summaryrefslogtreecommitdiffstats
path: root/xlators/mgmt/glusterd
Commit message (Collapse)AuthorAgeFilesLines
...
* multiple files: ensure nfs_svc functions are availableYaniv Kaul2020-01-237-25/+48
| | | | | | | | | | | | When gNFS is disabled in build time, we have to ensure taht the .stop(), .start() and other functions of the nfs_svc are not called, otherwise we'd crash. In addition, #ifdef more code that is gNFS related. updates: bz#1793995 Signed-off-by: Yaniv Kaul <ykaul@redhat.com> Change-Id: I5081f1670c04ca306aeaab7208829b0f2f149a42
* dictionary: remove the 'extra_free' parameterYaniv Kaul2020-01-212-4/+6
| | | | | | | | | | | | | | | This parameter may have been used in the past, but is no longer needed. Removing it and the few locations it was actually referenced. This allows to remove an extra memdup as well, that was not needed in the 1st place in server_setvolume() and unserialize_rsp_direntp() functions. A followup separate patch will remove extra_stdfree parmeter from the dictionary structure. Change-Id: Ica0ff0a330672373aaa60e808b7e76ec489a0fe3 updates: bz#1193929 Signed-off-by: Yaniv Kaul <ykaul@redhat.com>
* glusterd: fix use after freeXie Changlong2020-01-201-1/+3
| | | | | | updates: bz#789278 Change-Id: I652d8d4428cf6ce61b712a66d309e78030a5f911 Signed-off-by: Xie Changlong <xiechanglong@cmss.chinamobile.com>
* multiple xlators: reduce key lengthYaniv Kaul2020-01-1411-87/+84
| | | | | | | | | | | | | | | In many cases, we were freely allocating long keys with no need. Smaller char arrays are just fine almost anywhere, so just went ahead and looked where they we can use smaller ones. In some cases, annotated the functions as static and the prefixes passed as const as it was easier to read and understand. Where relevant, converted the dict functions to use known key length. Change-Id: I882ab33ea20d90b63278336cd1370c09ffdab7f2 updates: bz#1193929 Signed-off-by: Yaniv Kaul <ykaul@redhat.com>
* glusterd-volgen.c: fix to disable property when gNFS disabled.Yaniv Kaul2020-01-131-8/+10
| | | | | | | | | | | | | | | | | | | The code was not ifdef'ed properly when gNFS was not enabled. Strangely, I could not reproduce the failure on my system (Fedora 31), but it was reproduced elsehwere and the fix was verified to correct it. The failure: gluster volume create testvol replica 3 127.0.0.2:/tests/brick{1..3} force gluster v set testvol write-behind off grep -rne write-behind /var/lib/glusterd/vols/testvol/trusted-testvol.tcp-fuse.vol The last grep was supposed to come out empty. The issue was that perfxl_option_handler may not have been called when it should have been. Change-Id: Ie9f8ec87dabeef6624527c2266ddf9af01ca7373 updates: bz#1193929 Signed-off-by: Yaniv Kaul <ykaul@redhat.com>
* afr: expose cluster.optimistic-change-log to CLI.Ravishankar N2020-01-071-0/+5
| | | | | | | | | | | This volume option was not made avaialble to `gluster volume set` CLI. Reported-by: epolakis(https://github.com/kinsu) in https://github.com/gluster/glusterfs/issues/781 fixes: bz#1787554 Change-Id: I7141bdd4e53ee99e22b354edde8d023bfc0b2cd7 Signed-off-by: Ravishankar N <ravishankar@redhat.com>
* mgmt/glusterd: Correct the log path of valgrind.He Min2020-01-061-1/+1
| | | | | | | | | | | | | | | | | Problem: Currently enabling run-with-valgrind will cause the gnfs and quota to fail to start. The phenomenon is as follows. ------------------------------------------------------------------------------ NFS Server on localhost 2049 0 N 48406 Quota Daemon on localhost N/A N/A N 48428 ------------------------------------------------------------------------------ Solution: The cause of the above phenomenon is that the log path of valgrind is set incorrectly. Gnfs and quota can start with valgrind normally after correcting the log path. Updates: #788 Change-Id: Ib91408c08522ff66afff908fbff3fce4b93ea770 Signed-off-by: He Min <hemin@cmss.chinamobile.com>
* Remove some compiler warnings when not compiling gNFSYaniv Kaul2020-01-022-4/+4
| | | | | | | | | | As a follow up to https://review.gluster.org/#/c/glusterfs/+/23799/ When compiling without gNFS, there were some 'unused' warnings by the compiler. This patch fixes them. Change-Id: I621562261f53950e821a450e0e7da304d00ae557 updates: bz#1193929 Signed-off-by: Yaniv Kaul <ykaul@redhat.com>
* glusterd: deafult options after volume resetSanju Rakonde2020-01-011-0/+23
| | | | | | | | | | | | | | | | | | | Problem: default option itransport.address-family is disappered in volume info output after a volume reset. Cause: with 3.8.0 onwards volume option transport.address-family has default value, any volume which is created will have this option set. So, volume info will show this in its output. But, with reset volume, this option is not handled. Solution: In glusterd_enable_default_options(), we should add this option along with other default options. This function is called by glusterd_options_reset() with volume reset command. fixes: bz#1786478 Change-Id: I58f7aa24cf01f308c4efe6cae748cc3bc8b99b1d Signed-off-by: Sanju Rakonde <srakonde@redhat.com>
* [RFC]#ifdef gNFS related code if we are not compiling gNFSYaniv Kaul2019-12-1815-27/+61
| | | | | | | | | | | | | | | | If we are not compiling gNFS (--enable-gnfs is not given in the ./configure script params), there is little point in compiling code that is related to it. This patch tries to eliminate it. My hope (and it's not clear from the code ) is that I did not break the NFS Ganesha support as well. Other than that, tried to compile with and without anad it looks sane. Change-Id: I8d6c98066b9fceab4ec10fc6f5e81ab069e853bd updates: bz#1193929 Signed-off-by: Yaniv Kaul <ykaul@redhat.com>
* glusterd: unlink the file after killing the processSanju Rakonde2019-12-181-0/+2
| | | | | | | | | | In glusterd_proc_stop(), after killing the pid we should remove the pidfile. fixes: bz#1784375 Change-Id: Ib6367aed590932c884b0f6f892fc40542aa19686 Signed-off-by: Sanju Rakonde <srakonde@redhat.com>
* glusterd: set xaatrs after checking the brick orderSanju Rakonde2019-12-051-30/+30
| | | | | | | | | | | | | | | | | | | | | | Problem: When volume creation fails complaining about the bricks from same hosts for replica volumes, the bricks can't be re-used to create any volume without using force at the end. It says, brick is already part of a volume. Reason: When volume create opeartion issued, we set xattrs on the bricks. If the transaction fails in later checks, the xattrs will remain on the brick. When the brick is re-used, by looking at the xattrs, glusterd thinks it is already part of volume. Solution: Check the brick order for replica and disperse volumes before setting the xattrs. fixes: bz#1776801 Change-Id: I44a971b37f520e5a20dc9fad6520286d315063b9 Signed-off-by: Sanju Rakonde <srakonde@redhat.com>
* glusterd-store: fix Coverity CID 1407767Xie Changlong2019-12-041-0/+2
| | | | | | | | To avoid memory leak. updates: bz#789278 Change-Id: I321c65a66c9ac372d059aa3ca89a989c6089979e Signed-off-by: Xie Changlong <xiechanglong@cmss.chinamobile.com>
* glusterd-op-sm.c (and others) - improve glusterd_op_stage_set_volume()Yaniv Kaul2019-11-295-306/+219
| | | | | | | | | | | | | | | | | | | | | | Multiple changes to the function in the hope to make it somewhat faster. 1. Checking for key length against constant strings before calling strcmp() to save some calls. 2. Verifying if a match was already made against the key to reduce yet more checks. 3. Alignment of error message when they can fit on less lines - just makes 'grep' on the code for error messages easier and it's more readable. 4. Multiple functions where call _gd_get_vmep() one by one. Instead, extracted it to be callable (it was static) and re-used its result, instead of calling it again and again. 5. Removed some unneeded include statement. 6. Removed redundant null checks. Hopefully, no functional changes. Change-Id: Id281224e49adeca6757f96653b4cb13c7c9ba8c9 updates: bz#1193929 Signed-off-by: Yaniv Kaul <ykaul@redhat.com>
* debug/io-stats: add an option to set volume-idAmar Tumballi2019-11-291-0/+9
| | | | | | | | | | | 'volume-id' is good to have for a graph for uniquely identifying it. Add it to graph->volume_id while generating volfile itself. This can be further used in many other places. Updates: #763 Change-Id: I80516d62d28a284e8ff4707841570ced97a37e73 Signed-off-by: Amar Tumballi <amar@kadalu.io>
* store.c/glusterd-store.c: remove sys_stat callsYaniv Kaul2019-11-271-3/+2
| | | | | | | | | | | | | | | Instead of querying for the file size and allocating a char array according to its size, let's just use a fixed size. Those calls are not really needed, and are either expensive or cached anyway. Since we do dynamic allocation/free, let's just use a fixed array instead. I'll see if there are other sys_stat() calls that are not really useful and try to eliminate them in separate patches. Change-Id: I76b40e78a52ab38f613fc0cdef4be60e6253bf20 updates: bz#1193929 Signed-off-by: Yaniv Kaul <ykaul@redhat.com>
* glusterd-volgen.c: improve volgen_graph_set_options_generic()Yaniv Kaul2019-11-181-11/+12
| | | | | | | | | | | | | | | Skip fetching "skip-CLIOT" unconditionally on every invocation of volgen_graph_set_options_generic(). Instead, fetch only if the vme->key matches to it. We calculate the length of vme->key (but we would have done it anyway in dict_get() later on, so now we can use dict_getn() instead and re-use that key length) and check if the lengths match before doing a strcmp() between them. Lastly, if they match, we actually do the fetch. Change-Id: I9d9a7104f9e920bf81477128adb5fc87f5d30627 updates: bz#1193929 Signed-off-by: Yaniv Kaul <ykaul@redhat.com>
* [WIP]gluster-volgen.c: remove more of JBR and FDL xlatorsYaniv Kaul2019-11-132-139/+1
| | | | | | | | | the JBR and FDL experimental xlators were apparently removed. Removed additional leftovers scattered in the code. Change-Id: I78b6fa5fd9044dc48cdcb1fb094b8c267c2d1323 updates: bz#1193929 Signed-off-by: Yaniv Kaul <ykaul@redhat.com>
* glusterd: Client Handling of Elastic ClustersMohit Agrawal2019-11-121-0/+49
| | | | | | | | | | | | | | Configure the list of gluster servers in the key GLUSTERD_BRICK_SERVERS at the time of GETSPEC RPC CALL and access the value in client side to update volfile serve list so that client would be able to connect next volfile server in case of current volfile server is down Updates #741 Signed-off-by: Mohit Agrawal <moagrawa@redhat.com> Change-Id: I23f36ddb92982bb02ffd83937a8bd8a2c97e8104
* glusterd/ganesha: fix Coverity CID 1405785Xie Changlong2019-10-301-1/+1
| | | | | | | | | To avoid double free updates: bz#789278 Change-Id: I15ae54ed696295d4cb015668722e77983b062ccb Signed-off-by: Xie Changlong <xiechanglong@cmss.chinamobile.com>
* glusterd: display correct rebalance data size after glusterd restartSanju Rakonde2019-10-181-5/+5
| | | | | | | | | | | | | | | | | | | Problem: After completion of rebalance, if glusterd is restarted, rebalance status displays wrong rebalance data size in its output. Cause: While glusterd restoring the information from /var/lib/glusterd/ into its memory, glusterd fetches rebalance_data from /var/lib/glusterd/vols/volname/node_state.info. This value is converted into an integer using atoi(), which is returning incorrect value for larger values. Solution: use sscanf() instead of atoi() to convert string to integer(in this case it is unsigned long) fixes: bz#1762438 Change-Id: Icbdb096919612b4a1d6fb0e315f09d38900abf4e Signed-off-by: Sanju Rakonde <srakonde@redhat.com>
* rpc: align structsYaniv Kaul2019-10-174-203/+197
| | | | | | | | | | | | | squash tens of warnings on padding of structs in afr structures. The warnings were found by manually added '-Wpadded' to the GCC command line. Also made relevant structs and definitions static, where it was applicable. Change-Id: Ib71a7e9c6179378f072d796d11172d086c343e53 updates: bz#1193929 Signed-off-by: Yaniv Kaul <ykaul@redhat.com>
* glusterd-volgen.c: fix setting the key export-dirYaniv Kaul2019-10-142-108/+41
| | | | | | | | | | | | | | | | | | | | | | | | The return value of gf_asprintf() was used as the key length, and it was stored into ret. ret was re-used before the dict function was called, therefore was invalid as the real key length and contained a different value. That was masked by the fact the key length was only used in key hash, so while it was incorrect, it was harmless. The hash was consistent and the key length was re-calculated anyway in dict_set_lk(). This patch fixes it, so later on we can use the key length also in dict_set_lk() to save another strlen() (sent in a different patch). However, in the course of this patch I've also decided to reduce the copy-pasta in this code path and put all NFS options in an array and iterate through them. This makes the code shorter and easier to read. (It's also more efficient since once an entry was found, there's no additional strcmp()'s as the previous code did). Change-Id: I968ed50a55f3b1a7ad027c72b06e0fa3788eaa9b updates: bz#1193929 Signed-off-by: Yaniv Kaul <ykaul@redhat.com>
* glusterd: Re-alignment of structure attributes in glusterd.Purna Pavan Chandra Aekkaladevi2019-10-147-13/+13
| | | | | | | | | Structures defined in few files of glusterd src have been modified into memory efficient manner. Change-Id: Ie712b41cc74b04d5169bfd2402bff3b9c0d8c95b Updates: bz#1754448 Signed-off-by: Purna Pavan Chandra Aekkaladevi <paekkala@redhat.com>
* glusterd-utils.c: attach_brick() - remove dead codeYaniv Kaul2019-10-111-5/+0
| | | | | | | | | pidfile1 and pidfile2 were not used anywhere. Removed the assignment and the variables. Change-Id: Ic5fe091ba28bb500c370410a63440953048fd0b7 updates: bz#1193929 Signed-off-by: Yaniv Kaul <ykaul@redhat.com>
* glusterd: rebalance start should fail when quorum is not metSanju Rakonde2019-10-101-1/+2
| | | | | | | | | | rebalance start should not succeed if quorum is not met. this patch adds a condition to check whether quorum is met in pre-validation stage. fixes: bz#1760467 Change-Id: Ic7d0d08f69e4bc6d5e7abae713ec1881531c8ad4 Signed-off-by: Sanju Rakonde <srakonde@redhat.com>
* glusterd: improve loggingSanju Rakonde2019-10-011-3/+3
| | | | | | | updates: bz#1193929 Change-Id: I5b4a39fbdaa43642a322440d550ca24df815cae9 Signed-off-by: Sanju Rakonde <srakonde@redhat.com>
* read-ahead/io-cache: turn off by defaultRaghavendra Gowdappa2019-09-261-2/+2
| | | | | | | | | | | | | | | | We've found perf xlators io-cache and read-ahead not adding any performance improvement. At best read-ahead is redundant due to kernel read-ahead and at worst io-cache is degrading the performance for workloads that doesn't involve re-read. Given that VFS already have both these functionalities, this patch makes these two translators turned off by default for native fuse mounts. For non-native fuse mounts like gfapi (NFS-ganesha/samba) we can have these xlators on by having custom profiles. Change-Id: Ie7535788909d4c741844473696f001274dc0bb60 Signed-off-by: Raghavendra Gowdappa <rgowdapp@redhat.com> fixes: bz#1676479
* glusterd/ganesha: fixing resource leak in tear_down_cluster()Jiffin Tony Thottan2019-09-231-0/+8
| | | | | | | | CID: 1370947 Updates: bz#789278 Change-Id: Ib694056430ff0536ed705a0e77e5ace22486891e Signed-off-by: Jiffin Tony Thottan <jthottan@redhat.com>
* mgmt/glusterd: Fixed typos and reworded logsN Balachandran2019-09-232-8/+8
| | | | | | | | Fixed typos and reworded log messages for clarity. Change-Id: I46f616ce7d3eb993c77a5812e8bc044e5f283354 Fixes: bz#1753859 Signed-off-by: N Balachandran <nbalacha@redhat.com>
* protocol/client: don't reopen fds on which POSIX locks are held after a ↵Raghavendra G2019-09-121-0/+13
| | | | | | | | | | | | | | | | | | | reconnect Bricks cleanup any granted locks after a client disconnects and currently these locks are not healed after a reconnect. This means post reconnect a competing process could be granted a lock even though the first process which was granted locks has not unlocked. By not re-opening fds, subsequent operations on such fds will fail forcing the application to close the current fd and reopen a new one. This way we prevent any silent corruption. A new option "client.strict-locks" is introduced to control this behaviour. This option is set to "off" by default. Change-Id: Ieed545efea466cb5e8f5a36199aa26380c301b9e Signed-off-by: Raghavendra G <rgowdapp@redhat.com> updates: bz#1694920
* glusterd, rpc, glusterfsd: fix coverity defects and put required annotationsAtin Mukherjee2019-09-103-1/+4
| | | | | | | | | | | 1404965 - Null pointer dereference 1404316 - Program hangs 1401715 - Program hangs 1401713 - Program hangs Updates: bz#789278 Change-Id: I6e6575daafcb067bc910445f82a9d564f43b75a2 Signed-off-by: Atin Mukherjee <amukherj@redhat.com>
* cluster/ec: quorum-count implementationPranith Kumar K2019-09-081-0/+46
| | | | | | fixes: #721 Change-Id: I5333540e3c635ccf441cf1f4696e4c8986e38ea8 Signed-off-by: Pranith Kumar K <pkarampu@redhat.com>
* glusterd: IPV6 hostname address is not parsed correctlyMohit Agrawal2019-09-061-5/+11
| | | | | | | | | | | | Problem: IPV6 hostname address is not parsed correctly in function glusterd_check_brick_order Solution: Update the code to parse hostname address Change-Id: Ifb2f83f9c6e987b2292070e048e97eeb51b728ab Fixes: bz#1747746 Credits: Amgad Saleh <amgad.saleh@nokia.com> Signed-off-by: Mohit Agrawal <moagrawal@redhat.com>
* glusterd-store.c: remove of dead codeYaniv Kaul2019-09-051-130/+0
| | | | | | | | These functions do not seem to be in use. Change-Id: Ie76baf2a9727b9ba0e66f234226b1e62788245f2 updates: bz#1193929 Signed-off-by: Yaniv Kaul <ykaul@redhat.com>
* build: Fix libglusterd Makefile targetAnoop C S2019-08-301-4/+0
| | | | | | | | | | | | | * Fix libglusterd.la target path in cli/src/Makefile.am * Like libglusterfs, libgfxdr and libgfrpc, libglusterd is also expected to be ready by the time xlators/mgmt/glusterd sources are compiled. Therefore this change removes the additional mentioning of libglusterd.la target in Makefile.am Change-Id: I1b787316cfb6cd7487f49e661490b9788a0b80b3 Updates: bz#1193929 Signed-off-by: Anoop C S <anoopcs@redhat.com>
* glusterd: Fixed incorrect size argumentN Balachandran2019-08-271-2/+3
| | | | | | | | | | An incorrect size argument to snprintf caused the glusterd process to crash on startup. This has been fixed. Change-Id: Iddafb5468866d0182cd8239210c92c893e643285 Fixes: bz#1745965 Signed-off-by: N Balachandran <nbalacha@redhat.com>
* glusterd: Unused value coverity fixSanju Rakonde2019-08-261-0/+5
| | | | | | | | CID: 1288765 updates: bz#789278 Change-Id: Ie6b01f81339769f44d82fd7c32ad0ed1a697c69c Signed-off-by: Sanju Rakonde <srakonde@redhat.com>
* glusterd: stop stale bricks during handshaking in brick mux modeAtin Mukherjee2019-08-254-9/+55
| | | | | | | | | | | | | | | | | | | | This patch addresses two problems: 1. During friend handshaking, if a volume is imported due to change in the version, the old bricks were not stopped which would lead to a situation where bricks will run with old volfiles. 2. As part of attaching shd service in glusterd_attach_svc, there might be a case that the volume for which we're attempting to attach a shd service might become stale and in the process of deletion and hence in every retrials (if the rpc connection isn't ready) check for the existance of the volume and then only attempt the further attach request. Fixes: bz#1733425 Change-Id: I6bac6b871f7e31cb5bf277db979289dec196a03e Signed-off-by: Atin Mukherjee <amukherj@redhat.com> Signed-off-by: Mohammed Rafi KC <rkavunga@redhat.com>
* glusterd: Add warning and abort in case of failures in migration during ↵heal-infoVishal Pandey2019-08-251-0/+11
| | | | | | | | | | | | | | | | | | remove-brick commit Problem - Currently remove-brick commit goes through even though there were files that failed to migrate or were skipped. There is no warning raised to the user. Solution- Add a check in the remove brick staging phase to verify if the status of the rebalnce process is complete but there has been failures or some skipped files while migration, In this case user will be given a warning and remove-brick commit. User will need to use the force option to remove the bricks. Fixes: bz#1514683 Signed-offby- Vishal Pandey <vpandey@redhat.com> Change-Id: I014d0f0afb4b2fac35ab0de52227f98dbae079d5
* Revert "glusterd: (storhaug) remove ganesha (843e1b0)"Jiffin Tony Thottan2019-08-2413-15/+1320
| | | | | | | | | please note as an additional change, macro GLUSTERD_GET_SNAP_DIR moved from glusterd-store.c to glusterd-snapshot-utils.h Change-Id: I811efefc148453fe32e4f0d322e80455447cec71 updates: #663 Signed-off-by: Jiffin Tony Thottan <jthottan@redhat.com>
* posix: In brick_mux brick is crashed while start/stop volume in loopMohit Agrawal2019-08-201-2/+3
| | | | | | | | | | | | | | | Problem: In brick_mux environment sometime brick is crashed while volume stop/start in a loop.Brick is crashed in janitor task at the time of accessing priv.If posix priv is cleaned up before call janitor task then janitor task is crashed. Solution: To avoid the crash in brick_mux environment introduce a new flag janitor_task_stop in posix_private and before send CHILD_DOWN event wait for update the flag by janitor_task_done Change-Id: Id9fa5d183a463b2b682774ab5cb9868357d139a4 fixes: bz#1730409 Signed-off-by: Mohit Agrawal <moagrawal@redhat.com>
* libglusterfs: remove dependency of rpcAmar Tumballi2019-08-162-2/+11
| | | | | | | | | | | | | | | | | | Goal: 'libglusterfs' files shouldn't have any dependency outside of the tree, specially the header files, shouldn't have '#include' from outside the tree. Fixes: * Had to introduce libglusterd so, methods and structures required for only mgmt/glusterd, and cli/ are separated from 'libglusterfs/' * Remove rpc/xdr/gen from build, which was used mainly so dependency for libglusterfs could be properly satisfied. * Move rpcsvc_auth_data to client_t.h, so all dependencies could be handled. Updates: bz#1636297 Change-Id: I0e80243a5a3f4615e6fac6e1b947ad08a9363fce Signed-off-by: Amar Tumballi <amarts@redhat.com>
* glusterd: create separate logdirs for cluster.rc instancesN Balachandran2019-08-1415-130/+180
| | | | | | | | | | Create a separate logdir for each host instance created by cluster.rc. This makes it easier to determine the files belonging to a particular instance. Change-Id: Ic8321f83f98995412b7d5f095b3d3f0391767a8b Fixes: bz#1733042 Signed-off-by: N Balachandran <nbalacha@redhat.com>
* glusterd/shd: Return null proc if process is not running.Mohammed Rafi KC2019-08-054-18/+65
| | | | | | | | | | | | | We were ruturning first proc entry even if it is not running. This was in an assumption that the process could have just started and not updated the pidfile. Now we that we have introduced the states for process state, we can take decision based on that. Change-Id: Ibfc11c966b0db599a8d6a08d8b975233b2bbfb8c Fixes: bz#1728766 Signed-off-by: Mohammed Rafi KC <rkavunga@redhat.com>
* multiple files: reduce minor work under RCU_READ_LOCKYaniv Kaul2019-08-0512-240/+261
| | | | | | | | | 1. Try to unlock faster - in error paths. 2. Remove memory allocations - do them before the lock. Change-Id: I1e9ddd80b99de45ad0f557d62a5f28951dfd54c8 updates: bz#1193929 Signed-off-by: Yaniv Kaul <ykaul@redhat.com>
* event: rename event_XXX with gf_ prefixedXiubo Li2019-07-291-1/+1
| | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | I hit one crash issue when using the libgfapi. In the libgfapi it will call glfs_poller() --> event_dispatch() in file api/src/glfs.c:721, and the event_dispatch() is defined by libgluster locally, the problem is the name of event_dispatch() is the extremly the same with the one from libevent package form the OS. For example, if a executable program Foo, which will also use and link the libevent and the libgfapi at the same time, I can hit the crash, like: kernel: glfs_glfspoll[68486]: segfault at 1c0 ip 00007fef006fd2b8 sp 00007feeeaffce30 error 4 in libevent-2.0.so.5.1.9[7fef006ed000+46000] The link for Foo is: lib_foo_LADD = -levent $(GFAPI_LIBS) It will crash. This is because the glfs_poller() is calling the event_dispatch() from the libevent, not the libglsuter. The gfapi link info : GFAPI_LIBS = -lacl -lgfapi -lglusterfs -lgfrpc -lgfxdr -luuid If I link Foo like: lib_foo_LADD = $(GFAPI_LIBS) -levent It will works well without any problem. And if Foo call one private lib, such as handler_glfs.so, and the handler_glfs.so will link the GFAPI_LIBS directly, while the Foo won't and it will dlopen(handler_glfs.so), then the crash will be hit everytime. The link info will be: foo_LADD = -levent libhandler_glfs_LIBADD = $(GFAPI_LIBS) I can avoid the crash temporarily by linking the GFAPI_LIBS in Foo too like: foo_LADD = $(GFAPI_LIBS) -levent libhandler_glfs_LIBADD = $(GFAPI_LIBS) But this is ugly since the Foo won't use any APIs from the GFAPI_LIBS. And in some cases when the --as-needed link option is added(on many dists it is added as default), then the crash is back again, the above workaround won't work. Fixes: #699 Change-Id: I38f0200b941bd1cff4bf3066fca2fc1f9a5263aa Signed-off-by: Xiubo Li <xiubli@redhat.com>
* glusterd: write voldir once in glusterd-store and don't attempt again.Yaniv Kaul2019-07-291-29/+16
| | | | | | | | | | | | | | | | | | glusterd_store_brickinfos() is calling per each brick the function glusterd_store_brickinfo(). In it, we call: ret = glusterd_store_create_brick_dir(volinfo); However, volinfo is the same for all those bricks - no need to again and again call it (which tries to mkdir that dir). We can do it once above the loops in glusterd_store_brickinfos() While at, combine two similar functions that write additional dirs. Change-Id: I5858cf7783f088ea13a8fa20115118efa816f4cb updates: bz#1193929 Signed-off-by: Yaniv Kaul <ykaul@redhat.com>
* core: use more restrictive mode while creating the directoriesSanju Rakonde2019-07-2310-41/+41
| | | | | | | fixes: bz#1724024 Change-Id: I539fb7248b2cfc037ec29f1413ea648f9ec21ef2 Signed-off-by: Sanju Rakonde <srakonde@redhat.com>
* glusterd: do not mark skip_locking as true for geo-rep operationsSanju Rakonde2019-07-141-2/+7
| | | | | | | | | | | | | | | | | We need to send the commit req to peers in case of geo-rep operations even though it is a no volname operation. In commit phase peers try to set the txn_opinfo which will fail because it is a no volname operation where we don't require a commit phase. We mark skip_locking as true for no volname operations, but we have to give an exception to geo-rep operations, so that they can set txn_opinfo in commit phase. Please refer to detailed RCA at the bug: 1729463 fixes: bz#1729463 Change-Id: I9f2478b12a281f6e052035c0563c40543493a3fc Signed-off-by: Sanju Rakonde <srakonde@redhat.com>
generated by cgit (git 2.34.1) at 2024-06-17 13:12:50 +0000