Whamcloud - gitweb
tappro [Wed, 15 Jul 2009 11:42:33 +0000 (11:42 +0000)]
Branch HEAD
b=19844
i=rread
i=zam
- fix the bug with timeout due to missed clients during recovery, introduce
mechanism to track such events and re-charge timer.
- fix incorrect handling of some counters in class_unlink_export
- correct procfs output information about recovery
- add regression tests
liangzhen [Wed, 15 Jul 2009 04:15:10 +0000 (04:15 +0000)]
Branch HEAD
ordering write_lock to avoid deadlock
b=19518
i=adilger
i=zhenyu.xu
manoj [Tue, 14 Jul 2009 21:11:44 +0000 (21:11 +0000)]
b=19894
r=grev
r=nathan.rutman
Kill and cleanup iozone before replicating data.
zhanghc [Tue, 14 Jul 2009 13:34:03 +0000 (13:34 +0000)]
b=18539
don't allocate new files on degraded OSTs
i=adilger@sun.com
i=hongchao.zhang@sun.com
yangsheng [Tue, 14 Jul 2009 08:12:49 +0000 (08:12 +0000)]
Branch HEAD
b=19869
i=johann, shadow
Config check for the U64 type independent in user & kernel space.
jxiong [Tue, 14 Jul 2009 06:51:00 +0000 (06:51 +0000)]
b=18881
r=vitaly,eric.mei
- cleanup the cl_sync_io
- use cl_sync_io to write pages when the lock is being canceled
shadow [Tue, 14 Jul 2009 06:18:38 +0000 (06:18 +0000)]
Land first part of new dcache handling (bug 16654).
make DCACHE_LUSTRE_INVALID always defined.
Branch HEAD
b=16417
i=green
i=johann
nathan [Mon, 13 Jul 2009 16:53:47 +0000 (16:53 +0000)]
b=18144
i=ericm
i=wangybin
hold lock while assigning OST indicies
yujian [Mon, 13 Jul 2009 02:08:50 +0000 (02:08 +0000)]
Branch HEAD
b=19432
i=robert.read
i=grev
1) refresh Kerberos 5 TGT for mpirun user
2) move MPIRUN and MPI_USER from ncli.sh to local.sh
3) remove MPIBIN
4) redirect the stderr of "which mpirun" to /dev/null
5) run MPI_USER_{UID,GID} initialization only for MPI tests
girish [Sun, 12 Jul 2009 16:33:39 +0000 (16:33 +0000)]
b=19872
i=adilger
i=dmitry
fix fiemap test 130e for multi-platform tests
pravins [Sat, 11 Jul 2009 16:24:57 +0000 (16:24 +0000)]
b=18857
reverting patch due to patchless client issue.
pravins [Sat, 11 Jul 2009 09:29:37 +0000 (09:29 +0000)]
b=18857
i=green
i=tappro
fix new inode sanity check for HEAD.
anserper [Fri, 10 Jul 2009 20:04:56 +0000 (20:04 +0000)]
b=18695
i=Andreas Dilger
regression test, the updated version
maxim [Fri, 10 Jul 2009 07:19:47 +0000 (07:19 +0000)]
b=20079
i=isaac
i=liang
Minor bugfix of usocklnd - calling usocklnd_assign_ni_nid unconditionally
(both for user-mode and server-mode cases) to let lnet layer know our
actual nid.
tappro [Thu, 9 Jul 2009 18:19:58 +0000 (18:19 +0000)]
Branch HEAD
b=19669
i=adilger
i=rread
Handle compat/incompat flags properly
yangsheng [Thu, 9 Jul 2009 10:09:35 +0000 (10:09 +0000)]
Branch b1_8
b=20009
i=yangsheng, wangyb
Other fix to cover the reuse and build case.
Author: brian
grev [Thu, 9 Jul 2009 08:35:24 +0000 (08:35 +0000)]
b=20094
i=Alexander.Zarochentsev (Zam)
test_121: ignore useless dd output
grev [Thu, 9 Jul 2009 08:27:53 +0000 (08:27 +0000)]
b=20074
i=Alexey.Lyashkov
i=Robert.Read
process tar's rc correctly
grev [Thu, 9 Jul 2009 08:23:13 +0000 (08:23 +0000)]
b=19215
i=Alexander.Zarochentsev (Zam)
i=Robert.Read
CMD sanityN test_33a fix, verbose do_nodes
grev [Thu, 9 Jul 2009 08:10:23 +0000 (08:10 +0000)]
b=19737
i=Robert.Read
acc-sm, t-f NFSCLIENT changes
vitaly [Wed, 8 Jul 2009 23:30:02 +0000 (23:30 +0000)]
Branch HEAD
b=18619
i=green
i=panda
oap_async_flags protection
isaac [Wed, 8 Jul 2009 19:33:31 +0000 (19:33 +0000)]
i=liang,b=13065:
- port router pinger to userspace.
nathan [Wed, 8 Jul 2009 17:04:53 +0000 (17:04 +0000)]
b=19889
i=rread
i=manoj
return error codes from pool (and other) commands
jxiong [Wed, 8 Jul 2009 06:32:13 +0000 (06:32 +0000)]
b=20038
i=eric.mei
i=rread
- Cleanup client_obd_list_lock
- When a cl_page is to be freed, the page has to clear Uptodate bit as well because it may be NOT under the protection of a ldlm lock
- Add more debug code
dzogin [Wed, 8 Jul 2009 06:01:49 +0000 (06:01 +0000)]
Branch HEAD
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
yujian [Wed, 8 Jul 2009 05:53:44 +0000 (05:53 +0000)]
Branch HEAD
b=20047
i=robert.read
i=yong.fan
reset debug level if lustre was already mounted
yujian [Wed, 8 Jul 2009 05:44:36 +0000 (05:44 +0000)]
Branch HEAD
b=19967
i=yong.fan
i=eric.mei
add test 2, 5, 6 to the ALWAYS_EXCEPT list
tappro [Wed, 8 Jul 2009 04:21:21 +0000 (04:21 +0000)]
Branch HEAD
b=20049
i=shadow
i=fanyong
remove wrong assertion
shadow [Wed, 8 Jul 2009 03:38:54 +0000 (03:38 +0000)]
move ldlm namespace creation in setup phase to avoid grab
_lprocfs_lock with cli_sem held.
Branch HEAD
b=18380
i=rread
zam [Tue, 7 Jul 2009 09:43:43 +0000 (09:43 +0000)]
Branch HEAD
b=19630
i=alexander.zarochentsev
i=alexey.lyashkov
fail truncate syscall if one ost is inactive.
deshmukh [Tue, 7 Jul 2009 06:56:06 +0000 (06:56 +0000)]
b=18504
i=adilger
i=pravin
i=girish
Moved IAM code from ldiskfs to OSD.
girish [Tue, 7 Jul 2009 03:14:36 +0000 (03:14 +0000)]
b=19872
i=adilger
i=johann
fix test for last block
yangsheng [Tue, 7 Jul 2009 02:38:57 +0000 (02:38 +0000)]
Branch b1_8
b=20009
i=brian, wangyb
Fix for sles11 patchless build.
fanyong [Tue, 7 Jul 2009 02:23:47 +0000 (02:23 +0000)]
Branch HEAD
b=19430
i=eric.mei
i=robert.read
Add the missed mdd operation interfaces for ".lustre" object.
fanyong [Tue, 7 Jul 2009 01:59:33 +0000 (01:59 +0000)]
Branch HEAD
b=19802
i=vitaly.fertman
i=robert.read
mdt_quota_off should process "-EALREADY" case.
b=20002
i=tianzy
i=robert.read
check "auto_quota_on()" return value before "build_lqs()" called
b=20017
i=vitaly.fertman
i=tappro
MDS should return some error when quota on with SOM conflict found.
robert.read [Fri, 3 Jul 2009 16:13:40 +0000 (16:13 +0000)]
Branch HEAD
b=200067
i=jay
i=rread
a=fanyong
Check tsk before using it.
yangsheng [Fri, 3 Jul 2009 10:57:06 +0000 (10:57 +0000)]
Branch b1_8
b=20009
i=yangsheng, wangyb
Fixed previous patch.
robert.read [Fri, 3 Jul 2009 06:49:30 +0000 (06:49 +0000)]
Branch HEAD
b=19507
i=johann
HEAD port of original 1.8.1 patch.
tappro [Fri, 3 Jul 2009 06:33:40 +0000 (06:33 +0000)]
Branch HEAD
b=18554
i=rread
Fix test to use also local hostname
liangzhen [Fri, 3 Jul 2009 06:32:48 +0000 (06:32 +0000)]
Branch HEAD
b=18451
Growing TX/FMR/PMR pool at runtime for o2iblnd, these runtime allocated
pools can be freed if not accessed for long time.
i=isaac
tianzy [Fri, 3 Jul 2009 04:47:09 +0000 (04:47 +0000)]
Branch HEAD
fix quota spurious message
b=19898
i=tianzy
i=panda
robert.read [Thu, 2 Jul 2009 18:26:39 +0000 (18:26 +0000)]
Branch HEAD
b=20054
i=girish
i=ericm
Fix lfs getstripe to print the stripe data.
yangsheng [Thu, 2 Jul 2009 15:36:00 +0000 (15:36 +0000)]
Branch b1_8
b=19969
Fixed for rpm install dependencies failed.
zhanghc [Thu, 2 Jul 2009 01:07:38 +0000 (01:07 +0000)]
b=19870
print correct pid/comm in libcfs_debug_dumpstack
i=sheng.yang@sun.com
i=nathan.rutman@sun.com
robert.read [Wed, 1 Jul 2009 23:16:58 +0000 (23:16 +0000)]
Branch HEAD
b=18442
Fix test breakage caused by whitespace changes made by 18442.
johann [Wed, 1 Jul 2009 20:46:45 +0000 (20:46 +0000)]
Branch b1_8
b=19875
i=girish
Fix ext4_free_inode() vs. ext4_claim_inode() race.
Backport
http://git.kernel.org/?p=linux/kernel/git/tytso/ext4.git;a=commit;h=
7ce9d5d1f3c8736511daa413c64985a05b2feee3
johann [Wed, 1 Jul 2009 20:43:31 +0000 (20:43 +0000)]
Branch b1_8
b=20009
i=wangyb (patch from yangsheng)
Fix "invalid module format" error on patchless clients.
girish [Wed, 1 Jul 2009 10:07:44 +0000 (10:07 +0000)]
b=17569
i=adilger
i=johann
add check for >8TB filesystem
tappro [Wed, 1 Jul 2009 08:56:23 +0000 (08:56 +0000)]
Branch HEAD
b=17765
i=zam
i=fanyong
Update server boot epoch even if recovery didn't start
girish [Wed, 1 Jul 2009 06:57:31 +0000 (06:57 +0000)]
b=18442
i=adilger
i=rread
o=james simmons
integrate "ll_getstripe_info" into "lfs getstripe"
deshmukh [Wed, 1 Jul 2009 06:50:11 +0000 (06:50 +0000)]
b=18518
i=adilger,zhenyu.xu
remove ext3-ialloc-2.6.patch from series files
jxiong [Wed, 1 Jul 2009 05:35:27 +0000 (05:35 +0000)]
b=19663,19910
r=eric.mei,rread
Fixed bug 19663 & 19910.
yangsheng [Wed, 1 Jul 2009 04:37:35 +0000 (04:37 +0000)]
Branch HEAD
b=19553
i=johann, shadow
Workaround for OFED-1.4.1 on SLES10.
isaac [Wed, 1 Jul 2009 04:06:50 +0000 (04:06 +0000)]
b=19520,i=brian,i=liang:
- add AC_MSG_CHECKING to OFED checks.
bobijam [Wed, 1 Jul 2009 01:17:42 +0000 (01:17 +0000)]
Branch HEAD
b=18702
i=nathan.rutman
i=johann
Server failover cleanup needs destroy llog.
brian [Tue, 30 Jun 2009 21:15:16 +0000 (21:15 +0000)]
(Finally!) remove the additional -lustre tag from the Lustre kernel RPM
names. The code is still there to do it, although it's now optionalized
with the default being to remove the tag.
robert.read [Mon, 29 Jun 2009 21:30:28 +0000 (21:30 +0000)]
Branch HEAD
b=19280
Remove conf-sanity 50e and 50f from the skip list.
vitaly [Mon, 29 Jun 2009 16:06:40 +0000 (16:06 +0000)]
Branch HEAD
b=19207
i=zam
i=panda
direct I/O code checks whether data can fit user buffer now
girish [Mon, 29 Jun 2009 04:34:56 +0000 (04:34 +0000)]
b=17569
i=adilger
i=johann
add check for >8TB filesystems. Don't allo mounting them until force_over_8tb mount option is specified
shadow [Sat, 27 Jun 2009 05:28:28 +0000 (05:28 +0000)]
Add async create into HEAD (port from 1.x)
Branch HEAD
b=18357
i=rread
shadow [Sat, 27 Jun 2009 04:56:03 +0000 (04:56 +0000)]
with AT enabled, we can have connect request timeout ~ reconnect
timeout and test can't see real disconnect
Branch HEAD
b=19280
i=rread
stinson1947 [Fri, 26 Jun 2009 20:22:17 +0000 (20:22 +0000)]
Updated version string for 2.0.0 build 21.
tappro [Fri, 26 Jun 2009 17:49:58 +0000 (17:49 +0000)]
Branch HEAD
b=19550
i=rread
i=shadow
Fix small issue with incorrect check for flags
tappro [Fri, 26 Jun 2009 17:48:12 +0000 (17:48 +0000)]
Branch HEAD
b=19550
i=rread
i=zam
Add lock counter to export and output it in dump_exports
robert.read [Fri, 26 Jun 2009 17:01:01 +0000 (17:01 +0000)]
Branch HEAD
b=19933
i=grev
a=johann
Add error messages to sanity test 103
brian [Fri, 26 Jun 2009 13:20:03 +0000 (13:20 +0000)]
b=19927
i=yangsheng
i=wangyb
If the $LUSTRE_VERS environment variable is set, don't bother doing the
legwork to determine the version, and just exit early with it.
vs [Fri, 26 Jun 2009 13:18:26 +0000 (13:18 +0000)]
b=19086
i=rread,tappro
- remove call of df in backgroup from facet_failover()
- use client_df in fail() to wait for recovery completion
- replay_single:test_70b: use fail() instead of facet_failover
- replay_single:test_74: umount/mount all clients
yangsheng [Fri, 26 Jun 2009 11:30:49 +0000 (11:30 +0000)]
Branch b1_8
b=19848
i=brian, wangyb
Change for sles11 new kernel build.
grev [Fri, 26 Jun 2009 11:04:59 +0000 (11:04 +0000)]
b=19451
i=Brian
IOR client load for recovery scale tests
grev [Fri, 26 Jun 2009 09:14:58 +0000 (09:14 +0000)]
o=Adilger
i=Robert
i=grev
report patchless_client for kernel version
fanyong [Fri, 26 Jun 2009 06:31:01 +0000 (06:31 +0000)]
Branch HEAD
b=19435
i=tianzy
i=andrew.perepechko
Process quota on/off/invalid serially.
fanyong [Fri, 26 Jun 2009 06:01:04 +0000 (06:01 +0000)]
Branch HEAD
b=18317
i=tianzy
i=andrew.perepechko
Wait until all lqs have been freed on cleanup.
bobijam [Fri, 26 Jun 2009 01:07:52 +0000 (01:07 +0000)]
Branch HEAD
b=16317
i=adilger
add conf-sanity test_44.
grev [Thu, 25 Jun 2009 16:58:20 +0000 (16:58 +0000)]
b=18021
i=Adilger
i=Nathan
i=Robert
t-f CLIENTONLY does not depend on *_HOST
vitaly [Thu, 25 Jun 2009 09:58:38 +0000 (09:58 +0000)]
Branch HEAD
b=18619
i=vitaly
i=johann
max rpc in flight panda's fix
yujian [Thu, 25 Jun 2009 01:42:57 +0000 (01:42 +0000)]
Branch HEAD
b=19413
i=eric.mei
specify $MOUNT for $LFS flushctx
fanyong [Wed, 24 Jun 2009 23:04:49 +0000 (23:04 +0000)]
Branch HEAD
b=18574
i=robert.read
typo fix for the srcipts.
johann [Wed, 24 Jun 2009 22:48:19 +0000 (22:48 +0000)]
Branch b1_8
b=19875
backport patches to RHEL5/ext4.
johann [Wed, 24 Jun 2009 21:33:44 +0000 (21:33 +0000)]
Branch b1_8
remove ext4-ialloc-2.6-rhel5.patch from patch series.
tappro [Wed, 24 Jun 2009 19:12:57 +0000 (19:12 +0000)]
Branch HEAD
b=19550
i=rread
i=zam
Move unlinked exports to the new list until they become 'zombie'. Wait for these
exports during umount. Output in dump_exports() is exteneded and contain more
fields to help with debugging. The new exp_cb_count was added to track number of
callbacks keeping the export references.
brian [Wed, 24 Jun 2009 18:45:06 +0000 (18:45 +0000)]
b=19853
i=jack
i=yangsheng
The SLES 11 kernel spec has what is IMHO a bug in that it tries to write to
a file in RPMS/i386 even though you are building i686 RPMs.
brian [Wed, 24 Jun 2009 18:27:19 +0000 (18:27 +0000)]
b=19850
i=jack
i=wangyb
Shouldn't disable srp-target-mod after enabling it.
johann [Wed, 24 Jun 2009 15:39:18 +0000 (15:39 +0000)]
Branch b1_8
b=19875
i=adilger
i=yansgheng
- use mb_find_next_bit() instead of find_next_bit() in
ext4_mb_check_ondisk_bitmap() as done in the ext4 mainline
- convert ext4_lock_group to use sb_bgl_lock, patch already
merged upstream.
yangsheng [Wed, 24 Jun 2009 13:51:18 +0000 (13:51 +0000)]
Branch b1_8
b=19869
i=shadow, johann, tianzy
Fixes for Werror flag enabled.
pravins [Tue, 23 Jun 2009 21:52:20 +0000 (21:52 +0000)]
b=17670
fix build failure from last commit
pravins [Tue, 23 Jun 2009 21:03:20 +0000 (21:03 +0000)]
b=17670
i=tom.wang
i=yong.fan
osd cleanup.
-- remove unneccessary forward declaration.
-- reorg osd_ea_fid_get(); req for fid_in_dirent patch.
-- remove oie_curr_pos and oie_next_pos
pravins [Tue, 23 Jun 2009 20:59:16 +0000 (20:59 +0000)]
b=17670
i=tom.wang
i=yong.fan
build and return lu_dirent from osd iterator.
use type info available in ext3 dirent
tappro [Tue, 23 Jun 2009 17:36:09 +0000 (17:36 +0000)]
Branch HEAD
b=19885
i=rread
i=zam
Simpified interop case. When 2.0 server is started over old 1.8 FS it evicts
clients by truncating last_rcvd file. Therefore clients will use 2.0-formatted RPC after
reconnect
tianzy [Tue, 23 Jun 2009 06:25:18 +0000 (06:25 +0000)]
Branch HEAD
use lqs to judge which uid/gid are set quota limitaton and
which uid/gid are not, and treat them differently.
I found a few issues after landing the patch for 18574.(some are caused by it, some are easier to
be reproduced by it)
1. in setquota, quota slaves will set their quota limitation to 1 and then get more quota. But in
this time, it's lqs size is the default value. If we have set a small quota limitation, we will
meet -EDQUOT. So we should adjust qunit firstly in this situation.
2. filter_quota_ctl(), we should promise there is no qunit is calculated or sent between
qctxt_wait_pending_dqacq() and fsfilt_quotactl(). Now I did this via a "fake" qunit.
b=18574
i=johann
i=panda
aVS: ----------------------------------------------------------------------
tianzy [Tue, 23 Jun 2009 06:20:26 +0000 (06:20 +0000)]
Branch HEAD
part I
1. mainly fix the lqs_key of lustre_qunit_size.
2. clean quota_compute_lqs().
3. only supply quota_search_lqs() to outside and make quota_create_lqs() internal.
b=18616
i=johann
i=panda
isaac [Tue, 23 Jun 2009 04:06:17 +0000 (04:06 +0000)]
i=wangdi,i=rread,b=19920:
- LIBCFS_ALLOC_GFP should not print 'kmalloced' message at allocation failures.
dzogin [Tue, 23 Jun 2009 01:01:29 +0000 (01:01 +0000)]
Branch HEAD
b=18624
i=zhenyu.xu
i=dmitry.zogin
Description: Unable to run several mkfs.lustre on loop devices at the same time.
jxiong [Mon, 22 Jun 2009 23:23:51 +0000 (23:23 +0000)]
b=19188
r=rread,fanyong
cleanup the variables with `_', rename them to `unused'
ericm [Mon, 22 Jun 2009 19:44:51 +0000 (19:44 +0000)]
branch: HEAD
server save opc in replay_state thus we don't need to access repmsg after
reply being submmitted to ptlrpc.
b=19386
r=wangdi
r=fanyong
vitaly [Mon, 22 Jun 2009 19:21:18 +0000 (19:21 +0000)]
Branch HEAD
b=19154
i=tappro
i=vs
let the statfs cache to get old enough on replay-single 20b
jxiong [Mon, 22 Jun 2009 18:39:46 +0000 (18:39 +0000)]
b=19188
r=pravin,rread
More debug info are added to catch bugs.
jxiong [Mon, 22 Jun 2009 18:34:53 +0000 (18:34 +0000)]
b=19188
r=pravin,rread
It is bad to access the ols_lock if the lock is detached.
deshmukh [Mon, 22 Jun 2009 18:04:37 +0000 (18:04 +0000)]
b=19398
i=tom.wang
i=nathan.rutman
i=jinshan.xiong
Fix for soft lockup - CPU#13 stuck for 10s! [umount:16789]
bobijam [Mon, 22 Jun 2009 01:49:08 +0000 (01:49 +0000)]
Branch HEAD
b=19913
o=grev
i=johann
skip test_151 for remote ost && no_dsh
bobijam [Mon, 22 Jun 2009 01:46:52 +0000 (01:46 +0000)]
Branch HEAD
b=19913
o=johann
i=grev
i=bobijam
Add support remote ost in test 151.