Whamcloud - gitweb
alex [Fri, 14 Oct 2005 10:20:29 +0000 (10:20 +0000)]
b=9516
r=alex
- limit number of in-flight async destroy rpcs MDS issues
to destroy OST objects
nic [Thu, 6 Oct 2005 20:51:49 +0000 (20:51 +0000)]
1.2.8 and 1.4.0 did not have build/ tagged for them, so pull HEAD
nic [Thu, 6 Oct 2005 18:13:00 +0000 (18:13 +0000)]
allow vars to be overriden from the command line
ericm [Wed, 5 Oct 2005 17:22:34 +0000 (17:22 +0000)]
b1_4_xattr don't need special requirement for lnet since it's re-based
on b1_4 now.
ericm [Wed, 5 Oct 2005 05:12:17 +0000 (05:12 +0000)]
b1_4_xattr need same lnet as b1_4_newconfig
pjkirner [Thu, 29 Sep 2005 20:04:40 +0000 (20:04 +0000)]
* make scheduler threads balance rx and tx more evenly
* move peer allocation outside spin lock
* fix bug with irqs not being disabled on spinlocks that need syncronzation with callbacks
adilger [Thu, 29 Sep 2005 06:03:30 +0000 (06:03 +0000)]
Branch b1_4
Add "make srpm" target for <drum roll>making .src.rpm files</drum roll>.
lsy [Mon, 26 Sep 2005 08:19:04 +0000 (08:19 +0000)]
add makefile for standalone quota.
brian [Sat, 24 Sep 2005 03:25:58 +0000 (03:25 +0000)]
Group lock code to pass test 58.
brian [Sat, 24 Sep 2005 03:06:38 +0000 (03:06 +0000)]
Backout changes made to day in efforts to make test 41 work. It passes with
code checked out prior to all of these changes.
alex [Sat, 24 Sep 2005 01:23:23 +0000 (01:23 +0000)]
- minor optimization to avoid extra expenses related to OST A/M/D in the
path of LOV EA revalidation
alex [Sat, 24 Sep 2005 00:33:55 +0000 (00:33 +0000)]
- re-fill size/blocks attributes in open reply if ea lov has been changed
- yank granted update locks in the same case in order to invalidate
cached size/blocks attributes on clients
alex [Sat, 24 Sep 2005 00:21:24 +0000 (00:21 +0000)]
- missed bit from test 41 debugging
alex [Sat, 24 Sep 2005 00:17:35 +0000 (00:17 +0000)]
- OST A/M/D should work OK in non-CMD setup
brian [Fri, 23 Sep 2005 22:45:21 +0000 (22:45 +0000)]
r=alex
Possible fix for file size mismatch between mds and osts.
wangdi [Fri, 23 Sep 2005 13:47:31 +0000 (13:47 +0000)]
HEAD
remove debug patch for test41
wangdi [Fri, 23 Sep 2005 12:57:23 +0000 (12:57 +0000)]
Branch: HEAD
try tmp patch for failed test41
alex [Thu, 22 Sep 2005 20:04:01 +0000 (20:04 +0000)]
- don't enable capabilities by default
alex [Thu, 22 Sep 2005 19:09:45 +0000 (19:09 +0000)]
b=7148
r=alex (original patch by don milos)
- revalide i_size/i_blocks along with LOV EA in case of OSS removal
alex [Wed, 21 Sep 2005 23:09:14 +0000 (23:09 +0000)]
- drop ll_capa_stat flag in order to preven flood
alex [Wed, 21 Sep 2005 23:00:34 +0000 (23:00 +0000)]
- make default key timelife 3 days long
alex [Wed, 21 Sep 2005 21:14:26 +0000 (21:14 +0000)]
- generate capability in open resend case
- regression test for the case
alex [Wed, 21 Sep 2005 20:42:37 +0000 (20:42 +0000)]
- additional debug to catch the case when client gets openhandle w/o capa
ericm [Wed, 21 Sep 2005 18:52:32 +0000 (18:52 +0000)]
add debug msg for gss seq window.
alex [Wed, 21 Sep 2005 18:17:14 +0000 (18:17 +0000)]
- drop needless capas upon close
lsy [Wed, 21 Sep 2005 17:46:04 +0000 (17:46 +0000)]
drop truncate capa after ll_truncate.
lsy [Wed, 21 Sep 2005 15:45:19 +0000 (15:45 +0000)]
more debug message.
nathan [Wed, 21 Sep 2005 14:29:51 +0000 (14:29 +0000)]
add b1_4_mountconf using lnet
lsy [Wed, 21 Sep 2005 12:40:34 +0000 (12:40 +0000)]
message cleanup.
alex [Wed, 21 Sep 2005 12:40:32 +0000 (12:40 +0000)]
- swab capa data in mdc_replay_open(), the sequence can be:
1) ll_file_open() has a request with capa
2) the inode is already opened, so we close just received openhandle
3) mds fails, recovers, ptlrpcd replays open request
4) ll_set_capa() finds capa in replayed open request non-swabbed - BUG
lsy [Wed, 21 Sep 2005 12:02:28 +0000 (12:02 +0000)]
put capa before calling ll_renew_capa, in case it might get updated.
alex [Wed, 21 Sep 2005 11:07:08 +0000 (11:07 +0000)]
- don't make noise if we found needed capa
- minor cleanups
alex [Wed, 21 Sep 2005 11:01:02 +0000 (11:01 +0000)]
- nobody needs these warnings any more
lsy [Wed, 21 Sep 2005 10:08:19 +0000 (10:08 +0000)]
don't make too much noise.
lsy [Wed, 21 Sep 2005 09:51:50 +0000 (09:51 +0000)]
export symbol.
lsy [Wed, 21 Sep 2005 09:39:53 +0000 (09:39 +0000)]
make noise when capa not found.
ericm [Wed, 21 Sep 2005 08:37:50 +0000 (08:37 +0000)]
missed i_gen
alex [Wed, 21 Sep 2005 08:36:11 +0000 (08:36 +0000)]
- pack fake reply if mds_getattr() couldn't lookup fid. otherwise
mds_handle() will oops.
alex [Wed, 21 Sep 2005 07:16:25 +0000 (07:16 +0000)]
- in order to prevent flood during renewing capa for deleted inodes:
- mds_getattr() returns regular errors in rq_status
- mdc_interpret_getattr() interprets it properly
lsy [Wed, 21 Sep 2005 06:58:38 +0000 (06:58 +0000)]
client capa list is sorted.
new capa for mds and oss should be added at the end. (so it should be sorted too)
ericm [Wed, 21 Sep 2005 05:06:59 +0000 (05:06 +0000)]
the missing part of last capa fix.
ericm [Wed, 21 Sep 2005 01:19:14 +0000 (01:19 +0000)]
remove stale capa structure from mds/filter obd.
ericm [Wed, 21 Sep 2005 01:01:01 +0000 (01:01 +0000)]
on OSS compare the whole capa structure because we might have 2 capa only
have difference in expiry time.
adilger [Wed, 21 Sep 2005 00:02:40 +0000 (00:02 +0000)]
Branch b1_4
Update the CVS commit script to add a reminder about HLD/DLD updates.
r=design meeting
alex [Tue, 20 Sep 2005 23:41:33 +0000 (23:41 +0000)]
- one more DEBUG_CAPA() to make sure we use proper capability to check
permissions
alex [Tue, 20 Sep 2005 23:01:21 +0000 (23:01 +0000)]
- typo fixed
alex [Tue, 20 Sep 2005 22:52:22 +0000 (22:52 +0000)]
- bump capa cache size on MDS/OSS to 3000 to decrease cache refill load
alex [Tue, 20 Sep 2005 22:51:19 +0000 (22:51 +0000)]
- allocate tfm per every sign procedure (tfm isn't thread-safe)
lsy [Tue, 20 Sep 2005 21:39:06 +0000 (21:39 +0000)]
cleanup debug message.
lsy [Tue, 20 Sep 2005 16:23:46 +0000 (16:23 +0000)]
typo.
lsy [Tue, 20 Sep 2005 15:52:39 +0000 (15:52 +0000)]
add debug for capa_key.
alex [Tue, 20 Sep 2005 13:25:11 +0000 (13:25 +0000)]
- warn if HMAC's don't match
lsy [Tue, 20 Sep 2005 12:37:20 +0000 (12:37 +0000)]
add debug msg, and some cleanup.
capa_renew should use tcapa instead of capa.
alex [Tue, 20 Sep 2005 10:57:25 +0000 (10:57 +0000)]
- needless \n grrrrr
alex [Tue, 20 Sep 2005 10:53:44 +0000 (10:53 +0000)]
- additional debug
alex [Tue, 20 Sep 2005 09:38:36 +0000 (09:38 +0000)]
- minor typo: CDEBUG() doesn't add \n at the end
lsy [Tue, 20 Sep 2005 08:25:05 +0000 (08:25 +0000)]
change debug level.
lsy [Tue, 20 Sep 2005 06:26:23 +0000 (06:26 +0000)]
llite doesn't put it.
lsy [Tue, 20 Sep 2005 06:03:08 +0000 (06:03 +0000)]
capa_renew should increase the refc.
lsy [Tue, 20 Sep 2005 03:26:04 +0000 (03:26 +0000)]
update struct lustre_capa_data too with the introduction of igen.
ericm [Tue, 20 Sep 2005 03:05:40 +0000 (03:05 +0000)]
add inode generation in capability.
alex [Mon, 19 Sep 2005 19:55:27 +0000 (19:55 +0000)]
- make noise if capa hasn't been passed even
alex [Mon, 19 Sep 2005 19:00:52 +0000 (19:00 +0000)]
- additional debug for capa denies
wangdi [Mon, 19 Sep 2005 07:31:08 +0000 (07:31 +0000)]
Branch: HEAD
typo fix about ldlm flags
ericm [Mon, 19 Sep 2005 07:11:07 +0000 (07:11 +0000)]
roundup gss credential timeout down to avoid expiration between accept &
authorize on server, or check & encrypt rpc on client.
alex [Sun, 18 Sep 2005 16:06:52 +0000 (16:06 +0000)]
- protect i_nlink in smfs_mkdir() as well
wangdi [Sun, 18 Sep 2005 04:15:57 +0000 (04:15 +0000)]
Branch: HEAD
uuid of client mdc and osc will be modified in mds ost add
alex [Sat, 17 Sep 2005 22:48:46 +0000 (22:48 +0000)]
- protect i_nlink from concurrent changing
alex [Sat, 17 Sep 2005 22:33:52 +0000 (22:33 +0000)]
b=7268
- increase timeout for bulk PUT/GET a little bit
ericm [Sat, 17 Sep 2005 07:49:51 +0000 (07:49 +0000)]
temporarily disable sanity 57b which lead to LBUG.
thanks tappro.
wangdi [Sat, 17 Sep 2005 06:40:20 +0000 (06:40 +0000)]
Branch: HEAD
tmp fix about ost add/del, reset osc uuid in mds, but not reset mdc uuid in mds
tappro [Sat, 17 Sep 2005 01:09:18 +0000 (01:09 +0000)]
prevent null_audit case
tappro [Fri, 16 Sep 2005 23:30:58 +0000 (23:30 +0000)]
b=7230
update mtime as well as size on MDS and in llite
alex [Fri, 16 Sep 2005 22:23:31 +0000 (22:23 +0000)]
- obd_disconnect(sbi->ll_dt_exp) clears namespace and lets inodes go, we
shouldn't release crypto subsystem before, because it's used in ll_clear_inode()
alex [Fri, 16 Sep 2005 22:18:40 +0000 (22:18 +0000)]
- revert debugging bits back - it breaks everything
niu [Fri, 16 Sep 2005 18:41:24 +0000 (18:41 +0000)]
simple fix for inconsistent mtime on different client, commit to test if
it's helpful.
- b 7230
tappro [Fri, 16 Sep 2005 16:12:08 +0000 (16:12 +0000)]
add command null_audit in lctl
tappro [Fri, 16 Sep 2005 16:11:27 +0000 (16:11 +0000)]
do LASSERT if *name is not NULL
wangdi [Fri, 16 Sep 2005 15:52:26 +0000 (15:52 +0000)]
Branch: HEAD
missing update lvb_mtime
tappro [Fri, 16 Sep 2005 15:48:53 +0000 (15:48 +0000)]
add null_audit command to lctl which will make audit silent. For testing purposes.
wangdi [Fri, 16 Sep 2005 15:27:35 +0000 (15:27 +0000)]
Branch: HEAD
fix LDLM flags defination confilct
tappro [Fri, 16 Sep 2005 12:45:20 +0000 (12:45 +0000)]
r=alex
do d_delete() for backfs dentry as vfs_unlink/rmdir do
alex [Fri, 16 Sep 2005 12:35:02 +0000 (12:35 +0000)]
- dump i_mode as well
alex [Fri, 16 Sep 2005 12:34:33 +0000 (12:34 +0000)]
- check for non-zero inum
yury [Fri, 16 Sep 2005 10:46:07 +0000 (10:46 +0000)]
- removed not needed error message
lsy [Fri, 16 Sep 2005 10:27:13 +0000 (10:27 +0000)]
revert part of last change.
lsy [Fri, 16 Sep 2005 10:15:53 +0000 (10:15 +0000)]
igrab() before ll_renew_capa() in case inode is cleared.
alex [Fri, 16 Sep 2005 09:49:24 +0000 (09:49 +0000)]
- we don't need this LBUG during acceptance
alex [Fri, 16 Sep 2005 09:47:58 +0000 (09:47 +0000)]
b=7340
- setup cr_replayid immediately because it might take long to get to
mdc_set_open_replay_data() and replay can happen in meantime
ericm [Fri, 16 Sep 2005 07:11:57 +0000 (07:11 +0000)]
hack: wait rsi really be cleaned up before module unload.
yury [Fri, 16 Sep 2005 06:21:49 +0000 (06:21 +0000)]
- added some more noise in unlink time when errors to see what happening
wangdi [Fri, 16 Sep 2005 06:18:34 +0000 (06:18 +0000)]
Branch: HEAD
revert back ldlm flags and make them compatiable with b1_4, according to andreas comments.
alex [Fri, 16 Sep 2005 01:05:26 +0000 (01:05 +0000)]
- check for bad inode in smfs_do_lookup()
alex [Fri, 16 Sep 2005 01:04:41 +0000 (01:04 +0000)]
b=6866
r=adilger(original patch)
Bad interaction between the ea-in-inode patch and the "no-read"
code in the 2.6 kernel caused the inode and/or EA data not to
be read from disk, causing single-file corruption.
ericm [Fri, 16 Sep 2005 00:14:21 +0000 (00:14 +0000)]
add an assert, from alex.
wangdi [Thu, 15 Sep 2005 17:53:50 +0000 (17:53 +0000)]
Branch: HEAD
add flock fix patch of Donm
fix about build warning in lustre_dlm.h
wangdi [Thu, 15 Sep 2005 17:19:21 +0000 (17:19 +0000)]
Branch: HEAD
land posix flock patch of Don.
alex [Thu, 15 Sep 2005 16:34:55 +0000 (16:34 +0000)]
b=7272
- very messy debug bits to check for double direntries
tappro [Thu, 15 Sep 2005 13:42:19 +0000 (13:42 +0000)]
b=8654
revert back last changes after testing.
tappro [Thu, 15 Sep 2005 10:29:27 +0000 (10:29 +0000)]
b=8654
turn on stat auditing on client for testing purpose