Whamcloud - gitweb
10 years agob=20581 MDS returns full hash for readdir to decrease hash collision 2.0.59-llnl2-base
nasf [Tue, 1 Mar 2011 07:59:40 +0000 (15:59 +0800)]
b=20581 MDS returns full hash for readdir to decrease hash collision

1) MDS returns full hash (both major and minor, 64bit) for readdir to decrease
   hash collision.
2) Synchronization control between traversing thread and statahead thread when
Issues: LU-71
Change-Id: I043784d58607d474a501111c9690fdab89ce8a4a
Signed-off-by: nasf <yong.fan@whamcloud.com>
Reviewed-on: http://review.whamcloud.com/281
Tested-by: Hudson
Reviewed-by: Cliff White <cliffw@whamcloud.com>
Reviewed-by: Oleg Drokin <green@whamcloud.com>
10 years agoLU-68 Fix a race between lock cancel and write
Oleg Drokin [Mon, 14 Mar 2011 22:46:17 +0000 (18:46 -0400)]
LU-68 Fix a race between lock cancel and write

A race in osc_lock_detach due to disconnect between reading
kms from lock resource and write updating kms in the resource
could lead to loss of the update.
Lock the object across entire kms access to avert the issue.

Change-Id: I4b15e611b91b7a4a6b69f5c9fa9b6bd0a1f8cac9
Signed-off-by: Oleg Drokin <green@whamcloud.com>
Reviewed-on: http://review.whamcloud.com/327
Tested-by: Hudson
Reviewed-by: wangdi <di.wang@whamcloud.com>
Reviewed-by: Jinshan Xiong <jay@whamcloud.com>
10 years agoLU-82 Remove useless clio locks
Jinshan Xiong [Mon, 7 Mar 2011 18:56:17 +0000 (10:56 -0800)]
LU-82 Remove useless clio locks

A CLS_NEW toplock must be deleted also if the last sublock is being deleted.

Signed-off-by: Jinshan Xiong <jay@whamcloud.com>
Change-Id: I9831ab534ae222fe74ebaf34faafe00bf3d18cb7
Reviewed-on: http://review.whamcloud.com/305
Tested-by: Hudson
Reviewed-by: Oleg Drokin <green@whamcloud.com>
10 years agoLU-83 Kernel BUG in ext4 if wrong mount option is given
Oleg Drokin [Fri, 11 Mar 2011 01:30:17 +0000 (20:30 -0500)]
LU-83 Kernel BUG in ext4 if wrong mount option is given

Original attempt at landing a fix for this bug landed hopelessly
broken patch.
Replace the patch with proper code.

Change-Id: Iddc38c4254b1f981b10e272b671fad232e20fd70
Signed-off-by: Oleg Drokin <green@whamcloud.com>
Reviewed-on: http://review.whamcloud.com/314
Tested-by: Hudson
Reviewed-by: Johann Lombardi <johann@whamcloud.com>
10 years agob=24375 Fix lvb updating race in enqueue vs completion case
Oleg Drokin [Fri, 11 Feb 2011 03:28:15 +0000 (22:28 -0500)]
b=24375 Fix lvb updating race in enqueue vs completion case

ldlm_enqueue_tail checked for lock mode and updated lvb without
taking appropriate locks.
Take res and lock around mode check and lvb updating.

Issue: LU-67
Change-Id: I8fb764313326b8bf8f813e7cf77c050dd68afe45
Reviewed-on: http://review.whamcloud.com/229
Tested-by: Hudson
Reviewed-by: Liang Zhen <liang@whamcloud.com>
Reviewed-by: Andreas Dilger <adilger@whamcloud.com>
Reviewed-by: Alex Zhuravlev <bzzz@whamcloud.com>
10 years agob=22980 improve obdfilter-survey performance on multi-core system
Niu Yawei [Fri, 18 Feb 2011 03:46:53 +0000 (19:46 -0800)]
b=22980 improve obdfilter-survey performance on multi-core system

1) Add unlock_ioctl support.
2) Land OBD_ALLOC_LARGE() patch from b4424 to avoid vmalloc in ioctl path.
3) Minor fix in class_handle_ioctl() to avoid double free.

Issue: LU-66
Change-Id: Id965cd13fdaeb63197c30e557708afbfa9ee9f39
Signed-off-by: Niu Yawei <niu@whamcloud.com>
Reviewed-on: http://review.whamcloud.com/247
Tested-by: Hudson
Reviewed-by: Oleg Drokin <green@whamcloud.com>
Reviewed-by: Andreas Dilger <adilger@whamcloud.com>
10 years agoLU-116 fix mmp test 9 and test 10
Yu Jian [Mon, 7 Mar 2011 10:44:02 +0000 (18:44 +0800)]
LU-116 fix mmp test 9 and test 10

1) fix the sleep time in test 9 to abort e2fsck before it's complete
2) fix test 10 to allow e2fsck run on mounted filesystem with -n option

Signed-off-by: Yu Jian <yujian@whamcloud.com>
Change-Id: I2605d75f47e44db659ced2e5435e3b51cc8897c6
Reviewed-on: http://review.whamcloud.com/301
Tested-by: Hudson
Reviewed-by: Brian J. Murrell <brian@whamcloud.com>
Reviewed-by: Oleg Drokin <green@whamcloud.com>
10 years agob=22703 no error when close if application has known write failure already
nasf [Wed, 2 Mar 2011 10:23:08 +0000 (18:23 +0800)]
b=22703 no error when close if application has known write failure already

Do not return error when close if application has known write failure already.

Issues: LU-65
Change-Id: I25b9628fcdd3bf5b7db9d3f8460e06c18226b2e0
Signed-off-by: nasf <yong.fan@whamcloud.com>
Reviewed-on: http://review.whamcloud.com/286
Tested-by: Hudson
Reviewed-by: Jinshan Xiong <jay@whamcloud.com>
Reviewed-by: Oleg Drokin <green@whamcloud.com>
10 years agoMake llite_lloop.ko build conditional
Christopher J. Morrone [Thu, 24 Feb 2011 22:11:45 +0000 (14:11 -0800)]
Make llite_lloop.ko build conditional

The llite_lloop.ko module only works with kernel page sizes smaller
than 64K, so test for the kernel page size and only build the module

Issue: LU-94
Change-Id: I78e416832767bbe29529d152e45addec7a999506
Signed-off-by: Christopher J. Morrone <morrone2@llnl.gov>
Reviewed-on: http://review.whamcloud.com/266
Tested-by: Hudson
Reviewed-by: Brian J. Murrell <brian@whamcloud.com>
Reviewed-by: Oleg Drokin <green@whamcloud.com>
10 years agoORNL-2 Enable OBD_CONNECT_ATTRFID supporting on lustre-2.x
nasf [Mon, 7 Mar 2011 15:23:42 +0000 (23:23 +0800)]
ORNL-2 Enable OBD_CONNECT_ATTRFID supporting on lustre-2.x

To drop redundant MDS-side RPC for root object revalidation when
stats each item for traversing directory based on absolute path.

Signed-off-by: nasf <yong.fan@whamcloud.com>
Change-Id: I5f7c7755b4142ac1f23c66517e5919c1398e15b8
Reviewed-on: http://review.whamcloud.com/302
Tested-by: Hudson
Reviewed-by: Lai Siyao <laisiyao@whamcloud.com>
Reviewed-by: Oleg Drokin <green@whamcloud.com>
10 years agob=24417 Update Build-Depends
Brian J. Murrell [Fri, 21 Jan 2011 22:47:58 +0000 (17:47 -0500)]
b=24417 Update Build-Depends

- remove texlive-latex-recommended as a build requirement
- add missing "| automake1.7 | automake1.8 | automake1.9" to

Issue: LU-52
Signed-off-by: Brian J. Murrell <brian@whamcloud.com>
Change-Id: Ief22c678014836e8372a18dd3ed48a95631177b5
Reviewed-on: http://review.whamcloud.com/212
Tested-by: Hudson
Reviewed-by: Robert Read <rread@whamcloud.com>
Reviewed-by: Oleg Drokin <green@whamcloud.com>
10 years agoLU-120 the perfect storm
Brian J. Murrell [Mon, 7 Mar 2011 22:37:44 +0000 (17:37 -0500)]
LU-120 the perfect storm

There is a corner case in the debian patching/unpatching process
where a patch converts the use of a given file to a file.in that is
to be substed.  The problem is that between the configure and the
make debs, a file, which a debian patch wants to remove already
exists (it was created by the configure) and dpkg-buildpackage's
unpatch step fails becuase it wants to try to create a file which
configure already created.

Signed-off-by: Brian J. Murrell <brian@whamcloud.com>
Change-Id: I185e496b82687d21ea3d744e939536c106b702f9
Reviewed-on: http://review.whamcloud.com/308
Tested-by: Hudson
Reviewed-by: Robert Read <rread@whamcloud.com>
Reviewed-by: Oleg Drokin <green@whamcloud.com>
10 years agoLU-97 lustre-modules should depend on ldiskfs.
Brian J. Murrell [Thu, 24 Feb 2011 18:06:24 +0000 (13:06 -0500)]
LU-97 lustre-modules should depend on ldiskfs.

Only for the server of course.
But depend on a generically named tag, and have ldiskfs provide
that tag so that any future backing stores can be used as a
replacement for ldiskfs.

Signed-off-by: Brian J. Murrell <brian@whamcloud.com>
Change-Id: Ibef2b8c2b2a8c7339143dc32d32b59ace94f00fd
Reviewed-on: http://review.whamcloud.com/265
Reviewed-by: Robert Read <rread@whamcloud.com>
Tested-by: Hudson
Reviewed-by: Andreas Dilger <adilger@whamcloud.com>
Reviewed-by: Oleg Drokin <green@whamcloud.com>
10 years agoLU-103 Add support for test groups.
Robert Read [Tue, 1 Mar 2011 00:46:16 +0000 (16:46 -0800)]
LU-103 Add support for test groups.

Added an -g GROUP_FILE option to auster. The tests listed in the
group file will be used instead of any passed on the command line. I also added
a couple of test-groups to get things started, although
I'm sure we'll want to refine these.

Took care of a few cleanups in run_suite as well.

Signed-off-by: Robert Read <rread@whamcloud.com>
Change-Id: I5000c2b95a6eb1a49f5305637d4b55de5edc6d5c
Reviewed-on: http://review.whamcloud.com/278
Tested-by: Hudson
Reviewed-by: Chris Gearing <chris@whamcloud.com>
Reviewed-by: Wei Liu <sarah@whamcloud.com>
Reviewed-by: Oleg Drokin <green@whamcloud.com>
10 years agoLU-119 Remove code added by a conflict resolution
Brian J. Murrell [Mon, 7 Mar 2011 17:47:05 +0000 (12:47 -0500)]
LU-119 Remove code added by a conflict resolution

I erroneously re-added some code that was removed in
62345145fa7cae1846f96078bac43ddf58cea640 while resolving
a conflict in LU-85's landing of commit

Signed-off-by: Brian J. Murrell <brian@whamcloud.com>
Change-Id: Id2489a3aa5560cdf4e7cf4755ddc16437558a2fc
Reviewed-on: http://review.whamcloud.com/304
Tested-by: Hudson
Reviewed-by: Christopher J. Morrone <morrone2@llnl.gov>
Reviewed-by: Robert Read <rread@whamcloud.com>
Reviewed-by: Oleg Drokin <green@whamcloud.com>
10 years agoNo bugzilla for this. One of the test failing was config-sanity.sh
James Simmons [Fri, 18 Feb 2011 18:58:41 +0000 (13:58 -0500)]
No bugzilla for this. One of the test failing was config-sanity.sh
test number 37. So I tracked it down to way. The reason was the mounts
were failing. Upon investigation I discovered the test caught a error
reported back by the mount. I tracked it down to the section of code in
the patch. Basically mount_lustre.c attempts to set the stripe_cache_size
for /dev/mdX devices. The problem is this extra feature is present for
certain classes of md devices. In the case of our test bed we are one
of the classes that lacks this enhancement. So instead of failing to mount
in that case we just print a warning that the special param is not present
but still succed to mount.

Change-Id: Ib8a743e77b5e6994e2640c51b781f7d0d345d5a7
Signed-off-by: James Simmons <uja.ornl@gmail.com>
Reviewed-on: http://review.whamcloud.com/254
Tested-by: Hudson
Reviewed-by: Brian J. Murrell <brian@whamcloud.com>
Reviewed-by: Oleg Drokin <green@whamcloud.com>
10 years agoLU-85 Update some URLs and text to reflect our releases. 60/260/6
Brian J. Murrell [Wed, 23 Feb 2011 21:38:40 +0000 (16:38 -0500)]
LU-85 Update some URLs and text to reflect our releases.

The text and URLs in the debian packaging metadata were old
(circa Sun Microsystems), so update them and take the opportunity
to inform about the source of the release being installed.

Signed-off-by: Brian J. Murrell <brian@whamcloud.com>
Change-Id: I7860e8ada7757d8a991a472af293749fd4c89f5a

10 years agoMerge "b=21581 too long file / path names for old tar"
Oleg Drokin [Fri, 4 Mar 2011 23:51:32 +0000 (15:51 -0800)]
Merge "b=21581 too long file / path names for old tar"

10 years agoLU-111 Fix force options parsing 97/297/2
Oleg Drokin [Fri, 4 Mar 2011 16:15:41 +0000 (11:15 -0500)]
LU-111 Fix force options parsing

parse_options had incorrect check for force option that matched
all force options like force_over_16tb and as a result those did
not get propagated down.

Change-Id: I46d07a6d505acc9150e2655a63be55fb7ef6954d
Signed-off-by: Oleg Drokin <green@whamcloud.com>
10 years agoMerge "Remove rpm spec code and message related to "alias lustre llite""
Oleg Drokin [Fri, 4 Mar 2011 22:37:12 +0000 (14:37 -0800)]
Merge "Remove rpm spec code and message related to "alias lustre llite""

10 years agoMerge "Remove obsolete nodelustre script"
Oleg Drokin [Fri, 4 Mar 2011 04:42:13 +0000 (20:42 -0800)]
Merge "Remove obsolete nodelustre script"

10 years agoAdd -D logdir option to Auster so that the whole logdir can be specified 88/288/2
Chris [Thu, 3 Mar 2011 00:38:21 +0000 (00:38 +0000)]
Add -D logdir option to Auster so that the whole logdir can be specified

Added a -D logdir so that Auster does not append a time based suffix to the dir
this means that the location of the logs can be known by the parent of Auster

Use Case: The job manager needs to place console logs into the logs directory
and is a parent of Auster, therefor the job manage creates a directory and passes
it to auster using the -D option

Issue: LU-105
Change-Id: Idbd258677b9c6624fdac43aa6223fe6f516bc989
Signed-off-by: Chris Gearing <chris@whamcloud.com>
10 years agoRemove obsolete nodelustre script 91/291/1
Christopher J. Morrone [Thu, 3 Mar 2011 01:05:17 +0000 (17:05 -0800)]
Remove obsolete nodelustre script

Change-Id: I1cf7e933225b773e95210f0bc5088ecbd7ab0037
Signed-off-by: Christopher J. Morrone <morrone2@llnl.gov>
10 years agoMerge "Remove EXPORT_SYMBOL for static function."
Oleg Drokin [Tue, 1 Mar 2011 04:07:59 +0000 (20:07 -0800)]
Merge "Remove EXPORT_SYMBOL for static function."

10 years agob=21581 too long file / path names for old tar 22/222/5
Brian J. Murrell [Mon, 24 Jan 2011 20:53:53 +0000 (15:53 -0500)]
b=21581 too long file / path names for old tar

Instruct automake to use tar's ustar format to prevent errors when
pathnames are longer than 99 chars.
- this requires automake >= 1.9, so adjust accordingly
  - including dealing with multiple versions of automake installed
Per andreas' inspection, provide a hint to the user on how to install
a missing automake or autoconf.
Per mjmac's inspection, give Debian users help on how to install
- note that I am *assuming" that Debian's lsb distribution id is "Debian"


Issue: LU-47
Signed-off-by: Brian J. Murrell <brian@whamcloud.com>
Change-Id: Ica6419196d2a6910824e8f2326537d9b46612c15

10 years agoAdd location argument to $upload_script call 69/269/1
John Spray [Mon, 28 Feb 2011 16:33:08 +0000 (16:33 +0000)]
Add location argument to $upload_script call

Issue: LU-100
Change-Id: Idbe262f15f0915ef9ffe8f6f05256eaf324d8a51
Signed-off-by: John Spray <john@whamcloud.com>
10 years agoMerge "b=24413 fix for automake > 1.9.6"
Oleg Drokin [Sat, 26 Feb 2011 01:55:40 +0000 (17:55 -0800)]
Merge "b=24413 fix for automake > 1.9.6"

10 years agob=24413 fix for automake > 1.9.6 07/207/5
Brian J. Murrell [Fri, 21 Jan 2011 18:54:06 +0000 (13:54 -0500)]
b=24413 fix for automake > 1.9.6

We seem to be using a Makefile variable that does not exist in more
recent versions of automake.  This fixes that problem.

Issue: LU-48
Signed-off-by: Brian J. Murrell <brian@whamcloud.com>
Change-Id: I84d0656e9bee0878c7ac25183476e39dca63366f

10 years agoRemove rpm spec code and message related to "alias lustre llite" 64/264/1
Christopher J. Morrone [Sat, 19 Feb 2011 01:52:43 +0000 (17:52 -0800)]
Remove rpm spec code and message related to "alias lustre llite"

It has been quite some time since we used the module alias:

  alias lustre llite

Remove conversion code from the spec file.
Remove the confusing error message that make reference to the alias.

Change-Id: I412ce738243e3d9824b4a94f8a4fd95811cfae0d
Signed-off-by: Christopher J. Morrone <morrone2@llnl.gov>
10 years agoRemove EXPORT_SYMBOL for static function. 62/262/1
Christopher J. Morrone [Wed, 16 Feb 2011 21:13:17 +0000 (13:13 -0800)]
Remove EXPORT_SYMBOL for static function.

gcc 4.4.4 in RHEL6 on ppc64 does not allow EXPORT_SYMBOL for
a static function (which seems reasonable).  The following error
is seen on the EXPORT_SYMBOL line:

error: __ksymtab_obd_device_alloc causes a section type conflict

Change-Id: I6ad5ae13346b82bd723e05545d1e2a765b7460a8
Signed-off-by: Christopher J. Morrone <morrone2@llnl.gov>
10 years agoDon't upload to maloo by default 44/244/2
Robert Read [Thu, 17 Feb 2011 06:43:17 +0000 (22:43 -0800)]
Don't upload to maloo by default

- Main change is to invert meaning of -l flag. It now turns *on* uploading
  logs to maloo, instead of disabling it.

- Also added check to find the absolute path for maloo_upload.sh, just like
  we do for the test scripts we run.

- The verbose option now dumps the output of the script to the console (and
  also saves it to the log file).

- The -h (help) option will now page the help text if needed.

Issue: LU-86
Change-Id: Ieccb3f431b87d39007cd4f267b6e3e61a41a024a
Signed-off-by: Robert Read <rread@whamcloud.com>
10 years agoUpdates to test scripts: 23/223/4
John Spray [Tue, 8 Feb 2011 15:25:06 +0000 (15:25 +0000)]
Updates to test scripts:
 * Add auster test-runner script
 * Add integration with maloo test DB

Change-Id: Ifb4903ffad3fafc989e3261d72d41051b171a90e

10 years agoProperly pick module parameters from modules.conf 02/202/4
Oleg Drokin [Thu, 20 Jan 2011 02:55:00 +0000 (21:55 -0500)]
Properly pick module parameters from modules.conf

Fix incorrect regexp that assumes "options <module> params" format
instead of proper "options module params".
Also use baseline to extract just the name of the module.
Now depends on GNU grep for -P (perl_ regexps so that '+' glob works.

Issue: LU-46
Change-Id: Id4d939bf2d9a412d1d5f3f2e7fcbf466ce436e28

10 years agob=24416 debian packaging fixes 13/213/6
Brian J. Murrell [Sat, 22 Jan 2011 11:16:55 +0000 (06:16 -0500)]
b=24416 debian packaging fixes

- don't make a patch out of anything in /debian
- exclude noise files from the debian built source tarball
- fake debian/patche{s,d} for make dist
- a few more reasons to run autogen.sh
- figure out if dist tarball needs autogen.shs and include it if so
- look for and run autogen.sh in the build subdir
- make debdiff as part of make dist
- add a debian/source/format file
- mv the orig tarball and the debdiff to the debs dir
- don't try to dist /debian for non-dpkg-using build targets

Issue: LU-51
Change-Id: I041aaef217e107def86ce808d0e96fc6891e1dcd

10 years agob=24378 wrong error code conversion in mdd_path_current() 2.0.59-llnl-base
Vitaly Fertman [Tue, 28 Dec 2010 20:52:23 +0000 (23:52 +0300)]
b=24378 wrong error code conversion in mdd_path_current()


10 years agob=24357 thread_sanity() fix
Elena Gryaznova [Fri, 24 Dec 2010 17:22:20 +0000 (20:22 +0300)]
b=24357 thread_sanity() fix


10 years agob=24037 Remove iopen ldiskfs patch for rhel5.
yangsheng [Fri, 24 Dec 2010 03:53:48 +0000 (11:53 +0800)]
b=24037 Remove iopen ldiskfs patch for rhel5.

10 years agob=23368 MMP: Watchdog detected LOCKUP on CPU 1 RIP: __list_add+0x6/0x68
Rahul Deshmukh [Mon, 27 Dec 2010 07:29:52 +0000 (12:59 +0530)]
b=23368 MMP: Watchdog detected LOCKUP on CPU 1 RIP: __list_add+0x6/0x68

10 years agob=24294 test_pios: take the ost-s sizes into account
Elena Gryaznova [Thu, 23 Dec 2010 23:39:13 +0000 (02:39 +0300)]
b=24294 test_pios: take the ost-s sizes into account


10 years agob=24255 move seq_client_alloc_seq out of lcs_sem
Vitaly Fertman [Thu, 23 Dec 2010 22:42:46 +0000 (01:42 +0300)]
b=24255 move seq_client_alloc_seq out of lcs_sem


do not send RPC under seq->lcs_sem, because in a case of recovery
the recovery thread takes it too and deadlocks.

10 years agob=23793 MOUNTOPT "-o" cleanup
Elena Gryaznova [Thu, 23 Dec 2010 21:23:27 +0000 (00:23 +0300)]
b=23793 MOUNTOPT "-o" cleanup


10 years agob=21122 fix a race between page fault and lock cancel.
Eric Mei [Thu, 23 Dec 2010 01:48:57 +0000 (18:48 -0700)]
b=21122 fix a race between page fault and lock cancel.


10 years agob=16359 _debug_req fix
cliff.white@oracle.com [Thu, 23 Dec 2010 22:38:39 +0000 (01:38 +0300)]
b=16359 _debug_req fix


10 years agob=23409 add -i to the setstripe usage and man page
Andreas Dilger [Tue, 10 Aug 2010 08:43:49 +0000 (04:43 -0400)]
b=23409 add -i to the setstripe usage and man page

Add the "-i" option to the "lfs setstripe" usage and man page.
Fix nroff formatting in the "lfs setstripe" and "lfs getstripe".


10 years agoRevert "b=20288 kiblnd_check_conns can deadlock"
Vitaly Fertman [Thu, 23 Dec 2010 22:49:21 +0000 (01:49 +0300)]
Revert "b=20288 kiblnd_check_conns can deadlock"

This reverts commit 9887737999bb2db2bc2f10b77854dee7f471ba62.

10 years agob=24291 Ignore PARAM_NETWORK in mgs_write_log_param()
Zhiqi Tao [Thu, 16 Dec 2010 10:28:29 +0000 (21:28 +1100)]
b=24291 Ignore PARAM_NETWORK in mgs_write_log_param()


10 years agob=24320 do not fork a new thread in mem pressure
Vitaly Fertman [Tue, 21 Dec 2010 13:07:45 +0000 (16:07 +0300)]
b=24320 do not fork a new thread in mem pressure


we already check for PF_MEMALLOC in ldlm shrinker and pass this flag
to the blocking thread, but a new thread start was still done with
no check for this flag.

10 years agob=24128 mgs_write_log_ost() should not update client's log
Vitaly Fertman [Tue, 21 Dec 2010 13:04:22 +0000 (16:04 +0300)]
b=24128 mgs_write_log_ost() should not update client's log

o=Lai Siyao

when an OST is set to an existed index, don't update client logs.
otherwise client will have duplicate OSC logs to the same OST.

10 years agob=24017 make llverfs more robust against errors during testing
Jian Yu [Mon, 20 Dec 2010 10:33:40 +0000 (18:33 +0800)]
b=24017 make llverfs more robust against errors during testing

Make llverfs more robust about continuing in the face of errors.

o=Andreas Dilger

10 years agob=23051 improve summary of acc-sm to include test times
Elena Gryaznova [Mon, 20 Dec 2010 18:41:36 +0000 (21:41 +0300)]
b=23051 improve summary of acc-sm to include test times


10 years agob=22771 Patch to disable mbcache
Girish Shilamkar [Sun, 19 Dec 2010 12:25:43 +0000 (17:55 +0530)]
b=22771 Patch to disable mbcache


10 years agob=16890 llapi_quotactl man page update
Andrew Perepechko [Mon, 20 Dec 2010 21:37:32 +0000 (00:37 +0300)]
b=16890 llapi_quotactl man page update

additional explanation of igrace/bgrace/itime/btime

10 years agob24336 ldlm_resource::lr_lvb_data is protected by wrong lock
Liang Zhen [Tue, 21 Dec 2010 01:26:08 +0000 (09:26 +0800)]
b24336 ldlm_resource::lr_lvb_data is protected by wrong lock

- ldlm_resource::lr_lvb_data should always be protected by lr_lvb_sem
- cleanup some unnecessary lock dance


10 years agob=22117 Patch to limit mmp interval
Girish Shilamkar [Sun, 19 Dec 2010 12:28:51 +0000 (17:58 +0530)]
b=22117 Patch to limit mmp interval


10 years agoRevert "b=22176 Add .sync_fs super block handler"
Vitaly Fertman [Wed, 22 Dec 2010 00:58:11 +0000 (03:58 +0300)]
Revert "b=22176 Add .sync_fs super block handler"

This reverts commit 860404143dafbf696b4d38bb9135eabfd471cc36.

10 years agoUpdated for 2.1.0 build 10. v2_0_59_0
Terry Rutledge [Sat, 18 Dec 2010 00:30:56 +0000 (16:30 -0800)]
Updated for 2.1.0 build 10.

10 years agob=24324 no new lock on disconnected export
Vitaly Fertman [Thu, 16 Dec 2010 00:19:35 +0000 (03:19 +0300)]
b=24324 no new lock on disconnected export


check if export is being disconnected and do not add new locks to it

10 years agob=24218 fix race on ksock_tx_t::tx_resid
Vitaly Fertman [Thu, 16 Dec 2010 00:19:37 +0000 (03:19 +0300)]
b=24218 fix race on ksock_tx_t::tx_resid


If a connection is closed before ksocknal_transmit() returns to
ksocknal_process_transmit(), then nobody has refcount on conn::ksnc_sock
and all pending ZC requests will be finalized by ksocknal_connsock_decref->
ksocknal_finalize_zcreq, ksocknal_finalize_zcreq will mark not-acked ZC
request as error by setting tx::tx_reside = -1.
This is race because ksocknal_process_transmit() will check tx::tx_resid
right after calling ksocknal_transmit(), and it can get tx->tx_resid != 0
and rc == 0 then hit later LASSERT(rc < 0).

10 years agob=24078 replay barrier checks sequence allocation
Mikhail Pershin [Wed, 15 Dec 2010 15:07:12 +0000 (18:07 +0300)]
b=24078 replay barrier checks sequence allocation


10 years agob=23289 revert patch on 21828
Liang Zhen [Wed, 15 Dec 2010 15:04:30 +0000 (23:04 +0800)]
b=23289 revert patch on 21828

this patch reverted change on bug 21828, because svc::srv_n_active_reqs
is dropped by ptlrpc_server_finish_request() now which is used to be
dropped by last calling of ptlrpc_server_drop_request, so recovery
thread has no refcount on svc::srv_n_active_reqs and we should just
remove these refcount changes.
This should also resovled issue of bug 21140, because logic in
ptlrpc_server_request_get() can be totally screwed up by

10 years agob=23289 don't start more than minimum MDT threads
Andreas Dilger [Tue, 14 Dec 2010 07:33:11 +0000 (00:33 -0700)]
b=23289 don't start more than minimum MDT threads

Because there are many different portals running on the MDS node (MDT,
readpage, setattr, SEQ, SEQ server, FLD, DLM), starting more than the
minimum number of threads can result in thousands of threads running
on large MDS nodes, even if many of those threads are not needed.

Start only the minimum number of threads at mount time, and start more
as required by the load.


10 years agob=23072 avoid duplicate cfs_size_round declarations
Andreas Dilger [Tue, 14 Dec 2010 06:54:49 +0000 (23:54 -0700)]
b=23072 avoid duplicate cfs_size_round declarations

Use HAVE_CFS_SIZE_ROUND to determine if libcfs_private.h or lustre_user.h has
declared the cfs_size_round() helper function.  This avoids conflicts if both
of these headers are included in different orders.

10 years agob=22376 accmode fix
Vitaly Fertman [Thu, 16 Dec 2010 00:19:36 +0000 (03:19 +0300)]
b=22376 accmode fix


accmode is fixed to not let other flags to be set together with MAY_EXEC
test is added

10 years agob=21244 add man page for ll_decode_filter_fid tool
Andreas Dilger [Fri, 17 Dec 2010 00:17:22 +0000 (17:17 -0700)]
b=21244 add man page for ll_decode_filter_fid tool

Add man page for new ll_decode_filter_fid(8) tool.
Add references to lustre(7) man page.

10 years agob=20288 kiblnd_check_conns can deadlock
Isaac Huang [Wed, 15 Dec 2010 15:35:21 +0000 (08:35 -0700)]
b=20288 kiblnd_check_conns can deadlock

Avoid dropping global lock in kiblnd_check_conns


10 years agob=16605 verify object group before using it
Andreas Dilger [Wed, 15 Dec 2010 06:44:39 +0000 (23:44 -0700)]
b=16605 verify object group before using it

Avoid trying to access an invalid/unknown group sent from the client.
Don't return -EINVAL, since it is possible that the problem is the OST
code is old and this object actually exists.  Instead, return -EBADF
so that we can more easily diagnose this problem.

10 years agob=24048 set body->easize in mdc_getattr_pack()
Mikhail Pershin [Wed, 15 Dec 2010 14:51:41 +0000 (17:51 +0300)]
b=24048 set body->easize in mdc_getattr_pack()


This allows client to supply the server with enough info about reply buffers.
Server may use this to allocate buffers and don't guess with max_mdsize

10 years agob=24171 check the init() iterator results properly
Mikhail Pershin [Wed, 15 Dec 2010 14:50:39 +0000 (17:50 +0300)]
b=24171 check the init() iterator results properly


10 years agob=24300 Don't run autogen.sh in the spl and zfs repos.
Ricardo M. Correia [Mon, 13 Dec 2010 15:56:14 +0000 (15:56 +0000)]
b=24300 Don't run autogen.sh in the spl and zfs repos.

The SPL and ZFS repositories already come with the configure script ready to
run, therefore it's not necessary to run autogen.sh on those directories.

Furthermore, in SPL version 0.5.0 this was causing a version conflict when
regenerating the files with the autotools/libtool that come with RHEL5.

10 years agob=24266 increase replay-single test_70d dbench duration for HARD failure mode
Elena Gryaznova [Mon, 13 Dec 2010 17:37:49 +0000 (20:37 +0300)]
b=24266 increase replay-single test_70d dbench duration for HARD failure mode


10 years agob=24037 Using cfs_curproc_umask() instead of reference directly.
yangsheng [Mon, 13 Dec 2010 18:56:38 +0000 (02:56 +0800)]
b=24037 Using cfs_curproc_umask() instead of reference directly.


10 years agob=22755 Don't consume grant twice on recoverable resend
Vitaly Fertman [Tue, 14 Dec 2010 23:07:51 +0000 (02:07 +0300)]
b=22755 Don't consume grant twice on recoverable resend


10 years agob=22556 should check # of remained RPCs before aborting
Vitaly Fertman [Tue, 14 Dec 2010 23:00:07 +0000 (02:00 +0300)]
b=22556 should check # of remained RPCs before aborting


lstcon_rpc_trans_postwait() calls lstcon_rpc_trans_abort() only when the
transaction is timeout, so if we got "end_session" to interrupt waiting
on transaction, then we can hit ASSERTION(crpc->crp_stamp != 0).

10 years agob=23003 sanity test is added, checks how we work with timestamps before Jan 1 1970
Vladimir Saveliev [Mon, 13 Dec 2010 19:51:45 +0000 (22:51 +0300)]
b=23003 sanity test is added, checks how we work with timestamps before Jan 1 1970


10 years agob=24270 need to mkdir mntpt before mount
minhdiep [Mon, 13 Dec 2010 17:01:05 +0000 (10:01 -0700)]
b=24270 need to mkdir mntpt before mount


10 years agob=24198 set strictatime to default.
yangsheng [Mon, 13 Dec 2010 19:05:33 +0000 (03:05 +0800)]
b=24198 set strictatime to default.

RELATIME set to default in upstream kernel. We set strictatime when
relatime or noatime not given explicitly to keep update atime.


10 years agob=24175 avoid reporting 'No such file or directory' error
Vladimir Saveliev [Thu, 9 Dec 2010 20:09:39 +0000 (23:09 +0300)]
b=24175 avoid reporting 'No such file or directory' error


10 years agob=24037 avoid conflicting DCACHE_LUSTRE_INVALID
Andreas Dilger [Tue, 14 Dec 2010 08:21:45 +0000 (01:21 -0700)]
b=24037 avoid conflicting DCACHE_LUSTRE_INVALID

In newer kernels our definition of DCACHE_LUSTRE_INVALID needs
to be changed to avoid conflictint with other DCACHE flags.
Use a high value to avoid conflict for a long time, maybe until
we can actually get rid of the need for it entirely.


10 years agob=24037 Include backing-dev.h header
Brian Behlendorf [Thu, 2 Dec 2010 20:20:57 +0000 (12:20 -0800)]
b=24037 Include backing-dev.h header

With ldiskfs disabled we fail to include the backing-dev.h header
which is required for the lustre_sb_info structure.  Presumably,
this was not as issue previously because it was included through
one of the ldiskfs headers.  The backing-dev.h header exists all
the way back to 2.6.0 kernels so I have included it unconditionally
in the linux/linux-fs.h header.  This header is included as part
of libcfs.h which is in turn included by lustre_disk.h.

  In file included from
  /home/behlendo/src/git/lustre-kdmu/lustre/ofd/ofd_internal.h:12, from
  error: field ‘bdi’ has incomplete type

Signed-off-by: Brian Behlendorf <behlendorf1@llnl.gov>

10 years agob=23821 Limit bio_alloc() to BIO_MAX_PAGES iovecs.
Andreas Dilger [Tue, 14 Dec 2010 07:37:39 +0000 (00:37 -0700)]
b=23821 Limit bio_alloc() to BIO_MAX_PAGES iovecs.

Fix logic error when patch was originally landed from b=9945.


10 years agob=22176 Add .sync_fs super block handler
Rahul Deshmukh [Tue, 14 Dec 2010 06:19:09 +0000 (11:49 +0530)]
b=22176 Add .sync_fs super block handler


This patch adds .sync_fs super block handler i.e.
on issuing sync command, the dirty loi's are marked
urgent and synced.

10 years agob=24239 use SAMPLE_FILE instead of termcap
Elena Gryaznova [Wed, 8 Dec 2010 19:17:53 +0000 (22:17 +0300)]
b=24239 use SAMPLE_FILE instead of termcap


10 years agob=24230 always allow to handle normal request for liblustre
Vitaly Fertman [Fri, 10 Dec 2010 12:53:44 +0000 (15:53 +0300)]
b=24230 always allow to handle normal request for liblustre


as liblustre is 1-threaded we should not prevert normal requests
from being processed if only 1 thread is left

10 years agob=24055 a patch to detect if quota is turned on properly
Landen [Fri, 10 Dec 2010 16:47:38 +0000 (00:47 +0800)]
b=24055 a patch to detect if quota is turned on properly


10 years agob=24039 lfs setstripe --pool broken
Elena Gryaznova [Wed, 8 Dec 2010 13:35:29 +0000 (16:35 +0300)]
b=24039 lfs setstripe --pool broken


fix llapi_search_fsname() to handle relative pathname

10 years agob=24037 Remove iopen patch.
yangsheng [Wed, 8 Dec 2010 18:50:03 +0000 (02:50 +0800)]
b=24037 Remove iopen patch.


10 years agob=23961 fix for setup with several network interfaces
Elena Gryaznova [Wed, 8 Dec 2010 18:41:08 +0000 (21:41 +0300)]
b=23961 fix for setup with several network interfaces


- metadata-updates fix for setup when several interfaces are UP on host;
  hostname could be assigned to IP which is different from lnet network used,
  the hostname-s of NODES_TO_USE are now stored in HOSTS
- new SHUTDOWN_ATTEMPTS: the tunable number of attepts to shutdown node
- shutdown_node_hard () fix: do not call "power off" each time,
  wait that the node is not pingable before the next "power off" attempt
- unused check_port() is removed

10 years agob=23848 mdc_clear_open_replay_data() ASSERTION(mod != LP_POISON && mod != NULL)
Vitaly Fertman [Fri, 10 Dec 2010 13:01:42 +0000 (16:01 +0300)]
b=23848 mdc_clear_open_replay_data() ASSERTION(mod != LP_POISON && mod != NULL)

it is possible to have no md_open_data on close for some recovery cases.
also avoid ELC for recovery tests which rely on lost blocking AST rpc.

10 years agob=23827 fix brw_stats issue
Cliff White [Fri, 10 Dec 2010 07:13:45 +0000 (23:13 -0800)]
b=23827 fix brw_stats issue


Correct issue with brw_stats - fix typo

10 years agob=23820 Handle unsent requests with rq_net_err in ptlrpc_check_set()
Vitaly Fertman [Fri, 10 Dec 2010 23:14:33 +0000 (02:14 +0300)]
b=23820 Handle unsent requests with rq_net_err in ptlrpc_check_set()


ptlrpc_check_set()) ASSERTION(req->rq_phase == RQ_PHASE_BULK) failed

10 years agob=23781 fix obdo leak issue
hongchao.zhang [Fri, 10 Dec 2010 12:40:25 +0000 (20:40 +0800)]
b=23781 fix obdo leak issue

in ll_set_attr_raw, if ll_extent_lock call fails, the newly allocated
obdo has not been freed before return


10 years agob=23701 uninline part of ptlrpc_main to reduce stack usage
Andrew Perepechko [Fri, 10 Dec 2010 17:36:13 +0000 (20:36 +0300)]
b=23701 uninline part of ptlrpc_main to reduce stack usage

i=Oleg Drokin
i=Alexander Zarochentsev

10 years agob=23193 Land ISER performance patch for Sundance
Zhiqi Tao [Thu, 2 Dec 2010 05:50:28 +0000 (13:50 +0800)]
b=23193 Land ISER performance patch for Sundance

The purpose of this patch is to make ISER support 1MB RDMAs instead of splitting 2 512K RDMAs and increase the maximum number of ISCSI commands.


10 years agob=23277 Increase kernel log buffer size to 256k
Cliff White [Mon, 6 Dec 2010 18:29:32 +0000 (10:29 -0800)]
b=23277 Increase kernel log buffer size to 256k


Change CONFIG_LOG_BUF_SHIFT to 18 (256k) for ia64/x84_64

10 years agob=21804 protect rpc by refcount in ptlrpc_server_handle_req_in()
Vitaly Fertman [Tue, 7 Dec 2010 22:00:13 +0000 (01:00 +0300)]
b=21804 protect rpc by refcount in ptlrpc_server_handle_req_in()


ptlrpc_hpreq_init() inserts rpc to the export list and by the time
of ptlrpc_server_request_add() it could be already handled and
released. To not lose request in between, take an extra reference
on the request.

10 years agob=24243 Invalid WNC intent lock makes server LBUG
Matt Wu [Tue, 7 Dec 2010 11:25:54 +0000 (19:25 +0800)]
b=24243 Invalid WNC intent lock makes server LBUG


do not grant PDO locks for non-dir objects, but return -ENOTDIR instead.

10 years agob=20563 fix fid_flatten32().
Eric Mei [Mon, 6 Dec 2010 22:50:07 +0000 (15:50 -0700)]
b=20563 fix fid_flatten32().


10 years agob=20563 fix fid_flatten().
Eric Mei [Mon, 6 Dec 2010 22:49:54 +0000 (15:49 -0700)]
b=20563 fix fid_flatten().


10 years agoNeeded to create a new tag for 2.1.0 build 09. For some reason the v2_0_58_0
Terry Rutledge [Sat, 4 Dec 2010 05:19:03 +0000 (21:19 -0800)]
Needed to create a new tag for 2.1.0 build 09.  For some reason the
previous tag was not being propagated out, so v2_0_58_0 is being
created for this build.

10 years agoUpdated for 2.1.0 build 09.
Terry Rutledge [Sat, 4 Dec 2010 05:14:57 +0000 (21:14 -0800)]
Updated for 2.1.0 build 09.