Whamcloud - gitweb
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
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
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.
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.
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.
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.
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.
vitaly [Tue, 27 Oct 2009 12:25:52 +0000 (12:25 +0000)]
Branch HEAD
b=20098
reverting patches as they lead to some problems
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
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
robert.read [Fri, 23 Oct 2009 19:26:30 +0000 (19:26 +0000)]
Branch HEAD
b=20790
i=nathan
Add ltrack_stats to HEAD.
bobijam [Fri, 23 Oct 2009 06:09:35 +0000 (06:09 +0000)]
Branch HEAD
b=20149
add more debug info.
stinson1947 [Thu, 22 Oct 2009 20:14:21 +0000 (20:14 +0000)]
Updaged version string for 2.0.0 build 29.
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.
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
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.
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.
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
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.
ericm [Thu, 22 Oct 2009 03:15:20 +0000 (03:15 +0000)]
branch: HEAD
wait before test file size.
b=20801
r=rread
fanyong [Thu, 22 Oct 2009 02:38:49 +0000 (02:38 +0000)]
Branch HEAD
b=20811
i=tianzy
i=robert.read
Do not try to acquire quota from master repeatly if mismatched quota configure was found.
grev [Tue, 20 Oct 2009 14:42:38 +0000 (14:42 +0000)]
b=20863
i=Brian
sanity test_200* cleanup
vitaly [Tue, 20 Oct 2009 12:37:38 +0000 (12:37 +0000)]
Branch HEAD
b=20098
i=green
i=tappro
pin open rpc properly in memory
vitaly [Tue, 20 Oct 2009 12:35:03 +0000 (12:35 +0000)]
Branch HEAD
b=20098
i=tappro
i=green
adding refcount to md_open_data
jxiong [Tue, 20 Oct 2009 09:20:45 +0000 (09:20 +0000)]
b=20824
r=wangdi,eric.mei
Reimplement cl_lock_peek().
shadow [Tue, 20 Oct 2009 05:25:58 +0000 (05:25 +0000)]
remove assert for replay data.
Branch HEAD
b=21044
i=zam
dzogin [Tue, 20 Oct 2009 03:21:38 +0000 (03:21 +0000)]
Branch HEAD
b=20302
i=johann
----------------------------------------------------------------------
Modified Files:
lustre/ChangeLog lustre/ptlrpc/client.c
---------------------------------------------------------------------
Description: Fix in ptlrpc_expire_one_request() to print the signed time
difference.
fanyong [Tue, 20 Oct 2009 02:05:09 +0000 (02:05 +0000)]
Branch HEAD
b=15691
i=eric.mei
i=jian.chen
Repeating test when found performance dropped to reduce the affect of external events.
ericm [Mon, 19 Oct 2009 21:21:48 +0000 (21:21 +0000)]
branch: HEAD
fix ost buil read/write busy waiting.
b=20444
r=rread
r=wangdi
zam [Mon, 19 Oct 2009 20:05:53 +0000 (20:05 +0000)]
Branch HEAD
b=20136
i=johann
i=alexey.lyashkov
mdt_sendpage: exit l_wait_event and abort bulk on reconnect.
isaac [Thu, 15 Oct 2009 04:14:37 +0000 (04:14 +0000)]
b=20818,i=rread,i=liang:
- user options libcfs_console_(max|min)_delay ignored by libcfs.
isaac [Thu, 15 Oct 2009 04:05:36 +0000 (04:05 +0000)]
b=17103,i=nikita,i=green,i=isaac:
- Rewrite ptlrpc_queue_wait() using sets (from rread).
zhanghc [Thu, 15 Oct 2009 02:49:09 +0000 (02:49 +0000)]
b=20584
set "IT_CREAT" in the default intent(oit) in function
ll_file_open(llite/file.c) if "it_flags" contains O_CREAT
i=johann@sun.com
i=alexey.lyashkov@sun.com
shadow [Thu, 15 Oct 2009 02:42:36 +0000 (02:42 +0000)]
simplify disconnecting client export and avoid sending requests after client is
dead.
Branch HEAD
b=19557
i=tappro
i=shadow
jxiong [Thu, 15 Oct 2009 02:07:16 +0000 (02:07 +0000)]
b=20394
r=tappro,rread
Fix an over-strict assertion
fanyong [Thu, 15 Oct 2009 01:45:51 +0000 (01:45 +0000)]
Branch HEAD
b=21028
i=tappro
i=robert
Initialize "req->rq_timed_list" anyway when it is unlinked from the list.
grev [Wed, 14 Oct 2009 21:55:27 +0000 (21:55 +0000)]
b=20719
i=Alexey.Lyashkov
i=Robert.Read
test_68a: load llite_loop if not loaded
module_loaded fn
brian [Wed, 14 Oct 2009 21:46:13 +0000 (21:46 +0000)]
Add missing changelog entry.
ericm [Wed, 14 Oct 2009 21:42:13 +0000 (21:42 +0000)]
branch: HEAD
fix the OST grant number synchronization.
b=20278
r=wangdi
r=vs
johann [Wed, 14 Oct 2009 20:59:10 +0000 (20:59 +0000)]
Branch HEAD
b=19345
i=shadow
i=wangdi
use shrink_dcache_sb() instead of shrink_dcache_parent() since
the latter can hold the cpu for a long time.
johann [Tue, 13 Oct 2009 21:01:57 +0000 (21:01 +0000)]
Branch b1_8
b=19763
i=adilger
i=girish
speed up quotacheck. take into account itable_unused_count.
zam [Tue, 13 Oct 2009 11:40:48 +0000 (11:40 +0000)]
Branch HEAD
b=20740
i=alexey.lyashkov
i=andrew.perepechko
replace client_obd_list_is_locked and the corresponding assertions
in the osc code by ASSERT_SPIN_LOCKED() because the first doesn't
work correctly w/o CONFIG_SMP on.
zam [Tue, 13 Oct 2009 11:33:00 +0000 (11:33 +0000)]
Branch HEAD
b=20642
i=zhen.liang
i=alexander.zarochentsev
fix for already freed rs object access in commit_replies:
1 . Introduce rs->rs_committed flag to indicate that commit_replies() has done
all access to the rs object.
2. check rs->rs_committed instead of list_empty_careful() to sure that
commit_replies does not access the rs object in parallel with
ptlrpc_handle_rs and it is safe not to use exp_uncommitted_replies_lock.
tappro [Tue, 13 Oct 2009 07:07:29 +0000 (07:07 +0000)]
Branch HEAD
b=20887
i=nathan
i=rread
use AT naturally for recovery requests. Their deadline is set to at_extra value
so early replay will be set back if request is in recovery queue.
tappro [Tue, 13 Oct 2009 06:02:55 +0000 (06:02 +0000)]
Branch HEAD
b=19884,20274
i=vitaly
i=rread
wait for mds-ost sync and for destroy is done before calculating free space
after tests
nathan [Mon, 12 Oct 2009 23:26:13 +0000 (23:26 +0000)]
b=20766
i=adilger
i=thomas.leibovici@cea.fr
changelog records using epoch time instead of CURRENT_SECONDS
nathan [Mon, 12 Oct 2009 23:20:39 +0000 (23:20 +0000)]
b=20689
i=manoj
i=yujian
use the correct osc name for add_failnid
eeb [Mon, 12 Oct 2009 19:03:12 +0000 (19:03 +0000)]
* fixed comment empty LN_CONFIG_USERSPACE macro definition
eeb [Mon, 12 Oct 2009 19:01:16 +0000 (19:01 +0000)]
* Defined empty LN_CONFIG_USERSPACE macro to avoid a warning when
running ./configure sincew it's called from build scripts shared
by all branches.
robert.read [Fri, 9 Oct 2009 23:14:52 +0000 (23:14 +0000)]
b=18857
Skip both 33a and 33b.
brian [Fri, 9 Oct 2009 13:47:37 +0000 (13:47 +0000)]
b=19720
i=robert.read
i=dmitry.zogin
Clean up various warnings stemming from unused variables and unchecked
function return values.
eeb [Fri, 9 Oct 2009 12:24:19 +0000 (12:24 +0000)]
* Fixed comment describing how cl_lock_operations::clo_modify() is called
grev [Fri, 9 Oct 2009 11:33:23 +0000 (11:33 +0000)]
b=18150
i=Johann
tets_22: correct check of mount client without ost
jxiong [Fri, 9 Oct 2009 11:21:30 +0000 (11:21 +0000)]
b=20826
r=wangdi,jay
Rework the patch of bug 20305.
eeb [Fri, 9 Oct 2009 09:25:33 +0000 (09:25 +0000)]
- Removed confusing comment about an invalid assertion that was removed.
stinson1947 [Thu, 8 Oct 2009 23:31:52 +0000 (23:31 +0000)]
Updated for 2.0 build 28.
ericm [Thu, 8 Oct 2009 21:51:14 +0000 (21:51 +0000)]
branch: HEAD
really send setattr cookies to OSTs.
b=20756
r=wangdi
r=tappro
ericm [Thu, 8 Oct 2009 21:06:14 +0000 (21:06 +0000)]
branch: HEAD
use separate lprocfs counter for softirq context.
b=20650
r=eeb
r=rread
ericm [Thu, 8 Oct 2009 21:00:29 +0000 (21:00 +0000)]
branch: HEAD
speed up mdsrate test.
b=20501
r=rread
r=grev
panda [Thu, 8 Oct 2009 20:39:20 +0000 (20:39 +0000)]
b=19113
o=Johann Lombardi
i=Nathan Rutman
i=Andrew Perepechko
avoid stopping race with ptlrpc thread
deshmukh [Thu, 8 Oct 2009 06:49:19 +0000 (06:49 +0000)]
Fixed a typo.
robert.read [Wed, 7 Oct 2009 17:48:59 +0000 (17:48 +0000)]
Branch HEAD
b=18857
Disable replay-single test 33b.
maxim [Wed, 7 Oct 2009 10:05:39 +0000 (10:05 +0000)]
b=20784
i=rread
Minor fix of regexp in test_215. Re-enabling test_215. Minor change of test_36f to restore locale settings after test.
vitaly [Wed, 7 Oct 2009 09:54:26 +0000 (09:54 +0000)]
Branch HEAD
b=18629
i=rread
reconnect the client back
johann [Wed, 7 Oct 2009 09:50:02 +0000 (09:50 +0000)]
Branch HEAD
b=20903
i=rread
i=dmitry
Prevent deadlock if obd_zombie_barrier() is called from the zombie thread context.
deshmukh [Wed, 7 Oct 2009 07:47:12 +0000 (07:47 +0000)]
b=20672
i=adilger
i=bzzz
Details:
The pre-allocation tracker "bb_prealloc_nr" was not descrimented
even if pre-allocation was free. This problem is fixed with this
patch. Also some cleanup.
nathan [Tue, 6 Oct 2009 22:45:23 +0000 (22:45 +0000)]
add cvsignore to tests/mpi
nathan [Tue, 6 Oct 2009 22:43:44 +0000 (22:43 +0000)]
b=20603
i=rahul
i=rread
use sync output for changelog so we don't overflow buffer
vitaly [Tue, 6 Oct 2009 19:18:04 +0000 (19:18 +0000)]
Branch HEAD
b=20098
reverting
vitaly [Tue, 6 Oct 2009 13:12:42 +0000 (13:12 +0000)]
Branch HEAD
b=20929
i=tappro
i=rread
dequeue req if needed on error
nathan [Mon, 5 Oct 2009 21:19:26 +0000 (21:19 +0000)]
doc improvements
ericm [Mon, 5 Oct 2009 15:00:05 +0000 (15:00 +0000)]
branch: HEAD
fix quota for the patch of bz19778.
a=tianzy
i=fanyong
i=andrew.perepechko
pravins [Mon, 5 Oct 2009 04:23:27 +0000 (04:23 +0000)]
b=17670
fix build failure.
pravins [Sun, 4 Oct 2009 18:13:14 +0000 (18:13 +0000)]
b=17670
i=adilger
i=girish
add "FID in dirent" support in ext4
pravins [Sun, 4 Oct 2009 18:05:00 +0000 (18:05 +0000)]
b=17670
i=adilger
i=rahul
i=girish
add "FID in dirent" support in ext3
robert.read [Fri, 2 Oct 2009 23:50:08 +0000 (23:50 +0000)]
Branch HEAD
b=20784,20719
Disable test 214 because of bug 20784.
Revert skip_env changes in test 68a,b.
nathan [Fri, 2 Oct 2009 23:47:09 +0000 (23:47 +0000)]
b=20878
i=rread
temporarily disable 5a, 5b until 20878 is fixed
dzogin [Fri, 2 Oct 2009 19:47:03 +0000 (19:47 +0000)]
Branch HEAD
b=20902
i=grev
Modified Files:
lustre/tests/conf-sanity.sh
Description: Conf-sanity test 50g cleanup
eeb [Fri, 2 Oct 2009 12:53:31 +0000 (12:53 +0000)]
* Added make target (mkid) to build idutils database
* Change cscope/ctags/etags targets to skip ldiskfs/linux-stage
directory to avoid duplicate symbol matches.
shadow [Fri, 2 Oct 2009 11:16:49 +0000 (11:16 +0000)]
Branch HEAD
b=19662
i=adilger
i=shadow
grev [Thu, 1 Oct 2009 21:27:03 +0000 (21:27 +0000)]
b=20698
i=Robert.Read
new base for new test_[a-z] tests
grev [Thu, 1 Oct 2009 13:37:14 +0000 (13:37 +0000)]
b=17873
i=Brian
use skip_env instead of skip
vitaly [Thu, 1 Oct 2009 10:03:42 +0000 (10:03 +0000)]
Branch HEAD
b=20098
i=tappro
i=green
mod refcount is added to avoid usage after freeing
johann [Thu, 1 Oct 2009 09:44:06 +0000 (09:44 +0000)]
Branch HEAD
b=20205
i=panda
i=oleg
i=vitaly
don't call ost_rw_prolong_locks() for lockless i/o.
shadow [Thu, 1 Oct 2009 06:18:00 +0000 (06:18 +0000)]
correctly handle too big reply message.
Branch HEAD
b=19526
i=ericm
i=rread
shadow [Thu, 1 Oct 2009 04:24:28 +0000 (04:24 +0000)]
don't try dump ldlm lock buffer for llog cancel requests
and avoid error
LustreError: 2505:0:(layout.c:1469:__req_capsule_get()) @@@ Wrong buffer for field `dlm_req' (1 of 2) in format `LDLM_CALLBACK': 32 vs. 104 (client)
Branch HEAD
b=20349
i=rread
dzogin [Thu, 1 Oct 2009 01:14:07 +0000 (01:14 +0000)]
Branch HEAD
b=19917
----------------------------------------------------------------------
Modified Files:
lustre/ChangeLog lustre/obdfilter/filter_io.c
----------------------------------------------------------------------
Description: Drop unnecessary __GFP_NOMEMALLOC flag from filter_get_page()
brian [Wed, 30 Sep 2009 13:57:51 +0000 (13:57 +0000)]
b=20877
i=tianzy
i=grev
A better test that does not make assumptions about performance (i.e. that
a given I/O operation will complete in a given amount of time).
Per Tian's suggestion, probably this could be functionalized to deal with
other situations of the same nature.
yangsheng [Wed, 30 Sep 2009 06:29:21 +0000 (06:29 +0000)]
Branch HEAD
b=20539
Add the target file for OEL5.
yangsheng [Wed, 30 Sep 2009 05:01:42 +0000 (05:01 +0000)]
Branch HEAD
b=20539
i=johann, bobijam
Add OEL5 support.
yangsheng [Wed, 30 Sep 2009 04:13:00 +0000 (04:13 +0000)]
Branch HEAD
b=19990
i=adilger, johann
Add patch come from upstream kernel to handle the bitmap counter overflow.
yangsheng [Wed, 30 Sep 2009 04:06:35 +0000 (04:06 +0000)]
Branch HEAD
b=17153
i=johann, bobijam
Remove the BKL usage.
ericm [Wed, 30 Sep 2009 02:27:15 +0000 (02:27 +0000)]
branch: HEAD
move common client helper function to lclient.
b=19688
o=adilger
r=tappro
r=ericm
eeb [Tue, 29 Sep 2009 13:40:39 +0000 (13:40 +0000)]
* Removed spurious LASSERT from lob_sublock_env_get() - the preceding 'if'
expression will SEGV before the LASSERT fails.
dzogin [Fri, 25 Sep 2009 18:17:01 +0000 (18:17 +0000)]
Branch HEAD
b=20482
i=adilger
----------------------------------------------------------------------
Modified Files:
lustre/ChangeLog lustre/tests/conf-sanity.sh
----------------------------------------------------------------------
Description: Conf-sanity.sh 50g test - deactivated OST should not cause a panic.
grev [Fri, 25 Sep 2009 15:44:43 +0000 (15:44 +0000)]
b=19242
i=Robert.Read
add MOUNT2 check to check_and_setup_lustre fn
minor fix
grev [Fri, 25 Sep 2009 15:04:17 +0000 (15:04 +0000)]
b=19242
i=Robert.Read
add MOUNT2 check to check_and_setup_lustre fn
johann [Fri, 25 Sep 2009 13:52:49 +0000 (13:52 +0000)]
Branch b1_8
b=20205
i=girish
Disable write barrier on ext4/SLES11 as we did for ext3/SLES10, since
it seems to hurt write performance badly.
bobijam [Fri, 25 Sep 2009 07:53:36 +0000 (07:53 +0000)]
Branch HEAD
'./lustrecvs -r xxxxx' should checkout lustre root directory using tag xxxxx,
while current lustrecvs will checkout lustre root directory using HEAD tag.
shadow [Fri, 25 Sep 2009 07:20:45 +0000 (07:20 +0000)]
revert fix - because this hide real problem.
Branch HEAD
b=18857