Whamcloud - gitweb
anserper [Wed, 19 Aug 2009 15:48:44 +0000 (15:48 +0000)]
b=16267
i=Vitaly Fertman
liblustre man pages
anserper [Mon, 17 Aug 2009 19:20:54 +0000 (19:20 +0000)]
ChangeLog
anserper [Mon, 17 Aug 2009 17:43:14 +0000 (17:43 +0000)]
revert incidental commit
anserper [Mon, 17 Aug 2009 17:38:36 +0000 (17:38 +0000)]
b=20008
i=Oleg Drokin
i=Alex Tomas
pin the last page before starting transaction in setattr to avoid GFP_FS allocation under transaction
isaac [Wed, 12 Aug 2009 17:53:54 +0000 (17:53 +0000)]
b=19782,i=isaac:
- fixed a typo in LST sample test script.
shadow [Mon, 10 Aug 2009 09:32:29 +0000 (09:32 +0000)]
correctly handle get_lock vs lock cancel race
Branch b1_8
b=20162
i=gree
i=panda
bobijam [Mon, 10 Aug 2009 02:27:28 +0000 (02:27 +0000)]
Branch b1_8
b=19640
i=johann
i=adilger
Avoid calling vfs_link upon a file with link count of 0 in mds_orphan_add_link().
dzogin [Sun, 9 Aug 2009 17:16:57 +0000 (17:16 +0000)]
b=20200
i=andrew.perepechko
i=tianzy
Description: Enhancement for lfs(1) command to use numeric uid/gid.
dzogin [Sun, 9 Aug 2009 17:01:39 +0000 (17:01 +0000)]
b=18674
i=alexey.lyashkov
i=dmitry.zogin
Modified Files:
Tag: b1_8
lustre/ChangeLog lustre/ost/ost_handler.c
Description: Do not start bulk transfer, if client tried to reconnect.
Details : Do not start bulk transfer, if client tried to reconnect,
but failed due to active rpcs. Check for exp_abort_active_req
has been implemented.
wangdi [Sat, 8 Aug 2009 13:27:30 +0000 (13:27 +0000)]
Branch b1_8
Fix for basetest to avoid skipping test like test_300ca when
runing sanity with ONLY=300
b=20385
o=adilger (patch from Andreas)
i=wangdi
fanyong [Sat, 8 Aug 2009 12:40:59 +0000 (12:40 +0000)]
Branch b1_8
b=18902
i=alexey.lyashkov
i=vladimir.saveliev
1) do not reset statahead if conflict was found when opendir
2) do not release "sai" if someone else want to use it for race condition
3) patch from bug 20139
4) increase timeout value for racer killing threads to process client evicted case
There are race conditions for statahead:
1) just before statahead thread is triggered, the first item of the dentry many moved to other
directory, which will cause the statahead thread to be triggered for different directory, and then
cause the original "ls -l" thread which is waiting for the expected directory can not be waken up.
2) when some items are in statahead, such items maybe move to other directory, then the interpret
call maybe find different dentry under such changed directory, and left the original original "ls
-l" thread which is waiting for the expected directory can not be waken up.
3) lli_fid should be set just after vfs inode created on client-side, to prevent the same inode was
created multiple times by statahead thread and other stat thread.
bobijam [Mon, 3 Aug 2009 01:47:57 +0000 (01:47 +0000)]
Branch b1_8
b=20219
o=johann
i=andrew.perepechko (panda)
i=alexander.zarochentsev (zam)
Release lock ref for lockless i/o if allocation failure.
dzogin [Sun, 2 Aug 2009 13:55:49 +0000 (13:55 +0000)]
Branch b1_8
b=20321
i=dmitry.zogin
i=adilger
Description: Deadlock between filter_destroy() and filter_commitrw_write().
Details : filter_destroy() does not hold the DLM lock over the whole
operation. If the DLM lock is dropped, filter_commitrw() can go
through, causing the deadlock between page lock and i_mutex.
adilger [Sat, 1 Aug 2009 19:00:20 +0000 (19:00 +0000)]
Branch b1_8
Add ChangeLog entry
b=18539
anserper [Sat, 1 Aug 2009 16:49:50 +0000 (16:49 +0000)]
added a comment wrt unaligned dio and write grants
anserper [Fri, 31 Jul 2009 16:21:57 +0000 (16:21 +0000)]
b=20314
i=Johann Lombardi
i=Oleg Drokin
properly handle rc<0 from osc_brw_prep_request
anserper [Fri, 31 Jul 2009 16:03:55 +0000 (16:03 +0000)]
b=20317
i=Johann Lombardi
i=Alexander Zarochentsev
do not update kms without extent locks on DIO path
grev [Fri, 31 Jul 2009 15:33:43 +0000 (15:33 +0000)]
b=16476
i=Brian
wait_for_host fix for nodes which have no lustre-tests installed
grev [Fri, 31 Jul 2009 15:03:39 +0000 (15:03 +0000)]
b=19887
i=Nathan
test_26b: increase wait client evicted value
yangsheng [Fri, 31 Jul 2009 10:38:09 +0000 (10:38 +0000)]
Branch b1_8
b=19808
Add a entry of ChangLog to notice new kernel support.
shadow [Fri, 31 Jul 2009 10:28:03 +0000 (10:28 +0000)]
correctly handling request reference in error cases.
Branch b1_8
b=20122
i=adilger
i=panda
anserper [Thu, 30 Jul 2009 21:17:07 +0000 (21:17 +0000)]
b=18362
i=Johann Lombardi
cleanup HAVE_I_ALLOC_SEM ifndef as we dont have any autoconf test for it anymore
grev [Thu, 30 Jul 2009 19:25:24 +0000 (19:25 +0000)]
b=20259
i=Brian
get rid of useless rm PI*
anserper [Thu, 30 Jul 2009 16:59:18 +0000 (16:59 +0000)]
b=18801
o=Christopher Morrone
i=Andrew Perepechko
i=Johann Lombardi
introduce a Lustre option to switch I/O to direct I/O (LLNL needs this)
shadow [Thu, 30 Jul 2009 10:06:14 +0000 (10:06 +0000)]
checking permission should be part of 'open' part of mds_open, not 'lookup'
part, so server should be set DISP_OPEN_OPEN disposition before perform
permission check.
Also don not need call revalidate dentry if client already have LOOKUP lock.
Branch b1_8
b=17545
i=green
i=panda
tianzy [Thu, 30 Jul 2009 06:22:02 +0000 (06:22 +0000)]
add a changelog entry for 18793
bobijam [Thu, 30 Jul 2009 03:00:19 +0000 (03:00 +0000)]
Branch b1_8
b=19854
Add a changelog entry.
anserper [Wed, 29 Jul 2009 18:09:59 +0000 (18:09 +0000)]
*** empty log message ***
anserper [Wed, 29 Jul 2009 18:03:25 +0000 (18:03 +0000)]
*** empty log message ***
anserper [Wed, 29 Jul 2009 18:00:16 +0000 (18:00 +0000)]
*** empty log message ***
johann [Wed, 29 Jul 2009 17:43:13 +0000 (17:43 +0000)]
Branch b1_8
sync up changelog between b_release_1_8_1 & b1_8.
johann [Wed, 29 Jul 2009 17:21:18 +0000 (17:21 +0000)]
Branch b1_8
b=19785
disable shrink grant feature.
anserper [Wed, 29 Jul 2009 09:03:47 +0000 (09:03 +0000)]
b=16267
i=Johann Lombardi
llapi_file_get_stripe man page
anserper [Wed, 29 Jul 2009 08:53:36 +0000 (08:53 +0000)]
b=16890
i=Johann Lombardi
i=ZhiYong Tian
tiny cleanup of the quota related commands output
vs [Wed, 29 Jul 2009 00:39:45 +0000 (00:39 +0000)]
b=17807
i=h.huang,nikita.danilov
do not return 0 from ll_revalidate_nd in case of error:
invalidated dentry may get recreated and
real_lookup will revalidate agai instead of lookup. That revalidate oopses
derefencing error code in the line:
nd->intent.open.file->private_data = it; because nd->intent.open.file holds
error from first revalidate.
green [Tue, 28 Jul 2009 23:28:19 +0000 (23:28 +0000)]
b=19854
r=johann,bobijam
When a client reconnects from another NID, properly update export nid hash position and ldlm reverse import.
yangsheng [Tue, 28 Jul 2009 12:42:40 +0000 (12:42 +0000)]
Branch b1_8
b=19808
i=adilger, shadow, green
Landed for 2.6.30 kernel patchless support.
yangsheng [Tue, 28 Jul 2009 08:18:51 +0000 (08:18 +0000)]
Branch b1_8
b=19808
i=shadow, yangsheng
Remove obsolete config check.
shadow [Tue, 28 Jul 2009 06:31:04 +0000 (06:31 +0000)]
fix buffer number for acl packing.
Branch b1_8
b=20020
yangsheng [Tue, 28 Jul 2009 06:29:59 +0000 (06:29 +0000)]
Branch b1_8
b=19847
Update kernel to SLES10 SP2 2.6.16.60-0.39.3.
bobijam [Mon, 27 Jul 2009 09:21:56 +0000 (09:21 +0000)]
Branch b1_8
fix 2.6-rhel5.series.
bobijam [Mon, 27 Jul 2009 01:37:03 +0000 (01:37 +0000)]
Branch b1_8
b=20164
i=adilger
i=grev
When echo client attached to OST device, o_brw_async() ops is not supported,
we turn to o_brw() instead.
adilger [Fri, 24 Jul 2009 22:54:19 +0000 (22:54 +0000)]
Branch b1_8
Remove ll_getstripe_info.c file, which is now part of "lfs getstripe".
b=18442
i=girish.shilamkar
i=rread
anserper [Fri, 24 Jul 2009 20:05:18 +0000 (20:05 +0000)]
b=20142
i=Oleg Drokin
i=Johann Lombardi
remove racy i_size_reads from filter_preprw_read; those could lead to certain inconsistencies during page preparations, assertions in ost_brw_read following them
grev [Fri, 24 Jul 2009 17:59:52 +0000 (17:59 +0000)]
b=20124
i=Jian.Yu
wait_for host for HARD failure mode
shadow [Fri, 24 Jul 2009 14:00:10 +0000 (14:00 +0000)]
looks i was land old version of patch :(
commit right version.
Branch b1_8
b=20020
tianzy [Fri, 24 Jul 2009 06:26:34 +0000 (06:26 +0000)]
Branch b1_8
make test_11 of sanity_quota.sh invalid
b=19835
i=johann
green [Thu, 23 Jul 2009 19:19:48 +0000 (19:19 +0000)]
b=19860
r=adilger,shadow
get rid of ptlrpc_daemonize
remove unneeded exports
grev [Thu, 23 Jul 2009 13:43:15 +0000 (13:43 +0000)]
b=19415
i=Robert.Read
restart_client_loads fix related to new client load var name
shadow [Thu, 23 Jul 2009 13:19:46 +0000 (13:19 +0000)]
correctly shrink reply for avoid send too big message to client.
Branch b1_8
b=20020
i=adilger
i=tappro
anserper [Thu, 23 Jul 2009 10:14:41 +0000 (10:14 +0000)]
b=19778
a=24644
i=Alexey Lyashkov
i=ZhiYong Tian
do not loop infinitely in client_quota_ctl, retry several times after timeout
bobijam [Thu, 23 Jul 2009 09:32:22 +0000 (09:32 +0000)]
Branch b1_8
b=19854
o=johann
i=oleg.drokin (green)
i=robert.read
Check for conflicting UUID only on initial connect.
liangzhen [Thu, 23 Jul 2009 04:45:34 +0000 (04:45 +0000)]
Branch b1_8
ordering write_lock to avoid deadlock
b=19518
i=adilger
i=zhenyu.xu
tianzy [Wed, 22 Jul 2009 13:41:11 +0000 (13:41 +0000)]
Branch b1_8
fix "Cannot mount mdt or osts if --param xxx.quota_type in mkfs command"
b=20236
i=johann
brian [Tue, 21 Jul 2009 19:32:53 +0000 (19:32 +0000)]
b=19248
o=rread
i=brian
Fix mpicc compile error. I guess mpicc is using some of gcc's more strict
flags.
anserper [Tue, 21 Jul 2009 14:03:56 +0000 (14:03 +0000)]
b=18948
i=Nathan Rutman
i=Alexander Zarochentsev
tunably recovery patch
tappro [Sat, 18 Jul 2009 08:10:16 +0000 (08:10 +0000)]
Branch b1_8
b=19885
i=adilger
i=shadow
Handle downgrade case from 2.0 to 1.8 and evict all client in case of it
grev [Fri, 17 Jul 2009 16:28:02 +0000 (16:28 +0000)]
b=19312
i=Robert.Read
new test: test_parallel_grouplock
grev [Fri, 17 Jul 2009 15:59:34 +0000 (15:59 +0000)]
b=19312
i=Brian
i=Oleg
port parallel_grouplock.c from cmd3
grev [Fri, 17 Jul 2009 15:33:20 +0000 (15:33 +0000)]
b=20179
i=Robert.Read
process TESTSUITE log to fail if error
yangsheng [Fri, 17 Jul 2009 13:30:27 +0000 (13:30 +0000)]
Branch b1_8
b=19615
i=shadow, johann
Disable server build for rhel4 & sles9.
grev [Thu, 16 Jul 2009 15:37:43 +0000 (15:37 +0000)]
b=19415
i=Robert.Read
fixes related to new load var
anserper [Thu, 16 Jul 2009 14:27:01 +0000 (14:27 +0000)]
b=18801
the lost part of the previous commit
grev [Thu, 16 Jul 2009 10:40:59 +0000 (10:40 +0000)]
b=19215
i=Robert.Read
verbose do_nodes
do_node print HOSTNAME for no_dsh
tianzy [Thu, 16 Jul 2009 09:55:52 +0000 (09:55 +0000)]
Branch b1_8
fix a typo for LQUOTA_QAQ_CEATE_LQS
b=20031
anserper [Thu, 16 Jul 2009 09:44:20 +0000 (09:44 +0000)]
add time.h include to avoid warning
anserper [Thu, 16 Jul 2009 09:03:28 +0000 (09:03 +0000)]
b=18801
i=Oleg Drokin
i=Alexander Zarochentsev
atime, mtime and additional POSIX compliance stuff for direct I/O and lockless I/O
girish [Thu, 16 Jul 2009 07:35:57 +0000 (07:35 +0000)]
b=19872
i=adilger
i=dmitry.zogin
fix test 130e to write 64k blocks with 128k stripes so FIEMAP can work properly across mixed platform clusters
tianzy [Thu, 16 Jul 2009 06:40:12 +0000 (06:40 +0000)]
Branch b1_8
Fix group quota "too many blocks" OSS crashes.
In filter_commitrw_write():
1. lquota_chkquota();
2. obdo_from_inode();
3. lquota_pending_commit();
2. has changed uid and gid in this case, then caused
lquota_chkquota() and lquota_pending_commit() to use
different uid and gid. That is the root of the bug.
b=18793
i=yong.fan
i=panda
zhanghc [Thu, 16 Jul 2009 05:22:21 +0000 (05:22 +0000)]
b=19870
fix a bug of processing export->exp_rpc_count during recovery
i=tappro@sun.com
i=alexey.lyashkov@sun.com
zhanghc [Thu, 16 Jul 2009 04:46:30 +0000 (04:46 +0000)]
b=18539
don't allocate new files on degraded OSTs
i=adilger@sun.com
i=hongchao.zhang@sun.com
shadow [Thu, 16 Jul 2009 04:46:01 +0000 (04:46 +0000)]
revert attachment 24367 due compilation problems.
Branch b1_8
b=18801
yujian [Thu, 16 Jul 2009 02:01:58 +0000 (02:01 +0000)]
Branch b1_8
b=20047
i=robert.read
i=yong.fan
reset debug level if lustre was already mounted
yujian [Thu, 16 Jul 2009 01:47:12 +0000 (01:47 +0000)]
Branch b1_8
b=19432
i=robert.read
i=grev
1) move MPIRUN and MPI_USER from ncli.sh to local.sh
2) remove MPIBIN
3) redirect the stderr of "which mpirun" to /dev/null
bobijam [Thu, 16 Jul 2009 01:32:15 +0000 (01:32 +0000)]
Branch b1_8
b=17682
i=alexey.lyashkov (shadow)
i=sheng.yang
DLM debug message improvement.
w.li [Thu, 16 Jul 2009 01:03:30 +0000 (01:03 +0000)]
Branch b1_8
b=19999
i=grev
i=johann
Remove PID reading from multiop_bg_pause. Turn to $! for multiop PID in
runmultiop_bg_pause. The result looks more consistent with HEAD.
yangsheng [Wed, 15 Jul 2009 15:27:53 +0000 (15:27 +0000)]
Branch b1_8
b=19205
i=adilger, Jacques-Charles
'lfs df' for pool support.
fanyong [Wed, 15 Jul 2009 15:04:51 +0000 (15:04 +0000)]
Branch b1_8
b=20002
i=tianzy
i=robert.read
Check "auto_quota_on()" return value before "build_lqs()" called.
grev [Wed, 15 Jul 2009 13:35:12 +0000 (13:35 +0000)]
b=19150
i=Robert.Read
cleanup: work in TESTDIR instead of DIR
anserper [Wed, 15 Jul 2009 13:28:47 +0000 (13:28 +0000)]
b=19544
i=Robert Read
i=Dmitry Zogin
move to static on-stack allocations
dzogin [Wed, 15 Jul 2009 13:21:11 +0000 (13:21 +0000)]
b=19529
i=andrew.perepechko
i=adilger
Description: Avoid deadlock for local client writes
Details : Use new OBD_BRW_MEMALLOC flag to notify OST about writes in the
memory freeing context. This allows OST threads to set the
PF_MEMALLOC flag on task structures in order to allocate memory
from reserved pools and complete IO.
Use GFP_HIGHUSER for OST allocations for non-local client writes,
so that the OST threads generate memory pressure and allow
inactive pages to be reclaimed.
w.li [Wed, 15 Jul 2009 13:19:41 +0000 (13:19 +0000)]
Branch b1_8
b=19980
i=alexey.lyashkov
i=andrew.perepechko
Subtract evicted clients from "completed_clients" in recovery_status.
anserper [Wed, 15 Jul 2009 12:47:33 +0000 (12:47 +0000)]
b=18801
i=Oleg Drokin
i=Alexander Zarochentsev
atime, mtime and addition POSIX compliance stuff for direct I/O and lockless I/o
anserper [Wed, 15 Jul 2009 12:41:15 +0000 (12:41 +0000)]
b=16267
added the man page to the manpage list
anserper [Wed, 15 Jul 2009 12:27:00 +0000 (12:27 +0000)]
b=16267
i=ZhiYong Tian
a man page for llapi_file_open and llapi_file_create
anserper [Wed, 15 Jul 2009 12:23:39 +0000 (12:23 +0000)]
file llapi_file_open.3 was added on branch b1_8 on 2009-07-15 12:27:00 +0000
grev [Wed, 15 Jul 2009 11:47:33 +0000 (11:47 +0000)]
b=20074
i=Alexey.Lyashkov
i=Robert.Read
process tar's rc correctly
shadow [Tue, 14 Jul 2009 06:23:12 +0000 (06:23 +0000)]
first part of new dcache handling (bug 16654)
make DCACHE_LUSTRE_INVALID is always defined.
Branch b1_8
b=16417
i=green
i=johann
grev [Thu, 9 Jul 2009 08:44:23 +0000 (08:44 +0000)]
b=20094
i=Alexander.Zarochentsev (Zam)
test_121: ignore useless dd output
nathan [Wed, 8 Jul 2009 19:51:03 +0000 (19:51 +0000)]
b=19889
i=rread
i=manoj
return error codes from pool_cmd (and others)
shadow [Wed, 8 Jul 2009 03:45:44 +0000 (03:45 +0000)]
lock ordering violation between &cli->cl_sem and _lprocfs_lock
Branch b1_8
b=18380
i=jay
i=deen
brian [Tue, 7 Jul 2009 19:26:09 +0000 (19:26 +0000)]
b=20055
i=green
i=zhanghc
o=johann
Fix for connectathon, NFSv4: ASSERTION(last_discon == NULL) failed
zhanghc [Tue, 7 Jul 2009 16:09:41 +0000 (16:09 +0000)]
b=16774
cache the extent lock gotten before doing I/O in "ll_file_aio_read"
and "ll_file_aio_write", then "ll_read_page", "ll_read_ahead_page"
and "ll_writepage" will save much time spent on "ldlm_lock_match"
(in cache_add_extent) to find the appropriate extent lock
1, using current->journal_info to store the locks(struct ll_thread_data)
2, replace o_reget_short_lock, o_release_short_lock with o_get_lock in obd_ops
i=johann@sun.com
i=oleg.drokin@sun.com
grev [Mon, 6 Jul 2009 13:35:23 +0000 (13:35 +0000)]
b=19737
i=Robert.Read
acc-sm, t-f NFSCLIENT changes
girish [Sat, 4 Jul 2009 13:21:36 +0000 (13:21 +0000)]
b=20054
o=rread
i=girish
i=eric.mei
print out stripe data. regression caused by 18442
anserper [Fri, 3 Jul 2009 22:40:50 +0000 (22:40 +0000)]
b=16890
i=Elena
SHOW_GROUP_INFO call consisting of two commands was sometimes interpreted as a long command with
zam [Fri, 3 Jul 2009 15:04:45 +0000 (15:04 +0000)]
Branch b1_8
b=19963
i=johann
i=andrew.perepechko
remove temporary file upon successful completion of replay-single test 13;
check whether it is deletable and can it cause an EEXIST failure in
replay-dual test 13 (see bug 19963).
zam [Fri, 3 Jul 2009 14:55:46 +0000 (14:55 +0000)]
Branch b1_8
b=19963
i=johann
mds_lookup: misplaced EXIT debug marker.
girish [Fri, 3 Jul 2009 05:39:34 +0000 (05:39 +0000)]
b=18442
i=girish
o=rread
fix test 200d to check for new output pattern
anserper [Thu, 2 Jul 2009 21:59:21 +0000 (21:59 +0000)]
b=19980
i=Andrew Perepechko
i=Alexey Lyashkov
o=Mike Pershin
fix for recovery messages