Whamcloud - gitweb
adilger [Fri, 10 Feb 2006 19:03:16 +0000 (19:03 +0000)]
Branch b_release_1_4_6
Add ChangeLog entry for fixes previously committed.
b=9562, b=5681
eeb [Fri, 10 Feb 2006 07:10:35 +0000 (07:10 +0000)]
* Better console error messages if 'ip2nets' can't match an IP address.
* Fixed the LNET ping to handle signals properly.
* Fixed overflow/use-before-set bugs in linux-time.h
NB ???-time.h should also be checked; e.g. the same bugs had been
cut and pasted into winnt-time.h; I've fixed the use-before-set
but not the overflow issue there so I left a #error fttb.
* Fixed ptllnd bug that wasn't initialising rx descriptors completely
(sizeof(ptr) instead of sizeof(*ptr)). This wasn't evident until the
paths that catch malformed packets are exercised, so there are undoubtedly
more to come.
* Fixed a bug in LNET teardown that failed an assertion about the route
table being empty if there was an error while it was being built.
* Fixed a crash in LNetEQPoll(<invalid handle>)
ericm [Thu, 9 Feb 2006 18:49:49 +0000 (18:49 +0000)]
branch: b_release_1_4_6
add acl test scripts into EXTRA_DIST.
b=9667
r=green,brian
scjody [Thu, 9 Feb 2006 18:10:08 +0000 (18:10 +0000)]
Branch b_release_1_4_6
Update SLES 9 target and configs to 2.6.5-244. Boston has been fixed to work
with this kernel, so I'm reapplying the change.
adilger [Wed, 8 Feb 2006 18:45:33 +0000 (18:45 +0000)]
Branch b_release_1_4_6
Using mmap + lustre on a system running elan it was possible to race when
cleaning up the mmap and dereference a freed pointer.
Fix from Quadrics for the rhel-2.4.21 qsnet patch.
b=9869
adilger [Wed, 8 Feb 2006 08:51:46 +0000 (08:51 +0000)]
Branch b_release_1_4_6
Remove unused struct members.
zhanghc [Wed, 8 Feb 2006 07:54:50 +0000 (07:54 +0000)]
Land b1_4_atime_update on b_release_1_4_6
BUG: 5972
DESCRIPTION: Liblustre read not setting st_atime
Modified Files:
Tag: b_release_1_4_6
include/linux/lustre_compat25.h liblustre/rw.c
liblustre/tests/sanity.c llite/file.c llite/llite_lib.c
lov/lov_merge.c obdfilter/filter_io.c tests/replay-single.sh
tests/sanityN.sh
adilger [Wed, 8 Feb 2006 00:51:43 +0000 (00:51 +0000)]
Branch b_release_1_4_6
Disable test 39, which is failing on almost all runs. Was added for CROW.
b=9930
eeb [Tue, 7 Feb 2006 15:48:01 +0000 (15:48 +0000)]
* landing future protocol compatibility work (b_rls146_lnetprotovrsn)
on b_release_1_4_6
zhanghc [Tue, 7 Feb 2006 10:34:40 +0000 (10:34 +0000)]
land bug9972(b1_4_lov_lvb_cleanup) on b_release_1_4_6
DESCRIPTION: replace loi_rss,loi_mtime,loi_ctime,loi_ctime with struct ost_lvb.
on the same time, add obd_merge_lvb(lov_merge_lvb) into obd_ops
(lov_obd_ops)
BUG: 9972
Modified Files:
Tag: b_release_1_4_6
include/linux/lustre_lib.h include/linux/obd.h
include/linux/obd_class.h liblustre/llite_lib.h liblustre/rw.c
liblustre/super.c llite/file.c llite/llite_mmap.c llite/rw.c
lov/lov_internal.h lov/lov_merge.c lov/lov_obd.c
lov/lov_request.c obdclass/lprocfs_status.c
obdfilter/filter_lvb.c osc/osc_request.c
adilger [Tue, 7 Feb 2006 01:10:35 +0000 (01:10 +0000)]
Branch b_release_1_4_6
Description: OST may deadlock under high load on fragmented files
Details : If there was a heavy load and highly-fragmented OST filesystems
it was possible to have all the OST threads deadlock waiting on
allocation of biovecs, because the biovecs were not released
until the entire RPC IO was completed. Instead, release biovecs
as soon as they are complete to ensure forward IO progress.
b=10076
r=adilger, r=alex (original patch from Brian B.)
adilger [Mon, 6 Feb 2006 23:31:06 +0000 (23:31 +0000)]
Branch b_release_1_4_6
Don't LASSERT on client-supplied data. Likely will be avoided by recent
commit for bug 3555, 9574, et. al, but avoidable nonetheless.
b=3934, b=7384, b=9910
adilger [Mon, 6 Feb 2006 22:25:01 +0000 (22:25 +0000)]
Branch b_release_1_4_6
- Verify the client-supplied oa->o_dirty value validity before we assert on
it. Getting a negative o_dirty would cause the OST to LBUG. b=10125
- improve precreate path log messages. b=10123
adilger [Mon, 6 Feb 2006 21:05:43 +0000 (21:05 +0000)]
Branch b_release_1_4_6
Remove unused "rc" variable.
adilger [Mon, 6 Feb 2006 20:57:28 +0000 (20:57 +0000)]
Branch b_release_1_4_6
Fix unlikely bugs in mballoc found by Coverity:
- NULL bh dereferenced in ext3_mb_init_cache() if kmalloc(bh) fails
- don't free on-stack bhs in ext3_mb_init_cache() if groups_per_page == 1
- remove unnecessary "if (err)" check in ext3_mb_new_blocks()
- remove redundant "if (s == NULL)" check in ext3_mb_seq_history_open()
b=10093
adilger [Mon, 6 Feb 2006 20:44:51 +0000 (20:44 +0000)]
Branch b_release_1_4_6
When unlinking block-mapped files with the delete thread on a filesystem
with extents enabled it will trip an assertion. This is not a configuration
that CFS supports, since extents+mballoc are only used on 2.6 kernels with
CFS Lustre, and there is no 2.6 asyncdel patch. HP SFS can run with extents
and mballoc enabled on 2.4 kernels.
b=10134
r=fergal (original patch)
adilger [Mon, 6 Feb 2006 20:22:45 +0000 (20:22 +0000)]
Branch b_release_1_4_6
Don't dereference data=NULL in case of old liblustre client connections.
b=8007
nathan [Mon, 6 Feb 2006 19:40:28 +0000 (19:40 +0000)]
Branch b_release_1_4_6
b=10139
r=adilger
(original patch by fergal)
don't try to ping '*' nids
zhanghc [Sun, 5 Feb 2006 10:27:54 +0000 (10:27 +0000)]
Land b1_4_lfs_df on b_release_1_4_6
BUG: 9461
DESCRIPTION: Implement 'lfs df' to report actual free space on per-OST basis
Modified Files:
include/lustre/liblustreapi.h include/lustre/lustre_user.h
liblustre/tests/sanity.c llite/dir.c llite/file.c
llite/llite_internal.h llite/llite_lib.c mds/mds_join.c
tests/sanity.sh utils/lfs.c utils/liblustreapi.c
alex [Sat, 4 Feb 2006 14:00:02 +0000 (14:00 +0000)]
b=9811
r=nikita
- free export when all OSCs are cleaned
scjody [Sat, 4 Feb 2006 01:09:17 +0000 (01:09 +0000)]
Branch b_release_1_4_6
Revert suse kernel update for now, since it doesn't work on Boston (upstream
issue, not ours.)
adilger [Fri, 3 Feb 2006 23:10:51 +0000 (23:10 +0000)]
Branch b_release_1_4_6
2.6.12 series is actually for 2.6.12.6
adilger [Fri, 3 Feb 2006 08:16:29 +0000 (08:16 +0000)]
Branch b_release_1_4_6
Remove ancient, obsolete portals block device #define.
adilger [Thu, 2 Feb 2006 22:34:43 +0000 (22:34 +0000)]
Branch b_release_1_4_6
ALlow default paths for patch-verification scripts.
adilger [Wed, 1 Feb 2006 23:23:43 +0000 (23:23 +0000)]
Branch b_release_1_4_6
Remove duplicate ChangeLog entries.
green [Wed, 1 Feb 2006 20:21:29 +0000 (20:21 +0000)]
Branch: b_release_1_4_6
Merge from b1_4
adilger [Wed, 1 Feb 2006 18:59:45 +0000 (18:59 +0000)]
Branch b_release_1_4_6
Add 2.6.12 kernel series, and clarify usage of ldiskfs series files.
nathan [Wed, 1 Feb 2006 18:03:26 +0000 (18:03 +0000)]
Branch b_release_1_4_6
b=10130
r=adilger
Set a fake non-0 NAL in 1.4.6, so if someone runs a pre-LNET version without
--write-conf, the ASSERT in class_add_uuid doesn't crash us.
adilger [Wed, 1 Feb 2006 00:39:56 +0000 (00:39 +0000)]
Branch b_release_1_4_6
LLNL has updated the BG/L clients to 2.6.9 and have stopped using 2.4.19.
eeb [Tue, 31 Jan 2006 15:16:22 +0000 (15:16 +0000)]
* Fixed 2.6.x sk_??? isms (required by zerocopy tcp)
* Future protocol version compatibility support....
* consolidated PROTO MAGIC decls in lib-types.h
* changed qswlnd's protocol to encapsulate the LNET protocol within its
own magic+version tagged packets and added a response when incoming
packets have new magic/version to tell peer which version I am.
'portals_compatibility' works as described in the documentation to
inter-operate with portals.
* Changed block_debug_check() to print EVERY error (it's only called by
obdecho so it can't be a DOS vector without root).
* Fixed a bug in osc_request: when a single oversize BRW was chopped up, the
subsequent RPCs used the 'oa' that had been overwritten by the previous
ones. This turned off checking in obdecho.
The fix I applied was to restore the original 'oa' from a saved copy on
all subsequent RPCs.
adilger [Tue, 31 Jan 2006 05:50:40 +0000 (05:50 +0000)]
Branch b_release_1_4_6
Quiet overly-verbose error message.
adilger [Tue, 31 Jan 2006 05:16:44 +0000 (05:16 +0000)]
Branch b_release_1_4_6
Disable test 15c, as it is failing repeatedly, and was only added during
CROW, so it may depend on crow-specific changes (though it passes in UML).
b=10124
adilger [Mon, 30 Jan 2006 22:48:41 +0000 (22:48 +0000)]
Branch b_release_1_4_6
Add journal statistics patch to CVS.
Not currently referenced by any kernel series, but useful for debugging
performance problems.
b=9387
r=alex (original patch)
adilger [Sun, 29 Jan 2006 11:59:17 +0000 (11:59 +0000)]
Branch b_release_1_4_6
Fix unused variable warning.
adilger [Sat, 28 Jan 2006 01:14:38 +0000 (01:14 +0000)]
Branch b_release_1_4_6
Description: Random seed of liblustre clients was not sufficiently random
Details : Improve initial seed of liblustre random number generator.
b=9794, tested by Cray
adilger [Sat, 28 Jan 2006 00:48:38 +0000 (00:48 +0000)]
Branch b_release_1_4_6
Remove hard-coded test directory from 63b.
SFS 3760
adilger [Sat, 28 Jan 2006 00:29:11 +0000 (00:29 +0000)]
Branch b_release_1_4_6
Include lock flags in the LDLM namespace dump.
b=9573 (SFS 4389)
adilger [Fri, 27 Jan 2006 09:29:07 +0000 (09:29 +0000)]
Branch b_release_1_4_6
Description: MDS or OST may oops/LBUG if a client is connecting multiple times
Details : The client ptlrpc code may be trying to reconnect to a down
server before a previous connection attempt has timed out.
Increase the reconnect interval to be longer than the connection
timeout interval to avoid sending duplicate connections to
servers.
This first part of the fix simply increases the reconnect interval to be
longer than the connection request timeout, to avoid having multiple
connect requests in flight from the same client at the same time.
b=9635
r=nathan
nathan [Fri, 27 Jan 2006 02:10:54 +0000 (02:10 +0000)]
Branch b_release_1_4_6
b=9860
r=adilger
Include ost index from mds connect, so we can set it in the last_rcvd file.
This will be useful when upgrading to 1.6
eeb [Thu, 26 Jan 2006 14:20:17 +0000 (14:20 +0000)]
* Adding a temporary branch tag to buildcvs for LNET protocol
compatibility work
adilger [Wed, 25 Jan 2006 18:32:54 +0000 (18:32 +0000)]
Branch b_release_1_4_6
Add missing (old) ChangeLog entry.
b=6422
eeb [Wed, 25 Jan 2006 17:08:50 +0000 (17:08 +0000)]
* Fixes needed to build for XT3/catamount
adilger [Wed, 25 Jan 2006 12:24:29 +0000 (12:24 +0000)]
Branch b_release_1_4_6
Bump build version.
b=10074
adilger [Wed, 25 Jan 2006 12:03:41 +0000 (12:03 +0000)]
Branch b_release_1_4_6
Fix compile warning.
adilger [Wed, 25 Jan 2006 12:01:33 +0000 (12:01 +0000)]
Branch b_release_1_4_6
Check return value from fstat(). (COV:629)
b=10098
adilger [Wed, 25 Jan 2006 11:59:26 +0000 (11:59 +0000)]
Branch b_release_1_4_6
Remove bogus check for NULL dchild in mds_open_by_fid(). We will always get
either a valid dentry or ERR_PTR() back from ll_lookup_one_len(). This
extra check provokes a false deref-before-NULL error in Coverity (COV:717,718)
b=10098
r=nikita
adilger [Wed, 25 Jan 2006 11:16:26 +0000 (11:16 +0000)]
Branch b_release_1_4_6
Fix potential NULL deref in ldlm_resource_foreach() (COV:707)
b=10098
adilger [Wed, 25 Jan 2006 11:14:11 +0000 (11:14 +0000)]
Branch b_release_1_4_6
Fix potential NULL obd deref in client_disconnect_export() (COV:708)
b=10098
adilger [Wed, 25 Jan 2006 05:51:04 +0000 (05:51 +0000)]
Merge b_release_1_4_6 from b_hd_newconfig (20060124_1325)
Revert "core" part of winnt and darwin compatibility code, as this
apparently does not compile for catamount. The arch-specific code
is still there as it does not affect anything outside those arches.
b=10074
adilger [Wed, 25 Jan 2006 00:33:39 +0000 (00:33 +0000)]
Merge b_release_1_4_6 from b_hd_newconfig (20060124_1325)
- cross-platform compiling for winnt (r=mattwu, original patch)
- improve use of compat macros for darwin (r=nikita, original patch)
Description: fix use of portals/lnet pid
Details : incorrect use of portals/lnet pid caused them to get out of
sync and would result in silent dropping of RPC messages
b=10074
r=eeb (original patch)
Severity : major
Frequency : Infiniband IB LND only
Description: iiblnd wasn't mapping all memory
Details : iiblnd wasn't mapping all memory, resulting in comms errors
on some architectures/memory configs
b=9776
r=eeb (original patch)
adilger [Tue, 24 Jan 2006 20:23:38 +0000 (20:23 +0000)]
Merge b_release_1_4_6 from HEAD (20060124_1302)
Minor fixes in "build" directory merged from parent.
adilger [Sat, 21 Jan 2006 00:54:32 +0000 (00:54 +0000)]
Branch b_release_1_4_6
Add explanitory comment.
adilger [Sat, 21 Jan 2006 00:53:46 +0000 (00:53 +0000)]
Branch b_release_1_4_6
Fix memory leak found by Coverity - we were not freeing any of the
other allocated structs if there was a later allocation error.
adilger [Fri, 20 Jan 2006 23:01:23 +0000 (23:01 +0000)]
Branch b_release_1_4_6
Fix errors spotted by Coverity - check for NULL after dereference
niu [Fri, 20 Jan 2006 08:23:25 +0000 (08:23 +0000)]
check return value of quota off, though it can only fail for network problem.
adilger [Wed, 18 Jan 2006 17:57:48 +0000 (17:57 +0000)]
Branch b_release_1_4_6
Finish lconf fixup.
b=10079
brian [Tue, 17 Jan 2006 22:38:50 +0000 (22:38 +0000)]
Do the same for the non-smp configs too.
brian [Tue, 17 Jan 2006 22:21:39 +0000 (22:21 +0000)]
Make kernel configs look like make oldconfig will make.
brian [Tue, 17 Jan 2006 22:15:15 +0000 (22:15 +0000)]
Use sendmail and set the sender address to ensure that mail is not sent from a
domain name that is neither an A or MX record (and thusly gets refused by
mail servers).
adilger [Tue, 17 Jan 2006 19:08:22 +0000 (19:08 +0000)]
Branch b_release_1_4_6
Use the "fid" EA name when storing FID data on the OST. This avoids
confusion on the OST when using the "lov" EA name and would certainly
be a problem with lfsck.
b=10072
r=yury
adilger [Tue, 17 Jan 2006 10:41:35 +0000 (10:41 +0000)]
Branch b_release_1_4_6
Quiet compile warning.
adilger [Sat, 14 Jan 2006 01:08:14 +0000 (01:08 +0000)]
Branch b_release_1_4_6
Merge changes from b1_4 that were missed in recent b_release_1_4_6 (b2) update
SWIOTLB = run x86_64 kernel on nocona
EXT3_MOUNT = avoid flag conflict with newer 2.6.12 kernel, make patches closer
adilger [Fri, 13 Jan 2006 21:34:18 +0000 (21:34 +0000)]
Branch b_release_1_4_6
Ignore generated file.
phil [Fri, 13 Jan 2006 17:24:44 +0000 (17:24 +0000)]
5047 looks like a major, if not critical, data corruption bug. Updating the
changelog accordingly.
adilger [Thu, 12 Jan 2006 07:50:21 +0000 (07:50 +0000)]
Branch b_release_1_4_6
Dbench failing to run properly because the clients.txt file is truncated.
rundbench verifies that the file has a non-zero size instead of just existing.
r=brian
houfeng [Thu, 12 Jan 2006 04:53:32 +0000 (04:53 +0000)]
remove needless lwizard; that break rpm building
cliffw [Thu, 12 Jan 2006 00:19:25 +0000 (00:19 +0000)]
Remove lwizard and associated documents
adilger [Wed, 11 Jan 2006 09:59:57 +0000 (09:59 +0000)]
Merge b_release_1_4_6 from b1_4 (20060110_0557)
Description: OST becomes very slow and/or deadlocked during object unlink
Details : filter_destroy() was holding onto the parent directory lock
while truncating+unlinking objects. For very large objects this
may block other threads for a long time and slow overall OST
responsiveness. It may also be possible to get a lock ordering
deadlock in this case, or run out of journal credits because of
the combined truncate+unlink. Solution is to do object truncate
first in one transaction without parent lock, and then do the
final unlink in a new transaction with the parent lock. This
reduces the lock hold time dramatically.
b=4180, b=6984, b=7171, b=9963, b=9331
r=alex
Description: MDS or OST cleanup may trip kernel BUG when dropping kernel lock
Details : mds_cleanup() and filter_cleanup() need to drop the kernel lock
before unmounting their filesystem in order to avoid deadlock.
The kernel_locked() function in 2.4 kernels only checks whether
the kernel lock is held, not whether it is this process that is
holding it as 2.6 kernels do.
b=9967
r=nathan
Description: unlinked inodes were kept in memory on the client
Details : If a client is repeatedly creating and unlinking files it
can accumulate a lot of stale inodes in the inode slab cache.
If there is no other client load running this can cause the
client node to run out of memory. Instead flush old inodes
from client cache that have the same inode number as a new inode.
b=1443
r=niu (original patch)
Description: Referencing freed requests leading to crash, memleask with NFS.
Details : We used to require that call to ll_revalidate_it was always
followed by ll_lookup_it. Also with revalidate_special() it is
possible to call ll_revalidate_it() twice for the same dentry
even if first occurence returned success. This fix changes semantic
between DISP_ENQ_COMPLETE disposition flag to mean there is extra
reference on a request referred from the intent.
ll_intent_release() then releases such a request.
b=9942, b=9903
r=green (original patch)
cliffw [Tue, 10 Jan 2006 23:53:18 +0000 (23:53 +0000)]
Added the <url> for the Knowledge Base article
alex [Tue, 10 Jan 2006 20:21:52 +0000 (20:21 +0000)]
b=9969
- skip invalidated dentries via ll_dcompare()
- one more patch to the suse series to embed d_name structure into dentry unconditionally
adilger [Tue, 10 Jan 2006 17:58:52 +0000 (17:58 +0000)]
Merge b_release_1_4_6 from b1_4 (20060110_0557)
- fix lustre/tests build problem before doing full merge
adilger [Tue, 10 Jan 2006 12:48:27 +0000 (12:48 +0000)]
Update b_release_1_4_6 from b_hd_newconfig
- add timestamps for liblustre debug messages
- use wait_event_interruptible_exclusive() to LNDs so we don't get
too-many-thread wakeups and lock contention when waiting for events
- fix tx credits handling
- fix incorrect error for multiple networks and portals_compatible option
- add 2.4 compatibility macro for wait_event_interruptible_exclusive()
brian [Mon, 9 Jan 2006 19:14:25 +0000 (19:14 +0000)]
Update the ChangeLog and version info for 1.4.6 beta2.
wangdi [Sat, 7 Jan 2006 05:18:22 +0000 (05:18 +0000)]
Branch: b_release_1_4_6
b=9926
1) shrink the reply buffer for unlink and orphan close
2) retrieve max mdsize and cookiesize in getattr, because in join, max_mdsize and
cookiesize might change, and all the clients should know this info.
r=adilger
alex [Fri, 6 Jan 2006 23:57:11 +0000 (23:57 +0000)]
b=5681
r=nikita
- don't unhash in-use dentries because it breaks sys_getcwd()
green [Fri, 6 Jan 2006 06:21:27 +0000 (06:21 +0000)]
b=9942,9903
r=adilger
Treat DISP_ENQ_COMPLETE as a flag that request referenced from this intent
has extra reference that needs to be released, if still set at intent freeing
time.
green [Fri, 6 Jan 2006 06:05:29 +0000 (06:05 +0000)]
b=9942
r=adilger
In link_raw() always call iget before doing d_instantiate
green [Fri, 6 Jan 2006 05:49:12 +0000 (05:49 +0000)]
b=10027
r=adilger
Do not segfault in 'lfs check' if no Lustre fs mounted.
adilger [Thu, 5 Jan 2006 22:37:06 +0000 (22:37 +0000)]
Branch b_release_1_4_6
Don't call _sysio_shutdown() on catamount, Lustre isn't the only user of
libsysio there.
b=10024, b=7193
green [Thu, 5 Jan 2006 21:45:29 +0000 (21:45 +0000)]
b=9780
Check that we actually encountered a symlink before returning error.
green [Thu, 5 Jan 2006 21:12:19 +0000 (21:12 +0000)]
b=9781
r=adilger
When we cancel a lock in completion or blocking callback 0 we must return
-ERESTART so that resource would be reprocessed again.
green [Thu, 5 Jan 2006 21:08:55 +0000 (21:08 +0000)]
b=7313
r=adilger
Latest version of reply-less lock callbacks patch with proper replay
support - for testing by Cray.
alex [Thu, 5 Jan 2006 19:46:14 +0000 (19:46 +0000)]
b=9678
r=adilger
- free pool after client_obd_cleanup() because the latter may abort in-flight requests from the pool
adilger [Thu, 5 Jan 2006 09:58:33 +0000 (09:58 +0000)]
Branch b_release_1_4_6
Don't update thread->t_id as we exit from ptlrpc_main() after waking
t_ctl_waitq, as it is possible the thread struct is being freed by
ptlrpc_stop_thread() that was waiting on t_ctl_waitq.
r=alex
green [Wed, 4 Jan 2006 22:21:51 +0000 (22:21 +0000)]
Branch: b_release_1_4_6
b=9780
Better version proposed by Lee.
nic [Wed, 4 Jan 2006 21:06:17 +0000 (21:06 +0000)]
update the fc3 kernel to be more like the newer rhel4 kernel to allow the rhel4
patches to apply
adilger [Tue, 3 Jan 2006 22:50:45 +0000 (22:50 +0000)]
Branch b_release_1_4_6
Ensure that statfs does not come out of local cached value.
b=9691
green [Tue, 3 Jan 2006 22:48:44 +0000 (22:48 +0000)]
b=9780
properly handle O_NOFOLLOW
green [Tue, 3 Jan 2006 22:46:30 +0000 (22:46 +0000)]
More sync-up with sysio head
alex [Tue, 3 Jan 2006 19:10:41 +0000 (19:10 +0000)]
b=9919
r=nikita (original fix)
hijacks pages from struct brw_page (allocated by echo_client_kbrw()) by filter_brw(). This
- fixes problem introduced by 5137 (filter_preprw_*() assumes that pages are preallocated),
- avoids duplicate allocation,
- avoids data copying.
Also fixes small defect in filter_iobuf_get().
adilger [Mon, 2 Jan 2006 08:04:54 +0000 (08:04 +0000)]
Branch b_release_1_4_6
Define $tdir and $tfile for sanityN.sh, bringing us one tiny step closer to
being able to use test-framework.sh for sanityN.sh.
adilger [Mon, 2 Jan 2006 07:48:32 +0000 (07:48 +0000)]
Branch b_release_1_4_6
Description: two objects could be created on the same OST for a single file
Details : If an OST is down, in some cases it was possible to create two
objects on a single OST for a single file. No problems other
than potential performance impact and spurious error messages.
b=9923
r=green
adilger [Mon, 2 Jan 2006 07:38:42 +0000 (07:38 +0000)]
Branch b_release_1_4_6
We should always be starting service threads when we call
ptlrpc_start_threads(). This caused a very tricky to notice "communications"
problem when an OST didn't start any service threads. This can only be a
code problem, and not from module parameter input.
r=hwartens
phil [Sat, 31 Dec 2005 16:33:00 +0000 (16:33 +0000)]
a local-tree-cleanup commit full of empty calories
green [Sat, 31 Dec 2005 13:44:12 +0000 (13:44 +0000)]
b=10018
Fix 64 bit compile problem
niu [Sat, 31 Dec 2005 05:03:09 +0000 (05:03 +0000)]
acquire block quota on owner of child also, in case of ea block counted in some kernel version.
b: 10017
yury [Fri, 30 Dec 2005 07:58:39 +0000 (07:58 +0000)]
b=9926
r=adilger
- do not confuse developer looking at logs by saying that namespace is freeing on last lock put.
- cleanups in mds_init_lov_desc()
- using stripes = min(tgt_count, LOV_MAX_STRIPE_COUNT) as a basis for calculating easize/cookiesize on MDS and client.
- removed redundant obd_set_info("mds_conn") in mds_notify()
green [Thu, 29 Dec 2005 22:00:01 +0000 (22:00 +0000)]
Change some CERROR()s that are not errors to CWARN()s in administrative
evictions path.
green [Thu, 29 Dec 2005 21:57:30 +0000 (21:57 +0000)]
Switch b_release_1_4_6 to use new libsysio.
green [Thu, 29 Dec 2005 21:50:49 +0000 (21:50 +0000)]
Branch: b_release_1_4_6
new libsysio from Sandia.