Whamcloud - gitweb
fs/lustre-release.git
14 years ago- corrected a trivial comment typo.
isaac [Wed, 25 Nov 2009 20:17:31 +0000 (20:17 +0000)]
- corrected a trivial comment typo.

14 years agoBranch HEAD
robert.read [Wed, 25 Nov 2009 17:35:00 +0000 (17:35 +0000)]
Branch HEAD

Update version number.

14 years agob=16818
zhanghc [Wed, 25 Nov 2009 10:08:32 +0000 (10:08 +0000)]
b=16818

1, the test will be skipped if the MGS and OST1 are on the same node
2, the test will be interrputed if "zconf_mount $MOUNT2" failed

i=nathan.rutman@sun.com
i=robert.read@sun.com
i=grev@sun.com

14 years agoBranch HEAD
zam [Wed, 25 Nov 2009 09:11:33 +0000 (09:11 +0000)]
Branch HEAD
b=20816
i=robert.read
i=tappro
i=johann

improve simulation of late reply: ignore obd_fail_timeout for ping replies.

14 years agoBranch HEAD
robert.read [Wed, 25 Nov 2009 03:10:51 +0000 (03:10 +0000)]
Branch HEAD
b=19742

The FIEMAP tests need to be updated, so there's seems to be no point in running them right now.

14 years agoBranch HEAD
tianzy [Wed, 25 Nov 2009 03:03:00 +0000 (03:03 +0000)]
Branch HEAD
fix some problem in test_23 of ost_pools.sh
b=20696
i=manoj.joseph
i=johann

14 years agob=19906
jxiong [Wed, 25 Nov 2009 02:20:28 +0000 (02:20 +0000)]
b=19906

cancel wait lock - small fixes.

14 years agob=19906
jxiong [Wed, 25 Nov 2009 02:10:39 +0000 (02:10 +0000)]
b=19906
r=wangdi,ericm

Revise osc_lock_cancel_wait. In clio, we tend to cancel the conflicting locks
at the same client before enqueuing a new lock. In the old implementation, it
waits at osc layer with parent lock held, this is not good - so I moved the
wait to lov layer.

14 years agob=21178
jxiong [Wed, 25 Nov 2009 01:27:52 +0000 (01:27 +0000)]
b=21178
r=wangdi,ericm

Fix a race condition in cl_sync_io_note.

14 years agob=19928
jxiong [Wed, 25 Nov 2009 01:05:36 +0000 (01:05 +0000)]
b=19928
r=wangdi,ericm

A workaround fix for 19928 where we found that sometimes MDS takes much long time to finish setattr, which then causes clients can't take inode mutex on time after holding extent lock.

I just added ->lli_trunc_sem to make the above process atomic.

14 years agoBranch HEAD
adilger [Tue, 24 Nov 2009 17:48:38 +0000 (17:48 +0000)]
Branch HEAD
Fix whitespace (tabs, trailing spaces).

14 years agoBranch HEAD
robert.read [Tue, 24 Nov 2009 00:22:12 +0000 (00:22 +0000)]
Branch HEAD
b=21240,19242

Disabling replay-single 44a, which is using obd_race, which appears to be racy itself, and
also appear to be incorrectly checking the return code of lclt recover, which appears be supposed
to fail in this test. Also disabling 44b which looks like it has the same issue with lctl recover.

14 years agodon't fail import after eviction
shadow [Thu, 19 Nov 2009 14:18:49 +0000 (14:18 +0000)]
don't fail import after eviction

Branch HEAD
b=20765
i=tappro
i=zam

14 years agoavoid unnecessary allocation.
shadow [Thu, 19 Nov 2009 12:05:53 +0000 (12:05 +0000)]
avoid unnecessary allocation.

Branch HEAD
b=19557
i=vitaly

14 years agob=21150
panda [Thu, 19 Nov 2009 11:28:16 +0000 (11:28 +0000)]
b=21150
i=Johann Lombardi
i=ZhiYong Tian

avoid mds-lov cleanup during dqacq

14 years agob=21064
fanyong [Thu, 19 Nov 2009 08:39:57 +0000 (08:39 +0000)]
b=21064
i=robert.read
i=adilger

1) Break reqeust process when service to be stopped.
2) Recovery the process for "ptlrpc_stop_thread()" as suggestion.

14 years ago*** empty log message ***
tappro [Thu, 19 Nov 2009 07:08:38 +0000 (07:08 +0000)]
*** empty log message ***

14 years agob=19906
jxiong [Thu, 19 Nov 2009 05:58:04 +0000 (05:58 +0000)]
b=19906
r=wangdi,ericm

Revert a bit of cleanup patch.

14 years agob=20354
zhanghc [Thu, 19 Nov 2009 04:08:03 +0000 (04:08 +0000)]
b=20354

the client_uuid(lcd_uuid) stored in last_rcvd on server(MDT or OST)
could be corrupted, so it is needed to check during loading from disk

i=johann@sun.com
i=tappro@sun.com

14 years agob=19906
jxiong [Thu, 19 Nov 2009 03:39:07 +0000 (03:39 +0000)]
b=19906
r=wangdi,ericm

Make racer work at the client side.

14 years agob=19906
jxiong [Thu, 19 Nov 2009 03:06:29 +0000 (03:06 +0000)]
b=19906
r=wangdi,ericm

IO segfault fix

14 years agob=19906
jxiong [Thu, 19 Nov 2009 03:01:03 +0000 (03:01 +0000)]
b=19906
r=wangdi,ericm

cleanup patch, add more debug info.

14 years agoBack out landing that was made in error of flags being mis-set.
brian [Wed, 18 Nov 2009 21:12:21 +0000 (21:12 +0000)]
Back out landing that was made in error of flags being mis-set.

14 years agob=20617
brian [Wed, 18 Nov 2009 14:44:01 +0000 (14:44 +0000)]
b=20617

One hunk missed in landing.  Result is benign though as both constructs
result in the same value.

14 years agob=20617
brian [Wed, 18 Nov 2009 13:36:13 +0000 (13:36 +0000)]
b=20617
i=wangyb
i=yangsheng

Build the RDAC and mptlinux drivers with the Lustre kernel in lbuild.

14 years agob=21188
pravins [Tue, 17 Nov 2009 21:59:20 +0000 (21:59 +0000)]
b=21188
i=johann
i=girish
fix typo.

14 years agob=20619
brian [Fri, 13 Nov 2009 22:02:33 +0000 (22:02 +0000)]
b=20619

Fix typo.

14 years agoUpdated content to point to valid wiki page.
stinson1947 [Fri, 13 Nov 2009 19:58:17 +0000 (19:58 +0000)]
Updated content to point to valid wiki page.

14 years agobuild fixes for posix 2008 enviroment
shadow [Fri, 13 Nov 2009 10:05:58 +0000 (10:05 +0000)]
build fixes for posix 2008 enviroment

Branch HEAD
b=19808
i=tappro
i=lee@sandia

14 years agoBranch b1_8
yangsheng [Fri, 13 Nov 2009 06:06:59 +0000 (06:06 +0000)]
Branch b1_8
b=20773

Fork the patch ext4-hash-indexed-dir-dotdot-update.patch for rhel5.4.

14 years agoBranch HEAD
robert.read [Fri, 13 Nov 2009 00:35:50 +0000 (00:35 +0000)]
Branch HEAD
b=21188

This is needed temporarily because ext3_data_in_dirent.patch has been disabled for RHEL5 in ldiskfs.

14 years agob=19721
brian [Thu, 12 Nov 2009 21:33:52 +0000 (21:33 +0000)]
b=19721
i=brian
o=adilger

A "debs" make target so that one simply has to do "make debs" in a configured
source tree to get both userspace and kernel modules debian packages.
- this includes automatically updating the debian/changelog if not up-to-date
  with respect to the current version of lustre
- I'm not convinced that all of the gyrations to use m-a to build the kernel
  modules is worth it and if there are any non-trivial problems found with it
  I might just rip it all out and have debian/rules build and package up the
  kernel modules package
Some cleanups:
- fix automake requirement so that it's much more flexible
- allow both Ubuntu and Debian headers packages to be required
- comment out the body of the autogen-stamp
  + if somebody adds a patch to debian/patches that requires an
    autogen, they can supply a patch to create the autogen.sh and
    uncommment the autogen-stamp body
    * this is ideally and hopefull a very rare event
    * we would love feedback from packages that find creating patches necessary
- workaround in version_tag.pl for some (Debian?) systems where, for whatever
  reason, MODULES_{TRUE|FALSE} does not get included in the {auto,}Makefile.in
  files

14 years agoBranch b1_8
johann [Thu, 12 Nov 2009 21:13:39 +0000 (21:13 +0000)]
Branch b1_8
b=21188

Remove ext3/4_data_in_dirent patch since it causes spurious assertion
failure in ldiskfs.

14 years agob=20619
brian [Thu, 12 Nov 2009 20:22:43 +0000 (20:22 +0000)]
b=20619
i=wangyb
i=yangsheng

Allows a command-line argument to override the OFED version being built.
A "daily" build can be used by supplying "--ofed-version daily-M.m-YYYYMMDD"
(where M.m is major.minor, such as 1.5) to lbuild.

This allows very easy buiding with OFED daily snapshots to streamline testing
in advance of OFA's betas, RCs and GAs.

14 years agob=21306
nathan [Thu, 12 Nov 2009 19:55:27 +0000 (19:55 +0000)]
b=21306
i=manoj
insure pool additions propagate before striping

14 years agoBranch HEAD
yangsheng [Thu, 12 Nov 2009 17:31:19 +0000 (17:31 +0000)]
Branch HEAD
b=20773

i=adilger, shadow, girish, brian, wangyb

Update kernel to RHEL5 U4.

14 years agoBranch b1_8
yangsheng [Thu, 12 Nov 2009 17:14:18 +0000 (17:14 +0000)]
Branch b1_8
b=20773

i=adilger, shadow, girish, brian, wangyb

Update kernel to RHEL5 U4 2.6.18-164.6.1.el5.

14 years agob=21108
deshmukh [Thu, 12 Nov 2009 08:20:06 +0000 (08:20 +0000)]
b=21108
i=rread

Increasing the timeout for df pdsh to complete.

14 years agoBranch HEAD
liuy [Thu, 12 Nov 2009 00:59:53 +0000 (00:59 +0000)]
Branch HEAD
b=19518
o=liangzhen
i=adilger
i=bobijam

first step change to move lustre hashes to libcfs:
1. no access to lh_rwlock if LH_REHASH is not set
2. buckets are allocated one by one, to avoid unnecessary cacheline
   conflict (on bucket::lhb_rwlock) while searching.
3. OBD_ALLOC/FREE are replaced by LIBCFS_ALLOC/FREE

14 years agoa=Johann Lombardi
panda [Wed, 11 Nov 2009 23:52:05 +0000 (23:52 +0000)]
a=Johann Lombardi
i=Andrew Perepechko
i=ZhiYong Tian

get lov objid count from MDS

14 years ago Branch HEAD
dzogin [Wed, 11 Nov 2009 22:33:30 +0000 (22:33 +0000)]
 Branch HEAD
 b=21049
 i=adilger
 i=robert.read
 ----------------------------------------------------------------------
 Modified Files:
  lustre/ChangeLog lustre/obdfilter/filter.c
  lustre/ost/ost_handler.c lustre/tests/sanity.sh
  lustre/tests/test-framework.sh
 ----------------------------------------------------------------------

14 years agob=21304
nathan [Wed, 11 Nov 2009 21:51:36 +0000 (21:51 +0000)]
b=21304
i=manoj
compare full list of uuids

14 years agoBranch HEAD
nicolas.williams [Wed, 11 Nov 2009 21:23:25 +0000 (21:23 +0000)]
Branch HEAD
b=21243
i=robert.read@sun.com
i=adilger@sun.com

14 years ago* Removed unused struct member ccc_page::cpg_checksum
eeb [Tue, 10 Nov 2009 21:55:34 +0000 (21:55 +0000)]
* Removed unused struct member ccc_page::cpg_checksum

14 years agoBranch b1_8
johann [Tue, 10 Nov 2009 19:57:24 +0000 (19:57 +0000)]
Branch b1_8
b=21224
i=girish
i=panda

Export ext4_get_group_desc() instead of reimplementing it
in fsfilt.
Speed up quotacheck by using find_next_bit() and stopping
scanning once we have know that no more inodes in-use remain.

14 years agob=20198
pravins [Tue, 10 Nov 2009 06:58:07 +0000 (06:58 +0000)]
b=20198
i=rread
get extra debug info for test 37

14 years agob=20645
jxiong [Tue, 10 Nov 2009 02:33:10 +0000 (02:33 +0000)]
b=20645
r=jay

A temporal fix for osc_trunc_check.

14 years agoBranch b1_8
adilger [Tue, 10 Nov 2009 00:23:40 +0000 (00:23 +0000)]
Branch b1_8
Ignore generated files.

14 years agoRevert the patch for 19906
jxiong [Mon, 9 Nov 2009 06:05:16 +0000 (06:05 +0000)]
Revert the patch for 19906

14 years agoBranch HEAD
tappro [Fri, 6 Nov 2009 22:16:21 +0000 (22:16 +0000)]
Branch HEAD
b=21208
i=rread
i=zam

do request cleanup outside handle_recovery_req()

14 years agoUpdated for 2.0.0 build 30. Please note that a new version string is being used...
stinson1947 [Fri, 6 Nov 2009 21:34:40 +0000 (21:34 +0000)]
Updated for 2.0.0 build 30.  Please note that a new version string is being used, LUSTRE_MINOR is now 10 (up from 9), LUSTRE_PATCH is reset to 0, and builds will now be represented with the build number in the LUSTRE_FIX field.

14 years agob=19919
nathan [Fri, 6 Nov 2009 21:13:41 +0000 (21:13 +0000)]
b=19919
i=manoj
i=adilger
 Make sure there is a valid pool when creating a new file with llapi
 Fixes for 19406
 Various pools test fixes

14 years agob=21281
walter [Fri, 6 Nov 2009 18:07:44 +0000 (18:07 +0000)]
b=21281
o=walter (walter@routingdynamics.com)
i=johann
i=grev
gcc-4.4.0 RHEL-5.3 x86_64 build error fix

14 years agob=21102
brian [Fri, 6 Nov 2009 17:50:43 +0000 (17:50 +0000)]
b=21102
i=yangsheng
i=wangyb

Make the reuse cache hash value for a release kernel different than for a
non-release kernel.  This is because the name of the kernel is embedded in
the package and non-release kernels have a datestring included in them which
is really not suitable for release kernels.

14 years agob=20757
maxim [Fri, 6 Nov 2009 14:09:49 +0000 (14:09 +0000)]
b=20757
i=liang
i=maxim
Minor change of lnet-sysctl.h to allow LND sysctl even if
CFS_SYSFS_MODULE_PARM enabled.

14 years agob=20592
deshmukh [Fri, 6 Nov 2009 13:55:08 +0000 (13:55 +0000)]
b=20592
i=johann
i=andrew.perepechko (panda)

use kthread_run() instead of kernel_thread() so that log dumper won't stuck
in do_exit(). (patch ported to head)

14 years agoFix comment in autoconf/configure test output.
adilger [Fri, 6 Nov 2009 09:58:09 +0000 (09:58 +0000)]
Fix comment in autoconf/configure test output.

14 years agob=15504
grev [Thu, 5 Nov 2009 23:45:18 +0000 (23:45 +0000)]
b=15504
i=Brian
get rid of dash in var name

14 years agob=19669
nathan [Thu, 5 Nov 2009 23:33:58 +0000 (23:33 +0000)]
b=19669
i=adilger
i=vitaly
- get and set LMA in mdd layer
- change LMA endianness and handle swabbing

14 years agoBranch HEAD
dzogin [Thu, 5 Nov 2009 22:10:32 +0000 (22:10 +0000)]
Branch HEAD
b=21031
i=andrew.perpechko
i=girish.shilamkar
----------------------------------------------------------------------
 Modified Files:
  ldiskfs/ChangeLog
  ldiskfs/kernel_patches/series/ldiskfs-2.6-rhel5.series
  ldiskfs/kernel_patches/series/ldiskfs-2.6-sles10.series
 Added Files:
  ldiskfs/kernel_patches/patches/ext3-corrupted-orphans-2.6.patch
----------------------------------------------------------------------
Description: ext3-corrupted-orphans-2.6.patch added.
Details    : Backport of ext3 patch to handle corrupted orphans.
-------------------------------------------------------------------------------

14 years agob=21231
brian [Thu, 5 Nov 2009 22:04:19 +0000 (22:04 +0000)]
b=21231
i=brian
o=Christopher Morrone

Fix the --with-lustre-iokit configure macro.

14 years agob=19808
brian [Thu, 5 Nov 2009 21:52:47 +0000 (21:52 +0000)]
b=19808
i=yangsheng

Fix the asm reorganization for i?86 as has been done previously for the
other arches.

14 years agoBranch:HEAD
wangdi [Thu, 5 Nov 2009 21:45:51 +0000 (21:45 +0000)]
Branch:HEAD

b=19906
1. Fix osc_lock_fits_into to make sure only the granted locks can be matched by
   the enqueue sublocks.
2. Add loc_lock_cancel to Unuse all sublocks once use(for glimpse) failed.
3. some other small fixes.
i=Jay,Ericm

14 years agoBranch HEAD
robert.read [Thu, 5 Nov 2009 21:13:06 +0000 (21:13 +0000)]
Branch HEAD

Correct typo in example cvsrc.

14 years agob=18631
nicolas.williams [Thu, 5 Nov 2009 20:56:12 +0000 (20:56 +0000)]
b=18631
i=robert.read@sun.com
i=tom.wang@sun.com

14 years agoBranch HEAD
vitaly [Thu, 5 Nov 2009 20:37:53 +0000 (20:37 +0000)]
Branch HEAD
b=18478
i=jay

sanity test_180, specify node for the test properly

14 years agoBranch HEAD
zam [Wed, 4 Nov 2009 18:44:10 +0000 (18:44 +0000)]
Branch HEAD
b=20498
i=tappro
i=robert.read

More debugging output for hanging unmount: dump locks
for the exports with extra references.

14 years agob=20529
jxiong [Wed, 4 Nov 2009 13:25:35 +0000 (13:25 +0000)]
b=20529
r=jay,wangdi

- fix memory map issue of group lock
- revise liblustre group lock implementation

14 years agoBranch HEAD
tianzy [Wed, 4 Nov 2009 07:15:38 +0000 (07:15 +0000)]
Branch HEAD
don't reveal unnecessary message to users
b=21068
i=rebort
i=yong.fan

14 years agob=20703,i=wangdi:
isaac [Wed, 4 Nov 2009 06:26:55 +0000 (06:26 +0000)]
b=20703,i=wangdi:
- fixed a wrong error message.

14 years agob=19200
fanyong [Wed, 4 Nov 2009 05:29:17 +0000 (05:29 +0000)]
b=19200
i=tom.wang
i=oleg.drokin
i=adilger

Drop join file, do not support it any more.

14 years agob=19128
green [Wed, 4 Nov 2009 04:17:26 +0000 (04:17 +0000)]
b=19128
r=adilger,tappro,rread

Properly add a request to the start of a replay list if there are no
requests with smaller tid in the list already.

14 years agob=20530
fanyong [Wed, 4 Nov 2009 02:20:51 +0000 (02:20 +0000)]
b=20530
i=johann
i=tianzy

In quota slaves, quota_chk_acq_common() shouldn't retry indefinitely
if relative clients are evicted or reconnect already.

14 years agob=20975
manoj [Tue, 3 Nov 2009 22:17:51 +0000 (22:17 +0000)]
b=20975
r=robert.read
r=nathan.rutman

Change forced stop of MDS to a normal stop.

14 years agoBranch HEAD
robert.read [Tue, 3 Nov 2009 20:55:09 +0000 (20:55 +0000)]
Branch HEAD
b=21255,9787

Temporarily disable long running test.

14 years agoBranch: b_cmd_cray
wangdi [Fri, 30 Oct 2009 18:58:30 +0000 (18:58 +0000)]
Branch: b_cmd_cray
Use HEAD version ldiskfs on b_cmd_cray branch.

14 years agoBranch HEAD
vitaly [Fri, 30 Oct 2009 10:42:14 +0000 (10:42 +0000)]
Branch HEAD
b=21168
i=jian.yu

ensure all OST get reconnected after the client remount before proceeding

14 years agoBranch HEAD
vitaly [Fri, 30 Oct 2009 10:37:28 +0000 (10:37 +0000)]
Branch HEAD
b=18478
i=jay
i=liang

a new osc is added for echo client purpose (instead of re-using an existing one)
a test is added

14 years agob=19242
grev [Thu, 29 Oct 2009 14:12:04 +0000 (14:12 +0000)]
b=19242
i=Robert.Read
additional mdcdev check

14 years agoBranch HEAD
tappro [Thu, 29 Oct 2009 12:13:43 +0000 (12:13 +0000)]
Branch HEAD
b=20802
i=adilger
i=rread

Exit with error from class_new_export() if export is being stopped instead of
assertion. Properly clean up export data in error path.

14 years agoBranch HEAD
tappro [Thu, 29 Oct 2009 12:10:04 +0000 (12:10 +0000)]
Branch HEAD
b=20802
i=adilger
i=rread

delete client slot early, during export disconnect to avoid duplicated records
in last_rcvd file about the same export uuid

14 years agoBranch HEAD
tappro [Thu, 29 Oct 2009 12:07:24 +0000 (12:07 +0000)]
Branch HEAD
b=20802
i=adilger
i=alex

take export reference for transaction callback during client addition

14 years agoBranch HEAD
tappro [Thu, 29 Oct 2009 12:02:07 +0000 (12:02 +0000)]
Branch HEAD
b=20802
i=adilger
i=alex

don't sync upon client slot deletion, do server update at first instead.

14 years agob=19906
jxiong [Thu, 29 Oct 2009 02:47:03 +0000 (02:47 +0000)]
b=19906
r=rread

remove noisy messages in cl_lock_get/put/get_trust.

14 years agoBranch HEAD
fanyong [Thu, 29 Oct 2009 02:16:02 +0000 (02:16 +0000)]
Branch HEAD
b=21064
i=robert.read
i=rahul.deshmukh

1) do not wait service thread in "ptlrpc_stop_thread()" if it has stopped already.
2) give more debug information if service thread can not stop.

14 years agoBranch HEAD
fanyong [Thu, 29 Oct 2009 02:00:48 +0000 (02:00 +0000)]
Branch HEAD
b=20984
i=robert.read
i=tom.wang

1) increase sleep interval for "sleep_maxage()".
2) use "O_LOV_DELAY_CREATE" mode when create in "get_mds_dir()" to detect MDT index,to avoid unnecessary "enospc" error.
3) only obtain the first line if the output contain many OSTs result.
4) other fixes from b1_8.

14 years agoBranch HEAD
fanyong [Thu, 29 Oct 2009 01:40:54 +0000 (01:40 +0000)]
Branch HEAD
b=21056
i=tappro
i=robert.read

Hold lov obd_device reference count before access lower OSC target.

14 years agoBranch HEAD
vitaly [Tue, 27 Oct 2009 12:25:52 +0000 (12:25 +0000)]
Branch HEAD
b=20098

reverting patches as they lead to some problems

14 years agob=20308
grev [Tue, 27 Oct 2009 10:33:19 +0000 (10:33 +0000)]
b=20308
i=Robert.Read
statahead test: N mountpoints on all available clients

14 years agoadd 2.6.27 kernel support
shadow [Mon, 26 Oct 2009 08:09:35 +0000 (08:09 +0000)]
add 2.6.27 kernel support

Branch HEAD
b=14250
i=rread
i=jay

14 years agoBranch HEAD
robert.read [Fri, 23 Oct 2009 19:26:30 +0000 (19:26 +0000)]
Branch HEAD
b=20790
i=nathan

Add ltrack_stats to HEAD.

14 years agoBranch HEAD
bobijam [Fri, 23 Oct 2009 06:09:35 +0000 (06:09 +0000)]
Branch HEAD
b=20149

add more debug info.

14 years agoUpdaged version string for 2.0.0 build 29.
stinson1947 [Thu, 22 Oct 2009 20:14:21 +0000 (20:14 +0000)]
Updaged version string for 2.0.0 build 29.

14 years agoBranch HEAD
johann [Thu, 22 Oct 2009 12:08:51 +0000 (12:08 +0000)]
Branch HEAD
b=20633
i=green
i=eeb

disable irq in trace_lock_tcd only when needed.

14 years agoBranch HEAD
vitaly [Thu, 22 Oct 2009 11:16:03 +0000 (11:16 +0000)]
Branch HEAD
b=20571
i=green
i=deen

handle openflags in liblustre properly

14 years agob=20897
maxim [Thu, 22 Oct 2009 10:40:20 +0000 (10:40 +0000)]
b=20897
i=isaac
i=maxim
Fix for conn race when create_conn creates conn and add it to peer's list,
releases global lock, gets a comm error, acquires the lock again and attempts
to close the conn while it could be closed by some other thread.

14 years agoBranch HEAD
zam [Thu, 22 Oct 2009 07:35:58 +0000 (07:35 +0000)]
Branch HEAD
b=20831
i=nathan.rutman
i=robert.read

Fix a race between do_requeue and client_disconnect_export by protecting
the check for cl_conn_count by the client semaphore.

14 years agoBranch HEAD
tianzy [Thu, 22 Oct 2009 05:39:31 +0000 (05:39 +0000)]
Branch HEAD
clean the code relative to lqs_refcount
b=18782
i=johann
i=panda

14 years agoBranch HEAD
bobijam [Thu, 22 Oct 2009 03:26:03 +0000 (03:26 +0000)]
Branch HEAD
b=19325
i=vitaly.fertman
i=oleg.drokin (green)

Description: 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.

14 years agobranch: HEAD
ericm [Thu, 22 Oct 2009 03:15:20 +0000 (03:15 +0000)]
branch: HEAD
wait before test file size.
b=20801
r=rread