Whamcloud - gitweb
adilger [Wed, 17 Mar 2004 11:46:23 +0000 (11:46 +0000)]
Revert the reservation part of the bug 2059 changes. This seems to have
been causing problems on the OSTs and wasn't the primary reason for applying
this patch to 1.2.
b=2059
adilger [Wed, 17 Mar 2004 10:53:44 +0000 (10:53 +0000)]
Clean up properly if llog_lvfs_close() fails.
b=2325
adilger [Wed, 17 Mar 2004 09:33:32 +0000 (09:33 +0000)]
Fix assertion again, sigh.
b=2925
adilger [Wed, 17 Mar 2004 08:21:51 +0000 (08:21 +0000)]
Fix assertion for end of loop.
b=2925
adilger [Wed, 17 Mar 2004 05:37:49 +0000 (05:37 +0000)]
Reduce number of credits for an UNLINK op (update quota files at most once).
Limit our transaction requests to the maximum transaction size regardless.
Handle error case in journal start.
b=2059
adilger [Wed, 17 Mar 2004 04:08:51 +0000 (04:08 +0000)]
Fix reservation vs. grant locking.
b=2059
adilger [Wed, 17 Mar 2004 01:33:04 +0000 (01:33 +0000)]
Update build version.
adilger [Wed, 17 Mar 2004 01:32:17 +0000 (01:32 +0000)]
Automatically generate fsfilt_extN.c from fsfilt_ext3.c
b=2059
cvs2svn [Wed, 17 Mar 2004 01:04:44 +0000 (01:04 +0000)]
This commit was manufactured by cvs2svn to create branch 'b1_2'.
nic [Wed, 17 Mar 2004 01:04:43 +0000 (01:04 +0000)]
land b_v26 (20040316_1603)
adilger [Wed, 17 Mar 2004 01:02:59 +0000 (01:02 +0000)]
Update b1_2 from HEAD (20040316_1727)
b=2939
rread [Wed, 17 Mar 2004 00:26:38 +0000 (00:26 +0000)]
b=2936
Skip closing files for the self_export, and add a test.
nic [Wed, 17 Mar 2004 00:21:47 +0000 (00:21 +0000)]
add RHBUILD to targets and spec to distinguish files that come from using a redhat patched kernel
typo fix to have lbuild error on building rpms
removed erroneous EXTRA_VERSION manipulation
adilger [Tue, 16 Mar 2004 23:37:34 +0000 (23:37 +0000)]
Don't evict page beyond end of each stripe. In rare cases this page was
being waited on for a read and the VFS is very unhappy when it isn't
uptodate when the read completes. Rightfully so. This was introduced during
bug 2765 changes to fix the extent calculations.
b=2925
adilger [Tue, 16 Mar 2004 23:13:19 +0000 (23:13 +0000)]
This changes the fsfilt transaction calculations to take into account the
cases where we are updating multiple log files for a single transaction
(esp. unlink of files). For small numbers of stripes this isn't very
likely to cause problems, but after updating to the new log format for 2306,
we happened to run single-file IORs with 32 stripes as the first IO to the
filesystem, and this caused 32 new logs to be created in a single transaction
when that file was unlinked. This caused the unlink journal transaction to
repeatedly run out of credits and BUG the MDS until this patch was applied.
Removed fsfilt_extN.c since it is perpetually out-of-date and any time it
needs to be used it is easier to just run fsfilt_ext3.c through sed.
b=2059, b=2931
zab [Tue, 16 Mar 2004 23:06:38 +0000 (23:06 +0000)]
file pers.c was initially added on branch b_cray_portals_merge.
adilger [Tue, 16 Mar 2004 22:37:33 +0000 (22:37 +0000)]
Hold dentry reference over log close so it is valid for unlink.
b=2325
r=robert
adilger [Tue, 16 Mar 2004 21:13:48 +0000 (21:13 +0000)]
Don't spew errors during readahead.
nic [Tue, 16 Mar 2004 18:45:26 +0000 (18:45 +0000)]
remove unneccessary debugging
adilger [Tue, 16 Mar 2004 16:05:48 +0000 (16:05 +0000)]
Correct number of journal credits calculated for FSFILT_OP_CANCEL_UNLINK_LOG
so that we don't have transaction errors on small MDS filesystems.
b=2931
r=tianying
adilger [Tue, 16 Mar 2004 08:48:13 +0000 (08:48 +0000)]
b=2926
wangchao [Tue, 16 Mar 2004 04:42:55 +0000 (04:42 +0000)]
b=2939
r=shaver
let lustre could be mounted with the same name for node and mds
buffalo-pull [Tue, 16 Mar 2004 01:06:33 +0000 (01:06 +0000)]
- tagging RC_CURRENT
adilger [Mon, 15 Mar 2004 21:58:20 +0000 (21:58 +0000)]
Update b1_2 from HEAD (20040315_1445)
b=2898, b=2926
adilger [Mon, 15 Mar 2004 21:43:35 +0000 (21:43 +0000)]
b=2464
adilger [Mon, 15 Mar 2004 21:38:50 +0000 (21:38 +0000)]
Compile warning fixes for ia64.
adilger [Mon, 15 Mar 2004 20:58:22 +0000 (20:58 +0000)]
Don't spew LustreErrors for problems during readahead.
green [Mon, 15 Mar 2004 20:55:58 +0000 (20:55 +0000)]
b=2926
r=phik
Make mds_reint_create to use mds_lock_new_child everywhere instead of taking
the lock (and incorrect one) on its own.
yury [Mon, 15 Mar 2004 16:38:55 +0000 (16:38 +0000)]
*** empty log message ***
yury [Mon, 15 Mar 2004 07:17:46 +0000 (07:17 +0000)]
*** empty log message ***
zab [Sat, 13 Mar 2004 02:30:47 +0000 (02:30 +0000)]
- upping the spec to 1.2.0.4 in preparation for tagging
rread [Sat, 13 Mar 2004 02:18:48 +0000 (02:18 +0000)]
b=2898
Update ChangeLog.
buffalo-pull [Sat, 13 Mar 2004 02:14:35 +0000 (02:14 +0000)]
- tagging RC_CURRENT
rread [Sat, 13 Mar 2004 02:13:23 +0000 (02:13 +0000)]
b=2898
Protect conn_cnt check and update with exp_lock in
target_handle_connect.
Add a test to replay since, and add an OBD_FAIL style macro to create
a race condition in a specific location called OBD_RACE.
zab [Fri, 12 Mar 2004 19:50:27 +0000 (19:50 +0000)]
file kpr.h was initially added on branch b_cray_portals_merge.
cvs2svn [Fri, 12 Mar 2004 17:50:37 +0000 (17:50 +0000)]
This commit was manufactured by cvs2svn to create branch 'b1_2'.
nic [Fri, 12 Mar 2004 17:50:36 +0000 (17:50 +0000)]
add target and configuration files for hp_pnnl on ia64
nic [Fri, 12 Mar 2004 17:47:32 +0000 (17:47 +0000)]
fix lmake to link to correct ia64 kernels
allow lustre-kernel-2.4.spec to be built on ia64
buffalo-pull [Fri, 12 Mar 2004 04:46:14 +0000 (04:46 +0000)]
- tagging RC_CURRENT
yury [Thu, 11 Mar 2004 17:59:27 +0000 (17:59 +0000)]
Removed fsfilt_tmpfs.c from HEAD branch as it was got there mistakenly before its time came ;) This is because it was cvs add'ed before creating b_tmpfs branch.
yury [Thu, 11 Mar 2004 17:44:12 +0000 (17:44 +0000)]
Cleanups about readdir stuff. ext2_dirent and macros has moved to lustre_dir.h. Related chnages in llite and fsfilt_tmpfs.c
yury [Thu, 11 Mar 2004 08:11:43 +0000 (08:11 +0000)]
Cleanups and fixes at Andreas' review results.
jacob [Thu, 11 Mar 2004 04:29:11 +0000 (04:29 +0000)]
Merge HEAD to b1_2.
b=2306, 2325, 2264
Fixes for supporting 8k llog chunks:
- fix destroying of named logs (2325)
- overwrite old logs when running lconf --write_conf (2264)
- bump LLOG_CHUNKSIZE to 8k to allow for larger clusters (2306)
jacob [Thu, 11 Mar 2004 03:41:37 +0000 (03:41 +0000)]
merge error
jacob [Wed, 10 Mar 2004 22:31:06 +0000 (22:31 +0000)]
b=2306
r=phil, andreas
bump LLOG_CHUNKSIZE to 8k to allow for larger clusters
jacob [Wed, 10 Mar 2004 22:28:29 +0000 (22:28 +0000)]
b=2264
r=robert
overwrite log when running lconf --write_conf
jacob [Wed, 10 Mar 2004 22:25:05 +0000 (22:25 +0000)]
b=2325
r=robert
fix llog_destroy to work with named logs
nic [Wed, 10 Mar 2004 20:01:20 +0000 (20:01 +0000)]
b=2914
r=jacob
remove __attribute__ ((packed)) from ptl_event_t to fix 'kernel unaligned
access...' messages on ia64
cvs2svn [Wed, 10 Mar 2004 07:50:31 +0000 (07:50 +0000)]
This commit was manufactured by cvs2svn to create branch 'b1_2'.
yury [Wed, 10 Mar 2004 07:50:30 +0000 (07:50 +0000)]
added tmpfs related stuff
buffalo-pull [Wed, 10 Mar 2004 05:27:40 +0000 (05:27 +0000)]
- tagging RC_CURRENT
adilger [Wed, 10 Mar 2004 01:19:30 +0000 (01:19 +0000)]
Update b1_2 from HEAD (20040309_1747)
b=2818, b=2901, b=2663, b=2908, b=2530, b=2464, b=2306 (socknal zc part)
adilger [Wed, 10 Mar 2004 01:02:13 +0000 (01:02 +0000)]
Fix desctription of configure option.
rread [Tue, 9 Mar 2004 22:47:17 +0000 (22:47 +0000)]
Don't LBUG the MDS if orphan cleanup requests times out; instead allow
the individual OSC to recover on it's own.
b=1530
adilger [Tue, 9 Mar 2004 20:06:10 +0000 (20:06 +0000)]
Delete obd_self_export from exp_obd_chain while we are disconnecting exports
during a forced cleanup. Otherwise, we will end up pointing to work_list,
which is on the stack, and a list_del_init() on exp_obd_chain will end up
stomping on some innocent variable on the stack, ending its already-short
life just a bit too early.
b=2908
r=robert
hitao [Tue, 9 Mar 2004 15:41:55 +0000 (15:41 +0000)]
file suse-2.6.3 was initially added on branch b_v26.
hitao [Tue, 9 Mar 2004 15:41:49 +0000 (15:41 +0000)]
file vfs_nointent_2.6.3-suse.patch was initially added on branch b_v26.
hitao [Tue, 9 Mar 2004 15:41:48 +0000 (15:41 +0000)]
file vfs_intent-2.6.3-suse.patch was initially added on branch b_v26.
hitao [Tue, 9 Mar 2004 15:41:47 +0000 (15:41 +0000)]
file nfs-cifs-intent-2.6.3-suse.patch was initially added on branch b_v26.
adilger [Tue, 9 Mar 2004 09:36:08 +0000 (09:36 +0000)]
Don't allocate zero-length LVBs on the MDS. Verified that all users of
l_lvb_data check l_lvb_len != 0 before using, and tested of course.
b=1012
uid707 [Tue, 9 Mar 2004 08:47:20 +0000 (08:47 +0000)]
Set loi_kms_valid even if we get a lock back that has rss == kms. This
happens for zero-length files, and without this we will revoke our own
lock the next time we write to the file because osc_enqueue() won't match.
b=2818
adilger [Tue, 9 Mar 2004 08:39:54 +0000 (08:39 +0000)]
Stop error spew if no real error, doh.
b=2633
adilger [Tue, 9 Mar 2004 08:37:35 +0000 (08:37 +0000)]
Call test_foo() functions directly instead of via run_one() so they do setup
even if a test is excluded.
Rename "test_foo" files to "f<testno>" so they are deleted at the end of
the sanity.sh run.
adilger [Tue, 9 Mar 2004 07:56:33 +0000 (07:56 +0000)]
Drop the filter export grants under the same lock as the removal of the
export from the device from the export list. This still isn't possible
for --force disconnects, so we don't do filter_grant_sanity_check() at
that time.
Also, change the LASSERTs to CERRORs in case this happens again. First of
all to see if these errors are just transient during export disconnection
which means they are just overly-aggressive assertions, or if there is a
permanent accounting error that indicates some other problem. Secondly, it
isn't possible to stop the use of existing export references, so there will
always be some small race during client eviction as completing operations
might update the export before it is destroyed.
b=2663
ericm [Tue, 9 Mar 2004 04:37:54 +0000 (04:37 +0000)]
file 2.6.3-mm4 was initially added on branch b_v26.
ericm [Tue, 9 Mar 2004 04:37:53 +0000 (04:37 +0000)]
file vfs_nointent-2.6.3-mm4.patch was initially added on branch b_v26.
ericm [Tue, 9 Mar 2004 04:37:52 +0000 (04:37 +0000)]
file iopen-2.6.3-mm4.patch was initially added on branch b_v26.
ericm [Tue, 9 Mar 2004 04:37:51 +0000 (04:37 +0000)]
file fs-intent-2.6.3-mm4.patch was initially added on branch b_v26.
jacob [Tue, 9 Mar 2004 03:32:50 +0000 (03:32 +0000)]
b=2295
r=adilger
* clean up bugs in previous patch
* print basename(argv[0]) instead of "mount" with error messages
youfeng [Tue, 9 Mar 2004 02:37:28 +0000 (02:37 +0000)]
b 2295
r Jacob
add a missing arg to printf, make the program exit with 1 on args error
youfeng [Tue, 9 Mar 2004 01:23:03 +0000 (01:23 +0000)]
b 2295
r Adilger
fix a llmount, mount.lustre segment fault bug
jacob [Tue, 9 Mar 2004 01:08:30 +0000 (01:08 +0000)]
fixup last commit
jacob [Tue, 9 Mar 2004 00:15:38 +0000 (00:15 +0000)]
r=adilger,phil
b=2464
Land initial b_bgl fixes. Includes compile and a couple of swabbing
fixes.
adilger [Mon, 8 Mar 2004 20:33:26 +0000 (20:33 +0000)]
Copy the extent lock policy data only in the completion callback, or in the
enqueue reply if the extent was changed and granted immediately. This
avoids overwriting the policy data if it was updated in the completion callback
before the reply was processed.
b=2901
r=phil
adilger [Mon, 8 Mar 2004 20:08:55 +0000 (20:08 +0000)]
Handle glimpse races where inode is evicted by client but still holding lock.
b=2818
eeb [Mon, 8 Mar 2004 17:46:43 +0000 (17:46 +0000)]
* Made echo_commitrw() pay attention to the final 'rc' parameter (bug 1450)
adilger [Mon, 8 Mar 2004 17:12:40 +0000 (17:12 +0000)]
Fix munlink messages to not make it appear like there was an unlink failure.
b=2167
adilger [Sun, 7 Mar 2004 03:17:59 +0000 (03:17 +0000)]
Increment the ost_start_idx inside the object creation loop even if there is
a failed object creation instead of in advance. Otherwise, if there is a
down OST the following one will have twice as many objects allocated on it
(observed with watch-lustre on ALC while running IOR with one OST inactive
on the MDS).
buffalo-pull [Sat, 6 Mar 2004 00:16:17 +0000 (00:16 +0000)]
- tagging RC_CURRENT
adilger [Fri, 5 Mar 2004 18:10:38 +0000 (18:10 +0000)]
Merge documentation, liblustre, obdecho, smfs changes that landed on HEAD
after b1_2 was branched to avoid merge problems later.
Things not yet landed from this period:
- bug 2308 changes to portals (including kiov changes all over the place)
- GMNAL updates from llp2 (mixed in with 2308)
Those will be landed separately.
adilger [Fri, 5 Mar 2004 18:04:40 +0000 (18:04 +0000)]
Merge documentation, liblustre, obdecho, smfs changes that landed on HEAD
after b1_2 was branched to avoid merge problems later.
Things not yet landed from this period:
- bug 2308 changes to portals (including kiov changes all over the place)
- GMNAL updates from llp2 (mixed in with 2308)
Those will be landed separately.
buffalo-pull [Fri, 5 Mar 2004 04:39:35 +0000 (04:39 +0000)]
- tagging RC_CURRENT
adilger [Fri, 5 Mar 2004 03:52:30 +0000 (03:52 +0000)]
Land b1_2 onto HEAD (20040304_171022)
b=2733, b=2773, b=2529, b=2730, b=2819, b=2814, b=2822, b=1450, b=2676, b=2681
b=2817, b=2706, b=2816, b=1987, b=2884, b=1191, b=2809, b=2765, b=2805, b=1972
adilger [Fri, 5 Mar 2004 00:58:51 +0000 (00:58 +0000)]
Remove bogus patch hunk, all is well.
adilger [Fri, 5 Mar 2004 00:44:16 +0000 (00:44 +0000)]
Fix bug in HP kernel with calling sync_fs() under the superblock lock.
b=2681
adilger [Thu, 4 Mar 2004 23:58:26 +0000 (23:58 +0000)]
Update build version.
adilger [Thu, 4 Mar 2004 23:49:27 +0000 (23:49 +0000)]
Add fields (currently unused) to ost_lvb for glimpse time/blocks updates.
b=1021
r=phil
nic [Thu, 4 Mar 2004 23:15:20 +0000 (23:15 +0000)]
b=2681
r=jacob,adilger
update kernel series for hp_pnnl, add invalidate-show patch and update
ext3-delete_thread patch for this series
phil [Thu, 4 Mar 2004 19:07:47 +0000 (19:07 +0000)]
Enlarge the LVB in advance, to avoid an obvious 1.2.0 -> 1.2.1
protocol change
adilger [Thu, 4 Mar 2004 18:29:10 +0000 (18:29 +0000)]
Quiet portals compiler warning on ia64.
b=2886
r=eeb
ericm [Thu, 4 Mar 2004 12:29:44 +0000 (12:29 +0000)]
file vanilla-2.6.3-nfs4 was initially added on branch b_v26.
ericm [Thu, 4 Mar 2004 11:07:43 +0000 (11:07 +0000)]
liblustre:
- don't build mpi programs by default, as Bob's suggestion
adilger [Thu, 4 Mar 2004 09:06:21 +0000 (09:06 +0000)]
Additional debugging.
b=2765
adilger [Thu, 4 Mar 2004 09:02:23 +0000 (09:02 +0000)]
Update build version.
adilger [Thu, 4 Mar 2004 08:33:01 +0000 (08:33 +0000)]
Fix three problems related to lock extent revocation:
- we were incorrectly calculating the "end" index value to be at the beginning
of the next stripe instead of at the end of the current stripe (extra +1 for
end, instead of using <= end for page indices)
- when comparing end << PAGE_CACHE_SHIFT for overflow against extent->end we
used an unsigned long (which overflowed during the test ;-) so we always
evicted pages to EOF (may have been a culprit in bug 2106 also)
- when matching against other locks (to see if page is under 2 locks) we
were doing so in the file offset space (from page->index) instead of in
the osc/extent space so we were matching against the wrong locks, if any
Change LL_CDEBUG_PAGE() to take a debug mask, so we can use it for error cases.
b=2765
adilger [Thu, 4 Mar 2004 08:10:13 +0000 (08:10 +0000)]
Don't call the completion AST if we can already match this lock.
We could fix some callers of obd_match() and ldlm_lock_match() to be a
bit more efficient, but we won't right now.
r=phil
buffalo-pull [Thu, 4 Mar 2004 00:04:24 +0000 (00:04 +0000)]
- tagging RC_CURRENT
ericm [Wed, 3 Mar 2004 16:52:32 +0000 (16:52 +0000)]
liblustre:
- another inode drop fix
- fix a stupid error of last checkin
ericm [Wed, 3 Mar 2004 14:37:03 +0000 (14:37 +0000)]
liblustre:
- supress a warning of glimpse callback
- remove the debug code accidently checked in sometime before
ericm [Wed, 3 Mar 2004 14:33:55 +0000 (14:33 +0000)]
liblustre:
- don't blindly drop inode
- fix iget
- root inode checking