Whamcloud - gitweb
Elena Gryaznova [Wed, 26 May 2010 06:46:11 +0000 (10:46 +0400)]
b=22841 local.sh mgs mkfs options MGSSIZE fix
i=Dmitry.Zoguine
Elena Gryaznova [Wed, 26 May 2010 06:36:35 +0000 (10:36 +0400)]
b=20918 report max recovery time estimated
i=Andrew.Perepechko
Nicolas Williams [Tue, 25 May 2010 23:51:58 +0000 (18:51 -0500)]
b22131 ASSERTION(request->rq_repdata == NULL) failed
rq_repdata needs to be reset on free.
i=eric.mei@oracle.com
Alexander.Zarochentsev [Tue, 25 May 2010 07:22:10 +0000 (11:22 +0400)]
b=22598 Disable COS by default
i=robert.read
Mikhail Pershin [Tue, 25 May 2010 05:00:28 +0000 (09:00 +0400)]
b=22830 correct interop warning message
i=zam
Dmitry Zogin [Mon, 24 May 2010 18:38:53 +0000 (14:38 -0400)]
b=17382 obdfilter-survey gives unreasonably high numbers
Wait for all threads to complete when running test_brw.
i=andreas.dilger
i=oleg.drokin
Vitaly Fertman [Mon, 24 May 2010 19:23:45 +0000 (23:23 +0400)]
b=19325 adjust waiting extent locks during 1st enqueue
o=bobijam
i=vitaly
i=green
Re-landing. Adjust locks' extents on their first enqueue, so that at the time
they get granted, there is no need for another pass through the queues since
they are already shaped into the proper forms.
Rahul Deshmukh [Wed, 26 May 2010 08:48:17 +0000 (14:18 +0530)]
b=21815 per-nid stats should not access lustre hash internal structures directly
Fixed Yang Sheng's patch as per Robert's comment.
i=rread
Fan Yong [Tue, 25 May 2010 02:08:25 +0000 (10:08 +0800)]
b=21636 drop noisy debug messages
Drop noisy debug messages
i=robert.read
i=rahul.deshmukh
Fan Yong [Tue, 25 May 2010 02:01:39 +0000 (10:01 +0800)]
b=22342 invalidate dentry can not be counted as the same
Invalidate dentry can not be counted as the same.
i=robert.read
i=eric.mei
Robert Read [Thu, 20 May 2010 04:35:36 +0000 (21:35 -0700)]
b=22040 Don't run connectathon lock tests on nfsv4
We won't support flock on nfsv4 in 2.x until 14080 has been fixed.
i=green
Brian J. Murrell [Tue, 25 May 2010 21:06:43 +0000 (17:06 -0400)]
b=17258 remove resurrected files
build/{lmake,lustre-kernel-2.4.spec.in} died with commit
9aa2968b965a5acdf127a5635dfde8950bc8d3a2 but a cherry-pick seems to
have resurrected them. ~sigh~ Bad cherry-pick, bad cherry-pick.
Let's try a wooden stake this time.
Brian J. Murrell [Tue, 25 May 2010 17:48:13 +0000 (13:48 -0400)]
b=22847 separate format and content in ext[34]_warning()
ext{3,4}_warning should not try to overload the message into the format
but should instead, pass a "%s" as the format and the string as an argument
for the %s.
i=panda
i=whitebear
Brian J. Murrell [Tue, 25 May 2010 17:48:12 +0000 (13:48 -0400)]
b=22642 ldiskfs to figure out ext3/4 base itself
Ldiskfs should figure out whether to base itself on ext3 or ext4
by itself and not rely on lustre's configure to tell it.
i=mjmac
i=johann
Brian J. Murrell [Tue, 25 May 2010 17:48:11 +0000 (13:48 -0400)]
b=22787 update to ofed to 1.5.1
For O/Ses where we don't use the vendor supplied OFED, update the built
OFED to 1.5.1.
i=johann
Brian J. Murrell [Tue, 25 May 2010 17:48:10 +0000 (13:48 -0400)]
b=17258 give the BUILD_TESTS love to ldiskfs as well
Because ldiskfs re-uses so (too?) much of the lustre auto* goop we need
to stub the BUILD_TESTS assignment into it's autoMakefile.am, even though
it's completely unused/unneed there.
i=wangyb
i=yangsheng
Brian J. Murrell [Tue, 25 May 2010 17:48:09 +0000 (13:48 -0400)]
b=17258 fix error with make rpms after configure --disable-tests
If one configures lustre with "--disable-tests" a subsequent "make rpms"
will fail as it would still try to package up the lustre-tests RPM.
Fixing this provided the opportunity to fix another wart, that being the
subst'ing the configure arguments into the lustre.spec. Now they are
passed as value with "--define 'configure_args ...'" when calling rpmbuild.
i=sheng.yang
i=yibin.wang
Brian J. Murrell [Fri, 21 May 2010 15:06:45 +0000 (11:06 -0400)]
b=20619 be flexible in version format
The version format for the OFED release needs to be more flexible than
the previous M.m format that was required. Now it can be any dot-separated
format. i.e. 1 or 1.4 or 1.5.1 or 1.5.1.2, etc. Of course, it has to be a
valid and unique OFED version (so the "1" example would be a valid version
string but invalid because there is no OFED version "1". Ditto for 1.5.1.2
at the time of writing this comment.).
i=michael.macdonald
i=minh.diep
Vladimir Saveliev [Fri, 21 May 2010 11:54:03 +0000 (15:54 +0400)]
b=21109 mds_lov_read_objids cleanup and conf-sanity tests
calculate mds->mds_lov_objid_lastidx and mds->mds_lov_objid_lastpage correctly
have index strings big enough to store indicies in decimal
test to check size of lov_objid and
test to check configuraion with big indicies
i=adilger
i=panda
i=nathan
Rahul Deshmukh [Fri, 21 May 2010 06:20:30 +0000 (11:50 +0530)]
b=21951 2.6.32-fc13 patchless client support for HEAD
b=21951 2.6.32-fc13 patchless client support for HEAD
i=sheng.yang
i=sebastien.buisson
Improvement changes for proc_handler, head port
Dmitry Zogin [Fri, 21 May 2010 03:35:32 +0000 (23:35 -0400)]
b=17086 LSI Fusion MPT driver hacks to improve performance
patch to set CONFIG_FUSION_MAX_SGE=256 for Rhel5
i=johann
Nathan Rutman [Thu, 20 May 2010 19:08:50 +0000 (12:08 -0700)]
b=21900 verify that we can write to the OST
i=dmitry.zoguine
a=nathan
Vitaly Fertman [Thu, 20 May 2010 15:45:43 +0000 (19:45 +0400)]
b=22244 delegate lock cancel to blocking thread
i=adilger
i=green
instead of cancelling locks locally in the shrinking thread,
deletate it to a separate blocking thread.
Vitaly Fertman [Thu, 20 May 2010 15:45:42 +0000 (19:45 +0400)]
b=22244 ldlm cancel flags cleanup
i=adilger
i=green
cleanup of cancel flags passed to ldlm lock cancel code
Fan Yong [Thu, 20 May 2010 15:18:32 +0000 (23:18 +0800)]
b=22560 fix "obd_connect_names[]" for "OBD_CONNECT_FULL20" introduced
fix "obd_connect_names[]" for "OBD_CONNECT_FULL20" introduced
i=robert.read
i=andreas.dilger
Mikhail Pershin [Thu, 20 May 2010 15:03:05 +0000 (19:03 +0400)]
b=21485 allocate lcd inside obd_init_export()
i=rread
i=zam
Fan Yong [Thu, 20 May 2010 05:54:12 +0000 (13:54 +0800)]
b=22637 MDS returns OBD_MD_FLSIZE to client only when no OSS object allocated
MDS returns OBD_MD_FLSIZE to client only when no OSS object allocated.
i=robert.read
i=andreas.dilger
Alexander.Zarochentsev [Thu, 20 May 2010 07:33:48 +0000 (11:33 +0400)]
b=22827 restore locking for thread->t_flags
i=tappro
i=robert.read
Robert Read [Thu, 13 May 2010 23:59:42 +0000 (16:59 -0700)]
b=20938 Add liblustreapi.a to dependencies lists
I think this is why we occasionally file to build lustre on LBATS
with a missing liblustreapi.a.
i=brian
Eric Mei [Tue, 18 May 2010 14:59:35 +0000 (08:59 -0600)]
b=22458 fix concurrent mgs lock revocation.
r=nathan
r=rread
Mikhail Pershin [Tue, 18 May 2010 05:52:31 +0000 (09:52 +0400)]
b=15587 ignore security.capability xattr on client side
i=adilger
i=johann
Mikhail Pershin [Tue, 18 May 2010 05:43:29 +0000 (09:43 +0400)]
b=21681 Quiet bogus previously committed transno error
i=zhang,panda
LiuYing [Tue, 18 May 2010 02:48:24 +0000 (10:48 +0800)]
b=22455 remove "lnet." prefix from lctl params display
remove "lnet." prefix from lctl params display and change one
"memused" to "lnet_memused".
o=adilger
i=johann
i=emoly.liu
i=rread
Fan Yong [Tue, 18 May 2010 02:35:09 +0000 (10:35 +0800)]
b=22731 server should not fall into LBUG if client send invalid parameter
server should not fall into LBUG if client send invalid parameter
i=robert.read
i=di.wang
Fan Yong [Tue, 18 May 2010 02:24:44 +0000 (10:24 +0800)]
b=22560 introduce "OBD_CONNECT_FULL20" to distinguish 1.8 client from 2.0 one for different checksum policy
Introduce "OBD_CONNECT_FULL20" to distinguish 1.8 client from 2.0 one for different checksum policy:
1) for 1.8 client, use fixed first 88 bytes of ptlrpc_body
2) for 2.0 client, use lm_buflens
i=andreas.dilger
i=robert.read
Nathan Rutman [Fri, 14 May 2010 16:18:43 +0000 (09:18 -0700)]
b=15253 add conf_param -d to remove permanent settings
i=adilger
i=rread
Rahul Deshmukh [Fri, 14 May 2010 10:29:39 +0000 (15:59 +0530)]
b=22625 Fix libcfs_debug_file_path module option
i=rahul
Landing patch by Brian Behlendorf <behlendorf1@llnl.gov>,
fix libcfs_debug_file_path module option
Maxim Patlasov [Fri, 14 May 2010 09:11:39 +0000 (13:11 +0400)]
b=21945 Adding WIRE_ATTR to lnet structures traversing network
i=liang
i=isaac
LST passed some lnet structures via network neglecting the lack of WIRE_ATTR attribute. This resulted in incompatibility of LST-s running on different platforms.
LiuYing [Fri, 14 May 2010 00:47:07 +0000 (08:47 +0800)]
b=22455 add "list_param -R"
list parameters recursively with the "-R" option
o=adilger
i=emoly.liu
i=nathan
Terry Rutledge [Thu, 13 May 2010 20:30:07 +0000 (13:30 -0700)]
Updated for build 42.
Robert Read [Wed, 12 May 2010 19:07:34 +0000 (12:07 -0700)]
Revert "b=22637 MDS returns OBD_MD_FLSIZE to client only when no OSS object allocated"
Hit ASSERTION(attr->la_blocks == 0), see bug 22802.
This reverts commit
33b4bafea13bd2cfe90dba3a8651a175683f3999.
Vitaly Fertman [Wed, 12 May 2010 11:04:16 +0000 (15:04 +0400)]
b=22573 do not skip previously granted locks in osc_lock_enqueue_wait().
i=eric.mei
i=wangdi
as CLIO adds new locks to the tail, walk through the head of the queue
to cancel overlapping conflicting locks on enqueue.
Mikhail Pershin [Wed, 12 May 2010 16:25:20 +0000 (20:25 +0400)]
b=22518 mount client2 at the start of tests, disable COS, fix test 10
i=grev
pravin [Wed, 12 May 2010 14:25:24 +0000 (19:55 +0530)]
b=18857 enhance seq allocation scalability by updating seq data asynchronously.
this patch also removes seq replay. ref bug for details.
a=pravin,tappro
i=tappro
i=alexander.zarochentsev
i=pravin
Alexander.Zarochentsev [Wed, 12 May 2010 10:29:19 +0000 (14:29 +0400)]
b=21140 Fix srv_threads_running counting.
It was possible to overload n_active_request processing incoming requests and
break the thread reservation logic. Likely, it was responsible to the long
processing of requests.
The patch makes srv_threads_running to exactly count only running not sleeping
threads. All threads accounting and comparing/reservation of threads are done
under the service spinlock so it produce a reliable result. The thread
reservation logic is based on new srv_threads_running value and cannot be
confused by not active sleeping threads. The thread reservation logic is
concentrated now in one place, where the wakeup condition is checked (now in
ptlrpc_main_check_event), once a thread is woken up, it is counted as running
and does further work w/o additional checks.
i=zhen.liang
i=robert.read
Wang Di [Tue, 11 May 2010 17:58:48 +0000 (13:58 -0400)]
b=22683 remove unnecessary check and assert in the cfs_hash function.
o=Eric.Barton
i=Robert.Read
i=Di.Wang
Vladimir Saveliev [Tue, 11 May 2010 06:40:46 +0000 (10:40 +0400)]
b=13698 allow e2fsck part of lfsck.sh to be run without lfsck
this combines initial patch from Andreas (https://bugzilla.lustre.org/attachment.cgi?id=29696)
and several necessary fixes (https://bugzilla.lustre.org/attachment.cgi?id=29747)
i=adilger
Rahul Deshmukh [Tue, 11 May 2010 06:25:30 +0000 (11:55 +0530)]
b=20562 increase LUSTRE_SEQ_META_WIDTH to keep FLD compact
Fixed the build error for fc11 and fc12 patchless client.
i=rread
Fan Yong [Tue, 11 May 2010 03:01:45 +0000 (11:01 +0800)]
b=22598 diagnostic patch for lock cancel callback error processing
diagnostic patch for lock cancel callback error processing.
i=robert
i=di.wang
Fan Yong [Tue, 11 May 2010 01:55:54 +0000 (09:55 +0800)]
b=19986 cleanup lock to eliminate former test cases effect before replay-single test_53
cleanup lock to eliminate former test cases effect before replay-single test_53
i=robert
i=di.wang
Mikhail Pershin [Sun, 9 May 2010 09:09:23 +0000 (13:09 +0400)]
b=18143 Make VBR compatible with pdirops.
i=zam
i=bzzz
Nathan Rutman [Fri, 7 May 2010 22:33:29 +0000 (15:33 -0700)]
b=22283 clarify writeconf in man page
Nathan Rutman [Fri, 7 May 2010 19:29:07 +0000 (12:29 -0700)]
b=22671 Check for modules directly instead of keeping state
i=nico
i=rread
Elena Gryaznova [Fri, 7 May 2010 16:07:54 +0000 (20:07 +0400)]
b=22581 LOADS env var in ncli.sh should allow overwrite
i=Minh.Diep
Oleg Drokin [Wed, 5 May 2010 23:35:58 +0000 (19:35 -0400)]
b=22522 do not remove from res_list without locks
Patch in bug 21501 moved list manipulation of res_list outside or res_lock
introducing a race window in flock code. Move it back under the lock.
i=rread
i=adilger
Eric Mei [Wed, 5 May 2010 01:14:34 +0000 (19:14 -0600)]
b=22669 fix fault page index handler in newer kernel.
r=wangdi
r=rread
Rahul Deshmukh [Tue, 4 May 2010 15:16:01 +0000 (20:46 +0530)]
b=21502 symlink compatibility between 1.6 and 2.0
Fixed the sleep in spin lock, inode->i_sb->s_op-> dirty_inode(inode)
was called with spin lock held.
i=bzzz
i=pravin
Eric Mei [Tue, 4 May 2010 14:30:09 +0000 (08:30 -0600)]
b=22683 don't manipulate hash in lov_sub_enter/lov_sub_exit.
r=wangdi
r=rread
Robert Read [Fri, 30 Apr 2010 16:23:28 +0000 (09:23 -0700)]
Revert "b=19427 correct lmm_object_id and reserve fids for fid-on-OST."
This reverts commit
4c01e64e0a72c1682ebf0a8bd4cccf99fd04cd88.
This caused the interop issue seen in bug 22730.
Robert Read [Thu, 29 Apr 2010 22:09:34 +0000 (15:09 -0700)]
Prepare for Build 41
Fan Yong [Thu, 29 Apr 2010 06:06:37 +0000 (14:06 +0800)]
b=16680 remove some noisy debug messages
Remove some noisy debug messages.
i=robert.read
i=rahul.deshmukh
Jian Yu [Thu, 29 Apr 2010 05:59:13 +0000 (13:59 +0800)]
b=20326 Test suite for MMP feature
Tests for multiple mount protection (MMP) feature.
i=andreas.dilger
i=grev
Fan Yong [Thu, 29 Apr 2010 05:37:03 +0000 (13:37 +0800)]
b=22069 port "llapi_get_connect_flags()" API from b1_8 to master
Port "llapi_get_connect_flags()" API from b1_8 to master.
i=robert.read
i=landen
Manoj Joseph [Thu, 29 Apr 2010 03:27:21 +0000 (21:27 -0600)]
b=22075 buffalo-v2 should detect test timeouts
buffalo-v2 now detects test timeouts. It now generates status entries
in results.yml after a sub-test completes. If the test status is missing
a timeout is assumed to have occured.
i=robert.read
i=grev
Manoj Joseph [Thu, 29 Apr 2010 03:27:20 +0000 (21:27 -0600)]
b=21962 Quote the error message in results.yaml
Quote and escape the error message in results.yaml
i=robert.read
i=nicolas.williams
Nathan Rutman [Wed, 28 Apr 2010 18:04:59 +0000 (11:04 -0700)]
b=22582 remove leading / from fid2path results print full path if given mountpoint
i=manoj
i=rread
Robert Read [Wed, 28 Apr 2010 17:06:59 +0000 (10:06 -0700)]
b=22456 Remove files for unsupported kernels
Removes support for fc3, fc5, rhel4, sles10, 2.6.18 vanilla and 2.6.22 vanilla.
i=adilger
Elena Gryaznova [Tue, 27 Apr 2010 15:32:56 +0000 (19:32 +0400)]
b=18649 set wait_recovery_complete() MAX value to max recovery time estimated
i=Mikhail.Pershin
Elena Gryaznova [Tue, 27 Apr 2010 15:23:04 +0000 (19:23 +0400)]
b=20918 t-f max recovery time estimation
i=Nathan.Rutman
i=Brian.Murrell
inform user about the estimated maximum recovery time value
to help him to set the server FAILOVER period properly
Fan Yong [Tue, 27 Apr 2010 05:23:03 +0000 (13:23 +0800)]
b=22342 process racer condition between statahead and rename/unlink operation
1) process racer condition between statahead and rename/unlink operation
2) replace "lli_lock" with "lli_sa_lock" for statahead related processing
i=robert.read
i=tom.wang
Fan Yong [Tue, 27 Apr 2010 03:36:53 +0000 (11:36 +0800)]
b=22634 hold "mds_qonoff_sem" when call "lustre_read_quota()", and check parameter properly in such function
1) replace "cfs_semaphore_t" with "cfs_rw_semaphore_t" for "mds_qonoff_sem" to enhance the parallel processing of quota related operations
2) hold "mds_qonoff_sem" when call "lustre_read_quota()", and check parameter properly in such function
i=robert.read
i=landen
Fan Yong [Tue, 27 Apr 2010 02:51:27 +0000 (10:51 +0800)]
b=22614 enlarge MDSSIZE/OSTSIZE to increase default journal size for conf-sanity
1) enlarge MDSSIZE/OSTSIZE to increase default journal size for conf-sanity
2) journal handler error process in lustre_commit_dquot
i=robert.read
i=landen
Li Wei [Tue, 27 Apr 2010 02:28:09 +0000 (10:28 +0800)]
b=21251 ha.sh: Fix ha_wait_loads and ha_dump_logs
Report existence of hanging workloads. Ignore "lctl dk" failures, since
some nodes may be down.
i=robert.read
Li Wei [Tue, 27 Apr 2010 02:28:08 +0000 (10:28 +0800)]
b=21251 Add lustre/tests/ha.sh
This is a simple failover test script that works with configurations
controlled by a CRM and have multiple targets per server.
i=robert.read
i=grev
Robert Read [Mon, 26 Apr 2010 22:11:52 +0000 (15:11 -0700)]
Revert "b=21379 Fix orphans proceeding in osc_create"
This reverts commit
2deb4f149f4601f9128fd39efd4705573520f277.
Eric Mei [Mon, 26 Apr 2010 14:59:52 +0000 (08:59 -0600)]
b=22458 move lcw_dump out of softirq context.
Now the message dump is done in thread context in lc_watchdogd.
r=rread
r=nathan
pravin [Mon, 26 Apr 2010 13:15:26 +0000 (18:45 +0530)]
b=21128 run sync ldlm_bl_to_thread_list() in separate thread to save stack space.
i=oleg.drokin
i=rahul
Fan Yong [Mon, 26 Apr 2010 07:38:37 +0000 (15:38 +0800)]
b=22637 MDS returns OBD_MD_FLSIZE to client only when no OSS object allocated
i=robert.read
i=landen
Wang Di [Mon, 26 Apr 2010 03:57:44 +0000 (23:57 -0400)]
b=22513 Remove unecessary lock in read-ahead process.
i=Robert.Read
i=Eric.Mei
hongchao.zhang [Wed, 21 Apr 2010 00:54:53 +0000 (08:54 +0800)]
b=21938 use the same set during replay
some requests use its own ptlrpc_request_set to process its requests, but Lustre
will use a specific ptlrpc_request_set to process the requests during recovery.
this patch fixes this problem to allow the requests to use its own set if it have
one
i=johann@sun.com
i=tappro@sun.com
Mikhail Pershin [Sun, 25 Apr 2010 12:09:48 +0000 (16:09 +0400)]
b=15936 Unified target cleanups v2
i=rread
i=andreas
Cliff White [Tue, 6 Apr 2010 06:40:15 +0000 (23:40 -0700)]
b=20373 Putting parent lock for rep-ack on create is wasteful
Do not put locks if no create.
i=robert.read
i=tappro
Eric Mei [Fri, 23 Apr 2010 18:48:07 +0000 (12:48 -0600)]
b=22310 add a little more comment.
r=adilger
Isaac Huang [Fri, 23 Apr 2010 04:03:14 +0000 (00:03 -0400)]
b=21678 Add more debug info to lnd_query code path
The peer health code lacked some important debugging info in lnd_query
code paths. This patch added necessary debug prints, not just for bug
21678, but also for future troubleshooting.
i=liang
i=maxim
Wang Di [Fri, 23 Apr 2010 19:57:04 +0000 (12:57 -0700)]
b=19427 correct lmm_object_id and reserve fids for fid-on-OST.
1. Change lmm_object_id to fid.
2. Cleanup fid spaces reservation (for fid-on-OST).http://arch.lustre.org/index.php?title=Interoperability_fids_zfs#NEW.0
3. Rename group to Seq.
i= Andreas.diger
i= pravin.shelar
Vladimir Saveliev [Thu, 22 Apr 2010 19:15:32 +0000 (12:15 -0700)]
b=22615 fixes for regressions caused by 11063
set atime to past under PW EOF extent lock
fix truncate in liblustre
i=vitaly
i=ericm
Manoj Joseph [Thu, 22 Apr 2010 19:15:28 +0000 (12:15 -0700)]
b=22507 rm -rf not replicated
Support replication of recursive directory removal.
i=nathan.rutman
i=robert.read
Wang Di [Thu, 22 Apr 2010 19:15:24 +0000 (12:15 -0700)]
b=22520 set the thread to be uninterrupt before add to waitq
In lov_subobject_kill, if the thread needs to wait the object being
freed, it should set the thread to be uninterrupt, otherwise, the thread
might spin there.
i=Eric.mei
i=Robert
Wang Di [Thu, 22 Apr 2010 19:15:20 +0000 (12:15 -0700)]
b=22296 Fix script problem for recovery-double-scale
Force the test threads stopped before shutdown the clients in
recovery-double-scale.
i=Jack.Chen
i=WangDi
Mikhail Pershin [Wed, 21 Apr 2010 18:43:20 +0000 (11:43 -0700)]
b=22161 Use LCK_PW for parent lock in mdt_link(). Pdirops test set
i=adilger
i=bzzz
i=rread
yangsheng [Wed, 21 Apr 2010 18:43:12 +0000 (11:43 -0700)]
b=19919 Supply a absolute path.
i=andreas
i=johann
Mikhail Pershin [Wed, 21 Apr 2010 18:43:11 +0000 (11:43 -0700)]
b=22190 Apply 19195 patch to add tls data for recovery thread.
This will be needed anyway when sync journal will be ported
i=zam
i=oleg
Rahul Deshmukh [Wed, 21 Apr 2010 18:42:26 +0000 (11:42 -0700)]
fixed for bug 22237
b=22237 replay-single test-13 mmp failure, BUG: warning at
fs/proc/generic.c:764/remove_proc_entry()
The proc entry EXT4_MAX_DIR_SIZE_NAME was not remove in cleanup path. It
is now fixed.
i=johann
Wang Di [Wed, 21 Apr 2010 18:42:24 +0000 (11:42 -0700)]
b=22129 add debug patch for 22129
i=Robert
i=Fangyong
Wang Di [Wed, 21 Apr 2010 18:42:20 +0000 (11:42 -0700)]
b=21379 Fix orphans proceeding in osc_create
in "osc_create", there is no need to proceed after the orphans are deleted, which cause the
oscc->oscc_next_id is increased by 1 and break the balance between it and the last objid stored
in "mds->mds_lov_page_array"
o=Zhang Hongchao
i=Johann
I=Tappro
Dmitry Zogin [Mon, 19 Apr 2010 23:29:32 +0000 (16:29 -0700)]
b=22409 Spurious error messages from smp_processor_id() on preemptible kernel
Disable a preemption by grabbing the lock in fs_trace_get_tcd() first.
The function fs_trace_get_tcd() was moved up.
o=andreas.dilger
i=johann
i=dmitry.zogin
i=nathan.rutman
Andrew Perepechko [Mon, 19 Apr 2010 23:29:30 +0000 (16:29 -0700)]
b=20953 sanity-quota test 30 fixes
i=Yong Fan
Eric Mei [Mon, 19 Apr 2010 23:29:27 +0000 (16:29 -0700)]
b=22310 temporary fix: align readahead window end to 1M rpc boundary.
r=wangdi
r=rread
Eric Mei [Mon, 19 Apr 2010 23:29:25 +0000 (16:29 -0700)]
b=22310 do not break early in osc_io_submit() which might cause sub-optimal rpc size for read and synchronous write.
i=wangdi
i=rread
Vitaly Fertman [Mon, 19 Apr 2010 23:29:21 +0000 (16:29 -0700)]
b=21949 echo code fixes.
i=eric.may
i=bzzz
1. access to a wrong page on echo server commit;
2. wrong lock type enqueue in echo client test_brw code;
3. wrong thread offset initialization in lctl test_brw code;