Whamcloud - gitweb
bobijam [Mon, 10 Mar 2008 02:03:33 +0000 (02:03 +0000)]
Branch b1_6
b=14321
o=Brian Behlendorf(behlendorf1@llnl.gov)
i=bobijam
i=shadow
Description: lustre_mgs: operation 101 on unconnected MGS
Details : When MGC is disconnected from MGS long enough, MGS will evict the
MGC, and late on MGC cannot successfully connect to MGS and a lot
of the error messages complaining that MGS is not connected.
komaln [Sat, 8 Mar 2008 20:12:59 +0000 (20:12 +0000)]
b=14471
r=Adilger
Changed recovery-small.sh to use 'lctl {get,set}_param'
adilger [Fri, 7 Mar 2008 20:59:49 +0000 (20:59 +0000)]
Branch b1_6
Make description of bug 14775 more clear.
anserper [Fri, 7 Mar 2008 19:08:07 +0000 (19:08 +0000)]
Branch b1_6
b=15036
i=johann
i=tianzy
update local vars types to handle 64-bit data
anserper [Fri, 7 Mar 2008 19:05:46 +0000 (19:05 +0000)]
Branch b1_6
b=15036
i=johann
i=tianzy
update local vars types to handle with 64-bit limits
komaln [Fri, 7 Mar 2008 11:05:37 +0000 (11:05 +0000)]
b=14471
A few corrections in earlier patch to change sanity.sh to use 'lctl {get,set}_param'
jxiong [Fri, 7 Mar 2008 04:29:14 +0000 (04:29 +0000)]
Forget to initialize the li_group.
grev [Thu, 6 Mar 2008 21:38:32 +0000 (21:38 +0000)]
b=13557 (author Adilger)
print the stripecount as a signed value
grev [Thu, 6 Mar 2008 17:49:58 +0000 (17:49 +0000)]
b=14957
i=Adilger
test_23a fix
wangdi [Thu, 6 Mar 2008 08:11:55 +0000 (08:11 +0000)]
Branch: b1_6
b=14483
detect stride IO mode in read-ahead.
i=adilger, nikita
komaln [Thu, 6 Mar 2008 05:29:32 +0000 (05:29 +0000)]
b=14971
r=Adilger
to solve RHEL5/ia64 build error related to 'lctl {get,set}_param' patch
shadow [Thu, 6 Mar 2008 04:19:28 +0000 (04:19 +0000)]
Client not clear own cache if answer to reconnect is lost.
b=14775
i=green
i=nathan
i=johann
jxiong [Thu, 6 Mar 2008 03:03:53 +0000 (03:03 +0000)]
b=11300
r=oleg,adilger
Fix interval tree issues at the customer's side.
bobijam [Thu, 6 Mar 2008 02:24:08 +0000 (02:24 +0000)]
Branch b1_6
b=13715
i=johann
i=nathan
i=adilger
protect accessing imp->generation w/ imp_lock.
bobijam [Thu, 6 Mar 2008 02:09:09 +0000 (02:09 +0000)]
Branch b1_6
b=12584
i=adilger
There is a issue with getfacl which just literally find the symbol link
target with `pwd`+[symbol content], regardless where the link file is
located.
This fix avoids the issue by chdir to the symbol link's dir before getting it's
acl list.
deen [Wed, 5 Mar 2008 18:09:51 +0000 (18:09 +0000)]
When the failover node is the primary node, it is possible
to have two identical connections in imp_conn_list. We must
compare not conn's pointers but NIDs, otherwise we can defeat
connection throttling.
b=14774
i=adilger
i=johann
bobijam [Wed, 5 Mar 2008 07:24:49 +0000 (07:24 +0000)]
Branch b1_6
b=14873
i=johann
i=shadow
protect competitive adding to obd->obd_nid_stats.
green [Wed, 5 Mar 2008 06:22:46 +0000 (06:22 +0000)]
b=13677
r=eeb,adilger
wait for async rpc completion in liblustre
grev [Tue, 4 Mar 2008 19:25:40 +0000 (19:25 +0000)]
b=13297 (comment 36)
add 27u to ALWAYS_EXCEPT list
adilger [Mon, 3 Mar 2008 21:16:40 +0000 (21:16 +0000)]
Branch b1_6
Fix incorrect build version.
adilger [Mon, 3 Mar 2008 17:19:47 +0000 (17:19 +0000)]
Branch b1_6
Update build version for b1_6.
green [Mon, 3 Mar 2008 02:33:47 +0000 (02:33 +0000)]
b=14748
r=adilger,jay
When traversing extent wating list queue for PR lock and meeting same or wider
lock without AST_SENT, we are guaranteed there is no other one like this down
the line, so we can return immediatelly
komaln [Sun, 2 Mar 2008 20:23:26 +0000 (20:23 +0000)]
b=14471
r=Adilger
Changed conf-sanity.sh to use 'lctl {get,set}_param'
komaln [Sun, 2 Mar 2008 19:23:42 +0000 (19:23 +0000)]
b=14471
r=Adilger
Changed sanityN.sh to use 'lctl {get,set}_param'
komaln [Sun, 2 Mar 2008 19:19:15 +0000 (19:19 +0000)]
b=14471
r=Adilger
Changed sanity.sh to use 'lctl {get,set}_param'
girish [Thu, 28 Feb 2008 10:40:16 +0000 (10:40 +0000)]
Fix for sanity quota failure. (From Kalpak)
acc-sm passed.
b=13487
tianzy [Thu, 28 Feb 2008 07:07:46 +0000 (07:07 +0000)]
Branch b1_6
problem:
when a quota request is delayed or dropped, the quota slave who has sent it
will wait for it forever.
solution:
When a quota request time out, dqacq_interpret will hanle it in case the
situation above happened.
b=14840
i=johann
i=andrew.perepechko
yangsheng [Thu, 28 Feb 2008 06:34:21 +0000 (06:34 +0000)]
Branch b1_6
b=13397
i=shadow
i=johann
Add kernel config for vanilla-2.6.22.14.
tianzy [Thu, 28 Feb 2008 05:33:09 +0000 (05:33 +0000)]
Branch b1_6
After an unnecessary block in osts for waiting in-flight quota request is
deleted, a few places should be fixed in order to work with it.
b=14783
i=johann
i=andrew.perepechko
grev [Wed, 27 Feb 2008 21:42:42 +0000 (21:42 +0000)]
b=14978
i=Adilger
check whether proc files exist; error if they do not exist
wangdi [Wed, 27 Feb 2008 20:45:14 +0000 (20:45 +0000)]
Branch: b1_6
b=13464
Set import disconnection before mgc reconnection, and refine the previous
13464 fix.
i=nathan,adiger
grev [Wed, 27 Feb 2008 19:38:14 +0000 (19:38 +0000)]
b=14229
i=Adilger
fix test_59 to fail if write/read failed
grev [Wed, 27 Feb 2008 19:20:11 +0000 (19:20 +0000)]
b=14928
i=Adilger
don't re-write mtab, wait it will be updated
grev [Wed, 27 Feb 2008 19:06:35 +0000 (19:06 +0000)]
b=14528
i=Adilger
add wiretest to lustre rpm
grev [Tue, 26 Feb 2008 19:33:18 +0000 (19:33 +0000)]
b=13571
i=Adilger
fs2' tests cleanup; renumbering:
- dup test_23 -> 23a and 23b
- 33->33a; 33a->33b
wangdi [Tue, 26 Feb 2008 01:45:40 +0000 (01:45 +0000)]
Branch:b1_6
Missing clear "clear" entry in the filter setup error handler.
b=14370
i=adilger,ericm
bwzhou [Mon, 25 Feb 2008 11:36:22 +0000 (11:36 +0000)]
Branch b1_6
b=13285
r=johann, deen
print messages to console about locks.
yangsheng [Sat, 23 Feb 2008 16:20:27 +0000 (16:20 +0000)]
Branch b1_6
b=14958
i=adilger
i=yangsheng
Change CONFIG_NR_CPUS from 8 to 32 in a several config file.
anserper [Fri, 22 Feb 2008 23:09:55 +0000 (23:09 +0000)]
Branch b1_6
b=14852
i=shadow
i=adilger
additional id checks
deen [Fri, 22 Feb 2008 22:03:12 +0000 (22:03 +0000)]
Adding total number of sampled request for an MDS node in snmp support.
b=14729
i=tappro
anserper [Fri, 22 Feb 2008 21:07:34 +0000 (21:07 +0000)]
Branch b1_6
b=14899
i=adilger
fix missing \n (which caused error messages)
adilger [Fri, 22 Feb 2008 19:28:17 +0000 (19:28 +0000)]
Branch b1_6
Make the mi_cbdata match the HEAD version of the statahead patch.
i=fanyong
b=11401
komaln [Fri, 22 Feb 2008 10:23:18 +0000 (10:23 +0000)]
b=14471
r=Adilger, Wangdi
to add lctl {get,set}_param functionality.
adilger [Fri, 22 Feb 2008 00:01:43 +0000 (00:01 +0000)]
Branch b1_6
Fix comment slightly.
Add a bit more information if debugging is enabled.
adilger [Thu, 21 Feb 2008 23:51:21 +0000 (23:51 +0000)]
Branch b1_6
Trivial whitespace cleanup.
johann [Thu, 21 Feb 2008 13:28:18 +0000 (13:28 +0000)]
Branch b1_6
b=14844
i=adilger
i=deen
fix compile issue on PPC. add configure check for asm-ppc/segment.h.
girish [Thu, 21 Feb 2008 12:51:56 +0000 (12:51 +0000)]
2.6.18.8 vanilla series should use sd_iostats-2.6-rhel5.patch
i=adilger
i=bzzz
b=14891
yangsheng [Thu, 21 Feb 2008 08:29:56 +0000 (08:29 +0000)]
Branch b1_6
b=13397
i=adilger
i=shadow
Add server support for vanilla-2.6.22.14 kernel.
--Landed the kernel patches.
--ldiskfs patches already landed.
huanghua [Thu, 21 Feb 2008 02:02:05 +0000 (02:02 +0000)]
Branch b1_6
b=14149
a=15414
i=yong.fan
i=yury.umanets
define OBD_CONNECT_FID flag.
grev [Wed, 20 Feb 2008 19:01:24 +0000 (19:01 +0000)]
b=14944
i=Adilger
test_38 cleanup
bobijam [Wed, 20 Feb 2008 02:17:38 +0000 (02:17 +0000)]
Branch b1_6
b=14529
i=adilger
i=panda
Description: MDS or OSS nodes crash due to stack overflow
Details : Code changes in 1.6.4 increased the stack usage of some functions.
In some cases, in conjunction with device drivers that use a lot
of stack the MDS (or possibly OSS) service threads could overflow
the stack. One change which was identified to consume additional
stack has been reworked to avoid the extra stack usage.
grev [Tue, 19 Feb 2008 23:10:50 +0000 (23:10 +0000)]
b=14842
i=Adilger
move IOZONE=no after $NAME.sh included for SLOW=no default
grev [Tue, 19 Feb 2008 16:25:02 +0000 (16:25 +0000)]
b=14842
i=Adilger
more fixes to reduce acc-sm time; SLOW=no is default now.
deen [Mon, 18 Feb 2008 14:43:36 +0000 (14:43 +0000)]
Better handle error messages in extents code.
b=14864
i=adilger
i=deen
vitaly [Mon, 18 Feb 2008 12:43:20 +0000 (12:43 +0000)]
Branch b1_6
b=14379
i=green
i=tappro
increase the amount of buffers for ELC RPCs when there is no conflicting
locks but we cancel locks from lru.
johann [Mon, 18 Feb 2008 11:12:25 +0000 (11:12 +0000)]
Branch b1_6
b=13223
i=alexey.lyashkov
i=yong.fan
check for -ENODEV in ptlrpc_replay_interpret()
johann [Mon, 18 Feb 2008 09:34:19 +0000 (09:34 +0000)]
Branch b1_6
b=13400
fix indenting and remove trailing whitespaces in the quota code.
johann [Mon, 18 Feb 2008 08:56:07 +0000 (08:56 +0000)]
Branch b1_6
b=14648
i=adilger
i=bobijam
when available, use the Adler-32 algorithm instead of CRC32 for
page checksums.
johann [Mon, 18 Feb 2008 08:07:21 +0000 (08:07 +0000)]
Branch b1_6
b=13805
i=adilger
i=bobijam
add support for several checksum algorithms. Currently, CRC32 and
Adler-32 are supported. The checksum type can be changed on the fly
through /proc/fs/lustre/osc/*/checksum_type.
yangsheng [Mon, 18 Feb 2008 05:57:44 +0000 (05:57 +0000)]
Branch b1_6
b=14876
i=adilger
i=johann
Kernel security update for RHEL5 2.6.18-53.1.13.el5.
yangsheng [Mon, 18 Feb 2008 05:52:45 +0000 (05:52 +0000)]
Branch b1_6
b=14858
i=adilger
i=johann
Update SLES10 SP1 kernel to 2.6.16.54-0.2.5.
johann [Sun, 17 Feb 2008 19:11:45 +0000 (19:11 +0000)]
Branch b1_6
b=14845
i=shadow
i=deen
fix duplicated definitions between asm-ppc64/types.h and lustre_types.h on PPC.
johann [Sun, 17 Feb 2008 09:32:07 +0000 (09:32 +0000)]
Branch b1_6
b=14618
Downgrad the MPT Fusion driver in the RHEL4 kernel from 3.02.99.00 to 3.02.73 until
the problem regression is fixed.
yury [Sat, 16 Feb 2008 16:23:37 +0000 (16:23 +0000)]
b=13766
r=vitaly,shadow
- fixes abd cleanups in lru resize. Main idea is to avoid dropping SLV much in various situations/loads. Completely reworked memory pressure handler. Fixes in SLV calculations, etc. Please see bug 13766 for more details.
bwzhou [Fri, 15 Feb 2008 09:53:49 +0000 (09:53 +0000)]
Branch b1_6
b=13380
r=adilger, johann
fix a potential deadlock condition in netpoll code of the sles9 kernel
bwzhou [Fri, 15 Feb 2008 09:51:09 +0000 (09:51 +0000)]
Branch b1_6
b=14763
r=adilger
fix a problem with conf-sanity
bwzhou [Thu, 14 Feb 2008 08:51:46 +0000 (08:51 +0000)]
file netpoll_xmit_lock-2.6-suse.patch was added on branch b1_6 on 2008-02-15 09:53:48 +0000
grev [Thu, 14 Feb 2008 00:54:21 +0000 (00:54 +0000)]
b=13726
i=Adilger
fix locale for tests where exact output is required
anserper [Wed, 13 Feb 2008 23:50:29 +0000 (23:50 +0000)]
Branch b1_6
b=14591
i=tianzy
i=andrew.perepechko
anserper [Wed, 13 Feb 2008 23:46:24 +0000 (23:46 +0000)]
temporarily remove "sticked" file
anserper [Wed, 13 Feb 2008 23:34:25 +0000 (23:34 +0000)]
Branch b1_6
b=14591
i=tianzy
i=andrew.perepechko
kernel patch to avoid oops in dquot code (refer to https://bugzilla.lustre.org/attachment.cgi?id=15058&action=edit)
anserper [Wed, 13 Feb 2008 21:54:36 +0000 (21:54 +0000)]
Branch b1_6
b=13285
i=johann
i=deen
handle the NULL return case from ptlrpc_prep_req() (refer to https://bugzilla.lustre.org/attachment.cgi?id=13101&action=edit)
anserper [Wed, 13 Feb 2008 21:05:42 +0000 (21:05 +0000)]
Branch b1_6
b=13285
i=johann
i=deen
fixes some return code issues (refer to https://bugzilla.lustre.org/attachment.cgi?id=15278&action=edit)
tianzy [Wed, 13 Feb 2008 11:51:34 +0000 (11:51 +0000)]
Branch b1_6
Reset error code if we are retrying to insert the qdata into the hashtable
to avoid later assertion.
b=14785
i=johann
i=andrew.perepechko
jxiong [Wed, 13 Feb 2008 11:51:09 +0000 (11:51 +0000)]
Porting raid5 improvements to rhel5 kernels.
b=13648
r=alex,andreas
tianzy [Wed, 13 Feb 2008 11:38:02 +0000 (11:38 +0000)]
Branch b1_6
In filter_commitrw_write, lqs is freed between lquota_chkquota and
lquota_pending_commit which led to "quota_pending_commit()) there are too many
blocks!"
b=14784
i=Johann
i=andrew.perepechko
tianzy [Wed, 13 Feb 2008 07:59:51 +0000 (07:59 +0000)]
Branch b1_6
an unnecessary block in osts for waiting in-flight quota request
b=14783
i=Johann
i=andrew.perepechko
yangsheng [Wed, 13 Feb 2008 07:05:55 +0000 (07:05 +0000)]
Branch b1_6
b=14720
i=scjody
i=johann
Update RHEL5 kernel to 2.6.18-53.1.6.el5.
bobijam [Wed, 13 Feb 2008 01:55:24 +0000 (01:55 +0000)]
Branch b1_6
b=13173
i=shadow
make "lctl lov_getconfig" handle the case which "default_stripe_count == -1" by
returning -1 instead of 65535.
grev [Tue, 12 Feb 2008 22:08:19 +0000 (22:08 +0000)]
b=14620
re-vert att. 14680
grev [Tue, 12 Feb 2008 10:47:16 +0000 (10:47 +0000)]
b=14330
i=Adilger
RUNAS dbench and iozone
grev [Fri, 8 Feb 2008 14:54:59 +0000 (14:54 +0000)]
b=14330 (comment 14)
i=Adilger
remove obsolete cfg scripts
grev [Fri, 8 Feb 2008 14:53:21 +0000 (14:53 +0000)]
b=14330
i=Adilger
remove obsolete insanity-ltest.sh
grev [Fri, 8 Feb 2008 11:57:27 +0000 (11:57 +0000)]
b=14330
i=Adilger
move RUNAS to local.sh; run BONNIE as RUNAS_ID user
b=14474
i=Adilger
don't allow sanity to see MOUNT2; remove chown of MOUNT
anserper [Thu, 7 Feb 2008 15:47:06 +0000 (15:47 +0000)]
Branch b1_6
b=14421
Changelog update
anserper [Thu, 7 Feb 2008 10:22:56 +0000 (10:22 +0000)]
Branch b1_6
b=14421
i=nikita.danilov
i=alexey.lyashkov
Removes unnecessary PageDirty check (dirty page is perfectly valid here).
grev [Wed, 6 Feb 2008 18:55:03 +0000 (18:55 +0000)]
b=12651
i=Adilger
skip test_78 if we have no enough space; make the test file slightly smaller
yangsheng [Tue, 5 Feb 2008 12:53:21 +0000 (12:53 +0000)]
Branch b1_6
b=14793
i=adilger
i=johann
Update RHEL4 kernel to 2.6.9-67.0.4.
shadow [Tue, 5 Feb 2008 11:52:01 +0000 (11:52 +0000)]
Don't allow skipping OSTs if index has been specified.
b=13196
i=adilger
i=green
jxiong [Tue, 5 Feb 2008 08:38:56 +0000 (08:38 +0000)]
Patch to remove the warning of compiling for interval_tree.c
b=11300
r=shadow
girish [Mon, 4 Feb 2008 16:54:30 +0000 (16:54 +0000)]
JBD2_HAS_COMPAT_FEATURE was used instead of JBD2_HAS_INCOMPAT_FEATURE for
checking _INCOMPAT_ASYNC_COMMIT.
Fixed it.
b=10657
yangsheng [Mon, 4 Feb 2008 09:30:00 +0000 (09:30 +0000)]
Branch b1_6
b=14714
Fixed previous commit error.
tianzy [Mon, 4 Feb 2008 07:23:16 +0000 (07:23 +0000)]
Branch b1_6
handle lustre.fail_loc issue when run sanity-quota.sh on different nodes
b=14700
i=johann
i=andrew.perepechko
yangsheng [Mon, 4 Feb 2008 02:44:15 +0000 (02:44 +0000)]
Branch b1_6
b=14714
i=yangsheng
i=scjody
Use d_move_locked & d_rehash_cond instead of __d_move & __d_rehash in kernel patches.
girish [Sat, 2 Feb 2008 11:51:04 +0000 (11:51 +0000)]
Check in a fix from Johann Lombardi about missing get_bh().
And few small changes as per the upstream patch.
b=10657
jxiong [Thu, 31 Jan 2008 08:02:20 +0000 (08:02 +0000)]
Using interval tree for scalable handling of many extent locks.
b=11300
r=vitaly,nikita
komaln [Thu, 31 Jan 2008 05:41:48 +0000 (05:41 +0000)]
b=14184
r=Adilger, Wangdi
To detail ldlm_enqueue and mds_reint stats.
bwzhou [Thu, 31 Jan 2008 05:36:58 +0000 (05:36 +0000)]
Branch b1_6
b=14721
r=johann, scjody
Quiet "excessive revalidate_it loops" message
bobijam [Thu, 31 Jan 2008 02:24:11 +0000 (02:24 +0000)]
Branch b1_6
b=13999
o=Philippe Bernadat(philippe.bernadat@hp.com)
i=johann
i=shadow
Don't attempt to wake up evictor thread if it hasn't started yet.
grev [Thu, 31 Jan 2008 02:03:25 +0000 (02:03 +0000)]
b=12499
run sysctl properly (found by Huang Hua)