Whamcloud - gitweb
jxiong [Thu, 27 Sep 2007 05:13:51 +0000 (05:13 +0000)]
restore this file. wrongly uploaded just now.
jxiong [Thu, 27 Sep 2007 05:09:24 +0000 (05:09 +0000)]
b=5498
r=andreas,wangdi
w
Add support for lustre loop device emulator.
mjmac [Thu, 27 Sep 2007 03:39:03 +0000 (03:39 +0000)]
b=13541
Reverting broken commit.
bwzhou [Thu, 27 Sep 2007 03:15:17 +0000 (03:15 +0000)]
Branch b1_6
b=12840
i=green
i=johann
prevent self-evicting through procfs
bobijam [Thu, 27 Sep 2007 02:52:19 +0000 (02:52 +0000)]
Branch b1_6
b=12948
i= wangdi, johann
Description: buffer overruns could theoretically occur
Details : llapi_semantic_traverse() modifies the "path" argument by
appending values to the end of the origin string, and a buffer
overrun may occur. Adding buffer overrun check in liblustreapi.
bwzhou [Thu, 27 Sep 2007 02:09:48 +0000 (02:09 +0000)]
Branch b1_6
b=13720
i=johann
check for existence of $LUSTRE/utils/l_getgroups before using it as group upcall
nathan [Wed, 26 Sep 2007 23:36:58 +0000 (23:36 +0000)]
b=13716
i=adilger
i=green
clear invalid exp ref, take real references in ptlrpc_at_send_early_reply
walter [Wed, 26 Sep 2007 21:00:47 +0000 (21:00 +0000)]
bz-13541
adilger [Wed, 26 Sep 2007 20:05:40 +0000 (20:05 +0000)]
Branch b1_6
Move patches that landed for 1.6.3 after branching into 1.6.3 section.
Remove entry for reverted patch b=10657.
Move 11270 out of 1.6.3 section.
zam [Wed, 26 Sep 2007 19:10:08 +0000 (19:10 +0000)]
ll_setattr_raw/truncate: there should be no
ll_extent_lock after a successful obd_match.
b=11270
i=vitaly
i=green
yury [Wed, 26 Sep 2007 18:49:54 +0000 (18:49 +0000)]
b=13696
r=adilger,nathan
- adds "lruresize" and "nolruresize" mount options;
- adds lru resize performance test.
yury [Wed, 26 Sep 2007 17:21:56 +0000 (17:21 +0000)]
b=2262
r=adilger,nathan
- fixed possible divide by zero in ldlm_pools_shrink();
- in ldlm_cancel_lru_local() cancel not more than requested locks in case count is specified (not lru resize case, proc one).
grev [Wed, 26 Sep 2007 02:00:13 +0000 (02:00 +0000)]
b=13557
i=Adilger
set stripecount correctly
adilger [Tue, 25 Sep 2007 19:38:22 +0000 (19:38 +0000)]
Branch b1_6
Remove ancient tag file.
adilger [Tue, 25 Sep 2007 18:17:14 +0000 (18:17 +0000)]
Branch b1_6
Move $RUNAS initialization until after cfg parsing so that $RUNAS_ID can
be set in the cfg file. Use "error" instead of "echo".
b=12609
adilger [Tue, 25 Sep 2007 16:39:32 +0000 (16:39 +0000)]
Branch b1_6
Description: e2fsck reports "invalid unused inodes count"
Details : If a new ldiskfs filesystem is created with the "uninit_groups"
feature and only a single inode is created in a group then the
"bg_unused_inodes" count is incorrectly updated. Creating a
second inode in that group would update it correctly.
i=behlendorf
i=valerie
walter [Tue, 25 Sep 2007 16:36:06 +0000 (16:36 +0000)]
bz-13516
walter [Tue, 25 Sep 2007 15:18:42 +0000 (15:18 +0000)]
convert uint64_t to __u64
grev [Tue, 25 Sep 2007 09:34:22 +0000 (09:34 +0000)]
b=13504
i=Nathan
give multiop chan��e to start before umount client.
huanghua [Tue, 25 Sep 2007 02:42:36 +0000 (02:42 +0000)]
Branch b1_6
b=13685
b=11401
i=adilger
i=tappro
fix race condition in stopping statahead thread.
huanghua [Tue, 25 Sep 2007 02:37:34 +0000 (02:37 +0000)]
Branch b1_6
b=13653
b=11401
i=adilger
i=green
dir entry may have been removed before statahead starts.
nathan [Mon, 24 Sep 2007 16:53:05 +0000 (16:53 +0000)]
b=13537
b=3055
in some cases we wouldn't schedule at ptlrpc_invalidate_import
cap reconnect ping interval
use simple reconnect backoff instead of "drain"
bwzhou [Mon, 24 Sep 2007 09:46:03 +0000 (09:46 +0000)]
Branch b1_6
b=12609
i=johann
i=tianzy
set l_getgroups as the default group upcall for MDS;
a minor fix on sanity.sh
yangsheng [Mon, 24 Sep 2007 03:44:51 +0000 (03:44 +0000)]
Branch b1_6
b=13616
i=johann
i=scjody
Modify target file & which_kernel for RHEL5 2.6.18-8.1.10.el5 kernel.
nathan [Thu, 20 Sep 2007 17:04:50 +0000 (17:04 +0000)]
b=13669
i=adilger
i=green
send early replies even after we've packed the final reply.
yury [Thu, 20 Sep 2007 12:15:45 +0000 (12:15 +0000)]
b=2262
r=vitaly,nikita
- slv recalc fixes and cleanups in ldlm_pool.c;
- fixes and cleanups in lru resize sanity test 124.
yury [Thu, 20 Sep 2007 07:46:47 +0000 (07:46 +0000)]
b=13632
r=adilger, nikita
- Do not take namespace list lock while doing pool shrink or recalc, so that locks cancel may be done without ns list lock.
huanghua [Thu, 20 Sep 2007 07:06:27 +0000 (07:06 +0000)]
Branch b1_6
b=11401
i=shadow
i=tappro
- use cfs_* macros and functions instead of linux primitives (a=12609)
- cast fixing (a=12706)
yury [Thu, 20 Sep 2007 06:57:39 +0000 (06:57 +0000)]
b=13664
r=adilger, nathan
- use cancel_lru_locks in replay-ost-single.sh test_4 to cancel cached locks instead of using buggy echo -n clear > $i/lru_size
bwzhou [Thu, 20 Sep 2007 06:33:50 +0000 (06:33 +0000)]
Branch b1_6
b=12234
i=adilger
i=johann
correct the cast of 'index' in obd_device_list_seq_show()
grev [Thu, 20 Sep 2007 03:05:33 +0000 (03:05 +0000)]
b=13499
i=Adilger
Wait recovery complete before AFTERUSED is checked.
grev [Thu, 20 Sep 2007 00:46:20 +0000 (00:46 +0000)]
b=13587
i=Scjody
set stripe count over all OSTs; calculate SIZE based on min kbytesfree.
grev [Thu, 20 Sep 2007 00:08:39 +0000 (00:08 +0000)]
b=13628
i=Adilger
i=Nathan
do not create recursive dirs
grev [Wed, 19 Sep 2007 23:48:34 +0000 (23:48 +0000)]
b=12613
i=Nathan
remove 18 from A_E list
grev [Wed, 19 Sep 2007 23:01:13 +0000 (23:01 +0000)]
b=12613
i=Adilger
i=Nathan
test_18 MDS_MKFS_OPTS fix; cfg/local.sh MKFSOPT fix.
grev [Wed, 19 Sep 2007 22:15:05 +0000 (22:15 +0000)]
b=10825
i=Adilger
make ALWAYS_EXCEPT usage more consistent.
grev [Wed, 19 Sep 2007 21:52:15 +0000 (21:52 +0000)]
b=13078
new 1.4 image from Andreas.
grev [Wed, 19 Sep 2007 21:11:49 +0000 (21:11 +0000)]
b=13078
i=Adilger
b=13649
i=Nathan
SAVE_PWD fix; 32a, 32b 35 fixes: disk1_4 correct path, correct FSNAME.
zam [Wed, 19 Sep 2007 20:22:26 +0000 (20:22 +0000)]
corrention patch for the lockless i/o code addesses the following problems:
. "false conflicts" may cause file contention and switching to lockless i/o
. compilation warning on a 64-bit platform
. contention reports are recorded as lustre errors
. llap_from_page complains about wrong llap_checksums for lockless pages
. assertions needed to be sure that ll_releasepage is called on pages with
locks only
. NS_DEFAULT_CONTENDED_LOCKS=0 was considered as wrong default value
and, the lockless read/write is disabled by default now.
b=11270
i=green
i=adilger
johann [Wed, 19 Sep 2007 18:59:23 +0000 (18:59 +0000)]
Branch b1_6
b=13358
i=nikita
add regression test.
johann [Wed, 19 Sep 2007 18:57:13 +0000 (18:57 +0000)]
Branch b1_6
b=13358
i=nikita
i=bobijam
Don't LASSERT correctness of data obtained through the network.
grev [Tue, 18 Sep 2007 16:00:48 +0000 (16:00 +0000)]
b=12900
i=Adilger
export DIR2 tdir
adilger [Mon, 17 Sep 2007 17:10:34 +0000 (17:10 +0000)]
Branch b1_6
Add OBD_CONNECT_MDS_MDS and OBD_CONNECT_REAL from HEAD so we don't get any
conflicts in the future.
adilger [Mon, 17 Sep 2007 16:24:16 +0000 (16:24 +0000)]
Branch b1_6
Add OBD_CONNECT_MDS_MDS and OBD_CONNECT_REAL from HEAD so we don't get any
conflicts in the future.
yury [Sat, 15 Sep 2007 13:01:54 +0000 (13:01 +0000)]
- fixed arguments order in ldlm_dump_all_namespaces().
kalpak [Sat, 15 Sep 2007 07:14:43 +0000 (07:14 +0000)]
b=13610
Fix typo in last commit.
kalpak [Sat, 15 Sep 2007 07:06:58 +0000 (07:06 +0000)]
b=13610
i=andreas, johann
The changed prototype of ext_prepare_callback will be used in lustre if HAVE_EXT_PREPARE_CB_EXTENT is declared.
nathan [Fri, 14 Sep 2007 22:38:38 +0000 (22:38 +0000)]
b=13553
i=adilger
df after clearing fail_loc
adilger [Fri, 14 Sep 2007 21:58:43 +0000 (21:58 +0000)]
Branch b1_6
Disable client page cache checksumming by default. This does not affect
the wire checksumming, and only adds a lot of CPU overhead that is mostly
useful only for debugging rare client memory corruption vs. the more
common network transfer corruption. It can be enabled at runtime via
"echo 1 > $LPROC/llite/*/checksum_pages".
b=13579
i=scjody
i=nathan
adilger [Fri, 14 Sep 2007 21:52:16 +0000 (21:52 +0000)]
Branch b1_6
b=13547, b=13627
Description: Data corruption for OSTs that are formatted larger than 2TB
on 32-bit servers.
Details : When generating the bio request for lustre file writes the
sector number would overflow a temporary variable before being
used for the IO. The data reads correctly from Lustre (which
will overflow in a similar manner) but other file data or
filesystem metadata may be corrupted in some cases.
grev [Fri, 14 Sep 2007 17:49:51 +0000 (17:49 +0000)]
b=13394
i=Nathan
get rid of "./"
shadow [Fri, 14 Sep 2007 17:42:58 +0000 (17:42 +0000)]
fix small typo in last commit.
b=13304
cliffw [Fri, 14 Sep 2007 17:33:14 +0000 (17:33 +0000)]
b=13457
i=nathan
i=brian
fixes option syntax for lfs
vitaly [Fri, 14 Sep 2007 17:05:34 +0000 (17:05 +0000)]
Branch b1_6
b=13563
i=adilger
i=green
1) cancel lru locks in ldlm_cli_cancel() and send 1 batched cancel RPC;
2) pass the whole list of canceled locks to the async blocking thread, not
lock-by-lock, and send cancels in batched cancel RPCs.
yury [Fri, 14 Sep 2007 16:14:11 +0000 (16:14 +0000)]
b=13595,13608
r=nathan,adilger,shadow,green
- separates client and server namespaces. Each "side" has own list and own lock;
- separate pool shrinker to client and server shrinkers which work each with own list. This is needed to avoid mixing up server and client pool cached resources which are too different. Client's locks may be canceled in sync manner and we can return to VM number of still cached resources. And server resources (locks) are not removed in sync way, we just change SLV and expect that client will cancel something. To VM we return 0 as number of canceled locks;
- in ldlm_pools_shrink() use down_trylock() to avoid locking ns sem when it is already locked. This fixes hang up in test 116 if memory pressure comes. This issue is due to deadlock bewteen shrinker and pool thread if client and server run on same host;
- move lru add stuff into separate func;
- change l_last_used and move lock to tail of lru for case of FL_TEST_LOCK to make sure that it will still hang for some time in lru afer that. So that, if we looked for look even with FL_TEST_LOCK this means that we may need its resourse yet some time and better to stay lock in cache.
johann [Fri, 14 Sep 2007 13:36:02 +0000 (13:36 +0000)]
Branch b1_6
b=13358
i=wangdi
i=nikita
Impossible (EPROTO is returned) to access a directory that has a
non-default striping and ACLs.
johann [Fri, 14 Sep 2007 12:57:01 +0000 (12:57 +0000)]
Branch b1_6
b=13181
i=alex
i=nikita
Don't take the BKL in fsfilt_ext3_setattr() for 2.6 kernels.
It causes scheduling issues when removing large files.
grev [Fri, 14 Sep 2007 12:15:17 +0000 (12:15 +0000)]
b=12460
i=Adilger
Skip "cd ." part for patchless.
yury [Fri, 14 Sep 2007 10:25:19 +0000 (10:25 +0000)]
- cleanups in ldlm_pool.c - all pools recalc is moved to separate function ldlm_pools_recalc();
- grant_speed, grant_rate, garnt_plan, cancel_rate have own procfs functions which take into account possible schedule lag for pools thread.
shadow [Fri, 14 Sep 2007 07:58:20 +0000 (07:58 +0000)]
Kernels after 2.6.16 not allow resotre s_dev from put_super, fix it.
b=13304
i=johann
i=green
deen [Thu, 13 Sep 2007 18:45:05 +0000 (18:45 +0000)]
Never resend AST requests.
b=13596
i=adilger
i=deen
original patch by green
yury [Thu, 13 Sep 2007 18:09:55 +0000 (18:09 +0000)]
- fixed issue with long time not being scheduled pools thread. We need to compensate that in lprocfs_rd_pool_state()
kalpak [Thu, 13 Sep 2007 11:35:05 +0000 (11:35 +0000)]
b=10555
Use correct structure name.
bobijam [Thu, 13 Sep 2007 03:11:44 +0000 (03:11 +0000)]
Branch b1_6
b=13600
i=kalpak, bobijam
original patch by adilger.
Description: "lfs find -obd UUID" prints directories
Details : "lfs find -obd UUID" will return all directory names instead
of just file names. It is incorrect because the directories
do not reside on the OSTs.
liuy [Thu, 13 Sep 2007 02:24:50 +0000 (02:24 +0000)]
Branch b1_6
b=2369
i=johann, deen
replace inode->i_size access with i_size_read/write
scjody [Thu, 13 Sep 2007 01:38:27 +0000 (01:38 +0000)]
Branch b1_6
Remove client patches from SLES 10 kernel. This causes SLES 10 clients to
behave as patchless clients even on a Lustre-patched (server) kernel.
b=12411
i=adilger
i=johann
grev [Wed, 12 Sep 2007 23:21:43 +0000 (23:21 +0000)]
b=13486
i=Nathan
Increase a set of files to choose from.
adilger [Wed, 12 Sep 2007 21:50:35 +0000 (21:50 +0000)]
Branch b1_6
Move bug 13547 to 1.6.3 section instead of 1.6.4
cliffw [Wed, 12 Sep 2007 19:50:28 +0000 (19:50 +0000)]
b=12186
Update ChangeLog
cliffw [Wed, 12 Sep 2007 19:42:44 +0000 (19:42 +0000)]
b=12186
i=Adilger
Fixes man pages for lfs
nathan [Wed, 12 Sep 2007 19:18:47 +0000 (19:18 +0000)]
b=3055
i=green
changes from eeb's inspection
adilger [Wed, 12 Sep 2007 18:36:37 +0000 (18:36 +0000)]
Branch b1_6
Allow dbench to find client.txt in /usr/share/dbench.
b=12467
i=nathan
i=adilger (original patch)
scjody [Wed, 12 Sep 2007 16:42:12 +0000 (16:42 +0000)]
Branch b1_6
Add
--------------------------------------------------------------------------------
separator
komaln [Wed, 12 Sep 2007 12:31:08 +0000 (12:31 +0000)]
b=11230
Quiet a warning.
komaln [Wed, 12 Sep 2007 11:55:04 +0000 (11:55 +0000)]
b=11230
r=Nathan, Adilger
To tune the kernel for good SCSI performance.
yury [Wed, 12 Sep 2007 09:15:08 +0000 (09:15 +0000)]
- make lru resize stuff in sanity.sh less verbose;
- cleanups in tests 120* and test 124 about disable/enable lru resize;
- make test_124 sleep for 30s to not make it too long.
liuy [Wed, 12 Sep 2007 03:29:20 +0000 (03:29 +0000)]
Branch b1_6
b=13125
replace stripe_cnt_min with *stripe_cnt to keep good object distribution
bobijam [Wed, 12 Sep 2007 02:22:58 +0000 (02:22 +0000)]
Branch b1_6
b=13555
disable test_36 for now.
nathan [Wed, 12 Sep 2007 00:33:14 +0000 (00:33 +0000)]
b=13552
wait longer for remote clients
scjody [Tue, 11 Sep 2007 23:33:20 +0000 (23:33 +0000)]
Branch b1_6
Update versions for 1.6.4
scjody [Tue, 11 Sep 2007 23:17:17 +0000 (23:17 +0000)]
Branch b1_6
Update versions and dates.
wangdi [Tue, 11 Sep 2007 21:00:02 +0000 (21:00 +0000)]
Branch: b1_6
b=13436
Only disconnect errors(ENOTCONNECT & ENODEV) should be returned by rq_status.
i=green
i=adiger
johann [Tue, 11 Sep 2007 18:35:34 +0000 (18:35 +0000)]
Branch b1_6
Disable journal checksum patch due to release-blocking problems with
quota and other errors (LTS does not support # in the series files).
b=13487
b=13449
r=scjody
johann [Tue, 11 Sep 2007 18:08:23 +0000 (18:08 +0000)]
Branch b1_6
Disable journal checksum patch due to release-blocking problems with
quota and other errors
b=13487
b=13449
r=scjody
yury [Tue, 11 Sep 2007 11:15:27 +0000 (11:15 +0000)]
- zero out GS, GR and CR for client pools also in recalc.
yury [Tue, 11 Sep 2007 09:34:16 +0000 (09:34 +0000)]
- disable/enable all namespace pools matched with pass argument in sanity.sh;
- disable all pools (and OST ones too) in ELC tests;
- fixes in test_124:
- make test_124 less stressing, create only 1000 locks in it;
- calculate LVF correctly, that is, based on lock number created, desired sleep time and current limit instead of cardcoded value;
- make test suitable for running CMD environment, where desired set of data may be shared between few MDSes.
yury [Mon, 10 Sep 2007 15:10:23 +0000 (15:10 +0000)]
- cleanup in ldlm_cancel_lru_local()
yury [Mon, 10 Sep 2007 14:49:10 +0000 (14:49 +0000)]
- add OBD_CONNECT_LRU_RESIZE for mds->ost connections. MDS does not take locks on OST, but this flag will allow all namespaces look same way if LRU_RESIZE is supported.
yury [Mon, 10 Sep 2007 14:00:20 +0000 (14:00 +0000)]
- fixed bug in ldlm_cancel_lru_local(). It should cancel locks according to SLV only in the case of passed zero @count. If @count is not zero - cancel requested number of locks regardless SLV. Found by failure of sanityN test_20;
- raise error and return -EINVAL if ldlm_cancel_lru() did not cancel requested number of locks in case of sync cancel if lru resize is supported.
adilger [Mon, 10 Sep 2007 08:10:01 +0000 (08:10 +0000)]
Branch b1_6
Add "df" so the statfs data will be updated on disk (when statfs patches
are landed).
b=13198
adilger [Sat, 8 Sep 2007 18:36:40 +0000 (18:36 +0000)]
Branch b1_6
Update ancient versioning doc to match current policy.
yury [Sat, 8 Sep 2007 17:13:20 +0000 (17:13 +0000)]
- fixed disabling pool shrinker in test 120a.
adilger [Fri, 7 Sep 2007 19:40:12 +0000 (19:40 +0000)]
Branch b1_6
Add linefeed so message isn't confusingly overwritten.
b=13547
nikita [Fri, 7 Sep 2007 18:09:58 +0000 (18:09 +0000)]
Severity : cleanup
Bugzilla : 13532
Description: rewrite ext2-derived code in llite/dir.c and obdclass/uuid.c
Details : rewrite inherited code (uuid parsing code from ext2 utils and
readdir code from ext3) from scratch preserving functionality.
b=13532
i=adilger
i=alex
i=green
nathan [Fri, 7 Sep 2007 17:43:25 +0000 (17:43 +0000)]
b=13542
i=deen
i=tianzy
the reply state was leaked in the no_reply case
mjmac [Fri, 7 Sep 2007 10:37:57 +0000 (10:37 +0000)]
b=12348
i=brian
i=scjody
* Include disk1_4.zip in lustre-tests RPM
yury [Fri, 7 Sep 2007 09:32:25 +0000 (09:32 +0000)]
b=2262
r=adilger,vitaly,nikita
- landed lru resize.
kalpak [Fri, 7 Sep 2007 08:53:48 +0000 (08:53 +0000)]
b=13128
i=adilger,shadow
Minor cleanups in lfs find and lfs getstripe
scjody [Thu, 6 Sep 2007 19:03:06 +0000 (19:03 +0000)]
Branch b1_6
Strip trailing whitespace.
vitaly [Thu, 6 Sep 2007 12:51:09 +0000 (12:51 +0000)]
Branch b1_6
b=13060
i=huanghua
i=tappro
put the lock after unlocking the resource