Whamcloud - gitweb
rread [Thu, 22 Apr 2004 21:49:13 +0000 (21:49 +0000)]
b=2945
It's possible for df to fail here, instead of returning a smaller fs
size. In either case, it means (most likely) we've been evicted.
rread [Thu, 22 Apr 2004 21:43:41 +0000 (21:43 +0000)]
b=2410
Check for NULL reply in mdc_close.
cvs2svn [Tue, 20 Apr 2004 14:34:10 +0000 (14:34 +0000)]
This commit was manufactured by cvs2svn to create branch 'b1_2'.
eeb [Tue, 20 Apr 2004 14:34:09 +0000 (14:34 +0000)]
* Landed b_cray_portals_merge (3148, 3158)
buffalo-pull [Tue, 20 Apr 2004 02:48:09 +0000 (02:48 +0000)]
- tagging RC_CURRENT
ccooper [Mon, 19 Apr 2004 19:31:39 +0000 (19:31 +0000)]
- target and config files for recent chaos kernels (for tinderbox)
ccooper [Mon, 19 Apr 2004 14:29:47 +0000 (14:29 +0000)]
- fix syntax error
adilger [Sun, 18 Apr 2004 22:50:33 +0000 (22:50 +0000)]
Don't try to put a hyphen into the Lustre RPM release version.
phil [Sat, 17 Apr 2004 05:05:06 +0000 (05:05 +0000)]
2.4.21-suse-171 patches and series
jacob [Sat, 17 Apr 2004 04:33:25 +0000 (04:33 +0000)]
This is HEAD.
jacob [Sat, 17 Apr 2004 04:17:31 +0000 (04:17 +0000)]
Fix --disable-modules build.
zab [Sat, 17 Apr 2004 04:11:08 +0000 (04:11 +0000)]
- add a cute script that uses 'ploticus' to graph the rpcs from a D_RPCTRACE
log. the script should be easy to use and has pointers to ploticus.
http://tetsuo.zabbo.net/~zab/rpc-graph.ps
is an example.
buffalo-pull [Sat, 17 Apr 2004 03:20:36 +0000 (03:20 +0000)]
- tagging RC_CURRENT
adilger [Fri, 16 Apr 2004 23:39:49 +0000 (23:39 +0000)]
Land b_smallfix onto HEAD (20040416_1638) (more 2.6 build fixes)
b=2912/2922, b=3145 too
adilger [Fri, 16 Apr 2004 17:38:58 +0000 (17:38 +0000)]
Landing HEAD changes for 2.6 onto b1_2 - kernel patches only.
This gets the diff under half a million lines ;-).
adilger [Fri, 16 Apr 2004 16:55:11 +0000 (16:55 +0000)]
Fix braino.
b=3159
r=zab
adilger [Fri, 16 Apr 2004 01:32:44 +0000 (01:32 +0000)]
Update build version.
adilger [Fri, 16 Apr 2004 01:02:14 +0000 (01:02 +0000)]
Print out lock handle that causes assertion failure.
b=3159
buffalo-pull [Thu, 15 Apr 2004 03:18:25 +0000 (03:18 +0000)]
- tagging RC_CURRENT
cvs2svn [Wed, 14 Apr 2004 20:39:46 +0000 (20:39 +0000)]
This commit was manufactured by cvs2svn to create branch 'b1_2'.
adilger [Wed, 14 Apr 2004 20:39:45 +0000 (20:39 +0000)]
Land b_smallfix onto HEAD (20040414_1359)
b=3116, b=3064, b=3107, b=3122, b=1519, 2.6 kernel fixes, group lock fixes
adilger [Wed, 14 Apr 2004 20:36:54 +0000 (20:36 +0000)]
Fix the RPC lprocfs stats. The lprocfs entries need to be created before
we call ptlrpc_init_svc() or it will silently not create any "stats" entry.
b=1519
adilger [Wed, 14 Apr 2004 20:29:30 +0000 (20:29 +0000)]
Add a common error handling function for AST errors. This fixes a case where
we don't drop locks on failed completion ASTs and end up evicting the client.
b=3145
nic [Wed, 14 Apr 2004 19:01:38 +0000 (19:01 +0000)]
fix compiler warning
adilger [Wed, 14 Apr 2004 16:33:17 +0000 (16:33 +0000)]
Don't dereference negative dentry.
b=3106
buffalo-pull [Tue, 13 Apr 2004 23:26:40 +0000 (23:26 +0000)]
- tagging RC_CURRENT
adilger [Tue, 13 Apr 2004 19:08:29 +0000 (19:08 +0000)]
Update build version.
adilger [Tue, 13 Apr 2004 18:38:14 +0000 (18:38 +0000)]
Add missing part of bug 2706 fix.
b=2706
r=phil
adilger [Tue, 13 Apr 2004 18:34:28 +0000 (18:34 +0000)]
Only create lprocfs entries after a device is set up (instead of at attach
time) so that they don't try to use initialized data.
b=1519
buffalo-pull [Tue, 13 Apr 2004 04:04:01 +0000 (04:04 +0000)]
- tagging RC_CURRENT
eeb [Tue, 13 Apr 2004 01:58:28 +0000 (01:58 +0000)]
* Landed b_cray_portals_merge.
Superbug 2871:
Sub-bugs fixed 2869 2870 3034 3036 3037 3035 3038 3135
nic [Mon, 12 Apr 2004 21:45:14 +0000 (21:45 +0000)]
land b_inodebits
adilger [Mon, 12 Apr 2004 21:29:37 +0000 (21:29 +0000)]
Fix resend race during recovery (don't set rq_timeout = 0).
b=2950
r=robert
adilger [Mon, 12 Apr 2004 19:54:43 +0000 (19:54 +0000)]
Don't wait forever in sleep_on() if two threads call portals_debug_dumplog()
at the same time.
b=3107
r=zab
adilger [Thu, 8 Apr 2004 19:46:57 +0000 (19:46 +0000)]
Disable ll_readpage() log dumping (for debugging bug 2805), it caused bug 3122.
b=3122
adilger [Thu, 8 Apr 2004 19:12:01 +0000 (19:12 +0000)]
Don't dereference an ERR_PTR() in the mds_reint_link() error handling path.
b=3107
r=phil
phil [Thu, 8 Apr 2004 18:44:12 +0000 (18:44 +0000)]
Point people at the autolustre rpms, too
nic [Thu, 8 Apr 2004 17:19:51 +0000 (17:19 +0000)]
add entry for 3118
nic [Thu, 8 Apr 2004 16:55:55 +0000 (16:55 +0000)]
b=3118
r=adilger
fix badly initialized nameidata. Bug found and fix reported by 'johann lombardi <johann.lombardi@bull.net>'
adilger [Wed, 7 Apr 2004 21:53:25 +0000 (21:53 +0000)]
Update build tag.
adilger [Wed, 7 Apr 2004 21:53:00 +0000 (21:53 +0000)]
Fix patch breakage.
adilger [Wed, 7 Apr 2004 21:52:05 +0000 (21:52 +0000)]
Fix patch breakage.
adilger [Wed, 7 Apr 2004 21:46:19 +0000 (21:46 +0000)]
Fix duplicate merge of this hunk.
jacob [Wed, 7 Apr 2004 19:12:05 +0000 (19:12 +0000)]
land b_smallfix 20040407_1414:
b=3057, 2947, 2925, 3050, 3041, 2849, 1662
zab [Wed, 7 Apr 2004 17:31:23 +0000 (17:31 +0000)]
- put the exit(1)s back in after MPI_Abort() in the hopes that MPI
implementations which don't exit in MPI_Abort() will honor the exit(1)s.
eeb [Tue, 6 Apr 2004 22:45:41 +0000 (22:45 +0000)]
* changed ptlrpc_unregister_reply(), ptlrpc_unregister_bulk() and
ptlrpc_abort_bulk() to always l_wait_event() after doing
PtlMDUnlink() on the relevant MD. Previous code assumed that when
PtlMDUnlink() returned PTL_MD_INVALID, the relevant callback handler
must have run already, which isn't true for liblustre. The
l_wait_event() will be a NOOP for kernel lustre (condition already
true) but gives the liblustre event handler a chance to call all
the callbacks.
adilger [Tue, 6 Apr 2004 22:29:15 +0000 (22:29 +0000)]
In ll_file_read() use (*ppos + count - 1) for end of read instead of the value
in policy->l_extent.end. This makes the intent of the comparison more clear,
and avoids any problems if the policy data is ever changed (e.g. to return the
granted extent or similar).
Useful debugging for lock extent problems.
b=2925
adilger [Tue, 6 Apr 2004 22:24:54 +0000 (22:24 +0000)]
Minor correctness fixes for EA compat support:
- print the right size in the error message
- work correctly some day in the future when we have v2 EAs
b=3059
adilger [Tue, 6 Apr 2004 19:10:44 +0000 (19:10 +0000)]
Use the highest extent to determine the kms value, not the last one.
Try not to grant overlapping locks for the same client, to speed lock cancel.
b=2925
r=phil
nic [Mon, 5 Apr 2004 19:43:35 +0000 (19:43 +0000)]
b=3054
r=zab
add case for 2.6 and /proc/sys/vm/dirty_writeback_centisecs where 2.4 uses kupdated
eeb [Mon, 5 Apr 2004 19:26:16 +0000 (19:26 +0000)]
* Fix for 2895
adilger [Mon, 5 Apr 2004 18:36:21 +0000 (18:36 +0000)]
Update version tag.
adilger [Mon, 5 Apr 2004 18:19:03 +0000 (18:19 +0000)]
Debugging for mdc_close oops.
b=2410
r=robert
adilger [Mon, 5 Apr 2004 18:14:48 +0000 (18:14 +0000)]
Don't update to the kms value for a lock that is just in the process of
being cancelled.
b=3057
r=phil
phil [Sun, 4 Apr 2004 20:36:38 +0000 (20:36 +0000)]
b=3106
Don't deref NULL inode if we make it into filter_lvbo_init for (in
this case) the wrong object group
adilger [Sat, 3 Apr 2004 19:28:33 +0000 (19:28 +0000)]
NULL the filter arrays as they are freed to avoid potential problems on cleanup.
r=phil
adilger [Sat, 3 Apr 2004 19:27:21 +0000 (19:27 +0000)]
Initialize lock handle so that we don't fail ASSERT on error cleanup.
b=3057
r=phil
adilger [Sat, 3 Apr 2004 04:40:53 +0000 (04:40 +0000)]
Quiet compiler warning.
b=1557
adilger [Sat, 3 Apr 2004 04:39:18 +0000 (04:39 +0000)]
Don't dereference uninitialized variable.
b=2543
buffalo-pull [Sat, 3 Apr 2004 04:23:14 +0000 (04:23 +0000)]
- tagging RC_CURRENT
nic [Fri, 2 Apr 2004 21:21:53 +0000 (21:21 +0000)]
create linux/configs if not present
buffalo-pull [Fri, 2 Apr 2004 02:11:34 +0000 (02:11 +0000)]
- tagging RC_CURRENT
adilger [Thu, 1 Apr 2004 19:50:19 +0000 (19:50 +0000)]
Change default stripe size based on ALC/ADEV testing.
adilger [Thu, 1 Apr 2004 12:04:48 +0000 (12:04 +0000)]
Minor changes to sync b1_2 with HEAD that shouldn't affect anything.
adilger [Thu, 1 Apr 2004 12:03:07 +0000 (12:03 +0000)]
Minor changes to sync b1_2 with HEAD that shouldn't affect anything.
adilger [Thu, 1 Apr 2004 11:41:45 +0000 (11:41 +0000)]
Disable tests for which fixes are not currently in b1_2.
adilger [Thu, 1 Apr 2004 06:42:31 +0000 (06:42 +0000)]
Fix sanity test.
b=1414
jacob [Thu, 1 Apr 2004 04:09:23 +0000 (04:09 +0000)]
b=3041
re-add things CVS mistakenly removed.
jacob [Thu, 1 Apr 2004 04:06:36 +0000 (04:06 +0000)]
b=3104
Do not cvs add/rm on one branch, then cvs up -r to another branch, and
expect CVS to cope.
jacob [Thu, 1 Apr 2004 03:34:59 +0000 (03:34 +0000)]
b=3041
b=3058
Clean up the autoconf and automake bits:
* now liblustre and kernel modules can be built in the same tree
use --disable-modules, --disable-liblustre to disable those
bits
* 2.6 modules are now built as .ko, and lconf can deal with them
* fixes for builddir != srcdir
buffalo-pull [Thu, 1 Apr 2004 03:25:01 +0000 (03:25 +0000)]
- tagging RC_CURRENT
adilger [Thu, 1 Apr 2004 00:52:40 +0000 (00:52 +0000)]
Don't allow setstripe on a file/directory without write permission for same.
Regression tests for setstripe permission check.
b=1414
adilger [Thu, 1 Apr 2004 00:48:48 +0000 (00:48 +0000)]
Fix mismerge from HEAD.
nic [Thu, 1 Apr 2004 00:31:26 +0000 (00:31 +0000)]
b=3082
r=andreas
sys_fchmod was missing setattr_raw magic, add chmod_common for both sys_fchmod and sys_chmod
adjust locking around chown_common() setattr_raw RPC, as i_sem need to be held over setattr
adilger [Wed, 31 Mar 2004 22:36:42 +0000 (22:36 +0000)]
Merge from HEAD to b1_2 select fixes for 1.2.2
b=2922, b=2948, b=1414
adilger [Wed, 31 Mar 2004 22:26:54 +0000 (22:26 +0000)]
b=2948
adilger [Wed, 31 Mar 2004 20:37:02 +0000 (20:37 +0000)]
Merge fixes from b_smallfix to b1_2 for 1.2.2
b=2947, b=2543, b=1557, b=2009
adilger [Wed, 31 Mar 2004 19:38:09 +0000 (19:38 +0000)]
Keep tests in sync with HEAD.
adilger [Wed, 31 Mar 2004 19:30:21 +0000 (19:30 +0000)]
Bump kernel patch version to 34, to match HEAD.
b=2517, b=2399 are the changes for this kernel version
adilger [Wed, 31 Mar 2004 18:18:04 +0000 (18:18 +0000)]
Land b_smallfix onto HEAD (20040330_2312)
b=2947, b=2543, b=1557, b=2009
adilger [Wed, 31 Mar 2004 04:24:00 +0000 (04:24 +0000)]
Fix test errors for 6 & 7 (wasn't waiting for async delete to complete).
b=2921
adilger [Wed, 31 Mar 2004 04:11:48 +0000 (04:11 +0000)]
Remove ancient files.
buffalo-pull [Wed, 31 Mar 2004 01:17:54 +0000 (01:17 +0000)]
- tagging RC_CURRENT
adilger [Wed, 31 Mar 2004 01:02:01 +0000 (01:02 +0000)]
Don't copy lvb into possibly NULL reply on ldlm_handle_enqueue() error
client won't use it anyways. Regression test for same.
Fix secondary bug hit during regression tests in mds_mfd_close() cleanup.
b=2983
adilger [Tue, 30 Mar 2004 21:28:34 +0000 (21:28 +0000)]
b=1414
buffalo-pull [Tue, 30 Mar 2004 03:26:44 +0000 (03:26 +0000)]
- tagging RC_CURRENT
nic [Mon, 29 Mar 2004 16:57:21 +0000 (16:57 +0000)]
change function declarations to correctly reflect the MPI_Abort exiting, remove
exits. this is what should have been done in the first place...yeah.
ericm [Mon, 29 Mar 2004 06:02:28 +0000 (06:02 +0000)]
liblustre:
- support VPATH build (still some unclean stuff in liblustre build)
buffalo-pull [Sat, 27 Mar 2004 03:06:18 +0000 (03:06 +0000)]
- tagging RC_CURRENT
nic [Fri, 26 Mar 2004 16:35:50 +0000 (16:35 +0000)]
fix compiler warnings
hitao [Fri, 26 Mar 2004 08:33:13 +0000 (08:33 +0000)]
to get Lustre built with 2.6.3+mm4, LTIME_S(dentry->d_inode->i_size) is not correct
buffalo-pull [Fri, 26 Mar 2004 00:39:39 +0000 (00:39 +0000)]
- tagging RC_CURRENT
adilger [Thu, 25 Mar 2004 20:16:58 +0000 (20:16 +0000)]
Don't free resource lvb if there is a lookup error (causes later assertion).
Don't dereference dentry after dputting it.
Remove use of obdo in filter_lvb, since we don't actually need it.
b=2922
adilger [Thu, 25 Mar 2004 20:08:53 +0000 (20:08 +0000)]
loff_t is a long long in the kernel, liblustre should use the same to
avoid compiler warnings on 64-bit (this #define is for windows port only).
adilger [Thu, 25 Mar 2004 18:42:47 +0000 (18:42 +0000)]
Move test 66[ab] from sanity.sh to recovery-small.sh because we need to do
ops on the OST, and that isn't possible with sanity.sh if it is remote.
Fix test replay-ost-single test 7 for bug 2123 also.
b=2983, b=2123
buffalo-pull [Thu, 25 Mar 2004 04:33:11 +0000 (04:33 +0000)]
- tagging RC_CURRENT
zab [Thu, 25 Mar 2004 01:33:16 +0000 (01:33 +0000)]
b=2985
r=zab
use MPI_Abort() in the exiting error paths of write_* so that other processes
are pulled out of MPI_Barrier(). I added a -f <filename> argument to
write_disjoint while I was veryfying this.
adilger [Wed, 24 Mar 2004 09:21:03 +0000 (09:21 +0000)]
Fix for bug 2921 (test failure only). We had coincidentally created a file
in test 4 with the same size, and if we didn't wait for its deletion then
it could so happen that the "kbytesfree before" would equal "kbytesfree after"
since that file was freed while we wrote the new one.
b=2921
adilger [Wed, 24 Mar 2004 05:40:47 +0000 (05:40 +0000)]
Don't copy lvb into reply message on error, since that message might not have
allocated enough space for the lvb and the client will ignore it anyways.
Also fixes a NULL request deref in mds_mfd_close() when cleaning up opens
on the MDS. The initial sanity test (66b) triggered a new bug 2986, but
another test with multiop tests this properly.
b=2983
buffalo-pull [Wed, 24 Mar 2004 03:08:43 +0000 (03:08 +0000)]
- tagging RC_CURRENT