr_local_t *local = NULL;
+ int32_t callcnt = 0;
+ int ret = 0;
+ char *xattr = NULL;
+ char *xattr_serz = NULL;
+ char xattr_cky[1024] = {0,};
+ dict_t *nxattr = NULL;
+ long cky = 0;
+ int32_t padding = 0;
+ int32_t tlen = 0;
+
+ if (!frame || !frame->local || !this) {
+ gf_log ("", GF_LOG_ERROR, "possible NULL deref");
+ goto out;
+ }
+
+ local = frame->local;
+ cky = (long) cookie;
+
+ LOCK (&frame->lock);
+ {
+ callcnt = --local->call_count;
+
+ if (!dict || (op_ret < 0))
+ goto out;
+
+ if (!local->dict)
+ local->dict = dict_new ();
+
+ if (local->dict) {
+ ret = dict_get_str (dict,
+ local->cont.getxattr.name,
+ &xattr);
+ if (ret)
+ goto out;
+
+ xattr = gf_strdup (xattr);
+
+ (void)snprintf (xattr_cky, 1024, "%s-%ld",
+ local->cont.getxattr.name, cky);
+ ret = dict_set_dynstr (local->dict,
+ xattr_cky, xattr);
+ if (ret) {
+ gf_log (this->name, GF_LOG_ERROR,
+ "Cannot set xattr cookie key");
+ goto out;
+ }
+
+ local->cont.getxattr.xattr_len
+ += strlen (xattr) + 1;
+ }
+ }
+out:
+ UNLOCK (&frame->lock);
+
+ if (!callcnt) {
+ if (!local->cont.getxattr.xattr_len)
+ goto unwind;
+
+ nxattr = dict_new ();
+ if (!nxattr)
+ goto unwind;
+
+ /* extra bytes for decorations (brackets and <>'s) */
+ padding += strlen (this->name)
+ + strlen (AFR_PATHINFO_HEADER) + 4;
+ local->cont.getxattr.xattr_len += (padding + 2);
+
+ xattr_serz = GF_CALLOC (local->cont.getxattr.xattr_len,
+ sizeof (char), gf_common_mt_char);
+
+ if (!xattr_serz)
+ goto unwind;
+
+ /* the xlator info */
+ (void) sprintf (xattr_serz, "(<"AFR_PATHINFO_HEADER"%s> ",
+ this->name);
+
+ /* actual series of pathinfo */
+ ret = dict_serialize_value_with_delim (local->dict,
+ xattr_serz
+ + strlen (xattr_serz),
+ &tlen, ' ');
+ if (ret) {
+ gf_log (this->name, GF_LOG_ERROR, "Error serializing"
+ " dictionary");
+ goto unwind;
+ }
+
+ /* closing part */
+ *(xattr_serz + padding + tlen) = ')';
+ *(xattr_serz + padding + tlen + 1) = '\0';
+
+ ret = dict_set_dynstr (nxattr, local->cont.getxattr.name,
+ xattr_serz);
+ if (ret)
+ gf_log (this->name, GF_LOG_ERROR, "Cannot set pathinfo"
+ " key in dict");
+
+ unwind:
+ AFR_STACK_UNWIND (fgetxattr, frame, op_ret, op_errno, nxattr,
+ xdata);
+
+ if (nxattr)
+ dict_unref (nxattr);
+ }
+
+ return ret;
+}
+
+int32_t
afr_getxattr_pathinfo_cbk (call_frame_t *frame, void *cookie,
xlator_t *this, int32_t op_ret, int32_t op_errno,
- dict_t *dict)
+ dict_t *dict, dict_t *xdata)
{
- afr_local_t *local = NULL;
- int32_t callcnt = 0;
- int ret = 0;
- char *pathinfo = NULL;
- char *pathinfo_serz = NULL;
- char pathinfo_cky[1024] = {0,};
- dict_t *xattr = NULL;
- long cky = 0;
- int32_t padding = 0;
- int32_t tlen = 0;
+ afr_local_t *local = NULL;
+ int32_t callcnt = 0;
+ int ret = 0;
+ char *xattr = NULL;
+ char *xattr_serz = NULL;
+ char xattr_cky[1024] = {0,};
+ dict_t *nxattr = NULL;
+ long cky = 0;
+ int32_t padding = 0;
+ int32_t tlen = 0;
if (!frame || !frame->local || !this) {
- gf_log (this->name, GF_LOG_ERROR, "possible NULL deref");
+ gf_log ("", GF_LOG_ERROR, "possible NULL deref");
goto out;
}
@@ -666,90 +1249,222 @@ afr_getxattr_pathinfo_cbk (call_frame_t *frame, void *cookie,
local->dict = dict_new ();
if (local->dict) {
- ret = dict_get_str (dict, GF_XATTR_PATHINFO_KEY, &pathinfo);
+ ret = dict_get_str (dict,
+ local->cont.getxattr.name,
+ &xattr);
if (ret)
goto out;
- pathinfo = gf_strdup (pathinfo);
+ xattr = gf_strdup (xattr);
- snprintf (pathinfo_cky, 1024, "%s-%ld", GF_XATTR_PATHINFO_KEY, cky);
- ret = dict_set_dynstr (local->dict, pathinfo_cky, pathinfo);
+ (void)snprintf (xattr_cky, 1024, "%s-%ld",
+ local->cont.getxattr.name, cky);
+ ret = dict_set_dynstr (local->dict,
+ xattr_cky, xattr);
if (ret) {
- gf_log (this->name, GF_LOG_ERROR, "Cannot set pathinfo cookie key");
+ gf_log (this->name, GF_LOG_ERROR,
+ "Cannot set xattr cookie key");
goto out;
}
- local->cont.getxattr.pathinfo_len += strlen (pathinfo) + 1;
+ local->cont.getxattr.xattr_len += strlen (xattr) + 1;
}
}
out:
UNLOCK (&frame->lock);
if (!callcnt) {
- if (!local->cont.getxattr.pathinfo_len)
+ if (!local->cont.getxattr.xattr_len)
goto unwind;
- xattr = dict_new ();
- if (!xattr)
+ nxattr = dict_new ();
+ if (!nxattr)
goto unwind;
/* extra bytes for decorations (brackets and <>'s) */
- padding = strlen (this->name) + strlen (AFR_PATHINFO_HEADER) + 4;
- local->cont.getxattr.pathinfo_len += (padding + 2);
+ padding += strlen (this->name) + strlen (AFR_PATHINFO_HEADER) + 4;
+ local->cont.getxattr.xattr_len += (padding + 2);
- pathinfo_serz = GF_CALLOC (local->cont.getxattr.pathinfo_len, sizeof (char),
- gf_common_mt_char);
+ xattr_serz = GF_CALLOC (local->cont.getxattr.xattr_len,
+ sizeof (char), gf_common_mt_char);
- if (!pathinfo_serz)
+ if (!xattr_serz)
goto unwind;
/* the xlator info */
- sprintf (pathinfo_serz, "(<"AFR_PATHINFO_HEADER"%s> ", this->name);
+ (void) sprintf (xattr_serz, "(<"AFR_PATHINFO_HEADER"%s> ",
+ this->name);
/* actual series of pathinfo */
- ret = dict_serialize_value_with_delim (local->dict, pathinfo_serz + strlen (pathinfo_serz),
+ ret = dict_serialize_value_with_delim (local->dict,
+ xattr_serz + strlen (xattr_serz),
&tlen, ' ');
if (ret) {
- gf_log (this->name, GF_LOG_ERROR, "Error serializing dictionary");
+ gf_log (this->name, GF_LOG_ERROR, "Error serializing"
+ " dictionary");
goto unwind;
}
/* closing part */
- *(pathinfo_serz + padding + tlen) = ')';
- *(pathinfo_serz + padding + tlen + 1) = '\0';
+ *(xattr_serz + padding + tlen) = ')';
+ *(xattr_serz + padding + tlen + 1) = '\0';
- ret = dict_set_dynstr (xattr, GF_XATTR_PATHINFO_KEY, pathinfo_serz);
+ ret = dict_set_dynstr (nxattr, local->cont.getxattr.name,
+ xattr_serz);
if (ret)
- gf_log (this->name, GF_LOG_ERROR, "Cannot set pathinfo key in dict");
+ gf_log (this->name, GF_LOG_ERROR, "Cannot set pathinfo"
+ " key in dict");
unwind:
- AFR_STACK_UNWIND (getxattr, frame, op_ret, op_errno, xattr);
+ AFR_STACK_UNWIND (getxattr, frame, op_ret, op_errno, nxattr,
+ xdata);
- if (local->dict)
- dict_unref (local->dict);
-
- if (xattr)
- dict_unref (xattr);
+ if (nxattr)
+ dict_unref (nxattr);
}
return ret;
}
+static int
+afr_aggregate_stime_xattr (dict_t *this, char *key, data_t *value, void *data)
+{
+ int ret = 0;
+
+ if (fnmatch (GF_XATTR_STIME_PATTERN, key, FNM_NOESCAPE) == 0)
+ ret = gf_get_min_stime (THIS, data, key, value);
+
+ return ret;
+}
+
int32_t
-afr_getxattr (call_frame_t *frame, xlator_t *this,
- loc_t *loc, const char *name)
+afr_common_getxattr_stime_cbk (call_frame_t *frame, void *cookie,
+ xlator_t *this, int32_t op_ret, int32_t op_errno,
+ dict_t *dict, dict_t *xdata)
+{
+ afr_local_t *local = NULL;
+ int32_t callcnt = 0;
+
+ if (!frame || !frame->local || !this) {
+ gf_log ("", GF_LOG_ERROR, "possible NULL deref");
+ goto out;
+ }
+
+ local = frame->local;
+
+ LOCK (&frame->lock);
+ {
+ callcnt = --local->call_count;
+
+ if (!dict || (op_ret < 0)) {
+ local->op_errno = op_errno;
+ goto cleanup;
+ }
+
+ if (!local->dict)
+ local->dict = dict_copy_with_ref (dict, NULL);
+ else
+ dict_foreach (dict, afr_aggregate_stime_xattr,
+ local->dict);
+ local->op_ret = 0;
+ }
+
+cleanup:
+ UNLOCK (&frame->lock);
+
+ if (!callcnt) {
+ AFR_STACK_UNWIND (getxattr, frame, local->op_ret,
+ local->op_errno, local->dict, xdata);
+ }
+
+out:
+ return 0;
+}
+
+
+static gf_boolean_t
+afr_is_special_xattr (const char *name, fop_getxattr_cbk_t *cbk,
+ gf_boolean_t is_fgetxattr)
{
- afr_private_t *priv = NULL;
- xlator_t **children = NULL;
- int call_child = 0;
- afr_local_t *local = NULL;
- xlator_list_t *trav = NULL;
- xlator_t **sub_volumes = NULL;
- int i = 0;
- int32_t op_errno = 0;
- int32_t read_child = -1;
- int ret = -1;
+ gf_boolean_t is_spl = _gf_true;
+ GF_ASSERT (cbk);
+ if (!cbk) {
+ is_spl = _gf_false;
+ goto out;
+ }
+
+ if (!strcmp (name, GF_XATTR_PATHINFO_KEY)) {
+ if (is_fgetxattr) {
+ *cbk = afr_fgetxattr_pathinfo_cbk;
+ } else {
+ *cbk = afr_getxattr_pathinfo_cbk;
+ }
+ } else if (!strncmp (name, GF_XATTR_CLRLK_CMD,
+ strlen (GF_XATTR_CLRLK_CMD))) {
+ if (is_fgetxattr) {
+ *cbk = afr_fgetxattr_clrlk_cbk;
+ } else {
+ *cbk = afr_getxattr_clrlk_cbk;
+ }
+ } else if (!strncmp (name, GF_XATTR_LOCKINFO_KEY,
+ strlen (GF_XATTR_LOCKINFO_KEY))) {
+ if (is_fgetxattr) {
+ *cbk = afr_fgetxattr_lockinfo_cbk;
+ } else {
+ *cbk = afr_getxattr_lockinfo_cbk;
+ }
+ } else if (fnmatch (GF_XATTR_STIME_PATTERN, name, FNM_NOESCAPE) == 0) {
+ *cbk = afr_common_getxattr_stime_cbk;
+ } else {
+ is_spl = _gf_false;
+ }
+
+out:
+ return is_spl;
+}
+
+static void
+afr_getxattr_frm_all_children (xlator_t *this, call_frame_t *frame,
+ const char *name, loc_t *loc,
+ fop_getxattr_cbk_t cbk)
+{
+ afr_private_t *priv = NULL;
+ afr_local_t *local = NULL;
+ xlator_t **children = NULL;
+ int i = 0;
+
+ priv = this->private;
+ children = priv->children;
+
+ local = frame->local;
+ local->call_count = priv->child_count;
+
+ for (i = 0; i < priv->child_count; i++) {
+ STACK_WIND_COOKIE (frame, cbk,
+ (void *) (long) i,
+ children[i], children[i]->fops->getxattr,
+ loc, name, NULL);
+ }
+ return;
+}
+
+int32_t
+afr_getxattr (call_frame_t *frame, xlator_t *this,
+ loc_t *loc, const char *name, dict_t *xdata)
+{
+ afr_private_t *priv = NULL;
+ xlator_t **children = NULL;
+ int call_child = 0;
+ afr_local_t *local = NULL;
+ xlator_list_t *trav = NULL;
+ xlator_t **sub_volumes = NULL;
+ int i = 0;
+ int32_t op_errno = 0;
+ int32_t read_child = -1;
+ int ret = -1;
+ fop_getxattr_cbk_t cbk = NULL;
+ int afr_xtime_gauge[MCNT_MAX] = {0,};
VALIDATE_OR_GOTO (frame, out);
VALIDATE_OR_GOTO (this, out);
@@ -760,7 +1475,9 @@ afr_getxattr (call_frame_t *frame, xlator_t *this,
children = priv->children;
- ALLOC_OR_GOTO (frame->local, afr_local_t, out);
+ AFR_SBRAIN_CHECK_LOC (loc, out);
+
+ AFR_LOCAL_ALLOC_OR_GOTO (frame->local, out);
local = frame->local;
ret = afr_local_init (local, priv, &op_errno);
@@ -768,38 +1485,98 @@ afr_getxattr (call_frame_t *frame, xlator_t *this,
goto out;
loc_copy (&local->loc, loc);
- if (name)
- local->cont.getxattr.name = gf_strdup (name);
+ if (!name)
+ goto no_name;
+
+ local->cont.getxattr.name = gf_strdup (name);
+
+ if (!strncmp (name, AFR_XATTR_PREFIX,
+ strlen (AFR_XATTR_PREFIX))) {
+ gf_log (this->name, GF_LOG_INFO,
+ "%s: no data present for key %s",
+ loc->path, name);
+ op_errno = ENODATA;
+ goto out;
+ }
+ if ((strcmp (GF_XATTR_MARKER_KEY, name) == 0)
+ && (GF_CLIENT_PID_GSYNCD == frame->root->pid)) {
+ local->marker.call_count = priv->child_count;
+
+ sub_volumes = alloca ( priv->child_count * sizeof (xlator_t *));
+ for (i = 0, trav = this->children; trav ;
+ trav = trav->next, i++) {
+
+ *(sub_volumes + i) = trav->xlator;
+ }
+
+ if (cluster_getmarkerattr (frame, this, loc, name,
+ local, afr_getxattr_unwind,
+ sub_volumes,
+ priv->child_count,
+ MARKER_UUID_TYPE,
+ marker_uuid_default_gauge,
+ priv->vol_uuid)) {
- if (name) {
- if (!strncmp (name, AFR_XATTR_PREFIX,
- strlen (AFR_XATTR_PREFIX))) {
gf_log (this->name, GF_LOG_INFO,
- "%s: no data present for key %s",
+ "%s: failed to get marker attr (%s)",
loc->path, name);
- op_errno = ENODATA;
+ op_errno = EINVAL;
goto out;
}
- if ((strcmp (GF_XATTR_MARKER_KEY, name) == 0)
- && (-1 == frame->root->pid)) {
+ return 0;
+ }
+
+ /*
+ * if we are doing getxattr with pathinfo as the key then we
+ * collect information from all childs
+ */
+ if (afr_is_special_xattr (name, &cbk, 0)) {
+ afr_getxattr_frm_all_children (this, frame, name,
+ loc, cbk);
+ return 0;
+ }
+
+ if (XATTR_IS_NODE_UUID (name)) {
+ i = 0;
+ STACK_WIND_COOKIE (frame, afr_getxattr_node_uuid_cbk,
+ (void *) (long) i,
+ children[i],
+ children[i]->fops->getxattr,
+ loc, name, xdata);
+ return 0;
+ }
+
+ if (*priv->vol_uuid) {
+ if ((match_uuid_local (name, priv->vol_uuid) == 0)
+ && (GF_CLIENT_PID_GSYNCD == frame->root->pid)) {
local->marker.call_count = priv->child_count;
- sub_volumes = alloca ( priv->child_count * sizeof (xlator_t *));
+ sub_volumes = alloca ( priv->child_count
+ * sizeof (xlator_t *));
for (i = 0, trav = this->children; trav ;
trav = trav->next, i++) {
*(sub_volumes + i) = trav->xlator;
+
}
- if (cluster_getmarkerattr (frame, this, loc, name,
- local, afr_getxattr_unwind,
+ /* don't err out on getting ENOTCONN (brick down)
+ * from a subset of the bricks
+ */
+ memcpy (afr_xtime_gauge, marker_xtime_default_gauge,
+ sizeof (afr_xtime_gauge));
+ afr_xtime_gauge[MCNT_NOTFOUND] = 0;
+ afr_xtime_gauge[MCNT_ENOTCONN] = 0;
+ if (cluster_getmarkerattr (frame, this, loc,
+ name, local,
+ afr_getxattr_unwind,
sub_volumes,
priv->child_count,
- MARKER_UUID_TYPE,
+ MARKER_XTIME_TYPE,
+ afr_xtime_gauge,
priv->vol_uuid)) {
-
gf_log (this->name, GF_LOG_INFO,
"%s: failed to get marker attr (%s)",
loc->path, name);
@@ -809,84 +1586,209 @@ afr_getxattr (call_frame_t *frame, xlator_t *this,
return 0;
}
+ }
- /*
- * if we are doing getxattr with pathinfo as the key then we
- * collect information from all childs
- */
- if (strncmp (name, GF_XATTR_PATHINFO_KEY,
- strlen (GF_XATTR_PATHINFO_KEY)) == 0) {
-
- local->call_count = priv->child_count;
- for (i = 0; i < priv->child_count; i++) {
- STACK_WIND_COOKIE (frame, afr_getxattr_pathinfo_cbk,
- (void *) (long) i,
- children[i], children[i]->fops->getxattr,
- loc, name);
- }
+no_name:
+ local->fresh_children = afr_children_create (priv->child_count);
+ if (!local->fresh_children) {
+ op_errno = ENOMEM;
+ goto out;
+ }
- return 0;
- }
+ read_child = afr_inode_get_read_ctx (this, loc->inode,
+ local->fresh_children);
+ ret = afr_get_call_child (this, local->child_up, read_child,
+ local->fresh_children,
+ &call_child,
+ &local->cont.getxattr.last_index);
+ if (ret < 0) {
+ op_errno = -ret;
+ goto out;
+ }
- if (*priv->vol_uuid) {
- if ((match_uuid_local (name, priv->vol_uuid) == 0)
- && (-1 == frame->root->pid)) {
+ STACK_WIND_COOKIE (frame, afr_getxattr_cbk,
+ (void *) (long) call_child,
+ children[call_child],
+ children[call_child]->fops->getxattr,
+ loc, name, xdata);
- local->marker.call_count = priv->child_count;
+ ret = 0;
+out:
+ if (ret < 0)
+ AFR_STACK_UNWIND (getxattr, frame, -1, op_errno, NULL, NULL);
+ return 0;
+}
- sub_volumes = alloca ( priv->child_count * sizeof (xlator_t *));
- for (i = 0, trav = this->children; trav ;
- trav = trav->next, i++) {
+/* {{{ fgetxattr */
- *(sub_volumes + i) = trav->xlator;
- }
+int32_t
+afr_fgetxattr_cbk (call_frame_t *frame, void *cookie,
+ xlator_t *this, int32_t op_ret, int32_t op_errno,
+ dict_t *dict, dict_t *xdata)
+{
+ afr_private_t * priv = NULL;
+ afr_local_t * local = NULL;
+ xlator_t ** children = NULL;
+ int unwind = 1;
+ int32_t *last_index = NULL;
+ int32_t next_call_child = -1;
+ int32_t read_child = -1;
+ int32_t *fresh_children = NULL;
- if (cluster_getmarkerattr (frame, this, loc,
- name, local,
- afr_getxattr_unwind,
- sub_volumes,
- priv->child_count,
- MARKER_XTIME_TYPE,
- priv->vol_uuid)) {
- gf_log (this->name, GF_LOG_INFO,
- "%s: failed to get marker attr (%s)",
- loc->path, name);
- op_errno = EINVAL;
- goto out;
- }
+ priv = this->private;
+ children = priv->children;
- return 0;
- }
- }
+ local = frame->local;
+
+ read_child = (long) cookie;
+
+ if (op_ret == -1) {
+ last_index = &local->cont.getxattr.last_index;
+ fresh_children = local->fresh_children;
+ next_call_child = afr_next_call_child (fresh_children,
+ local->child_up,
+ priv->child_count,
+ last_index, read_child);
+ if (next_call_child < 0)
+ goto out;
+
+ unwind = 0;
+ STACK_WIND_COOKIE (frame, afr_fgetxattr_cbk,
+ (void *) (long) read_child,
+ children[next_call_child],
+ children[next_call_child]->fops->fgetxattr,
+ local->fd,
+ local->cont.getxattr.name,
+ NULL);
+ }
+
+out:
+ if (unwind) {
+ if (op_ret >= 0 && dict)
+ __filter_xattrs (dict);
+
+ AFR_STACK_UNWIND (fgetxattr, frame, op_ret, op_errno, dict,
+ xdata);
+ }
+
+ return 0;
+}
+
+int32_t
+afr_fgetxattr_unwind (call_frame_t *frame,
+ int op_ret, int op_errno, dict_t *dict, dict_t *xdata)
+
+{
+ AFR_STACK_UNWIND (fgetxattr, frame, op_ret, op_errno, dict, xdata);
+ return 0;
+}
+
+static void
+afr_fgetxattr_frm_all_children (xlator_t *this, call_frame_t *frame,
+ const char *name, fd_t *fd,
+ fop_fgetxattr_cbk_t cbk)
+{
+ afr_private_t *priv = NULL;
+ afr_local_t *local = NULL;
+ xlator_t **children = NULL;
+ int i = 0;
+
+ priv = this->private;
+ children = priv->children;
+
+ local = frame->local;
+ local->call_count = priv->child_count;
+
+ for (i = 0; i < priv->child_count; i++) {
+ STACK_WIND_COOKIE (frame, cbk,
+ (void *) (long) i,
+ children[i], children[i]->fops->fgetxattr,
+ fd, name, NULL);
}
+ return;
+}
+
+int32_t
+afr_fgetxattr (call_frame_t *frame, xlator_t *this,
+ fd_t *fd, const char *name, dict_t *xdata)
+{
+ afr_private_t *priv = NULL;
+ xlator_t **children = NULL;
+ int call_child = 0;
+ afr_local_t *local = NULL;
+ int32_t op_ret = -1;
+ int32_t op_errno = 0;
+ int32_t read_child = -1;
+ fop_fgetxattr_cbk_t cbk = NULL;
+
+ VALIDATE_OR_GOTO (frame, out);
+ VALIDATE_OR_GOTO (this, out);
+ VALIDATE_OR_GOTO (this->private, out);
+
+ priv = this->private;
+ VALIDATE_OR_GOTO (priv->children, out);
+
+ children = priv->children;
+
+ AFR_SBRAIN_CHECK_FD (fd, out);
+
+ AFR_LOCAL_ALLOC_OR_GOTO (local, out);
+ frame->local = local;
+
+ op_ret = afr_local_init (local, priv, &op_errno);
+ if (op_ret < 0) {
+ op_errno = -op_ret;
+ goto out;
+ }
+
+ local->fd = fd_ref (fd);
+ if (name)
+ local->cont.getxattr.name = gf_strdup (name);
+
+ /* pathinfo gets handled only in getxattr(), but we need to handle
+ * lockinfo.
+ * If we are doing fgetxattr with lockinfo as the key then we
+ * collect information from all children.
+ */
+ if (afr_is_special_xattr (name, &cbk, 1)) {
+ afr_fgetxattr_frm_all_children (this, frame, name,
+ fd, cbk);
+ return 0;
+ }
+
+
local->fresh_children = afr_children_create (priv->child_count);
if (!local->fresh_children) {
op_errno = ENOMEM;
goto out;
}
- read_child = afr_inode_get_read_ctx (this, loc->inode, local->fresh_children);
- ret = afr_get_call_child (this, local->child_up, read_child,
+ read_child = afr_inode_get_read_ctx (this, fd->inode,
+ local->fresh_children);
+ op_ret = afr_get_call_child (this, local->child_up, read_child,
local->fresh_children,
&call_child,
&local->cont.getxattr.last_index);
- if (ret < 0) {
- op_errno = -ret;
+ if (op_ret < 0) {
+ op_errno = -op_ret;
+ op_ret = -1;
goto out;
}
- STACK_WIND_COOKIE (frame, afr_getxattr_cbk,
+ STACK_WIND_COOKIE (frame, afr_fgetxattr_cbk,
(void *) (long) call_child,
children[call_child],
- children[call_child]->fops->getxattr,
- loc, name);
+ children[call_child]->fops->fgetxattr,
+ fd, name, xdata);
- ret = 0;
+ op_ret = 0;
out:
- if (ret < 0)
- AFR_STACK_UNWIND (getxattr, frame, -1, op_errno, NULL);
+ if (op_ret == -1) {
+ AFR_STACK_UNWIND (fgetxattr, frame, op_ret, op_errno, NULL,
+ NULL);
+ }
return 0;
}
@@ -911,7 +1813,7 @@ int32_t
afr_readv_cbk (call_frame_t *frame, void *cookie,
xlator_t *this, int32_t op_ret, int32_t op_errno,
struct iovec *vector, int32_t count, struct iatt *buf,
- struct iobref *iobref)
+ struct iobref *iobref, dict_t *xdata)
{
afr_private_t * priv = NULL;
afr_local_t * local = NULL;
@@ -952,13 +1854,15 @@ afr_readv_cbk (call_frame_t *frame, void *cookie,
children[next_call_child],
children[next_call_child]->fops->readv,
local->fd, local->cont.readv.size,
- local->cont.readv.offset);
+ local->cont.readv.offset,
+ local->cont.readv.flags,
+ NULL);
}
out:
if (unwind) {
AFR_STACK_UNWIND (readv, frame, op_ret, op_errno,
- vector, count, buf, iobref);
+ vector, count, buf, iobref, xdata);
}
return 0;
@@ -967,7 +1871,7 @@ out:
int32_t
afr_readv (call_frame_t *frame, xlator_t *this,
- fd_t *fd, size_t size, off_t offset)
+ fd_t *fd, size_t size, off_t offset, uint32_t flags, dict_t *xdata)
{
afr_private_t * priv = NULL;
afr_local_t * local = NULL;
@@ -985,7 +1889,9 @@ afr_readv (call_frame_t *frame, xlator_t *this,
priv = this->private;
children = priv->children;
- ALLOC_OR_GOTO (frame->local, afr_local_t, out);
+ AFR_SBRAIN_CHECK_FD (fd, out);
+
+ AFR_LOCAL_ALLOC_OR_GOTO (frame->local, out);
local = frame->local;
ret = afr_local_init (local, priv, &op_errno);
@@ -1012,23 +1918,21 @@ afr_readv (call_frame_t *frame, xlator_t *this,
local->cont.readv.size = size;
local->cont.readv.offset = offset;
+ local->cont.readv.flags = flags;
+
+ afr_open_fd_fix (fd, this);
- ret = afr_open_fd_fix (frame, this, _gf_false);
- if (ret) {
- op_errno = -ret;
- goto out;
- }
STACK_WIND_COOKIE (frame, afr_readv_cbk,
(void *) (long) call_child,
children[call_child],
children[call_child]->fops->readv,
- fd, size, offset);
+ fd, size, offset, flags, xdata);
ret = 0;
out:
if (ret < 0) {
AFR_STACK_UNWIND (readv, frame, -1, op_errno, NULL, 0, NULL,
- NULL);
+ NULL, NULL);
}
return 0;
}
diff --git a/xlators/cluster/afr/src/afr-inode-read.h b/xlators/cluster/afr/src/afr-inode-read.h
index 5479cfbd5..e4091a793 100644
--- a/xlators/cluster/afr/src/afr-inode-read.h
+++ b/xlators/cluster/afr/src/afr-inode-read.h
@@ -1,20 +1,11 @@
/*
- Copyright (c) 2007-2011 Gluster, Inc. <http://www.gluster.com>
- This file is part of GlusterFS.
-
- GlusterFS is free software; you can redistribute it and/or modify
- it under the terms of the GNU General Public License as published
- by the Free Software Foundation; either version 3 of the License,
- or (at your option) any later version.
-
- GlusterFS is distributed in the hope that it will be useful, but
- WITHOUT ANY WARRANTY; without even the implied warranty of
- MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU
- General Public License for more details.
-
- You should have received a copy of the GNU General Public License
- along with this program. If not, see
- <http://www.gnu.org/licenses/>.
+ Copyright (c) 2008-2012 Red Hat, Inc. <http://www.redhat.com>
+ This file is part of GlusterFS.
+
+ This file is licensed to you under your choice of the GNU Lesser
+ General Public License, version 3 or any later version (LGPLv3 or
+ later), or the GNU General Public License, version 2 (GPLv2), in all
+ cases as published by the Free Software Foundation.
*/
#ifndef __INODE_READ_H__
@@ -22,26 +13,30 @@
int32_t
afr_access (call_frame_t *frame, xlator_t *this,
- loc_t *loc, int32_t mask);
+ loc_t *loc, int32_t mask, dict_t *xdata);
int32_t
afr_stat (call_frame_t *frame, xlator_t *this,
- loc_t *loc);
+ loc_t *loc, dict_t *xdata);
int32_t
afr_fstat (call_frame_t *frame, xlator_t *this,
- fd_t *fd);
+ fd_t *fd, dict_t *xdata);
int32_t
afr_readlink (call_frame_t *frame, xlator_t *this,
- loc_t *loc, size_t size);
+ loc_t *loc, size_t size, dict_t *xdata);
int32_t
afr_readv (call_frame_t *frame, xlator_t *this,
- fd_t *fd, size_t size, off_t offset);
+ fd_t *fd, size_t size, off_t offset, uint32_t flags, dict_t *xdata);
int32_t
afr_getxattr (call_frame_t *frame, xlator_t *this,
- loc_t *loc, const char *name);
+ loc_t *loc, const char *name, dict_t *xdata);
+
+int32_t
+afr_fgetxattr (call_frame_t *frame, xlator_t *this,
+ fd_t *fd, const char *name, dict_t *xdata);
#endif /* __INODE_READ_H__ */
diff --git a/xlators/cluster/afr/src/afr-inode-write.c b/xlators/cluster/afr/src/afr-inode-write.c
index bb8b5f0fe..c1ec69a55 100644
--- a/xlators/cluster/afr/src/afr-inode-write.c
+++ b/xlators/cluster/afr/src/afr-inode-write.c
@@ -1,20 +1,11 @@
/*
- Copyright (c) 2007-2011 Gluster, Inc. <http://www.gluster.com>
+ Copyright (c) 2008-2012 Red Hat, Inc. <http://www.redhat.com>
This file is part of GlusterFS.
- GlusterFS is free software; you can redistribute it and/or modify
- it under the terms of the GNU General Public License as published
- by the Free Software Foundation; either version 3 of the License,
- or (at your option) any later version.
-
- GlusterFS is distributed in the hope that it will be useful, but
- WITHOUT ANY WARRANTY; without even the implied warranty of
- MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU
- General Public License for more details.
-
- You should have received a copy of the GNU General Public License
- along with this program. If not, see
- <http://www.gnu.org/licenses/>.
+ This file is licensed to you under your choice of the GNU Lesser
+ General Public License, version 3 or any later version (LGPLv3 or
+ later), or the GNU General Public License, version 2 (GPLv2), in all
+ cases as published by the Free Software Foundation.
*/
@@ -48,45 +39,151 @@
#include "afr-transaction.h"
#include "afr-self-heal-common.h"
+void
+__inode_write_fop_cbk (call_frame_t *frame, int child_index, int read_child,
+ xlator_t *this, int32_t *op_ret, int32_t *op_errno,
+ struct iatt *prebuf, struct iatt *postbuf, dict_t *xdata)
+{
+ afr_local_t *local = NULL;
+
+ local = frame->local;
+
+ if (afr_fop_failed (*op_ret, *op_errno)) {
+ local->child_errno[child_index] = *op_errno;
+
+ switch (local->op) {
+ case GF_FOP_TRUNCATE:
+ case GF_FOP_FTRUNCATE:
+ if (*op_errno != EFBIG)
+ afr_transaction_fop_failed (frame, this,
+ child_index);
+ break;
+ default:
+ afr_transaction_fop_failed (frame, this, child_index);
+ break;
+ }
+ local->op_errno = *op_errno;
+ goto out;
+ }
+
+ if ((local->success_count == 0) || (read_child == child_index)) {
+ local->op_ret = *op_ret;
+ if (prebuf)
+ local->cont.inode_wfop.prebuf = *prebuf;
+ if (postbuf)
+ local->cont.inode_wfop.postbuf = *postbuf;
+ }
+
+ local->success_count++;
+out:
+ return;
+}
+
/* {{{ writev */
-int
+void
+afr_writev_copy_outvars (call_frame_t *src_frame, call_frame_t *dst_frame)
+{
+ afr_local_t *src_local = NULL;
+ afr_local_t *dst_local = NULL;
+
+ src_local = src_frame->local;
+ dst_local = dst_frame->local;
+
+ dst_local->op_ret = src_local->op_ret;
+ dst_local->op_errno = src_local->op_errno;
+ dst_local->cont.inode_wfop.prebuf = src_local->cont.inode_wfop.prebuf;
+ dst_local->cont.inode_wfop.postbuf = src_local->cont.inode_wfop.postbuf;
+}
+
+void
afr_writev_unwind (call_frame_t *frame, xlator_t *this)
{
afr_local_t * local = NULL;
- call_frame_t *main_frame = NULL;
+ local = frame->local;
+
+ AFR_STACK_UNWIND (writev, frame,
+ local->op_ret, local->op_errno,
+ &local->cont.inode_wfop.prebuf,
+ &local->cont.inode_wfop.postbuf,
+ NULL);
+}
+
+call_frame_t*
+afr_transaction_detach_fop_frame (call_frame_t *frame)
+{
+ afr_local_t * local = NULL;
+ call_frame_t *fop_frame = NULL;
local = frame->local;
LOCK (&frame->lock);
{
- if (local->transaction.main_frame)
- main_frame = local->transaction.main_frame;
+ fop_frame = local->transaction.main_frame;
local->transaction.main_frame = NULL;
}
UNLOCK (&frame->lock);
- if (main_frame) {
- AFR_STACK_UNWIND (writev, main_frame,
- local->op_ret, local->op_errno,
- &local->cont.writev.prebuf,
- &local->cont.writev.postbuf);
+ return fop_frame;
+}
+
+int
+afr_transaction_writev_unwind (call_frame_t *frame, xlator_t *this)
+{
+ call_frame_t *fop_frame = NULL;
+
+ fop_frame = afr_transaction_detach_fop_frame (frame);
+
+ if (fop_frame) {
+ afr_writev_copy_outvars (frame, fop_frame);
+ afr_writev_unwind (fop_frame, this);
}
return 0;
}
+static void
+afr_writev_handle_short_writes (call_frame_t *frame, xlator_t *this)
+{
+ afr_local_t *local = NULL;
+ afr_private_t *priv = NULL;
+ int i = 0;
+
+ local = frame->local;
+ priv = this->private;
+ /*
+ * We already have the best case result of the writev calls staged
+ * as the return value. Any writev that returns some value less
+ * than the best case is now out of sync, so mark the fop as
+ * failed. Note that fops that have returned with errors have
+ * already been marked as failed.
+ */
+ for (i = 0; i < priv->child_count; i++) {
+ if ((!local->replies[i].valid) ||
+ (local->replies[i].op_ret == -1))
+ continue;
+
+ if (local->replies[i].op_ret < local->op_ret)
+ afr_transaction_fop_failed(frame, this, i);
+ }
+}
int
afr_writev_wind_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
int32_t op_ret, int32_t op_errno, struct iatt *prebuf,
- struct iatt *postbuf)
+ struct iatt *postbuf, dict_t *xdata)
{
afr_local_t * local = NULL;
+ afr_private_t *priv = NULL;
+ call_frame_t *fop_frame = NULL;
int child_index = (long) cookie;
int call_count = -1;
int read_child = 0;
+ int ret = 0;
+ uint32_t open_fd_count = 0;
+ uint32_t write_is_append = 0;
local = frame->local;
+ priv = this->private;
read_child = afr_inode_get_read_ctx (this, local->fd->inode, NULL);
@@ -96,32 +193,81 @@ afr_writev_wind_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
local->read_child_returned = _gf_true;
}
- if (afr_fop_failed (op_ret, op_errno))
- afr_transaction_fop_failed (frame, this, child_index);
-
- if (op_ret != -1) {
- if (local->success_count == 0) {
- local->op_ret = op_ret;
- local->cont.writev.prebuf = *prebuf;
- local->cont.writev.postbuf = *postbuf;
+ __inode_write_fop_cbk (frame, child_index, read_child, this,
+ &op_ret, &op_errno, prebuf, postbuf,
+ xdata);
+
+ local->replies[child_index].valid = 1;
+ local->replies[child_index].op_ret = op_ret;
+ local->replies[child_index].op_errno = op_errno;
+
+
+ /* stage the best case return value for unwind */
+ if ((local->success_count == 0) || (op_ret > local->op_ret)) {
+ local->op_ret = op_ret;
+ local->op_errno = op_errno;
+ }
+
+ if (op_ret != -1) {
+ if (xdata) {
+ ret = dict_get_uint32 (xdata,
+ GLUSTERFS_OPEN_FD_COUNT,
+ &open_fd_count);
+ if ((ret == 0) &&
+ (open_fd_count > local->open_fd_count)) {
+ local->open_fd_count = open_fd_count;
+ local->update_open_fd_count = _gf_true;
+ }
+
+ write_is_append = 0;
+ ret = dict_get_uint32 (xdata,
+ GLUSTERFS_WRITE_IS_APPEND,
+ &write_is_append);
+ if (ret || !write_is_append)
+ local->append_write = _gf_false;
}
- if (child_index == read_child) {
- local->cont.writev.prebuf = *prebuf;
- local->cont.writev.postbuf = *postbuf;
- }
- }
-
- local->op_errno = op_errno;
+ }
}
UNLOCK (&frame->lock);
call_count = afr_frame_return (frame);
if (call_count == 0) {
- local->transaction.unwind (frame, this);
- local->transaction.resume (frame, this);
+ if (local->update_open_fd_count)
+ afr_handle_open_fd_count (frame, this);
+
+ if (!local->stable_write && !local->append_write)
+ /* An appended write removes the necessity to
+ fsync() the file. This is because self-heal
+ has the logic to check for larger file when
+ the xattrs are not reliably pointing at
+ a stale file.
+ */
+ afr_fd_report_unstable_write (this, local->fd);
+
+ afr_writev_handle_short_writes (frame, this);
+ if (afr_any_fops_failed (local, priv)) {
+ //Don't unwind until post-op is complete
+ local->transaction.resume (frame, this);
+ } else {
+ /*
+ * Generally inode-write fops do transaction.unwind then
+ * transaction.resume, but writev needs to make sure that
+ * delayed post-op frame is placed in fdctx before unwind
+ * happens. This prevents the race of flush doing the
+ * changelog wakeup first in fuse thread and then this
+ * writev placing its delayed post-op frame in fdctx.
+ * This helps flush make sure all the delayed post-ops are
+ * completed.
+ */
+
+ fop_frame = afr_transaction_detach_fop_frame (frame);
+ afr_writev_copy_outvars (frame, fop_frame);
+ local->transaction.resume (frame, this);
+ afr_writev_unwind (fop_frame, this);
+ }
}
return 0;
}
@@ -133,6 +279,8 @@ afr_writev_wind (call_frame_t *frame, xlator_t *this)
afr_private_t *priv = NULL;
int i = 0;
int call_count = -1;
+ dict_t *xdata = NULL;
+ GF_UNUSED int ret = 0;
local = frame->local;
priv = this->private;
@@ -146,6 +294,28 @@ afr_writev_wind (call_frame_t *frame, xlator_t *this)
}
local->call_count = call_count;
+ local->replies = GF_CALLOC(priv->child_count, sizeof(*local->replies),
+ gf_afr_mt_reply_t);
+ if (!local->replies) {
+ local->op_ret = -1;
+ local->op_errno = ENOMEM;
+ local->transaction.unwind(frame, this);
+ local->transaction.resume(frame, this);
+ return 0;
+ }
+
+ xdata = dict_new ();
+ if (xdata) {
+ ret = dict_set_uint32 (xdata, GLUSTERFS_OPEN_FD_COUNT,
+ sizeof (uint32_t));
+ ret = dict_set_uint32 (xdata, GLUSTERFS_WRITE_IS_APPEND,
+ 0);
+ /* Set append_write to be true speculatively. If on any
+ server it turns not be true, we unset it in the
+ callback.
+ */
+ local->append_write = _gf_true;
+ }
for (i = 0; i < priv->child_count; i++) {
if (local->transaction.pre_op[i]) {
@@ -157,13 +327,18 @@ afr_writev_wind (call_frame_t *frame, xlator_t *this)
local->cont.writev.vector,
local->cont.writev.count,
local->cont.writev.offset,
- local->cont.writev.iobref);
+ local->cont.writev.flags,
+ local->cont.writev.iobref,
+ xdata);
if (!--call_count)
break;
}
}
+ if (xdata)
+ dict_unref (xdata);
+
return 0;
}
@@ -203,7 +378,7 @@ afr_do_writev (call_frame_t *frame, xlator_t *this)
}
transaction_frame->local = local;
- frame->local = NULL;
+ AFR_LOCAL_ALLOC_OR_GOTO (frame->local, out);
local->op = GF_FOP_WRITE;
@@ -211,10 +386,17 @@ afr_do_writev (call_frame_t *frame, xlator_t *this)
local->transaction.fop = afr_writev_wind;
local->transaction.done = afr_writev_done;
- local->transaction.unwind = afr_writev_unwind;
+ local->transaction.unwind = afr_transaction_writev_unwind;
local->transaction.main_frame = frame;
if (local->fd->flags & O_APPEND) {
+ /*
+ * Backend vfs ignores the 'offset' for append mode fd so
+ * locking just the region provided for the writev does not
+ * give consistency gurantee. The actual write may happen at a
+ * completely different range than the one provided by the
+ * offset, len in the fop. So lock the entire file.
+ */
local->transaction.start = 0;
local->transaction.len = 0;
} else {
@@ -223,162 +405,91 @@ afr_do_writev (call_frame_t *frame, xlator_t *this)
local->cont.writev.count);
}
- afr_transaction (transaction_frame, this, AFR_DATA_TRANSACTION);
+ op_ret = afr_transaction (transaction_frame, this, AFR_DATA_TRANSACTION);
+ if (op_ret < 0) {
+ op_errno = -op_ret;
+ goto out;
+ }
op_ret = 0;
out:
- if (op_ret == -1) {
+ if (op_ret < 0) {
if (transaction_frame)
AFR_STACK_DESTROY (transaction_frame);
- AFR_STACK_UNWIND (writev, frame, op_ret, op_errno, NULL, NULL);
+ AFR_STACK_UNWIND (writev, frame, op_ret, op_errno, NULL, NULL, NULL);
}
return 0;
}
-static int
-afr_prepare_loc (call_frame_t *frame, fd_t *fd)
+static void
+afr_trigger_open_fd_self_heal (fd_t *fd, xlator_t *this)
{
- afr_local_t *local = NULL;
- char *name = NULL;
- char *path = NULL;
- int ret = 0;
-
- if ((!fd) || (!fd->inode))
- return -1;
-
- local = frame->local;
- ret = inode_path (fd->inode, NULL, (char **)&path);
- if (ret <= 0) {
- gf_log (frame->this->name, GF_LOG_DEBUG,
- "Unable to get path for gfid: %s",
- uuid_utoa (fd->inode->gfid));
- return -1;
- }
-
- if (local->loc.path) {
- if (strcmp (path, local->loc.path))
- gf_log (frame->this->name, GF_LOG_DEBUG,
- "overwriting old loc->path %s with %s",
- local->loc.path, path);
- GF_FREE ((char *)local->loc.path);
- }
- local->loc.path = path;
-
- name = strrchr (local->loc.path, '/');
- if (name)
- name++;
- local->loc.name = name;
-
- if (local->loc.inode) {
- inode_unref (local->loc.inode);
- }
- local->loc.inode = inode_ref (fd->inode);
-
- if (local->loc.parent) {
- inode_unref (local->loc.parent);
+ call_frame_t *frame = NULL;
+ afr_local_t *local = NULL;
+ afr_self_heal_t *sh = NULL;
+ char *reason = NULL;
+ int32_t op_errno = 0;
+ int ret = 0;
+
+ if (!fd || !fd->inode || uuid_is_null (fd->inode->gfid)) {
+ gf_log_callingfn (this->name, GF_LOG_ERROR, "Invalid args: "
+ "fd: %p, inode: %p", fd,
+ fd ? fd->inode : NULL);
+ goto out;
}
- local->loc.parent = inode_parent (local->loc.inode, 0, NULL);
-
- return 0;
-}
-
-afr_fd_paused_call_t*
-afr_paused_call_create (call_frame_t *frame)
-{
- afr_local_t *local = NULL;
- afr_fd_paused_call_t *paused_call = NULL;
+ frame = create_frame (this, this->ctx->pool);
+ if (!frame)
+ goto out;
+ AFR_LOCAL_ALLOC_OR_GOTO (frame->local, out);
local = frame->local;
- GF_ASSERT (local->fop_call_continue);
-
- paused_call = GF_CALLOC (1, sizeof (*paused_call),
- gf_afr_fd_paused_call_t);
- if (paused_call) {
- INIT_LIST_HEAD (&paused_call->call_list);
- paused_call->frame = frame;
- }
-
- return paused_call;
-}
-
-static int
-afr_pause_fd_fop (call_frame_t *frame, xlator_t *this, afr_fd_ctx_t *fd_ctx)
-{
- afr_fd_paused_call_t *paused_call = NULL;
- int ret = 0;
-
- paused_call = afr_paused_call_create (frame);
- if (paused_call)
- list_add (&paused_call->call_list, &fd_ctx->paused_calls);
- else
- ret = -ENOMEM;
-
- return ret;
-}
+ ret = afr_local_init (local, this->private, &op_errno);
+ if (ret < 0)
+ goto out;
-static void
-afr_trigger_open_fd_self_heal (call_frame_t *frame, xlator_t *this)
-{
- afr_local_t *local = NULL;
- afr_self_heal_t *sh = NULL;
- inode_t *inode = NULL;
- char *reason = NULL;
+ local->loc.inode = inode_ref (fd->inode);
+ ret = loc_path (&local->loc, NULL);
+ if (ret < 0)
+ goto out;
- local = frame->local;
sh = &local->self_heal;
- inode = local->fd->inode;
-
- sh->do_missing_entry_self_heal = _gf_true;
- sh->do_gfid_self_heal = _gf_true;
- sh->do_data_self_heal = _gf_true;
+ sh->do_metadata_self_heal = _gf_true;
+ if (fd->inode->ia_type == IA_IFREG)
+ sh->do_data_self_heal = _gf_true;
+ else if (fd->inode->ia_type == IA_IFDIR)
+ sh->do_entry_self_heal = _gf_true;
reason = "subvolume came online";
- afr_launch_self_heal (frame, this, inode, _gf_true, inode->ia_type,
- reason, NULL, NULL);
+ afr_launch_self_heal (frame, this, fd->inode, _gf_true,
+ fd->inode->ia_type, reason, NULL, NULL);
+ return;
+out:
+ AFR_STACK_DESTROY (frame);
}
-int
-afr_open_fd_fix (call_frame_t *frame, xlator_t *this, gf_boolean_t pause_fop)
-{
- int ret = 0;
- int i = 0;
- afr_fd_ctx_t *fd_ctx = NULL;
- gf_boolean_t need_self_heal = _gf_false;
- int *need_open = NULL;
- int need_open_count = 0;
- afr_local_t *local = NULL;
- afr_private_t *priv = NULL;
- gf_boolean_t fop_continue = _gf_true;
+void
+afr_open_fd_fix (fd_t *fd, xlator_t *this)
+{
+ int ret = 0;
+ int i = 0;
+ afr_fd_ctx_t *fd_ctx = NULL;
+ gf_boolean_t need_self_heal = _gf_false;
+ int *need_open = NULL;
+ size_t need_open_count = 0;
+ afr_private_t *priv = NULL;
- local = frame->local;
priv = this->private;
- GF_ASSERT (local->fd);
-
- if (fd_is_anonymous (local->fd)) {
- fop_continue = _gf_true;
- goto out;
- }
-
- if (pause_fop)
- GF_ASSERT (local->fop_call_continue);
-
- ret = afr_prepare_loc (frame, local->fd);
- if (ret < 0) {
- //File does not exist we cant open it.
- ret = 0;
+ if (!afr_is_fd_fixable (fd))
goto out;
- }
- fd_ctx = afr_fd_ctx_get (local->fd, this);
- if (!fd_ctx) {
- ret = -EINVAL;
+ fd_ctx = afr_fd_ctx_get (fd, this);
+ if (!fd_ctx)
goto out;
- }
- LOCK (&local->fd->lock);
+ LOCK (&fd->lock);
{
if (fd_ctx->up_count < priv->up_count) {
need_self_heal = _gf_true;
@@ -386,62 +497,40 @@ afr_open_fd_fix (call_frame_t *frame, xlator_t *this, gf_boolean_t pause_fop)
fd_ctx->down_count = priv->down_count;
}
+ need_open = alloca (priv->child_count * sizeof (*need_open));
for (i = 0; i < priv->child_count; i++) {
- if ((fd_ctx->opened_on[i] == AFR_FD_NOT_OPENED) &&
- local->child_up[i]) {
- fd_ctx->opened_on[i] = AFR_FD_OPENING;
- if (!need_open)
- need_open = GF_CALLOC (priv->child_count,
- sizeof (*need_open),
- gf_afr_mt_int32_t);
- need_open[i] = 1;
- need_open_count++;
- } else if (pause_fop && local->child_up[i] &&
- (fd_ctx->opened_on[i] == AFR_FD_OPENING)) {
- local->fop_paused = _gf_true;
- }
- }
+ need_open[i] = 0;
+ if (fd_ctx->opened_on[i] != AFR_FD_NOT_OPENED)
+ continue;
+
+ if (!priv->child_up[i])
+ continue;
+
+ fd_ctx->opened_on[i] = AFR_FD_OPENING;
- if (local->fop_paused) {
- GF_ASSERT (pause_fop);
- gf_log (this->name, GF_LOG_INFO, "Pause fd %p",
- local->fd);
- ret = afr_pause_fd_fop (frame, this, fd_ctx);
- if (ret)
- goto unlock;
- fop_continue = _gf_false;
+ need_open[i] = 1;
+ need_open_count++;
}
}
-unlock:
- UNLOCK (&local->fd->lock);
- if (ret) {
- gf_log (this->name, GF_LOG_ERROR, "Failed to fix fd for %s",
- local->loc.path);
- fop_continue = _gf_false;
+ UNLOCK (&fd->lock);
+ if (ret)
goto out;
- }
if (need_self_heal)
- afr_trigger_open_fd_self_heal (frame, this);
+ afr_trigger_open_fd_self_heal (fd, this);
if (!need_open_count)
goto out;
- gf_log (this->name, GF_LOG_INFO, "Opening fd %p", local->fd);
- afr_fix_open (frame, this, fd_ctx, need_open_count, need_open);
- fop_continue = _gf_false;
+ afr_fix_open (this, fd, need_open_count, need_open);
out:
- if (need_open)
- GF_FREE (need_open);
- if (fop_continue && local->fop_call_continue)
- local->fop_call_continue (frame, this);
- return ret;
+ return;
}
int
afr_writev (call_frame_t *frame, xlator_t *this, fd_t *fd,
struct iovec *vector, int32_t count, off_t offset,
- struct iobref *iobref)
+ uint32_t flags, struct iobref *iobref, dict_t *xdata)
{
afr_private_t * priv = NULL;
afr_local_t * local = NULL;
@@ -454,9 +543,14 @@ afr_writev (call_frame_t *frame, xlator_t *this, fd_t *fd,
priv = this->private;
+ if (afr_is_split_brain (this, fd->inode)) {
+ op_errno = EIO;
+ goto out;
+ }
+
QUORUM_CHECK(writev,out);
- ALLOC_OR_GOTO (frame->local, afr_local_t, out);
+ AFR_LOCAL_ALLOC_OR_GOTO (frame->local, out);
local = frame->local;
ret = afr_local_init (local, priv, &op_errno);
@@ -466,21 +560,24 @@ afr_writev (call_frame_t *frame, xlator_t *this, fd_t *fd,
local->cont.writev.vector = iov_dup (vector, count);
local->cont.writev.count = count;
local->cont.writev.offset = offset;
+ local->cont.writev.flags = flags;
local->cont.writev.iobref = iobref_ref (iobref);
local->fd = fd_ref (fd);
- local->fop_call_continue = afr_do_writev;
- ret = afr_open_fd_fix (frame, this, _gf_true);
- if (ret) {
- op_errno = -ret;
- goto out;
- }
+ /* detect here, but set it in writev_wind_cbk *after* the unstable
+ write is performed
+ */
+ local->stable_write = !!((fd->flags|flags)&(O_SYNC|O_DSYNC));
+
+ afr_open_fd_fix (fd, this);
+
+ afr_do_writev (frame, this);
ret = 0;
out:
if (ret < 0)
- AFR_STACK_UNWIND (writev, frame, -1, op_errno, NULL, NULL);
+ AFR_STACK_UNWIND (writev, frame, -1, op_errno, NULL, NULL, NULL);
return 0;
}
@@ -509,8 +606,9 @@ afr_truncate_unwind (call_frame_t *frame, xlator_t *this)
if (main_frame) {
AFR_STACK_UNWIND (truncate, main_frame, local->op_ret,
local->op_errno,
- &local->cont.truncate.prebuf,
- &local->cont.truncate.postbuf);
+ &local->cont.inode_wfop.prebuf,
+ &local->cont.inode_wfop.postbuf,
+ NULL);
}
return 0;
@@ -520,17 +618,14 @@ afr_truncate_unwind (call_frame_t *frame, xlator_t *this)
int
afr_truncate_wind_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
int32_t op_ret, int32_t op_errno, struct iatt *prebuf,
- struct iatt *postbuf)
+ struct iatt *postbuf, dict_t *xdata)
{
afr_local_t * local = NULL;
- afr_private_t * priv = NULL;
int child_index = (long) cookie;
int read_child = 0;
int call_count = -1;
- int need_unwind = 0;
local = frame->local;
- priv = this->private;
read_child = afr_inode_get_read_ctx (this, local->loc.inode, NULL);
@@ -540,38 +635,22 @@ afr_truncate_wind_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
local->read_child_returned = _gf_true;
}
- if (afr_fop_failed (op_ret, op_errno) && op_errno != EFBIG)
- afr_transaction_fop_failed (frame, this, child_index);
-
if (op_ret != -1) {
- if (local->success_count == 0) {
- local->op_ret = op_ret;
- local->cont.truncate.prebuf = *prebuf;
- local->cont.truncate.postbuf = *postbuf;
- }
-
- if (child_index == read_child) {
- local->cont.truncate.prebuf = *prebuf;
- local->cont.truncate.postbuf = *postbuf;
- }
-
- local->success_count++;
-
- if ((local->success_count >= priv->wait_count)
- && local->read_child_returned) {
- need_unwind = 1;
- }
+ if (prebuf->ia_size != postbuf->ia_size)
+ local->stable_write = _gf_false;
}
- local->op_errno = op_errno;
+ __inode_write_fop_cbk (frame, child_index, read_child, this,
+ &op_ret, &op_errno, prebuf, postbuf,
+ xdata);
}
UNLOCK (&frame->lock);
- if (need_unwind)
- local->transaction.unwind (frame, this);
-
call_count = afr_frame_return (frame);
if (call_count == 0) {
+ if (local->stable_write && afr_txn_nothing_failed (frame, this))
+ local->transaction.unwind (frame, this);
+
local->transaction.resume (frame, this);
}
@@ -599,6 +678,7 @@ afr_truncate_wind (call_frame_t *frame, xlator_t *this)
}
local->call_count = call_count;
+ local->stable_write = _gf_true;
for (i = 0; i < priv->child_count; i++) {
if (local->transaction.pre_op[i]) {
@@ -607,7 +687,8 @@ afr_truncate_wind (call_frame_t *frame, xlator_t *this)
priv->children[i],
priv->children[i]->fops->truncate,
&local->loc,
- local->cont.truncate.offset);
+ local->cont.truncate.offset,
+ NULL);
if (!--call_count)
break;
@@ -635,7 +716,7 @@ afr_truncate_done (call_frame_t *frame, xlator_t *this)
int
afr_truncate (call_frame_t *frame, xlator_t *this,
- loc_t *loc, off_t offset)
+ loc_t *loc, off_t offset, dict_t *xdata)
{
afr_private_t * priv = NULL;
afr_local_t * local = NULL;
@@ -657,7 +738,7 @@ afr_truncate (call_frame_t *frame, xlator_t *this,
goto out;
}
- ALLOC_OR_GOTO (transaction_frame->local, afr_local_t, out);
+ AFR_LOCAL_ALLOC_OR_GOTO (transaction_frame->local, out);
local = transaction_frame->local;
ret = afr_local_init (local, priv, &op_errno);
@@ -676,14 +757,18 @@ afr_truncate (call_frame_t *frame, xlator_t *this,
local->transaction.start = offset;
local->transaction.len = 0;
- afr_transaction (transaction_frame, this, AFR_DATA_TRANSACTION);
+ ret = afr_transaction (transaction_frame, this, AFR_DATA_TRANSACTION);
+ if (ret < 0) {
+ op_errno = -ret;
+ goto out;
+ }
ret = 0;
out:
if (ret < 0) {
if (transaction_frame)
AFR_STACK_DESTROY (transaction_frame);
- AFR_STACK_UNWIND (truncate, frame, -1, op_errno, NULL, NULL);
+ AFR_STACK_UNWIND (truncate, frame, -1, op_errno, NULL, NULL, NULL);
}
return 0;
@@ -714,8 +799,9 @@ afr_ftruncate_unwind (call_frame_t *frame, xlator_t *this)
if (main_frame) {
AFR_STACK_UNWIND (ftruncate, main_frame, local->op_ret,
local->op_errno,
- &local->cont.ftruncate.prebuf,
- &local->cont.ftruncate.postbuf);
+ &local->cont.inode_wfop.prebuf,
+ &local->cont.inode_wfop.postbuf,
+ NULL);
}
return 0;
}
@@ -724,17 +810,14 @@ afr_ftruncate_unwind (call_frame_t *frame, xlator_t *this)
int
afr_ftruncate_wind_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
int32_t op_ret, int32_t op_errno, struct iatt *prebuf,
- struct iatt *postbuf)
+ struct iatt *postbuf, dict_t *xdata)
{
afr_local_t * local = NULL;
- afr_private_t * priv = NULL;
int child_index = (long) cookie;
int call_count = -1;
- int need_unwind = 0;
int read_child = 0;
local = frame->local;
- priv = this->private;
read_child = afr_inode_get_read_ctx (this, local->fd->inode, NULL);
@@ -744,38 +827,22 @@ afr_ftruncate_wind_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
local->read_child_returned = _gf_true;
}
- if (afr_fop_failed (op_ret, op_errno))
- afr_transaction_fop_failed (frame, this, child_index);
-
if (op_ret != -1) {
- if (local->success_count == 0) {
- local->op_ret = op_ret;
- local->cont.ftruncate.prebuf = *prebuf;
- local->cont.ftruncate.postbuf = *postbuf;
- }
-
- if (child_index == read_child) {
- local->cont.ftruncate.prebuf = *prebuf;
- local->cont.ftruncate.postbuf = *postbuf;
- }
-
- local->success_count++;
-
- if ((local->success_count >= priv->wait_count)
- && local->read_child_returned) {
- need_unwind = 1;
- }
+ if (prebuf->ia_size != postbuf->ia_size)
+ local->stable_write = _gf_false;
}
- local->op_errno = op_errno;
+ __inode_write_fop_cbk (frame, child_index, read_child, this,
+ &op_ret, &op_errno, prebuf, postbuf,
+ xdata);
}
UNLOCK (&frame->lock);
- if (need_unwind)
- local->transaction.unwind (frame, this);
-
call_count = afr_frame_return (frame);
if (call_count == 0) {
+ if (local->stable_write && afr_txn_nothing_failed (frame, this))
+ local->transaction.unwind (frame, this);
+
local->transaction.resume (frame, this);
}
@@ -803,6 +870,7 @@ afr_ftruncate_wind (call_frame_t *frame, xlator_t *this)
}
local->call_count = call_count;
+ local->stable_write = _gf_true;
for (i = 0; i < priv->child_count; i++) {
if (local->transaction.pre_op[i]) {
@@ -810,7 +878,9 @@ afr_ftruncate_wind (call_frame_t *frame, xlator_t *this)
(void *) (long) i,
priv->children[i],
priv->children[i]->fops->ftruncate,
- local->fd, local->cont.ftruncate.offset);
+ local->fd,
+ local->cont.ftruncate.offset,
+ NULL);
if (!--call_count)
break;
@@ -865,14 +935,19 @@ afr_do_ftruncate (call_frame_t *frame, xlator_t *this)
local->transaction.start = local->cont.ftruncate.offset;
local->transaction.len = 0;
- afr_transaction (transaction_frame, this, AFR_DATA_TRANSACTION);
+ op_ret = afr_transaction (transaction_frame, this, AFR_DATA_TRANSACTION);
+ if (op_ret < 0) {
+ op_errno = -op_ret;
+ goto out;
+ }
op_ret = 0;
out:
- if (op_ret == -1) {
+ if (op_ret < 0) {
if (transaction_frame)
AFR_STACK_DESTROY (transaction_frame);
- AFR_STACK_UNWIND (ftruncate, frame, op_ret, op_errno, NULL, NULL);
+ AFR_STACK_UNWIND (ftruncate, frame, op_ret, op_errno, NULL,
+ NULL, NULL);
}
return 0;
@@ -881,7 +956,7 @@ out:
int
afr_ftruncate (call_frame_t *frame, xlator_t *this,
- fd_t *fd, off_t offset)
+ fd_t *fd, off_t offset, dict_t *xdata)
{
afr_private_t * priv = NULL;
afr_local_t * local = NULL;
@@ -895,9 +970,13 @@ afr_ftruncate (call_frame_t *frame, xlator_t *this,
priv = this->private;
+ if (afr_is_split_brain (this, fd->inode)) {
+ op_errno = EIO;
+ goto out;
+ }
QUORUM_CHECK(ftruncate,out);
- ALLOC_OR_GOTO (frame->local, afr_local_t, out);
+ AFR_LOCAL_ALLOC_OR_GOTO (frame->local, out);
local = frame->local;
ret = afr_local_init (local, priv, &op_errno);
@@ -907,20 +986,17 @@ afr_ftruncate (call_frame_t *frame, xlator_t *this,
local->cont.ftruncate.offset = offset;
local->fd = fd_ref (fd);
- local->fop_call_continue = afr_do_ftruncate;
- ret = afr_open_fd_fix (frame, this, _gf_true);
- if (ret) {
- op_errno = -ret;
- goto out;
- }
+ afr_open_fd_fix (fd, this);
+
+ afr_do_ftruncate (frame, this);
ret = 0;
out:
if (ret < 0) {
if (transaction_frame)
AFR_STACK_DESTROY (transaction_frame);
- AFR_STACK_UNWIND (ftruncate, frame, -1, op_errno, NULL, NULL);
+ AFR_STACK_UNWIND (ftruncate, frame, -1, op_errno, NULL, NULL, NULL);
}
return 0;
@@ -949,8 +1025,9 @@ afr_setattr_unwind (call_frame_t *frame, xlator_t *this)
if (main_frame) {
AFR_STACK_UNWIND (setattr, main_frame, local->op_ret,
local->op_errno,
- &local->cont.setattr.preop_buf,
- &local->cont.setattr.postop_buf);
+ &local->cont.inode_wfop.prebuf,
+ &local->cont.inode_wfop.postbuf,
+ NULL);
}
return 0;
@@ -960,7 +1037,7 @@ afr_setattr_unwind (call_frame_t *frame, xlator_t *this)
int
afr_setattr_wind_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
int32_t op_ret, int32_t op_errno,
- struct iatt *preop, struct iatt *postop)
+ struct iatt *preop, struct iatt *postop, dict_t *xdata)
{
afr_local_t * local = NULL;
afr_private_t * priv = NULL;
@@ -980,29 +1057,14 @@ afr_setattr_wind_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
local->read_child_returned = _gf_true;
}
- if (afr_fop_failed (op_ret, op_errno))
- afr_transaction_fop_failed (frame, this, child_index);
-
- if (op_ret != -1) {
- if (local->success_count == 0) {
- local->op_ret = op_ret;
- local->cont.setattr.preop_buf = *preop;
- local->cont.setattr.postop_buf = *postop;
- }
-
- if (child_index == read_child) {
- local->cont.setattr.preop_buf = *preop;
- local->cont.setattr.postop_buf = *postop;
- }
-
- local->success_count++;
+ __inode_write_fop_cbk (frame, child_index, read_child, this,
+ &op_ret, &op_errno, preop, postop,
+ xdata);
- if ((local->success_count >= priv->wait_count)
- && local->read_child_returned) {
- need_unwind = 1;
- }
+ if ((local->success_count >= priv->wait_count)
+ && local->read_child_returned) {
+ need_unwind = 1;
}
- local->op_errno = op_errno;
}
UNLOCK (&frame->lock);
@@ -1048,7 +1110,8 @@ afr_setattr_wind (call_frame_t *frame, xlator_t *this)
priv->children[i]->fops->setattr,
&local->loc,
&local->cont.setattr.in_buf,
- local->cont.setattr.valid);
+ local->cont.setattr.valid,
+ NULL);
if (!--call_count)
break;
@@ -1076,7 +1139,7 @@ afr_setattr_done (call_frame_t *frame, xlator_t *this)
int
afr_setattr (call_frame_t *frame, xlator_t *this,
- loc_t *loc, struct iatt *buf, int32_t valid)
+ loc_t *loc, struct iatt *buf, int32_t valid, dict_t *xdata)
{
afr_private_t * priv = NULL;
afr_local_t * local = NULL;
@@ -1098,7 +1161,7 @@ afr_setattr (call_frame_t *frame, xlator_t *this,
goto out;
}
- ALLOC_OR_GOTO (transaction_frame->local, afr_local_t, out);
+ AFR_LOCAL_ALLOC_OR_GOTO (transaction_frame->local, out);
local = transaction_frame->local;
ret = afr_local_init (local, priv, &op_errno);
@@ -1118,14 +1181,18 @@ afr_setattr (call_frame_t *frame, xlator_t *this,
local->transaction.start = LLONG_MAX - 1;
local->transaction.len = 0;
- afr_transaction (transaction_frame, this, AFR_METADATA_TRANSACTION);
+ ret = afr_transaction (transaction_frame, this, AFR_METADATA_TRANSACTION);
+ if (ret < 0) {
+ op_errno = -ret;
+ goto out;
+ }
ret = 0;
out:
if (ret < 0) {
if (transaction_frame)
AFR_STACK_DESTROY (transaction_frame);
- AFR_STACK_UNWIND (setattr, frame, -1, op_errno, NULL, NULL);
+ AFR_STACK_UNWIND (setattr, frame, -1, op_errno, NULL, NULL, NULL);
}
return 0;
@@ -1152,8 +1219,9 @@ afr_fsetattr_unwind (call_frame_t *frame, xlator_t *this)
if (main_frame) {
AFR_STACK_UNWIND (fsetattr, main_frame, local->op_ret,
local->op_errno,
- &local->cont.fsetattr.preop_buf,
- &local->cont.fsetattr.postop_buf);
+ &local->cont.inode_wfop.prebuf,
+ &local->cont.inode_wfop.postbuf,
+ NULL);
}
return 0;
@@ -1163,7 +1231,7 @@ afr_fsetattr_unwind (call_frame_t *frame, xlator_t *this)
int
afr_fsetattr_wind_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
int32_t op_ret, int32_t op_errno,
- struct iatt *preop, struct iatt *postop)
+ struct iatt *preop, struct iatt *postop, dict_t *xdata)
{
afr_local_t * local = NULL;
afr_private_t * priv = NULL;
@@ -1183,29 +1251,14 @@ afr_fsetattr_wind_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
local->read_child_returned = _gf_true;
}
- if (afr_fop_failed (op_ret, op_errno))
- afr_transaction_fop_failed (frame, this, child_index);
+ __inode_write_fop_cbk (frame, child_index, read_child, this,
+ &op_ret, &op_errno, preop, postop,
+ xdata);
- if (op_ret != -1) {
- if (local->success_count == 0) {
- local->op_ret = op_ret;
- local->cont.fsetattr.preop_buf = *preop;
- local->cont.fsetattr.postop_buf = *postop;
- }
-
- if (child_index == read_child) {
- local->cont.fsetattr.preop_buf = *preop;
- local->cont.fsetattr.postop_buf = *postop;
- }
-
- local->success_count++;
-
- if ((local->success_count >= priv->wait_count)
- && local->read_child_returned) {
- need_unwind = 1;
- }
+ if ((local->success_count >= priv->wait_count)
+ && local->read_child_returned) {
+ need_unwind = 1;
}
- local->op_errno = op_errno;
}
UNLOCK (&frame->lock);
@@ -1251,7 +1304,8 @@ afr_fsetattr_wind (call_frame_t *frame, xlator_t *this)
priv->children[i]->fops->fsetattr,
local->fd,
&local->cont.fsetattr.in_buf,
- local->cont.fsetattr.valid);
+ local->cont.fsetattr.valid,
+ NULL);
if (!--call_count)
break;
@@ -1278,7 +1332,7 @@ afr_fsetattr_done (call_frame_t *frame, xlator_t *this)
int
afr_fsetattr (call_frame_t *frame, xlator_t *this,
- fd_t *fd, struct iatt *buf, int32_t valid)
+ fd_t *fd, struct iatt *buf, int32_t valid, dict_t *xdata)
{
afr_private_t * priv = NULL;
afr_local_t * local = NULL;
@@ -1292,6 +1346,11 @@ afr_fsetattr (call_frame_t *frame, xlator_t *this,
priv = this->private;
+ if (afr_is_split_brain (this, fd->inode)) {
+ op_errno = EIO;
+ goto out;
+ }
+
QUORUM_CHECK(fsetattr,out);
transaction_frame = copy_frame (frame);
@@ -1300,7 +1359,7 @@ afr_fsetattr (call_frame_t *frame, xlator_t *this,
goto out;
}
- ALLOC_OR_GOTO (transaction_frame->local, afr_local_t, out);
+ AFR_LOCAL_ALLOC_OR_GOTO (transaction_frame->local, out);
local = transaction_frame->local;
ret = afr_local_init (local, priv, &op_errno);
@@ -1316,24 +1375,24 @@ afr_fsetattr (call_frame_t *frame, xlator_t *this,
local->fd = fd_ref (fd);
- ret = afr_open_fd_fix (transaction_frame, this, _gf_false);
- if (ret) {
- op_errno = -ret;
- goto out;
- }
+ afr_open_fd_fix (fd, this);
local->transaction.main_frame = frame;
local->transaction.start = LLONG_MAX - 1;
local->transaction.len = 0;
- afr_transaction (transaction_frame, this, AFR_METADATA_TRANSACTION);
+ ret = afr_transaction (transaction_frame, this, AFR_METADATA_TRANSACTION);
+ if (ret < 0) {
+ op_errno = -ret;
+ goto out;
+ }
ret = 0;
out:
if (ret < 0) {
if (transaction_frame)
AFR_STACK_DESTROY (transaction_frame);
- AFR_STACK_UNWIND (fsetattr, frame, -1, op_errno, NULL, NULL);
+ AFR_STACK_UNWIND (fsetattr, frame, -1, op_errno, NULL, NULL, NULL);
}
return 0;
@@ -1361,7 +1420,8 @@ afr_setxattr_unwind (call_frame_t *frame, xlator_t *this)
if (main_frame) {
AFR_STACK_UNWIND (setxattr, main_frame,
- local->op_ret, local->op_errno);
+ local->op_ret, local->op_errno,
+ NULL);
}
return 0;
}
@@ -1369,30 +1429,25 @@ afr_setxattr_unwind (call_frame_t *frame, xlator_t *this)
int
afr_setxattr_wind_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
- int32_t op_ret, int32_t op_errno)
+ int32_t op_ret, int32_t op_errno, dict_t *xdata)
{
- afr_local_t *local = NULL;
- afr_private_t *priv = NULL;
- int call_count = -1;
- int need_unwind = 0;
+ afr_local_t *local = NULL;
+ afr_private_t *priv = NULL;
+ int call_count = -1;
+ int need_unwind = 0;
+ int child_index = (long) cookie;
local = frame->local;
priv = this->private;
LOCK (&frame->lock);
{
- if (op_ret != -1) {
- if (local->success_count == 0) {
- local->op_ret = op_ret;
- }
- local->success_count++;
-
- if (local->success_count == priv->child_count) {
- need_unwind = 1;
- }
+ __inode_write_fop_cbk (frame, child_index, -1, this,
+ &op_ret, &op_errno, NULL, NULL,
+ xdata);
+ if (local->success_count == priv->child_count) {
+ need_unwind = 1;
}
-
- local->op_errno = op_errno;
}
UNLOCK (&frame->lock);
@@ -1438,7 +1493,8 @@ afr_setxattr_wind (call_frame_t *frame, xlator_t *this)
priv->children[i]->fops->setxattr,
&local->loc,
local->cont.setxattr.dict,
- local->cont.setxattr.flags);
+ local->cont.setxattr.flags,
+ NULL);
if (!--call_count)
break;
@@ -1463,24 +1519,24 @@ afr_setxattr_done (call_frame_t *frame, xlator_t *this)
int
afr_setxattr (call_frame_t *frame, xlator_t *this,
- loc_t *loc, dict_t *dict, int32_t flags)
+ loc_t *loc, dict_t *dict, int32_t flags, dict_t *xdata)
{
afr_private_t *priv = NULL;
afr_local_t *local = NULL;
call_frame_t *transaction_frame = NULL;
- data_pair_t *trav = NULL;
int ret = -1;
int op_errno = EINVAL;
- VALIDATE_OR_GOTO (frame, out);
VALIDATE_OR_GOTO (this, out);
- VALIDATE_OR_GOTO (this->private, out);
GF_IF_INTERNAL_XATTR_GOTO ("trusted.afr.*", dict,
- trav, op_errno, out);
+ op_errno, out);
GF_IF_INTERNAL_XATTR_GOTO ("trusted.glusterfs.afr.*", dict,
- trav, op_errno, out);
+ op_errno, out);
+
+ VALIDATE_OR_GOTO (frame, out);
+ VALIDATE_OR_GOTO (this->private, out);
priv = this->private;
@@ -1491,7 +1547,7 @@ afr_setxattr (call_frame_t *frame, xlator_t *this,
goto out;
}
- ALLOC_OR_GOTO (transaction_frame->local, afr_local_t, out);
+ AFR_LOCAL_ALLOC_OR_GOTO (transaction_frame->local, out);
local = transaction_frame->local;
ret = afr_local_init (local, priv, &op_errno);
@@ -1511,14 +1567,18 @@ afr_setxattr (call_frame_t *frame, xlator_t *this,
local->transaction.start = LLONG_MAX - 1;
local->transaction.len = 0;
- afr_transaction (transaction_frame, this, AFR_METADATA_TRANSACTION);
+ ret = afr_transaction (transaction_frame, this, AFR_METADATA_TRANSACTION);
+ if (ret < 0) {
+ op_errno = -ret;
+ goto out;
+ }
ret = 0;
out:
if (ret < 0) {
if (transaction_frame)
AFR_STACK_DESTROY (transaction_frame);
- AFR_STACK_UNWIND (setxattr, frame, -1, op_errno);
+ AFR_STACK_UNWIND (setxattr, frame, -1, op_errno, NULL);
}
return 0;
@@ -1545,7 +1605,8 @@ afr_fsetxattr_unwind (call_frame_t *frame, xlator_t *this)
if (main_frame) {
AFR_STACK_UNWIND (fsetxattr, main_frame,
- local->op_ret, local->op_errno);
+ local->op_ret, local->op_errno,
+ NULL);
}
return 0;
}
@@ -1553,30 +1614,26 @@ afr_fsetxattr_unwind (call_frame_t *frame, xlator_t *this)
int
afr_fsetxattr_wind_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
- int32_t op_ret, int32_t op_errno)
+ int32_t op_ret, int32_t op_errno, dict_t *xdata)
{
- afr_local_t *local = NULL;
- afr_private_t *priv = NULL;
- int call_count = -1;
- int need_unwind = 0;
+ afr_local_t *local = NULL;
+ afr_private_t *priv = NULL;
+ int call_count = -1;
+ int need_unwind = 0;
+ int child_index = (long) cookie;
local = frame->local;
priv = this->private;
LOCK (&frame->lock);
{
- if (op_ret != -1) {
- if (local->success_count == 0) {
- local->op_ret = op_ret;
- }
- local->success_count++;
- if (local->success_count == priv->child_count) {
- need_unwind = 1;
- }
+ __inode_write_fop_cbk (frame, child_index, -1, this,
+ &op_ret, &op_errno, NULL, NULL,
+ xdata);
+ if (local->success_count == priv->child_count) {
+ need_unwind = 1;
}
-
- local->op_errno = op_errno;
}
UNLOCK (&frame->lock);
@@ -1622,7 +1679,8 @@ afr_fsetxattr_wind (call_frame_t *frame, xlator_t *this)
priv->children[i]->fops->fsetxattr,
local->fd,
local->cont.fsetxattr.dict,
- local->cont.fsetxattr.flags);
+ local->cont.fsetxattr.flags,
+ NULL);
if (!--call_count)
break;
@@ -1647,33 +1705,34 @@ afr_fsetxattr_done (call_frame_t *frame, xlator_t *this)
int
afr_fsetxattr (call_frame_t *frame, xlator_t *this,
- fd_t *fd, dict_t *dict, int32_t flags)
+ fd_t *fd, dict_t *dict, int32_t flags, dict_t *xdata)
{
afr_private_t *priv = NULL;
afr_local_t *local = NULL;
call_frame_t *transaction_frame = NULL;
int ret = -1;
int op_errno = EINVAL;
- data_pair_t *trav = NULL;
VALIDATE_OR_GOTO (frame, out);
VALIDATE_OR_GOTO (this, out);
VALIDATE_OR_GOTO (this->private, out);
GF_IF_INTERNAL_XATTR_GOTO ("trusted.afr.*", dict,
- trav, op_errno, out);
+ op_errno, out);
GF_IF_INTERNAL_XATTR_GOTO ("trusted.glusterfs.afr.*", dict,
- trav, op_errno, out);
-
- if (ret)
- goto out;
+ op_errno, out);
priv = this->private;
+ if (afr_is_split_brain (this, fd->inode)) {
+ op_errno = EIO;
+ goto out;
+ }
+
QUORUM_CHECK(fsetxattr,out);
- ALLOC_OR_GOTO (local, afr_local_t, out);
+ AFR_LOCAL_ALLOC_OR_GOTO (local, out);
ret = afr_local_init (local, priv, &op_errno);
if (ret < 0)
@@ -1701,14 +1760,18 @@ afr_fsetxattr (call_frame_t *frame, xlator_t *this,
local->transaction.start = LLONG_MAX - 1;
local->transaction.len = 0;
- afr_transaction (transaction_frame, this, AFR_METADATA_TRANSACTION);
+ ret = afr_transaction (transaction_frame, this, AFR_METADATA_TRANSACTION);
+ if (ret < 0) {
+ op_errno = -ret;
+ goto out;
+ }
ret = 0;
out:
if (ret < 0) {
if (transaction_frame)
AFR_STACK_DESTROY (transaction_frame);
- AFR_STACK_UNWIND (fsetxattr, frame, -1, op_errno);
+ AFR_STACK_UNWIND (fsetxattr, frame, -1, op_errno, NULL);
}
return 0;
@@ -1716,6 +1779,7 @@ out:
/* }}} */
+
/* {{{ removexattr */
@@ -1737,38 +1801,34 @@ afr_removexattr_unwind (call_frame_t *frame, xlator_t *this)
if (main_frame) {
AFR_STACK_UNWIND (removexattr, main_frame,
- local->op_ret, local->op_errno)
- }
+ local->op_ret, local->op_errno,
+ NULL);
+ }
return 0;
}
int
afr_removexattr_wind_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
- int32_t op_ret, int32_t op_errno)
+ int32_t op_ret, int32_t op_errno, dict_t *xdata)
{
- afr_local_t * local = NULL;
- afr_private_t * priv = NULL;
- int call_count = -1;
- int need_unwind = 0;
+ afr_local_t *local = NULL;
+ afr_private_t *priv = NULL;
+ int call_count = -1;
+ int need_unwind = 0;
+ int child_index = (long) cookie;
local = frame->local;
priv = this->private;
LOCK (&frame->lock);
{
- if (op_ret != -1) {
- if (local->success_count == 0) {
- local->op_ret = op_ret;
- }
- local->success_count++;
-
- if (local->success_count == priv->wait_count) {
- need_unwind = 1;
- }
+ __inode_write_fop_cbk (frame, child_index, -1, this,
+ &op_ret, &op_errno, NULL, NULL,
+ xdata);
+ if (local->success_count == priv->wait_count) {
+ need_unwind = 1;
}
-
- local->op_errno = op_errno;
}
UNLOCK (&frame->lock);
@@ -1813,7 +1873,8 @@ afr_removexattr_wind (call_frame_t *frame, xlator_t *this)
priv->children[i],
priv->children[i]->fops->removexattr,
&local->loc,
- local->cont.removexattr.name);
+ local->cont.removexattr.name,
+ NULL);
if (!--call_count)
break;
@@ -1839,16 +1900,23 @@ afr_removexattr_done (call_frame_t *frame, xlator_t *this)
int
afr_removexattr (call_frame_t *frame, xlator_t *this,
- loc_t *loc, const char *name)
+ loc_t *loc, const char *name, dict_t *xdata)
{
- afr_private_t * priv = NULL;
- afr_local_t * local = NULL;
- call_frame_t *transaction_frame = NULL;
- int ret = -1;
- int op_errno = 0;
+ afr_private_t *priv = NULL;
+ afr_local_t *local = NULL;
+ call_frame_t *transaction_frame = NULL;
+ int ret = -1;
+ int op_errno = 0;
- VALIDATE_OR_GOTO (frame, out);
VALIDATE_OR_GOTO (this, out);
+
+ GF_IF_NATIVE_XATTR_GOTO ("trusted.afr.*",
+ name, op_errno, out);
+
+ GF_IF_NATIVE_XATTR_GOTO ("trusted.glusterfs.afr.*",
+ name, op_errno, out);
+
+ VALIDATE_OR_GOTO (frame, out);
VALIDATE_OR_GOTO (this->private, out);
VALIDATE_OR_GOTO (loc, out);
@@ -1862,7 +1930,7 @@ afr_removexattr (call_frame_t *frame, xlator_t *this,
goto out;
}
- ALLOC_OR_GOTO (transaction_frame->local, afr_local_t, out);
+ AFR_LOCAL_ALLOC_OR_GOTO (transaction_frame->local, out);
local = transaction_frame->local;
ret = afr_local_init (local, priv, &op_errno);
@@ -1881,15 +1949,913 @@ afr_removexattr (call_frame_t *frame, xlator_t *this,
local->transaction.start = LLONG_MAX - 1;
local->transaction.len = 0;
- afr_transaction (transaction_frame, this, AFR_METADATA_TRANSACTION);
+ ret = afr_transaction (transaction_frame, this, AFR_METADATA_TRANSACTION);
+ if (ret < 0) {
+ op_errno = -ret;
+ goto out;
+ }
ret = 0;
out:
if (ret < 0) {
if (transaction_frame)
AFR_STACK_DESTROY (transaction_frame);
- AFR_STACK_UNWIND (removexattr, frame, -1, op_errno);
+ AFR_STACK_UNWIND (removexattr, frame, -1, op_errno, NULL);
+ }
+
+ return 0;
+}
+
+/* ffremovexattr */
+int
+afr_fremovexattr_unwind (call_frame_t *frame, xlator_t *this)
+{
+ afr_local_t * local = NULL;
+ call_frame_t *main_frame = NULL;
+
+ local = frame->local;
+
+ LOCK (&frame->lock);
+ {
+ if (local->transaction.main_frame)
+ main_frame = local->transaction.main_frame;
+ local->transaction.main_frame = NULL;
+ }
+ UNLOCK (&frame->lock);
+
+ if (main_frame) {
+ AFR_STACK_UNWIND (fremovexattr, main_frame,
+ local->op_ret, local->op_errno,
+ NULL);
+ }
+ return 0;
+}
+
+
+int
+afr_fremovexattr_wind_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
+ int32_t op_ret, int32_t op_errno, dict_t *xdata)
+{
+ afr_local_t * local = NULL;
+ afr_private_t * priv = NULL;
+ int call_count = -1;
+ int need_unwind = 0;
+ int child_index = (long) cookie;
+
+ local = frame->local;
+ priv = this->private;
+
+ LOCK (&frame->lock);
+ {
+ __inode_write_fop_cbk (frame, child_index, -1, this,
+ &op_ret, &op_errno, NULL, NULL,
+ xdata);
+
+ if (local->success_count == priv->wait_count) {
+ need_unwind = 1;
+ }
+ }
+ UNLOCK (&frame->lock);
+
+ if (need_unwind)
+ local->transaction.unwind (frame, this);
+
+ call_count = afr_frame_return (frame);
+
+ if (call_count == 0) {
+ local->transaction.resume (frame, this);
+ }
+
+ return 0;
+}
+
+
+int32_t
+afr_fremovexattr_wind (call_frame_t *frame, xlator_t *this)
+{
+ afr_local_t *local = NULL;
+ afr_private_t *priv = NULL;
+ int call_count = -1;
+ int i = 0;
+
+ local = frame->local;
+ priv = this->private;
+
+ call_count = afr_pre_op_done_children_count (local->transaction.pre_op,
+ priv->child_count);
+
+ if (call_count == 0) {
+ local->transaction.resume (frame, this);
+ return 0;
+ }
+
+ local->call_count = call_count;
+
+ for (i = 0; i < priv->child_count; i++) {
+ if (local->transaction.pre_op[i]) {
+ STACK_WIND_COOKIE (frame, afr_fremovexattr_wind_cbk,
+ (void *) (long) i,
+ priv->children[i],
+ priv->children[i]->fops->fremovexattr,
+ local->fd,
+ local->cont.removexattr.name,
+ NULL);
+
+ if (!--call_count)
+ break;
+ }
+ }
+
+ return 0;
+}
+
+
+int
+afr_fremovexattr_done (call_frame_t *frame, xlator_t *this)
+{
+ afr_local_t * local = frame->local;
+
+ local->transaction.unwind (frame, this);
+
+ AFR_STACK_DESTROY (frame);
+
+ return 0;
+}
+
+
+int
+afr_fremovexattr (call_frame_t *frame, xlator_t *this,
+ fd_t *fd, const char *name, dict_t *xdata)
+{
+ afr_private_t * priv = NULL;
+ afr_local_t * local = NULL;
+ call_frame_t *transaction_frame = NULL;
+ int ret = -1;
+ int op_ret = -1;
+ int op_errno = 0;
+
+ VALIDATE_OR_GOTO (this, out);
+
+ GF_IF_NATIVE_XATTR_GOTO ("trusted.afr.*",
+ name, op_errno, out);
+
+ GF_IF_NATIVE_XATTR_GOTO ("trusted.glusterfs.afr.*",
+ name, op_errno, out);
+
+ VALIDATE_OR_GOTO (frame, out);
+ VALIDATE_OR_GOTO (this->private, out);
+
+ priv = this->private;
+ if (afr_is_split_brain (this, fd->inode)) {
+ op_errno = EIO;
+ goto out;
+ }
+
+ QUORUM_CHECK(fremovexattr, out);
+
+ transaction_frame = copy_frame (frame);
+ if (!transaction_frame) {
+ goto out;
+ }
+
+ AFR_LOCAL_ALLOC_OR_GOTO (local, out);
+
+ ret = afr_local_init (local, priv, &op_errno);
+ if (ret < 0) {
+ op_errno = -ret;
+ goto out;
+ }
+
+ transaction_frame->local = local;
+
+ local->op_ret = -1;
+
+ local->cont.removexattr.name = gf_strdup (name);
+
+ local->transaction.fop = afr_fremovexattr_wind;
+ local->transaction.done = afr_fremovexattr_done;
+ local->transaction.unwind = afr_fremovexattr_unwind;
+
+ local->fd = fd_ref (fd);
+
+ local->transaction.main_frame = frame;
+ local->transaction.start = LLONG_MAX - 1;
+ local->transaction.len = 0;
+
+ op_ret = afr_transaction (transaction_frame, this, AFR_METADATA_TRANSACTION);
+ if (op_ret < 0) {
+ op_errno = -op_ret;
+ goto out;
+ }
+
+ op_ret = 0;
+out:
+ if (op_ret < 0) {
+ if (transaction_frame)
+ AFR_STACK_DESTROY (transaction_frame);
+ AFR_STACK_UNWIND (fremovexattr, frame, op_ret, op_errno, NULL);
+ }
+
+ return 0;
+}
+
+static int
+afr_fallocate_unwind (call_frame_t *frame, xlator_t *this)
+{
+ afr_local_t * local = NULL;
+ call_frame_t *main_frame = NULL;
+
+ local = frame->local;
+
+ LOCK (&frame->lock);
+ {
+ if (local->transaction.main_frame)
+ main_frame = local->transaction.main_frame;
+ local->transaction.main_frame = NULL;
}
+ UNLOCK (&frame->lock);
+ if (main_frame) {
+ AFR_STACK_UNWIND (fallocate, main_frame, local->op_ret,
+ local->op_errno,
+ &local->cont.inode_wfop.prebuf,
+ &local->cont.inode_wfop.postbuf,
+ NULL);
+ }
return 0;
}
+
+static int
+afr_fallocate_wind_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
+ int32_t op_ret, int32_t op_errno, struct iatt *prebuf,
+ struct iatt *postbuf, dict_t *xdata)
+{
+ afr_local_t * local = NULL;
+ afr_private_t * priv = NULL;
+ int child_index = (long) cookie;
+ int call_count = -1;
+ int need_unwind = 0;
+ int read_child = 0;
+
+ local = frame->local;
+ priv = this->private;
+
+ read_child = afr_inode_get_read_ctx (this, local->fd->inode, NULL);
+
+ LOCK (&frame->lock);
+ {
+ if (child_index == read_child) {
+ local->read_child_returned = _gf_true;
+ }
+
+ __inode_write_fop_cbk (frame, child_index, read_child, this,
+ &op_ret, &op_errno, prebuf, postbuf,
+ xdata);
+
+ if ((local->success_count >= priv->wait_count)
+ && local->read_child_returned) {
+ need_unwind = 1;
+ }
+ }
+ UNLOCK (&frame->lock);
+
+ if (need_unwind)
+ local->transaction.unwind (frame, this);
+
+ call_count = afr_frame_return (frame);
+
+ if (call_count == 0) {
+ local->transaction.resume (frame, this);
+ }
+
+ return 0;
+}
+
+static int
+afr_fallocate_wind (call_frame_t *frame, xlator_t *this)
+{
+ afr_local_t *local = NULL;
+ afr_private_t *priv = NULL;
+ int call_count = -1;
+ int i = 0;
+
+ local = frame->local;
+ priv = this->private;
+
+ call_count = afr_pre_op_done_children_count (local->transaction.pre_op,
+ priv->child_count);
+
+ if (call_count == 0) {
+ local->transaction.resume (frame, this);
+ return 0;
+ }
+
+ local->call_count = call_count;
+
+ for (i = 0; i < priv->child_count; i++) {
+ if (local->transaction.pre_op[i]) {
+ STACK_WIND_COOKIE (frame, afr_fallocate_wind_cbk,
+ (void *) (long) i,
+ priv->children[i],
+ priv->children[i]->fops->fallocate,
+ local->fd,
+ local->cont.fallocate.mode,
+ local->cont.fallocate.offset,
+ local->cont.fallocate.len,
+ NULL);
+
+ if (!--call_count)
+ break;
+ }
+ }
+
+ return 0;
+}
+
+static int
+afr_fallocate_done (call_frame_t *frame, xlator_t *this)
+{
+ afr_local_t *local = NULL;
+
+ local = frame->local;
+
+ local->transaction.unwind (frame, this);
+
+ AFR_STACK_DESTROY (frame);
+
+ return 0;
+}
+
+static int
+afr_do_fallocate (call_frame_t *frame, xlator_t *this)
+{
+ call_frame_t * transaction_frame = NULL;
+ afr_local_t * local = NULL;
+ int op_ret = -1;
+ int op_errno = 0;
+
+ local = frame->local;
+
+ transaction_frame = copy_frame (frame);
+ if (!transaction_frame) {
+ goto out;
+ }
+
+ transaction_frame->local = local;
+ frame->local = NULL;
+
+ local->op = GF_FOP_FALLOCATE;
+
+ local->transaction.fop = afr_fallocate_wind;
+ local->transaction.done = afr_fallocate_done;
+ local->transaction.unwind = afr_fallocate_unwind;
+
+ local->transaction.main_frame = frame;
+
+ local->transaction.start = local->cont.fallocate.offset;
+ local->transaction.len = 0;
+
+ /* fallocate can modify the file size */
+ op_ret = afr_transaction (transaction_frame, this, AFR_DATA_TRANSACTION);
+ if (op_ret < 0) {
+ op_errno = -op_ret;
+ goto out;
+ }
+
+ op_ret = 0;
+out:
+ if (op_ret < 0) {
+ if (transaction_frame)
+ AFR_STACK_DESTROY (transaction_frame);
+ AFR_STACK_UNWIND (fallocate, frame, op_ret, op_errno, NULL,
+ NULL, NULL);
+ }
+
+ return 0;
+}
+
+int
+afr_fallocate (call_frame_t *frame, xlator_t *this, fd_t *fd, int32_t mode,
+ off_t offset, size_t len, dict_t *xdata)
+{
+ afr_private_t * priv = NULL;
+ afr_local_t * local = NULL;
+ call_frame_t *transaction_frame = NULL;
+ int ret = -1;
+ int op_errno = 0;
+
+ VALIDATE_OR_GOTO (frame, out);
+ VALIDATE_OR_GOTO (this, out);
+ VALIDATE_OR_GOTO (this->private, out);
+
+ priv = this->private;
+
+ if (afr_is_split_brain (this, fd->inode)) {
+ op_errno = EIO;
+ goto out;
+ }
+ QUORUM_CHECK(fallocate,out);
+
+ AFR_LOCAL_ALLOC_OR_GOTO (frame->local, out);
+ local = frame->local;
+
+ ret = afr_local_init (local, priv, &op_errno);
+ if (ret < 0)
+ goto out;
+
+ local->cont.fallocate.mode = mode;
+ local->cont.fallocate.offset = offset;
+ local->cont.fallocate.len = len;
+
+ local->fd = fd_ref (fd);
+
+ afr_open_fd_fix (fd, this);
+
+ afr_do_fallocate (frame, this);
+
+ ret = 0;
+out:
+ if (ret < 0) {
+ if (transaction_frame)
+ AFR_STACK_DESTROY (transaction_frame);
+ AFR_STACK_UNWIND (fallocate, frame, -1, op_errno, NULL, NULL, NULL);
+ }
+
+ return 0;
+}
+
+/* }}} */
+
+/* {{{ discard */
+
+static int
+afr_discard_unwind (call_frame_t *frame, xlator_t *this)
+{
+ afr_local_t * local = NULL;
+ call_frame_t *main_frame = NULL;
+
+ local = frame->local;
+
+ LOCK (&frame->lock);
+ {
+ if (local->transaction.main_frame)
+ main_frame = local->transaction.main_frame;
+ local->transaction.main_frame = NULL;
+ }
+ UNLOCK (&frame->lock);
+
+ if (main_frame) {
+ AFR_STACK_UNWIND (discard, main_frame, local->op_ret,
+ local->op_errno,
+ &local->cont.inode_wfop.prebuf,
+ &local->cont.inode_wfop.postbuf,
+ NULL);
+ }
+ return 0;
+}
+
+static int
+afr_discard_wind_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
+ int32_t op_ret, int32_t op_errno, struct iatt *prebuf,
+ struct iatt *postbuf, dict_t *xdata)
+{
+ afr_local_t * local = NULL;
+ afr_private_t * priv = NULL;
+ int child_index = (long) cookie;
+ int call_count = -1;
+ int need_unwind = 0;
+ int read_child = 0;
+
+ local = frame->local;
+ priv = this->private;
+
+ read_child = afr_inode_get_read_ctx (this, local->fd->inode, NULL);
+
+ LOCK (&frame->lock);
+ {
+ if (child_index == read_child) {
+ local->read_child_returned = _gf_true;
+ }
+
+ __inode_write_fop_cbk (frame, child_index, read_child, this,
+ &op_ret, &op_errno, prebuf, postbuf,
+ xdata);
+
+ if ((local->success_count >= priv->wait_count)
+ && local->read_child_returned) {
+ need_unwind = 1;
+ }
+ }
+ UNLOCK (&frame->lock);
+
+ if (need_unwind)
+ local->transaction.unwind (frame, this);
+
+ call_count = afr_frame_return (frame);
+
+ if (call_count == 0) {
+ local->transaction.resume (frame, this);
+ }
+
+ return 0;
+}
+
+static int
+afr_discard_wind (call_frame_t *frame, xlator_t *this)
+{
+ afr_local_t *local = NULL;
+ afr_private_t *priv = NULL;
+ int call_count = -1;
+ int i = 0;
+
+ local = frame->local;
+ priv = this->private;
+
+ call_count = afr_pre_op_done_children_count (local->transaction.pre_op,
+ priv->child_count);
+
+ if (call_count == 0) {
+ local->transaction.resume (frame, this);
+ return 0;
+ }
+
+ local->call_count = call_count;
+
+ for (i = 0; i < priv->child_count; i++) {
+ if (local->transaction.pre_op[i]) {
+ STACK_WIND_COOKIE (frame, afr_discard_wind_cbk,
+ (void *) (long) i,
+ priv->children[i],
+ priv->children[i]->fops->discard,
+ local->fd,
+ local->cont.discard.offset,
+ local->cont.discard.len,
+ NULL);
+
+ if (!--call_count)
+ break;
+ }
+ }
+
+ return 0;
+}
+
+static int
+afr_discard_done (call_frame_t *frame, xlator_t *this)
+{
+ afr_local_t *local = NULL;
+
+ local = frame->local;
+
+ local->transaction.unwind (frame, this);
+
+ AFR_STACK_DESTROY (frame);
+
+ return 0;
+}
+
+static int
+afr_do_discard (call_frame_t *frame, xlator_t *this)
+{
+ call_frame_t * transaction_frame = NULL;
+ afr_local_t * local = NULL;
+ int op_ret = -1;
+ int op_errno = 0;
+
+ local = frame->local;
+
+ transaction_frame = copy_frame (frame);
+ if (!transaction_frame) {
+ goto out;
+ }
+
+ transaction_frame->local = local;
+ frame->local = NULL;
+
+ local->op = GF_FOP_DISCARD;
+
+ local->transaction.fop = afr_discard_wind;
+ local->transaction.done = afr_discard_done;
+ local->transaction.unwind = afr_discard_unwind;
+
+ local->transaction.main_frame = frame;
+
+ local->transaction.start = local->cont.discard.offset;
+ local->transaction.len = 0;
+
+ op_ret = afr_transaction (transaction_frame, this, AFR_DATA_TRANSACTION);
+ if (op_ret < 0) {
+ op_errno = -op_ret;
+ goto out;
+ }
+
+ op_ret = 0;
+out:
+ if (op_ret < 0) {
+ if (transaction_frame)
+ AFR_STACK_DESTROY (transaction_frame);
+ AFR_STACK_UNWIND (discard, frame, op_ret, op_errno, NULL,
+ NULL, NULL);
+ }
+
+ return 0;
+}
+
+int
+afr_discard (call_frame_t *frame, xlator_t *this, fd_t *fd, off_t offset,
+ size_t len, dict_t *xdata)
+{
+ afr_private_t * priv = NULL;
+ afr_local_t * local = NULL;
+ call_frame_t *transaction_frame = NULL;
+ int ret = -1;
+ int op_errno = 0;
+
+ VALIDATE_OR_GOTO (frame, out);
+ VALIDATE_OR_GOTO (this, out);
+ VALIDATE_OR_GOTO (this->private, out);
+
+ priv = this->private;
+
+ if (afr_is_split_brain (this, fd->inode)) {
+ op_errno = EIO;
+ goto out;
+ }
+ QUORUM_CHECK(discard, out);
+
+ AFR_LOCAL_ALLOC_OR_GOTO (frame->local, out);
+ local = frame->local;
+
+ ret = afr_local_init (local, priv, &op_errno);
+ if (ret < 0)
+ goto out;
+
+ local->cont.discard.offset = offset;
+ local->cont.discard.len = len;
+
+ local->fd = fd_ref (fd);
+
+ afr_open_fd_fix (fd, this);
+
+ afr_do_discard(frame, this);
+
+ ret = 0;
+out:
+ if (ret < 0) {
+ if (transaction_frame)
+ AFR_STACK_DESTROY (transaction_frame);
+ AFR_STACK_UNWIND (discard, frame, -1, op_errno, NULL, NULL, NULL);
+ }
+
+ return 0;
+}
+
+
+/* {{{ zerofill */
+
+static int
+afr_zerofill_unwind (call_frame_t *frame, xlator_t *this)
+{
+ afr_local_t *local = NULL;
+ call_frame_t *main_frame = NULL;
+
+ local = frame->local;
+
+ LOCK (&frame->lock);
+ {
+ if (local->transaction.main_frame) {
+ main_frame = local->transaction.main_frame;
+ }
+ local->transaction.main_frame = NULL;
+ }
+ UNLOCK (&frame->lock);
+
+ if (main_frame) {
+ AFR_STACK_UNWIND (zerofill, main_frame, local->op_ret,
+ local->op_errno,
+ &local->cont.zerofill.prebuf,
+ &local->cont.zerofill.postbuf,
+ NULL);
+ }
+ return 0;
+}
+
+static int
+afr_zerofill_wind_cbk(call_frame_t *frame, void *cookie, xlator_t *this,
+ int32_t op_ret, int32_t op_errno, struct iatt *prebuf,
+ struct iatt *postbuf, dict_t *xdata)
+{
+ afr_local_t *local = NULL;
+ afr_private_t *priv = NULL;
+ int child_index = (long) cookie;
+ int call_count = -1;
+ int need_unwind = 0;
+ int read_child = 0;
+
+ local = frame->local;
+ priv = this->private;
+
+ read_child = afr_inode_get_read_ctx (this, local->fd->inode, NULL);
+
+ LOCK (&frame->lock);
+ {
+ if (child_index == read_child) {
+ local->read_child_returned = _gf_true;
+ }
+
+ if (afr_fop_failed (op_ret, op_errno)) {
+ afr_transaction_fop_failed (frame, this, child_index);
+ }
+
+ if (op_ret != -1) {
+ if (local->success_count == 0) {
+ local->op_ret = op_ret;
+ local->cont.zerofill.prebuf = *prebuf;
+ local->cont.zerofill.postbuf = *postbuf;
+ }
+
+ if (child_index == read_child) {
+ local->cont.zerofill.prebuf = *prebuf;
+ local->cont.zerofill.postbuf = *postbuf;
+ }
+
+ local->success_count++;
+
+ if ((local->success_count >= priv->wait_count)
+ && local->read_child_returned) {
+ need_unwind = 1;
+ }
+ }
+ local->op_errno = op_errno;
+ }
+ UNLOCK (&frame->lock);
+
+ if (need_unwind) {
+ local->transaction.unwind (frame, this);
+ }
+ call_count = afr_frame_return (frame);
+
+ if (call_count == 0) {
+ local->transaction.resume (frame, this);
+ }
+
+ return 0;
+}
+
+static int
+afr_zerofill_wind (call_frame_t *frame, xlator_t *this)
+{
+ afr_local_t *local = NULL;
+ afr_private_t *priv = NULL;
+ int call_count = -1;
+ int i = 0;
+
+ local = frame->local;
+ priv = this->private;
+
+ call_count = afr_pre_op_done_children_count (local->transaction.pre_op,
+ priv->child_count);
+
+ if (call_count == 0) {
+ local->transaction.resume (frame, this);
+ return 0;
+ }
+
+ local->call_count = call_count;
+
+ for (i = 0; i < priv->child_count; i++) {
+ if (local->transaction.pre_op[i]) {
+ STACK_WIND_COOKIE (frame, afr_zerofill_wind_cbk,
+ (void *) (long) i,
+ priv->children[i],
+ priv->children[i]->fops->zerofill,
+ local->fd,
+ local->cont.zerofill.offset,
+ local->cont.zerofill.len,
+ NULL);
+
+ if (!--call_count)
+ break;
+ }
+ }
+
+ return 0;
+}
+
+static int
+afr_zerofill_done (call_frame_t *frame, xlator_t *this)
+{
+ afr_local_t *local = NULL;
+
+ local = frame->local;
+
+ local->transaction.unwind (frame, this);
+
+ AFR_STACK_DESTROY (frame);
+
+ return 0;
+}
+
+static int
+afr_do_zerofill(call_frame_t *frame, xlator_t *this)
+{
+ call_frame_t *transaction_frame = NULL;
+ afr_local_t *local = NULL;
+ int op_ret = -1;
+ int op_errno = 0;
+
+ local = frame->local;
+
+ transaction_frame = copy_frame (frame);
+ if (!transaction_frame) {
+ goto out;
+ }
+
+ transaction_frame->local = local;
+ frame->local = NULL;
+
+ local->op = GF_FOP_ZEROFILL;
+
+ local->transaction.fop = afr_zerofill_wind;
+ local->transaction.done = afr_zerofill_done;
+ local->transaction.unwind = afr_zerofill_unwind;
+
+ local->transaction.main_frame = frame;
+
+ local->transaction.start = local->cont.zerofill.offset;
+ local->transaction.len = 0;
+
+ op_ret = afr_transaction (transaction_frame, this,
+ AFR_DATA_TRANSACTION);
+ if (op_ret < 0) {
+ op_errno = -op_ret;
+ goto out;
+ }
+
+ op_ret = 0;
+out:
+ if (op_ret < 0) {
+ if (transaction_frame) {
+ AFR_STACK_DESTROY (transaction_frame);
+ }
+ AFR_STACK_UNWIND (zerofill, frame, op_ret, op_errno, NULL,
+ NULL, NULL);
+ }
+
+ return 0;
+}
+
+int
+afr_zerofill (call_frame_t *frame, xlator_t *this, fd_t *fd, off_t offset,
+ size_t len, dict_t *xdata)
+{
+ afr_private_t *priv = NULL;
+ afr_local_t *local = NULL;
+ call_frame_t *transaction_frame = NULL;
+ int ret = -1;
+ int op_errno = 0;
+
+ VALIDATE_OR_GOTO (frame, out);
+ VALIDATE_OR_GOTO (this, out);
+ VALIDATE_OR_GOTO (this->private, out);
+
+ priv = this->private;
+
+ if (afr_is_split_brain (this, fd->inode)) {
+ op_errno = EIO;
+ goto out;
+ }
+ QUORUM_CHECK(zerofill, out);
+
+ AFR_LOCAL_ALLOC_OR_GOTO (frame->local, out);
+ local = frame->local;
+
+ ret = afr_local_init (local, priv, &op_errno);
+ if (ret < 0) {
+ goto out;
+ }
+ local->cont.zerofill.offset = offset;
+ local->cont.zerofill.len = len;
+
+ local->fd = fd_ref (fd);
+
+ afr_open_fd_fix (fd, this);
+
+ afr_do_zerofill(frame, this);
+
+ ret = 0;
+out:
+ if (ret < 0) {
+ if (transaction_frame) {
+ AFR_STACK_DESTROY (transaction_frame);
+ }
+ AFR_STACK_UNWIND (zerofill, frame, -1, op_errno, NULL,
+ NULL, NULL);
+ }
+
+ return 0;
+}
+
+/* }}} */
+
+
diff --git a/xlators/cluster/afr/src/afr-inode-write.h b/xlators/cluster/afr/src/afr-inode-write.h
index a2c3520f9..8e93ca44a 100644
--- a/xlators/cluster/afr/src/afr-inode-write.h
+++ b/xlators/cluster/afr/src/afr-inode-write.h
@@ -1,20 +1,11 @@
/*
- Copyright (c) 2007-2011 Gluster, Inc. <http://www.gluster.com>
- This file is part of GlusterFS.
-
- GlusterFS is free software; you can redistribute it and/or modify
- it under the terms of the GNU General Public License as published
- by the Free Software Foundation; either version 3 of the License,
- or (at your option) any later version.
-
- GlusterFS is distributed in the hope that it will be useful, but
- WITHOUT ANY WARRANTY; without even the implied warranty of
- MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU
- General Public License for more details.
-
- You should have received a copy of the GNU General Public License
- along with this program. If not, see
- <http://www.gnu.org/licenses/>.
+ Copyright (c) 2008-2012 Red Hat, Inc. <http://www.redhat.com>
+ This file is part of GlusterFS.
+
+ This file is licensed to you under your choice of the GNU Lesser
+ General Public License, version 3 or any later version (LGPLv3 or
+ later), or the GNU General Public License, version 2 (GPLv2), in all
+ cases as published by the Free Software Foundation.
*/
#ifndef __INODE_WRITE_H__
@@ -22,55 +13,70 @@
int32_t
afr_chmod (call_frame_t *frame, xlator_t *this,
- loc_t *loc, mode_t mode);
+ loc_t *loc, mode_t mode, dict_t *xdata);
int32_t
afr_chown (call_frame_t *frame, xlator_t *this,
- loc_t *loc, uid_t uid, gid_t gid);
+ loc_t *loc, uid_t uid, gid_t gid, dict_t *xdata);
int
afr_fchown (call_frame_t *frame, xlator_t *this,
- fd_t *fd, uid_t uid, gid_t gid);
+ fd_t *fd, uid_t uid, gid_t gid, dict_t *xdata);
int32_t
afr_fchmod (call_frame_t *frame, xlator_t *this,
- fd_t *fd, mode_t mode);
+ fd_t *fd, mode_t mode, dict_t *xdata);
int32_t
-afr_writev (call_frame_t *frame, xlator_t *this, fd_t *fd,
+afr_writev (call_frame_t *frame, xlator_t *this, fd_t *fd,
struct iovec *vector, int32_t count, off_t offset,
- struct iobref *iobref);
+ uint32_t flags, struct iobref *iobref, dict_t *xdata);
int32_t
afr_truncate (call_frame_t *frame, xlator_t *this,
- loc_t *loc, off_t offset);
+ loc_t *loc, off_t offset, dict_t *xdata);
int32_t
afr_ftruncate (call_frame_t *frame, xlator_t *this,
- fd_t *fd, off_t offset);
+ fd_t *fd, off_t offset, dict_t *xdata);
int32_t
afr_utimens (call_frame_t *frame, xlator_t *this,
- loc_t *loc, struct timespec tv[2]);
+ loc_t *loc, struct timespec tv[2], dict_t *xdata);
int
afr_setattr (call_frame_t *frame, xlator_t *this,
- loc_t *loc, struct iatt *buf, int32_t valid);
+ loc_t *loc, struct iatt *buf, int32_t valid, dict_t *xdata);
int
afr_fsetattr (call_frame_t *frame, xlator_t *this,
- fd_t *fd, struct iatt *buf, int32_t valid);
+ fd_t *fd, struct iatt *buf, int32_t valid, dict_t *xdata);
int32_t
afr_setxattr (call_frame_t *frame, xlator_t *this,
- loc_t *loc, dict_t *dict, int32_t flags);
+ loc_t *loc, dict_t *dict, int32_t flags, dict_t *xdata);
int32_t
afr_fsetxattr (call_frame_t *frame, xlator_t *this,
- fd_t *fd, dict_t *dict, int32_t flags);
+ fd_t *fd, dict_t *dict, int32_t flags, dict_t *xdata);
int32_t
afr_removexattr (call_frame_t *frame, xlator_t *this,
- loc_t *loc, const char *name);
+ loc_t *loc, const char *name, dict_t *xdata);
+
+int32_t
+afr_fremovexattr (call_frame_t *frame, xlator_t *this,
+ fd_t *fd, const char *name, dict_t *xdata);
+
+int
+afr_discard (call_frame_t *frame, xlator_t *this, fd_t *fd, off_t offset,
+ size_t len, dict_t *xdata);
+
+int
+afr_fallocate (call_frame_t *frame, xlator_t *this, fd_t *fd, int32_t mode,
+ off_t offset, size_t len, dict_t *xdata);
+int
+afr_zerofill(call_frame_t *frame, xlator_t *this, fd_t *fd, off_t offset,
+ size_t len, dict_t *xdata);
#endif /* __INODE_WRITE_H__ */
diff --git a/xlators/cluster/afr/src/afr-lk-common.c b/xlators/cluster/afr/src/afr-lk-common.c
index 2fe134990..060d78f35 100644
--- a/xlators/cluster/afr/src/afr-lk-common.c
+++ b/xlators/cluster/afr/src/afr-lk-common.c
@@ -1,20 +1,11 @@
/*
- Copyright (c) 2007-2011 Gluster, Inc. <http://www.gluster.com>
+ Copyright (c) 2008-2012 Red Hat, Inc. <http://www.redhat.com>
This file is part of GlusterFS.
- GlusterFS is free software; you can redistribute it and/or modify
- it under the terms of the GNU General Public License as published
- by the Free Software Foundation; either version 3 of the License,
- or (at your option) any later version.
-
- GlusterFS is distributed in the hope that it will be useful, but
- WITHOUT ANY WARRANTY; without even the implied warranty of
- MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU
- General Public License for more details.
-
- You should have received a copy of the GNU General Public License
- along with this program. If not, see
- <http://www.gnu.org/licenses/>.
+ This file is licensed to you under your choice of the GNU Lesser
+ General Public License, version 3 or any later version (LGPLv3 or
+ later), or the GNU General Public License, version 2 (GPLv2), in all
+ cases as published by the Free Software Foundation.
*/
#include "dict.h"
@@ -31,8 +22,69 @@
#define LOCKED_YES 0x1 /* for DATA, METADATA, ENTRY and higher_path */
#define LOCKED_LOWER 0x2 /* for lower path */
+#define AFR_TRACE_INODELK_IN(frame, this, params ...) \
+ do { \
+ afr_private_t *_priv = this->private; \
+ if (!_priv->inodelk_trace) \
+ break; \
+ afr_trace_inodelk_in (frame, this, params); \
+ } while (0);
+
+#define AFR_TRACE_INODELK_OUT(frame, this, params ...) \
+ do { \
+ afr_private_t *_priv = this->private; \
+ if (!_priv->inodelk_trace) \
+ break; \
+ afr_trace_inodelk_out (frame, this, params); \
+ } while (0);
+
+#define AFR_TRACE_ENTRYLK_IN(frame, this, params ...) \
+ do { \
+ afr_private_t *_priv = this->private; \
+ if (!_priv->entrylk_trace) \
+ break; \
+ afr_trace_entrylk_in (frame, this, params); \
+ } while (0);
+
+#define AFR_TRACE_ENTRYLK_OUT(frame, this, params ...) \
+ do { \
+ afr_private_t *_priv = this->private; \
+ if (!_priv->entrylk_trace) \
+ break; \
+ afr_trace_entrylk_out (frame, this, params); \
+ } while (0);
+
int
-afr_lock_blocking (call_frame_t *frame, xlator_t *this, int child_index);
+afr_entry_lockee_cmp (const void *l1, const void *l2)
+{
+ const afr_entry_lockee_t *r1 = l1;
+ const afr_entry_lockee_t *r2 = l2;
+ int ret = 0;
+ uuid_t gfid1 = {0};
+ uuid_t gfid2 = {0};
+
+ loc_gfid ((loc_t*)&r1->loc, gfid1);
+ loc_gfid ((loc_t*)&r2->loc, gfid2);
+ ret = uuid_compare (gfid1, gfid2);
+ /*Entrylks with NULL basename are the 'smallest'*/
+ if (ret == 0) {
+ if (!r1->basename)
+ return -1;
+ if (!r2->basename)
+ return 1;
+ ret = strcmp (r1->basename, r2->basename);
+ }
+
+ if (ret <= 0)
+ return -1;
+ else
+ return 1;
+}
+
+int afr_lock_blocking (call_frame_t *frame, xlator_t *this, int child_index);
+
+static int
+afr_copy_locked_nodes (call_frame_t *frame, xlator_t *this);
static uint64_t afr_lock_number = 1;
@@ -57,13 +109,13 @@ afr_set_lock_number (call_frame_t *frame, xlator_t *this)
}
void
-afr_set_lk_owner (call_frame_t *frame, xlator_t *this)
+afr_set_lk_owner (call_frame_t *frame, xlator_t *this, void *lk_owner)
{
gf_log (this->name, GF_LOG_TRACE,
"Setting lk-owner=%llu",
- (unsigned long long) (unsigned long)frame->root);
+ (unsigned long long) (unsigned long)lk_owner);
- set_lk_owner_from_ptr (&frame->root->lk_owner, frame->root);
+ set_lk_owner_from_ptr (&frame->root->lk_owner, lk_owner);
}
static int
@@ -99,16 +151,9 @@ internal_lock_count (call_frame_t *frame, xlator_t *this)
local = frame->local;
priv = this->private;
- if (local->fd) {
- for (i = 0; i < priv->child_count; i++) {
- if (local->child_up[i] && local->fd_open_on[i])
- ++call_count;
- }
- } else {
- for (i = 0; i < priv->child_count; i++) {
- if (local->child_up[i])
- ++call_count;
- }
+ for (i = 0; i < priv->child_count; i++) {
+ if (local->child_up[i])
+ ++call_count;
}
return call_count;
@@ -242,27 +287,20 @@ afr_set_lock_call_type (afr_lock_call_type_t lock_call_type,
}
static void
-afr_trace_inodelk_out (call_frame_t *frame, afr_lock_call_type_t lock_call_type,
+afr_trace_inodelk_out (call_frame_t *frame, xlator_t *this,
+ afr_lock_call_type_t lock_call_type,
afr_lock_op_type_t lk_op_type, struct gf_flock *flock,
int op_ret, int op_errno, int32_t child_index)
{
- xlator_t *this = NULL;
afr_internal_lock_t *int_lock = NULL;
afr_local_t *local = NULL;
- afr_private_t *priv = NULL;
char lockee[256];
char lock_call_type_str[256];
char verdict[16];
- this = THIS;
local = frame->local;
int_lock = &local->internal_lock;
- priv = this->private;
-
- if (!priv->inodelk_trace) {
- return;
- }
afr_print_lockee (lockee, 256, &local->loc, local->fd, child_index);
@@ -271,37 +309,29 @@ afr_trace_inodelk_out (call_frame_t *frame, afr_lock_call_type_t lock_call_type,
afr_print_verdict (op_ret, op_errno, verdict);
gf_log (this->name, GF_LOG_INFO,
- "[%s %s] [%s] Lockee={%s} Number={%llu}",
+ "[%s %s] [%s] lk-owner=%s Lockee={%s} Number={%llu}",
lock_call_type_str,
lk_op_type == AFR_LOCK_OP ? "LOCK REPLY" : "UNLOCK REPLY",
- verdict,
- lockee,
+ verdict, lkowner_utoa (&frame->root->lk_owner), lockee,
(unsigned long long) int_lock->lock_number);
}
static void
-afr_trace_inodelk_in (call_frame_t *frame, afr_lock_call_type_t lock_call_type,
+afr_trace_inodelk_in (call_frame_t *frame, xlator_t *this,
+ afr_lock_call_type_t lock_call_type,
afr_lock_op_type_t lk_op_type, struct gf_flock *flock,
int32_t cmd, int32_t child_index)
{
- xlator_t *this = NULL;
afr_local_t *local = NULL;
afr_internal_lock_t *int_lock = NULL;
- afr_private_t *priv = NULL;
char lock[256];
char lockee[256];
char lock_call_type_str[256];
- this = THIS;
local = frame->local;
int_lock = &local->internal_lock;
- priv = this->private;
-
- if (!priv->inodelk_trace) {
- return;
- }
afr_print_inodelk (lock, 256, cmd, flock, &frame->root->lk_owner);
afr_print_lockee (lockee, 256, &local->loc, local->fd, child_index);
@@ -318,20 +348,21 @@ afr_trace_inodelk_in (call_frame_t *frame, afr_lock_call_type_t lock_call_type,
}
static void
-afr_trace_entrylk_in (call_frame_t *frame, afr_lock_call_type_t lock_call_type,
+afr_trace_entrylk_in (call_frame_t *frame, xlator_t *this,
+ afr_lock_call_type_t lock_call_type,
afr_lock_op_type_t lk_op_type, const char *basename,
- int32_t child_index)
+ int32_t cookie)
{
- xlator_t *this = NULL;
afr_local_t *local = NULL;
afr_internal_lock_t *int_lock = NULL;
afr_private_t *priv = NULL;
+ int child_index = 0;
+ int lockee_no = 0;
char lock[256];
char lockee[256];
char lock_call_type_str[256];
- this = THIS;
local = frame->local;
int_lock = &local->internal_lock;
priv = this->private;
@@ -339,36 +370,41 @@ afr_trace_entrylk_in (call_frame_t *frame, afr_lock_call_type_t lock_call_type,
if (!priv->entrylk_trace) {
return;
}
+ lockee_no = cookie / priv->child_count;
+ child_index = cookie % priv->child_count;
afr_print_entrylk (lock, 256, basename, &frame->root->lk_owner);
- afr_print_lockee (lockee, 256, &local->loc, local->fd, child_index);
+ afr_print_lockee (lockee, 256, &int_lock->lockee[lockee_no].loc, local->fd,
+ child_index);
afr_set_lock_call_type (lock_call_type, lock_call_type_str, int_lock);
gf_log (this->name, GF_LOG_INFO,
- "[%s %s] Lock={%s} Lockee={%s} Number={%llu}",
+ "[%s %s] Lock={%s} Lockee={%s} Number={%llu}, Cookie={%d}",
lock_call_type_str,
lk_op_type == AFR_LOCK_OP ? "LOCK REQUEST" : "UNLOCK REQUEST",
lock, lockee,
- (unsigned long long) int_lock->lock_number);
+ (unsigned long long) int_lock->lock_number,
+ cookie);
}
static void
-afr_trace_entrylk_out (call_frame_t *frame, afr_lock_call_type_t lock_call_type,
- afr_lock_op_type_t lk_op_type, const char *basename, int op_ret,
- int op_errno, int32_t child_index)
+afr_trace_entrylk_out (call_frame_t *frame, xlator_t *this,
+ afr_lock_call_type_t lock_call_type,
+ afr_lock_op_type_t lk_op_type, const char *basename,
+ int op_ret, int op_errno, int32_t cookie)
{
- xlator_t *this = NULL;
afr_internal_lock_t *int_lock = NULL;
afr_local_t *local = NULL;
afr_private_t *priv = NULL;
+ int lockee_no = 0;
+ int child_index = 0;
char lock[256];
char lockee[256];
char lock_call_type_str[256];
char verdict[16];
- this = THIS;
local = frame->local;
int_lock = &local->internal_lock;
priv = this->private;
@@ -376,20 +412,25 @@ afr_trace_entrylk_out (call_frame_t *frame, afr_lock_call_type_t lock_call_type,
if (!priv->entrylk_trace) {
return;
}
+ lockee_no = cookie / priv->child_count;
+ child_index = cookie % priv->child_count;
- afr_print_lockee (lockee, 256, &local->loc, local->fd, child_index);
+ afr_print_entrylk (lock, 256, basename, &frame->root->lk_owner);
+ afr_print_lockee (lockee, 256, &int_lock->lockee[lockee_no].loc, local->fd,
+ child_index);
afr_set_lock_call_type (lock_call_type, lock_call_type_str, int_lock);
afr_print_verdict (op_ret, op_errno, verdict);
gf_log (this->name, GF_LOG_INFO,
- "[%s %s] [%s] Lock={%s} Lockee={%s} Number={%llu}",
+ "[%s %s] [%s] Lock={%s} Lockee={%s} Number={%llu} Cookie={%d}",
lock_call_type_str,
lk_op_type == AFR_LOCK_OP ? "LOCK REPLY" : "UNLOCK REPLY",
verdict,
lock, lockee,
- (unsigned long long) int_lock->lock_number);
+ (unsigned long long) int_lock->lock_number,
+ cookie);
}
@@ -442,6 +483,47 @@ is_afr_lock_transaction (afr_local_t *local)
return ret;
}
+int
+afr_init_entry_lockee (afr_entry_lockee_t *lockee, afr_local_t *local,
+ loc_t *loc, char *basename, int child_count)
+{
+ int ret = -1;
+
+ loc_copy (&lockee->loc, loc);
+ lockee->basename = (basename)? gf_strdup (basename): NULL;
+ if (basename && !lockee->basename)
+ goto out;
+
+ lockee->locked_count = 0;
+ lockee->locked_nodes = GF_CALLOC (child_count,
+ sizeof (*lockee->locked_nodes),
+ gf_afr_mt_afr_node_character);
+
+ if (!lockee->locked_nodes)
+ goto out;
+
+ ret = 0;
+out:
+ return ret;
+
+}
+
+void
+afr_entry_lockee_cleanup (afr_internal_lock_t *int_lock)
+{
+ int i = 0;
+
+ for (i = 0; i < int_lock->lockee_count; i++) {
+ loc_wipe (&int_lock->lockee[i].loc);
+ if (int_lock->lockee[i].basename)
+ GF_FREE (int_lock->lockee[i].basename);
+ if (int_lock->lockee[i].locked_nodes)
+ GF_FREE (int_lock->lockee[i].locked_nodes);
+ }
+
+ return;
+}
+
static int
initialize_entrylk_variables (call_frame_t *frame, xlator_t *this)
{
@@ -459,8 +541,13 @@ initialize_entrylk_variables (call_frame_t *frame, xlator_t *this)
int_lock->lock_op_ret = -1;
int_lock->lock_op_errno = 0;
- for (i = 0; i < priv->child_count; i++) {
- int_lock->entry_locked_nodes[i] = 0;
+ for (i = 0; i < AFR_LOCKEE_COUNT_MAX; i++) {
+ if (!int_lock->lockee[i].locked_nodes)
+ break;
+ int_lock->lockee[i].locked_count = 0;
+ memset (int_lock->lockee[i].locked_nodes, 0,
+ sizeof (*int_lock->lockee[i].locked_nodes) *
+ priv->child_count);
}
return 0;
@@ -472,19 +559,23 @@ initialize_inodelk_variables (call_frame_t *frame, xlator_t *this)
afr_local_t *local = NULL;
afr_internal_lock_t *int_lock = NULL;
afr_private_t *priv = NULL;
- int i = 0;
+ afr_inodelk_t *inodelk = NULL;
priv = this->private;
local = frame->local;
int_lock = &local->internal_lock;
- int_lock->inodelk_lock_count = 0;
- int_lock->lock_op_ret = -1;
- int_lock->lock_op_errno = 0;
+ inodelk = afr_get_inodelk (int_lock, int_lock->domain);
- for (i = 0; i < priv->child_count; i++) {
- int_lock->inode_locked_nodes[i] = 0;
- }
+ inodelk->lock_count = 0;
+ int_lock->lk_attempted_count = 0;
+ int_lock->lock_op_ret = -1;
+ int_lock->lock_op_errno = 0;
+
+ memset (inodelk->locked_nodes, 0,
+ sizeof (*inodelk->locked_nodes) * priv->child_count);
+ memset (int_lock->locked_nodes, 0,
+ sizeof (*int_lock->locked_nodes) * priv->child_count);
return 0;
}
@@ -494,7 +585,7 @@ lower_path (loc_t *l1, const char *b1, loc_t *l2, const char *b2)
{
int ret = 0;
- ret = strcmp (l1->path, l2->path);
+ ret = uuid_compare (l1->inode->gfid, l2->inode->gfid);
if (ret == 0)
ret = strcmp (b1, b2);
@@ -506,6 +597,18 @@ lower_path (loc_t *l1, const char *b1, loc_t *l2, const char *b2)
}
int
+afr_lockee_locked_nodes_count (afr_internal_lock_t *int_lock)
+{
+ int call_count = 0;
+ int i = 0;
+
+ for (i = 0; i < int_lock->lockee_count; i++)
+ call_count += int_lock->lockee[i].locked_count;
+
+ return call_count;
+}
+
+int
afr_locked_nodes_count (unsigned char *locked_nodes, int child_count)
{
@@ -523,7 +626,7 @@ afr_locked_nodes_count (unsigned char *locked_nodes, int child_count)
/* FIXME: What if UNLOCK fails */
static int32_t
afr_unlock_common_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
- int32_t op_ret, int32_t op_errno)
+ int32_t op_ret, int32_t op_errno, dict_t *xdata)
{
afr_local_t *local = NULL;
afr_internal_lock_t *int_lock = NULL;
@@ -549,33 +652,37 @@ afr_unlock_common_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
static int32_t
afr_unlock_inodelk_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
- int32_t op_ret, int32_t op_errno)
+ int32_t op_ret, int32_t op_errno, dict_t *xdata)
{
afr_local_t *local = NULL;
afr_internal_lock_t *int_lock = NULL;
+ afr_inodelk_t *inodelk = NULL;
int32_t child_index = (long)cookie;
+ afr_private_t *priv = NULL;
local = frame->local;
int_lock = &local->internal_lock;
- afr_trace_inodelk_out (frame, AFR_INODELK_TRANSACTION,
+ AFR_TRACE_INODELK_OUT (frame, this, AFR_INODELK_TRANSACTION,
AFR_UNLOCK_OP, NULL, op_ret,
op_errno, child_index);
+ priv = this->private;
+
if (op_ret < 0 && op_errno != ENOTCONN && op_errno != EBADFD) {
- gf_log (this->name, GF_LOG_ERROR,
- "%s: unlock failed on %d, reason: %s",
- local->loc.path, child_index, strerror (op_errno));
+ gf_log (this->name, GF_LOG_INFO, "%s: unlock failed on subvolume %s "
+ "with lock owner %s", local->loc.path,
+ priv->children[child_index]->name,
+ lkowner_utoa (&frame->root->lk_owner));
}
- int_lock->inode_locked_nodes[child_index] &= LOCKED_NO;
-
- if (op_ret == 1) {
+ inodelk = afr_get_inodelk (int_lock, int_lock->domain);
+ inodelk->locked_nodes[child_index] &= LOCKED_NO;
+ if (local->transaction.eager_lock)
local->transaction.eager_lock[child_index] = 0;
- }
- afr_unlock_common_cbk (frame, cookie, this, op_ret, op_errno);
+ afr_unlock_common_cbk (frame, cookie, this, op_ret, op_errno, xdata);
return 0;
@@ -585,9 +692,12 @@ static int
afr_unlock_inodelk (call_frame_t *frame, xlator_t *this)
{
afr_internal_lock_t *int_lock = NULL;
+ afr_inodelk_t *inodelk = NULL;
afr_local_t *local = NULL;
afr_private_t *priv = NULL;
struct gf_flock flock = {0,};
+ struct gf_flock full_flock = {0,};
+ struct gf_flock *flock_use = NULL;
int call_count = 0;
int i = 0;
int piggyback = 0;
@@ -598,15 +708,14 @@ afr_unlock_inodelk (call_frame_t *frame, xlator_t *this)
int_lock = &local->internal_lock;
priv = this->private;
- flock.l_start = int_lock->lk_flock.l_start;
- flock.l_len = int_lock->lk_flock.l_len;
- flock.l_type = F_UNLCK;
+ inodelk = afr_get_inodelk (int_lock, int_lock->domain);
- gf_log (this->name, GF_LOG_DEBUG, "attempting data unlock range %"PRIu64
- " %"PRIu64" by %s", flock.l_start, flock.l_len,
- lkowner_utoa (&frame->root->lk_owner));
+ flock.l_start = inodelk->flock.l_start;
+ flock.l_len = inodelk->flock.l_len;
+ flock.l_type = F_UNLCK;
- call_count = afr_locked_nodes_count (int_lock->inode_locked_nodes,
+ full_flock.l_type = F_UNLCK;
+ call_count = afr_locked_nodes_count (inodelk->locked_nodes,
priv->child_count);
int_lock->lk_call_count = call_count;
@@ -622,11 +731,11 @@ afr_unlock_inodelk (call_frame_t *frame, xlator_t *this)
fd_ctx = afr_fd_ctx_get (local->fd, this);
for (i = 0; i < priv->child_count; i++) {
- if ((int_lock->inode_locked_nodes[i] & LOCKED_YES)
- != LOCKED_YES)
+ if ((inodelk->locked_nodes[i] & LOCKED_YES) != LOCKED_YES)
continue;
if (local->fd) {
+ flock_use = &flock;
if (!local->transaction.eager_lock[i]) {
goto wind;
}
@@ -638,43 +747,48 @@ afr_unlock_inodelk (call_frame_t *frame, xlator_t *this)
if (fd_ctx->lock_piggyback[i]) {
fd_ctx->lock_piggyback[i]--;
piggyback = 1;
+ } else {
+ fd_ctx->lock_acquired[i]--;
}
}
UNLOCK (&local->fd->lock);
if (piggyback) {
afr_unlock_inodelk_cbk (frame, (void *) (long) i,
- this, 1, 0);
+ this, 1, 0, NULL);
if (!--call_count)
break;
continue;
}
- fd_ctx->lock_acquired[i]--;
+ flock_use = &full_flock;
wind:
- afr_trace_inodelk_in (frame, AFR_INODELK_TRANSACTION,
- AFR_UNLOCK_OP, &flock, F_SETLK, i);
+ AFR_TRACE_INODELK_IN (frame, this,
+ AFR_INODELK_TRANSACTION,
+ AFR_UNLOCK_OP, flock_use, F_SETLK,
+ i);
STACK_WIND_COOKIE (frame, afr_unlock_inodelk_cbk,
(void *) (long)i,
priv->children[i],
priv->children[i]->fops->finodelk,
- this->name, local->fd,
- F_SETLK, &flock);
+ int_lock->domain, local->fd,
+ F_SETLK, flock_use, NULL);
if (!--call_count)
break;
} else {
- afr_trace_inodelk_in (frame, AFR_INODELK_TRANSACTION,
+ AFR_TRACE_INODELK_IN (frame, this,
+ AFR_INODELK_TRANSACTION,
AFR_UNLOCK_OP, &flock, F_SETLK, i);
STACK_WIND_COOKIE (frame, afr_unlock_inodelk_cbk,
(void *) (long)i,
priv->children[i],
priv->children[i]->fops->inodelk,
- this->name, &local->loc,
- F_SETLK, &flock);
+ int_lock->domain, &local->loc,
+ F_SETLK, &flock, NULL);
if (!--call_count)
break;
@@ -686,24 +800,34 @@ out:
static int32_t
afr_unlock_entrylk_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
- int32_t op_ret, int32_t op_errno)
+ int32_t op_ret, int32_t op_errno, dict_t *xdata)
{
afr_local_t *local = NULL;
- int32_t child_index = (long)cookie;
+ afr_private_t *priv = NULL;
+ afr_internal_lock_t *int_lock = NULL;
+ int32_t child_index = 0;
+ int lockee_no = 0;
+
+ priv = this->private;
+ lockee_no = (int)((long) cookie) / priv->child_count;
+ child_index = (int) ((long) cookie) % priv->child_count;
local = frame->local;
+ int_lock = &local->internal_lock;
- afr_trace_entrylk_out (frame, AFR_ENTRYLK_TRANSACTION,
- AFR_UNLOCK_OP, NULL, op_ret,
- op_errno, child_index);
+ AFR_TRACE_ENTRYLK_OUT (frame, this, AFR_ENTRYLK_TRANSACTION,
+ AFR_UNLOCK_OP,
+ int_lock->lockee[lockee_no].basename, op_ret,
+ op_errno, (int) ((long)cookie));
- if (op_ret < 0 && op_errno != ENOTCONN && op_errno != EBADFD) {
+ if (op_ret < 0) {
gf_log (this->name, GF_LOG_ERROR,
"%s: unlock failed on %d, reason: %s",
local->loc.path, child_index, strerror (op_errno));
}
- afr_unlock_common_cbk (frame, cookie, this, op_ret, op_errno);
+ int_lock->lockee[lockee_no].locked_nodes[child_index] &= LOCKED_NO;
+ afr_unlock_common_cbk (frame, cookie, this, op_ret, op_errno, NULL);
return 0;
}
@@ -711,24 +835,22 @@ afr_unlock_entrylk_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
static int
afr_unlock_entrylk (call_frame_t *frame, xlator_t *this)
{
- afr_internal_lock_t *int_lock = NULL;
- afr_local_t *local = NULL;
- afr_private_t *priv = NULL;
- const char *basename = NULL;
- loc_t *loc = NULL;
- int call_count = 0;
- int i = -1;
+ afr_internal_lock_t *int_lock = NULL;
+ afr_local_t *local = NULL;
+ afr_private_t *priv = NULL;
+ int call_count = 0;
+ int index = 0;
+ int lockee_no = 0;
+ int copies = 0;
+ int i = -1;
local = frame->local;
int_lock = &local->internal_lock;
priv = this->private;
+ copies = priv->child_count;
- basename = int_lock->lk_basename;
- if (int_lock->lk_loc)
- loc = int_lock->lk_loc;
+ call_count = afr_lockee_locked_nodes_count (int_lock);
- call_count = afr_locked_nodes_count (int_lock->entry_locked_nodes,
- priv->child_count);
int_lock->lk_call_count = call_count;
if (!call_count){
@@ -738,18 +860,23 @@ afr_unlock_entrylk (call_frame_t *frame, xlator_t *this)
goto out;
}
- for (i = 0; i < priv->child_count; i++) {
- if (int_lock->entry_locked_nodes[i] & LOCKED_YES) {
- afr_trace_entrylk_in (frame, AFR_ENTRYLK_NB_TRANSACTION,
- AFR_UNLOCK_OP, basename, i);
+ for (i = 0; i < int_lock->lockee_count * priv->child_count; i++) {
+ lockee_no = i / copies;
+ index = i % copies;
+ if (int_lock->lockee[lockee_no].locked_nodes[index] & LOCKED_YES) {
+ AFR_TRACE_ENTRYLK_IN (frame, this, AFR_ENTRYLK_NB_TRANSACTION,
+ AFR_UNLOCK_OP,
+ int_lock->lockee[lockee_no].basename,
+ i);
STACK_WIND_COOKIE (frame, afr_unlock_entrylk_cbk,
(void *) (long) i,
- priv->children[i],
- priv->children[i]->fops->entrylk,
- this->name,
- loc, basename,
- ENTRYLK_UNLOCK, ENTRYLK_WRLCK);
+ priv->children[index],
+ priv->children[index]->fops->entrylk,
+ int_lock->domain,
+ &int_lock->lockee[lockee_no].loc,
+ int_lock->lockee[lockee_no].basename,
+ ENTRYLK_UNLOCK, ENTRYLK_WRLCK, NULL);
if (!--call_count)
break;
@@ -763,15 +890,22 @@ out:
static int32_t
afr_lock_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
- int32_t op_ret, int32_t op_errno)
+ int32_t op_ret, int32_t op_errno, dict_t *xdata)
{
- afr_internal_lock_t *int_lock = NULL;
- afr_local_t *local = NULL;
- int child_index = (long) cookie;
+ afr_internal_lock_t *int_lock = NULL;
+ afr_local_t *local = NULL;
+ afr_private_t *priv = NULL;
+ int cky = (long) cookie;
+ int child_index = 0;
+ int lockee_no = 0;
+ priv = this->private;
local = frame->local;
int_lock = &local->internal_lock;
+ child_index = ((int)cky) % priv->child_count;
+ lockee_no = ((int)cky) / priv->child_count;
+
LOCK (&frame->lock);
{
if (op_ret == -1) {
@@ -787,6 +921,8 @@ afr_lock_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
local->op_errno = op_errno;
int_lock->lock_op_errno = op_errno;
}
+
+ int_lock->lk_attempted_count++;
}
UNLOCK (&frame->lock);
@@ -795,10 +931,17 @@ afr_lock_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
afr_unlock (frame, this);
} else {
if (op_ret == 0) {
- int_lock->locked_nodes[child_index] |= LOCKED_YES;
- int_lock->lock_count++;
+ if (local->transaction.type == AFR_ENTRY_TRANSACTION ||
+ local->transaction.type == AFR_ENTRY_RENAME_TRANSACTION) {
+ int_lock->lockee[lockee_no].locked_nodes[child_index] |= LOCKED_YES;
+ int_lock->lockee[lockee_no].locked_count++;
+ int_lock->entrylk_lock_count++;
+ } else {
+ int_lock->locked_nodes[child_index] |= LOCKED_YES;
+ int_lock->lock_count++;
+ }
}
- afr_lock_blocking (frame, this, child_index + 1);
+ afr_lock_blocking (frame, this, cky + 1);
}
return 0;
@@ -806,98 +949,26 @@ afr_lock_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
static int32_t
afr_blocking_inodelk_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
- int32_t op_ret, int32_t op_errno)
+ int32_t op_ret, int32_t op_errno, dict_t *xdata)
{
- afr_trace_inodelk_out (frame, AFR_INODELK_TRANSACTION,
+ AFR_TRACE_INODELK_OUT (frame, this, AFR_INODELK_TRANSACTION,
AFR_LOCK_OP, NULL, op_ret,
op_errno, (long) cookie);
- afr_lock_cbk (frame, cookie, this, op_ret, op_errno);
+ afr_lock_cbk (frame, cookie, this, op_ret, op_errno, xdata);
return 0;
}
static int32_t
-afr_lock_lower_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
- int32_t op_ret, int32_t op_errno)
-{
- afr_internal_lock_t *int_lock = NULL;
- afr_private_t *priv = NULL;
- afr_local_t *local = NULL;
- loc_t *lower = NULL;
- loc_t *higher = NULL;
- const char *higher_name = NULL;
- int child_index = (long) cookie;
-
- priv = this->private;
- local = frame->local;
- int_lock = &local->internal_lock;
-
- LOCK (&frame->lock);
- {
- if (op_ret == -1) {
- if (op_errno == ENOSYS) {
- /* return ENOTSUP */
-
- gf_log (this->name, GF_LOG_ERROR,
- "subvolume does not support locking. "
- "please load features/locks xlator on server");
-
- local->op_ret = op_ret;
- }
-
- local->op_errno = op_errno;
- }
- }
- UNLOCK (&frame->lock);
-
- if (op_ret != 0) {
- afr_unlock (frame, this);
- goto out;
- } else {
- int_lock->lower_locked_nodes[child_index] |= LOCKED_LOWER;
- int_lock->lock_count++;
- }
-
- /* The lower path has been locked. Now lock the higher path */
-
- lower = lower_path (&local->transaction.parent_loc,
- local->transaction.basename,
- &local->transaction.new_parent_loc,
- local->transaction.new_basename);
-
- higher = (lower == &local->transaction.parent_loc ?
- &local->transaction.new_parent_loc :
- &local->transaction.parent_loc);
-
- higher_name = (higher == &local->transaction.parent_loc ?
- local->transaction.basename :
- local->transaction.new_basename);
-
- afr_trace_entrylk_in (frame, AFR_ENTRYLK_TRANSACTION,
- AFR_LOCK_OP, higher_name, child_index);
-
-
- STACK_WIND_COOKIE (frame, afr_lock_cbk,
- (void *) (long) child_index,
- priv->children[child_index],
- priv->children[child_index]->fops->entrylk,
- this->name, higher, higher_name,
- ENTRYLK_LOCK, ENTRYLK_WRLCK);
-
-out:
- return 0;
-}
-
-static int32_t
afr_blocking_entrylk_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
- int32_t op_ret, int32_t op_errno)
+ int32_t op_ret, int32_t op_errno, dict_t *xdata)
{
- afr_trace_entrylk_out (frame, AFR_ENTRYLK_TRANSACTION,
+ AFR_TRACE_ENTRYLK_OUT (frame, this, AFR_ENTRYLK_TRANSACTION,
AFR_LOCK_OP, NULL, op_ret,
op_errno, (long)cookie);
- afr_lock_cbk (frame, cookie, this, op_ret, op_errno);
+ afr_lock_cbk (frame, cookie, this, op_ret, op_errno, xdata);
return 0;
}
@@ -905,6 +976,7 @@ static int
afr_copy_locked_nodes (call_frame_t *frame, xlator_t *this)
{
afr_internal_lock_t *int_lock = NULL;
+ afr_inodelk_t *inodelk = NULL;
afr_local_t *local = NULL;
afr_private_t *priv = NULL;
@@ -915,18 +987,16 @@ afr_copy_locked_nodes (call_frame_t *frame, xlator_t *this)
switch (local->transaction.type) {
case AFR_DATA_TRANSACTION:
case AFR_METADATA_TRANSACTION:
- memcpy (int_lock->inode_locked_nodes,
- int_lock->locked_nodes,
- priv->child_count);
- int_lock->inodelk_lock_count = int_lock->lock_count;
+ inodelk = afr_get_inodelk (int_lock, int_lock->domain);
+ memcpy (inodelk->locked_nodes, int_lock->locked_nodes,
+ sizeof (*inodelk->locked_nodes) * priv->child_count);
+ inodelk->lock_count = int_lock->lock_count;
break;
case AFR_ENTRY_RENAME_TRANSACTION:
case AFR_ENTRY_TRANSACTION:
- memcpy (int_lock->entry_locked_nodes,
- int_lock->locked_nodes,
- priv->child_count);
- int_lock->entrylk_lock_count = int_lock->lock_count;
+ /*entrylk_count is being used in both non-blocking and blocking
+ * modes */
break;
}
@@ -934,25 +1004,67 @@ afr_copy_locked_nodes (call_frame_t *frame, xlator_t *this)
}
+static inline gf_boolean_t
+afr_is_entrylk (afr_internal_lock_t *int_lock,
+ afr_transaction_type trans_type)
+{
+ gf_boolean_t is_entrylk = _gf_false;
+
+ if ((int_lock->transaction_lk_type == AFR_SELFHEAL_LK) &&
+ int_lock->selfheal_lk_type == AFR_ENTRY_SELF_HEAL_LK) {
+
+ is_entrylk = _gf_true;
+
+ } else if ((int_lock->transaction_lk_type == AFR_TRANSACTION_LK) &&
+ (trans_type == AFR_ENTRY_TRANSACTION ||
+ trans_type == AFR_ENTRY_RENAME_TRANSACTION)) {
+
+ is_entrylk = _gf_true;
+
+ } else {
+ is_entrylk = _gf_false;
+ }
+
+ return is_entrylk;
+}
+
+static gf_boolean_t
+_is_lock_wind_needed (afr_local_t *local, int child_index)
+{
+ if (!local->child_up[child_index])
+ return _gf_false;
+
+ return _gf_true;
+}
+
int
-afr_lock_blocking (call_frame_t *frame, xlator_t *this, int child_index)
+afr_lock_blocking (call_frame_t *frame, xlator_t *this, int cookie)
{
afr_internal_lock_t *int_lock = NULL;
+ afr_inodelk_t *inodelk = NULL;
afr_local_t *local = NULL;
afr_private_t *priv = NULL;
- loc_t *lower = NULL;
- const char *lower_name = NULL;
struct gf_flock flock = {0,};
uint64_t ctx = 0;
int ret = 0;
+ int child_index = 0;
+ int lockee_no = 0;
+ gf_boolean_t is_entrylk = _gf_false;
- local = frame->local;
- int_lock = &local->internal_lock;
- priv = this->private;
+ local = frame->local;
+ int_lock = &local->internal_lock;
+ priv = this->private;
+ child_index = cookie % priv->child_count;
+ lockee_no = cookie / priv->child_count;
+ is_entrylk = afr_is_entrylk (int_lock, local->transaction.type);
- flock.l_start = int_lock->lk_flock.l_start;
- flock.l_len = int_lock->lk_flock.l_len;
- flock.l_type = int_lock->lk_flock.l_type;
+
+ if (!is_entrylk) {
+ inodelk = afr_get_inodelk (int_lock, int_lock->domain);
+ flock.l_start = inodelk->flock.l_start;
+ flock.l_len = inodelk->flock.l_len;
+ flock.l_type = inodelk->flock.l_type;
+ }
if (local->fd) {
ret = fd_ctx_get (local->fd, this, &ctx);
@@ -971,42 +1083,26 @@ afr_lock_blocking (call_frame_t *frame, xlator_t *this, int child_index)
return 0;
}
-
- /* skip over children that or down
- or don't have the fd open */
-
- while ((child_index < priv->child_count)
- && (!local->child_up[child_index] ||
- !local->fd_open_on[child_index]))
-
- child_index++;
- } else {
- /* skip over children that are down */
- while ((child_index < priv->child_count)
- && !local->child_up[child_index])
- child_index++;
}
- if ((child_index == priv->child_count) &&
- int_lock->lock_count == 0) {
-
- gf_log (this->name, GF_LOG_INFO,
- "unable to lock on even one child");
-
- local->op_ret = -1;
- int_lock->lock_op_ret = -1;
+ if (int_lock->lk_expected_count == int_lock->lk_attempted_count) {
+ if ((is_entrylk && int_lock->entrylk_lock_count == 0) ||
+ (!is_entrylk && int_lock->lock_count == 0)) {
+ gf_log (this->name, GF_LOG_INFO,
+ "unable to lock on even one child");
- afr_copy_locked_nodes (frame, this);
+ local->op_ret = -1;
+ int_lock->lock_op_ret = -1;
- afr_unlock(frame, this);
+ afr_copy_locked_nodes (frame, this);
- return 0;
+ afr_unlock(frame, this);
+ return 0;
+ }
}
- if ((child_index == priv->child_count)
- || (int_lock->lock_count == int_lock->lk_expected_count)) {
-
+ if (int_lock->lk_expected_count == int_lock->lk_attempted_count) {
/* we're done locking */
gf_log (this->name, GF_LOG_DEBUG,
@@ -1019,12 +1115,18 @@ afr_lock_blocking (call_frame_t *frame, xlator_t *this, int child_index)
return 0;
}
+ if (!_is_lock_wind_needed (local, child_index)) {
+ afr_lock_blocking (frame, this, cookie + 1);
+ return 0;
+ }
+
switch (local->transaction.type) {
case AFR_DATA_TRANSACTION:
case AFR_METADATA_TRANSACTION:
if (local->fd) {
- afr_trace_inodelk_in (frame, AFR_INODELK_TRANSACTION,
+ AFR_TRACE_INODELK_IN (frame, this,
+ AFR_INODELK_TRANSACTION,
AFR_LOCK_OP, &flock, F_SETLKW,
child_index);
@@ -1032,11 +1134,12 @@ afr_lock_blocking (call_frame_t *frame, xlator_t *this, int child_index)
(void *) (long) child_index,
priv->children[child_index],
priv->children[child_index]->fops->finodelk,
- this->name, local->fd,
- F_SETLKW, &flock);
+ int_lock->domain, local->fd,
+ F_SETLKW, &flock, NULL);
} else {
- afr_trace_inodelk_in (frame, AFR_INODELK_TRANSACTION,
+ AFR_TRACE_INODELK_IN (frame, this,
+ AFR_INODELK_TRANSACTION,
AFR_LOCK_OP, &flock, F_SETLKW,
child_index);
@@ -1044,63 +1147,44 @@ afr_lock_blocking (call_frame_t *frame, xlator_t *this, int child_index)
(void *) (long) child_index,
priv->children[child_index],
priv->children[child_index]->fops->inodelk,
- this->name, &local->loc,
- F_SETLKW, &flock);
+ int_lock->domain, &local->loc,
+ F_SETLKW, &flock, NULL);
}
break;
case AFR_ENTRY_RENAME_TRANSACTION:
- {
- lower = lower_path (&local->transaction.parent_loc,
- local->transaction.basename,
- &local->transaction.new_parent_loc,
- local->transaction.new_basename);
-
- lower_name = (lower == &local->transaction.parent_loc ?
- local->transaction.basename :
- local->transaction.new_basename);
-
- afr_trace_entrylk_in (frame, AFR_ENTRYLK_TRANSACTION,
- AFR_LOCK_OP, lower_name, child_index);
-
-
- STACK_WIND_COOKIE (frame, afr_lock_lower_cbk,
- (void *) (long) child_index,
- priv->children[child_index],
- priv->children[child_index]->fops->entrylk,
- this->name, lower, lower_name,
- ENTRYLK_LOCK, ENTRYLK_WRLCK);
-
- break;
- }
-
case AFR_ENTRY_TRANSACTION:
+ /*Accounting for child_index increments on 'down'
+ *and 'fd-less' children */
+
if (local->fd) {
- afr_trace_entrylk_in (frame, AFR_ENTRYLK_TRANSACTION,
- AFR_LOCK_OP, local->transaction.basename,
- child_index);
+ AFR_TRACE_ENTRYLK_IN (frame, this, AFR_ENTRYLK_TRANSACTION,
+ AFR_LOCK_OP,
+ int_lock->lockee[lockee_no].basename,
+ cookie);
STACK_WIND_COOKIE (frame, afr_blocking_entrylk_cbk,
- (void *) (long) child_index,
+ (void *) (long) cookie,
priv->children[child_index],
priv->children[child_index]->fops->fentrylk,
- this->name, local->fd,
- local->transaction.basename,
- ENTRYLK_LOCK, ENTRYLK_WRLCK);
+ int_lock->domain, local->fd,
+ int_lock->lockee[lockee_no].basename,
+ ENTRYLK_LOCK, ENTRYLK_WRLCK, NULL);
} else {
- afr_trace_entrylk_in (frame, AFR_ENTRYLK_TRANSACTION,
+ AFR_TRACE_ENTRYLK_IN (frame, this,
+ AFR_ENTRYLK_TRANSACTION,
AFR_LOCK_OP, local->transaction.basename,
child_index);
STACK_WIND_COOKIE (frame, afr_blocking_entrylk_cbk,
- (void *) (long) child_index,
+ (void *) (long) cookie,
priv->children[child_index],
priv->children[child_index]->fops->entrylk,
- this->name,
- &local->transaction.parent_loc,
- local->transaction.basename,
- ENTRYLK_LOCK, ENTRYLK_WRLCK);
+ int_lock->domain,
+ &int_lock->lockee[lockee_no].loc,
+ int_lock->lockee[lockee_no].basename,
+ ENTRYLK_LOCK, ENTRYLK_WRLCK, NULL);
}
break;
@@ -1128,11 +1212,12 @@ afr_blocking_lock (call_frame_t *frame, xlator_t *this)
break;
case AFR_ENTRY_RENAME_TRANSACTION:
+ case AFR_ENTRY_TRANSACTION:
up_count = afr_up_children_count (local->child_up,
priv->child_count);
- int_lock->lk_expected_count = 2 * up_count;
- //fallthrough
- case AFR_ENTRY_TRANSACTION:
+ int_lock->lk_call_count = int_lock->lk_expected_count
+ = (int_lock->lockee_count *
+ up_count);
initialize_entrylk_variables (frame, this);
break;
}
@@ -1144,42 +1229,55 @@ afr_blocking_lock (call_frame_t *frame, xlator_t *this)
static int32_t
afr_nonblocking_entrylk_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
- int32_t op_ret, int32_t op_errno)
+ int32_t op_ret, int32_t op_errno, dict_t *xdata)
{
afr_internal_lock_t *int_lock = NULL;
afr_local_t *local = NULL;
int call_count = 0;
int child_index = (long) cookie;
+ int copies = 0;
+ int index = 0;
+ int lockee_no = 0;
+ afr_private_t *priv = NULL;
+
+ priv = this->private;
+
+ copies = priv->child_count;
+ index = child_index % copies;
+ lockee_no = child_index / copies;
local = frame->local;
int_lock = &local->internal_lock;
- afr_trace_entrylk_out (frame, AFR_ENTRYLK_TRANSACTION,
- AFR_LOCK_OP, NULL, op_ret,
+ AFR_TRACE_ENTRYLK_OUT (frame, this, AFR_ENTRYLK_TRANSACTION,
+ AFR_LOCK_OP,
+ int_lock->lockee[lockee_no].basename, op_ret,
op_errno, (long) cookie);
- LOCK (&frame->lock);
- {
- call_count = --int_lock->lk_call_count;
- }
- UNLOCK (&frame->lock);
-
- if (op_ret < 0 ) {
- if (op_errno == ENOSYS) {
+ LOCK (&frame->lock);
+ {
+ if (op_ret < 0 ) {
+ if (op_errno == ENOSYS) {
/* return ENOTSUP */
- gf_log (this->name, GF_LOG_ERROR,
- "subvolume does not support locking. "
- "please load features/locks xlator on server");
- local->op_ret = op_ret;
- int_lock->lock_op_ret = op_ret;
+ gf_log (this->name, GF_LOG_ERROR,
+ "subvolume does not support locking. "
+ "please load features/locks xlator on server");
+ local->op_ret = op_ret;
+ int_lock->lock_op_ret = op_ret;
+
+ int_lock->lock_op_errno = op_errno;
+ local->op_errno = op_errno;
+ }
+ } else if (op_ret == 0) {
+ int_lock->lockee[lockee_no].locked_nodes[index] |= \
+ LOCKED_YES;
+ int_lock->lockee[lockee_no].locked_count++;
+ int_lock->entrylk_lock_count++;
+ }
- int_lock->lock_op_errno = op_errno;
- local->op_errno = op_errno;
- }
- } else if (op_ret == 0) {
- int_lock->entry_locked_nodes[child_index] |= LOCKED_YES;
- int_lock->entrylk_lock_count++;
+ call_count = --int_lock->lk_call_count;
}
+ UNLOCK (&frame->lock);
if (call_count == 0) {
gf_log (this->name, GF_LOG_TRACE,
@@ -1206,42 +1304,26 @@ afr_nonblocking_entrylk_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
return 0;
}
-void
-afr_mark_fd_open_on (afr_local_t *local, afr_fd_ctx_t *fd_ctx,
- size_t child_count)
-{
- int i = 0;
-
- GF_ASSERT (local->fd_open_on);
-
- memset (local->fd_open_on, 0, sizeof (*local->fd_open_on)*child_count);
- for (i = 0; i < child_count; i++)
- if (fd_ctx->opened_on[i] == AFR_FD_OPENED)
- local->fd_open_on[i] = 1;
-}
-
int
afr_nonblocking_entrylk (call_frame_t *frame, xlator_t *this)
{
- afr_internal_lock_t *int_lock = NULL;
- afr_local_t *local = NULL;
- afr_private_t *priv = NULL;
- afr_fd_ctx_t *fd_ctx = NULL;
- const char *basename = NULL;
- loc_t *loc = NULL;
- int32_t call_count = 0;
+ afr_internal_lock_t *int_lock = NULL;
+ afr_local_t *local = NULL;
+ afr_private_t *priv = NULL;
+ afr_fd_ctx_t *fd_ctx = NULL;
+ int copies = 0;
+ int index = 0;
+ int lockee_no = 0;
+ int32_t call_count = 0;
int i = 0;
local = frame->local;
int_lock = &local->internal_lock;
priv = this->private;
+ copies = priv->child_count;
initialize_entrylk_variables (frame, this);
- basename = int_lock->lk_basename;
- if (int_lock->lk_loc)
- loc = int_lock->lk_loc;
-
if (local->fd) {
fd_ctx = afr_fd_ctx_get (local->fd, this);
if (!fd_ctx) {
@@ -1254,11 +1336,11 @@ afr_nonblocking_entrylk (call_frame_t *frame, xlator_t *this)
local->op_errno = EINVAL;
int_lock->lock_op_errno = EINVAL;
+ afr_unlock (frame, this);
return -1;
}
- afr_mark_fd_open_on (local, fd_ctx, priv->child_count);
- call_count = internal_lock_count (frame, this);
+ call_count = int_lock->lockee_count * internal_lock_count (frame, this);
int_lock->lk_call_count = call_count;
int_lock->lk_expected_count = call_count;
@@ -1271,42 +1353,52 @@ afr_nonblocking_entrylk (call_frame_t *frame, xlator_t *this)
/* Send non-blocking entrylk calls only on up children
and where the fd has been opened */
- for (i = 0; i < priv->child_count; i++) {
- if (local->child_up[i] && local->fd_open_on[i]) {
- afr_trace_entrylk_in (frame, AFR_ENTRYLK_NB_TRANSACTION,
- AFR_LOCK_OP, basename, i);
+ for (i = 0; i < int_lock->lockee_count*priv->child_count; i++) {
+ index = i%copies;
+ lockee_no = i/copies;
+ if (local->child_up[index]) {
+ AFR_TRACE_ENTRYLK_IN (frame, this, AFR_ENTRYLK_NB_TRANSACTION,
+ AFR_LOCK_OP,
+ int_lock->lockee[lockee_no].basename,
+ i);
STACK_WIND_COOKIE (frame, afr_nonblocking_entrylk_cbk,
(void *) (long) i,
- priv->children[i],
- priv->children[i]->fops->fentrylk,
+ priv->children[index],
+ priv->children[index]->fops->fentrylk,
this->name, local->fd,
- basename,
- ENTRYLK_LOCK_NB, ENTRYLK_WRLCK);
+ int_lock->lockee[lockee_no].basename,
+ ENTRYLK_LOCK_NB, ENTRYLK_WRLCK,
+ NULL);
+ if (!--call_count)
+ break;
}
}
} else {
- GF_ASSERT (loc);
-
- call_count = internal_lock_count (frame, this);
+ call_count = int_lock->lockee_count * internal_lock_count (frame, this);
int_lock->lk_call_count = call_count;
int_lock->lk_expected_count = call_count;
- for (i = 0; i < priv->child_count; i++) {
- if (local->child_up[i]) {
- afr_trace_entrylk_in (frame, AFR_ENTRYLK_NB_TRANSACTION,
- AFR_LOCK_OP, basename, i);
+ for (i = 0; i < int_lock->lockee_count*priv->child_count; i++) {
+ index = i%copies;
+ lockee_no = i/copies;
+ if (local->child_up[index]) {
+ AFR_TRACE_ENTRYLK_IN (frame, this, AFR_ENTRYLK_NB_TRANSACTION,
+ AFR_LOCK_OP,
+ int_lock->lockee[lockee_no].basename,
+ i);
STACK_WIND_COOKIE (frame, afr_nonblocking_entrylk_cbk,
(void *) (long) i,
- priv->children[i],
- priv->children[i]->fops->entrylk,
- this->name, loc, basename,
- ENTRYLK_LOCK_NB, ENTRYLK_WRLCK);
+ priv->children[index],
+ priv->children[index]->fops->entrylk,
+ this->name, &int_lock->lockee[lockee_no].loc,
+ int_lock->lockee[lockee_no].basename,
+ ENTRYLK_LOCK_NB, ENTRYLK_WRLCK,
+ NULL);
if (!--call_count)
break;
-
}
}
}
@@ -1316,70 +1408,69 @@ out:
int32_t
afr_nonblocking_inodelk_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
- int32_t op_ret, int32_t op_errno)
+ int32_t op_ret, int32_t op_errno, dict_t *xdata)
{
afr_internal_lock_t *int_lock = NULL;
+ afr_inodelk_t *inodelk = NULL;
afr_local_t *local = NULL;
int call_count = 0;
int child_index = (long) cookie;
afr_fd_ctx_t *fd_ctx = NULL;
- afr_private_t *priv = NULL;
- priv = this->private;
local = frame->local;
int_lock = &local->internal_lock;
+ inodelk = afr_get_inodelk (int_lock, int_lock->domain);
- afr_trace_inodelk_out (frame, AFR_INODELK_NB_TRANSACTION,
+ AFR_TRACE_INODELK_OUT (frame, this, AFR_INODELK_NB_TRANSACTION,
AFR_LOCK_OP, NULL, op_ret,
op_errno, (long) cookie);
+ if (local->fd)
+ fd_ctx = afr_fd_ctx_get (local->fd, this);
+
LOCK (&frame->lock);
{
+ if (op_ret < 0) {
+ if (op_errno == ENOSYS) {
+ /* return ENOTSUP */
+ gf_log (this->name, GF_LOG_ERROR,
+ "subvolume does not support locking. "
+ "please load features/locks xlator on "
+ "server");
+ local->op_ret = op_ret;
+ int_lock->lock_op_ret = op_ret;
+ int_lock->lock_op_errno = op_errno;
+ local->op_errno = op_errno;
+ }
+ if (local->transaction.eager_lock)
+ local->transaction.eager_lock[child_index] = 0;
+ } else {
+ inodelk->locked_nodes[child_index] |= LOCKED_YES;
+ inodelk->lock_count++;
+
+ if (local->transaction.eager_lock &&
+ local->transaction.eager_lock[child_index] &&
+ local->fd) {
+ /* piggybacked */
+ if (op_ret == 1) {
+ /* piggybacked */
+ } else if (op_ret == 0) {
+ /* lock acquired from server */
+ fd_ctx->lock_acquired[child_index]++;
+ }
+ }
+ }
+
call_count = --int_lock->lk_call_count;
}
UNLOCK (&frame->lock);
- if (op_ret < 0) {
- if (op_errno == ENOSYS) {
- /* return ENOTSUP */
- gf_log (this->name, GF_LOG_ERROR,
- "subvolume does not support locking. "
- "please load features/locks xlator on server");
- local->op_ret = op_ret;
- int_lock->lock_op_ret = op_ret;
- int_lock->lock_op_errno = op_errno;
- local->op_errno = op_errno;
- }
- } else {
- int_lock->inode_locked_nodes[child_index]
- |= LOCKED_YES;
- int_lock->inodelk_lock_count++;
-
- if (priv->eager_lock && local->fd) {
- fd_ctx = afr_fd_ctx_get (local->fd, this);
- local->transaction.eager_lock[child_index] = 1;
- /* piggybacked */
-
- if (op_ret == 1) {
- /* piggybacked */
- } else if (op_ret == 0) {
- /* lock acquired from server */
- LOCK (&local->fd->lock);
- {
- fd_ctx->lock_acquired[child_index]++;
- }
- UNLOCK (&local->fd->lock);
- }
- }
- }
-
if (call_count == 0) {
gf_log (this->name, GF_LOG_TRACE,
"Last inode locking reply received");
/* all locks successful. Proceed to call FOP */
- if (int_lock->inodelk_lock_count ==
- int_lock->lk_expected_count) {
+ if (inodelk->lock_count == int_lock->lk_expected_count) {
gf_log (this->name, GF_LOG_TRACE,
"All servers locked. Calling the cbk");
int_lock->lock_op_ret = 0;
@@ -1403,30 +1494,29 @@ int
afr_nonblocking_inodelk (call_frame_t *frame, xlator_t *this)
{
afr_internal_lock_t *int_lock = NULL;
+ afr_inodelk_t *inodelk = NULL;
afr_local_t *local = NULL;
afr_private_t *priv = NULL;
afr_fd_ctx_t *fd_ctx = NULL;
- int32_t call_count = 0;
- int i = 0;
- int ret = 0;
- struct gf_flock flock = {0,};
- struct gf_flock full_flock = {0,};
- struct gf_flock *flock_use = &flock;
- int piggyback = 0;
+ int32_t call_count = 0;
+ int i = 0;
+ int ret = 0;
+ struct gf_flock flock = {0,};
+ struct gf_flock full_flock = {0,};
+ struct gf_flock *flock_use = NULL;
+ int piggyback = 0;
local = frame->local;
int_lock = &local->internal_lock;
priv = this->private;
- flock.l_start = int_lock->lk_flock.l_start;
- flock.l_len = int_lock->lk_flock.l_len;
- flock.l_type = int_lock->lk_flock.l_type;
+ inodelk = afr_get_inodelk (int_lock, int_lock->domain);
- gf_log (this->name, GF_LOG_DEBUG, "attempting data lock range %"PRIu64
- " %"PRIu64" by %s", flock.l_start, flock.l_len,
- lkowner_utoa (&frame->root->lk_owner));
+ flock.l_start = inodelk->flock.l_start;
+ flock.l_len = inodelk->flock.l_len;
+ flock.l_type = inodelk->flock.l_type;
- full_flock.l_type = int_lock->lk_flock.l_type;
+ full_flock.l_type = inodelk->flock.l_type;
initialize_inodelk_variables (frame, this);
@@ -1442,11 +1532,11 @@ afr_nonblocking_inodelk (call_frame_t *frame, xlator_t *this)
local->op_errno = EINVAL;
int_lock->lock_op_errno = EINVAL;
+ afr_unlock (frame, this);
ret = -1;
goto out;
}
- afr_mark_fd_open_on (local, fd_ctx, priv->child_count);
call_count = internal_lock_count (frame, this);
int_lock->lk_call_count = call_count;
int_lock->lk_expected_count = call_count;
@@ -1461,14 +1551,18 @@ afr_nonblocking_inodelk (call_frame_t *frame, xlator_t *this)
/* Send non-blocking inodelk calls only on up children
and where the fd has been opened */
for (i = 0; i < priv->child_count; i++) {
- if (!local->child_up[i] || !local->fd_open_on[i])
+ if (!local->child_up[i])
continue;
- if (!priv->eager_lock)
+ flock_use = &flock;
+ if (!local->transaction.eager_lock_on) {
goto wind;
+ }
- flock_use = &full_flock;
piggyback = 0;
+ local->transaction.eager_lock[i] = 1;
+
+ afr_set_delayed_post_op (frame, this);
LOCK (&local->fd->lock);
{
@@ -1482,21 +1576,23 @@ afr_nonblocking_inodelk (call_frame_t *frame, xlator_t *this)
if (piggyback) {
/* (op_ret == 1) => indicate piggybacked lock */
afr_nonblocking_inodelk_cbk (frame, (void *) (long) i,
- this, 1, 0);
+ this, 1, 0, NULL);
if (!--call_count)
break;
continue;
}
+ flock_use = &full_flock;
wind:
- afr_trace_inodelk_in (frame, AFR_INODELK_NB_TRANSACTION,
+ AFR_TRACE_INODELK_IN (frame, this,
+ AFR_INODELK_NB_TRANSACTION,
AFR_LOCK_OP, flock_use, F_SETLK, i);
STACK_WIND_COOKIE (frame, afr_nonblocking_inodelk_cbk,
(void *) (long) i,
priv->children[i],
priv->children[i]->fops->finodelk,
- this->name, local->fd,
- F_SETLK, flock_use);
+ int_lock->domain, local->fd,
+ F_SETLK, flock_use, NULL);
if (!--call_count)
break;
@@ -1509,15 +1605,16 @@ afr_nonblocking_inodelk (call_frame_t *frame, xlator_t *this)
for (i = 0; i < priv->child_count; i++) {
if (!local->child_up[i])
continue;
- afr_trace_inodelk_in (frame, AFR_INODELK_NB_TRANSACTION,
+ AFR_TRACE_INODELK_IN (frame, this,
+ AFR_INODELK_NB_TRANSACTION,
AFR_LOCK_OP, &flock, F_SETLK, i);
STACK_WIND_COOKIE (frame, afr_nonblocking_inodelk_cbk,
(void *) (long) i,
priv->children[i],
priv->children[i]->fops->inodelk,
- this->name, &local->loc,
- F_SETLK, &flock);
+ int_lock->domain, &local->loc,
+ F_SETLK, &flock, NULL);
if (!--call_count)
break;
@@ -1527,200 +1624,6 @@ out:
return ret;
}
-static int
-__is_lower_locked (call_frame_t *frame, xlator_t *this)
-{
- afr_internal_lock_t *int_lock = NULL;
- afr_private_t *priv = NULL;
- afr_local_t *local = NULL;
- int count = 0;
- int i = 0;
-
- local = frame->local;
- int_lock = &local->internal_lock;
- priv = this->private;
-
- for (i = 0; i < priv->child_count; i++) {
- if (int_lock->lower_locked_nodes[i] & LOCKED_LOWER)
- count++;
- }
-
- return count;
-
-}
-
-static int
-__is_higher_locked (call_frame_t *frame, xlator_t *this)
-{
- afr_internal_lock_t *int_lock = NULL;
- afr_private_t *priv = NULL;
- afr_local_t *local = NULL;
- int count = 0;
- int i = 0;
-
- local = frame->local;
- int_lock = &local->internal_lock;
- priv = this->private;
-
- for (i = 0; i < priv->child_count; i++) {
- if (int_lock->locked_nodes[i] & LOCKED_YES)
- count++;
- }
-
- return count;
-
-}
-
-static int
-afr_unlock_lower_entrylk (call_frame_t *frame, xlator_t *this)
-{
- afr_internal_lock_t *int_lock = NULL;
- afr_local_t *local = NULL;
- afr_private_t *priv = NULL;
- const char *basename = NULL;
- loc_t *loc = NULL;
- int call_count = 0;
- int i = -1;
-
- local = frame->local;
- int_lock = &local->internal_lock;
- priv = this->private;
-
- basename = int_lock->lk_basename;
- if (int_lock->lk_loc)
- loc = int_lock->lk_loc;
-
- call_count = __is_lower_locked (frame, this);
- int_lock->lk_call_count = call_count;
-
- if (!call_count){
- gf_log (this->name, GF_LOG_TRACE,
- "No internal locks unlocked");
- int_lock->lock_cbk (frame, this);
- goto out;
- }
-
- for (i = 0; i < priv->child_count; i++) {
- if (int_lock->lower_locked_nodes[i] & LOCKED_LOWER) {
- afr_trace_entrylk_in (frame, AFR_ENTRYLK_NB_TRANSACTION,
- AFR_UNLOCK_OP, basename, i);
-
- STACK_WIND_COOKIE (frame, afr_unlock_entrylk_cbk,
- (void *) (long) i,
- priv->children[i],
- priv->children[i]->fops->entrylk,
- this->name,
- loc, basename,
- ENTRYLK_UNLOCK, ENTRYLK_WRLCK);
-
- if (!--call_count)
- break;
-
- }
- }
-
-out:
- return 0;
-
-}
-
-
-static int
-afr_post_unlock_higher_cbk (call_frame_t *frame, xlator_t *this)
-{
- afr_local_t *local = NULL;
-
- local = frame->local;
-
- local->transaction.done (frame, this);
- return 0;
-}
-
-static int
-afr_post_unlock_lower_cbk (call_frame_t *frame, xlator_t *this)
-{
- afr_internal_lock_t *int_lock = NULL;
- afr_local_t *local = NULL;
- loc_t *lower = NULL;
- loc_t *higher = NULL;
- const char *higher_name = NULL;
-
- local = frame->local;
- int_lock = &local->internal_lock;
-
- lower = lower_path (&local->transaction.parent_loc,
- local->transaction.basename,
- &local->transaction.new_parent_loc,
- local->transaction.new_basename);
-
- higher = (lower == &local->transaction.parent_loc ?
- &local->transaction.new_parent_loc :
- &local->transaction.parent_loc);
-
- higher_name = (higher == &local->transaction.parent_loc ?
- local->transaction.basename :
- local->transaction.new_basename);
-
- if (__is_higher_locked (frame, this)) {
- gf_log (this->name, GF_LOG_DEBUG,
- "unlocking higher");
- int_lock->lk_basename = higher_name;
- int_lock->lk_loc = higher;
- int_lock->lock_cbk = afr_post_unlock_higher_cbk;
-
- afr_unlock_entrylk (frame, this);
- } else
- local->transaction.done (frame, this);
-
- return 0;
-}
-
-static int
-afr_rename_unlock (call_frame_t *frame, xlator_t *this)
-{
- afr_internal_lock_t *int_lock = NULL;
- afr_local_t *local = NULL;
- loc_t *lower = NULL;
- const char *lower_name = NULL;
-
- local = frame->local;
- int_lock = &local->internal_lock;
-
- lower = lower_path (&local->transaction.parent_loc,
- local->transaction.basename,
- &local->transaction.new_parent_loc,
- local->transaction.new_basename);
-
- lower_name = (lower == &local->transaction.parent_loc ?
- local->transaction.basename :
- local->transaction.new_basename);
-
- if (__is_lower_locked (frame, this)) {
- gf_log (this->name, GF_LOG_DEBUG,
- "unlocking lower");
- int_lock->lk_basename = lower_name;
- int_lock->lk_loc = lower;
- int_lock->lock_cbk = afr_post_unlock_lower_cbk;
-
- afr_unlock_lower_entrylk (frame, this);
- } else
- afr_post_unlock_lower_cbk (frame, this);
-
- return 0;
-}
-
-static int
-afr_rename_transaction (call_frame_t *frame, xlator_t *this)
-{
- afr_local_t *local = NULL;
-
- local = frame->local;
-
- return (local->transaction.type ==
- AFR_ENTRY_RENAME_TRANSACTION);
-
-}
-
int32_t
afr_unlock (call_frame_t *frame, xlator_t *this)
{
@@ -1732,10 +1635,8 @@ afr_unlock (call_frame_t *frame, xlator_t *this)
if (is_afr_lock_transaction (local))
afr_unlock_inodelk (frame, this);
else
- if (!afr_rename_transaction (frame, this))
- afr_unlock_entrylk (frame, this);
- else
- afr_rename_unlock (frame, this);
+ afr_unlock_entrylk (frame, this);
+
} else {
if (is_afr_lock_selfheal (local))
afr_unlock_inodelk (frame, this);
@@ -1904,10 +1805,12 @@ out:
int32_t
afr_get_locks_fd_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
- int32_t op_ret, int32_t op_errno, struct gf_flock *lock);
+ int32_t op_ret, int32_t op_errno, struct gf_flock *lock,
+ dict_t *xdata);
int32_t
afr_recover_lock_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
- int32_t op_ret, int32_t op_errno, struct gf_flock *lock)
+ int32_t op_ret, int32_t op_errno, struct gf_flock *lock,
+ dict_t *xdata)
{
afr_local_t *local = NULL;
afr_private_t *priv = NULL;
@@ -1931,7 +1834,7 @@ afr_recover_lock_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
(void *) (long) source_child,
priv->children[source_child],
priv->children[source_child]->fops->lk,
- local->fd, F_GETLK_FD, &flock);
+ local->fd, F_GETLK_FD, &flock, NULL);
return 0;
@@ -1959,7 +1862,7 @@ afr_recover_lock (call_frame_t *frame, xlator_t *this,
(void *) (long) lock_recovery_child,
priv->children[lock_recovery_child],
priv->children[lock_recovery_child]->fops->lk,
- local->fd, F_SETLK, flock);
+ local->fd, F_SETLK, flock, NULL);
return 0;
}
@@ -1977,7 +1880,8 @@ is_afr_lock_eol (struct gf_flock *lock)
int32_t
afr_get_locks_fd_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
- int32_t op_ret, int32_t op_errno, struct gf_flock *lock)
+ int32_t op_ret, int32_t op_errno, struct gf_flock *lock,
+ dict_t *xdata)
{
if (op_ret) {
gf_log (this->name, GF_LOG_INFO,
@@ -2037,7 +1941,7 @@ afr_lock_recovery (call_frame_t *frame, xlator_t *this)
(void *) (long) source_child,
priv->children[source_child],
priv->children[source_child]->fops->lk,
- local->fd, F_GETLK_FD, &flock);
+ local->fd, F_GETLK_FD, &flock, NULL);
out:
return ret;
@@ -2065,7 +1969,8 @@ out:
int32_t
afr_lock_recovery_preopen_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
- int32_t op_ret, int32_t op_errno, fd_t *fd)
+ int32_t op_ret, int32_t op_errno, fd_t *fd,
+ dict_t *xdata)
{
int32_t child_index = (long )cookie;
int ret = 0;
@@ -2137,8 +2042,7 @@ afr_lock_recovery_preopen (call_frame_t *frame, xlator_t *this)
(void *)(long) child_index,
priv->children[child_index],
priv->children[child_index]->fops->open,
- &loc, fdctx->flags, local->fd,
- fdctx->wbflags);
+ &loc, fdctx->flags, local->fd, NULL);
return 0;
}
@@ -2187,7 +2091,7 @@ afr_attempt_lock_recovery (xlator_t *this, int32_t child_index)
goto out;
}
- ALLOC_OR_GOTO (frame->local, afr_local_t, out);
+ AFR_LOCAL_ALLOC_OR_GOTO (frame->local, out);
local = frame->local;
ret = afr_local_init (local, priv, &op_errno);
if (ret < 0) {
@@ -2233,27 +2137,38 @@ out:
return ret;
}
-void
-afr_lk_transfer_datalock (call_frame_t *dst, call_frame_t *src,
+int
+afr_lk_transfer_datalock (call_frame_t *dst, call_frame_t *src, char *dom,
unsigned int child_count)
{
- afr_local_t *dst_local = NULL;
- afr_local_t *src_local = NULL;
- afr_internal_lock_t *dst_lock = NULL;
- afr_internal_lock_t *src_lock = NULL;
+ afr_local_t *dst_local = NULL;
+ afr_local_t *src_local = NULL;
+ afr_internal_lock_t *dst_lock = NULL;
+ afr_internal_lock_t *src_lock = NULL;
+ afr_inodelk_t *dst_inodelk = NULL;
+ afr_inodelk_t *src_inodelk = NULL;
+ int ret = -1;
- dst_local = dst->local;
- dst_lock = &dst_local->internal_lock;
src_local = src->local;
src_lock = &src_local->internal_lock;
- if (src_lock->inode_locked_nodes) {
- memcpy (dst_lock->inode_locked_nodes,
- src_lock->inode_locked_nodes,
- sizeof (*dst_lock->inode_locked_nodes) * child_count);
- memset (src_lock->inode_locked_nodes, 0,
- sizeof (*src_lock->inode_locked_nodes) * child_count);
- }
-
- dst_lock->inodelk_lock_count = src_lock->inodelk_lock_count;
- src_lock->inodelk_lock_count = 0;
+ src_inodelk = afr_get_inodelk (src_lock, dom);
+ dst_local = dst->local;
+ dst_lock = &dst_local->internal_lock;
+ dst_inodelk = afr_get_inodelk (dst_lock, dom);
+ if (!dst_inodelk || !src_inodelk)
+ goto out;
+ if (src_inodelk->locked_nodes) {
+ memcpy (dst_inodelk->locked_nodes, src_inodelk->locked_nodes,
+ sizeof (*dst_inodelk->locked_nodes) * child_count);
+ memset (src_inodelk->locked_nodes, 0,
+ sizeof (*src_inodelk->locked_nodes) * child_count);
+ }
+
+ dst_lock->transaction_lk_type = src_lock->transaction_lk_type;
+ dst_lock->selfheal_lk_type = src_lock->selfheal_lk_type;
+ dst_inodelk->lock_count = src_inodelk->lock_count;
+ src_inodelk->lock_count = 0;
+ ret = 0;
+out:
+ return ret;
}
diff --git a/xlators/cluster/afr/src/afr-mem-types.h b/xlators/cluster/afr/src/afr-mem-types.h
index ebe189c35..73594f265 100644
--- a/xlators/cluster/afr/src/afr-mem-types.h
+++ b/xlators/cluster/afr/src/afr-mem-types.h
@@ -1,20 +1,11 @@
/*
- Copyright (c) 2008-2011 Gluster, Inc. <http://www.gluster.com>
- This file is part of GlusterFS.
+ Copyright (c) 2008-2012 Red Hat, Inc. <http://www.redhat.com>
+ This file is part of GlusterFS.
- GlusterFS is free software; you can redistribute it and/or modify
- it under the terms of the GNU General Public License as published
- by the Free Software Foundation; either version 3 of the License,
- or (at your option) any later version.
-
- GlusterFS is distributed in the hope that it will be useful, but
- WITHOUT ANY WARRANTY; without even the implied warranty of
- MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU
- General Public License for more details.
-
- You should have received a copy of the GNU General Public License
- along with this program. If not, see
- <http://www.gnu.org/licenses/>.
+ This file is licensed to you under your choice of the GNU Lesser
+ General Public License, version 3 or any later version (LGPLv3 or
+ later), or the GNU General Public License, version 2 (GPLv2), in all
+ cases as published by the Free Software Foundation.
*/
@@ -26,7 +17,6 @@
enum gf_afr_mem_types_ {
gf_afr_mt_iovec = gf_common_mt_end + 1,
gf_afr_mt_afr_fd_ctx_t,
- gf_afr_mt_afr_local_t,
gf_afr_mt_afr_private_t,
gf_afr_mt_int32_t,
gf_afr_mt_char,
@@ -44,8 +34,17 @@ enum gf_afr_mem_types_ {
gf_afr_mt_locked_fd,
gf_afr_mt_inode_ctx_t,
gf_afr_fd_paused_call_t,
- gf_afr_mt_afr_crawl_data_t,
- gf_afr_mt_afr_brick_pos_t,
+ gf_afr_mt_crawl_data_t,
+ gf_afr_mt_brick_pos_t,
+ gf_afr_mt_shd_bool_t,
+ gf_afr_mt_shd_timer_t,
+ gf_afr_mt_shd_event_t,
+ gf_afr_mt_time_t,
+ gf_afr_mt_pos_data_t,
+ gf_afr_mt_reply_t,
+ gf_afr_mt_stats_t,
+ gf_afr_mt_shd_crawl_event_t,
+ gf_afr_mt_uint64_t,
gf_afr_mt_end
};
#endif
diff --git a/xlators/cluster/afr/src/afr-open.c b/xlators/cluster/afr/src/afr-open.c
index 813b3c451..643a5d692 100644
--- a/xlators/cluster/afr/src/afr-open.c
+++ b/xlators/cluster/afr/src/afr-open.c
@@ -1,20 +1,11 @@
/*
- Copyright (c) 2007-2011 Gluster, Inc. <http://www.gluster.com>
+ Copyright (c) 2008-2012 Red Hat, Inc. <http://www.redhat.com>
This file is part of GlusterFS.
- GlusterFS is free software; you can redistribute it and/or modify
- it under the terms of the GNU General Public License as published
- by the Free Software Foundation; either version 3 of the License,
- or (at your option) any later version.
-
- GlusterFS is distributed in the hope that it will be useful, but
- WITHOUT ANY WARRANTY; without even the implied warranty of
- MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU
- General Public License for more details.
-
- You should have received a copy of the GNU General Public License
- along with this program. If not, see
- <http://www.gnu.org/licenses/>.
+ This file is licensed to you under your choice of the GNU Lesser
+ General Public License, version 3 or any later version (LGPLv3 or
+ later), or the GNU General Public License, version 2 (GPLv2), in all
+ cases as published by the Free Software Foundation.
*/
#include <libgen.h>
@@ -123,7 +114,7 @@ out:
int
afr_open_ftruncate_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
int32_t op_ret, int32_t op_errno, struct iatt *prebuf,
- struct iatt *postbuf)
+ struct iatt *postbuf, dict_t *xdata)
{
afr_local_t * local = frame->local;
afr_private_t *priv = NULL;
@@ -132,7 +123,7 @@ afr_open_ftruncate_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
if (afr_open_only_data_self_heal (priv->data_self_heal))
afr_perform_data_self_heal (frame, this);
AFR_STACK_UNWIND (open, frame, local->op_ret, local->op_errno,
- local->fd);
+ local->fd, xdata);
return 0;
}
@@ -140,7 +131,7 @@ afr_open_ftruncate_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
int
afr_open_cbk (call_frame_t *frame, void *cookie,
xlator_t *this, int32_t op_ret, int32_t op_errno,
- fd_t *fd)
+ fd_t *fd, dict_t *xdata)
{
afr_local_t * local = NULL;
int ret = 0;
@@ -162,8 +153,7 @@ afr_open_cbk (call_frame_t *frame, void *cookie,
local->success_count++;
ret = afr_child_fd_ctx_set (this, fd, child_index,
- local->cont.open.flags,
- local->cont.open.wbflags);
+ local->cont.open.flags);
if (ret) {
local->op_ret = -1;
local->op_errno = -ret;
@@ -181,12 +171,12 @@ unlock:
&& (local->op_ret >= 0)) {
STACK_WIND (frame, afr_open_ftruncate_cbk,
this, this->fops->ftruncate,
- fd, 0);
+ fd, 0, NULL);
} else {
if (afr_open_only_data_self_heal (priv->data_self_heal))
afr_perform_data_self_heal (frame, this);
AFR_STACK_UNWIND (open, frame, local->op_ret,
- local->op_errno, local->fd);
+ local->op_errno, local->fd, xdata);
}
}
@@ -195,7 +185,7 @@ unlock:
int
afr_open (call_frame_t *frame, xlator_t *this, loc_t *loc, int32_t flags,
- fd_t *fd, int32_t wbflags)
+ fd_t *fd, dict_t *xdata)
{
afr_private_t * priv = NULL;
afr_local_t * local = NULL;
@@ -225,7 +215,7 @@ afr_open (call_frame_t *frame, xlator_t *this, loc_t *loc, int32_t flags,
goto out;
}
- ALLOC_OR_GOTO (frame->local, afr_local_t, out);
+ AFR_LOCAL_ALLOC_OR_GOTO (frame->local, out);
local = frame->local;
ret = afr_local_init (local, priv, &op_errno);
@@ -236,7 +226,6 @@ afr_open (call_frame_t *frame, xlator_t *this, loc_t *loc, int32_t flags,
loc_copy (&local->loc, loc);
local->cont.open.flags = flags;
- local->cont.open.wbflags = wbflags;
local->fd = fd_ref (fd);
@@ -245,7 +234,7 @@ afr_open (call_frame_t *frame, xlator_t *this, loc_t *loc, int32_t flags,
STACK_WIND_COOKIE (frame, afr_open_cbk, (void *) (long) i,
priv->children[i],
priv->children[i]->fops->open,
- loc, wind_flags, fd, wbflags);
+ loc, wind_flags, fd, xdata);
if (!--call_count)
break;
@@ -255,75 +244,35 @@ afr_open (call_frame_t *frame, xlator_t *this, loc_t *loc, int32_t flags,
ret = 0;
out:
if (ret < 0)
- AFR_STACK_UNWIND (open, frame, -1, op_errno, fd);
+ AFR_STACK_UNWIND (open, frame, -1, op_errno, fd, xdata);
return 0;
}
-//NOTE: this function should be called with holding the lock on
-//fd to which fd_ctx belongs
-void
-afr_get_resumable_calls (xlator_t *this, afr_fd_ctx_t *fd_ctx,
- struct list_head *list)
-{
- afr_fd_paused_call_t *paused_call = NULL;
- afr_fd_paused_call_t *tmp = NULL;
- afr_local_t *call_local = NULL;
- afr_private_t *priv = NULL;
- int i = 0;
- gf_boolean_t call = _gf_false;
-
- priv = this->private;
- list_for_each_entry_safe (paused_call, tmp, &fd_ctx->paused_calls,
- call_list) {
- call = _gf_true;
- call_local = paused_call->frame->local;
- for (i = 0; i < priv->child_count; i++) {
- if (call_local->child_up[i] &&
- (fd_ctx->opened_on[i] == AFR_FD_OPENING))
- call = _gf_false;
- }
-
- if (call) {
- list_del_init (&paused_call->call_list);
- list_add (&paused_call->call_list, list);
- }
- }
-}
-
-void
-afr_resume_calls (xlator_t *this, struct list_head *list)
-{
- afr_fd_paused_call_t *paused_call = NULL;
- afr_fd_paused_call_t *tmp = NULL;
- afr_local_t *call_local = NULL;
-
- list_for_each_entry_safe (paused_call, tmp, list, call_list) {
- list_del_init (&paused_call->call_list);
- call_local = paused_call->frame->local;
- call_local->fop_call_continue (paused_call->frame, this);
- GF_FREE (paused_call);
- }
-}
-
int
afr_openfd_fix_open_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
- int32_t op_ret, int32_t op_errno, fd_t *fd)
+ int32_t op_ret, int32_t op_errno, fd_t *fd,
+ dict_t *xdata)
{
- afr_local_t *local = NULL;
- afr_private_t *priv = NULL;
- afr_fd_ctx_t *fd_ctx = NULL;
- int call_count = 0;
- int child_index = (long) cookie;
- struct list_head paused_calls = {0};
- gf_boolean_t fop_paused = _gf_false;
+ afr_local_t *local = NULL;
+ afr_private_t *priv = NULL;
+ afr_fd_ctx_t *fd_ctx = NULL;
+ int call_count = 0;
+ int child_index = (long) cookie;
priv = this->private;
local = frame->local;
- call_count = afr_frame_return (frame);
+ if (op_ret >= 0) {
+ gf_log (this->name, GF_LOG_DEBUG, "fd for %s opened "
+ "successfully on subvolume %s", local->loc.path,
+ priv->children[child_index]->name);
+ } else {
+ gf_log (this->name, GF_LOG_ERROR, "Failed to open %s "
+ "on subvolume %s", local->loc.path,
+ priv->children[child_index]->name);
+ }
- //Note: No frame locking needed for this block of code
fd_ctx = afr_fd_ctx_get (local->fd, this);
if (!fd_ctx) {
gf_log (this->name, GF_LOG_WARNING,
@@ -331,90 +280,73 @@ afr_openfd_fix_open_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
goto out;
}
- fop_paused = local->fop_paused;
LOCK (&local->fd->lock);
{
if (op_ret >= 0) {
fd_ctx->opened_on[child_index] = AFR_FD_OPENED;
- gf_log (this->name, GF_LOG_INFO, "fd for %s opened "
- "successfully on subvolume %s", local->loc.path,
- priv->children[child_index]->name);
} else {
- //Change open status from OPENING to NOT OPENED.
fd_ctx->opened_on[child_index] = AFR_FD_NOT_OPENED;
}
- if (call_count == 0) {
- INIT_LIST_HEAD (&paused_calls);
- afr_get_resumable_calls (this, fd_ctx, &paused_calls);
- }
}
UNLOCK (&local->fd->lock);
out:
- if (call_count == 0) {
- afr_resume_calls (this, &paused_calls);
- //If the fop is paused then resume_calls will continue the fop
- if (fop_paused)
- goto done;
-
- if (local->fop_call_continue)
- local->fop_call_continue (frame, this);
- else
- AFR_STACK_DESTROY (frame);
- }
+ call_count = afr_frame_return (frame);
+ if (call_count == 0)
+ AFR_STACK_DESTROY (frame);
-done:
return 0;
}
-int
-afr_fix_open (call_frame_t *frame, xlator_t *this, afr_fd_ctx_t *fd_ctx,
- int need_open_count, int *need_open)
+void
+afr_fix_open (xlator_t *this, fd_t *fd, size_t need_open_count, int *need_open)
{
- afr_local_t *local = NULL;
- afr_private_t *priv = NULL;
- int i = 0;
- call_frame_t *open_frame = NULL;
- afr_local_t *open_local = NULL;
- int ret = -1;
- ia_type_t ia_type = IA_INVAL;
- int32_t op_errno = 0;
-
- GF_ASSERT (fd_ctx);
- GF_ASSERT (need_open_count > 0);
- GF_ASSERT (need_open);
+ afr_private_t *priv = NULL;
+ int i = 0;
+ call_frame_t *frame = NULL;
+ afr_local_t *local = NULL;
+ int ret = -1;
+ int32_t op_errno = 0;
+ afr_fd_ctx_t *fd_ctx = NULL;
- local = frame->local;
priv = this->private;
- if (!local->fop_call_continue) {
- open_frame = copy_frame (frame);
- if (!open_frame) {
- ret = -ENOMEM;
- goto out;
- }
- ALLOC_OR_GOTO (open_frame->local, afr_local_t, out);
- open_local = open_frame->local;
- ret = afr_local_init (open_local, priv, &op_errno);
- if (ret < 0)
- goto out;
- loc_copy (&open_local->loc, &local->loc);
- open_local->fd = fd_ref (local->fd);
- } else {
- ret = 0;
- open_frame = frame;
- open_local = local;
+
+ if (!afr_is_fd_fixable (fd) || !need_open || !need_open_count)
+ goto out;
+
+ fd_ctx = afr_fd_ctx_get (fd, this);
+ if (!fd_ctx) {
+ ret = -1;
+ goto out;
+ }
+
+ frame = create_frame (this, this->ctx->pool);
+ if (!frame) {
+ ret = -1;
+ goto out;
}
- open_local->call_count = need_open_count;
+ AFR_LOCAL_ALLOC_OR_GOTO (frame->local, out);
+ local = frame->local;
+ ret = afr_local_init (local, priv, &op_errno);
+ if (ret < 0)
+ goto out;
+
+ local->loc.inode = inode_ref (fd->inode);
+ ret = loc_path (&local->loc, NULL);
+ if (ret < 0)
+ goto out;
+
+ local->fd = fd_ref (fd);
+ local->call_count = need_open_count;
- gf_log (this->name, GF_LOG_DEBUG, "need open count: %d",
+ gf_log (this->name, GF_LOG_DEBUG, "need open count: %zd",
need_open_count);
- ia_type = open_local->fd->inode->ia_type;
- GF_ASSERT (ia_type != IA_INVAL);
for (i = 0; i < priv->child_count; i++) {
if (!need_open[i])
continue;
- if (IA_IFDIR == ia_type) {
+
+ if (IA_IFDIR == fd->inode->ia_type) {
gf_log (this->name, GF_LOG_DEBUG,
"opening fd for dir %s on subvolume %s",
local->loc.path, priv->children[i]->name);
@@ -423,25 +355,28 @@ afr_fix_open (call_frame_t *frame, xlator_t *this, afr_fd_ctx_t *fd_ctx,
(void*) (long) i,
priv->children[i],
priv->children[i]->fops->opendir,
- &open_local->loc, open_local->fd);
+ &local->loc, local->fd,
+ NULL);
} else {
gf_log (this->name, GF_LOG_DEBUG,
"opening fd for file %s on subvolume %s",
local->loc.path, priv->children[i]->name);
- STACK_WIND_COOKIE (open_frame, afr_openfd_fix_open_cbk,
+ STACK_WIND_COOKIE (frame, afr_openfd_fix_open_cbk,
(void *)(long) i,
priv->children[i],
priv->children[i]->fops->open,
- &open_local->loc, fd_ctx->flags,
- open_local->fd, fd_ctx->wbflags);
+ &local->loc,
+ fd_ctx->flags & (~O_TRUNC),
+ local->fd, NULL);
}
}
+ op_errno = 0;
+ ret = 0;
out:
if (op_errno)
- ret = -op_errno;
- if (ret && open_frame)
- AFR_STACK_DESTROY (open_frame);
- return ret;
+ ret = -1; //For handling ALLOC_OR_GOTO
+ if (ret && frame)
+ AFR_STACK_DESTROY (frame);
}
diff --git a/xlators/cluster/afr/src/afr-self-heal-algorithm.c b/xlators/cluster/afr/src/afr-self-heal-algorithm.c
index cb1516d84..83846f152 100644
--- a/xlators/cluster/afr/src/afr-self-heal-algorithm.c
+++ b/xlators/cluster/afr/src/afr-self-heal-algorithm.c
@@ -1,23 +1,15 @@
/*
- Copyright (c) 2010-2011 Gluster, Inc. <http://www.gluster.com>
+ Copyright (c) 2008-2012 Red Hat, Inc. <http://www.redhat.com>
This file is part of GlusterFS.
- GlusterFS is free software; you can redistribute it and/or modify
- it under the terms of the GNU General Public License as published
- by the Free Software Foundation; either version 3 of the License,
- or (at your option) any later version.
-
- GlusterFS is distributed in the hope that it will be useful, but
- WITHOUT ANY WARRANTY; without even the implied warranty of
- MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU
- General Public License for more details.
-
- You should have received a copy of the GNU General Public License
- along with this program. If not, see
- <http://www.gnu.org/licenses/>.
+ This file is licensed to you under your choice of the GNU Lesser
+ General Public License, version 3 or any later version (LGPLv3 or
+ later), or the GNU General Public License, version 2 (GPLv2), in all
+ cases as published by the Free Software Foundation.
*/
+#include <openssl/md5.h>
#include "glusterfs.h"
#include "afr.h"
#include "xlator.h"
@@ -33,7 +25,6 @@
#include "compat-errno.h"
#include "compat.h"
#include "byte-order.h"
-#include "md5.h"
#include "afr-transaction.h"
#include "afr-self-heal.h"
@@ -72,8 +63,7 @@ sh_private_cleanup (call_frame_t *frame, xlator_t *this)
sh = &local->self_heal;
sh_priv = sh->private;
- if (sh_priv)
- GF_FREE (sh_priv);
+ GF_FREE (sh_priv);
}
static int
@@ -110,10 +100,10 @@ sh_loop_driver_done (call_frame_t *sh_frame, xlator_t *this,
}
sh_private_cleanup (sh_frame, this);
- if (sh->op_failed) {
+ if (is_self_heal_failed (sh, AFR_CHECK_SPECIFIC)) {
GF_ASSERT (!last_loop_frame);
//loop_finish should have happened and the old_loop should be NULL
- gf_log (this->name, GF_LOG_INFO,
+ gf_log (this->name, GF_LOG_DEBUG,
"self-heal aborting on %s",
local->loc.path);
@@ -121,10 +111,10 @@ sh_loop_driver_done (call_frame_t *sh_frame, xlator_t *this,
} else {
GF_ASSERT (last_loop_frame);
if (diff_blocks == total_blocks) {
- gf_log (this->name, GF_LOG_INFO, "full self-heal "
+ gf_log (this->name, GF_LOG_DEBUG, "full self-heal "
"completed on %s",local->loc.path);
} else {
- gf_log (this->name, GF_LOG_INFO,
+ gf_log (this->name, GF_LOG_DEBUG,
"diff self-heal on %s: completed. "
"(%d blocks of %d were different (%.2f%%))",
local->loc.path, diff_blocks, total_blocks,
@@ -153,7 +143,7 @@ sh_loop_finish (call_frame_t *loop_frame, xlator_t *this)
}
if (loop_sh && loop_sh->data_lock_held) {
- afr_sh_data_unlock (loop_frame, this,
+ afr_sh_data_unlock (loop_frame, this, this->name,
sh_destroy_frame);
} else {
sh_destroy_frame (loop_frame, this);
@@ -174,7 +164,7 @@ sh_loop_lock_success (call_frame_t *loop_frame, xlator_t *this)
sh_loop_finish (loop_sh->old_loop_frame, this);
loop_sh->old_loop_frame = NULL;
- gf_log (this->name, GF_LOG_DEBUG, "Aquired lock for range %"PRIu64
+ gf_log (this->name, GF_LOG_DEBUG, "Acquired lock for range %"PRIu64
" %"PRIu64, loop_sh->offset, loop_sh->block_size);
loop_sh->data_lock_held = _gf_true;
loop_sh->sh_data_algo_start (loop_frame, this);
@@ -224,7 +214,7 @@ sh_loop_frame_create (call_frame_t *sh_frame, xlator_t *this,
goto out;
//We want the frame to have same lk_owner as sh_frame
//so that locks translator allows conflicting locks
- new_loop_local = afr_local_copy (local, this);
+ new_loop_local = afr_self_heal_local_init (local, this);
if (!new_loop_local)
goto out;
new_loop_frame->local = new_loop_local;
@@ -239,7 +229,7 @@ sh_loop_frame_create (call_frame_t *sh_frame, xlator_t *this,
gf_afr_mt_char);
if (!new_loop_sh->write_needed)
goto out;
- new_loop_sh->checksum = GF_CALLOC (priv->child_count, MD5_DIGEST_LEN,
+ new_loop_sh->checksum = GF_CALLOC (priv->child_count, MD5_DIGEST_LENGTH,
gf_afr_mt_uint8_t);
if (!new_loop_sh->checksum)
goto out;
@@ -283,10 +273,10 @@ sh_loop_start (call_frame_t *sh_frame, xlator_t *this, off_t offset,
new_loop_sh->offset = offset;
new_loop_sh->block_size = sh->block_size;
afr_sh_data_lock (new_loop_frame, this, offset, new_loop_sh->block_size,
- sh_loop_lock_success, sh_loop_lock_failure);
+ _gf_true, this->name, sh_loop_lock_success, sh_loop_lock_failure);
return 0;
out:
- sh->op_failed = 1;
+ afr_set_self_heal_status (sh, AFR_SELF_HEAL_FAILED);
if (old_loop_frame)
sh_loop_finish (old_loop_frame, this);
sh_loop_return (sh_frame, this, new_loop_frame, -1, ENOMEM);
@@ -317,8 +307,9 @@ sh_loop_driver (call_frame_t *sh_frame, xlator_t *this,
sh_priv->loops_running--;
offset = sh_priv->offset;
block_size = sh->block_size;
- while ((!sh->eof_reached) && (0 == sh->op_failed) &&
- (sh_priv->loops_running < priv->data_self_heal_window_size)
+ while ((!sh->eof_reached) &&
+ (!is_self_heal_failed (sh, AFR_CHECK_SPECIFIC)) &&
+ (sh_priv->loops_running < priv->data_self_heal_window_size)
&& (sh_priv->offset < sh->file_size)) {
loop++;
@@ -337,7 +328,8 @@ sh_loop_driver (call_frame_t *sh_frame, xlator_t *this,
if (0 == loop) {
//loop finish does unlock, but the erasing of the pending
//xattrs needs to happen before that so do not finish the loop
- if (is_driver_done && !sh->op_failed)
+ if (is_driver_done &&
+ !is_self_heal_failed (sh, AFR_CHECK_SPECIFIC))
goto driver_done;
if (old_loop_frame) {
sh_loop_finish (old_loop_frame, this);
@@ -348,7 +340,7 @@ sh_loop_driver (call_frame_t *sh_frame, xlator_t *this,
//If we have more loops to form we should finish previous loop after
//the next loop lock
while (loop--) {
- if (sh->op_failed) {
+ if (is_self_heal_failed (sh, AFR_CHECK_SPECIFIC)) {
// op failed in other loop, stop spawning more loops
if (old_loop_frame) {
sh_loop_finish (old_loop_frame, this);
@@ -394,7 +386,7 @@ sh_loop_return (call_frame_t *sh_frame, xlator_t *this, call_frame_t *loop_frame
}
if (op_ret == -1) {
- sh->op_failed = 1;
+ afr_set_self_heal_status (sh, AFR_SELF_HEAL_FAILED);
afr_sh_set_error (sh, op_errno);
if (loop_frame) {
sh_loop_finish (loop_frame, this);
@@ -410,7 +402,7 @@ sh_loop_return (call_frame_t *sh_frame, xlator_t *this, call_frame_t *loop_frame
static int
sh_loop_write_cbk (call_frame_t *loop_frame, void *cookie, xlator_t *this,
int32_t op_ret, int32_t op_errno, struct iatt *buf,
- struct iatt *postbuf)
+ struct iatt *postbuf, dict_t *xdata)
{
afr_private_t * priv = NULL;
afr_local_t * loop_local = NULL;
@@ -442,13 +434,22 @@ sh_loop_write_cbk (call_frame_t *loop_frame, void *cookie, xlator_t *this,
priv->children[child_index]->name,
strerror (op_errno));
- sh->op_failed = 1;
+ afr_set_self_heal_status (sh, AFR_SELF_HEAL_FAILED);
afr_sh_set_error (loop_sh, op_errno);
+ } else if (op_ret < loop_local->cont.writev.vector->iov_len) {
+ gf_log (this->name, GF_LOG_ERROR,
+ "incomplete write to %s on subvolume %s "
+ "(expected %lu, returned %d)", sh_local->loc.path,
+ priv->children[child_index]->name,
+ loop_local->cont.writev.vector->iov_len, op_ret);
+ afr_set_self_heal_status (sh, AFR_SELF_HEAL_FAILED);
}
call_count = afr_frame_return (loop_frame);
if (call_count == 0) {
+ iobref_unref(loop_local->cont.writev.iobref);
+
sh_loop_return (sh_frame, this, loop_frame,
loop_sh->op_ret, loop_sh->op_errno);
}
@@ -456,12 +457,41 @@ sh_loop_write_cbk (call_frame_t *loop_frame, void *cookie, xlator_t *this,
return 0;
}
+static void
+sh_prune_writes_needed (call_frame_t *sh_frame, call_frame_t *loop_frame,
+ afr_private_t *priv)
+{
+ afr_local_t *sh_local = NULL;
+ afr_self_heal_t *sh = NULL;
+ afr_local_t *loop_local = NULL;
+ afr_self_heal_t *loop_sh = NULL;
+ int i = 0;
+
+ sh_local = sh_frame->local;
+ sh = &sh_local->self_heal;
+
+ if (!strcmp (sh->algo->name, "diff"))
+ return;
+
+ loop_local = loop_frame->local;
+ loop_sh = &loop_local->self_heal;
+
+ /* full self-heal guarantees there exists atleast 1 file with size 0
+ * That means for other files we can preserve holes that come after
+ * its size before 'trim'
+ */
+ for (i = 0; i < priv->child_count; i++) {
+ if (loop_sh->write_needed[i] &&
+ ((loop_sh->offset + 1) > sh->buf[i].ia_size))
+ loop_sh->write_needed[i] = 0;
+ }
+}
static int
sh_loop_read_cbk (call_frame_t *loop_frame, void *cookie,
xlator_t *this, int32_t op_ret, int32_t op_errno,
struct iovec *vector, int32_t count, struct iatt *buf,
- struct iobref *iobref)
+ struct iobref *iobref, dict_t *xdata)
{
afr_private_t * priv = NULL;
afr_local_t * loop_local = NULL;
@@ -486,7 +516,7 @@ sh_loop_read_cbk (call_frame_t *loop_frame, void *cookie,
if (op_ret <= 0) {
if (op_ret < 0) {
- sh->op_failed = 1;
+ afr_set_self_heal_status (sh, AFR_SELF_HEAL_FAILED);
gf_log (this->name, GF_LOG_ERROR, "read failed on %d "
"for %s reason :%s", sh->source,
sh_local->loc.path, strerror (errno));
@@ -499,18 +529,26 @@ sh_loop_read_cbk (call_frame_t *loop_frame, void *cookie,
goto out;
}
- if (loop_sh->file_has_holes && iov_0filled (vector, count) == 0) {
- gf_log (this->name, GF_LOG_DEBUG, "0 filled block");
- sh_loop_return (sh_frame, this, loop_frame,
- op_ret, op_errno);
- goto out;
- }
+ if (loop_sh->file_has_holes && iov_0filled (vector, count) == 0)
+ sh_prune_writes_needed (sh_frame, loop_frame, priv);
call_count = sh_number_of_writes_needed (loop_sh->write_needed,
priv->child_count);
- GF_ASSERT (call_count > 0);
+ if (call_count == 0) {
+ sh_loop_return (sh_frame, this, loop_frame, 0, 0);
+ goto out;
+ }
+
loop_local->call_count = call_count;
+ /*
+ * We only really need the request size at the moment, but the buffer
+ * is required if we want to issue a retry in the event of a short write.
+ * Therefore, we duplicate the vector and ref the iobref here...
+ */
+ loop_local->cont.writev.vector = iov_dup(vector, count);
+ loop_local->cont.writev.iobref = iobref_ref(iobref);
+
for (i = 0; i < priv->child_count; i++) {
if (!loop_sh->write_needed[i])
continue;
@@ -519,7 +557,7 @@ sh_loop_read_cbk (call_frame_t *loop_frame, void *cookie,
priv->children[i],
priv->children[i]->fops->writev,
loop_sh->healing_fd, vector, count,
- loop_sh->offset, iobref);
+ loop_sh->offset, 0, iobref, NULL);
if (!--call_count)
break;
@@ -546,7 +584,7 @@ sh_loop_read (call_frame_t *loop_frame, xlator_t *this)
priv->children[loop_sh->source],
priv->children[loop_sh->source]->fops->readv,
loop_sh->healing_fd, loop_sh->block_size,
- loop_sh->offset);
+ loop_sh->offset, 0, NULL);
return 0;
}
@@ -555,7 +593,8 @@ sh_loop_read (call_frame_t *loop_frame, xlator_t *this)
static int
sh_diff_checksum_cbk (call_frame_t *loop_frame, void *cookie, xlator_t *this,
int32_t op_ret, int32_t op_errno,
- uint32_t weak_checksum, uint8_t *strong_checksum)
+ uint32_t weak_checksum, uint8_t *strong_checksum,
+ dict_t *xdata)
{
afr_private_t *priv = NULL;
afr_local_t *loop_local = NULL;
@@ -587,10 +626,10 @@ sh_diff_checksum_cbk (call_frame_t *loop_frame, void *cookie, xlator_t *this,
"checksum on %s failed on subvolume %s (%s)",
sh_local->loc.path, priv->children[child_index]->name,
strerror (op_errno));
- sh->op_failed = 1;
+ afr_set_self_heal_status (sh, AFR_SELF_HEAL_FAILED);
} else {
- memcpy (loop_sh->checksum + child_index * MD5_DIGEST_LEN,
- strong_checksum, MD5_DIGEST_LEN);
+ memcpy (loop_sh->checksum + child_index * MD5_DIGEST_LENGTH,
+ strong_checksum, MD5_DIGEST_LENGTH);
}
call_count = afr_frame_return (loop_frame);
@@ -600,9 +639,9 @@ sh_diff_checksum_cbk (call_frame_t *loop_frame, void *cookie, xlator_t *this,
if (sh->sources[i] || !sh_local->child_up[i])
continue;
- if (memcmp (loop_sh->checksum + (i * MD5_DIGEST_LEN),
- loop_sh->checksum + (sh->source * MD5_DIGEST_LEN),
- MD5_DIGEST_LEN)) {
+ if (memcmp (loop_sh->checksum + (i * MD5_DIGEST_LENGTH),
+ loop_sh->checksum + (sh->source * MD5_DIGEST_LENGTH),
+ MD5_DIGEST_LENGTH)) {
/*
Checksums differ, so this block
must be written to this sink
@@ -625,7 +664,8 @@ sh_diff_checksum_cbk (call_frame_t *loop_frame, void *cookie, xlator_t *this,
}
UNLOCK (&sh_priv->lock);
- if (write_needed && !sh->op_failed) {
+ if (write_needed &&
+ !is_self_heal_failed (sh, AFR_CHECK_SPECIFIC)) {
sh_loop_read (loop_frame, this);
} else {
sh_loop_return (sh_frame, this, loop_frame,
@@ -658,7 +698,7 @@ sh_diff_checksum (call_frame_t *loop_frame, xlator_t *this)
priv->children[loop_sh->source],
priv->children[loop_sh->source]->fops->rchecksum,
loop_sh->healing_fd,
- loop_sh->offset, loop_sh->block_size);
+ loop_sh->offset, loop_sh->block_size, NULL);
for (i = 0; i < priv->child_count; i++) {
if (loop_sh->sources[i] || !loop_local->child_up[i])
@@ -669,7 +709,7 @@ sh_diff_checksum (call_frame_t *loop_frame, xlator_t *this)
priv->children[i],
priv->children[i]->fops->rchecksum,
loop_sh->healing_fd,
- loop_sh->offset, loop_sh->block_size);
+ loop_sh->offset, loop_sh->block_size, NULL);
if (!--call_count)
break;
@@ -714,14 +754,15 @@ out:
return sh_priv;
}
-void
-afr_sh_transfer_lock (call_frame_t *dst, call_frame_t *src,
+int
+afr_sh_transfer_lock (call_frame_t *dst, call_frame_t *src, char *dom,
unsigned int child_count)
{
afr_local_t *dst_local = NULL;
afr_self_heal_t *dst_sh = NULL;
afr_local_t *src_local = NULL;
afr_self_heal_t *src_sh = NULL;
+ int ret = -1;
dst_local = dst->local;
dst_sh = &dst_local->self_heal;
@@ -729,9 +770,12 @@ afr_sh_transfer_lock (call_frame_t *dst, call_frame_t *src,
src_sh = &src_local->self_heal;
GF_ASSERT (src_sh->data_lock_held);
GF_ASSERT (!dst_sh->data_lock_held);
- afr_lk_transfer_datalock (dst, src, child_count);
+ ret = afr_lk_transfer_datalock (dst, src, dom, child_count);
+ if (ret)
+ return ret;
src_sh->data_lock_held = _gf_false;
dst_sh->data_lock_held = _gf_true;
+ return 0;
}
int
@@ -753,7 +797,10 @@ afr_sh_start_loops (call_frame_t *sh_frame, xlator_t *this,
ret = sh_loop_frame_create (sh_frame, this, NULL, &first_loop_frame);
if (ret)
goto out;
- afr_sh_transfer_lock (first_loop_frame, sh_frame, priv->child_count);
+ ret = afr_sh_transfer_lock (first_loop_frame, sh_frame, this->name,
+ priv->child_count);
+ if (ret)
+ goto out;
sh->private = afr_sh_priv_init ();
if (!sh->private) {
ret = -1;
@@ -763,7 +810,7 @@ afr_sh_start_loops (call_frame_t *sh_frame, xlator_t *this,
ret = 0;
out:
if (ret) {
- sh->op_failed = 1;
+ afr_set_self_heal_status (sh, AFR_SELF_HEAL_FAILED);
sh_loop_driver_done (sh_frame, this, NULL);
}
return 0;
diff --git a/xlators/cluster/afr/src/afr-self-heal-algorithm.h b/xlators/cluster/afr/src/afr-self-heal-algorithm.h
index 04d8e8a6c..6b20789b1 100644
--- a/xlators/cluster/afr/src/afr-self-heal-algorithm.h
+++ b/xlators/cluster/afr/src/afr-self-heal-algorithm.h
@@ -1,26 +1,16 @@
/*
- Copyright (c) 2010-2011 Gluster, Inc. <http://www.gluster.com>
- This file is part of GlusterFS.
+ Copyright (c) 2008-2012 Red Hat, Inc. <http://www.redhat.com>
+ This file is part of GlusterFS.
- GlusterFS is free software; you can redistribute it and/or modify
- it under the terms of the GNU General Public License as published
- by the Free Software Foundation; either version 3 of the License,
- or (at your option) any later version.
-
- GlusterFS is distributed in the hope that it will be useful, but
- WITHOUT ANY WARRANTY; without even the implied warranty of
- MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU
- General Public License for more details.
-
- You should have received a copy of the GNU General Public License
- along with this program. If not, see
- <http://www.gnu.org/licenses/>.
+ This file is licensed to you under your choice of the GNU Lesser
+ General Public License, version 3 or any later version (LGPLv3 or
+ later), or the GNU General Public License, version 2 (GPLv2), in all
+ cases as published by the Free Software Foundation.
*/
#ifndef __AFR_SELF_HEAL_ALGORITHM_H__
#define __AFR_SELF_HEAL_ALGORITHM_H__
-
typedef int (*afr_sh_algo_fn) (call_frame_t *frame,
xlator_t *this);
diff --git a/xlators/cluster/afr/src/afr-self-heal-common.c b/xlators/cluster/afr/src/afr-self-heal-common.c
index 5acbf90aa..ef92b4205 100644
--- a/xlators/cluster/afr/src/afr-self-heal-common.c
+++ b/xlators/cluster/afr/src/afr-self-heal-common.c
@@ -1,20 +1,11 @@
/*
- Copyright (c) 2008-2011 Gluster, Inc. <http://www.gluster.com>
+ Copyright (c) 2008-2012 Red Hat, Inc. <http://www.redhat.com>
This file is part of GlusterFS.
- GlusterFS is free software; you can redistribute it and/or modify
- it under the terms of the GNU General Public License as published
- by the Free Software Foundation; either version 3 of the License,
- or (at your option) any later version.
-
- GlusterFS is distributed in the hope that it will be useful, but
- WITHOUT ANY WARRANTY; without even the implied warranty of
- MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU
- General Public License for more details.
-
- You should have received a copy of the GNU General Public License
- along with this program. If not, see
- <http://www.gnu.org/licenses/>.
+ This file is licensed to you under your choice of the GNU Lesser
+ General Public License, version 3 or any later version (LGPLv3 or
+ later), or the GNU General Public License, version 2 (GPLv2), in all
+ cases as published by the Free Software Foundation.
*/
#include "glusterfs.h"
@@ -27,6 +18,52 @@
#include "afr-self-heal.h"
#include "pump.h"
+#define ADD_FMT_STRING(msg, off, sh_str, status, print_log) \
+ do { \
+ if (AFR_SELF_HEAL_NOT_ATTEMPTED != status) { \
+ off += snprintf (msg + off, sizeof (msg) - off, \
+ " "sh_str" self heal %s,", \
+ get_sh_completion_status (status));\
+ print_log = 1; \
+ } \
+ } while (0)
+
+#define ADD_FMT_STRING_SYNC(msg, off, sh_str, status, print_log) \
+ do { \
+ if (AFR_SELF_HEAL_SYNC_BEGIN == status || \
+ AFR_SELF_HEAL_FAILED == status) { \
+ off += snprintf (msg + off, sizeof (msg) - off, \
+ " "sh_str" self heal %s,", \
+ get_sh_completion_status (status));\
+ print_log = 1; \
+ } \
+ } while (0)
+
+
+void
+afr_sh_reset (call_frame_t *frame, xlator_t *this)
+{
+ afr_local_t *local = NULL;
+ afr_self_heal_t *sh = NULL;
+ afr_private_t *priv = NULL;
+
+ local = frame->local;
+ sh = &local->self_heal;
+ priv = this->private;
+
+ memset (sh->child_errno, 0,
+ sizeof (*sh->child_errno) * priv->child_count);
+ memset (sh->buf, 0, sizeof (*sh->buf) * priv->child_count);
+ memset (sh->parentbufs, 0,
+ sizeof (*sh->parentbufs) * priv->child_count);
+ memset (sh->success, 0, sizeof (*sh->success) * priv->child_count);
+ memset (sh->locked_nodes, 0,
+ sizeof (*sh->locked_nodes) * priv->child_count);
+ sh->active_sinks = 0;
+
+ afr_reset_xattr (sh->xattr, priv->child_count);
+}
+
//Intersection[child]=1 if child is part of intersection
void
afr_children_intersection_get (int32_t *set1, int32_t *set2,
@@ -97,8 +134,8 @@ void
afr_sh_set_error (afr_self_heal_t *sh, int32_t op_errno)
{
sh->op_ret = -1;
- if (afr_error_more_important (sh->op_errno, op_errno))
- sh->op_errno = op_errno;
+ sh->op_errno = afr_most_important_error(sh->op_errno, op_errno,
+ _gf_false);
}
void
@@ -120,13 +157,85 @@ afr_sh_print_pending_matrix (int32_t *pending_matrix[], xlator_t *this)
ptr += sprintf (ptr, "%d ", pending_matrix[i][j]);
}
sprintf (ptr, "]");
- gf_log (this->name, GF_LOG_DEBUG,
- "pending_matrix: %s", buf);
+ gf_log (this->name, GF_LOG_DEBUG, "pending_matrix: %s", buf);
}
GF_FREE (buf);
}
+char*
+afr_get_pending_matrix_str (int32_t *pending_matrix[], xlator_t *this)
+{
+ afr_private_t * priv = this->private;
+ char *buf = NULL;
+ char *ptr = NULL;
+ int i = 0;
+ int j = 0;
+ int child_count = priv->child_count;
+ char *matrix_begin = "[ [ ";
+ char *matrix_end = "] ]";
+ char *seperator = "] [ ";
+ int pending_entry_strlen = 12; //Including space after entry
+ int matrix_begin_strlen = 0;
+ int matrix_end_strlen = 0;
+ int seperator_strlen = 0;
+ int string_length = 0;
+ char *msg = "- Pending matrix: ";
+
+ /*
+ * for a list of lists of [ [ a b ] [ c d ] ]
+ * */
+
+ matrix_begin_strlen = strlen (matrix_begin);
+ matrix_end_strlen = strlen (matrix_end);
+ seperator_strlen = strlen (seperator);
+ string_length = matrix_begin_strlen + matrix_end_strlen
+ + (child_count -1) * seperator_strlen
+ + (child_count * child_count * pending_entry_strlen);
+
+ buf = GF_CALLOC (1, 1 + strlen (msg) + string_length , gf_afr_mt_char);
+ if (!buf)
+ goto out;
+
+ ptr = buf;
+ ptr += sprintf (ptr, "%s", msg);
+ ptr += sprintf (ptr, "%s", matrix_begin);
+ for (i = 0; i < priv->child_count; i++) {
+ for (j = 0; j < priv->child_count; j++) {
+ ptr += sprintf (ptr, "%d ", pending_matrix[i][j]);
+ }
+ if (i < priv->child_count -1)
+ ptr += sprintf (ptr, "%s", seperator);
+ }
+
+ ptr += sprintf (ptr, "%s", matrix_end);
+
+out:
+ return buf;
+}
+
+void
+afr_sh_print_split_brain_log (int32_t *pending_matrix[], xlator_t *this,
+ const char *loc)
+{
+ char *buf = NULL;
+ char *free_ptr = NULL;
+
+ buf = afr_get_pending_matrix_str (pending_matrix, this);
+ if (buf)
+ free_ptr = buf;
+ else
+ buf = "";
+
+
+ gf_log (this->name, GF_LOG_ERROR, "Unable to self-heal contents of '%s'"
+ " (possible split-brain). Please delete the file from all but "
+ "the preferred subvolume.%s", loc, buf);
+ GF_FREE (free_ptr);
+ return;
+}
+
+
void
afr_init_pending_matrix (int32_t **pending_matrix, size_t child_count)
{
@@ -392,6 +501,8 @@ afr_find_biggest_witness_among_fools (int32_t *witnesses,
{
int i = 0;
int biggest_witness = -1;
+ int biggest_witness_idx = -1;
+ int biggest_witness_cnt = -1;
GF_ASSERT (witnesses);
GF_ASSERT (characters);
@@ -401,10 +512,21 @@ afr_find_biggest_witness_among_fools (int32_t *witnesses,
if (characters[i].type != AFR_NODE_FOOL)
continue;
- if (biggest_witness < witnesses[i])
+ if (biggest_witness < witnesses[i]) {
biggest_witness = witnesses[i];
+ biggest_witness_idx = i;
+ biggest_witness_cnt = 1;
+ continue;
+ }
+
+ if (biggest_witness == witnesses[i])
+ biggest_witness_cnt++;
}
- return biggest_witness;
+
+ if (biggest_witness_cnt != 1)
+ return -1;
+
+ return biggest_witness_idx;
}
int
@@ -432,10 +554,84 @@ afr_mark_fool_as_source_by_witness (int32_t *sources, int32_t *witnesses,
return nsources;
}
+
+int
+afr_mark_fool_as_source_by_idx (int32_t *sources, int child_count, int idx)
+{
+ if (idx >= 0 && idx < child_count) {
+ sources[idx] = 1;
+ return 1;
+ }
+ return 0;
+}
+
+
+static int
+afr_find_largest_file_size (struct iatt *bufs, int32_t *success_children,
+ int child_count)
+{
+ int idx = -1;
+ int i = -1;
+ int child = -1;
+ uint64_t max_size = 0;
+ uint64_t min_size = 0;
+ int num_children = 0;
+
+ for (i = 0; i < child_count; i++) {
+ if (success_children[i] == -1)
+ break;
+
+ child = success_children[i];
+ if (bufs[child].ia_size > max_size) {
+ max_size = bufs[child].ia_size;
+ idx = child;
+ }
+
+ if ((num_children == 0) || (bufs[child].ia_size < min_size)) {
+ min_size = bufs[child].ia_size;
+ }
+
+ num_children++;
+ }
+
+ /* If sizes are same for all of them, finding sources will have to
+ * happen with pending changelog. So return -1
+ */
+ if ((num_children > 1) && (min_size == max_size))
+ return -1;
+ return idx;
+}
+
+
+static int
+afr_find_newest_file (struct iatt *bufs, int32_t *success_children,
+ int child_count)
+{
+ int idx = -1;
+ int i = -1;
+ int child = -1;
+ uint64_t max_ctime = 0;
+
+ for (i = 0; i < child_count; i++) {
+ if (success_children[i] == -1)
+ break;
+
+ child = success_children[i];
+ if (bufs[child].ia_ctime > max_ctime) {
+ max_ctime = bufs[child].ia_ctime;
+ idx = child;
+ }
+ }
+
+ return idx;
+}
+
+
static int
afr_mark_biggest_of_fools_as_source (int32_t *sources, int32_t **pending_matrix,
afr_node_character *characters,
- int child_count)
+ int32_t *success_children,
+ int child_count, struct iatt *bufs)
{
int32_t biggest_witness = 0;
int nsources = 0;
@@ -443,6 +639,11 @@ afr_mark_biggest_of_fools_as_source (int32_t *sources, int32_t **pending_matrix,
GF_ASSERT (child_count > 0);
+ biggest_witness = afr_find_largest_file_size (bufs, success_children,
+ child_count);
+ if (biggest_witness != -1)
+ goto found;
+
witnesses = GF_CALLOC (child_count, sizeof (*witnesses),
gf_afr_mt_int32_t);
if (NULL == witnesses) {
@@ -455,12 +656,17 @@ afr_mark_biggest_of_fools_as_source (int32_t *sources, int32_t **pending_matrix,
biggest_witness = afr_find_biggest_witness_among_fools (witnesses,
characters,
child_count);
- nsources = afr_mark_fool_as_source_by_witness (sources, witnesses,
- characters, child_count,
- biggest_witness);
+ if (biggest_witness != -1)
+ goto found;
+
+ biggest_witness = afr_find_newest_file (bufs, success_children,
+ child_count);
+
+found:
+ nsources = afr_mark_fool_as_source_by_idx (sources, child_count,
+ biggest_witness);
out:
- if (witnesses)
- GF_FREE (witnesses);
+ GF_FREE (witnesses);
return nsources;
}
@@ -526,6 +732,78 @@ out:
return nsources;
}
+int
+afr_get_no_xattr_dir_read_child (xlator_t *this, int32_t *success_children,
+ struct iatt *bufs)
+{
+ afr_private_t *priv = NULL;
+ int i = 0;
+ int child = -1;
+ int read_child = -1;
+
+ priv = this->private;
+ for (i = 0; i < priv->child_count; i++) {
+ child = success_children[i];
+ if (child < 0)
+ break;
+ if (read_child < 0)
+ read_child = child;
+ else if (bufs[read_child].ia_size < bufs[child].ia_size)
+ read_child = child;
+ }
+ return read_child;
+}
+
+int
+afr_sh_mark_zero_size_file_as_sink (struct iatt *bufs, int32_t *success_children,
+ int child_count, int32_t *sources)
+{
+ int nsources = 0;
+ int i = 0;
+ int child = 0;
+ gf_boolean_t sink_exists = _gf_false;
+ gf_boolean_t source_exists = _gf_false;
+ int source = -1;
+
+ for (i = 0; i < child_count; i++) {
+ child = success_children[i];
+ if (child < 0)
+ break;
+ if (!bufs[child].ia_size) {
+ sink_exists = _gf_true;
+ continue;
+ }
+ if (!source_exists) {
+ source_exists = _gf_true;
+ source = child;
+ continue;
+ }
+ if (bufs[source].ia_size != bufs[child].ia_size) {
+ nsources = -1;
+ goto out;
+ }
+ }
+ if (!source_exists && !sink_exists) {
+ nsources = -1;
+ goto out;
+ }
+
+ if (!source_exists || !sink_exists)
+ goto out;
+
+ for (i = 0; i < child_count; i++) {
+ child = success_children[i];
+ if (child < 0)
+ break;
+ if (bufs[child].ia_size) {
+ sources[child] = 1;
+ nsources++;
+ }
+ }
+out:
+ return nsources;
+}
+
char *
afr_get_character_str (afr_node_type type)
{
@@ -691,11 +969,24 @@ afr_mark_sources (xlator_t *this, int32_t *sources, int32_t **pending_matrix,
afr_find_character_types (characters, pending_matrix, success_children,
child_count);
if (afr_sh_all_nodes_innocent (characters, child_count)) {
- if (type == AFR_SELF_HEAL_METADATA)
+ switch (type) {
+ case AFR_SELF_HEAL_METADATA:
nsources = afr_sh_mark_lowest_uid_as_source (bufs,
success_children,
child_count,
sources);
+ break;
+ case AFR_SELF_HEAL_DATA:
+ nsources = afr_sh_mark_zero_size_file_as_sink (bufs,
+ success_children,
+ child_count,
+ sources);
+ if ((nsources < 0) && subvol_status)
+ *subvol_status |= SPLIT_BRAIN;
+ break;
+ default:
+ break;
+ }
goto out;
}
@@ -717,7 +1008,8 @@ afr_mark_sources (xlator_t *this, int32_t *sources, int32_t **pending_matrix,
nsources = afr_mark_biggest_of_fools_as_source (sources,
pending_matrix,
characters,
- child_count);
+ success_children,
+ child_count, bufs);
}
out:
@@ -735,50 +1027,108 @@ afr_sh_pending_to_delta (afr_private_t *priv, dict_t **xattr,
int32_t *delta_matrix[], unsigned char success[],
int child_count, afr_transaction_type type)
{
- int i = 0;
- int j = 0;
+ int tgt = 0;
+ int src = 0;
+ int value = 0;
afr_build_pending_matrix (priv->pending_key, delta_matrix, NULL,
xattr, type, priv->child_count);
- for (i = 0; i < priv->child_count; i++)
- for (j = 0; j < priv->child_count; j++)
- delta_matrix[i][j] = -delta_matrix[i][j];
+
+ /*
+ * The algorithm here has two parts. First, for each subvol indexed
+ * as tgt, we try to figure out what count everyone should have for it.
+ * If the self-heal succeeded, that's easy; the value is zero.
+ * Otherwise, the value is the maximum of the succeeding nodes' counts.
+ * Once we know the value, we loop through (possibly for a second time)
+ * setting each count to the difference so that when we're done all
+ * succeeding nodes will have the same count for tgt.
+ */
+ for (tgt = 0; tgt < priv->child_count; ++tgt) {
+ value = 0;
+ if (!success[tgt]) {
+ /* Find the maximum. */
+ for (src = 0; src < priv->child_count; ++src) {
+ if (!success[src]) {
+ continue;
+ }
+ if (delta_matrix[src][tgt] > value) {
+ value = delta_matrix[src][tgt];
+ }
+ }
+ }
+ /* Force everyone who succeeded to the chosen value. */
+ for (src = 0; src < priv->child_count; ++src) {
+ if (success[src]) {
+ delta_matrix[src][tgt] = value
+ - delta_matrix[src][tgt];
+ }
+ else {
+ delta_matrix[src][tgt] = 0;
+ }
+ }
+ }
}
int
-afr_sh_delta_to_xattr (afr_private_t *priv,
+afr_sh_delta_to_xattr (xlator_t *this,
int32_t *delta_matrix[], dict_t *xattr[],
int child_count, afr_transaction_type type)
{
- int i = 0;
- int j = 0;
- int k = 0;
- int ret = 0;
- int32_t *pending = NULL;
+ int i = 0;
+ int j = 0;
+ int k = 0;
+ int ret = 0;
+ int32_t *pending = NULL;
+ int32_t *local_pending = NULL;
+ afr_private_t *priv = NULL;
+ priv = this->private;
for (i = 0; i < child_count; i++) {
if (!xattr[i])
continue;
+ local_pending = NULL;
for (j = 0; j < child_count; j++) {
pending = GF_CALLOC (sizeof (int32_t), 3,
gf_afr_mt_int32_t);
- if (!pending)
+ if (!pending) {
+ gf_log (this->name, GF_LOG_ERROR,
+ "failed to allocate pending entry "
+ "for %s[%d] on %s",
+ priv->pending_key[j], type,
+ priv->children[i]->name);
continue;
+ }
/* 3 = data+metadata+entry */
k = afr_index_for_transaction_type (type);
pending[k] = hton32 (delta_matrix[i][j]);
+ if (j == i) {
+ local_pending = pending;
+ continue;
+ }
ret = dict_set_bin (xattr[i], priv->pending_key[j],
pending,
- 3 * sizeof (int32_t));
- if (ret < 0)
- gf_log (THIS->name, GF_LOG_WARNING,
+ AFR_NUM_CHANGE_LOGS * sizeof (int32_t));
+ if (ret < 0) {
+ gf_log (this->name, GF_LOG_WARNING,
"Unable to set dict value.");
+ GF_FREE (pending);
+ }
+ }
+ if (local_pending) {
+ ret = dict_set_bin (xattr[i], priv->pending_key[i],
+ local_pending,
+ AFR_NUM_CHANGE_LOGS * sizeof (int32_t));
+ if (ret < 0) {
+ gf_log (this->name, GF_LOG_WARNING,
+ "Unable to set dict value.");
+ GF_FREE (local_pending);
+ }
}
}
return 0;
@@ -786,127 +1136,23 @@ afr_sh_delta_to_xattr (afr_private_t *priv,
int
-afr_sh_has_metadata_pending (dict_t *xattr, xlator_t *this)
-{
- /* Indexable by result of afr_index_for_transaction_type(): 0 -- 2. */
- int32_t pending[3] = {0,};
- void *pending_raw = NULL;
- afr_private_t *priv = NULL;
- int ret = -1;
- int i = 0;
- int j = 0;
-
- priv = this->private;
-
- for (i = 0; i < priv->child_count; i++) {
- ret = dict_get_ptr (xattr, priv->pending_key[i],
- &pending_raw);
-
- if (ret != 0)
- return 0;
-
- memcpy (pending, pending_raw, sizeof(pending));
- j = afr_index_for_transaction_type (AFR_METADATA_TRANSACTION);
-
- if (pending[j])
- return 1;
- }
-
- return 0;
-}
-
-
-int
-afr_sh_has_data_pending (dict_t *xattr, xlator_t *this)
-{
- /* Indexable by result of afr_index_for_transaction_type(): 0 -- 2. */
- int32_t pending[3] = {0,};
- void *pending_raw = NULL;
- afr_private_t *priv = NULL;
- int ret = -1;
- int i = 0;
- int j = 0;
-
- priv = this->private;
-
- for (i = 0; i < priv->child_count; i++) {
- ret = dict_get_ptr (xattr, priv->pending_key[i],
- &pending_raw);
-
- if (ret != 0)
- return 0;
-
- memcpy (pending, pending_raw, sizeof(pending));
- j = afr_index_for_transaction_type (AFR_DATA_TRANSACTION);
-
- if (pending[j])
- return 1;
- }
-
- return 0;
-}
-
-
-int
-afr_sh_has_entry_pending (dict_t *xattr, xlator_t *this)
-{
- /* Indexable by result of afr_index_for_transaction_type(): 0 -- 2. */
- int32_t pending[3] = {0,};
- void *pending_raw = NULL;
- afr_private_t *priv = NULL;
- int ret = -1;
- int i = 0;
- int j = 0;
-
- priv = this->private;
-
- for (i = 0; i < priv->child_count; i++) {
- ret = dict_get_ptr (xattr, priv->pending_key[i],
- &pending_raw);
-
- if (ret != 0)
- return 0;
-
- memcpy (pending, pending_raw, sizeof(pending));
- j = afr_index_for_transaction_type (AFR_ENTRY_TRANSACTION);
-
- if (pending[j])
- return 1;
- }
-
- return 0;
-}
-
-int
afr_sh_missing_entries_done (call_frame_t *frame, xlator_t *this)
{
afr_local_t *local = NULL;
afr_self_heal_t *sh = NULL;
- afr_private_t *priv = NULL;
- int i = 0;
local = frame->local;
sh = &local->self_heal;
- priv = this->private;
-
-// memset (sh->child_errno, 0, sizeof (int) * priv->child_count);
- memset (sh->buf, 0, sizeof (struct iatt) * priv->child_count);
- for (i = 0; i < priv->child_count; i++) {
- sh->locked_nodes[i] = 0;
- }
+ afr_sh_reset (frame, this);
- for (i = 0; i < priv->child_count; i++) {
- if (sh->xattr[i])
- dict_unref (sh->xattr[i]);
- sh->xattr[i] = NULL;
- }
-
- if (local->govinda_gOvinda || sh->op_failed) {
- gf_log (this->name, GF_LOG_INFO,
+ if (local->unhealable) {
+ gf_log (this->name, GF_LOG_DEBUG,
"split brain found, aborting selfheal of %s",
local->loc.path);
- sh->op_failed = 1;
+ }
+
+ if (is_self_heal_failed (sh, AFR_CHECK_SPECIFIC)) {
sh->completion_cbk (frame, this);
} else {
gf_log (this->name, GF_LOG_TRACE,
@@ -992,7 +1238,7 @@ afr_sh_common_lookup_resp_handler (call_frame_t *frame, void *cookie,
sh->success_count++;
sh->xattr[child_index] = dict_ref (xattr);
} else {
- gf_log (this->name, GF_LOG_ERROR, "path %s on subvolume"
+ gf_log (this->name, GF_LOG_DEBUG, "path %s on subvolume"
" %s => -1 (%s)", loc->path,
priv->children[child_index]->name,
strerror (op_errno));
@@ -1023,8 +1269,7 @@ afr_valid_ia_type (ia_type_t ia_type)
int
afr_impunge_frame_create (call_frame_t *frame, xlator_t *this,
- int active_source, int ret_child, mode_t entry_mode,
- call_frame_t **impunge_frame)
+ int active_source, call_frame_t **impunge_frame)
{
afr_local_t *local = NULL;
afr_local_t *impunge_local = NULL;
@@ -1041,21 +1286,24 @@ afr_impunge_frame_create (call_frame_t *frame, xlator_t *this,
goto out;
}
- ALLOC_OR_GOTO (impunge_local, afr_local_t, out);
+ AFR_LOCAL_ALLOC_OR_GOTO (impunge_local, out);
local = frame->local;
new_frame->local = impunge_local;
impunge_sh = &impunge_local->self_heal;
impunge_sh->sh_frame = frame;
impunge_sh->active_source = active_source;
- impunge_sh->impunge_ret_child = ret_child;
- impunge_sh->impunging_entry_mode = entry_mode;
impunge_local->child_up = memdup (local->child_up,
sizeof (*local->child_up) *
priv->child_count);
if (!impunge_local->child_up)
goto out;
+ impunge_local->pending = afr_matrix_create (priv->child_count,
+ AFR_NUM_CHANGE_LOGS);
+ if (!impunge_local->pending)
+ goto out;
+
ret = afr_sh_common_create (impunge_sh, priv->child_count);
if (ret) {
op_errno = -ret;
@@ -1070,54 +1318,89 @@ out:
}
void
-afr_sh_call_entry_impunge_recreate (call_frame_t *frame, xlator_t *this,
- int child_index, struct iatt *buf,
- struct iatt *postparent,
- afr_impunge_done_cbk_t impunge_done)
+afr_sh_missing_entry_call_impunge_recreate (call_frame_t *frame, xlator_t *this,
+ struct iatt *buf,
+ struct iatt *postparent,
+ afr_impunge_done_cbk_t impunge_done)
{
call_frame_t *impunge_frame = NULL;
afr_local_t *local = NULL;
afr_local_t *impunge_local = NULL;
afr_self_heal_t *sh = NULL;
+ afr_self_heal_t *impunge_sh = NULL;
int ret = 0;
- mode_t mode = 0;
+ unsigned int enoent_count = 0;
+ afr_private_t *priv = NULL;
+ int i = 0;
+ int32_t op_errno = 0;
local = frame->local;
sh = &local->self_heal;
- mode = st_mode_from_ia (buf->ia_prot, buf->ia_type);
- ret = afr_impunge_frame_create (frame, this, sh->source, child_index,
- mode, &impunge_frame);
+ priv = this->private;
+
+ enoent_count = afr_errno_count (NULL, sh->child_errno,
+ priv->child_count, ENOENT);
+ if (!enoent_count) {
+ gf_log (this->name, GF_LOG_INFO,
+ "no missing files - %s. proceeding to metadata check",
+ local->loc.path);
+ goto out;
+ }
+ sh->impunge_done = impunge_done;
+ ret = afr_impunge_frame_create (frame, this, sh->source, &impunge_frame);
if (ret)
goto out;
impunge_local = impunge_frame->local;
+ impunge_sh = &impunge_local->self_heal;
loc_copy (&impunge_local->loc, &local->loc);
- sh->impunge_done = impunge_done;
- impunge_local->call_count = 1;
- afr_sh_entry_impunge_create (impunge_frame, this, child_index, buf,
- postparent);
+ ret = afr_build_parent_loc (&impunge_sh->parent_loc,
+ &impunge_local->loc, &op_errno);
+ if (ret) {
+ ret = -op_errno;
+ goto out;
+ }
+ impunge_local->call_count = enoent_count;
+ impunge_sh->entrybuf = sh->buf[sh->source];
+ impunge_sh->parentbuf = sh->parentbufs[sh->source];
+ for (i = 0; i < priv->child_count; i++) {
+ if (!impunge_local->child_up[i]) {
+ impunge_sh->child_errno[i] = ENOTCONN;
+ continue;
+ }
+ if (sh->child_errno[i] != ENOENT) {
+ impunge_sh->child_errno[i] = EEXIST;
+ continue;
+ }
+ }
+ for (i = 0; i < priv->child_count; i++) {
+ if (sh->child_errno[i] != ENOENT)
+ continue;
+ afr_sh_entry_impunge_create (impunge_frame, this, i);
+ enoent_count--;
+ }
+ GF_ASSERT (!enoent_count);
return;
out:
- gf_log (this->name, GF_LOG_ERROR, "impunge of %s failed, reason: %s",
- local->loc.path, strerror (-ret));
- impunge_done (frame, this, child_index, -1, -ret);
+ if (ret) {
+ gf_log (this->name, GF_LOG_ERROR, "impunge of %s failed, "
+ "reason: %s", local->loc.path, strerror (-ret));
+ afr_set_self_heal_status (sh, AFR_SELF_HEAL_FAILED);
+ }
+ afr_sh_missing_entries_finish (frame, this);
}
int
-afr_sh_create_entry_cbk (call_frame_t *frame, xlator_t *this, int child,
+afr_sh_create_entry_cbk (call_frame_t *frame, xlator_t *this,
int32_t op_ret, int32_t op_errno)
{
- int call_count = 0;
afr_local_t *local = NULL;
+ afr_self_heal_t *sh = NULL;
local = frame->local;
-
- if (op_ret == -1)
- gf_log (this->name, GF_LOG_ERROR,
- "create entry %s failed, on child %d reason, %s",
- local->loc.path, child, strerror (op_errno));
- call_count = afr_frame_return (frame);
- if (call_count == 0)
- afr_sh_missing_entries_finish (frame, this);
+ sh = &local->self_heal;
+ if (op_ret < 0)
+ afr_set_self_heal_status (sh, AFR_SELF_HEAL_FAILED);
+ afr_sh_missing_entries_finish (frame, this);
return 0;
}
@@ -1127,26 +1410,11 @@ sh_missing_entries_create (call_frame_t *frame, xlator_t *this)
afr_local_t *local = NULL;
afr_self_heal_t *sh = NULL;
int type = 0;
- afr_private_t *priv = NULL;
- int enoent_count = 0;
- int i = 0;
struct iatt *buf = NULL;
struct iatt *postparent = NULL;
local = frame->local;
sh = &local->self_heal;
- priv = this->private;
-
- enoent_count = afr_errno_count (NULL, sh->child_errno,
- priv->child_count, ENOENT);
- if (enoent_count == 0) {
- gf_log (this->name, GF_LOG_INFO,
- "no missing files - %s. proceeding to metadata check",
- local->loc.path);
- /* proceed to next step - metadata self-heal */
- afr_sh_missing_entries_finish (frame, this);
- return 0;
- }
buf = &sh->buf[sh->source];
postparent = &sh->parentbufs[sh->source];
@@ -1155,22 +1423,14 @@ sh_missing_entries_create (call_frame_t *frame, xlator_t *this)
if (!afr_valid_ia_type (type)) {
gf_log (this->name, GF_LOG_ERROR,
"%s: unknown file type: 0%o", local->loc.path, type);
- local->govinda_gOvinda = 1;
+ afr_set_local_for_unhealable (local);
afr_sh_missing_entries_finish (frame, this);
goto out;
}
- local->call_count = enoent_count;
- for (i = 0; i < priv->child_count; i++) {
- //If !child_up errno will be zero
- if (sh->child_errno[i] != ENOENT)
- continue;
- afr_sh_call_entry_impunge_recreate (frame, this, i,
+ afr_sh_missing_entry_call_impunge_recreate (frame, this,
buf, postparent,
afr_sh_create_entry_cbk);
- enoent_count--;
- }
- GF_ASSERT (enoent_count == 0);
out:
return 0;
}
@@ -1185,6 +1445,10 @@ afr_sh_missing_entries_lookup_done (call_frame_t *frame, xlator_t *this,
ia_type_t ia_type = IA_INVAL;
int32_t nsources = 0;
loc_t *loc = NULL;
+ int32_t subvol_status = 0;
+ afr_transaction_type txn_type = AFR_DATA_TRANSACTION;
+ gf_boolean_t split_brain = _gf_false;
+ int read_child = -1;
local = frame->local;
sh = &local->self_heal;
@@ -1192,25 +1456,47 @@ afr_sh_missing_entries_lookup_done (call_frame_t *frame, xlator_t *this,
loc = &local->loc;
if (op_ret < 0) {
- if (op_errno == EIO)
- local->govinda_gOvinda = 1;
+ if (op_errno == EIO) {
+ afr_set_local_for_unhealable (local);
+ }
// EIO can happen if finding the fresh parent dir failed
goto out;
}
//now No chance for the ia_type to conflict
ia_type = sh->buf[sh->success_children[0]].ia_type;
+ txn_type = afr_transaction_type_get (ia_type);
nsources = afr_build_sources (this, sh->xattr, sh->buf,
sh->pending_matrix, sh->sources,
- sh->success_children,
- afr_transaction_type_get (ia_type),
- NULL, _gf_false);
+ sh->success_children, txn_type,
+ &subvol_status, _gf_false);
if (nsources < 0) {
gf_log (this->name, GF_LOG_INFO, "No sources for dir of %s,"
" in missing entry self-heal, continuing with the rest"
" of the self-heals", local->loc.path);
- op_errno = EIO;
- goto out;
+ if (subvol_status & SPLIT_BRAIN) {
+ split_brain = _gf_true;
+ switch (txn_type) {
+ case AFR_DATA_TRANSACTION:
+ nsources = 1;
+ sh->sources[sh->success_children[0]] = 1;
+ break;
+ case AFR_ENTRY_TRANSACTION:
+ read_child = afr_get_no_xattr_dir_read_child
+ (this,
+ sh->success_children,
+ sh->buf);
+ sh->sources[read_child] = 1;
+ nsources = 1;
+ break;
+ default:
+ op_errno = EIO;
+ goto out;
+ }
+ } else {
+ op_errno = EIO;
+ goto out;
+ }
}
afr_get_fresh_children (sh->success_children, sh->sources,
@@ -1227,10 +1513,14 @@ afr_sh_missing_entries_lookup_done (call_frame_t *frame, xlator_t *this,
sh->type = sh->buf[sh->source].ia_type;
if (uuid_is_null (loc->inode->gfid))
uuid_copy (loc->gfid, sh->buf[sh->source].ia_gfid);
- sh_missing_entries_create (frame, this);
+ if (split_brain) {
+ afr_sh_missing_entries_finish (frame, this);
+ } else {
+ sh_missing_entries_create (frame, this);
+ }
return;
out:
- sh->op_failed = 1;
+ afr_set_self_heal_status (sh, AFR_SELF_HEAL_FAILED);
afr_sh_set_error (sh, op_errno);
afr_sh_missing_entries_finish (frame, this);
return;
@@ -1314,7 +1604,7 @@ afr_sh_remove_entry_cbk (call_frame_t *frame, xlator_t *this, int child,
LOCK (&frame->lock);
{
afr_sh_set_error (sh, EIO);
- sh->op_failed = 1;
+ afr_set_self_heal_status (sh, AFR_SELF_HEAL_FAILED);
}
UNLOCK (&frame->lock);
}
@@ -1336,13 +1626,14 @@ afr_sh_call_entry_expunge_remove (call_frame_t *frame, xlator_t *this,
afr_self_heal_t *sh = NULL;
afr_self_heal_t *expunge_sh = NULL;
int32_t op_errno = 0;
+ int ret = 0;
expunge_frame = copy_frame (frame);
if (!expunge_frame) {
goto out;
}
- ALLOC_OR_GOTO (expunge_local, afr_local_t, out);
+ AFR_LOCAL_ALLOC_OR_GOTO (expunge_local, out);
local = frame->local;
sh = &local->self_heal;
@@ -1350,6 +1641,12 @@ afr_sh_call_entry_expunge_remove (call_frame_t *frame, xlator_t *this,
expunge_sh = &expunge_local->self_heal;
expunge_sh->sh_frame = frame;
loc_copy (&expunge_local->loc, &local->loc);
+ ret = afr_build_parent_loc (&expunge_sh->parent_loc,
+ &expunge_local->loc, &op_errno);
+ if (ret) {
+ ret = -op_errno;
+ goto out;
+ }
sh->expunge_done = expunge_done;
afr_sh_entry_expunge_remove (expunge_frame, this, child_index, buf,
parentbuf);
@@ -1389,7 +1686,7 @@ afr_sh_purge_stale_entries_done (call_frame_t *frame, xlator_t *this)
sh = &local->self_heal;
priv = this->private;
- if (sh->op_failed) {
+ if (is_self_heal_failed (sh, AFR_CHECK_SPECIFIC)) {
afr_sh_missing_entries_finish (frame, this);
} else {
if (afr_gfid_missing_count (this->name, sh->fresh_children,
@@ -1399,7 +1696,8 @@ afr_sh_purge_stale_entries_done (call_frame_t *frame, xlator_t *this)
afr_sh_missing_entries_lookup_done,
sh->sh_gfid_req,
AFR_LOOKUP_FAIL_CONFLICTS|
- AFR_LOOKUP_FAIL_MISSING_GFIDS);
+ AFR_LOOKUP_FAIL_MISSING_GFIDS,
+ NULL);
} else {
//No need to set gfid so goto missing entries lookup done
//Behave as if you have done the lookup
@@ -1481,7 +1779,7 @@ afr_sh_purge_entry_common (call_frame_t *frame, xlator_t *this,
if (!purge_condition (local, priv, i))
continue;
gf_log (this->name, GF_LOG_INFO, "purging the stale entry %s "
- "on %d", local->loc.path, i);
+ "on %s", local->loc.path, priv->children[i]->name);
afr_sh_call_entry_expunge_remove (frame, this,
(long) i, &sh->buf[i],
&sh->parentbufs[i],
@@ -1601,10 +1899,8 @@ afr_sh_children_lookup_done (call_frame_t *frame, xlator_t *this,
sh->child_errno,
priv->child_count, ENOENT);
if (fresh_child_enoents == fresh_parent_count) {
- gf_log (this->name, GF_LOG_INFO, "Deleting stale file %s",
- local->loc.path);
afr_sh_set_error (sh, ENOENT);
- sh->op_failed = 1;
+ afr_set_self_heal_status (sh, AFR_SELF_HEAL_FAILED);
afr_sh_purge_entry (frame, this);
} else if (!afr_conflicting_iattrs (sh->buf, sh->fresh_children,
priv->child_count, local->loc.path,
@@ -1618,14 +1914,14 @@ afr_sh_children_lookup_done (call_frame_t *frame, xlator_t *this,
afr_sh_purge_stale_entry (frame, this);
} else {
op_errno = EIO;
- local->govinda_gOvinda = 1;
+ afr_set_local_for_unhealable (local);
goto fail;
}
return;
fail:
- sh->op_failed = 1;
+ afr_set_self_heal_status (sh, AFR_SELF_HEAL_FAILED);
afr_sh_set_error (sh, op_errno);
afr_sh_missing_entries_finish (frame, this);
return;
@@ -1647,25 +1943,17 @@ afr_sh_find_fresh_parents (call_frame_t *frame, xlator_t *this,
sh = &local->self_heal;
priv = this->private;
- /* If We can't find a fresh parent directory here,
- * we wont know which subvol is correct without finding a parent dir
- * upwards which has correct xattrs, for that we may have to
- * do lookups till root, we dont wanna do that,
- * instead make sure that if there are conflicting gfid
- * parent dirs, self-heal thus lookup is failed with EIO.
- * if there are missing entries we dont know whether to delete or
- * create so fail with EIO,
- * If there are conflicting xattr fail with EIO.
- */
if (op_ret < 0)
goto out;
enoent_count = afr_errno_count (NULL, sh->child_errno,
priv->child_count, ENOENT);
if (enoent_count > 0) {
- gf_log (this->name, GF_LOG_ERROR, "Parent dir missing for %s,"
- " in missing entry self-heal, aborting self-heal",
+ gf_log (this->name, GF_LOG_INFO, "Parent dir missing for %s,"
+ " in missing entry self-heal, aborting missing-entry "
+ "self-heal",
local->loc.path);
- goto out;
+ afr_sh_missing_entries_finish (frame, this);
+ return;
}
nsources = afr_build_sources (this, sh->xattr, sh->buf,
@@ -1684,6 +1972,7 @@ afr_sh_find_fresh_parents (call_frame_t *frame, xlator_t *this,
gf_log (this->name, GF_LOG_ERROR, "No sources for dir "
"of %s, in missing entry self-heal, aborting "
"self-heal", local->loc.path);
+ op_errno = EIO;
goto out;
}
@@ -1691,18 +1980,20 @@ afr_sh_find_fresh_parents (call_frame_t *frame, xlator_t *this,
if (source == -1) {
GF_ASSERT (0);
gf_log (this->name, GF_LOG_DEBUG, "No active sources found.");
+ op_errno = EIO;
goto out;
}
afr_get_fresh_children (sh->success_children, sh->sources,
sh->fresh_parent_dirs, priv->child_count);
afr_sh_common_lookup (frame, this, &local->loc,
- afr_sh_children_lookup_done, NULL, 0);
+ afr_sh_children_lookup_done, NULL, 0,
+ NULL);
return;
out:
- afr_sh_set_error (sh, EIO);
- sh->op_failed = 1;
- afr_sh_missing_entries_finish (frame, this);
+ afr_sh_set_error (sh, op_errno);
+ afr_set_self_heal_status (sh, AFR_SELF_HEAL_FAILED);
+ afr_sh_missing_entries_finish (frame, this);
return;
}
@@ -1730,7 +2021,7 @@ afr_sh_common_reset (afr_self_heal_t *sh, unsigned int child_count)
int
afr_sh_common_lookup (call_frame_t *frame, xlator_t *this, loc_t *loc,
afr_lookup_done_cbk_t lookup_done , uuid_t gfid,
- int32_t flags)
+ int32_t flags, dict_t *xdata)
{
afr_local_t *local = NULL;
int i = 0;
@@ -1791,7 +2082,8 @@ afr_sh_common_lookup (call_frame_t *frame, xlator_t *this, loc_t *loc,
int
-afr_sh_post_nb_entrylk_conflicting_sh_cbk (call_frame_t *frame, xlator_t *this)
+afr_sh_post_nb_entrylk_missing_entry_sh_cbk (call_frame_t *frame,
+ xlator_t *this)
{
afr_internal_lock_t *int_lock = NULL;
afr_local_t *local = NULL;
@@ -1804,7 +2096,7 @@ afr_sh_post_nb_entrylk_conflicting_sh_cbk (call_frame_t *frame, xlator_t *this)
if (int_lock->lock_op_ret < 0) {
gf_log (this->name, GF_LOG_INFO,
"Non blocking entrylks failed.");
- sh->op_failed = -1;
+ afr_set_self_heal_status (sh, AFR_SELF_HEAL_FAILED);
afr_sh_missing_entries_done (frame, this);
} else {
@@ -1812,34 +2104,8 @@ afr_sh_post_nb_entrylk_conflicting_sh_cbk (call_frame_t *frame, xlator_t *this)
"Non blocking entrylks done. Proceeding to FOP");
afr_sh_common_lookup (frame, this, &sh->parent_loc,
afr_sh_find_fresh_parents,
- NULL, AFR_LOOKUP_FAIL_CONFLICTS);
- }
-
- return 0;
-}
-
-int
-afr_sh_post_nb_entrylk_gfid_sh_cbk (call_frame_t *frame, xlator_t *this)
-{
- afr_internal_lock_t *int_lock = NULL;
- afr_local_t *local = NULL;
- afr_self_heal_t *sh = NULL;
-
- local = frame->local;
- sh = &local->self_heal;
- int_lock = &local->internal_lock;
-
- if (int_lock->lock_op_ret < 0) {
- gf_log (this->name, GF_LOG_INFO,
- "Non blocking entrylks failed.");
- afr_sh_missing_entries_done (frame, this);
- } else {
- gf_log (this->name, GF_LOG_DEBUG,
- "Non blocking entrylks done. Proceeding to FOP");
- afr_sh_common_lookup (frame, this, &local->loc,
- afr_sh_missing_entries_lookup_done,
- sh->sh_gfid_req, AFR_LOOKUP_FAIL_CONFLICTS|
- AFR_LOOKUP_FAIL_MISSING_GFIDS);
+ NULL, AFR_LOOKUP_FAIL_CONFLICTS,
+ NULL);
}
return 0;
@@ -1851,7 +2117,9 @@ afr_sh_entrylk (call_frame_t *frame, xlator_t *this, loc_t *loc,
{
afr_internal_lock_t *int_lock = NULL;
afr_local_t *local = NULL;
+ afr_private_t *priv = NULL;
+ priv = this->private;
local = frame->local;
int_lock = &local->internal_lock;
@@ -1863,7 +2131,12 @@ afr_sh_entrylk (call_frame_t *frame, xlator_t *this, loc_t *loc,
int_lock->lk_basename = base_name;
int_lock->lk_loc = loc;
int_lock->lock_cbk = lock_cbk;
+ int_lock->domain = this->name;
+ int_lock->lockee_count = 0;
+ afr_init_entry_lockee (&int_lock->lockee[0], local, loc,
+ base_name, priv->child_count);
+ int_lock->lockee_count++;
afr_nonblocking_entrylk (frame, this);
return 0;
@@ -1875,6 +2148,9 @@ afr_self_heal_parent_entrylk (call_frame_t *frame, xlator_t *this,
{
afr_local_t *local = NULL;
afr_self_heal_t *sh = NULL;
+ afr_internal_lock_t *int_lock = NULL;
+ int ret = -1;
+ int32_t op_errno = 0;
local = frame->local;
sh = &local->self_heal;
@@ -1883,43 +2159,52 @@ afr_self_heal_parent_entrylk (call_frame_t *frame, xlator_t *this,
"attempting to recreate missing entries for path=%s",
local->loc.path);
- GF_ASSERT (local->loc.parent);
- afr_build_parent_loc (&sh->parent_loc, &local->loc);
+ ret = afr_build_parent_loc (&sh->parent_loc, &local->loc, &op_errno);
+ if (ret)
+ goto out;
afr_sh_entrylk (frame, this, &sh->parent_loc, NULL,
lock_cbk);
return 0;
-}
-
-static int
-afr_self_heal_conflicting_entries (call_frame_t *frame, xlator_t *this)
-{
- afr_self_heal_parent_entrylk (frame, this,
- afr_sh_post_nb_entrylk_conflicting_sh_cbk);
+out:
+ int_lock = &local->internal_lock;
+ int_lock->lock_op_ret = -1;
+ lock_cbk (frame, this);
return 0;
}
static int
-afr_self_heal_gfids (call_frame_t *frame, xlator_t *this)
+afr_self_heal_missing_entries (call_frame_t *frame, xlator_t *this)
{
+ afr_local_t *local = NULL;
+ afr_self_heal_t *sh = NULL;
+
+ local = frame->local;
+ sh = &local->self_heal;
+
+ sh->sh_type_in_action = AFR_SELF_HEAL_GFID_OR_MISSING_ENTRY;
+
+ afr_set_self_heal_status (sh, AFR_SELF_HEAL_STARTED);
+
afr_self_heal_parent_entrylk (frame, this,
- afr_sh_post_nb_entrylk_gfid_sh_cbk);
+ afr_sh_post_nb_entrylk_missing_entry_sh_cbk);
return 0;
}
-afr_local_t *afr_local_copy (afr_local_t *l, xlator_t *this)
+afr_local_t*
+afr_self_heal_local_init (afr_local_t *l, xlator_t *this)
{
- afr_private_t *priv = NULL;
- afr_local_t *lc = NULL;
- afr_self_heal_t *sh = NULL;
- afr_self_heal_t *shc = NULL;
+ afr_private_t *priv = NULL;
+ afr_local_t *lc = NULL;
+ afr_self_heal_t *sh = NULL;
+ afr_self_heal_t *shc = NULL;
+ int ret = 0;
priv = this->private;
sh = &l->self_heal;
- lc = GF_CALLOC (1, sizeof (afr_local_t),
- gf_afr_mt_afr_local_t);
+ lc = mem_get0 (this->local_pool);
if (!lc)
goto out;
@@ -1932,16 +2217,27 @@ afr_local_t *afr_local_copy (afr_local_t *l, xlator_t *this)
shc->do_data_self_heal = sh->do_data_self_heal;
shc->do_metadata_self_heal = sh->do_metadata_self_heal;
shc->do_entry_self_heal = sh->do_entry_self_heal;
+ shc->force_confirm_spb = sh->force_confirm_spb;
shc->forced_merge = sh->forced_merge;
shc->background = sh->background;
shc->type = sh->type;
+ shc->data_sh_info = "";
+ shc->metadata_sh_info = "";
uuid_copy (shc->sh_gfid_req, sh->sh_gfid_req);
- if (l->loc.path)
- loc_copy (&lc->loc, &l->loc);
+ if (l->loc.path) {
+ ret = loc_copy (&lc->loc, &l->loc);
+ if (ret < 0)
+ goto out;
+ }
lc->child_up = memdup (l->child_up,
sizeof (*lc->child_up) * priv->child_count);
+ if (!lc->child_up) {
+ ret = -1;
+ goto out;
+ }
+
if (l->xattr_req)
lc->xattr_req = dict_ref (l->xattr_req);
@@ -1949,40 +2245,25 @@ afr_local_t *afr_local_copy (afr_local_t *l, xlator_t *this)
lc->cont.lookup.inode = inode_ref (l->cont.lookup.inode);
if (l->cont.lookup.xattr)
lc->cont.lookup.xattr = dict_ref (l->cont.lookup.xattr);
- if (l->internal_lock.inode_locked_nodes)
- lc->internal_lock.inode_locked_nodes =
- memdup (l->internal_lock.inode_locked_nodes,
- sizeof (*lc->internal_lock.inode_locked_nodes) * priv->child_count);
- else
- lc->internal_lock.inode_locked_nodes =
- GF_CALLOC (sizeof (*l->internal_lock.inode_locked_nodes),
- priv->child_count,
- gf_afr_mt_char);
- if (l->internal_lock.entry_locked_nodes)
- lc->internal_lock.entry_locked_nodes =
- memdup (l->internal_lock.entry_locked_nodes,
- sizeof (*lc->internal_lock.entry_locked_nodes) * priv->child_count);
- else
- lc->internal_lock.entry_locked_nodes =
- GF_CALLOC (sizeof (*l->internal_lock.entry_locked_nodes),
- priv->child_count,
- gf_afr_mt_char);
- if (l->internal_lock.locked_nodes)
- lc->internal_lock.locked_nodes =
- memdup (l->internal_lock.locked_nodes,
- sizeof (*lc->internal_lock.locked_nodes) * priv->child_count);
- else
- lc->internal_lock.locked_nodes =
- GF_CALLOC (sizeof (*l->internal_lock.locked_nodes),
- priv->child_count,
- gf_afr_mt_char);
- lc->internal_lock.inodelk_lock_count =
- l->internal_lock.inodelk_lock_count;
- lc->internal_lock.entrylk_lock_count =
- l->internal_lock.entrylk_lock_count;
+ lc->internal_lock.locked_nodes =
+ GF_CALLOC (sizeof (*l->internal_lock.locked_nodes),
+ priv->child_count, gf_afr_mt_char);
+ if (!lc->internal_lock.locked_nodes) {
+ ret = -1;
+ goto out;
+ }
+
+ ret = afr_inodelk_init (&lc->internal_lock.inodelk[0],
+ this->name, priv->child_count);
+ if (ret)
+ goto out;
out:
+ if (ret) {
+ afr_local_cleanup (lc, this);
+ lc = NULL;
+ }
return lc;
}
@@ -1992,32 +2273,39 @@ afr_self_heal_completion_cbk (call_frame_t *bgsh_frame, xlator_t *this)
afr_private_t * priv = NULL;
afr_local_t * local = NULL;
afr_self_heal_t * sh = NULL;
+ afr_local_t * orig_frame_local = NULL;
+ afr_self_heal_t * orig_frame_sh = NULL;
char sh_type_str[256] = {0,};
- gf_boolean_t split_brain = _gf_false;
+ gf_loglevel_t loglevel = 0;
priv = this->private;
local = bgsh_frame->local;
sh = &local->self_heal;
- if (local->govinda_gOvinda)
- split_brain = _gf_true;
-
- afr_set_split_brain (this, sh->inode, split_brain);
+ if (local->unhealable) {
+ afr_set_split_brain (this, sh->inode, SPB, SPB);
+ }
afr_self_heal_type_str_get (sh, sh_type_str,
sizeof(sh_type_str));
- if (sh->op_failed) {
- gf_log (this->name, GF_LOG_ERROR, "background %s self-heal "
- "failed on %s", sh_type_str, local->loc.path);
+ if (is_self_heal_failed (sh, AFR_CHECK_ALL) && !priv->shd.iamshd) {
+ loglevel = GF_LOG_ERROR;
+ } else if (!is_self_heal_failed (sh, AFR_CHECK_ALL)) {
+ loglevel = GF_LOG_INFO;
} else {
- gf_log (this->name, GF_LOG_INFO, "background %s self-heal "
- "completed on %s", sh_type_str, local->loc.path);
+ loglevel = GF_LOG_DEBUG;
}
+ afr_log_self_heal_completion_status (local, loglevel);
+
FRAME_SU_UNDO (bgsh_frame, afr_local_t);
if (!sh->unwound && sh->unwind) {
- sh->unwind (sh->orig_frame, this, sh->op_ret, sh->op_errno);
+ orig_frame_local = sh->orig_frame->local;
+ orig_frame_sh = &orig_frame_local->self_heal;
+ orig_frame_sh->actual_sh_started = _gf_true;
+ sh->unwind (sh->orig_frame, this, sh->op_ret, sh->op_errno,
+ is_self_heal_failed (sh, AFR_CHECK_ALL));
}
if (sh->background) {
@@ -2039,7 +2327,6 @@ afr_self_heal (call_frame_t *frame, xlator_t *this, inode_t *inode)
afr_local_t *local = NULL;
afr_self_heal_t *sh = NULL;
afr_private_t *priv = NULL;
- int i = 0;
int32_t op_errno = 0;
int ret = 0;
afr_self_heal_t *orig_sh = NULL;
@@ -2060,21 +2347,20 @@ afr_self_heal (call_frame_t *frame, xlator_t *this, inode_t *inode)
local->self_heal.do_data_self_heal,
local->self_heal.do_entry_self_heal);
- op_errno = ENOMEM;
+ op_errno = ENOMEM;
sh_frame = copy_frame (frame);
if (!sh_frame)
goto out;
- afr_set_lk_owner (sh_frame, this);
+ afr_set_lk_owner (sh_frame, this, sh_frame->root);
afr_set_low_priority (sh_frame);
- sh_local = afr_local_copy (local, this);
+ sh_local = afr_self_heal_local_init (local, this);
if (!sh_local)
goto out;
sh_frame->local = sh_local;
sh = &sh_local->self_heal;
sh->inode = inode_ref (inode);
-
sh->orig_frame = frame;
sh->completion_cbk = afr_self_heal_completion_cbk;
@@ -2093,30 +2379,16 @@ afr_self_heal (call_frame_t *frame, xlator_t *this, inode_t *inode)
if (!sh->locked_nodes)
goto out;
- sh->pending_matrix = GF_CALLOC (sizeof (int32_t *), priv->child_count,
- gf_afr_mt_int32_t);
+ sh->pending_matrix = afr_matrix_create (priv->child_count,
+ priv->child_count);
if (!sh->pending_matrix)
goto out;
- for (i = 0; i < priv->child_count; i++) {
- sh->pending_matrix[i] = GF_CALLOC (sizeof (int32_t),
- priv->child_count,
- gf_afr_mt_int32_t);
- if (!sh->pending_matrix[i])
- goto out;
- }
-
- sh->delta_matrix = GF_CALLOC (sizeof (int32_t *), priv->child_count,
- gf_afr_mt_int32_t);
+ sh->delta_matrix = afr_matrix_create (priv->child_count,
+ priv->child_count);
if (!sh->delta_matrix)
goto out;
- for (i = 0; i < priv->child_count; i++) {
- sh->delta_matrix[i] = GF_CALLOC (sizeof (int32_t),
- priv->child_count,
- gf_afr_mt_int32_t);
- if (!sh->delta_matrix)
- goto out;
- }
+
sh->fresh_parent_dirs = afr_children_create (priv->child_count);
if (!sh->fresh_parent_dirs)
goto out;
@@ -2142,18 +2414,16 @@ afr_self_heal (call_frame_t *frame, xlator_t *this, inode_t *inode)
UNLOCK (&priv->lock);
}
- if (!local->loc.name) {
- /* nameless lookup */
+ if (!local->loc.parent) {
sh->do_missing_entry_self_heal = _gf_false;
sh->do_gfid_self_heal = _gf_false;
}
+ sh->sh_type_in_action = AFR_SELF_HEAL_INVALID;
+
FRAME_SU_DO (sh_frame, afr_local_t);
- if (sh->do_missing_entry_self_heal) {
- afr_self_heal_conflicting_entries (sh_frame, this);
- } else if (sh->do_gfid_self_heal) {
- GF_ASSERT (!uuid_is_null (sh->sh_gfid_req));
- afr_self_heal_gfids (sh_frame, this);
+ if (sh->do_missing_entry_self_heal || sh->do_gfid_self_heal) {
+ afr_self_heal_missing_entries (sh_frame, this);
} else {
loc = &sh_local->loc;
if (uuid_is_null (loc->inode->gfid) && uuid_is_null (loc->gfid)) {
@@ -2172,7 +2442,9 @@ afr_self_heal (call_frame_t *frame, xlator_t *this, inode_t *inode)
out:
if (op_errno) {
- orig_sh->unwind (frame, this, -1, op_errno);
+ orig_sh->unwind (frame, this, -1, op_errno, 1);
+ if (sh_frame)
+ AFR_STACK_DESTROY (sh_frame);
}
return 0;
}
@@ -2229,13 +2501,21 @@ afr_self_heal_type_for_transaction (afr_transaction_type type)
}
int
-afr_build_child_loc (xlator_t *this, loc_t *child, loc_t *parent, char *name, uuid_t gfid)
+afr_build_child_loc (xlator_t *this, loc_t *child, loc_t *parent, char *name)
{
int ret = -1;
+ uuid_t pargfid = {0};
- if (!child) {
+ if (!child)
+ goto out;
+
+ if (!uuid_is_null (parent->inode->gfid))
+ uuid_copy (pargfid, parent->inode->gfid);
+ else if (!uuid_is_null (parent->gfid))
+ uuid_copy (pargfid, parent->gfid);
+
+ if (uuid_is_null (pargfid))
goto out;
- }
if (strcmp (parent->path, "/") == 0)
ret = gf_asprintf ((char **)&child->path, "/%s", name);
@@ -2248,27 +2528,285 @@ afr_build_child_loc (xlator_t *this, loc_t *child, loc_t *parent, char *name, uu
"asprintf failed while setting child path");
}
- if (!child->path) {
- goto out;
- }
-
child->name = strrchr (child->path, '/');
if (child->name)
child->name++;
child->parent = inode_ref (parent->inode);
child->inode = inode_new (parent->inode->table);
+ uuid_copy (child->pargfid, pargfid);
if (!child->inode) {
ret = -1;
goto out;
}
- uuid_copy (child->gfid, gfid);
ret = 0;
out:
- if (ret == -1)
+ if ((ret == -1) && child)
loc_wipe (child);
return ret;
}
+
+int
+afr_sh_erase_pending (call_frame_t *frame, xlator_t *this,
+ afr_transaction_type type, afr_fxattrop_cbk_t cbk,
+ int (*finish)(call_frame_t *frame, xlator_t *this))
+{
+ afr_local_t *local = NULL;
+ afr_self_heal_t *sh = NULL;
+ afr_private_t *priv = NULL;
+ int call_count = 0;
+ int i = 0;
+ dict_t **erase_xattr = NULL;
+ int ret = -1;
+
+ local = frame->local;
+ sh = &local->self_heal;
+ priv = this->private;
+
+ afr_sh_pending_to_delta (priv, sh->xattr, sh->delta_matrix,
+ sh->success, priv->child_count, type);
+
+ erase_xattr = GF_CALLOC (sizeof (*erase_xattr), priv->child_count,
+ gf_afr_mt_dict_t);
+ if (!erase_xattr)
+ goto out;
+
+ for (i = 0; i < priv->child_count; i++) {
+ if (sh->xattr[i]) {
+ call_count++;
+ erase_xattr[i] = dict_new ();
+ if (!erase_xattr[i])
+ goto out;
+ }
+ }
+
+ afr_sh_delta_to_xattr (this, sh->delta_matrix, erase_xattr,
+ priv->child_count, type);
+
+ gf_log (this->name, GF_LOG_DEBUG, "Delta matrix for: %s",
+ lkowner_utoa (&frame->root->lk_owner));
+ afr_sh_print_pending_matrix (sh->delta_matrix, this);
+ local->call_count = call_count;
+ if (call_count == 0) {
+ ret = 0;
+ finish (frame, this);
+ goto out;
+ }
+
+ for (i = 0; i < priv->child_count; i++) {
+ if (!erase_xattr[i])
+ continue;
+
+ if (sh->healing_fd) {//true for ENTRY, reg file DATA transaction
+ STACK_WIND_COOKIE (frame, cbk, (void *) (long) i,
+ priv->children[i],
+ priv->children[i]->fops->fxattrop,
+ sh->healing_fd,
+ GF_XATTROP_ADD_ARRAY, erase_xattr[i],
+ NULL);
+ } else {
+ STACK_WIND_COOKIE (frame, cbk, (void *) (long) i,
+ priv->children[i],
+ priv->children[i]->fops->xattrop,
+ &local->loc,
+ GF_XATTROP_ADD_ARRAY, erase_xattr[i],
+ NULL);
+ }
+ }
+
+ ret = 0;
+out:
+ if (erase_xattr) {
+ for (i = 0; i < priv->child_count; i++) {
+ if (erase_xattr[i]) {
+ dict_unref (erase_xattr[i]);
+ }
+ }
+ }
+
+ GF_FREE (erase_xattr);
+
+ if (ret < 0) {
+ afr_set_self_heal_status (sh, AFR_SELF_HEAL_FAILED);
+ finish (frame, this);
+ }
+
+ return 0;
+}
+
+void
+afr_set_self_heal_status(afr_self_heal_t *sh, afr_self_heal_status status)
+{
+ xlator_t *this = NULL;
+ afr_sh_status_for_all_type *sh_status = &(sh->afr_all_sh_status);
+ afr_self_heal_type sh_type_in_action = sh->sh_type_in_action;
+ this = THIS;
+
+ if (!sh) {
+ gf_log_callingfn (this->name, GF_LOG_ERROR, "Null self heal"
+ "Structure");
+ goto out;
+ }
+
+ switch (sh_type_in_action) {
+ case AFR_SELF_HEAL_GFID_OR_MISSING_ENTRY:
+ sh_status->gfid_or_missing_entry_self_heal = status;
+ break;
+ case AFR_SELF_HEAL_METADATA:
+ sh_status->metadata_self_heal = status;
+ break;
+ case AFR_SELF_HEAL_DATA:
+ sh_status->data_self_heal = status;
+ break;
+ case AFR_SELF_HEAL_ENTRY:
+ sh_status->entry_self_heal = status;
+ break;
+ case AFR_SELF_HEAL_INVALID:
+ gf_log_callingfn (this->name, GF_LOG_ERROR, "Invalid"
+ "self heal type in action");
+ break;
+ }
+out:
+ return;
+}
+
+void
+afr_set_local_for_unhealable (afr_local_t *local)
+{
+ afr_self_heal_t *sh = NULL;
+
+ sh = &local->self_heal;
+
+ local->unhealable = 1;
+ afr_set_self_heal_status (sh, AFR_SELF_HEAL_FAILED);
+}
+
+int
+is_self_heal_failed (afr_self_heal_t *sh, afr_sh_fail_check_type type)
+{
+ afr_sh_status_for_all_type sh_status = sh->afr_all_sh_status;
+ afr_self_heal_type sh_type_in_action = AFR_SELF_HEAL_INVALID;
+ afr_self_heal_status status = AFR_SELF_HEAL_FAILED;
+ xlator_t *this = NULL;
+ int sh_failed = 0;
+
+ this = THIS;
+
+ if (!sh) {
+ gf_log_callingfn (this->name, GF_LOG_ERROR, "Null self heal "
+ "structure");
+ sh_failed = 1;
+ goto out;
+ }
+
+ if (type == AFR_CHECK_ALL) {
+ if ((sh_status.gfid_or_missing_entry_self_heal == AFR_SELF_HEAL_FAILED)
+ || (sh_status.metadata_self_heal == AFR_SELF_HEAL_FAILED)
+ || (sh_status.data_self_heal == AFR_SELF_HEAL_FAILED)
+ || (sh_status.entry_self_heal == AFR_SELF_HEAL_FAILED))
+ sh_failed = 1;
+ } else if (type == AFR_CHECK_SPECIFIC) {
+ sh_type_in_action = sh->sh_type_in_action;
+ switch (sh_type_in_action) {
+ case AFR_SELF_HEAL_GFID_OR_MISSING_ENTRY:
+ status = sh_status.gfid_or_missing_entry_self_heal;
+ break;
+ case AFR_SELF_HEAL_METADATA:
+ status = sh_status.metadata_self_heal;
+ break;
+ case AFR_SELF_HEAL_ENTRY:
+ status = sh_status.entry_self_heal;
+ break;
+ case AFR_SELF_HEAL_DATA:
+ status = sh_status.data_self_heal;
+ break;
+ case AFR_SELF_HEAL_INVALID:
+ status = AFR_SELF_HEAL_NOT_ATTEMPTED;
+ break;
+ }
+ if (status == AFR_SELF_HEAL_FAILED)
+ sh_failed = 1;
+
+ }
+
+out:
+ return sh_failed;
+}
+
+char *
+get_sh_completion_status (afr_self_heal_status status)
+{
+
+ char *not_attempted = " is not attempted";
+ char *failed = " failed";
+ char *started = " is started";
+ char *sync_begin = " is successfully completed";
+ char *result = " has unknown status";
+
+ switch (status)
+ {
+ case AFR_SELF_HEAL_NOT_ATTEMPTED:
+ result = not_attempted;
+ break;
+ case AFR_SELF_HEAL_FAILED:
+ result = failed;
+ break;
+ case AFR_SELF_HEAL_STARTED:
+ result = started;
+ break;
+ case AFR_SELF_HEAL_SYNC_BEGIN:
+ result = sync_begin;
+ break;
+ }
+
+ return result;
+
+}
+
+void
+afr_log_self_heal_completion_status (afr_local_t *local, gf_loglevel_t loglvl)
+{
+
+ char sh_log[4096] = {0};
+ afr_self_heal_t *sh = &local->self_heal;
+ afr_sh_status_for_all_type all_status = sh->afr_all_sh_status;
+ xlator_t *this = NULL;
+ size_t off = 0;
+ int data_sh = 0;
+ int metadata_sh = 0;
+ int print_log = 0;
+
+ this = THIS;
+
+ ADD_FMT_STRING (sh_log, off, "gfid or missing entry",
+ all_status.gfid_or_missing_entry_self_heal, print_log);
+ ADD_FMT_STRING_SYNC (sh_log, off, "metadata",
+ all_status.metadata_self_heal, print_log);
+ if (sh->background) {
+ ADD_FMT_STRING_SYNC (sh_log, off, "backgroung data",
+ all_status.data_self_heal, print_log);
+ } else {
+ ADD_FMT_STRING_SYNC (sh_log, off, "foreground data",
+ all_status.data_self_heal, print_log);
+ }
+ ADD_FMT_STRING_SYNC (sh_log, off, "entry", all_status.entry_self_heal,
+ print_log);
+
+ if (AFR_SELF_HEAL_SYNC_BEGIN == all_status.data_self_heal &&
+ strcmp (sh->data_sh_info, "") && sh->data_sh_info )
+ data_sh = 1;
+ if (AFR_SELF_HEAL_SYNC_BEGIN == all_status.metadata_self_heal &&
+ strcmp (sh->metadata_sh_info, "") && sh->metadata_sh_info)
+ metadata_sh = 1;
+
+ if (!print_log)
+ return;
+
+ gf_log (this->name, loglvl, "%s %s %s on %s", sh_log,
+ ((data_sh == 1) ? sh->data_sh_info : ""),
+ ((metadata_sh == 1) ? sh->metadata_sh_info : ""),
+ local->loc.path);
+}
diff --git a/xlators/cluster/afr/src/afr-self-heal-common.h b/xlators/cluster/afr/src/afr-self-heal-common.h
index b7a736a74..473264776 100644
--- a/xlators/cluster/afr/src/afr-self-heal-common.h
+++ b/xlators/cluster/afr/src/afr-self-heal-common.h
@@ -1,33 +1,18 @@
/*
- Copyright (c) 2008-2011 Gluster, Inc. <http://www.gluster.com>
+ Copyright (c) 2008-2012 Red Hat, Inc. <http://www.redhat.com>
This file is part of GlusterFS.
- GlusterFS is free software; you can redistribute it and/or modify
- it under the terms of the GNU General Public License as published
- by the Free Software Foundation; either version 3 of the License,
- or (at your option) any later version.
-
- GlusterFS is distributed in the hope that it will be useful, but
- WITHOUT ANY WARRANTY; without even the implied warranty of
- MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU
- General Public License for more details.
-
- You should have received a copy of the GNU General Public License
- along with this program. If not, see
- <http://www.gnu.org/licenses/>.
+ This file is licensed to you under your choice of the GNU Lesser
+ General Public License, version 3 or any later version (LGPLv3 or
+ later), or the GNU General Public License, version 2 (GPLv2), in all
+ cases as published by the Free Software Foundation.
*/
#ifndef __AFR_SELF_HEAL_COMMON_H__
#define __AFR_SELF_HEAL_COMMON_H__
#define FILE_HAS_HOLES(buf) (((buf)->ia_size) > ((buf)->ia_blocks * 512))
-
-typedef enum {
- AFR_SELF_HEAL_ENTRY,
- AFR_SELF_HEAL_METADATA,
- AFR_SELF_HEAL_DATA,
- AFR_SELF_HEAL_INVALID = -1,
-} afr_self_heal_type;
+#define AFR_SH_MIN_PARTICIPANTS 2
typedef enum {
AFR_LOOKUP_FAIL_CONFLICTS = 1,
@@ -43,6 +28,10 @@ afr_sh_source_count (int sources[], int child_count);
void
afr_sh_print_pending_matrix (int32_t *pending_matrix[], xlator_t *this);
+void
+afr_sh_print_split_brain_log (int32_t *pending_matrix[], xlator_t *this,
+ const char *loc);
+
int
afr_build_pending_matrix (char **pending_key, int32_t **pending_matrix,
unsigned char *ignorant_subvols,
@@ -60,7 +49,7 @@ afr_mark_sources (xlator_t *this, int32_t *sources, int32_t **pending_matrix,
int32_t *success_children, int32_t *subvol_status);
int
-afr_sh_delta_to_xattr (afr_private_t *priv,
+afr_sh_delta_to_xattr (xlator_t *this,
int32_t *delta_matrix[], dict_t *xattr[],
int child_count, afr_transaction_type type);
@@ -90,7 +79,7 @@ afr_sh_common_lookup_resp_handler (call_frame_t *frame, void *cookie,
int
afr_sh_common_lookup (call_frame_t *frame, xlator_t *this, loc_t *loc,
afr_lookup_done_cbk_t lookup_cbk, uuid_t uuid,
- int32_t flags);
+ int32_t flags, dict_t *xdata);
int
afr_sh_entry_expunge_remove (call_frame_t *expunge_frame, xlator_t *this,
int active_src, struct iatt *buf,
@@ -100,16 +89,15 @@ afr_sh_entrylk (call_frame_t *frame, xlator_t *this, loc_t *loc,
char *base_name, afr_lock_cbk_t lock_cbk);
int
afr_sh_entry_impunge_create (call_frame_t *impunge_frame, xlator_t *this,
- int child_index, struct iatt *buf,
- struct iatt *postparent);
+ int child_index);
int
-afr_sh_data_unlock (call_frame_t *frame, xlator_t *this,
+afr_sh_data_unlock (call_frame_t *frame, xlator_t *this, char *dom,
afr_lock_cbk_t lock_cbk);
afr_local_t *
-afr_local_copy (afr_local_t *l, xlator_t *this);
+afr_self_heal_local_init (afr_local_t *l, xlator_t *this);
int
afr_sh_data_lock (call_frame_t *frame, xlator_t *this,
- off_t start, off_t len,
+ off_t start, off_t len, gf_boolean_t block, char *dom,
afr_lock_cbk_t success_handler,
afr_lock_cbk_t failure_handler);
void
@@ -119,12 +107,38 @@ afr_sh_mark_source_sinks (call_frame_t *frame, xlator_t *this);
typedef int
(*afr_fxattrop_cbk_t) (call_frame_t *frame, void *cookie,
xlator_t *this, int32_t op_ret, int32_t op_errno,
- dict_t *xattr);
+ dict_t *xattr, dict_t *xdata);
int
-afr_build_child_loc (xlator_t *this, loc_t *child, loc_t *parent, char *name,
- uuid_t gfid);
+afr_build_child_loc (xlator_t *this, loc_t *child, loc_t *parent, char *name);
int
afr_impunge_frame_create (call_frame_t *frame, xlator_t *this,
- int active_source, int ret_child, mode_t entry_mode,
- call_frame_t **impunge_frame);
+ int active_source, call_frame_t **impunge_frame);
+void
+afr_sh_reset (call_frame_t *frame, xlator_t *this);
+
+void
+afr_children_intersection_get (int32_t *set1, int32_t *set2,
+ int *intersection, unsigned int child_count);
+int
+afr_get_no_xattr_dir_read_child (xlator_t *this, int32_t *success_children,
+ struct iatt *bufs);
+int
+afr_sh_erase_pending (call_frame_t *frame, xlator_t *this,
+ afr_transaction_type type, afr_fxattrop_cbk_t cbk,
+ int (*finish)(call_frame_t *frame, xlator_t *this));
+
+void
+afr_set_local_for_unhealable (afr_local_t *local);
+
+int
+is_self_heal_failed (afr_self_heal_t *sh, afr_sh_fail_check_type type);
+
+void
+afr_set_self_heal_status (afr_self_heal_t *sh, afr_self_heal_status status);
+
+void
+afr_log_self_heal_completion_status (afr_local_t *local, gf_loglevel_t logl);
+
+char*
+afr_get_pending_matrix_str (int32_t *pending_matrix[], xlator_t *this);
#endif /* __AFR_SELF_HEAL_COMMON_H__ */
diff --git a/xlators/cluster/afr/src/afr-self-heal-data.c b/xlators/cluster/afr/src/afr-self-heal-data.c
index c1c1d483e..9de26ee56 100644
--- a/xlators/cluster/afr/src/afr-self-heal-data.c
+++ b/xlators/cluster/afr/src/afr-self-heal-data.c
@@ -1,20 +1,11 @@
/*
- Copyright (c) 2008-2011 Gluster, Inc. <http://www.gluster.com>
+ Copyright (c) 2008-2012 Red Hat, Inc. <http://www.redhat.com>
This file is part of GlusterFS.
- GlusterFS is free software; you can redistribute it and/or modify
- it under the terms of the GNU General Public License as published
- by the Free Software Foundation; either version 3 of the License,
- or (at your option) any later version.
-
- GlusterFS is distributed in the hope that it will be useful, but
- WITHOUT ANY WARRANTY; without even the implied warranty of
- MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU
- General Public License for more details.
-
- You should have received a copy of the GNU General Public License
- along with this program. If not, see
- <http://www.gnu.org/licenses/>.
+ This file is licensed to you under your choice of the GNU Lesser
+ General Public License, version 3 or any later version (LGPLv3 or
+ later), or the GNU General Public License, version 2 (GPLv2), in all
+ cases as published by the Free Software Foundation.
*/
#include <libgen.h>
@@ -49,6 +40,14 @@
#include "afr-self-heal-common.h"
#include "afr-self-heal-algorithm.h"
+int
+afr_sh_data_fail (call_frame_t *frame, xlator_t *this);
+
+static inline gf_boolean_t
+afr_sh_data_proceed (unsigned int success_count)
+{
+ return (success_count >= AFR_SH_MIN_PARTICIPANTS);
+}
extern int
sh_loop_finish (call_frame_t *loop_frame, xlator_t *this);
@@ -63,15 +62,6 @@ int
afr_sh_data_finish (call_frame_t *frame, xlator_t *this);
int
-afr_sh_data_fxattrop (call_frame_t *frame, xlator_t *this,
- afr_fxattrop_cbk_t fxattrop_cbk);
-
-int
-afr_post_sh_data_fxattrop_cbk (call_frame_t *frame, void *cookie,
- xlator_t *this, int32_t op_ret, int32_t op_errno,
- dict_t *xattr);
-
-int
afr_sh_data_done (call_frame_t *frame, xlator_t *this)
{
afr_local_t *local = NULL;
@@ -88,7 +78,7 @@ afr_sh_data_done (call_frame_t *frame, xlator_t *this)
int
afr_sh_data_flush_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
- int32_t op_ret, int32_t op_errno)
+ int32_t op_ret, int32_t op_errno, dict_t *xdata)
{
afr_local_t *local = NULL;
afr_private_t *priv = NULL;
@@ -101,7 +91,7 @@ afr_sh_data_flush_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
LOCK (&frame->lock);
{
if (op_ret == -1) {
- gf_log (this->name, GF_LOG_INFO,
+ gf_log (this->name, GF_LOG_ERROR,
"flush failed on %s on subvolume %s: %s",
local->loc.path, priv->children[child_index]->name,
strerror (op_errno));
@@ -131,6 +121,11 @@ afr_sh_data_close (call_frame_t *frame, xlator_t *this)
sh = &local->self_heal;
priv = this->private;
+ if (!sh->healing_fd) {
+ //This happens when file is non-reg
+ afr_sh_data_done (frame, this);
+ return 0;
+ }
call_count = afr_set_elem_count_get (sh->success,
priv->child_count);
local->call_count = call_count;
@@ -151,7 +146,7 @@ afr_sh_data_close (call_frame_t *frame, xlator_t *this)
(void *) (long) i,
priv->children[i],
priv->children[i]->fops->flush,
- sh->healing_fd);
+ sh->healing_fd, NULL);
if (!--call_count)
break;
@@ -161,9 +156,28 @@ afr_sh_data_close (call_frame_t *frame, xlator_t *this)
}
int
+afr_sh_dom_unlock (call_frame_t *frame, xlator_t *this)
+{
+ afr_local_t *local = NULL;
+ afr_self_heal_t *sh = NULL;
+ afr_private_t *priv = NULL;
+
+ local = frame->local;
+ sh = &local->self_heal;
+ priv = this->private;
+
+ if (sh->sh_dom_lock_held)
+ afr_sh_data_unlock (frame, this, priv->sh_domain,
+ afr_sh_data_close);
+ else
+ afr_sh_data_close (frame, this);
+ return 0;
+}
+
+int
afr_sh_data_setattr_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
int32_t op_ret, int32_t op_errno, struct iatt *statpre,
- struct iatt *statpost)
+ struct iatt *statpost, dict_t *xdata)
{
afr_local_t *local = NULL;
@@ -195,29 +209,20 @@ afr_sh_data_setattr_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
}
int
-afr_sh_data_setattr (call_frame_t *frame, xlator_t *this)
+afr_sh_data_setattr (call_frame_t *frame, xlator_t *this, struct iatt* stbuf)
{
afr_local_t *local = NULL;
afr_private_t *priv = NULL;
afr_self_heal_t *sh = NULL;
int i = 0;
int call_count = 0;
- int source = 0;
int32_t valid = 0;
- struct iatt stbuf = {0,};
local = frame->local;
sh = &local->self_heal;
priv = this->private;
- source = sh->source;
-
- valid |= (GF_SET_ATTR_ATIME | GF_SET_ATTR_MTIME);
-
- stbuf.ia_atime = sh->buf[source].ia_atime;
- stbuf.ia_atime_nsec = sh->buf[source].ia_atime_nsec;
- stbuf.ia_mtime = sh->buf[source].ia_mtime;
- stbuf.ia_mtime_nsec = sh->buf[source].ia_mtime_nsec;
+ valid = (GF_SET_ATTR_ATIME | GF_SET_ATTR_MTIME);
call_count = afr_set_elem_count_get (sh->success,
priv->child_count);
@@ -237,7 +242,7 @@ afr_sh_data_setattr (call_frame_t *frame, xlator_t *this)
(void *) (long) i,
priv->children[i],
priv->children[i]->fops->setattr,
- &local->loc, &stbuf, valid);
+ &local->loc, stbuf, valid, NULL);
if (!--call_count)
break;
@@ -249,7 +254,7 @@ afr_sh_data_setattr (call_frame_t *frame, xlator_t *this)
int
afr_sh_data_setattr_fstat_cbk (call_frame_t *frame, void *cookie,
xlator_t *this, int32_t op_ret, int32_t op_errno,
- struct iatt *buf)
+ struct iatt *buf, dict_t *xdata)
{
afr_local_t *local = NULL;
afr_self_heal_t *sh = NULL;
@@ -259,9 +264,14 @@ afr_sh_data_setattr_fstat_cbk (call_frame_t *frame, void *cookie,
sh = &local->self_heal;
GF_ASSERT (sh->source == child_index);
- if (op_ret != -1)
+ if (op_ret != -1) {
sh->buf[child_index] = *buf;
- afr_sh_data_setattr (frame, this);
+ afr_sh_data_setattr (frame, this, buf);
+ } else {
+ gf_log (this->name, GF_LOG_ERROR, "%s: Failed to set "
+ "time-stamps after self-heal", local->loc.path);
+ afr_sh_data_fail (frame, this);
+ }
return 0;
}
@@ -286,37 +296,51 @@ afr_sh_set_timestamps (call_frame_t *frame, xlator_t *this)
(void *) (long) sh->source,
priv->children[sh->source],
priv->children[sh->source]->fops->fstat,
- sh->healing_fd);
+ sh->healing_fd, NULL);
return 0;
}
//Fun fact, lock_cbk is being used for both lock & unlock
int
-afr_sh_data_unlock (call_frame_t *frame, xlator_t *this,
+afr_sh_data_unlock (call_frame_t *frame, xlator_t *this, char *dom,
afr_lock_cbk_t lock_cbk)
{
afr_local_t *local = NULL;
afr_internal_lock_t *int_lock = NULL;
afr_self_heal_t *sh = NULL;
+ afr_private_t *priv = NULL;
+ int ret = 0;
local = frame->local;
int_lock = &local->internal_lock;
sh = &local->self_heal;
+ priv = this->private;
- GF_ASSERT (sh->data_lock_held);
-
- sh->data_lock_held = _gf_false;
+ if (strcmp (dom, this->name) == 0) {
+ sh->data_lock_held = _gf_false;
+ } else if (strcmp (dom, priv->sh_domain) == 0) {
+ sh->sh_dom_lock_held = _gf_false;
+ } else {
+ ret = -1;
+ goto out;
+ }
int_lock->lock_cbk = lock_cbk;
+ int_lock->domain = dom;
afr_unlock (frame, this);
+out:
+ if (ret) {
+ int_lock->lock_op_ret = -1;
+ int_lock->lock_cbk (frame, this);
+ }
return 0;
}
int
afr_sh_data_finish (call_frame_t *frame, xlator_t *this)
{
- afr_local_t *local = NULL;
- afr_self_heal_t *sh = NULL;
+ afr_local_t *local = NULL;
+ afr_self_heal_t *sh = NULL;
local = frame->local;
sh = &local->self_heal;
@@ -325,9 +349,9 @@ afr_sh_data_finish (call_frame_t *frame, xlator_t *this)
"finishing data selfheal of %s", local->loc.path);
if (sh->data_lock_held)
- afr_sh_data_unlock (frame, this, afr_sh_data_close);
+ afr_sh_data_unlock (frame, this, this->name, afr_sh_dom_unlock);
else
- afr_sh_data_close (frame, this);
+ afr_sh_dom_unlock (frame, this);
return 0;
}
@@ -344,34 +368,49 @@ afr_sh_data_fail (call_frame_t *frame, xlator_t *this)
gf_log (this->name, GF_LOG_DEBUG,
"finishing failed data selfheal of %s", local->loc.path);
- sh->op_failed = 1;
- if (sh->data_lock_held)
- afr_sh_data_unlock (frame, this, afr_sh_data_close);
- else
- afr_sh_data_close (frame, this);
+ afr_set_self_heal_status (sh, AFR_SELF_HEAL_FAILED);
+ afr_sh_data_finish (frame, this);
return 0;
}
int
afr_sh_data_erase_pending_cbk (call_frame_t *frame, void *cookie,
xlator_t *this, int32_t op_ret,
- int32_t op_errno, dict_t *xattr)
+ int32_t op_errno, dict_t *xattr, dict_t *xdata)
{
int call_count = 0;
afr_local_t *local = NULL;
afr_self_heal_t *sh = NULL;
+ afr_private_t *priv = NULL;
+ int32_t child_index = (long) cookie;
+
+ priv = this->private;
+ local = frame->local;
+ sh = &local->self_heal;
+ if (op_ret < 0) {
+ gf_log (this->name, GF_LOG_ERROR, "Erasing of pending change "
+ "log failed on %s for subvol %s, reason: %s",
+ local->loc.path, priv->children[child_index]->name,
+ strerror (op_errno));
+ afr_set_self_heal_status (sh, AFR_SELF_HEAL_FAILED);
+ }
call_count = afr_frame_return (frame);
if (call_count == 0) {
- local = frame->local;
- sh = &local->self_heal;
+ if (is_self_heal_failed (sh, AFR_CHECK_SPECIFIC)) {
+ if (sh->old_loop_frame)
+ sh_loop_finish (sh->old_loop_frame, this);
+ sh->old_loop_frame = NULL;
+ afr_sh_data_fail (frame, this);
+ goto out;
+ }
if (!IA_ISREG (sh->type)) {
afr_sh_data_finish (frame, this);
goto out;
}
GF_ASSERT (sh->old_loop_frame);
- afr_sh_data_lock (frame, this, 0, 0,
+ afr_sh_data_lock (frame, this, 0, 0, _gf_true, this->name,
afr_post_sh_big_lock_success,
afr_post_sh_big_lock_failure);
}
@@ -382,74 +421,95 @@ out:
int
afr_sh_data_erase_pending (call_frame_t *frame, xlator_t *this)
{
- afr_local_t *local = NULL;
- afr_self_heal_t *sh = NULL;
- afr_private_t *priv = NULL;
- int call_count = 0;
- int i = 0;
- dict_t **erase_xattr = NULL;
+ afr_sh_erase_pending (frame, this, AFR_DATA_TRANSACTION,
+ afr_sh_data_erase_pending_cbk,
+ afr_sh_data_finish);
+ return 0;
+}
+
+int
+afr_sh_data_fsync_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
+ int op_ret, int op_errno, struct iatt *pre,
+ struct iatt *post, dict_t *xdata)
+{
+ afr_local_t *local = NULL;
+ afr_private_t *priv = NULL;
+ afr_self_heal_t *sh = NULL;
+ int call_count = 0;
+ int child_index = (long) cookie;
local = frame->local;
- sh = &local->self_heal;
priv = this->private;
+ sh = &local->self_heal;
+
+ if (op_ret < 0) {
+ gf_log (this->name, GF_LOG_ERROR, "%s: Failed to fsync on "
+ "%s - %s", local->loc.path,
+ priv->children[child_index]->name, strerror (op_errno));
+ LOCK (&frame->lock);
+ {
+ afr_set_self_heal_status (sh, AFR_SELF_HEAL_FAILED);
+ }
+ UNLOCK (&frame->lock);
+ if (sh->old_loop_frame)
+ sh_loop_finish (sh->old_loop_frame, this);
+ sh->old_loop_frame = NULL;
+ }
- afr_sh_pending_to_delta (priv, sh->xattr, sh->delta_matrix, sh->success,
- priv->child_count, AFR_DATA_TRANSACTION);
- gf_log (this->name, GF_LOG_DEBUG, "Delta matrix for: %s",
- lkowner_utoa (&frame->root->lk_owner));
- afr_sh_print_pending_matrix (sh->delta_matrix, this);
+ call_count = afr_frame_return (frame);
+ if (call_count == 0) {
+ if (is_self_heal_failed (sh, AFR_CHECK_SPECIFIC))
+ afr_sh_data_fail (frame, this);
+ else
+ afr_sh_data_erase_pending (frame, this);
+ }
+ return 0;
+}
- erase_xattr = GF_CALLOC (sizeof (*erase_xattr), priv->child_count,
- gf_afr_mt_dict_t);
+/*
+ * Before erasing xattrs, make sure the data is written to disk
+ */
+int
+afr_sh_data_fsync (call_frame_t *frame, xlator_t *this)
+{
+ afr_local_t *local = NULL;
+ afr_private_t *priv = NULL;
+ afr_self_heal_t *sh = NULL;
+ int i = 0;
+ int call_count = 0;
- for (i = 0; i < priv->child_count; i++) {
- if (sh->xattr[i]) {
- call_count++;
+ local = frame->local;
+ priv = this->private;
+ sh = &local->self_heal;
- erase_xattr[i] = get_new_dict();
- dict_ref (erase_xattr[i]);
- }
+ call_count = sh->active_sinks;
+ if (call_count == 0) {
+ afr_sh_data_erase_pending (frame, this);
+ return 0;
}
- afr_sh_delta_to_xattr (priv, sh->delta_matrix, erase_xattr,
- priv->child_count, AFR_DATA_TRANSACTION);
-
- GF_ASSERT (call_count);
local->call_count = call_count;
for (i = 0; i < priv->child_count; i++) {
- if (!erase_xattr[i])
+ if (!sh->success[i] || sh->sources[i])
continue;
- gf_log (this->name, GF_LOG_DEBUG,
- "erasing pending flags from %s on %s",
- local->loc.path, priv->children[i]->name);
-
- STACK_WIND_COOKIE (frame, afr_sh_data_erase_pending_cbk,
- (void *) (long) i,
- priv->children[i],
- priv->children[i]->fops->fxattrop,
- sh->healing_fd,
- GF_XATTROP_ADD_ARRAY, erase_xattr[i]);
- if (!--call_count)
- break;
- }
-
- for (i = 0; i < priv->child_count; i++) {
- if (erase_xattr[i]) {
- dict_unref (erase_xattr[i]);
- }
+ STACK_WIND_COOKIE (frame, afr_sh_data_fsync_cbk,
+ (void *) (long) i, priv->children[i],
+ priv->children[i]->fops->fsync,
+ sh->healing_fd, 1, NULL);
}
- GF_FREE (erase_xattr);
return 0;
}
-
static struct afr_sh_algorithm *
sh_algo_from_name (xlator_t *this, char *name)
{
int i = 0;
+ if (name == NULL)
+ goto out;
+
while (afr_self_heal_algorithms[i].name) {
if (!strcmp (name, afr_self_heal_algorithms[i].name)) {
return &afr_self_heal_algorithms[i];
@@ -458,17 +518,22 @@ sh_algo_from_name (xlator_t *this, char *name)
i++;
}
+out:
return NULL;
}
static int
-sh_zero_byte_files_exist (afr_self_heal_t *sh, int child_count)
+sh_zero_byte_files_exist (afr_local_t *local, int child_count)
{
- int i;
- int ret = 0;
+ int i = 0;
+ int ret = 0;
+ afr_self_heal_t *sh = NULL;
+ sh = &local->self_heal;
for (i = 0; i < child_count; i++) {
+ if (!local->child_up[i] || sh->child_errno[i])
+ continue;
if (sh->buf[i].ia_size == 0) {
ret = 1;
break;
@@ -495,8 +560,7 @@ afr_sh_data_pick_algo (call_frame_t *frame, xlator_t *this)
if (algo == NULL) {
/* option not set, so fall back on heuristics */
- if ((local->enoent_count != 0)
- || sh_zero_byte_files_exist (sh, priv->child_count)
+ if (sh_zero_byte_files_exist (local, priv->child_count)
|| (sh->file_size <= (priv->data_self_heal_window_size *
this->ctx->page_size))) {
@@ -534,11 +598,12 @@ afr_sh_data_sync_prepare (call_frame_t *frame, xlator_t *this)
local = frame->local;
sh = &local->self_heal;
- sh->algo_completion_cbk = afr_sh_data_erase_pending;
+ sh->algo_completion_cbk = afr_sh_data_fsync;
sh->algo_abort_cbk = afr_sh_data_fail;
sh_algo = afr_sh_data_pick_algo (frame, this);
+ sh->algo = sh_algo;
sh_algo->fn (frame, this);
return 0;
@@ -547,38 +612,46 @@ afr_sh_data_sync_prepare (call_frame_t *frame, xlator_t *this)
int
afr_sh_data_trim_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
int32_t op_ret, int32_t op_errno, struct iatt *prebuf,
- struct iatt *postbuf)
+ struct iatt *postbuf, dict_t *xdata)
{
- afr_private_t * priv = NULL;
- afr_local_t * local = NULL;
int call_count = 0;
int child_index = 0;
+ afr_private_t *priv = NULL;
+ afr_local_t *local = NULL;
+ afr_self_heal_t *sh = NULL;
- priv = this->private;
+ priv = this->private;
local = frame->local;
+ sh = &local->self_heal;
child_index = (long) cookie;
LOCK (&frame->lock);
{
- if (op_ret == -1)
- gf_log (this->name, GF_LOG_INFO,
+ if (op_ret == -1) {
+ gf_log (this->name, GF_LOG_ERROR,
"ftruncate of %s on subvolume %s failed (%s)",
local->loc.path,
priv->children[child_index]->name,
strerror (op_errno));
- else
+ afr_set_self_heal_status (sh, AFR_SELF_HEAL_FAILED);
+ } else {
gf_log (this->name, GF_LOG_DEBUG,
"ftruncate of %s on subvolume %s completed",
local->loc.path,
priv->children[child_index]->name);
+ }
}
UNLOCK (&frame->lock);
call_count = afr_frame_return (frame);
- if (call_count == 0)
- afr_sh_data_sync_prepare (frame, this);
+ if (call_count == 0) {
+ if (is_self_heal_failed (sh, AFR_CHECK_SPECIFIC))
+ afr_sh_data_fail (frame, this);
+ else
+ afr_sh_data_sync_prepare (frame, this);
+ }
return 0;
}
@@ -612,7 +685,8 @@ afr_sh_data_trim_sinks (call_frame_t *frame, xlator_t *this)
(void *) (long) i,
priv->children[i],
priv->children[i]->fops->ftruncate,
- sh->healing_fd, sh->file_size);
+ sh->healing_fd, sh->file_size,
+ NULL);
if (!--call_count)
break;
@@ -626,6 +700,7 @@ afr_sh_inode_set_read_ctx (afr_self_heal_t *sh, xlator_t *this)
{
afr_private_t *priv = NULL;
int ret = 0;
+ int i = 0;
priv = this->private;
sh->source = afr_sh_select_source (sh->sources, priv->child_count);
@@ -634,6 +709,15 @@ afr_sh_inode_set_read_ctx (afr_self_heal_t *sh, xlator_t *this)
goto out;
}
+ /* detect changes not visible through pending flags -- JIC */
+ for (i = 0; i < priv->child_count; i++) {
+ if (i == sh->source || sh->child_errno[i])
+ continue;
+
+ if (SIZE_DIFFERS (&sh->buf[i], &sh->buf[sh->source]))
+ sh->sources[i] = 0;
+ }
+
afr_reset_children (sh->fresh_children, priv->child_count);
afr_get_fresh_children (sh->success_children, sh->sources,
sh->fresh_children, priv->child_count);
@@ -643,73 +727,211 @@ out:
return ret;
}
-int
-afr_sh_data_fix (call_frame_t *frame, xlator_t *this)
+char*
+afr_get_sizes_str (afr_local_t *local, struct iatt *bufs, xlator_t *this)
{
- afr_local_t *local = NULL;
- afr_self_heal_t *sh = NULL;
- afr_private_t *priv = NULL;
- int nsources = 0;
- int source = 0;
- int i = 0;
- int ret = 0;
+ afr_private_t *priv = NULL;
+ int i = 0;
+ char num[1024] = {0};
+ size_t len = 0;
+ char *sizes_str = NULL;
+ size_t off = 0;
+ char *fmt_str = "%llu bytes on %s, ";
+ char *child_down = " %s,";
+ char *child_unknown = " %s,";
+ int down_child_present = 0;
+ int down_count = 0;
+ int unknown_count = 0;
+ int unknown_child_present = 0;
+ char *down_subvol_1 = " down subvolume is ";
+ char *unknown_subvol_1 = " unknown subvolume is ";
+ char *down_subvol_2 = " down subvolumes are ";
+ char *unknown_subvol_2 = " unknown subvolumes are ";
- local = frame->local;
- sh = &local->self_heal;
priv = this->private;
- gf_log (this->name, GF_LOG_DEBUG, "Pending matrix for: %s",
- lkowner_utoa (&frame->root->lk_owner));
+ for (i = 0; i < priv->child_count; i++) {
+ if (local->child_up[i] == 1) {
+ len += snprintf (num, sizeof (num), fmt_str,
+ (unsigned long long) bufs[i].ia_size,
+ priv->children[i]->name);
+ } else if (local->child_up[i] == 0) {
+ len += snprintf (num, sizeof (num), child_down,
+ priv->children[i]->name);
+ if (!down_child_present)
+ down_child_present = 1;
+ down_count ++;
+ } else if (local->child_up[i] == -1) {
+ len += snprintf (num, sizeof (num), child_unknown,
+ priv->children[i]->name);
+ if (!unknown_child_present)
+ unknown_child_present = 1;
+ unknown_count++;
+ }
- nsources = afr_build_sources (this, sh->xattr, sh->buf, sh->pending_matrix,
- sh->sources, sh->success_children,
- AFR_DATA_TRANSACTION, NULL, _gf_false);
- if (nsources == 0) {
- gf_log (this->name, GF_LOG_DEBUG,
- "No self-heal needed for %s",
- local->loc.path);
+ }
- afr_sh_data_finish (frame, this);
- return 0;
+ if (down_child_present) {
+ if (down_count > 1)
+ len += snprintf (num, sizeof (num), "%s",
+ down_subvol_2);
+ else
+ len += snprintf (num, sizeof (num), "%s",
+ down_subvol_1);
+ }
+ if (unknown_child_present) {
+ if (unknown_count > 1)
+ len += snprintf (num, sizeof (num), "%s",
+ unknown_subvol_2);
+ else
+ len += snprintf (num, sizeof (num), "%s",
+ unknown_subvol_1);
}
- if ((nsources == -1)
- && (priv->favorite_child != -1)
- && (sh->child_errno[priv->favorite_child] == 0)) {
+ len++;//for '\0'
- gf_log (this->name, GF_LOG_DEBUG,
- "Picking favorite child %s as authentic source to "
- "resolve conflicting data of %s",
- priv->children[priv->favorite_child]->name,
- local->loc.path);
+ sizes_str = GF_CALLOC (len, sizeof (char), gf_common_mt_char);
- sh->sources[priv->favorite_child] = 1;
+ if (!sizes_str)
+ return NULL;
- nsources = afr_sh_source_count (sh->sources,
- priv->child_count);
+ for (i = 0; i < priv->child_count; i++) {
+ if (local->child_up[i] == 1) {
+ off += snprintf (sizes_str + off, len - off, fmt_str,
+ (unsigned long long) bufs[i].ia_size,
+ priv->children[i]->name);
+ }
}
- if (nsources == -1) {
- gf_log (this->name, GF_LOG_ERROR,
- "Unable to self-heal contents of '%s' (possible "
- "split-brain). Please delete the file from all but "
- "the preferred subvolume.", local->loc.path);
+ if (down_child_present) {
+ if (down_count > 1) {
+ off += snprintf (sizes_str + off, len - off, "%s",
+ down_subvol_2);
+ } else {
+ off += snprintf (sizes_str + off, len - off, "%s",
+ down_subvol_1);
+ }
+ }
- local->govinda_gOvinda = 1;
+ for (i = 0; i < priv->child_count; i++) {
+ if (local->child_up[i] == 0) {
+ off += snprintf (sizes_str + off, len - off, child_down,
+ priv->children[i]->name);
+ }
+ }
- afr_sh_data_fail (frame, this);
- return 0;
+ if (unknown_child_present) {
+ if (unknown_count > 1) {
+ off += snprintf (sizes_str + off, len - off, "%s",
+ unknown_subvol_2);
+ } else {
+ off += snprintf (sizes_str + off, len - off, "%s",
+ unknown_subvol_1);
+ }
}
- ret = afr_sh_inode_set_read_ctx (sh, this);
- if (ret) {
- gf_log (this->name, GF_LOG_DEBUG,
- "No active sources found.");
+ for (i = 0; i < priv->child_count; i++) {
+ if (local->child_up[i] == -1) {
+ off += snprintf (sizes_str + off, len - off,
+ child_unknown,
+ priv->children[i]->name);
- afr_sh_data_fail (frame, this);
- return 0;
+ }
}
+ return sizes_str;
+}
+
+char*
+afr_get_sinks_str (xlator_t *this, afr_local_t *local, afr_self_heal_t *sh)
+{
+ afr_private_t *priv = NULL;
+ int i = 0;
+ char num[1024] = {0};
+ size_t len = 0;
+ char *sinks_str = NULL;
+ char *temp_str = " to sinks ";
+ char *str_format = " %s,";
+ char off = 0;
+
+ priv = this->private;
+
+ len += snprintf (num, sizeof (num), "%s", temp_str);
+ for (i = 0; i < priv->child_count; i++) {
+ if ((sh->sources[i] == 0) && (local->child_up[i] == 1)) {
+ len += snprintf (num, sizeof (num), str_format,
+ priv->children[i]->name);
+ }
+ }
+
+ len ++;
+
+ sinks_str = GF_CALLOC (len, sizeof (char), gf_common_mt_char);
+
+ if (!sinks_str)
+ return NULL;
+
+ off += snprintf (sinks_str + off, len - off, "%s", temp_str);
+
+ for (i = 0; i < priv->child_count; i++) {
+ if ((sh->sources[i] == 0) && (local->child_up[i] == 1)) {
+ off += snprintf (sinks_str + off, len - off,
+ str_format,
+ priv->children[i]->name);
+ }
+ }
+
+ return sinks_str;
+
+}
+
+
+void
+afr_set_data_sh_info_str (afr_local_t *local, afr_self_heal_t *sh, xlator_t *this)
+{
+ char *pending_matrix_str = NULL;
+ char *sizes_str = NULL;
+ char *sinks_str = NULL;
+ afr_private_t *priv = NULL;
+
+ priv = this->private;
+
+ pending_matrix_str = afr_get_pending_matrix_str (sh->pending_matrix,
+ this);
+ if (!pending_matrix_str)
+ pending_matrix_str = "";
+
+ sizes_str = afr_get_sizes_str (local, sh->buf, this);
+ if (!sizes_str)
+ sizes_str = "";
+
+ sinks_str = afr_get_sinks_str (this, local, sh);
+ if (!sinks_str)
+ sinks_str = "";
+
+ gf_asprintf (&sh->data_sh_info, " data self heal from %s %s with "
+ "%s data %s", priv->children[sh->source]->name, sinks_str,
+ sizes_str, pending_matrix_str);
+
+ if (pending_matrix_str && strcmp (pending_matrix_str, ""))
+ GF_FREE (pending_matrix_str);
+
+ if (sizes_str && strcmp (sizes_str, ""))
+ GF_FREE (sizes_str);
+}
+
+void
+afr_sh_data_fix (call_frame_t *frame, xlator_t *this)
+{
+ int source = 0;
+ afr_local_t *local = NULL;
+ afr_self_heal_t *sh = NULL;
+ afr_private_t *priv = NULL;
+
+ local = frame->local;
+ sh = &local->self_heal;
+ priv = this->private;
+
source = sh->source;
sh->block_size = this->ctx->page_size;
sh->file_size = sh->buf[source].ia_size;
@@ -717,17 +939,9 @@ afr_sh_data_fix (call_frame_t *frame, xlator_t *this)
if (FILE_HAS_HOLES (&sh->buf[source]))
sh->file_has_holes = 1;
- /* detect changes not visible through pending flags -- JIC */
- for (i = 0; i < priv->child_count; i++) {
- if (i == source || sh->child_errno[i])
- continue;
-
- if (SIZE_DIFFERS (&sh->buf[i], &sh->buf[source]))
- sh->sources[i] = 0;
- }
-
- if (sh->background && sh->unwind) {
- sh->unwind (sh->orig_frame, this, sh->op_ret, sh->op_errno);
+ if (sh->background && sh->unwind && !sh->unwound) {
+ sh->unwind (sh->orig_frame, this, sh->op_ret, sh->op_errno,
+ is_self_heal_failed (sh, AFR_CHECK_SPECIFIC));
sh->unwound = _gf_true;
}
@@ -737,66 +951,120 @@ afr_sh_data_fix (call_frame_t *frame, xlator_t *this)
"no active sinks for performing self-heal on file %s",
local->loc.path);
afr_sh_data_finish (frame, this);
- return 0;
+ return;
}
gf_log (this->name, GF_LOG_DEBUG,
"self-healing file %s from subvolume %s to %d other",
local->loc.path, priv->children[sh->source]->name,
sh->active_sinks);
- afr_sh_data_trim_sinks (frame, this);
- return 0;
+ sh->actual_sh_started = _gf_true;
+ afr_set_self_heal_status (sh, AFR_SELF_HEAL_SYNC_BEGIN);
+ afr_sh_data_trim_sinks (frame, this);
}
-static void
-afr_destroy_pending_matrix (int32_t **pending_matrix, int32_t child_count)
+int
+afr_sh_data_fxattrop_fstat_done (call_frame_t *frame, xlator_t *this)
{
+ afr_local_t *local = NULL;
+ afr_self_heal_t *sh = NULL;
+ afr_private_t *priv = NULL;
+ int nsources = 0;
+ int ret = 0;
+ int *old_sources = NULL;
+ int tstamp_source = 0;
int i = 0;
- GF_ASSERT (child_count > 0);
- if (pending_matrix) {
- for (i = 0; i < child_count; i++) {
- if (pending_matrix[i])
- GF_FREE (pending_matrix[i]);
- }
- GF_FREE (pending_matrix);
+
+ local = frame->local;
+ sh = &local->self_heal;
+ priv = this->private;
+
+ gf_log (this->name, GF_LOG_DEBUG, "Pending matrix for: %s",
+ lkowner_utoa (&frame->root->lk_owner));
+ if (sh->sync_done) {
+ //store sources before sync so that mtime can be set using the
+ //iatt buf from one of them.
+ old_sources = alloca (priv->child_count*sizeof (*old_sources));
+ memcpy (old_sources, sh->sources,
+ priv->child_count * sizeof (*old_sources));
}
-}
-static int32_t**
-afr_create_pending_matrix (int32_t child_count)
-{
- gf_boolean_t cleanup = _gf_false;
- int32_t **pending_matrix = NULL;
- int i = 0;
+ nsources = afr_build_sources (this, sh->xattr, sh->buf, sh->pending_matrix,
+ sh->sources, sh->success_children,
+ AFR_DATA_TRANSACTION, NULL, _gf_true);
+ if ((nsources == -1)
+ && (priv->favorite_child != -1)
+ && (sh->child_errno[priv->favorite_child] == 0)) {
- GF_ASSERT (child_count > 0);
+ gf_log (this->name, GF_LOG_DEBUG,
+ "Picking favorite child %s as authentic source to "
+ "resolve conflicting data of %s",
+ priv->children[priv->favorite_child]->name,
+ local->loc.path);
- pending_matrix = GF_CALLOC (sizeof (*pending_matrix), child_count,
- gf_afr_mt_int32_t);
- if (NULL == pending_matrix)
- goto out;
- for (i = 0; i < child_count; i++) {
- pending_matrix[i] = GF_CALLOC (sizeof (**pending_matrix),
- child_count,
- gf_afr_mt_int32_t);
- if (NULL == pending_matrix[i]) {
- cleanup = _gf_true;
- goto out;
- }
+ sh->sources[priv->favorite_child] = 1;
+
+ nsources = afr_sh_source_count (sh->sources,
+ priv->child_count);
}
-out:
- if (_gf_true == cleanup) {
- afr_destroy_pending_matrix (pending_matrix, child_count);
- pending_matrix = NULL;
+
+ if (nsources == -1) {
+ afr_sh_print_split_brain_log (sh->pending_matrix, this,
+ local->loc.path);
+ afr_set_split_brain (this, sh->inode, DONT_KNOW, SPB);
+
+ afr_sh_data_fail (frame, this);
+ return 0;
+ }
+
+ afr_set_split_brain (this, sh->inode, DONT_KNOW, NO_SPB);
+
+ ret = afr_sh_inode_set_read_ctx (sh, this);
+ if (ret) {
+ gf_log (this->name, GF_LOG_DEBUG,
+ "No active sources found.");
+
+ afr_sh_data_fail (frame, this);
+ return 0;
}
- return pending_matrix;
+
+ if (sh->sync_done) {
+ /* Perform setattr from one of the old_sources if possible
+ * Because only they have the correct mtime, the new sources
+ * (i.e. old sinks) have mtime from last writev in sync.
+ */
+ tstamp_source = sh->source;
+ for (i = 0; i < priv->child_count; i++) {
+ if (old_sources[i] && sh->sources[i])
+ tstamp_source = i;
+ }
+ afr_sh_data_setattr (frame, this, &sh->buf[tstamp_source]);
+ } else {
+ afr_set_data_sh_info_str (local, sh, this);
+ if (nsources == 0) {
+ gf_log (this->name, GF_LOG_DEBUG,
+ "No self-heal needed for %s",
+ local->loc.path);
+
+ afr_sh_data_finish (frame, this);
+ return 0;
+ }
+
+ if (sh->do_data_self_heal &&
+ afr_data_self_heal_enabled (priv->data_self_heal))
+ afr_sh_data_fix (frame, this);
+ else
+ afr_sh_data_finish (frame, this);
+ }
+ return 0;
}
int
afr_lookup_select_read_child_by_txn_type (xlator_t *this, afr_local_t *local,
dict_t **xattr,
- afr_transaction_type txn_type)
+ afr_transaction_type txn_type,
+ uuid_t gfid)
{
afr_private_t *priv = NULL;
int read_child = -1;
@@ -813,19 +1081,33 @@ afr_lookup_select_read_child_by_txn_type (xlator_t *this, afr_local_t *local,
bufs = local->cont.lookup.bufs;
success_children = local->cont.lookup.success_children;
- pending_matrix = afr_create_pending_matrix (priv->child_count);
- if (NULL == pending_matrix)
- goto out;
-
+ pending_matrix = local->cont.lookup.pending_matrix;
sources = local->cont.lookup.sources;
memset (sources, 0, sizeof (*sources) * priv->child_count);
nsources = afr_build_sources (this, xattr, bufs, pending_matrix,
sources, success_children, txn_type,
&subvol_status, _gf_false);
- if (subvol_status & SPLIT_BRAIN)
- gf_log (this->name, GF_LOG_WARNING, "%s: Possible split-brain",
+ if (subvol_status & SPLIT_BRAIN) {
+ gf_log (this->name, GF_LOG_DEBUG, "%s: Possible split-brain",
local->loc.path);
+ switch (txn_type) {
+ case AFR_DATA_TRANSACTION:
+ local->cont.lookup.possible_spb = _gf_true;
+ nsources = 1;
+ sources[success_children[0]] = 1;
+ break;
+ case AFR_ENTRY_TRANSACTION:
+ read_child = afr_get_no_xattr_dir_read_child (this,
+ success_children,
+ bufs);
+ sources[read_child] = 1;
+ nsources = 1;
+ break;
+ default:
+ break;
+ }
+ }
if (nsources < 0)
goto out;
@@ -835,39 +1117,18 @@ afr_lookup_select_read_child_by_txn_type (xlator_t *this, afr_local_t *local,
priv->child_count,
prev_read_child,
config_read_child,
- sources);
+ sources,
+ priv->hash_mode, gfid);
out:
- afr_destroy_pending_matrix (pending_matrix, priv->child_count);
gf_log (this->name, GF_LOG_DEBUG, "returning read_child: %d",
read_child);
return read_child;
}
int
-afr_sh_data_special_file_fix (call_frame_t *frame, xlator_t *this)
-{
- afr_private_t *priv = NULL;
- afr_self_heal_t *sh = NULL;
- afr_local_t *local = NULL;
- int i = 0;
-
- local = frame->local;
- sh = &local->self_heal;
- priv = this->private;
-
- for (i = 0; i < priv->child_count ; i++)
- if (1 == local->child_up[i])
- sh->success[i] = 1;
-
- afr_sh_data_erase_pending (frame, this);
-
- return 0;
-}
-
-int
afr_sh_data_fstat_cbk (call_frame_t *frame, void *cookie,
xlator_t *this, int32_t op_ret, int32_t op_errno,
- struct iatt *buf)
+ struct iatt *buf, dict_t *xdata)
{
afr_private_t *priv = NULL;
afr_local_t *local = NULL;
@@ -890,6 +1151,12 @@ afr_sh_data_fstat_cbk (call_frame_t *frame, void *cookie,
sh->buf[child_index] = *buf;
sh->success_children[sh->success_count] = child_index;
sh->success_count++;
+ } else {
+ gf_log (this->name, GF_LOG_ERROR, "%s: fstat failed "
+ "on %s, reason %s", local->loc.path,
+ priv->children[child_index]->name,
+ strerror (op_errno));
+ sh->child_errno[child_index] = op_errno;
}
}
UNLOCK (&frame->lock);
@@ -900,13 +1167,17 @@ afr_sh_data_fstat_cbk (call_frame_t *frame, void *cookie,
/* Previous versions of glusterfs might have set
* the pending data xattrs which need to be erased
*/
- if (IA_ISREG (buf->ia_type))
- afr_sh_data_fix (frame, this);
- else
- afr_sh_data_special_file_fix (frame, this);
-
+ if (!afr_sh_data_proceed (sh->success_count)) {
+ gf_log (this->name, GF_LOG_ERROR, "inspecting metadata "
+ "succeeded on < %d children, aborting "
+ "self-heal for %s", AFR_SH_MIN_PARTICIPANTS,
+ local->loc.path);
+ afr_sh_data_fail (frame, this);
+ goto out;
+ }
+ afr_sh_data_fxattrop_fstat_done (frame, this);
}
-
+out:
return 0;
}
@@ -917,33 +1188,41 @@ afr_sh_data_fstat (call_frame_t *frame, xlator_t *this)
afr_self_heal_t *sh = NULL;
afr_local_t *local = NULL;
afr_private_t *priv = NULL;
- int call_count = 0;
- int i = 0;
+ int call_count = 0;
+ int i = 0;
+ int child = 0;
+ int32_t *fstat_children = NULL;
priv = this->private;
local = frame->local;
sh = &local->self_heal;
- call_count = afr_up_children_count (local->child_up,
- priv->child_count);
-
+ fstat_children = memdup (sh->success_children,
+ sizeof (*fstat_children) * priv->child_count);
+ if (!fstat_children) {
+ afr_sh_data_fail (frame, this);
+ goto out;
+ }
+ call_count = sh->success_count;
local->call_count = call_count;
+ memset (sh->buf, 0, sizeof (*sh->buf) * priv->child_count);
afr_reset_children (sh->success_children, priv->child_count);
sh->success_count = 0;
for (i = 0; i < priv->child_count; i++) {
- if (local->child_up[i]) {
- STACK_WIND_COOKIE (frame, afr_sh_data_fstat_cbk,
- (void *) (long) i,
- priv->children[i],
- priv->children[i]->fops->fstat,
- sh->healing_fd);
-
- if (!--call_count)
- break;
- }
+ child = fstat_children[i];
+ if (child == -1)
+ break;
+ STACK_WIND_COOKIE (frame, afr_sh_data_fstat_cbk,
+ (void *) (long) child,
+ priv->children[child],
+ priv->children[child]->fops->fstat,
+ sh->healing_fd, NULL);
+ --call_count;
}
-
+ GF_ASSERT (!call_count);
+out:
+ GF_FREE (fstat_children);
return 0;
}
@@ -972,73 +1251,60 @@ afr_sh_common_fxattrop_resp_handler (call_frame_t *frame, void *cookie,
sh->xattr[child_index] = dict_ref (xattr);
sh->success_children[sh->success_count] = child_index;
sh->success_count++;
+ } else {
+ gf_log (this->name, GF_LOG_ERROR, "fxattrop of %s "
+ "failed on %s, reason %s", local->loc.path,
+ priv->children[child_index]->name,
+ strerror (op_errno));
+ sh->child_errno[child_index] = op_errno;
}
}
UNLOCK (&frame->lock);
}
int
-afr_post_sh_data_fxattrop_cbk (call_frame_t *frame, void *cookie,
- xlator_t *this, int32_t op_ret, int32_t op_errno,
- dict_t *xattr)
+afr_sh_data_fxattrop_cbk (call_frame_t *frame, void *cookie,
+ xlator_t *this, int32_t op_ret, int32_t op_errno,
+ dict_t *xattr, dict_t *xdata)
{
int call_count = -1;
- int ret = 0;
- afr_local_t *local = NULL;
- afr_self_heal_t *sh = NULL;
-
- afr_sh_common_fxattrop_resp_handler (frame, cookie, this, op_ret,
- op_errno, xattr);
+ afr_local_t *local = NULL;
+ afr_self_heal_t *sh = NULL;
local = frame->local;
- sh = &local->self_heal;
- call_count = afr_frame_return (frame);
- if (call_count == 0) {
- (void) afr_build_sources (this, sh->xattr, NULL,
- sh->pending_matrix,
- sh->sources, sh->success_children,
- AFR_DATA_TRANSACTION, NULL, _gf_false);
- ret = afr_sh_inode_set_read_ctx (sh, this);
- if (ret)
- afr_sh_data_fail (frame, this);
- else
- afr_sh_set_timestamps (frame, this);
- }
-
- return 0;
-}
-
-int
-afr_sh_data_fxattrop_cbk (call_frame_t *frame, void *cookie,
- xlator_t *this, int32_t op_ret, int32_t op_errno,
- dict_t *xattr)
-{
- int call_count = -1;
+ sh = &local->self_heal;
afr_sh_common_fxattrop_resp_handler (frame, cookie, this, op_ret,
op_errno, xattr);
call_count = afr_frame_return (frame);
if (call_count == 0) {
+ if (!afr_sh_data_proceed (sh->success_count)) {
+ gf_log (this->name, GF_LOG_ERROR, "%s, inspecting "
+ "change log succeeded on < %d children",
+ local->loc.path, AFR_SH_MIN_PARTICIPANTS);
+ afr_sh_data_fail (frame, this);
+ goto out;
+ }
afr_sh_data_fstat (frame, this);
}
-
+out:
return 0;
}
int
-afr_sh_data_fxattrop (call_frame_t *frame, xlator_t *this,
- afr_fxattrop_cbk_t fxattrop_cbk)
+afr_sh_data_fxattrop (call_frame_t *frame, xlator_t *this)
{
afr_self_heal_t *sh = NULL;
afr_local_t *local = NULL;
afr_private_t *priv = NULL;
- dict_t *xattr_req = NULL;
+ dict_t **xattr_req;
int32_t *zero_pending = NULL;
int call_count = 0;
int i = 0;
int ret = 0;
+ int j;
priv = this->private;
local = frame->local;
@@ -1049,42 +1315,53 @@ afr_sh_data_fxattrop (call_frame_t *frame, xlator_t *this,
local->call_count = call_count;
- xattr_req = dict_new();
- if (!xattr_req) {
- ret = -1;
- goto out;
- }
-
- for (i = 0; i < priv->child_count; i++) {
- zero_pending = GF_CALLOC (3, sizeof (*zero_pending),
- gf_afr_mt_int32_t);
- if (!zero_pending) {
- ret = -1;
- goto out;
- }
- ret = dict_set_dynptr (xattr_req, priv->pending_key[i],
- zero_pending,
- 3 * sizeof (*zero_pending));
- if (ret < 0) {
- gf_log (this->name, GF_LOG_WARNING,
- "Unable to set dict value");
- goto out;
- } else {
- zero_pending = NULL;
- }
- }
+ xattr_req = GF_CALLOC(priv->child_count, sizeof(struct dict_t *),
+ gf_afr_mt_dict_t);
+ if (!xattr_req)
+ goto out;
+
+ for (i = 0; i < priv->child_count; i++) {
+ xattr_req[i] = dict_new();
+ if (!xattr_req[i]) {
+ ret = -1;
+ goto out;
+ }
+ }
+
+ for (i = 0; i < priv->child_count; i++) {
+ for (j = 0; j < priv->child_count; j++) {
+ zero_pending = GF_CALLOC (3, sizeof (*zero_pending),
+ gf_afr_mt_int32_t);
+ if (!zero_pending) {
+ ret = -1;
+ goto out;
+ }
+ ret = dict_set_dynptr (xattr_req[i], priv->pending_key[j],
+ zero_pending,
+ 3 * sizeof (*zero_pending));
+ if (ret < 0) {
+ gf_log (this->name, GF_LOG_WARNING,
+ "Unable to set dict value");
+ goto out;
+ } else {
+ zero_pending = NULL;
+ }
+ }
+ }
afr_reset_xattr (sh->xattr, priv->child_count);
afr_reset_children (sh->success_children, priv->child_count);
+ memset (sh->child_errno, 0,
+ sizeof (*sh->child_errno) * priv->child_count);
sh->success_count = 0;
for (i = 0; i < priv->child_count; i++) {
if (local->child_up[i]) {
- STACK_WIND_COOKIE (frame, fxattrop_cbk,
+ STACK_WIND_COOKIE (frame, afr_sh_data_fxattrop_cbk,
(void *) (long) i,
priv->children[i],
priv->children[i]->fops->fxattrop,
sh->healing_fd, GF_XATTROP_ADD_ARRAY,
- xattr_req);
+ xattr_req[i], NULL);
if (!--call_count)
break;
@@ -1092,14 +1369,16 @@ afr_sh_data_fxattrop (call_frame_t *frame, xlator_t *this,
}
out:
- if (xattr_req)
- dict_unref (xattr_req);
+ if (xattr_req) {
+ for (i = 0; i < priv->child_count; i++)
+ if (xattr_req[i])
+ dict_unref(xattr_req[i]);
+ GF_FREE(xattr_req);
+ }
if (ret) {
- if (zero_pending)
- GF_FREE (zero_pending);
- sh->op_failed = 1;
- afr_sh_data_done (frame, this);
+ GF_FREE (zero_pending);
+ afr_sh_data_fail (frame, this);
}
return 0;
@@ -1115,7 +1394,23 @@ afr_sh_data_big_lock_success (call_frame_t *frame, xlator_t *this)
sh = &local->self_heal;
sh->data_lock_held = _gf_true;
- afr_sh_data_fxattrop (frame, this, afr_sh_data_fxattrop_cbk);
+ afr_sh_data_fxattrop (frame, this);
+ return 0;
+}
+
+int
+afr_sh_dom_lock_success (call_frame_t *frame, xlator_t *this)
+{
+ afr_local_t *local = NULL;
+ afr_self_heal_t *sh = NULL;
+
+ local = frame->local;
+ sh = &local->self_heal;
+
+ sh->sh_dom_lock_held = _gf_true;
+ afr_sh_data_lock (frame, this, 0, 0, _gf_true, this->name,
+ afr_sh_data_big_lock_success,
+ afr_sh_data_fail);
return 0;
}
@@ -1164,8 +1459,13 @@ afr_sh_data_post_nonblocking_inodelk_cbk (call_frame_t *frame, xlator_t *this)
"failed for %s. by %s",
local->loc.path, lkowner_utoa (&frame->root->lk_owner));
- int_lock->lock_cbk = afr_sh_data_post_blocking_inodelk_cbk;
- afr_blocking_lock (frame, this);
+ if (!sh->data_lock_block) {
+ sh->data_lock_failure_handler(frame, this);
+ } else {
+ int_lock->lock_cbk =
+ afr_sh_data_post_blocking_inodelk_cbk;
+ afr_blocking_lock (frame, this);
+ }
} else {
gf_log (this->name, GF_LOG_DEBUG, "Non Blocking data inodelks "
@@ -1178,9 +1478,11 @@ afr_sh_data_post_nonblocking_inodelk_cbk (call_frame_t *frame, xlator_t *this)
}
int
-afr_sh_data_lock_rec (call_frame_t *frame, xlator_t *this, off_t start, off_t len)
+afr_sh_data_lock_rec (call_frame_t *frame, xlator_t *this, char *dom,
+ off_t start, off_t len)
{
afr_internal_lock_t *int_lock = NULL;
+ afr_inodelk_t *inodelk = NULL;
afr_local_t *local = NULL;
local = frame->local;
@@ -1191,11 +1493,14 @@ afr_sh_data_lock_rec (call_frame_t *frame, xlator_t *this, off_t start, off_t le
afr_set_lock_number (frame, this);
- int_lock->lk_flock.l_start = start;
- int_lock->lk_flock.l_len = len;
- int_lock->lk_flock.l_type = F_WRLCK;
int_lock->lock_cbk = afr_sh_data_post_nonblocking_inodelk_cbk;
+ int_lock->domain = dom;
+ inodelk = afr_get_inodelk (int_lock, int_lock->domain);
+ inodelk->flock.l_start = start;
+ inodelk->flock.l_len = len;
+ inodelk->flock.l_type = F_WRLCK;
+
afr_nonblocking_inodelk (frame, this);
return 0;
@@ -1214,7 +1519,8 @@ afr_post_sh_big_lock_success (call_frame_t *frame, xlator_t *this)
sh_loop_finish (sh->old_loop_frame, this);
sh->old_loop_frame = NULL;
sh->data_lock_held = _gf_true;
- afr_sh_data_fxattrop (frame, this, afr_post_sh_data_fxattrop_cbk);
+ sh->sync_done = _gf_true;
+ afr_sh_data_fxattrop (frame, this);
return 0;
}
@@ -1237,8 +1543,8 @@ afr_post_sh_big_lock_failure (call_frame_t *frame, xlator_t *this)
int
afr_sh_data_lock (call_frame_t *frame, xlator_t *this,
- off_t start, off_t len,
- afr_lock_cbk_t success_handler,
+ off_t start, off_t len, gf_boolean_t block,
+ char *dom, afr_lock_cbk_t success_handler,
afr_lock_cbk_t failure_handler)
{
afr_local_t * local = NULL;
@@ -1249,12 +1555,13 @@ afr_sh_data_lock (call_frame_t *frame, xlator_t *this,
sh->data_lock_success_handler = success_handler;
sh->data_lock_failure_handler = failure_handler;
- return afr_sh_data_lock_rec (frame, this, start, len);
+ sh->data_lock_block = block;
+ return afr_sh_data_lock_rec (frame, this, dom, start, len);
}
int
afr_sh_data_open_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
- int32_t op_ret, int32_t op_errno, fd_t *fd)
+ int32_t op_ret, int32_t op_errno, fd_t *fd, dict_t *xdata)
{
afr_local_t *local = NULL;
afr_self_heal_t *sh = NULL;
@@ -1280,20 +1587,20 @@ afr_sh_data_open_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
local->loc.path,
priv->children[child_index]->name,
strerror (op_errno));
- sh->op_failed = 1;
+ afr_set_self_heal_status (sh, AFR_SELF_HEAL_FAILED);
+ } else {
+ gf_log (this->name, GF_LOG_TRACE,
+ "open of %s succeeded on child %s",
+ local->loc.path,
+ priv->children[child_index]->name);
}
-
- gf_log (this->name, GF_LOG_TRACE,
- "open of %s succeeded on child %s",
- local->loc.path,
- priv->children[child_index]->name);
}
UNLOCK (&frame->lock);
call_count = afr_frame_return (frame);
if (call_count == 0) {
- if (sh->op_failed) {
+ if (is_self_heal_failed (sh, AFR_CHECK_SPECIFIC)) {
afr_sh_data_fail (frame, this);
return 0;
}
@@ -1302,9 +1609,8 @@ afr_sh_data_open_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
"fd for %s opened, commencing sync",
local->loc.path);
- afr_sh_data_lock (frame, this, 0, 0,
- afr_sh_data_big_lock_success,
- afr_sh_data_fail);
+ afr_sh_data_lock (frame, this, 0, 0, _gf_true, priv->sh_domain,
+ afr_sh_dom_lock_success, afr_sh_data_fail);
}
return 0;
@@ -1341,7 +1647,7 @@ afr_sh_data_open (call_frame_t *frame, xlator_t *this)
priv->children[i],
priv->children[i]->fops->open,
&local->loc,
- O_RDWR|O_LARGEFILE, fd, 0);
+ O_RDWR|O_LARGEFILE, fd, NULL);
if (!--call_count)
break;
@@ -1350,20 +1656,93 @@ afr_sh_data_open (call_frame_t *frame, xlator_t *this)
return 0;
}
+void
+afr_sh_non_reg_fix (call_frame_t *frame, xlator_t *this,
+ int32_t op_ret, int32_t op_errno)
+{
+ afr_private_t *priv = NULL;
+ afr_self_heal_t *sh = NULL;
+ afr_local_t *local = NULL;
+ int i = 0;
+
+ if (op_ret < 0) {
+ afr_sh_data_fail (frame, this);
+ return;
+ }
+
+ local = frame->local;
+ sh = &local->self_heal;
+ priv = this->private;
+
+ for (i = 0; i < priv->child_count ; i++) {
+ if (1 == local->child_up[i])
+ sh->success[i] = 1;
+ }
+
+ afr_sh_erase_pending (frame, this, AFR_DATA_TRANSACTION,
+ afr_sh_data_erase_pending_cbk,
+ afr_sh_data_finish);
+}
int
-afr_self_heal_data (call_frame_t *frame, xlator_t *this)
+afr_sh_non_reg_lock_success (call_frame_t *frame, xlator_t *this)
{
afr_local_t *local = NULL;
afr_self_heal_t *sh = NULL;
- afr_private_t *priv = this->private;
local = frame->local;
sh = &local->self_heal;
+ sh->data_lock_held = _gf_true;
+ afr_sh_common_lookup (frame, this, &local->loc,
+ afr_sh_non_reg_fix, NULL,
+ AFR_LOOKUP_FAIL_CONFLICTS |
+ AFR_LOOKUP_FAIL_MISSING_GFIDS,
+ NULL);
+ return 0;
+}
+gf_boolean_t
+afr_can_start_data_self_heal (afr_self_heal_t *sh, afr_private_t *priv)
+{
+ if (sh->force_confirm_spb)
+ return _gf_true;
if (sh->do_data_self_heal &&
- afr_data_self_heal_enabled (priv->data_self_heal)) {
- afr_sh_data_open (frame, this);
+ afr_data_self_heal_enabled (priv->data_self_heal))
+ return _gf_true;
+ return _gf_false;
+}
+
+int
+afr_self_heal_data (call_frame_t *frame, xlator_t *this)
+{
+ afr_local_t *local = NULL;
+ afr_self_heal_t *sh = NULL;
+ afr_private_t *priv = this->private;
+ int ret = -1;
+
+ local = frame->local;
+ sh = &local->self_heal;
+
+ sh->sh_type_in_action = AFR_SELF_HEAL_DATA;
+
+ if (afr_can_start_data_self_heal (sh, priv)) {
+ afr_set_self_heal_status (sh, AFR_SELF_HEAL_STARTED);
+ ret = afr_inodelk_init (&local->internal_lock.inodelk[1],
+ priv->sh_domain, priv->child_count);
+ if (ret < 0) {
+ afr_set_self_heal_status (sh, AFR_SELF_HEAL_FAILED);
+ afr_sh_data_done (frame, this);
+ return 0;
+ }
+
+ if (IA_ISREG (sh->type)) {
+ afr_sh_data_open (frame, this);
+ } else {
+ afr_sh_data_lock (frame, this, 0, 0, _gf_true,
+ this->name,
+ afr_sh_non_reg_lock_success,
+ afr_sh_data_fail);
+ }
} else {
gf_log (this->name, GF_LOG_TRACE,
"not doing data self heal on %s",
diff --git a/xlators/cluster/afr/src/afr-self-heal-entry.c b/xlators/cluster/afr/src/afr-self-heal-entry.c
index ed1c51a21..53491a1d7 100644
--- a/xlators/cluster/afr/src/afr-self-heal-entry.c
+++ b/xlators/cluster/afr/src/afr-self-heal-entry.c
@@ -1,20 +1,11 @@
/*
- Copyright (c) 2008-2011 Gluster, Inc. <http://www.gluster.com>
+ Copyright (c) 2008-2012 Red Hat, Inc. <http://www.redhat.com>
This file is part of GlusterFS.
- GlusterFS is free software; you can redistribute it and/or modify
- it under the terms of the GNU General Public License as published
- by the Free Software Foundation; either version 3 of the License,
- or (at your option) any later version.
-
- GlusterFS is distributed in the hope that it will be useful, but
- WITHOUT ANY WARRANTY; without even the implied warranty of
- MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU
- General Public License for more details.
-
- You should have received a copy of the GNU General Public License
- along with this program. If not, see
- <http://www.gnu.org/licenses/>.
+ This file is licensed to you under your choice of the GNU Lesser
+ General Public License, version 3 or any later version (LGPLv3 or
+ later), or the GNU General Public License, version 2 (GPLv2), in all
+ cases as published by the Free Software Foundation.
*/
#include <libgen.h>
@@ -59,6 +50,9 @@
} while (0);
int
+afr_sh_entry_impunge_create_file (call_frame_t *impunge_frame, xlator_t *this,
+ int child_index);
+int
afr_sh_entry_done (call_frame_t *frame, xlator_t *this)
{
afr_local_t *local = NULL;
@@ -67,10 +61,6 @@ afr_sh_entry_done (call_frame_t *frame, xlator_t *this)
local = frame->local;
sh = &local->self_heal;
- if (sh->healing_fd)
- fd_unref (sh->healing_fd);
- sh->healing_fd = NULL;
-
sh->completion_cbk (frame, this);
return 0;
@@ -112,7 +102,7 @@ afr_sh_entry_finish (call_frame_t *frame, xlator_t *this)
int
afr_sh_entry_erase_pending_cbk (call_frame_t *frame, void *cookie,
xlator_t *this, int32_t op_ret,
- int32_t op_errno, dict_t *xattr)
+ int32_t op_errno, dict_t *xattr, dict_t *xdata)
{
long i = 0;
int call_count = 0;
@@ -167,72 +157,20 @@ afr_sh_entry_erase_pending (call_frame_t *frame, xlator_t *this)
{
afr_local_t *local = NULL;
afr_self_heal_t *sh = NULL;
- afr_private_t *priv = NULL;
- int call_count = 0;
- int i = 0;
- dict_t **erase_xattr = NULL;
- int need_unwind = 0;
local = frame->local;
sh = &local->self_heal;
- priv = this->private;
if (sh->entries_skipped) {
- need_unwind = 1;
- sh->op_failed = _gf_true;
+ afr_set_self_heal_status (sh, AFR_SELF_HEAL_FAILED);
goto out;
}
- afr_sh_pending_to_delta (priv, sh->xattr, sh->delta_matrix, sh->success,
- priv->child_count, AFR_ENTRY_TRANSACTION);
-
- erase_xattr = GF_CALLOC (sizeof (*erase_xattr), priv->child_count,
- gf_afr_mt_dict_t);
-
- for (i = 0; i < priv->child_count; i++) {
- if (sh->xattr[i]) {
- call_count++;
-
- erase_xattr[i] = get_new_dict();
- dict_ref (erase_xattr[i]);
- }
- }
-
- if (call_count == 0)
- need_unwind = 1;
-
- afr_sh_delta_to_xattr (priv, sh->delta_matrix, erase_xattr,
- priv->child_count, AFR_ENTRY_TRANSACTION);
-
- local->call_count = call_count;
- for (i = 0; i < priv->child_count; i++) {
- if (!erase_xattr[i])
- continue;
-
- gf_log (this->name, GF_LOG_TRACE,
- "erasing pending flags from %s on %s",
- local->loc.path, priv->children[i]->name);
-
- STACK_WIND_COOKIE (frame, afr_sh_entry_erase_pending_cbk,
- (void *) (long) i,
- priv->children[i],
- priv->children[i]->fops->xattrop,
- &local->loc,
- GF_XATTROP_ADD_ARRAY, erase_xattr[i]);
- if (!--call_count)
- break;
- }
-
- for (i = 0; i < priv->child_count; i++) {
- if (erase_xattr[i]) {
- dict_unref (erase_xattr[i]);
- }
- }
- GF_FREE (erase_xattr);
-
+ afr_sh_erase_pending (frame, this, AFR_ENTRY_TRANSACTION,
+ afr_sh_entry_erase_pending_cbk,
+ afr_sh_entry_finish);
+ return 0;
out:
- if (need_unwind)
- afr_sh_entry_finish (frame, this);
-
+ afr_sh_entry_finish (frame, this);
return 0;
}
@@ -317,8 +255,7 @@ int
afr_sh_entry_impunge_all (call_frame_t *frame, xlator_t *this);
int
-afr_sh_entry_impunge_subvol (call_frame_t *frame, xlator_t *this,
- int active_src);
+afr_sh_entry_impunge_subvol (call_frame_t *frame, xlator_t *this);
int
afr_sh_entry_expunge_all (call_frame_t *frame, xlator_t *this);
@@ -346,7 +283,8 @@ int
afr_sh_entry_expunge_parent_setattr_cbk (call_frame_t *expunge_frame,
void *cookie, xlator_t *this,
int32_t op_ret, int32_t op_errno,
- struct iatt *preop, struct iatt *postop)
+ struct iatt *preop, struct iatt *postop,
+ dict_t *xdata)
{
afr_private_t *priv = NULL;
afr_local_t *expunge_local = NULL;
@@ -382,7 +320,7 @@ afr_sh_entry_expunge_remove_cbk (call_frame_t *expunge_frame, void *cookie,
xlator_t *this,
int32_t op_ret, int32_t op_errno,
struct iatt *preparent,
- struct iatt *postparent)
+ struct iatt *postparent, dict_t *xdata)
{
afr_private_t *priv = NULL;
afr_local_t *expunge_local = NULL;
@@ -410,7 +348,6 @@ afr_sh_entry_expunge_remove_cbk (call_frame_t *expunge_frame, void *cookie,
}
valid = GF_SET_ATTR_ATIME | GF_SET_ATTR_MTIME;
- afr_build_parent_loc (&expunge_sh->parent_loc, &expunge_local->loc);
STACK_WIND_COOKIE (expunge_frame, afr_sh_entry_expunge_parent_setattr_cbk,
(void *) (long) active_src,
@@ -418,7 +355,7 @@ afr_sh_entry_expunge_remove_cbk (call_frame_t *expunge_frame, void *cookie,
priv->children[active_src]->fops->setattr,
&expunge_sh->parent_loc,
&expunge_sh->parentbuf,
- valid);
+ valid, NULL);
return 0;
}
@@ -442,7 +379,7 @@ afr_sh_entry_expunge_unlink (call_frame_t *expunge_frame, xlator_t *this,
(void *) (long) active_src,
priv->children[active_src],
priv->children[active_src]->fops->unlink,
- &expunge_local->loc);
+ &expunge_local->loc, 0, NULL);
return 0;
}
@@ -467,7 +404,7 @@ afr_sh_entry_expunge_rmdir (call_frame_t *expunge_frame, xlator_t *this,
(void *) (long) active_src,
priv->children[active_src],
priv->children[active_src]->fops->rmdir,
- &expunge_local->loc, 1);
+ &expunge_local->loc, 1, NULL);
return 0;
}
@@ -591,7 +528,7 @@ afr_sh_entry_expunge_purge (call_frame_t *expunge_frame, xlator_t *this,
(void *) (long) active_src,
priv->children[active_src],
priv->children[active_src]->fops->lookup,
- &expunge_local->loc, 0);
+ &expunge_local->loc, NULL);
return 0;
}
@@ -643,7 +580,8 @@ afr_sh_entry_expunge_entry_cbk (call_frame_t *expunge_frame, void *cookie,
if (need_expunge) {
gf_log (this->name, GF_LOG_INFO,
- "missing entry %s on %s",
+ "Entry %s is missing on %s and deleting from "
+ "replica's other bricks",
expunge_local->loc.path,
priv->children[source]->name);
@@ -675,6 +613,19 @@ out:
return 0;
}
+static gf_boolean_t
+can_skip_entry_self_heal (char *name, loc_t *parent_loc)
+{
+ if (strcmp (name, ".") == 0) {
+ return _gf_true;
+ } else if (strcmp (name, "..") == 0) {
+ return _gf_true;
+ } else if (loc_is_root (parent_loc) &&
+ (strcmp (name, GF_REPLICATE_TRASH_DIR) == 0)) {
+ return _gf_true;
+ }
+ return _gf_false;
+}
int
afr_sh_entry_expunge_entry (call_frame_t *frame, xlator_t *this,
@@ -702,15 +653,7 @@ afr_sh_entry_expunge_entry (call_frame_t *frame, xlator_t *this,
sh->expunge_done = afr_sh_entry_expunge_entry_done;
name = entry->d_name;
-
- if ((strcmp (name, ".") == 0)
- || (strcmp (name, "..") == 0)
- || ((strcmp (local->loc.path, "/") == 0)
- && (strcmp (name, GF_REPLICATE_TRASH_DIR) == 0))) {
-
- gf_log (this->name, GF_LOG_TRACE,
- "skipping inspection of %s under %s",
- name, local->loc.path);
+ if (can_skip_entry_self_heal (name, &local->loc)) {
op_ret = 0;
goto out;
}
@@ -725,16 +668,17 @@ afr_sh_entry_expunge_entry (call_frame_t *frame, xlator_t *this,
goto out;
}
- ALLOC_OR_GOTO (expunge_local, afr_local_t, out);
+ AFR_LOCAL_ALLOC_OR_GOTO (expunge_local, out);
expunge_frame->local = expunge_local;
expunge_sh = &expunge_local->self_heal;
expunge_sh->sh_frame = frame;
expunge_sh->active_source = active_src;
expunge_sh->entrybuf = entry->d_stat;
+ loc_copy (&expunge_sh->parent_loc, &local->loc);
ret = afr_build_child_loc (this, &expunge_local->loc, &local->loc,
- name, entry->d_stat.ia_gfid);
+ name);
if (ret != 0) {
op_errno = EINVAL;
goto out;
@@ -749,7 +693,7 @@ afr_sh_entry_expunge_entry (call_frame_t *frame, xlator_t *this,
(void *) (long) source,
priv->children[source],
priv->children[source]->fops->lookup,
- &expunge_local->loc, 0);
+ &expunge_local->loc, NULL);
ret = 0;
out:
@@ -764,7 +708,7 @@ int
afr_sh_entry_expunge_readdir_cbk (call_frame_t *frame, void *cookie,
xlator_t *this,
int32_t op_ret, int32_t op_errno,
- gf_dirent_t *entries)
+ gf_dirent_t *entries, dict_t *xdata)
{
afr_private_t *priv = NULL;
afr_local_t *local = NULL;
@@ -862,7 +806,7 @@ afr_sh_entry_expunge_all (call_frame_t *frame, xlator_t *this)
active_src = next_active_sink (frame, this, sh->active_source);
sh->active_source = active_src;
- if (sh->op_failed) {
+ if (is_self_heal_failed (sh, AFR_CHECK_SPECIFIC)) {
goto out;
}
@@ -887,15 +831,19 @@ out:
int
afr_sh_entry_impunge_entry_done (call_frame_t *frame, xlator_t *this,
- int active_src, int32_t op_ret,
- int32_t op_errno)
+ int32_t op_ret, int32_t op_errno)
{
int call_count = 0;
+ afr_local_t *local = NULL;
+ afr_self_heal_t *sh = NULL;
+ local = frame->local;
+ sh = &local->self_heal;
+ if (op_ret < 0)
+ sh->entries_skipped = _gf_true;
call_count = afr_frame_return (frame);
-
if (call_count == 0)
- afr_sh_entry_impunge_subvol (frame, this, active_src);
+ afr_sh_entry_impunge_subvol (frame, this);
return 0;
}
@@ -909,22 +857,20 @@ afr_sh_entry_call_impunge_done (call_frame_t *impunge_frame, xlator_t *this,
afr_self_heal_t *sh = NULL;
afr_self_heal_t *impunge_sh = NULL;
call_frame_t *frame = NULL;
- int32_t impunge_ret_child = 0;
AFR_INIT_SH_FRAME_VALS (impunge_frame, impunge_local, impunge_sh,
frame, local, sh);
- impunge_ret_child = impunge_sh->impunge_ret_child;
AFR_STACK_DESTROY (impunge_frame);
- sh->impunge_done (frame, this, impunge_ret_child, op_ret,
- op_errno);
+ sh->impunge_done (frame, this, op_ret, op_errno);
}
int
afr_sh_entry_impunge_setattr_cbk (call_frame_t *impunge_frame, void *cookie,
xlator_t *this,
int32_t op_ret, int32_t op_errno,
- struct iatt *preop, struct iatt *postop)
+ struct iatt *preop, struct iatt *postop,
+ dict_t *xdata)
{
int call_count = 0;
afr_private_t *priv = NULL;
@@ -936,7 +882,7 @@ afr_sh_entry_impunge_setattr_cbk (call_frame_t *impunge_frame, void *cookie,
child_index = (long) cookie;
if (op_ret == 0) {
- gf_log (this->name, GF_LOG_TRACE,
+ gf_log (this->name, GF_LOG_DEBUG,
"setattr done for %s on %s",
impunge_local->loc.path,
priv->children[child_index]->name);
@@ -948,31 +894,114 @@ afr_sh_entry_impunge_setattr_cbk (call_frame_t *impunge_frame, void *cookie,
strerror (op_errno));
}
- LOCK (&impunge_frame->lock);
- {
- call_count = --impunge_local->call_count;
+ call_count = afr_frame_return (impunge_frame);
+ if (call_count == 0) {
+ afr_sh_entry_call_impunge_done (impunge_frame, this,
+ 0, op_errno);
}
- UNLOCK (&impunge_frame->lock);
- if (call_count == 0)
- afr_sh_entry_call_impunge_done (impunge_frame, this,
- op_ret, op_errno);
+ return 0;
+}
+
+int
+afr_sh_entry_impunge_parent_setattr_cbk (call_frame_t *setattr_frame,
+ void *cookie, xlator_t *this,
+ int32_t op_ret, int32_t op_errno,
+ struct iatt *preop, struct iatt *postop,
+ dict_t *xdata)
+{
+ int call_count = 0;
+ afr_local_t *setattr_local = NULL;
+
+ setattr_local = setattr_frame->local;
+ if (op_ret != 0) {
+ gf_log (this->name, GF_LOG_INFO,
+ "setattr on parent directory (%s) failed: %s",
+ setattr_local->loc.path, strerror (op_errno));
+ }
+ call_count = afr_frame_return (setattr_frame);
+ if (call_count == 0)
+ AFR_STACK_DESTROY (setattr_frame);
return 0;
}
+int
+afr_sh_entry_impunge_setattr (call_frame_t *impunge_frame, xlator_t *this)
+{
+ afr_private_t *priv = NULL;
+ afr_local_t *impunge_local = NULL;
+ afr_local_t *setattr_local = NULL;
+ afr_self_heal_t *impunge_sh = NULL;
+ call_frame_t *setattr_frame = NULL;
+ int32_t valid = 0;
+ int32_t op_errno = 0;
+ int child_index = 0;
+ int call_count = 0;
+ int i = 0;
+
+ priv = this->private;
+ impunge_local = impunge_frame->local;
+ impunge_sh = &impunge_local->self_heal;
+
+ gf_log (this->name, GF_LOG_DEBUG,
+ "setting ownership of %s on %s to %d/%d",
+ impunge_local->loc.path,
+ priv->children[child_index]->name,
+ impunge_sh->entrybuf.ia_uid,
+ impunge_sh->entrybuf.ia_gid);
+
+ setattr_frame = copy_frame (impunge_frame);
+ if (!setattr_frame) {
+ op_errno = ENOMEM;
+ goto out;
+ }
+ AFR_LOCAL_ALLOC_OR_GOTO (setattr_frame->local, out);
+ setattr_local = setattr_frame->local;
+ call_count = afr_errno_count (NULL, impunge_sh->child_errno,
+ priv->child_count, 0);
+ loc_copy (&setattr_local->loc, &impunge_sh->parent_loc);
+ impunge_local->call_count = call_count;
+ setattr_local->call_count = call_count;
+ for (i = 0; i < priv->child_count; i++) {
+ if (impunge_sh->child_errno[i])
+ continue;
+ valid = GF_SET_ATTR_ATIME | GF_SET_ATTR_MTIME;
+ STACK_WIND_COOKIE (setattr_frame,
+ afr_sh_entry_impunge_parent_setattr_cbk,
+ (void *) (long) i, priv->children[i],
+ priv->children[i]->fops->setattr,
+ &setattr_local->loc,
+ &impunge_sh->parentbuf, valid, NULL);
+
+ valid = GF_SET_ATTR_UID | GF_SET_ATTR_GID |
+ GF_SET_ATTR_ATIME | GF_SET_ATTR_MTIME;
+ STACK_WIND_COOKIE (impunge_frame,
+ afr_sh_entry_impunge_setattr_cbk,
+ (void *) (long) i, priv->children[i],
+ priv->children[i]->fops->setattr,
+ &impunge_local->loc,
+ &impunge_sh->entrybuf, valid, NULL);
+ call_count--;
+ }
+ GF_ASSERT (!call_count);
+ return 0;
+out:
+ if (setattr_frame)
+ AFR_STACK_DESTROY (setattr_frame);
+ afr_sh_entry_call_impunge_done (impunge_frame, this, 0, op_errno);
+ return 0;
+}
int
afr_sh_entry_impunge_xattrop_cbk (call_frame_t *impunge_frame, void *cookie,
xlator_t *this,
int32_t op_ret, int32_t op_errno,
- dict_t *xattr)
+ dict_t *xattr, dict_t *xdata)
{
afr_private_t *priv = NULL;
afr_local_t *impunge_local = NULL;
int child_index = 0;
- struct iatt stbuf = {0};
- int32_t valid = 0;
priv = this->private;
impunge_local = impunge_frame->local;
@@ -985,55 +1014,59 @@ afr_sh_entry_impunge_xattrop_cbk (call_frame_t *impunge_frame, void *cookie,
impunge_local->loc.path,
priv->children[child_index]->name,
strerror (op_errno));
+ goto out;
}
- gf_log (this->name, GF_LOG_TRACE,
- "setting ownership of %s on %s to %d/%d",
- impunge_local->loc.path,
- priv->children[child_index]->name,
- impunge_local->cont.lookup.buf.ia_uid,
- impunge_local->cont.lookup.buf.ia_gid);
-
- stbuf.ia_atime = impunge_local->cont.lookup.buf.ia_atime;
- stbuf.ia_atime_nsec = impunge_local->cont.lookup.buf.ia_atime_nsec;
- stbuf.ia_mtime = impunge_local->cont.lookup.buf.ia_mtime;
- stbuf.ia_mtime_nsec = impunge_local->cont.lookup.buf.ia_mtime_nsec;
-
- stbuf.ia_uid = impunge_local->cont.lookup.buf.ia_uid;
- stbuf.ia_gid = impunge_local->cont.lookup.buf.ia_gid;
-
- valid = GF_SET_ATTR_UID | GF_SET_ATTR_GID |
- GF_SET_ATTR_ATIME | GF_SET_ATTR_MTIME;
-
- STACK_WIND_COOKIE (impunge_frame, afr_sh_entry_impunge_setattr_cbk,
- (void *) (long) child_index,
- priv->children[child_index],
- priv->children[child_index]->fops->setattr,
- &impunge_local->loc,
- &stbuf, valid);
+ afr_sh_entry_impunge_setattr (impunge_frame, this);
+ return 0;
+out:
+ afr_sh_entry_call_impunge_done (impunge_frame, this,
+ -1, op_errno);
return 0;
}
-
int
-afr_sh_entry_impunge_parent_setattr_cbk (call_frame_t *setattr_frame,
- void *cookie, xlator_t *this,
- int32_t op_ret, int32_t op_errno,
- struct iatt *preop, struct iatt *postop)
+afr_sh_entry_impunge_perform_xattrop (call_frame_t *impunge_frame,
+ xlator_t *this)
{
- loc_t *parent_loc = cookie;
+ int active_src = 0;
+ dict_t *xattr = NULL;
+ afr_private_t *priv = NULL;
+ afr_local_t *impunge_local = NULL;
+ afr_self_heal_t *impunge_sh = NULL;
+ int32_t op_errno = 0;
- if (op_ret != 0) {
- gf_log (this->name, GF_LOG_INFO,
- "setattr on parent directory (%s) failed: %s",
- parent_loc->path, strerror (op_errno));
+ priv = this->private;
+ impunge_local = impunge_frame->local;
+ impunge_sh = &impunge_local->self_heal;
+ active_src = impunge_sh->active_source;
+
+ afr_prepare_new_entry_pending_matrix (impunge_local->pending,
+ afr_is_errno_unset,
+ impunge_sh->child_errno,
+ &impunge_sh->entrybuf,
+ priv->child_count);
+ xattr = dict_new ();
+ if (!xattr) {
+ op_errno = ENOMEM;
+ goto out;
}
- loc_wipe (parent_loc);
+ afr_set_pending_dict (priv, xattr, impunge_local->pending, active_src,
+ LOCAL_LAST);
- GF_FREE (parent_loc);
+ STACK_WIND_COOKIE (impunge_frame, afr_sh_entry_impunge_xattrop_cbk,
+ (void *) (long) active_src,
+ priv->children[active_src],
+ priv->children[active_src]->fops->xattrop,
+ &impunge_local->loc, GF_XATTROP_ADD_ARRAY, xattr, NULL);
- AFR_STACK_DESTROY (setattr_frame);
+ if (xattr)
+ dict_unref (xattr);
+ return 0;
+out:
+ afr_sh_entry_call_impunge_done (impunge_frame, this,
+ -1, op_errno);
return 0;
}
@@ -1043,124 +1076,165 @@ afr_sh_entry_impunge_newfile_cbk (call_frame_t *impunge_frame, void *cookie,
int32_t op_ret, int32_t op_errno,
inode_t *inode, struct iatt *stbuf,
struct iatt *preparent,
- struct iatt *postparent)
+ struct iatt *postparent, dict_t *xdata)
{
int call_count = 0;
afr_private_t *priv = NULL;
afr_local_t *impunge_local = NULL;
afr_self_heal_t *impunge_sh = NULL;
- int active_src = 0;
int child_index = 0;
- int32_t *pending_array = NULL;
- dict_t *xattr = NULL;
- int ret = 0;
- int idx = 0;
- call_frame_t *setattr_frame = NULL;
- int32_t valid = 0;
- loc_t *parent_loc = NULL;
- struct iatt parentbuf = {0,};
priv = this->private;
impunge_local = impunge_frame->local;
impunge_sh = &impunge_local->self_heal;
- active_src = impunge_sh->active_source;
child_index = (long) cookie;
if (op_ret == -1) {
- ret = -1;
+ impunge_sh->child_errno[child_index] = op_errno;
gf_log (this->name, GF_LOG_ERROR,
"creation of %s on %s failed (%s)",
impunge_local->loc.path,
priv->children[child_index]->name,
strerror (op_errno));
- goto out;
+ } else {
+ impunge_sh->child_errno[child_index] = 0;
}
- inode->ia_type = stbuf->ia_type;
-
- xattr = dict_new ();
- if (!xattr) {
- ret = -1;
- goto out;
+ call_count = afr_frame_return (impunge_frame);
+ if (call_count == 0) {
+ if (!afr_errno_count (NULL, impunge_sh->child_errno,
+ priv->child_count, 0)) {
+ // new_file creation failed every where
+ afr_sh_entry_call_impunge_done (impunge_frame, this,
+ -1, op_errno);
+ goto out;
+ }
+ afr_sh_entry_impunge_perform_xattrop (impunge_frame, this);
}
+out:
+ return 0;
+}
- pending_array = (int32_t*) GF_CALLOC (3, sizeof (*pending_array),
- gf_afr_mt_int32_t);
+int
+afr_sh_entry_impunge_hardlink_cbk (call_frame_t *impunge_frame, void *cookie,
+ xlator_t *this, int32_t op_ret,
+ int32_t op_errno, inode_t *inode,
+ struct iatt *buf, struct iatt *preparent,
+ struct iatt *postparent, dict_t *xdata)
+{
+ int call_count = 0;
+ afr_local_t *impunge_local = NULL;
+ afr_self_heal_t *impunge_sh = NULL;
- if (!pending_array) {
- ret = -1;
- goto out;
- }
+ impunge_local = impunge_frame->local;
+ impunge_sh = &impunge_local->self_heal;
- /* Pending data xattrs shouldn't be set for special files
- */
- idx = afr_index_for_transaction_type (AFR_METADATA_TRANSACTION);
- pending_array[idx] = hton32 (1);
- if (IA_ISDIR (stbuf->ia_type))
- idx = afr_index_for_transaction_type (AFR_ENTRY_TRANSACTION);
- else if (IA_ISREG (stbuf->ia_type))
- idx = afr_index_for_transaction_type (AFR_DATA_TRANSACTION);
- else
- goto cont;
- pending_array[idx] = hton32 (1);
-
-cont:
- ret = dict_set_dynptr (xattr, priv->pending_key[child_index],
- pending_array,
- 3 * sizeof (*pending_array));
- if (ret < 0) {
- gf_log (this->name, GF_LOG_WARNING,
- "Unable to set dict value.");
- } else {
- pending_array = NULL;
+ if (IA_IFLNK == impunge_sh->entrybuf.ia_type) {
+ //For symlinks impunge is attempted un-conditionally
+ //So the file can already exist.
+ if ((op_ret < 0) && (op_errno == EEXIST))
+ op_ret = 0;
}
- valid = GF_SET_ATTR_ATIME | GF_SET_ATTR_MTIME;
- parentbuf = impunge_sh->parentbuf;
- setattr_frame = copy_frame (impunge_frame);
- parent_loc = GF_CALLOC (1, sizeof (*parent_loc),
- gf_afr_mt_loc_t);
- if (!parent_loc) {
- ret = -1;
- goto out;
- }
- afr_build_parent_loc (parent_loc, &impunge_local->loc);
+ call_count = afr_frame_return (impunge_frame);
+ if (call_count == 0)
+ afr_sh_entry_call_impunge_done (impunge_frame, this,
+ op_ret, op_errno);
- STACK_WIND_COOKIE (impunge_frame, afr_sh_entry_impunge_xattrop_cbk,
- (void *) (long) child_index,
- priv->children[active_src],
- priv->children[active_src]->fops->xattrop,
- &impunge_local->loc, GF_XATTROP_ADD_ARRAY, xattr);
+ return 0;
+}
- STACK_WIND_COOKIE (setattr_frame, afr_sh_entry_impunge_parent_setattr_cbk,
- (void *) (long) parent_loc,
- priv->children[child_index],
- priv->children[child_index]->fops->setattr,
- parent_loc, &parentbuf, valid);
+int
+afr_sh_entry_impunge_hardlink (call_frame_t *impunge_frame, xlator_t *this,
+ int child_index)
+{
+ afr_private_t *priv = NULL;
+ afr_local_t *impunge_local = NULL;
+ afr_self_heal_t *impunge_sh = NULL;
+ loc_t *loc = NULL;
+ struct iatt *buf = NULL;
+ loc_t oldloc = {0};
-out:
- if (xattr)
- dict_unref (xattr);
+ priv = this->private;
+ impunge_local = impunge_frame->local;
+ impunge_sh = &impunge_local->self_heal;
+ loc = &impunge_local->loc;
+ buf = &impunge_sh->entrybuf;
- if (ret) {
- if (pending_array)
- GF_FREE (pending_array);
+ oldloc.inode = inode_ref (loc->inode);
+ uuid_copy (oldloc.gfid, buf->ia_gfid);
+ gf_log (this->name, GF_LOG_DEBUG, "linking missing file %s on %s",
+ loc->path, priv->children[child_index]->name);
- LOCK (&impunge_frame->lock);
- {
- call_count = --impunge_local->call_count;
- }
- UNLOCK (&impunge_frame->lock);
+ STACK_WIND_COOKIE (impunge_frame, afr_sh_entry_impunge_hardlink_cbk,
+ (void *) (long) child_index,
+ priv->children[child_index],
+ priv->children[child_index]->fops->link,
+ &oldloc, loc, NULL);
+ loc_wipe (&oldloc);
- if (call_count == 0)
- afr_sh_entry_call_impunge_done (impunge_frame, this,
- -1, op_errno);
- }
+ return 0;
+}
+int
+afr_sh_nameless_lookup_cbk (call_frame_t *impunge_frame, void *cookie,
+ xlator_t *this,
+ int32_t op_ret, int32_t op_errno, inode_t *inode,
+ struct iatt *buf, dict_t *xattr,
+ struct iatt *postparent)
+{
+ if (op_ret < 0) {
+ afr_sh_entry_impunge_create_file (impunge_frame, this,
+ (long)cookie);
+ } else {
+ afr_sh_entry_impunge_hardlink (impunge_frame, this,
+ (long)cookie);
+ }
return 0;
}
+int
+afr_sh_entry_impunge_check_hardlink (call_frame_t *impunge_frame,
+ xlator_t *this,
+ int child_index, struct iatt *stbuf)
+{
+ afr_private_t *priv = NULL;
+ call_frame_t *frame = NULL;
+ afr_local_t *impunge_local = NULL;
+ afr_local_t *local = NULL;
+ afr_self_heal_t *impunge_sh = NULL;
+ afr_self_heal_t *sh = NULL;
+ loc_t *loc = NULL;
+ dict_t *xattr_req = NULL;
+ loc_t oldloc = {0};
+ int ret = -1;
+
+ priv = this->private;
+ AFR_INIT_SH_FRAME_VALS (impunge_frame, impunge_local, impunge_sh,
+ frame, local, sh);
+ loc = &impunge_local->loc;
+
+ xattr_req = dict_new ();
+ if (!xattr_req)
+ goto out;
+ oldloc.inode = inode_ref (loc->inode);
+ uuid_copy (oldloc.gfid, stbuf->ia_gfid);
+
+ STACK_WIND_COOKIE (impunge_frame, afr_sh_nameless_lookup_cbk,
+ (void *) (long) child_index,
+ priv->children[child_index],
+ priv->children[child_index]->fops->lookup,
+ &oldloc, xattr_req);
+ ret = 0;
+out:
+ if (xattr_req)
+ dict_unref (xattr_req);
+ loc_wipe (&oldloc);
+ if (ret)
+ sh->impunge_done (frame, this, -1, ENOMEM);
+ return 0;
+}
int
afr_sh_entry_impunge_mknod (call_frame_t *impunge_frame, xlator_t *this,
@@ -1189,6 +1263,35 @@ afr_sh_entry_impunge_mknod (call_frame_t *impunge_frame, xlator_t *this,
gf_log (this->name, GF_LOG_INFO, "%s: gfid set failed",
impunge_local->loc.path);
+ /*
+ * Reason for adding GLUSTERFS_INTERNAL_FOP_KEY :
+ *
+ * Problem:
+ * While a brick is down in a replica pair, lets say the user creates
+ * one file(file-A) and a hard link to that file(h-file-A). After the
+ * brick comes back up, entry self-heal is attempted on parent dir of
+ * these two files. As part of readdir in self-heal it reads both the
+ * entries file-A and h-file-A for both of them it does name less lookup
+ * to check if there are any hardlinks already present in the
+ * destination brick. It finds that there are no hard links already
+ * present for files file-A, h-file-A. Self-heal does mknods for both
+ * file-A and h-file-A. This leads to file-A and h-file-A not being
+ * hardlinks anymore.
+ *
+ * Fix: (More like shrinking of race-window, the race itself is still
+ * present in posix-mknod).
+ * If mknod comes with the presence of GLUSTERFS_INTERNAL_FOP_KEY then
+ * posix_mknod checks if there are already any gfid-links and does
+ * link() instead of mknod. There still can be a race where two
+ * posix_mknods same gfid see that
+ * gfid-link file is not present and proceeds with mknods and result in
+ * two different files with same gfid.
+ */
+ ret = dict_set_str (dict, GLUSTERFS_INTERNAL_FOP_KEY, "yes");
+ if (ret)
+ gf_log (this->name, GF_LOG_INFO, "%s: %s set failed",
+ impunge_local->loc.path, GLUSTERFS_INTERNAL_FOP_KEY);
+
STACK_WIND_COOKIE (impunge_frame, afr_sh_entry_impunge_newfile_cbk,
(void *) (long) child_index,
priv->children[child_index],
@@ -1196,7 +1299,7 @@ afr_sh_entry_impunge_mknod (call_frame_t *impunge_frame, xlator_t *this,
&impunge_local->loc,
st_mode_from_ia (stbuf->ia_prot, stbuf->ia_type),
makedev (ia_major (stbuf->ia_rdev),
- ia_minor (stbuf->ia_rdev)), dict);
+ ia_minor (stbuf->ia_rdev)), 0, dict);
if (dict)
dict_unref (dict);
@@ -1243,7 +1346,7 @@ afr_sh_entry_impunge_mkdir (call_frame_t *impunge_frame, xlator_t *this,
priv->children[child_index]->fops->mkdir,
&impunge_local->loc,
st_mode_from_ia (stbuf->ia_prot, stbuf->ia_type),
- dict);
+ 0, dict);
if (dict)
dict_unref (dict);
@@ -1265,7 +1368,7 @@ afr_sh_entry_impunge_symlink (call_frame_t *impunge_frame, xlator_t *this,
priv = this->private;
impunge_local = impunge_frame->local;
- buf = &impunge_local->cont.symlink.buf;
+ buf = &impunge_local->cont.dir_fop.buf;
dict = dict_new ();
if (!dict) {
@@ -1290,7 +1393,7 @@ afr_sh_entry_impunge_symlink (call_frame_t *impunge_frame, xlator_t *this,
(void *) (long) child_index,
priv->children[child_index],
priv->children[child_index]->fops->symlink,
- linkname, &impunge_local->loc, dict);
+ linkname, &impunge_local->loc, 0, dict);
if (dict)
dict_unref (dict);
@@ -1304,7 +1407,7 @@ afr_sh_entry_impunge_symlink_unlink_cbk (call_frame_t *impunge_frame,
void *cookie, xlator_t *this,
int32_t op_ret, int32_t op_errno,
struct iatt *preparent,
- struct iatt *postparent)
+ struct iatt *postparent, dict_t *xdata)
{
afr_private_t *priv = NULL;
afr_local_t *impunge_local = NULL;
@@ -1365,7 +1468,7 @@ afr_sh_entry_impunge_symlink_unlink (call_frame_t *impunge_frame, xlator_t *this
(void *) (long) child_index,
priv->children[child_index],
priv->children[child_index]->fops->unlink,
- &impunge_local->loc);
+ &impunge_local->loc, 0, NULL);
return 0;
}
@@ -1375,7 +1478,7 @@ int
afr_sh_entry_impunge_readlink_sink_cbk (call_frame_t *impunge_frame, void *cookie,
xlator_t *this,
int32_t op_ret, int32_t op_errno,
- const char *linkname, struct iatt *sbuf)
+ const char *linkname, struct iatt *sbuf, dict_t *xdata)
{
afr_private_t *priv = NULL;
afr_local_t *impunge_local = NULL;
@@ -1457,7 +1560,7 @@ afr_sh_entry_impunge_readlink_sink (call_frame_t *impunge_frame, xlator_t *this,
(void *) (long) child_index,
priv->children[child_index],
priv->children[child_index]->fops->readlink,
- &impunge_local->loc, 4096);
+ &impunge_local->loc, 4096, NULL);
return 0;
}
@@ -1467,7 +1570,7 @@ int
afr_sh_entry_impunge_readlink_cbk (call_frame_t *impunge_frame, void *cookie,
xlator_t *this,
int32_t op_ret, int32_t op_errno,
- const char *linkname, struct iatt *sbuf)
+ const char *linkname, struct iatt *sbuf, dict_t *xdata)
{
afr_private_t *priv = NULL;
afr_local_t *impunge_local = NULL;
@@ -1525,36 +1628,84 @@ afr_sh_entry_impunge_readlink (call_frame_t *impunge_frame, xlator_t *this,
impunge_local = impunge_frame->local;
impunge_sh = &impunge_local->self_heal;
active_src = impunge_sh->active_source;
- impunge_local->cont.symlink.buf = *stbuf;
+ impunge_local->cont.dir_fop.buf = *stbuf;
STACK_WIND_COOKIE (impunge_frame, afr_sh_entry_impunge_readlink_cbk,
(void *) (long) child_index,
priv->children[active_src],
priv->children[active_src]->fops->readlink,
- &impunge_local->loc, 4096);
+ &impunge_local->loc, 4096, NULL);
return 0;
}
int
afr_sh_entry_impunge_create (call_frame_t *impunge_frame, xlator_t *this,
- int child_index, struct iatt *buf,
- struct iatt *postparent)
+ int child_index)
{
- afr_local_t *impunge_local = NULL;
- afr_self_heal_t *impunge_sh = NULL;
+ call_frame_t *frame = NULL;
+ afr_local_t *impunge_local = NULL;
+ afr_local_t *local = NULL;
+ afr_self_heal_t *impunge_sh = NULL;
+ afr_self_heal_t *sh = NULL;
afr_private_t *priv = NULL;
ia_type_t type = IA_INVAL;
- int ret = 0;
int active_src = 0;
+ struct iatt *buf = NULL;
- impunge_local = impunge_frame->local;
- impunge_sh = &impunge_local->self_heal;
- impunge_sh->parentbuf = *postparent;
+ AFR_INIT_SH_FRAME_VALS (impunge_frame, impunge_local, impunge_sh,
+ frame, local, sh);
active_src = impunge_sh->active_source;
- impunge_local->cont.lookup.buf = *buf;
- afr_update_loc_gfids (&impunge_local->loc, buf, postparent);
+ afr_update_loc_gfids (&impunge_local->loc, &impunge_sh->entrybuf,
+ &impunge_sh->parentbuf);
+
+ buf = &impunge_sh->entrybuf;
+ type = buf->ia_type;
+
+ switch (type) {
+ case IA_IFSOCK:
+ case IA_IFREG:
+ case IA_IFBLK:
+ case IA_IFCHR:
+ case IA_IFIFO:
+ case IA_IFLNK:
+ afr_sh_entry_impunge_check_hardlink (impunge_frame, this,
+ child_index, buf);
+ break;
+ case IA_IFDIR:
+ afr_sh_entry_impunge_mkdir (impunge_frame, this,
+ child_index, buf);
+ break;
+ default:
+ gf_log (this->name, GF_LOG_ERROR,
+ "%s has unknown file type on %s: 0%o",
+ impunge_local->loc.path,
+ priv->children[active_src]->name, type);
+ sh->impunge_done (frame, this, -1, EINVAL);
+ break;
+ }
+
+ return 0;
+}
+
+int
+afr_sh_entry_impunge_create_file (call_frame_t *impunge_frame, xlator_t *this,
+ int child_index)
+{
+ call_frame_t *frame = NULL;
+ afr_local_t *impunge_local = NULL;
+ afr_local_t *local = NULL;
+ afr_self_heal_t *impunge_sh = NULL;
+ afr_self_heal_t *sh = NULL;
+ afr_private_t *priv = NULL;
+ ia_type_t type = IA_INVAL;
+ int active_src = 0;
+ struct iatt *buf = NULL;
+ AFR_INIT_SH_FRAME_VALS (impunge_frame, impunge_local, impunge_sh,
+ frame, local, sh);
+ active_src = impunge_sh->active_source;
+ buf = &impunge_sh->entrybuf;
type = buf->ia_type;
switch (type) {
@@ -1570,41 +1721,30 @@ afr_sh_entry_impunge_create (call_frame_t *impunge_frame, xlator_t *this,
afr_sh_entry_impunge_readlink (impunge_frame, this,
child_index, buf);
break;
- case IA_IFDIR:
- afr_sh_entry_impunge_mkdir (impunge_frame, this,
- child_index, buf);
- break;
default:
gf_log (this->name, GF_LOG_ERROR,
"%s has unknown file type on %s: 0%o",
impunge_local->loc.path,
priv->children[active_src]->name, type);
- ret = -1;
+ sh->impunge_done (frame, this, -1, EINVAL);
break;
}
- return ret;
+ return 0;
}
gf_boolean_t
-afr_sh_need_recreate (afr_self_heal_t *impunge_sh, int *sources,
- unsigned int child, unsigned int child_count)
+afr_sh_need_recreate (afr_self_heal_t *impunge_sh, unsigned int child,
+ unsigned int child_count)
{
- int32_t *success_children = NULL;
gf_boolean_t recreate = _gf_false;
- GF_ASSERT (impunge_sh->impunging_entry_mode);
GF_ASSERT (impunge_sh->child_errno);
- GF_ASSERT (sources);
- success_children = impunge_sh->success_children;
- if (child == impunge_sh->active_source) {
- GF_ASSERT (afr_is_child_present (success_children,
- child_count, child));
+ if (child == impunge_sh->active_source)
goto out;
- }
- if (IA_ISLNK (impunge_sh->impunging_entry_mode)) {
+ if (IA_IFLNK == impunge_sh->entrybuf.ia_type) {
recreate = _gf_true;
goto out;
}
@@ -1623,7 +1763,7 @@ afr_sh_recreate_count (afr_self_heal_t *impunge_sh, int *sources,
int i = 0;
for (i = 0; i < child_count; i++) {
- if (afr_sh_need_recreate (impunge_sh, sources, i, child_count))
+ if (afr_sh_need_recreate (impunge_sh, i, child_count))
count++;
}
@@ -1640,8 +1780,6 @@ afr_sh_entry_call_impunge_recreate (call_frame_t *impunge_frame,
call_frame_t *frame = NULL;
afr_local_t *local = NULL;
afr_self_heal_t *sh = NULL;
- struct iatt *buf = NULL;
- struct iatt *postparent = NULL;
unsigned int recreate_count = 0;
int i = 0;
int active_src = 0;
@@ -1649,24 +1787,34 @@ afr_sh_entry_call_impunge_recreate (call_frame_t *impunge_frame,
priv = this->private;
AFR_INIT_SH_FRAME_VALS (impunge_frame, impunge_local, impunge_sh,
frame, local, sh);
- active_src = impunge_sh->active_source;
- buf = &impunge_sh->buf[active_src];
- postparent = &impunge_sh->parentbufs[active_src];
-
+ active_src = impunge_sh->active_source;
+ impunge_sh->entrybuf = impunge_sh->buf[active_src];
+ impunge_sh->parentbuf = impunge_sh->parentbufs[active_src];
recreate_count = afr_sh_recreate_count (impunge_sh, sh->sources,
priv->child_count);
- GF_ASSERT (recreate_count);
+ if (!recreate_count) {
+ afr_sh_entry_call_impunge_done (impunge_frame, this, 0, 0);
+ goto out;
+ }
impunge_local->call_count = recreate_count;
for (i = 0; i < priv->child_count; i++) {
- if (afr_sh_need_recreate (impunge_sh, sh->sources, i,
- priv->child_count)) {
- (void)afr_sh_entry_impunge_create (impunge_frame, this,
- i, buf,
- postparent);
- recreate_count--;
+ if (!impunge_local->child_up[i]) {
+ impunge_sh->child_errno[i] = ENOTCONN;
+ continue;
+ }
+ if (!afr_sh_need_recreate (impunge_sh, i, priv->child_count)) {
+ impunge_sh->child_errno[i] = EEXIST;
+ continue;
}
}
+ for (i = 0; i < priv->child_count; i++) {
+ if (!afr_sh_need_recreate (impunge_sh, i, priv->child_count))
+ continue;
+ (void)afr_sh_entry_impunge_create (impunge_frame, this, i);
+ recreate_count--;
+ }
GF_ASSERT (!recreate_count);
+out:
return 0;
}
@@ -1680,7 +1828,6 @@ afr_sh_entry_common_lookup_done (call_frame_t *impunge_frame, xlator_t *this,
call_frame_t *frame = NULL;
afr_local_t *local = NULL;
afr_self_heal_t *sh = NULL;
- unsigned int recreate_count = 0;
unsigned int gfid_miss_count = 0;
unsigned int children_up_count = 0;
uuid_t gfid = {0};
@@ -1729,15 +1876,9 @@ afr_sh_entry_common_lookup_done (call_frame_t *impunge_frame, xlator_t *this,
afr_sh_common_lookup (impunge_frame, this, &impunge_local->loc,
afr_sh_entry_common_lookup_done, gfid,
AFR_LOOKUP_FAIL_CONFLICTS |
- AFR_LOOKUP_FAIL_MISSING_GFIDS);
+ AFR_LOOKUP_FAIL_MISSING_GFIDS,
+ NULL);
} else {
- recreate_count = afr_sh_recreate_count (impunge_sh, sh->sources,
- priv->child_count);
- if (!recreate_count) {
- op_ret = 0;
- op_errno = 0;
- goto done;
- }
afr_sh_entry_call_impunge_recreate (impunge_frame, this);
}
return;
@@ -1753,13 +1894,13 @@ afr_sh_entry_impunge_entry (call_frame_t *frame, xlator_t *this,
{
afr_local_t *local = NULL;
afr_self_heal_t *sh = NULL;
+ afr_self_heal_t *impunge_sh = NULL;
int ret = -1;
call_frame_t *impunge_frame = NULL;
afr_local_t *impunge_local = NULL;
int active_src = 0;
int op_errno = 0;
int op_ret = -1;
- mode_t entry_mode = 0;
local = frame->local;
sh = &local->self_heal;
@@ -1767,14 +1908,7 @@ afr_sh_entry_impunge_entry (call_frame_t *frame, xlator_t *this,
active_src = sh->active_source;
sh->impunge_done = afr_sh_entry_impunge_entry_done;
- if ((strcmp (entry->d_name, ".") == 0)
- || (strcmp (entry->d_name, "..") == 0)
- || ((strcmp (local->loc.path, "/") == 0)
- && (strcmp (entry->d_name, GF_REPLICATE_TRASH_DIR) == 0))) {
-
- gf_log (this->name, GF_LOG_TRACE,
- "skipping inspection of %s under %s",
- entry->d_name, local->loc.path);
+ if (can_skip_entry_self_heal (entry->d_name, &local->loc)) {
op_ret = 0;
goto out;
}
@@ -1783,18 +1917,18 @@ afr_sh_entry_impunge_entry (call_frame_t *frame, xlator_t *this,
"inspecting existence of %s under %s",
entry->d_name, local->loc.path);
- entry_mode = st_mode_from_ia (entry->d_stat.ia_prot,
- entry->d_stat.ia_type);
- ret = afr_impunge_frame_create (frame, this, active_src, active_src,
- entry_mode, &impunge_frame);
+ ret = afr_impunge_frame_create (frame, this, active_src,
+ &impunge_frame);
if (ret) {
op_errno = -ret;
goto out;
}
impunge_local = impunge_frame->local;
+ impunge_sh = &impunge_local->self_heal;
ret = afr_build_child_loc (this, &impunge_local->loc, &local->loc,
- entry->d_name, entry->d_stat.ia_gfid);
+ entry->d_name);
+ loc_copy (&impunge_sh->parent_loc, &local->loc);
if (ret != 0) {
op_errno = ENOMEM;
goto out;
@@ -1802,14 +1936,14 @@ afr_sh_entry_impunge_entry (call_frame_t *frame, xlator_t *this,
afr_sh_common_lookup (impunge_frame, this, &impunge_local->loc,
afr_sh_entry_common_lookup_done, NULL,
- AFR_LOOKUP_FAIL_CONFLICTS);
+ AFR_LOOKUP_FAIL_CONFLICTS, NULL);
op_ret = 0;
out:
if (ret) {
if (impunge_frame)
AFR_STACK_DESTROY (impunge_frame);
- sh->impunge_done (frame, this, active_src, op_ret, op_errno);
+ sh->impunge_done (frame, this, op_ret, op_errno);
}
return 0;
@@ -1820,7 +1954,7 @@ int
afr_sh_entry_impunge_readdir_cbk (call_frame_t *frame, void *cookie,
xlator_t *this,
int32_t op_ret, int32_t op_errno,
- gf_dirent_t *entries)
+ gf_dirent_t *entries, dict_t *xdata)
{
afr_private_t *priv = NULL;
afr_local_t *local = NULL;
@@ -1843,6 +1977,7 @@ afr_sh_entry_impunge_readdir_cbk (call_frame_t *frame, void *cookie,
local->loc.path,
priv->children[active_src]->name,
strerror (op_errno));
+ afr_set_self_heal_status (sh, AFR_SELF_HEAL_FAILED);
} else {
gf_log (this->name, GF_LOG_TRACE,
"readdir of %s on subvolume %s complete",
@@ -1859,7 +1994,7 @@ afr_sh_entry_impunge_readdir_cbk (call_frame_t *frame, void *cookie,
entry_count++;
}
- gf_log (this->name, GF_LOG_TRACE,
+ gf_log (this->name, GF_LOG_DEBUG,
"readdir'ed %d entries from %s",
entry_count, priv->children[active_src]->name);
@@ -1875,21 +2010,24 @@ afr_sh_entry_impunge_readdir_cbk (call_frame_t *frame, void *cookie,
int
-afr_sh_entry_impunge_subvol (call_frame_t *frame, xlator_t *this,
- int active_src)
+afr_sh_entry_impunge_subvol (call_frame_t *frame, xlator_t *this)
{
afr_private_t *priv = NULL;
afr_local_t *local = NULL;
afr_self_heal_t *sh = NULL;
+ int32_t active_src = 0;
priv = this->private;
local = frame->local;
sh = &local->self_heal;
+ active_src = sh->active_source;
+ gf_log (this->name, GF_LOG_DEBUG, "%s: readdir from offset %zd",
+ local->loc.path, sh->offset);
STACK_WIND (frame, afr_sh_entry_impunge_readdir_cbk,
priv->children[active_src],
priv->children[active_src]->fops->readdirp,
- sh->healing_fd, sh->block_size, sh->offset, 0);
+ sh->healing_fd, sh->block_size, sh->offset, NULL);
return 0;
}
@@ -1912,7 +2050,7 @@ afr_sh_entry_impunge_all (call_frame_t *frame, xlator_t *this)
active_src = next_active_source (frame, this, sh->active_source);
sh->active_source = active_src;
- if (sh->op_failed) {
+ if (is_self_heal_failed (sh, AFR_CHECK_SPECIFIC)) {
afr_sh_entry_finish (frame, this);
return 0;
}
@@ -1927,7 +2065,7 @@ afr_sh_entry_impunge_all (call_frame_t *frame, xlator_t *this)
"impunging entries of %s on %s to other sinks",
local->loc.path, priv->children[active_src]->name);
- afr_sh_entry_impunge_subvol (frame, this, active_src);
+ afr_sh_entry_impunge_subvol (frame, this);
return 0;
}
@@ -1935,7 +2073,7 @@ afr_sh_entry_impunge_all (call_frame_t *frame, xlator_t *this)
int
afr_sh_entry_opendir_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
- int32_t op_ret, int32_t op_errno, fd_t *fd)
+ int32_t op_ret, int32_t op_errno, fd_t *fd, dict_t *xdata)
{
afr_local_t *local = NULL;
afr_self_heal_t *sh = NULL;
@@ -1961,7 +2099,7 @@ afr_sh_entry_opendir_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
local->loc.path,
priv->children[child_index]->name,
strerror (op_errno));
- sh->op_failed = 1;
+ afr_set_self_heal_status (sh, AFR_SELF_HEAL_FAILED);
}
}
UNLOCK (&frame->lock);
@@ -1969,7 +2107,7 @@ afr_sh_entry_opendir_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
call_count = afr_frame_return (frame);
if (call_count == 0) {
- if (sh->op_failed) {
+ if (is_self_heal_failed (sh, AFR_CHECK_SPECIFIC)) {
afr_sh_entry_finish (frame, this);
return 0;
}
@@ -2007,7 +2145,7 @@ afr_sh_entry_open (call_frame_t *frame, xlator_t *this)
source = local->self_heal.source;
sources = local->self_heal.sources;
- sh->block_size = 65536; //131072
+ sh->block_size = priv->sh_readdir_size;
sh->offset = 0;
call_count = sh->active_sinks;
@@ -2029,7 +2167,7 @@ afr_sh_entry_open (call_frame_t *frame, xlator_t *this)
(void *) (long) source,
priv->children[source],
priv->children[source]->fops->opendir,
- &local->loc, fd);
+ &local->loc, fd, NULL);
call_count--;
}
@@ -2046,7 +2184,7 @@ afr_sh_entry_open (call_frame_t *frame, xlator_t *this)
(void *) (long) i,
priv->children[i],
priv->children[i]->fops->opendir,
- &local->loc, fd);
+ &local->loc, fd, NULL);
if (!--call_count)
break;
@@ -2101,6 +2239,8 @@ afr_sh_entry_sync_prepare (call_frame_t *frame, xlator_t *this)
"merging all entries as a conservative decision",
local->loc.path);
+ sh->actual_sh_started = _gf_true;
+ afr_set_self_heal_status (sh, AFR_SELF_HEAL_SYNC_BEGIN);
afr_sh_entry_open (frame, this);
return 0;
@@ -2123,7 +2263,7 @@ afr_sh_entry_fix (call_frame_t *frame, xlator_t *this,
priv = this->private;
if (op_ret < 0) {
- sh->op_failed = 1;
+ afr_set_self_heal_status (sh, AFR_SELF_HEAL_FAILED);
afr_sh_set_error (sh, op_errno);
afr_sh_entry_finish (frame, this);
goto out;
@@ -2186,7 +2326,7 @@ afr_sh_post_nonblocking_entry_cbk (call_frame_t *frame, xlator_t *this)
if (int_lock->lock_op_ret < 0) {
gf_log (this->name, GF_LOG_ERROR, "Non Blocking entrylks "
"failed for %s.", local->loc.path);
- sh->op_failed = 1;
+ afr_set_self_heal_status (sh, AFR_SELF_HEAL_FAILED);
afr_sh_entry_done (frame, this);
} else {
@@ -2195,7 +2335,8 @@ afr_sh_post_nonblocking_entry_cbk (call_frame_t *frame, xlator_t *this)
afr_sh_common_lookup (frame, this, &local->loc,
afr_sh_entry_fix, NULL,
AFR_LOOKUP_FAIL_CONFLICTS |
- AFR_LOOKUP_FAIL_MISSING_GFIDS);
+ AFR_LOOKUP_FAIL_MISSING_GFIDS,
+ NULL);
}
return 0;
@@ -2204,14 +2345,18 @@ afr_sh_post_nonblocking_entry_cbk (call_frame_t *frame, xlator_t *this)
int
afr_self_heal_entry (call_frame_t *frame, xlator_t *this)
{
- afr_local_t *local = NULL;
+ afr_local_t *local = NULL;
afr_private_t *priv = NULL;
-
+ afr_self_heal_t *sh = NULL;
priv = this->private;
local = frame->local;
+ sh = &local->self_heal;
+
+ sh->sh_type_in_action = AFR_SELF_HEAL_ENTRY;
if (local->self_heal.do_entry_self_heal && priv->entry_self_heal) {
+ afr_set_self_heal_status (sh, AFR_SELF_HEAL_STARTED);
afr_sh_entrylk (frame, this, &local->loc, NULL,
afr_sh_post_nonblocking_entry_cbk);
} else {
diff --git a/xlators/cluster/afr/src/afr-self-heal-metadata.c b/xlators/cluster/afr/src/afr-self-heal-metadata.c
index 148e858bd..fd5da6cfd 100644
--- a/xlators/cluster/afr/src/afr-self-heal-metadata.c
+++ b/xlators/cluster/afr/src/afr-self-heal-metadata.c
@@ -1,20 +1,11 @@
/*
- Copyright (c) 2008-2011 Gluster, Inc. <http://www.gluster.com>
+ Copyright (c) 2008-2012 Red Hat, Inc. <http://www.redhat.com>
This file is part of GlusterFS.
- GlusterFS is free software; you can redistribute it and/or modify
- it under the terms of the GNU General Public License as published
- by the Free Software Foundation; either version 3 of the License,
- or (at your option) any later version.
-
- GlusterFS is distributed in the hope that it will be useful, but
- WITHOUT ANY WARRANTY; without even the implied warranty of
- MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU
- General Public License for more details.
-
- You should have received a copy of the GNU General Public License
- along with this program. If not, see
- <http://www.gnu.org/licenses/>.
+ This file is licensed to you under your choice of the GNU Lesser
+ General Public License, version 3 or any later version (LGPLv3 or
+ later), or the GNU General Public License, version 2 (GPLv2), in all
+ cases as published by the Free Software Foundation.
*/
#include <libgen.h>
@@ -54,31 +45,17 @@ afr_sh_metadata_done (call_frame_t *frame, xlator_t *this)
{
afr_local_t *local = NULL;
afr_self_heal_t *sh = NULL;
- afr_private_t *priv = NULL;
local = frame->local;
sh = &local->self_heal;
- priv = this->private;
-
-// memset (sh->child_errno, 0, sizeof (int) * priv->child_count);
- memset (sh->buf, 0, sizeof (struct iatt) * priv->child_count);
- memset (sh->success, 0, sizeof (*sh->success) * priv->child_count);
- afr_reset_xattr (sh->xattr, priv->child_count);
- if (local->govinda_gOvinda) {
- gf_log (this->name, GF_LOG_INFO,
- "split-brain detected, aborting selfheal of %s",
+ afr_sh_reset (frame, this);
+ if (IA_ISDIR (sh->type)) {
+ gf_log (this->name, GF_LOG_DEBUG,
+ "proceeding to entry check on %s",
local->loc.path);
- sh->op_failed = 1;
- sh->completion_cbk (frame, this);
+ afr_self_heal_entry (frame, this);
} else {
- if (IA_ISDIR (sh->type)) {
- gf_log (this->name, GF_LOG_DEBUG,
- "proceeding to entry check on %s",
- local->loc.path);
- afr_self_heal_entry (frame, this);
- return 0;
- }
gf_log (this->name, GF_LOG_DEBUG,
"proceeding to data check on %s",
local->loc.path);
@@ -111,11 +88,24 @@ afr_sh_metadata_finish (call_frame_t *frame, xlator_t *this)
return 0;
}
+int
+afr_sh_metadata_fail (call_frame_t *frame, xlator_t *this)
+{
+ afr_local_t *local = NULL;
+ afr_self_heal_t *sh = NULL;
+
+ local = frame->local;
+ sh = &local->self_heal;
+
+ afr_set_self_heal_status (sh, AFR_SELF_HEAL_FAILED);
+ afr_sh_metadata_finish (frame, this);
+ return 0;
+}
int
afr_sh_metadata_erase_pending_cbk (call_frame_t *frame, void *cookie,
xlator_t *this, int32_t op_ret,
- int32_t op_errno, dict_t *xattr)
+ int32_t op_errno, dict_t *xattr, dict_t *xdata)
{
afr_local_t *local = NULL;
int call_count = 0;
@@ -147,85 +137,19 @@ afr_sh_metadata_erase_pending_cbk (call_frame_t *frame, void *cookie,
return 0;
}
-
int
afr_sh_metadata_erase_pending (call_frame_t *frame, xlator_t *this)
{
- afr_local_t *local = NULL;
- afr_self_heal_t *sh = NULL;
- afr_private_t *priv = NULL;
- int call_count = 0;
- int i = 0;
- dict_t **erase_xattr = NULL;
-
-
- local = frame->local;
- sh = &local->self_heal;
- priv = this->private;
-
- afr_sh_pending_to_delta (priv, sh->xattr, sh->delta_matrix,
- sh->success, priv->child_count,
- AFR_METADATA_TRANSACTION);
-
- erase_xattr = GF_CALLOC (sizeof (*erase_xattr), priv->child_count,
- gf_afr_mt_dict_t);
- if (!erase_xattr)
- return -ENOMEM;
-
- for (i = 0; i < priv->child_count; i++) {
- if (sh->xattr[i]) {
- call_count++;
-
- erase_xattr[i] = get_new_dict();
- dict_ref (erase_xattr[i]);
- }
- }
-
- afr_sh_delta_to_xattr (priv, sh->delta_matrix, erase_xattr,
- priv->child_count, AFR_METADATA_TRANSACTION);
-
- local->call_count = call_count;
-
- if (call_count == 0) {
- gf_log (this->name, GF_LOG_INFO,
- "metadata of %s not healed on any subvolume",
- local->loc.path);
-
- afr_sh_metadata_finish (frame, this);
- }
-
- for (i = 0; i < priv->child_count; i++) {
- if (!erase_xattr[i])
- continue;
-
- gf_log (this->name, GF_LOG_TRACE,
- "erasing pending flags from %s on %s",
- local->loc.path, priv->children[i]->name);
-
- STACK_WIND_COOKIE (frame, afr_sh_metadata_erase_pending_cbk,
- (void *) (long) i,
- priv->children[i],
- priv->children[i]->fops->xattrop,
- &local->loc,
- GF_XATTROP_ADD_ARRAY, erase_xattr[i]);
- if (!--call_count)
- break;
- }
-
- for (i = 0; i < priv->child_count; i++) {
- if (erase_xattr[i]) {
- dict_unref (erase_xattr[i]);
- }
- }
- GF_FREE (erase_xattr);
-
- return 0;
+ afr_sh_erase_pending (frame, this, AFR_METADATA_TRANSACTION,
+ afr_sh_metadata_erase_pending_cbk,
+ afr_sh_metadata_finish);
+ return 0;
}
int
afr_sh_metadata_sync_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
- int32_t op_ret, int32_t op_errno)
+ int32_t op_ret, int32_t op_errno, dict_t *xdata)
{
afr_local_t *local = NULL;
afr_self_heal_t *sh = NULL;
@@ -256,8 +180,13 @@ afr_sh_metadata_sync_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
call_count = afr_frame_return (frame);
- if (call_count == 0)
+ if (call_count == 0) {
+ if (local->xattr_req) {
+ dict_unref (local->xattr_req);
+ local->xattr_req = NULL;
+ }
afr_sh_metadata_erase_pending (frame, this);
+ }
return 0;
}
@@ -266,9 +195,9 @@ afr_sh_metadata_sync_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
int
afr_sh_metadata_setattr_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
int32_t op_ret, int32_t op_errno,
- struct iatt *preop, struct iatt *postop)
+ struct iatt *preop, struct iatt *postop, dict_t *xdata)
{
- afr_sh_metadata_sync_cbk (frame, cookie, this, op_ret, op_errno);
+ afr_sh_metadata_sync_cbk (frame, cookie, this, op_ret, op_errno, xdata);
return 0;
}
@@ -276,13 +205,93 @@ afr_sh_metadata_setattr_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
int
afr_sh_metadata_xattr_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
- int32_t op_ret, int32_t op_errno)
+ int32_t op_ret, int32_t op_errno, dict_t *xdata)
+{
+ afr_sh_metadata_sync_cbk (frame, cookie, this, op_ret, op_errno, xdata);
+
+ return 0;
+}
+
+int
+afr_sh_removexattr_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
+ int32_t op_ret, int32_t op_errno,
+ dict_t *xdata)
{
- afr_sh_metadata_sync_cbk (frame, cookie, this, op_ret, op_errno);
+ int i = 0;
+ afr_private_t *priv = NULL;
+ afr_local_t *local = NULL;
+ priv = this->private;
+ local = frame->local;
+
+ if (op_ret < 0) {
+ afr_sh_metadata_sync_cbk (frame, cookie,
+ this, -1, op_errno, xdata);
+ goto out;
+ }
+
+ i = (long) cookie;
+
+ STACK_WIND_COOKIE (frame, afr_sh_metadata_xattr_cbk,
+ (void *) (long) i,
+ priv->children[i],
+ priv->children[i]->fops->setxattr,
+ &local->loc, local->xattr_req, 0, NULL);
+
+ out:
return 0;
}
+inline void
+afr_prune_special_keys (dict_t *xattr_dict)
+{
+ dict_del (xattr_dict, GF_SELINUX_XATTR_KEY);
+}
+
+inline void
+afr_prune_pending_keys (dict_t *xattr_dict, afr_private_t *priv)
+{
+ int i = 0;
+
+ for (; i < priv->child_count; i++) {
+ dict_del (xattr_dict, priv->pending_key[i]);
+ }
+}
+
+int
+afr_sh_getxattr_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
+ int32_t op_ret, int32_t op_errno, dict_t *xattr,
+ dict_t *xdata)
+{
+ int i = 0;
+ afr_private_t *priv = NULL;
+ afr_local_t *local = NULL;
+
+ priv = this->private;
+ local = frame->local;
+
+ if (op_ret < 0) {
+ afr_sh_metadata_sync_cbk (frame, cookie,
+ this, -1, op_errno, xdata);
+ goto out;
+ }
+
+ afr_prune_pending_keys (xattr, priv);
+
+ afr_prune_special_keys (xattr);
+
+ i = (long) cookie;
+
+ /* send removexattr in bulk via xdata */
+ STACK_WIND_COOKIE (frame, afr_sh_removexattr_cbk,
+ cookie,
+ priv->children[i],
+ priv->children[i]->fops->removexattr,
+ &local->loc, "", xattr);
+
+ out:
+ return 0;
+}
int
afr_sh_metadata_sync (call_frame_t *frame, xlator_t *this, dict_t *xattr)
@@ -308,9 +317,10 @@ afr_sh_metadata_sync (call_frame_t *frame, xlator_t *this, dict_t *xattr)
/*
* 2 calls per sink - setattr, setxattr
*/
- if (xattr)
+ if (xattr) {
call_count = active_sinks * 2;
- else
+ local->xattr_req = dict_ref (xattr);
+ } else
call_count = active_sinks;
local->call_count = call_count;
@@ -346,18 +356,18 @@ afr_sh_metadata_sync (call_frame_t *frame, xlator_t *this, dict_t *xattr)
(void *) (long) i,
priv->children[i],
priv->children[i]->fops->setattr,
- &local->loc, &stbuf, valid);
+ &local->loc, &stbuf, valid, NULL);
call_count--;
if (!xattr)
continue;
- STACK_WIND_COOKIE (frame, afr_sh_metadata_xattr_cbk,
+ STACK_WIND_COOKIE (frame, afr_sh_getxattr_cbk,
(void *) (long) i,
priv->children[i],
- priv->children[i]->fops->setxattr,
- &local->loc, xattr, 0);
+ priv->children[i]->fops->getxattr,
+ &local->loc, NULL, NULL);
call_count--;
}
@@ -366,17 +376,15 @@ afr_sh_metadata_sync (call_frame_t *frame, xlator_t *this, dict_t *xattr)
int
-afr_sh_metadata_getxattr_cbk (call_frame_t *frame, void *cookie,
- xlator_t *this,
- int32_t op_ret, int32_t op_errno, dict_t *xattr)
+afr_sh_metadata_getxattr_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
+ int32_t op_ret, int32_t op_errno, dict_t *xattr,
+ dict_t *xdata)
{
afr_local_t *local = NULL;
afr_self_heal_t *sh = NULL;
afr_private_t *priv = NULL;
int source = 0;
- int i;
-
local = frame->local;
sh = &local->self_heal;
priv = this->private;
@@ -391,16 +399,147 @@ afr_sh_metadata_getxattr_cbk (call_frame_t *frame, void *cookie,
afr_sh_metadata_sync (frame, this, NULL);
} else {
- for (i = 0; i < priv->child_count; i++) {
- dict_del (xattr, priv->pending_key[i]);
- }
-
+ afr_prune_pending_keys (xattr, priv);
afr_sh_metadata_sync (frame, this, xattr);
}
return 0;
}
+static void
+afr_set_metadata_sh_info_str (afr_local_t *local, afr_self_heal_t *sh,
+ xlator_t *this)
+{
+ afr_private_t *priv = NULL;
+ int i = 0;
+ char num[1024] = {0};
+ size_t len = 0;
+ char *string = NULL;
+ size_t off = 0;
+ char *source_child = " from source %s to";
+ char *format = " %s, ";
+ char *string_msg = " metadata self heal";
+ char *pending_matrix_str = NULL;
+ int down_child_present = 0;
+ int unknown_child_present = 0;
+ char *down_subvol_1 = " down subvolume is ";
+ char *unknown_subvol_1 = " unknown subvolume is";
+ char *down_subvol_2 = " down subvolumes are ";
+ char *unknown_subvol_2 = " unknown subvolumes are ";
+ int down_count = 0;
+ int unknown_count = 0;
+
+ priv = this->private;
+
+ pending_matrix_str = afr_get_pending_matrix_str (sh->pending_matrix,
+ this);
+
+ if (!pending_matrix_str)
+ pending_matrix_str = "";
+
+ len += snprintf (num, sizeof (num), "%s", string_msg);
+
+ for (i = 0; i < priv->child_count; i++) {
+ if ((sh->source == i) && (local->child_up[i] == 1)) {
+ len += snprintf (num, sizeof (num), source_child,
+ priv->children[i]->name);
+ } else if ((local->child_up[i] == 1) && (sh->sources[i] == 0)) {
+ len += snprintf (num, sizeof (num), format,
+ priv->children[i]->name);
+ } else if (local->child_up[i] == 0) {
+ len += snprintf (num, sizeof (num), format,
+ priv->children[i]->name);
+ if (!down_child_present)
+ down_child_present = 1;
+ down_count++;
+ } else if (local->child_up[i] == -1) {
+ len += snprintf (num, sizeof (num), format,
+ priv->children[i]->name);
+ if (!unknown_child_present)
+ unknown_child_present = 1;
+ unknown_count++;
+ }
+ }
+
+ if (down_child_present) {
+ if (down_count > 1) {
+ len += snprintf (num, sizeof (num), "%s",
+ down_subvol_2);
+ } else {
+ len += snprintf (num, sizeof (num), "%s",
+ down_subvol_1);
+ }
+ }
+ if (unknown_child_present) {
+ if (unknown_count > 1) {
+ len += snprintf (num, sizeof (num), "%s",
+ unknown_subvol_2);
+ } else {
+ len += snprintf (num, sizeof (num), "%s",
+ unknown_subvol_1);
+ }
+ }
+
+ len ++;
+
+ string = GF_CALLOC (len, sizeof (char), gf_common_mt_char);
+ if (!string)
+ return;
+
+ off += snprintf (string + off, len - off, "%s", string_msg);
+ for (i=0; i < priv->child_count; i++) {
+ if ((sh->source == i) && (local->child_up[i] == 1))
+ off += snprintf (string + off, len - off, source_child,
+ priv->children[i]->name);
+ }
+
+ for (i = 0; i < priv->child_count; i++) {
+ if ((local->child_up[i] == 1)&& (sh->sources[i] == 0))
+ off += snprintf (string + off, len - off, format,
+ priv->children[i]->name);
+ }
+
+ if (down_child_present) {
+ if (down_count > 1) {
+ off += snprintf (string + off, len - off, "%s",
+ down_subvol_2);
+ } else {
+ off += snprintf (string + off, len - off, "%s",
+ down_subvol_1);
+ }
+ }
+
+ for (i = 0; i < priv->child_count; i++) {
+ if (local->child_up[i] == 0)
+ off += snprintf (string + off, len - off, format,
+ priv->children[i]->name);
+ }
+
+ if (unknown_child_present) {
+ if (unknown_count > 1) {
+ off += snprintf (string + off, len - off, "%s",
+ unknown_subvol_2);
+ } else {
+ off += snprintf (string + off, len - off, "%s",
+ unknown_subvol_1);
+ }
+ }
+
+ for (i = 0; i < priv->child_count; i++) {
+ if (local->child_up[i] == -1)
+ off += snprintf (string + off, len - off, format,
+ priv->children[i]->name);
+ }
+
+ gf_asprintf (&sh->metadata_sh_info, "%s metadata %s,", string,
+ pending_matrix_str);
+
+ if (pending_matrix_str && strcmp (pending_matrix_str, ""))
+ GF_FREE (pending_matrix_str);
+
+ if (string && strcmp (string, ""))
+ GF_FREE (string);
+}
int
afr_sh_metadata_sync_prepare (call_frame_t *frame, xlator_t *this)
@@ -430,10 +569,13 @@ afr_sh_metadata_sync_prepare (call_frame_t *frame, xlator_t *this)
local->loc.path, priv->children[source]->name,
sh->active_sinks);
+ sh->actual_sh_started = _gf_true;
+ afr_set_self_heal_status (sh, AFR_SELF_HEAL_SYNC_BEGIN);
+ afr_set_metadata_sh_info_str (local, sh, this);
STACK_WIND (frame, afr_sh_metadata_getxattr_cbk,
priv->children[source],
priv->children[source]->fops->getxattr,
- &local->loc, NULL);
+ &local->loc, NULL, NULL);
return 0;
}
@@ -455,7 +597,7 @@ afr_sh_metadata_fix (call_frame_t *frame, xlator_t *this,
priv = this->private;
if (op_ret < 0) {
- sh->op_failed = 1;
+ afr_set_self_heal_status (sh, AFR_SELF_HEAL_FAILED);
afr_sh_set_error (sh, op_errno);
afr_sh_metadata_finish (frame, this);
goto out;
@@ -464,15 +606,6 @@ afr_sh_metadata_fix (call_frame_t *frame, xlator_t *this,
sh->pending_matrix, sh->sources,
sh->success_children,
AFR_METADATA_TRANSACTION, NULL, _gf_false);
- if (nsources == 0) {
- gf_log (this->name, GF_LOG_TRACE,
- "No self-heal needed for %s",
- local->loc.path);
-
- afr_sh_metadata_finish (frame, this);
- goto out;
- }
-
if ((nsources == -1)
&& (priv->favorite_child != -1)
&& (sh->child_errno[priv->favorite_child] == 0)) {
@@ -489,12 +622,18 @@ afr_sh_metadata_fix (call_frame_t *frame, xlator_t *this,
}
if (nsources == -1) {
- gf_log (this->name, GF_LOG_ERROR,
- "Unable to self-heal permissions/ownership of '%s' "
- "(possible split-brain). Please fix the file on "
- "all backend volumes", local->loc.path);
+ afr_sh_print_split_brain_log (sh->pending_matrix, this,
+ local->loc.path);
+ afr_set_split_brain (this, sh->inode, SPB, DONT_KNOW);
+ afr_sh_metadata_fail (frame, this);
+ goto out;
+ }
- local->govinda_gOvinda = 1;
+ afr_set_split_brain (this, sh->inode, NO_SPB, DONT_KNOW);
+ if (nsources == 0) {
+ gf_log (this->name, GF_LOG_TRACE,
+ "No self-heal needed for %s",
+ local->loc.path);
afr_sh_metadata_finish (frame, this);
goto out;
@@ -533,7 +672,10 @@ afr_sh_metadata_fix (call_frame_t *frame, xlator_t *this,
sh->fresh_children);
}
- afr_sh_metadata_sync_prepare (frame, this);
+ if (sh->do_metadata_self_heal && priv->metadata_self_heal)
+ afr_sh_metadata_sync_prepare (frame, this);
+ else
+ afr_sh_metadata_finish (frame, this);
out:
return;
}
@@ -549,9 +691,9 @@ afr_sh_metadata_post_nonblocking_inodelk_cbk (call_frame_t *frame,
int_lock = &local->internal_lock;
if (int_lock->lock_op_ret < 0) {
- gf_log (this->name, GF_LOG_ERROR, "Non Blocking metadata "
+ gf_log (this->name, GF_LOG_DEBUG, "Non Blocking metadata "
"inodelks failed for %s.", local->loc.path);
- gf_log (this->name, GF_LOG_ERROR, "Metadata self-heal "
+ gf_log (this->name, GF_LOG_DEBUG, "Metadata self-heal "
"failed for %s.", local->loc.path);
afr_sh_metadata_done (frame, this);
} else {
@@ -562,7 +704,8 @@ afr_sh_metadata_post_nonblocking_inodelk_cbk (call_frame_t *frame,
afr_sh_common_lookup (frame, this, &local->loc,
afr_sh_metadata_fix, NULL,
AFR_LOOKUP_FAIL_CONFLICTS |
- AFR_LOOKUP_FAIL_MISSING_GFIDS);
+ AFR_LOOKUP_FAIL_MISSING_GFIDS,
+ NULL);
}
return 0;
@@ -572,19 +715,22 @@ int
afr_sh_metadata_lock (call_frame_t *frame, xlator_t *this)
{
afr_internal_lock_t *int_lock = NULL;
+ afr_inodelk_t *inodelk = NULL;
afr_local_t *local = NULL;
local = frame->local;
int_lock = &local->internal_lock;
+ int_lock->domain = this->name;
+ inodelk = afr_get_inodelk (int_lock, int_lock->domain);
int_lock->transaction_lk_type = AFR_SELFHEAL_LK;
int_lock->selfheal_lk_type = AFR_METADATA_SELF_HEAL_LK;
afr_set_lock_number (frame, this);
- int_lock->lk_flock.l_start = LLONG_MAX - 1;
- int_lock->lk_flock.l_len = 0;
- int_lock->lk_flock.l_type = F_WRLCK;
+ inodelk->flock.l_start = LLONG_MAX - 1;
+ inodelk->flock.l_len = 0;
+ inodelk->flock.l_type = F_WRLCK;
int_lock->lock_cbk = afr_sh_metadata_post_nonblocking_inodelk_cbk;
afr_nonblocking_inodelk (frame, this);
@@ -592,17 +738,29 @@ afr_sh_metadata_lock (call_frame_t *frame, xlator_t *this)
return 0;
}
+gf_boolean_t
+afr_can_start_metadata_self_heal (afr_self_heal_t *sh, afr_private_t *priv)
+{
+ if (sh->force_confirm_spb)
+ return _gf_true;
+ if (sh->do_metadata_self_heal && priv->metadata_self_heal)
+ return _gf_true;
+ return _gf_false;
+}
int
afr_self_heal_metadata (call_frame_t *frame, xlator_t *this)
{
afr_local_t *local = NULL;
afr_private_t *priv = this->private;
-
+ afr_self_heal_t *sh = &local->self_heal;
local = frame->local;
+ sh = &local->self_heal;
+ sh->sh_type_in_action = AFR_SELF_HEAL_METADATA;
- if (local->self_heal.do_metadata_self_heal && priv->metadata_self_heal) {
+ if (afr_can_start_metadata_self_heal (sh, priv)) {
+ afr_set_self_heal_status (sh, AFR_SELF_HEAL_STARTED);
afr_sh_metadata_lock (frame, this);
} else {
afr_sh_metadata_done (frame, this);
diff --git a/xlators/cluster/afr/src/afr-self-heal.h b/xlators/cluster/afr/src/afr-self-heal.h
index f40c06faa..7c9bc8111 100644
--- a/xlators/cluster/afr/src/afr-self-heal.h
+++ b/xlators/cluster/afr/src/afr-self-heal.h
@@ -1,20 +1,11 @@
/*
- Copyright (c) 2008-2011 Gluster, Inc. <http://www.gluster.com>
- This file is part of GlusterFS.
-
- GlusterFS is free software; you can redistribute it and/or modify
- it under the terms of the GNU General Public License as published
- by the Free Software Foundation; either version 3 of the License,
- or (at your option) any later version.
-
- GlusterFS is distributed in the hope that it will be useful, but
- WITHOUT ANY WARRANTY; without even the implied warranty of
- MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU
- General Public License for more details.
-
- You should have received a copy of the GNU General Public License
- along with this program. If not, see
- <http://www.gnu.org/licenses/>.
+ Copyright (c) 2008-2012 Red Hat, Inc. <http://www.redhat.com>
+ This file is part of GlusterFS.
+
+ This file is licensed to you under your choice of the GNU Lesser
+ General Public License, version 3 or any later version (LGPLv3 or
+ later), or the GNU General Public License, version 2 (GPLv2), in all
+ cases as published by the Free Software Foundation.
*/
#ifndef __AFR_SELF_HEAL_H__
@@ -30,13 +21,6 @@
#define SIZE_GREATER(buf1,buf2) ((buf1)->ia_size > (buf2)->ia_size)
int
-afr_sh_has_metadata_pending (dict_t *xattr, xlator_t *this);
-int
-afr_sh_has_entry_pending (dict_t *xattr, xlator_t *this);
-int
-afr_sh_has_data_pending (dict_t *xattr, xlator_t *this);
-
-int
afr_self_heal_entry (call_frame_t *frame, xlator_t *this);
int
@@ -54,5 +38,6 @@ afr_self_heal (call_frame_t *frame, xlator_t *this, inode_t *inode);
int
afr_lookup_select_read_child_by_txn_type (xlator_t *this, afr_local_t *local,
dict_t **xattr,
- afr_transaction_type txn_type);
+ afr_transaction_type txn_type,
+ uuid_t gfid);
#endif /* __AFR_SELF_HEAL_H__ */
diff --git a/xlators/cluster/afr/src/afr-self-heald.c b/xlators/cluster/afr/src/afr-self-heald.c
index 186d7dd26..1b48a1bca 100644
--- a/xlators/cluster/afr/src/afr-self-heald.c
+++ b/xlators/cluster/afr/src/afr-self-heald.c
@@ -1,20 +1,11 @@
/*
- Copyright (c) 2010-2011 Gluster, Inc. <http://www.gluster.com>
- This file is part of GlusterFS.
-
- GlusterFS is free software; you can redistribute it and/or modify
- it under the terms of the GNU General Public License as published
- by the Free Software Foundation; either version 3 of the License,
- or (at your option) any later version.
-
- GlusterFS is distributed in the hope that it will be useful, but
- WITHOUT ANY WARRANTY; without even the implied warranty of
- MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU
- General Public License for more details.
-
- You should have received a copy of the GNU General Public License
- along with this program. If not, see
- <http://www.gnu.org/licenses/>.
+ Copyright (c) 2008-2012 Red Hat, Inc. <http://www.redhat.com>
+ This file is part of GlusterFS.
+
+ This file is licensed to you under your choice of the GNU Lesser
+ General Public License, version 3 or any later version (LGPLv3 or
+ later), or the GNU General Public License, version 2 (GPLv2), in all
+ cases as published by the Free Software Foundation.
*/
#ifndef _CONFIG_H
@@ -25,9 +16,1076 @@
#include "syncop.h"
#include "afr-self-heald.h"
#include "afr-self-heal-common.h"
+#include "protocol-common.h"
+#include "event-history.h"
+
+typedef enum {
+ STOP_CRAWL_ON_SINGLE_SUBVOL = 1
+} afr_crawl_flags_t;
+
+typedef enum {
+ HEAL = 1,
+ INFO,
+ STATISTICS_TO_BE_HEALED,
+} shd_crawl_op;
+
+typedef struct shd_dump {
+ dict_t *dict;
+ xlator_t *this;
+ int child;
+} shd_dump_t;
+
+typedef struct shd_event_ {
+ int child;
+ char *path;
+} shd_event_t;
+
+typedef struct shd_pos_ {
+ int child;
+ xlator_t *this;
+ afr_child_pos_t pos;
+} shd_pos_t;
+
+typedef int
+(*afr_crawl_done_cbk_t) (int ret, call_frame_t *sync_frame, void *crawl_data);
+
+void
+afr_start_crawl (xlator_t *this, int idx, afr_crawl_type_t crawl,
+ process_entry_cbk_t process_entry, void *op_data,
+ gf_boolean_t exclusive, int crawl_flags,
+ afr_crawl_done_cbk_t crawl_done);
static int
-_crawl_directory (loc_t *loc, pid_t pid);
+_crawl_directory (fd_t *fd, loc_t *loc, afr_crawl_data_t *crawl_data);
+
+/* For calling straight through (e.g. already in a synctask). */
+int
+afr_find_child_position (xlator_t *this, int child, afr_child_pos_t *pos);
+
+/* For deferring through a new synctask. */
+int
+afr_syncop_find_child_position (void *data);
+
+static int
+_loc_assign_gfid_path (loc_t *loc)
+{
+ int ret = -1;
+ char gfid_path[64] = {0};
+
+ if (loc->inode && !uuid_is_null (loc->inode->gfid)) {
+ ret = inode_path (loc->inode, NULL, (char**)&loc->path);
+ } else if (!uuid_is_null (loc->gfid)) {
+ snprintf (gfid_path, sizeof (gfid_path), "<gfid:%s>",
+ uuid_utoa (loc->gfid));
+ loc->path = gf_strdup (gfid_path);
+ if (loc->path)
+ ret = 0;
+ }
+ return ret;
+}
+
+void
+_destroy_crawl_event_data (void *data)
+{
+ shd_crawl_event_t *crawl_event = NULL;
+
+ if (!data)
+ goto out;
+
+ crawl_event = (shd_crawl_event_t *)data;
+ GF_FREE (crawl_event->start_time_str);
+ GF_FREE (crawl_event->end_time_str);
+
+out:
+ return;
+}
+
+void
+_destroy_shd_event_data (void *data)
+{
+ shd_event_t *event = NULL;
+ if (!data)
+ goto out;
+ event = (shd_event_t*)data;
+ GF_FREE (event->path);
+out:
+ return;
+}
+void
+shd_cleanup_event (void *event)
+{
+ shd_event_t *shd_event = event;
+
+ if (!shd_event)
+ goto out;
+ GF_FREE (shd_event->path);
+ GF_FREE (shd_event);
+out:
+ return;
+}
+
+int
+afr_get_local_child (afr_self_heald_t *shd, unsigned int child_count)
+{
+ int i = 0;
+ int ret = -1;
+ for (i = 0; i < child_count; i++) {
+ if (shd->pos[i] == AFR_POS_LOCAL) {
+ ret = i;
+ break;
+ }
+ }
+ return ret;
+}
+
+static int
+_build_index_loc (xlator_t *this, loc_t *loc, char *name, loc_t *parent)
+{
+ int ret = 0;
+
+ uuid_copy (loc->pargfid, parent->inode->gfid);
+ loc->path = "";
+ loc->name = name;
+ loc->parent = inode_ref (parent->inode);
+ if (!loc->parent) {
+ loc->path = NULL;
+ loc_wipe (loc);
+ ret = -1;
+ }
+ return ret;
+}
+
+int
+_add_crawl_stats_to_dict (xlator_t *this, dict_t *output, int child,
+ shd_crawl_event_t *shd_event, struct timeval *tv)
+{
+ int ret = 0;
+ uint64_t count = 0;
+ char key[256] = {0};
+ int xl_id = 0;
+ uint64_t healed_count = 0;
+ uint64_t split_brain_count = 0;
+ uint64_t heal_failed_count = 0;
+ char *start_time_str = NULL;
+ char *end_time_str = NULL;
+ char *crawl_type = NULL;
+ int progress = -1;
+
+ healed_count = shd_event->healed_count;
+ split_brain_count = shd_event->split_brain_count;
+ heal_failed_count = shd_event->heal_failed_count;
+ start_time_str = shd_event->start_time_str;
+ end_time_str = shd_event->end_time_str;
+ crawl_type = shd_event->crawl_type;
+
+ if (!start_time_str) {
+ ret = -1;
+ goto out;
+ }
+
+
+ ret = dict_get_int32 (output, this->name, &xl_id);
+ if (ret) {
+ gf_log (this->name, GF_LOG_ERROR, "xl does not have id");
+ goto out;
+ }
+
+ snprintf (key, sizeof (key), "statistics-%d-%d-count", xl_id, child);
+ ret = dict_get_uint64 (output, key, &count);
+
+ snprintf (key, sizeof (key), "statistics_healed_cnt-%d-%d-%"PRIu64,
+ xl_id, child, count);
+ ret = dict_set_uint64(output, key, healed_count);
+ if (ret) {
+ gf_log (this->name, GF_LOG_ERROR, "Could not add statistics_"
+ "healed_count to outout");
+ goto out;
+ }
+ snprintf (key, sizeof (key), "statistics_sb_cnt-%d-%d-%"PRIu64,
+ xl_id, child, count);
+ ret = dict_set_uint64 (output, key, split_brain_count);
+ if (ret) {
+ gf_log (this->name, GF_LOG_ERROR, "Could not add statistics_"
+ "split_brain_count to outout");
+ goto out;
+ }
+ snprintf (key, sizeof (key), "statistics_crawl_type-%d-%d-%"PRIu64,
+ xl_id, child, count);
+ ret = dict_set_dynstr (output, key, gf_strdup (crawl_type));
+ if (ret) {
+ gf_log (this->name, GF_LOG_ERROR, "Could not add statistics_"
+ "crawl_type to output");
+ goto out;
+ }
+ snprintf (key, sizeof (key), "statistics_heal_failed_cnt-%d-%d-%"PRIu64,
+ xl_id, child, count);
+ ret = dict_set_uint64 (output, key, heal_failed_count);
+ if (ret) {
+ gf_log (this->name, GF_LOG_ERROR, "Could not add statistics_"
+ "healed_failed_count to outout");
+ goto out;
+ }
+ snprintf (key, sizeof (key), "statistics_strt_time-%d-%d-%"PRIu64,
+ xl_id, child, count);
+ ret = dict_set_dynstr (output, key, gf_strdup(start_time_str));
+
+ if (ret) {
+ gf_log (this->name, GF_LOG_ERROR, "Could not add statistics_"
+ "crawl_start_time to outout");
+ goto out;
+ }
+
+ snprintf (key, sizeof (key), "statistics_end_time-%d-%d-%"PRIu64,
+ xl_id, child, count);
+
+ if (!end_time_str)
+ end_time_str = "Could not determine the end time";
+ ret = dict_set_dynstr (output, key, gf_strdup(end_time_str));
+ if (ret) {
+ gf_log (this->name, GF_LOG_ERROR, "Could not add statistics_"
+ "crawl_end_time to outout");
+ goto out;
+ }
+ snprintf (key, sizeof (key), "statistics_inprogress-%d-%d-%"PRIu64,
+ xl_id, child, count);
+
+ if (shd_event->crawl_inprogress == _gf_true)
+ progress = 1;
+ else
+ progress = 0;
+
+ ret = dict_set_int32 (output, key, progress);
+ if (ret) {
+ gf_log (this->name, GF_LOG_ERROR, "Could not add statistics_"
+ "inprogress to outout");
+ goto out;
+ }
+
+ snprintf (key, sizeof (key), "statistics-%d-%d-count",xl_id, child);
+ ret = dict_set_uint64 (output, key, count + 1);
+ if (ret) {
+ gf_log (this->name, GF_LOG_ERROR, "Could not increment the "
+ "counter.");
+ goto out;
+ }
+out:
+ return ret;
+}
+
+int
+_add_path_to_dict (xlator_t *this, dict_t *output, int child, char *path,
+ struct timeval *tv, gf_boolean_t dyn)
+{
+ //subkey not used for now
+ int ret = -1;
+ uint64_t count = 0;
+ char key[256] = {0};
+ int xl_id = 0;
+
+ ret = dict_get_int32 (output, this->name, &xl_id);
+ if (ret) {
+ gf_log (this->name, GF_LOG_ERROR, "xl does not have id");
+ goto out;
+ }
+
+ snprintf (key, sizeof (key), "%d-%d-count", xl_id, child);
+ ret = dict_get_uint64 (output, key, &count);
+
+ snprintf (key, sizeof (key), "%d-%d-%"PRIu64, xl_id, child, count);
+ if (dyn)
+ ret = dict_set_dynstr (output, key, path);
+ else
+ ret = dict_set_str (output, key, path);
+ if (ret) {
+ gf_log (this->name, GF_LOG_ERROR, "%s: Could not add to output",
+ path);
+ goto out;
+ }
+
+ if (!tv)
+ goto inc_count;
+ snprintf (key, sizeof (key), "%d-%d-%"PRIu64"-time", xl_id,
+ child, count);
+ ret = dict_set_uint32 (output, key, tv->tv_sec);
+ if (ret) {
+ gf_log (this->name, GF_LOG_ERROR, "%s: Could not set time",
+ path);
+ goto out;
+ }
+
+inc_count:
+ snprintf (key, sizeof (key), "%d-%d-count", xl_id, child);
+ ret = dict_set_uint64 (output, key, count + 1);
+ if (ret) {
+ gf_log (this->name, GF_LOG_ERROR, "Could not increment count");
+ goto out;
+ }
+ ret = 0;
+out:
+ return ret;
+}
+
+int
+_get_path_from_gfid_loc (xlator_t *this, xlator_t *readdir_xl, loc_t *child,
+ char **fpath, gf_boolean_t *missing)
+{
+ dict_t *xattr = NULL;
+ char *path = NULL;
+ int ret = -1;
+
+ ret = syncop_getxattr (readdir_xl, child, &xattr, GFID_TO_PATH_KEY);
+ if (ret < 0) {
+ if ((errno == ENOENT) && missing)
+ *missing = _gf_true;
+ goto out;
+ }
+ ret = dict_get_str (xattr, GFID_TO_PATH_KEY, &path);
+ if (ret) {
+ gf_log (this->name, GF_LOG_ERROR, "Failed to get path for "
+ "gfid %s", uuid_utoa (child->gfid));
+ goto out;
+ }
+ path = gf_strdup (path);
+ if (!path) {
+ ret = -1;
+ goto out;
+ }
+ ret = 0;
+out:
+ if (!ret)
+ *fpath = path;
+ if (xattr)
+ dict_unref (xattr);
+ return ret;
+}
+
+int
+_add_event_to_dict (circular_buffer_t *cb, void *data)
+{
+ int ret = 0;
+ shd_dump_t *dump_data = NULL;
+ shd_event_t *shd_event = NULL;
+
+ dump_data = data;
+ shd_event = cb->data;
+ if (shd_event->child != dump_data->child)
+ goto out;
+ ret = _add_path_to_dict (dump_data->this, dump_data->dict,
+ dump_data->child, shd_event->path, &cb->tv,
+ _gf_false);
+out:
+ return ret;
+}
+
+int
+_add_crawl_event_statistics_to_dict (circular_buffer_t *cb, void *data)
+{
+ int ret = 0;
+ shd_dump_t *dump_data = NULL;
+ shd_crawl_event_t *shd_event = NULL;
+
+ dump_data = data;
+ shd_event = cb->data;
+ ret = _add_crawl_stats_to_dict (dump_data->this, dump_data->dict,
+ dump_data->child, shd_event, &cb->tv);
+ return ret;
+}
+
+int
+_add_eh_to_dict (xlator_t *this, eh_t *eh, dict_t *dict, int child)
+{
+ shd_dump_t dump_data = {0};
+
+ dump_data.this = this;
+ dump_data.dict = dict;
+ dump_data.child = child;
+ eh_dump (eh, &dump_data, _add_event_to_dict);
+ return 0;
+}
+
+
+int
+_add_statistics_to_dict (xlator_t *this, dict_t *dict, int child)
+{
+ shd_dump_t dump_data = {0};
+ afr_private_t *priv = NULL;
+ afr_self_heald_t *shd = NULL;
+
+ priv = this->private;
+ shd = &priv->shd;
+
+ dump_data.this = this;
+ dump_data.dict = dict;
+ dump_data.child = child;
+ eh_dump (shd->statistics[child], &dump_data,
+ _add_crawl_event_statistics_to_dict);
+ return 0;
+
+}
+
+void
+_remove_stale_index (xlator_t *this, xlator_t *readdir_xl,
+ loc_t *parent, char *fname)
+{
+ int ret = 0;
+ loc_t index_loc = {0};
+
+ ret = _build_index_loc (this, &index_loc, fname, parent);
+ if (ret)
+ goto out;
+ gf_log (this->name, GF_LOG_DEBUG, "Removing stale index "
+ "for %s on %s", index_loc.name, readdir_xl->name);
+ ret = syncop_unlink (readdir_xl, &index_loc);
+ if(ret && (errno != ENOENT)) {
+ gf_log(this->name, GF_LOG_ERROR, "%s: Failed to remove index "
+ "on %s - %s",index_loc.name, readdir_xl->name,
+ strerror (errno));
+ }
+ index_loc.path = NULL;
+ loc_wipe (&index_loc);
+out:
+ return;
+}
+
+int
+_count_hard_links_under_base_indices_dir (xlator_t *this,
+ afr_crawl_data_t *crawl_data,
+ gf_dirent_t *entry, loc_t *childloc,
+ loc_t *parentloc, struct iatt *iattr)
+{
+ xlator_t *readdir_xl = crawl_data->readdir_xl;
+ struct iatt parent = {0};
+ int ret = 0;
+ dict_t *output = NULL;
+ int xl_id = 0;
+ char key[256] = {0};
+ int child = -1;
+ uint64_t hardlinks = 0;
+
+ output = crawl_data->op_data;
+ child = crawl_data->child;
+
+ ret = syncop_lookup (readdir_xl, childloc, NULL, iattr, NULL, &parent);
+ if (ret)
+ goto out;
+
+ ret = dict_get_int32 (output, this->name, &xl_id);
+ if (ret)
+ goto out;
+
+ snprintf (key, sizeof (key), "%d-%d-hardlinks", xl_id, child);
+ ret = dict_get_uint64 (output, key, &hardlinks);
+
+ /*Removing the count of base_entry under indices/base_indicies and
+ * entry under indices/xattrop */
+ hardlinks = hardlinks + iattr->ia_nlink - 2;
+ ret = dict_set_uint64 (output, key, hardlinks);
+ if (ret)
+ goto out;
+
+out:
+ return ret;
+}
+
+int
+_add_summary_to_dict (xlator_t *this, afr_crawl_data_t *crawl_data,
+ gf_dirent_t *entry,
+ loc_t *childloc, loc_t *parentloc, struct iatt *iattr)
+{
+ dict_t *output = NULL;
+ xlator_t *readdir_xl = NULL;
+ int ret = -1;
+ char *path = NULL;
+ gf_boolean_t missing = _gf_false;
+ char gfid_str[64] = {0};
+
+ if (uuid_is_null (childloc->gfid))
+ goto out;
+
+ output = crawl_data->op_data;
+ readdir_xl = crawl_data->readdir_xl;
+
+ ret = _get_path_from_gfid_loc (this, readdir_xl, childloc, &path,
+ &missing);
+ if (ret == 0) {
+ ret = _add_path_to_dict (this, output, crawl_data->child, path,
+ NULL, _gf_true);
+ } else if (missing) {
+ _remove_stale_index (this, readdir_xl, parentloc,
+ uuid_utoa_r (childloc->gfid, gfid_str));
+ }
+
+out:
+ if (ret && path)
+ GF_FREE (path);
+ return ret;
+}
+
+void
+_crawl_post_sh_action (xlator_t *this, loc_t *parent, loc_t *child,
+ int32_t op_ret, int32_t op_errno, dict_t *xattr_rsp,
+ afr_crawl_data_t *crawl_data)
+{
+ int ret = 0;
+ afr_private_t *priv = NULL;
+ afr_self_heald_t *shd = NULL;
+ eh_t *eh = NULL;
+ char *path = NULL;
+ char gfid_str[64] = {0};
+ shd_event_t *event = NULL;
+ int32_t sh_failed = 0;
+ gf_boolean_t split_brain = 0;
+ int32_t actual_sh_done = 0;
+ shd_crawl_event_t **shd_crawl_event = NULL;
+
+ priv = this->private;
+ shd = &priv->shd;
+ if (crawl_data->crawl == INDEX) {
+ if ((op_ret < 0) && (op_errno == ENOENT)) {
+ _remove_stale_index (this, crawl_data->readdir_xl,
+ parent, uuid_utoa_r (child->gfid,
+ gfid_str));
+ goto out;
+ }
+ ret = _get_path_from_gfid_loc (this, crawl_data->readdir_xl,
+ child, &path, NULL);
+ if (ret)
+ goto out;
+ } else {
+ path = gf_strdup (child->path);
+ if (!path) {
+ ret = -1;
+ goto out;
+ }
+ }
+
+ if (xattr_rsp) {
+ ret = dict_get_int32 (xattr_rsp, "sh-failed", &sh_failed);
+ ret = dict_get_int32 (xattr_rsp, "actual-sh-done", &actual_sh_done);
+ }
+
+ shd_crawl_event = (shd_crawl_event_t**)(shd->crawl_events);
+
+ split_brain = afr_is_split_brain (this, child->inode);
+ if ((op_ret < 0 && op_errno == EIO) || split_brain) {
+ eh = shd->split_brain;
+ shd_crawl_event[crawl_data->child]->split_brain_count += 1;
+ } else if ((op_ret < 0) || sh_failed) {
+ eh = shd->heal_failed;
+ shd_crawl_event[crawl_data->child]->heal_failed_count += 1;
+ } else if (actual_sh_done == 1) {
+ eh = shd->healed;
+ shd_crawl_event[crawl_data->child]->healed_count += 1;
+ }
+ ret = -1;
+
+ if (eh != NULL) {
+ event = GF_CALLOC (1, sizeof (*event), gf_afr_mt_shd_event_t);
+ if (!event)
+ goto out;
+ event->child = crawl_data->child;
+ event->path = path;
+
+ ret = eh_save_history (eh, event);
+ if (ret < 0) {
+ gf_log (this->name, GF_LOG_ERROR, "%s:Failed to save "
+ "to event history, (%d, %s)", path, op_ret,
+ strerror (op_errno));
+
+ goto out;
+ }
+ } else {
+ gf_log (this->name, GF_LOG_DEBUG, "%s:Self heal already done ",
+ path);
+
+ }
+ ret = 0;
+out:
+ if (ret && path)
+ GF_FREE (path);
+ return;
+}
+
+int
+_link_inode_update_loc (xlator_t *this, loc_t *loc, struct iatt *iattr)
+{
+ inode_t *link_inode = NULL;
+ int ret = -1;
+
+ link_inode = inode_link (loc->inode, NULL, NULL, iattr);
+ if (link_inode == NULL) {
+ gf_log (this->name, GF_LOG_ERROR, "inode link failed "
+ "on the inode (%s)", uuid_utoa (iattr->ia_gfid));
+ goto out;
+ }
+ inode_unref (loc->inode);
+ loc->inode = link_inode;
+ ret = 0;
+out:
+ return ret;
+}
+
+int
+_self_heal_entry (xlator_t *this, afr_crawl_data_t *crawl_data, gf_dirent_t *entry,
+ loc_t *child, loc_t *parent, struct iatt *iattr)
+{
+ struct iatt parentbuf = {0};
+ int ret = 0;
+ dict_t *xattr_rsp = NULL;
+ dict_t *xattr_req = NULL;
+
+ xattr_req = dict_new ();
+ if (!xattr_req) {
+ errno = ENOMEM;
+ ret = -1;
+ goto out;
+ }
+
+ ret = dict_set_int32 (xattr_req, "allow-sh-for-running-transaction", 1);
+
+ gf_log (this->name, GF_LOG_DEBUG, "lookup %s", child->path);
+
+ ret = syncop_lookup (this, child, xattr_req,
+ iattr, &xattr_rsp, &parentbuf);
+ _crawl_post_sh_action (this, parent, child, ret, errno, xattr_rsp,
+ crawl_data);
+ if (xattr_rsp)
+ dict_unref (xattr_rsp);
+ if (ret == 0)
+ ret = _link_inode_update_loc (this, child, iattr);
+
+out:
+ if (xattr_req)
+ dict_unref(xattr_req);
+ return ret;
+}
+
+static int
+afr_crawl_done (int ret, call_frame_t *sync_frame, void *data)
+{
+ GF_FREE (data);
+ STACK_DESTROY (sync_frame->root);
+ return 0;
+}
+
+void
+_do_self_heal_on_subvol (xlator_t *this, int child, afr_crawl_type_t crawl)
+{
+ afr_start_crawl (this, child, crawl, _self_heal_entry,
+ NULL, _gf_true, STOP_CRAWL_ON_SINGLE_SUBVOL,
+ afr_crawl_done);
+}
+
+gf_boolean_t
+_crawl_proceed (xlator_t *this, int child, int crawl_flags, char **reason)
+{
+ afr_private_t *priv = NULL;
+ afr_self_heald_t *shd = NULL;
+ gf_boolean_t proceed = _gf_false;
+ char *msg = NULL;
+
+ priv = this->private;
+ shd = &priv->shd;
+ if (!shd->enabled) {
+ msg = "Self-heal daemon is not enabled";
+ gf_log (this->name, GF_LOG_DEBUG, "%s", msg);
+ goto out;
+ }
+ if (!priv->child_up[child]) {
+ gf_log (this->name, GF_LOG_DEBUG, "Stopping crawl for %s , "
+ "subvol went down", priv->children[child]->name);
+ msg = "Brick is Not connected";
+ goto out;
+ }
+
+ if (crawl_flags & STOP_CRAWL_ON_SINGLE_SUBVOL) {
+ if (afr_up_children_count (priv->child_up,
+ priv->child_count) < 2) {
+ gf_log (this->name, GF_LOG_DEBUG, "Stopping crawl as "
+ "< 2 children are up");
+ msg = "< 2 bricks in replica are running";
+ goto out;
+ }
+ }
+ proceed = _gf_true;
+out:
+ if (reason)
+ *reason = msg;
+ return proceed;
+}
+
+int
+_do_crawl_op_on_local_subvols (xlator_t *this, afr_crawl_type_t crawl,
+ shd_crawl_op op, dict_t *output)
+{
+ afr_private_t *priv = NULL;
+ char *status = NULL;
+ char *subkey = NULL;
+ char key[256] = {0};
+ shd_pos_t pos_data = {0};
+ int op_ret = -1;
+ int xl_id = -1;
+ int i = 0;
+ int ret = 0;
+ int crawl_flags = 0;
+
+ priv = this->private;
+ if (op == HEAL)
+ crawl_flags |= STOP_CRAWL_ON_SINGLE_SUBVOL;
+
+ if (output) {
+ ret = dict_get_int32 (output, this->name, &xl_id);
+ if (ret) {
+ gf_log (this->name, GF_LOG_ERROR, "Invalid input, "
+ "translator-id is not available");
+ goto out;
+ }
+ }
+ pos_data.this = this;
+ subkey = "status";
+ for (i = 0; i < priv->child_count; i++) {
+ if (_crawl_proceed (this, i, crawl_flags, &status)) {
+ pos_data.child = i;
+ /*
+ * We're already in a synctask in this case, so we
+ * don't need to defer through a second (and in fact
+ * that can cause deadlock). Just call straight
+ * through instead.
+ */
+ ret = afr_find_child_position(pos_data.this,
+ pos_data.child,
+ &pos_data.pos);
+ if (ret) {
+ status = "Not able to find brick location";
+ } else if (pos_data.pos == AFR_POS_REMOTE) {
+ status = "brick is remote";
+ } else {
+ op_ret = 0;
+ if (op == HEAL) {
+ status = "Started self-heal";
+ _do_self_heal_on_subvol (this, i,
+ crawl);
+ } else if (output && (op == INFO)) {
+ status = "";
+ afr_start_crawl (this, i, INDEX,
+ _add_summary_to_dict,
+ output, _gf_false, 0,
+ NULL);
+ } else if (output &&
+ (op == STATISTICS_TO_BE_HEALED)) {
+ status = "";
+ afr_start_crawl (this, i,
+ INDEX_TO_BE_HEALED,
+ _count_hard_links_under_base_indices_dir,
+ output, _gf_false,
+ 0, NULL);
+ }
+ }
+ if (output) {
+ snprintf (key, sizeof (key), "%d-%d-%s", xl_id,
+ i, subkey);
+ ret = dict_set_str (output, key, status);
+ }
+ if (!op_ret && (crawl == FULL))
+ break;
+ }
+ if (output) {
+ snprintf (key, sizeof (key), "%d-%d-%s", xl_id, i,
+ subkey);
+ ret = dict_set_str (output, key, status);
+ }
+ }
+out:
+ return op_ret;
+}
+
+int
+_do_self_heal_on_local_subvols (xlator_t *this, afr_crawl_type_t crawl,
+ dict_t *output)
+{
+ return _do_crawl_op_on_local_subvols (this, crawl, HEAL, output);
+}
+
+int
+_get_index_summary_on_local_subvols (xlator_t *this, dict_t *output)
+{
+ return _do_crawl_op_on_local_subvols (this, INDEX, INFO, output);
+}
+
+void
+afr_fill_completed_crawl_statistics_to_dict (xlator_t *this, dict_t *dict)
+{
+ afr_private_t *priv = NULL;
+ afr_self_heald_t *shd = NULL;
+ int i = 0;
+ priv = this->private;
+ shd= &priv->shd;
+ for (i = 0; i < priv->child_count; i++) {
+ if (shd->pos[i] != AFR_POS_LOCAL)
+ continue;
+ _add_statistics_to_dict (this, dict, i);
+ }
+
+ return ;
+}
+
+static void
+reset_crawl_event (shd_crawl_event_t *crawl_event)
+{
+ crawl_event->healed_count = 0;
+ crawl_event->split_brain_count = 0;
+ crawl_event->heal_failed_count = 0;
+ GF_FREE (crawl_event->start_time_str);
+ crawl_event->start_time_str = NULL;
+ crawl_event->end_time_str = NULL;
+ crawl_event->crawl_type = NULL;
+ crawl_event->crawl_inprogress = _gf_false;
+ return;
+}
+
+static void
+afr_copy_crawl_event_struct (shd_crawl_event_t *src, shd_crawl_event_t *dst)
+{
+ dst->healed_count = src->healed_count;
+ dst->split_brain_count = src->split_brain_count;
+ dst->heal_failed_count = src->heal_failed_count;
+ dst->start_time_str = gf_strdup (src->start_time_str);
+ dst->end_time_str = "Crawl is already in progress";
+ dst->crawl_type = src->crawl_type;
+ dst->crawl_inprogress = _gf_true;
+ return;
+}
+
+static int
+afr_fill_crawl_statistics_of_running_crawl(xlator_t *this, dict_t *dict)
+{
+ shd_crawl_event_t *evnt = NULL;
+ int ret = 0;
+ afr_private_t *priv = NULL;
+ afr_self_heald_t *shd = NULL;
+ int i = 0;
+ priv = this->private;
+ shd = &priv->shd;
+
+ evnt = GF_CALLOC (1, sizeof (shd_crawl_event_t),
+ gf_afr_mt_shd_crawl_event_t);
+ if (!evnt) {
+ ret = -1;
+ goto out;
+ }
+ LOCK (&priv->lock);
+ {
+ for (i = 0; i < priv->child_count; i++) {
+ if (shd->pos[i] != AFR_POS_LOCAL)
+ continue;
+
+ reset_crawl_event (evnt);
+
+ if (!shd->crawl_events[i]) {
+ continue;
+ }
+
+ afr_copy_crawl_event_struct (shd->crawl_events[i],
+ evnt);
+ _add_crawl_stats_to_dict (this, dict, i, evnt, NULL);
+
+ }
+ }
+ UNLOCK (&priv->lock);
+ reset_crawl_event (evnt);
+ GF_FREE (evnt);
+
+out:
+ return ret;
+}
+
+static int
+_add_local_subvols_crawl_statistics_to_dict (xlator_t *this, dict_t *dict)
+{
+ int ret = 0;
+ afr_fill_completed_crawl_statistics_to_dict (this, dict);
+ ret = afr_fill_crawl_statistics_of_running_crawl (this, dict);
+ return ret;
+}
+int
+_add_local_subvols_eh_to_dict (xlator_t *this, eh_t *eh, dict_t *dict)
+{
+ afr_private_t *priv = NULL;
+ afr_self_heald_t *shd = NULL;
+ int i = 0;
+
+ priv = this->private;
+ shd = &priv->shd;
+
+ for (i = 0; i < priv->child_count; i++) {
+ if (shd->pos[i] != AFR_POS_LOCAL)
+ continue;
+ _add_eh_to_dict (this, eh, dict, i);
+ }
+ return 0;
+}
+
+int
+afr_xl_op (xlator_t *this, dict_t *input, dict_t *output)
+{
+ gf_xl_afr_op_t op = GF_AFR_OP_INVALID;
+ int ret = 0;
+ afr_private_t *priv = NULL;
+ afr_self_heald_t *shd = NULL;
+ int xl_id = 0;
+
+ priv = this->private;
+ shd = &priv->shd;
+
+ ret = dict_get_int32 (input, "xl-op", (int32_t*)&op);
+ if (ret)
+ goto out;
+ ret = dict_get_int32 (input, this->name, &xl_id);
+ if (ret)
+ goto out;
+ ret = dict_set_int32 (output, this->name, xl_id);
+ if (ret)
+ goto out;
+ switch (op) {
+ case GF_AFR_OP_HEAL_INDEX:
+ ret = _do_self_heal_on_local_subvols (this, INDEX, output);
+ break;
+ case GF_AFR_OP_HEAL_FULL:
+ ret = _do_self_heal_on_local_subvols (this, FULL, output);
+ break;
+ case GF_AFR_OP_INDEX_SUMMARY:
+ (void)_get_index_summary_on_local_subvols (this, output);
+ ret = 0;
+ break;
+ case GF_AFR_OP_HEALED_FILES:
+ ret = _add_local_subvols_eh_to_dict (this, shd->healed, output);
+ break;
+ case GF_AFR_OP_HEAL_FAILED_FILES:
+ ret = _add_local_subvols_eh_to_dict (this, shd->heal_failed,
+ output);
+ break;
+ case GF_AFR_OP_SPLIT_BRAIN_FILES:
+ ret = _add_local_subvols_eh_to_dict (this, shd->split_brain,
+ output);
+ break;
+ case GF_AFR_OP_STATISTICS:
+ ret = _add_local_subvols_crawl_statistics_to_dict (this, output);
+ break;
+ case GF_AFR_OP_STATISTICS_HEAL_COUNT:
+ case GF_AFR_OP_STATISTICS_HEAL_COUNT_PER_REPLICA:
+ ret = _do_crawl_op_on_local_subvols (this, INDEX_TO_BE_HEALED,
+ STATISTICS_TO_BE_HEALED,
+ output);
+ break;
+ default:
+ gf_log (this->name, GF_LOG_ERROR, "Unknown set op %d", op);
+ break;
+ }
+out:
+ dict_del (output, this->name);
+ return ret;
+}
+
+void
+afr_poll_self_heal (void *data)
+{
+ afr_private_t *priv = NULL;
+ afr_self_heald_t *shd = NULL;
+ struct timespec timeout = {0};
+ xlator_t *this = NULL;
+ long child = (long)data;
+ gf_timer_t *old_timer = NULL;
+ gf_timer_t *new_timer = NULL;
+ shd_pos_t pos_data = {0};
+ int ret = 0;
+
+ this = THIS;
+ priv = this->private;
+ shd = &priv->shd;
+
+ if (shd->pos[child] == AFR_POS_UNKNOWN) {
+ pos_data.this = this;
+ pos_data.child = child;
+ ret = synctask_new (this->ctx->env,
+ afr_syncop_find_child_position,
+ NULL, NULL, &pos_data);
+ if (!ret)
+ shd->pos[child] = pos_data.pos;
+ }
+ if (shd->enabled && (shd->pos[child] == AFR_POS_LOCAL))
+ _do_self_heal_on_subvol (this, child, INDEX);
+ timeout.tv_sec = shd->timeout;
+ timeout.tv_nsec = 0;
+ //notify and previous timer should be synchronized.
+ LOCK (&priv->lock);
+ {
+ old_timer = shd->timer[child];
+ if (shd->pos[child] == AFR_POS_REMOTE)
+ goto unlock;
+ shd->timer[child] = gf_timer_call_after (this->ctx, timeout,
+ afr_poll_self_heal,
+ data);
+ new_timer = shd->timer[child];
+ }
+unlock:
+ UNLOCK (&priv->lock);
+
+ if (old_timer)
+ gf_timer_call_cancel (this->ctx, old_timer);
+ if (!new_timer && (shd->pos[child] != AFR_POS_REMOTE)) {
+ gf_log (this->name, GF_LOG_WARNING,
+ "Could not create self-heal polling timer for %s",
+ priv->children[child]->name);
+ }
+ return;
+}
+
+static int
+afr_handle_child_up (int ret, call_frame_t *sync_frame, void *data)
+{
+ afr_self_heald_t *shd = NULL;
+ shd_pos_t *pos_data = data;
+ afr_private_t *priv = NULL;
+
+ if (ret)
+ goto out;
+
+ priv = pos_data->this->private;
+ shd = &priv->shd;
+ shd->pos[pos_data->child] = pos_data->pos;
+ if (pos_data->pos != AFR_POS_REMOTE)
+ afr_poll_self_heal ((void*)(long)pos_data->child);
+ _do_self_heal_on_local_subvols (THIS, INDEX, NULL);
+out:
+ GF_FREE (data);
+ return 0;
+}
+
+void
+afr_proactive_self_heal (void *data)
+{
+ xlator_t *this = NULL;
+ long child = (long)data;
+ shd_pos_t *pos_data = NULL;
+ int ret = 0;
+
+ this = THIS;
+
+ //Position of brick could have changed and it could be local now.
+ //Compute the position again
+ pos_data = GF_CALLOC (1, sizeof (*pos_data), gf_afr_mt_pos_data_t);
+ if (!pos_data)
+ goto out;
+ pos_data->this = this;
+ pos_data->child = child;
+ ret = synctask_new (this->ctx->env, afr_syncop_find_child_position,
+ afr_handle_child_up, NULL, pos_data);
+ if (ret)
+ goto out;
+out:
+ return;
+}
+
static int
get_pathinfo_host (char *pathinfo, char *hostname, size_t size)
{
@@ -84,30 +1142,206 @@ out:
return ret;
}
-inline void
-afr_fill_loc_info (loc_t *loc, struct iatt *iatt, struct iatt *parent)
+int
+afr_crawl_build_start_loc (xlator_t *this, afr_crawl_data_t *crawl_data,
+ loc_t *dirloc)
+{
+ afr_private_t *priv = NULL;
+ dict_t *xattr = NULL;
+ void *index_gfid = NULL;
+ void *base_indices_holder_vgfid = NULL;
+ loc_t rootloc = {0};
+ struct iatt iattr = {0};
+ struct iatt parent = {0};
+ int ret = 0;
+ xlator_t *readdir_xl = crawl_data->readdir_xl;
+
+ priv = this->private;
+ if (crawl_data->crawl == FULL) {
+ afr_build_root_loc (this, dirloc);
+ } else if (crawl_data->crawl == INDEX) {
+ afr_build_root_loc (this, &rootloc);
+ ret = syncop_getxattr (readdir_xl, &rootloc, &xattr,
+ GF_XATTROP_INDEX_GFID);
+ if (ret < 0)
+ goto out;
+ ret = dict_get_ptr (xattr, GF_XATTROP_INDEX_GFID, &index_gfid);
+ if (ret < 0) {
+ gf_log (this->name, GF_LOG_ERROR, "failed to get index "
+ "dir gfid on %s", readdir_xl->name);
+ goto out;
+ }
+ if (!index_gfid) {
+ gf_log (this->name, GF_LOG_ERROR, "index gfid empty "
+ "on %s", readdir_xl->name);
+ ret = -1;
+ goto out;
+ }
+ uuid_copy (dirloc->gfid, index_gfid);
+ dirloc->path = "";
+ dirloc->inode = inode_new (priv->root_inode->table);
+ ret = syncop_lookup (readdir_xl, dirloc, NULL,
+ &iattr, NULL, &parent);
+ if (ret < 0) {
+ if (errno != ENOENT) {
+ gf_log (this->name, GF_LOG_ERROR, "lookup "
+ "failed on index dir on %s - (%s)",
+ readdir_xl->name, strerror (errno));
+ }
+ goto out;
+ }
+ ret = _link_inode_update_loc (this, dirloc, &iattr);
+ if (ret)
+ goto out;
+ } else if (crawl_data->crawl == INDEX_TO_BE_HEALED) {
+ afr_build_root_loc (this, &rootloc);
+ ret = syncop_getxattr (readdir_xl, &rootloc, &xattr,
+ GF_BASE_INDICES_HOLDER_GFID);
+ if (ret < 0)
+ goto out;
+ ret = dict_get_ptr (xattr, GF_BASE_INDICES_HOLDER_GFID,
+ &base_indices_holder_vgfid);
+ if (ret < 0) {
+ gf_log (this->name, GF_LOG_ERROR, "index gfid empty "
+ "on %s", readdir_xl->name);
+ ret = -1;
+ goto out;
+ }
+ if (!base_indices_holder_vgfid) {
+ gf_log (this->name, GF_LOG_ERROR, "Base indices holder"
+ "virtual gfid is null on %s", readdir_xl->name);
+ ret = -1;
+ goto out;
+ }
+ uuid_copy (dirloc->gfid, base_indices_holder_vgfid);
+ dirloc->path = "";
+ dirloc->inode = inode_new (priv->root_inode->table);
+ ret = syncop_lookup (readdir_xl, dirloc, NULL, &iattr, NULL,
+ &parent);
+ if (ret < 0) {
+ if (errno != ENOENT) {
+ gf_log (this->name, GF_LOG_ERROR, "lookup "
+ "failed for base_indices_holder dir"
+ " on %s - (%s)", readdir_xl->name,
+ strerror (errno));
+
+ } else {
+ gf_log (this->name, GF_LOG_ERROR, "base_indices"
+ "_holder is not yet created.");
+ }
+ goto out;
+ }
+ ret = _link_inode_update_loc (this, dirloc, &iattr);
+ if (ret)
+ goto out;
+ }
+ ret = 0;
+out:
+ if (xattr)
+ dict_unref (xattr);
+ loc_wipe (&rootloc);
+ return ret;
+}
+
+int
+afr_crawl_opendir (xlator_t *this, afr_crawl_data_t *crawl_data, fd_t **dirfd,
+ loc_t *dirloc)
{
- afr_update_loc_gfids (loc, iatt, parent);
- uuid_copy (loc->inode->gfid, iatt->ia_gfid);
+ fd_t *fd = NULL;
+ int ret = 0;
+
+ if (crawl_data->crawl == FULL) {
+ fd = fd_create (dirloc->inode, crawl_data->pid);
+ if (!fd) {
+ gf_log (this->name, GF_LOG_ERROR,
+ "Failed to create fd for %s", dirloc->path);
+ ret = -1;
+ goto out;
+ }
+
+ ret = syncop_opendir (crawl_data->readdir_xl, dirloc, fd);
+ if (ret < 0) {
+ gf_log (this->name, GF_LOG_ERROR,
+ "opendir failed on %s", dirloc->path);
+ goto out;
+ }
+ } else {
+ fd = fd_anonymous (dirloc->inode);
+ }
+ ret = 0;
+out:
+ if (!ret)
+ *dirfd = fd;
+ return ret;
+}
+
+xlator_t*
+afr_crawl_readdir_xl_get (xlator_t *this, afr_crawl_data_t *crawl_data)
+{
+ afr_private_t *priv = this->private;
+
+ if (crawl_data->crawl == FULL) {
+ return this;
+ } else {
+ return priv->children[crawl_data->child];
+ }
+ return NULL;
+}
+
+int
+afr_crawl_build_child_loc (xlator_t *this, loc_t *child, loc_t *parent,
+ gf_dirent_t *entry, afr_crawl_data_t *crawl_data)
+{
+ int ret = -1;
+ afr_private_t *priv = NULL;
+
+ priv = this->private;
+ if (crawl_data->crawl == FULL) {
+ ret = afr_build_child_loc (this, child, parent, entry->d_name);
+ } else if (crawl_data->crawl == INDEX_TO_BE_HEALED) {
+ ret = _build_index_loc (this, child, entry->d_name, parent);
+ if (ret)
+ goto out;
+ child->inode = inode_new (priv->root_inode->table);
+ if (!child->inode) {
+ ret = -1;
+ goto out;
+ }
+ child->path = NULL;
+ } else {
+ child->inode = inode_new (priv->root_inode->table);
+ if (!child->inode)
+ goto out;
+ uuid_parse (entry->d_name, child->gfid);
+ ret = _loc_assign_gfid_path (child);
+ }
+out:
+ return ret;
}
static int
-_perform_self_heal (xlator_t *this, loc_t *parentloc, gf_dirent_t *entries,
- off_t *offset, pid_t pid)
+_process_entries (xlator_t *this, loc_t *parentloc, gf_dirent_t *entries,
+ off_t *offset, afr_crawl_data_t *crawl_data)
{
gf_dirent_t *entry = NULL;
gf_dirent_t *tmp = NULL;
- struct iatt iatt = {0};
- struct iatt parent = {0};;
int ret = 0;
loc_t entry_loc = {0};
+ fd_t *fd = NULL;
+ struct iatt iattr = {0};
list_for_each_entry_safe (entry, tmp, &entries->list, list) {
+ if (!_crawl_proceed (this, crawl_data->child,
+ crawl_data->crawl_flags, NULL)) {
+ ret = -1;
+ goto out;
+ }
*offset = entry->d_off;
if (IS_ENTRY_CWD (entry->d_name) ||
IS_ENTRY_PARENT (entry->d_name))
continue;
- if (uuid_is_null (entry->d_stat.ia_gfid)) {
+ if ((crawl_data->crawl == FULL) &&
+ uuid_is_null (entry->d_stat.ia_gfid)) {
gf_log (this->name, GF_LOG_WARNING, "%s/%s: No "
"gfid present skipping",
parentloc->path, entry->d_name);
@@ -115,88 +1349,93 @@ _perform_self_heal (xlator_t *this, loc_t *parentloc, gf_dirent_t *entries,
}
loc_wipe (&entry_loc);
- ret = afr_build_child_loc (this, &entry_loc, parentloc,
- entry->d_name, entry->d_stat.ia_gfid);
+ ret = afr_crawl_build_child_loc (this, &entry_loc, parentloc,
+ entry, crawl_data);
if (ret)
goto out;
- gf_log (this->name, GF_LOG_DEBUG, "lookup %s", entry_loc.path);
+ ret = crawl_data->process_entry (this, crawl_data, entry,
+ &entry_loc, parentloc, &iattr);
- ret = syncop_lookup (this, &entry_loc, NULL,
- &iatt, NULL, &parent);
- //Don't fail the crawl if lookup fails as it
- //could be because of split-brain
- if (ret || (!IA_ISDIR (iatt.ia_type)))
+ if (crawl_data->crawl == INDEX_TO_BE_HEALED && ret) {
+ goto out;
+ } else if (ret) {
continue;
- afr_fill_loc_info (&entry_loc, &iatt, &parent);
- ret = _crawl_directory (&entry_loc, pid);
+ }
+
+ if ((crawl_data->crawl == INDEX) ||
+ (crawl_data->crawl == INDEX_TO_BE_HEALED))
+ continue;
+
+ if (!IA_ISDIR (iattr.ia_type))
+ continue;
+ fd = NULL;
+ ret = afr_crawl_opendir (this, crawl_data, &fd, &entry_loc);
+ if (ret)
+ continue;
+ ret = _crawl_directory (fd, &entry_loc, crawl_data);
+ if (fd)
+ fd_unref (fd);
}
ret = 0;
out:
- if (entry_loc.path)
- loc_wipe (&entry_loc);
+ if ((crawl_data->crawl == INDEX_TO_BE_HEALED) && ret) {
+ gf_log (this->name, GF_LOG_ERROR,"Failed to get the hardlink "
+ "count");
+ }
+ loc_wipe (&entry_loc);
return ret;
}
static int
-_crawl_directory (loc_t *loc, pid_t pid)
+_crawl_directory (fd_t *fd, loc_t *loc, afr_crawl_data_t *crawl_data)
{
xlator_t *this = NULL;
- afr_private_t *priv = NULL;
- fd_t *fd = NULL;
off_t offset = 0;
gf_dirent_t entries;
- struct iatt iatt = {0};
- struct iatt parent = {0};;
int ret = 0;
gf_boolean_t free_entries = _gf_false;
+ xlator_t *readdir_xl = crawl_data->readdir_xl;
INIT_LIST_HEAD (&entries.list);
this = THIS;
- priv = this->private;
GF_ASSERT (loc->inode);
- gf_log (this->name, GF_LOG_DEBUG, "crawling %s", loc->path);
- fd = fd_create (loc->inode, pid);
- if (!fd) {
- gf_log (this->name, GF_LOG_ERROR,
- "Failed to create fd for %s", loc->path);
- goto out;
- }
-
- if (!loc->parent) {
- ret = syncop_lookup (this, loc, NULL,
- &iatt, NULL, &parent);
- }
-
- ret = syncop_opendir (this, loc, fd);
- if (ret < 0) {
- gf_log (this->name, GF_LOG_ERROR,
- "opendir failed on %s", loc->path);
- goto out;
- }
+ if (crawl_data->crawl == FULL)
+ gf_log (this->name, GF_LOG_DEBUG, "crawling %s", loc->path);
+ else
+ gf_log (this->name, GF_LOG_DEBUG, "crawling INDEX %s",
+ uuid_utoa (loc->gfid));
- while (syncop_readdirp (this, fd, 131072, offset, NULL, &entries)) {
+ while (1) {
+ if (crawl_data->crawl == FULL)
+ ret = syncop_readdirp (readdir_xl, fd, 131072, offset,
+ NULL, &entries);
+ else
+ ret = syncop_readdir (readdir_xl, fd, 131072, offset,
+ &entries);
+ if (ret <= 0)
+ break;
ret = 0;
free_entries = _gf_true;
- if (afr_up_children_count (priv->child_up,
- priv->child_count) < 2) {
- gf_log (this->name, GF_LOG_ERROR, "Stopping crawl as "
- "< 2 children are up");
+
+ if (!_crawl_proceed (this, crawl_data->child,
+ crawl_data->crawl_flags, NULL)) {
ret = -1;
goto out;
}
-
if (list_empty (&entries.list))
goto out;
- ret = _perform_self_heal (this, loc, &entries, &offset, pid);
+ ret = _process_entries (this, loc, &entries, &offset,
+ crawl_data);
+ if ((ret < 0) && (crawl_data->crawl == INDEX_TO_BE_HEALED)) {
+ goto out;
+ }
gf_dirent_free (&entries);
free_entries = _gf_false;
}
- if (fd)
- fd_unref (fd);
ret = 0;
out:
if (free_entries)
@@ -204,279 +1443,332 @@ out:
return ret;
}
+static char*
+position_str_get (afr_child_pos_t pos)
+{
+ switch (pos) {
+ case AFR_POS_UNKNOWN:
+ return "unknown";
+ case AFR_POS_LOCAL:
+ return "local";
+ case AFR_POS_REMOTE:
+ return "remote";
+ }
+ return NULL;
+}
+
int
-afr_find_child_position (xlator_t *this, int child)
+afr_find_child_position (xlator_t *this, int child, afr_child_pos_t *pos)
{
afr_private_t *priv = NULL;
+ afr_self_heald_t *shd = NULL;
dict_t *xattr_rsp = NULL;
loc_t loc = {0};
int ret = 0;
- gf_boolean_t local = _gf_false;
- char *pathinfo = NULL;
- afr_child_pos_t *pos = NULL;
- inode_table_t *itable = NULL;
+ char *node_uuid = NULL;
priv = this->private;
- pos = &priv->shd.pos[child];
+ shd = &priv->shd;
- if (*pos != AFR_POS_UNKNOWN) {
- goto out;
- }
-
- //TODO: Hack to make the root_loc hack work
- LOCK (&priv->lock);
- {
- if (!priv->root_inode) {
- itable = inode_table_new (0, this);
- if (!itable)
- goto unlock;
- priv->root_inode = inode_new (itable);
- if (!priv->root_inode)
- goto unlock;
- }
- }
-unlock:
- UNLOCK (&priv->lock);
-
- if (!priv->root_inode) {
- ret = -1;
- goto out;
- }
- afr_build_root_loc (priv->root_inode, &loc);
+ afr_build_root_loc (this, &loc);
ret = syncop_getxattr (priv->children[child], &loc, &xattr_rsp,
- GF_XATTR_PATHINFO_KEY);
- if (ret) {
- gf_log (this->name, GF_LOG_ERROR, "getxattr failed on child "
- "%d", child);
+ GF_XATTR_NODE_UUID_KEY);
+ if (ret < 0) {
+ gf_log (this->name, GF_LOG_ERROR, "getxattr failed on %s - "
+ "(%s)", priv->children[child]->name, strerror (errno));
goto out;
}
- ret = dict_get_str (xattr_rsp, GF_XATTR_PATHINFO_KEY, &pathinfo);
+ ret = dict_get_str (xattr_rsp, GF_XATTR_NODE_UUID_KEY, &node_uuid);
if (ret) {
- gf_log (this->name, GF_LOG_ERROR, "Pathinfo key not found on "
- "child %d", child);
+ gf_log (this->name, GF_LOG_ERROR, "node-uuid key not found on "
+ "child %s", priv->children[child]->name);
goto out;
}
- ret = afr_local_pathinfo (pathinfo, &local);
- if (ret)
- goto out;
- if (local)
+ if (!strcmp (node_uuid, shd->node_uuid))
*pos = AFR_POS_LOCAL;
else
*pos = AFR_POS_REMOTE;
- gf_log (this->name, GF_LOG_INFO, "child %d is %d", child, *pos);
+ gf_log (this->name, GF_LOG_DEBUG, "child %s is %s",
+ priv->children[child]->name, position_str_get (*pos));
out:
+ if (ret)
+ *pos = AFR_POS_UNKNOWN;
+ loc_wipe (&loc);
return ret;
}
-static int
-afr_crawl_done (int ret, call_frame_t *sync_frame, void *data)
+int
+afr_syncop_find_child_position (void *data)
{
- GF_FREE (data);
- STACK_DESTROY (sync_frame->root);
- return 0;
+ shd_pos_t *pos_data = data;
+ int ret = 0;
+
+ ret = afr_find_child_position (pos_data->this, pos_data->child,
+ &pos_data->pos);
+ return ret;
}
static int
-afr_find_all_children_postions (xlator_t *this)
+afr_dir_crawl (void *data)
{
- int ret = -1;
- int i = 0;
- gf_boolean_t succeeded = _gf_false;
- afr_private_t *priv = NULL;
+ xlator_t *this = NULL;
+ int ret = -1;
+ xlator_t *readdir_xl = NULL;
+ fd_t *fd = NULL;
+ loc_t dirloc = {0};
+ afr_crawl_data_t *crawl_data = data;
- priv = this->private;
- for (i = 0; i < priv->child_count; i++) {
- if (priv->child_up[i] != 1)
- continue;
- ret = afr_find_child_position (this, i);
- if (ret) {
- gf_log (this->name, GF_LOG_ERROR,
- "Failed to determine if the "
- "child %s is local.",
- priv->children[i]->name);
- continue;
+ this = THIS;
+
+ if (!_crawl_proceed (this, crawl_data->child, crawl_data->crawl_flags,
+ NULL))
+ goto out;
+
+ readdir_xl = afr_crawl_readdir_xl_get (this, crawl_data);
+ if (!readdir_xl)
+ goto out;
+ crawl_data->readdir_xl = readdir_xl;
+
+ ret = afr_crawl_build_start_loc (this, crawl_data, &dirloc);
+ if (ret)
+ goto out;
+
+ ret = afr_crawl_opendir (this, crawl_data, &fd, &dirloc);
+ if (ret) {
+ if (crawl_data->crawl == INDEX_TO_BE_HEALED) {
+ gf_log (this->name, GF_LOG_ERROR, "Failed to open base_"
+ "indices_holder");
}
- succeeded = _gf_true;
+ goto out;
}
- if (succeeded)
- ret = 0;
+
+ ret = _crawl_directory (fd, &dirloc, crawl_data);
+ if (ret)
+ gf_log (this->name, GF_LOG_ERROR, "Crawl failed on %s",
+ readdir_xl->name);
+ else
+ gf_log (this->name, GF_LOG_DEBUG, "Crawl completed "
+ "on %s", readdir_xl->name);
+ if (crawl_data->crawl == INDEX)
+ dirloc.path = NULL;
+out:
+ if (fd)
+ fd_unref (fd);
+ if ((crawl_data->crawl == INDEX) ||
+ (crawl_data->crawl == INDEX_TO_BE_HEALED ))
+ dirloc.path = NULL;
+ loc_wipe (&dirloc);
return ret;
}
-static gf_boolean_t
-afr_local_child_exists (afr_child_pos_t *pos, unsigned int child_count)
+char *
+get_crawl_type_in_string (afr_crawl_type_t crawl)
{
- int i = 0;
- gf_boolean_t local = _gf_false;
-
- for (i = 0; i < child_count; i++, pos++) {
- if (*pos == AFR_POS_LOCAL) {
- local = _gf_true;
- break;
- }
+ char *index = "INDEX";
+ char *full = "FULL";
+ char *crawl_type = NULL;
+
+ if (crawl == INDEX){
+ crawl_type = index;
+ } else if (crawl == FULL) {
+ crawl_type = full;
}
- return local;
+
+ return crawl_type;
}
-int
-afr_init_child_position (xlator_t *this, int child)
+static int
+afr_allocate_crawl_event (xlator_t *this, int child, afr_crawl_type_t crawl)
{
- int ret = 0;
+ afr_private_t *priv = NULL;
+ afr_self_heald_t *shd = NULL;
+ int ret = 0;
+ shd_crawl_event_t *crawl_event = NULL;
+ time_t get_time = 0;
- if (child == AFR_ALL_CHILDREN) {
- ret = afr_find_all_children_postions (this);
- } else {
- ret = afr_find_child_position (this, child);
+ priv = this->private;
+ shd = &priv->shd;
+
+ crawl_event = GF_CALLOC (sizeof (shd_crawl_event_t), 1,
+ gf_afr_mt_shd_crawl_event_t);
+ if (!crawl_event) {
+ ret = -1;
+ goto out;
+ }
+
+ get_time = time(NULL);
+ if (get_time == ((time_t)-1)) {
+ ret = -1;
+ goto out;
}
+
+ crawl_event->start_time_str = gf_strdup (ctime(&get_time));
+
+ crawl_event->crawl_type = get_crawl_type_in_string (crawl);
+ if (!crawl_event->crawl_type) {
+ ret = -1;
+ goto out;
+ }
+ LOCK (&priv->lock);
+ {
+ shd->crawl_events[child] = crawl_event;
+ }
+ UNLOCK (&priv->lock);
+ ret = 0;
+out:
return ret;
+
}
-int
-afr_is_local_child (afr_self_heald_t *shd, int child, unsigned int child_count)
+static int
+afr_put_crawl_event_in_eh (xlator_t *this, int child)
{
- gf_boolean_t local = _gf_false;
+ afr_private_t *priv = NULL;
+ afr_self_heald_t *shd = NULL;
+ int ret = 0;
+ time_t get_time = 0;
+ shd_crawl_event_t **crawl_event = NULL;
- if (child == AFR_ALL_CHILDREN)
- local = afr_local_child_exists (shd->pos, child_count);
- else
- local = (shd->pos[child] == AFR_POS_LOCAL);
+ priv = this->private;
+ shd = &priv->shd;
- return local;
+ get_time = time(NULL);
+ if (get_time == ((time_t)-1)) {
+ ret = -1;
+ goto out;
+ }
+ crawl_event = (shd_crawl_event_t**)shd->crawl_events;
+ LOCK (&priv->lock);
+ {
+ crawl_event[child]->end_time_str = gf_strdup (ctime(&get_time));
+ ret = eh_save_history (shd->statistics[child],
+ crawl_event[child]);
+ crawl_event[child] = NULL;
+ }
+ UNLOCK (&priv->lock);
+out:
+ return ret;
}
static int
-afr_crawl_directory (xlator_t *this, pid_t pid)
+afr_dir_exclusive_crawl (void *data)
{
afr_private_t *priv = NULL;
afr_self_heald_t *shd = NULL;
- loc_t loc = {0};
gf_boolean_t crawl = _gf_false;
- int ret = 0;
+ int ret = 0;
+ int child = -1;
+ xlator_t *this = NULL;
+ afr_crawl_data_t *crawl_data = data;
+ this = THIS;
priv = this->private;
shd = &priv->shd;
-
+ child = crawl_data->child;
LOCK (&priv->lock);
{
- if (shd->inprogress) {
- shd->pending = _gf_true;
+ if (shd->inprogress[child]) {
+ if (shd->pending[child] != FULL)
+ shd->pending[child] = crawl_data->crawl;
} else {
- shd->inprogress = _gf_true;
+ shd->inprogress[child] = _gf_true;
crawl = _gf_true;
}
}
UNLOCK (&priv->lock);
- if (!priv->root_inode) {
- ret = -1;
+ if (!crawl) {
+ gf_log (this->name, GF_LOG_INFO, "Another crawl is in progress "
+ "for %s", priv->children[child]->name);
goto out;
}
- if (!crawl)
- goto out;
-
- afr_build_root_loc (priv->root_inode, &loc);
- while (crawl) {
- ret = _crawl_directory (&loc, pid);
+ do {
+ ret = afr_allocate_crawl_event (this, child, crawl_data->crawl);
if (ret)
- gf_log (this->name, GF_LOG_ERROR, "Crawl failed");
- else
- gf_log (this->name, GF_LOG_INFO, "Crawl completed");
+ goto out;
+ afr_dir_crawl (data);
+
+ ret = afr_put_crawl_event_in_eh (this, child);
+ if (ret < 0)
+ goto out;
+
LOCK (&priv->lock);
{
- if (shd->pending) {
- shd->pending = _gf_false;
+ if (shd->pending[child] != NONE) {
+ crawl_data->crawl = shd->pending[child];
+ shd->pending[child] = NONE;
} else {
- shd->inprogress = _gf_false;
+ shd->inprogress[child] = _gf_false;
crawl = _gf_false;
}
}
UNLOCK (&priv->lock);
- }
-out:
- return ret;
-}
-
-static int
-afr_crawl (void *data)
-{
- xlator_t *this = NULL;
- afr_private_t *priv = NULL;
- afr_self_heald_t *shd = NULL;
- int ret = -1;
- afr_crawl_data_t *crawl_data = data;
-
- this = THIS;
- priv = this->private;
- shd = &priv->shd;
-
- ret = afr_init_child_position (this, crawl_data->child);
- if (ret)
- goto out;
-
- if (!afr_is_local_child (shd, crawl_data->child, priv->child_count))
- goto out;
-
- ret = afr_crawl_directory (this, crawl_data->pid);
+ } while (crawl);
out:
return ret;
}
void
-afr_proactive_self_heal (xlator_t *this, int idx)
+afr_start_crawl (xlator_t *this, int idx, afr_crawl_type_t crawl,
+ process_entry_cbk_t process_entry, void *op_data,
+ gf_boolean_t exclusive, int crawl_flags,
+ afr_crawl_done_cbk_t crawl_done)
{
afr_private_t *priv = NULL;
- afr_self_heald_t *shd = NULL;
call_frame_t *frame = NULL;
afr_crawl_data_t *crawl_data = NULL;
int ret = 0;
+ int (*crawler) (void*) = NULL;
priv = this->private;
- shd = &priv->shd;
- if (!shd->enabled)
- goto out;
-
- if ((idx != AFR_ALL_CHILDREN) &&
- (shd->pos[idx] == AFR_POS_REMOTE))
- goto out;
frame = create_frame (this, this->ctx->pool);
if (!frame)
goto out;
- afr_set_lk_owner (frame, this);
+ afr_set_lk_owner (frame, this, frame->root);
afr_set_low_priority (frame);
crawl_data = GF_CALLOC (1, sizeof (*crawl_data),
- gf_afr_mt_afr_crawl_data_t);
+ gf_afr_mt_crawl_data_t);
if (!crawl_data)
goto out;
+ crawl_data->process_entry = process_entry;
crawl_data->child = idx;
crawl_data->pid = frame->root->pid;
- gf_log (this->name, GF_LOG_INFO, "starting crawl for %d", idx);
- ret = synctask_new (this->ctx->env, afr_crawl,
- afr_crawl_done, frame, crawl_data);
+ crawl_data->crawl = crawl;
+ crawl_data->op_data = op_data;
+ crawl_data->crawl_flags = crawl_flags;
+ gf_log (this->name, GF_LOG_DEBUG, "starting crawl %d for %s",
+ crawl_data->crawl, priv->children[idx]->name);
+
+ if (exclusive)
+ crawler = afr_dir_exclusive_crawl;
+ else
+ crawler = afr_dir_crawl;
+ ret = synctask_new (this->ctx->env, crawler,
+ crawl_done, frame, crawl_data);
if (ret)
- gf_log (this->name, GF_LOG_ERROR, "Could not create the "
- "task for %d ret %d", idx, ret);
+ gf_log (this->name, GF_LOG_ERROR, "afr crawl failed for child"
+ " %d with ret %d", idx, ret);
out:
return;
}
-//TODO: This is a hack
void
-afr_build_root_loc (inode_t *inode, loc_t *loc)
+afr_build_root_loc (xlator_t *this, loc_t *loc)
{
- loc->path = "/";
+ afr_private_t *priv = NULL;
+
+ priv = this->private;
+ loc->path = gf_strdup ("/");
loc->name = "";
- loc->inode = inode;
- loc->inode->ia_type = IA_IFDIR;
- memset (loc->inode->gfid, 0, 16);
- loc->inode->gfid[15] = 1;
+ loc->inode = inode_ref (priv->root_inode);
uuid_copy (loc->gfid, loc->inode->gfid);
}
@@ -493,4 +1785,3 @@ afr_set_root_gfid (dict_t *dict)
return ret;
}
-
diff --git a/xlators/cluster/afr/src/afr-self-heald.h b/xlators/cluster/afr/src/afr-self-heald.h
index 6eb119b07..e0c083754 100644
--- a/xlators/cluster/afr/src/afr-self-heald.h
+++ b/xlators/cluster/afr/src/afr-self-heald.h
@@ -1,20 +1,11 @@
/*
- Copyright (c) 2010-2011 Gluster, Inc. <http://www.gluster.com>
- This file is part of GlusterFS.
-
- GlusterFS is free software; you can redistribute it and/or modify
- it under the terms of the GNU General Public License as published
- by the Free Software Foundation; either version 3 of the License,
- or (at your option) any later version.
-
- GlusterFS is distributed in the hope that it will be useful, but
- WITHOUT ANY WARRANTY; without even the implied warranty of
- MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU
- General Public License for more details.
-
- You should have received a copy of the GNU General Public License
- along with this program. If not, see
- <http://www.gnu.org/licenses/>.
+ Copyright (c) 2008-2012 Red Hat, Inc. <http://www.redhat.com>
+ This file is part of GlusterFS.
+
+ This file is licensed to you under your choice of the GNU Lesser
+ General Public License, version 3 or any later version (LGPLv3 or
+ later), or the GNU General Public License, version 2 (GPLv2), in all
+ cases as published by the Free Software Foundation.
*/
#ifndef __AFR_SELF_HEALD_H__
@@ -27,17 +18,48 @@
#define AFR_ALL_CHILDREN -1
typedef struct afr_crawl_data_ {
- int child;
- pid_t pid;
+ int child;
+ pid_t pid;
+ afr_crawl_type_t crawl;
+ xlator_t *readdir_xl;
+ void *op_data;
+ int crawl_flags;
+ int (*process_entry) (xlator_t *this, struct afr_crawl_data_ *crawl_data,
+ gf_dirent_t *entry, loc_t *child, loc_t *parent,
+ struct iatt *iattr);
} afr_crawl_data_t;
-void afr_proactive_self_heal (xlator_t *this, int idx);
+typedef struct crawl_event_stats_ {
+ uint64_t healed_count;
+ uint64_t split_brain_count;
+ uint64_t heal_failed_count;
+ char *start_time_str;
+ char *end_time_str;
+ char *crawl_type;
+ gf_boolean_t crawl_inprogress;
+} shd_crawl_event_t;
+
+void _destroy_crawl_event_data (void *data);
+void _destroy_shd_event_data (void *data);
+
+typedef int (*process_entry_cbk_t) (xlator_t *this, afr_crawl_data_t *crawl_data,
+ gf_dirent_t *entry, loc_t *child, loc_t *parent,
+ struct iatt *iattr);
-void afr_build_root_loc (inode_t *inode, loc_t *loc);
+void afr_build_root_loc (xlator_t *this, loc_t *loc);
int afr_set_root_gfid (dict_t *dict);
-inline void
-afr_fill_loc_info (loc_t *loc, struct iatt *iatt, struct iatt *parent);
+void
+afr_proactive_self_heal (void *data);
+int
+afr_xl_op (xlator_t *this, dict_t *input, dict_t *output);
+
+/*
+ * In addition to its self-heal use, this is used to find a local default
+ * read_child.
+ */
+int
+afr_local_pathinfo (char *pathinfo, gf_boolean_t *local);
#endif /* __AFR_SELF_HEALD_H__ */
diff --git a/xlators/cluster/afr/src/afr-transaction.c b/xlators/cluster/afr/src/afr-transaction.c
index 36d74aed8..20306e469 100644
--- a/xlators/cluster/afr/src/afr-transaction.c
+++ b/xlators/cluster/afr/src/afr-transaction.c
@@ -1,25 +1,17 @@
/*
- Copyright (c) 2007-2011 Gluster, Inc. <http://www.gluster.com>
+ Copyright (c) 2008-2012 Red Hat, Inc. <http://www.redhat.com>
This file is part of GlusterFS.
- GlusterFS is free software; you can redistribute it and/or modify
- it under the terms of the GNU General Public License as published
- by the Free Software Foundation; either version 3 of the License,
- or (at your option) any later version.
-
- GlusterFS is distributed in the hope that it will be useful, but
- WITHOUT ANY WARRANTY; without even the implied warranty of
- MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU
- General Public License for more details.
-
- You should have received a copy of the GNU General Public License
- along with this program. If not, see
- <http://www.gnu.org/licenses/>.
+ This file is licensed to you under your choice of the GNU Lesser
+ General Public License, version 3 or any later version (LGPLv3 or
+ later), or the GNU General Public License, version 2 (GPLv2), in all
+ cases as published by the Free Software Foundation.
*/
#include "dict.h"
#include "byte-order.h"
#include "common-utils.h"
+#include "timer.h"
#include "afr.h"
#include "afr-transaction.h"
@@ -32,7 +24,6 @@
of RENAME */
#define LOCKED_LOWER 0x2 /* for lower_path of RENAME */
-
afr_fd_ctx_t *
__afr_fd_ctx_get (fd_t *fd, xlator_t *this)
{
@@ -82,27 +73,26 @@ afr_fd_ctx_get (fd_t *fd, xlator_t *this)
static void
-afr_pid_save (call_frame_t *frame)
+afr_save_lk_owner (call_frame_t *frame)
{
afr_local_t * local = NULL;
local = frame->local;
- local->saved_pid = frame->root->pid;
+ local->saved_lk_owner = frame->root->lk_owner;
}
static void
-afr_pid_restore (call_frame_t *frame)
+afr_restore_lk_owner (call_frame_t *frame)
{
afr_local_t * local = NULL;
local = frame->local;
- frame->root->pid = local->saved_pid;
+ frame->root->lk_owner = local->saved_lk_owner;
}
-
static void
__mark_all_pending (int32_t *pending[], int child_count,
afr_transaction_type type)
@@ -155,51 +145,23 @@ out:
return;
}
-
-static void
-__mark_pre_op_undone_on_fd (call_frame_t *frame, xlator_t *this, int child_index)
-{
- afr_local_t *local = NULL;
- afr_fd_ctx_t *fd_ctx = NULL;
-
- local = frame->local;
-
- if (!local->fd)
- return;
-
- fd_ctx = afr_fd_ctx_get (local->fd, this);
-
- if (!fd_ctx)
- goto out;
-
- LOCK (&local->fd->lock);
- {
- if (local->transaction.type == AFR_DATA_TRANSACTION)
- fd_ctx->pre_op_done[child_index]--;
- }
- UNLOCK (&local->fd->lock);
-out:
- return;
-}
-
-
static void
-__mark_down_children (int32_t *pending[], int child_count,
- unsigned char *child_up, afr_transaction_type type)
+__mark_non_participant_children (int32_t *pending[], int child_count,
+ unsigned char *participants,
+ afr_transaction_type type)
{
int i = 0;
int j = 0;
+ j = afr_index_for_transaction_type (type);
for (i = 0; i < child_count; i++) {
- j = afr_index_for_transaction_type (type);
-
- if (!child_up[i])
+ if (!participants[i])
pending[i][j] = 0;
}
}
-static void
+void
__mark_all_success (int32_t *pending[], int child_count,
afr_transaction_type type)
{
@@ -212,6 +174,54 @@ __mark_all_success (int32_t *pending[], int child_count,
}
}
+void
+_set_all_child_errno (int *child_errno, unsigned int child_count)
+{
+ int i = 0;
+
+ for (i = 0; i < child_count; i++)
+ if (child_errno[i] == 0)
+ child_errno[i] = ENOTCONN;
+}
+
+void
+afr_transaction_perform_fop (call_frame_t *frame, xlator_t *this)
+{
+ afr_local_t *local = NULL;
+ afr_private_t *priv = NULL;
+ fd_t *fd = NULL;
+
+ local = frame->local;
+ priv = this->private;
+ fd = local->fd;
+
+ __mark_all_success (local->pending, priv->child_count,
+ local->transaction.type);
+
+ _set_all_child_errno (local->child_errno, priv->child_count);
+
+ /* Perform fops with the lk-owner from top xlator.
+ * Eg: lk-owner of posix-lk and flush should be same,
+ * flush cant clear the posix-lks without that lk-owner.
+ */
+ afr_save_lk_owner (frame);
+ frame->root->lk_owner =
+ local->transaction.main_frame->root->lk_owner;
+
+
+ /* The wake up needs to happen independent of
+ what type of fop arrives here. If it was
+ a write, then it has already inherited the
+ lock and changelog. If it was not a write,
+ then the presumption of the optimization (of
+ optimizing for successive write operations)
+ fails.
+ */
+ if (fd)
+ afr_delayed_changelog_wake_up (this, fd);
+ local->transaction.fop (frame, this);
+}
+
static int
__changelog_enabled (afr_private_t *priv, afr_transaction_type type)
@@ -275,64 +285,42 @@ __fop_changelog_needed (call_frame_t *frame, xlator_t *this)
return op_ret;
}
-
-static int
-afr_set_pending_dict (afr_private_t *priv, dict_t *xattr, int32_t **pending)
+int
+afr_set_pending_dict (afr_private_t *priv, dict_t *xattr, int32_t **pending,
+ int child, afr_xattrop_type_t op)
{
int i = 0;
int ret = 0;
+ if (op == LOCAL_FIRST) {
+ ret = dict_set_static_bin (xattr, priv->pending_key[child],
+ pending[child],
+ AFR_NUM_CHANGE_LOGS * sizeof (int32_t));
+ if (ret)
+ goto out;
+ }
for (i = 0; i < priv->child_count; i++) {
+ if (i == child)
+ continue;
ret = dict_set_static_bin (xattr, priv->pending_key[i],
- pending[i], 3 * sizeof (int32_t));
+ pending[i],
+ AFR_NUM_CHANGE_LOGS * sizeof (int32_t));
/* 3 = data+metadata+entry */
if (ret < 0)
goto out;
}
-
-out:
- return ret;
-}
-
-
-static int
-afr_set_piggyback_dict (afr_private_t *priv, dict_t *xattr, int32_t **pending,
- afr_transaction_type type)
-{
- int i = 0;
- int ret = 0;
- int *arr = NULL;
- int index = 0;
- size_t pending_xattr_size = 3 * sizeof (int32_t);
- /* 3 = data+metadata+entry */
-
- index = afr_index_for_transaction_type (type);
-
- for (i = 0; i < priv->child_count; i++) {
- arr = GF_CALLOC (1, pending_xattr_size,
- gf_afr_mt_char);
- if (!arr) {
- ret = -1;
- goto out;
- }
-
- memcpy (arr, pending[i], pending_xattr_size);
-
- arr[index] = hton32 (ntoh32(arr[index]) + 1);
-
- ret = dict_set_bin (xattr, priv->pending_key[i],
- arr, pending_xattr_size);
-
- if (ret < 0)
+ if (op == LOCAL_LAST) {
+ ret = dict_set_static_bin (xattr, priv->pending_key[child],
+ pending[child],
+ AFR_NUM_CHANGE_LOGS * sizeof (int32_t));
+ if (ret)
goto out;
}
-
out:
return ret;
}
-
int
afr_lock_server_count (afr_private_t *priv, afr_transaction_type type)
{
@@ -360,7 +348,8 @@ afr_lock_server_count (afr_private_t *priv, afr_transaction_type type)
int32_t
afr_changelog_post_op_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
- int32_t op_ret, int32_t op_errno, dict_t *xattr)
+ int32_t op_ret, int32_t op_errno, dict_t *xattr,
+ dict_t *xdata)
{
afr_internal_lock_t *int_lock = NULL;
afr_private_t *priv = NULL;
@@ -378,6 +367,11 @@ afr_changelog_post_op_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
UNLOCK (&frame->lock);
if (call_count == 0) {
+ if (local->transaction.resume_stub) {
+ call_resume (local->transaction.resume_stub);
+ local->transaction.resume_stub = NULL;
+ }
+
if (afr_lock_server_count (priv, local->transaction.type) == 0) {
local->transaction.done (frame, this);
} else {
@@ -444,24 +438,41 @@ afr_transaction_rm_stale_children (call_frame_t *frame, xlator_t *this,
afr_inode_rm_stale_children (this, inode, stale_children);
out:
- if (stale_children)
- GF_FREE (stale_children);
+ GF_FREE (stale_children);
return;
}
+afr_inodelk_t*
+afr_get_inodelk (afr_internal_lock_t *int_lock, char *dom)
+{
+ afr_inodelk_t *inodelk = NULL;
+ int i = 0;
+
+ for (i = 0; int_lock->inodelk[i].domain; i++) {
+ inodelk = &int_lock->inodelk[i];
+ if (strcmp (dom, inodelk->domain) == 0)
+ return inodelk;
+ }
+ return NULL;
+}
+
unsigned char*
afr_locked_nodes_get (afr_transaction_type type, afr_internal_lock_t *int_lock)
{
unsigned char *locked_nodes = NULL;
+ afr_inodelk_t *inodelk = NULL;
switch (type) {
case AFR_DATA_TRANSACTION:
case AFR_METADATA_TRANSACTION:
- locked_nodes = int_lock->inode_locked_nodes;
+ inodelk = afr_get_inodelk (int_lock, int_lock->domain);
+ locked_nodes = inodelk->locked_nodes;
break;
case AFR_ENTRY_TRANSACTION:
case AFR_ENTRY_RENAME_TRANSACTION:
- locked_nodes = int_lock->entry_locked_nodes;
+ /*Because same set of subvols participate in all lockee
+ * entities*/
+ locked_nodes = int_lock->lockee[0].locked_nodes;
break;
}
return locked_nodes;
@@ -499,12 +510,141 @@ afr_changelog_post_op_call_count (afr_transaction_type type,
return call_count;
}
+void
+afr_compute_txn_changelog (afr_local_t *local, afr_private_t *priv)
+{
+ int i = 0;
+ int index = 0;
+ int32_t postop = 0;
+ int32_t preop = 1;
+ int32_t **txn_changelog = NULL;
+
+ txn_changelog = local->transaction.txn_changelog;
+ index = afr_index_for_transaction_type (local->transaction.type);
+ for (i = 0; i < priv->child_count; i++) {
+ postop = ntoh32 (local->pending[i][index]);
+ txn_changelog[i][index] = hton32 (postop + preop);
+ }
+}
+
+afr_xattrop_type_t
+afr_get_postop_xattrop_type (int32_t **pending, int optimized, int child,
+ afr_transaction_type type)
+{
+ int index = 0;
+ afr_xattrop_type_t op = LOCAL_LAST;
+
+ index = afr_index_for_transaction_type (type);
+ if (optimized && !pending[child][index])
+ op = LOCAL_FIRST;
+ return op;
+}
+
+void
+afr_set_postop_dict (afr_local_t *local, xlator_t *this, dict_t *xattr,
+ int optimized, int child)
+{
+ int32_t **txn_changelog = NULL;
+ int32_t **changelog = NULL;
+ afr_private_t *priv = NULL;
+ int ret = 0;
+ afr_xattrop_type_t op = LOCAL_LAST;
+
+ priv = this->private;
+ txn_changelog = local->transaction.txn_changelog;
+ op = afr_get_postop_xattrop_type (local->pending, optimized, child,
+ local->transaction.type);
+ if (optimized)
+ changelog = txn_changelog;
+ else
+ changelog = local->pending;
+ ret = afr_set_pending_dict (priv, xattr, changelog, child, op);
+ if (ret < 0)
+ gf_log (this->name, GF_LOG_INFO,
+ "failed to set pending entry");
+}
+
+
+gf_boolean_t
+afr_txn_nothing_failed (call_frame_t *frame, xlator_t *this)
+{
+ afr_private_t *priv = NULL;
+ afr_local_t *local = NULL;
+ int index = -1;
+ int i = 0;
+
+ local = frame->local;
+ priv = this->private;
+
+ index = afr_index_for_transaction_type (local->transaction.type);
+
+ for (i = 0; i < priv->child_count; i++) {
+ if (local->pending[i][index] == 0)
+ return _gf_false;
+ }
+
+ return _gf_true;
+}
+
+static void
+afr_dir_fop_handle_all_fop_failures (call_frame_t *frame)
+{
+ xlator_t *this = NULL;
+ afr_local_t *local = NULL;
+ afr_private_t *priv = NULL;
+
+ this = frame->this;
+ local = frame->local;
+ priv = this->private;
+
+ if ((local->transaction.type != AFR_ENTRY_TRANSACTION) &&
+ (local->transaction.type != AFR_ENTRY_RENAME_TRANSACTION))
+ return;
+
+ if (local->op_ret >= 0)
+ goto out;
+
+ __mark_all_success (local->pending, priv->child_count,
+ local->transaction.type);
+out:
+ return;
+}
+
+static void
+afr_data_handle_quota_errors (call_frame_t *frame, xlator_t *this)
+{
+ int i = 0;
+ afr_private_t *priv = NULL;
+ afr_local_t *local = NULL;
+ gf_boolean_t all_quota_failures = _gf_false;
+
+ local = frame->local;
+ priv = this->private;
+ if (local->transaction.type != AFR_DATA_TRANSACTION)
+ return;
+ /*
+ * Idea is to not leave the file in FOOL-FOOL scenario in case on
+ * all the bricks data transaction failed with EDQUOT to avoid
+ * increasing un-necessary load of self-heals in the system.
+ */
+ all_quota_failures = _gf_true;
+ for (i = 0; i < priv->child_count; i++) {
+ if (local->transaction.pre_op[i] &&
+ (local->child_errno[i] != EDQUOT)) {
+ all_quota_failures = _gf_false;
+ break;
+ }
+ }
+ if (all_quota_failures)
+ __mark_all_success (local->pending, priv->child_count,
+ local->transaction.type);
+}
+
int
-afr_changelog_post_op (call_frame_t *frame, xlator_t *this)
+afr_changelog_post_op_now (call_frame_t *frame, xlator_t *this)
{
afr_private_t * priv = this->private;
afr_internal_lock_t *int_lock = NULL;
- int ret = 0;
int i = 0;
int call_count = 0;
@@ -512,14 +652,17 @@ afr_changelog_post_op (call_frame_t *frame, xlator_t *this)
afr_fd_ctx_t *fdctx = NULL;
dict_t **xattr = NULL;
int piggyback = 0;
- int index = 0;
int nothing_failed = 1;
local = frame->local;
int_lock = &local->internal_lock;
- __mark_down_children (local->pending, priv->child_count,
- local->child_up, local->transaction.type);
+ __mark_non_participant_children (local->pending, priv->child_count,
+ local->transaction.pre_op,
+ local->transaction.type);
+
+ afr_data_handle_quota_errors (frame, this);
+ afr_dir_fop_handle_all_fop_failures (frame);
if (local->fd)
afr_transaction_rm_stale_children (frame, this,
@@ -547,82 +690,65 @@ afr_changelog_post_op (call_frame_t *frame, xlator_t *this)
goto out;
}
- /* check if something has failed, to handle piggybacking */
- nothing_failed = 1;
- index = afr_index_for_transaction_type (local->transaction.type);
- for (i = 0; i < priv->child_count; i++) {
- if (local->pending[i][index] == 0) {
- nothing_failed = 0;
- break;
- }
- }
+ nothing_failed = afr_txn_nothing_failed (frame, this);
- index = afr_index_for_transaction_type (local->transaction.type);
- if (local->optimistic_change_log &&
- local->transaction.type != AFR_DATA_TRANSACTION) {
- /* if nothing_failed, then local->pending[..] == {0 .. 0} */
- for (i = 0; i < priv->child_count; i++)
- local->pending[i][index]++;
- }
+ afr_compute_txn_changelog (local , priv);
for (i = 0; i < priv->child_count; i++) {
if (!local->transaction.pre_op[i])
continue;
- ret = afr_set_pending_dict (priv, xattr[i],
- local->pending);
-
- if (ret < 0)
- gf_log (this->name, GF_LOG_INFO,
- "failed to set pending entry");
-
+ if (local->transaction.type != AFR_DATA_TRANSACTION)
+ afr_set_postop_dict (local, this, xattr[i],
+ local->optimistic_change_log, i);
switch (local->transaction.type) {
case AFR_DATA_TRANSACTION:
{
if (!fdctx) {
+ afr_set_postop_dict (local, this, xattr[i],
+ 0, i);
STACK_WIND (frame, afr_changelog_post_op_cbk,
priv->children[i],
priv->children[i]->fops->xattrop,
&local->loc,
- GF_XATTROP_ADD_ARRAY, xattr[i]);
+ GF_XATTROP_ADD_ARRAY, xattr[i],
+ NULL);
break;
}
- LOCK (&local->fd->lock);
- {
- piggyback = 0;
- if (fdctx->pre_op_piggyback[i]) {
- fdctx->pre_op_piggyback[i]--;
- piggyback = 1;
- }
- }
- UNLOCK (&local->fd->lock);
+ /* local->transaction.postop_piggybacked[] was
+ precomputed in is_piggyback_postop() when called from
+ afr_changelog_post_op_safe()
+ */
- if (piggyback && !nothing_failed)
- ret = afr_set_piggyback_dict (priv, xattr[i],
- local->pending,
- local->transaction.type);
+ piggyback = 0;
+ if (local->transaction.postop_piggybacked[i])
+ piggyback = 1;
+
+ afr_set_postop_dict (local, this, xattr[i],
+ piggyback, i);
if (nothing_failed && piggyback) {
afr_changelog_post_op_cbk (frame, (void *)(long)i,
- this, 1, 0, xattr[i]);
+ this, 1, 0, xattr[i], NULL);
} else {
- __mark_pre_op_undone_on_fd (frame, this, i);
STACK_WIND_COOKIE (frame,
afr_changelog_post_op_cbk,
(void *) (long) i,
priv->children[i],
priv->children[i]->fops->fxattrop,
local->fd,
- GF_XATTROP_ADD_ARRAY, xattr[i]);
+ GF_XATTROP_ADD_ARRAY, xattr[i],
+ NULL);
}
}
break;
case AFR_METADATA_TRANSACTION:
{
- if (nothing_failed) {
+ if (nothing_failed && local->optimistic_change_log) {
afr_changelog_post_op_cbk (frame, (void *)(long)i,
- this, 1, 0, xattr[i]);
+ this, 1, 0, xattr[i],
+ NULL);
break;
}
@@ -631,28 +757,32 @@ afr_changelog_post_op (call_frame_t *frame, xlator_t *this)
priv->children[i],
priv->children[i]->fops->fxattrop,
local->fd,
- GF_XATTROP_ADD_ARRAY, xattr[i]);
+ GF_XATTROP_ADD_ARRAY, xattr[i],
+ NULL);
else
STACK_WIND (frame, afr_changelog_post_op_cbk,
priv->children[i],
priv->children[i]->fops->xattrop,
&local->loc,
- GF_XATTROP_ADD_ARRAY, xattr[i]);
+ GF_XATTROP_ADD_ARRAY, xattr[i],
+ NULL);
}
break;
case AFR_ENTRY_RENAME_TRANSACTION:
{
- if (nothing_failed) {
+ if (nothing_failed && local->optimistic_change_log) {
afr_changelog_post_op_cbk (frame, (void *)(long)i,
- this, 1, 0, xattr[i]);
+ this, 1, 0, xattr[i],
+ NULL);
} else {
STACK_WIND_COOKIE (frame, afr_changelog_post_op_cbk,
(void *) (long) i,
priv->children[i],
priv->children[i]->fops->xattrop,
&local->transaction.new_parent_loc,
- GF_XATTROP_ADD_ARRAY, xattr[i]);
+ GF_XATTROP_ADD_ARRAY, xattr[i],
+ NULL);
}
call_count--;
}
@@ -665,20 +795,17 @@ afr_changelog_post_op (call_frame_t *frame, xlator_t *this)
value
*/
- ret = afr_set_pending_dict (priv, xattr[i],
- local->pending);
-
- if (ret < 0)
- gf_log (this->name, GF_LOG_INFO,
- "failed to set pending entry");
+ afr_set_postop_dict (local, this, xattr[i],
+ local->optimistic_change_log, i);
/* fall through */
case AFR_ENTRY_TRANSACTION:
{
- if (nothing_failed) {
+ if (nothing_failed && local->optimistic_change_log) {
afr_changelog_post_op_cbk (frame, (void *)(long)i,
- this, 1, 0, xattr[i]);
+ this, 1, 0, xattr[i],
+ NULL);
break;
}
@@ -687,13 +814,15 @@ afr_changelog_post_op (call_frame_t *frame, xlator_t *this)
priv->children[i],
priv->children[i]->fops->fxattrop,
local->fd,
- GF_XATTROP_ADD_ARRAY, xattr[i]);
+ GF_XATTROP_ADD_ARRAY, xattr[i],
+ NULL);
else
STACK_WIND (frame, afr_changelog_post_op_cbk,
priv->children[i],
priv->children[i]->fops->xattrop,
&local->transaction.parent_loc,
- GF_XATTROP_ADD_ARRAY, xattr[i]);
+ GF_XATTROP_ADD_ARRAY, xattr[i],
+ NULL);
}
break;
}
@@ -713,7 +842,8 @@ out:
int32_t
afr_changelog_pre_op_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
- int32_t op_ret, int32_t op_errno, dict_t *xattr)
+ int32_t op_ret, int32_t op_errno, dict_t *xattr,
+ dict_t *xdata)
{
afr_local_t * local = NULL;
afr_private_t * priv = this->private;
@@ -758,12 +888,7 @@ afr_changelog_pre_op_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
(local->op_errno == ENOTSUP)) {
local->transaction.resume (frame, this);
} else {
- __mark_all_success (local->pending, priv->child_count,
- local->transaction.type);
-
- afr_pid_restore (frame);
-
- local->transaction.fop (frame, this);
+ afr_transaction_perform_fop (frame, this);
}
}
@@ -816,8 +941,8 @@ afr_changelog_pre_op (call_frame_t *frame, xlator_t *this)
for (i = 0; i < priv->child_count; i++) {
if (!locked_nodes[i])
continue;
- ret = afr_set_pending_dict (priv, xattr[i],
- local->pending);
+ ret = afr_set_pending_dict (priv, xattr[i], local->pending,
+ i, LOCAL_FIRST);
if (ret < 0)
gf_log (this->name, GF_LOG_INFO,
@@ -834,7 +959,8 @@ afr_changelog_pre_op (call_frame_t *frame, xlator_t *this)
priv->children[i],
priv->children[i]->fops->xattrop,
&(local->loc),
- GF_XATTROP_ADD_ARRAY, xattr[i]);
+ GF_XATTROP_ADD_ARRAY, xattr[i],
+ NULL);
break;
}
@@ -851,9 +977,12 @@ afr_changelog_pre_op (call_frame_t *frame, xlator_t *this)
}
UNLOCK (&local->fd->lock);
+ afr_set_delayed_post_op (frame, this);
+
if (piggyback)
afr_changelog_pre_op_cbk (frame, (void *)(long)i,
- this, 1, 0, xattr[i]);
+ this, 1, 0, xattr[i],
+ NULL);
else
STACK_WIND_COOKIE (frame,
afr_changelog_pre_op_cbk,
@@ -861,14 +990,16 @@ afr_changelog_pre_op (call_frame_t *frame, xlator_t *this)
priv->children[i],
priv->children[i]->fops->fxattrop,
local->fd,
- GF_XATTROP_ADD_ARRAY, xattr[i]);
+ GF_XATTROP_ADD_ARRAY, xattr[i],
+ NULL);
}
break;
case AFR_METADATA_TRANSACTION:
{
if (local->optimistic_change_log) {
afr_changelog_pre_op_cbk (frame, (void *)(long)i,
- this, 1, 0, xattr[i]);
+ this, 1, 0, xattr[i],
+ NULL);
break;
}
@@ -879,7 +1010,8 @@ afr_changelog_pre_op (call_frame_t *frame, xlator_t *this)
priv->children[i],
priv->children[i]->fops->fxattrop,
local->fd,
- GF_XATTROP_ADD_ARRAY, xattr[i]);
+ GF_XATTROP_ADD_ARRAY, xattr[i],
+ NULL);
else
STACK_WIND_COOKIE (frame,
afr_changelog_pre_op_cbk,
@@ -887,7 +1019,8 @@ afr_changelog_pre_op (call_frame_t *frame, xlator_t *this)
priv->children[i],
priv->children[i]->fops->xattrop,
&(local->loc),
- GF_XATTROP_ADD_ARRAY, xattr[i]);
+ GF_XATTROP_ADD_ARRAY, xattr[i],
+ NULL);
}
break;
@@ -895,7 +1028,8 @@ afr_changelog_pre_op (call_frame_t *frame, xlator_t *this)
{
if (local->optimistic_change_log) {
afr_changelog_pre_op_cbk (frame, (void *)(long)i,
- this, 1, 0, xattr[i]);
+ this, 1, 0, xattr[i],
+ NULL);
} else {
STACK_WIND_COOKIE (frame,
afr_changelog_pre_op_cbk,
@@ -903,7 +1037,8 @@ afr_changelog_pre_op (call_frame_t *frame, xlator_t *this)
priv->children[i],
priv->children[i]->fops->xattrop,
&local->transaction.new_parent_loc,
- GF_XATTROP_ADD_ARRAY, xattr[i]);
+ GF_XATTROP_ADD_ARRAY, xattr[i],
+ NULL);
}
call_count--;
@@ -918,8 +1053,8 @@ afr_changelog_pre_op (call_frame_t *frame, xlator_t *this)
value
*/
- ret = afr_set_pending_dict (priv, xattr[i],
- local->pending);
+ ret = afr_set_pending_dict (priv, xattr[i], local->pending,
+ i, LOCAL_FIRST);
if (ret < 0)
gf_log (this->name, GF_LOG_INFO,
@@ -931,7 +1066,8 @@ afr_changelog_pre_op (call_frame_t *frame, xlator_t *this)
{
if (local->optimistic_change_log) {
afr_changelog_pre_op_cbk (frame, (void *)(long)i,
- this, 1, 0, xattr[i]);
+ this, 1, 0, xattr[i],
+ NULL);
break;
}
@@ -942,7 +1078,8 @@ afr_changelog_pre_op (call_frame_t *frame, xlator_t *this)
priv->children[i],
priv->children[i]->fops->fxattrop,
local->fd,
- GF_XATTROP_ADD_ARRAY, xattr[i]);
+ GF_XATTROP_ADD_ARRAY, xattr[i],
+ NULL);
else
STACK_WIND_COOKIE (frame,
afr_changelog_pre_op_cbk,
@@ -950,7 +1087,8 @@ afr_changelog_pre_op (call_frame_t *frame, xlator_t *this)
priv->children[i],
priv->children[i]->fops->xattrop,
&local->transaction.parent_loc,
- GF_XATTROP_ADD_ARRAY, xattr[i]);
+ GF_XATTROP_ADD_ARRAY, xattr[i],
+ NULL);
}
break;
}
@@ -1112,12 +1250,14 @@ int
afr_set_transaction_flock (afr_local_t *local)
{
afr_internal_lock_t *int_lock = NULL;
+ afr_inodelk_t *inodelk = NULL;
int_lock = &local->internal_lock;
+ inodelk = afr_get_inodelk (int_lock, int_lock->domain);
- int_lock->lk_flock.l_len = local->transaction.len;
- int_lock->lk_flock.l_start = local->transaction.start;
- int_lock->lk_flock.l_type = F_WRLCK;
+ inodelk->flock.l_len = local->transaction.len;
+ inodelk->flock.l_start = local->transaction.start;
+ inodelk->flock.l_type = F_WRLCK;
return 0;
}
@@ -1132,6 +1272,7 @@ afr_lock_rec (call_frame_t *frame, xlator_t *this)
int_lock = &local->internal_lock;
int_lock->transaction_lk_type = AFR_TRANSACTION_LK;
+ int_lock->domain = this->name;
switch (local->transaction.type) {
case AFR_DATA_TRANSACTION:
@@ -1145,8 +1286,8 @@ afr_lock_rec (call_frame_t *frame, xlator_t *this)
case AFR_ENTRY_RENAME_TRANSACTION:
- int_lock->lock_cbk = afr_post_blocking_rename_cbk;
- afr_blocking_lock (frame, this);
+ int_lock->lock_cbk = afr_post_nonblocking_entrylk_cbk;
+ afr_nonblocking_entrylk (frame, this);
break;
case AFR_ENTRY_TRANSACTION:
@@ -1168,12 +1309,6 @@ afr_lock_rec (call_frame_t *frame, xlator_t *this)
int
afr_lock (call_frame_t *frame, xlator_t *this)
{
- afr_pid_save (frame);
-
- frame->root->pid = (long) frame->root;
-
- afr_set_lk_owner (frame, this);
-
afr_set_lock_number (frame, this);
return afr_lock_rec (frame, this);
@@ -1185,28 +1320,463 @@ afr_lock (call_frame_t *frame, xlator_t *this)
int
afr_internal_lock_finish (call_frame_t *frame, xlator_t *this)
{
- afr_local_t *local = NULL;
- afr_private_t *priv = NULL;
+ if (__fop_changelog_needed (frame, this)) {
+ afr_changelog_pre_op (frame, this);
+ } else {
+ afr_transaction_perform_fop (frame, this);
+ }
+
+ return 0;
+}
+
+
+void
+afr_set_delayed_post_op (call_frame_t *frame, xlator_t *this)
+{
+ afr_local_t *local = NULL;
+ afr_private_t *priv = NULL;
+
+ /* call this function from any of the related optimizations
+ which benefit from delaying post op are enabled, namely:
+
+ - changelog piggybacking
+ - eager locking
+ */
+
+ priv = this->private;
+ if (!priv)
+ return;
+
+ if (!priv->post_op_delay_secs)
+ return;
+
+ local = frame->local;
+ if (!local->transaction.eager_lock_on)
+ return;
+
+ if (!local)
+ return;
+
+ if (!local->fd)
+ return;
+
+ if (local->op == GF_FOP_WRITE)
+ local->delayed_post_op = _gf_true;
+}
+
+gf_boolean_t
+afr_are_multiple_fds_opened (inode_t *inode, xlator_t *this)
+{
+ afr_inode_ctx_t *ictx = NULL;
+
+ if (!inode) {
+ /* If false is returned, it may keep on taking eager-lock
+ * which may lead to starvation, so return true to avoid that.
+ */
+ gf_log_callingfn (this->name, GF_LOG_ERROR, "Invalid inode");
+ return _gf_true;
+ }
+ /* Lets say mount1 has eager-lock(full-lock) and after the eager-lock
+ * is taken mount2 opened the same file, it won't be able to
+ * perform any data operations until mount1 releases eager-lock.
+ * To avoid such scenario do not enable eager-lock for this transaction
+ * if open-fd-count is > 1
+ */
+
+ ictx = afr_inode_ctx_get (inode, this);
+ if (!ictx)
+ return _gf_true;
+
+ if (ictx->open_fd_count > 1)
+ return _gf_true;
+
+ return _gf_false;
+}
+
+gf_boolean_t
+afr_any_fops_failed (afr_local_t *local, afr_private_t *priv)
+{
+ if (local->success_count != priv->child_count)
+ return _gf_true;
+ return _gf_false;
+}
+
+gf_boolean_t
+is_afr_delayed_changelog_post_op_needed (call_frame_t *frame, xlator_t *this)
+{
+ afr_local_t *local = NULL;
+ gf_boolean_t res = _gf_false;
+ afr_private_t *priv = NULL;
priv = this->private;
+
local = frame->local;
+ if (!local)
+ goto out;
- if (__fop_changelog_needed (frame, this)) {
- afr_changelog_pre_op (frame, this);
- } else {
- __mark_all_success (local->pending, priv->child_count,
- local->transaction.type);
+ if (!local->delayed_post_op)
+ goto out;
+
+ //Mark pending changelog ASAP
+ if (afr_any_fops_failed (local, priv))
+ goto out;
+
+ if (local->fd && afr_are_multiple_fds_opened (local->fd->inode, this))
+ goto out;
+
+ res = _gf_true;
+out:
+ return res;
+}
+
+
+void
+afr_delayed_changelog_post_op (xlator_t *this, call_frame_t *frame, fd_t *fd,
+ call_stub_t *stub);
+
+void
+afr_delayed_changelog_wake_up_cbk (void *data)
+{
+ fd_t *fd = NULL;
+
+ fd = data;
+
+ afr_delayed_changelog_wake_up (THIS, fd);
+}
+
+
+/*
+ Check if the frame is destined to get optimized away
+ with changelog piggybacking
+*/
+static gf_boolean_t
+is_piggyback_post_op (call_frame_t *frame, fd_t *fd)
+{
+ afr_fd_ctx_t *fdctx = NULL;
+ afr_local_t *local = NULL;
+ gf_boolean_t piggyback = _gf_true;
+ afr_private_t *priv = NULL;
+ int i = 0;
+
+ priv = frame->this->private;
+ local = frame->local;
+ fdctx = afr_fd_ctx_get (fd, frame->this);
- afr_pid_restore (frame);
+ LOCK(&fd->lock);
+ {
+ piggyback = _gf_true;
+
+ for (i = 0; i < priv->child_count; i++) {
+ if (!local->transaction.pre_op[i])
+ continue;
+ if (fdctx->pre_op_piggyback[i]) {
+ fdctx->pre_op_piggyback[i]--;
+ local->transaction.postop_piggybacked[i] = 1;
+ } else {
+ /* For at least _one_ subvolume we cannot
+ piggyback on the changelog, and have to
+ perform a hard POST-OP and therefore fsync
+ if necesssary
+ */
+ piggyback = _gf_false;
+ GF_ASSERT (fdctx->pre_op_done[i]);
+ fdctx->pre_op_done[i]--;
+ }
+ }
+ }
+ UNLOCK(&fd->lock);
- local->transaction.fop (frame, this);
+ if (!afr_txn_nothing_failed (frame, frame->this)) {
+ /* something failed in this transaction,
+ we will be performing a hard post-op
+ */
+ return _gf_false;
}
+ return piggyback;
+}
+
+
+/* SET operation */
+int
+afr_fd_report_unstable_write (xlator_t *this, fd_t *fd)
+{
+ afr_fd_ctx_t *fdctx = NULL;
+
+ fdctx = afr_fd_ctx_get (fd, this);
+
+ LOCK(&fd->lock);
+ {
+ fdctx->witnessed_unstable_write = _gf_true;
+ }
+ UNLOCK(&fd->lock);
+
return 0;
}
+/* TEST and CLEAR operation */
+gf_boolean_t
+afr_fd_has_witnessed_unstable_write (xlator_t *this, fd_t *fd)
+{
+ afr_fd_ctx_t *fdctx = NULL;
+ gf_boolean_t witness = _gf_false;
+
+ fdctx = afr_fd_ctx_get (fd, this);
+ if (!fdctx)
+ return _gf_true;
+
+ LOCK(&fd->lock);
+ {
+ if (fdctx->witnessed_unstable_write) {
+ witness = _gf_true;
+ fdctx->witnessed_unstable_write = _gf_false;
+ }
+ }
+ UNLOCK (&fd->lock);
+
+ return witness;
+}
+
int
+afr_changelog_fsync_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
+ int op_ret, int op_errno, struct iatt *pre,
+ struct iatt *post, dict_t *xdata)
+{
+ afr_private_t *priv = NULL;
+ int child_index = (long) cookie;
+ int call_count = -1;
+ afr_local_t *local = NULL;
+
+ priv = this->private;
+ local = frame->local;
+
+ if (afr_fop_failed (op_ret, op_errno)) {
+ /* Failure of fsync() is as good as failure of previous
+ write(). So treat it like one.
+ */
+ gf_log (this->name, GF_LOG_WARNING,
+ "fsync(%s) failed on subvolume %s. Transaction was %s",
+ uuid_utoa (local->fd->inode->gfid),
+ priv->children[child_index]->name,
+ gf_fop_list[local->op]);
+
+ afr_transaction_fop_failed (frame, this, child_index);
+ }
+
+ call_count = afr_frame_return (frame);
+
+ if (call_count == 0)
+ afr_changelog_post_op_now (frame, this);
+
+ return 0;
+}
+
+
+int
+afr_changelog_fsync (call_frame_t *frame, xlator_t *this)
+{
+ afr_local_t *local = NULL;
+ int i = 0;
+ int call_count = 0;
+ afr_private_t *priv = NULL;
+ dict_t *xdata = NULL;
+ GF_UNUSED int ret = -1;
+
+ local = frame->local;
+ priv = this->private;
+
+ call_count = afr_pre_op_done_children_count (local->transaction.pre_op,
+ priv->child_count);
+
+ if (!call_count) {
+ /* will go straight to unlock */
+ afr_changelog_post_op_now (frame, this);
+ return 0;
+ }
+
+ local->call_count = call_count;
+
+ xdata = dict_new();
+ if (xdata)
+ ret = dict_set_int32 (xdata, "batch-fsync", 1);
+
+ for (i = 0; i < priv->child_count; i++) {
+ if (!local->transaction.pre_op[i])
+ continue;
+
+ STACK_WIND_COOKIE (frame, afr_changelog_fsync_cbk,
+ (void *) (long) i, priv->children[i],
+ priv->children[i]->fops->fsync, local->fd,
+ 1, xdata);
+ if (!--call_count)
+ break;
+ }
+
+ if (xdata)
+ dict_unref (xdata);
+
+ return 0;
+}
+
+
+ int
+afr_changelog_post_op_safe (call_frame_t *frame, xlator_t *this)
+{
+ afr_local_t *local = NULL;
+ afr_private_t *priv = NULL;
+
+ local = frame->local;
+ priv = this->private;
+
+ if (!local->fd || local->transaction.type != AFR_DATA_TRANSACTION) {
+ afr_changelog_post_op_now (frame, this);
+ return 0;
+ }
+
+ if (is_piggyback_post_op (frame, local->fd)) {
+ /* just detected that this post-op is about to
+ be optimized away as a new write() has
+ already piggybacked on this frame's changelog.
+ */
+ afr_changelog_post_op_now (frame, this);
+ return 0;
+ }
+
+ /* Calling afr_changelog_post_op_now() now will result in
+ issuing ->[f]xattrop().
+
+ Performing a hard POST-OP (->[f]xattrop() FOP) is a more
+ responsible operation that what it might appear on the surface.
+
+ The changelog of a file (in the xattr of the file on the server)
+ stores information (pending count) about the state of the file
+ on the OTHER server. This changelog is blindly trusted, and must
+ therefore be updated in such a way it remains trustworthy. This
+ implies that decrementing the pending count (essentially "clearing
+ the dirty flag") must be done STRICTLY after we are sure that the
+ operation on the other server has reached stable storage.
+
+ While the backend filesystem on that server will eventually flush
+ it to stable storage, we (being in userspace) have no mechanism
+ to get notified when the write became "stable".
+
+ This means we need take matter into our own hands and issue an
+ fsync() EVEN IF THE APPLICATION WAS PERFORMING UNSTABLE WRITES,
+ and get an acknowledgement for it. And we need to wait for the
+ fsync() acknowledgement before initiating the hard POST-OP.
+
+ However if the FD itself was opened in O_SYNC or O_DSYNC then
+ we are already guaranteed that the writes were made stable as
+ part of the FOP itself. The same holds true for NFS stable
+ writes which happen on an anonymous FD with O_DSYNC or O_SYNC
+ flag set in the writev() @flags param. For all other write types,
+ mark a flag in the fdctx whenever an unstable write is witnessed.
+ */
+
+ if (!afr_fd_has_witnessed_unstable_write (this, local->fd)) {
+ afr_changelog_post_op_now (frame, this);
+ return 0;
+ }
+
+ /* Check whether users want durability and perform fsync/post-op
+ * accordingly.
+ */
+ if (priv->ensure_durability) {
+ /* Time to fsync() */
+ afr_changelog_fsync (frame, this);
+ } else {
+ afr_changelog_post_op_now (frame, this);
+ }
+
+ return 0;
+}
+
+
+void
+afr_delayed_changelog_post_op (xlator_t *this, call_frame_t *frame, fd_t *fd,
+ call_stub_t *stub)
+{
+ afr_fd_ctx_t *fd_ctx = NULL;
+ call_frame_t *prev_frame = NULL;
+ struct timespec delta = {0, };
+ afr_private_t *priv = NULL;
+ afr_local_t *local = NULL;
+
+ priv = this->private;
+
+ fd_ctx = afr_fd_ctx_get (fd, this);
+ if (!fd_ctx)
+ goto out;
+
+ delta.tv_sec = priv->post_op_delay_secs;
+ delta.tv_nsec = 0;
+
+ pthread_mutex_lock (&fd_ctx->delay_lock);
+ {
+ prev_frame = fd_ctx->delay_frame;
+ fd_ctx->delay_frame = NULL;
+ if (fd_ctx->delay_timer)
+ gf_timer_call_cancel (this->ctx, fd_ctx->delay_timer);
+ fd_ctx->delay_timer = NULL;
+ if (!frame)
+ goto unlock;
+ fd_ctx->delay_timer = gf_timer_call_after (this->ctx, delta,
+ afr_delayed_changelog_wake_up_cbk,
+ fd);
+ fd_ctx->delay_frame = frame;
+ }
+unlock:
+ pthread_mutex_unlock (&fd_ctx->delay_lock);
+
+out:
+ if (prev_frame) {
+ local = prev_frame->local;
+ local->transaction.resume_stub = stub;
+ afr_changelog_post_op_safe (prev_frame, this);
+ } else if (stub) {
+ call_resume (stub);
+ }
+}
+
+
+void
+afr_changelog_post_op (call_frame_t *frame, xlator_t *this)
+{
+ afr_local_t *local = NULL;
+
+ local = frame->local;
+
+ if (is_afr_delayed_changelog_post_op_needed (frame, this))
+ afr_delayed_changelog_post_op (this, frame, local->fd, NULL);
+ else
+ afr_changelog_post_op_safe (frame, this);
+}
+
+
+
+/* Wake up the sleeping/delayed post-op, and also register
+ a stub to have it resumed after this transaction
+ completely finishes.
+
+ The @stub gets saved in @local and gets resumed in
+ afr_local_cleanup()
+ */
+void
+afr_delayed_changelog_wake_resume (xlator_t *this, fd_t *fd, call_stub_t *stub)
+{
+ afr_delayed_changelog_post_op (this, NULL, fd, stub);
+}
+
+
+void
+afr_delayed_changelog_wake_up (xlator_t *this, fd_t *fd)
+{
+ afr_delayed_changelog_post_op (this, NULL, fd, NULL);
+}
+
+
+ int
afr_transaction_resume (call_frame_t *frame, xlator_t *this)
{
afr_internal_lock_t *int_lock = NULL;
@@ -1217,6 +1787,19 @@ afr_transaction_resume (call_frame_t *frame, xlator_t *this)
int_lock = &local->internal_lock;
priv = this->private;
+ if (local->transaction.eager_lock_on) {
+ /* We don't need to retain "local" in the
+ fd list anymore, writes to all subvols
+ are finished by now */
+ LOCK (&local->fd->lock);
+ {
+ list_del_init (&local->transaction.eager_locked);
+ }
+ UNLOCK (&local->fd->lock);
+ }
+
+ afr_restore_lk_owner (frame);
+
if (__fop_changelog_needed (frame, this)) {
afr_changelog_post_op (frame, this);
} else {
@@ -1237,7 +1820,8 @@ afr_transaction_resume (call_frame_t *frame, xlator_t *this)
*/
void
-afr_transaction_fop_failed (call_frame_t *frame, xlator_t *this, int child_index)
+afr_transaction_fop_failed (call_frame_t *frame, xlator_t *this,
+ int child_index)
{
afr_local_t * local = NULL;
afr_private_t * priv = NULL;
@@ -1246,7 +1830,89 @@ afr_transaction_fop_failed (call_frame_t *frame, xlator_t *this, int child_index
priv = this->private;
__mark_child_dead (local->pending, priv->child_count,
- child_index, local->transaction.type);
+ child_index, local->transaction.type);
+}
+
+
+
+ static gf_boolean_t
+afr_locals_overlap (afr_local_t *local1, afr_local_t *local2)
+{
+ uint64_t start1 = local1->transaction.start;
+ uint64_t start2 = local2->transaction.start;
+ uint64_t end1 = 0;
+ uint64_t end2 = 0;
+
+ if (local1->transaction.len)
+ end1 = start1 + local1->transaction.len - 1;
+ else
+ end1 = ULLONG_MAX;
+
+ if (local2->transaction.len)
+ end2 = start2 + local2->transaction.len - 1;
+ else
+ end2 = ULLONG_MAX;
+
+ return ((end1 >= start2) && (end2 >= start1));
+}
+
+void
+afr_transaction_eager_lock_init (afr_local_t *local, xlator_t *this)
+{
+ afr_private_t *priv = NULL;
+ afr_fd_ctx_t *fdctx = NULL;
+ afr_local_t *each = NULL;
+
+ priv = this->private;
+
+ if (!local->fd)
+ return;
+
+ if (local->transaction.type != AFR_DATA_TRANSACTION)
+ return;
+
+ if (!priv->eager_lock)
+ return;
+
+ fdctx = afr_fd_ctx_get (local->fd, this);
+ if (!fdctx)
+ return;
+
+ if (afr_are_multiple_fds_opened (local->fd->inode, this))
+ return;
+ /*
+ * Once full file lock is acquired in eager-lock phase, overlapping
+ * writes do not compete for inode-locks, instead are transferred to the
+ * next writes. Because of this overlapping writes are not ordered.
+ * This can cause inconsistencies in replication.
+ * Example:
+ * Two overlapping writes w1, w2 are sent in parallel on same fd
+ * in two threads t1, t2.
+ * Both threads can execute afr_writev_wind in the following manner.
+ * t1 winds w1 on brick-0
+ * t2 winds w2 on brick-0
+ * t2 winds w2 on brick-1
+ * t1 winds w1 on brick-1
+ *
+ * This check makes sure the locks are not transferred for
+ * overlapping writes.
+ */
+ LOCK (&local->fd->lock);
+ {
+ list_for_each_entry (each, &fdctx->eager_locked,
+ transaction.eager_locked) {
+ if (afr_locals_overlap (each, local)) {
+ local->transaction.eager_lock_on = _gf_false;
+ goto unlock;
+ }
+ }
+
+ local->transaction.eager_lock_on = _gf_true;
+ list_add_tail (&local->transaction.eager_locked,
+ &fdctx->eager_locked);
+ }
+unlock:
+ UNLOCK (&local->fd->lock);
}
@@ -1255,20 +1921,43 @@ afr_transaction (call_frame_t *frame, xlator_t *this, afr_transaction_type type)
{
afr_local_t * local = NULL;
afr_private_t * priv = NULL;
+ fd_t *fd = NULL;
+ int ret = -1;
local = frame->local;
priv = this->private;
- afr_transaction_local_init (local, this);
-
local->transaction.resume = afr_transaction_resume;
local->transaction.type = type;
+ ret = afr_transaction_local_init (local, this);
+ if (ret < 0)
+ goto out;
+
+ afr_transaction_eager_lock_init (local, this);
+
+ if (local->fd && local->transaction.eager_lock_on)
+ afr_set_lk_owner (frame, this, local->fd);
+ else
+ afr_set_lk_owner (frame, this, frame->root);
+
+ if (!local->transaction.eager_lock_on && local->loc.inode) {
+ fd = fd_lookup (local->loc.inode, frame->root->pid);
+ if (fd == NULL)
+ fd = fd_lookup_anonymous (local->loc.inode);
+
+ if (fd) {
+ afr_delayed_changelog_wake_up (this, fd);
+ fd_unref (fd);
+ }
+ }
+
if (afr_lock_server_count (priv, local->transaction.type) == 0) {
afr_internal_lock_finish (frame, this);
} else {
afr_lock (frame, this);
}
-
- return 0;
+ ret = 0;
+out:
+ return ret;
}
diff --git a/xlators/cluster/afr/src/afr-transaction.h b/xlators/cluster/afr/src/afr-transaction.h
index 10f274fec..fa626fd0d 100644
--- a/xlators/cluster/afr/src/afr-transaction.h
+++ b/xlators/cluster/afr/src/afr-transaction.h
@@ -1,25 +1,21 @@
/*
- Copyright (c) 2007-2011 Gluster, Inc. <http://www.gluster.com>
- This file is part of GlusterFS.
-
- GlusterFS is free software; you can redistribute it and/or modify
- it under the terms of the GNU General Public License as published
- by the Free Software Foundation; either version 3 of the License,
- or (at your option) any later version.
-
- GlusterFS is distributed in the hope that it will be useful, but
- WITHOUT ANY WARRANTY; without even the implied warranty of
- MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU
- General Public License for more details.
-
- You should have received a copy of the GNU General Public License
- along with this program. If not, see
- <http://www.gnu.org/licenses/>.
+ Copyright (c) 2008-2012 Red Hat, Inc. <http://www.redhat.com>
+ This file is part of GlusterFS.
+
+ This file is licensed to you under your choice of the GNU Lesser
+ General Public License, version 3 or any later version (LGPLv3 or
+ later), or the GNU General Public License, version 2 (GPLv2), in all
+ cases as published by the Free Software Foundation.
*/
#ifndef __TRANSACTION_H__
#define __TRANSACTION_H__
+typedef enum {
+ LOCAL_FIRST = 1,
+ LOCAL_LAST = 2
+} afr_xattrop_type_t;
+
void
afr_transaction_fop_failed (call_frame_t *frame, xlator_t *this,
int child_index);
@@ -27,9 +23,29 @@ afr_transaction_fop_failed (call_frame_t *frame, xlator_t *this,
int
afr_lock_server_count (afr_private_t *priv, afr_transaction_type type);
+afr_inodelk_t*
+afr_get_inodelk (afr_internal_lock_t *int_lock, char *dom);
+
int32_t
afr_transaction (call_frame_t *frame, xlator_t *this, afr_transaction_type type);
afr_fd_ctx_t *
afr_fd_ctx_get (fd_t *fd, xlator_t *this);
+int
+afr_set_pending_dict (afr_private_t *priv, dict_t *xattr, int32_t **pending,
+ int child, afr_xattrop_type_t op);
+void
+afr_set_delayed_post_op (call_frame_t *frame, xlator_t *this);
+
+void
+afr_delayed_changelog_wake_up (xlator_t *this, fd_t *fd);
+
+void
+__mark_all_success (int32_t *pending[], int child_count,
+ afr_transaction_type type);
+gf_boolean_t
+afr_any_fops_failed (afr_local_t *local, afr_private_t *priv);
+
+gf_boolean_t
+afr_txn_nothing_failed (call_frame_t *frame, xlator_t *this);
#endif /* __TRANSACTION_H__ */
diff --git a/xlators/cluster/afr/src/afr.c b/xlators/cluster/afr/src/afr.c
index 22be68a65..c724eb2ae 100644
--- a/xlators/cluster/afr/src/afr.c
+++ b/xlators/cluster/afr/src/afr.c
@@ -1,20 +1,11 @@
/*
- Copyright (c) 2007-2011 Gluster, Inc. <http://www.gluster.com>
+ Copyright (c) 2008-2012 Red Hat, Inc. <http://www.redhat.com>
This file is part of GlusterFS.
- GlusterFS is free software; you can redistribute it and/or modify
- it under the terms of the GNU General Public License as published
- by the Free Software Foundation; either version 3 of the License,
- or (at your option) any later version.
-
- GlusterFS is distributed in the hope that it will be useful, but
- WITHOUT ANY WARRANTY; without even the implied warranty of
- MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU
- General Public License for more details.
-
- You should have received a copy of the GNU General Public License
- along with this program. If not, see
- <http://www.gnu.org/licenses/>.
+ This file is licensed to you under your choice of the GNU Lesser
+ General Public License, version 3 or any later version (LGPLv3 or
+ later), or the GNU General Public License, version 2 (GPLv2), in all
+ cases as published by the Free Software Foundation.
*/
#include <libgen.h>
@@ -30,6 +21,11 @@
#endif
#include "afr-common.c"
+#define SHD_INODE_LRU_LIMIT 2048
+#define AFR_EH_HEALED_LIMIT 1024
+#define AFR_EH_HEAL_FAIL_LIMIT 1024
+#define AFR_EH_SPLIT_BRAIN_LIMIT 1024
+
struct volume_options options[];
int32_t
@@ -37,8 +33,13 @@ notify (xlator_t *this, int32_t event,
void *data, ...)
{
int ret = -1;
+ va_list ap;
+ void *data2 = NULL;
- ret = afr_notify (this, event, data);
+ va_start (ap, data);
+ data2 = va_arg (ap, dict_t*);
+ va_end (ap);
+ ret = afr_notify (this, event, data, data2);
return ret;
}
@@ -106,6 +107,7 @@ reconfigure (xlator_t *this, dict_t *options)
{
afr_private_t *priv = NULL;
xlator_t *read_subvol = NULL;
+ int read_subvol_index = -1;
int ret = -1;
int index = -1;
char *qtype = NULL;
@@ -148,6 +150,9 @@ reconfigure (xlator_t *this, dict_t *options)
GF_OPTION_RECONF ("read-subvolume", read_subvol, options, xlator, out);
+ GF_OPTION_RECONF ("read-hash-mode", priv->hash_mode,
+ options, uint32, out);
+
if (read_subvol) {
index = xlator_subvolume_index (this, read_subvol);
if (index == -1) {
@@ -158,10 +163,37 @@ reconfigure (xlator_t *this, dict_t *options)
priv->read_child = index;
}
+ GF_OPTION_RECONF ("read-subvolume-index",read_subvol_index, options,int32,out);
+
+ if (read_subvol_index >-1) {
+ index=read_subvol_index;
+ if (index >= priv->child_count) {
+ gf_log (this->name, GF_LOG_ERROR, "%d not a subvolume-index",
+ index);
+ goto out;
+ }
+ priv->read_child = index;
+ }
+
+ GF_OPTION_RECONF ("eager-lock", priv->eager_lock, options, bool, out);
GF_OPTION_RECONF ("quorum-type", qtype, options, str, out);
GF_OPTION_RECONF ("quorum-count", priv->quorum_count, options,
uint32, out);
fix_quorum_options(this,priv,qtype);
+ GF_OPTION_RECONF ("heal-timeout", priv->shd.timeout, options,
+ int32, out);
+
+ GF_OPTION_RECONF ("post-op-delay-secs", priv->post_op_delay_secs, options,
+ uint32, out);
+
+ GF_OPTION_RECONF (AFR_SH_READDIR_SIZE_KEY, priv->sh_readdir_size,
+ options, size, out);
+ /* Reset this so we re-discover in case the topology changed. */
+ GF_OPTION_RECONF ("readdir-failover", priv->readdir_failover, options,
+ bool, out);
+ GF_OPTION_RECONF ("ensure-durability", priv->ensure_durability, options,
+ bool, out);
+ priv->did_discovery = _gf_false;
ret = 0;
out:
@@ -189,6 +221,7 @@ init (xlator_t *this)
int ret = -1;
GF_UNUSED int op_errno = 0;
xlator_t *read_subvol = NULL;
+ int read_subvol_index = -1;
xlator_t *fav_child = NULL;
char *qtype = NULL;
@@ -204,7 +237,10 @@ init (xlator_t *this)
"Volume is dangling.");
}
- ALLOC_OR_GOTO (this->private, afr_private_t, out);
+ this->private = GF_CALLOC (1, sizeof (afr_private_t),
+ gf_afr_mt_afr_private_t);
+ if (!this->private)
+ goto out;
priv = this->private;
LOCK_INIT (&priv->lock);
@@ -217,7 +253,6 @@ init (xlator_t *this)
priv->child_count = child_count;
-
priv->read_child = -1;
GF_OPTION_INIT ("read-subvolume", read_subvol, xlator, out);
@@ -229,6 +264,18 @@ init (xlator_t *this)
goto out;
}
}
+ GF_OPTION_INIT ("read-subvolume-index",read_subvol_index,int32,out);
+ if (read_subvol_index > -1) {
+ if (read_subvol_index >= priv->child_count) {
+ gf_log (this->name, GF_LOG_ERROR, "%d not a subvolume-index",
+ read_subvol_index);
+ goto out;
+ }
+ priv->read_child = read_subvol_index;
+ }
+ GF_OPTION_INIT ("choose-local", priv->choose_local, bool, out);
+
+ GF_OPTION_INIT ("read-hash-mode", priv->hash_mode, uint32, out);
priv->favorite_child = -1;
GF_OPTION_INIT ("favorite-child", fav_child, xlator, out);
@@ -263,6 +310,8 @@ init (xlator_t *this)
GF_OPTION_INIT ("self-heal-daemon", priv->shd.enabled, bool, out);
+ GF_OPTION_INIT ("iam-self-heal-daemon", priv->shd.iamshd, bool, out);
+
GF_OPTION_INIT ("data-change-log", priv->data_change_log, bool, out);
GF_OPTION_INIT ("metadata-change-log", priv->metadata_change_log, bool,
@@ -279,10 +328,18 @@ init (xlator_t *this)
GF_OPTION_INIT ("strict-readdir", priv->strict_readdir, bool, out);
+ GF_OPTION_INIT ("eager-lock", priv->eager_lock, bool, out);
GF_OPTION_INIT ("quorum-type", qtype, str, out);
GF_OPTION_INIT ("quorum-count", priv->quorum_count, uint32, out);
+ GF_OPTION_INIT (AFR_SH_READDIR_SIZE_KEY, priv->sh_readdir_size, size,
+ out);
fix_quorum_options(this,priv,qtype);
+ GF_OPTION_INIT ("post-op-delay-secs", priv->post_op_delay_secs, uint32, out);
+ GF_OPTION_INIT ("readdir-failover", priv->readdir_failover, bool, out);
+ GF_OPTION_INIT ("ensure-durability", priv->ensure_durability, bool,
+ out);
+
priv->wait_count = 1;
priv->child_up = GF_CALLOC (sizeof (unsigned char), child_count,
@@ -323,8 +380,6 @@ init (xlator_t *this)
AFR_XATTR_PREFIX,
trav->xlator->name);
if (-1 == ret) {
- gf_log (this->name, GF_LOG_ERROR,
- "asprintf failed to set pending key");
ret = -ENOMEM;
goto out;
}
@@ -333,6 +388,13 @@ init (xlator_t *this)
i++;
}
+ ret = gf_asprintf (&priv->sh_domain, AFR_SH_DATA_DOMAIN_FMT,
+ this->name);
+ if (-1 == ret) {
+ ret = -ENOMEM;
+ goto out;
+ }
+
priv->last_event = GF_CALLOC (child_count, sizeof (*priv->last_event),
gf_afr_mt_int32_t);
if (!priv->last_event) {
@@ -340,16 +402,67 @@ init (xlator_t *this)
goto out;
}
- priv->shd.pos = GF_CALLOC (sizeof (*priv->shd.pos), child_count,
- gf_afr_mt_afr_brick_pos_t);
- if (!priv->shd.pos) {
- ret = -ENOMEM;
+ /* keep more local here as we may need them for self-heal etc */
+ this->local_pool = mem_pool_new (afr_local_t, 512);
+ if (!this->local_pool) {
+ ret = -1;
+ gf_log (this->name, GF_LOG_ERROR,
+ "failed to create local_t's memory pool");
goto out;
}
priv->first_lookup = 1;
priv->root_inode = NULL;
+ if (!priv->shd.iamshd) {
+ ret = 0;
+ goto out;
+ }
+
+ ret = -ENOMEM;
+ priv->shd.pos = GF_CALLOC (sizeof (*priv->shd.pos), child_count,
+ gf_afr_mt_brick_pos_t);
+ if (!priv->shd.pos)
+ goto out;
+
+ priv->shd.pending = GF_CALLOC (sizeof (*priv->shd.pending), child_count,
+ gf_afr_mt_int32_t);
+ if (!priv->shd.pending)
+ goto out;
+
+ priv->shd.inprogress = GF_CALLOC (sizeof (*priv->shd.inprogress),
+ child_count, gf_afr_mt_shd_bool_t);
+ if (!priv->shd.inprogress)
+ goto out;
+ priv->shd.timer = GF_CALLOC (sizeof (*priv->shd.timer), child_count,
+ gf_afr_mt_shd_timer_t);
+ if (!priv->shd.timer)
+ goto out;
+
+ priv->shd.healed = eh_new (AFR_EH_HEALED_LIMIT, _gf_false,
+ _destroy_shd_event_data);
+ if (!priv->shd.healed)
+ goto out;
+
+ priv->shd.heal_failed = eh_new (AFR_EH_HEAL_FAIL_LIMIT, _gf_false,
+ _destroy_shd_event_data);
+ if (!priv->shd.heal_failed)
+ goto out;
+
+ priv->shd.split_brain = eh_new (AFR_EH_SPLIT_BRAIN_LIMIT, _gf_false,
+ _destroy_shd_event_data);
+ if (!priv->shd.split_brain)
+ goto out;
+
+ this->itable = inode_table_new (SHD_INODE_LRU_LIMIT, this);
+ if (!this->itable)
+ goto out;
+ priv->root_inode = inode_ref (this->itable->root);
+ GF_OPTION_INIT ("node-uuid", priv->shd.node_uuid, str, out);
+ GF_OPTION_INIT ("heal-timeout", priv->shd.timeout, int32, out);
+ ret = afr_initialise_statistics (this);
+ if (ret)
+ goto out;
ret = 0;
out:
return ret;
@@ -364,6 +477,8 @@ fini (xlator_t *this)
priv = this->private;
this->private = NULL;
afr_priv_destroy (priv);
+ if (this->itable);//I dont see any destroy func
+
return 0;
}
@@ -382,6 +497,9 @@ struct xlator_fops fops = {
.finodelk = afr_finodelk,
.entrylk = afr_entrylk,
.fentrylk = afr_fentrylk,
+ .fallocate = afr_fallocate,
+ .discard = afr_discard,
+ .zerofill = afr_zerofill,
/* inode read */
.access = afr_access,
@@ -389,6 +507,7 @@ struct xlator_fops fops = {
.fstat = afr_fstat,
.readlink = afr_readlink,
.getxattr = afr_getxattr,
+ .fgetxattr = afr_fgetxattr,
.readv = afr_readv,
/* inode write */
@@ -400,6 +519,7 @@ struct xlator_fops fops = {
.setattr = afr_setattr,
.fsetattr = afr_fsetattr,
.removexattr = afr_removexattr,
+ .fremovexattr = afr_fremovexattr,
/* dir read */
.opendir = afr_opendir,
@@ -432,33 +552,79 @@ struct xlator_cbks cbks = {
struct volume_options options[] = {
{ .key = {"read-subvolume" },
- .type = GF_OPTION_TYPE_XLATOR
+ .type = GF_OPTION_TYPE_XLATOR,
+ .description = "inode-read fops happen only on one of the bricks in "
+ "replicate. Afr will prefer the one specified using "
+ "this option if it is not stale. Option value must be "
+ "one of the xlator names of the children. "
+ "Ex: <volname>-client-0 till "
+ "<volname>-client-<number-of-bricks - 1>"
+ },
+ { .key = {"read-subvolume-index" },
+ .type = GF_OPTION_TYPE_INT,
+ .default_value = "-1",
+ .description = "inode-read fops happen only on one of the bricks in "
+ "replicate. AFR will prefer the one specified using "
+ "this option if it is not stale. allowed options"
+ " include -1 till replica-count - 1"
+ },
+ { .key = {"read-hash-mode" },
+ .type = GF_OPTION_TYPE_INT,
+ .min = 0,
+ .max = 2,
+ .default_value = "0",
+ .description = "inode-read fops happen only on one of the bricks in "
+ "replicate. AFR will prefer the one computed using "
+ "the method specified using this option"
+ "0 = first responder, "
+ "1 = hash by GFID of file (all clients use "
+ "same subvolume), "
+ "2 = hash by GFID of file and client PID",
+ },
+ { .key = {"choose-local" },
+ .type = GF_OPTION_TYPE_BOOL,
+ .default_value = "true",
+ .description = "Choose a local subvolume(i.e. Brick) to read from if "
+ "read-subvolume is not explicitly set.",
},
{ .key = {"favorite-child"},
- .type = GF_OPTION_TYPE_XLATOR
+ .type = GF_OPTION_TYPE_XLATOR,
+ .description = "If a split-brain happens choose subvol/brick set by "
+ "this option as source."
},
{ .key = {"background-self-heal-count"},
.type = GF_OPTION_TYPE_INT,
.min = 0,
.default_value = "16",
+ .validate = GF_OPT_VALIDATE_MIN,
+ .description = "This specifies the number of self-heals that can be "
+ " performed in background without blocking the fop"
},
{ .key = {"data-self-heal"},
.type = GF_OPTION_TYPE_STR,
- .default_value = "",
.value = {"1", "on", "yes", "true", "enable",
"0", "off", "no", "false", "disable",
"open"},
.default_value = "on",
+ .description = "Using this option we can enable/disable data "
+ "self-heal on the file. \"open\" means data "
+ "self-heal action will only be triggered by file "
+ "open operations."
},
{ .key = {"data-self-heal-algorithm"},
.type = GF_OPTION_TYPE_STR,
- .default_value = "",
.description = "Select between \"full\", \"diff\". The "
"\"full\" algorithm copies the entire file from "
"source to sink. The \"diff\" algorithm copies to "
"sink only those blocks whose checksums don't match "
- "with those of source.",
- .value = { "diff", "full", "" }
+ "with those of source. If no option is configured "
+ "the option is chosen dynamically as follows: "
+ "If the file does not exist on one of the sinks "
+ "or empty file exists or if the source file size is "
+ "about the same as page size the entire file will "
+ "be read and written i.e \"full\" algo, "
+ "otherwise \"diff\" algo is chosen.",
+ .value = { "diff", "full"}
},
{ .key = {"data-self-heal-window-size"},
.type = GF_OPTION_TYPE_INT,
@@ -471,26 +637,43 @@ struct volume_options options[] = {
{ .key = {"metadata-self-heal"},
.type = GF_OPTION_TYPE_BOOL,
.default_value = "on",
+ .description = "Using this option we can enable/disable metadata "
+ "i.e. Permissions, ownerships, xattrs self-heal on "
+ "the file/directory."
},
{ .key = {"entry-self-heal"},
.type = GF_OPTION_TYPE_BOOL,
.default_value = "on",
+ .description = "Using this option we can enable/disable entry "
+ "self-heal on the directory."
},
{ .key = {"data-change-log"},
.type = GF_OPTION_TYPE_BOOL,
.default_value = "on",
+ .description = "Data fops like write/truncate will not perform "
+ "pre/post fop changelog operations in afr transaction "
+ "if this option is disabled"
},
{ .key = {"metadata-change-log"},
.type = GF_OPTION_TYPE_BOOL,
.default_value = "on",
+ .description = "Metadata fops like setattr/setxattr will not perform "
+ "pre/post fop changelog operations in afr transaction "
+ "if this option is disabled"
},
{ .key = {"entry-change-log"},
.type = GF_OPTION_TYPE_BOOL,
.default_value = "on",
+ .description = "Entry fops like create/unlink will not perform "
+ "pre/post fop changelog operations in afr transaction "
+ "if this option is disabled"
},
{ .key = {"optimistic-change-log"},
.type = GF_OPTION_TYPE_BOOL,
.default_value = "on",
+ .description = "Entry/Metadata fops will not perform "
+ "pre fop changelog operations in afr transaction "
+ "if this option is enabled."
},
{ .key = {"strict-readdir"},
.type = GF_OPTION_TYPE_BOOL,
@@ -499,18 +682,55 @@ struct volume_options options[] = {
{ .key = {"inodelk-trace"},
.type = GF_OPTION_TYPE_BOOL,
.default_value = "off",
+ .description = "Enabling this option logs inode lock/unlocks"
},
{ .key = {"entrylk-trace"},
.type = GF_OPTION_TYPE_BOOL,
.default_value = "off",
+ .description = "Enabling this option logs entry lock/unlocks"
+ },
+ { .key = {"eager-lock"},
+ .type = GF_OPTION_TYPE_BOOL,
+ .default_value = "on",
+ .description = "Lock phase of a transaction has two sub-phases. "
+ "First is an attempt to acquire locks in parallel by "
+ "broadcasting non-blocking lock requests. If lock "
+ "aquistion fails on any server, then the held locks "
+ "are unlocked and revert to a blocking locked mode "
+ "sequentially on one server after another. If this "
+ "option is enabled the initial broadcasting lock "
+ "request attempt to acquire lock on the entire file. "
+ "If this fails, we revert back to the sequential "
+ "\"regional\" blocking lock as before. In the case "
+ "where such an \"eager\" lock is granted in the "
+ "non-blocking phase, it gives rise to an opportunity "
+ "for optimization. i.e, if the next write transaction "
+ "on the same FD arrives before the unlock phase of "
+ "the first transaction, it \"takes over\" the full "
+ "file lock. Similarly if yet another data transaction "
+ "arrives before the unlock phase of the \"optimized\" "
+ "transaction, that in turn \"takes over\" the lock as "
+ "well. The actual unlock now happens at the end of "
+ "the last \"optimzed\" transaction."
+
},
{ .key = {"self-heal-daemon"},
.type = GF_OPTION_TYPE_BOOL,
.default_value = "off",
+ .description = "This option applies to only self-heal-daemon. "
+ "Index directory crawl and automatic healing of files"
+ "will not be performed if this option is turned off."
+ },
+ { .key = {"iam-self-heal-daemon"},
+ .type = GF_OPTION_TYPE_BOOL,
+ .default_value = "off",
+ .description = "This option differentiates if the replicate "
+ "translator is running as part of self-heal-daemon "
+ "or not."
},
{ .key = {"quorum-type"},
.type = GF_OPTION_TYPE_STR,
- .value = { "none", "auto", "fixed", "" },
+ .value = { "none", "auto", "fixed"},
.default_value = "none",
.description = "If value is \"fixed\" only allow writes if "
"quorum-count bricks are present. If value is "
@@ -527,5 +747,47 @@ struct volume_options options[] = {
"this many bricks or present. Other quorum types "
"will OVERWRITE this value.",
},
+ { .key = {"node-uuid"},
+ .type = GF_OPTION_TYPE_STR,
+ .description = "Local glusterd uuid string, used in starting "
+ "self-heal-daemon so that it can crawl only on "
+ "local index directories.",
+ },
+ { .key = {"heal-timeout"},
+ .type = GF_OPTION_TYPE_INT,
+ .min = 60,
+ .max = INT_MAX,
+ .default_value = "600",
+ .description = "time interval for checking the need to self-heal "
+ "in self-heal-daemon"
+ },
+ { .key = {"post-op-delay-secs"},
+ .type = GF_OPTION_TYPE_INT,
+ .min = 0,
+ .max = INT_MAX,
+ .default_value = "1",
+ .description = "Time interval induced artificially before "
+ "post-operation phase of the transaction to "
+ "enhance overlap of adjacent write operations.",
+ },
+ { .key = {AFR_SH_READDIR_SIZE_KEY},
+ .type = GF_OPTION_TYPE_SIZET,
+ .description = "readdirp size for performing entry self-heal",
+ .min = 1024,
+ .max = 131072,
+ .default_value = "1KB",
+ },
+ { .key = {"readdir-failover"},
+ .type = GF_OPTION_TYPE_BOOL,
+ .description = "readdir(p) will not failover if this option is off",
+ .default_value = "on",
+ },
+ { .key = {"ensure-durability"},
+ .type = GF_OPTION_TYPE_BOOL,
+ .description = "Afr performs fsyncs for transactions if this "
+ "option is on to make sure the changelogs/data is "
+ "written to the disk",
+ .default_value = "on",
+ },
{ .key = {NULL} },
};
diff --git a/xlators/cluster/afr/src/afr.h b/xlators/cluster/afr/src/afr.h
index 889828a6f..21064db58 100644
--- a/xlators/cluster/afr/src/afr.h
+++ b/xlators/cluster/afr/src/afr.h
@@ -1,20 +1,11 @@
/*
- Copyright (c) 2008-2011 Gluster, Inc. <http://www.gluster.com>
+ Copyright (c) 2008-2012 Red Hat, Inc. <http://www.redhat.com>
This file is part of GlusterFS.
- GlusterFS is free software; you can redistribute it and/or modify
- it under the terms of the GNU General Public License as published
- by the Free Software Foundation; either version 3 of the License,
- or (at your option) any later version.
-
- GlusterFS is distributed in the hope that it will be useful, but
- WITHOUT ANY WARRANTY; without even the implied warranty of
- MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU
- General Public License for more details.
-
- You should have received a copy of the GNU General Public License
- along with this program. If not, see
- <http://www.gnu.org/licenses/>.
+ This file is licensed to you under your choice of the GNU Lesser
+ General Public License, version 3 or any later version (LGPLv3 or
+ later), or the GNU General Public License, version 2 (GPLv2), in all
+ cases as published by the Free Software Foundation.
*/
@@ -32,9 +23,15 @@
#include "afr-self-heal-algorithm.h"
#include "libxlator.h"
+#include "timer.h"
#define AFR_XATTR_PREFIX "trusted.afr"
#define AFR_PATHINFO_HEADER "REPLICATE:"
+#define AFR_SH_READDIR_SIZE_KEY "self-heal-readdir-size"
+#define AFR_SH_DATA_DOMAIN_FMT "%s:self-heal"
+
+#define AFR_LOCKEE_COUNT_MAX 3
+#define AFR_DOM_COUNT_MAX 3
struct _pump_private;
@@ -43,8 +40,7 @@ typedef int (*afr_expunge_done_cbk_t) (call_frame_t *frame, xlator_t *this,
int32_t op_errno);
typedef int (*afr_impunge_done_cbk_t) (call_frame_t *frame, xlator_t *this,
- int child, int32_t op_error,
- int32_t op_errno);
+ int32_t op_error, int32_t op_errno);
typedef int (*afr_post_remove_call_t) (call_frame_t *frame, xlator_t *this);
typedef int (*afr_lock_cbk_t) (call_frame_t *frame, xlator_t *this);
@@ -66,10 +62,8 @@ typedef enum {
AFR_INODE_SET_READ_CTX = 1,
AFR_INODE_RM_STALE_CHILDREN,
AFR_INODE_SET_OPENDIR_DONE,
- AFR_INODE_SET_SPLIT_BRAIN,
AFR_INODE_GET_READ_CTX,
AFR_INODE_GET_OPENDIR_DONE,
- AFR_INODE_GET_SPLIT_BRAIN,
} afr_inode_op_t;
typedef struct afr_inode_params_ {
@@ -83,16 +77,41 @@ typedef struct afr_inode_params_ {
} u;
} afr_inode_params_t;
+typedef enum afr_spb_state {
+ DONT_KNOW,
+ SPB,
+ NO_SPB
+} afr_spb_state_t;
+
typedef struct afr_inode_ctx_ {
uint64_t masks;
int32_t *fresh_children;//increasing order of latency
+ afr_spb_state_t mdata_spb;
+ afr_spb_state_t data_spb;
+ uint32_t open_fd_count;
} afr_inode_ctx_t;
+typedef enum {
+ NONE,
+ INDEX,
+ INDEX_TO_BE_HEALED,
+ FULL,
+} afr_crawl_type_t;
+
typedef struct afr_self_heald_ {
- gf_boolean_t enabled;
- gf_boolean_t pending;
- gf_boolean_t inprogress;
- afr_child_pos_t *pos;
+ gf_boolean_t enabled;
+ gf_boolean_t iamshd;
+ afr_crawl_type_t *pending;
+ gf_boolean_t *inprogress;
+ afr_child_pos_t *pos;
+ gf_timer_t **timer;
+ eh_t *healed;
+ eh_t *heal_failed;
+ eh_t *split_brain;
+ eh_t **statistics;
+ void **crawl_events;
+ char *node_uuid;
+ int timeout;
} afr_self_heald_t;
typedef struct _afr_private {
@@ -126,6 +145,7 @@ typedef struct _afr_private {
gf_boolean_t entry_change_log; /* on/off */
int read_child; /* read-subvolume */
+ unsigned int hash_mode; /* for when read_child is not set */
int favorite_child; /* subvolume to be preferred in resolving
split-brain cases */
@@ -146,14 +166,48 @@ typedef struct _afr_private {
struct list_head saved_fds; /* list of fds on which locks have succeeded */
gf_boolean_t optimistic_change_log;
gf_boolean_t eager_lock;
+ uint32_t post_op_delay_secs;
unsigned int quorum_count;
char vol_uuid[UUID_SIZE + 1];
int32_t *last_event;
afr_self_heald_t shd;
+ gf_boolean_t choose_local;
+ gf_boolean_t did_discovery;
+ gf_boolean_t readdir_failover;
+ uint64_t sh_readdir_size;
+ gf_boolean_t ensure_durability;
+ char *sh_domain;
} afr_private_t;
+typedef enum {
+ AFR_SELF_HEAL_NOT_ATTEMPTED,
+ AFR_SELF_HEAL_STARTED,
+ AFR_SELF_HEAL_FAILED,
+ AFR_SELF_HEAL_SYNC_BEGIN,
+} afr_self_heal_status;
+
typedef struct {
+ afr_self_heal_status gfid_or_missing_entry_self_heal;
+ afr_self_heal_status metadata_self_heal;
+ afr_self_heal_status data_self_heal;
+ afr_self_heal_status entry_self_heal;
+} afr_sh_status_for_all_type;
+
+typedef enum {
+ AFR_SELF_HEAL_ENTRY,
+ AFR_SELF_HEAL_METADATA,
+ AFR_SELF_HEAL_DATA,
+ AFR_SELF_HEAL_GFID_OR_MISSING_ENTRY,
+ AFR_SELF_HEAL_INVALID = -1,
+} afr_self_heal_type;
+
+typedef enum {
+ AFR_CHECK_ALL,
+ AFR_CHECK_SPECIFIC,
+} afr_sh_fail_check_type;
+
+struct afr_self_heal_ {
/* External interface: These are variables (some optional) that
are set by whoever has triggered self-heal */
@@ -162,6 +216,8 @@ typedef struct {
gf_boolean_t do_entry_self_heal;
gf_boolean_t do_gfid_self_heal;
gf_boolean_t do_missing_entry_self_heal;
+ gf_boolean_t force_confirm_spb; /* Check for split-brains even when
+ self-heal is turned off */
gf_boolean_t forced_merge; /* Is this a self-heal triggered to
forcibly merge the directories? */
@@ -179,7 +235,7 @@ typedef struct {
background, this function will be called as soon as possible. */
int (*unwind) (call_frame_t *frame, xlator_t *this, int32_t op_ret,
- int32_t op_errno);
+ int32_t op_errno, int32_t sh_failed);
/* End of external interface members */
@@ -192,7 +248,6 @@ typedef struct {
afr_expunge_done_cbk_t expunge_done;
afr_impunge_done_cbk_t impunge_done;
- int32_t impunge_ret_child;
/* array of xattr's, one for each child */
dict_t **xattr;
@@ -226,13 +281,13 @@ typedef struct {
unsigned char *locked_nodes;
int lock_count;
- mode_t impunging_entry_mode;
const char *linkname;
gf_boolean_t entries_skipped;
- int op_failed;
-
+ gf_boolean_t actual_sh_started;
+ gf_boolean_t sync_done;
gf_boolean_t data_lock_held;
+ gf_boolean_t sh_dom_lock_held;
gf_boolean_t eof_reached;
fd_t *healing_fd;
int file_has_holes;
@@ -243,16 +298,22 @@ typedef struct {
uint8_t *checksum;
afr_post_remove_call_t post_remove_call;
- loc_t parent_loc;
+ char *data_sh_info;
+ char *metadata_sh_info;
+ loc_t parent_loc;
call_frame_t *orig_frame;
call_frame_t *old_loop_frame;
gf_boolean_t unwound;
afr_sh_algo_private_t *private;
+ afr_sh_status_for_all_type afr_all_sh_status;
+ afr_self_heal_type sh_type_in_action;
+ struct afr_sh_algorithm *algo;
afr_lock_cbk_t data_lock_success_handler;
afr_lock_cbk_t data_lock_failure_handler;
+ gf_boolean_t data_lock_block;
int (*completion_cbk) (call_frame_t *frame, xlator_t *this);
int (*sh_data_algo_start) (call_frame_t *frame, xlator_t *this);
int (*algo_completion_cbk) (call_frame_t *frame, xlator_t *this);
@@ -260,7 +321,9 @@ typedef struct {
void (*gfid_sh_success_cbk) (call_frame_t *sh_frame, xlator_t *this);
call_frame_t *sh_frame;
-} afr_self_heal_t;
+};
+
+typedef struct afr_self_heal_ afr_self_heal_t;
typedef enum {
AFR_DATA_TRANSACTION, /* truncate, write, ... */
@@ -322,11 +385,31 @@ afr_index_for_transaction_type (afr_transaction_type type)
return -1; /* make gcc happy */
}
+typedef struct {
+ loc_t loc;
+ char *basename;
+ unsigned char *locked_nodes;
+ int locked_count;
+
+} afr_entry_lockee_t;
+
+int
+afr_entry_lockee_cmp (const void *l1, const void *l2);
+
+typedef struct {
+ char *domain; /* Domain on which inodelk is taken */
+ struct gf_flock flock;
+ unsigned char *locked_nodes;
+ int32_t lock_count;
+} afr_inodelk_t;
typedef struct {
loc_t *lk_loc;
- struct gf_flock lk_flock;
+ int lockee_count;
+ afr_entry_lockee_t lockee[AFR_LOCKEE_COUNT_MAX];
+
+ afr_inodelk_t inodelk[AFR_DOM_COUNT_MAX];
const char *lk_basename;
const char *lower_basename;
const char *higher_basename;
@@ -335,23 +418,22 @@ typedef struct {
unsigned char *locked_nodes;
unsigned char *lower_locked_nodes;
- unsigned char *inode_locked_nodes;
- unsigned char *entry_locked_nodes;
selfheal_lk_type_t selfheal_lk_type;
transaction_lk_type_t transaction_lk_type;
int32_t lock_count;
- int32_t inodelk_lock_count;
int32_t entrylk_lock_count;
uint64_t lock_number;
int32_t lk_call_count;
int32_t lk_expected_count;
+ int32_t lk_attempted_count;
int32_t lock_op_ret;
int32_t lock_op_errno;
afr_lock_cbk_t lock_cbk;
+ char *domain; /* Domain on which inode/entry lock/unlock in progress.*/
} afr_internal_lock_t;
typedef struct _afr_locked_fd {
@@ -359,21 +441,29 @@ typedef struct _afr_locked_fd {
struct list_head list;
} afr_locked_fd_t;
+struct afr_reply {
+ int valid;
+ int32_t op_ret;
+ int32_t op_errno;
+};
+
typedef struct _afr_local {
int uid;
int gid;
unsigned int call_count;
unsigned int success_count;
unsigned int enoent_count;
+ uint32_t open_fd_count;
+ gf_boolean_t update_open_fd_count;
- unsigned int govinda_gOvinda;
+ unsigned int unhealable;
unsigned int read_child_index;
unsigned char read_child_returned;
unsigned int first_up_child;
- pid_t saved_pid;
+ gf_lkowner_t saved_lk_owner;
int32_t op_ret;
int32_t op_errno;
@@ -384,7 +474,6 @@ typedef struct _afr_local {
loc_t newloc;
fd_t *fd;
- unsigned char *fd_open_on;
glusterfs_fop_t fop;
@@ -406,13 +495,25 @@ typedef struct _afr_local {
dict_t *dict;
int optimistic_change_log;
+ gf_boolean_t delayed_post_op;
+
- gf_boolean_t fop_paused;
- int (*fop_call_continue) (call_frame_t *frame, xlator_t *this);
+ /* Is the current writev() going to perform a stable write?
+ i.e, is fd->flags or @flags writev param have O_SYNC or
+ O_DSYNC?
+ */
+ gf_boolean_t stable_write;
- /*
- This struct contains the arguments for the "continuation"
- (scheme-like) of fops
+ /* This write appended to the file. Nnot necessarily O_APPEND,
+ just means the offset of write was at the end of file.
+ */
+ gf_boolean_t append_write;
+
+ int allow_sh_for_running_transaction;
+
+
+ /* This struct contains the arguments for the "continuation"
+ (scheme-like) of fops
*/
int op;
@@ -423,6 +524,7 @@ typedef struct _afr_local {
} statfs;
struct {
+ uint32_t parent_entrylk;
uuid_t gfid_req;
inode_t *inode;
struct iatt buf;
@@ -434,11 +536,13 @@ typedef struct _afr_local {
int32_t read_child;
int32_t *sources;
int32_t *success_children;
+ int32_t **pending_matrix;
+ gf_boolean_t fresh_lookup;
+ gf_boolean_t possible_spb;
} lookup;
struct {
int32_t flags;
- int32_t wbflags;
} open;
struct {
@@ -471,13 +575,14 @@ typedef struct _afr_local {
struct {
char *name;
int last_index;
- long pathinfo_len;
+ long xattr_len;
} getxattr;
struct {
size_t size;
off_t offset;
int last_index;
+ uint32_t flags;
} readv;
/* dir read */
@@ -504,44 +609,34 @@ typedef struct _afr_local {
struct {
struct iatt prebuf;
struct iatt postbuf;
+ } inode_wfop; //common structure for all inode-write-fops
+ struct {
int32_t op_ret;
struct iovec *vector;
struct iobref *iobref;
int32_t count;
off_t offset;
+ uint32_t flags;
} writev;
struct {
- struct iatt prebuf;
- struct iatt postbuf;
- } fsync;
-
- struct {
off_t offset;
- struct iatt prebuf;
- struct iatt postbuf;
} truncate;
struct {
off_t offset;
- struct iatt prebuf;
- struct iatt postbuf;
} ftruncate;
struct {
struct iatt in_buf;
int32_t valid;
- struct iatt preop_buf;
- struct iatt postop_buf;
} setattr;
struct {
struct iatt in_buf;
int32_t valid;
- struct iatt preop_buf;
- struct iatt postop_buf;
} fsetattr;
struct {
@@ -558,93 +653,77 @@ typedef struct _afr_local {
char *name;
} removexattr;
+ struct {
+ dict_t *xattr;
+ } xattrop;
+
+ struct {
+ dict_t *xattr;
+ } fxattrop;
+
/* dir write */
struct {
- fd_t *fd;
- dict_t *params;
- int32_t flags;
- mode_t mode;
inode_t *inode;
struct iatt buf;
struct iatt preparent;
struct iatt postparent;
- struct iatt read_child_buf;
+ struct iatt prenewparent;
+ struct iatt postnewparent;
+ } dir_fop; //common structure for all dir fops
+
+ struct {
+ fd_t *fd;
+ dict_t *params;
+ int32_t flags;
+ mode_t mode;
} create;
struct {
dev_t dev;
mode_t mode;
dict_t *params;
- inode_t *inode;
- struct iatt buf;
- struct iatt preparent;
- struct iatt postparent;
- struct iatt read_child_buf;
} mknod;
struct {
int32_t mode;
dict_t *params;
- inode_t *inode;
- struct iatt buf;
- struct iatt read_child_buf;
- struct iatt preparent;
- struct iatt postparent;
} mkdir;
struct {
- int32_t op_ret;
- int32_t op_errno;
- struct iatt preparent;
- struct iatt postparent;
- } unlink;
-
- struct {
- int flags;
- int32_t op_ret;
- int32_t op_errno;
- struct iatt preparent;
- struct iatt postparent;
+ int flags;
} rmdir;
struct {
- struct iatt buf;
- struct iatt read_child_buf;
- struct iatt preoldparent;
- struct iatt prenewparent;
- struct iatt postoldparent;
- struct iatt postnewparent;
- } rename;
-
- struct {
- inode_t *inode;
- struct iatt buf;
- struct iatt read_child_buf;
- struct iatt preparent;
- struct iatt postparent;
- } link;
-
- struct {
- inode_t *inode;
dict_t *params;
- struct iatt buf;
- struct iatt read_child_buf;
char *linkpath;
- struct iatt preparent;
- struct iatt postparent;
} symlink;
+ struct {
+ int32_t mode;
+ off_t offset;
+ size_t len;
+ } fallocate;
+
+ struct {
+ off_t offset;
+ size_t len;
+ } discard;
+
struct {
- int32_t flags;
- dir_entry_t *entries;
- int32_t count;
- } setdents;
+ off_t offset;
+ size_t len;
+ struct iatt prebuf;
+ struct iatt postbuf;
+ } zerofill;
+
+
} cont;
struct {
off_t start, len;
+ gf_boolean_t eager_lock_on;
int *eager_lock;
char *basename;
@@ -655,12 +734,18 @@ typedef struct _afr_local {
afr_transaction_type type;
- int success_count;
- int erase_pending;
- int failure_count;
+ /* pre-compute the post piggyback status before
+ entering POST-OP phase
+ */
+ int *postop_piggybacked;
+
+ /* stub to resume on destruction
+ of the transaction frame */
+ call_stub_t *resume_stub;
+
+ struct list_head eager_locked;
- int last_tried;
- int32_t *child_errno;
+ int32_t **txn_changelog;//changelog after pre+post ops
unsigned char *pre_op;
call_frame_t *main_frame;
@@ -679,6 +764,15 @@ typedef struct _afr_local {
afr_self_heal_t self_heal;
struct marker_str marker;
+
+ /* extra data for fops */
+ dict_t *xdata_req;
+ dict_t *xdata_rsp;
+
+ mode_t umask;
+ int xflag;
+ gf_boolean_t do_discovery;
+ struct afr_reply *replies;
} afr_local_t;
typedef enum {
@@ -688,11 +782,6 @@ typedef enum {
} afr_fd_open_status_t;
typedef struct {
- struct list_head call_list;
- call_frame_t *frame;
-} afr_fd_paused_call_t;
-
-typedef struct {
unsigned int *pre_op_done;
afr_fd_open_status_t *opened_on; /* which subvolumes the fd is open on */
unsigned int *pre_op_piggyback;
@@ -701,7 +790,6 @@ typedef struct {
unsigned int *lock_acquired;
int flags;
- int32_t wbflags;
uint64_t up_count; /* number of CHILD_UPs this fd has seen */
uint64_t down_count; /* number of CHILD_DOWNs this fd has seen */
@@ -712,20 +800,32 @@ typedef struct {
struct list_head entries; /* needed for readdir failover */
unsigned char *locked_on; /* which subvolumes locks have been successful */
- struct list_head paused_calls; /* queued calls while fix_open happens */
+
+ /* used for delayed-post-op optimization */
+ pthread_mutex_t delay_lock;
+ gf_timer_t *delay_timer;
+ call_frame_t *delay_frame;
+ int call_child;
+
+ /* set if any write on this fd was a non stable write
+ (i.e, without O_SYNC or O_DSYNC)
+ */
+ gf_boolean_t witnessed_unstable_write;
+
+ /* list of frames currently in progress */
+ struct list_head eager_locked;
} afr_fd_ctx_t;
/* try alloc and if it fails, goto label */
-#define ALLOC_OR_GOTO(var, type, label) do { \
- var = GF_CALLOC (sizeof (type), 1, \
- gf_afr_mt_##type); \
- if (!var) { \
- gf_log (this->name, GF_LOG_ERROR, \
- "out of memory :("); \
- op_errno = ENOMEM; \
- goto label; \
- } \
+#define AFR_LOCAL_ALLOC_OR_GOTO(var, label) do { \
+ var = mem_get0 (THIS->local_pool); \
+ if (!var) { \
+ gf_log (this->name, GF_LOG_ERROR, \
+ "out of memory :("); \
+ op_errno = ENOMEM; \
+ goto label; \
+ } \
} while (0);
@@ -746,8 +846,14 @@ int
pump_command_reply (call_frame_t *frame, xlator_t *this);
int32_t
-afr_notify (xlator_t *this, int32_t event,
- void *data, ...);
+afr_notify (xlator_t *this, int32_t event, void *data, void *data2);
+
+int
+afr_init_entry_lockee (afr_entry_lockee_t *lockee, afr_local_t *local,
+ loc_t *loc, char *basename, int child_count);
+
+void
+afr_entry_lockee_cleanup (afr_internal_lock_t *int_lock);
int
afr_attempt_lock_recovery (xlator_t *this, int32_t child_index);
@@ -760,7 +866,7 @@ afr_mark_locked_nodes (xlator_t *this, fd_t *fd,
unsigned char *locked_nodes);
void
-afr_set_lk_owner (call_frame_t *frame, xlator_t *this);
+afr_set_lk_owner (call_frame_t *frame, xlator_t *this, void *lk_owner);
int
afr_set_lock_number (call_frame_t *frame, xlator_t *this);
@@ -784,8 +890,8 @@ afr_blocking_lock (call_frame_t *frame, xlator_t *this);
int
afr_internal_lock_finish (call_frame_t *frame, xlator_t *this);
-void
-afr_lk_transfer_datalock (call_frame_t *dst, call_frame_t *src,
+int
+afr_lk_transfer_datalock (call_frame_t *dst, call_frame_t *src, char *dom,
unsigned int child_count);
int pump_start (call_frame_t *frame, xlator_t *this);
@@ -803,8 +909,8 @@ void
afr_inode_set_read_ctx (xlator_t *this, inode_t *inode, int32_t read_child,
int32_t *fresh_children);
-void
-afr_build_parent_loc (loc_t *parent, loc_t *child);
+int
+afr_build_parent_loc (loc_t *parent, loc_t *child, int32_t *op_errno);
unsigned int
afr_up_children_count (unsigned char *child_up, unsigned int child_count);
@@ -835,11 +941,12 @@ gf_boolean_t
afr_is_split_brain (xlator_t *this, inode_t *inode);
void
-afr_set_split_brain (xlator_t *this, inode_t *inode, gf_boolean_t set);
+afr_set_split_brain (xlator_t *this, inode_t *inode, afr_spb_state_t mdata_spb,
+ afr_spb_state_t data_spb);
int
afr_open (call_frame_t *frame, xlator_t *this, loc_t *loc, int32_t flags,
- fd_t *fd, int32_t wbflags);
+ fd_t *fd, dict_t *xdata);
void
afr_set_opendir_done (xlator_t *this, inode_t *inode);
@@ -866,22 +973,27 @@ afr_launch_openfd_self_heal (call_frame_t *frame, xlator_t *this, fd_t *fd);
frame->local = NULL; \
} \
STACK_UNWIND_STRICT (fop, frame, params); \
- afr_local_cleanup (__local, __this); \
- GF_FREE (__local); \
- } while (0);
+ if (__local) { \
+ afr_local_cleanup (__local, __this); \
+ mem_put (__local); \
+ } \
+ } while (0)
-#define AFR_STACK_DESTROY(frame) \
- do { \
- afr_local_t *__local = NULL; \
- xlator_t *__this = NULL; \
- __local = frame->local; \
- __this = frame->this; \
- frame->local = NULL; \
- STACK_DESTROY (frame->root); \
- afr_local_cleanup (__local, __this); \
- GF_FREE (__local); \
+#define AFR_STACK_DESTROY(frame) \
+ do { \
+ afr_local_t *__local = NULL; \
+ xlator_t *__this = NULL; \
+ __local = frame->local; \
+ __this = frame->this; \
+ frame->local = NULL; \
+ STACK_DESTROY (frame->root); \
+ if (__local) { \
+ afr_local_cleanup (__local, __this); \
+ mem_put (__local); \
+ } \
} while (0);
+#define AFR_NUM_CHANGE_LOGS 3 /*data + metadata + entry*/
/* allocate and return a string that is the basename of argument */
static inline char *
AFR_BASENAME (const char *str)
@@ -917,12 +1029,13 @@ afr_first_up_child (unsigned char *child_up, size_t child_count);
int
afr_select_read_child_from_policy (int32_t *fresh_children, int32_t child_count,
int32_t prev_read_child,
- int32_t config_read_child, int32_t *sources);
+ int32_t config_read_child, int32_t *sources,
+ unsigned int hmode, uuid_t gfid);
void
afr_set_read_ctx_from_policy (xlator_t *this, inode_t *inode,
int32_t *fresh_children, int32_t prev_read_child,
- int32_t config_read_child);
+ int32_t config_read_child, uuid_t gfid);
int32_t
afr_get_call_child (xlator_t *this, unsigned char *child_up, int32_t read_child,
@@ -944,8 +1057,9 @@ afr_children_rm_child (int32_t *children, int32_t child,
int32_t child_count);
void
afr_reset_children (int32_t *children, int32_t child_count);
-gf_boolean_t
-afr_error_more_important (int32_t old_errno, int32_t new_errno);
+int32_t
+afr_most_important_error(int32_t old_errno, int32_t new_errno,
+ gf_boolean_t eio);
int
afr_errno_count (int32_t *children, int *child_errno,
unsigned int child_count, int32_t op_errno);
@@ -986,12 +1100,13 @@ afr_launch_self_heal (call_frame_t *frame, xlator_t *this, inode_t *inode,
void (*gfid_sh_success_cbk) (call_frame_t *sh_frame,
xlator_t *this),
int (*unwind) (call_frame_t *frame, xlator_t *this,
- int32_t op_ret, int32_t op_errno));
-int
-afr_fix_open (call_frame_t *frame, xlator_t *this, afr_fd_ctx_t *fd_ctx,
- int need_open_count, int *need_open);
-int
-afr_open_fd_fix (call_frame_t *frame, xlator_t *this, gf_boolean_t pause_fop);
+ int32_t op_ret, int32_t op_errno,
+ int32_t sh_failed));
+void
+afr_fix_open (xlator_t *this, fd_t *fd, size_t need_open_count, int *need_open);
+
+void
+afr_open_fd_fix (fd_t *fd, xlator_t *this);
int
afr_set_elem_count_get (unsigned char *elems, int child_count);
@@ -1008,11 +1123,33 @@ void
afr_set_low_priority (call_frame_t *frame);
int
afr_child_fd_ctx_set (xlator_t *this, fd_t *fd, int32_t child,
- int flags, int32_t wb_flags);
+ int flags);
gf_boolean_t
afr_have_quorum (char *logname, afr_private_t *priv);
+void
+afr_matrix_cleanup (int32_t **pending, unsigned int m);
+
+int32_t**
+afr_matrix_create (unsigned int m, unsigned int n);
+
+gf_boolean_t
+afr_is_errno_set (int *child_errno, int child);
+
+gf_boolean_t
+afr_is_errno_unset (int *child_errno, int child);
+
+gf_boolean_t
+afr_is_fd_fixable (fd_t *fd);
+
+void
+afr_prepare_new_entry_pending_matrix (int32_t **pending,
+ gf_boolean_t (*is_pending) (int *, int),
+ int *ctx, struct iatt *buf,
+ unsigned int child_count);
+void
+afr_xattr_array_destroy (dict_t **xattr, unsigned int child_count);
/*
* Special value indicating we should use the "auto" quorum method instead of
* a fixed value (including zero to turn off quorum enforcement).
@@ -1032,4 +1169,44 @@ afr_have_quorum (char *logname, afr_private_t *priv);
} \
} while (0);
+
+#define AFR_SBRAIN_MSG "Failed on %s as split-brain is seen. Returning EIO."
+
+#define AFR_SBRAIN_CHECK_FD(fd, label) do { \
+ if (fd->inode && afr_is_split_brain (this, fd->inode)) { \
+ op_errno = EIO; \
+ gf_log (this->name, GF_LOG_WARNING, \
+ AFR_SBRAIN_MSG ,uuid_utoa (fd->inode->gfid)); \
+ goto label; \
+ } \
+} while (0)
+
+#define AFR_SBRAIN_CHECK_LOC(loc, label) do { \
+ if (loc->inode && afr_is_split_brain (this, loc->inode)) { \
+ op_errno = EIO; \
+ loc_path (loc, NULL); \
+ gf_log (this->name, GF_LOG_WARNING, \
+ AFR_SBRAIN_MSG , loc->path); \
+ goto label; \
+ } \
+} while (0)
+
+int
+afr_fd_report_unstable_write (xlator_t *this, fd_t *fd);
+
+gf_boolean_t
+afr_fd_has_witnessed_unstable_write (xlator_t *this, fd_t *fd);
+
+void
+afr_delayed_changelog_wake_resume (xlator_t *this, fd_t *fd, call_stub_t *stub);
+
+int
+afr_inodelk_init (afr_inodelk_t *lk, char *dom, size_t child_count);
+
+void
+afr_handle_open_fd_count (call_frame_t *frame, xlator_t *this);
+
+afr_inode_ctx_t*
+afr_inode_ctx_get (inode_t *inode, xlator_t *this);
+
#endif /* __AFR_H__ */
diff --git a/xlators/cluster/afr/src/pump.c b/xlators/cluster/afr/src/pump.c
index 357aa9f29..a7f72fb30 100644
--- a/xlators/cluster/afr/src/pump.c
+++ b/xlators/cluster/afr/src/pump.c
@@ -1,20 +1,11 @@
/*
- Copyright (c) 2007-2011 Gluster, Inc. <http://www.gluster.com>
- This file is part of GlusterFS.
-
- GlusterFS is free software; you can redistribute it and/or modify
- it under the terms of the GNU General Public License as published
- by the Free Software Foundation; either version 3 of the License,
- or (at your option) any later version.
-
- GlusterFS is distributed in the hope that it will be useful, but
- WITHOUT ANY WARRANTY; without even the implied warranty of
- MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU
- General Public License for more details.
-
- You should have received a copy of the GNU General Public License
- along with this program. If not, see
- <http://www.gnu.org/licenses/>.
+ Copyright (c) 2008-2012 Red Hat, Inc. <http://www.redhat.com>
+ This file is part of GlusterFS.
+
+ This file is licensed to you under your choice of the GNU Lesser
+ General Public License, version 3 or any later version (LGPLv3 or
+ later), or the GNU General Public License, version 2 (GPLv2), in all
+ cases as published by the Free Software Foundation.
*/
#include <unistd.h>
@@ -32,6 +23,13 @@
#include "glusterfs.h"
static uint64_t pump_pid = 0;
+static inline void
+pump_fill_loc_info (loc_t *loc, struct iatt *iatt, struct iatt *parent)
+{
+ afr_update_loc_gfids (loc, iatt, parent);
+ uuid_copy (loc->inode->gfid, iatt->ia_gfid);
+}
+
static int
pump_mark_start_pending (xlator_t *this)
{
@@ -167,7 +165,7 @@ pump_save_path (xlator_t *this, const char *path)
GF_ASSERT (priv->root_inode);
- afr_build_root_loc (priv->root_inode, &loc);
+ afr_build_root_loc (this, &loc);
dict = dict_new ();
dict_ret = dict_set_str (dict, PUMP_PATH, (char *)path);
@@ -187,6 +185,7 @@ pump_save_path (xlator_t *this, const char *path)
dict_unref (dict);
+ loc_wipe (&loc);
return 0;
}
@@ -384,8 +383,7 @@ gf_pump_traverse_directory (loc_t *loc)
}
loc_wipe (&entry_loc);
ret = afr_build_child_loc (this, &entry_loc, loc,
- entry->d_name,
- entry->d_stat.ia_gfid);
+ entry->d_name);
if (ret)
goto out;
@@ -407,7 +405,7 @@ gf_pump_traverse_directory (loc_t *loc)
entry_loc.path);
continue;
}
- afr_fill_loc_info (&entry_loc, &iatt,
+ pump_fill_loc_info (&entry_loc, &iatt,
&parent);
pump_update_resume_state (this, entry_loc.path);
@@ -430,7 +428,7 @@ gf_pump_traverse_directory (loc_t *loc)
gf_pump_traverse_directory (&entry_loc);
}
}
- }
+ }
}
gf_dirent_free (&entries);
@@ -441,6 +439,10 @@ gf_pump_traverse_directory (loc_t *loc)
}
+ ret = syncop_close (fd);
+ if (ret < 0)
+ gf_log (this->name, GF_LOG_DEBUG, "closing the fd failed");
+
if (is_directory_empty && IS_ROOT_PATH (loc->path)) {
pump_change_state (this, PUMP_STATE_RUNNING);
gf_log (this->name, GF_LOG_INFO, "Empty source brick. "
@@ -480,7 +482,7 @@ pump_update_resume_path (xlator_t *this)
static int32_t
pump_xattr_cleaner (call_frame_t *frame, void *cookie, xlator_t *this,
- int32_t op_ret, int32_t op_errno)
+ int32_t op_ret, int32_t op_errno, dict_t *xdata)
{
afr_private_t *priv = NULL;
loc_t loc = {0};
@@ -491,7 +493,7 @@ pump_xattr_cleaner (call_frame_t *frame, void *cookie, xlator_t *this,
priv = this->private;
- afr_build_root_loc (priv->root_inode, &loc);
+ afr_build_root_loc (this, &loc);
ret = syncop_removexattr (priv->children[source], &loc,
PUMP_PATH);
@@ -507,6 +509,7 @@ pump_xattr_cleaner (call_frame_t *frame, void *cookie, xlator_t *this,
"failed with %s", strerror (errno));
}
+ loc_wipe (&loc);
return pump_command_reply (frame, this);
}
@@ -526,7 +529,7 @@ pump_complete_migration (xlator_t *this)
GF_ASSERT (priv->root_inode);
- afr_build_root_loc (priv->root_inode, &loc);
+ afr_build_root_loc (this, &loc);
dict = dict_new ();
@@ -568,6 +571,7 @@ pump_complete_migration (xlator_t *this)
call_resume (pump_priv->cleaner);
}
+ loc_wipe (&loc);
return 0;
}
@@ -623,7 +627,7 @@ pump_task (void *data)
GF_ASSERT (priv->root_inode);
- afr_build_root_loc (priv->root_inode, &loc);
+ afr_build_root_loc (this, &loc);
xattr_req = dict_new ();
if (!xattr_req) {
gf_log (this->name, GF_LOG_DEBUG,
@@ -661,6 +665,7 @@ out:
if (xattr_req)
dict_unref (xattr_req);
+ loc_wipe (&loc);
return 0;
}
@@ -694,7 +699,7 @@ pump_start (call_frame_t *pump_frame, xlator_t *this)
priv = this->private;
pump_priv = priv->pump_private;
- afr_set_lk_owner (pump_frame, this);
+ afr_set_lk_owner (pump_frame, this, pump_frame->root);
pump_pid = (uint64_t) (unsigned long)pump_frame->root;
ret = synctask_new (pump_priv->env, pump_task,
@@ -743,7 +748,7 @@ pump_cmd_start_setxattr_cbk (call_frame_t *frame,
void *cookie,
xlator_t *this,
int32_t op_ret,
- int32_t op_errno)
+ int32_t op_errno, dict_t *xdata)
{
call_frame_t *prev = NULL;
@@ -795,7 +800,7 @@ pump_initiate_sink_connect (call_frame_t *frame, xlator_t *this)
GF_ASSERT (priv->root_inode);
- afr_build_root_loc (priv->root_inode, &loc);
+ afr_build_root_loc (this, &loc);
data = data_ref (dict_get (local->dict, RB_PUMP_CMD_START));
if (!data) {
@@ -836,7 +841,7 @@ pump_initiate_sink_connect (call_frame_t *frame, xlator_t *this)
PUMP_SINK_CHILD(this)->fops->setxattr,
&loc,
dict,
- 0);
+ 0, NULL);
ret = 0;
@@ -850,6 +855,7 @@ out:
if (ret && clnt_cmd)
GF_FREE (clnt_cmd);
+ loc_wipe (&loc);
return ret;
}
@@ -869,7 +875,7 @@ pump_cmd_start_getxattr_cbk (call_frame_t *frame,
xlator_t *this,
int32_t op_ret,
int32_t op_errno,
- dict_t *dict)
+ dict_t *dict, dict_t *xdata)
{
afr_local_t *local = NULL;
char *path = NULL;
@@ -936,6 +942,7 @@ pump_execute_status (call_frame_t *frame, xlator_t *this)
uint64_t number_files = 0;
char filename[PATH_MAX];
+ char summary[PATH_MAX+256];
char *dict_str = NULL;
int32_t op_ret = 0;
@@ -964,12 +971,15 @@ pump_execute_status (call_frame_t *frame, xlator_t *this)
}
if (pump_priv->pump_finished) {
- snprintf (dict_str, PATH_MAX + 256, "Number of files migrated = %"PRIu64" Migration complete ",
- number_files);
+ snprintf (summary, PATH_MAX+256,
+ "no_of_files=%"PRIu64, number_files);
} else {
- snprintf (dict_str, PATH_MAX + 256, "Number of files migrated = %"PRIu64" Current file= %s ",
- number_files, filename);
+ snprintf (summary, PATH_MAX+256,
+ "no_of_files=%"PRIu64":current_file=%s",
+ number_files, filename);
}
+ snprintf (dict_str, PATH_MAX+256, "status=%d:%s",
+ (pump_priv->pump_finished)?1:0, summary);
dict = dict_new ();
@@ -985,13 +995,12 @@ pump_execute_status (call_frame_t *frame, xlator_t *this)
out:
- AFR_STACK_UNWIND (getxattr, frame, op_ret, op_errno, dict);
+ AFR_STACK_UNWIND (getxattr, frame, op_ret, op_errno, dict, NULL);
if (dict)
dict_unref (dict);
- if (dict_str)
- GF_FREE (dict_str);
+ GF_FREE (dict_str);
return 0;
}
@@ -1033,14 +1042,14 @@ pump_execute_start (call_frame_t *frame, xlator_t *this)
GF_ASSERT (priv->root_inode);
- afr_build_root_loc (priv->root_inode, &loc);
+ afr_build_root_loc (this, &loc);
STACK_WIND (frame,
pump_cmd_start_getxattr_cbk,
PUMP_SOURCE_CHILD(this),
PUMP_SOURCE_CHILD(this)->fops->getxattr,
&loc,
- PUMP_PATH);
+ PUMP_PATH, NULL);
ret = 0;
@@ -1050,6 +1059,7 @@ out:
pump_command_reply (frame, this);
}
+ loc_wipe (&loc);
return 0;
}
@@ -1057,7 +1067,7 @@ static int
pump_cleanup_helper (void *data) {
call_frame_t *frame = data;
- pump_xattr_cleaner (frame, 0, frame->this, 0, 0);
+ pump_xattr_cleaner (frame, 0, frame->this, 0, 0, NULL);
return 0;
}
@@ -1139,7 +1149,7 @@ pump_execute_abort (call_frame_t *frame, xlator_t *this)
} else {
pump_priv->cleaner = fop_setxattr_cbk_stub (frame,
pump_xattr_cleaner,
- 0, 0);
+ 0, 0, NULL);
}
return 0;
@@ -1270,7 +1280,7 @@ struct _xattr_key {
struct list_head list;
};
-static void
+static int
__gather_xattr_keys (dict_t *dict, char *key, data_t *value,
void *data)
{
@@ -1282,13 +1292,14 @@ __gather_xattr_keys (dict_t *dict, char *key, data_t *value,
xkey = GF_CALLOC (1, sizeof (*xkey), gf_afr_mt_xattr_key);
if (!xkey)
- return;
+ return -1;
xkey->key = key;
INIT_LIST_HEAD (&xkey->list);
list_add_tail (&xkey->list, list);
}
+ return 0;
}
static void
@@ -1316,7 +1327,7 @@ __filter_xattrs (dict_t *dict)
int32_t
pump_getxattr_cbk (call_frame_t *frame, void *cookie,
xlator_t *this, int32_t op_ret, int32_t op_errno,
- dict_t *dict)
+ dict_t *dict, dict_t *xdata)
{
afr_private_t *priv = NULL;
afr_local_t *local = NULL;
@@ -1351,7 +1362,7 @@ pump_getxattr_cbk (call_frame_t *frame, void *cookie,
children[next_call_child],
children[next_call_child]->fops->getxattr,
&local->loc,
- local->cont.getxattr.name);
+ local->cont.getxattr.name, NULL);
}
out:
@@ -1359,7 +1370,7 @@ out:
if (op_ret >= 0 && dict)
__filter_xattrs (dict);
- AFR_STACK_UNWIND (getxattr, frame, op_ret, op_errno, dict);
+ AFR_STACK_UNWIND (getxattr, frame, op_ret, op_errno, dict, NULL);
}
return 0;
@@ -1367,7 +1378,7 @@ out:
int32_t
pump_getxattr (call_frame_t *frame, xlator_t *this,
- loc_t *loc, const char *name)
+ loc_t *loc, const char *name, dict_t *xdata)
{
afr_private_t * priv = NULL;
xlator_t ** children = NULL;
@@ -1390,12 +1401,12 @@ pump_getxattr (call_frame_t *frame, xlator_t *this,
STACK_WIND (frame, default_getxattr_cbk,
FIRST_CHILD (this),
(FIRST_CHILD (this))->fops->getxattr,
- loc, name);
+ loc, name, xdata);
return 0;
}
- ALLOC_OR_GOTO (frame->local, afr_local_t, out);
+ AFR_LOCAL_ALLOC_OR_GOTO (frame->local, out);
local = frame->local;
ret = afr_local_init (local, priv, &op_errno);
@@ -1444,12 +1455,12 @@ pump_getxattr (call_frame_t *frame, xlator_t *this,
STACK_WIND_COOKIE (frame, pump_getxattr_cbk,
(void *) (long) call_child,
children[call_child], children[call_child]->fops->getxattr,
- loc, name);
+ loc, name, xdata);
ret = 0;
out:
if (ret < 0)
- AFR_STACK_UNWIND (getxattr, frame, -1, op_errno, NULL);
+ AFR_STACK_UNWIND (getxattr, frame, -1, op_errno, NULL, NULL);
return 0;
}
@@ -1471,14 +1482,14 @@ afr_setxattr_unwind (call_frame_t *frame, xlator_t *this)
if (main_frame) {
AFR_STACK_UNWIND (setxattr, main_frame,
- local->op_ret, local->op_errno)
+ local->op_ret, local->op_errno, NULL);
}
return 0;
}
static int
afr_setxattr_wind_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
- int32_t op_ret, int32_t op_errno)
+ int32_t op_ret, int32_t op_errno, dict_t *xdata)
{
afr_local_t * local = NULL;
afr_private_t * priv = NULL;
@@ -1547,7 +1558,7 @@ afr_setxattr_wind (call_frame_t *frame, xlator_t *this)
priv->children[i]->fops->setxattr,
&local->loc,
local->cont.setxattr.dict,
- local->cont.setxattr.flags);
+ local->cont.setxattr.flags, NULL);
if (!--call_count)
break;
@@ -1575,11 +1586,9 @@ pump_setxattr_cbk (call_frame_t *frame,
void *cookie,
xlator_t *this,
int32_t op_ret,
- int32_t op_errno)
+ int32_t op_errno, dict_t *xdata)
{
- STACK_UNWIND (frame,
- op_ret,
- op_errno);
+ AFR_STACK_UNWIND (setxattr, frame, op_ret, op_errno, xdata);
return 0;
}
@@ -1597,12 +1606,10 @@ pump_command_reply (call_frame_t *frame, xlator_t *this)
gf_log (this->name, GF_LOG_INFO,
"Command succeeded");
- dict_unref (local->dict);
-
AFR_STACK_UNWIND (setxattr,
frame,
local->op_ret,
- local->op_errno);
+ local->op_errno, NULL);
return 0;
}
@@ -1639,12 +1646,11 @@ pump_parse_command (call_frame_t *frame, xlator_t *this,
int
pump_setxattr (call_frame_t *frame, xlator_t *this,
- loc_t *loc, dict_t *dict, int32_t flags)
+ loc_t *loc, dict_t *dict, int32_t flags, dict_t *xdata)
{
afr_private_t * priv = NULL;
afr_local_t * local = NULL;
call_frame_t *transaction_frame = NULL;
- data_pair_t * trav = NULL;
int ret = -1;
int op_errno = 0;
@@ -1653,19 +1659,19 @@ pump_setxattr (call_frame_t *frame, xlator_t *this,
VALIDATE_OR_GOTO (this->private, out);
GF_IF_INTERNAL_XATTR_GOTO ("trusted.glusterfs.pump*", dict,
- trav, op_errno, out);
+ op_errno, out);
priv = this->private;
if (!priv->use_afr_in_pump) {
STACK_WIND (frame, default_setxattr_cbk,
FIRST_CHILD (this),
(FIRST_CHILD (this))->fops->setxattr,
- loc, dict, flags);
+ loc, dict, flags, xdata);
return 0;
}
- ALLOC_OR_GOTO (local, afr_local_t, out);
+ AFR_LOCAL_ALLOC_OR_GOTO (local, out);
ret = afr_local_init (local, priv, &op_errno);
if (ret < 0) {
@@ -1714,7 +1720,7 @@ out:
if (ret < 0) {
if (transaction_frame)
AFR_STACK_DESTROY (transaction_frame);
- AFR_STACK_UNWIND (setxattr, frame, -1, op_errno);
+ AFR_STACK_UNWIND (setxattr, frame, -1, op_errno, NULL);
}
return 0;
@@ -1748,7 +1754,7 @@ static int32_t
pump_truncate (call_frame_t *frame,
xlator_t *this,
loc_t *loc,
- off_t offset)
+ off_t offset, dict_t *xdata)
{
afr_private_t *priv = NULL;
priv = this->private;
@@ -1758,11 +1764,11 @@ pump_truncate (call_frame_t *frame,
FIRST_CHILD(this),
FIRST_CHILD(this)->fops->truncate,
loc,
- offset);
+ offset, xdata);
return 0;
}
- afr_truncate (frame, this, loc, offset);
+ afr_truncate (frame, this, loc, offset, xdata);
return 0;
}
@@ -1771,7 +1777,7 @@ static int32_t
pump_ftruncate (call_frame_t *frame,
xlator_t *this,
fd_t *fd,
- off_t offset)
+ off_t offset, dict_t *xdata)
{
afr_private_t *priv = NULL;
priv = this->private;
@@ -1781,11 +1787,11 @@ pump_ftruncate (call_frame_t *frame,
FIRST_CHILD(this),
FIRST_CHILD(this)->fops->ftruncate,
fd,
- offset);
+ offset, xdata);
return 0;
}
- afr_ftruncate (frame, this, fd, offset);
+ afr_ftruncate (frame, this, fd, offset, xdata);
return 0;
}
@@ -1794,7 +1800,7 @@ pump_ftruncate (call_frame_t *frame,
int
pump_mknod (call_frame_t *frame, xlator_t *this,
- loc_t *loc, mode_t mode, dev_t rdev, dict_t *parms)
+ loc_t *loc, mode_t mode, dev_t rdev, mode_t umask, dict_t *xdata)
{
afr_private_t *priv = NULL;
priv = this->private;
@@ -1802,10 +1808,10 @@ pump_mknod (call_frame_t *frame, xlator_t *this,
STACK_WIND (frame, default_mknod_cbk,
FIRST_CHILD(this),
FIRST_CHILD(this)->fops->mknod,
- loc, mode, rdev, parms);
+ loc, mode, rdev, umask, xdata);
return 0;
}
- afr_mknod (frame, this, loc, mode, rdev, parms);
+ afr_mknod (frame, this, loc, mode, rdev, umask, xdata);
return 0;
}
@@ -1814,7 +1820,7 @@ pump_mknod (call_frame_t *frame, xlator_t *this,
int
pump_mkdir (call_frame_t *frame, xlator_t *this,
- loc_t *loc, mode_t mode, dict_t *params)
+ loc_t *loc, mode_t mode, mode_t umask, dict_t *xdata)
{
afr_private_t *priv = NULL;
priv = this->private;
@@ -1822,10 +1828,10 @@ pump_mkdir (call_frame_t *frame, xlator_t *this,
STACK_WIND (frame, default_mkdir_cbk,
FIRST_CHILD(this),
FIRST_CHILD(this)->fops->mkdir,
- loc, mode, params);
+ loc, mode, umask, xdata);
return 0;
}
- afr_mkdir (frame, this, loc, mode, params);
+ afr_mkdir (frame, this, loc, mode, umask, xdata);
return 0;
}
@@ -1834,7 +1840,7 @@ pump_mkdir (call_frame_t *frame, xlator_t *this,
static int32_t
pump_unlink (call_frame_t *frame,
xlator_t *this,
- loc_t *loc)
+ loc_t *loc, int xflag, dict_t *xdata)
{
afr_private_t *priv = NULL;
priv = this->private;
@@ -1843,10 +1849,10 @@ pump_unlink (call_frame_t *frame,
default_unlink_cbk,
FIRST_CHILD(this),
FIRST_CHILD(this)->fops->unlink,
- loc);
+ loc, xflag, xdata);
return 0;
}
- afr_unlink (frame, this, loc);
+ afr_unlink (frame, this, loc, xflag, xdata);
return 0;
}
@@ -1854,7 +1860,7 @@ pump_unlink (call_frame_t *frame,
static int
pump_rmdir (call_frame_t *frame, xlator_t *this,
- loc_t *loc, int flags)
+ loc_t *loc, int flags, dict_t *xdata)
{
afr_private_t *priv = NULL;
@@ -1864,11 +1870,11 @@ pump_rmdir (call_frame_t *frame, xlator_t *this,
STACK_WIND (frame, default_rmdir_cbk,
FIRST_CHILD(this),
FIRST_CHILD(this)->fops->rmdir,
- loc, flags);
+ loc, flags, xdata);
return 0;
}
- afr_rmdir (frame, this, loc, flags);
+ afr_rmdir (frame, this, loc, flags, xdata);
return 0;
}
@@ -1877,7 +1883,7 @@ pump_rmdir (call_frame_t *frame, xlator_t *this,
int
pump_symlink (call_frame_t *frame, xlator_t *this,
- const char *linkpath, loc_t *loc, dict_t *params)
+ const char *linkpath, loc_t *loc, mode_t umask, dict_t *xdata)
{
afr_private_t *priv = NULL;
priv = this->private;
@@ -1885,10 +1891,10 @@ pump_symlink (call_frame_t *frame, xlator_t *this,
STACK_WIND (frame, default_symlink_cbk,
FIRST_CHILD(this),
FIRST_CHILD(this)->fops->symlink,
- linkpath, loc, params);
+ linkpath, loc, umask, xdata);
return 0;
}
- afr_symlink (frame, this, linkpath, loc, params);
+ afr_symlink (frame, this, linkpath, loc, umask, xdata);
return 0;
}
@@ -1898,7 +1904,7 @@ static int32_t
pump_rename (call_frame_t *frame,
xlator_t *this,
loc_t *oldloc,
- loc_t *newloc)
+ loc_t *newloc, dict_t *xdata)
{
afr_private_t *priv = NULL;
priv = this->private;
@@ -1907,10 +1913,10 @@ pump_rename (call_frame_t *frame,
default_rename_cbk,
FIRST_CHILD(this),
FIRST_CHILD(this)->fops->rename,
- oldloc, newloc);
+ oldloc, newloc, xdata);
return 0;
}
- afr_rename (frame, this, oldloc, newloc);
+ afr_rename (frame, this, oldloc, newloc, xdata);
return 0;
}
@@ -1920,7 +1926,7 @@ static int32_t
pump_link (call_frame_t *frame,
xlator_t *this,
loc_t *oldloc,
- loc_t *newloc)
+ loc_t *newloc, dict_t *xdata)
{
afr_private_t *priv = NULL;
priv = this->private;
@@ -1929,10 +1935,10 @@ pump_link (call_frame_t *frame,
default_link_cbk,
FIRST_CHILD(this),
FIRST_CHILD(this)->fops->link,
- oldloc, newloc);
+ oldloc, newloc, xdata);
return 0;
}
- afr_link (frame, this, oldloc, newloc);
+ afr_link (frame, this, oldloc, newloc, xdata);
return 0;
}
@@ -1941,7 +1947,7 @@ pump_link (call_frame_t *frame,
static int32_t
pump_create (call_frame_t *frame, xlator_t *this,
loc_t *loc, int32_t flags, mode_t mode,
- fd_t *fd, dict_t *params)
+ mode_t umask, fd_t *fd, dict_t *xdata)
{
afr_private_t *priv = NULL;
priv = this->private;
@@ -1949,10 +1955,10 @@ pump_create (call_frame_t *frame, xlator_t *this,
STACK_WIND (frame, default_create_cbk,
FIRST_CHILD(this),
FIRST_CHILD(this)->fops->create,
- loc, flags, mode, fd, params);
+ loc, flags, mode, umask, fd, xdata);
return 0;
}
- afr_create (frame, this, loc, flags, mode, fd, params);
+ afr_create (frame, this, loc, flags, mode, umask, fd, xdata);
return 0;
}
@@ -1962,8 +1968,7 @@ static int32_t
pump_open (call_frame_t *frame,
xlator_t *this,
loc_t *loc,
- int32_t flags, fd_t *fd,
- int32_t wbflags)
+ int32_t flags, fd_t *fd, dict_t *xdata)
{
afr_private_t *priv = NULL;
priv = this->private;
@@ -1972,10 +1977,10 @@ pump_open (call_frame_t *frame,
default_open_cbk,
FIRST_CHILD(this),
FIRST_CHILD(this)->fops->open,
- loc, flags, fd, wbflags);
+ loc, flags, fd, xdata);
return 0;
}
- afr_open (frame, this, loc, flags, fd, wbflags);
+ afr_open (frame, this, loc, flags, fd, xdata);
return 0;
}
@@ -1987,8 +1992,8 @@ pump_writev (call_frame_t *frame,
fd_t *fd,
struct iovec *vector,
int32_t count,
- off_t off,
- struct iobref *iobref)
+ off_t off, uint32_t flags,
+ struct iobref *iobref, dict_t *xdata)
{
afr_private_t *priv = NULL;
priv = this->private;
@@ -2000,20 +2005,20 @@ pump_writev (call_frame_t *frame,
fd,
vector,
count,
- off,
- iobref);
+ off, flags,
+ iobref, xdata);
return 0;
}
- afr_writev (frame, this, fd, vector, count, off, iobref);
- return 0;
+ afr_writev (frame, this, fd, vector, count, off, flags, iobref, xdata);
+ return 0;
}
static int32_t
pump_flush (call_frame_t *frame,
xlator_t *this,
- fd_t *fd)
+ fd_t *fd, dict_t *xdata)
{
afr_private_t *priv = NULL;
priv = this->private;
@@ -2022,10 +2027,10 @@ pump_flush (call_frame_t *frame,
default_flush_cbk,
FIRST_CHILD(this),
FIRST_CHILD(this)->fops->flush,
- fd);
+ fd, xdata);
return 0;
}
- afr_flush (frame, this, fd);
+ afr_flush (frame, this, fd, xdata);
return 0;
}
@@ -2035,7 +2040,7 @@ static int32_t
pump_fsync (call_frame_t *frame,
xlator_t *this,
fd_t *fd,
- int32_t flags)
+ int32_t flags, dict_t *xdata)
{
afr_private_t *priv = NULL;
priv = this->private;
@@ -2045,10 +2050,10 @@ pump_fsync (call_frame_t *frame,
FIRST_CHILD(this),
FIRST_CHILD(this)->fops->fsync,
fd,
- flags);
+ flags, xdata);
return 0;
}
- afr_fsync (frame, this, fd, flags);
+ afr_fsync (frame, this, fd, flags, xdata);
return 0;
}
@@ -2057,7 +2062,7 @@ pump_fsync (call_frame_t *frame,
static int32_t
pump_opendir (call_frame_t *frame,
xlator_t *this,
- loc_t *loc, fd_t *fd)
+ loc_t *loc, fd_t *fd, dict_t *xdata)
{
afr_private_t *priv = NULL;
priv = this->private;
@@ -2066,10 +2071,10 @@ pump_opendir (call_frame_t *frame,
default_opendir_cbk,
FIRST_CHILD(this),
FIRST_CHILD(this)->fops->opendir,
- loc, fd);
+ loc, fd, xdata);
return 0;
}
- afr_opendir (frame, this, loc, fd);
+ afr_opendir (frame, this, loc, fd, xdata);
return 0;
}
@@ -2079,7 +2084,7 @@ static int32_t
pump_fsyncdir (call_frame_t *frame,
xlator_t *this,
fd_t *fd,
- int32_t flags)
+ int32_t flags, dict_t *xdata)
{
afr_private_t *priv = NULL;
priv = this->private;
@@ -2089,10 +2094,10 @@ pump_fsyncdir (call_frame_t *frame,
FIRST_CHILD(this),
FIRST_CHILD(this)->fops->fsyncdir,
fd,
- flags);
+ flags, xdata);
return 0;
}
- afr_fsyncdir (frame, this, fd, flags);
+ afr_fsyncdir (frame, this, fd, flags, xdata);
return 0;
}
@@ -2103,7 +2108,7 @@ pump_xattrop (call_frame_t *frame,
xlator_t *this,
loc_t *loc,
gf_xattrop_flags_t flags,
- dict_t *dict)
+ dict_t *dict, dict_t *xdata)
{
afr_private_t *priv = NULL;
priv = this->private;
@@ -2114,10 +2119,10 @@ pump_xattrop (call_frame_t *frame,
FIRST_CHILD(this)->fops->xattrop,
loc,
flags,
- dict);
+ dict, xdata);
return 0;
}
- afr_xattrop (frame, this, loc, flags, dict);
+ afr_xattrop (frame, this, loc, flags, dict, xdata);
return 0;
}
@@ -2127,7 +2132,7 @@ pump_fxattrop (call_frame_t *frame,
xlator_t *this,
fd_t *fd,
gf_xattrop_flags_t flags,
- dict_t *dict)
+ dict_t *dict, dict_t *xdata)
{
afr_private_t *priv = NULL;
priv = this->private;
@@ -2138,10 +2143,10 @@ pump_fxattrop (call_frame_t *frame,
FIRST_CHILD(this)->fops->fxattrop,
fd,
flags,
- dict);
+ dict, xdata);
return 0;
}
- afr_fxattrop (frame, this, fd, flags, dict);
+ afr_fxattrop (frame, this, fd, flags, dict, xdata);
return 0;
}
@@ -2151,9 +2156,17 @@ static int32_t
pump_removexattr (call_frame_t *frame,
xlator_t *this,
loc_t *loc,
- const char *name)
+ const char *name, dict_t *xdata)
{
- afr_private_t *priv = NULL;
+ afr_private_t *priv = NULL;
+ int op_errno = -1;
+
+ VALIDATE_OR_GOTO (this, out);
+
+ GF_IF_NATIVE_XATTR_GOTO ("trusted.glusterfs.pump*",
+ name, op_errno, out);
+
+ op_errno = 0;
priv = this->private;
if (!priv->use_afr_in_pump) {
STACK_WIND (frame,
@@ -2161,10 +2174,14 @@ pump_removexattr (call_frame_t *frame,
FIRST_CHILD(this),
FIRST_CHILD(this)->fops->removexattr,
loc,
- name);
+ name, xdata);
return 0;
}
- afr_removexattr (frame, this, loc, name);
+ afr_removexattr (frame, this, loc, name, xdata);
+
+ out:
+ if (op_errno)
+ AFR_STACK_UNWIND (removexattr, frame, -1, op_errno, NULL);
return 0;
}
@@ -2176,7 +2193,7 @@ pump_readdir (call_frame_t *frame,
xlator_t *this,
fd_t *fd,
size_t size,
- off_t off)
+ off_t off, dict_t *xdata)
{
afr_private_t *priv = NULL;
priv = this->private;
@@ -2185,10 +2202,10 @@ pump_readdir (call_frame_t *frame,
default_readdir_cbk,
FIRST_CHILD(this),
FIRST_CHILD(this)->fops->readdir,
- fd, size, off);
+ fd, size, off, xdata);
return 0;
}
- afr_readdir (frame, this, fd, size, off);
+ afr_readdir (frame, this, fd, size, off, xdata);
return 0;
}
@@ -2239,13 +2256,24 @@ pump_release (xlator_t *this,
}
+static int32_t
+pump_forget (xlator_t *this, inode_t *inode)
+{
+ afr_private_t *priv = NULL;
+
+ priv = this->private;
+ if (priv->use_afr_in_pump)
+ afr_forget (this, inode);
+
+ return 0;
+}
static int32_t
pump_setattr (call_frame_t *frame,
xlator_t *this,
loc_t *loc,
struct iatt *stbuf,
- int32_t valid)
+ int32_t valid, dict_t *xdata)
{
afr_private_t *priv = NULL;
priv = this->private;
@@ -2254,10 +2282,10 @@ pump_setattr (call_frame_t *frame,
default_setattr_cbk,
FIRST_CHILD (this),
FIRST_CHILD (this)->fops->setattr,
- loc, stbuf, valid);
+ loc, stbuf, valid, xdata);
return 0;
}
- afr_setattr (frame, this, loc, stbuf, valid);
+ afr_setattr (frame, this, loc, stbuf, valid, xdata);
return 0;
}
@@ -2268,7 +2296,7 @@ pump_fsetattr (call_frame_t *frame,
xlator_t *this,
fd_t *fd,
struct iatt *stbuf,
- int32_t valid)
+ int32_t valid, dict_t *xdata)
{
afr_private_t *priv = NULL;
priv = this->private;
@@ -2277,10 +2305,10 @@ pump_fsetattr (call_frame_t *frame,
default_fsetattr_cbk,
FIRST_CHILD (this),
FIRST_CHILD (this)->fops->fsetattr,
- fd, stbuf, valid);
+ fd, stbuf, valid, xdata);
return 0;
}
- afr_fsetattr (frame, this, fd, stbuf, valid);
+ afr_fsetattr (frame, this, fd, stbuf, valid, xdata);
return 0;
}
@@ -2329,7 +2357,7 @@ notify (xlator_t *this, int32_t event,
child_xl = (xlator_t *) data;
- ret = afr_notify (this, event, data);
+ ret = afr_notify (this, event, data, NULL);
switch (event) {
case GF_EVENT_CHILD_DOWN:
@@ -2380,7 +2408,10 @@ init (xlator_t *this)
"Volume is dangling.");
}
- ALLOC_OR_GOTO (this->private, afr_private_t, out);
+ this->private = GF_CALLOC (1, sizeof (afr_private_t),
+ gf_afr_mt_afr_private_t);
+ if (!this->private)
+ goto out;
priv = this->private;
LOCK_INIT (&priv->lock);
@@ -2406,14 +2437,13 @@ init (xlator_t *this)
priv->metadata_self_heal = 1;
priv->entry_self_heal = 1;
- priv->data_self_heal_algorithm = "";
-
priv->data_self_heal_window_size = 16;
priv->data_change_log = 1;
priv->metadata_change_log = 1;
priv->entry_change_log = 1;
priv->use_afr_in_pump = 1;
+ priv->sh_readdir_size = 65536;
/* Locking options */
@@ -2471,6 +2501,12 @@ init (xlator_t *this)
i++;
}
+ ret = gf_asprintf (&priv->sh_domain, "%s-self-heal", this->name);
+ if (-1 == ret) {
+ op_errno = ENOMEM;
+ goto out;
+ }
+
priv->first_lookup = 1;
priv->root_inode = NULL;
@@ -2502,7 +2538,7 @@ init (xlator_t *this)
goto out;
}
- pump_priv->env = syncenv_new (0);
+ pump_priv->env = this->ctx->env;
if (!pump_priv->env) {
gf_log (this->name, GF_LOG_ERROR,
"Could not create new sync-environment");
@@ -2510,6 +2546,15 @@ init (xlator_t *this)
goto out;
}
+ /* keep more local here as we may need them for self-heal etc */
+ this->local_pool = mem_pool_new (afr_local_t, 128);
+ if (!this->local_pool) {
+ ret = -1;
+ gf_log (this->name, GF_LOG_ERROR,
+ "failed to create local_t's memory pool");
+ goto out;
+ }
+
priv->pump_private = pump_priv;
pump_change_state (this, PUMP_STATE_ABORT);
@@ -2534,9 +2579,6 @@ fini (xlator_t *this)
if (!pump_priv)
goto afr_priv;
- if (pump_priv->env)
- syncenv_destroy (pump_priv->env);
-
GF_FREE (pump_priv->resume_path);
LOCK_DESTROY (&pump_priv->resume_path_lock);
LOCK_DESTROY (&pump_priv->pump_state_lock);
@@ -2591,6 +2633,7 @@ struct xlator_dumpops dumpops = {
struct xlator_cbks cbks = {
.release = pump_release,
.releasedir = pump_releasedir,
+ .forget = pump_forget,
};
struct volume_options options[] = {
diff --git a/xlators/cluster/afr/src/pump.h b/xlators/cluster/afr/src/pump.h
index 4213a635e..bc4c31a78 100644
--- a/xlators/cluster/afr/src/pump.h
+++ b/xlators/cluster/afr/src/pump.h
@@ -1,20 +1,11 @@
/*
- Copyright (c) 2007-2011 Gluster, Inc. <http://www.gluster.com>
- This file is part of GlusterFS.
-
- GlusterFS is free software; you can redistribute it and/or modify
- it under the terms of the GNU General Public License as published
- by the Free Software Foundation; either version 3 of the License,
- or (at your option) any later version.
-
- GlusterFS is distributed in the hope that it will be useful, but
- WITHOUT ANY WARRANTY; without even the implied warranty of
- MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU
- General Public License for more details.
-
- You should have received a copy of the GNU General Public License
- along with this program. If not, see
- <http://www.gnu.org/licenses/>.
+ Copyright (c) 2008-2012 Red Hat, Inc. <http://www.redhat.com>
+ This file is part of GlusterFS.
+
+ This file is licensed to you under your choice of the GNU Lesser
+ General Public License, version 3 or any later version (LGPLv3 or
+ later), or the GNU General Public License, version 2 (GPLv2), in all
+ cases as published by the Free Software Foundation.
*/
#ifndef __PUMP_H__
diff --git a/xlators/cluster/dht/src/Makefile.am b/xlators/cluster/dht/src/Makefile.am
index e35058d65..174bea841 100644
--- a/xlators/cluster/dht/src/Makefile.am
+++ b/xlators/cluster/dht/src/Makefile.am
@@ -4,7 +4,7 @@ xlatordir = $(libdir)/glusterfs/$(PACKAGE_VERSION)/xlator/cluster
dht_common_source = dht-layout.c dht-helper.c dht-linkfile.c dht-rebalance.c \
dht-selfheal.c dht-rename.c dht-hashfn.c dht-diskusage.c \
- dht-common.c dht-inode-write.c dht-inode-read.c \
+ dht-common.c dht-inode-write.c dht-inode-read.c dht-shared.c \
$(top_builddir)/xlators/lib/src/libxlator.c
dht_la_SOURCES = $(dht_common_source) dht.c
@@ -12,22 +12,23 @@ dht_la_SOURCES = $(dht_common_source) dht.c
nufa_la_SOURCES = $(dht_common_source) nufa.c
switch_la_SOURCES = $(dht_common_source) switch.c
-dht_la_LDFLAGS = -module -avoidversion
+dht_la_LDFLAGS = -module -avoid-version
dht_la_LIBADD = $(top_builddir)/libglusterfs/src/libglusterfs.la
-nufa_la_LDFLAGS = -module -avoidversion
+nufa_la_LDFLAGS = -module -avoid-version
nufa_la_LIBADD = $(top_builddir)/libglusterfs/src/libglusterfs.la
-switch_la_LDFLAGS = -module -avoidversion
+switch_la_LDFLAGS = -module -avoid-version
switch_la_LIBADD = $(top_builddir)/libglusterfs/src/libglusterfs.la
noinst_HEADERS = dht-common.h dht-mem-types.h \
$(top_builddir)/xlators/lib/src/libxlator.h
-AM_CFLAGS = -fPIC -D_FILE_OFFSET_BITS=64 -D_GNU_SOURCE -Wall -D$(GF_HOST_OS) \
- -I$(top_srcdir)/libglusterfs/src -shared -nostartfiles $(GF_CFLAGS) \
+AM_CPPFLAGS = $(GF_CPPFLAGS) -I$(top_srcdir)/libglusterfs/src \
-I$(top_srcdir)/xlators/lib/src
+AM_CFLAGS = -Wall $(GF_CFLAGS)
+
CLEANFILES =
uninstall-local:
diff --git a/xlators/cluster/dht/src/dht-common.c b/xlators/cluster/dht/src/dht-common.c
index a16ed7b6c..8f61339e6 100644
--- a/xlators/cluster/dht/src/dht-common.c
+++ b/xlators/cluster/dht/src/dht-common.c
@@ -1,20 +1,11 @@
/*
- Copyright (c) 2009-2011 Gluster, Inc. <http://www.gluster.com>
+ Copyright (c) 2008-2012 Red Hat, Inc. <http://www.redhat.com>
This file is part of GlusterFS.
- GlusterFS is free software; you can redistribute it and/or modify
- it under the terms of the GNU General Public License as published
- by the Free Software Foundation; either version 3 of the License,
- or (at your option) any later version.
-
- GlusterFS is distributed in the hope that it will be useful, but
- WITHOUT ANY WARRANTY; without even the implied warranty of
- MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU
- General Public License for more details.
-
- You should have received a copy of the GNU General Public License
- along with this program. If not, see
- <http://www.gnu.org/licenses/>.
+ This file is licensed to you under your choice of the GNU Lesser
+ General Public License, version 3 or any later version (LGPLv3 or
+ later), or the GNU General Public License, version 2 (GPLv2), in all
+ cases as published by the Free Software Foundation.
*/
@@ -31,17 +22,18 @@
#include "dht-common.h"
#include "defaults.h"
#include "byte-order.h"
+#include "glusterfs-acl.h"
#include <sys/time.h>
#include <libgen.h>
-void
+int
dht_aggregate (dict_t *this, char *key, data_t *value, void *data)
{
dict_t *dst = NULL;
int64_t *ptr = 0, *size = NULL;
int32_t ret = -1;
- data_pair_t *data_pair = NULL;
+ data_t *dict_data = NULL;
dst = data;
@@ -53,32 +45,37 @@ dht_aggregate (dict_t *this, char *key, data_t *value, void *data)
if (size == NULL) {
gf_log ("dht", GF_LOG_WARNING,
"memory allocation failed");
- return;
+ return -1;
}
ret = dict_set_bin (dst, key, size, sizeof (int64_t));
if (ret < 0) {
gf_log ("dht", GF_LOG_WARNING,
"dht aggregate dict set failed");
GF_FREE (size);
- return;
+ return -1;
}
}
ptr = data_to_bin (value);
if (ptr == NULL) {
gf_log ("dht", GF_LOG_WARNING, "data to bin failed");
- return;
+ return -1;
}
*size = hton64 (ntoh64 (*size) + ntoh64 (*ptr));
+
+ } else if (fnmatch (GF_XATTR_STIME_PATTERN, key, FNM_NOESCAPE) == 0) {
+ ret = gf_get_min_stime (THIS, dst, key, value);
+ if (ret < 0)
+ return ret;
} else {
/* compare user xattrs only */
if (!strncmp (key, "user.", strlen ("user."))) {
- ret = dict_lookup (dst, key, &data_pair);
- if (!ret && data_pair && value) {
- ret = is_data_equal (data_pair->value, value);
+ ret = dict_lookup (dst, key, &dict_data);
+ if (!ret && dict_data && value) {
+ ret = is_data_equal (dict_data, value);
if (!ret)
- gf_log ("dht", GF_LOG_WARNING,
+ gf_log ("dht", GF_LOG_DEBUG,
"xattr mismatch for %s", key);
}
}
@@ -87,7 +84,7 @@ dht_aggregate (dict_t *this, char *key, data_t *value, void *data)
gf_log ("dht", GF_LOG_WARNING, "xattr dict set failed");
}
- return;
+ return 0;
}
@@ -114,7 +111,7 @@ out:
int
dht_lookup_selfheal_cbk (call_frame_t *frame, void *cookie,
xlator_t *this,
- int op_ret, int op_errno)
+ int op_ret, int op_errno, dict_t *xdata)
{
dht_local_t *local = NULL;
dht_layout_t *layout = NULL;
@@ -134,7 +131,10 @@ dht_lookup_selfheal_cbk (call_frame_t *frame, void *cookie,
ret = dht_layout_set (this, local->inode, layout);
}
- WIPE (&local->postparent);
+ if (local->loc.parent) {
+ dht_inode_ctx_time_update (local->loc.parent, this,
+ &local->postparent, 1);
+ }
DHT_STRIP_PHASE1_FLAGS (&local->stbuf);
@@ -154,9 +154,11 @@ dht_discover_complete (xlator_t *this, call_frame_t *discover_frame)
int op_errno = 0;
int ret = -1;
dht_layout_t *layout = NULL;
+ dht_conf_t *conf = NULL;
local = discover_frame->local;
layout = local->layout;
+ conf = this->private;
LOCK(&discover_frame->lock);
{
@@ -190,16 +192,23 @@ dht_discover_complete (xlator_t *this, call_frame_t *discover_frame)
}
} else {
ret = dht_layout_normalize (this, &local->loc, layout);
-
- if (ret != 0) {
+ if ((ret < 0) || ((ret > 0) && (local->op_ret != 0))) {
+ /* either the layout is incorrect or the directory is
+ * not found even in one subvolume.
+ */
gf_log (this->name, GF_LOG_DEBUG,
- "normalizing failed on %s",
- local->loc.path);
- op_errno = EINVAL;
- goto out;
+ "normalizing failed on %s "
+ "(overlaps/holes present: %s, "
+ "ENOENT errors: %d)", local->loc.path,
+ (ret < 0) ? "yes" : "no", (ret > 0) ? ret : 0);
+ if ((ret > 0) && (ret == conf->subvolume_cnt)) {
+ op_errno = ESTALE;
+ goto out;
+ }
}
- dht_layout_set (this, local->inode, layout);
+ if (local->inode)
+ dht_layout_set (this, local->inode, layout);
}
DHT_STACK_UNWIND (lookup, main_frame, local->op_ret, local->op_errno,
@@ -228,6 +237,7 @@ dht_discover_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
int is_dir = 0;
int is_linkfile = 0;
int attempt_unwind = 0;
+ dht_conf_t *conf = 0;
GF_VALIDATE_OR_GOTO ("dht", frame, out);
GF_VALIDATE_OR_GOTO ("dht", this, out);
@@ -237,6 +247,7 @@ dht_discover_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
local = frame->local;
prev = cookie;
+ conf = this->private;
layout = local->layout;
@@ -271,7 +282,8 @@ dht_discover_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
goto unlock;
}
- is_linkfile = check_is_linkfile (inode, stbuf, xattr);
+ is_linkfile = check_is_linkfile (inode, stbuf, xattr,
+ conf->link_xattr_name);
is_dir = check_is_dir (inode, stbuf, xattr);
if (is_dir) {
@@ -330,23 +342,20 @@ dht_discover (call_frame_t *frame, xlator_t *this, loc_t *loc)
int i = 0;
call_frame_t *discover_frame = NULL;
-
conf = this->private;
local = frame->local;
- ret = dict_set_uint32 (local->xattr_req,
- "trusted.glusterfs.dht", 4 * 4);
+ ret = dict_set_uint32 (local->xattr_req, conf->xattr_name, 4 * 4);
if (ret)
gf_log (this->name, GF_LOG_WARNING,
- "%s: failed to set 'trusted.glusterfs.dht' key",
- loc->path);
+ "%s: failed to set '%s' key",
+ loc->path, conf->xattr_name);
- ret = dict_set_uint32 (local->xattr_req,
- "trusted.glusterfs.dht.linkto", 256);
+ ret = dict_set_uint32 (local->xattr_req, conf->link_xattr_name, 256);
if (ret)
gf_log (this->name, GF_LOG_WARNING,
- "%s: failed to set 'trusted.glusterfs.dht.linkto' key",
- loc->path);
+ "%s: failed to set '%s' key",
+ loc->path, conf->link_xattr_name);
call_cnt = conf->subvolume_cnt;
local->call_cnt = call_cnt;
@@ -380,7 +389,8 @@ dht_discover (call_frame_t *frame, xlator_t *this, loc_t *loc)
return 0;
err:
- DHT_STACK_UNWIND (lookup, frame, -1, op_errno, NULL, NULL, NULL, NULL);
+ DHT_STACK_UNWIND (lookup, frame, -1, op_errno, NULL, NULL, NULL,
+ NULL);
return 0;
}
@@ -431,7 +441,7 @@ dht_lookup_dir_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
op_ret, op_errno, xattr);
if (op_ret == -1) {
- local->op_errno = ENOENT;
+ local->op_errno = op_errno;
gf_log (this->name, GF_LOG_DEBUG,
"lookup of %s on %s returned error (%s)",
local->loc.path, prev->this->name,
@@ -490,6 +500,11 @@ unlock:
dht_layout_set (this, local->inode, layout);
}
+ if (local->loc.parent) {
+ dht_inode_ctx_time_update (local->loc.parent, this,
+ &local->postparent, 1);
+ }
+
DHT_STRIP_PHASE1_FLAGS (&local->stbuf);
DHT_STACK_UNWIND (lookup, frame, local->op_ret, local->op_errno,
local->inode, &local->stbuf, local->xattr,
@@ -521,6 +536,8 @@ dht_revalidate_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
int ret = -1;
int is_dir = 0;
int is_linkfile = 0;
+ call_frame_t *copy = NULL;
+ dht_local_t *copy_local = NULL;
GF_VALIDATE_OR_GOTO ("dht", frame, err);
GF_VALIDATE_OR_GOTO ("dht", this, err);
@@ -579,7 +596,8 @@ dht_revalidate_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
layout = local->layout;
is_dir = check_is_dir (inode, stbuf, xattr);
- is_linkfile = check_is_linkfile (inode, stbuf, xattr);
+ is_linkfile = check_is_linkfile (inode, stbuf, xattr,
+ conf->link_xattr_name);
if (is_linkfile) {
gf_log (this->name, GF_LOG_INFO,
@@ -591,6 +609,23 @@ dht_revalidate_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
}
if (is_dir) {
+ ret = dht_dir_has_layout (xattr, conf->xattr_name);
+ if (ret >= 0) {
+ if (is_greater_time(local->stbuf.ia_ctime,
+ local->stbuf.ia_ctime_nsec,
+ stbuf->ia_ctime,
+ stbuf->ia_ctime_nsec)) {
+ local->prebuf.ia_gid = stbuf->ia_gid;
+ local->prebuf.ia_uid = stbuf->ia_uid;
+ }
+ }
+ if (local->stbuf.ia_type != IA_INVAL)
+ {
+ if ((local->stbuf.ia_gid != stbuf->ia_gid) ||
+ (local->stbuf.ia_uid != stbuf->ia_uid)) {
+ local->need_selfheal = 1;
+ }
+ }
ret = dht_layout_dir_mismatch (this, layout,
prev->this, &local->loc,
xattr);
@@ -629,7 +664,28 @@ out:
&& (conf && conf->unhashed_sticky_bit)) {
local->stbuf.ia_prot.sticky = 1;
}
- if (local->layout_mismatch) {
+ if (local->need_selfheal) {
+ local->need_selfheal = 0;
+ uuid_copy (local->gfid, local->stbuf.ia_gfid);
+ local->stbuf.ia_gid = local->prebuf.ia_gid;
+ local->stbuf.ia_uid = local->prebuf.ia_uid;
+ copy = create_frame (this, this->ctx->pool);
+ if (copy) {
+ copy_local = dht_local_init (copy, &local->loc,
+ NULL, 0);
+ if (!copy_local)
+ goto cont;
+ copy_local->stbuf = local->stbuf;
+ copy->local = copy_local;
+ FRAME_SU_DO (copy, dht_local_t);
+ ret = synctask_new (this->ctx->env,
+ dht_dir_attr_heal,
+ dht_dir_attr_heal_done,
+ copy, copy);
+ }
+ }
+cont:
+ if (local->layout_mismatch) {
/* Found layout mismatch in the directory, need to
fix this in the inode context */
dht_layout_unref (this, local->layout);
@@ -655,7 +711,10 @@ out:
local->op_errno = ESTALE;
}
- WIPE (&local->postparent);
+ if (local->loc.parent) {
+ dht_inode_ctx_time_update (local->loc.parent, this,
+ &local->postparent, 1);
+ }
DHT_STRIP_PHASE1_FLAGS (&local->stbuf);
DHT_STACK_UNWIND (lookup, frame, local->op_ret, local->op_errno,
@@ -673,7 +732,8 @@ dht_lookup_linkfile_create_cbk (call_frame_t *frame, void *cookie,
xlator_t *this,
int32_t op_ret, int32_t op_errno,
inode_t *inode, struct iatt *stbuf,
- struct iatt *preparent, struct iatt *postparent)
+ struct iatt *preparent, struct iatt *postparent,
+ dict_t *xdata)
{
dht_local_t *local = NULL;
xlator_t *cached_subvol = NULL;
@@ -706,8 +766,14 @@ dht_lookup_linkfile_create_cbk (call_frame_t *frame, void *cookie,
local->stbuf.ia_prot.sticky = 1;
}
+ if (local->loc.parent) {
+ dht_inode_ctx_time_update (local->loc.parent, this,
+ postparent, 1);
+ }
+
unwind:
- WIPE (&local->postparent);
+ if (local->linked == _gf_true)
+ dht_linkfile_attr_heal (frame, this);
DHT_STRIP_PHASE1_FLAGS (&local->stbuf);
DHT_STACK_UNWIND (lookup, frame, local->op_ret, local->op_errno,
@@ -780,7 +846,10 @@ dht_lookup_everywhere_done (call_frame_t *frame, xlator_t *this)
"<nil>"));
}
- WIPE (&local->postparent);
+ if (local->loc.parent) {
+ dht_inode_ctx_time_update (local->loc.parent, this,
+ &local->postparent, 1);
+ }
DHT_STRIP_PHASE1_FLAGS (&local->stbuf);
DHT_STACK_UNWIND (lookup, frame, local->op_ret,
@@ -810,7 +879,10 @@ dht_lookup_everywhere_done (call_frame_t *frame, xlator_t *this)
local->op_errno = EINVAL;
}
- WIPE (&local->postparent);
+ if (local->loc.parent) {
+ dht_inode_ctx_time_update (local->loc.parent, this,
+ &local->postparent, 1);
+ }
DHT_STRIP_PHASE1_FLAGS (&local->stbuf);
DHT_STACK_UNWIND (lookup, frame, local->op_ret,
@@ -826,7 +898,7 @@ dht_lookup_everywhere_done (call_frame_t *frame, xlator_t *this)
hashed_subvol->name);
ret = dht_linkfile_create (frame,
- dht_lookup_linkfile_create_cbk,
+ dht_lookup_linkfile_create_cbk, this,
cached_subvol, hashed_subvol, &local->loc);
return ret;
@@ -836,7 +908,8 @@ dht_lookup_everywhere_done (call_frame_t *frame, xlator_t *this)
int
dht_lookup_unlink_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
int op_ret, int op_errno,
- struct iatt *preparent, struct iatt *postparent)
+ struct iatt *preparent, struct iatt *postparent,
+ dict_t *xdata)
{
int this_call_cnt = 0;
@@ -863,8 +936,9 @@ dht_lookup_everywhere_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
xlator_t *subvol = NULL;
loc_t *loc = NULL;
xlator_t *link_subvol = NULL;
- int ret = -1;
- int32_t fd_count = 0;
+ int ret = -1;
+ int32_t fd_count = 0;
+ dht_conf_t *conf = NULL;
GF_VALIDATE_OR_GOTO ("dht", frame, out);
GF_VALIDATE_OR_GOTO ("dht", this, out);
@@ -874,6 +948,7 @@ dht_lookup_everywhere_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
local = frame->local;
loc = &local->loc;
+ conf = this->private;
prev = cookie;
subvol = prev->this;
@@ -895,7 +970,8 @@ dht_lookup_everywhere_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
loc->path, prev->this->name);
}
- is_linkfile = check_is_linkfile (inode, buf, xattr);
+ is_linkfile = check_is_linkfile (inode, buf, xattr,
+ conf->link_xattr_name);
is_dir = check_is_dir (inode, buf, xattr);
if (is_linkfile) {
@@ -955,7 +1031,7 @@ unlock:
"deleting stale linkfile %s on %s",
loc->path, subvol->name);
STACK_WIND (frame, dht_lookup_unlink_cbk,
- subvol, subvol->fops->unlink, loc);
+ subvol, subvol->fops->unlink, loc, 0, NULL);
return 0;
}
}
@@ -1037,7 +1113,16 @@ dht_lookup_linkfile_cbk (call_frame_t *frame, void *cookie,
gf_log (this->name, GF_LOG_INFO,
"lookup of %s on %s (following linkfile) failed (%s)",
local->loc.path, subvol->name, strerror (op_errno));
- goto err;
+
+ /* If cached subvol returned ENOTCONN, do not do
+ lookup_everywhere. We need to make sure linkfile does not get
+ removed, which can take away the namespace, and subvol is
+ anyways down. */
+
+ if (op_errno != ENOTCONN)
+ goto err;
+ else
+ goto unwind;
}
if (check_is_dir (inode, stbuf, xattr)) {
@@ -1047,7 +1132,7 @@ dht_lookup_linkfile_cbk (call_frame_t *frame, void *cookie,
goto err;
}
- if (check_is_linkfile (inode, stbuf, xattr)) {
+ if (check_is_linkfile (inode, stbuf, xattr, conf->link_xattr_name)) {
gf_log (this->name, GF_LOG_INFO,
"lookup of %s on %s (following linkfile) reached link",
local->loc.path, subvol->name);
@@ -1075,9 +1160,12 @@ dht_lookup_linkfile_cbk (call_frame_t *frame, void *cookie,
op_errno = EINVAL;
}
-unwind:
- WIPE (postparent);
+ if (local->loc.parent) {
+ dht_inode_ctx_time_update (local->loc.parent, this,
+ postparent, 1);
+ }
+unwind:
DHT_STRIP_PHASE1_FLAGS (stbuf);
DHT_STACK_UNWIND (lookup, frame, op_ret, op_errno, inode, stbuf, xattr,
postparent);
@@ -1159,7 +1247,6 @@ dht_lookup_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
loc_t *loc = NULL;
call_frame_t *prev = NULL;
int ret = 0;
- uint64_t tmp_layout = 0;
dht_layout_t *parent_layout = NULL;
GF_VALIDATE_OR_GOTO ("dht", frame, err);
@@ -1190,8 +1277,10 @@ dht_lookup_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
}
if ((conf->search_unhashed == GF_DHT_LOOKUP_UNHASHED_AUTO) &&
(loc->parent)) {
- ret = inode_ctx_get (loc->parent, this, &tmp_layout);
- parent_layout = (dht_layout_t *)(long)tmp_layout;
+ ret = dht_inode_ctx_layout_get (loc->parent, this,
+ &parent_layout);
+ if (ret || !parent_layout)
+ goto out;
if (parent_layout->search_unhashed) {
local->op_errno = ENOENT;
dht_lookup_everywhere (frame, this, loc);
@@ -1220,7 +1309,8 @@ dht_lookup_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
goto out;
}
- is_linkfile = check_is_linkfile (inode, stbuf, xattr);
+ is_linkfile = check_is_linkfile (inode, stbuf, xattr,
+ conf->link_xattr_name);
if (!is_linkfile) {
/* non-directory and not a linkfile */
@@ -1260,7 +1350,10 @@ out:
* from each of the subvolume. See dht_iatt_merge for reference.
*/
- WIPE (postparent);
+ if (!op_ret && local->loc.parent) {
+ dht_inode_ctx_time_update (local->loc.parent, this,
+ postparent, 1);
+ }
DHT_STRIP_PHASE1_FLAGS (stbuf);
DHT_STACK_UNWIND (lookup, frame, op_ret, op_errno, inode, stbuf, xattr,
@@ -1269,6 +1362,39 @@ err:
return 0;
}
+/* For directories, check if acl xattrs have been requested (by the acl xlator),
+ * if not, request for them. These xattrs are needed for dht dir self-heal to
+ * perform proper self-healing of dirs
+ */
+void
+dht_check_and_set_acl_xattr_req (inode_t *inode, dict_t *xattr_req)
+{
+ int ret = 0;
+
+ GF_ASSERT (inode);
+ GF_ASSERT (xattr_req);
+
+ if (inode->ia_type != IA_IFDIR)
+ return;
+
+ if (!dict_get (xattr_req, POSIX_ACL_ACCESS_XATTR)) {
+ ret = dict_set_int8 (xattr_req, POSIX_ACL_ACCESS_XATTR, 0);
+ if (ret)
+ gf_log (THIS->name, GF_LOG_WARNING,
+ "failed to set key %s",
+ POSIX_ACL_ACCESS_XATTR);
+ }
+
+ if (!dict_get (xattr_req, POSIX_ACL_DEFAULT_XATTR)) {
+ ret = dict_set_int8 (xattr_req, POSIX_ACL_DEFAULT_XATTR, 0);
+ if (ret)
+ gf_log (THIS->name, GF_LOG_WARNING,
+ "failed to set key %s",
+ POSIX_ACL_DEFAULT_XATTR);
+ }
+
+ return;
+}
int
dht_lookup (call_frame_t *frame, xlator_t *this,
@@ -1289,7 +1415,6 @@ dht_lookup (call_frame_t *frame, xlator_t *this,
VALIDATE_OR_GOTO (this, err);
VALIDATE_OR_GOTO (loc, err);
VALIDATE_OR_GOTO (loc->inode, err);
- VALIDATE_OR_GOTO (loc->path, err);
conf = this->private;
if (!conf)
@@ -1364,7 +1489,7 @@ dht_lookup (call_frame_t *frame, xlator_t *this,
* revalidates directly go to the cached-subvolume.
*/
ret = dict_set_uint32 (local->xattr_req,
- "trusted.glusterfs.dht", 4 * 4);
+ conf->xattr_name, 4 * 4);
if (IA_ISDIR (local->inode->ia_type)) {
local->call_cnt = call_cnt = conf->subvolume_cnt;
@@ -1384,6 +1509,9 @@ dht_lookup (call_frame_t *frame, xlator_t *this,
ret = dict_set_uint32 (local->xattr_req,
GLUSTERFS_OPEN_FD_COUNT, 4);
+ /* need it for dir self-heal */
+ dht_check_and_set_acl_xattr_req (loc->inode, local->xattr_req);
+
for (i = 0; i < call_cnt; i++) {
subvol = layout->list[i].xlator;
@@ -1396,16 +1524,19 @@ dht_lookup (call_frame_t *frame, xlator_t *this,
do_fresh_lookup:
/* TODO: remove the hard-coding */
ret = dict_set_uint32 (local->xattr_req,
- "trusted.glusterfs.dht", 4 * 4);
+ conf->xattr_name, 4 * 4);
ret = dict_set_uint32 (local->xattr_req,
- DHT_LINKFILE_KEY, 256);
+ conf->link_xattr_name, 256);
/* need it for self-healing linkfiles which is
'in-migration' state */
ret = dict_set_uint32 (local->xattr_req,
GLUSTERFS_OPEN_FD_COUNT, 4);
+ /* need it for dir self-heal */
+ dht_check_and_set_acl_xattr_req (loc->inode, local->xattr_req);
+
if (!hashed_subvol) {
gf_log (this->name, GF_LOG_DEBUG,
"no subvolume in layout for path=%s, "
@@ -1439,7 +1570,8 @@ dht_lookup (call_frame_t *frame, xlator_t *this,
err:
op_errno = (op_errno == -1) ? errno : op_errno;
- DHT_STACK_UNWIND (lookup, frame, -1, op_errno, NULL, NULL, NULL, NULL);
+ DHT_STACK_UNWIND (lookup, frame, -1, op_errno, NULL, NULL, NULL,
+ NULL);
return 0;
}
@@ -1447,7 +1579,7 @@ err:
int
dht_unlink_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
int op_ret, int op_errno, struct iatt *preparent,
- struct iatt *postparent)
+ struct iatt *postparent, dict_t *xdata)
{
dht_local_t *local = NULL;
call_frame_t *prev = NULL;
@@ -1471,14 +1603,18 @@ dht_unlink_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
local->postparent = *postparent;
local->preparent = *preparent;
- WIPE (&local->postparent);
- WIPE (&local->preparent);
+ if (local->loc.parent) {
+ dht_inode_ctx_time_update (local->loc.parent, this,
+ &local->preparent, 0);
+ dht_inode_ctx_time_update (local->loc.parent, this,
+ &local->postparent, 1);
+ }
}
unlock:
UNLOCK (&frame->lock);
DHT_STACK_UNWIND (unlink, frame, local->op_ret, local->op_errno,
- &local->preparent, &local->postparent);
+ &local->preparent, &local->postparent, NULL);
return 0;
}
@@ -1487,7 +1623,7 @@ unlock:
int
dht_unlink_linkfile_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
int op_ret, int op_errno, struct iatt *preparent,
- struct iatt *postparent)
+ struct iatt *postparent, dict_t *xdata)
{
dht_local_t *local = NULL;
call_frame_t *prev = NULL;
@@ -1499,7 +1635,8 @@ dht_unlink_linkfile_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
LOCK (&frame->lock);
{
- if (op_ret == -1) {
+ if ((op_ret == -1) && !((op_errno == ENOENT) ||
+ (op_errno == ENOTCONN))) {
local->op_errno = op_errno;
gf_log (this->name, GF_LOG_DEBUG,
"subvolume %s returned -1 (%s)",
@@ -1512,7 +1649,7 @@ dht_unlink_linkfile_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
unlock:
UNLOCK (&frame->lock);
- if (op_ret == -1)
+ if (local->op_ret == -1)
goto err;
cached_subvol = dht_subvol_get_cached (this, local->loc.inode);
@@ -1526,53 +1663,19 @@ unlock:
STACK_WIND (frame, dht_unlink_cbk,
cached_subvol, cached_subvol->fops->unlink,
- &local->loc);
+ &local->loc, local->flags, NULL);
return 0;
err:
DHT_STACK_UNWIND (unlink, frame, -1, local->op_errno,
- NULL, NULL);
+ NULL, NULL, NULL);
return 0;
}
-static int
-dht_ufo_xattr_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
- int op_ret, int op_errno)
-{
- dht_local_t *local = NULL;
- int this_call_cnt = 0;
- call_frame_t *prev = NULL;
-
- local = frame->local;
- prev = cookie;
-
- LOCK (&frame->lock);
- {
- if (op_ret == -1) {
- local->op_ret = -1;
- local->op_errno = op_errno;
- gf_log (this->name, GF_LOG_DEBUG,
- "subvolume %s returned -1 (%s)",
- prev->this->name, strerror (op_errno));
- goto unlock;
- }
- }
-unlock:
- UNLOCK (&frame->lock);
-
- this_call_cnt = dht_frame_return (frame);
- if (is_last_call (this_call_cnt)) {
- DHT_STACK_UNWIND (setxattr, frame, local->op_ret, local->op_errno);
- }
-
- return 0;
-}
-
-
int
dht_err_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
- int op_ret, int op_errno)
+ int op_ret, int op_errno, dict_t *xdata)
{
dht_local_t *local = NULL;
int this_call_cnt = 0;
@@ -1598,7 +1701,8 @@ unlock:
this_call_cnt = dht_frame_return (frame);
if (is_last_call (this_call_cnt)) {
- DHT_STACK_UNWIND (setxattr, frame, local->op_ret, local->op_errno);
+ DHT_STACK_UNWIND (setxattr, frame, local->op_ret,
+ local->op_errno, NULL);
}
return 0;
@@ -1621,114 +1725,249 @@ fill_layout_info (dht_layout_t *layout, char *buf)
}
}
+void
+dht_fill_pathinfo_xattr (xlator_t *this, dht_local_t *local,
+ char *xattr_buf, int32_t alloc_len,
+ int flag, char *layout_buf)
+{
+ if (flag && local->xattr_val)
+ snprintf (xattr_buf, alloc_len,
+ "((<"DHT_PATHINFO_HEADER"%s> %s) (%s-layout %s))",
+ this->name, local->xattr_val, this->name,
+ layout_buf);
+ else if (local->xattr_val)
+ snprintf (xattr_buf, alloc_len,
+ "(<"DHT_PATHINFO_HEADER"%s> %s)",
+ this->name, local->xattr_val);
+ else if (flag)
+ snprintf (xattr_buf, alloc_len, "(%s-layout %s)",
+ this->name, layout_buf);
+}
+
int
-dht_pathinfo_getxattr_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
- int op_ret, int op_errno, dict_t *xattr)
+dht_vgetxattr_alloc_and_fill (dht_local_t *local, dict_t *xattr, xlator_t *this,
+ int op_errno)
+{
+ int ret = -1;
+ char *value = NULL;
+ int32_t plen = 0;
+
+ ret = dict_get_str (xattr, local->xsel, &value);
+ if (ret) {
+ gf_log (this->name, GF_LOG_ERROR,
+ "Subvolume %s returned -1 (%s)", this->name,
+ strerror (op_errno));
+ local->op_ret = -1;
+ local->op_errno = op_errno;
+ goto out;
+ }
+
+ local->alloc_len += strlen(value);
+
+ if (!local->xattr_val) {
+ local->alloc_len += (strlen (DHT_PATHINFO_HEADER) + 10);
+ local->xattr_val = GF_CALLOC (local->alloc_len, sizeof (char),
+ gf_common_mt_char);
+ if (!local->xattr_val) {
+ ret = -1;
+ goto out;
+ }
+ }
+
+ if (local->xattr_val) {
+ plen = strlen (local->xattr_val);
+ if (plen) {
+ /* extra byte(s) for \0 to be safe */
+ local->alloc_len += (plen + 2);
+ local->xattr_val = GF_REALLOC (local->xattr_val,
+ local->alloc_len);
+ if (!local->xattr_val) {
+ ret = -1;
+ goto out;
+ }
+ }
+
+ (void) strcat (local->xattr_val, value);
+ (void) strcat (local->xattr_val, " ");
+ local->op_ret = 0;
+ }
+
+ ret = 0;
+
+ out:
+ return ret;
+}
+
+int
+dht_vgetxattr_fill_and_set (dht_local_t *local, dict_t **dict, xlator_t *this,
+ gf_boolean_t flag)
+{
+ int ret = -1;
+ char *xattr_buf = NULL;
+ char layout_buf[8192] = {0,};
+
+ if (flag)
+ fill_layout_info (local->layout, layout_buf);
+
+ *dict = dict_new ();
+ if (!*dict)
+ goto out;
+
+ local->xattr_val[strlen (local->xattr_val) - 1] = '\0';
+
+ /* we would need max this many bytes to create xattr string
+ * extra 40 bytes is just an estimated amount of additional
+ * space required as we include translator name and some
+ * spaces, brackets etc. when forming the pathinfo string.
+ *
+ * For node-uuid we just don't have all the pretty formatting,
+ * but since this is a generic routine for pathinfo & node-uuid
+ * we dont have conditional space allocation and try to be
+ * generic
+ */
+ local->alloc_len += (2 * strlen (this->name))
+ + strlen (layout_buf)
+ + 40;
+ xattr_buf = GF_CALLOC (local->alloc_len, sizeof (char),
+ gf_common_mt_char);
+ if (!xattr_buf)
+ goto out;
+
+ if (XATTR_IS_PATHINFO (local->xsel)) {
+ (void) dht_fill_pathinfo_xattr (this, local, xattr_buf,
+ local->alloc_len, flag,
+ layout_buf);
+ } else if (XATTR_IS_NODE_UUID (local->xsel)) {
+ (void) snprintf (xattr_buf, local->alloc_len, "%s",
+ local->xattr_val);
+ } else {
+ gf_log (this->name, GF_LOG_WARNING,
+ "Unknown local->xsel (%s)", local->xsel);
+ goto out;
+ }
+
+ ret = dict_set_dynstr (*dict, local->xsel, xattr_buf);
+ GF_FREE (local->xattr_val);
+
+ out:
+ return ret;
+}
+
+int
+dht_vgetxattr_dir_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
+ int op_ret, int op_errno, dict_t *xattr, dict_t *xdata)
{
- dht_local_t *local = NULL;
int ret = 0;
- int flag = 0;
+ dht_local_t *local = NULL;
int this_call_cnt = 0;
- char *value_got = NULL;
- char layout_buf[8192] = {0,};
- char *xattr_buf = NULL;
dict_t *dict = NULL;
- int32_t alloc_len = 0;
- int32_t plen = 0;
- local = frame->local;
+ VALIDATE_OR_GOTO (frame, out);
+ VALIDATE_OR_GOTO (frame->local, out);
- if (op_ret != -1) {
- ret = dict_get_str (xattr, GF_XATTR_PATHINFO_KEY, &value_got);
- if (!ret) {
- alloc_len = strlen (value_got);
+ local = frame->local;
- /**
- * allocate the buffer:- we allocate 10 bytes extra in case we need to
- * append ' Link: ' in the buffer for another STACK_WIND
- */
- if (!local->pathinfo) {
- alloc_len += (strlen (DHT_PATHINFO_HEADER) + 10);
- local->pathinfo = GF_CALLOC (alloc_len, sizeof (char), gf_common_mt_char);
+ LOCK (&frame->lock);
+ {
+ this_call_cnt = --local->call_cnt;
+ if (op_ret < 0) {
+ if (op_errno != ENOTCONN) {
+ gf_log (this->name, GF_LOG_ERROR,
+ "getxattr err (%s) for dir",
+ strerror (op_errno));
+ local->op_ret = -1;
+ local->op_errno = op_errno;
}
- if (local->pathinfo) {
- plen = strlen (local->pathinfo);
- if (plen) {
- /* extra byte(s) for \0 to be safe */
- alloc_len += (plen + 2);
- local->pathinfo = GF_REALLOC (local->pathinfo,
- alloc_len);
- if (!local->pathinfo)
- goto out;
- }
-
- strcat (local->pathinfo, value_got);
- }
+ goto unlock;
}
+
+ ret = dht_vgetxattr_alloc_and_fill (local, xattr, this,
+ op_errno);
+ if (ret)
+ gf_log (this->name, GF_LOG_ERROR,
+ "alloc or fill failure");
}
+ unlock:
+ UNLOCK (&frame->lock);
- out:
- this_call_cnt = dht_frame_return (frame);
- if (is_last_call (this_call_cnt)) {
- if (local->layout->cnt > 1) {
- /* Set it for directory */
- fill_layout_info (local->layout, layout_buf);
- flag = 1;
- }
+ if (!is_last_call (this_call_cnt))
+ goto out;
- dict = dict_new ();
+ /* -- last call: do patch ups -- */
- /* we would need max-to-max this many bytes to create pathinfo string */
- alloc_len += (2 * strlen (this->name)) + strlen (layout_buf) + 40;
- xattr_buf = GF_CALLOC (alloc_len, sizeof (char), gf_common_mt_char);
+ if (local->op_ret == -1) {
+ goto unwind;
+ }
- if (flag && local->pathinfo)
- snprintf (xattr_buf, alloc_len, "((<"DHT_PATHINFO_HEADER"%s> %s) (%s-layout %s))",
- this->name, local->pathinfo, this->name,
- layout_buf);
- else if (local->pathinfo)
- snprintf (xattr_buf, alloc_len, "(<"DHT_PATHINFO_HEADER"%s> %s)",
- this->name, local->pathinfo);
- else if (flag)
- snprintf (xattr_buf, alloc_len, "(%s-layout %s)",
- this->name, layout_buf);
+ ret = dht_vgetxattr_fill_and_set (local, &dict, this, _gf_true);
+ if (ret)
+ goto unwind;
- ret = dict_set_dynstr (dict, GF_XATTR_PATHINFO_KEY,
- xattr_buf);
+ DHT_STACK_UNWIND (getxattr, frame, 0, 0, dict, xdata);
+ goto cleanup;
- if (local->pathinfo)
- GF_FREE (local->pathinfo);
+ unwind:
+ DHT_STACK_UNWIND (getxattr, frame, -1, local->op_errno, NULL, NULL);
+ cleanup:
+ if (dict)
+ dict_unref (dict);
+ out:
+ return 0;
+}
- DHT_STACK_UNWIND (getxattr, frame, op_ret, op_errno, dict);
+int
+dht_vgetxattr_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
+ int op_ret, int op_errno, dict_t *xattr, dict_t *xdata)
+{
+ dht_local_t *local = NULL;
+ int ret = 0;
+ dict_t *dict = NULL;
+ call_frame_t *prev = NULL;
+ gf_boolean_t flag = _gf_true;
- if (dict)
- dict_unref (dict);
+ local = frame->local;
+ prev = cookie;
- return 0;
+ if (op_ret < 0) {
+ local->op_ret = -1;
+ local->op_errno = op_errno;
+ gf_log (this->name, GF_LOG_ERROR, "Subvolume %s returned -1 "
+ "(%s)", prev->this->name, strerror (op_errno));
+ goto unwind;
}
- if (local->pathinfo)
- strcat (local->pathinfo, " Link: ");
- if (local->hashed_subvol) {
- /* This will happen if there pending */
- STACK_WIND (frame, dht_pathinfo_getxattr_cbk, local->hashed_subvol,
- local->hashed_subvol->fops->getxattr,
- &local->loc, local->key);
-
- return 0;
+ ret = dht_vgetxattr_alloc_and_fill (local, xattr, this,
+ op_errno);
+ if (ret) {
+ gf_log (this->name, GF_LOG_ERROR,
+ "alloc or fill failure");
+ goto unwind;
}
- gf_log ("this->name", GF_LOG_ERROR, "Unable to find hashed_subvol for path"
- " %s", local->pathinfo);
+ flag = (local->layout->cnt > 1) ? _gf_true : _gf_false;
+
+ ret = dht_vgetxattr_fill_and_set (local, &dict, this, flag);
+ if (ret)
+ goto unwind;
+
+ DHT_STACK_UNWIND (getxattr, frame, 0, 0, dict, xdata);
+ goto cleanup;
+
+ unwind:
+ DHT_STACK_UNWIND (getxattr, frame, -1, local->op_errno,
+ NULL, NULL);
+ cleanup:
+ if (dict)
+ dict_unref (dict);
- DHT_STACK_UNWIND (getxattr, frame, -1, op_errno, dict);
return 0;
}
int
dht_linkinfo_getxattr_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
- int op_ret, int op_errno, dict_t *xattr)
+ int op_ret, int op_errno, dict_t *xattr,
+ dict_t *xdata)
{
int ret = 0;
char *value = NULL;
@@ -1743,21 +1982,24 @@ dht_linkinfo_getxattr_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
}
}
- DHT_STACK_UNWIND (getxattr, frame, op_ret, op_errno, xattr);
+ DHT_STACK_UNWIND (getxattr, frame, op_ret, op_errno, xattr, xdata);
return 0;
}
int
dht_getxattr_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
- int op_ret, int op_errno, dict_t *xattr)
+ int op_ret, int op_errno, dict_t *xattr, dict_t *xdata)
{
int this_call_cnt = 0;
dht_local_t *local = NULL;
+ dht_conf_t *conf = NULL;
VALIDATE_OR_GOTO (frame, out);
VALIDATE_OR_GOTO (frame->local, out);
+ VALIDATE_OR_GOTO (this->private, out);
+ conf = this->private;
local = frame->local;
this_call_cnt = dht_frame_return (frame);
@@ -1765,8 +2007,8 @@ dht_getxattr_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
if (!xattr || (op_ret == -1))
goto out;
- if (dict_get (xattr, "trusted.glusterfs.dht")) {
- dict_del (xattr, "trusted.glusterfs.dht");
+ if (dict_get (xattr, conf->xattr_name)) {
+ dict_del (xattr, conf->xattr_name);
}
local->op_ret = 0;
@@ -1783,24 +2025,88 @@ dht_getxattr_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
}
out:
if (is_last_call (this_call_cnt)) {
- DHT_STACK_UNWIND (getxattr, frame, local->op_ret, op_errno, local->xattr);
+ DHT_STACK_UNWIND (getxattr, frame, local->op_ret, op_errno,
+ local->xattr, NULL);
}
return 0;
}
int32_t
dht_getxattr_unwind (call_frame_t *frame,
- int op_ret, int op_errno, dict_t *dict)
+ int op_ret, int op_errno, dict_t *dict, dict_t *xdata)
{
- DHT_STACK_UNWIND (getxattr, frame, op_ret, op_errno, dict);
+ DHT_STACK_UNWIND (getxattr, frame, op_ret, op_errno, dict, xdata);
return 0;
}
int
+dht_getxattr_get_real_filename_cbk (call_frame_t *frame, void *cookie,
+ xlator_t *this, int op_ret, int op_errno,
+ dict_t *xattr, dict_t *xdata)
+{
+ int this_call_cnt = 0;
+ dht_local_t *local = NULL;
+
+
+ local = frame->local;
+
+ if (op_ret != -1) {
+ if (local->xattr)
+ dict_unref (local->xattr);
+ local->xattr = dict_ref (xattr);
+
+ if (local->xattr_req)
+ dict_unref (local->xattr_req);
+ local->xattr_req = dict_ref (xdata);
+ }
+
+ this_call_cnt = dht_frame_return (frame);
+ if (is_last_call (this_call_cnt)) {
+ DHT_STACK_UNWIND (getxattr, frame, local->op_ret, op_errno,
+ local->xattr, local->xattr_req);
+ }
+
+ return 0;
+}
+
+
+int
+dht_getxattr_get_real_filename (call_frame_t *frame, xlator_t *this,
+ loc_t *loc, const char *key, dict_t *xdata)
+{
+ dht_local_t *local = NULL;
+ int i = 0;
+ dht_layout_t *layout = NULL;
+ int cnt = 0;
+ xlator_t *subvol = NULL;
+
+
+ local = frame->local;
+ layout = local->layout;
+
+ cnt = local->call_cnt = layout->cnt;
+
+ local->op_ret = -1;
+ local->op_errno = ENODATA;
+
+ for (i = 0; i < cnt; i++) {
+ subvol = layout->list[i].xlator;
+ STACK_WIND (frame, dht_getxattr_get_real_filename_cbk,
+ subvol, subvol->fops->getxattr,
+ loc, key, xdata);
+ }
+
+ return 0;
+}
+
+
+int
dht_getxattr (call_frame_t *frame, xlator_t *this,
- loc_t *loc, const char *key)
+ loc_t *loc, const char *key, dict_t *xdata)
+#define DHT_IS_DIR(layout) (layout->cnt > 1)
{
+
xlator_t *subvol = NULL;
xlator_t *hashed_subvol = NULL;
xlator_t *cached_subvol = NULL;
@@ -1816,7 +2122,6 @@ dht_getxattr (call_frame_t *frame, xlator_t *this,
VALIDATE_OR_GOTO (this, err);
VALIDATE_OR_GOTO (loc, err);
VALIDATE_OR_GOTO (loc->inode, err);
- VALIDATE_OR_GOTO (loc->path, err);
VALIDATE_OR_GOTO (this->private, err);
conf = this->private;
@@ -1844,24 +2149,67 @@ dht_getxattr (call_frame_t *frame, xlator_t *this,
}
}
- if (key && (strcmp (key, GF_XATTR_PATHINFO_KEY) == 0)) {
- hashed_subvol = dht_subvol_get_hashed (this, loc);
+ if (key &&
+ (strncmp (key, GF_XATTR_GET_REAL_FILENAME_KEY,
+ strlen (GF_XATTR_GET_REAL_FILENAME_KEY)) == 0)
+ && DHT_IS_DIR(layout)) {
+ dht_getxattr_get_real_filename (frame, this, loc, key, xdata);
+ return 0;
+ }
+
+ /* for file use cached subvolume (obviously!): see if {}
+ * below
+ * for directory:
+ * wind to all subvolumes and exclude subvolumes which
+ * return ENOTCONN (in callback)
+ *
+ * NOTE: Don't trust inode here, as that may not be valid
+ * (until inode_link() happens)
+ */
+ if (key && DHT_IS_DIR(layout) &&
+ ((strcmp (key, GF_XATTR_PATHINFO_KEY) == 0)
+ || (strcmp (key, GF_XATTR_NODE_UUID_KEY) == 0))) {
+ (void) strncpy (local->xsel, key, 256);
+ cnt = local->call_cnt = layout->cnt;
+ for (i = 0; i < cnt; i++) {
+ subvol = layout->list[i].xlator;
+ STACK_WIND (frame, dht_vgetxattr_dir_cbk,
+ subvol, subvol->fops->getxattr,
+ loc, key, NULL);
+ }
+ return 0;
+ }
+
+ /* node-uuid or pathinfo for files */
+ if (key && ((strcmp (key, GF_XATTR_NODE_UUID_KEY) == 0)
+ || (strcmp (key, GF_XATTR_PATHINFO_KEY) == 0))) {
cached_subvol = local->cached_subvol;
+ (void) strncpy (local->xsel, key, 256);
local->call_cnt = 1;
- if (hashed_subvol != cached_subvol) {
- local->call_cnt = 2;
- local->hashed_subvol = hashed_subvol;
- }
-
- STACK_WIND (frame, dht_pathinfo_getxattr_cbk, cached_subvol,
- cached_subvol->fops->getxattr, loc, key);
+ STACK_WIND (frame, dht_vgetxattr_cbk, cached_subvol,
+ cached_subvol->fops->getxattr, loc, key, NULL);
return 0;
}
+
if (key && (strcmp (key, GF_XATTR_LINKINFO_KEY) == 0)) {
hashed_subvol = dht_subvol_get_hashed (this, loc);
+ if (!hashed_subvol) {
+ gf_log (this->name, GF_LOG_ERROR, "Failed to get"
+ "hashed subvol for %s", loc->path);
+ op_errno = EINVAL;
+ goto err;
+ }
+
cached_subvol = dht_subvol_get_cached (this, loc->inode);
+ if (!cached_subvol) {
+ gf_log (this->name, GF_LOG_ERROR, "Failed to get"
+ "cached subvol for %s", loc->path);
+ op_errno = EINVAL;
+ goto err;
+ }
+
if (hashed_subvol == cached_subvol) {
op_errno = ENODATA;
goto err;
@@ -1869,7 +2217,7 @@ dht_getxattr (call_frame_t *frame, xlator_t *this,
if (hashed_subvol) {
STACK_WIND (frame, dht_linkinfo_getxattr_cbk, hashed_subvol,
hashed_subvol->fops->getxattr, loc,
- GF_XATTR_PATHINFO_KEY);
+ GF_XATTR_PATHINFO_KEY, NULL);
return 0;
}
op_errno = ENODATA;
@@ -1877,13 +2225,13 @@ dht_getxattr (call_frame_t *frame, xlator_t *this,
}
if (key && (!strcmp (GF_XATTR_MARKER_KEY, key))
- && (-1 == frame->root->pid)) {
-
- if (loc->inode-> ia_type == IA_IFDIR) {
+ && (GF_CLIENT_PID_GSYNCD == frame->root->pid)) {
+ if (DHT_IS_DIR(layout)) {
cnt = layout->cnt;
} else {
cnt = 1;
}
+
sub_volumes = alloca ( cnt * sizeof (xlator_t *));
for (i = 0; i < cnt; i++)
*(sub_volumes + i) = layout->list[i].xlator;
@@ -1891,7 +2239,8 @@ dht_getxattr (call_frame_t *frame, xlator_t *this,
if (cluster_getmarkerattr (frame, this, loc, key,
local, dht_getxattr_unwind,
sub_volumes, cnt,
- MARKER_UUID_TYPE, conf->vol_uuid)) {
+ MARKER_UUID_TYPE, marker_uuid_default_gauge,
+ conf->vol_uuid)) {
op_errno = EINVAL;
goto err;
}
@@ -1901,8 +2250,8 @@ dht_getxattr (call_frame_t *frame, xlator_t *this,
if (key && *conf->vol_uuid) {
if ((match_uuid_local (key, conf->vol_uuid) == 0) &&
- (-1 == frame->root->pid)) {
- if (loc->inode-> ia_type == IA_IFDIR) {
+ (GF_CLIENT_PID_GSYNCD == frame->root->pid)) {
+ if (DHT_IS_DIR(layout)) {
cnt = layout->cnt;
} else {
cnt = 1;
@@ -1915,6 +2264,7 @@ dht_getxattr (call_frame_t *frame, xlator_t *this,
local, dht_getxattr_unwind,
sub_volumes, cnt,
MARKER_XTIME_TYPE,
+ marker_xtime_default_gauge,
conf->vol_uuid)) {
op_errno = EINVAL;
goto err;
@@ -1924,7 +2274,7 @@ dht_getxattr (call_frame_t *frame, xlator_t *this,
}
}
- if (loc->inode-> ia_type == IA_IFDIR) {
+ if (DHT_IS_DIR(layout)) {
cnt = local->call_cnt = layout->cnt;
} else {
cnt = local->call_cnt = 1;
@@ -1934,33 +2284,100 @@ dht_getxattr (call_frame_t *frame, xlator_t *this,
subvol = layout->list[i].xlator;
STACK_WIND (frame, dht_getxattr_cbk,
subvol, subvol->fops->getxattr,
- loc, key);
+ loc, key, NULL);
}
return 0;
err:
op_errno = (op_errno == -1) ? errno : op_errno;
- DHT_STACK_UNWIND (getxattr, frame, -1, op_errno, NULL);
+ DHT_STACK_UNWIND (getxattr, frame, -1, op_errno, NULL, NULL);
+
+ return 0;
+}
+#undef DHT_IS_DIR
+
+int
+dht_fgetxattr (call_frame_t *frame, xlator_t *this,
+ fd_t *fd, const char *key, dict_t *xdata)
+{
+ xlator_t *subvol = NULL;
+ dht_local_t *local = NULL;
+ dht_layout_t *layout = NULL;
+ int op_errno = -1;
+ int i = 0;
+ int cnt = 0;
+
+ VALIDATE_OR_GOTO (frame, err);
+ VALIDATE_OR_GOTO (this, err);
+ VALIDATE_OR_GOTO (fd, err);
+ VALIDATE_OR_GOTO (fd->inode, err);
+ VALIDATE_OR_GOTO (this->private, err);
+
+ local = dht_local_init (frame, NULL, fd, GF_FOP_FGETXATTR);
+ if (!local) {
+ op_errno = ENOMEM;
+
+ goto err;
+ }
+
+ layout = local->layout;
+ if (!layout) {
+ gf_log (this->name, GF_LOG_ERROR,
+ "layout is NULL");
+ op_errno = ENOENT;
+ goto err;
+ }
+
+ if (key) {
+ local->key = gf_strdup (key);
+ if (!local->key) {
+ op_errno = ENOMEM;
+ goto err;
+ }
+ }
+
+ if ((fd->inode->ia_type == IA_IFDIR)
+ && (strncmp (key, GF_XATTR_LOCKINFO_KEY,
+ strlen (GF_XATTR_LOCKINFO_KEY) != 0))) {
+ cnt = local->call_cnt = layout->cnt;
+ } else {
+ cnt = local->call_cnt = 1;
+ }
+
+ for (i = 0; i < cnt; i++) {
+ subvol = layout->list[i].xlator;
+ STACK_WIND (frame, dht_getxattr_cbk,
+ subvol, subvol->fops->fgetxattr,
+ fd, key, NULL);
+ }
+ return 0;
+
+err:
+ op_errno = (op_errno == -1) ? errno : op_errno;
+ DHT_STACK_UNWIND (fgetxattr, frame, -1, op_errno, NULL, NULL);
return 0;
}
int
dht_fsetxattr (call_frame_t *frame, xlator_t *this,
- fd_t *fd, dict_t *xattr, int flags)
+ fd_t *fd, dict_t *xattr, int flags, dict_t *xdata)
{
xlator_t *subvol = NULL;
dht_local_t *local = NULL;
int op_errno = EINVAL;
- data_pair_t *trav = NULL;
+ dht_conf_t *conf = NULL;
VALIDATE_OR_GOTO (frame, err);
VALIDATE_OR_GOTO (this, err);
VALIDATE_OR_GOTO (fd, err);
VALIDATE_OR_GOTO (fd->inode, err);
+ VALIDATE_OR_GOTO (this->private, err);
- GF_IF_INTERNAL_XATTR_GOTO ("trusted.glusterfs.dht*", xattr,
- trav, op_errno, err);
+ conf = this->private;
+
+ GF_IF_INTERNAL_XATTR_GOTO (conf->wild_xattr_name, xattr,
+ op_errno, err);
local = dht_local_init (frame, NULL, fd, GF_FOP_FSETXATTR);
if (!local) {
@@ -1979,13 +2396,13 @@ dht_fsetxattr (call_frame_t *frame, xlator_t *this,
local->call_cnt = 1;
STACK_WIND (frame, dht_err_cbk, subvol, subvol->fops->fsetxattr,
- fd, xattr, flags);
+ fd, xattr, flags, NULL);
return 0;
err:
op_errno = (op_errno == -1) ? errno : op_errno;
- DHT_STACK_UNWIND (fsetxattr, frame, -1, op_errno);
+ DHT_STACK_UNWIND (fsetxattr, frame, -1, op_errno, NULL);
return 0;
}
@@ -1993,16 +2410,18 @@ err:
static int
dht_common_setxattr_cbk (call_frame_t *frame, void *cookie,
- xlator_t *this, int32_t op_ret, int32_t op_errno)
+ xlator_t *this, int32_t op_ret, int32_t op_errno,
+ dict_t *xdata)
{
- DHT_STACK_UNWIND (setxattr, frame, op_ret, op_errno);
+ DHT_STACK_UNWIND (setxattr, frame, op_ret, op_errno, xdata);
return 0;
}
int
dht_checking_pathinfo_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
- int op_ret, int op_errno, dict_t *xattr)
+ int op_ret, int op_errno, dict_t *xattr,
+ dict_t *xdata)
{
int i = -1;
int ret = -1;
@@ -2034,7 +2453,7 @@ dht_checking_pathinfo_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
out:
this_call_cnt = dht_frame_return (frame);
if (is_last_call (this_call_cnt)) {
- DHT_STACK_UNWIND (setxattr, frame, local->op_ret, ENOTSUP);
+ DHT_STACK_UNWIND (setxattr, frame, local->op_ret, ENOTSUP, NULL);
}
return 0;
@@ -2042,7 +2461,7 @@ out:
int
dht_setxattr (call_frame_t *frame, xlator_t *this,
- loc_t *loc, dict_t *xattr, int flags)
+ loc_t *loc, dict_t *xattr, int flags, dict_t *xdata)
{
xlator_t *subvol = NULL;
dht_local_t *local = NULL;
@@ -2054,21 +2473,19 @@ dht_setxattr (call_frame_t *frame, xlator_t *this,
data_t *tmp = NULL;
uint32_t dir_spread = 0;
char value[4096] = {0,};
- int forced_rebalance = 0;
+ gf_dht_migrate_data_type_t forced_rebalance = GF_DHT_MIGRATE_DATA;
int call_cnt = 0;
- data_pair_t *trav = NULL;
-
VALIDATE_OR_GOTO (frame, err);
VALIDATE_OR_GOTO (this, err);
VALIDATE_OR_GOTO (loc, err);
VALIDATE_OR_GOTO (loc->inode, err);
- VALIDATE_OR_GOTO (loc->path, err);
-
- GF_IF_INTERNAL_XATTR_GOTO ("trusted.glusterfs.dht*", xattr,
- trav, op_errno, err);
conf = this->private;
+
+ GF_IF_INTERNAL_XATTR_GOTO (conf->wild_xattr_name, xattr,
+ op_errno, err);
+
local = dht_local_init (frame, loc, NULL, GF_FOP_SETXATTR);
if (!local) {
op_errno = ENOMEM;
@@ -2093,25 +2510,6 @@ dht_setxattr (call_frame_t *frame, xlator_t *this,
local->call_cnt = call_cnt = layout->cnt;
- /* This key is sent by Unified File and Object storage
- * to test xattr support in backend.
- */
- tmp = dict_get (xattr, "user.ufo-test");
- if (tmp) {
- if (IA_ISREG (loc->inode->ia_type)) {
- op_errno = ENOTSUP;
- goto err;
- }
- local->op_ret = 0;
- for (i = 0; i < call_cnt; i++) {
- STACK_WIND (frame, dht_ufo_xattr_cbk,
- layout->list[i].xlator,
- layout->list[i].xlator->fops->setxattr,
- loc, xattr, flags);
- }
- return 0;
- }
-
tmp = dict_get (xattr, "distribute.migrate-data");
if (tmp) {
if (IA_ISDIR (loc->inode->ia_type)) {
@@ -2123,9 +2521,20 @@ dht_setxattr (call_frame_t *frame, xlator_t *this,
(ie, 'target' subvolume given there, etc) */
memcpy (value, tmp->data, tmp->len);
if (strcmp (value, "force") == 0)
- forced_rebalance = 1;
+ forced_rebalance =
+ GF_DHT_MIGRATE_DATA_EVEN_IF_LINK_EXISTS;
+
+ if (conf->decommission_in_progress)
+ forced_rebalance = GF_DHT_MIGRATE_HARDLINK;
local->rebalance.target_node = dht_subvol_get_hashed (this, loc);
+ if (!local->rebalance.target_node) {
+ gf_log (this->name, GF_LOG_ERROR, "Failed to get "
+ "hashed subvol for %s", loc->path);
+ op_errno = EINVAL;
+ goto err;
+ }
+
local->rebalance.from_subvol = local->cached_subvol;
if (local->rebalance.target_node == local->rebalance.from_subvol) {
@@ -2165,7 +2574,7 @@ dht_setxattr (call_frame_t *frame, xlator_t *this,
STACK_WIND (frame, dht_checking_pathinfo_cbk,
conf->subvolumes[i],
conf->subvolumes[i]->fops->getxattr,
- loc, GF_XATTR_PATHINFO_KEY);
+ loc, GF_XATTR_PATHINFO_KEY, NULL);
}
return 0;
}
@@ -2175,9 +2584,13 @@ dht_setxattr (call_frame_t *frame, xlator_t *this,
gf_log (this->name, GF_LOG_INFO,
"fixing the layout of %s", loc->path);
- dht_fix_directory_layout (frame, dht_common_setxattr_cbk,
- layout);
- return 0;
+ ret = dht_fix_directory_layout (frame, dht_common_setxattr_cbk,
+ layout);
+ if (ret) {
+ op_errno = ENOTCONN;
+ goto err;
+ }
+ return ret;
}
tmp = dict_get (xattr, "distribute.directory-spread-count");
@@ -2189,10 +2602,14 @@ dht_setxattr (call_frame_t *frame, xlator_t *this,
(dir_spread > 0))) {
layout->spread_cnt = dir_spread;
- dht_fix_directory_layout (frame,
- dht_common_setxattr_cbk,
- layout);
- return 0;
+ ret = dht_fix_directory_layout (frame,
+ dht_common_setxattr_cbk,
+ layout);
+ if (ret) {
+ op_errno = ENOTCONN;
+ goto err;
+ }
+ return ret;
}
gf_log (this->name, GF_LOG_ERROR,
"wrong 'directory-spread-count' value (%s)", value);
@@ -2204,14 +2621,14 @@ dht_setxattr (call_frame_t *frame, xlator_t *this,
STACK_WIND (frame, dht_err_cbk,
layout->list[i].xlator,
layout->list[i].xlator->fops->setxattr,
- loc, xattr, flags);
+ loc, xattr, flags, xdata);
}
return 0;
err:
op_errno = (op_errno == -1) ? errno : op_errno;
- DHT_STACK_UNWIND (setxattr, frame, -1, op_errno);
+ DHT_STACK_UNWIND (setxattr, frame, -1, op_errno, NULL);
return 0;
}
@@ -2219,7 +2636,7 @@ err:
int
dht_removexattr_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
- int op_ret, int op_errno)
+ int op_ret, int op_errno, dict_t *xdata)
{
dht_local_t *local = NULL;
int this_call_cnt = 0;
@@ -2245,7 +2662,8 @@ unlock:
this_call_cnt = dht_frame_return (frame);
if (is_last_call (this_call_cnt)) {
- DHT_STACK_UNWIND (removexattr, frame, local->op_ret, local->op_errno);
+ DHT_STACK_UNWIND (removexattr, frame, local->op_ret,
+ local->op_errno, NULL);
}
return 0;
@@ -2254,21 +2672,27 @@ unlock:
int
dht_removexattr (call_frame_t *frame, xlator_t *this,
- loc_t *loc, const char *key)
+ loc_t *loc, const char *key, dict_t *xdata)
{
xlator_t *subvol = NULL;
int op_errno = -1;
dht_local_t *local = NULL;
dht_layout_t *layout = NULL;
int call_cnt = 0;
+ dht_conf_t *conf = NULL;
int i;
- VALIDATE_OR_GOTO (frame, err);
VALIDATE_OR_GOTO (this, err);
+ VALIDATE_OR_GOTO (this->private, err);
+
+ conf = this->private;
+
+ GF_IF_NATIVE_XATTR_GOTO (conf->wild_xattr_name, key, op_errno, err);
+
+ VALIDATE_OR_GOTO (frame, err);
VALIDATE_OR_GOTO (loc, err);
VALIDATE_OR_GOTO (loc->inode, err);
- VALIDATE_OR_GOTO (loc->path, err);
local = dht_local_init (frame, loc, NULL, GF_FOP_REMOVEXATTR);
if (!local) {
@@ -2299,14 +2723,78 @@ dht_removexattr (call_frame_t *frame, xlator_t *this,
STACK_WIND (frame, dht_removexattr_cbk,
layout->list[i].xlator,
layout->list[i].xlator->fops->removexattr,
- loc, key);
+ loc, key, NULL);
+ }
+
+ return 0;
+
+err:
+ op_errno = (op_errno == -1) ? errno : op_errno;
+ DHT_STACK_UNWIND (removexattr, frame, -1, op_errno, NULL);
+
+ return 0;
+}
+
+int
+dht_fremovexattr (call_frame_t *frame, xlator_t *this,
+ fd_t *fd, const char *key, dict_t *xdata)
+{
+ xlator_t *subvol = NULL;
+ int op_errno = -1;
+ dht_local_t *local = NULL;
+ dht_layout_t *layout = NULL;
+ int call_cnt = 0;
+ dht_conf_t *conf = 0;
+
+ int i;
+
+ VALIDATE_OR_GOTO (this, err);
+ VALIDATE_OR_GOTO (this->private, err);
+
+ conf = this->private;
+
+ GF_IF_NATIVE_XATTR_GOTO (conf->wild_xattr_name, key, op_errno, err);
+
+ VALIDATE_OR_GOTO (frame, err);
+
+ local = dht_local_init (frame, NULL, fd, GF_FOP_FREMOVEXATTR);
+ if (!local) {
+ op_errno = ENOMEM;
+ goto err;
+ }
+
+ subvol = local->cached_subvol;
+ if (!subvol) {
+ gf_log (this->name, GF_LOG_DEBUG,
+ "no cached subvolume for inode=%s",
+ uuid_utoa (fd->inode->gfid));
+ op_errno = EINVAL;
+ goto err;
+ }
+
+ layout = local->layout;
+ if (!local->layout) {
+ gf_log (this->name, GF_LOG_DEBUG,
+ "no layout for inode=%s", uuid_utoa (fd->inode->gfid));
+ op_errno = EINVAL;
+ goto err;
+ }
+
+ local->call_cnt = call_cnt = layout->cnt;
+ local->key = gf_strdup (key);
+
+ for (i = 0; i < call_cnt; i++) {
+ STACK_WIND (frame, dht_removexattr_cbk,
+ layout->list[i].xlator,
+ layout->list[i].xlator->fops->fremovexattr,
+ fd, key, NULL);
}
return 0;
err:
op_errno = (op_errno == -1) ? errno : op_errno;
- DHT_STACK_UNWIND (removexattr, frame, -1, op_errno);
+ DHT_STACK_UNWIND (fremovexattr, frame, -1, op_errno, NULL);
return 0;
}
@@ -2314,7 +2802,7 @@ err:
int
dht_fd_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
- int op_ret, int op_errno, fd_t *fd)
+ int op_ret, int op_errno, fd_t *fd, dict_t *xdata)
{
dht_local_t *local = NULL;
int this_call_cnt = 0;
@@ -2341,7 +2829,7 @@ unlock:
this_call_cnt = dht_frame_return (frame);
if (is_last_call (this_call_cnt))
DHT_STACK_UNWIND (open, frame, local->op_ret, local->op_errno,
- local->fd);
+ local->fd, NULL);
return 0;
}
@@ -2371,7 +2859,7 @@ dht_normalize_stats (struct statvfs *buf, unsigned long bsize,
int
dht_statfs_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
- int op_ret, int op_errno, struct statvfs *statvfs)
+ int op_ret, int op_errno, struct statvfs *statvfs, dict_t *xdata)
{
dht_local_t *local = NULL;
int this_call_cnt = 0;
@@ -2417,14 +2905,14 @@ unlock:
this_call_cnt = dht_frame_return (frame);
if (is_last_call (this_call_cnt))
DHT_STACK_UNWIND (statfs, frame, local->op_ret, local->op_errno,
- &local->statvfs);
+ &local->statvfs, xdata);
return 0;
}
int
-dht_statfs (call_frame_t *frame, xlator_t *this, loc_t *loc)
+dht_statfs (call_frame_t *frame, xlator_t *this, loc_t *loc, dict_t *xdata)
{
xlator_t *subvol = NULL;
dht_local_t *local = NULL;
@@ -2436,7 +2924,6 @@ dht_statfs (call_frame_t *frame, xlator_t *this, loc_t *loc)
VALIDATE_OR_GOTO (this, err);
VALIDATE_OR_GOTO (loc, err);
VALIDATE_OR_GOTO (loc->inode, err);
- VALIDATE_OR_GOTO (loc->path, err);
VALIDATE_OR_GOTO (this->private, err);
conf = this->private;
@@ -2453,7 +2940,8 @@ dht_statfs (call_frame_t *frame, xlator_t *this, loc_t *loc)
for (i = 0; i < conf->subvolume_cnt; i++) {
STACK_WIND (frame, dht_statfs_cbk,
conf->subvolumes[i],
- conf->subvolumes[i]->fops->statfs, loc);
+ conf->subvolumes[i]->fops->statfs, loc,
+ xdata);
}
return 0;
}
@@ -2469,20 +2957,21 @@ dht_statfs (call_frame_t *frame, xlator_t *this, loc_t *loc)
local->call_cnt = 1;
STACK_WIND (frame, dht_statfs_cbk,
- subvol, subvol->fops->statfs, loc);
+ subvol, subvol->fops->statfs, loc, xdata);
return 0;
err:
op_errno = (op_errno == -1) ? errno : op_errno;
- DHT_STACK_UNWIND (statfs, frame, -1, op_errno, NULL);
+ DHT_STACK_UNWIND (statfs, frame, -1, op_errno, NULL, NULL);
return 0;
}
int
-dht_opendir (call_frame_t *frame, xlator_t *this, loc_t *loc, fd_t *fd)
+dht_opendir (call_frame_t *frame, xlator_t *this, loc_t *loc, fd_t *fd,
+ dict_t *xdata)
{
dht_local_t *local = NULL;
dht_conf_t *conf = NULL;
@@ -2509,14 +2998,14 @@ dht_opendir (call_frame_t *frame, xlator_t *this, loc_t *loc, fd_t *fd)
STACK_WIND (frame, dht_fd_cbk,
conf->subvolumes[i],
conf->subvolumes[i]->fops->opendir,
- loc, fd);
+ loc, fd, xdata);
}
return 0;
err:
op_errno = (op_errno == -1) ? errno : op_errno;
- DHT_STACK_UNWIND (opendir, frame, -1, op_errno, NULL);
+ DHT_STACK_UNWIND (opendir, frame, -1, op_errno, NULL, NULL);
return 0;
}
@@ -2524,7 +3013,7 @@ err:
int
dht_readdirp_cbk (call_frame_t *frame, void *cookie, xlator_t *this, int op_ret,
- int op_errno, gf_dirent_t *orig_entries)
+ int op_errno, gf_dirent_t *orig_entries, dict_t *xdata)
{
dht_local_t *local = NULL;
gf_dirent_t entries;
@@ -2537,6 +3026,7 @@ dht_readdirp_cbk (call_frame_t *frame, void *cookie, xlator_t *this, int op_ret,
dht_layout_t *layout = 0;
dht_conf_t *conf = NULL;
xlator_t *subvol = 0;
+ int ret = 0;
INIT_LIST_HEAD (&entries.list);
prev = cookie;
@@ -2553,10 +3043,13 @@ dht_readdirp_cbk (call_frame_t *frame, void *cookie, xlator_t *this, int op_ret,
list_for_each_entry (orig_entry, (&orig_entries->list), list) {
next_offset = orig_entry->d_off;
-
- if (check_is_linkfile_wo_dict (NULL, (&orig_entry->d_stat))
- || (check_is_dir (NULL, (&orig_entry->d_stat), NULL)
- && (prev->this != dht_first_up_subvol (this)))) {
+ if (check_is_dir (NULL, (&orig_entry->d_stat), NULL) &&
+ (prev->this != local->first_up_subvol)) {
+ continue;
+ }
+ if (check_is_linkfile (NULL, (&orig_entry->d_stat),
+ orig_entry->dict,
+ conf->link_xattr_name)) {
continue;
}
@@ -2585,6 +3078,24 @@ dht_readdirp_cbk (call_frame_t *frame, void *cookie, xlator_t *this, int op_ret,
entry->d_type = orig_entry->d_type;
entry->d_len = orig_entry->d_len;
+ if (orig_entry->dict)
+ entry->dict = dict_ref (orig_entry->dict);
+
+ /* making sure we set the inode ctx right with layout,
+ currently possible only for non-directories, so for
+ directories don't set entry inodes */
+ if (!IA_ISDIR(entry->d_stat.ia_type)) {
+ ret = dht_layout_preset (this, prev->this,
+ orig_entry->inode);
+ if (ret)
+ gf_log (this->name, GF_LOG_WARNING,
+ "failed to link the layout in inode");
+ entry->inode = inode_ref (orig_entry->inode);
+ } else if (orig_entry->inode) {
+ dht_inode_ctx_time_update (orig_entry->inode, this,
+ &entry->d_stat, 1);
+ }
+
list_add_tail (&entry->list, &entries.list);
count++;
}
@@ -2614,10 +3125,23 @@ done:
goto unwind;
}
+ if (conf->readdir_optimize == _gf_true) {
+ if (next_subvol != local->first_up_subvol) {
+ ret = dict_set_int32 (local->xattr,
+ GF_READDIR_SKIP_DIRS, 1);
+ if (ret)
+ gf_log (this->name, GF_LOG_ERROR,
+ "dict set failed");
+ } else {
+ dict_del (local->xattr,
+ GF_READDIR_SKIP_DIRS);
+ }
+ }
+
STACK_WIND (frame, dht_readdirp_cbk,
next_subvol, next_subvol->fops->readdirp,
local->fd, local->size, next_offset,
- local->xattr_req);
+ local->xattr);
return 0;
}
@@ -2625,7 +3149,7 @@ unwind:
if (op_ret < 0)
op_ret = 0;
- DHT_STACK_UNWIND (readdirp, frame, op_ret, op_errno, &entries);
+ DHT_STACK_UNWIND (readdirp, frame, op_ret, op_errno, &entries, NULL);
gf_dirent_free (&entries);
@@ -2636,7 +3160,8 @@ unwind:
int
dht_readdir_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
- int op_ret, int op_errno, gf_dirent_t *orig_entries)
+ int op_ret, int op_errno, gf_dirent_t *orig_entries,
+ dict_t *xdata)
{
dht_local_t *local = NULL;
gf_dirent_t entries;
@@ -2713,7 +3238,7 @@ done:
STACK_WIND (frame, dht_readdir_cbk,
next_subvol, next_subvol->fops->readdir,
- local->fd, local->size, next_offset);
+ local->fd, local->size, next_offset, NULL);
return 0;
}
@@ -2721,7 +3246,7 @@ unwind:
if (op_ret < 0)
op_ret = 0;
- DHT_STACK_UNWIND (readdir, frame, op_ret, op_errno, &entries);
+ DHT_STACK_UNWIND (readdir, frame, op_ret, op_errno, &entries, NULL);
gf_dirent_free (&entries);
@@ -2737,11 +3262,15 @@ dht_do_readdir (call_frame_t *frame, xlator_t *this, fd_t *fd, size_t size,
int op_errno = -1;
xlator_t *xvol = NULL;
off_t xoff = 0;
-
+ int ret = 0;
+ dht_conf_t *conf = NULL;
VALIDATE_OR_GOTO (frame, err);
VALIDATE_OR_GOTO (this, err);
VALIDATE_OR_GOTO (fd, err);
+ VALIDATE_OR_GOTO (this->private, err);
+
+ conf = this->private;
local = dht_local_init (frame, NULL, NULL, whichop);
if (!local) {
@@ -2752,22 +3281,51 @@ dht_do_readdir (call_frame_t *frame, xlator_t *this, fd_t *fd, size_t size,
local->fd = fd_ref (fd);
local->size = size;
local->xattr_req = (dict)? dict_ref (dict) : NULL;
+ local->first_up_subvol = dht_first_up_subvol (this);
dht_deitransform (this, yoff, &xvol, (uint64_t *)&xoff);
/* TODO: do proper readdir */
- if (whichop == GF_FOP_READDIR)
- STACK_WIND (frame, dht_readdir_cbk, xvol, xvol->fops->readdir,
- fd, size, xoff);
- else
+ if (whichop == GF_FOP_READDIRP) {
+ if (dict)
+ local->xattr = dict_ref (dict);
+ else
+ local->xattr = dict_new ();
+
+ if (local->xattr) {
+ ret = dict_set_uint32 (local->xattr,
+ conf->link_xattr_name, 256);
+ if (ret)
+ gf_log (this->name, GF_LOG_WARNING,
+ "failed to set '%s' key",
+ conf->link_xattr_name);
+ if (conf->readdir_optimize == _gf_true) {
+ if (xvol != local->first_up_subvol) {
+ ret = dict_set_int32 (local->xattr,
+ GF_READDIR_SKIP_DIRS, 1);
+ if (ret)
+ gf_log (this->name,
+ GF_LOG_ERROR,
+ "Dict set failed");
+ } else {
+ dict_del (local->xattr,
+ GF_READDIR_SKIP_DIRS);
+ }
+ }
+ }
+
STACK_WIND (frame, dht_readdirp_cbk, xvol, xvol->fops->readdirp,
- fd, size, xoff, dict);
+ fd, size, xoff, local->xattr);
+ } else {
+ STACK_WIND (frame, dht_readdir_cbk, xvol, xvol->fops->readdir,
+ fd, size, xoff, local->xattr);
+ }
return 0;
err:
op_errno = (op_errno == -1) ? errno : op_errno;
- DHT_STACK_UNWIND (readdir, frame, -1, op_errno, NULL);
+ DHT_STACK_UNWIND (readdir, frame, -1, op_errno, NULL, NULL);
return 0;
}
@@ -2775,7 +3333,7 @@ err:
int
dht_readdir (call_frame_t *frame, xlator_t *this, fd_t *fd, size_t size,
- off_t yoff)
+ off_t yoff, dict_t *xdata)
{
int op = GF_FOP_READDIR;
dht_conf_t *conf = NULL;
@@ -2812,7 +3370,7 @@ dht_readdirp (call_frame_t *frame, xlator_t *this, fd_t *fd, size_t size,
int
dht_fsyncdir_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
- int op_ret, int op_errno)
+ int op_ret, int op_errno, dict_t *xdata)
{
dht_local_t *local = NULL;
int this_call_cnt = 0;
@@ -2832,14 +3390,16 @@ dht_fsyncdir_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
this_call_cnt = dht_frame_return (frame);
if (is_last_call (this_call_cnt))
- DHT_STACK_UNWIND (fsyncdir, frame, local->op_ret, local->op_errno);
+ DHT_STACK_UNWIND (fsyncdir, frame, local->op_ret,
+ local->op_errno, xdata);
return 0;
}
int
-dht_fsyncdir (call_frame_t *frame, xlator_t *this, fd_t *fd, int datasync)
+dht_fsyncdir (call_frame_t *frame, xlator_t *this, fd_t *fd,
+ int datasync, dict_t *xdata)
{
dht_local_t *local = NULL;
dht_conf_t *conf = NULL;
@@ -2866,14 +3426,14 @@ dht_fsyncdir (call_frame_t *frame, xlator_t *this, fd_t *fd, int datasync)
STACK_WIND (frame, dht_fsyncdir_cbk,
conf->subvolumes[i],
conf->subvolumes[i]->fops->fsyncdir,
- fd, datasync);
+ fd, datasync, xdata);
}
return 0;
err:
op_errno = (op_errno == -1) ? errno : op_errno;
- DHT_STACK_UNWIND (fsyncdir, frame, -1, op_errno);
+ DHT_STACK_UNWIND (fsyncdir, frame, -1, op_errno, NULL);
return 0;
}
@@ -2883,9 +3443,9 @@ int
dht_newfile_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
int op_ret, int op_errno,
inode_t *inode, struct iatt *stbuf, struct iatt *preparent,
- struct iatt *postparent)
+ struct iatt *postparent, dict_t *xdata)
{
- call_frame_t *prev = NULL;
+ xlator_t *prev = NULL;
int ret = -1;
dht_local_t *local = NULL;
@@ -2903,19 +3463,24 @@ dht_newfile_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
prev = cookie;
if (local->loc.parent) {
- WIPE (preparent);
- WIPE (postparent);
+
+ dht_inode_ctx_time_update (local->loc.parent, this,
+ preparent, 0);
+ dht_inode_ctx_time_update (local->loc.parent, this,
+ postparent, 1);
}
- ret = dht_layout_preset (this, prev->this, inode);
+ ret = dht_layout_preset (this, prev, inode);
if (ret < 0) {
gf_log (this->name, GF_LOG_DEBUG,
"could not set pre-set layout for subvolume %s",
- prev->this->name);
+ prev? prev->name: NULL);
op_ret = -1;
op_errno = EINVAL;
goto out;
}
+ if (local->linked == _gf_true)
+ dht_linkfile_attr_heal (frame, this);
out:
/*
* FIXME: ia_size and st_blocks of preparent and postparent do not have
@@ -2924,10 +3489,9 @@ out:
* corresponding values from each of the subvolume.
* See dht_iatt_merge for reference.
*/
-
DHT_STRIP_PHASE1_FLAGS (stbuf);
- DHT_STACK_UNWIND (mknod, frame, op_ret, op_errno, inode, stbuf, preparent,
- postparent);
+ DHT_STACK_UNWIND (mknod, frame, op_ret, op_errno, inode, stbuf,
+ preparent, postparent, xdata);
return 0;
}
@@ -2936,7 +3500,8 @@ dht_mknod_linkfile_create_cbk (call_frame_t *frame, void *cookie,
xlator_t *this,
int32_t op_ret, int32_t op_errno,
inode_t *inode, struct iatt *stbuf,
- struct iatt *preparent, struct iatt *postparent)
+ struct iatt *preparent, struct iatt *postparent,
+ dict_t *xdata)
{
dht_local_t *local = NULL;
xlator_t *cached_subvol = NULL;
@@ -2945,22 +3510,28 @@ dht_mknod_linkfile_create_cbk (call_frame_t *frame, void *cookie,
goto err;
local = frame->local;
+ if (!local || !local->cached_subvol) {
+ op_errno = EINVAL;
+ goto err;
+ }
+
cached_subvol = local->cached_subvol;
- STACK_WIND (frame, dht_newfile_cbk,
- cached_subvol, cached_subvol->fops->mknod,
- &local->loc, local->mode, local->rdev,
- local->params);
+ STACK_WIND_COOKIE (frame, dht_newfile_cbk, (void *)cached_subvol,
+ cached_subvol, cached_subvol->fops->mknod,
+ &local->loc, local->mode, local->rdev, local->umask,
+ local->params);
return 0;
err:
- DHT_STACK_UNWIND (mknod, frame, -1, op_errno, NULL, NULL, NULL, NULL);
+ DHT_STACK_UNWIND (mknod, frame, -1, op_errno, NULL, NULL, NULL, NULL,
+ NULL);
return 0;
}
int
dht_mknod (call_frame_t *frame, xlator_t *this,
- loc_t *loc, mode_t mode, dev_t rdev, dict_t *params)
+ loc_t *loc, mode_t mode, dev_t rdev, mode_t umask, dict_t *params)
{
xlator_t *subvol = NULL;
int op_errno = -1;
@@ -2992,11 +3563,13 @@ dht_mknod (call_frame_t *frame, xlator_t *this,
gf_log (this->name, GF_LOG_TRACE,
"creating %s on %s", loc->path, subvol->name);
- STACK_WIND (frame, dht_newfile_cbk,
- subvol, subvol->fops->mknod,
- loc, mode, rdev, params);
+ STACK_WIND_COOKIE (frame, dht_newfile_cbk, (void *)subvol,
+ subvol, subvol->fops->mknod, loc, mode,
+ rdev, umask, params);
} else {
- avail_subvol = dht_free_disk_available_subvol (this, subvol);
+
+ avail_subvol = dht_free_disk_available_subvol (this, subvol,
+ local);
if (avail_subvol != subvol) {
/* Choose the minimum filled volume, and create the
files there */
@@ -3005,17 +3578,18 @@ dht_mknod (call_frame_t *frame, xlator_t *this,
local->cached_subvol = avail_subvol;
local->mode = mode;
local->rdev = rdev;
-
+ local->umask = umask;
dht_linkfile_create (frame,
dht_mknod_linkfile_create_cbk,
- avail_subvol, subvol, loc);
+ this, avail_subvol, subvol, loc);
} else {
gf_log (this->name, GF_LOG_TRACE,
"creating %s on %s", loc->path, subvol->name);
- STACK_WIND (frame, dht_newfile_cbk,
- subvol, subvol->fops->mknod,
- loc, mode, rdev, params);
+ STACK_WIND_COOKIE (frame, dht_newfile_cbk,
+ (void *)subvol, subvol,
+ subvol->fops->mknod, loc, mode,
+ rdev, umask, params);
}
}
@@ -3024,7 +3598,7 @@ dht_mknod (call_frame_t *frame, xlator_t *this,
err:
op_errno = (op_errno == -1) ? errno : op_errno;
DHT_STACK_UNWIND (mknod, frame, -1, op_errno,
- NULL, NULL, NULL, NULL);
+ NULL, NULL, NULL, NULL, NULL);
return 0;
}
@@ -3032,7 +3606,7 @@ err:
int
dht_symlink (call_frame_t *frame, xlator_t *this,
- const char *linkname, loc_t *loc, dict_t *params)
+ const char *linkname, loc_t *loc, mode_t umask, dict_t *params)
{
xlator_t *subvol = NULL;
int op_errno = -1;
@@ -3060,23 +3634,24 @@ dht_symlink (call_frame_t *frame, xlator_t *this,
gf_log (this->name, GF_LOG_TRACE,
"creating %s on %s", loc->path, subvol->name);
- STACK_WIND (frame, dht_newfile_cbk,
- subvol, subvol->fops->symlink,
- linkname, loc, params);
+ STACK_WIND_COOKIE (frame, dht_newfile_cbk, (void *)subvol, subvol,
+ subvol->fops->symlink, linkname, loc, umask,
+ params);
return 0;
err:
op_errno = (op_errno == -1) ? errno : op_errno;
DHT_STACK_UNWIND (link, frame, -1, op_errno,
- NULL, NULL, NULL, NULL);
+ NULL, NULL, NULL, NULL, NULL);
return 0;
}
int
-dht_unlink (call_frame_t *frame, xlator_t *this, loc_t *loc)
+dht_unlink (call_frame_t *frame, xlator_t *this, loc_t *loc, int xflag,
+ dict_t *xdata)
{
xlator_t *cached_subvol = NULL;
xlator_t *hashed_subvol = NULL;
@@ -3094,7 +3669,7 @@ dht_unlink (call_frame_t *frame, xlator_t *this, loc_t *loc)
local->loc.path, cached_subvol->name, loc->path);
STACK_WIND (frame, dht_unlink_cbk,
cached_subvol, cached_subvol->fops->unlink,
- &local->loc);
+ &local->loc, xflag, xdata);
goto done;
}
@@ -3122,18 +3697,21 @@ dht_unlink (call_frame_t *frame, xlator_t *this, loc_t *loc)
goto err;
}
+ local->flags = xflag;
if (hashed_subvol != cached_subvol) {
STACK_WIND (frame, dht_unlink_linkfile_cbk,
- hashed_subvol, hashed_subvol->fops->unlink, loc);
+ hashed_subvol, hashed_subvol->fops->unlink, loc,
+ xflag, xdata);
} else {
STACK_WIND (frame, dht_unlink_cbk,
- cached_subvol, cached_subvol->fops->unlink, loc);
+ cached_subvol, cached_subvol->fops->unlink, loc,
+ xflag, xdata);
}
done:
return 0;
err:
op_errno = (op_errno == -1) ? errno : op_errno;
- DHT_STACK_UNWIND (unlink, frame, -1, op_errno, NULL, NULL);
+ DHT_STACK_UNWIND (unlink, frame, -1, op_errno, NULL, NULL, NULL);
return 0;
}
@@ -3143,13 +3721,16 @@ int
dht_link_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
int op_ret, int op_errno,
inode_t *inode, struct iatt *stbuf, struct iatt *preparent,
- struct iatt *postparent)
+ struct iatt *postparent, dict_t *xdata)
{
call_frame_t *prev = NULL;
dht_layout_t *layout = NULL;
+ dht_local_t *local = NULL;
prev = cookie;
+ local = frame->local;
+
if (op_ret == -1)
goto out;
@@ -3163,13 +3744,20 @@ dht_link_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
goto out;
}
- WIPE (preparent);
- WIPE (postparent);
-
+ if (local->loc.parent) {
+ dht_inode_ctx_time_update (local->loc.parent, this,
+ preparent, 0);
+ dht_inode_ctx_time_update (local->loc.parent, this,
+ postparent, 1);
+ }
+ if (local->linked == _gf_true) {
+ local->stbuf = *stbuf;
+ dht_linkfile_attr_heal (frame, this);
+ }
out:
DHT_STRIP_PHASE1_FLAGS (stbuf);
DHT_STACK_UNWIND (link, frame, op_ret, op_errno, inode, stbuf, preparent,
- postparent);
+ postparent, NULL);
return 0;
}
@@ -3179,7 +3767,8 @@ int
dht_link_linkfile_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
int op_ret, int op_errno,
inode_t *inode, struct iatt *stbuf,
- struct iatt *preparent, struct iatt *postparent)
+ struct iatt *preparent, struct iatt *postparent,
+ dict_t *xdata)
{
dht_local_t *local = NULL;
xlator_t *srcvol = NULL;
@@ -3191,14 +3780,14 @@ dht_link_linkfile_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
srcvol = local->linkfile.srcvol;
STACK_WIND (frame, dht_link_cbk, srcvol, srcvol->fops->link,
- &local->loc, &local->loc2);
+ &local->loc, &local->loc2, xdata);
return 0;
err:
DHT_STRIP_PHASE1_FLAGS (stbuf);
DHT_STACK_UNWIND (link, frame, op_ret, op_errno, inode, stbuf, preparent,
- postparent);
+ postparent, NULL);
return 0;
}
@@ -3206,7 +3795,7 @@ err:
int
dht_link (call_frame_t *frame, xlator_t *this,
- loc_t *oldloc, loc_t *newloc)
+ loc_t *oldloc, loc_t *newloc, dict_t *xdata)
{
xlator_t *cached_subvol = NULL;
xlator_t *hashed_subvol = NULL;
@@ -3251,19 +3840,19 @@ dht_link (call_frame_t *frame, xlator_t *this,
if (hashed_subvol != cached_subvol) {
uuid_copy (local->gfid, oldloc->inode->gfid);
- dht_linkfile_create (frame, dht_link_linkfile_cbk,
+ dht_linkfile_create (frame, dht_link_linkfile_cbk, this,
cached_subvol, hashed_subvol, newloc);
} else {
STACK_WIND (frame, dht_link_cbk,
cached_subvol, cached_subvol->fops->link,
- oldloc, newloc);
+ oldloc, newloc, xdata);
}
return 0;
err:
op_errno = (op_errno == -1) ? errno : op_errno;
- DHT_STACK_UNWIND (link, frame, -1, op_errno, NULL, NULL, NULL, NULL);
+ DHT_STACK_UNWIND (link, frame, -1, op_errno, NULL, NULL, NULL, NULL, NULL);
return 0;
}
@@ -3273,7 +3862,7 @@ int
dht_create_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
int op_ret, int op_errno,
fd_t *fd, inode_t *inode, struct iatt *stbuf,
- struct iatt *preparent, struct iatt *postparent)
+ struct iatt *preparent, struct iatt *postparent, dict_t *xdata)
{
call_frame_t *prev = NULL;
int ret = -1;
@@ -3292,8 +3881,11 @@ dht_create_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
prev = cookie;
if (local->loc.parent) {
- WIPE (preparent);
- WIPE (postparent);
+ dht_inode_ctx_time_update (local->loc.parent, this,
+ preparent, 0);
+
+ dht_inode_ctx_time_update (local->loc.parent, this,
+ postparent, 1);
}
ret = dht_layout_preset (this, prev->this, inode);
@@ -3305,11 +3897,14 @@ dht_create_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
op_errno = EINVAL;
goto out;
}
-
+ if (local->linked == _gf_true) {
+ local->stbuf = *stbuf;
+ dht_linkfile_attr_heal (frame, this);
+ }
out:
DHT_STRIP_PHASE1_FLAGS (stbuf);
DHT_STACK_UNWIND (create, frame, op_ret, op_errno, fd, inode, stbuf, preparent,
- postparent);
+ postparent, NULL);
return 0;
}
@@ -3319,7 +3914,8 @@ dht_create_linkfile_create_cbk (call_frame_t *frame, void *cookie,
xlator_t *this,
int32_t op_ret, int32_t op_errno,
inode_t *inode, struct iatt *stbuf,
- struct iatt *preparent, struct iatt *postparent)
+ struct iatt *preparent, struct iatt *postparent,
+ dict_t *xdata)
{
dht_local_t *local = NULL;
xlator_t *cached_subvol = NULL;
@@ -3333,18 +3929,19 @@ dht_create_linkfile_create_cbk (call_frame_t *frame, void *cookie,
STACK_WIND (frame, dht_create_cbk,
cached_subvol, cached_subvol->fops->create,
&local->loc, local->flags, local->mode,
- local->fd, local->params);
+ local->umask, local->fd, local->params);
return 0;
err:
- DHT_STACK_UNWIND (create, frame, -1, op_errno, NULL, NULL, NULL, NULL, NULL);
+ DHT_STACK_UNWIND (create, frame, -1, op_errno, NULL, NULL, NULL,
+ NULL, NULL, NULL);
return 0;
}
int
dht_create (call_frame_t *frame, xlator_t *this,
loc_t *loc, int32_t flags, mode_t mode,
- fd_t *fd, dict_t *params)
+ mode_t umask, fd_t *fd, dict_t *params)
{
int op_errno = -1;
xlator_t *subvol = NULL;
@@ -3370,7 +3967,7 @@ dht_create (call_frame_t *frame, xlator_t *this,
local->loc.path, subvol->name, loc->path);
STACK_WIND (frame, dht_create_cbk,
subvol, subvol->fops->create,
- &local->loc, flags, mode, fd, params);
+ &local->loc, flags, mode, umask, fd, params);
goto done;
}
@@ -3388,38 +3985,38 @@ dht_create (call_frame_t *frame, xlator_t *this,
"creating %s on %s", loc->path, subvol->name);
STACK_WIND (frame, dht_create_cbk,
subvol, subvol->fops->create,
- loc, flags, mode, fd, params);
+ loc, flags, mode, umask, fd, params);
goto done;
}
/* Choose the minimum filled volume, and create the
files there */
- avail_subvol = dht_free_disk_available_subvol (this, subvol);
+ avail_subvol = dht_free_disk_available_subvol (this, subvol, local);
if (avail_subvol != subvol) {
local->params = dict_ref (params);
local->flags = flags;
local->mode = mode;
-
+ local->umask = umask;
local->cached_subvol = avail_subvol;
local->hashed_subvol = subvol;
gf_log (this->name, GF_LOG_TRACE,
"creating %s on %s (link at %s)", loc->path,
avail_subvol->name, subvol->name);
- dht_linkfile_create (frame,
- dht_create_linkfile_create_cbk,
- avail_subvol, subvol, loc);
+ dht_linkfile_create (frame, dht_create_linkfile_create_cbk,
+ this, avail_subvol, subvol, loc);
goto done;
}
gf_log (this->name, GF_LOG_TRACE,
"creating %s on %s", loc->path, subvol->name);
STACK_WIND (frame, dht_create_cbk,
subvol, subvol->fops->create,
- loc, flags, mode, fd, params);
+ loc, flags, mode, umask, fd, params);
done:
return 0;
err:
op_errno = (op_errno == -1) ? errno : op_errno;
- DHT_STACK_UNWIND (create, frame, -1, op_errno, NULL, NULL, NULL, NULL, NULL);
+ DHT_STACK_UNWIND (create, frame, -1, op_errno, NULL, NULL, NULL,
+ NULL, NULL, NULL);
return 0;
}
@@ -3428,7 +4025,7 @@ err:
int
dht_mkdir_selfheal_cbk (call_frame_t *frame, void *cookie,
xlator_t *this,
- int32_t op_ret, int32_t op_errno)
+ int32_t op_ret, int32_t op_errno, dict_t *xdata)
{
dht_local_t *local = NULL;
dht_layout_t *layout = NULL;
@@ -3439,14 +4036,17 @@ dht_mkdir_selfheal_cbk (call_frame_t *frame, void *cookie,
if (op_ret == 0) {
dht_layout_set (this, local->inode, layout);
if (local->loc.parent) {
- WIPE (&local->preparent);
- WIPE (&local->postparent);
+ dht_inode_ctx_time_update (local->loc.parent, this,
+ &local->preparent, 0);
+
+ dht_inode_ctx_time_update (local->loc.parent, this,
+ &local->postparent, 1);
}
}
DHT_STACK_UNWIND (mkdir, frame, op_ret, op_errno,
local->inode, &local->stbuf, &local->preparent,
- &local->postparent);
+ &local->postparent, NULL);
return 0;
}
@@ -3454,7 +4054,7 @@ dht_mkdir_selfheal_cbk (call_frame_t *frame, void *cookie,
int
dht_mkdir_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
int op_ret, int op_errno, inode_t *inode, struct iatt *stbuf,
- struct iatt *preparent, struct iatt *postparent)
+ struct iatt *preparent, struct iatt *postparent, dict_t *xdata)
{
dht_local_t *local = NULL;
int this_call_cnt = 0;
@@ -3475,6 +4075,15 @@ dht_mkdir_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
ret = dht_layout_merge (this, layout, prev->this,
-1, ENOSPC, NULL);
} else {
+ if (op_ret == -1 && op_errno == EEXIST)
+ /* Very likely just a race between mkdir and
+ self-heal (from lookup of a concurrent mkdir
+ attempt).
+ Ignore error for now. layout setting will
+ anyways fail if this was a different (old)
+ pre-existing different directory.
+ */
+ op_ret = 0;
ret = dht_layout_merge (this, layout, prev->this,
op_ret, op_errno, NULL);
}
@@ -3507,7 +4116,8 @@ int
dht_mkdir_hashed_cbk (call_frame_t *frame, void *cookie,
xlator_t *this, int op_ret, int op_errno,
inode_t *inode, struct iatt *stbuf,
- struct iatt *preparent, struct iatt *postparent)
+ struct iatt *preparent, struct iatt *postparent,
+ dict_t *xdata)
{
dht_local_t *local = NULL;
int ret = -1;
@@ -3564,19 +4174,20 @@ dht_mkdir_hashed_cbk (call_frame_t *frame, void *cookie,
continue;
STACK_WIND (frame, dht_mkdir_cbk,
conf->subvolumes[i],
- conf->subvolumes[i]->fops->mkdir,
- &local->loc, local->mode, local->params);
+ conf->subvolumes[i]->fops->mkdir, &local->loc,
+ local->mode, local->umask, local->params);
}
return 0;
err:
- DHT_STACK_UNWIND (mkdir, frame, -1, op_errno, NULL, NULL, NULL, NULL);
+ DHT_STACK_UNWIND (mkdir, frame, -1, op_errno, NULL, NULL, NULL,
+ NULL, NULL);
return 0;
}
-int
+ int
dht_mkdir (call_frame_t *frame, xlator_t *this,
- loc_t *loc, mode_t mode, dict_t *params)
+ loc_t *loc, mode_t mode, mode_t umask, dict_t *params)
{
dht_local_t *local = NULL;
dht_conf_t *conf = NULL;
@@ -3612,6 +4223,7 @@ dht_mkdir (call_frame_t *frame, xlator_t *this,
local->hashed_subvol = hashed_subvol;
local->mode = mode;
+ local->umask = umask;
local->params = dict_ref (params);
local->inode = inode_ref (loc->inode);
@@ -3624,13 +4236,14 @@ dht_mkdir (call_frame_t *frame, xlator_t *this,
STACK_WIND (frame, dht_mkdir_hashed_cbk,
hashed_subvol,
hashed_subvol->fops->mkdir,
- loc, mode, params);
+ loc, mode, umask, params);
return 0;
err:
op_errno = (op_errno == -1) ? errno : op_errno;
- DHT_STACK_UNWIND (mkdir, frame, -1, op_errno, NULL, NULL, NULL, NULL);
+ DHT_STACK_UNWIND (mkdir, frame, -1, op_errno, NULL, NULL, NULL,
+ NULL, NULL);
return 0;
}
@@ -3638,14 +4251,87 @@ err:
int
dht_rmdir_selfheal_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
- int op_ret, int op_errno)
+ int op_ret, int op_errno, dict_t *xdata)
{
dht_local_t *local = NULL;
local = frame->local;
DHT_STACK_UNWIND (rmdir, frame, local->op_ret, local->op_errno,
- &local->preparent, &local->postparent);
+ &local->preparent, &local->postparent, NULL);
+
+ return 0;
+}
+
+
+int
+dht_rmdir_hashed_subvol_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
+ int op_ret, int op_errno, struct iatt *preparent,
+ struct iatt *postparent, dict_t *xdata)
+{
+ dht_local_t *local = NULL;
+ int this_call_cnt = 0;
+ call_frame_t *prev = NULL;
+
+ local = frame->local;
+ prev = cookie;
+
+ LOCK (&frame->lock);
+ {
+ if (op_ret == -1) {
+ local->op_errno = op_errno;
+ local->op_ret = -1;
+ if (op_errno != ENOENT && op_errno != EACCES) {
+ local->need_selfheal = 1;
+ }
+
+
+ gf_log (this->name, GF_LOG_DEBUG,
+ "rmdir on %s for %s failed (%s)",
+ prev->this->name, local->loc.path,
+ strerror (op_errno));
+ goto unlock;
+ }
+
+ dht_iatt_merge (this, &local->preparent, preparent, prev->this);
+ dht_iatt_merge (this, &local->postparent, postparent,
+ prev->this);
+
+ }
+unlock:
+ UNLOCK (&frame->lock);
+
+ this_call_cnt = dht_frame_return (frame);
+ if (is_last_call (this_call_cnt)) {
+ if (local->need_selfheal) {
+ local->layout =
+ dht_layout_get (this, local->loc.inode);
+
+ /* TODO: neater interface needed below */
+ local->stbuf.ia_type = local->loc.inode->ia_type;
+
+ uuid_copy (local->gfid, local->loc.inode->gfid);
+ dht_selfheal_restore (frame, dht_rmdir_selfheal_cbk,
+ &local->loc, local->layout);
+ } else {
+
+ if (local->loc.parent) {
+ dht_inode_ctx_time_update (local->loc.parent,
+ this,
+ &local->preparent,
+ 0);
+
+ dht_inode_ctx_time_update (local->loc.parent,
+ this,
+ &local->postparent,
+ 1);
+ }
+
+ DHT_STACK_UNWIND (rmdir, frame, local->op_ret,
+ local->op_errno, &local->preparent,
+ &local->postparent, NULL);
+ }
+ }
return 0;
}
@@ -3654,11 +4340,12 @@ dht_rmdir_selfheal_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
int
dht_rmdir_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
int op_ret, int op_errno, struct iatt *preparent,
- struct iatt *postparent)
+ struct iatt *postparent, dict_t *xdata)
{
dht_local_t *local = NULL;
int this_call_cnt = 0;
call_frame_t *prev = NULL;
+ int done = 0;
local = frame->local;
prev = cookie;
@@ -3691,7 +4378,16 @@ unlock:
this_call_cnt = dht_frame_return (frame);
- if (is_last_call (this_call_cnt)) {
+
+ /* if local->hashed_subvol, we are yet to wind to hashed_subvol. */
+ if (local->hashed_subvol && (this_call_cnt == 1)) {
+ done = 1;
+ } else if (!local->hashed_subvol && !this_call_cnt) {
+ done = 1;
+ }
+
+
+ if (done) {
if (local->need_selfheal && local->fop_succeeded) {
local->layout =
dht_layout_get (this, local->loc.inode);
@@ -3702,15 +4398,34 @@ unlock:
uuid_copy (local->gfid, local->loc.inode->gfid);
dht_selfheal_restore (frame, dht_rmdir_selfheal_cbk,
&local->loc, local->layout);
- } else {
+ } else if (this_call_cnt) {
+ /* If non-hashed subvol's have responded, proceed */
+
+ local->need_selfheal = 0;
+ STACK_WIND (frame, dht_rmdir_hashed_subvol_cbk,
+ local->hashed_subvol,
+ local->hashed_subvol->fops->rmdir,
+ &local->loc, local->flags, NULL);
+ } else if (!this_call_cnt) {
+ /* All subvol's have responded, proceed */
+
if (local->loc.parent) {
- WIPE (&local->preparent);
- WIPE (&local->postparent);
+
+ dht_inode_ctx_time_update (local->loc.parent,
+ this,
+ &local->preparent,
+ 0);
+
+ dht_inode_ctx_time_update (local->loc.parent,
+ this,
+ &local->postparent,
+ 1);
+
}
DHT_STACK_UNWIND (rmdir, frame, local->op_ret,
local->op_errno, &local->preparent,
- &local->postparent);
+ &local->postparent, NULL);
}
}
@@ -3724,6 +4439,7 @@ dht_rmdir_do (call_frame_t *frame, xlator_t *this)
dht_local_t *local = NULL;
dht_conf_t *conf = NULL;
int i = 0;
+ xlator_t *hashed_subvol = NULL;
VALIDATE_OR_GOTO (this->private, err);
@@ -3735,18 +4451,41 @@ dht_rmdir_do (call_frame_t *frame, xlator_t *this)
local->call_cnt = conf->subvolume_cnt;
+ /* first remove from non-hashed_subvol */
+ hashed_subvol = dht_subvol_get_hashed (this, &local->loc);
+
+ if (!hashed_subvol) {
+ gf_log (this->name, GF_LOG_WARNING, "failed to get hashed "
+ "subvol for %s",local->loc.path);
+ } else {
+ local->hashed_subvol = hashed_subvol;
+ }
+
+ /* When DHT has only 1 child */
+ if (conf->subvolume_cnt == 1) {
+ STACK_WIND (frame, dht_rmdir_hashed_subvol_cbk,
+ conf->subvolumes[0],
+ conf->subvolumes[0]->fops->rmdir,
+ &local->loc, local->flags, NULL);
+ return 0;
+ }
+
for (i = 0; i < conf->subvolume_cnt; i++) {
+ if (hashed_subvol &&
+ (hashed_subvol == conf->subvolumes[i]))
+ continue;
+
STACK_WIND (frame, dht_rmdir_cbk,
conf->subvolumes[i],
conf->subvolumes[i]->fops->rmdir,
- &local->loc, local->flags);
+ &local->loc, local->flags, NULL);
}
return 0;
err:
DHT_STACK_UNWIND (rmdir, frame, local->op_ret, local->op_errno,
- &local->preparent, &local->postparent);
+ &local->preparent, &local->postparent, NULL);
return 0;
}
@@ -3754,7 +4493,7 @@ err:
int
dht_rmdir_linkfile_unlink_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
int op_ret, int op_errno, struct iatt *preparent,
- struct iatt *postparent)
+ struct iatt *postparent, dict_t *xdata)
{
dht_local_t *local = NULL;
call_frame_t *prev = NULL;
@@ -3802,6 +4541,7 @@ dht_rmdir_lookup_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
call_frame_t *main_frame = NULL;
dht_local_t *main_local = NULL;
int this_call_cnt = 0;
+ dht_conf_t *conf = this->private;
local = frame->local;
prev = cookie;
@@ -3813,7 +4553,7 @@ dht_rmdir_lookup_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
if (op_ret != 0)
goto err;
- if (check_is_linkfile (inode, stbuf, xattr) == 0) {
+ if (!check_is_linkfile (inode, stbuf, xattr, conf->link_xattr_name)) {
main_local->op_ret = -1;
main_local->op_errno = ENOTEMPTY;
@@ -3824,7 +4564,7 @@ dht_rmdir_lookup_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
}
STACK_WIND (frame, dht_rmdir_linkfile_unlink_cbk,
- src, src->fops->unlink, &local->loc);
+ src, src->fops->unlink, &local->loc, 0, NULL);
return 0;
err:
@@ -3847,6 +4587,8 @@ dht_rmdir_is_subvol_empty (call_frame_t *frame, xlator_t *this,
call_frame_t *lookup_frame = NULL;
dht_local_t *lookup_local = NULL;
dht_local_t *local = NULL;
+ dict_t *xattrs = NULL;
+ dht_conf_t *conf = this->private;
local = frame->local;
@@ -3855,7 +4597,8 @@ dht_rmdir_is_subvol_empty (call_frame_t *frame, xlator_t *this,
continue;
if (strcmp (trav->d_name, "..") == 0)
continue;
- if (check_is_linkfile (NULL, (&trav->d_stat), NULL) == 1) {
+ if (check_is_linkfile (NULL, (&trav->d_stat), trav->dict,
+ conf->link_xattr_name)) {
ret++;
continue;
}
@@ -3867,6 +4610,21 @@ dht_rmdir_is_subvol_empty (call_frame_t *frame, xlator_t *this,
return 0;
}
+ xattrs = dict_new ();
+ if (!xattrs) {
+ gf_log (this->name, GF_LOG_ERROR, "dict_new failed");
+ return -1;
+ }
+
+ ret = dict_set_uint32 (xattrs, conf->link_xattr_name, 256);
+ if (ret) {
+ gf_log (this->name, GF_LOG_ERROR, "failed to set linkto key"
+ " in dict");
+ if (xattrs)
+ dict_unref (xattrs);
+ return -1;
+ }
+
list_for_each_entry (trav, &entries->list, list) {
if (strcmp (trav->d_name, ".") == 0)
continue;
@@ -3883,8 +4641,7 @@ dht_rmdir_is_subvol_empty (call_frame_t *frame, xlator_t *this,
goto err;
}
- lookup_local = GF_CALLOC (sizeof (*local), 1,
- gf_dht_mt_dht_local_t);
+ lookup_local = mem_get0 (this->local_pool);
if (!lookup_local) {
goto err;
}
@@ -3911,12 +4668,18 @@ dht_rmdir_is_subvol_empty (call_frame_t *frame, xlator_t *this,
STACK_WIND (lookup_frame, dht_rmdir_lookup_cbk,
src, src->fops->lookup,
- &lookup_local->loc, NULL);
+ &lookup_local->loc, xattrs);
ret++;
}
+ if (xattrs)
+ dict_unref (xattrs);
+
return ret;
err:
+ if (xattrs)
+ dict_unref (xattrs);
+
DHT_STACK_DESTROY (lookup_frame);
return 0;
}
@@ -3924,7 +4687,8 @@ err:
int
dht_rmdir_readdirp_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
- int op_ret, int op_errno, gf_dirent_t *entries)
+ int op_ret, int op_errno, gf_dirent_t *entries,
+ dict_t *xdata)
{
dht_local_t *local = NULL;
int this_call_cnt = -1;
@@ -3968,12 +4732,14 @@ dht_rmdir_readdirp_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
int
dht_rmdir_opendir_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
- int op_ret, int op_errno, fd_t *fd)
+ int op_ret, int op_errno, fd_t *fd, dict_t *xdata)
{
dht_local_t *local = NULL;
int this_call_cnt = -1;
call_frame_t *prev = NULL;
-
+ dict_t *dict = NULL;
+ int ret = 0;
+ dht_conf_t *conf = this->private;
local = frame->local;
prev = cookie;
@@ -3983,14 +4749,32 @@ dht_rmdir_opendir_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
"opendir on %s for %s failed (%s)",
prev->this->name, local->loc.path,
strerror (op_errno));
+ if (op_errno != ENOENT) {
+ local->op_ret = -1;
+ local->op_errno = op_errno;
+ }
+ goto err;
+ }
+
+ dict = dict_new ();
+ if (!dict) {
local->op_ret = -1;
- local->op_errno = op_errno;
+ local->op_errno = ENOMEM;
goto err;
}
+ ret = dict_set_uint32 (dict, conf->link_xattr_name, 256);
+ if (ret)
+ gf_log (this->name, GF_LOG_WARNING,
+ "%s: failed to set '%s' key",
+ local->loc.path, conf->link_xattr_name);
+
STACK_WIND (frame, dht_rmdir_readdirp_cbk,
prev->this, prev->this->fops->readdirp,
- local->fd, 4096, 0, NULL);
+ local->fd, 4096, 0, dict);
+
+ if (dict)
+ dict_unref (dict);
return 0;
@@ -4006,7 +4790,8 @@ err:
int
-dht_rmdir (call_frame_t *frame, xlator_t *this, loc_t *loc, int flags)
+dht_rmdir (call_frame_t *frame, xlator_t *this, loc_t *loc, int flags,
+ dict_t *xdata)
{
dht_local_t *local = NULL;
dht_conf_t *conf = NULL;
@@ -4045,7 +4830,7 @@ dht_rmdir (call_frame_t *frame, xlator_t *this, loc_t *loc, int flags)
STACK_WIND (frame, dht_rmdir_opendir_cbk,
conf->subvolumes[i],
conf->subvolumes[i]->fops->opendir,
- loc, local->fd);
+ loc, local->fd, NULL);
}
return 0;
@@ -4053,17 +4838,17 @@ dht_rmdir (call_frame_t *frame, xlator_t *this, loc_t *loc, int flags)
err:
op_errno = (op_errno == -1) ? errno : op_errno;
DHT_STACK_UNWIND (rmdir, frame, -1, op_errno,
- NULL, NULL);
+ NULL, NULL, NULL);
return 0;
}
int
dht_entrylk_cbk (call_frame_t *frame, void *cookie,
- xlator_t *this, int32_t op_ret, int32_t op_errno)
+ xlator_t *this, int32_t op_ret, int32_t op_errno, dict_t *xdata)
{
- DHT_STACK_UNWIND (entrylk, frame, op_ret, op_errno);
+ DHT_STACK_UNWIND (entrylk, frame, op_ret, op_errno, xdata);
return 0;
}
@@ -4071,7 +4856,7 @@ dht_entrylk_cbk (call_frame_t *frame, void *cookie,
int
dht_entrylk (call_frame_t *frame, xlator_t *this,
const char *volume, loc_t *loc, const char *basename,
- entrylk_cmd cmd, entrylk_type type)
+ entrylk_cmd cmd, entrylk_type type, dict_t *xdata)
{
xlator_t *subvol = NULL;
int op_errno = -1;
@@ -4081,7 +4866,6 @@ dht_entrylk (call_frame_t *frame, xlator_t *this,
VALIDATE_OR_GOTO (this, err);
VALIDATE_OR_GOTO (loc, err);
VALIDATE_OR_GOTO (loc->inode, err);
- VALIDATE_OR_GOTO (loc->path, err);
local = dht_local_init (frame, loc, NULL, GF_FOP_ENTRYLK);
if (!local) {
@@ -4101,13 +4885,13 @@ dht_entrylk (call_frame_t *frame, xlator_t *this,
STACK_WIND (frame, dht_entrylk_cbk,
subvol, subvol->fops->entrylk,
- volume, loc, basename, cmd, type);
+ volume, loc, basename, cmd, type, xdata);
return 0;
err:
op_errno = (op_errno == -1) ? errno : op_errno;
- DHT_STACK_UNWIND (entrylk, frame, -1, op_errno);
+ DHT_STACK_UNWIND (entrylk, frame, -1, op_errno, NULL);
return 0;
}
@@ -4115,10 +4899,10 @@ err:
int
dht_fentrylk_cbk (call_frame_t *frame, void *cookie,
- xlator_t *this, int32_t op_ret, int32_t op_errno)
+ xlator_t *this, int32_t op_ret, int32_t op_errno, dict_t *xdata)
{
- DHT_STACK_UNWIND (fentrylk, frame, op_ret, op_errno);
+ DHT_STACK_UNWIND (fentrylk, frame, op_ret, op_errno, NULL);
return 0;
}
@@ -4126,7 +4910,7 @@ dht_fentrylk_cbk (call_frame_t *frame, void *cookie,
int
dht_fentrylk (call_frame_t *frame, xlator_t *this,
const char *volume, fd_t *fd, const char *basename,
- entrylk_cmd cmd, entrylk_type type)
+ entrylk_cmd cmd, entrylk_type type, dict_t *xdata)
{
xlator_t *subvol = NULL;
int op_errno = -1;
@@ -4145,13 +4929,13 @@ dht_fentrylk (call_frame_t *frame, xlator_t *this,
STACK_WIND (frame, dht_fentrylk_cbk,
subvol, subvol->fops->fentrylk,
- volume, fd, basename, cmd, type);
+ volume, fd, basename, cmd, type, xdata);
return 0;
err:
op_errno = (op_errno == -1) ? errno : op_errno;
- DHT_STACK_UNWIND (fentrylk, frame, -1, op_errno);
+ DHT_STACK_UNWIND (fentrylk, frame, -1, op_errno, NULL);
return 0;
}
@@ -4160,16 +4944,21 @@ err:
int
dht_forget (xlator_t *this, inode_t *inode)
{
- uint64_t tmp_layout = 0;
+ uint64_t ctx_int = 0;
+ dht_inode_ctx_t *ctx = NULL;
dht_layout_t *layout = NULL;
- inode_ctx_del (inode, this, &tmp_layout);
+ inode_ctx_del (inode, this, &ctx_int);
- if (!tmp_layout)
+ if (!ctx_int)
return 0;
- layout = (dht_layout_t *)(long)tmp_layout;
+ ctx = (dht_inode_ctx_t *) (long) ctx_int;
+
+ layout = ctx->layout;
+ ctx->layout = NULL;
dht_layout_unref (this, layout);
+ GF_FREE (ctx);
return 0;
}
@@ -4178,16 +4967,22 @@ dht_forget (xlator_t *this, inode_t *inode)
int
dht_notify (xlator_t *this, int event, void *data, ...)
{
- xlator_t *subvol = NULL;
- int cnt = -1;
- int i = -1;
- dht_conf_t *conf = NULL;
- int ret = -1;
- int propagate = 0;
+ xlator_t *subvol = NULL;
+ int cnt = -1;
+ int i = -1;
+ dht_conf_t *conf = NULL;
+ int ret = -1;
+ int propagate = 0;
+
+ int had_heard_from_all = 0;
+ int have_heard_from_all = 0;
+ struct timeval time = {0,};
+ gf_defrag_info_t *defrag = NULL;
+ dict_t *dict = NULL;
+ gf_defrag_type cmd = 0;
+ dict_t *output = NULL;
+ va_list ap;
- int had_heard_from_all = 0;
- int have_heard_from_all = 0;
- struct timeval time = {0,};
conf = this->private;
if (!conf)
@@ -4249,7 +5044,11 @@ dht_notify (xlator_t *this, int event, void *data, ...)
if (conf->assert_no_child_down) {
gf_log (this->name, GF_LOG_WARNING,
"Received CHILD_DOWN. Exiting");
- exit(0);
+ if (conf->defrag) {
+ gf_defrag_stop (conf->defrag, NULL);
+ } else {
+ kill (getpid(), SIGTERM);
+ }
}
for (i = 0; i < conf->subvolume_cnt; i++) {
@@ -4300,6 +5099,36 @@ dht_notify (xlator_t *this, int event, void *data, ...)
UNLOCK (&conf->subvolume_lock);
break;
+ case GF_EVENT_VOLUME_DEFRAG:
+ {
+ if (!conf->defrag) {
+ return ret;
+ }
+ defrag = conf->defrag;
+
+ dict = data;
+ va_start (ap, data);
+ output = va_arg (ap, dict_t*);
+
+ ret = dict_get_int32 (dict, "rebalance-command",
+ (int32_t*)&cmd);
+ if (ret)
+ return ret;
+ LOCK (&defrag->lock);
+ {
+ if (defrag->is_exiting)
+ goto unlock;
+ if (cmd == GF_DEFRAG_CMD_STATUS)
+ gf_defrag_status_get (defrag, output);
+ else if (cmd == GF_DEFRAG_CMD_STOP)
+ gf_defrag_stop (defrag, output);
+ }
+unlock:
+ UNLOCK (&defrag->lock);
+ return 0;
+ break;
+ }
+
default:
propagate = 1;
break;
@@ -4315,9 +5144,12 @@ dht_notify (xlator_t *this, int event, void *data, ...)
/* if all subvols have reported status, no need to hide anything
or wait for anything else. Just propagate blindly */
- if (have_heard_from_all)
+ if (have_heard_from_all) {
propagate = 1;
+ }
+
+
if (!had_heard_from_all && have_heard_from_all) {
/* This is the first event which completes aggregation
of events from all subvolumes. If at least one subvol
@@ -4336,6 +5168,19 @@ dht_notify (xlator_t *this, int event, void *data, ...)
/* continue to check other events for CHILD_UP */
}
}
+
+ /* rebalance is started with assert_no_child_down. So we do
+ * not need to handle CHILD_DOWN event here.
+ */
+ if (conf->defrag) {
+ ret = gf_thread_create (&conf->defrag->th, NULL,
+ gf_defrag_start, this);
+ if (ret) {
+ conf->defrag = NULL;
+ GF_FREE (conf->defrag);
+ kill (getpid(), SIGTERM);
+ }
+ }
}
ret = 0;
@@ -4344,3 +5189,24 @@ dht_notify (xlator_t *this, int event, void *data, ...)
return ret;
}
+
+int
+dht_inode_ctx_layout_get (inode_t *inode, xlator_t *this, dht_layout_t **layout)
+{
+ dht_inode_ctx_t *ctx = NULL;
+ int ret = -1;
+
+ ret = dht_inode_ctx_get (inode, this, &ctx);
+
+ if (!ret && ctx) {
+ if (ctx->layout) {
+ if (layout)
+ *layout = ctx->layout;
+ ret = 0;
+ } else {
+ ret = -1;
+ }
+ }
+
+ return ret;
+}
diff --git a/xlators/cluster/dht/src/dht-common.h b/xlators/cluster/dht/src/dht-common.h
index 8a0765b63..5ccd66799 100644
--- a/xlators/cluster/dht/src/dht-common.h
+++ b/xlators/cluster/dht/src/dht-common.h
@@ -1,20 +1,11 @@
/*
- Copyright (c) 2008-2011 Gluster, Inc. <http://www.gluster.com>
+ Copyright (c) 2008-2012 Red Hat, Inc. <http://www.redhat.com>
This file is part of GlusterFS.
- GlusterFS is free software; you can redistribute it and/or modify
- it under the terms of the GNU General Public License as published
- by the Free Software Foundation; either version 3 of the License,
- or (at your option) any later version.
-
- GlusterFS is distributed in the hope that it will be useful, but
- WITHOUT ANY WARRANTY; without even the implied warranty of
- MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU
- General Public License for more details.
-
- You should have received a copy of the GNU General Public License
- along with this program. If not, see
- <http://www.gnu.org/licenses/>.
+ This file is licensed to you under your choice of the GNU Lesser
+ General Public License, version 3 or any later version (LGPLv3 or
+ later), or the GNU General Public License, version 2 (GPLv2), in all
+ cases as published by the Free Software Foundation.
*/
#ifndef _CONFIG_H
@@ -22,6 +13,8 @@
#include "config.h"
#endif
+#include <regex.h>
+
#include "dht-mem-types.h"
#include "libxlator.h"
#include "syncop.h"
@@ -38,7 +31,8 @@
typedef int (*dht_selfheal_dir_cbk_t) (call_frame_t *frame, void *cookie,
xlator_t *this,
- int32_t op_ret, int32_t op_errno);
+ int32_t op_ret, int32_t op_errno,
+ dict_t *xdata);
typedef int (*dht_defrag_cbk_fn_t) (xlator_t *this, call_frame_t *frame,
int ret);
@@ -61,20 +55,38 @@ struct dht_layout {
uint32_t start;
uint32_t stop;
xlator_t *xlator;
- } list[0];
+ } list[];
};
typedef struct dht_layout dht_layout_t;
+struct dht_stat_time {
+ uint32_t atime;
+ uint32_t atime_nsec;
+ uint32_t ctime;
+ uint32_t ctime_nsec;
+ uint32_t mtime;
+ uint32_t mtime_nsec;
+};
+
+typedef struct dht_stat_time dht_stat_time_t;
+
+struct dht_inode_ctx {
+ dht_layout_t *layout;
+ dht_stat_time_t time;
+};
+
+typedef struct dht_inode_ctx dht_inode_ctx_t;
+
typedef enum {
DHT_HASH_TYPE_DM,
+ DHT_HASH_TYPE_DM_USER,
} dht_hashfn_type_t;
/* rebalance related */
struct dht_rebalance_ {
xlator_t *from_subvol;
xlator_t *target_node;
- int32_t wbflags;
off_t offset;
size_t size;
int32_t flags;
@@ -83,6 +95,7 @@ struct dht_rebalance_ {
struct iovec *vector;
struct iatt stbuf;
dht_defrag_cbk_fn_t target_op_fn;
+ dict_t *xdata;
};
struct dht_local {
@@ -129,7 +142,6 @@ struct dht_local {
struct {
uint32_t hole_cnt;
uint32_t overlaps_cnt;
- uint32_t missing;
uint32_t down;
uint32_t misc;
dht_selfheal_dir_cbk_t dir_cbk;
@@ -142,11 +154,16 @@ struct dht_local {
int32_t flags;
mode_t mode;
dev_t rdev;
+ mode_t umask;
/* need for file-info */
- char *pathinfo;
+ char *xattr_val;
char *key;
+ /* which xattr request? */
+ char xsel[256];
+ int32_t alloc_len;
+
char *newpath;
/* gfid related */
@@ -162,7 +179,12 @@ struct dht_local {
glusterfs_fop_t fop;
+ gf_boolean_t linked;
+ xlator_t *link_subvol;
+
struct dht_rebalance_ rebalance;
+ xlator_t *first_up_subvol;
+
};
typedef struct dht_local dht_local_t;
@@ -175,6 +197,59 @@ struct dht_du {
};
typedef struct dht_du dht_du_t;
+enum gf_defrag_type {
+ GF_DEFRAG_CMD_START = 1,
+ GF_DEFRAG_CMD_STOP = 1 + 1,
+ GF_DEFRAG_CMD_STATUS = 1 + 2,
+ GF_DEFRAG_CMD_START_LAYOUT_FIX = 1 + 3,
+ GF_DEFRAG_CMD_START_FORCE = 1 + 4,
+};
+typedef enum gf_defrag_type gf_defrag_type;
+
+enum gf_defrag_status_t {
+ GF_DEFRAG_STATUS_NOT_STARTED,
+ GF_DEFRAG_STATUS_STARTED,
+ GF_DEFRAG_STATUS_STOPPED,
+ GF_DEFRAG_STATUS_COMPLETE,
+ GF_DEFRAG_STATUS_FAILED,
+ GF_DEFRAG_STATUS_LAYOUT_FIX_STARTED,
+ GF_DEFRAG_STATUS_LAYOUT_FIX_STOPPED,
+ GF_DEFRAG_STATUS_LAYOUT_FIX_COMPLETE,
+ GF_DEFRAG_STATUS_LAYOUT_FIX_FAILED,
+};
+typedef enum gf_defrag_status_t gf_defrag_status_t;
+
+typedef struct gf_defrag_pattern_list gf_defrag_pattern_list_t;
+
+struct gf_defrag_pattern_list {
+ char path_pattern[256];
+ uint64_t size;
+ gf_defrag_pattern_list_t *next;
+};
+
+struct gf_defrag_info_ {
+ uint64_t total_files;
+ uint64_t total_data;
+ uint64_t num_files_lookedup;
+ uint64_t total_failures;
+ uint64_t skipped;
+ gf_lock_t lock;
+ int cmd;
+ pthread_t th;
+ gf_defrag_status_t defrag_status;
+ struct rpc_clnt *rpc;
+ uint32_t connected;
+ uint32_t is_exiting;
+ pid_t pid;
+ inode_t *root_inode;
+ uuid_t node_uuid;
+ struct timeval start_time;
+ gf_boolean_t stats;
+ gf_defrag_pattern_list_t *defrag_pattern;
+};
+
+typedef struct gf_defrag_info_ gf_defrag_info_t;
+
struct dht_conf {
gf_lock_t subvolume_lock;
int subvolume_cnt;
@@ -186,8 +261,8 @@ struct dht_conf {
gf_boolean_t search_unhashed;
int gen;
dht_du_t *du_stats;
- uint64_t min_free_disk;
- uint32_t min_free_inodes;
+ double min_free_disk;
+ double min_free_inodes;
char disk_unit;
int32_t refresh_interval;
gf_boolean_t unhashed_sticky_bit;
@@ -204,10 +279,28 @@ struct dht_conf {
/* Will be a global flag to control the layout spread count */
uint32_t dir_spread_cnt;
- struct syncenv *env; /* The env pointer to the rebalance synctask */
-
/* to keep track of nodes which are decomissioned */
xlator_t **decommissioned_bricks;
+ int decommission_in_progress;
+ int decommission_subvols_cnt;
+
+ /* defrag related */
+ gf_defrag_info_t *defrag;
+
+ /* Request to filter directory entries in readdir request */
+
+ gf_boolean_t readdir_optimize;
+
+ /* Support regex-based name reinterpretation. */
+ regex_t rsync_regex;
+ gf_boolean_t rsync_regex_valid;
+ regex_t extra_regex;
+ gf_boolean_t extra_regex_valid;
+
+ /* Support variable xattr names. */
+ char *xattr_name;
+ char *link_xattr_name;
+ char *wild_xattr_name;
};
typedef struct dht_conf dht_conf_t;
@@ -222,29 +315,28 @@ struct dht_disk_layout {
};
typedef struct dht_disk_layout dht_disk_layout_t;
+typedef enum {
+ GF_DHT_MIGRATE_DATA,
+ GF_DHT_MIGRATE_DATA_EVEN_IF_LINK_EXISTS,
+ GF_DHT_MIGRATE_HARDLINK,
+ GF_DHT_MIGRATE_HARDLINK_IN_PROGRESS
+} gf_dht_migrate_data_type_t;
#define ENTRY_MISSING(op_ret, op_errno) (op_ret == -1 && op_errno == ENOENT)
-#define is_fs_root(loc) (strcmp (loc->path, "/") == 0)
-
-#define is_revalidate(loc) (inode_ctx_get (loc->inode, this, NULL) == 0)
+#define is_revalidate(loc) (dht_inode_ctx_layout_get (loc->inode, this, NULL) == 0)
#define is_last_call(cnt) (cnt == 0)
#define DHT_MIGRATION_IN_PROGRESS 1
#define DHT_MIGRATION_COMPLETED 2
-#define DHT_LINKFILE_KEY "trusted.glusterfs.dht.linkto"
#define DHT_LINKFILE_MODE (S_ISVTX)
-#define check_is_linkfile(i,s,x) ( \
+#define check_is_linkfile(i,s,x,n) ( \
((st_mode_from_ia ((s)->ia_prot, (s)->ia_type) & ~S_IFMT) \
- == DHT_LINKFILE_MODE) && \
- dict_get (x, DHT_LINKFILE_KEY))
-
-#define check_is_linkfile_wo_dict(i,s) ( \
- ((st_mode_from_ia ((s)->ia_prot, (s)->ia_type) & ~S_IFMT) \
- == DHT_LINKFILE_MODE))
+ == DHT_LINKFILE_MODE) && \
+ dict_get (x, n))
#define IS_DHT_MIGRATION_PHASE2(buf) ( \
IA_ISREG ((buf)->ia_type) && \
@@ -289,6 +381,25 @@ typedef struct dht_disk_layout dht_disk_layout_t;
dht_local_wipe (__xl, __local); \
} while (0)
+#define DHT_UPDATE_TIME(ctx_sec, ctx_nsec, new_sec, new_nsec, inode, post) do {\
+ int32_t sec = 0; \
+ sec = new_sec; \
+ LOCK (&inode->lock); \
+ { \
+ new_sec = max(new_sec, ctx_sec); \
+ if (sec < new_sec) \
+ new_nsec = ctx_nsec; \
+ if (sec == new_sec) \
+ new_nsec = max (new_nsec, ctx_nsec); \
+ if (post) { \
+ ctx_sec = new_sec; \
+ ctx_nsec = new_nsec; \
+ } \
+ } \
+ UNLOCK (&inode->lock); \
+ } while (0)
+
+#define is_greater_time(a, an, b, bn) (((a) < (b)) || (((a) == (b)) && ((an) < (bn))))
dht_layout_t *dht_layout_new (xlator_t *this, int cnt);
dht_layout_t *dht_layout_get (xlator_t *this, inode_t *inode);
dht_layout_t *dht_layout_for_subvol (xlator_t *this, xlator_t *subvol);
@@ -298,7 +409,7 @@ int dht_layout_normalize (xlator_t *this, l
int dht_layout_anomalies (xlator_t *this, loc_t *loc, dht_layout_t *layout,
uint32_t *holes_p, uint32_t *overlaps_p,
uint32_t *missing_p, uint32_t *down_p,
- uint32_t *misc_p);
+ uint32_t *misc_p, uint32_t *no_space_p);
int dht_layout_dir_mismatch (xlator_t *this, dht_layout_t *layout,
xlator_t *subvol, loc_t *loc, dict_t *xattr);
@@ -332,12 +443,14 @@ int dht_iatt_merge (xlator_t *this, struct iatt *to, struct iatt
xlator_t *dht_subvol_get_hashed (xlator_t *this, loc_t *loc);
xlator_t *dht_subvol_get_cached (xlator_t *this, inode_t *inode);
xlator_t *dht_subvol_next (xlator_t *this, xlator_t *prev);
+xlator_t *dht_subvol_next_available (xlator_t *this, xlator_t *prev);
int dht_subvol_cnt (xlator_t *this, xlator_t *subvol);
-int dht_hash_compute (int type, const char *name, uint32_t *hash_p);
+int dht_hash_compute (xlator_t *this, int type, const char *name, uint32_t *hash_p);
int dht_linkfile_create (call_frame_t *frame, fop_mknod_cbk_t linkfile_cbk,
- xlator_t *tovol, xlator_t *fromvol, loc_t *loc);
+ xlator_t *this, xlator_t *tovol,
+ xlator_t *fromvol, loc_t *loc);
int dht_lookup_directory (call_frame_t *frame, xlator_t *this, loc_t *loc);
int dht_lookup_everywhere (call_frame_t *frame, xlator_t *this, loc_t *loc);
int
@@ -352,17 +465,15 @@ dht_selfheal_restore (call_frame_t *frame, dht_selfheal_dir_cbk_t cbk,
int
dht_layout_sort_volname (dht_layout_t *layout);
-int dht_rename (call_frame_t *frame, xlator_t *this,
- loc_t *oldloc, loc_t *newloc);
-
int dht_get_du_info (call_frame_t *frame, xlator_t *this, loc_t *loc);
gf_boolean_t dht_is_subvol_filled (xlator_t *this, xlator_t *subvol);
-xlator_t *dht_free_disk_available_subvol (xlator_t *this, xlator_t *subvol);
+xlator_t *dht_free_disk_available_subvol (xlator_t *this, xlator_t *subvol,
+ dht_local_t *layout);
int dht_get_du_info_for_subvol (xlator_t *this, int subvol_idx);
int dht_layout_preset (xlator_t *this, xlator_t *subvol, inode_t *inode);
-int dht_layout_set (xlator_t *this, inode_t *inode, dht_layout_t *layout);
+int dht_layout_set (xlator_t *this, inode_t *inode, dht_layout_t *layout);;
void dht_layout_unref (xlator_t *this, dht_layout_t *layout);
dht_layout_t *dht_layout_ref (xlator_t *this, dht_layout_t *layout);
xlator_t *dht_first_up_subvol (xlator_t *this);
@@ -377,7 +488,8 @@ int dht_rename_cleanup (call_frame_t *frame)
int dht_rename_links_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
int32_t op_ret, int32_t op_errno,
inode_t *inode, struct iatt *stbuf,
- struct iatt *preparent, struct iatt *postparent);
+ struct iatt *preparent, struct iatt *postparent,
+ dict_t *xdata);
int dht_fix_directory_layout (call_frame_t *frame,
dht_selfheal_dir_cbk_t dir_cbk,
@@ -400,73 +512,73 @@ int32_t dht_lookup (call_frame_t *frame,
int32_t dht_stat (call_frame_t *frame,
xlator_t *this,
- loc_t *loc);
+ loc_t *loc, dict_t *xdata);
int32_t dht_fstat (call_frame_t *frame,
xlator_t *this,
- fd_t *fd);
+ fd_t *fd, dict_t *xdata);
int32_t dht_truncate (call_frame_t *frame,
xlator_t *this,
loc_t *loc,
- off_t offset);
+ off_t offset, dict_t *xdata);
int32_t dht_ftruncate (call_frame_t *frame,
xlator_t *this,
fd_t *fd,
- off_t offset);
+ off_t offset, dict_t *xdata);
int32_t dht_access (call_frame_t *frame,
xlator_t *this,
loc_t *loc,
- int32_t mask);
+ int32_t mask, dict_t *xdata);
int32_t dht_readlink (call_frame_t *frame,
xlator_t *this,
loc_t *loc,
- size_t size);
+ size_t size, dict_t *xdata);
-int32_t dht_mknod (call_frame_t *frame, xlator_t *this,
- loc_t *loc, mode_t mode, dev_t rdev, dict_t *params);
+int32_t dht_mknod (call_frame_t *frame, xlator_t *this, loc_t *loc,
+ mode_t mode, dev_t rdev, mode_t umask, dict_t *xdata);
int32_t dht_mkdir (call_frame_t *frame, xlator_t *this,
- loc_t *loc, mode_t mode, dict_t *params);
+ loc_t *loc, mode_t mode, mode_t umask, dict_t *xdata);
int32_t dht_unlink (call_frame_t *frame,
xlator_t *this,
- loc_t *loc);
+ loc_t *loc, int xflag, dict_t *xdata);
int32_t dht_rmdir (call_frame_t *frame, xlator_t *this,
- loc_t *loc, int flags);
+ loc_t *loc, int flags, dict_t *xdata);
int32_t dht_symlink (call_frame_t *frame, xlator_t *this,
- const char *linkpath, loc_t *loc, dict_t *params);
+ const char *linkpath, loc_t *loc, mode_t umask,
+ dict_t *xdata);
int32_t dht_rename (call_frame_t *frame,
xlator_t *this,
loc_t *oldloc,
- loc_t *newloc);
+ loc_t *newloc, dict_t *xdata);
int32_t dht_link (call_frame_t *frame,
xlator_t *this,
loc_t *oldloc,
- loc_t *newloc);
+ loc_t *newloc, dict_t *xdata);
int32_t dht_create (call_frame_t *frame, xlator_t *this,
loc_t *loc, int32_t flags, mode_t mode,
- fd_t *fd, dict_t *params);
+ mode_t umask, fd_t *fd, dict_t *params);
int32_t dht_open (call_frame_t *frame,
xlator_t *this,
loc_t *loc,
- int32_t flags, fd_t *fd,
- int32_t wbflags);
+ int32_t flags, fd_t *fd, dict_t *xdata);
int32_t dht_readv (call_frame_t *frame,
xlator_t *this,
fd_t *fd,
size_t size,
- off_t offset);
+ off_t offset, uint32_t flags, dict_t *xdata);
int32_t dht_writev (call_frame_t *frame,
xlator_t *this,
@@ -474,83 +586,88 @@ int32_t dht_writev (call_frame_t *frame,
struct iovec *vector,
int32_t count,
off_t offset,
- struct iobref *iobref);
+ uint32_t flags,
+ struct iobref *iobref, dict_t *xdata);
int32_t dht_flush (call_frame_t *frame,
xlator_t *this,
- fd_t *fd);
+ fd_t *fd, dict_t *xdata);
int32_t dht_fsync (call_frame_t *frame,
xlator_t *this,
fd_t *fd,
- int32_t datasync);
+ int32_t datasync, dict_t *xdata);
int32_t dht_opendir (call_frame_t *frame,
xlator_t *this,
- loc_t *loc, fd_t *fd);
+ loc_t *loc, fd_t *fd, dict_t *xdata);
int32_t dht_fsyncdir (call_frame_t *frame,
xlator_t *this,
fd_t *fd,
- int32_t datasync);
+ int32_t datasync, dict_t *xdata);
int32_t dht_statfs (call_frame_t *frame,
xlator_t *this,
- loc_t *loc);
+ loc_t *loc, dict_t *xdata);
int32_t dht_setxattr (call_frame_t *frame,
xlator_t *this,
loc_t *loc,
dict_t *dict,
- int32_t flags);
+ int32_t flags, dict_t *xdata);
int32_t dht_getxattr (call_frame_t *frame,
xlator_t *this,
loc_t *loc,
- const char *name);
+ const char *name, dict_t *xdata);
int32_t dht_fsetxattr (call_frame_t *frame,
xlator_t *this,
fd_t *fd,
dict_t *dict,
- int32_t flags);
+ int32_t flags, dict_t *xdata);
int32_t dht_fgetxattr (call_frame_t *frame,
xlator_t *this,
fd_t *fd,
- const char *name);
+ const char *name, dict_t *xdata);
int32_t dht_removexattr (call_frame_t *frame,
xlator_t *this,
loc_t *loc,
- const char *name);
+ const char *name, dict_t *xdata);
+int32_t dht_fremovexattr (call_frame_t *frame,
+ xlator_t *this,
+ fd_t *fd,
+ const char *name, dict_t *xdata);
int32_t dht_lk (call_frame_t *frame,
xlator_t *this,
fd_t *fd,
int32_t cmd,
- struct gf_flock *flock);
+ struct gf_flock *flock, dict_t *xdata);
int32_t dht_inodelk (call_frame_t *frame, xlator_t *this,
const char *volume, loc_t *loc, int32_t cmd,
- struct gf_flock *flock);
+ struct gf_flock *flock, dict_t *xdata);
int32_t dht_finodelk (call_frame_t *frame, xlator_t *this,
const char *volume, fd_t *fd, int32_t cmd,
- struct gf_flock *flock);
+ struct gf_flock *flock, dict_t *xdata);
int32_t dht_entrylk (call_frame_t *frame, xlator_t *this,
const char *volume, loc_t *loc, const char *basename,
- entrylk_cmd cmd, entrylk_type type);
+ entrylk_cmd cmd, entrylk_type type, dict_t *xdata);
int32_t dht_fentrylk (call_frame_t *frame, xlator_t *this,
const char *volume, fd_t *fd, const char *basename,
- entrylk_cmd cmd, entrylk_type type);
+ entrylk_cmd cmd, entrylk_type type, dict_t *xdata);
int32_t dht_readdir (call_frame_t *frame,
xlator_t *this,
fd_t *fd,
- size_t size, off_t off);
+ size_t size, off_t off, dict_t *xdata);
int32_t dht_readdirp (call_frame_t *frame,
xlator_t *this,
@@ -561,20 +678,29 @@ int32_t dht_xattrop (call_frame_t *frame,
xlator_t *this,
loc_t *loc,
gf_xattrop_flags_t flags,
- dict_t *dict);
+ dict_t *dict, dict_t *xdata);
int32_t dht_fxattrop (call_frame_t *frame,
xlator_t *this,
fd_t *fd,
gf_xattrop_flags_t flags,
- dict_t *dict);
+ dict_t *dict, dict_t *xdata);
int32_t dht_forget (xlator_t *this, inode_t *inode);
int32_t dht_setattr (call_frame_t *frame, xlator_t *this, loc_t *loc,
- struct iatt *stbuf, int32_t valid);
+ struct iatt *stbuf, int32_t valid, dict_t *xdata);
int32_t dht_fsetattr (call_frame_t *frame, xlator_t *this, fd_t *fd,
- struct iatt *stbuf, int32_t valid);
-
+ struct iatt *stbuf, int32_t valid, dict_t *xdata);
+int32_t dht_fallocate(call_frame_t *frame, xlator_t *this, fd_t *fd,
+ int32_t mode, off_t offset, size_t len, dict_t *xdata);
+int32_t dht_discard(call_frame_t *frame, xlator_t *this, fd_t *fd,
+ off_t offset, size_t len, dict_t *xdata);
+int32_t dht_zerofill(call_frame_t *frame, xlator_t *this, fd_t *fd,
+ off_t offset, size_t len, dict_t *xdata);
+
+int32_t dht_init (xlator_t *this);
+void dht_fini (xlator_t *this);
+int dht_reconfigure (xlator_t *this, dict_t *options);
int32_t dht_notify (xlator_t *this, int32_t event, void *data, ...);
/* definitions for nufa/switch */
@@ -597,12 +723,65 @@ int dht_lookup_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
int dht_create_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
int op_ret, int op_errno,
fd_t *fd, inode_t *inode, struct iatt *stbuf,
- struct iatt *preparent, struct iatt *postparent);
+ struct iatt *preparent, struct iatt *postparent,
+ dict_t *xdata);
int dht_newfile_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
int op_ret, int op_errno,
inode_t *inode, struct iatt *stbuf, struct iatt *preparent,
- struct iatt *postparent);
+ struct iatt *postparent, dict_t *xdata);
+
+int
+gf_defrag_status_get (gf_defrag_info_t *defrag, dict_t *dict);
+
+int
+gf_defrag_stop (gf_defrag_info_t *defrag, dict_t *output);
+void*
+gf_defrag_start (void *this);
+int32_t
+gf_defrag_handle_hardlink (xlator_t *this, loc_t *loc, dict_t *xattrs,
+ struct iatt *stbuf);
+int
+dht_migrate_file (xlator_t *this, loc_t *loc, xlator_t *from, xlator_t *to,
+ int flag);
+int
+dht_inode_ctx_layout_get (inode_t *inode, xlator_t *this,
+ dht_layout_t **layout_int);
+int
+dht_inode_ctx_layout_set (inode_t *inode, xlator_t *this,
+ dht_layout_t* layout_int);
+int
+dht_inode_ctx_time_update (inode_t *inode, xlator_t *this, struct iatt *stat,
+ int32_t update_ctx);
+
+int dht_inode_ctx_get (inode_t *inode, xlator_t *this, dht_inode_ctx_t **ctx);
+int dht_inode_ctx_set (inode_t *inode, xlator_t *this, dht_inode_ctx_t *ctx);
+int
+dht_dir_attr_heal (void *data);
+int
+dht_dir_attr_heal_done (int ret, call_frame_t *sync_frame, void *data);
+int
+dht_dir_has_layout (dict_t *xattr, char *name);
+gf_boolean_t
+dht_is_subvol_in_layout (dht_layout_t *layout, xlator_t *xlator);
+xlator_t *
+dht_subvol_with_free_space_inodes (xlator_t *this, xlator_t *subvol,
+ dht_layout_t *layout);
+xlator_t *
+dht_subvol_maxspace_nonzeroinode (xlator_t *this, xlator_t *subvol,
+ dht_layout_t *layout);
+int
+dht_linkfile_attr_heal (call_frame_t *frame, xlator_t *this);
+
+void
+dht_layout_dump (dht_layout_t *layout, const char *prefix);
+int32_t
+dht_priv_dump (xlator_t *this);
+int32_t
+dht_inodectx_dump (xlator_t *this, inode_t *inode);
+
+int
+dht_inode_ctx_get1 (xlator_t *this, inode_t *inode, xlator_t **subvol);
#endif/* _DHT_H */
diff --git a/xlators/cluster/dht/src/dht-diskusage.c b/xlators/cluster/dht/src/dht-diskusage.c
index 5453e3b10..fe3955ecb 100644
--- a/xlators/cluster/dht/src/dht-diskusage.c
+++ b/xlators/cluster/dht/src/dht-diskusage.c
@@ -1,20 +1,11 @@
/*
- Copyright (c) 2010-2011 Gluster, Inc. <http://www.gluster.com>
+ Copyright (c) 2008-2012 Red Hat, Inc. <http://www.redhat.com>
This file is part of GlusterFS.
- GlusterFS is free software; you can redistribute it and/or modify
- it under the terms of the GNU General Public License as published
- by the Free Software Foundation; either version 3 of the License,
- or (at your option) any later version.
-
- GlusterFS is distributed in the hope that it will be useful, but
- WITHOUT ANY WARRANTY; without even the implied warranty of
- MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU
- General Public License for more details.
-
- You should have received a copy of the GNU General Public License
- along with this program. If not, see
- <http://www.gnu.org/licenses/>.
+ This file is licensed to you under your choice of the GNU Lesser
+ General Public License, version 3 or any later version (LGPLv3 or
+ later), or the GNU General Public License, version 2 (GPLv2), in all
+ cases as published by the Free Software Foundation.
*/
@@ -35,7 +26,8 @@
int
dht_du_info_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
- int op_ret, int op_errno, struct statvfs *statvfs)
+ int op_ret, int op_errno, struct statvfs *statvfs,
+ dict_t *xdata)
{
dht_conf_t *conf = NULL;
call_frame_t *prev = NULL;
@@ -105,6 +97,7 @@ dht_get_du_info_for_subvol (xlator_t *this, int subvol_idx)
call_frame_t *statfs_frame = NULL;
dht_local_t *statfs_local = NULL;
call_pool_t *pool = NULL;
+ loc_t tmp_loc = {0,};
conf = this->private;
pool = this->ctx->pool;
@@ -121,15 +114,14 @@ dht_get_du_info_for_subvol (xlator_t *this, int subvol_idx)
goto err;
}
- loc_t tmp_loc = { .inode = NULL,
- .path = "/",
- };
+ /* make it root gfid, should be enough to get the proper info back */
+ tmp_loc.gfid[15] = 1;
statfs_local->call_cnt = 1;
STACK_WIND (statfs_frame, dht_du_info_cbk,
conf->subvolumes[subvol_idx],
conf->subvolumes[subvol_idx]->fops->statfs,
- &tmp_loc);
+ &tmp_loc, NULL);
return 0;
err:
@@ -142,15 +134,21 @@ err:
int
dht_get_du_info (call_frame_t *frame, xlator_t *this, loc_t *loc)
{
- int i = 0;
+ int i = 0;
dht_conf_t *conf = NULL;
call_frame_t *statfs_frame = NULL;
dht_local_t *statfs_local = NULL;
- struct timeval tv = {0,};
+ struct timeval tv = {0,};
+ loc_t tmp_loc = {0,};
conf = this->private;
gettimeofday (&tv, NULL);
+
+ /* make it root gfid, should be enough to get the proper
+ info back */
+ tmp_loc.gfid[15] = 1;
+
if (tv.tv_sec > (conf->refresh_interval
+ conf->last_stat_fetch.tv_sec)) {
@@ -166,16 +164,12 @@ dht_get_du_info (call_frame_t *frame, xlator_t *this, loc_t *loc)
goto err;
}
- loc_t tmp_loc = { .inode = NULL,
- .path = "/",
- };
-
statfs_local->call_cnt = conf->subvolume_cnt;
for (i = 0; i < conf->subvolume_cnt; i++) {
STACK_WIND (statfs_frame, dht_du_info_cbk,
conf->subvolumes[i],
conf->subvolumes[i]->fops->statfs,
- &tmp_loc);
+ &tmp_loc, NULL);
}
conf->last_stat_fetch.tv_sec = tv.tv_sec;
@@ -254,50 +248,167 @@ dht_is_subvol_filled (xlator_t *this, xlator_t *subvol)
return is_subvol_filled;
}
+
+/*Get the best subvolume to create the file in*/
xlator_t *
-dht_free_disk_available_subvol (xlator_t *this, xlator_t *subvol)
+dht_free_disk_available_subvol (xlator_t *this, xlator_t *subvol,
+ dht_local_t *local)
{
- int i = 0;
- double max = 0;
- double max_inodes = 0;
xlator_t *avail_subvol = NULL;
dht_conf_t *conf = NULL;
+ dht_layout_t *layout = NULL;
+ loc_t *loc = NULL;
conf = this->private;
-
- LOCK (&conf->subvolume_lock);
+ if (!local)
+ goto out;
+ loc = &local->loc;
+ if (!local->layout) {
+ layout = dht_layout_get (this, loc->parent);
+
+ if (!layout) {
+ gf_log (this->name, GF_LOG_DEBUG,
+ "layout missing path=%s parent=%s",
+ loc->path, uuid_utoa (loc->parent->gfid));
+ goto out;
+ }
+ } else {
+ layout = dht_layout_ref (this, local->layout);
+ }
+
+ LOCK (&conf->subvolume_lock);
{
- for (i = 0; i < conf->subvolume_cnt; i++) {
- if (conf->disk_unit == 'p') {
- if ((conf->du_stats[i].avail_percent > max)
- && (conf->du_stats[i].avail_inodes > max_inodes)) {
- max = conf->du_stats[i].avail_percent;
- max_inodes = conf->du_stats[i].avail_inodes;
- avail_subvol = conf->subvolumes[i];
- }
- } else {
- if ((conf->du_stats[i].avail_space > max)
- && (conf->du_stats[i].avail_inodes > max_inodes)) {
- max = conf->du_stats[i].avail_space;
- max_inodes = conf->du_stats[i].avail_inodes;
- avail_subvol = conf->subvolumes[i];
- }
+ avail_subvol = dht_subvol_with_free_space_inodes(this, subvol,
+ layout);
+ if(!avail_subvol)
+ {
+ avail_subvol = dht_subvol_maxspace_nonzeroinode(this,
+ subvol,
+ layout);
+ }
- }
- }
}
UNLOCK (&conf->subvolume_lock);
-
+out:
if (!avail_subvol) {
- gf_log (this->name, GF_LOG_DEBUG,
- "no subvolume has enough free space and inodes to create");
+ gf_log (this->name,
+ GF_LOG_DEBUG,
+ "no subvolume has enough free space and/or inodes\
+ to create");
+ avail_subvol = subvol;
}
- if ((max < conf->min_free_disk) && (max_inodes < conf->min_free_inodes))
- avail_subvol = subvol;
+ if (layout)
+ dht_layout_unref (this, layout);
+ return avail_subvol;
+}
- if (!avail_subvol)
- avail_subvol = subvol;
+static inline
+int32_t dht_subvol_has_err (xlator_t *this, dht_layout_t *layout)
+{
+ int ret = -1;
+ int i = 0;
+
+ if (!this || !layout)
+ goto out;
+
+ /* check if subvol has layout errors, before selecting it */
+ for (i = 0; i < layout->cnt; i++) {
+ if (!strcmp (layout->list[i].xlator->name, this->name) &&
+ (layout->list[i].err != 0)) {
+ ret = -1;
+ goto out;
+ }
+ }
+ ret = 0;
+out:
+ return ret;
+}
- return avail_subvol;
+/*Get subvolume which has both space and inodes more than the min criteria*/
+xlator_t *
+dht_subvol_with_free_space_inodes(xlator_t *this, xlator_t *subvol,
+ dht_layout_t *layout)
+{
+ int i = 0;
+ double max = 0;
+ double max_inodes = 0;
+ int ignore_subvol = 0;
+
+ xlator_t *avail_subvol = NULL;
+ dht_conf_t *conf = NULL;
+
+ conf = this->private;
+
+ for(i=0; i < conf->subvolume_cnt; i++) {
+ /* check if subvol has layout errors, before selecting it */
+ ignore_subvol = dht_subvol_has_err (conf->subvolumes[i],
+ layout);
+ if (ignore_subvol)
+ continue;
+
+ if ((conf->disk_unit == 'p') &&
+ (conf->du_stats[i].avail_percent > conf->min_free_disk) &&
+ (conf->du_stats[i].avail_inodes > conf->min_free_inodes)) {
+ if ((conf->du_stats[i].avail_inodes > max_inodes) ||
+ (conf->du_stats[i].avail_percent > max)) {
+ max = conf->du_stats[i].avail_percent;
+ max_inodes = conf->du_stats[i].avail_inodes;
+ avail_subvol = conf->subvolumes[i];
+ }
+ }
+
+ if ((conf->disk_unit != 'p') &&
+ (conf->du_stats[i].avail_space > conf->min_free_disk) &&
+ (conf->du_stats[i].avail_inodes > conf->min_free_inodes)) {
+ if ((conf->du_stats[i].avail_inodes > max_inodes) ||
+ (conf->du_stats[i].avail_space > max)) {
+ max = conf->du_stats[i].avail_space;
+ max_inodes = conf->du_stats[i].avail_inodes;
+ avail_subvol = conf->subvolumes[i];
+ }
+ }
+ }
+
+ return avail_subvol;
+}
+
+
+/* Get subvol which has atleast one inode and maximum space */
+xlator_t *
+dht_subvol_maxspace_nonzeroinode (xlator_t *this, xlator_t *subvol,
+ dht_layout_t *layout)
+{
+ int i = 0;
+ double max = 0;
+ int ignore_subvol = 0;
+
+ xlator_t *avail_subvol = NULL;
+ dht_conf_t *conf = NULL;
+
+ conf = this->private;
+
+ for (i = 0; i < conf->subvolume_cnt; i++) {
+ /* check if subvol has layout errors, before selecting it */
+ ignore_subvol = dht_subvol_has_err (conf->subvolumes[i],
+ layout);
+ if (ignore_subvol)
+ continue;
+
+ if (conf->disk_unit == 'p') {
+ if ((conf->du_stats[i].avail_percent > max)
+ && (conf->du_stats[i].avail_inodes > 0 )) {
+ max = conf->du_stats[i].avail_percent;
+ avail_subvol = conf->subvolumes[i];
+ }
+ } else {
+ if ((conf->du_stats[i].avail_space > max)
+ && (conf->du_stats[i].avail_inodes > 0)) {
+ max = conf->du_stats[i].avail_space;
+ avail_subvol = conf->subvolumes[i];
+ }
+ }
+ }
+
+ return avail_subvol;
}
diff --git a/xlators/cluster/dht/src/dht-hashfn.c b/xlators/cluster/dht/src/dht-hashfn.c
index c8ae74172..656cf23a0 100644
--- a/xlators/cluster/dht/src/dht-hashfn.c
+++ b/xlators/cluster/dht/src/dht-hashfn.c
@@ -1,20 +1,11 @@
/*
- Copyright (c) 2008-2011 Gluster, Inc. <http://www.gluster.com>
+ Copyright (c) 2008-2012 Red Hat, Inc. <http://www.redhat.com>
This file is part of GlusterFS.
- GlusterFS is free software; you can redistribute it and/or modify
- it under the terms of the GNU General Public License as published
- by the Free Software Foundation; either version 3 of the License,
- or (at your option) any later version.
-
- GlusterFS is distributed in the hope that it will be useful, but
- WITHOUT ANY WARRANTY; without even the implied warranty of
- MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU
- General Public License for more details.
-
- You should have received a copy of the GNU General Public License
- along with this program. If not, see
- <http://www.gnu.org/licenses/>.
+ This file is licensed to you under your choice of the GNU Lesser
+ General Public License, version 3 or any later version (LGPLv3 or
+ later), or the GNU General Public License, version 2 (GPLv2), in all
+ cases as published by the Free Software Foundation.
*/
#ifndef _CONFIG_H
@@ -37,6 +28,7 @@ dht_hash_compute_internal (int type, const char *name, uint32_t *hash_p)
switch (type) {
case DHT_HASH_TYPE_DM:
+ case DHT_HASH_TYPE_DM_USER:
hash = gf_dm_hashfn (name, strlen (name));
break;
default:
@@ -52,30 +44,68 @@ dht_hash_compute_internal (int type, const char *name, uint32_t *hash_p)
}
-#define MAKE_RSYNC_FRIENDLY_NAME(rsync_frndly_name, name) do { \
- rsync_frndly_name = (char *) name; \
- if (name[0] == '.') { \
- char *dot = 0; \
- int namelen = 0; \
- \
- dot = strrchr (name, '.'); \
- if (dot && dot > (name + 1) && *(dot + 1)) { \
- namelen = (dot - name); \
- rsync_frndly_name = alloca (namelen); \
- strncpy (rsync_frndly_name, name + 1, \
- namelen); \
- rsync_frndly_name[namelen - 1] = 0; \
- } \
- } \
- } while (0);
+static inline
+gf_boolean_t
+dht_munge_name (const char *original, char *modified, size_t len, regex_t *re)
+{
+ regmatch_t matches[2];
+ size_t new_len;
+
+ if (regexec(re,original,2,matches,0) != REG_NOMATCH) {
+ if (matches[1].rm_so != -1) {
+ new_len = matches[1].rm_eo - matches[1].rm_so;
+ /* Equal would fail due to the NUL at the end. */
+ if (new_len < len) {
+ memcpy (modified,original+matches[1].rm_so,
+ new_len);
+ modified[new_len] = '\0';
+ return _gf_true;
+ }
+ }
+ }
+ /* This is guaranteed safe because of how the dest was allocated. */
+ strcpy(modified,original);
+ return _gf_false;
+}
int
-dht_hash_compute (int type, const char *name, uint32_t *hash_p)
+dht_hash_compute (xlator_t *this, int type, const char *name, uint32_t *hash_p)
{
- char *rsync_friendly_name = NULL;
+ char *rsync_friendly_name = NULL;
+ dht_conf_t *priv = this->private;
+ size_t len = 0;
+ gf_boolean_t munged = _gf_false;
+
+ /*
+ * It wouldn't be safe to use alloca in an inline function that doesn't
+ * actually get inlined, and it wouldn't be efficient to do a real
+ * allocation, so we use alloca here (if needed) and pass that to the
+ * inline.
+ */
+
+ if (priv->extra_regex_valid) {
+ len = strlen(name) + 1;
+ rsync_friendly_name = alloca(len);
+ munged = dht_munge_name (name, rsync_friendly_name, len,
+ &priv->extra_regex);
+ }
+
+ if (!munged && priv->rsync_regex_valid) {
+ len = strlen(name) + 1;
+ rsync_friendly_name = alloca(len);
+ gf_log (this->name, GF_LOG_TRACE, "trying regex for %s", name);
+ munged = dht_munge_name (name, rsync_friendly_name, len,
+ &priv->rsync_regex);
+ if (munged) {
+ gf_log (this->name, GF_LOG_DEBUG,
+ "munged down to %s", rsync_friendly_name);
+ }
+ }
- MAKE_RSYNC_FRIENDLY_NAME (rsync_friendly_name, name);
+ if (!munged) {
+ rsync_friendly_name = (char *)name;
+ }
return dht_hash_compute_internal (type, rsync_friendly_name, hash_p);
}
diff --git a/xlators/cluster/dht/src/dht-helper.c b/xlators/cluster/dht/src/dht-helper.c
index 65c03b62f..311a48112 100644
--- a/xlators/cluster/dht/src/dht-helper.c
+++ b/xlators/cluster/dht/src/dht-helper.c
@@ -1,20 +1,11 @@
/*
- Copyright (c) 2008-2011 Gluster, Inc. <http://www.gluster.com>
+ Copyright (c) 2008-2012 Red Hat, Inc. <http://www.redhat.com>
This file is part of GlusterFS.
- GlusterFS is free software; you can redistribute it and/or modify
- it under the terms of the GNU General Public License as published
- by the Free Software Foundation; either version 3 of the License,
- or (at your option) any later version.
-
- GlusterFS is distributed in the hope that it will be useful, but
- WITHOUT ANY WARRANTY; without even the implied warranty of
- MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU
- General Public License for more details.
-
- You should have received a copy of the GNU General Public License
- along with this program. If not, see
- <http://www.gnu.org/licenses/>.
+ This file is licensed to you under your choice of the GNU Lesser
+ General Public License, version 3 or any later version (LGPLv3 or
+ later), or the GNU General Public License, version 2 (GPLv2), in all
+ cases as published by the Free Software Foundation.
*/
#ifndef _CONFIG_H
@@ -27,6 +18,28 @@
#include "xlator.h"
#include "dht-common.h"
+static inline int
+dht_inode_ctx_set1 (xlator_t *this, inode_t *inode, xlator_t *subvol)
+{
+ uint64_t tmp_subvol = 0;
+
+ tmp_subvol = (long)subvol;
+ return inode_ctx_set1 (inode, this, &tmp_subvol);
+}
+
+int
+dht_inode_ctx_get1 (xlator_t *this, inode_t *inode, xlator_t **subvol)
+{
+ int ret = -1;
+ uint64_t tmp_subvol = 0;
+
+ ret = inode_ctx_get1 (inode, this, &tmp_subvol);
+ if (tmp_subvol && subvol)
+ *subvol = (xlator_t *)tmp_subvol;
+
+ return ret;
+}
+
int
dht_frame_return (call_frame_t *frame)
@@ -49,6 +62,43 @@ dht_frame_return (call_frame_t *frame)
}
+static uint64_t
+dht_bits_for (uint64_t num)
+{
+ uint64_t bits = 0, ctrl = 1;
+
+ while (ctrl < num) {
+ ctrl *= 2;
+ bits ++;
+ }
+
+ return bits;
+}
+
+/*
+ * A slightly "updated" version of the algorithm described in the commit log
+ * is used here.
+ *
+ * The only enhancement is that:
+ *
+ * - The number of bits used by the backend filesystem for HUGE d_off which
+ * is described as 63, and
+ * - The number of bits used by the d_off presented by the transformation
+ * upwards which is described as 64, are both made "configurable."
+ */
+
+
+#define BACKEND_D_OFF_BITS 63
+#define PRESENT_D_OFF_BITS 63
+
+#define ONE 1ULL
+#define MASK (~0ULL)
+#define PRESENT_MASK (MASK >> (64 - PRESENT_D_OFF_BITS))
+#define BACKEND_MASK (MASK >> (64 - BACKEND_D_OFF_BITS))
+
+#define TOP_BIT (ONE << (PRESENT_D_OFF_BITS - 1))
+#define SHIFT_BITS (max (0, (BACKEND_D_OFF_BITS - PRESENT_D_OFF_BITS + 1)))
+
int
dht_itransform (xlator_t *this, xlator_t *subvol, uint64_t x, uint64_t *y_p)
{
@@ -56,6 +106,9 @@ dht_itransform (xlator_t *this, xlator_t *subvol, uint64_t x, uint64_t *y_p)
int cnt = 0;
int max = 0;
uint64_t y = 0;
+ uint64_t hi_mask = 0;
+ uint64_t off_mask = 0;
+ int max_bits = 0;
if (x == ((uint64_t) -1)) {
y = (uint64_t) -1;
@@ -69,7 +122,23 @@ dht_itransform (xlator_t *this, xlator_t *subvol, uint64_t x, uint64_t *y_p)
max = conf->subvolume_cnt;
cnt = dht_subvol_cnt (this, subvol);
- y = ((x * max) + cnt);
+ if (max == 1) {
+ y = x;
+ goto out;
+ }
+
+ max_bits = dht_bits_for (max);
+
+ hi_mask = ~(PRESENT_MASK >> (max_bits + 1));
+
+ if (x & hi_mask) {
+ /* HUGE d_off */
+ off_mask = MASK << max_bits;
+ y = TOP_BIT | ((x >> SHIFT_BITS) & off_mask) | cnt;
+ } else {
+ /* small d_off */
+ y = ((x * max) + cnt);
+ }
out:
if (y_p)
@@ -89,7 +158,7 @@ dht_filter_loc_subvol_key (xlator_t *this, loc_t *loc, loc_t *new_loc,
int ret = 0; /* not found */
/* Why do other tasks if first required 'char' itself is not there */
- if (!loc->name || !strchr (loc->name, '@'))
+ if (!new_loc || !loc || !loc->name || !strchr (loc->name, '@'))
goto out;
trav = this->children;
@@ -129,10 +198,8 @@ dht_filter_loc_subvol_key (xlator_t *this, loc_t *loc, loc_t *new_loc,
out:
if (!ret) {
/* !success */
- if (new_path)
- GF_FREE (new_path);
- if (new_name)
- GF_FREE (new_name);
+ GF_FREE (new_path);
+ GF_FREE (new_name);
}
return ret;
}
@@ -146,16 +213,38 @@ dht_deitransform (xlator_t *this, uint64_t y, xlator_t **subvol_p,
int max = 0;
uint64_t x = 0;
xlator_t *subvol = 0;
+ int max_bits = 0;
+ uint64_t off_mask = 0;
+ uint64_t host_mask = 0;
if (!this->private)
- goto out;
+ return -1;
conf = this->private;
max = conf->subvolume_cnt;
- cnt = y % max;
- x = y / max;
+ if (max == 1) {
+ x = y;
+ cnt = 0;
+ goto out;
+ }
+
+ if (y & TOP_BIT) {
+ /* HUGE d_off */
+ max_bits = dht_bits_for (max);
+ off_mask = (MASK << max_bits);
+ host_mask = ~(off_mask);
+
+ x = ((y & ~TOP_BIT) & off_mask) << SHIFT_BITS;
+
+ cnt = y & host_mask;
+ } else {
+ /* small d_off */
+ cnt = y % max;
+ x = y / max;
+ }
+out:
subvol = conf->subvolumes[cnt];
if (subvol_p)
@@ -164,7 +253,6 @@ dht_deitransform (xlator_t *this, uint64_t y, xlator_t **subvol_p,
if (x_p)
*x_p = x;
-out:
return 0;
}
@@ -215,21 +303,16 @@ dht_local_wipe (xlator_t *this, dht_local_t *local)
local->selfheal.layout = NULL;
}
- if (local->newpath) {
- GF_FREE (local->newpath);
- }
+ GF_FREE (local->newpath);
- if (local->key) {
- GF_FREE (local->key);
- }
+ GF_FREE (local->key);
- if (local->rebalance.vector)
- GF_FREE (local->rebalance.vector);
+ GF_FREE (local->rebalance.vector);
if (local->rebalance.iobref)
iobref_unref (local->rebalance.iobref);
- GF_FREE (local);
+ mem_put (local);
}
@@ -240,8 +323,7 @@ dht_local_init (call_frame_t *frame, loc_t *loc, fd_t *fd, glusterfs_fop_t fop)
inode_t *inode = NULL;
int ret = 0;
- /* TODO: use mem-pool */
- local = GF_CALLOC (1, sizeof (*local), gf_dht_mt_dht_local_t);
+ local = mem_get0 (THIS->local_pool);
if (!local)
goto out;
@@ -274,26 +356,12 @@ dht_local_init (call_frame_t *frame, loc_t *loc, fd_t *fd, glusterfs_fop_t fop)
out:
if (ret) {
if (local)
- GF_FREE (local);
+ mem_put (local);
local = NULL;
}
return local;
}
-
-char *
-basestr (const char *str)
-{
- char *basestr = NULL;
-
- basestr = strrchr (str, '/');
- if (basestr)
- basestr ++;
-
- return basestr;
-}
-
-
xlator_t *
dht_first_up_subvol (xlator_t *this)
{
@@ -358,11 +426,17 @@ dht_subvol_get_hashed (xlator_t *this, loc_t *loc)
dht_layout_t *layout = NULL;
xlator_t *subvol = NULL;
- if (is_fs_root (loc)) {
+ GF_VALIDATE_OR_GOTO ("dht", this, out);
+ GF_VALIDATE_OR_GOTO (this->name, loc, out);
+
+ if (__is_root_gfid (loc->gfid)) {
subvol = dht_first_up_subvol (this);
goto out;
}
+ GF_VALIDATE_OR_GOTO (this->name, loc->parent, out);
+ GF_VALIDATE_OR_GOTO (this->name, loc->name, out);
+
layout = dht_layout_get (this, loc->parent);
if (!layout) {
@@ -396,6 +470,8 @@ dht_subvol_get_cached (xlator_t *this, inode_t *inode)
dht_layout_t *layout = NULL;
xlator_t *subvol = NULL;
+ GF_VALIDATE_OR_GOTO (this->name, this, out);
+ GF_VALIDATE_OR_GOTO (this->name, inode, out);
layout = dht_layout_get (this, inode);
@@ -437,7 +513,36 @@ out:
return next;
}
+/* This func wraps around, if prev is actually the last subvol.
+ */
+xlator_t *
+dht_subvol_next_available (xlator_t *this, xlator_t *prev)
+{
+ dht_conf_t *conf = NULL;
+ int i = 0;
+ xlator_t *next = NULL;
+
+ conf = this->private;
+ if (!conf)
+ goto out;
+ for (i = 0; i < conf->subvolume_cnt; i++) {
+ if (conf->subvolumes[i] == prev) {
+ /* if prev is last in conf->subvolumes, then wrap
+ * around.
+ */
+ if ((i + 1) < conf->subvolume_cnt) {
+ next = conf->subvolumes[i + 1];
+ } else {
+ next = conf->subvolumes[0];
+ }
+ break;
+ }
+ }
+
+out:
+ return next;
+}
int
dht_subvol_cnt (xlator_t *this, xlator_t *subvol)
{
@@ -466,6 +571,15 @@ out:
(a) = (b); \
} while (0)
+
+#define set_if_greater_time(a, an, b, bn) do { \
+ if (((a) < (b)) || (((a) == (b)) && ((an) < (bn)))){ \
+ (a) = (b); \
+ (an) = (bn); \
+ } \
+ } while (0) \
+
+
int
dht_iatt_merge (xlator_t *this, struct iatt *to,
struct iatt *from, xlator_t *subvol)
@@ -489,9 +603,12 @@ dht_iatt_merge (xlator_t *this, struct iatt *to,
set_if_greater (to->ia_uid, from->ia_uid);
set_if_greater (to->ia_gid, from->ia_gid);
- set_if_greater (to->ia_atime, from->ia_atime);
- set_if_greater (to->ia_mtime, from->ia_mtime);
- set_if_greater (to->ia_ctime, from->ia_ctime);
+ set_if_greater_time(to->ia_atime, to->ia_atime_nsec,
+ from->ia_atime, from->ia_atime_nsec);
+ set_if_greater_time (to->ia_mtime, to->ia_mtime_nsec,
+ from->ia_mtime, from->ia_mtime_nsec);
+ set_if_greater_time (to->ia_ctime, to->ia_ctime_nsec,
+ from->ia_ctime, from->ia_ctime_nsec);
return 0;
}
@@ -617,20 +734,36 @@ dht_migration_complete_check_task (void *data)
call_frame_t *frame = NULL;
loc_t tmp_loc = {0,};
char *path = NULL;
+ dht_conf_t *conf = NULL;
+ inode_t *inode = NULL;
+ fd_t *iter_fd = NULL;
+ uint64_t tmp_subvol = 0;
+ int open_failed = 0;
this = THIS;
frame = data;
local = frame->local;
+ conf = this->private;
src_node = local->cached_subvol;
- /* getxattr on cached_subvol for 'linkto' value */
- if (!local->loc.inode)
+ if (!local->loc.inode && !local->fd)
+ goto out;
+
+ inode = (!local->fd) ? local->loc.inode : local->fd->inode;
+
+ /* getxattr on cached_subvol for 'linkto' value. Do path based getxattr
+ * as root:root. If a fd is already open, access check wont be done*/
+
+ if (!local->loc.inode) {
ret = syncop_fgetxattr (src_node, local->fd, &dict,
- DHT_LINKFILE_KEY);
- else
+ conf->link_xattr_name);
+ } else {
+ SYNCTASK_SETID (0, 0);
ret = syncop_getxattr (src_node, &local->loc, &dict,
- DHT_LINKFILE_KEY);
+ conf->link_xattr_name);
+ SYNCTASK_SETID (frame->root->uid, frame->root->gid);
+ }
if (!ret)
dst_node = dht_linkfile_subvol (this, NULL, NULL, dict);
@@ -681,10 +814,7 @@ dht_migration_complete_check_task (void *data)
/* update inode ctx (the layout) */
dht_layout_unref (this, local->layout);
- if (!local->loc.inode)
- ret = dht_layout_preset (this, dst_node, local->fd->inode);
- else
- ret = dht_layout_preset (this, dst_node, local->loc.inode);
+ ret = dht_layout_preset (this, dst_node, inode);
if (ret != 0) {
gf_log (this->name, GF_LOG_DEBUG,
"%s: could not set preset layout for subvol %s",
@@ -702,10 +832,7 @@ dht_migration_complete_check_task (void *data)
goto out;
}
- if (!local->loc.inode)
- ret = dht_layout_set (this, local->fd->inode, layout);
- else
- ret = dht_layout_set (this, local->loc.inode, layout);
+ ret = dht_layout_set (this, inode, layout);
if (ret) {
gf_log (this->name, GF_LOG_ERROR,
"%s: failed to set the new layout",
@@ -716,43 +843,46 @@ dht_migration_complete_check_task (void *data)
local->cached_subvol = dst_node;
ret = 0;
- if (!local->fd)
+ /* once we detect the migration complete, the inode-ctx2 is no more
+ required.. delete the ctx and also, it means, open() already
+ done on all the fd of inode */
+ ret = inode_ctx_reset1 (inode, this, &tmp_subvol);
+ if (tmp_subvol)
goto out;
- /* once we detect the migration complete, the fd-ctx is no more
- required.. delete the ctx, and do one extra 'fd_unref' for open fd */
- ret = fd_ctx_del (local->fd, this, NULL);
- if (!ret) {
- fd_unref (local->fd);
- ret = 0;
+ if (list_empty (&inode->fd_list))
goto out;
- }
- /* if 'local->fd' (ie, fd based operation), send a 'open()' on
- destination if not already done */
- if (local->loc.inode) {
- ret = syncop_open (dst_node, &local->loc,
- local->fd->flags, local->fd);
- } else {
- tmp_loc.inode = local->fd->inode;
- inode_path (local->fd->inode, NULL, &path);
- if (path)
- tmp_loc.path = path;
- ret = syncop_open (dst_node, &tmp_loc,
- local->fd->flags, local->fd);
- if (path)
- GF_FREE (path);
+ /* perform open as root:root. There is window between linkfile
+ * creation(root:root) and setattr with the correct uid/gid
+ */
+ SYNCTASK_SETID(0, 0);
+
+ /* perform 'open()' on all the fd's present on the inode */
+ tmp_loc.inode = inode;
+ inode_path (inode, NULL, &path);
+ if (path)
+ tmp_loc.path = path;
+ list_for_each_entry (iter_fd, &inode->fd_list, inode_list) {
+ if (fd_is_anonymous (iter_fd))
+ continue;
+ ret = syncop_open (dst_node, &tmp_loc,
+ iter_fd->flags, iter_fd);
+ if (ret == -1) {
+ gf_log (this->name, GF_LOG_ERROR, "failed to open "
+ "the fd (%p, flags=0%o) on file %s @ %s",
+ iter_fd, iter_fd->flags, path, dst_node->name);
+ open_failed = 1;
+ }
}
- if (ret == -1) {
- gf_log (this->name, GF_LOG_ERROR,
- "%s: failed to send open() on target file at %s",
- local->loc.path, dst_node->name);
+ GF_FREE (path);
+
+ SYNCTASK_SETID (frame->root->uid, frame->root->gid);
+ if (open_failed) {
+ ret = -1;
goto out;
}
-
- /* need this unref for the fd on src_node */
- fd_unref (local->fd);
ret = 0;
out:
@@ -763,11 +893,8 @@ int
dht_rebalance_complete_check (xlator_t *this, call_frame_t *frame)
{
int ret = -1;
- dht_conf_t *conf = NULL;
-
- conf = this->private;
- ret = synctask_new (conf->env, dht_migration_complete_check_task,
+ ret = synctask_new (this->ctx->env, dht_migration_complete_check_task,
dht_migration_complete_check_done,
frame, frame);
return ret;
@@ -799,20 +926,34 @@ dht_rebalance_inprogress_task (void *data)
char *path = NULL;
struct iatt stbuf = {0,};
loc_t tmp_loc = {0,};
+ dht_conf_t *conf = NULL;
+ inode_t *inode = NULL;
+ fd_t *iter_fd = NULL;
+ int open_failed = 0;
this = THIS;
frame = data;
local = frame->local;
+ conf = this->private;
src_node = local->cached_subvol;
- /* getxattr on cached_subvol for 'linkto' value */
- if (local->loc.inode)
+ if (!local->loc.inode && !local->fd)
+ goto out;
+
+ inode = (!local->fd) ? local->loc.inode : local->fd->inode;
+
+ /* getxattr on cached_subvol for 'linkto' value. Do path based getxattr
+ * as root:root. If a fd is already open, access check wont be done*/
+ if (local->loc.inode) {
+ SYNCTASK_SETID (0, 0);
ret = syncop_getxattr (src_node, &local->loc, &dict,
- DHT_LINKFILE_KEY);
- else
+ conf->link_xattr_name);
+ SYNCTASK_SETID (frame->root->uid, frame->root->gid);
+ } else {
ret = syncop_fgetxattr (src_node, local->fd, &dict,
- DHT_LINKFILE_KEY);
+ conf->link_xattr_name);
+ }
if (ret) {
gf_log (this->name, GF_LOG_ERROR,
@@ -854,34 +995,46 @@ dht_rebalance_inprogress_task (void *data)
ret = 0;
- if (!local->fd)
- goto out;
+ if (list_empty (&inode->fd_list))
+ goto done;
+
+ /* perform open as root:root. There is window between linkfile
+ * creation(root:root) and setattr with the correct uid/gid
+ */
+ SYNCTASK_SETID (0, 0);
+
+ tmp_loc.inode = inode;
+ inode_path (inode, NULL, &path);
+ if (path)
+ tmp_loc.path = path;
+
+ list_for_each_entry (iter_fd, &inode->fd_list, inode_list) {
+ if (fd_is_anonymous (iter_fd))
+ continue;
- if (local->loc.inode) {
- ret = syncop_open (dst_node, &local->loc,
- local->fd->flags, local->fd);
- } else {
- tmp_loc.inode = local->fd->inode;
- inode_path (local->fd->inode, NULL, &path);
- if (path)
- tmp_loc.path = path;
ret = syncop_open (dst_node, &tmp_loc,
- local->fd->flags, local->fd);
- if (path)
- GF_FREE (path);
+ iter_fd->flags, iter_fd);
+ if (ret == -1) {
+ gf_log (this->name, GF_LOG_ERROR, "failed to send open "
+ "the fd (%p, flags=0%o) on file %s @ %s",
+ iter_fd, iter_fd->flags, path, dst_node->name);
+ open_failed = 1;
+ }
}
+ GF_FREE (path);
- if (ret == -1) {
- gf_log (this->name, GF_LOG_ERROR,
- "%s: failed to send open() on target file at %s",
- local->loc.path, dst_node->name);
+ SYNCTASK_SETID (frame->root->uid, frame->root->gid);
+
+ if (open_failed) {
+ ret = -1;
goto out;
}
- ret = fd_ctx_set (local->fd, this, (uint64_t)(long)dst_node);
+done:
+ ret = dht_inode_ctx_set1 (this, inode, dst_node);
if (ret) {
gf_log (this->name, GF_LOG_ERROR,
- "%s: failed to set fd-ctx target file at %s",
+ "%s: failed to set inode-ctx target file at %s",
local->loc.path, dst_node->name);
goto out;
}
@@ -896,12 +1049,99 @@ dht_rebalance_in_progress_check (xlator_t *this, call_frame_t *frame)
{
int ret = -1;
- dht_conf_t *conf = NULL;
- conf = this->private;
-
- ret = synctask_new (conf->env, dht_rebalance_inprogress_task,
+ ret = synctask_new (this->ctx->env, dht_rebalance_inprogress_task,
dht_inprogress_check_done,
frame, frame);
return ret;
}
+
+int
+dht_inode_ctx_layout_set (inode_t *inode, xlator_t *this,
+ dht_layout_t *layout_int)
+{
+ dht_inode_ctx_t *ctx = NULL;
+ int ret = -1;
+
+ ret = dht_inode_ctx_get (inode, this, &ctx);
+ if (!ret && ctx) {
+ ctx->layout = layout_int;
+ } else {
+ ctx = GF_CALLOC (1, sizeof (*ctx), gf_dht_mt_inode_ctx_t);
+ if (!ctx)
+ return ret;
+ ctx->layout = layout_int;
+ }
+
+ ret = dht_inode_ctx_set (inode, this, ctx);
+
+ return ret;
+}
+
+int
+dht_inode_ctx_time_update (inode_t *inode, xlator_t *this, struct iatt *stat,
+ int32_t post)
+{
+ dht_inode_ctx_t *ctx = NULL;
+ dht_stat_time_t *time = 0;
+ int ret = -1;
+
+ GF_VALIDATE_OR_GOTO (this->name, stat, out);
+ GF_VALIDATE_OR_GOTO (this->name, inode, out);
+
+ ret = dht_inode_ctx_get (inode, this, &ctx);
+
+ if (ret) {
+ ctx = GF_CALLOC (1, sizeof (*ctx), gf_dht_mt_inode_ctx_t);
+ if (!ctx)
+ return -1;
+ }
+
+ time = &ctx->time;
+
+ DHT_UPDATE_TIME(time->mtime, time->mtime_nsec,
+ stat->ia_mtime, stat->ia_mtime_nsec, inode, post);
+ DHT_UPDATE_TIME(time->ctime, time->ctime_nsec,
+ stat->ia_ctime, stat->ia_ctime_nsec, inode, post);
+ DHT_UPDATE_TIME(time->atime, time->atime_nsec,
+ stat->ia_atime, stat->ia_atime_nsec, inode, post);
+
+ ret = dht_inode_ctx_set (inode, this, ctx);
+out:
+ return 0;
+}
+
+int
+dht_inode_ctx_get (inode_t *inode, xlator_t *this, dht_inode_ctx_t **ctx)
+{
+ int ret = -1;
+ uint64_t ctx_int = 0;
+
+ GF_VALIDATE_OR_GOTO ("dht", this, out);
+ GF_VALIDATE_OR_GOTO (this->name, inode, out);
+
+ ret = inode_ctx_get (inode, this, &ctx_int);
+
+ if (ret)
+ return ret;
+
+ if (ctx)
+ *ctx = (dht_inode_ctx_t *) ctx_int;
+out:
+ return ret;
+}
+
+int dht_inode_ctx_set (inode_t *inode, xlator_t *this, dht_inode_ctx_t *ctx)
+{
+ int ret = -1;
+ uint64_t ctx_int = 0;
+
+ GF_VALIDATE_OR_GOTO ("dht", this, out);
+ GF_VALIDATE_OR_GOTO (this->name, inode, out);
+ GF_VALIDATE_OR_GOTO (this->name, ctx, out);
+
+ ctx_int = (long)ctx;
+ ret = inode_ctx_set (inode, this, &ctx_int);
+out:
+ return ret;
+}
diff --git a/xlators/cluster/dht/src/dht-inode-read.c b/xlators/cluster/dht/src/dht-inode-read.c
index 7352e47bd..ece84151a 100644
--- a/xlators/cluster/dht/src/dht-inode-read.c
+++ b/xlators/cluster/dht/src/dht-inode-read.c
@@ -1,20 +1,11 @@
/*
- Copyright (c) 2011 Gluster, Inc. <http://www.gluster.com>
+ Copyright (c) 2008-2012 Red Hat, Inc. <http://www.redhat.com>
This file is part of GlusterFS.
- GlusterFS is free software; you can redistribute it and/or modify
- it under the terms of the GNU General Public License as published
- by the Free Software Foundation; either version 3 of the License,
- or (at your option) any later version.
-
- GlusterFS is distributed in the hope that it will be useful, but
- WITHOUT ANY WARRANTY; without even the implied warranty of
- MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU
- General Public License for more details.
-
- You should have received a copy of the GNU General Public License
- along with this program. If not, see
- <http://www.gnu.org/licenses/>.
+ This file is licensed to you under your choice of the GNU Lesser
+ General Public License, version 3 or any later version (LGPLv3 or
+ later), or the GNU General Public License, version 2 (GPLv2), in all
+ cases as published by the Free Software Foundation.
*/
#ifndef _CONFIG_H
@@ -34,7 +25,7 @@ int dht_fsync2 (xlator_t *this, call_frame_t *frame, int ret);
int
dht_open_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
- int op_ret, int op_errno, fd_t *fd)
+ int op_ret, int op_errno, fd_t *fd, dict_t *xdata)
{
dht_local_t *local = NULL;
call_frame_t *prev = NULL;
@@ -61,7 +52,7 @@ dht_open_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
return 0;
out:
- DHT_STACK_UNWIND (open, frame, op_ret, op_errno, local->fd);
+ DHT_STACK_UNWIND (open, frame, op_ret, op_errno, local->fd, xdata);
return 0;
}
@@ -86,17 +77,17 @@ dht_open2 (xlator_t *this, call_frame_t *frame, int op_ret)
STACK_WIND (frame, dht_open_cbk, subvol, subvol->fops->open,
&local->loc, local->rebalance.flags, local->fd,
- local->rebalance.wbflags);
+ NULL);
return 0;
out:
- DHT_STACK_UNWIND (stat, frame, -1, op_errno, NULL);
+ DHT_STACK_UNWIND (stat, frame, -1, op_errno, NULL, NULL);
return 0;
}
int
dht_open (call_frame_t *frame, xlator_t *this,
- loc_t *loc, int flags, fd_t *fd, int wbflags)
+ loc_t *loc, int flags, fd_t *fd, dict_t *xdata)
{
xlator_t *subvol = NULL;
int op_errno = -1;
@@ -120,30 +111,30 @@ dht_open (call_frame_t *frame, xlator_t *this,
goto err;
}
- local->rebalance.wbflags = wbflags;
local->rebalance.flags = flags;
local->call_cnt = 1;
STACK_WIND (frame, dht_open_cbk, subvol, subvol->fops->open,
- loc, flags, fd, wbflags);
+ loc, flags, fd, xdata);
return 0;
err:
op_errno = (op_errno == -1) ? errno : op_errno;
- DHT_STACK_UNWIND (open, frame, -1, op_errno, NULL);
+ DHT_STACK_UNWIND (open, frame, -1, op_errno, NULL, NULL);
return 0;
}
int
dht_file_attr_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
- int op_ret, int op_errno, struct iatt *stbuf)
+ int op_ret, int op_errno, struct iatt *stbuf, dict_t *xdata)
{
- uint64_t tmp_subvol = 0;
+ xlator_t *subvol = 0;
dht_local_t *local = NULL;
call_frame_t *prev = NULL;
int ret = -1;
+ inode_t *inode = NULL;
GF_VALIDATE_OR_GOTO ("dht", frame, err);
GF_VALIDATE_OR_GOTO ("dht", this, out);
@@ -164,26 +155,28 @@ dht_file_attr_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
if (local->call_cnt != 1)
goto out;
+ local->op_errno = op_errno;
/* Check if the rebalance phase2 is true */
if ((op_ret == -1) || IS_DHT_MIGRATION_PHASE2 (stbuf)) {
- if (local->fd)
- ret = fd_ctx_get (local->fd, this, &tmp_subvol);
- if (ret) {
+ inode = (local->fd) ? local->fd->inode : local->loc.inode;
+ ret = dht_inode_ctx_get1 (this, inode, &subvol);
+ if (!subvol) {
/* Phase 2 of migration */
local->rebalance.target_op_fn = dht_attr2;
ret = dht_rebalance_complete_check (this, frame);
+ if (!ret)
+ return 0;
} else {
/* value is already set in fd_ctx, that means no need
to check for whether its complete or not. */
dht_attr2 (this, frame, 0);
- }
- if (!ret)
return 0;
+ }
}
out:
DHT_STRIP_PHASE1_FLAGS (stbuf);
- DHT_STACK_UNWIND (stat, frame, op_ret, op_errno, stbuf);
+ DHT_STACK_UNWIND (stat, frame, op_ret, op_errno, stbuf, xdata);
err:
return 0;
}
@@ -208,21 +201,21 @@ dht_attr2 (xlator_t *this, call_frame_t *frame, int op_ret)
if (local->fop == GF_FOP_FSTAT) {
STACK_WIND (frame, dht_file_attr_cbk, subvol,
- subvol->fops->fstat, local->fd);
+ subvol->fops->fstat, local->fd, NULL);
} else {
STACK_WIND (frame, dht_file_attr_cbk, subvol,
- subvol->fops->stat, &local->loc);
+ subvol->fops->stat, &local->loc, NULL);
}
return 0;
out:
- DHT_STACK_UNWIND (stat, frame, -1, op_errno, NULL);
+ DHT_STACK_UNWIND (stat, frame, -1, op_errno, NULL, NULL);
return 0;
}
int
dht_attr_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
- int op_ret, int op_errno, struct iatt *stbuf)
+ int op_ret, int op_errno, struct iatt *stbuf, dict_t *xdata)
{
dht_local_t *local = NULL;
int this_call_cnt = 0;
@@ -257,14 +250,14 @@ out:
this_call_cnt = dht_frame_return (frame);
if (is_last_call (this_call_cnt)) {
DHT_STACK_UNWIND (stat, frame, local->op_ret, local->op_errno,
- &local->stbuf);
+ &local->stbuf, xdata);
}
err:
return 0;
}
int
-dht_stat (call_frame_t *frame, xlator_t *this, loc_t *loc)
+dht_stat (call_frame_t *frame, xlator_t *this, loc_t *loc, dict_t *xdata)
{
xlator_t *subvol = NULL;
int op_errno = -1;
@@ -300,7 +293,7 @@ dht_stat (call_frame_t *frame, xlator_t *this, loc_t *loc)
subvol = local->cached_subvol;
STACK_WIND (frame, dht_file_attr_cbk, subvol,
- subvol->fops->stat, loc);
+ subvol->fops->stat, loc, xdata);
return 0;
}
@@ -312,21 +305,21 @@ dht_stat (call_frame_t *frame, xlator_t *this, loc_t *loc)
STACK_WIND (frame, dht_attr_cbk,
subvol, subvol->fops->stat,
- loc);
+ loc, xdata);
}
return 0;
err:
op_errno = (op_errno == -1) ? errno : op_errno;
- DHT_STACK_UNWIND (stat, frame, -1, op_errno, NULL);
+ DHT_STACK_UNWIND (stat, frame, -1, op_errno, NULL, NULL);
return 0;
}
int
-dht_fstat (call_frame_t *frame, xlator_t *this, fd_t *fd)
+dht_fstat (call_frame_t *frame, xlator_t *this, fd_t *fd, dict_t *xdata)
{
xlator_t *subvol = NULL;
int op_errno = -1;
@@ -360,7 +353,7 @@ dht_fstat (call_frame_t *frame, xlator_t *this, fd_t *fd)
subvol = local->cached_subvol;
STACK_WIND (frame, dht_file_attr_cbk, subvol,
- subvol->fops->fstat, fd);
+ subvol->fops->fstat, fd, xdata);
return 0;
}
@@ -371,14 +364,14 @@ dht_fstat (call_frame_t *frame, xlator_t *this, fd_t *fd)
subvol = layout->list[i].xlator;
STACK_WIND (frame, dht_attr_cbk,
subvol, subvol->fops->fstat,
- fd);
+ fd, xdata);
}
return 0;
err:
op_errno = (op_errno == -1) ? errno : op_errno;
- DHT_STACK_UNWIND (fstat, frame, -1, op_errno, NULL);
+ DHT_STACK_UNWIND (fstat, frame, -1, op_errno, NULL, NULL);
return 0;
}
@@ -387,10 +380,12 @@ int
dht_readv_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
int op_ret, int op_errno,
struct iovec *vector, int count, struct iatt *stbuf,
- struct iobref *iobref)
+ struct iobref *iobref, dict_t *xdata)
{
dht_local_t *local = NULL;
int ret = 0;
+ inode_t *inode = NULL;
+ xlator_t *subvol = 0;
local = frame->local;
if (!local) {
@@ -406,25 +401,27 @@ dht_readv_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
if ((op_ret == -1) && (op_errno != ENOENT))
goto out;
+ local->op_errno = op_errno;
if ((op_ret == -1) || IS_DHT_MIGRATION_PHASE2 (stbuf)) {
/* File would be migrated to other node */
- ret = fd_ctx_get (local->fd, this, NULL);
- if (ret) {
+ ret = dht_inode_ctx_get1 (this, inode, &subvol);
+ if (!subvol) {
local->rebalance.target_op_fn = dht_readv2;
ret = dht_rebalance_complete_check (this, frame);
+ if (!ret)
+ return 0;
} else {
/* value is already set in fd_ctx, that means no need
to check for whether its complete or not. */
dht_readv2 (this, frame, 0);
- }
- if (!ret)
return 0;
+ }
}
out:
DHT_STRIP_PHASE1_FLAGS (stbuf);
DHT_STACK_UNWIND (readv, frame, op_ret, op_errno, vector, count, stbuf,
- iobref);
+ iobref, xdata);
return 0;
}
@@ -448,18 +445,19 @@ dht_readv2 (xlator_t *this, call_frame_t *frame, int op_ret)
subvol = local->cached_subvol;
STACK_WIND (frame, dht_readv_cbk, subvol, subvol->fops->readv,
- local->fd, local->rebalance.size, local->rebalance.offset);
+ local->fd, local->rebalance.size, local->rebalance.offset,
+ local->rebalance.flags, NULL);
return 0;
out:
- DHT_STACK_UNWIND (readv, frame, -1, op_errno, NULL, 0, NULL, NULL);
+ DHT_STACK_UNWIND (readv, frame, -1, op_errno, NULL, 0, NULL, NULL, NULL);
return 0;
}
int
dht_readv (call_frame_t *frame, xlator_t *this,
- fd_t *fd, size_t size, off_t off)
+ fd_t *fd, size_t size, off_t off, uint32_t flags, dict_t *xdata)
{
xlator_t *subvol = NULL;
int op_errno = -1;
@@ -485,35 +483,57 @@ dht_readv (call_frame_t *frame, xlator_t *this,
local->rebalance.offset = off;
local->rebalance.size = size;
+ local->rebalance.flags = flags;
local->call_cnt = 1;
STACK_WIND (frame, dht_readv_cbk,
subvol, subvol->fops->readv,
- fd, size, off);
+ fd, size, off, flags, xdata);
return 0;
err:
op_errno = (op_errno == -1) ? errno : op_errno;
- DHT_STACK_UNWIND (readv, frame, -1, op_errno, NULL, 0, NULL, NULL);
+ DHT_STACK_UNWIND (readv, frame, -1, op_errno, NULL, 0, NULL, NULL, NULL);
return 0;
}
int
dht_access_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
- int op_ret, int op_errno)
+ int op_ret, int op_errno, dict_t *xdata)
{
int ret = -1;
dht_local_t *local = NULL;
+ xlator_t *subvol = NULL;
+ call_frame_t *prev = NULL;
local = frame->local;
+ prev = cookie;
+ if (!prev || !prev->this)
+ goto out;
if (local->call_cnt != 1)
goto out;
+ if ((op_ret == -1) && (op_errno == ENOTCONN) &&
+ IA_ISDIR(local->loc.inode->ia_type)) {
+
+ subvol = dht_subvol_next_available (this, prev->this);
+ if (!subvol)
+ goto out;
+ /* check if we are done with visiting every node */
+ if (subvol == local->cached_subvol) {
+ goto out;
+ }
+
+ STACK_WIND (frame, dht_access_cbk, subvol, subvol->fops->access,
+ &local->loc, local->rebalance.flags, NULL);
+ return 0;
+ }
if ((op_ret == -1) && (op_errno == ENOENT)) {
/* File would be migrated to other node */
+ local->op_errno = op_errno;
local->rebalance.target_op_fn = dht_access2;
ret = dht_rebalance_complete_check (frame->this, frame);
if (!ret)
@@ -521,7 +541,7 @@ dht_access_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
}
out:
- DHT_STACK_UNWIND (access, frame, op_ret, op_errno);
+ DHT_STACK_UNWIND (access, frame, op_ret, op_errno, xdata);
return 0;
}
@@ -544,18 +564,19 @@ dht_access2 (xlator_t *this, call_frame_t *frame, int op_ret)
subvol = local->cached_subvol;
STACK_WIND (frame, dht_access_cbk, subvol, subvol->fops->access,
- &local->loc, local->rebalance.flags);
+ &local->loc, local->rebalance.flags, NULL);
return 0;
out:
- DHT_STACK_UNWIND (access, frame, -1, op_errno);
+ DHT_STACK_UNWIND (access, frame, -1, op_errno, NULL);
return 0;
}
int
-dht_access (call_frame_t *frame, xlator_t *this, loc_t *loc, int32_t mask)
+dht_access (call_frame_t *frame, xlator_t *this, loc_t *loc, int32_t mask,
+ dict_t *xdata)
{
xlator_t *subvol = NULL;
int op_errno = -1;
@@ -584,13 +605,13 @@ dht_access (call_frame_t *frame, xlator_t *this, loc_t *loc, int32_t mask)
}
STACK_WIND (frame, dht_access_cbk, subvol, subvol->fops->access,
- loc, mask);
+ loc, mask, xdata);
return 0;
err:
op_errno = (op_errno == -1) ? errno : op_errno;
- DHT_STACK_UNWIND (access, frame, -1, op_errno);
+ DHT_STACK_UNWIND (access, frame, -1, op_errno, NULL);
return 0;
}
@@ -598,10 +619,11 @@ err:
int
dht_flush_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
- int op_ret, int op_errno)
+ int op_ret, int op_errno, dict_t *xdata)
{
- dht_local_t *local = NULL;
- int ret = -1;
+ dht_local_t *local = NULL;
+ inode_t *inode = NULL;
+ xlator_t *subvol = 0;
local = frame->local;
@@ -611,14 +633,14 @@ dht_flush_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
goto out;
/* If context is set, then send flush() it to the destination */
- ret = fd_ctx_get (local->fd, this, NULL);
- if (!ret) {
+ dht_inode_ctx_get1 (this, inode, &subvol);
+ if (subvol) {
dht_flush2 (this, frame, 0);
return 0;
}
out:
- DHT_STACK_UNWIND (flush, frame, op_ret, op_errno);
+ DHT_STACK_UNWIND (flush, frame, op_ret, op_errno, xdata);
return 0;
}
@@ -628,14 +650,10 @@ dht_flush2 (xlator_t *this, call_frame_t *frame, int op_ret)
{
dht_local_t *local = NULL;
xlator_t *subvol = NULL;
- uint64_t tmp_subvol = 0;
- int ret = -1;
local = frame->local;
- ret = fd_ctx_get (local->fd, this, &tmp_subvol);
- if (!ret)
- subvol = (xlator_t *)(long)tmp_subvol;
+ dht_inode_ctx_get1 (this, local->fd->inode, &subvol);
if (!subvol)
subvol = local->cached_subvol;
@@ -643,20 +661,19 @@ dht_flush2 (xlator_t *this, call_frame_t *frame, int op_ret)
local->call_cnt = 2; /* This is the second attempt */
STACK_WIND (frame, dht_flush_cbk,
- subvol, subvol->fops->flush, local->fd);
+ subvol, subvol->fops->flush, local->fd, NULL);
return 0;
}
int
-dht_flush (call_frame_t *frame, xlator_t *this, fd_t *fd)
+dht_flush (call_frame_t *frame, xlator_t *this, fd_t *fd, dict_t *xdata)
{
xlator_t *subvol = NULL;
int op_errno = -1;
dht_local_t *local = NULL;
-
VALIDATE_OR_GOTO (frame, err);
VALIDATE_OR_GOTO (this, err);
VALIDATE_OR_GOTO (fd, err);
@@ -678,13 +695,13 @@ dht_flush (call_frame_t *frame, xlator_t *this, fd_t *fd)
local->call_cnt = 1;
STACK_WIND (frame, dht_flush_cbk,
- subvol, subvol->fops->flush, fd);
+ subvol, subvol->fops->flush, fd, xdata);
return 0;
err:
op_errno = (op_errno == -1) ? errno : op_errno;
- DHT_STACK_UNWIND (flush, frame, -1, op_errno);
+ DHT_STACK_UNWIND (flush, frame, -1, op_errno, NULL);
return 0;
}
@@ -692,17 +709,20 @@ err:
int
dht_fsync_cbk (call_frame_t *frame, void *cookie, xlator_t *this, int op_ret,
- int op_errno, struct iatt *prebuf, struct iatt *postbuf)
+ int op_errno, struct iatt *prebuf, struct iatt *postbuf,
+ dict_t *xdata)
{
dht_local_t *local = NULL;
call_frame_t *prev = NULL;
int ret = -1;
+ inode_t *inode = NULL;
+ xlator_t *subvol = 0;
local = frame->local;
prev = cookie;
local->op_errno = op_errno;
- if (op_ret == -1) {
+ if (op_ret == -1 && (op_errno != ENOENT)) {
gf_log (this->name, GF_LOG_DEBUG,
"subvolume %s returned -1 (%s)",
prev->this->name, strerror (op_errno));
@@ -717,8 +737,9 @@ dht_fsync_cbk (call_frame_t *frame, void *cookie, xlator_t *this, int op_ret,
goto out;
}
- ret = fd_ctx_get (local->fd, this, NULL);
- if (ret) {
+ local->op_errno = op_errno;
+ dht_inode_ctx_get1 (this, inode, &subvol);
+ if (!subvol) {
local->rebalance.target_op_fn = dht_fsync2;
/* Check if the rebalance phase1 is true */
@@ -733,17 +754,18 @@ dht_fsync_cbk (call_frame_t *frame, void *cookie, xlator_t *this, int op_ret,
if (IS_DHT_MIGRATION_PHASE2 (postbuf)) {
ret = dht_rebalance_complete_check (this, frame);
}
+ if (!ret)
+ return 0;
} else {
dht_fsync2 (this, frame, 0);
- }
- if (!ret)
return 0;
+ }
out:
DHT_STRIP_PHASE1_FLAGS (postbuf);
DHT_STRIP_PHASE1_FLAGS (prebuf);
DHT_STACK_UNWIND (fsync, frame, op_ret, op_errno,
- prebuf, postbuf);
+ prebuf, postbuf, xdata);
return 0;
}
@@ -753,34 +775,29 @@ dht_fsync2 (xlator_t *this, call_frame_t *frame, int op_ret)
{
dht_local_t *local = NULL;
xlator_t *subvol = NULL;
- uint64_t tmp_subvol = 0;
- int ret = -1;
local = frame->local;
- ret = fd_ctx_get (local->fd, this, &tmp_subvol);
- if (!ret)
- subvol = (xlator_t *)(long)tmp_subvol;
-
+ dht_inode_ctx_get1 (this, local->fd->inode, &subvol);
if (!subvol)
subvol = local->cached_subvol;
local->call_cnt = 2; /* This is the second attempt */
STACK_WIND (frame, dht_fsync_cbk, subvol, subvol->fops->fsync,
- local->fd, local->rebalance.flags);
+ local->fd, local->rebalance.flags, NULL);
return 0;
}
int
-dht_fsync (call_frame_t *frame, xlator_t *this, fd_t *fd, int datasync)
+dht_fsync (call_frame_t *frame, xlator_t *this, fd_t *fd, int datasync,
+ dict_t *xdata)
{
xlator_t *subvol = NULL;
int op_errno = -1;
dht_local_t *local = NULL;
-
VALIDATE_OR_GOTO (frame, err);
VALIDATE_OR_GOTO (this, err);
VALIDATE_OR_GOTO (fd, err);
@@ -798,13 +815,13 @@ dht_fsync (call_frame_t *frame, xlator_t *this, fd_t *fd, int datasync)
subvol = local->cached_subvol;
STACK_WIND (frame, dht_fsync_cbk, subvol, subvol->fops->fsync,
- fd, datasync);
+ fd, datasync, xdata);
return 0;
err:
op_errno = (op_errno == -1) ? errno : op_errno;
- DHT_STACK_UNWIND (fsync, frame, -1, op_errno, NULL, NULL);
+ DHT_STACK_UNWIND (fsync, frame, -1, op_errno, NULL, NULL, NULL);
return 0;
}
@@ -815,9 +832,9 @@ err:
phase 2 of migration */
int
dht_lk_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
- int op_ret, int op_errno, struct gf_flock *flock)
+ int op_ret, int op_errno, struct gf_flock *flock, dict_t *xdata)
{
- DHT_STACK_UNWIND (lk, frame, op_ret, op_errno, flock);
+ DHT_STACK_UNWIND (lk, frame, op_ret, op_errno, flock, xdata);
return 0;
}
@@ -825,7 +842,7 @@ dht_lk_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
int
dht_lk (call_frame_t *frame, xlator_t *this,
- fd_t *fd, int cmd, struct gf_flock *flock)
+ fd_t *fd, int cmd, struct gf_flock *flock, dict_t *xdata)
{
xlator_t *subvol = NULL;
int op_errno = -1;
@@ -845,13 +862,13 @@ dht_lk (call_frame_t *frame, xlator_t *this,
/* TODO: for rebalance, we need to preserve the fop arguments */
STACK_WIND (frame, dht_lk_cbk, subvol, subvol->fops->lk, fd,
- cmd, flock);
+ cmd, flock, xdata);
return 0;
err:
op_errno = (op_errno == -1) ? errno : op_errno;
- DHT_STACK_UNWIND (lk, frame, -1, op_errno, NULL);
+ DHT_STACK_UNWIND (lk, frame, -1, op_errno, NULL, NULL);
return 0;
}
@@ -859,7 +876,8 @@ err:
/* Symlinks are currently not migrated, so no need for any check here */
int
dht_readlink_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
- int op_ret, int op_errno, const char *path, struct iatt *stbuf)
+ int op_ret, int op_errno, const char *path,
+ struct iatt *stbuf, dict_t *xdata)
{
dht_local_t *local = NULL;
@@ -874,14 +892,15 @@ dht_readlink_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
err:
DHT_STRIP_PHASE1_FLAGS (stbuf);
- DHT_STACK_UNWIND (readlink, frame, op_ret, op_errno, path, stbuf);
+ DHT_STACK_UNWIND (readlink, frame, op_ret, op_errno, path, stbuf, xdata);
return 0;
}
int
-dht_readlink (call_frame_t *frame, xlator_t *this, loc_t *loc, size_t size)
+dht_readlink (call_frame_t *frame, xlator_t *this, loc_t *loc, size_t size,
+ dict_t *xdata)
{
xlator_t *subvol = NULL;
int op_errno = -1;
@@ -909,13 +928,13 @@ dht_readlink (call_frame_t *frame, xlator_t *this, loc_t *loc, size_t size)
STACK_WIND (frame, dht_readlink_cbk,
subvol, subvol->fops->readlink,
- loc, size);
+ loc, size, xdata);
return 0;
err:
op_errno = (op_errno == -1) ? errno : op_errno;
- DHT_STACK_UNWIND (readlink, frame, -1, op_errno, NULL, NULL);
+ DHT_STACK_UNWIND (readlink, frame, -1, op_errno, NULL, NULL, NULL);
return 0;
}
@@ -926,16 +945,16 @@ err:
int
dht_xattrop_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
- int32_t op_ret, int32_t op_errno, dict_t *dict)
+ int32_t op_ret, int32_t op_errno, dict_t *dict, dict_t *xdata)
{
- DHT_STACK_UNWIND (xattrop, frame, op_ret, op_errno, dict);
+ DHT_STACK_UNWIND (xattrop, frame, op_ret, op_errno, dict, xdata);
return 0;
}
int
dht_xattrop (call_frame_t *frame, xlator_t *this, loc_t *loc,
- gf_xattrop_flags_t flags, dict_t *dict)
+ gf_xattrop_flags_t flags, dict_t *dict, dict_t *xdata)
{
xlator_t *subvol = NULL;
int op_errno = -1;
@@ -966,13 +985,13 @@ dht_xattrop (call_frame_t *frame, xlator_t *this, loc_t *loc,
STACK_WIND (frame,
dht_xattrop_cbk,
subvol, subvol->fops->xattrop,
- loc, flags, dict);
+ loc, flags, dict, xdata);
return 0;
err:
op_errno = (op_errno == -1) ? errno : op_errno;
- DHT_STACK_UNWIND (xattrop, frame, -1, op_errno, NULL);
+ DHT_STACK_UNWIND (xattrop, frame, -1, op_errno, NULL, NULL);
return 0;
}
@@ -980,16 +999,16 @@ err:
int
dht_fxattrop_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
- int32_t op_ret, int32_t op_errno, dict_t *dict)
+ int32_t op_ret, int32_t op_errno, dict_t *dict, dict_t *xdata)
{
- DHT_STACK_UNWIND (fxattrop, frame, op_ret, op_errno, dict);
+ DHT_STACK_UNWIND (fxattrop, frame, op_ret, op_errno, dict, xdata);
return 0;
}
int
dht_fxattrop (call_frame_t *frame, xlator_t *this,
- fd_t *fd, gf_xattrop_flags_t flags, dict_t *dict)
+ fd_t *fd, gf_xattrop_flags_t flags, dict_t *dict, dict_t *xdata)
{
xlator_t *subvol = NULL;
int op_errno = -1;
@@ -1009,13 +1028,13 @@ dht_fxattrop (call_frame_t *frame, xlator_t *this,
STACK_WIND (frame,
dht_fxattrop_cbk,
subvol, subvol->fops->fxattrop,
- fd, flags, dict);
+ fd, flags, dict, xdata);
return 0;
err:
op_errno = (op_errno == -1) ? errno : op_errno;
- DHT_STACK_UNWIND (fxattrop, frame, -1, op_errno, NULL);
+ DHT_STACK_UNWIND (fxattrop, frame, -1, op_errno, NULL, NULL);
return 0;
}
@@ -1023,17 +1042,17 @@ err:
int
dht_inodelk_cbk (call_frame_t *frame, void *cookie,
- xlator_t *this, int32_t op_ret, int32_t op_errno)
+ xlator_t *this, int32_t op_ret, int32_t op_errno, dict_t *xdata)
{
- DHT_STACK_UNWIND (inodelk, frame, op_ret, op_errno);
+ DHT_STACK_UNWIND (inodelk, frame, op_ret, op_errno, xdata);
return 0;
}
int32_t
-dht_inodelk (call_frame_t *frame, xlator_t *this,
- const char *volume, loc_t *loc, int32_t cmd, struct gf_flock *lock)
+dht_inodelk (call_frame_t *frame, xlator_t *this, const char *volume,
+ loc_t *loc, int32_t cmd, struct gf_flock *lock, dict_t *xdata)
{
xlator_t *subvol = NULL;
int op_errno = -1;
@@ -1065,31 +1084,31 @@ dht_inodelk (call_frame_t *frame, xlator_t *this,
STACK_WIND (frame,
dht_inodelk_cbk,
subvol, subvol->fops->inodelk,
- volume, loc, cmd, lock);
+ volume, loc, cmd, lock, xdata);
return 0;
err:
op_errno = (op_errno == -1) ? errno : op_errno;
- DHT_STACK_UNWIND (inodelk, frame, -1, op_errno);
+ DHT_STACK_UNWIND (inodelk, frame, -1, op_errno, NULL);
return 0;
}
int
-dht_finodelk_cbk (call_frame_t *frame, void *cookie,
- xlator_t *this, int32_t op_ret, int32_t op_errno)
+dht_finodelk_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
+ int32_t op_ret, int32_t op_errno, dict_t *xdata)
{
- DHT_STACK_UNWIND (finodelk, frame, op_ret, op_errno);
+ DHT_STACK_UNWIND (finodelk, frame, op_ret, op_errno, xdata);
return 0;
}
int
-dht_finodelk (call_frame_t *frame, xlator_t *this,
- const char *volume, fd_t *fd, int32_t cmd, struct gf_flock *lock)
+dht_finodelk (call_frame_t *frame, xlator_t *this, const char *volume,
+ fd_t *fd, int32_t cmd, struct gf_flock *lock, dict_t *xdata)
{
xlator_t *subvol = NULL;
int op_errno = -1;
@@ -1107,16 +1126,14 @@ dht_finodelk (call_frame_t *frame, xlator_t *this,
}
- STACK_WIND (frame,
- dht_finodelk_cbk,
- subvol, subvol->fops->finodelk,
- volume, fd, cmd, lock);
+ STACK_WIND (frame, dht_finodelk_cbk, subvol, subvol->fops->finodelk,
+ volume, fd, cmd, lock, xdata);
return 0;
err:
op_errno = (op_errno == -1) ? errno : op_errno;
- DHT_STACK_UNWIND (finodelk, frame, -1, op_errno);
+ DHT_STACK_UNWIND (finodelk, frame, -1, op_errno, NULL);
return 0;
}
diff --git a/xlators/cluster/dht/src/dht-inode-write.c b/xlators/cluster/dht/src/dht-inode-write.c
index 4f65b4671..4b3f3a049 100644
--- a/xlators/cluster/dht/src/dht-inode-write.c
+++ b/xlators/cluster/dht/src/dht-inode-write.c
@@ -1,20 +1,11 @@
/*
- Copyright (c) 2011 Gluster, Inc. <http://www.gluster.com>
+ Copyright (c) 2008-2012 Red Hat, Inc. <http://www.redhat.com>
This file is part of GlusterFS.
- GlusterFS is free software; you can redistribute it and/or modify
- it under the terms of the GNU General Public License as published
- by the Free Software Foundation; either version 3 of the License,
- or (at your option) any later version.
-
- GlusterFS is distributed in the hope that it will be useful, but
- WITHOUT ANY WARRANTY; without even the implied warranty of
- MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU
- General Public License for more details.
-
- You should have received a copy of the GNU General Public License
- along with this program. If not, see
- <http://www.gnu.org/licenses/>.
+ This file is licensed to you under your choice of the GNU Lesser
+ General Public License, version 3 or any later version (LGPLv3 or
+ later), or the GNU General Public License, version 2 (GPLv2), in all
+ cases as published by the Free Software Foundation.
*/
@@ -28,16 +19,20 @@
int dht_writev2 (xlator_t *this, call_frame_t *frame, int ret);
int dht_truncate2 (xlator_t *this, call_frame_t *frame, int ret);
int dht_setattr2 (xlator_t *this, call_frame_t *frame, int ret);
+int dht_fallocate2(xlator_t *this, call_frame_t *frame, int op_ret);
+int dht_discard2(xlator_t *this, call_frame_t *frame, int op_ret);
+int dht_zerofill2(xlator_t *this, call_frame_t *frame, int op_ret);
int
dht_writev_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
int op_ret, int op_errno, struct iatt *prebuf,
- struct iatt *postbuf)
+ struct iatt *postbuf, dict_t *xdata)
{
dht_local_t *local = NULL;
int ret = -1;
+ xlator_t *subvol = NULL;
- if (op_ret == -1) {
+ if (op_ret == -1 && (op_errno != ENOENT)) {
goto out;
}
@@ -59,6 +54,7 @@ dht_writev_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
local->rebalance.target_op_fn = dht_writev2;
+ local->op_errno = op_errno;
/* Phase 2 of migration */
if (IS_DHT_MIGRATION_PHASE2 (postbuf)) {
ret = dht_rebalance_complete_check (this, frame);
@@ -71,8 +67,8 @@ dht_writev_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
dht_iatt_merge (this, &local->stbuf, postbuf, NULL);
dht_iatt_merge (this, &local->prebuf, prebuf, NULL);
- ret = fd_ctx_get (local->fd, this, NULL);
- if (!ret) {
+ ret = dht_inode_ctx_get1 (this, local->fd->inode, &subvol);
+ if (subvol) {
dht_writev2 (this, frame, 0);
return 0;
}
@@ -85,7 +81,8 @@ out:
DHT_STRIP_PHASE1_FLAGS (postbuf);
DHT_STRIP_PHASE1_FLAGS (prebuf);
- DHT_STACK_UNWIND (writev, frame, op_ret, op_errno, prebuf, postbuf);
+ DHT_STACK_UNWIND (writev, frame, op_ret, op_errno, prebuf, postbuf,
+ xdata);
return 0;
}
@@ -95,14 +92,10 @@ dht_writev2 (xlator_t *this, call_frame_t *frame, int op_ret)
{
dht_local_t *local = NULL;
xlator_t *subvol = NULL;
- uint64_t tmp_subvol = 0;
- int ret = -1;
local = frame->local;
- ret = fd_ctx_get (local->fd, this, &tmp_subvol);
- if (!ret)
- subvol = (xlator_t *)(long)tmp_subvol;
+ dht_inode_ctx_get1 (this, local->fd->inode, &subvol);
if (!subvol)
subvol = local->cached_subvol;
@@ -112,15 +105,16 @@ dht_writev2 (xlator_t *this, call_frame_t *frame, int op_ret)
STACK_WIND (frame, dht_writev_cbk,
subvol, subvol->fops->writev,
local->fd, local->rebalance.vector, local->rebalance.count,
- local->rebalance.offset, local->rebalance.iobref);
+ local->rebalance.offset, local->rebalance.flags,
+ local->rebalance.iobref, NULL);
return 0;
}
int
-dht_writev (call_frame_t *frame, xlator_t *this,
- fd_t *fd, struct iovec *vector, int count, off_t off,
- struct iobref *iobref)
+dht_writev (call_frame_t *frame, xlator_t *this, fd_t *fd,
+ struct iovec *vector, int count, off_t off, uint32_t flags,
+ struct iobref *iobref, dict_t *xdata)
{
xlator_t *subvol = NULL;
int op_errno = -1;
@@ -149,18 +143,19 @@ dht_writev (call_frame_t *frame, xlator_t *this,
local->rebalance.vector = iov_dup (vector, count);
local->rebalance.offset = off;
local->rebalance.count = count;
+ local->rebalance.flags = flags;
local->rebalance.iobref = iobref_ref (iobref);
local->call_cnt = 1;
STACK_WIND (frame, dht_writev_cbk,
subvol, subvol->fops->writev,
- fd, vector, count, off, iobref);
+ fd, vector, count, off, flags, iobref, xdata);
return 0;
err:
op_errno = (op_errno == -1) ? errno : op_errno;
- DHT_STACK_UNWIND (writev, frame, -1, op_errno, NULL, NULL);
+ DHT_STACK_UNWIND (writev, frame, -1, op_errno, NULL, NULL, NULL);
return 0;
}
@@ -170,11 +165,13 @@ err:
int
dht_truncate_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
int op_ret, int op_errno, struct iatt *prebuf,
- struct iatt *postbuf)
+ struct iatt *postbuf, dict_t *xdata)
{
dht_local_t *local = NULL;
call_frame_t *prev = NULL;
int ret = -1;
+ xlator_t *subvol = NULL;
+ inode_t *inode = NULL;
GF_VALIDATE_OR_GOTO ("dht", frame, err);
GF_VALIDATE_OR_GOTO ("dht", this, out);
@@ -204,6 +201,7 @@ dht_truncate_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
local->rebalance.target_op_fn = dht_truncate2;
+ local->op_errno = op_errno;
/* Phase 2 of migration */
if ((op_ret == -1) || IS_DHT_MIGRATION_PHASE2 (postbuf)) {
ret = dht_rebalance_complete_check (this, frame);
@@ -215,8 +213,9 @@ dht_truncate_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
if (IS_DHT_MIGRATION_PHASE1 (postbuf)) {
dht_iatt_merge (this, &local->stbuf, postbuf, NULL);
dht_iatt_merge (this, &local->prebuf, prebuf, NULL);
- ret = fd_ctx_get (local->fd, this, NULL);
- if (!ret) {
+ inode = (local->fd) ? local->fd->inode : local->loc.inode;
+ dht_inode_ctx_get1 (this, inode, &subvol);
+ if (subvol) {
dht_truncate2 (this, frame, 0);
return 0;
}
@@ -229,7 +228,7 @@ out:
DHT_STRIP_PHASE1_FLAGS (postbuf);
DHT_STRIP_PHASE1_FLAGS (prebuf);
DHT_STACK_UNWIND (truncate, frame, op_ret, op_errno,
- prebuf, postbuf);
+ prebuf, postbuf, xdata);
err:
return 0;
}
@@ -240,16 +239,13 @@ dht_truncate2 (xlator_t *this, call_frame_t *frame, int op_ret)
{
dht_local_t *local = NULL;
xlator_t *subvol = NULL;
- uint64_t tmp_subvol = 0;
- int ret = -1;
+ inode_t *inode = NULL;
local = frame->local;
- if (local->fd)
- ret = fd_ctx_get (local->fd, this, &tmp_subvol);
- if (!ret)
- subvol = (xlator_t *)(long)tmp_subvol;
+ inode = local->fd ? local->fd->inode : local->loc.inode;
+ dht_inode_ctx_get1 (this, inode, &subvol);
if (!subvol)
subvol = local->cached_subvol;
@@ -258,18 +254,19 @@ dht_truncate2 (xlator_t *this, call_frame_t *frame, int op_ret)
if (local->fop == GF_FOP_TRUNCATE) {
STACK_WIND (frame, dht_truncate_cbk, subvol,
subvol->fops->truncate, &local->loc,
- local->rebalance.offset);
+ local->rebalance.offset, NULL);
} else {
STACK_WIND (frame, dht_truncate_cbk, subvol,
subvol->fops->ftruncate, local->fd,
- local->rebalance.offset);
+ local->rebalance.offset, NULL);
}
return 0;
}
int
-dht_truncate (call_frame_t *frame, xlator_t *this, loc_t *loc, off_t offset)
+dht_truncate (call_frame_t *frame, xlator_t *this, loc_t *loc, off_t offset,
+ dict_t *xdata)
{
xlator_t *subvol = NULL;
int op_errno = -1;
@@ -299,19 +296,20 @@ dht_truncate (call_frame_t *frame, xlator_t *this, loc_t *loc, off_t offset)
STACK_WIND (frame, dht_truncate_cbk,
subvol, subvol->fops->truncate,
- loc, offset);
+ loc, offset, xdata);
return 0;
err:
op_errno = (op_errno == -1) ? errno : op_errno;
- DHT_STACK_UNWIND (truncate, frame, -1, op_errno, NULL, NULL);
+ DHT_STACK_UNWIND (truncate, frame, -1, op_errno, NULL, NULL, NULL);
return 0;
}
int
-dht_ftruncate (call_frame_t *frame, xlator_t *this, fd_t *fd, off_t offset)
+dht_ftruncate (call_frame_t *frame, xlator_t *this, fd_t *fd, off_t offset,
+ dict_t *xdata)
{
xlator_t *subvol = NULL;
int op_errno = -1;
@@ -339,22 +337,423 @@ dht_ftruncate (call_frame_t *frame, xlator_t *this, fd_t *fd, off_t offset)
STACK_WIND (frame, dht_truncate_cbk,
subvol, subvol->fops->ftruncate,
- fd, offset);
+ fd, offset, xdata);
return 0;
err:
op_errno = (op_errno == -1) ? errno : op_errno;
- DHT_STACK_UNWIND (ftruncate, frame, -1, op_errno, NULL, NULL);
+ DHT_STACK_UNWIND (ftruncate, frame, -1, op_errno, NULL, NULL, NULL);
return 0;
}
+
+int
+dht_fallocate_cbk(call_frame_t *frame, void *cookie, xlator_t *this,
+ int op_ret, int op_errno, struct iatt *prebuf,
+ struct iatt *postbuf, dict_t *xdata)
+{
+ dht_local_t *local = NULL;
+ call_frame_t *prev = NULL;
+ int ret = -1;
+ xlator_t *subvol = NULL;
+
+ GF_VALIDATE_OR_GOTO ("dht", frame, err);
+ GF_VALIDATE_OR_GOTO ("dht", this, out);
+ GF_VALIDATE_OR_GOTO ("dht", frame->local, out);
+ GF_VALIDATE_OR_GOTO ("dht", cookie, out);
+
+ local = frame->local;
+ prev = cookie;
+
+ if ((op_ret == -1) && (op_errno != ENOENT)) {
+ local->op_errno = op_errno;
+ local->op_ret = -1;
+ gf_log (this->name, GF_LOG_DEBUG,
+ "subvolume %s returned -1 (%s)",
+ prev->this->name, strerror (op_errno));
+
+ goto out;
+ }
+
+ if (local->call_cnt != 1) {
+ if (local->stbuf.ia_blocks) {
+ dht_iatt_merge (this, postbuf, &local->stbuf, NULL);
+ dht_iatt_merge (this, prebuf, &local->prebuf, NULL);
+ }
+ goto out;
+ }
+ local->rebalance.target_op_fn = dht_fallocate2;
+
+ /* Phase 2 of migration */
+ if ((op_ret == -1) || IS_DHT_MIGRATION_PHASE2 (postbuf)) {
+ ret = dht_rebalance_complete_check (this, frame);
+ if (!ret)
+ return 0;
+ }
+
+ /* Check if the rebalance phase1 is true */
+ if (IS_DHT_MIGRATION_PHASE1 (postbuf)) {
+ dht_iatt_merge (this, &local->stbuf, postbuf, NULL);
+ dht_iatt_merge (this, &local->prebuf, prebuf, NULL);
+ dht_inode_ctx_get1 (this, local->fd->inode, &subvol);
+ if (subvol) {
+ dht_fallocate2 (this, frame, 0);
+ return 0;
+ }
+ ret = dht_rebalance_in_progress_check (this, frame);
+ if (!ret)
+ return 0;
+ }
+
+out:
+ DHT_STRIP_PHASE1_FLAGS (postbuf);
+ DHT_STRIP_PHASE1_FLAGS (prebuf);
+ DHT_STACK_UNWIND (fallocate, frame, op_ret, op_errno,
+ prebuf, postbuf, xdata);
+err:
+ return 0;
+}
+
+int
+dht_fallocate2(xlator_t *this, call_frame_t *frame, int op_ret)
+{
+ dht_local_t *local = NULL;
+ xlator_t *subvol = NULL;
+
+ local = frame->local;
+
+ dht_inode_ctx_get1 (this, local->fd->inode, &subvol);
+
+ if (!subvol)
+ subvol = local->cached_subvol;
+
+ local->call_cnt = 2; /* This is the second attempt */
+
+ STACK_WIND(frame, dht_fallocate_cbk, subvol, subvol->fops->fallocate,
+ local->fd, local->rebalance.flags, local->rebalance.offset,
+ local->rebalance.size, NULL);
+
+ return 0;
+}
+
+int
+dht_fallocate(call_frame_t *frame, xlator_t *this, fd_t *fd, int32_t mode,
+ off_t offset, size_t len, dict_t *xdata)
+{
+ xlator_t *subvol = NULL;
+ int op_errno = -1;
+ dht_local_t *local = NULL;
+
+ VALIDATE_OR_GOTO (frame, err);
+ VALIDATE_OR_GOTO (this, err);
+ VALIDATE_OR_GOTO (fd, err);
+
+ local = dht_local_init (frame, NULL, fd, GF_FOP_FALLOCATE);
+ if (!local) {
+ op_errno = ENOMEM;
+ goto err;
+ }
+
+ local->rebalance.flags = mode;
+ local->rebalance.offset = offset;
+ local->rebalance.size = len;
+
+ local->call_cnt = 1;
+ subvol = local->cached_subvol;
+ if (!subvol) {
+ gf_log (this->name, GF_LOG_DEBUG,
+ "no cached subvolume for fd=%p", fd);
+ op_errno = EINVAL;
+ goto err;
+ }
+
+ STACK_WIND (frame, dht_fallocate_cbk,
+ subvol, subvol->fops->fallocate,
+ fd, mode, offset, len, xdata);
+
+ return 0;
+
+err:
+ op_errno = (op_errno == -1) ? errno : op_errno;
+ DHT_STACK_UNWIND (fallocate, frame, -1, op_errno, NULL, NULL, NULL);
+
+ return 0;
+}
+
+
+int
+dht_discard_cbk(call_frame_t *frame, void *cookie, xlator_t *this,
+ int op_ret, int op_errno, struct iatt *prebuf,
+ struct iatt *postbuf, dict_t *xdata)
+{
+ dht_local_t *local = NULL;
+ call_frame_t *prev = NULL;
+ int ret = -1;
+ xlator_t *subvol = NULL;
+
+ GF_VALIDATE_OR_GOTO ("dht", frame, err);
+ GF_VALIDATE_OR_GOTO ("dht", this, out);
+ GF_VALIDATE_OR_GOTO ("dht", frame->local, out);
+ GF_VALIDATE_OR_GOTO ("dht", cookie, out);
+
+ local = frame->local;
+ prev = cookie;
+
+ if ((op_ret == -1) && (op_errno != ENOENT)) {
+ local->op_errno = op_errno;
+ local->op_ret = -1;
+ gf_log (this->name, GF_LOG_DEBUG,
+ "subvolume %s returned -1 (%s)",
+ prev->this->name, strerror (op_errno));
+
+ goto out;
+ }
+
+ if (local->call_cnt != 1) {
+ if (local->stbuf.ia_blocks) {
+ dht_iatt_merge (this, postbuf, &local->stbuf, NULL);
+ dht_iatt_merge (this, prebuf, &local->prebuf, NULL);
+ }
+ goto out;
+ }
+ local->rebalance.target_op_fn = dht_discard2;
+
+ /* Phase 2 of migration */
+ if ((op_ret == -1) || IS_DHT_MIGRATION_PHASE2 (postbuf)) {
+ ret = dht_rebalance_complete_check (this, frame);
+ if (!ret)
+ return 0;
+ }
+
+ /* Check if the rebalance phase1 is true */
+ if (IS_DHT_MIGRATION_PHASE1 (postbuf)) {
+ dht_iatt_merge (this, &local->stbuf, postbuf, NULL);
+ dht_iatt_merge (this, &local->prebuf, prebuf, NULL);
+ dht_inode_ctx_get1 (this, local->fd->inode, &subvol);
+ if (subvol) {
+ dht_discard2 (this, frame, 0);
+ return 0;
+ }
+ ret = dht_rebalance_in_progress_check (this, frame);
+ if (!ret)
+ return 0;
+ }
+
+out:
+ DHT_STRIP_PHASE1_FLAGS (postbuf);
+ DHT_STRIP_PHASE1_FLAGS (prebuf);
+ DHT_STACK_UNWIND (discard, frame, op_ret, op_errno,
+ prebuf, postbuf, xdata);
+err:
+ return 0;
+}
+
+int
+dht_discard2(xlator_t *this, call_frame_t *frame, int op_ret)
+{
+ dht_local_t *local = NULL;
+ xlator_t *subvol = NULL;
+
+ local = frame->local;
+
+ dht_inode_ctx_get1 (this, local->fd->inode, &subvol);
+
+ if (!subvol)
+ subvol = local->cached_subvol;
+
+ local->call_cnt = 2; /* This is the second attempt */
+
+ STACK_WIND(frame, dht_discard_cbk, subvol, subvol->fops->discard,
+ local->fd, local->rebalance.offset, local->rebalance.size,
+ NULL);
+
+ return 0;
+}
+
+int
+dht_discard(call_frame_t *frame, xlator_t *this, fd_t *fd, off_t offset,
+ size_t len, dict_t *xdata)
+{
+ xlator_t *subvol = NULL;
+ int op_errno = -1;
+ dht_local_t *local = NULL;
+
+ VALIDATE_OR_GOTO (frame, err);
+ VALIDATE_OR_GOTO (this, err);
+ VALIDATE_OR_GOTO (fd, err);
+
+ local = dht_local_init (frame, NULL, fd, GF_FOP_DISCARD);
+ if (!local) {
+ op_errno = ENOMEM;
+ goto err;
+ }
+
+ local->rebalance.offset = offset;
+ local->rebalance.size = len;
+
+ local->call_cnt = 1;
+ subvol = local->cached_subvol;
+ if (!subvol) {
+ gf_log (this->name, GF_LOG_DEBUG,
+ "no cached subvolume for fd=%p", fd);
+ op_errno = EINVAL;
+ goto err;
+ }
+
+ STACK_WIND (frame, dht_discard_cbk, subvol, subvol->fops->discard,
+ fd, offset, len, xdata);
+
+ return 0;
+
+err:
+ op_errno = (op_errno == -1) ? errno : op_errno;
+ DHT_STACK_UNWIND (discard, frame, -1, op_errno, NULL, NULL, NULL);
+
+ return 0;
+}
+
+int
+dht_zerofill_cbk(call_frame_t *frame, void *cookie, xlator_t *this,
+ int op_ret, int op_errno, struct iatt *prebuf,
+ struct iatt *postbuf, dict_t *xdata)
+{
+ dht_local_t *local = NULL;
+ call_frame_t *prev = NULL;
+ int ret = -1;
+
+ GF_VALIDATE_OR_GOTO ("dht", frame, err);
+ GF_VALIDATE_OR_GOTO ("dht", this, out);
+ GF_VALIDATE_OR_GOTO ("dht", frame->local, out);
+ GF_VALIDATE_OR_GOTO ("dht", cookie, out);
+
+ local = frame->local;
+ prev = cookie;
+
+ if ((op_ret == -1) && (op_errno != ENOENT)) {
+ local->op_errno = op_errno;
+ local->op_ret = -1;
+ gf_log (this->name, GF_LOG_DEBUG,
+ "subvolume %s returned -1 (%s)",
+ prev->this->name, strerror (op_errno));
+ goto out;
+ }
+
+ if (local->call_cnt != 1) {
+ if (local->stbuf.ia_blocks) {
+ dht_iatt_merge (this, postbuf, &local->stbuf, NULL);
+ dht_iatt_merge (this, prebuf, &local->prebuf, NULL);
+ }
+ goto out;
+ }
+ local->rebalance.target_op_fn = dht_zerofill2;
+ /* Phase 2 of migration */
+ if ((op_ret == -1) || IS_DHT_MIGRATION_PHASE2 (postbuf)) {
+ ret = dht_rebalance_complete_check (this, frame);
+ if (!ret)
+ return 0;
+ }
+
+ /* Check if the rebalance phase1 is true */
+ if (IS_DHT_MIGRATION_PHASE1 (postbuf)) {
+ dht_iatt_merge (this, &local->stbuf, postbuf, NULL);
+ dht_iatt_merge (this, &local->prebuf, prebuf, NULL);
+ ret = fd_ctx_get (local->fd, this, NULL);
+ if (!ret) {
+ dht_zerofill2 (this, frame, 0);
+ return 0;
+ }
+ ret = dht_rebalance_in_progress_check (this, frame);
+ if (!ret)
+ return 0;
+ }
+
+out:
+ DHT_STRIP_PHASE1_FLAGS (postbuf);
+ DHT_STRIP_PHASE1_FLAGS (prebuf);
+ DHT_STACK_UNWIND (zerofill, frame, op_ret, op_errno,
+ prebuf, postbuf, xdata);
+err:
+ return 0;
+}
+
+int
+dht_zerofill2(xlator_t *this, call_frame_t *frame, int op_ret)
+{
+ dht_local_t *local = NULL;
+ xlator_t *subvol = NULL;
+ uint64_t tmp_subvol = 0;
+ int ret = -1;
+
+ local = frame->local;
+
+ if (local->fd)
+ ret = fd_ctx_get (local->fd, this, &tmp_subvol);
+ if (!ret)
+ subvol = (xlator_t *)(long)tmp_subvol;
+
+ if (!subvol)
+ subvol = local->cached_subvol;
+
+ local->call_cnt = 2; /* This is the second attempt */
+
+ STACK_WIND(frame, dht_zerofill_cbk, subvol, subvol->fops->zerofill,
+ local->fd, local->rebalance.offset, local->rebalance.size,
+ NULL);
+
+ return 0;
+}
+
+int
+dht_zerofill(call_frame_t *frame, xlator_t *this, fd_t *fd, off_t offset,
+ size_t len, dict_t *xdata)
+{
+ xlator_t *subvol = NULL;
+ int op_errno = -1;
+ dht_local_t *local = NULL;
+
+ VALIDATE_OR_GOTO (frame, err);
+ VALIDATE_OR_GOTO (this, err);
+ VALIDATE_OR_GOTO (fd, err);
+
+ local = dht_local_init (frame, NULL, fd, GF_FOP_ZEROFILL);
+ if (!local) {
+ op_errno = ENOMEM;
+ goto err;
+ }
+
+ local->rebalance.offset = offset;
+ local->rebalance.size = len;
+
+ local->call_cnt = 1;
+ subvol = local->cached_subvol;
+ if (!subvol) {
+ gf_log (this->name, GF_LOG_DEBUG,
+ "no cached subvolume for fd=%p", fd);
+ op_errno = EINVAL;
+ goto err;
+ }
+
+ STACK_WIND (frame, dht_zerofill_cbk, subvol, subvol->fops->zerofill,
+ fd, offset, len, xdata);
+
+ return 0;
+
+err:
+ op_errno = (op_errno == -1) ? errno : op_errno;
+ DHT_STACK_UNWIND (zerofill, frame, -1, op_errno, NULL, NULL, NULL);
+
+ return 0;
+}
+
+
+
/* handle cases of migration here for 'setattr()' calls */
int
dht_file_setattr_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
int op_ret, int op_errno, struct iatt *prebuf,
- struct iatt *postbuf)
+ struct iatt *postbuf, dict_t *xdata)
{
dht_local_t *local = NULL;
call_frame_t *prev = NULL;
@@ -391,7 +790,7 @@ out:
DHT_STRIP_PHASE1_FLAGS (postbuf);
DHT_STRIP_PHASE1_FLAGS (prebuf);
DHT_STACK_UNWIND (setattr, frame, op_ret, op_errno,
- prebuf, postbuf);
+ prebuf, postbuf, xdata);
return 0;
}
@@ -401,15 +800,13 @@ dht_setattr2 (xlator_t *this, call_frame_t *frame, int op_ret)
{
dht_local_t *local = NULL;
xlator_t *subvol = NULL;
- uint64_t tmp_subvol = 0;
- int ret = -1;
+ inode_t *inode = NULL;
local = frame->local;
- if (local->fd)
- ret = fd_ctx_get (local->fd, this, &tmp_subvol);
- if (!ret)
- subvol = (xlator_t *)(long)tmp_subvol;
+ inode = (local->fd) ? local->fd->inode : local->loc.inode;
+
+ dht_inode_ctx_get1 (this, inode, &subvol);
if (!subvol)
subvol = local->cached_subvol;
@@ -419,11 +816,13 @@ dht_setattr2 (xlator_t *this, call_frame_t *frame, int op_ret)
if (local->fop == GF_FOP_SETATTR) {
STACK_WIND (frame, dht_file_setattr_cbk, subvol,
subvol->fops->setattr, &local->loc,
- &local->rebalance.stbuf, local->rebalance.flags);
+ &local->rebalance.stbuf, local->rebalance.flags,
+ NULL);
} else {
STACK_WIND (frame, dht_file_setattr_cbk, subvol,
subvol->fops->fsetattr, local->fd,
- &local->rebalance.stbuf, local->rebalance.flags);
+ &local->rebalance.stbuf, local->rebalance.flags,
+ NULL);
}
return 0;
@@ -434,7 +833,7 @@ dht_setattr2 (xlator_t *this, call_frame_t *frame, int op_ret)
int
dht_setattr_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
int op_ret, int op_errno, struct iatt *statpre,
- struct iatt *statpost)
+ struct iatt *statpost, dict_t *xdata)
{
dht_local_t *local = NULL;
int this_call_cnt = 0;
@@ -465,7 +864,7 @@ unlock:
this_call_cnt = dht_frame_return (frame);
if (is_last_call (this_call_cnt))
DHT_STACK_UNWIND (setattr, frame, local->op_ret, local->op_errno,
- &local->prebuf, &local->stbuf);
+ &local->prebuf, &local->stbuf, xdata);
return 0;
}
@@ -473,7 +872,7 @@ unlock:
int
dht_setattr (call_frame_t *frame, xlator_t *this, loc_t *loc,
- struct iatt *stbuf, int32_t valid)
+ struct iatt *stbuf, int32_t valid, dict_t *xdata)
{
xlator_t *subvol = NULL;
dht_layout_t *layout = NULL;
@@ -519,7 +918,7 @@ dht_setattr (call_frame_t *frame, xlator_t *this, loc_t *loc,
STACK_WIND (frame, dht_file_setattr_cbk, subvol,
subvol->fops->setattr,
- loc, stbuf, valid);
+ loc, stbuf, valid, xdata);
return 0;
}
@@ -530,14 +929,14 @@ dht_setattr (call_frame_t *frame, xlator_t *this, loc_t *loc,
STACK_WIND (frame, dht_setattr_cbk,
layout->list[i].xlator,
layout->list[i].xlator->fops->setattr,
- loc, stbuf, valid);
+ loc, stbuf, valid, xdata);
}
return 0;
err:
op_errno = (op_errno == -1) ? errno : op_errno;
- DHT_STACK_UNWIND (setattr, frame, -1, op_errno, NULL, NULL);
+ DHT_STACK_UNWIND (setattr, frame, -1, op_errno, NULL, NULL, NULL);
return 0;
}
@@ -545,7 +944,7 @@ err:
int
dht_fsetattr (call_frame_t *frame, xlator_t *this, fd_t *fd, struct iatt *stbuf,
- int32_t valid)
+ int32_t valid, dict_t *xdata)
{
xlator_t *subvol = NULL;
dht_layout_t *layout = NULL;
@@ -590,7 +989,7 @@ dht_fsetattr (call_frame_t *frame, xlator_t *this, fd_t *fd, struct iatt *stbuf,
STACK_WIND (frame, dht_file_setattr_cbk, subvol,
subvol->fops->fsetattr,
- fd, stbuf, valid);
+ fd, stbuf, valid, xdata);
return 0;
}
@@ -601,14 +1000,14 @@ dht_fsetattr (call_frame_t *frame, xlator_t *this, fd_t *fd, struct iatt *stbuf,
STACK_WIND (frame, dht_setattr_cbk,
layout->list[i].xlator,
layout->list[i].xlator->fops->fsetattr,
- fd, stbuf, valid);
+ fd, stbuf, valid, xdata);
}
return 0;
err:
op_errno = (op_errno == -1) ? errno : op_errno;
- DHT_STACK_UNWIND (fsetattr, frame, -1, op_errno, NULL, NULL);
+ DHT_STACK_UNWIND (fsetattr, frame, -1, op_errno, NULL, NULL, NULL);
return 0;
}
diff --git a/xlators/cluster/dht/src/dht-layout.c b/xlators/cluster/dht/src/dht-layout.c
index 3c8f63f5b..38e9970a7 100644
--- a/xlators/cluster/dht/src/dht-layout.c
+++ b/xlators/cluster/dht/src/dht-layout.c
@@ -1,20 +1,11 @@
/*
- Copyright (c) 2008-2011 Gluster, Inc. <http://www.gluster.com>
+ Copyright (c) 2008-2012 Red Hat, Inc. <http://www.redhat.com>
This file is part of GlusterFS.
- GlusterFS is free software; you can redistribute it and/or modify
- it under the terms of the GNU General Public License as published
- by the Free Software Foundation; either version 3 of the License,
- or (at your option) any later version.
-
- GlusterFS is distributed in the hope that it will be useful, but
- WITHOUT ANY WARRANTY; without even the implied warranty of
- MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU
- General Public License for more details.
-
- You should have received a copy of the GNU General Public License
- along with this program. If not, see
- <http://www.gnu.org/licenses/>.
+ This file is licensed to you under your choice of the GNU Lesser
+ General Public License, version 3 or any later version (LGPLv3 or
+ later), or the GNU General Public License, version 2 (GPLv2), in all
+ cases as published by the Free Software Foundation.
*/
#ifndef _CONFIG_H
@@ -68,9 +59,7 @@ dht_layout_t *
dht_layout_get (xlator_t *this, inode_t *inode)
{
dht_conf_t *conf = NULL;
- uint64_t layout_int = 0;
dht_layout_t *layout = NULL;
- int ret = -1;
conf = this->private;
if (!conf)
@@ -78,9 +67,8 @@ dht_layout_get (xlator_t *this, inode_t *inode)
LOCK (&conf->layout_lock);
{
- ret = inode_ctx_get (inode, this, &layout_int);
- if (ret == 0) {
- layout = (dht_layout_t *) (unsigned long) layout_int;
+ dht_inode_ctx_layout_get (inode, this, &layout);
+ if (layout) {
layout->ref++;
}
}
@@ -98,7 +86,6 @@ dht_layout_set (xlator_t *this, inode_t *inode, dht_layout_t *layout)
int oldret = -1;
int ret = 0;
dht_layout_t *old_layout;
- uint64_t old_layout_int;
conf = this->private;
if (!conf)
@@ -106,16 +93,13 @@ dht_layout_set (xlator_t *this, inode_t *inode, dht_layout_t *layout)
LOCK (&conf->layout_lock);
{
- oldret = inode_ctx_get (inode, this, &old_layout_int);
-
+ oldret = dht_inode_ctx_layout_get (inode, this, &old_layout);
layout->ref++;
- ret = inode_ctx_put (inode, this, (uint64_t) (unsigned long)
- layout);
+ dht_inode_ctx_layout_set (inode, this, layout);
}
UNLOCK (&conf->layout_lock);
- if (oldret == 0) {
- old_layout = (dht_layout_t *) (unsigned long) old_layout_int;
+ if (!oldret) {
dht_layout_unref (this, old_layout);
}
@@ -130,7 +114,7 @@ dht_layout_unref (xlator_t *this, dht_layout_t *layout)
dht_conf_t *conf = NULL;
int ref = 0;
- if (layout->preset || !this->private)
+ if (!layout || layout->preset || !this->private)
return;
conf = this->private;
@@ -174,9 +158,9 @@ dht_layout_search (xlator_t *this, dht_layout_t *layout, const char *name)
int ret = 0;
- ret = dht_hash_compute (layout->type, name, &hash);
+ ret = dht_hash_compute (this, layout->type, name, &hash);
if (ret != 0) {
- gf_log (this->name, GF_LOG_INFO,
+ gf_log (this->name, GF_LOG_WARNING,
"hash computation failed for type=%d name=%s",
layout->type, name);
goto out;
@@ -191,7 +175,7 @@ dht_layout_search (xlator_t *this, dht_layout_t *layout, const char *name)
}
if (!subvol) {
- gf_log (this->name, GF_LOG_INFO,
+ gf_log (this->name, GF_LOG_WARNING,
"no subvolume for hash (value) = %u", hash);
}
@@ -280,6 +264,9 @@ dht_disk_layout_extract (xlator_t *this, dht_layout_t *layout,
if (disk_layout_p)
*disk_layout_p = disk_layout;
+ else
+ GF_FREE (disk_layout);
+
ret = 0;
out:
@@ -309,14 +296,18 @@ dht_disk_layout_merge (xlator_t *this, dht_layout_t *layout,
cnt = ntoh32 (disk_layout[0]);
if (cnt != 1) {
- gf_log (this->name, GF_LOG_INFO,
+ gf_log (this->name, GF_LOG_ERROR,
"disk layout has invalid count %d", cnt);
return -1;
}
- switch (disk_layout[1]) {
+ type = ntoh32 (disk_layout[1]);
+ switch (type) {
+ case DHT_HASH_TYPE_DM_USER:
+ gf_log (this->name, GF_LOG_DEBUG, "found user-set layout");
+ layout->type = type;
+ /* Fall through. */
case DHT_HASH_TYPE_DM:
- type = ntoh32 (disk_layout[1]);
break;
default:
gf_log (this->name, GF_LOG_CRITICAL,
@@ -344,11 +335,12 @@ int
dht_layout_merge (xlator_t *this, dht_layout_t *layout, xlator_t *subvol,
int op_ret, int op_errno, dict_t *xattr)
{
- int i = 0;
- int ret = -1;
- int err = -1;
- void *disk_layout_raw = NULL;
- int disk_layout_len = 0;
+ int i = 0;
+ int ret = -1;
+ int err = -1;
+ void *disk_layout_raw = NULL;
+ int disk_layout_len = 0;
+ dht_conf_t *conf = this->private;
if (op_ret != 0) {
err = op_errno;
@@ -369,12 +361,12 @@ dht_layout_merge (xlator_t *this, dht_layout_t *layout, xlator_t *subvol,
if (xattr) {
/* during lookup and not mkdir */
- ret = dict_get_ptr_and_len (xattr, "trusted.glusterfs.dht",
+ ret = dict_get_ptr_and_len (xattr, conf->xattr_name,
&disk_layout_raw, &disk_layout_len);
}
if (ret != 0) {
- layout->list[i].err = -1;
+ layout->list[i].err = 0;
gf_log (this->name, GF_LOG_TRACE,
"missing disk layout on %s. err = %d",
subvol->name, err);
@@ -385,7 +377,7 @@ dht_layout_merge (xlator_t *this, dht_layout_t *layout, xlator_t *subvol,
ret = dht_disk_layout_merge (this, layout, i, disk_layout_raw,
disk_layout_len);
if (ret != 0) {
- gf_log (this->name, GF_LOG_DEBUG,
+ gf_log (this->name, GF_LOG_WARNING,
"layout merge from subvolume %s failed",
subvol->name);
goto out;
@@ -421,6 +413,22 @@ dht_layout_entry_swap (dht_layout_t *layout, int i, int j)
layout->list[j].err = err_swap;
}
+void
+dht_layout_range_swap (dht_layout_t *layout, int i, int j)
+{
+ uint32_t start_swap = 0;
+ uint32_t stop_swap = 0;
+
+ start_swap = layout->list[i].start;
+ stop_swap = layout->list[i].stop;
+
+ layout->list[i].start = layout->list[j].start;
+ layout->list[i].stop = layout->list[j].stop;
+
+ layout->list[j].start = start_swap;
+ layout->list[j].stop = stop_swap;
+}
+
int64_t
dht_layout_entry_cmp_volname (dht_layout_t *layout, int i, int j)
{
@@ -428,17 +436,37 @@ dht_layout_entry_cmp_volname (dht_layout_t *layout, int i, int j)
layout->list[j].xlator->name));
}
+
+gf_boolean_t
+dht_is_subvol_in_layout (dht_layout_t *layout, xlator_t *xlator)
+{
+ int i = 0;
+
+ for (i = 0; i < layout->cnt; i++) {
+ if (!strcmp (layout->list[i].xlator->name, xlator->name))
+ return _gf_true;
+ }
+ return _gf_false;
+}
+
int64_t
dht_layout_entry_cmp (dht_layout_t *layout, int i, int j)
{
int64_t diff = 0;
+ /* swap zero'ed out layouts to front, if needed */
+ if (!layout->list[j].start && !layout->list[j].stop) {
+ diff = (int64_t) layout->list[i].stop
+ - (int64_t) layout->list[j].stop;
+ goto out;
+ }
if (layout->list[i].err || layout->list[j].err)
diff = layout->list[i].err - layout->list[j].err;
else
diff = (int64_t) layout->list[i].start
- (int64_t) layout->list[j].start;
+out:
return diff;
}
@@ -487,7 +515,8 @@ dht_layout_sort_volname (dht_layout_t *layout)
int
dht_layout_anomalies (xlator_t *this, loc_t *loc, dht_layout_t *layout,
uint32_t *holes_p, uint32_t *overlaps_p,
- uint32_t *missing_p, uint32_t *down_p, uint32_t *misc_p)
+ uint32_t *missing_p, uint32_t *down_p, uint32_t *misc_p,
+ uint32_t *no_space_p)
{
uint32_t overlaps = 0;
uint32_t missing = 0;
@@ -500,6 +529,7 @@ dht_layout_anomalies (xlator_t *this, loc_t *loc, dht_layout_t *layout,
uint32_t prev_stop = 0;
uint32_t last_stop = 0;
char is_virgin = 1;
+ uint32_t no_space = 0;
/* TODO: explain what is happening */
@@ -507,23 +537,30 @@ dht_layout_anomalies (xlator_t *this, loc_t *loc, dht_layout_t *layout,
prev_stop = last_stop;
for (i = 0; i < layout->cnt; i++) {
- if (layout->list[i].err) {
- switch (layout->list[i].err) {
- case -1:
- case ENOENT:
- missing++;
- break;
- case ENOTCONN:
- down++;
- break;
- case ENOSPC:
- down++;
- break;
- default:
- misc++;
+ switch (layout->list[i].err) {
+ case -1:
+ case ENOENT:
+ missing++;
+ continue;
+ case ENOTCONN:
+ down++;
+ continue;
+ case ENOSPC:
+ no_space++;
+ continue;
+ case 0:
+ /* if err == 0 and start == stop, then it is a non misc++;
+ * participating subvolume(spread-cnt). Then, do not
+ * check for anomalies. If start != stop, then treat it
+ * as misc err */
+ if (layout->list[i].start == layout->list[i].stop) {
+ continue;
}
+ break;
+ default:
+ misc++;
continue;
- }
+ }
is_virgin = 0;
@@ -556,6 +593,9 @@ dht_layout_anomalies (xlator_t *this, loc_t *loc, dht_layout_t *layout,
if (misc_p)
*misc_p = misc;
+ if (no_space_p)
+ *no_space_p = no_space;
+
return ret;
}
@@ -571,7 +611,6 @@ dht_layout_normalize (xlator_t *this, loc_t *loc, dht_layout_t *layout)
uint32_t down = 0;
uint32_t misc = 0;
-
ret = dht_layout_sort (layout);
if (ret == -1) {
gf_log (this->name, GF_LOG_WARNING,
@@ -581,7 +620,7 @@ dht_layout_normalize (xlator_t *this, loc_t *loc, dht_layout_t *layout)
ret = dht_layout_anomalies (this, loc, layout,
&holes, &overlaps,
- &missing, &down, &misc);
+ &missing, &down, &misc, NULL);
if (ret == -1) {
gf_log (this->name, GF_LOG_WARNING,
"error while finding anomalies in %s -- not good news",
@@ -599,43 +638,56 @@ dht_layout_normalize (xlator_t *this, loc_t *loc, dht_layout_t *layout)
"found anomalies in %s. holes=%d overlaps=%d",
loc->path, holes, overlaps);
}
- ret = 1;
+ ret = -1;
}
for (i = 0; i < layout->cnt; i++) {
- /* TODO During DHT selfheal rewrite (almost) find a better place to
- * detect this - probably in dht_layout_anomalies()
+ /* TODO During DHT selfheal rewrite (almost) find a better place
+ * to detect this - probably in dht_layout_anomalies()
*/
if (layout->list[i].err > 0) {
- gf_log (this->name, GF_LOG_DEBUG,
- "path=%s err=%s on subvol=%s",
- loc->path, strerror (layout->list[i].err),
- (layout->list[i].xlator ?
- layout->list[i].xlator->name : "<>"));
- if (layout->list[i].err == ENOENT)
- ret = 1;
+ gf_log_callingfn (this->name, GF_LOG_DEBUG,
+ "path=%s err=%s on subvol=%s",
+ loc->path,
+ strerror (layout->list[i].err),
+ (layout->list[i].xlator ?
+ layout->list[i].xlator->name
+ : "<>"));
+ if ((layout->list[i].err == ENOENT) && (ret >= 0)) {
+ ret++;
+ }
}
}
+
out:
return ret;
}
+int
+dht_dir_has_layout (dict_t *xattr, char *name)
+{
+
+ void *disk_layout_raw = NULL;
+
+ return dict_get_ptr (xattr, name, &disk_layout_raw);
+}
int
dht_layout_dir_mismatch (xlator_t *this, dht_layout_t *layout, xlator_t *subvol,
loc_t *loc, dict_t *xattr)
{
- int idx = 0;
- int pos = -1;
- int ret = 0;
- int err = 0;
- int dict_ret = 0;
- int32_t disk_layout[4];
- void *disk_layout_raw = NULL;
- int32_t count = -1;
- uint32_t start_off = -1;
- uint32_t stop_off = -1;
+ int idx = 0;
+ int pos = -1;
+ int ret = 0;
+ int err = 0;
+ int dict_ret = 0;
+ int32_t disk_layout[4];
+ void *disk_layout_raw = NULL;
+ int32_t count = -1;
+ uint32_t start_off = -1;
+ uint32_t stop_off = -1;
+ dht_conf_t *conf = this->private;
for (idx = 0; idx < layout->cnt; idx++) {
@@ -665,7 +717,7 @@ dht_layout_dir_mismatch (xlator_t *this, dht_layout_t *layout, xlator_t *subvol,
goto out;
}
- dict_ret = dict_get_ptr (xattr, "trusted.glusterfs.dht",
+ dict_ret = dict_get_ptr (xattr, conf->xattr_name,
&disk_layout_raw);
if (dict_ret < 0) {
@@ -681,7 +733,7 @@ dht_layout_dir_mismatch (xlator_t *this, dht_layout_t *layout, xlator_t *subvol,
count = ntoh32 (disk_layout[0]);
if (count != 1) {
- gf_log (this->name, GF_LOG_INFO,
+ gf_log (this->name, GF_LOG_ERROR,
"%s - disk layout has invalid count %d",
loc->path, count);
ret = -1;
@@ -730,7 +782,7 @@ dht_layout_preset (xlator_t *this, xlator_t *subvol, inode_t *inode)
LOCK (&conf->layout_lock);
{
- inode_ctx_put (inode, this, (uint64_t)(long)layout);
+ dht_inode_ctx_layout_set (inode, this, layout);
}
UNLOCK (&conf->layout_lock);
diff --git a/xlators/cluster/dht/src/dht-linkfile.c b/xlators/cluster/dht/src/dht-linkfile.c
index 2186b064a..dbc9d0b3c 100644
--- a/xlators/cluster/dht/src/dht-linkfile.c
+++ b/xlators/cluster/dht/src/dht-linkfile.c
@@ -1,20 +1,11 @@
/*
- Copyright (c) 2008-2011 Gluster, Inc. <http://www.gluster.com>
+ Copyright (c) 2008-2012 Red Hat, Inc. <http://www.redhat.com>
This file is part of GlusterFS.
- GlusterFS is free software; you can redistribute it and/or modify
- it under the terms of the GNU General Public License as published
- by the Free Software Foundation; either version 3 of the License,
- or (at your option) any later version.
-
- GlusterFS is distributed in the hope that it will be useful, but
- WITHOUT ANY WARRANTY; without even the implied warranty of
- MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU
- General Public License for more details.
-
- You should have received a copy of the GNU General Public License
- along with this program. If not, see
- <http://www.gnu.org/licenses/>.
+ This file is licensed to you under your choice of the GNU Lesser
+ General Public License, version 3 or any later version (LGPLv3 or
+ later), or the GNU General Public License, version 2 (GPLv2), in all
+ cases as published by the Free Software Foundation.
*/
#ifndef _CONFIG_H
@@ -28,37 +19,106 @@
#include "compat.h"
#include "dht-common.h"
+int
+dht_linkfile_lookup_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
+ int op_ret, int op_errno,
+ inode_t *inode, struct iatt *stbuf, dict_t *xattr,
+ struct iatt *postparent)
+{
+ char is_linkfile = 0;
+ dht_conf_t *conf = NULL;
+ dht_local_t *local = NULL;
+ call_frame_t *prev = NULL;
+ local = frame->local;
+ prev = cookie;
+ conf = this->private;
+
+ if (op_ret)
+ goto out;
+
+ is_linkfile = check_is_linkfile (inode, stbuf, xattr,
+ conf->link_xattr_name);
+ if (!is_linkfile)
+ gf_log (this->name, GF_LOG_WARNING, "got non-linkfile %s:%s",
+ prev->this->name, local->loc.path);
+out:
+ local->linkfile.linkfile_cbk (frame, cookie, this, op_ret, op_errno,
+ inode, stbuf, postparent, postparent,
+ xattr);
+ return 0;
+}
+#define is_equal(a, b) (a == b)
int
dht_linkfile_create_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
int op_ret, int op_errno, inode_t *inode,
struct iatt *stbuf, struct iatt *preparent,
- struct iatt *postparent)
+ struct iatt *postparent, dict_t *xdata)
{
dht_local_t *local = NULL;
+ xlator_t *subvol = NULL;
+ call_frame_t *prev = NULL;
+ dict_t *xattrs = NULL;
+ dht_conf_t *conf = NULL;
+ int ret = -1;
local = frame->local;
+ if (!op_ret)
+ local->linked = _gf_true;
+
+ FRAME_SU_UNDO (frame, dht_local_t);
+
+ if (op_ret && (op_errno == EEXIST)) {
+ conf = this->private;
+ prev = cookie;
+ subvol = prev->this;
+ if (!subvol)
+ goto out;
+ xattrs = dict_new ();
+ if (!xattrs)
+ goto out;
+ ret = dict_set_uint32 (xattrs, conf->link_xattr_name, 256);
+ if (ret) {
+ gf_log (this->name, GF_LOG_ERROR,
+ "Failed to set linkto key");
+ goto out;
+ }
+
+ STACK_WIND (frame, dht_linkfile_lookup_cbk, subvol,
+ subvol->fops->lookup, &local->loc, xattrs);
+ if (xattrs)
+ dict_unref (xattrs);
+ return 0;
+ }
+out:
local->linkfile.linkfile_cbk (frame, cookie, this, op_ret, op_errno,
- inode, stbuf, preparent, postparent);
+ inode, stbuf, preparent, postparent,
+ xdata);
+ if (xattrs)
+ dict_unref (xattrs);
return 0;
}
int
dht_linkfile_create (call_frame_t *frame, fop_mknod_cbk_t linkfile_cbk,
+ xlator_t *this,
xlator_t *tovol, xlator_t *fromvol, loc_t *loc)
{
dht_local_t *local = NULL;
dict_t *dict = NULL;
int need_unref = 0;
int ret = 0;
+ dht_conf_t *conf = this->private;
local = frame->local;
local->linkfile.linkfile_cbk = linkfile_cbk;
local->linkfile.srcvol = tovol;
+ local->linked = _gf_false;
+
dict = local->params;
if (!dict) {
dict = dict_new ();
@@ -74,8 +134,12 @@ dht_linkfile_create (call_frame_t *frame, fop_mknod_cbk_t linkfile_cbk,
"%s: gfid set failed", loc->path);
}
- ret = dict_set_str (dict, "trusted.glusterfs.dht.linkto",
- tovol->name);
+ ret = dict_set_str (dict, GLUSTERFS_INTERNAL_FOP_KEY, "yes");
+ if (ret)
+ gf_log ("dht-linkfile", GF_LOG_INFO,
+ "%s: internal-fop set failed", loc->path);
+
+ ret = dict_set_str (dict, conf->link_xattr_name, tovol->name);
if (ret < 0) {
gf_log (frame->this->name, GF_LOG_INFO,
@@ -84,9 +148,13 @@ dht_linkfile_create (call_frame_t *frame, fop_mknod_cbk_t linkfile_cbk,
goto out;
}
+ local->link_subvol = fromvol;
+ /* Always create as root:root. dht_linkfile_attr_heal fixes the
+ * ownsership */
+ FRAME_SU_DO (frame, dht_local_t);
STACK_WIND (frame, dht_linkfile_create_cbk,
fromvol, fromvol->fops->mknod, loc,
- S_IFREG | DHT_LINKFILE_MODE, 0, dict);
+ S_IFREG | DHT_LINKFILE_MODE, 0, 0, dict);
if (need_unref && dict)
dict_unref (dict);
@@ -94,7 +162,7 @@ dht_linkfile_create (call_frame_t *frame, fop_mknod_cbk_t linkfile_cbk,
return 0;
out:
local->linkfile.linkfile_cbk (frame, NULL, frame->this, -1, ENOMEM,
- loc->inode, NULL, NULL, NULL);
+ loc->inode, NULL, NULL, NULL, NULL);
if (need_unref && dict)
dict_unref (dict);
@@ -106,7 +174,8 @@ out:
int
dht_linkfile_unlink_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
int32_t op_ret, int32_t op_errno,
- struct iatt *preparent, struct iatt *postparent)
+ struct iatt *preparent, struct iatt *postparent,
+ dict_t *xdata)
{
dht_local_t *local = NULL;
call_frame_t *prev = NULL;
@@ -150,7 +219,7 @@ dht_linkfile_unlink (call_frame_t *frame, xlator_t *this,
STACK_WIND (unlink_frame, dht_linkfile_unlink_cbk,
subvol, subvol->fops->unlink,
- &unlink_local->loc);
+ &unlink_local->loc, 0, NULL);
return 0;
err:
@@ -175,7 +244,7 @@ dht_linkfile_subvol (xlator_t *this, inode_t *inode, struct iatt *stbuf,
if (!xattr)
goto out;
- ret = dict_get_ptr (xattr, "trusted.glusterfs.dht.linkto", &volname);
+ ret = dict_get_ptr (xattr, conf->link_xattr_name, &volname);
if ((-1 == ret) || !volname)
goto out;
@@ -190,3 +259,70 @@ dht_linkfile_subvol (xlator_t *this, inode_t *inode, struct iatt *stbuf,
out:
return subvol;
}
+
+int
+dht_linkfile_setattr_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
+ int op_ret, int op_errno, struct iatt *statpre,
+ struct iatt *statpost, dict_t *xdata)
+{
+ dht_local_t *local = NULL;
+ loc_t *loc = NULL;
+
+ local = frame->local;
+ loc = &local->loc;
+
+ if (op_ret)
+ gf_log (this->name, GF_LOG_ERROR, "setattr of uid/gid on %s"
+ " :<gfid:%s> failed (%s)",
+ (loc->path? loc->path: "NULL"),
+ uuid_utoa(local->gfid), strerror(op_errno));
+
+ DHT_STACK_DESTROY (frame);
+
+ return 0;
+}
+
+int
+dht_linkfile_attr_heal (call_frame_t *frame, xlator_t *this)
+{
+ int ret = -1;
+ call_frame_t *copy = NULL;
+ dht_local_t *local = NULL;
+ dht_local_t *copy_local = NULL;
+ xlator_t *subvol = NULL;
+ struct iatt stbuf = {0,};
+
+ local = frame->local;
+
+ GF_VALIDATE_OR_GOTO ("dht", local, out);
+ GF_VALIDATE_OR_GOTO ("dht", local->link_subvol, out);
+
+ if (local->stbuf.ia_type == IA_INVAL)
+ return 0;
+
+ uuid_copy (local->loc.gfid, local->stbuf.ia_gfid);
+
+ copy = copy_frame (frame);
+
+ if (!copy)
+ goto out;
+
+ copy_local = dht_local_init (copy, &local->loc, NULL, 0);
+
+ if (!copy_local)
+ goto out;
+
+ stbuf = local->stbuf;
+ subvol = local->link_subvol;
+
+ copy->local = copy_local;
+
+ FRAME_SU_DO (copy, dht_local_t);
+
+ STACK_WIND (copy, dht_linkfile_setattr_cbk, subvol,
+ subvol->fops->setattr, &copy_local->loc,
+ &stbuf, (GF_SET_ATTR_UID | GF_SET_ATTR_GID), NULL);
+ ret = 0;
+out:
+ return ret;
+}
diff --git a/xlators/cluster/dht/src/dht-mem-types.h b/xlators/cluster/dht/src/dht-mem-types.h
index 21fb5a7ca..e893eb48f 100644
--- a/xlators/cluster/dht/src/dht-mem-types.h
+++ b/xlators/cluster/dht/src/dht-mem-types.h
@@ -1,20 +1,11 @@
/*
- Copyright (c) 2008-2011 Gluster, Inc. <http://www.gluster.com>
+ Copyright (c) 2008-2012 Red Hat, Inc. <http://www.redhat.com>
This file is part of GlusterFS.
- GlusterFS is free software; you can redistribute it and/or modify
- it under the terms of the GNU General Public License as published
- by the Free Software Foundation; either version 3 of the License,
- or (at your option) any later version.
-
- GlusterFS is distributed in the hope that it will be useful, but
- WITHOUT ANY WARRANTY; without even the implied warranty of
- MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU
- General Public License for more details.
-
- You should have received a copy of the GNU General Public License
- along with this program. If not, see
- <http://www.gnu.org/licenses/>.
+ This file is licensed to you under your choice of the GNU Lesser
+ General Public License, version 3 or any later version (LGPLv3 or
+ later), or the GNU General Public License, version 2 (GPLv2), in all
+ cases as published by the Free Software Foundation.
*/
@@ -28,7 +19,6 @@ enum gf_dht_mem_types_ {
gf_dht_mt_dht_conf_t,
gf_dht_mt_char,
gf_dht_mt_int32_t,
- gf_dht_mt_dht_local_t,
gf_dht_mt_xlator_t,
gf_dht_mt_dht_layout_t,
gf_switch_mt_dht_conf_t,
@@ -37,6 +27,9 @@ enum gf_dht_mem_types_ {
gf_switch_mt_switch_struct,
gf_dht_mt_subvol_time,
gf_dht_mt_loc_t,
+ gf_defrag_info_mt,
+ gf_dht_mt_inode_ctx_t,
+ gf_dht_mt_ctx_stat_time_t,
gf_dht_mt_end
};
#endif
diff --git a/xlators/cluster/dht/src/dht-rebalance.c b/xlators/cluster/dht/src/dht-rebalance.c
index 70950e83e..bcb19f23e 100644
--- a/xlators/cluster/dht/src/dht-rebalance.c
+++ b/xlators/cluster/dht/src/dht-rebalance.c
@@ -1,20 +1,11 @@
/*
- Copyright (c) 2011 Gluster, Inc. <http://www.gluster.com>
+ Copyright (c) 2008-2012 Red Hat, Inc. <http://www.redhat.com>
This file is part of GlusterFS.
- GlusterFS is free software; you can redistribute it and/or modify
- it under the terms of the GNU General Public License as published
- by the Free Software Foundation; either version 3 of the License,
- or (at your option) any later version.
-
- GlusterFS is distributed in the hope that it will be useful, but
- WITHOUT ANY WARRANTY; without even the implied warranty of
- MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU
- General Public License for more details.
-
- You should have received a copy of the GNU General Public License
- along with this program. If not, see
- <http://www.gnu.org/licenses/>.
+ This file is licensed to you under your choice of the GNU Lesser
+ General Public License, version 3 or any later version (LGPLv3 or
+ later), or the GNU General Public License, version 2 (GPLv2), in all
+ cases as published by the Free Software Foundation.
*/
@@ -24,11 +15,12 @@
#endif
#include "dht-common.h"
+#include "xlator.h"
+#include <fnmatch.h>
#define GF_DISK_SECTOR_SIZE 512
#define DHT_REBALANCE_PID 4242 /* Change it if required */
#define DHT_REBALANCE_BLKSIZE (128 * 1024)
-#define DHT_MIGRATE_EVEN_IF_LINK_EXISTS 1
static int
dht_write_with_holes (xlator_t *to, fd_t *fd, struct iovec *vec, int count,
@@ -60,7 +52,7 @@ dht_write_with_holes (xlator_t *to, fd_t *fd, struct iovec *vec, int count,
ret = syncop_write (to, fd, (buf + tmp_offset),
(start_idx - tmp_offset),
(offset + tmp_offset),
- iobref);
+ iobref, 0);
/* 'path' will be logged in calling function */
if (ret < 0) {
gf_log (THIS->name, GF_LOG_WARNING,
@@ -78,7 +70,7 @@ dht_write_with_holes (xlator_t *to, fd_t *fd, struct iovec *vec, int count,
/* This means, last chunk is not yet written.. write it */
ret = syncop_write (to, fd, (buf + tmp_offset),
(buf_len - tmp_offset),
- (offset + tmp_offset), iobref);
+ (offset + tmp_offset), iobref, 0);
if (ret < 0) {
/* 'path' will be logged in calling function */
gf_log (THIS->name, GF_LOG_WARNING,
@@ -99,8 +91,118 @@ out:
}
+int32_t
+gf_defrag_handle_hardlink (xlator_t *this, loc_t *loc, dict_t *xattrs,
+ struct iatt *stbuf)
+{
+ int32_t ret = -1;
+ xlator_t *cached_subvol = NULL;
+ xlator_t *hashed_subvol = NULL;
+ xlator_t *linkto_subvol = NULL;
+ data_t *data = NULL;
+ struct iatt iatt = {0,};
+ int32_t op_errno = 0;
+ dht_conf_t *conf = NULL;
+
+ GF_VALIDATE_OR_GOTO ("defrag", loc, out);
+ GF_VALIDATE_OR_GOTO ("defrag", loc->name, out);
+ GF_VALIDATE_OR_GOTO ("defrag", stbuf, out);
+ GF_VALIDATE_OR_GOTO ("defrag", this, out);
+ GF_VALIDATE_OR_GOTO ("defrag", xattrs, out);
+ GF_VALIDATE_OR_GOTO ("defrag", this->private, out);
+
+ conf = this->private;
+
+ if (uuid_is_null (loc->pargfid)) {
+ gf_log ("", GF_LOG_ERROR, "loc->pargfid is NULL for "
+ "%s", loc->path);
+ goto out;
+ }
+
+ if (uuid_is_null (loc->gfid)) {
+ gf_log ("", GF_LOG_ERROR, "loc->gfid is NULL for "
+ "%s", loc->path);
+ goto out;
+ }
+
+ cached_subvol = dht_subvol_get_cached (this, loc->inode);
+ if (!cached_subvol) {
+ gf_log (this->name, GF_LOG_ERROR, "Failed to get cached subvol"
+ " for %s on %s", loc->name, this->name);
+ goto out;
+ }
+
+ hashed_subvol = dht_subvol_get_hashed (this, loc);
+ if (!hashed_subvol) {
+ gf_log (this->name, GF_LOG_ERROR, "Failed to get hashed subvol"
+ " for %s on %s", loc->name, this->name);
+ goto out;
+ }
+
+ gf_log (this->name, GF_LOG_INFO, "Attempting to migrate hardlink %s "
+ "with gfid %s from %s -> %s", loc->name, uuid_utoa (loc->gfid),
+ cached_subvol->name, hashed_subvol->name);
+ data = dict_get (xattrs, conf->link_xattr_name);
+ /* set linkto on cached -> hashed if not present, else link it */
+ if (!data) {
+ ret = dict_set_str (xattrs, conf->link_xattr_name,
+ hashed_subvol->name);
+ if (ret) {
+ gf_log (this->name, GF_LOG_ERROR, "Failed to set "
+ "linkto xattr in dict for %s", loc->name);
+ goto out;
+ }
+
+ ret = syncop_setxattr (cached_subvol, loc, xattrs, 0);
+ if (ret) {
+ gf_log (this->name, GF_LOG_ERROR, "Linkto setxattr "
+ "failed %s -> %s (%s)", cached_subvol->name,
+ loc->name, strerror (errno));
+ goto out;
+ }
+ goto out;
+ } else {
+ linkto_subvol = dht_linkfile_subvol (this, NULL, NULL, xattrs);
+ if (!linkto_subvol) {
+ gf_log (this->name, GF_LOG_ERROR, "Failed to get "
+ "linkto subvol for %s", loc->name);
+ } else {
+ hashed_subvol = linkto_subvol;
+ }
+
+ ret = syncop_link (hashed_subvol, loc, loc);
+ if (ret) {
+ op_errno = errno;
+ gf_log (this->name, GF_LOG_ERROR, "link of %s -> %s"
+ " failed on subvol %s (%s)", loc->name,
+ uuid_utoa(loc->gfid),
+ hashed_subvol->name, strerror (op_errno));
+ if (op_errno != EEXIST)
+ goto out;
+ }
+ }
+ ret = syncop_lookup (hashed_subvol, loc, NULL, &iatt, NULL, NULL);
+ if (ret) {
+ gf_log (this->name, GF_LOG_ERROR, "Failed lookup %s on %s (%s)"
+ , loc->name, hashed_subvol->name, strerror (errno));
+ goto out;
+ }
+
+ if (iatt.ia_nlink == stbuf->ia_nlink) {
+ ret = dht_migrate_file (this, loc, cached_subvol, hashed_subvol,
+ GF_DHT_MIGRATE_HARDLINK_IN_PROGRESS);
+ if (ret)
+ goto out;
+ }
+ ret = 0;
+out:
+ return ret;
+}
+
+
static inline int
-__is_file_migratable (xlator_t *this, loc_t *loc, struct iatt *stbuf)
+__is_file_migratable (xlator_t *this, loc_t *loc,
+ struct iatt *stbuf, dict_t *xattrs, int flags)
{
int ret = -1;
@@ -111,11 +213,25 @@ __is_file_migratable (xlator_t *this, loc_t *loc, struct iatt *stbuf)
goto out;
}
+ if (flags == GF_DHT_MIGRATE_HARDLINK_IN_PROGRESS) {
+ ret = 0;
+ goto out;
+ }
if (stbuf->ia_nlink > 1) {
- /* TODO : support migrating hardlinks */
- gf_log (this->name, GF_LOG_WARNING, "%s: file has hardlinks",
- loc->path);
- ret = -ENOTSUP;
+ /* support for decomission */
+ if (flags == GF_DHT_MIGRATE_HARDLINK) {
+ ret = gf_defrag_handle_hardlink (this, loc,
+ xattrs, stbuf);
+ if (ret) {
+ gf_log (this->name, GF_LOG_WARNING,
+ "%s: failed to migrate file with link",
+ loc->path);
+ }
+ } else {
+ gf_log (this->name, GF_LOG_WARNING,
+ "%s: file has hardlinks", loc->path);
+ }
+ ret = ENOTSUP;
goto out;
}
@@ -127,14 +243,16 @@ out:
static inline int
__dht_rebalance_create_dst_file (xlator_t *to, xlator_t *from, loc_t *loc, struct iatt *stbuf,
- dict_t *dict, fd_t **dst_fd)
+ dict_t *dict, fd_t **dst_fd, dict_t *xattr)
{
- xlator_t *this = NULL;
- int ret = -1;
- fd_t *fd = NULL;
- struct iatt new_stbuf = {0,};
+ xlator_t *this = NULL;
+ int ret = -1;
+ fd_t *fd = NULL;
+ struct iatt new_stbuf = {0,};
+ dht_conf_t *conf = NULL;
this = THIS;
+ conf = this->private;
ret = dict_set_static_bin (dict, "gfid-req", stbuf->ia_gfid, 16);
if (ret) {
@@ -143,7 +261,7 @@ __dht_rebalance_create_dst_file (xlator_t *to, xlator_t *from, loc_t *loc, struc
goto out;
}
- ret = dict_set_str (dict, DHT_LINKFILE_KEY, from->name);
+ ret = dict_set_str (dict, conf->link_xattr_name, from->name);
if (ret) {
gf_log (this->name, GF_LOG_ERROR,
"%s: failed to set gfid in dict for create", loc->path);
@@ -181,7 +299,7 @@ __dht_rebalance_create_dst_file (xlator_t *to, xlator_t *from, loc_t *loc, struc
/* Create the destination with LINKFILE mode, and linkto xattr,
if the linkfile already exists, it will just open the file */
ret = syncop_create (to, loc, O_RDWR, DHT_LINKFILE_MODE, fd,
- dict);
+ dict, &new_stbuf);
if (ret < 0) {
gf_log (this->name, GF_LOG_ERROR,
"failed to create %s on %s (%s)",
@@ -189,6 +307,26 @@ __dht_rebalance_create_dst_file (xlator_t *to, xlator_t *from, loc_t *loc, struc
goto out;
}
+ ret = syncop_fsetxattr (to, fd, xattr, 0);
+ if (ret == -1)
+ gf_log (this->name, GF_LOG_WARNING,
+ "%s: failed to set xattr on %s (%s)",
+ loc->path, to->name, strerror (errno));
+
+ ret = syncop_ftruncate (to, fd, stbuf->ia_size);
+ if (ret < 0)
+ gf_log (this->name, GF_LOG_ERROR,
+ "ftruncate failed for %s on %s (%s)",
+ loc->path, to->name, strerror (errno));
+
+ ret = syncop_fsetattr (to, fd, stbuf,
+ (GF_SET_ATTR_UID | GF_SET_ATTR_GID),
+ NULL, NULL);
+ if (ret < 0)
+ gf_log (this->name, GF_LOG_ERROR,
+ "chown failed for %s on %s (%s)",
+ loc->path, to->name, strerror (errno));
+
if (dst_fd)
*dst_fd = fd;
@@ -201,13 +339,16 @@ out:
static inline int
__dht_check_free_space (xlator_t *to, xlator_t *from, loc_t *loc,
- struct iatt *stbuf)
+ struct iatt *stbuf, int flag)
{
struct statvfs src_statfs = {0,};
struct statvfs dst_statfs = {0,};
int ret = -1;
xlator_t *this = NULL;
+ uint64_t src_statfs_blocks = 1;
+ uint64_t dst_statfs_blocks = 1;
+
this = THIS;
ret = syncop_statfs (from, loc, &src_statfs);
@@ -225,18 +366,47 @@ __dht_check_free_space (xlator_t *to, xlator_t *from, loc_t *loc,
loc->path, to->name, strerror (errno));
goto out;
}
- if (((dst_statfs.f_bavail *
- dst_statfs.f_bsize) / GF_DISK_SECTOR_SIZE) >
- (((src_statfs.f_bavail * src_statfs.f_bsize) /
- GF_DISK_SECTOR_SIZE) - stbuf->ia_blocks)) {
- gf_log (this->name, GF_LOG_WARNING,
- "data movement attempted from node (%s) with"
- " higher disk space to a node (%s) with "
- "lesser disk space (%s)", from->name,
- to->name, loc->path);
- /* this is not a 'failure', but we don't want to
- consider this as 'success' too :-/ */
+ /* if force option is given, do not check for space @ dst.
+ * Check only if space is avail for the file */
+ if (flag != GF_DHT_MIGRATE_DATA)
+ goto check_avail_space;
+
+ /* Check:
+ During rebalance `migrate-data` - Destination subvol experiences
+ a `reduction` in 'blocks' of free space, at the same time source
+ subvol gains certain 'blocks' of free space. A valid check is
+ necessary here to avoid errorneous move to destination where
+ the space could be scantily available.
+ */
+ if (stbuf) {
+ dst_statfs_blocks = ((dst_statfs.f_bavail *
+ dst_statfs.f_bsize) /
+ GF_DISK_SECTOR_SIZE);
+ src_statfs_blocks = ((src_statfs.f_bavail *
+ src_statfs.f_bsize) /
+ GF_DISK_SECTOR_SIZE);
+ if ((dst_statfs_blocks - stbuf->ia_blocks) <
+ (src_statfs_blocks + stbuf->ia_blocks)) {
+ gf_log (this->name, GF_LOG_WARNING,
+ "data movement attempted from node (%s) with"
+ " higher disk space to a node (%s) with "
+ "lesser disk space (%s)", from->name,
+ to->name, loc->path);
+
+ /* this is not a 'failure', but we don't want to
+ consider this as 'success' too :-/ */
+ ret = 1;
+ goto out;
+ }
+ }
+check_avail_space:
+ if (((dst_statfs.f_bavail * dst_statfs.f_bsize) /
+ GF_DISK_SECTOR_SIZE) < stbuf->ia_blocks) {
+ gf_log (this->name, GF_LOG_ERROR,
+ "data movement attempted from node (%s) with "
+ "to node (%s) which does not have required free space"
+ " for %s", from->name, to->name, loc->path);
ret = 1;
goto out;
}
@@ -263,7 +433,7 @@ __dht_rebalance_migrate_data (xlator_t *from, xlator_t *to, fd_t *src, fd_t *dst
read_size = (((ia_size - total) > DHT_REBALANCE_BLKSIZE) ?
DHT_REBALANCE_BLKSIZE : (ia_size - total));
ret = syncop_readv (from, src, read_size,
- offset, &vector, &count, &iobref);
+ offset, 0, &vector, &count, &iobref);
if (!ret || (ret < 0)) {
break;
}
@@ -273,15 +443,14 @@ __dht_rebalance_migrate_data (xlator_t *from, xlator_t *to, fd_t *src, fd_t *dst
ret, offset, iobref);
else
ret = syncop_writev (to, dst, vector, count,
- offset, iobref);
+ offset, iobref, 0);
if (ret < 0) {
break;
}
offset += ret;
total += ret;
- if (vector)
- GF_FREE (vector);
+ GF_FREE (vector);
if (iobref)
iobref_unref (iobref);
iobref = NULL;
@@ -289,8 +458,7 @@ __dht_rebalance_migrate_data (xlator_t *from, xlator_t *to, fd_t *src, fd_t *dst
}
if (iobref)
iobref_unref (iobref);
- if (vector)
- GF_FREE (vector);
+ GF_FREE (vector);
if (ret >= 0)
ret = 0;
@@ -308,8 +476,10 @@ __dht_rebalance_open_src_file (xlator_t *from, xlator_t *to, loc_t *loc,
dict_t *dict = NULL;
xlator_t *this = NULL;
struct iatt iatt = {0,};
+ dht_conf_t *conf = NULL;
this = THIS;
+ conf = this->private;
fd = fd_create (loc->inode, DHT_REBALANCE_PID);
if (!fd) {
@@ -332,7 +502,7 @@ __dht_rebalance_open_src_file (xlator_t *from, xlator_t *to, loc_t *loc,
if (!dict)
goto out;
- ret = dict_set_str (dict, DHT_LINKFILE_KEY, to->name);
+ ret = dict_set_str (dict, conf->link_xattr_name, to->name);
if (ret) {
gf_log (this->name, GF_LOG_ERROR,
"failed to set xattr in dict for %s (linkto:%s)",
@@ -385,12 +555,13 @@ migrate_special_files (xlator_t *this, xlator_t *from, xlator_t *to, loc_t *loc,
dict_t *dict = NULL;
char *link = NULL;
struct iatt stbuf = {0,};
+ dht_conf_t *conf = this->private;
dict = dict_new ();
if (!dict)
goto out;
- ret = dict_set_int32 (dict, DHT_LINKFILE_KEY, 256);
+ ret = dict_set_int32 (dict, conf->link_xattr_name, 256);
if (ret) {
gf_log (this->name, GF_LOG_ERROR,
"%s: failed to set 'linkto' key in dict", loc->path);
@@ -406,12 +577,13 @@ migrate_special_files (xlator_t *this, xlator_t *from, xlator_t *to, loc_t *loc,
}
/* we no more require this key */
- dict_del (dict, DHT_LINKFILE_KEY);
+ dict_del (dict, conf->link_xattr_name);
/* file exists in target node, only if it is 'linkfile' its valid,
otherwise, error out */
if (!ret) {
- if (!check_is_linkfile (loc->inode, &stbuf, rsp_dict)) {
+ if (!check_is_linkfile (loc->inode, &stbuf, rsp_dict,
+ conf->link_xattr_name)) {
gf_log (this->name, GF_LOG_WARNING,
"%s: file exists in destination", loc->path);
ret = -1;
@@ -447,7 +619,7 @@ migrate_special_files (xlator_t *this, xlator_t *from, xlator_t *to, loc_t *loc,
goto out;
}
- ret = syncop_symlink (to, loc, link, dict);
+ ret = syncop_symlink (to, loc, link, dict, 0);
if (ret) {
gf_log (this->name, GF_LOG_WARNING,
"%s: creating symlink failed (%s)",
@@ -461,7 +633,7 @@ migrate_special_files (xlator_t *this, xlator_t *from, xlator_t *to, loc_t *loc,
ret = syncop_mknod (to, loc, st_mode_from_ia (buf->ia_prot,
buf->ia_type),
makedev (ia_major (buf->ia_rdev),
- ia_minor (buf->ia_rdev)), dict);
+ ia_minor (buf->ia_rdev)), dict, 0);
if (ret) {
gf_log (this->name, GF_LOG_WARNING, "%s: mknod failed (%s)",
loc->path, strerror (errno));
@@ -469,6 +641,15 @@ migrate_special_files (xlator_t *this, xlator_t *from, xlator_t *to, loc_t *loc,
}
done:
+ ret = syncop_setattr (to, loc, buf,
+ (GF_SET_ATTR_UID | GF_SET_ATTR_GID |
+ GF_SET_ATTR_MODE), NULL, NULL);
+ if (ret) {
+ gf_log (this->name, GF_LOG_WARNING,
+ "%s: failed to perform setattr on %s (%s)",
+ loc->path, to->name, strerror (errno));
+ }
+
ret = syncop_unlink (from, loc);
if (ret)
gf_log (this->name, GF_LOG_WARNING, "%s: unlink failed (%s)",
@@ -504,7 +685,9 @@ dht_migrate_file (xlator_t *this, loc_t *loc, xlator_t *from, xlator_t *to,
fd_t *dst_fd = NULL;
dict_t *dict = NULL;
dict_t *xattr = NULL;
+ dict_t *xattr_rsp = NULL;
int file_has_holes = 0;
+ dht_conf_t *conf = this->private;
gf_log (this->name, GF_LOG_INFO, "%s: attempting to move from %s to %s",
loc->path, from->name, to->name);
@@ -513,19 +696,29 @@ dht_migrate_file (xlator_t *this, loc_t *loc, xlator_t *from, xlator_t *to,
if (!dict)
goto out;
+ ret = dict_set_int32 (dict, conf->link_xattr_name, 256);
+ if (ret) {
+ gf_log (this->name, GF_LOG_ERROR,
+ "%s: failed to set 'linkto' key in dict", loc->path);
+ goto out;
+ }
+
/* Phase 1 - Data migration is in progress from now on */
- ret = syncop_lookup (from, loc, NULL, &stbuf, NULL, NULL);
+ ret = syncop_lookup (from, loc, dict, &stbuf, &xattr_rsp, NULL);
if (ret) {
gf_log (this->name, GF_LOG_ERROR, "%s: lookup failed on %s (%s)",
loc->path, from->name, strerror (errno));
goto out;
}
+ /* we no more require this key */
+ dict_del (dict, conf->link_xattr_name);
+
/* preserve source mode, so set the same to the destination */
src_ia_prot = stbuf.ia_prot;
/* Check if file can be migrated */
- ret = __is_file_migratable (this, loc, &stbuf);
+ ret = __is_file_migratable (this, loc, &stbuf, xattr_rsp, flag);
if (ret)
goto out;
@@ -536,18 +729,22 @@ dht_migrate_file (xlator_t *this, loc_t *loc, xlator_t *from, xlator_t *to,
goto out;
}
+ /* TODO: move all xattr related operations to fd based operations */
+ ret = syncop_listxattr (from, loc, &xattr);
+ if (ret == -1)
+ gf_log (this->name, GF_LOG_WARNING,
+ "%s: failed to get xattr from %s (%s)",
+ loc->path, from->name, strerror (errno));
+
/* create the destination, with required modes/xattr */
ret = __dht_rebalance_create_dst_file (to, from, loc, &stbuf,
- dict, &dst_fd);
+ dict, &dst_fd, xattr);
if (ret)
goto out;
- /* Should happen on all files when 'force' option is not given */
- if (flag != DHT_MIGRATE_EVEN_IF_LINK_EXISTS) {
- ret = __dht_check_free_space (to, from, loc, &stbuf);
- if (ret) {
- goto out;
- }
+ ret = __dht_check_free_space (to, from, loc, &stbuf, flag);
+ if (ret) {
+ goto out;
}
/* Open the source, and also update mode/xattr */
@@ -558,6 +755,7 @@ dht_migrate_file (xlator_t *this, loc_t *loc, xlator_t *from, xlator_t *to,
goto out;
}
+
ret = syncop_fstat (from, src_fd, &stbuf);
if (ret) {
gf_log (this->name, GF_LOG_ERROR, "failed to lookup %s on %s (%s)",
@@ -587,22 +785,9 @@ dht_migrate_file (xlator_t *this, loc_t *loc, xlator_t *from, xlator_t *to,
goto out;
}
- /* TODO: move all xattr related operations to fd based operations */
- ret = syncop_listxattr (from, loc, &xattr);
- if (ret == -1)
- gf_log (this->name, GF_LOG_WARNING,
- "%s: failed to get xattr from %s (%s)",
- loc->path, from->name, strerror (errno));
-
- ret = syncop_setxattr (to, loc, xattr, 0);
- if (ret == -1)
- gf_log (this->name, GF_LOG_WARNING,
- "%s: failed to set xattr on %s (%s)",
- loc->path, to->name, strerror (errno));
-
/* TODO: Sync the locks */
- ret = syncop_fsync (to, dst_fd);
+ ret = syncop_fsync (to, dst_fd, 0);
if (ret)
gf_log (this->name, GF_LOG_WARNING,
"%s: failed to fsync on %s (%s)",
@@ -639,6 +824,7 @@ dht_migrate_file (xlator_t *this, loc_t *loc, xlator_t *from, xlator_t *to,
gf_log (this->name, GF_LOG_WARNING,
"%s: failed to perform setattr on %s (%s)",
loc->path, to->name, strerror (errno));
+ goto out;
}
/* Because 'futimes' is not portable */
@@ -659,6 +845,24 @@ dht_migrate_file (xlator_t *this, loc_t *loc, xlator_t *from, xlator_t *to,
gf_log (this->name, GF_LOG_WARNING, \
"%s: failed to perform setattr on %s (%s)",
loc->path, from->name, strerror (errno));
+ goto out;
+ }
+
+ /* Free up the data blocks on the source node, as the whole
+ file is migrated */
+ ret = syncop_ftruncate (from, src_fd, 0);
+ if (ret) {
+ gf_log (this->name, GF_LOG_WARNING,
+ "%s: failed to perform truncate on %s (%s)",
+ loc->path, from->name, strerror (errno));
+ }
+
+ /* remove the 'linkto' xattr from the destination */
+ ret = syncop_fremovexattr (to, dst_fd, conf->link_xattr_name);
+ if (ret) {
+ gf_log (this->name, GF_LOG_WARNING,
+ "%s: failed to perform removexattr on %s (%s)",
+ loc->path, to->name, strerror (errno));
}
/* Do a stat and check the gfid before unlink */
@@ -667,38 +871,23 @@ dht_migrate_file (xlator_t *this, loc_t *loc, xlator_t *from, xlator_t *to,
gf_log (this->name, GF_LOG_WARNING,
"%s: failed to do a stat on %s (%s)",
loc->path, from->name, strerror (errno));
+ goto out;
}
- if (uuid_compare (empty_iatt.ia_gfid, loc->inode->gfid) == 0) {
+ if (uuid_compare (empty_iatt.ia_gfid, loc->gfid) == 0) {
/* take out the source from namespace */
ret = syncop_unlink (from, loc);
if (ret) {
gf_log (this->name, GF_LOG_WARNING,
"%s: failed to perform unlink on %s (%s)",
loc->path, from->name, strerror (errno));
+ goto out;
}
}
- /* Free up the data blocks on the source node, as the whole
- file is migrated */
- ret = syncop_ftruncate (from, src_fd, 0);
- if (ret) {
- gf_log (this->name, GF_LOG_WARNING,
- "%s: failed to perform truncate on %s (%s)",
- loc->path, from->name, strerror (errno));
- }
-
- /* remove the 'linkto' xattr from the destination */
- ret = syncop_removexattr (to, loc, DHT_LINKFILE_KEY);
- if (ret) {
- gf_log (this->name, GF_LOG_WARNING,
- "%s: failed to perform removexattr on %s (%s)",
- loc->path, to->name, strerror (errno));
- }
-
ret = syncop_lookup (this, loc, NULL, NULL, NULL, NULL);
if (ret) {
- gf_log (this->name, GF_LOG_WARNING,
+ gf_log (this->name, GF_LOG_DEBUG,
"%s: failed to lookup the file on subvolumes (%s)",
loc->path, strerror (errno));
}
@@ -714,6 +903,8 @@ out:
if (xattr)
dict_unref (xattr);
+ if (xattr_rsp)
+ dict_unref (xattr_rsp);
if (dst_fd)
syncop_close (dst_fd);
@@ -788,7 +979,7 @@ rebalance_task_completion (int op_ret, call_frame_t *sync_frame, void *data)
op_errno = EPERM;
}
- DHT_STACK_UNWIND (setxattr, sync_frame, op_ret, op_errno);
+ DHT_STACK_UNWIND (setxattr, sync_frame, op_ret, op_errno, NULL);
return 0;
}
@@ -796,12 +987,829 @@ int
dht_start_rebalance_task (xlator_t *this, call_frame_t *frame)
{
int ret = -1;
- dht_conf_t *conf = NULL;
- conf = this->private;
-
- ret = synctask_new (conf->env, rebalance_task,
+ ret = synctask_new (this->ctx->env, rebalance_task,
rebalance_task_completion,
frame, frame);
return ret;
}
+
+int
+gf_listener_stop (xlator_t *this)
+{
+ glusterfs_ctx_t *ctx = NULL;
+ cmd_args_t *cmd_args = NULL;
+ int ret = 0;
+
+ ctx = this->ctx;
+ GF_ASSERT (ctx);
+ cmd_args = &ctx->cmd_args;
+ if (cmd_args->sock_file) {
+ ret = unlink (cmd_args->sock_file);
+ if (ret && (ENOENT == errno)) {
+ ret = 0;
+ }
+ }
+
+ if (ret) {
+ gf_log (this->name, GF_LOG_ERROR, "Failed to unlink listener "
+ "socket %s, error: %s", cmd_args->sock_file,
+ strerror (errno));
+ }
+ return ret;
+}
+
+void
+dht_build_root_inode (xlator_t *this, inode_t **inode)
+{
+ inode_table_t *itable = NULL;
+ uuid_t root_gfid = {0, };
+
+ itable = inode_table_new (0, this);
+ if (!itable)
+ return;
+
+ root_gfid[15] = 1;
+ *inode = inode_find (itable, root_gfid);
+}
+
+void
+dht_build_root_loc (inode_t *inode, loc_t *loc)
+{
+ loc->path = "/";
+ loc->inode = inode;
+ loc->inode->ia_type = IA_IFDIR;
+ memset (loc->gfid, 0, 16);
+ loc->gfid[15] = 1;
+}
+
+
+/* return values: 1 -> error, bug ignore and continue
+ 0 -> proceed
+ -1 -> error, handle it */
+int32_t
+gf_defrag_handle_migrate_error (int32_t op_errno, gf_defrag_info_t *defrag)
+{
+ /* if errno is not ENOSPC or ENOTCONN, we can still continue
+ with rebalance process */
+ if ((errno != ENOSPC) || (errno != ENOTCONN))
+ return 1;
+
+ if (errno == ENOTCONN) {
+ /* Most probably mount point went missing (mostly due
+ to a brick down), say rebalance failure to user,
+ let him restart it if everything is fine */
+ defrag->defrag_status = GF_DEFRAG_STATUS_FAILED;
+ return -1;
+ }
+
+ if (errno == ENOSPC) {
+ /* rebalance process itself failed, may be
+ remote brick went down, or write failed due to
+ disk full etc etc.. */
+ defrag->defrag_status = GF_DEFRAG_STATUS_FAILED;
+ return -1;
+ }
+
+ return 0;
+}
+
+static gf_boolean_t
+gf_defrag_pattern_match (gf_defrag_info_t *defrag, char *name, uint64_t size)
+{
+ gf_defrag_pattern_list_t *trav = NULL;
+ gf_boolean_t match = _gf_false;
+ gf_boolean_t ret = _gf_false;
+
+ GF_VALIDATE_OR_GOTO ("dht", defrag, out);
+
+ trav = defrag->defrag_pattern;
+ while (trav) {
+ if (!fnmatch (trav->path_pattern, name, FNM_NOESCAPE)) {
+ match = _gf_true;
+ break;
+ }
+ trav = trav->next;
+ }
+
+ if ((match == _gf_true) && (size >= trav->size))
+ ret = _gf_true;
+
+ out:
+ return ret;
+}
+
+/* We do a depth first traversal of directories. But before we move into
+ * subdirs, we complete the data migration of those directories whose layouts
+ * have been fixed
+ */
+
+int
+gf_defrag_migrate_data (xlator_t *this, gf_defrag_info_t *defrag, loc_t *loc,
+ dict_t *migrate_data)
+{
+ int ret = -1;
+ loc_t entry_loc = {0,};
+ fd_t *fd = NULL;
+ gf_dirent_t entries;
+ gf_dirent_t *tmp = NULL;
+ gf_dirent_t *entry = NULL;
+ gf_boolean_t free_entries = _gf_false;
+ off_t offset = 0;
+ dict_t *dict = NULL;
+ struct iatt iatt = {0,};
+ int32_t op_errno = 0;
+ char *uuid_str = NULL;
+ uuid_t node_uuid = {0,};
+ int readdir_operrno = 0;
+ struct timeval dir_start = {0,};
+ struct timeval end = {0,};
+ double elapsed = {0,};
+ struct timeval start = {0,};
+ int32_t err = 0;
+
+ gf_log (this->name, GF_LOG_INFO, "migrate data called on %s",
+ loc->path);
+ gettimeofday (&dir_start, NULL);
+
+ fd = fd_create (loc->inode, defrag->pid);
+ if (!fd) {
+ gf_log (this->name, GF_LOG_ERROR, "Failed to create fd");
+ goto out;
+ }
+
+ ret = syncop_opendir (this, loc, fd);
+ if (ret) {
+ gf_log (this->name, GF_LOG_ERROR, "Failed to open dir %s",
+ loc->path);
+ goto out;
+ }
+
+ INIT_LIST_HEAD (&entries.list);
+
+ while ((ret = syncop_readdirp (this, fd, 131072, offset, NULL,
+ &entries)) != 0) {
+
+ if (ret < 0) {
+
+ gf_log (this->name, GF_LOG_ERROR, "Readdir returned %s."
+ " Aborting migrate-data",
+ strerror(readdir_operrno));
+ goto out;
+ }
+
+ /* Need to keep track of ENOENT errno, that means, there is no
+ need to send more readdirp() */
+ readdir_operrno = errno;
+
+ if (list_empty (&entries.list))
+ break;
+
+ free_entries = _gf_true;
+
+ list_for_each_entry_safe (entry, tmp, &entries.list, list) {
+ if (defrag->defrag_status != GF_DEFRAG_STATUS_STARTED) {
+ ret = 1;
+ goto out;
+ }
+
+ offset = entry->d_off;
+
+ if (!strcmp (entry->d_name, ".") ||
+ !strcmp (entry->d_name, ".."))
+ continue;
+
+ if (IA_ISDIR (entry->d_stat.ia_type))
+ continue;
+
+ defrag->num_files_lookedup++;
+ if (defrag->stats == _gf_true) {
+ gettimeofday (&start, NULL);
+ }
+ if (defrag->defrag_pattern &&
+ (gf_defrag_pattern_match (defrag, entry->d_name,
+ entry->d_stat.ia_size)
+ == _gf_false)) {
+ continue;
+ }
+ loc_wipe (&entry_loc);
+ ret =dht_build_child_loc (this, &entry_loc, loc,
+ entry->d_name);
+ if (ret) {
+ gf_log (this->name, GF_LOG_ERROR, "Child loc"
+ " build failed");
+ goto out;
+ }
+
+ if (uuid_is_null (entry->d_stat.ia_gfid)) {
+ gf_log (this->name, GF_LOG_ERROR, "%s/%s"
+ " gfid not present", loc->path,
+ entry->d_name);
+ continue;
+ }
+
+ uuid_copy (entry_loc.gfid, entry->d_stat.ia_gfid);
+
+ if (uuid_is_null (loc->gfid)) {
+ gf_log (this->name, GF_LOG_ERROR, "%s/%s"
+ " gfid not present", loc->path,
+ entry->d_name);
+ continue;
+ }
+
+ uuid_copy (entry_loc.pargfid, loc->gfid);
+
+ entry_loc.inode->ia_type = entry->d_stat.ia_type;
+
+ ret = syncop_lookup (this, &entry_loc, NULL, &iatt,
+ NULL, NULL);
+ if (ret) {
+ gf_log (this->name, GF_LOG_ERROR, "%s"
+ " lookup failed", entry_loc.path);
+ continue;
+ }
+
+ ret = syncop_getxattr (this, &entry_loc, &dict,
+ GF_XATTR_NODE_UUID_KEY);
+ if(ret < 0) {
+ gf_log (this->name, GF_LOG_ERROR, "Failed to "
+ "get node-uuid for %s", entry_loc.path);
+ continue;
+ }
+
+ ret = dict_get_str (dict, GF_XATTR_NODE_UUID_KEY,
+ &uuid_str);
+ if(ret < 0) {
+ gf_log (this->name, GF_LOG_ERROR, "Failed to "
+ "get node-uuid from dict for %s",
+ entry_loc.path);
+ continue;
+ }
+
+ if (uuid_parse (uuid_str, node_uuid)) {
+ gf_log (this->name, GF_LOG_ERROR, "uuid_parse "
+ "failed for %s", entry_loc.path);
+ continue;
+ }
+
+ /* if file belongs to different node, skip migration
+ * the other node will take responsibility of migration
+ */
+ if (uuid_compare (node_uuid, defrag->node_uuid)) {
+ gf_log (this->name, GF_LOG_TRACE, "%s does not"
+ "belong to this node", entry_loc.path);
+ continue;
+ }
+
+ uuid_str = NULL;
+
+ dict_del (dict, GF_XATTR_NODE_UUID_KEY);
+
+
+ /* if distribute is present, it will honor this key.
+ * -1 is returned if distribute is not present or file
+ * doesn't have a link-file. If file has link-file, the
+ * path of link-file will be the value, and also that
+ * guarantees that file has to be mostly migrated */
+
+ ret = syncop_getxattr (this, &entry_loc, &dict,
+ GF_XATTR_LINKINFO_KEY);
+ if (ret < 0) {
+ gf_log (this->name, GF_LOG_TRACE, "failed to "
+ "get link-to key for %s",
+ entry_loc.path);
+ continue;
+ }
+
+ ret = syncop_setxattr (this, &entry_loc, migrate_data,
+ 0);
+ if (ret) {
+ err = op_errno;
+ /* errno is overloaded. See
+ * rebalance_task_completion () */
+ if (err != ENOSPC) {
+ gf_log (this->name, GF_LOG_DEBUG,
+ "migrate-data skipped for %s"
+ " due to space constraints",
+ entry_loc.path);
+ defrag->skipped +=1;
+ } else{
+ gf_log (this->name, GF_LOG_ERROR,
+ "migrate-data failed for %s",
+ entry_loc.path);
+ defrag->total_failures +=1;
+ }
+ }
+
+ if (ret == -1) {
+ op_errno = errno;
+ ret = gf_defrag_handle_migrate_error (op_errno,
+ defrag);
+
+ if (!ret)
+ gf_log (this->name, GF_LOG_DEBUG,
+ "migrate-data on %s failed: %s",
+ entry_loc.path,
+ strerror (op_errno));
+ else if (ret == 1)
+ continue;
+ else if (ret == -1)
+ goto out;
+ }
+
+ LOCK (&defrag->lock);
+ {
+ defrag->total_files += 1;
+ defrag->total_data += iatt.ia_size;
+ }
+ UNLOCK (&defrag->lock);
+ if (defrag->stats == _gf_true) {
+ gettimeofday (&end, NULL);
+ elapsed = (end.tv_sec - start.tv_sec) * 1e6 +
+ (end.tv_usec - start.tv_usec);
+ gf_log (this->name, GF_LOG_INFO, "Migration of "
+ "file:%s size:%"PRIu64" bytes took %.2f"
+ "secs", entry_loc.path, iatt.ia_size,
+ elapsed/1e6);
+ }
+ }
+
+ gf_dirent_free (&entries);
+ free_entries = _gf_false;
+ INIT_LIST_HEAD (&entries.list);
+
+ if (readdir_operrno == ENOENT)
+ break;
+ }
+
+ gettimeofday (&end, NULL);
+ elapsed = (end.tv_sec - dir_start.tv_sec) * 1e6 +
+ (end.tv_usec - dir_start.tv_usec);
+ gf_log (this->name, GF_LOG_INFO, "Migration operation on dir %s took "
+ "%.2f secs", loc->path, elapsed/1e6);
+ ret = 0;
+out:
+ if (free_entries)
+ gf_dirent_free (&entries);
+
+ loc_wipe (&entry_loc);
+
+ if (dict)
+ dict_unref(dict);
+
+ if (fd)
+ fd_unref (fd);
+ return ret;
+
+}
+
+
+int
+gf_defrag_fix_layout (xlator_t *this, gf_defrag_info_t *defrag, loc_t *loc,
+ dict_t *fix_layout, dict_t *migrate_data)
+{
+ int ret = -1;
+ loc_t entry_loc = {0,};
+ fd_t *fd = NULL;
+ gf_dirent_t entries;
+ gf_dirent_t *tmp = NULL;
+ gf_dirent_t *entry = NULL;
+ gf_boolean_t free_entries = _gf_false;
+ dict_t *dict = NULL;
+ off_t offset = 0;
+ struct iatt iatt = {0,};
+ int readdirp_errno = 0;
+
+ ret = syncop_lookup (this, loc, NULL, &iatt, NULL, NULL);
+ if (ret) {
+ gf_log (this->name, GF_LOG_ERROR, "Lookup failed on %s",
+ loc->path);
+ goto out;
+ }
+
+ if (defrag->cmd != GF_DEFRAG_CMD_START_LAYOUT_FIX) {
+ ret = gf_defrag_migrate_data (this, defrag, loc, migrate_data);
+ if (ret)
+ goto out;
+ }
+
+ gf_log (this->name, GF_LOG_TRACE, "fix layout called on %s", loc->path);
+
+ fd = fd_create (loc->inode, defrag->pid);
+ if (!fd) {
+ gf_log (this->name, GF_LOG_ERROR, "Failed to create fd");
+ ret = -1;
+ goto out;
+ }
+
+ ret = syncop_opendir (this, loc, fd);
+ if (ret) {
+ gf_log (this->name, GF_LOG_ERROR, "Failed to open dir %s",
+ loc->path);
+ ret = -1;
+ goto out;
+ }
+
+ INIT_LIST_HEAD (&entries.list);
+ while ((ret = syncop_readdirp (this, fd, 131072, offset, NULL,
+ &entries)) != 0)
+ {
+
+ if (ret < 0) {
+ gf_log (this->name, GF_LOG_ERROR, "Readdir returned %s"
+ ". Aborting fix-layout",strerror(errno));
+ goto out;
+ }
+
+ /* Need to keep track of ENOENT errno, that means, there is no
+ need to send more readdirp() */
+ readdirp_errno = errno;
+
+ if (list_empty (&entries.list))
+ break;
+
+ free_entries = _gf_true;
+
+ list_for_each_entry_safe (entry, tmp, &entries.list, list) {
+ if (defrag->defrag_status != GF_DEFRAG_STATUS_STARTED) {
+ ret = 1;
+ goto out;
+ }
+
+ offset = entry->d_off;
+
+ if (!strcmp (entry->d_name, ".") ||
+ !strcmp (entry->d_name, ".."))
+ continue;
+
+ if (!IA_ISDIR (entry->d_stat.ia_type))
+ continue;
+
+ loc_wipe (&entry_loc);
+ ret =dht_build_child_loc (this, &entry_loc, loc,
+ entry->d_name);
+ if (ret) {
+ gf_log (this->name, GF_LOG_ERROR, "Child loc"
+ " build failed");
+ goto out;
+ }
+
+ if (uuid_is_null (entry->d_stat.ia_gfid)) {
+ gf_log (this->name, GF_LOG_ERROR, "%s/%s"
+ " gfid not present", loc->path,
+ entry->d_name);
+ continue;
+ }
+
+ entry_loc.inode->ia_type = entry->d_stat.ia_type;
+
+ uuid_copy (entry_loc.gfid, entry->d_stat.ia_gfid);
+ if (uuid_is_null (loc->gfid)) {
+ gf_log (this->name, GF_LOG_ERROR, "%s/%s"
+ " gfid not present", loc->path,
+ entry->d_name);
+ continue;
+ }
+
+ uuid_copy (entry_loc.pargfid, loc->gfid);
+
+ ret = syncop_lookup (this, &entry_loc, NULL, &iatt,
+ NULL, NULL);
+ if (ret) {
+ gf_log (this->name, GF_LOG_ERROR, "%s"
+ " lookup failed", entry_loc.path);
+ continue;
+ }
+
+ ret = syncop_setxattr (this, &entry_loc, fix_layout,
+ 0);
+ if (ret) {
+ gf_log (this->name, GF_LOG_ERROR, "Setxattr "
+ "failed for %s", entry_loc.path);
+ defrag->defrag_status =
+ GF_DEFRAG_STATUS_FAILED;
+ defrag->total_failures ++;
+ goto out;
+ }
+ ret = gf_defrag_fix_layout (this, defrag, &entry_loc,
+ fix_layout, migrate_data);
+
+ if (ret) {
+ gf_log (this->name, GF_LOG_ERROR, "Fix layout "
+ "failed for %s", entry_loc.path);
+ defrag->total_failures++;
+ goto out;
+ }
+
+ }
+ gf_dirent_free (&entries);
+ free_entries = _gf_false;
+ INIT_LIST_HEAD (&entries.list);
+ if (readdirp_errno == ENOENT)
+ break;
+ }
+
+ ret = 0;
+out:
+ if (free_entries)
+ gf_dirent_free (&entries);
+
+ loc_wipe (&entry_loc);
+
+ if (dict)
+ dict_unref(dict);
+
+ if (fd)
+ fd_unref (fd);
+
+ return ret;
+
+}
+
+
+int
+gf_defrag_start_crawl (void *data)
+{
+ xlator_t *this = NULL;
+ dht_conf_t *conf = NULL;
+ gf_defrag_info_t *defrag = NULL;
+ int ret = -1;
+ loc_t loc = {0,};
+ struct iatt iatt = {0,};
+ struct iatt parent = {0,};
+ dict_t *fix_layout = NULL;
+ dict_t *migrate_data = NULL;
+ dict_t *status = NULL;
+ glusterfs_ctx_t *ctx = NULL;
+
+ this = data;
+ if (!this)
+ goto out;
+
+ ctx = this->ctx;
+ if (!ctx)
+ goto out;
+
+ conf = this->private;
+ if (!conf)
+ goto out;
+
+ defrag = conf->defrag;
+ if (!defrag)
+ goto out;
+
+ gettimeofday (&defrag->start_time, NULL);
+ dht_build_root_inode (this, &defrag->root_inode);
+ if (!defrag->root_inode)
+ goto out;
+
+ dht_build_root_loc (defrag->root_inode, &loc);
+
+ /* fix-layout on '/' first */
+
+ ret = syncop_lookup (this, &loc, NULL, &iatt, NULL, &parent);
+
+ if (ret) {
+ gf_log (this->name, GF_LOG_ERROR, "look up on / failed");
+ goto out;
+ }
+
+ fix_layout = dict_new ();
+ if (!fix_layout) {
+ ret = -1;
+ goto out;
+ }
+
+ ret = dict_set_str (fix_layout, GF_XATTR_FIX_LAYOUT_KEY, "yes");
+ if (ret) {
+ gf_log (this->name, GF_LOG_ERROR, "Failed to set dict str");
+ goto out;
+ }
+
+ ret = syncop_setxattr (this, &loc, fix_layout, 0);
+ if (ret) {
+ gf_log (this->name, GF_LOG_ERROR, "fix layout on %s failed",
+ loc.path);
+ defrag->total_failures++;
+ goto out;
+ }
+
+ if (defrag->cmd != GF_DEFRAG_CMD_START_LAYOUT_FIX) {
+ migrate_data = dict_new ();
+ if (!migrate_data) {
+ ret = -1;
+ goto out;
+ }
+ if (defrag->cmd == GF_DEFRAG_CMD_START_FORCE)
+ ret = dict_set_str (migrate_data,
+ "distribute.migrate-data", "force");
+ else
+ ret = dict_set_str (migrate_data,
+ "distribute.migrate-data",
+ "non-force");
+ if (ret)
+ goto out;
+ }
+ ret = gf_defrag_fix_layout (this, defrag, &loc, fix_layout,
+ migrate_data);
+ if ((defrag->defrag_status != GF_DEFRAG_STATUS_STOPPED) &&
+ (defrag->defrag_status != GF_DEFRAG_STATUS_FAILED)) {
+ defrag->defrag_status = GF_DEFRAG_STATUS_COMPLETE;
+ }
+
+
+
+out:
+ LOCK (&defrag->lock);
+ {
+ status = dict_new ();
+ gf_defrag_status_get (defrag, status);
+ if (ctx->notify)
+ ctx->notify (GF_EN_DEFRAG_STATUS, status);
+ if (status)
+ dict_unref (status);
+ defrag->is_exiting = 1;
+ }
+ UNLOCK (&defrag->lock);
+
+ if (defrag) {
+ GF_FREE (defrag);
+ conf->defrag = NULL;
+ }
+
+ return ret;
+}
+
+
+static int
+gf_defrag_done (int ret, call_frame_t *sync_frame, void *data)
+{
+ gf_listener_stop (sync_frame->this);
+
+ STACK_DESTROY (sync_frame->root);
+ kill (getpid(), SIGTERM);
+ return 0;
+}
+
+void *
+gf_defrag_start (void *data)
+{
+ int ret = -1;
+ call_frame_t *frame = NULL;
+ dht_conf_t *conf = NULL;
+ gf_defrag_info_t *defrag = NULL;
+ xlator_t *this = NULL;
+
+ this = data;
+ conf = this->private;
+ if (!conf)
+ goto out;
+
+ defrag = conf->defrag;
+ if (!defrag)
+ goto out;
+
+ frame = create_frame (this, this->ctx->pool);
+ if (!frame)
+ goto out;
+
+ frame->root->pid = GF_CLIENT_PID_DEFRAG;
+
+ defrag->pid = frame->root->pid;
+
+ defrag->defrag_status = GF_DEFRAG_STATUS_STARTED;
+
+ ret = synctask_new (this->ctx->env, gf_defrag_start_crawl,
+ gf_defrag_done, frame, this);
+
+ if (ret)
+ gf_log (this->name, GF_LOG_ERROR, "Could not create"
+ " task for rebalance");
+out:
+ return NULL;
+}
+
+int
+gf_defrag_status_get (gf_defrag_info_t *defrag, dict_t *dict)
+{
+ int ret = 0;
+ uint64_t files = 0;
+ uint64_t size = 0;
+ uint64_t lookup = 0;
+ uint64_t failures = 0;
+ uint64_t skipped = 0;
+ char *status = "";
+ double elapsed = 0;
+ struct timeval end = {0,};
+
+
+ if (!defrag)
+ goto out;
+
+ ret = 0;
+ if (defrag->defrag_status == GF_DEFRAG_STATUS_NOT_STARTED)
+ goto out;
+
+ files = defrag->total_files;
+ size = defrag->total_data;
+ lookup = defrag->num_files_lookedup;
+ failures = defrag->total_failures;
+ skipped = defrag->skipped;
+
+ gettimeofday (&end, NULL);
+
+ elapsed = end.tv_sec - defrag->start_time.tv_sec;
+
+ if (!dict)
+ goto log;
+
+ ret = dict_set_uint64 (dict, "files", files);
+ if (ret)
+ gf_log (THIS->name, GF_LOG_WARNING,
+ "failed to set file count");
+
+ ret = dict_set_uint64 (dict, "size", size);
+ if (ret)
+ gf_log (THIS->name, GF_LOG_WARNING,
+ "failed to set size of xfer");
+
+ ret = dict_set_uint64 (dict, "lookups", lookup);
+ if (ret)
+ gf_log (THIS->name, GF_LOG_WARNING,
+ "failed to set lookedup file count");
+
+
+ ret = dict_set_int32 (dict, "status", defrag->defrag_status);
+ if (ret)
+ gf_log (THIS->name, GF_LOG_WARNING,
+ "failed to set status");
+ if (elapsed) {
+ ret = dict_set_double (dict, "run-time", elapsed);
+ if (ret)
+ gf_log (THIS->name, GF_LOG_WARNING,
+ "failed to set run-time");
+ }
+
+ ret = dict_set_uint64 (dict, "failures", failures);
+ if (ret)
+ gf_log (THIS->name, GF_LOG_WARNING,
+ "failed to set failure count");
+
+ ret = dict_set_uint64 (dict, "skipped", skipped);
+ if (ret)
+ gf_log (THIS->name, GF_LOG_WARNING,
+ "failed to set skipped file count");
+log:
+ switch (defrag->defrag_status) {
+ case GF_DEFRAG_STATUS_NOT_STARTED:
+ status = "not started";
+ break;
+ case GF_DEFRAG_STATUS_STARTED:
+ status = "in progress";
+ break;
+ case GF_DEFRAG_STATUS_STOPPED:
+ status = "stopped";
+ break;
+ case GF_DEFRAG_STATUS_COMPLETE:
+ status = "completed";
+ break;
+ case GF_DEFRAG_STATUS_FAILED:
+ status = "failed";
+ break;
+ default:
+ break;
+ }
+
+ gf_log (THIS->name, GF_LOG_INFO, "Rebalance is %s. Time taken is %.2f "
+ "secs", status, elapsed);
+ gf_log (THIS->name, GF_LOG_INFO, "Files migrated: %"PRIu64", size: %"
+ PRIu64", lookups: %"PRIu64", failures: %"PRIu64", skipped: "
+ "%"PRIu64, files, size, lookup, failures, skipped);
+
+
+out:
+ return 0;
+}
+
+int
+gf_defrag_stop (gf_defrag_info_t *defrag, dict_t *output)
+{
+ /* TODO: set a variable 'stop_defrag' here, it should be checked
+ in defrag loop */
+ int ret = -1;
+ GF_ASSERT (defrag);
+
+ if (defrag->defrag_status == GF_DEFRAG_STATUS_NOT_STARTED) {
+ goto out;
+ }
+
+ gf_log ("", GF_LOG_INFO, "Received stop command on rebalance");
+ defrag->defrag_status = GF_DEFRAG_STATUS_STOPPED;
+
+ if (output)
+ gf_defrag_status_get (defrag, output);
+ ret = 0;
+out:
+ gf_log ("", GF_LOG_DEBUG, "Returning %d", ret);
+ return ret;
+}
diff --git a/xlators/cluster/dht/src/dht-rename.c b/xlators/cluster/dht/src/dht-rename.c
index fa06de711..5d6f4f232 100644
--- a/xlators/cluster/dht/src/dht-rename.c
+++ b/xlators/cluster/dht/src/dht-rename.c
@@ -1,20 +1,11 @@
/*
- Copyright (c) 2008-2011 Gluster, Inc. <http://www.gluster.com>
+ Copyright (c) 2008-2012 Red Hat, Inc. <http://www.redhat.com>
This file is part of GlusterFS.
- GlusterFS is free software; you can redistribute it and/or modify
- it under the terms of the GNU General Public License as published
- by the Free Software Foundation; either version 3 of the License,
- or (at your option) any later version.
-
- GlusterFS is distributed in the hope that it will be useful, but
- WITHOUT ANY WARRANTY; without even the implied warranty of
- MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU
- General Public License for more details.
-
- You should have received a copy of the GNU General Public License
- along with this program. If not, see
- <http://www.gnu.org/licenses/>.
+ This file is licensed to you under your choice of the GNU Lesser
+ General Public License, version 3 or any later version (LGPLv3 or
+ later), or the GNU General Public License, version 2 (GPLv2), in all
+ cases as published by the Free Software Foundation.
*/
/* TODO: link(oldpath, newpath) fails if newpath already exists. DHT should
@@ -35,7 +26,8 @@ int
dht_rename_dir_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
int32_t op_ret, int32_t op_errno, struct iatt *stbuf,
struct iatt *preoldparent, struct iatt *postoldparent,
- struct iatt *prenewparent, struct iatt *postnewparent)
+ struct iatt *prenewparent, struct iatt *postnewparent,
+ dict_t *xdata)
{
dht_local_t *local = NULL;
int this_call_cnt = 0;
@@ -84,7 +76,7 @@ unwind:
DHT_STACK_UNWIND (rename, frame, local->op_ret, local->op_errno,
&local->stbuf, &local->preoldparent,
&local->postoldparent,
- &local->preparent, &local->postparent);
+ &local->preparent, &local->postparent, xdata);
}
return 0;
@@ -97,7 +89,7 @@ dht_rename_hashed_dir_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
struct iatt *preoldparent,
struct iatt *postoldparent,
struct iatt *prenewparent,
- struct iatt *postnewparent)
+ struct iatt *postnewparent, dict_t *xdata)
{
dht_conf_t *conf = NULL;
dht_local_t *local = NULL;
@@ -147,7 +139,7 @@ dht_rename_hashed_dir_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
STACK_WIND (frame, dht_rename_dir_cbk,
conf->subvolumes[i],
conf->subvolumes[i]->fops->rename,
- &local->loc, &local->loc2);
+ &local->loc, &local->loc2, NULL);
if (!--call_cnt)
break;
}
@@ -164,7 +156,7 @@ unwind:
DHT_STACK_UNWIND (rename, frame, local->op_ret, local->op_errno,
&local->stbuf, &local->preoldparent,
&local->postoldparent,
- &local->preparent, &local->postparent);
+ &local->preparent, &local->postparent, NULL);
return 0;
}
@@ -185,19 +177,20 @@ dht_rename_dir_do (call_frame_t *frame, xlator_t *this)
STACK_WIND (frame, dht_rename_hashed_dir_cbk,
local->dst_hashed,
local->dst_hashed->fops->rename,
- &local->loc, &local->loc2);
+ &local->loc, &local->loc2, NULL);
return 0;
err:
DHT_STACK_UNWIND (rename, frame, local->op_ret, local->op_errno, NULL, NULL,
- NULL, NULL, NULL);
+ NULL, NULL, NULL, NULL);
return 0;
}
int
dht_rename_readdir_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
- int op_ret, int op_errno, gf_dirent_t *entries)
+ int op_ret, int op_errno, gf_dirent_t *entries,
+ dict_t *xdata)
{
dht_local_t *local = NULL;
int this_call_cnt = -1;
@@ -226,7 +219,7 @@ dht_rename_readdir_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
int
dht_rename_opendir_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
- int op_ret, int op_errno, fd_t *fd)
+ int op_ret, int op_errno, fd_t *fd, dict_t *xdata)
{
dht_local_t *local = NULL;
int this_call_cnt = -1;
@@ -246,7 +239,7 @@ dht_rename_opendir_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
STACK_WIND (frame, dht_rename_readdir_cbk,
prev->this, prev->this->fops->readdir,
- local->fd, 4096, 0);
+ local->fd, 4096, 0, NULL);
return 0;
@@ -302,22 +295,54 @@ dht_rename_dir (call_frame_t *frame, xlator_t *this)
STACK_WIND (frame, dht_rename_opendir_cbk,
conf->subvolumes[i],
conf->subvolumes[i]->fops->opendir,
- &local->loc2, local->fd);
+ &local->loc2, local->fd, NULL);
}
return 0;
err:
op_errno = (op_errno == -1) ? errno : op_errno;
- DHT_STACK_UNWIND (rename, frame, -1, op_errno, NULL, NULL, NULL, NULL, NULL);
+ DHT_STACK_UNWIND (rename, frame, -1, op_errno, NULL, NULL, NULL, NULL,
+ NULL, NULL);
return 0;
}
+#define DHT_MARK_FOP_INTERNAL(xattr) do { \
+ int tmp = -1; \
+ if (!xattr) { \
+ xattr = dict_new (); \
+ if (!xattr) \
+ break; \
+ } \
+ tmp = dict_set_str (xattr, GLUSTERFS_INTERNAL_FOP_KEY, "yes"); \
+ if (tmp) { \
+ gf_log (this->name, GF_LOG_ERROR, "Failed to set" \
+ " internal dict key for %s", local->loc.path); \
+ } \
+ }while (0)
+int
+dht_rename_done (call_frame_t *frame, xlator_t *this)
+{
+ dht_local_t *local = NULL;
+
+ local = frame->local;
+
+ if (local->linked == _gf_true) {
+ local->linked = _gf_false;
+ dht_linkfile_attr_heal (frame, this);
+ }
+ DHT_STRIP_PHASE1_FLAGS (&local->stbuf);
+ DHT_STACK_UNWIND (rename, frame, local->op_ret, local->op_errno,
+ &local->stbuf, &local->preoldparent,
+ &local->postoldparent, &local->preparent,
+ &local->postparent, NULL);
+ return 0;
+}
int
dht_rename_unlink_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
int32_t op_ret, int32_t op_errno, struct iatt *preparent,
- struct iatt *postparent)
+ struct iatt *postparent, dict_t *xdata)
{
dht_local_t *local = NULL;
call_frame_t *prev = NULL;
@@ -346,11 +371,7 @@ dht_rename_unlink_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
WIPE (&local->postparent);
if (is_last_call (this_call_cnt)) {
- DHT_STRIP_PHASE1_FLAGS (&local->stbuf);
- DHT_STACK_UNWIND (rename, frame, local->op_ret, local->op_errno,
- &local->stbuf, &local->preoldparent,
- &local->postoldparent, &local->preparent,
- &local->postparent);
+ dht_rename_done (frame, this);
}
out:
@@ -368,7 +389,7 @@ dht_rename_cleanup (call_frame_t *frame)
xlator_t *dst_hashed = NULL;
xlator_t *dst_cached = NULL;
int call_cnt = 0;
-
+ dict_t *xattr = NULL;
local = frame->local;
this = frame->this;
@@ -392,13 +413,15 @@ dht_rename_cleanup (call_frame_t *frame)
if (!call_cnt)
goto nolinks;
+ DHT_MARK_FOP_INTERNAL (xattr);
+
if (dst_hashed != src_hashed && dst_hashed != src_cached) {
gf_log (this->name, GF_LOG_TRACE,
"unlinking linkfile %s @ %s => %s",
local->loc.path, dst_hashed->name, src_cached->name);
STACK_WIND (frame, dht_rename_unlink_cbk,
dst_hashed, dst_hashed->fops->unlink,
- &local->loc);
+ &local->loc, 0, xattr);
}
if (src_cached != dst_hashed) {
@@ -407,9 +430,12 @@ dht_rename_cleanup (call_frame_t *frame)
local->loc2.path, src_cached->name);
STACK_WIND (frame, dht_rename_unlink_cbk,
src_cached, src_cached->fops->unlink,
- &local->loc2);
+ &local->loc2, 0, xattr);
}
+ if (xattr)
+ dict_unref (xattr);
+
return 0;
nolinks:
@@ -422,7 +448,7 @@ nolinks:
DHT_STACK_UNWIND (rename, frame, local->op_ret, local->op_errno,
&local->stbuf, &local->preoldparent,
&local->postoldparent, &local->preparent,
- &local->postparent);
+ &local->postparent, NULL);
return 0;
}
@@ -430,9 +456,10 @@ nolinks:
int
dht_rename_links_create_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
- int32_t op_ret, int32_t op_errno,
- inode_t *inode, struct iatt *stbuf,
- struct iatt *preparent, struct iatt *postparent)
+ int32_t op_ret, int32_t op_errno,
+ inode_t *inode, struct iatt *stbuf,
+ struct iatt *preparent, struct iatt *postparent,
+ dict_t *xdata)
{
call_frame_t *prev = NULL;
dht_local_t *local = NULL;
@@ -446,6 +473,10 @@ dht_rename_links_create_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
local->loc.path, prev->this->name, strerror (op_errno));
}
+ if (local->linked == _gf_true) {
+ local->linked = _gf_false;
+ dht_linkfile_attr_heal (frame, this);
+ }
DHT_STACK_DESTROY (frame);
return 0;
@@ -456,7 +487,8 @@ int
dht_rename_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
int32_t op_ret, int32_t op_errno, struct iatt *stbuf,
struct iatt *preoldparent, struct iatt *postoldparent,
- struct iatt *prenewparent, struct iatt *postnewparent)
+ struct iatt *prenewparent, struct iatt *postnewparent,
+ dict_t *xdata)
{
dht_local_t *local = NULL;
call_frame_t *prev = NULL;
@@ -467,6 +499,7 @@ dht_rename_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
xlator_t *rename_subvol = NULL;
call_frame_t *link_frame = NULL;
dht_local_t *link_local = NULL;
+ dict_t *xattr = NULL;
local = frame->local;
prev = cookie;
@@ -476,6 +509,8 @@ dht_rename_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
dst_hashed = local->dst_hashed;
dst_cached = local->dst_cached;
+ if (local->linked == _gf_true)
+ FRAME_SU_UNDO (frame, dht_local_t);
if (op_ret == -1) {
gf_log (this->name, GF_LOG_WARNING,
"%s: rename on %s failed (%s)", local->loc.path,
@@ -505,15 +540,25 @@ dht_rename_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
uuid_copy (link_local->gfid, local->loc.inode->gfid);
dht_linkfile_create (link_frame, dht_rename_links_create_cbk,
- src_cached, dst_hashed, &link_local->loc);
+ this, src_cached, dst_hashed,
+ &link_local->loc);
}
err:
- dht_iatt_merge (this, &local->stbuf, stbuf, prev->this);
- dht_iatt_merge (this, &local->preoldparent, preoldparent, prev->this);
- dht_iatt_merge (this, &local->postoldparent, postoldparent, prev->this);
- dht_iatt_merge (this, &local->preparent, prenewparent, prev->this);
- dht_iatt_merge (this, &local->postparent, postnewparent, prev->this);
+ /* Merge attrs only from src_cached. In case there of src_cached !=
+ * dst_hashed, this ignores linkfile attrs. */
+ if (prev->this == src_cached) {
+ dht_iatt_merge (this, &local->stbuf, stbuf, prev->this);
+ dht_iatt_merge (this, &local->preoldparent, preoldparent,
+ prev->this);
+ dht_iatt_merge (this, &local->postoldparent, postoldparent,
+ prev->this);
+ dht_iatt_merge (this, &local->preparent, prenewparent,
+ prev->this);
+ dht_iatt_merge (this, &local->postparent, postnewparent,
+ prev->this);
+ }
+
/* NOTE: rename_subvol is the same subvolume from which dht_rename_cbk
* is called. since rename has already happened on rename_subvol,
@@ -538,6 +583,8 @@ err:
if (local->call_cnt == 0)
goto unwind;
+ DHT_MARK_FOP_INTERNAL (xattr);
+
if (src_cached != dst_hashed && src_cached != dst_cached) {
gf_log (this->name, GF_LOG_TRACE,
"deleting old src datafile %s @ %s",
@@ -545,7 +592,7 @@ err:
STACK_WIND (frame, dht_rename_unlink_cbk,
src_cached, src_cached->fops->unlink,
- &local->loc);
+ &local->loc, 0, xattr);
}
if (src_hashed != rename_subvol && src_hashed != src_cached) {
@@ -555,7 +602,7 @@ err:
STACK_WIND (frame, dht_rename_unlink_cbk,
src_hashed, src_hashed->fops->unlink,
- &local->loc);
+ &local->loc, 0, xattr);
}
if (dst_cached
@@ -567,8 +614,10 @@ err:
STACK_WIND (frame, dht_rename_unlink_cbk,
dst_cached, dst_cached->fops->unlink,
- &local->loc2);
+ &local->loc2, 0, xattr);
}
+ if (xattr)
+ dict_unref (xattr);
return 0;
unwind:
@@ -576,16 +625,16 @@ unwind:
WIPE (&local->postoldparent);
WIPE (&local->preparent);
WIPE (&local->postparent);
+ if (xattr)
+ dict_unref (xattr);
- DHT_STRIP_PHASE1_FLAGS (&local->stbuf);
- DHT_STACK_UNWIND (rename, frame, local->op_ret, local->op_errno,
- &local->stbuf, &local->preoldparent,
- &local->postoldparent, &local->preparent,
- &local->postparent);
+ dht_rename_done (frame, this);
return 0;
cleanup:
+ if (xattr)
+ dict_unref (xattr);
dht_rename_cleanup (frame);
return 0;
@@ -619,9 +668,11 @@ dht_do_rename (call_frame_t *frame)
"renaming %s => %s (%s)",
local->loc.path, local->loc2.path, rename_subvol->name);
+ if (local->linked == _gf_true)
+ FRAME_SU_DO (frame, dht_local_t);
STACK_WIND (frame, dht_rename_cbk,
rename_subvol, rename_subvol->fops->rename,
- &local->loc, &local->loc2);
+ &local->loc, &local->loc2, NULL);
return 0;
}
@@ -631,7 +682,8 @@ int
dht_rename_links_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
int32_t op_ret, int32_t op_errno,
inode_t *inode, struct iatt *stbuf,
- struct iatt *preparent, struct iatt *postparent)
+ struct iatt *preparent, struct iatt *postparent,
+ dict_t *xdata)
{
dht_local_t *local = NULL;
call_frame_t *prev = NULL;
@@ -648,6 +700,9 @@ dht_rename_links_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
local->op_ret = -1;
if (op_errno != ENOENT)
local->op_errno = op_errno;
+ } else if (local->src_cached == prev->this) {
+ /* merge of attr returned only from linkfile creation */
+ dht_iatt_merge (this, &local->stbuf, stbuf, prev->this);
}
this_call_cnt = dht_frame_return (frame);
@@ -670,7 +725,8 @@ cleanup:
int
dht_rename_unlink_links_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
int32_t op_ret, int32_t op_errno,
- struct iatt *preparent, struct iatt *postparent)
+ struct iatt *preparent, struct iatt *postparent,
+ dict_t *xdata)
{
dht_local_t *local = NULL;
call_frame_t *prev = NULL;
@@ -712,6 +768,7 @@ dht_rename_create_links (call_frame_t *frame)
xlator_t *dst_hashed = NULL;
xlator_t *dst_cached = NULL;
int call_cnt = 0;
+ dict_t *xattr = NULL;
local = frame->local;
@@ -722,6 +779,7 @@ dht_rename_create_links (call_frame_t *frame)
dst_hashed = local->dst_hashed;
dst_cached = local->dst_cached;
+ DHT_MARK_FOP_INTERNAL (xattr);
if (src_cached == dst_cached) {
if (dst_hashed == dst_cached)
@@ -733,7 +791,7 @@ dht_rename_create_links (call_frame_t *frame)
STACK_WIND (frame, dht_rename_unlink_links_cbk,
dst_hashed, dst_hashed->fops->unlink,
- &local->loc2);
+ &local->loc2, 0, xattr);
return 0;
}
@@ -750,7 +808,7 @@ dht_rename_create_links (call_frame_t *frame)
"linkfile %s @ %s => %s",
local->loc.path, dst_hashed->name, src_cached->name);
memcpy (local->gfid, local->loc.inode->gfid, 16);
- dht_linkfile_create (frame, dht_rename_links_cbk,
+ dht_linkfile_create (frame, dht_rename_links_cbk, this,
src_cached, dst_hashed, &local->loc);
}
@@ -760,7 +818,7 @@ dht_rename_create_links (call_frame_t *frame)
local->loc2.path, src_cached->name);
STACK_WIND (frame, dht_rename_links_cbk,
src_cached, src_cached->fops->link,
- &local->loc, &local->loc2);
+ &local->loc, &local->loc2, xattr);
}
nolinks:
@@ -768,6 +826,8 @@ nolinks:
/* skip to next step */
dht_do_rename (frame);
}
+ if (xattr)
+ dict_unref (xattr);
return 0;
}
@@ -775,7 +835,7 @@ nolinks:
int
dht_rename (call_frame_t *frame, xlator_t *this,
- loc_t *oldloc, loc_t *newloc)
+ loc_t *oldloc, loc_t *newloc, dict_t *xdata)
{
xlator_t *src_cached = NULL;
xlator_t *src_hashed = NULL;
@@ -857,7 +917,8 @@ dht_rename (call_frame_t *frame, xlator_t *this,
err:
op_errno = (op_errno == -1) ? errno : op_errno;
- DHT_STACK_UNWIND (rename, frame, -1, op_errno, NULL, NULL, NULL, NULL, NULL);
+ DHT_STACK_UNWIND (rename, frame, -1, op_errno, NULL, NULL, NULL, NULL,
+ NULL, NULL);
return 0;
}
diff --git a/xlators/cluster/dht/src/dht-selfheal.c b/xlators/cluster/dht/src/dht-selfheal.c
index b87e0ab48..3fe96b1c7 100644
--- a/xlators/cluster/dht/src/dht-selfheal.c
+++ b/xlators/cluster/dht/src/dht-selfheal.c
@@ -1,20 +1,11 @@
/*
- Copyright (c) 2008-2011 Gluster, Inc. <http://www.gluster.com>
+ Copyright (c) 2008-2012 Red Hat, Inc. <http://www.redhat.com>
This file is part of GlusterFS.
- GlusterFS is free software; you can redistribute it and/or modify
- it under the terms of the GNU General Public License as published
- by the Free Software Foundation; either version 3 of the License,
- or (at your option) any later version.
-
- GlusterFS is distributed in the hope that it will be useful, but
- WITHOUT ANY WARRANTY; without even the implied warranty of
- MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU
- General Public License for more details.
-
- You should have received a copy of the GNU General Public License
- along with this program. If not, see
- <http://www.gnu.org/licenses/>.
+ This file is licensed to you under your choice of the GNU Lesser
+ General Public License, version 3 or any later version (LGPLv3 or
+ later), or the GNU General Public License, version 2 (GPLv2), in all
+ cases as published by the Free Software Foundation.
*/
#ifndef _CONFIG_H
@@ -26,7 +17,7 @@
#include "glusterfs.h"
#include "xlator.h"
#include "dht-common.h"
-
+#include "glusterfs-acl.h"
#define DHT_SET_LAYOUT_RANGE(layout,i,srt,chunk,cnt,path) do { \
layout->list[i].start = srt; \
@@ -38,43 +29,40 @@
layout->list[i].xlator->name, path); \
} while (0)
+#define DHT_RESET_LAYOUT_RANGE(layout) do { \
+ int cnt = 0; \
+ for (cnt = 0; cnt < layout->cnt; cnt++ ) { \
+ layout->list[cnt].start = 0; \
+ layout->list[cnt].stop = 0; \
+ } \
+ } while (0)
-static inline uint32_t
-dht_find_overlap (int idx, int cnk_idx, uint32_t start, uint32_t stop,
- uint32_t chunk_size)
+static uint32_t
+dht_overlap_calc (dht_layout_t *old, int o, dht_layout_t *new, int n)
{
- uint32_t overlap = 0;
- uint32_t chunk_begin = 0;
+ if (o >= old->cnt || n >= new->cnt)
+ return 0;
- chunk_begin = cnk_idx * chunk_size;
+ if (old->list[o].err > 0 || new->list[n].err > 0)
+ return 0;
- /* There is no chance of overlap */
- if ((chunk_begin > stop) ||
- ((chunk_begin + chunk_size) < start))
- goto out;
-
- if ((chunk_begin <= start) &&
- ((chunk_begin + chunk_size) <= stop)) {
- overlap = ((chunk_begin + chunk_size) - start);
- goto out;
+ if (old->list[o].start == old->list[o].stop) {
+ return 0;
}
- if ((chunk_begin <= start) &&
- ((chunk_begin + chunk_size) >= stop)) {
- overlap = (stop - start);
- goto out;
+ if (new->list[n].start == new->list[n].stop) {
+ return 0;
}
- if ((chunk_begin < stop) &&
- ((chunk_begin + chunk_size) >= stop)) {
- overlap = (stop - chunk_begin);
- goto out;
- }
+ if ((old->list[o].start > new->list[n].stop) ||
+ (old->list[o].stop < new->list[n].start))
+ return 0;
-out:
- return overlap;
+ return min (old->list[o].stop, new->list[n].stop) -
+ max (old->list[o].start, new->list[n].start) + 1;
}
+
int
dht_selfheal_dir_finish (call_frame_t *frame, xlator_t *this, int ret)
{
@@ -82,7 +70,7 @@ dht_selfheal_dir_finish (call_frame_t *frame, xlator_t *this, int ret)
local = frame->local;
local->selfheal.dir_cbk (frame, NULL, frame->this, ret,
- local->op_errno);
+ local->op_errno, NULL);
return 0;
}
@@ -90,7 +78,7 @@ dht_selfheal_dir_finish (call_frame_t *frame, xlator_t *this, int ret)
int
dht_selfheal_dir_xattr_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
- int op_ret, int op_errno)
+ int op_ret, int op_errno, dict_t *xdata)
{
dht_local_t *local = NULL;
call_frame_t *prev = NULL;
@@ -129,7 +117,8 @@ dht_selfheal_dir_xattr_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
int
dht_selfheal_dir_xattr_persubvol (call_frame_t *frame, loc_t *loc,
- dht_layout_t *layout, int i)
+ dht_layout_t *layout, int i,
+ xlator_t *req_subvol)
{
xlator_t *subvol = NULL;
dict_t *xattr = NULL;
@@ -137,12 +126,23 @@ dht_selfheal_dir_xattr_persubvol (call_frame_t *frame, loc_t *loc,
xlator_t *this = NULL;
int32_t *disk_layout = NULL;
dht_local_t *local = NULL;
-
+ dht_conf_t *conf = NULL;
local = frame->local;
- subvol = layout->list[i].xlator;
+ if (req_subvol)
+ subvol = req_subvol;
+ else
+ subvol = layout->list[i].xlator;
this = frame->this;
+ GF_VALIDATE_OR_GOTO ("", this, err);
+ GF_VALIDATE_OR_GOTO (this->name, layout, err);
+ GF_VALIDATE_OR_GOTO (this->name, local, err);
+ GF_VALIDATE_OR_GOTO (this->name, subvol, err);
+ VALIDATE_OR_GOTO (this->private, err);
+
+ conf = this->private;
+
xattr = get_new_dict ();
if (!xattr) {
goto err;
@@ -156,8 +156,7 @@ dht_selfheal_dir_xattr_persubvol (call_frame_t *frame, loc_t *loc,
goto err;
}
- ret = dict_set_bin (xattr, "trusted.glusterfs.dht",
- disk_layout, 4 * 4);
+ ret = dict_set_bin (xattr, conf->xattr_name, disk_layout, 4 * 4);
if (ret == -1) {
gf_log (this->name, GF_LOG_WARNING,
"%s: (subvol %s) failed to set xattr dictionary",
@@ -178,7 +177,7 @@ dht_selfheal_dir_xattr_persubvol (call_frame_t *frame, loc_t *loc,
STACK_WIND (frame, dht_selfheal_dir_xattr_cbk,
subvol, subvol->fops->setxattr,
- loc, xattr, 0);
+ loc, xattr, 0, NULL);
dict_unref (xattr);
@@ -188,11 +187,10 @@ err:
if (xattr)
dict_destroy (xattr);
- if (disk_layout)
- GF_FREE (disk_layout);
+ GF_FREE (disk_layout);
dht_selfheal_dir_xattr_cbk (frame, subvol, frame->this,
- -1, ENOMEM);
+ -1, ENOMEM, NULL);
return 0;
}
@@ -203,21 +201,42 @@ dht_fix_dir_xattr (call_frame_t *frame, loc_t *loc, dht_layout_t *layout)
int i = 0;
int count = 0;
xlator_t *this = NULL;
+ dht_conf_t *conf = NULL;
+ dht_layout_t *dummy = NULL;
local = frame->local;
this = frame->this;
+ conf = this->private;
gf_log (this->name, GF_LOG_DEBUG,
"writing the new range for all subvolumes");
- local->call_cnt = count = layout->cnt;
+ local->call_cnt = count = conf->subvolume_cnt;
for (i = 0; i < layout->cnt; i++) {
- dht_selfheal_dir_xattr_persubvol (frame, loc, layout, i);
+ dht_selfheal_dir_xattr_persubvol (frame, loc, layout, i, NULL);
if (--count == 0)
- break;
+ goto out;
+ }
+ /* if we are here, subvolcount > layout_count. subvols-per-directory
+ * option might be set here. We need to clear out layout from the
+ * non-participating subvolumes, else it will result in overlaps */
+ dummy = dht_layout_new (this, 1);
+ if (!dummy)
+ goto out;
+ for (i = 0; i < conf->subvolume_cnt; i++) {
+ if (_gf_false ==
+ dht_is_subvol_in_layout (layout, conf->subvolumes[i])) {
+ dht_selfheal_dir_xattr_persubvol (frame, loc, dummy, 0,
+ conf->subvolumes[i]);
+ if (--count == 0)
+ break;
+ }
}
+
+ dht_layout_unref (this, dummy);
+out:
return 0;
}
@@ -228,9 +247,12 @@ dht_selfheal_dir_xattr (call_frame_t *frame, loc_t *loc, dht_layout_t *layout)
int missing_xattr = 0;
int i = 0;
xlator_t *this = NULL;
+ dht_conf_t *conf = NULL;
+ dht_layout_t *dummy = NULL;
local = frame->local;
this = frame->this;
+ conf = this->private;
for (i = 0; i < layout->cnt; i++) {
if (layout->list[i].err != -1 || !layout->list[i].stop) {
@@ -259,18 +281,30 @@ dht_selfheal_dir_xattr (call_frame_t *frame, loc_t *loc, dht_layout_t *layout)
if (layout->list[i].err != -1 || !layout->list[i].stop)
continue;
- dht_selfheal_dir_xattr_persubvol (frame, loc, layout, i);
+ dht_selfheal_dir_xattr_persubvol (frame, loc, layout, i, NULL);
if (--missing_xattr == 0)
break;
}
+ dummy = dht_layout_new (this, 1);
+ if (!dummy)
+ goto out;
+ for (i = 0; i < conf->subvolume_cnt; i++) {
+ if (_gf_false ==
+ dht_is_subvol_in_layout (layout, conf->subvolumes[i])) {
+ dht_selfheal_dir_xattr_persubvol (frame, loc, dummy, 0,
+ conf->subvolumes[i]);
+ }
+ }
+ dht_layout_unref (this, dummy);
+out:
return 0;
}
int
dht_selfheal_dir_setattr_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
int op_ret, int op_errno, struct iatt *statpre,
- struct iatt *statpost)
+ struct iatt *statpost, dict_t *xdata)
{
dht_local_t *local = NULL;
dht_layout_t *layout = NULL;
@@ -324,7 +358,7 @@ dht_selfheal_dir_setattr (call_frame_t *frame, loc_t *loc, struct iatt *stbuf,
STACK_WIND (frame, dht_selfheal_dir_setattr_cbk,
layout->list[i].xlator,
layout->list[i].xlator->fops->setattr,
- loc, stbuf, valid);
+ loc, stbuf, valid, NULL);
}
}
@@ -335,7 +369,8 @@ int
dht_selfheal_dir_mkdir_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
int op_ret, int op_errno,
inode_t *inode, struct iatt *stbuf,
- struct iatt *preparent, struct iatt *postparent)
+ struct iatt *preparent, struct iatt *postparent,
+ dict_t *xdata)
{
dht_local_t *local = NULL;
dht_layout_t *layout = NULL;
@@ -381,6 +416,46 @@ out:
return 0;
}
+void
+dht_selfheal_dir_mkdir_setacl (dict_t *xattr, dict_t *dict)
+{
+ data_t *acl_default = NULL;
+ data_t *acl_access = NULL;
+ xlator_t *this = NULL;
+ int ret = -1;
+
+ GF_ASSERT (xattr);
+ GF_ASSERT (dict);
+
+ this = THIS;
+ GF_ASSERT (this);
+
+ acl_default = dict_get (xattr, POSIX_ACL_DEFAULT_XATTR);
+
+ if (!acl_default) {
+ gf_log (this->name, GF_LOG_DEBUG,
+ "ACL_DEFAULT xattr not present");
+ goto cont;
+ }
+ ret = dict_set (dict, POSIX_ACL_DEFAULT_XATTR, acl_default);
+ if (ret)
+ gf_log (this->name, GF_LOG_WARNING,
+ "Could not set ACL_DEFAULT xattr");
+cont:
+ acl_access = dict_get (xattr, POSIX_ACL_ACCESS_XATTR);
+ if (!acl_access) {
+ gf_log (this->name, GF_LOG_DEBUG,
+ "ACL_ACCESS xattr not present");
+ goto out;
+ }
+ ret = dict_set (dict, POSIX_ACL_ACCESS_XATTR, acl_access);
+ if (ret)
+ gf_log (this->name, GF_LOG_WARNING,
+ "Could not set ACL_ACCESS xattr");
+
+out:
+ return;
+}
int
dht_selfheal_dir_mkdir (call_frame_t *frame, loc_t *loc,
@@ -414,16 +489,19 @@ dht_selfheal_dir_mkdir (call_frame_t *frame, loc_t *loc,
ret = dict_set_static_bin (dict, "gfid-req", local->gfid, 16);
if (ret)
- gf_log (this->name, GF_LOG_INFO,
+ gf_log (this->name, GF_LOG_WARNING,
"%s: failed to set gfid in dict", loc->path);
} else if (local->params) {
/* Send the dictionary from higher layers directly */
dict = dict_ref (local->params);
}
+ /* Set acls */
+ if (local->xattr && dict)
+ dht_selfheal_dir_mkdir_setacl (local->xattr, dict);
if (!dict)
gf_log (this->name, GF_LOG_WARNING,
- "dict is NULL, need to make sure gfid's are same");
+ "dict is NULL, need to make sure gfids are same");
for (i = 0; i < layout->cnt; i++) {
if (layout->list[i].err == ENOENT || force) {
@@ -437,7 +515,7 @@ dht_selfheal_dir_mkdir (call_frame_t *frame, loc_t *loc,
loc,
st_mode_from_ia (local->stbuf.ia_prot,
local->stbuf.ia_type),
- dict);
+ 0, dict);
}
}
@@ -456,7 +534,7 @@ dht_selfheal_layout_alloc_start (xlator_t *this, loc_t *loc,
uint32_t hashval = 0;
int ret = 0;
- ret = dht_hash_compute (layout->type, loc->path, &hashval);
+ ret = dht_hash_compute (this, layout->type, loc->path, &hashval);
if (ret == 0) {
start = (hashval % layout->cnt);
}
@@ -479,7 +557,7 @@ dht_get_layout_count (xlator_t *this, dht_layout_t *layout, int new_layout)
for (j = 0; j < conf->subvolume_cnt; j++) {
if (conf->decommissioned_bricks[j] &&
conf->decommissioned_bricks[j] == layout->list[i].xlator) {
- layout->list[i].err = -EINVAL;
+ layout->list[i].err = EINVAL;
break;
}
}
@@ -487,9 +565,33 @@ dht_get_layout_count (xlator_t *this, dht_layout_t *layout, int new_layout)
for (i = 0; i < layout->cnt; i++) {
err = layout->list[i].err;
- if (err == -1 || err == 0) {
- layout->list[i].err = -1;
+ if (err == -1 || err == 0 || err == ENOENT) {
+ /* Setting list[i].err = -1 is an indication for
+ dht_selfheal_layout_new_directory() to assign
+ a range. We set it to -1 based on any one of
+ the three criteria:
+
+ - err == -1 already, which means directory
+ existed but layout was not set on it.
+
+ - err == 0, which means directory exists and
+ has an old layout piece which will be
+ overwritten now.
+
+ - err == ENOENT, which means directory does
+ not exist (possibly racing with mkdir or
+ finishing half done mkdir). The missing
+ directory will be attempted to be recreated.
+
+ It is important to note that it is safe
+ to race with mkdir() as self-heal and
+ mkdir are idempotent operations. Both will
+ strive to set the directory and layouts to
+ the same final state.
+ */
count++;
+ if (!err)
+ layout->list[i].err = -1;
}
}
@@ -504,49 +606,126 @@ dht_get_layout_count (xlator_t *this, dht_layout_t *layout, int new_layout)
}
}
- count = ((layout->spread_cnt) ? layout->spread_cnt :
- ((count) ? count : 1));
+ /* if layout->spread_cnt is set, check if it is <= available
+ * subvolumes (down brick and decommissioned bricks are considered
+ * un-availbale). Else return count (available up bricks) */
+ count = ((layout->spread_cnt &&
+ (layout->spread_cnt <= count)) ?
+ layout->spread_cnt : ((count) ? count : 1));
return count;
}
+void dht_selfheal_layout_new_directory (call_frame_t *frame, loc_t *loc,
+ dht_layout_t *new_layout);
+
+void dht_layout_entry_swap (dht_layout_t *layout, int i, int j);
+void dht_layout_range_swap (dht_layout_t *layout, int i, int j);
+
+/*
+ * It's a bit icky using local variables in a macro, but it makes the rest
+ * of the code a lot clearer.
+ */
+#define OV_ENTRY(x,y) table[x*new->cnt+y]
+
+void
+dht_selfheal_layout_maximize_overlap (call_frame_t *frame, loc_t *loc,
+ dht_layout_t *new, dht_layout_t *old)
+{
+ int i = 0;
+ int j = 0;
+ uint32_t curr_overlap = 0;
+ uint32_t max_overlap = 0;
+ int max_overlap_idx = -1;
+ uint32_t overlap = 0;
+ uint32_t *table = NULL;
+
+ dht_layout_sort_volname (old);
+ /* Now both old_layout->list[] and new_layout->list[]
+ are match the same xlators/subvolumes. i.e,
+ old_layout->[i] and new_layout->[i] are referring
+ to the same subvolumes
+ */
+
+ /* Build a table of overlaps between new[i] and old[j]. */
+ table = alloca(sizeof(overlap)*old->cnt*new->cnt);
+ if (!table) {
+ return;
+ }
+ memset(table,0,sizeof(overlap)*old->cnt*new->cnt);
+ for (i = 0; i < new->cnt; ++i) {
+ for (j = 0; j < old->cnt; ++j) {
+ OV_ENTRY(i,j) = dht_overlap_calc(old,j,new,i);
+ }
+ }
+
+ for (i = 0; i < new->cnt; i++) {
+ if (new->list[i].err > 0) {
+ /* Subvol might be marked for decommission
+ with EINVAL, or some other serious error
+ marked with positive errno.
+ */
+ continue;
+ }
+
+ max_overlap = 0;
+ max_overlap_idx = i;
+ for (j = (i + 1); j < new->cnt; ++j) {
+ if (new->list[j].err > 0) {
+ /* Subvol might be marked for decommission
+ with EINVAL, or some other serious error
+ marked with positive errno.
+ */
+ continue;
+ }
+ /* Calculate the overlap now. */
+ curr_overlap = OV_ENTRY(i,i) + OV_ENTRY(j,j);
+ /* Calculate the overlap after the proposed swap. */
+ overlap = OV_ENTRY(i,j) + OV_ENTRY(j,i);
+ /* Are we better than status quo? */
+ if (overlap > curr_overlap) {
+ overlap -= curr_overlap;
+ /* Are we better than the previous choice? */
+ if (overlap > max_overlap) {
+ max_overlap = overlap;
+ max_overlap_idx = j;
+ }
+ }
+ }
+
+ if (max_overlap_idx != i) {
+ dht_layout_range_swap (new, i, max_overlap_idx);
+ /* Need to swap the table values too. */
+ for (j = 0; j < old->cnt; ++j) {
+ overlap = OV_ENTRY(i,j);
+ OV_ENTRY(i,j) = OV_ENTRY(max_overlap_idx,j);
+ OV_ENTRY(max_overlap_idx,j) = overlap;
+ }
+ }
+ }
+}
+
+
dht_layout_t *
dht_fix_layout_of_directory (call_frame_t *frame, loc_t *loc,
dht_layout_t *layout)
{
- uint32_t chunk = 0;
- uint32_t start = 0;
- uint32_t stop = 0;
- uint32_t overlap = 0;
- uint32_t max_overlap = 0;
- uint32_t chunk_begin = 0;
- int count = 0;
- int cnt = 0;
int i = 0;
- int j = 0;
- int k = 0;
- int loop_cnt = 0;
- int start_subvol = 0;
- int *fix_array = NULL;
xlator_t *this = NULL;
dht_layout_t *new_layout = NULL;
dht_conf_t *priv = NULL;
dht_local_t *local = NULL;
+ uint32_t subvol_down = 0;
+ int ret = 0;
this = frame->this;
priv = this->private;
local = frame->local;
- count = cnt = dht_get_layout_count (this, layout, 0);
-
- chunk = ((unsigned long) 0xffffffff) / ((cnt) ? cnt : 1);
-
- start_subvol = dht_selfheal_layout_alloc_start (this, loc, layout);
-
- fix_array = GF_CALLOC (sizeof (int), layout->cnt, gf_common_mt_char);
- if (!fix_array) {
- /* No fix, use the existing layout itself */
+ if (layout->type == DHT_HASH_TYPE_DM_USER) {
+ gf_log (THIS->name, GF_LOG_DEBUG, "leaving %s alone",
+ loc->path);
goto done;
}
@@ -554,98 +733,33 @@ dht_fix_layout_of_directory (call_frame_t *frame, loc_t *loc,
if (!new_layout)
goto done;
- for (i = 0; i < new_layout->cnt; i++) {
- /* TODO: fix this in layout_alloc() itself */
- new_layout->list[i].err = -ENOENT;
- if (i < layout->cnt)
- new_layout->list[i].xlator = layout->list[i].xlator;
- }
-
- /* Check if there are any overlap in layout, and give the proper fix */
- for (i = 0; i < layout->cnt; i++) {
- /* No need to fix if 'err' is not '-1' */
- if (layout->list[i].err != -1)
- continue;
-
- /* If already existing layout is having no range, skip it */
- start = layout->list[i].start;
- stop = layout->list[i].stop;
- if ((stop - start) == 0)
- continue;
-
- max_overlap = 0;
-
- /* 'j' is used as starting point of each chunk */
- for (j = 1; j <= count; j++) {
- /* if chunk is already used, don't use it again */
- for (k = 0; k < i; k++)
- if (j == fix_array[k])
- break;
- if (k < i)
- continue;
-
- overlap = dht_find_overlap (i, (j-1), start, stop, chunk);
- if (max_overlap < overlap) {
- max_overlap = overlap;
- fix_array[i] = j;
- }
- }
-
- /* If we have any overlap, then use that itself as new
- layout for the subvolume */
- if (fix_array[i]) {
- chunk_begin = chunk * (fix_array[i] - 1);
- new_layout->list[i].err = -1;
- DHT_SET_LAYOUT_RANGE (new_layout, i, chunk_begin,
- chunk, cnt, loc->path);
- /* make sure to give (max - 1) as 'stop' range,
- if it is last chunk */
- if (fix_array[i] == count)
- new_layout->list[i].stop = 0xffffffff;
- if (--cnt == 0)
- goto done;
+ /* If a subvolume is down, do not re-write the layout. */
+ ret = dht_layout_anomalies (this, loc, layout, NULL, NULL, NULL,
+ &subvol_down, NULL, NULL);
- }
+ if (subvol_down || (ret == -1)) {
+ gf_log (this->name, GF_LOG_WARNING, "%u subvolume(s) are down"
+ ". Skipping fix layout.", subvol_down);
+ GF_FREE (new_layout);
+ return NULL;
}
- /* Now, look for layouts which are not having any overlaps
- and give it a fix */
- for (loop_cnt = 0, i = start_subvol; loop_cnt < new_layout->cnt;
- i++, loop_cnt++) {
- if (i == new_layout->cnt)
- i = 0;
-
- /* If 'fix_array[i]' is set, the layout is already fixed. */
- if (fix_array[i])
- continue;
+ for (i = 0; i < new_layout->cnt; i++) {
+ if (layout->list[i].err != ENOSPC)
+ new_layout->list[i].err = layout->list[i].err;
+ else
+ new_layout->list[i].err = -1;
- if (layout->list[i].err != -1) {
- new_layout->list[i].err = layout->list[i].err;
- continue;
- }
+ new_layout->list[i].xlator = layout->list[i].xlator;
+ }
- for (k = 1; k <= count; k++) {
- for (j = 0; j < new_layout->cnt; j++) {
- if (k == fix_array[j])
- break;
- }
- /* Didn't find any of the list begining with 'k' */
- if (j == new_layout->cnt)
- break;
- }
+ /* First give it a layout as though it is a new directory. This
+ ensures rotation to kick in */
+ dht_layout_sort_volname (new_layout);
+ dht_selfheal_layout_new_directory (frame, loc, new_layout);
- fix_array[i] = k;
- chunk_begin = (k - 1) * chunk;
- new_layout->list[i].err = -1;
- DHT_SET_LAYOUT_RANGE (new_layout, i, chunk_begin, chunk, cnt,
- loc->path);
- /* make sure to give (max - 1) as 'stop' range,
- if it is last chunk */
- if (k == count)
- new_layout->list[i].stop = 0xffffffff;
- if (--cnt == 0)
- goto done;
- }
+ /* Now selectively re-assign ranges only when it helps */
+ dht_selfheal_layout_maximize_overlap (frame, loc, new_layout, layout);
done:
if (new_layout) {
@@ -659,7 +773,7 @@ done:
local->layout = new_layout;
}
- return new_layout;
+ return local->layout;
}
@@ -683,9 +797,11 @@ dht_selfheal_layout_new_directory (call_frame_t *frame, loc_t *loc,
start_subvol = dht_selfheal_layout_alloc_start (this, loc, layout);
+ /* clear out the range, as we are re-computing here */
+ DHT_RESET_LAYOUT_RANGE (layout);
for (i = start_subvol; i < layout->cnt; i++) {
err = layout->list[i].err;
- if (err == -1) {
+ if (err == -1 || err == ENOENT) {
DHT_SET_LAYOUT_RANGE(layout, i, start, chunk,
cnt, loc->path);
if (--cnt == 0) {
@@ -698,7 +814,7 @@ dht_selfheal_layout_new_directory (call_frame_t *frame, loc_t *loc,
for (i = 0; i < start_subvol; i++) {
err = layout->list[i].err;
- if (err == -1) {
+ if (err == -1 || err == ENOENT) {
DHT_SET_LAYOUT_RANGE(layout, i, start, chunk,
cnt, loc->path);
if (--cnt == 0) {
@@ -717,35 +833,17 @@ int
dht_selfheal_dir_getafix (call_frame_t *frame, loc_t *loc,
dht_layout_t *layout)
{
- dht_conf_t *conf = NULL;
- xlator_t *this = NULL;
dht_local_t *local = NULL;
- int missing = -1;
- int down = -1;
- int holes = -1;
+ uint32_t holes = 0;
int ret = -1;
int i = -1;
- int overlaps = -1;
+ uint32_t overlaps = 0;
- this = frame->this;
- conf = this->private;
local = frame->local;
- missing = local->selfheal.missing;
- down = local->selfheal.down;
holes = local->selfheal.hole_cnt;
overlaps = local->selfheal.overlaps_cnt;
- if ((missing + down) == conf->subvolume_cnt) {
- dht_selfheal_layout_new_directory (frame, loc, layout);
- ret = 0;
- }
-
- if (holes <= down) {
- /* the down subvol might fill up the holes */
- ret = 0;
- }
-
if (holes || overlaps) {
dht_selfheal_layout_new_directory (frame, loc, layout);
ret = 0;
@@ -797,6 +895,9 @@ dht_fix_directory_layout (call_frame_t *frame,
/* No layout sorting required here */
tmp_layout = dht_fix_layout_of_directory (frame, &local->loc, layout);
+ if (!tmp_layout) {
+ return -1;
+ }
dht_fix_dir_xattr (frame, &local->loc, tmp_layout);
return 0;
@@ -819,9 +920,8 @@ dht_selfheal_directory (call_frame_t *frame, dht_selfheal_dir_cbk_t dir_cbk,
dht_layout_anomalies (this, loc, layout,
&local->selfheal.hole_cnt,
&local->selfheal.overlaps_cnt,
- &local->selfheal.missing,
- &local->selfheal.down,
- &local->selfheal.misc);
+ NULL, &local->selfheal.down,
+ &local->selfheal.misc, NULL);
down = local->selfheal.down;
misc = local->selfheal.misc;
@@ -830,14 +930,14 @@ dht_selfheal_directory (call_frame_t *frame, dht_selfheal_dir_cbk_t dir_cbk,
local->selfheal.layout = dht_layout_ref (this, layout);
if (down) {
- gf_log (this->name, GF_LOG_INFO,
+ gf_log (this->name, GF_LOG_WARNING,
"%d subvolumes down -- not fixing", down);
ret = 0;
goto sorry_no_fix;
}
if (misc) {
- gf_log (this->name, GF_LOG_INFO,
+ gf_log (this->name, GF_LOG_WARNING,
"%d subvolumes have unrecoverable errors", misc);
ret = 0;
goto sorry_no_fix;
@@ -847,7 +947,7 @@ dht_selfheal_directory (call_frame_t *frame, dht_selfheal_dir_cbk_t dir_cbk,
ret = dht_selfheal_dir_getafix (frame, loc, layout);
if (ret == -1) {
- gf_log (this->name, GF_LOG_INFO,
+ gf_log (this->name, GF_LOG_WARNING,
"not able to form layout for the directory");
goto sorry_no_fix;
}
@@ -880,3 +980,50 @@ dht_selfheal_restore (call_frame_t *frame, dht_selfheal_dir_cbk_t dir_cbk,
return ret;
}
+
+int
+dht_dir_attr_heal (void *data)
+{
+ call_frame_t *frame = NULL;
+ dht_local_t *local = NULL;
+ xlator_t *subvol = NULL;
+ xlator_t *this = NULL;
+ dht_conf_t *conf = NULL;
+ int call_cnt = 0;
+ int ret = -1;
+ int i = 0;
+
+ GF_VALIDATE_OR_GOTO ("dht", data, out);
+
+ frame = data;
+ local = frame->local;
+ this = frame->this;
+ GF_VALIDATE_OR_GOTO ("dht", this, out);
+ GF_VALIDATE_OR_GOTO ("dht", local, out);
+ conf = this->private;
+ GF_VALIDATE_OR_GOTO ("dht", conf, out);
+
+ call_cnt = conf->subvolume_cnt;
+
+ for (i = 0; i < call_cnt; i++) {
+ subvol = conf->subvolumes[i];
+ if (!subvol || (subvol == dht_first_up_subvol (this)))
+ continue;
+ ret = syncop_setattr (subvol, &local->loc, &local->stbuf,
+ (GF_SET_ATTR_UID | GF_SET_ATTR_GID),
+ NULL, NULL);
+ if (ret)
+ gf_log ("dht", GF_LOG_ERROR, "Failed to set uid/gid on"
+ " %s on %s subvol (%s)", local->loc.path,
+ subvol->name, strerror (errno));
+ }
+out:
+ return 0;
+}
+
+int
+dht_dir_attr_heal_done (int ret, call_frame_t *sync_frame, void *data)
+{
+ DHT_STACK_DESTROY (sync_frame);
+ return 0;
+}
diff --git a/xlators/cluster/dht/src/dht-shared.c b/xlators/cluster/dht/src/dht-shared.c
new file mode 100644
index 000000000..70aac7710
--- /dev/null
+++ b/xlators/cluster/dht/src/dht-shared.c
@@ -0,0 +1,758 @@
+/*
+ Copyright (c) 2008-2012 Red Hat, Inc. <http://www.redhat.com>
+ This file is part of GlusterFS.
+
+ This file is licensed to you under your choice of the GNU Lesser
+ General Public License, version 3 or any later version (LGPLv3 or
+ later), or the GNU General Public License, version 2 (GPLv2), in all
+ cases as published by the Free Software Foundation.
+*/
+
+
+#ifndef _CONFIG_H
+#define _CONFIG_H
+#include "config.h"
+#endif
+
+/* TODO: add NS locking */
+
+#include "statedump.h"
+#include "dht-common.h"
+
+/* TODO:
+ - use volumename in xattr instead of "dht"
+ - use NS locks
+ - handle all cases in self heal layout reconstruction
+ - complete linkfile selfheal
+*/
+struct volume_options options[];
+
+void
+dht_layout_dump (dht_layout_t *layout, const char *prefix)
+{
+
+ char key[GF_DUMP_MAX_BUF_LEN];
+ int i = 0;
+
+ if (!layout)
+ goto out;
+ if (!prefix)
+ goto out;
+
+ gf_proc_dump_build_key(key, prefix, "cnt");
+ gf_proc_dump_write(key, "%d", layout->cnt);
+ gf_proc_dump_build_key(key, prefix, "preset");
+ gf_proc_dump_write(key, "%d", layout->preset);
+ gf_proc_dump_build_key(key, prefix, "gen");
+ gf_proc_dump_write(key, "%d", layout->gen);
+ if (layout->type != IA_INVAL) {
+ gf_proc_dump_build_key(key, prefix, "inode type");
+ gf_proc_dump_write(key, "%d", layout->type);
+ }
+
+ if (!IA_ISDIR (layout->type))
+ goto out;
+
+ for (i = 0; i < layout->cnt; i++) {
+ gf_proc_dump_build_key(key, prefix,"list[%d].err", i);
+ gf_proc_dump_write(key, "%d", layout->list[i].err);
+ gf_proc_dump_build_key(key, prefix,"list[%d].start", i);
+ gf_proc_dump_write(key, "%u", layout->list[i].start);
+ gf_proc_dump_build_key(key, prefix,"list[%d].stop", i);
+ gf_proc_dump_write(key, "%u", layout->list[i].stop);
+ if (layout->list[i].xlator) {
+ gf_proc_dump_build_key(key, prefix,
+ "list[%d].xlator.type", i);
+ gf_proc_dump_write(key, "%s",
+ layout->list[i].xlator->type);
+ gf_proc_dump_build_key(key, prefix,
+ "list[%d].xlator.name", i);
+ gf_proc_dump_write(key, "%s",
+ layout->list[i].xlator->name);
+ }
+ }
+
+out:
+ return;
+}
+
+
+int32_t
+dht_priv_dump (xlator_t *this)
+{
+ char key_prefix[GF_DUMP_MAX_BUF_LEN];
+ char key[GF_DUMP_MAX_BUF_LEN];
+ int i = 0;
+ dht_conf_t *conf = NULL;
+ int ret = -1;
+
+ if (!this)
+ goto out;
+
+ conf = this->private;
+ if (!conf)
+ goto out;
+
+ ret = TRY_LOCK(&conf->subvolume_lock);
+ if (ret != 0) {
+ return ret;
+ }
+
+ gf_proc_dump_add_section("xlator.cluster.dht.%s.priv", this->name);
+ gf_proc_dump_build_key(key_prefix,"xlator.cluster.dht","%s.priv",
+ this->name);
+ gf_proc_dump_write("subvol_cnt","%d", conf->subvolume_cnt);
+ for (i = 0; i < conf->subvolume_cnt; i++) {
+ sprintf (key, "subvolumes[%d]", i);
+ gf_proc_dump_write(key, "%s.%s", conf->subvolumes[i]->type,
+ conf->subvolumes[i]->name);
+ if (conf->file_layouts && conf->file_layouts[i]){
+ sprintf (key, "file_layouts[%d]", i);
+ dht_layout_dump(conf->file_layouts[i], key);
+ }
+ if (conf->dir_layouts && conf->dir_layouts[i]) {
+ sprintf (key, "dir_layouts[%d]", i);
+ dht_layout_dump(conf->dir_layouts[i], key);
+ }
+ if (conf->subvolume_status) {
+
+ sprintf (key, "subvolume_status[%d]", i);
+ gf_proc_dump_write(key, "%d",
+ (int)conf->subvolume_status[i]);
+ }
+
+ }
+
+ gf_proc_dump_write("search_unhashed", "%d", conf->search_unhashed);
+ gf_proc_dump_write("gen", "%d", conf->gen);
+ gf_proc_dump_write("min_free_disk", "%lf", conf->min_free_disk);
+ gf_proc_dump_write("min_free_inodes", "%lf", conf->min_free_inodes);
+ gf_proc_dump_write("disk_unit", "%c", conf->disk_unit);
+ gf_proc_dump_write("refresh_interval", "%d", conf->refresh_interval);
+ gf_proc_dump_write("unhashed_sticky_bit", "%d", conf->unhashed_sticky_bit);
+ if (conf ->du_stats) {
+ gf_proc_dump_write("du_stats.avail_percent", "%lf",
+ conf->du_stats->avail_percent);
+ gf_proc_dump_write("du_stats.avail_space", "%lu",
+ conf->du_stats->avail_space);
+ gf_proc_dump_write("du_stats.avail_inodes", "%lf",
+ conf->du_stats->avail_inodes);
+ gf_proc_dump_write("du_stats.log", "%lu", conf->du_stats->log);
+ }
+
+ if (conf->last_stat_fetch.tv_sec)
+ gf_proc_dump_write("last_stat_fetch", "%s",
+ ctime(&conf->last_stat_fetch.tv_sec));
+
+ UNLOCK(&conf->subvolume_lock);
+
+out:
+ return ret;
+}
+
+int32_t
+dht_inodectx_dump (xlator_t *this, inode_t *inode)
+{
+ int ret = -1;
+ dht_layout_t *layout = NULL;
+
+ if (!this)
+ goto out;
+ if (!inode)
+ goto out;
+
+ ret = dht_inode_ctx_layout_get (inode, this, &layout);
+
+ if ((ret != 0) || !layout)
+ return ret;
+
+ gf_proc_dump_add_section("xlator.cluster.dht.%s.inode", this->name);
+ dht_layout_dump(layout, "layout");
+
+out:
+ return ret;
+}
+
+void
+dht_fini (xlator_t *this)
+{
+ int i = 0;
+ dht_conf_t *conf = NULL;
+
+ GF_VALIDATE_OR_GOTO ("dht", this, out);
+
+ conf = this->private;
+ this->private = NULL;
+ if (conf) {
+ if (conf->file_layouts) {
+ for (i = 0; i < conf->subvolume_cnt; i++) {
+ GF_FREE (conf->file_layouts[i]);
+ }
+ GF_FREE (conf->file_layouts);
+ }
+
+ GF_FREE (conf->subvolumes);
+
+ GF_FREE (conf->subvolume_status);
+
+ GF_FREE (conf);
+ }
+out:
+ return;
+}
+
+int32_t
+mem_acct_init (xlator_t *this)
+{
+ int ret = -1;
+
+ GF_VALIDATE_OR_GOTO ("dht", this, out);
+
+ ret = xlator_mem_acct_init (this, gf_dht_mt_end + 1);
+
+ if (ret != 0) {
+ gf_log (this->name, GF_LOG_ERROR, "Memory accounting init"
+ "failed");
+ return ret;
+ }
+out:
+ return ret;
+}
+
+
+int
+dht_parse_decommissioned_bricks (xlator_t *this, dht_conf_t *conf,
+ const char *bricks)
+{
+ int i = 0;
+ int ret = -1;
+ char *tmpstr = NULL;
+ char *dup_brick = NULL;
+ char *node = NULL;
+
+ if (!conf || !bricks)
+ goto out;
+
+ dup_brick = gf_strdup (bricks);
+ node = strtok_r (dup_brick, ",", &tmpstr);
+ while (node) {
+ for (i = 0; i < conf->subvolume_cnt; i++) {
+ if (!strcmp (conf->subvolumes[i]->name, node)) {
+ conf->decommissioned_bricks[i] =
+ conf->subvolumes[i];
+ conf->decommission_subvols_cnt++;
+ gf_log (this->name, GF_LOG_INFO,
+ "decommissioning subvolume %s",
+ conf->subvolumes[i]->name);
+ break;
+ }
+ }
+ if (i == conf->subvolume_cnt) {
+ /* Wrong node given. */
+ goto out;
+ }
+ node = strtok_r (NULL, ",", &tmpstr);
+ }
+
+ ret = 0;
+ conf->decommission_in_progress = 1;
+out:
+ GF_FREE (dup_brick);
+
+ return ret;
+}
+
+
+int
+dht_decommissioned_remove (xlator_t *this, dht_conf_t *conf)
+{
+ int i = 0;
+ int ret = -1;
+
+ if (!conf)
+ goto out;
+
+ for (i = 0; i < conf->subvolume_cnt; i++) {
+ if (conf->decommissioned_bricks[i]) {
+ conf->decommissioned_bricks[i] = NULL;
+ conf->decommission_subvols_cnt--;
+ }
+ }
+
+ ret = 0;
+out:
+
+ return ret;
+}
+void
+dht_init_regex (xlator_t *this, dict_t *odict, char *name,
+ regex_t *re, gf_boolean_t *re_valid)
+{
+ char *temp_str;
+
+ if (dict_get_str (odict, name, &temp_str) != 0) {
+ if (strcmp(name,"rsync-hash-regex")) {
+ return;
+ }
+ temp_str = "^\\.(.+)\\.[^.]+$";
+ }
+
+ if (*re_valid) {
+ regfree(re);
+ *re_valid = _gf_false;
+ }
+
+ if (!strcmp(temp_str,"none")) {
+ return;
+ }
+
+ if (regcomp(re,temp_str,REG_EXTENDED) == 0) {
+ gf_log (this->name, GF_LOG_INFO,
+ "using regex %s = %s", name, temp_str);
+ *re_valid = _gf_true;
+ }
+ else {
+ gf_log (this->name, GF_LOG_WARNING,
+ "compiling regex %s failed", temp_str);
+ }
+}
+
+int
+dht_reconfigure (xlator_t *this, dict_t *options)
+{
+ dht_conf_t *conf = NULL;
+ char *temp_str = NULL;
+ gf_boolean_t search_unhashed;
+ int ret = -1;
+
+ GF_VALIDATE_OR_GOTO ("dht", this, out);
+ GF_VALIDATE_OR_GOTO ("dht", options, out);
+
+ conf = this->private;
+ if (!conf)
+ return 0;
+
+ if (dict_get_str (options, "lookup-unhashed", &temp_str) == 0) {
+ /* If option is not "auto", other options _should_ be boolean*/
+ if (strcasecmp (temp_str, "auto")) {
+ if (!gf_string2boolean (temp_str, &search_unhashed)) {
+ gf_log(this->name, GF_LOG_DEBUG, "Reconfigure:"
+ " lookup-unhashed reconfigured (%s)",
+ temp_str);
+ conf->search_unhashed = search_unhashed;
+ } else {
+ gf_log(this->name, GF_LOG_ERROR, "Reconfigure:"
+ " lookup-unhashed should be boolean,"
+ " not (%s), defaulting to (%d)",
+ temp_str, conf->search_unhashed);
+ //return -1;
+ ret = -1;
+ goto out;
+ }
+ } else {
+ gf_log(this->name, GF_LOG_DEBUG, "Reconfigure:"
+ " lookup-unhashed reconfigured auto ");
+ conf->search_unhashed = GF_DHT_LOOKUP_UNHASHED_AUTO;
+ }
+ }
+
+ GF_OPTION_RECONF ("min-free-disk", conf->min_free_disk, options,
+ percent_or_size, out);
+ /* option can be any one of percent or bytes */
+ conf->disk_unit = 0;
+ if (conf->min_free_disk < 100.0)
+ conf->disk_unit = 'p';
+
+ GF_OPTION_RECONF ("min-free-inodes", conf->min_free_inodes, options,
+ percent, out);
+
+ GF_OPTION_RECONF ("directory-layout-spread", conf->dir_spread_cnt,
+ options, uint32, out);
+
+ GF_OPTION_RECONF ("readdir-optimize", conf->readdir_optimize, options,
+ bool, out);
+ if (conf->defrag) {
+ GF_OPTION_RECONF ("rebalance-stats", conf->defrag->stats,
+ options, bool, out);
+ }
+
+ if (dict_get_str (options, "decommissioned-bricks", &temp_str) == 0) {
+ ret = dht_parse_decommissioned_bricks (this, conf, temp_str);
+ if (ret == -1)
+ goto out;
+ } else {
+ ret = dht_decommissioned_remove (this, conf);
+ if (ret == -1)
+ goto out;
+ }
+
+ dht_init_regex (this, options, "rsync-hash-regex",
+ &conf->rsync_regex, &conf->rsync_regex_valid);
+ dht_init_regex (this, options, "extra-hash-regex",
+ &conf->extra_regex, &conf->extra_regex_valid);
+
+ ret = 0;
+out:
+ return ret;
+}
+
+static int
+gf_defrag_pattern_list_fill (xlator_t *this, gf_defrag_info_t *defrag, char *data)
+{
+ int ret = -1;
+ char *tmp_str = NULL;
+ char *tmp_str1 = NULL;
+ char *dup_str = NULL;
+ char *num = NULL;
+ char *pattern_str = NULL;
+ char *pattern = NULL;
+ gf_defrag_pattern_list_t *temp_list = NULL;
+ gf_defrag_pattern_list_t *pattern_list = NULL;
+
+ if (!this || !defrag || !data)
+ goto out;
+
+ /* Get the pattern for pattern list. "pattern:<optional-size>"
+ * eg: *avi, *pdf:10MB, *:1TB
+ */
+ pattern_str = strtok_r (data, ",", &tmp_str);
+ while (pattern_str) {
+ dup_str = gf_strdup (pattern_str);
+ pattern_list = GF_CALLOC (1, sizeof (gf_defrag_pattern_list_t),
+ 1);
+ if (!pattern_list) {
+ goto out;
+ }
+ pattern = strtok_r (dup_str, ":", &tmp_str1);
+ num = strtok_r (NULL, ":", &tmp_str1);
+ if (!pattern)
+ goto out;
+ if (!num) {
+ if (gf_string2bytesize(pattern, &pattern_list->size)
+ == 0) {
+ pattern = "*";
+ }
+ } else if (gf_string2bytesize (num, &pattern_list->size) != 0) {
+ gf_log (this->name, GF_LOG_ERROR,
+ "invalid number format \"%s\"", num);
+ goto out;
+ }
+ memcpy (pattern_list->path_pattern, pattern, strlen (dup_str));
+
+ if (!defrag->defrag_pattern)
+ temp_list = NULL;
+ else
+ temp_list = defrag->defrag_pattern;
+
+ pattern_list->next = temp_list;
+
+ defrag->defrag_pattern = pattern_list;
+ pattern_list = NULL;
+
+ GF_FREE (dup_str);
+ dup_str = NULL;
+
+ pattern_str = strtok_r (NULL, ",", &tmp_str);
+ }
+
+ ret = 0;
+out:
+ if (ret)
+ GF_FREE (pattern_list);
+ GF_FREE (dup_str);
+
+ return ret;
+}
+
+int
+dht_init (xlator_t *this)
+{
+ dht_conf_t *conf = NULL;
+ char *temp_str = NULL;
+ int ret = -1;
+ int i = 0;
+ gf_defrag_info_t *defrag = NULL;
+ int cmd = 0;
+ char *node_uuid = NULL;
+
+
+ GF_VALIDATE_OR_GOTO ("dht", this, err);
+
+ if (!this->children) {
+ gf_log (this->name, GF_LOG_CRITICAL,
+ "Distribute needs more than one subvolume");
+ return -1;
+ }
+
+ if (!this->parents) {
+ gf_log (this->name, GF_LOG_WARNING,
+ "dangling volume. check volfile");
+ }
+
+ conf = GF_CALLOC (1, sizeof (*conf), gf_dht_mt_dht_conf_t);
+ if (!conf) {
+ goto err;
+ }
+
+ ret = dict_get_int32 (this->options, "rebalance-cmd", &cmd);
+
+ if (cmd) {
+ defrag = GF_CALLOC (1, sizeof (gf_defrag_info_t),
+ gf_defrag_info_mt);
+
+ GF_VALIDATE_OR_GOTO (this->name, defrag, err);
+
+ LOCK_INIT (&defrag->lock);
+
+ defrag->is_exiting = 0;
+
+ conf->defrag = defrag;
+
+ ret = dict_get_str (this->options, "node-uuid", &node_uuid);
+ if (ret) {
+ gf_log (this->name, GF_LOG_ERROR, "node-uuid not "
+ "specified");
+ goto err;
+ }
+
+ if (uuid_parse (node_uuid, defrag->node_uuid)) {
+ gf_log (this->name, GF_LOG_ERROR, "Cannot parse "
+ "glusterd node uuid");
+ goto err;
+ }
+
+ defrag->cmd = cmd;
+
+ defrag->stats = _gf_false;
+ }
+
+ conf->search_unhashed = GF_DHT_LOOKUP_UNHASHED_ON;
+ if (dict_get_str (this->options, "lookup-unhashed", &temp_str) == 0) {
+ /* If option is not "auto", other options _should_ be boolean */
+ if (strcasecmp (temp_str, "auto"))
+ gf_string2boolean (temp_str, &conf->search_unhashed);
+ else
+ conf->search_unhashed = GF_DHT_LOOKUP_UNHASHED_AUTO;
+ }
+
+ GF_OPTION_INIT ("unhashed-sticky-bit", conf->unhashed_sticky_bit, bool,
+ err);
+
+ GF_OPTION_INIT ("use-readdirp", conf->use_readdirp, bool, err);
+
+ GF_OPTION_INIT ("min-free-disk", conf->min_free_disk, percent_or_size,
+ err);
+
+ GF_OPTION_INIT ("min-free-inodes", conf->min_free_inodes, percent,
+ err);
+
+ conf->dir_spread_cnt = conf->subvolume_cnt;
+ GF_OPTION_INIT ("directory-layout-spread", conf->dir_spread_cnt,
+ uint32, err);
+
+ GF_OPTION_INIT ("assert-no-child-down", conf->assert_no_child_down,
+ bool, err);
+
+ GF_OPTION_INIT ("readdir-optimize", conf->readdir_optimize, bool, err);
+
+ if (defrag) {
+ GF_OPTION_INIT ("rebalance-stats", defrag->stats, bool, err);
+ if (dict_get_str (this->options, "rebalance-filter", &temp_str)
+ == 0) {
+ if (gf_defrag_pattern_list_fill (this, defrag, temp_str)
+ == -1) {
+ gf_log (this->name, GF_LOG_ERROR, "Cannot parse"
+ " rebalance-filter (%s)", temp_str);
+ goto err;
+ }
+ }
+ }
+
+ /* option can be any one of percent or bytes */
+ conf->disk_unit = 0;
+ if (conf->min_free_disk < 100)
+ conf->disk_unit = 'p';
+
+ ret = dht_init_subvolumes (this, conf);
+ if (ret == -1) {
+ goto err;
+ }
+
+ if (dict_get_str (this->options, "decommissioned-bricks", &temp_str) == 0) {
+ ret = dht_parse_decommissioned_bricks (this, conf, temp_str);
+ if (ret == -1)
+ goto err;
+ }
+
+ dht_init_regex (this, this->options, "rsync-hash-regex",
+ &conf->rsync_regex, &conf->rsync_regex_valid);
+ dht_init_regex (this, this->options, "extra-hash-regex",
+ &conf->extra_regex, &conf->extra_regex_valid);
+
+ ret = dht_layouts_init (this, conf);
+ if (ret == -1) {
+ goto err;
+ }
+
+ LOCK_INIT (&conf->subvolume_lock);
+ LOCK_INIT (&conf->layout_lock);
+
+ conf->gen = 1;
+
+ this->local_pool = mem_pool_new (dht_local_t, 512);
+ if (!this->local_pool) {
+ gf_log (this->name, GF_LOG_ERROR,
+ "failed to create local_t's memory pool");
+ goto err;
+ }
+
+ GF_OPTION_INIT ("xattr-name", conf->xattr_name, str, err);
+ gf_asprintf (&conf->link_xattr_name, "%s.linkto", conf->xattr_name);
+ gf_asprintf (&conf->wild_xattr_name, "%s*", conf->xattr_name);
+ if (!conf->link_xattr_name || !conf->wild_xattr_name) {
+ goto err;
+ }
+
+ this->private = conf;
+
+ return 0;
+
+err:
+ if (conf) {
+ if (conf->file_layouts) {
+ for (i = 0; i < conf->subvolume_cnt; i++) {
+ GF_FREE (conf->file_layouts[i]);
+ }
+ GF_FREE (conf->file_layouts);
+ }
+
+ GF_FREE (conf->subvolumes);
+
+ GF_FREE (conf->subvolume_status);
+
+ GF_FREE (conf->du_stats);
+
+ GF_FREE (conf->defrag);
+
+ GF_FREE (conf->xattr_name);
+ GF_FREE (conf->link_xattr_name);
+ GF_FREE (conf->wild_xattr_name);
+
+ GF_FREE (conf);
+ }
+
+ return -1;
+}
+
+
+struct volume_options options[] = {
+ { .key = {"lookup-unhashed"},
+ .value = {"auto", "yes", "no", "enable", "disable", "1", "0",
+ "on", "off"},
+ .type = GF_OPTION_TYPE_STR,
+ .default_value = "on",
+ .description = "This option if set to ON, does a lookup through "
+ "all the sub-volumes, in case a lookup didn't return any result "
+ "from the hash subvolume. If set to OFF, it does not do a lookup "
+ "on the remaining subvolumes."
+ },
+ { .key = {"min-free-disk"},
+ .type = GF_OPTION_TYPE_PERCENT_OR_SIZET,
+ .default_value = "10%",
+ .description = "Percentage/Size of disk space, after which the "
+ "process starts balancing out the cluster, and logs will appear "
+ "in log files",
+ },
+ { .key = {"min-free-inodes"},
+ .type = GF_OPTION_TYPE_PERCENT,
+ .default_value = "5%",
+ .description = "after system has only N% of inodes, warnings "
+ "starts to appear in log files",
+ },
+ { .key = {"unhashed-sticky-bit"},
+ .type = GF_OPTION_TYPE_BOOL,
+ .default_value = "off",
+ },
+ { .key = {"use-readdirp"},
+ .type = GF_OPTION_TYPE_BOOL,
+ .default_value = "on",
+ .description = "This option if set to ON, forces the use of "
+ "readdirp, and hence also displays the stats of the files."
+ },
+ { .key = {"assert-no-child-down"},
+ .type = GF_OPTION_TYPE_BOOL,
+ .default_value = "off",
+ .description = "This option if set to ON, in the event of "
+ "CHILD_DOWN, will call exit."
+ },
+ { .key = {"directory-layout-spread"},
+ .type = GF_OPTION_TYPE_INT,
+ .min = 1,
+ .validate = GF_OPT_VALIDATE_MIN,
+ .description = "Specifies the directory layout spread."
+ },
+ { .key = {"decommissioned-bricks"},
+ .type = GF_OPTION_TYPE_ANY,
+ .description = "This option if set to ON, decommissions "
+ "the brick, so that no new data is allowed to be created "
+ "on that brick."
+ },
+ { .key = {"rebalance-cmd"},
+ .type = GF_OPTION_TYPE_INT,
+ },
+ { .key = {"node-uuid"},
+ .type = GF_OPTION_TYPE_STR,
+ },
+ { .key = {"rebalance-stats"},
+ .type = GF_OPTION_TYPE_BOOL,
+ .default_value = "off",
+ .description = "This option if set to ON displays and logs the "
+ " time taken for migration of each file, during the rebalance "
+ "process. If set to OFF, the rebalance logs will only display the "
+ "time spent in each directory."
+ },
+ { .key = {"readdir-optimize"},
+ .type = GF_OPTION_TYPE_BOOL,
+ .default_value = "off",
+ .description = "This option if set to ON enables the optimization "
+ "that allows DHT to requests non-first subvolumes to filter out "
+ "directory entries."
+ },
+ { .key = {"rsync-hash-regex"},
+ .type = GF_OPTION_TYPE_STR,
+ /* Setting a default here doesn't work. See dht_init_regex. */
+ .description = "Regular expression for stripping temporary-file "
+ "suffix and prefix used by rsync, to prevent relocation when the "
+ "file is renamed."
+ },
+ { .key = {"extra-hash-regex"},
+ .type = GF_OPTION_TYPE_STR,
+ /* Setting a default here doesn't work. See dht_init_regex. */
+ .description = "Regular expression for stripping temporary-file "
+ "suffix and prefix used by an application, to prevent relocation when "
+ "the file is renamed."
+ },
+ { .key = {"rebalance-filter"},
+ .type = GF_OPTION_TYPE_STR,
+ },
+
+ { .key = {"xattr-name"},
+ .type = GF_OPTION_TYPE_STR,
+ .default_value = "trusted.glusterfs.dht",
+ .description = "Base for extended attributes used by this "
+ "translator instance, to avoid conflicts with others above or "
+ "below it."
+ },
+
+ /* NUFA option */
+ { .key = {"local-volume-name"},
+ .type = GF_OPTION_TYPE_XLATOR
+ },
+
+ /* switch option */
+ { .key = {"pattern.switch.case"},
+ .type = GF_OPTION_TYPE_ANY
+ },
+
+ { .key = {NULL} },
+};
diff --git a/xlators/cluster/dht/src/dht.c b/xlators/cluster/dht/src/dht.c
index 8be573f51..fc0ca2f77 100644
--- a/xlators/cluster/dht/src/dht.c
+++ b/xlators/cluster/dht/src/dht.c
@@ -1,20 +1,11 @@
/*
- Copyright (c) 2008-2011 Gluster, Inc. <http://www.gluster.com>
+ Copyright (c) 2008-2012 Red Hat, Inc. <http://www.redhat.com>
This file is part of GlusterFS.
- GlusterFS is free software; you can redistribute it and/or modify
- it under the terms of the GNU General Public License as published
- by the Free Software Foundation; either version 3 of the License,
- or (at your option) any later version.
-
- GlusterFS is distributed in the hope that it will be useful, but
- WITHOUT ANY WARRANTY; without even the implied warranty of
- MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU
- General Public License for more details.
-
- You should have received a copy of the GNU General Public License
- along with this program. If not, see
- <http://www.gnu.org/licenses/>.
+ This file is licensed to you under your choice of the GNU Lesser
+ General Public License, version 3 or any later version (LGPLv3 or
+ later), or the GNU General Public License, version 2 (GPLv2), in all
+ cases as published by the Free Software Foundation.
*/
@@ -23,434 +14,15 @@
#include "config.h"
#endif
-/* TODO: add NS locking */
-
#include "statedump.h"
#include "dht-common.h"
-/* TODO:
- - use volumename in xattr instead of "dht"
- - use NS locks
- - handle all cases in self heal layout reconstruction
- - complete linkfile selfheal
-*/
-struct volume_options options[];
-
-void
-dht_layout_dump (dht_layout_t *layout, const char *prefix)
-{
-
- char key[GF_DUMP_MAX_BUF_LEN];
- int i = 0;
-
- GF_VALIDATE_OR_GOTO ("dht", layout, out);
- GF_VALIDATE_OR_GOTO ("dht", prefix, out);
-
- gf_proc_dump_build_key(key, prefix, "cnt");
- gf_proc_dump_write(key, "%d", layout->cnt);
- gf_proc_dump_build_key(key, prefix, "preset");
- gf_proc_dump_write(key, "%d", layout->preset);
- gf_proc_dump_build_key(key, prefix, "gen");
- gf_proc_dump_write(key, "%d", layout->gen);
- gf_proc_dump_build_key(key, prefix, "type");
- gf_proc_dump_write(key, "%d", layout->type);
-
- for (i = 0; i < layout->cnt; i++) {
- gf_proc_dump_build_key(key, prefix,"list[%d].err", i);
- gf_proc_dump_write(key, "%d", layout->list[i].err);
- gf_proc_dump_build_key(key, prefix,"list[%d].start", i);
- gf_proc_dump_write(key, "%u", layout->list[i].start);
- gf_proc_dump_build_key(key, prefix,"list[%d].stop", i);
- gf_proc_dump_write(key, "%u", layout->list[i].stop);
- if (layout->list[i].xlator) {
- gf_proc_dump_build_key(key, prefix,
- "list[%d].xlator.type", i);
- gf_proc_dump_write(key, "%s",
- layout->list[i].xlator->type);
- gf_proc_dump_build_key(key, prefix,
- "list[%d].xlator.name", i);
- gf_proc_dump_write(key, "%s",
- layout->list[i].xlator->name);
- }
- }
-
-out:
- return;
-}
-
-
-int32_t
-dht_priv_dump (xlator_t *this)
-{
- char key_prefix[GF_DUMP_MAX_BUF_LEN];
- char key[GF_DUMP_MAX_BUF_LEN];
- int i = 0;
- dht_conf_t *conf = NULL;
- int ret = -1;
-
- GF_VALIDATE_OR_GOTO ("dht", this, out);
-
- conf = this->private;
-
- if (!conf)
- return -1;
-
- ret = TRY_LOCK(&conf->subvolume_lock);
-
- if (ret != 0) {
- gf_log("", GF_LOG_WARNING, "Unable to lock dht subvolume %s",
- this->name);
- return ret;
- }
-
- gf_proc_dump_add_section("xlator.cluster.dht.%s.priv", this->name);
- gf_proc_dump_build_key(key_prefix,"xlator.cluster.dht","%s.priv",
- this->name);
- gf_proc_dump_write("subvol_cnt","%d", conf->subvolume_cnt);
- for (i = 0; i < conf->subvolume_cnt; i++) {
- sprintf (key, "subvolumes[%d]", i);
- gf_proc_dump_write(key, "%s.%s", conf->subvolumes[i]->type,
- conf->subvolumes[i]->name);
- if (conf->file_layouts && conf->file_layouts[i]){
- sprintf (key, "file_layouts[%d]", i);
- dht_layout_dump(conf->file_layouts[i], key);
- }
- if (conf->dir_layouts && conf->dir_layouts[i]) {
- sprintf (key, "dir_layouts[%d]", i);
- dht_layout_dump(conf->dir_layouts[i], key);
- }
- if (conf->subvolume_status) {
-
- sprintf (key, "subvolume_status[%d]", i);
- gf_proc_dump_write(key, "%d",
- (int)conf->subvolume_status[i]);
- }
-
- }
-
- gf_proc_dump_write("search_unhashed", "%d", conf->search_unhashed);
- gf_proc_dump_write("gen", "%d", conf->gen);
- gf_proc_dump_write("min_free_disk", "%lu", conf->min_free_disk);
- gf_proc_dump_write("min_free_inodes", "%lu", conf->min_free_inodes);
- gf_proc_dump_write("disk_unit", "%c", conf->disk_unit);
- gf_proc_dump_write("refresh_interval", "%d", conf->refresh_interval);
- gf_proc_dump_write("unhashed_sticky_bit", "%d", conf->unhashed_sticky_bit);
- if (conf ->du_stats) {
- gf_proc_dump_write("du_stats.avail_percent", "%lf",
- conf->du_stats->avail_percent);
- gf_proc_dump_write("du_stats.avail_space", "%lu",
- conf->du_stats->avail_space);
- gf_proc_dump_write("du_stats.avail_inodes", "%lf",
- conf->du_stats->avail_inodes);
- gf_proc_dump_write("du_stats.log", "%lu", conf->du_stats->log);
- }
- gf_proc_dump_write("last_stat_fetch", "%s", ctime(&conf->last_stat_fetch.tv_sec));
-
- UNLOCK(&conf->subvolume_lock);
-
-out:
- return ret;
-}
-
-int32_t
-dht_inodectx_dump (xlator_t *this, inode_t *inode)
-{
- int ret = -1;
- dht_layout_t *layout = NULL;
- uint64_t tmp_layout = 0;
-
- GF_VALIDATE_OR_GOTO ("dht", this, out);
- GF_VALIDATE_OR_GOTO ("dht", inode, out);
-
- ret = inode_ctx_get (inode, this, &tmp_layout);
-
- if (ret != 0)
- return ret;
-
- layout = (dht_layout_t *)(long)tmp_layout;
-
- if (!layout)
- return -1;
-
- gf_proc_dump_add_section("xlator.cluster.dht.%s.inode", this->name);
- dht_layout_dump(layout, "layout");
-
-out:
- return ret;
-}
-
-int
-notify (xlator_t *this, int event, void *data, ...)
-{
- int ret = -1;
-
- GF_VALIDATE_OR_GOTO ("dht", this, out);
-
- ret = dht_notify (this, event, data);
-
-out:
- return ret;
-}
-
-void
-fini (xlator_t *this)
-{
- int i = 0;
- dht_conf_t *conf = NULL;
-
- GF_VALIDATE_OR_GOTO ("dht", this, out);
-
- conf = this->private;
- this->private = NULL;
- if (conf) {
- if (conf->file_layouts) {
- for (i = 0; i < conf->subvolume_cnt; i++) {
- GF_FREE (conf->file_layouts[i]);
- }
- GF_FREE (conf->file_layouts);
- }
-
- if (conf->subvolumes)
- GF_FREE (conf->subvolumes);
-
- if (conf->subvolume_status)
- GF_FREE (conf->subvolume_status);
-
- GF_FREE (conf);
- }
-out:
- return;
-}
-
-int32_t
-mem_acct_init (xlator_t *this)
-{
- int ret = -1;
-
- GF_VALIDATE_OR_GOTO ("dht", this, out);
-
- ret = xlator_mem_acct_init (this, gf_dht_mt_end + 1);
-
- if (ret != 0) {
- gf_log (this->name, GF_LOG_ERROR, "Memory accounting init"
- "failed");
- return ret;
- }
-out:
- return ret;
-}
-
-
-int
-dht_parse_decommissioned_bricks (xlator_t *this, dht_conf_t *conf,
- const char *bricks)
-{
- int i = 0;
- int ret = -1;
- char *tmpstr = NULL;
- char *dup_brick = NULL;
- char *node = NULL;
-
- if (!conf || !bricks)
- goto out;
-
- dup_brick = gf_strdup (bricks);
- node = strtok_r (dup_brick, ",", &tmpstr);
- while (node) {
- for (i = 0; i < conf->subvolume_cnt; i++) {
- if (!strcmp (conf->subvolumes[i]->name, node)) {
- conf->decommissioned_bricks[i] =
- conf->subvolumes[i];
- gf_log (this->name, GF_LOG_INFO,
- "decommissioning subvolume %s",
- conf->subvolumes[i]->name);
- break;
- }
- }
- if (i == conf->subvolume_cnt) {
- /* Wrong node given. */
- goto out;
- }
- node = strtok_r (NULL, ",", &tmpstr);
- }
-
- ret = 0;
-out:
- if (dup_brick)
- GF_FREE (dup_brick);
-
- return ret;
-}
-
-int
-reconfigure (xlator_t *this, dict_t *options)
-{
- dht_conf_t *conf = NULL;
- char *temp_str = NULL;
- gf_boolean_t search_unhashed;
- int ret = -1;
-
- GF_VALIDATE_OR_GOTO ("dht", this, out);
- GF_VALIDATE_OR_GOTO ("dht", options, out);
-
- conf = this->private;
- if (!conf)
- return 0;
-
- if (dict_get_str (options, "lookup-unhashed", &temp_str) == 0) {
- /* If option is not "auto", other options _should_ be boolean*/
- if (strcasecmp (temp_str, "auto")) {
- if (!gf_string2boolean (temp_str, &search_unhashed)) {
- gf_log(this->name, GF_LOG_DEBUG, "Reconfigure:"
- " lookup-unhashed reconfigured (%s)",
- temp_str);
- conf->search_unhashed = search_unhashed;
- } else {
- gf_log(this->name, GF_LOG_ERROR, "Reconfigure:"
- " lookup-unhashed should be boolean,"
- " not (%s), defaulting to (%d)",
- temp_str, conf->search_unhashed);
- //return -1;
- ret = -1;
- goto out;
- }
- } else {
- gf_log(this->name, GF_LOG_DEBUG, "Reconfigure:"
- " lookup-unhashed reconfigured auto ");
- conf->search_unhashed = GF_DHT_LOOKUP_UNHASHED_AUTO;
- }
- }
-
- GF_OPTION_RECONF ("min-free-disk", conf->min_free_disk, options,
- percent_or_size, out);
- GF_OPTION_RECONF ("min-free-inodes", conf->min_free_inodes, options,
- percent, out);
- GF_OPTION_RECONF ("directory-layout-spread", conf->dir_spread_cnt,
- options, uint32, out);
-
- if (dict_get_str (options, "decommissioned-bricks", &temp_str) == 0) {
- ret = dht_parse_decommissioned_bricks (this, conf, temp_str);
- if (ret == -1)
- goto out;
- }
-
- ret = 0;
-out:
- return ret;
-}
-
-
-int
-init (xlator_t *this)
-{
- dht_conf_t *conf = NULL;
- char *temp_str = NULL;
- int ret = -1;
- int i = 0;
-
- GF_VALIDATE_OR_GOTO ("dht", this, err);
-
- if (!this->children) {
- gf_log (this->name, GF_LOG_CRITICAL,
- "Distribute needs more than one subvolume");
- return -1;
- }
-
- if (!this->parents) {
- gf_log (this->name, GF_LOG_WARNING,
- "dangling volume. check volfile");
- }
-
- conf = GF_CALLOC (1, sizeof (*conf), gf_dht_mt_dht_conf_t);
- if (!conf) {
- goto err;
- }
-
- conf->search_unhashed = GF_DHT_LOOKUP_UNHASHED_ON;
- if (dict_get_str (this->options, "lookup-unhashed", &temp_str) == 0) {
- /* If option is not "auto", other options _should_ be boolean */
- if (strcasecmp (temp_str, "auto"))
- gf_string2boolean (temp_str, &conf->search_unhashed);
- else
- conf->search_unhashed = GF_DHT_LOOKUP_UNHASHED_AUTO;
- }
-
- GF_OPTION_INIT ("unhashed-sticky-bit", conf->unhashed_sticky_bit, bool,
- err);
-
- GF_OPTION_INIT ("use-readdirp", conf->use_readdirp, bool, err);
-
- GF_OPTION_INIT ("min-free-disk", conf->min_free_disk, percent_or_size,
- err);
-
- GF_OPTION_INIT ("min-free-inodes", conf->min_free_inodes, percent,
- err);
-
- conf->dir_spread_cnt = conf->subvolume_cnt;
- GF_OPTION_INIT ("directory-layout-spread", conf->dir_spread_cnt,
- uint32, err);
-
- GF_OPTION_INIT ("assert-no-child-down", conf->assert_no_child_down,
- bool, err);
-
- ret = dht_init_subvolumes (this, conf);
- if (ret == -1) {
- goto err;
- }
-
- if (dict_get_str (this->options, "decommissioned-bricks", &temp_str) == 0) {
- ret = dht_parse_decommissioned_bricks (this, conf, temp_str);
- if (ret == -1)
- goto err;
- }
-
- ret = dht_layouts_init (this, conf);
- if (ret == -1) {
- goto err;
- }
-
- LOCK_INIT (&conf->subvolume_lock);
- LOCK_INIT (&conf->layout_lock);
-
- conf->gen = 1;
-
- /* Create 'syncop' environment */
- conf->env = syncenv_new (0);
- if (!conf->env) {
- gf_log (this->name, GF_LOG_ERROR,
- "failed to create sync environment %s",
- strerror (errno));
- goto err;
- }
-
- this->private = conf;
-
- return 0;
-
-err:
- if (conf) {
- if (conf->file_layouts) {
- for (i = 0; i < conf->subvolume_cnt; i++) {
- GF_FREE (conf->file_layouts[i]);
- }
- GF_FREE (conf->file_layouts);
- }
-
- if (conf->subvolumes)
- GF_FREE (conf->subvolumes);
-
- if (conf->subvolume_status)
- GF_FREE (conf->subvolume_status);
-
- if (conf->du_stats)
- GF_FREE (conf->du_stats);
-
- GF_FREE (conf);
- }
-
- return -1;
-}
-
+class_methods_t class_methods = {
+ .init = dht_init,
+ .fini = dht_fini,
+ .reconfigure = dht_reconfigure,
+ .notify = dht_notify
+};
struct xlator_fops fops = {
.lookup = dht_lookup,
@@ -478,6 +50,7 @@ struct xlator_fops fops = {
.access = dht_access,
.readlink = dht_readlink,
.getxattr = dht_getxattr,
+ .fgetxattr = dht_fgetxattr,
.readv = dht_readv,
.flush = dht_flush,
.fsync = dht_fsync,
@@ -486,6 +59,7 @@ struct xlator_fops fops = {
.lk = dht_lk,
/* Inode write operations */
+ .fremovexattr = dht_fremovexattr,
.removexattr = dht_removexattr,
.setxattr = dht_setxattr,
.fsetxattr = dht_fsetxattr,
@@ -496,6 +70,9 @@ struct xlator_fops fops = {
.fxattrop = dht_fxattrop,
.setattr = dht_setattr,
.fsetattr = dht_fsetattr,
+ .fallocate = dht_fallocate,
+ .discard = dht_discard,
+ .zerofill = dht_zerofill,
};
struct xlator_dumpops dumpops = {
@@ -509,44 +86,4 @@ struct xlator_cbks cbks = {
// .releasedir = dht_releasedir,
.forget = dht_forget
};
-
-
-struct volume_options options[] = {
- { .key = {"lookup-unhashed"},
- .value = {"auto", "yes", "no", "enable", "disable", "1", "0",
- "on", "off"},
- .type = GF_OPTION_TYPE_STR,
- .default_value = "on",
- },
- { .key = {"min-free-disk"},
- .type = GF_OPTION_TYPE_PERCENT_OR_SIZET,
- .default_value = "10%",
- .description = "Percentage/Size of disk space that must be "
- "kept free."
- },
- { .key = {"min-free-inodes"},
- .type = GF_OPTION_TYPE_PERCENT,
- .default_value = "5%",
- .description = "Percentage inodes that must be "
- "kept free."
- },
- { .key = {"unhashed-sticky-bit"},
- .type = GF_OPTION_TYPE_BOOL,
- .default_value = "off",
- },
- { .key = {"use-readdirp"},
- .type = GF_OPTION_TYPE_BOOL,
- .default_value = "on",
- },
- { .key = {"assert-no-child-down"},
- .type = GF_OPTION_TYPE_BOOL,
- .default_value = "off",
- },
- { .key = {"directory-layout-spread"},
- .type = GF_OPTION_TYPE_INT,
- },
- { .key = {"decommissioned-bricks"},
- .type = GF_OPTION_TYPE_ANY,
- },
- { .key = {NULL} },
-};
+;
diff --git a/xlators/cluster/dht/src/nufa.c b/xlators/cluster/dht/src/nufa.c
index 2f196951a..e934acdf0 100644
--- a/xlators/cluster/dht/src/nufa.c
+++ b/xlators/cluster/dht/src/nufa.c
@@ -1,20 +1,11 @@
/*
- Copyright (c) 2008-2011 Gluster, Inc. <http://www.gluster.com>
+ Copyright (c) 2008-2012 Red Hat, Inc. <http://www.redhat.com>
This file is part of GlusterFS.
- GlusterFS is free software; you can redistribute it and/or modify
- it under the terms of the GNU General Public License as published
- by the Free Software Foundation; either version 3 of the License,
- or (at your option) any later version.
-
- GlusterFS is distributed in the hope that it will be useful, but
- WITHOUT ANY WARRANTY; without even the implied warranty of
- MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU
- General Public License for more details.
-
- You should have received a copy of the GNU General Public License
- along with this program. If not, see
- <http://www.gnu.org/licenses/>.
+ This file is licensed to you under your choice of the GNU Lesser
+ General Public License, version 3 or any later version (LGPLv3 or
+ later), or the GNU General Public License, version 2 (GPLv2), in all
+ cases as published by the Free Software Foundation.
*/
@@ -27,6 +18,8 @@
/* TODO: all 'TODO's in dht.c holds good */
+extern struct volume_options options[];
+
int
nufa_local_lookup_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
int op_ret, int op_errno,
@@ -44,7 +37,6 @@ nufa_local_lookup_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
int call_cnt = 0;
int ret = 0;
-
conf = this->private;
prev = cookie;
@@ -62,7 +54,8 @@ nufa_local_lookup_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
if (op_ret == -1)
goto out;
- is_linkfile = check_is_linkfile (inode, stbuf, xattr);
+ is_linkfile = check_is_linkfile (inode, stbuf, xattr,
+ conf->link_xattr_name);
is_dir = check_is_dir (inode, stbuf, xattr);
if (!is_dir && !is_linkfile) {
@@ -141,7 +134,7 @@ out:
err:
DHT_STACK_UNWIND (lookup, frame, op_ret, op_errno,
- inode, stbuf, xattr, NULL);
+ inode, stbuf, xattr, postparent);
return 0;
}
@@ -211,7 +204,7 @@ nufa_lookup (call_frame_t *frame, xlator_t *this,
* revalidates directly go to the cached-subvolume.
*/
ret = dict_set_uint32 (local->xattr_req,
- "trusted.glusterfs.dht", 4 * 4);
+ conf->xattr_name, 4 * 4);
if (ret < 0) {
gf_log (this->name, GF_LOG_ERROR,
"Failed to set dict value.");
@@ -232,7 +225,7 @@ nufa_lookup (call_frame_t *frame, xlator_t *this,
} else {
do_fresh_lookup:
ret = dict_set_uint32 (local->xattr_req,
- "trusted.glusterfs.dht", 4 * 4);
+ conf->xattr_name, 4 * 4);
if (ret < 0) {
gf_log (this->name, GF_LOG_ERROR,
"Failed to set dict value.");
@@ -241,7 +234,7 @@ nufa_lookup (call_frame_t *frame, xlator_t *this,
}
ret = dict_set_uint32 (local->xattr_req,
- "trusted.glusterfs.dht.linkto", 256);
+ conf->link_xattr_name, 256);
if (ret < 0) {
gf_log (this->name, GF_LOG_ERROR,
"Failed to set dict value.");
@@ -260,7 +253,8 @@ nufa_lookup (call_frame_t *frame, xlator_t *this,
err:
op_errno = (op_errno == -1) ? errno : op_errno;
- DHT_STACK_UNWIND (lookup, frame, -1, op_errno, NULL, NULL, NULL, NULL);
+ DHT_STACK_UNWIND (lookup, frame, -1, op_errno, NULL, NULL, NULL,
+ NULL);
return 0;
}
@@ -269,7 +263,7 @@ nufa_create_linkfile_create_cbk (call_frame_t *frame, void *cookie,
xlator_t *this, int op_ret, int op_errno,
inode_t *inode, struct iatt *stbuf,
struct iatt *preparent,
- struct iatt *postparent)
+ struct iatt *postparent, dict_t *xdata)
{
dht_local_t *local = NULL;
@@ -280,21 +274,21 @@ nufa_create_linkfile_create_cbk (call_frame_t *frame, void *cookie,
STACK_WIND (frame, dht_create_cbk,
local->cached_subvol, local->cached_subvol->fops->create,
- &local->loc, local->flags, local->mode, local->fd,
- local->params);
+ &local->loc, local->flags, local->mode, local->umask,
+ local->fd, local->params);
return 0;
err:
DHT_STACK_UNWIND (create, frame, -1, op_errno,
- NULL, NULL, NULL, NULL, NULL);
+ NULL, NULL, NULL, NULL, NULL, NULL);
return 0;
}
int
nufa_create (call_frame_t *frame, xlator_t *this,
loc_t *loc, int32_t flags, mode_t mode,
- fd_t *fd, dict_t *params)
+ mode_t umask, fd_t *fd, dict_t *params)
{
dht_local_t *local = NULL;
dht_conf_t *conf = NULL;
@@ -329,7 +323,8 @@ nufa_create (call_frame_t *frame, xlator_t *this,
if (dht_is_subvol_filled (this, (xlator_t *)conf->private)) {
avail_subvol =
dht_free_disk_available_subvol (this,
- (xlator_t *)conf->private);
+ (xlator_t *)conf->private,
+ local);
}
if (subvol != avail_subvol) {
@@ -337,11 +332,10 @@ nufa_create (call_frame_t *frame, xlator_t *this,
local->params = dict_ref (params);
local->mode = mode;
local->flags = flags;
-
+ local->umask = umask;
local->cached_subvol = avail_subvol;
- dht_linkfile_create (frame,
- nufa_create_linkfile_create_cbk,
- avail_subvol, subvol, loc);
+ dht_linkfile_create (frame, nufa_create_linkfile_create_cbk,
+ this, avail_subvol, subvol, loc);
return 0;
}
@@ -350,14 +344,14 @@ nufa_create (call_frame_t *frame, xlator_t *this,
STACK_WIND (frame, dht_create_cbk,
subvol, subvol->fops->create,
- loc, flags, mode, fd, params);
+ loc, flags, mode, umask, fd, params);
return 0;
err:
op_errno = (op_errno == -1) ? errno : op_errno;
DHT_STACK_UNWIND (create, frame, -1, op_errno,
- NULL, NULL, NULL, NULL, NULL);
+ NULL, NULL, NULL, NULL, NULL, NULL);
return 0;
}
@@ -366,34 +360,39 @@ int
nufa_mknod_linkfile_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
int op_ret, int op_errno, inode_t *inode,
struct iatt *stbuf, struct iatt *preparent,
- struct iatt *postparent)
+ struct iatt *postparent, dict_t *xdata)
{
dht_local_t *local = NULL;
local = frame->local;
+ if (!local || !local->cached_subvol) {
+ op_errno = EINVAL;
+ op_ret = -1;
+ goto err;
+ }
if (op_ret >= 0) {
- STACK_WIND (frame, dht_newfile_cbk,
- local->cached_subvol,
+ STACK_WIND_COOKIE (frame, dht_newfile_cbk,
+ (void *)local->cached_subvol, local->cached_subvol,
local->cached_subvol->fops->mknod,
&local->loc, local->mode, local->rdev,
- local->params);
+ local->umask, local->params);
return 0;
}
-
+err:
WIPE (postparent);
WIPE (preparent);
DHT_STACK_UNWIND (link, frame, op_ret, op_errno,
- inode, stbuf, preparent, postparent);
+ inode, stbuf, preparent, postparent, xdata);
return 0;
}
int
nufa_mknod (call_frame_t *frame, xlator_t *this,
- loc_t *loc, mode_t mode, dev_t rdev, dict_t *params)
+ loc_t *loc, mode_t mode, dev_t rdev, mode_t umask, dict_t *params)
{
dht_local_t *local = NULL;
dht_conf_t *conf = NULL;
@@ -429,7 +428,8 @@ nufa_mknod (call_frame_t *frame, xlator_t *this,
if (dht_is_subvol_filled (this, (xlator_t *)conf->private)) {
avail_subvol =
dht_free_disk_available_subvol (this,
- (xlator_t *)conf->private);
+ (xlator_t *)conf->private,
+ local);
}
if (avail_subvol != subvol) {
@@ -437,10 +437,11 @@ nufa_mknod (call_frame_t *frame, xlator_t *this,
local->params = dict_ref (params);
local->mode = mode;
+ local->umask = umask;
local->rdev = rdev;
local->cached_subvol = avail_subvol;
- dht_linkfile_create (frame, nufa_mknod_linkfile_cbk,
+ dht_linkfile_create (frame, nufa_mknod_linkfile_cbk, this,
avail_subvol, subvol, loc);
return 0;
}
@@ -448,211 +449,185 @@ nufa_mknod (call_frame_t *frame, xlator_t *this,
gf_log (this->name, GF_LOG_TRACE,
"creating %s on %s", loc->path, subvol->name);
- STACK_WIND (frame, dht_newfile_cbk,
- subvol, subvol->fops->mknod,
- loc, mode, rdev, params);
+ STACK_WIND_COOKIE (frame, dht_newfile_cbk, (void *)subvol, subvol,
+ subvol->fops->mknod, loc, mode, rdev, umask,
+ params);
return 0;
err:
op_errno = (op_errno == -1) ? errno : op_errno;
DHT_STACK_UNWIND (mknod, frame, -1, op_errno,
- NULL, NULL, NULL, NULL);
+ NULL, NULL, NULL, NULL, NULL);
return 0;
}
-int
-notify (xlator_t *this, int event, void *data, ...)
+gf_boolean_t
+same_first_part (char *str1, char term1, char *str2, char term2)
{
- int ret = -1;
-
- ret = dht_notify (this, event, data);
-
- return ret;
+ gf_boolean_t ended1;
+ gf_boolean_t ended2;
+
+ for (;;) {
+ ended1 = ((*str1 == '\0') || (*str1 == term1));
+ ended2 = ((*str2 == '\0') || (*str2 == term2));
+ if (ended1 && ended2) {
+ return _gf_true;
+ }
+ if (ended1 || ended2 || (*str1 != *str2)) {
+ return _gf_false;
+ }
+ ++str1;
+ ++str2;
+ }
}
-void
-fini (xlator_t *this)
-{
- int i = 0;
- dht_conf_t *conf = NULL;
+typedef struct nufa_args {
+ xlator_t *this;
+ char *volname;
+ gf_boolean_t addr_match;
+} nufa_args_t;
- conf = this->private;
+static void
+nufa_find_local_brick (xlator_t *xl, void *data)
+{
+ nufa_args_t *args = data;
+ xlator_t *this = args->this;
+ char *local_volname = args->volname;
+ gf_boolean_t addr_match = args->addr_match;
+ char *brick_host = NULL;
+ dht_conf_t *conf = this->private;
+ int ret = -1;
+
+ /*This means a local subvol was already found. We pick the first brick
+ * that is local*/
+ if (conf->private)
+ return;
+
+ if (strcmp (xl->name, local_volname) == 0) {
+ conf->private = xl;
+ gf_log (this->name, GF_LOG_INFO, "Using specified subvol %s",
+ local_volname);
+ return;
+ }
- if (conf) {
- if (conf->file_layouts) {
- for (i = 0; i < conf->subvolume_cnt; i++) {
- GF_FREE (conf->file_layouts[i]);
- }
- GF_FREE (conf->file_layouts);
- }
+ if (!addr_match)
+ return;
- if (conf->subvolumes)
- GF_FREE (conf->subvolumes);
+ ret = dict_get_str (xl->options, "remote-host", &brick_host);
+ if ((ret == 0) &&
+ (gf_is_same_address (local_volname, brick_host) ||
+ gf_is_local_addr (brick_host))) {
+ conf->private = xl;
+ gf_log (this->name, GF_LOG_INFO, "Using the first local "
+ "subvol %s", xl->name);
+ return;
+ }
- if (conf->subvolume_status)
- GF_FREE (conf->subvolume_status);
+}
- GF_FREE (conf);
- }
+static void
+nufa_to_dht (xlator_t *this)
+{
+ GF_ASSERT (this);
+ GF_ASSERT (this->fops);
- return;
+ this->fops->lookup = dht_lookup;
+ this->fops->create = dht_create;
+ this->fops->mknod = dht_mknod;
}
int
-init (xlator_t *this)
+nufa_find_local_subvol (xlator_t *this,
+ void (*fn) (xlator_t *each, void* data), void *data)
{
- dht_conf_t *conf = NULL;
- xlator_list_t *trav = NULL;
- data_t *data = NULL;
- char *local_volname = NULL;
- char *temp_str = NULL;
- int ret = -1;
- int i = 0;
- char my_hostname[256];
- uint32_t temp_free_disk = 0;
-
- if (!this->children) {
- gf_log (this->name, GF_LOG_CRITICAL,
- "NUFA needs more than one subvolume");
+ int ret = -1;
+ dht_conf_t *conf = this->private;
+ xlator_list_t *trav = NULL;
+ xlator_t *parent = NULL;
+ xlator_t *candidate = NULL;
+
+ xlator_foreach_depth_first (this, fn, data);
+ if (!conf->private) {
+ gf_log (this->name, GF_LOG_ERROR, "Couldn't find a local "
+ "brick");
return -1;
}
- if (!this->parents) {
- gf_log (this->name, GF_LOG_WARNING,
- "dangling volume. check volfile");
- }
-
- conf = GF_CALLOC (1, sizeof (*conf),
- gf_dht_mt_dht_conf_t);
- if (!conf) {
- goto err;
- }
-
- conf->search_unhashed = GF_DHT_LOOKUP_UNHASHED_ON;
- if (dict_get_str (this->options, "lookup-unhashed", &temp_str) == 0) {
- /* If option is not "auto", other options _should_ be boolean */
- if (strcasecmp (temp_str, "auto"))
- gf_string2boolean (temp_str, &conf->search_unhashed);
- else
- conf->search_unhashed = GF_DHT_LOOKUP_UNHASHED_AUTO;
- }
+ candidate = conf->private;
+ trav = candidate->parents;
+ while (trav) {
- ret = dht_init_subvolumes (this, conf);
- if (ret == -1) {
- goto err;
- }
+ parent = trav->xlator;
+ if (strcmp (parent->type, "cluster/nufa") == 0) {
+ gf_log (this->name, GF_LOG_INFO, "Found local subvol, "
+ "%s", candidate->name);
+ ret = 0;
+ conf->private = candidate;
+ break;
+ }
- ret = dht_layouts_init (this, conf);
- if (ret == -1) {
- goto err;
+ candidate = parent;
+ trav = parent->parents;
}
- LOCK_INIT (&conf->subvolume_lock);
- LOCK_INIT (&conf->layout_lock);
+ return ret;
+}
- conf->gen = 1;
+int
+nufa_init (xlator_t *this)
+{
+ data_t *data = NULL;
+ char *local_volname = NULL;
+ int ret = -1;
+ char my_hostname[256];
+ gf_boolean_t addr_match = _gf_false;
+ nufa_args_t args = {0, };
- local_volname = "localhost";
- ret = gethostname (my_hostname, 256);
- if (ret < 0) {
- gf_log (this->name, GF_LOG_WARNING,
- "could not find hostname (%s)",
- strerror (errno));
+ ret = dht_init(this);
+ if (ret) {
+ return ret;
}
- if (ret == 0)
- local_volname = my_hostname;
-
- data = dict_get (this->options, "local-volume-name");
- if (data) {
+ if ((data = dict_get (this->options, "local-volume-name"))) {
local_volname = data->data;
- }
- trav = this->children;
- while (trav) {
- if (strcmp (trav->xlator->name, local_volname) == 0)
- break;
- trav = trav->next;
- }
+ } else {
+ addr_match = _gf_true;
+ local_volname = "localhost";
+ ret = gethostname (my_hostname, 256);
+ if (ret == 0)
+ local_volname = my_hostname;
- if (!trav) {
- gf_log (this->name, GF_LOG_ERROR,
- "Could not find subvolume named '%s'. "
- "Please define volume with the name as the hostname "
- "or override it with 'option local-volume-name'",
- local_volname);
- goto err;
- }
- /* The volume specified exists */
- conf->private = trav->xlator;
-
- conf->min_free_disk = 10;
- conf->disk_unit = 'p';
-
- if (dict_get_str (this->options, "min-free-disk",
- &temp_str) == 0) {
- if (gf_string2percent (temp_str,
- &temp_free_disk) == 0) {
- if (temp_free_disk > 100) {
- gf_string2bytesize (temp_str,
- &conf->min_free_disk);
- conf->disk_unit = 'b';
- } else {
- conf->min_free_disk = (uint64_t)temp_free_disk;
- conf->disk_unit = 'p';
- }
- } else {
- gf_string2bytesize (temp_str,
- &conf->min_free_disk);
- conf->disk_unit = 'b';
- }
- }
+ else
+ gf_log (this->name, GF_LOG_WARNING,
+ "could not find hostname (%s)",
+ strerror (errno));
- conf->du_stats = GF_CALLOC (conf->subvolume_cnt, sizeof (dht_du_t),
- gf_dht_mt_dht_du_t);
- if (!conf->du_stats) {
- goto err;
}
- /* Create 'syncop' environment */
- conf->env = syncenv_new (0);
- if (!conf->env) {
- gf_log (this->name, GF_LOG_ERROR,
- "failed to create sync environment %s",
- strerror (errno));
- goto err;
+ args.this = this;
+ args.volname = local_volname;
+ args.addr_match = addr_match;
+ ret = nufa_find_local_subvol (this, nufa_find_local_brick, &args);
+ if (ret) {
+ gf_log (this->name, GF_LOG_INFO,
+ "Unable to find local subvolume, switching "
+ "to dht mode");
+ nufa_to_dht (this);
}
-
- this->private = conf;
-
return 0;
+}
-err:
- if (conf) {
- if (conf->file_layouts) {
- for (i = 0; i < conf->subvolume_cnt; i++) {
- GF_FREE (conf->file_layouts[i]);
- }
- GF_FREE (conf->file_layouts);
- }
-
- if (conf->subvolumes)
- GF_FREE (conf->subvolumes);
-
- if (conf->subvolume_status)
- GF_FREE (conf->subvolume_status);
-
- if (conf->du_stats)
- GF_FREE (conf->du_stats);
-
- GF_FREE (conf);
- }
- return -1;
-}
+class_methods_t class_methods = {
+ .init = nufa_init,
+ .fini = dht_fini,
+ .reconfigure = dht_reconfigure,
+ .notify = dht_notify
+};
struct xlator_fops fops = {
@@ -699,19 +674,3 @@ struct xlator_fops fops = {
struct xlator_cbks cbks = {
.forget = dht_forget
};
-
-
-struct volume_options options[] = {
- { .key = {"lookup-unhashed"},
- .value = {"auto", "yes", "no", "enable", "disable", "1", "0",
- "on", "off"},
- .type = GF_OPTION_TYPE_STR
- },
- { .key = {"local-volume-name"},
- .type = GF_OPTION_TYPE_XLATOR
- },
- { .key = {"min-free-disk"},
- .type = GF_OPTION_TYPE_PERCENT_OR_SIZET,
- },
- { .key = {NULL} },
-};
diff --git a/xlators/cluster/dht/src/switch.c b/xlators/cluster/dht/src/switch.c
index fd3f22ea0..d3ea90ba8 100644
--- a/xlators/cluster/dht/src/switch.c
+++ b/xlators/cluster/dht/src/switch.c
@@ -1,20 +1,11 @@
/*
- Copyright (c) 2010-2011 Gluster, Inc. <http://www.gluster.com>
+ Copyright (c) 2008-2012 Red Hat, Inc. <http://www.redhat.com>
This file is part of GlusterFS.
- GlusterFS is free software; you can redistribute it and/or modify
- it under the terms of the GNU General Public License as published
- by the Free Software Foundation; either version 3 of the License,
- or (at your option) any later version.
-
- GlusterFS is distributed in the hope that it will be useful, but
- WITHOUT ANY WARRANTY; without even the implied warranty of
- MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU
- General Public License for more details.
-
- You should have received a copy of the GNU General Public License
- along with this program. If not, see
- <http://www.gnu.org/licenses/>.
+ This file is licensed to you under your choice of the GNU Lesser
+ General Public License, version 3 or any later version (LGPLv3 or
+ later), or the GNU General Public License, version 2 (GPLv2), in all
+ cases as published by the Free Software Foundation.
*/
@@ -31,6 +22,8 @@
#include <fnmatch.h>
#include <string.h>
+extern struct volume_options options[];
+
struct switch_sched_array {
xlator_t *xl;
int32_t eligible;
@@ -76,29 +69,37 @@ get_switch_matching_subvol (const char *path, dht_conf_t *conf,
struct switch_struct *cond = NULL;
struct switch_struct *trav = NULL;
char *pathname = NULL;
- int idx = 0;
+ int idx = 0;
+ xlator_t *subvol = NULL;
cond = conf->private;
+ subvol = hashed_subvol;
if (!cond)
- return hashed_subvol;
+ goto out;
- trav = cond;
pathname = gf_strdup (path);
+ if (!pathname)
+ goto out;
+
+ trav = cond;
while (trav) {
if (fnmatch (trav->path_pattern,
pathname, FNM_NOESCAPE) == 0) {
for (idx = 0; idx < trav->num_child; idx++) {
if (trav->array[idx].xl == hashed_subvol)
- return hashed_subvol;
+ goto out;
}
idx = trav->node_index++;
trav->node_index %= trav->num_child;
- return trav->array[idx].xl;
+ subvol = trav->array[idx].xl;
+ goto out;
}
trav = trav->next;
}
+out:
GF_FREE (pathname);
- return hashed_subvol;
+
+ return subvol;
}
@@ -136,7 +137,8 @@ switch_local_lookup_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
if (op_ret == -1)
goto out;
- is_linkfile = check_is_linkfile (inode, stbuf, xattr);
+ is_linkfile = check_is_linkfile (inode, stbuf, xattr,
+ conf->link_xattr_name);
is_dir = check_is_dir (inode, stbuf, xattr);
if (!is_dir && !is_linkfile) {
@@ -290,11 +292,11 @@ switch_lookup (call_frame_t *frame, xlator_t *this,
* attribute, revalidates directly go to the cached-subvolume.
*/
ret = dict_set_uint32 (local->xattr_req,
- "trusted.glusterfs.dht", 4 * 4);
+ conf->xattr_name, 4 * 4);
if (ret < 0)
gf_log (this->name, GF_LOG_WARNING,
- "failed to set dict value for "
- "trusted.glusterfs.dht");
+ "failed to set dict value for %s",
+ conf->xattr_name);
for (i = 0; i < layout->cnt; i++) {
subvol = layout->list[i].xlator;
@@ -309,18 +311,18 @@ switch_lookup (call_frame_t *frame, xlator_t *this,
} else {
do_fresh_lookup:
ret = dict_set_uint32 (local->xattr_req,
- "trusted.glusterfs.dht", 4 * 4);
+ conf->xattr_name, 4 * 4);
if (ret < 0)
gf_log (this->name, GF_LOG_WARNING,
- "failed to set dict value for "
- "trusted.glusterfs.dht");
+ "failed to set dict value for %s",
+ conf->xattr_name);
ret = dict_set_uint32 (local->xattr_req,
- "trusted.glusterfs.dht.linkto", 256);
+ conf->link_xattr_name, 256);
if (ret < 0)
gf_log (this->name, GF_LOG_WARNING,
- "failed to set dict value for "
- "trusted.glusterfs.dht.linkto");
+ "failed to set dict value for %s",
+ conf->link_xattr_name);
if (!hashed_subvol) {
gf_log (this->name, GF_LOG_DEBUG,
@@ -366,7 +368,8 @@ switch_lookup (call_frame_t *frame, xlator_t *this,
err:
op_errno = (op_errno == -1) ? errno : op_errno;
- DHT_STACK_UNWIND (lookup, frame, -1, op_errno, NULL, NULL, NULL, NULL);
+ DHT_STACK_UNWIND (lookup, frame, -1, op_errno,
+ NULL, NULL, NULL, NULL);
return 0;
}
@@ -375,7 +378,7 @@ switch_create_linkfile_create_cbk (call_frame_t *frame, void *cookie,
xlator_t *this, int op_ret, int op_errno,
inode_t *inode, struct iatt *stbuf,
struct iatt *preparent,
- struct iatt *postparent)
+ struct iatt *postparent, dict_t *xdata)
{
dht_local_t *local = NULL;
@@ -386,21 +389,21 @@ switch_create_linkfile_create_cbk (call_frame_t *frame, void *cookie,
STACK_WIND (frame, dht_create_cbk,
local->cached_subvol, local->cached_subvol->fops->create,
- &local->loc, local->flags, local->mode, local->fd,
- local->params);
+ &local->loc, local->flags, local->mode, local->umask,
+ local->fd, local->params);
return 0;
err:
DHT_STACK_UNWIND (create, frame, -1, op_errno,
- NULL, NULL, NULL, NULL, NULL);
+ NULL, NULL, NULL, NULL, NULL, NULL);
return 0;
}
int
switch_create (call_frame_t *frame, xlator_t *this,
loc_t *loc, int32_t flags, mode_t mode,
- fd_t *fd, dict_t *params)
+ mode_t umask, fd_t *fd, dict_t *params)
{
dht_local_t *local = NULL;
dht_conf_t *conf = NULL;
@@ -434,18 +437,18 @@ switch_create (call_frame_t *frame, xlator_t *this,
avail_subvol = get_switch_matching_subvol (loc->path, conf, subvol);
if (dht_is_subvol_filled (this, avail_subvol)) {
avail_subvol =
- dht_free_disk_available_subvol (this, avail_subvol);
+ dht_free_disk_available_subvol (this, avail_subvol,
+ local);
}
if (subvol != avail_subvol) {
/* create a link file instead of actual file */
local->mode = mode;
local->flags = flags;
-
+ local->umask = umask;
local->cached_subvol = avail_subvol;
- dht_linkfile_create (frame,
- switch_create_linkfile_create_cbk,
- avail_subvol, subvol, loc);
+ dht_linkfile_create (frame, switch_create_linkfile_create_cbk,
+ this, avail_subvol, subvol, loc);
return 0;
}
@@ -454,14 +457,14 @@ switch_create (call_frame_t *frame, xlator_t *this,
STACK_WIND (frame, dht_create_cbk,
subvol, subvol->fops->create,
- loc, flags, mode, fd, params);
+ loc, flags, mode, umask, fd, params);
return 0;
err:
op_errno = (op_errno == -1) ? errno : op_errno;
DHT_STACK_UNWIND (create, frame, -1, op_errno,
- NULL, NULL, NULL, NULL, NULL);
+ NULL, NULL, NULL, NULL, NULL, NULL);
return 0;
}
@@ -470,31 +473,36 @@ int
switch_mknod_linkfile_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
int op_ret, int op_errno, inode_t *inode,
struct iatt *stbuf, struct iatt *preparent,
- struct iatt *postparent)
+ struct iatt *postparent, dict_t *xdata)
{
dht_local_t *local = NULL;
local = frame->local;
+ if (!local || !local->cached_subvol) {
+ op_errno = EINVAL;
+ op_ret = -1;
+ goto err;
+ }
if (op_ret >= 0) {
- STACK_WIND (frame, dht_newfile_cbk,
- local->cached_subvol,
+ STACK_WIND_COOKIE (frame, dht_newfile_cbk,
+ (void *)local->cached_subvol, local->cached_subvol,
local->cached_subvol->fops->mknod,
&local->loc, local->mode, local->rdev,
- local->params);
+ local->umask, local->params);
return 0;
}
-
+err:
DHT_STACK_UNWIND (link, frame, op_ret, op_errno,
- inode, stbuf, preparent, postparent);
+ inode, stbuf, preparent, postparent, xdata);
return 0;
}
int
-switch_mknod (call_frame_t *frame, xlator_t *this,
- loc_t *loc, mode_t mode, dev_t rdev, dict_t *params)
+switch_mknod (call_frame_t *frame, xlator_t *this, loc_t *loc, mode_t mode,
+ dev_t rdev, mode_t umask, dict_t *params)
{
dht_local_t *local = NULL;
dht_conf_t *conf = NULL;
@@ -529,7 +537,8 @@ switch_mknod (call_frame_t *frame, xlator_t *this,
avail_subvol = get_switch_matching_subvol (loc->path, conf, subvol);
if (dht_is_subvol_filled (this, avail_subvol)) {
avail_subvol =
- dht_free_disk_available_subvol (this, avail_subvol);
+ dht_free_disk_available_subvol (this, avail_subvol,
+ local);
}
if (avail_subvol != subvol) {
@@ -537,46 +546,36 @@ switch_mknod (call_frame_t *frame, xlator_t *this,
local->params = dict_ref (params);
local->mode = mode;
+ local->umask = umask;
local->rdev = rdev;
local->cached_subvol = avail_subvol;
dht_linkfile_create (frame, switch_mknod_linkfile_cbk,
- avail_subvol, subvol, loc);
+ this, avail_subvol, subvol, loc);
return 0;
}
gf_log (this->name, GF_LOG_TRACE,
"creating %s on %s", loc->path, subvol->name);
- STACK_WIND (frame, dht_newfile_cbk,
- subvol, subvol->fops->mknod,
- loc, mode, rdev, params);
+ STACK_WIND_COOKIE (frame, dht_newfile_cbk, (void *)subvol, subvol,
+ subvol->fops->mknod, loc, mode, rdev, umask,
+ params);
return 0;
err:
op_errno = (op_errno == -1) ? errno : op_errno;
DHT_STACK_UNWIND (mknod, frame, -1, op_errno,
- NULL, NULL, NULL, NULL);
+ NULL, NULL, NULL, NULL, NULL);
return 0;
}
-int
-notify (xlator_t *this, int event, void *data, ...)
-{
- int ret = -1;
-
- ret = dht_notify (this, event, data);
-
- return ret;
-}
-
void
-fini (xlator_t *this)
+switch_fini (xlator_t *this)
{
- int i = 0;
dht_conf_t *conf = NULL;
struct switch_struct *trav = NULL;
struct switch_struct *prev = NULL;
@@ -587,30 +586,14 @@ fini (xlator_t *this)
trav = (struct switch_struct *)conf->private;
conf->private = NULL;
while (trav) {
- if (trav->array)
- GF_FREE (trav->array);
+ GF_FREE (trav->array);
prev = trav;
trav = trav->next;
GF_FREE (prev);
}
-
- if (conf->file_layouts) {
- for (i = 0; i < conf->subvolume_cnt; i++) {
- GF_FREE (conf->file_layouts[i]);
- }
- GF_FREE (conf->file_layouts);
- }
-
- if (conf->subvolumes)
- GF_FREE (conf->subvolumes);
-
- if (conf->subvolume_status)
- GF_FREE (conf->subvolume_status);
-
- GF_FREE (conf);
}
- return;
+ dht_fini(this);
}
int
@@ -670,8 +653,10 @@ set_switch_pattern (xlator_t *this, dht_conf_t *conf,
dup_str = gf_strdup (switch_str);
switch_opt = GF_CALLOC (1, sizeof (struct switch_struct),
gf_switch_mt_switch_struct);
- if (!switch_opt)
+ if (!switch_opt) {
+ GF_FREE (dup_str);
goto err;
+ }
pattern = strtok_r (dup_str, ":", &tmp_str1);
childs = strtok_r (NULL, ":", &tmp_str1);
@@ -681,6 +666,7 @@ set_switch_pattern (xlator_t *this, dht_conf_t *conf,
"for all the unconfigured child nodes,"
" hence neglecting current option");
switch_str = strtok_r (NULL, ";", &tmp_str);
+ GF_FREE (switch_opt);
GF_FREE (dup_str);
continue;
}
@@ -753,6 +739,7 @@ set_switch_pattern (xlator_t *this, dht_conf_t *conf,
/* First entry */
switch_buf = switch_opt;
}
+ switch_opt = NULL;
switch_str = strtok_r (NULL, ";", &tmp_str);
}
@@ -809,19 +796,20 @@ set_switch_pattern (xlator_t *this, dht_conf_t *conf,
/* First entry */
switch_buf = switch_opt;
}
+ switch_opt = NULL;
}
/* */
conf->private = switch_buf;
return 0;
err:
+ GF_FREE (switch_buf_array);
+ GF_FREE (switch_opt);
+
if (switch_buf) {
- if (switch_buf_array)
- GF_FREE (switch_buf_array);
trav = switch_buf;
while (trav) {
- if (trav->array)
- GF_FREE (trav->array);
+ GF_FREE (trav->array);
switch_opt = trav;
trav = trav->next;
GF_FREE (switch_opt);
@@ -831,68 +819,18 @@ err:
}
-int
-init (xlator_t *this)
+int32_t
+switch_init (xlator_t *this)
{
dht_conf_t *conf = NULL;
data_t *data = NULL;
- char *temp_str = NULL;
int ret = -1;
- int i = 0;
- uint32_t temp_free_disk = 0;
-
- if (!this->children) {
- gf_log (this->name, GF_LOG_CRITICAL,
- "SWITCH needs more than one subvolume");
- return -1;
- }
-
- if (!this->parents) {
- gf_log (this->name, GF_LOG_WARNING,
- "dangling volume. check volfile");
- }
-
- conf = GF_CALLOC (1, sizeof (*conf), gf_switch_mt_dht_conf_t);
- if (!conf) {
- goto err;
- }
-
- conf->search_unhashed = GF_DHT_LOOKUP_UNHASHED_ON;
- if (dict_get_str (this->options, "lookup-unhashed", &temp_str) == 0) {
- /* If option is not "auto", other options _should_ be boolean */
- if (strcasecmp (temp_str, "auto"))
- gf_string2boolean (temp_str, &conf->search_unhashed);
- else
- conf->search_unhashed = GF_DHT_LOOKUP_UNHASHED_AUTO;
- }
- conf->unhashed_sticky_bit = 0;
- if (dict_get_str (this->options, "unhashed-sticky-bit",
- &temp_str) == 0) {
- gf_string2boolean (temp_str, &conf->unhashed_sticky_bit);
- }
-
- conf->min_free_disk = 10;
- conf->disk_unit = 'p';
-
- if (dict_get_str (this->options, "min-free-disk",
- &temp_str) == 0) {
- if (gf_string2percent (temp_str,
- &temp_free_disk) == 0) {
- if (temp_free_disk > 100) {
- gf_string2bytesize (temp_str,
- &conf->min_free_disk);
- conf->disk_unit = 'b';
- } else {
- conf->min_free_disk = (uint64_t)temp_free_disk;
- conf->disk_unit = 'p';
- }
- } else {
- gf_string2bytesize (temp_str,
- &conf->min_free_disk);
- conf->disk_unit = 'b';
- }
+ ret = dht_init(this);
+ if (ret) {
+ return ret;
}
+ conf = this->private;
data = dict_get (this->options, "pattern.switch.case");
if (data) {
@@ -903,65 +841,23 @@ init (xlator_t *this)
}
}
- ret = dht_init_subvolumes (this, conf);
- if (ret == -1) {
- goto err;
- }
-
- ret = dht_layouts_init (this, conf);
- if (ret == -1) {
- goto err;
- }
-
- LOCK_INIT (&conf->subvolume_lock);
- LOCK_INIT (&conf->layout_lock);
-
- conf->gen = 1;
-
- conf->du_stats = GF_CALLOC (conf->subvolume_cnt, sizeof (dht_du_t),
- gf_switch_mt_dht_du_t);
- if (!conf->du_stats) {
- goto err;
- }
-
- /* Create 'syncop' environment */
- conf->env = syncenv_new (0);
- if (!conf->env) {
- gf_log (this->name, GF_LOG_ERROR,
- "failed to create sync environment %s",
- strerror (errno));
- goto err;
- }
-
this->private = conf;
-
return 0;
err:
- if (conf) {
- if (conf->file_layouts) {
- for (i = 0; i < conf->subvolume_cnt; i++) {
- GF_FREE (conf->file_layouts[i]);
- }
- GF_FREE (conf->file_layouts);
- }
-
- if (conf->subvolumes)
- GF_FREE (conf->subvolumes);
-
- if (conf->subvolume_status)
- GF_FREE (conf->subvolume_status);
-
- if (conf->du_stats)
- GF_FREE (conf->du_stats);
-
- GF_FREE (conf);
- }
-
+ dht_fini(this);
return -1;
}
+class_methods_t class_methods = {
+ .init = switch_init,
+ .fini = switch_fini,
+ .reconfigure = dht_reconfigure,
+ .notify = dht_notify
+};
+
+
struct xlator_fops fops = {
.lookup = switch_lookup,
.create = switch_create,
@@ -1006,19 +902,3 @@ struct xlator_fops fops = {
struct xlator_cbks cbks = {
.forget = dht_forget
};
-
-
-struct volume_options options[] = {
- { .key = {"lookup-unhashed"},
- .value = {"auto", "yes", "no", "enable", "disable", "1", "0",
- "on", "off"},
- .type = GF_OPTION_TYPE_STR
- },
- { .key = {"pattern.switch.case"},
- .type = GF_OPTION_TYPE_ANY
- },
- { .key = {"min-free-disk"},
- .type = GF_OPTION_TYPE_PERCENT_OR_SIZET,
- },
- { .key = {NULL} },
-};
diff --git a/xlators/cluster/ha/src/Makefile.am b/xlators/cluster/ha/src/Makefile.am
index 5f78a2965..5c1364b7f 100644
--- a/xlators/cluster/ha/src/Makefile.am
+++ b/xlators/cluster/ha/src/Makefile.am
@@ -1,15 +1,16 @@
xlator_LTLIBRARIES = ha.la
xlatordir = $(libdir)/glusterfs/$(PACKAGE_VERSION)/xlator/testing/cluster
-ha_la_LDFLAGS = -module -avoidversion
+ha_la_LDFLAGS = -module -avoid-version
ha_la_SOURCES = ha-helpers.c ha.c
ha_la_LIBADD = $(top_builddir)/libglusterfs/src/libglusterfs.la
noinst_HEADERS = ha.h
-AM_CFLAGS = -fPIC -D_FILE_OFFSET_BITS=64 -D_GNU_SOURCE -Wall -D$(GF_HOST_OS) \
- -I$(top_srcdir)/libglusterfs/src -shared -nostartfiles $(GF_CFLAGS)
+AM_CPPFLAGS = $(GF_CPPFLAGS) -I$(top_srcdir)/libglusterfs/src
+
+AM_CFLAGS = -Wall $(GF_CFLAGS)
CLEANFILES =
diff --git a/xlators/cluster/ha/src/ha-helpers.c b/xlators/cluster/ha/src/ha-helpers.c
index 1e4af1b62..19be1ed27 100644
--- a/xlators/cluster/ha/src/ha-helpers.c
+++ b/xlators/cluster/ha/src/ha-helpers.c
@@ -1,22 +1,12 @@
/*
- Copyright (c) 2008-2011 Gluster, Inc. <http://www.gluster.com>
- This file is part of GlusterFS.
-
- GlusterFS is free software; you can redistribute it and/or modify
- it under the terms of the GNU General Public License as published
- by the Free Software Foundation; either version 3 of the License,
- or (at your option) any later version.
-
- GlusterFS is distributed in the hope that it will be useful, but
- WITHOUT ANY WARRANTY; without even the implied warranty of
- MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU
- General Public License for more details.
-
- You should have received a copy of the GNU General Public License
- along with this program. If not, see
- <http://www.gnu.org/licenses/>.
-*/
+ Copyright (c) 2008-2012 Red Hat, Inc. <http://www.redhat.com>
+ This file is part of GlusterFS.
+ This file is licensed to you under your choice of the GNU Lesser
+ General Public License, version 3 or any later version (LGPLv3 or
+ later), or the GNU General Public License, version 2 (GPLv2), in all
+ cases as published by the Free Software Foundation.
+*/
#include "xlator.h"
#include "call-stub.h"
#include "defaults.h"
diff --git a/xlators/cluster/ha/src/ha-mem-types.h b/xlators/cluster/ha/src/ha-mem-types.h
index 9bfb3972b..e5e97d237 100644
--- a/xlators/cluster/ha/src/ha-mem-types.h
+++ b/xlators/cluster/ha/src/ha-mem-types.h
@@ -1,24 +1,13 @@
-
/*
- Copyright (c) 2008-2011 Gluster, Inc. <http://www.gluster.com>
+ Copyright (c) 2008-2012 Red Hat, Inc. <http://www.redhat.com>
This file is part of GlusterFS.
- GlusterFS is free software; you can redistribute it and/or modify
- it under the terms of the GNU General Public License as published
- by the Free Software Foundation; either version 3 of the License,
- or (at your option) any later version.
-
- GlusterFS is distributed in the hope that it will be useful, but
- WITHOUT ANY WARRANTY; without even the implied warranty of
- MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU
- General Public License for more details.
-
- You should have received a copy of the GNU General Public License
- along with this program. If not, see
- <http://www.gnu.org/licenses/>.
+ This file is licensed to you under your choice of the GNU Lesser
+ General Public License, version 3 or any later version (LGPLv3 or
+ later), or the GNU General Public License, version 2 (GPLv2), in all
+ cases as published by the Free Software Foundation.
*/
-
#ifndef __HA_MEM_TYPES_H__
#define __HA_MEM_TYPES_H__
diff --git a/xlators/cluster/ha/src/ha.c b/xlators/cluster/ha/src/ha.c
index 38d4229d3..3eccb516b 100644
--- a/xlators/cluster/ha/src/ha.c
+++ b/xlators/cluster/ha/src/ha.c
@@ -1,22 +1,12 @@
/*
- Copyright (c) 2008-2011 Gluster, Inc. <http://www.gluster.com>
- This file is part of GlusterFS.
-
- GlusterFS is free software; you can redistribute it and/or modify
- it under the terms of the GNU General Public License as published
- by the Free Software Foundation; either version 3 of the License,
- or (at your option) any later version.
-
- GlusterFS is distributed in the hope that it will be useful, but
- WITHOUT ANY WARRANTY; without even the implied warranty of
- MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU
- General Public License for more details.
-
- You should have received a copy of the GNU General Public License
- along with this program. If not, see
- <http://www.gnu.org/licenses/>.
-*/
+ Copyright (c) 2008-2012 Red Hat, Inc. <http://www.redhat.com>
+ This file is part of GlusterFS.
+ This file is licensed to you under your choice of the GNU Lesser
+ General Public License, version 3 or any later version (LGPLv3 or
+ later), or the GNU General Public License, version 2 (GPLv2), in all
+ cases as published by the Free Software Foundation.
+*/
/* generate errors randomly, code is simple now, better alogorithm
* can be written to decide what error to be returned and when
*/
@@ -1876,13 +1866,9 @@ err:
}
if (hafdp) {
- if (hafdp->fdstate) {
- GF_FREE (hafdp->fdstate);
- }
+ GF_FREE (hafdp->fdstate);
- if (hafdp->path) {
- GF_FREE (hafdp->path);
- }
+ GF_FREE (hafdp->path);
GF_FREE (hafdp);
}
diff --git a/xlators/cluster/ha/src/ha.h b/xlators/cluster/ha/src/ha.h
index 39b6851e7..e2ed7eaa6 100644
--- a/xlators/cluster/ha/src/ha.h
+++ b/xlators/cluster/ha/src/ha.h
@@ -1,22 +1,12 @@
/*
- Copyright (c) 2008-2011 Gluster, Inc. <http://www.gluster.com>
- This file is part of GlusterFS.
-
- GlusterFS is free software; you can redistribute it and/or modify
- it under the terms of the GNU General Public License as published
- by the Free Software Foundation; either version 3 of the License,
- or (at your option) any later version.
-
- GlusterFS is distributed in the hope that it will be useful, but
- WITHOUT ANY WARRANTY; without even the implied warranty of
- MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU
- General Public License for more details.
-
- You should have received a copy of the GNU General Public License
- along with this program. If not, see
- <http://www.gnu.org/licenses/>.
-*/
+ Copyright (c) 2008-2012 Red Hat, Inc. <http://www.redhat.com>
+ This file is part of GlusterFS.
+ This file is licensed to you under your choice of the GNU Lesser
+ General Public License, version 3 or any later version (LGPLv3 or
+ later), or the GNU General Public License, version 2 (GPLv2), in all
+ cases as published by the Free Software Foundation.
+*/
#ifndef __HA_H_
#define __HA_H_
diff --git a/xlators/cluster/map/src/Makefile.am b/xlators/cluster/map/src/Makefile.am
index 26e19137a..a278b05e2 100644
--- a/xlators/cluster/map/src/Makefile.am
+++ b/xlators/cluster/map/src/Makefile.am
@@ -1,15 +1,16 @@
xlator_LTLIBRARIES = map.la
xlatordir = $(libdir)/glusterfs/$(PACKAGE_VERSION)/xlator/testing/cluster
-map_la_LDFLAGS = -module -avoidversion
+map_la_LDFLAGS = -module -avoid-version
map_la_SOURCES = map.c map-helper.c
map_la_LIBADD = $(top_builddir)/libglusterfs/src/libglusterfs.la
noinst_HEADERS = map.h
-AM_CFLAGS = -fPIC -D_FILE_OFFSET_BITS=64 -D_GNU_SOURCE -Wall -D$(GF_HOST_OS) \
- -I$(top_srcdir)/libglusterfs/src -shared -nostartfiles $(GF_CFLAGS)
+AM_CPPFLAGS = $(GF_CPPFLAGS) -I$(top_srcdir)/libglusterfs/src
+
+AM_CFLAGS = -Wall $(GF_CFLAGS)
CLEANFILES =
diff --git a/xlators/cluster/map/src/map-helper.c b/xlators/cluster/map/src/map-helper.c
index 81212fcfd..851397b68 100644
--- a/xlators/cluster/map/src/map-helper.c
+++ b/xlators/cluster/map/src/map-helper.c
@@ -1,22 +1,12 @@
/*
- Copyright (c) 2009-2011 Gluster, Inc. <http://www.gluster.com>
- This file is part of GlusterFS.
-
- GlusterFS is free software; you can redistribute it and/or modify
- it under the terms of the GNU General Public License as published
- by the Free Software Foundation; either version 3 of the License,
- or (at your option) any later version.
-
- GlusterFS is distributed in the hope that it will be useful, but
- WITHOUT ANY WARRANTY; without even the implied warranty of
- MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU
- General Public License for more details.
-
- You should have received a copy of the GNU General Public License
- along with this program. If not, see
- <http://www.gnu.org/licenses/>.
-*/
+ Copyright (c) 2009-2012 Red Hat, Inc. <http://www.redhat.com>
+ This file is part of GlusterFS.
+ This file is licensed to you under your choice of the GNU Lesser
+ General Public License, version 3 or any later version (LGPLv3 or
+ later), or the GNU General Public License, version 2 (GPLv2), in all
+ cases as published by the Free Software Foundation.
+*/
#ifndef _CONFIG_H
#define _CONFIG_H
#include "config.h"
diff --git a/xlators/cluster/map/src/map-mem-types.h b/xlators/cluster/map/src/map-mem-types.h
index 669b93dc2..3e89f4736 100644
--- a/xlators/cluster/map/src/map-mem-types.h
+++ b/xlators/cluster/map/src/map-mem-types.h
@@ -1,24 +1,13 @@
-
/*
- Copyright (c) 2008-2011 Gluster, Inc. <http://www.gluster.com>
+ Copyright (c) 2008-2012 Red Hat, Inc. <http://www.redhat.com>
This file is part of GlusterFS.
- GlusterFS is free software; you can redistribute it and/or modify
- it under the terms of the GNU General Public License as published
- by the Free Software Foundation; either version 3 of the License,
- or (at your option) any later version.
-
- GlusterFS is distributed in the hope that it will be useful, but
- WITHOUT ANY WARRANTY; without even the implied warranty of
- MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU
- General Public License for more details.
-
- You should have received a copy of the GNU General Public License
- along with this program. If not, see
- <http://www.gnu.org/licenses/>.
+ This file is licensed to you under your choice of the GNU Lesser
+ General Public License, version 3 or any later version (LGPLv3 or
+ later), or the GNU General Public License, version 2 (GPLv2), in all
+ cases as published by the Free Software Foundation.
*/
-
#ifndef __MAP_MEM_TYPES_H__
#define __MAP_MEM_TYPES_H__
diff --git a/xlators/cluster/map/src/map.c b/xlators/cluster/map/src/map.c
index ead9da0b9..6150a33ce 100644
--- a/xlators/cluster/map/src/map.c
+++ b/xlators/cluster/map/src/map.c
@@ -1,22 +1,12 @@
/*
- Copyright (c) 2010-2011 Gluster, Inc. <http://www.gluster.com>
- This file is part of GlusterFS.
-
- GlusterFS is free software; you can redistribute it and/or modify
- it under the terms of the GNU General Public License as published
- by the Free Software Foundation; either version 3 of the License,
- or (at your option) any later version.
-
- GlusterFS is distributed in the hope that it will be useful, but
- WITHOUT ANY WARRANTY; without even the implied warranty of
- MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU
- General Public License for more details.
-
- You should have received a copy of the GNU General Public License
- along with this program. If not, see
- <http://www.gnu.org/licenses/>.
-*/
+ Copyright (c) 2010-2012 Red Hat, Inc. <http://www.redhat.com>
+ This file is part of GlusterFS.
+ This file is licensed to you under your choice of the GNU Lesser
+ General Public License, version 3 or any later version (LGPLv3 or
+ later), or the GNU General Public License, version 2 (GPLv2), in all
+ cases as published by the Free Software Foundation.
+*/
#ifndef _CONFIG_H
#define _CONFIG_H
#include "config.h"
@@ -2375,8 +2365,7 @@ fini (xlator_t *this)
priv = this->private;
if (priv) {
- if (priv->xlarray)
- GF_FREE (priv->xlarray);
+ GF_FREE (priv->xlarray);
trav_map = priv->map;
while (trav_map) {
diff --git a/xlators/cluster/map/src/map.h b/xlators/cluster/map/src/map.h
index bccac437c..7703a543e 100644
--- a/xlators/cluster/map/src/map.h
+++ b/xlators/cluster/map/src/map.h
@@ -1,22 +1,12 @@
/*
- Copyright (c) 2008-2011 Gluster, Inc. <http://www.gluster.com>
- This file is part of GlusterFS.
+ Copyright (c) 2008-2012 Red Hat, Inc. <http://www.redhat.com>
+ This file is part of GlusterFS.
- GlusterFS is free software; you can redistribute it and/or modify
- it under the terms of the GNU General Public License as published
- by the Free Software Foundation; either version 3 of the License,
- or (at your option) any later version.
-
- GlusterFS is distributed in the hope that it will be useful, but
- WITHOUT ANY WARRANTY; without even the implied warranty of
- MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU
- General Public License for more details.
-
- You should have received a copy of the GNU General Public License
- along with this program. If not, see
- <http://www.gnu.org/licenses/>.
+ This file is licensed to you under your choice of the GNU Lesser
+ General Public License, version 3 or any later version (LGPLv3 or
+ later), or the GNU General Public License, version 2 (GPLv2), in all
+ cases as published by the Free Software Foundation.
*/
-
#ifndef __MAP_H__
#define __MAP_H__
diff --git a/xlators/cluster/stripe/src/Makefile.am b/xlators/cluster/stripe/src/Makefile.am
index 0db3c9eeb..2d151422a 100644
--- a/xlators/cluster/stripe/src/Makefile.am
+++ b/xlators/cluster/stripe/src/Makefile.am
@@ -2,16 +2,19 @@
xlator_LTLIBRARIES = stripe.la
xlatordir = $(libdir)/glusterfs/$(PACKAGE_VERSION)/xlator/cluster
-stripe_la_LDFLAGS = -module -avoidversion
+stripe_la_LDFLAGS = -module -avoid-version
+
+stripe_la_SOURCES = stripe.c stripe-helpers.c \
+ $(top_builddir)/xlators/lib/src/libxlator.c
-stripe_la_SOURCES = stripe.c $(top_builddir)/xlators/lib/src/libxlator.c
stripe_la_LIBADD = $(top_builddir)/libglusterfs/src/libglusterfs.la
noinst_HEADERS = stripe.h stripe-mem-types.h $(top_builddir)/xlators/lib/src/libxlator.h
-AM_CFLAGS = -fPIC -D_FILE_OFFSET_BITS=64 -D_GNU_SOURCE -Wall -D$(GF_HOST_OS)\
- -I$(top_srcdir)/libglusterfs/src -shared -nostartfiles $(GF_CFLAGS) \
+AM_CPPFLAGS = $(GF_CPPFLAGS) -I$(top_srcdir)/libglusterfs/src \
-I$(top_srcdir)/xlators/lib/src
+AM_CFLAGS = -Wall $(GF_CFLAGS)
+
CLEANFILES =
diff --git a/xlators/cluster/stripe/src/stripe-helpers.c b/xlators/cluster/stripe/src/stripe-helpers.c
new file mode 100644
index 000000000..a83abdc72
--- /dev/null
+++ b/xlators/cluster/stripe/src/stripe-helpers.c
@@ -0,0 +1,675 @@
+/*
+ Copyright (c) 2008-2012 Red Hat, Inc. <http://www.redhat.com>
+ This file is part of GlusterFS.
+
+ This file is licensed to you under your choice of the GNU Lesser
+ General Public License, version 3 or any later version (LGPLv3 or
+ later), or the GNU General Public License, version 2 (GPLv2), in all
+ cases as published by the Free Software Foundation.
+*/
+
+#include <fnmatch.h>
+
+#include "stripe.h"
+#include "byte-order.h"
+#include "mem-types.h"
+
+void
+stripe_local_wipe (stripe_local_t *local)
+{
+ if (!local)
+ goto out;
+
+ loc_wipe (&local->loc);
+ loc_wipe (&local->loc2);
+
+ if (local->fd)
+ fd_unref (local->fd);
+
+ if (local->inode)
+ inode_unref (local->inode);
+
+ if (local->xattr)
+ dict_unref (local->xattr);
+
+ if (local->xdata)
+ dict_unref (local->xdata);
+
+out:
+ return;
+}
+
+
+
+int
+stripe_aggregate (dict_t *this, char *key, data_t *value, void *data)
+{
+ dict_t *dst = NULL;
+ int64_t *ptr = 0, *size = NULL;
+ int32_t ret = -1;
+
+ dst = data;
+
+ if (strcmp (key, GF_XATTR_QUOTA_SIZE_KEY) == 0) {
+ ret = dict_get_bin (dst, key, (void **)&size);
+ if (ret < 0) {
+ size = GF_CALLOC (1, sizeof (int64_t),
+ gf_common_mt_char);
+ if (size == NULL) {
+ gf_log ("stripe", GF_LOG_WARNING,
+ "memory allocation failed");
+ goto out;
+ }
+ ret = dict_set_bin (dst, key, size, sizeof (int64_t));
+ if (ret < 0) {
+ gf_log ("stripe", GF_LOG_WARNING,
+ "stripe aggregate dict set failed");
+ GF_FREE (size);
+ goto out;
+ }
+ }
+
+ ptr = data_to_bin (value);
+ if (ptr == NULL) {
+ gf_log ("stripe", GF_LOG_WARNING, "data to bin failed");
+ goto out;
+ }
+
+ *size = hton64 (ntoh64 (*size) + ntoh64 (*ptr));
+ } else if (strcmp (key, GF_CONTENT_KEY)) {
+ /* No need to aggregate 'CONTENT' data */
+ ret = dict_set (dst, key, value);
+ if (ret)
+ gf_log ("stripe", GF_LOG_WARNING, "xattr dict set failed");
+ }
+
+out:
+ return 0;
+}
+
+
+void
+stripe_aggregate_xattr (dict_t *dst, dict_t *src)
+{
+ if ((dst == NULL) || (src == NULL)) {
+ goto out;
+ }
+
+ dict_foreach (src, stripe_aggregate, dst);
+out:
+ return;
+}
+
+
+int32_t
+stripe_xattr_aggregate (char *buffer, stripe_local_t *local, int32_t *total)
+{
+ int32_t i = 0;
+ int32_t ret = -1;
+ int32_t len = 0;
+ char *sbuf = NULL;
+ stripe_xattr_sort_t *xattr = NULL;
+
+ if (!buffer || !local || !local->xattr_list)
+ goto out;
+
+ sbuf = buffer;
+
+ for (i = 0; i < local->nallocs; i++) {
+ xattr = local->xattr_list + i;
+ len = xattr->xattr_len;
+
+ if (len && xattr && xattr->xattr_value) {
+ memcpy (buffer, xattr->xattr_value, len);
+ buffer += len;
+ *buffer++ = ' ';
+ }
+ }
+
+ *--buffer = '\0';
+ if (total)
+ *total = buffer - sbuf;
+ ret = 0;
+
+ out:
+ return ret;
+}
+
+int32_t
+stripe_free_xattr_str (stripe_local_t *local)
+{
+ int32_t i = 0;
+ int32_t ret = -1;
+ stripe_xattr_sort_t *xattr = NULL;
+
+ if (!local || !local->xattr_list)
+ goto out;
+
+ for (i = 0; i < local->nallocs; i++) {
+ xattr = local->xattr_list + i;
+
+ if (xattr && xattr->xattr_value)
+ GF_FREE (xattr->xattr_value);
+ }
+
+ ret = 0;
+ out:
+ return ret;
+}
+
+
+int32_t
+stripe_fill_lockinfo_xattr (xlator_t *this, stripe_local_t *local,
+ void **xattr_serz)
+{
+ int32_t ret = -1, i = 0, len = 0;
+ dict_t *tmp1 = NULL, *tmp2 = NULL;
+ char *buf = NULL;
+ stripe_xattr_sort_t *xattr = NULL;
+
+ if (xattr_serz == NULL) {
+ goto out;
+ }
+
+ tmp2 = dict_new ();
+
+ if (tmp2 == NULL) {
+ goto out;
+ }
+
+ for (i = 0; i < local->nallocs; i++) {
+ xattr = local->xattr_list + i;
+ len = xattr->xattr_len;
+
+ if (len && xattr && xattr->xattr_value) {
+ ret = dict_reset (tmp2);
+ if (ret < 0) {
+ gf_log (this->name, GF_LOG_DEBUG,
+ "dict_reset failed (%s)",
+ strerror (-ret));
+ }
+
+ ret = dict_unserialize (xattr->xattr_value,
+ xattr->xattr_len,
+ &tmp2);
+ if (ret < 0) {
+ gf_log (this->name, GF_LOG_WARNING,
+ "dict_unserialize failed (%s)",
+ strerror (-ret));
+ ret = -1;
+ goto out;
+ }
+
+ tmp1 = dict_copy (tmp2, tmp1);
+ if (tmp1 == NULL) {
+ gf_log (this->name, GF_LOG_WARNING,
+ "dict_copy failed (%s)",
+ strerror (-ret));
+ ret = -1;
+ goto out;
+ }
+ }
+ }
+
+ len = dict_serialized_length (tmp1);
+ if (len > 0) {
+ buf = GF_CALLOC (1, len, gf_common_mt_dict_t);
+ if (buf == NULL) {
+ ret = -1;
+ goto out;
+ }
+
+ ret = dict_serialize (tmp1, buf);
+ if (ret < 0) {
+ gf_log (this->name, GF_LOG_WARNING,
+ "dict_serialize failed (%s)", strerror (-ret));
+ ret = -1;
+ goto out;
+ }
+
+ *xattr_serz = buf;
+ }
+
+ ret = 0;
+out:
+ if (tmp1 != NULL) {
+ dict_unref (tmp1);
+ }
+
+ if (tmp2 != NULL) {
+ dict_unref (tmp2);
+ }
+
+ return ret;
+}
+
+
+int32_t
+stripe_fill_pathinfo_xattr (xlator_t *this, stripe_local_t *local,
+ char **xattr_serz)
+{
+ int ret = -1;
+ int32_t padding = 0;
+ int32_t tlen = 0;
+ char stripe_size_str[20] = {0,};
+ char *pathinfo_serz = NULL;
+
+ if (!local) {
+ gf_log (this->name, GF_LOG_ERROR, "Possible NULL deref");
+ goto out;
+ }
+
+ (void) snprintf (stripe_size_str, 20, "%ld",
+ (local->fctx) ? local->fctx->stripe_size : 0);
+
+ /* extra bytes for decorations (brackets and <>'s) */
+ padding = strlen (this->name) + strlen (STRIPE_PATHINFO_HEADER)
+ + strlen (stripe_size_str) + 7;
+ local->xattr_total_len += (padding + 2);
+
+ pathinfo_serz = GF_CALLOC (local->xattr_total_len, sizeof (char),
+ gf_common_mt_char);
+ if (!pathinfo_serz)
+ goto out;
+
+ /* xlator info */
+ (void) sprintf (pathinfo_serz, "(<"STRIPE_PATHINFO_HEADER"%s:[%s]> ",
+ this->name, stripe_size_str);
+
+ ret = stripe_xattr_aggregate (pathinfo_serz + padding, local, &tlen);
+ if (ret) {
+ gf_log (this->name, GF_LOG_ERROR,
+ "Cannot aggregate pathinfo list");
+ goto out;
+ }
+
+ *(pathinfo_serz + padding + tlen) = ')';
+ *(pathinfo_serz + padding + tlen + 1) = '\0';
+
+ *xattr_serz = pathinfo_serz;
+
+ ret = 0;
+ out:
+ return ret;
+}
+
+/**
+ * stripe_get_matching_bs - Get the matching block size for the given path.
+ */
+int32_t
+stripe_get_matching_bs (const char *path, stripe_private_t *priv)
+{
+ struct stripe_options *trav = NULL;
+ uint64_t block_size = 0;
+
+ GF_VALIDATE_OR_GOTO ("stripe", priv, out);
+ GF_VALIDATE_OR_GOTO ("stripe", path, out);
+
+ LOCK (&priv->lock);
+ {
+ block_size = priv->block_size;
+ trav = priv->pattern;
+ while (trav) {
+ if (!fnmatch (trav->path_pattern, path, FNM_NOESCAPE)) {
+ block_size = trav->block_size;
+ break;
+ }
+ trav = trav->next;
+ }
+ }
+ UNLOCK (&priv->lock);
+
+out:
+ return block_size;
+}
+
+int32_t
+stripe_ctx_handle (xlator_t *this, call_frame_t *prev, stripe_local_t *local,
+ dict_t *dict)
+{
+ char key[256] = {0,};
+ data_t *data = NULL;
+ int32_t index = 0;
+ stripe_private_t *priv = NULL;
+
+ priv = this->private;
+
+
+ if (!local->fctx) {
+ local->fctx = GF_CALLOC (1, sizeof (stripe_fd_ctx_t),
+ gf_stripe_mt_stripe_fd_ctx_t);
+ if (!local->fctx) {
+ local->op_errno = ENOMEM;
+ local->op_ret = -1;
+ goto out;
+ }
+
+ local->fctx->static_array = 0;
+ }
+ /* Stripe block size */
+ sprintf (key, "trusted.%s.stripe-size", this->name);
+ data = dict_get (dict, key);
+ if (!data) {
+ local->xattr_self_heal_needed = 1;
+ gf_log (this->name, GF_LOG_ERROR,
+ "Failed to get stripe-size");
+ goto out;
+ } else {
+ if (!local->fctx->stripe_size) {
+ local->fctx->stripe_size =
+ data_to_int64 (data);
+ }
+
+ if (local->fctx->stripe_size != data_to_int64 (data)) {
+ gf_log (this->name, GF_LOG_WARNING,
+ "stripe-size mismatch in blocks");
+ local->xattr_self_heal_needed = 1;
+ }
+ }
+
+ /* Stripe count */
+ sprintf (key, "trusted.%s.stripe-count", this->name);
+ data = dict_get (dict, key);
+
+ if (!data) {
+ local->xattr_self_heal_needed = 1;
+ gf_log (this->name, GF_LOG_ERROR,
+ "Failed to get stripe-count");
+ goto out;
+ }
+ if (!local->fctx->xl_array) {
+ local->fctx->stripe_count = data_to_int32 (data);
+ if (!local->fctx->stripe_count) {
+ gf_log (this->name, GF_LOG_ERROR,
+ "error with stripe-count xattr");
+ local->op_ret = -1;
+ local->op_errno = EIO;
+ goto out;
+ }
+
+ local->fctx->xl_array = GF_CALLOC (local->fctx->stripe_count,
+ sizeof (xlator_t *),
+ gf_stripe_mt_xlator_t);
+
+ if (!local->fctx->xl_array) {
+ local->op_errno = ENOMEM;
+ local->op_ret = -1;
+ goto out;
+ }
+ }
+ if (local->fctx->stripe_count != data_to_int32 (data)) {
+ gf_log (this->name, GF_LOG_ERROR,
+ "error with stripe-count xattr (%d != %d)",
+ local->fctx->stripe_count, data_to_i