Whamcloud - gitweb
fs/lustre-release.git
18 years ago- complain if kernel doesn't support 1MB request
alex [Mon, 5 Sep 2005 14:47:38 +0000 (14:47 +0000)]
- complain if kernel doesn't support 1MB request

18 years ago- set SCSI_MAX_PHYS_SEGMENTS to 256 for 1MB requests
alex [Mon, 5 Sep 2005 14:09:29 +0000 (14:09 +0000)]
- set SCSI_MAX_PHYS_SEGMENTS to 256 for 1MB requests

18 years ago- disable debug
alex [Mon, 5 Sep 2005 13:39:19 +0000 (13:39 +0000)]
- disable debug

18 years ago- debug info to investigate a source of small writes
alex [Mon, 5 Sep 2005 12:56:32 +0000 (12:56 +0000)]
- debug info to investigate a source of small writes

18 years agorename confusing mdc_uuid to tgt_uuid.
niu [Mon, 5 Sep 2005 09:46:37 +0000 (09:46 +0000)]
rename confusing mdc_uuid to tgt_uuid.

18 years agoBranch: HEAD
wangdi [Mon, 5 Sep 2005 08:58:54 +0000 (08:58 +0000)]
Branch: HEAD
remove obsolete lgs.sh

18 years agoadd some debug message for capbility key update.
lsy [Mon, 5 Sep 2005 06:05:07 +0000 (06:05 +0000)]
add some debug message for capbility key update.
a small fix for key expiry.

18 years ago- typo fixed
alex [Sun, 4 Sep 2005 23:19:20 +0000 (23:19 +0000)]
- typo fixed

18 years ago- mds_postrecov_common() can return number of cleaned orphans as a success
alex [Sun, 4 Sep 2005 22:23:38 +0000 (22:23 +0000)]
- mds_postrecov_common() can return number of cleaned orphans as a success

18 years agob=7273
alex [Sun, 4 Sep 2005 22:21:50 +0000 (22:21 +0000)]
b=7273

 - additional debug info

18 years ago- make gcc happy
alex [Sun, 4 Sep 2005 14:38:34 +0000 (14:38 +0000)]
- make gcc happy

18 years agoBranch: HEAD
wangdi [Sun, 4 Sep 2005 03:35:45 +0000 (03:35 +0000)]
Branch: HEAD
add ost add/del test script

18 years agoBranch: HEAD
wangdi [Sun, 4 Sep 2005 03:21:20 +0000 (03:21 +0000)]
Branch: HEAD
remove lov_tgt_pending, which is useless for ost add/del

18 years ago- can hit during forced umount
alex [Sat, 3 Sep 2005 16:16:31 +0000 (16:16 +0000)]
- can hit during forced umount

18 years ago- i think we don't need that now
alex [Sat, 3 Sep 2005 16:12:11 +0000 (16:12 +0000)]
- i think we don't need that now

18 years ago- disable extents/mballoc for benchmarking purposes
alex [Sat, 3 Sep 2005 12:13:46 +0000 (12:13 +0000)]
- disable extents/mballoc for benchmarking purposes

18 years ago- serialize allocations for 2.1, 2.2 and 2.3
alex [Sat, 3 Sep 2005 09:37:02 +0000 (09:37 +0000)]
- serialize allocations for 2.1, 2.2 and 2.3

18 years ago- revert hacks back
alex [Fri, 2 Sep 2005 21:15:46 +0000 (21:15 +0000)]
- revert hacks back

18 years agob=7272
alex [Fri, 2 Sep 2005 19:31:56 +0000 (19:31 +0000)]
b=7272

 - more debug to catch the case

18 years ago- one more field in struct mds_rec_create to replay i/o epoch file was open in
alex [Fri, 2 Sep 2005 19:29:21 +0000 (19:29 +0000)]
- one more field in struct mds_rec_create to replay i/o epoch file was open in

18 years agofixes:
lsy [Fri, 2 Sep 2005 13:31:46 +0000 (13:31 +0000)]
fixes:
* also round for short expiry, in case timer might miss.
* add mds_mfd_put() because mds_handle2mfd() will add ref.
* filter_verify_capa() logic error.

18 years ago- few fixes in comments (typos)
yury [Fri, 2 Sep 2005 12:46:00 +0000 (12:46 +0000)]
- few fixes in comments (typos)
- removed wrong stuff in class_add_conn(). Apparently it is result of wrong merge when newcfg stuff was landed. Issues related to it were observed in #48.

18 years ago- eeb's sd iostat bits have been added
alex [Thu, 1 Sep 2005 23:15:39 +0000 (23:15 +0000)]
- eeb's sd iostat bits have been added

18 years ago- disable fo_alloc_sem
alex [Thu, 1 Sep 2005 22:53:41 +0000 (22:53 +0000)]
- disable fo_alloc_sem
- don't wait for transaction commit
  NOTE: to evaluate impact

18 years ago- add fo_alloc_lock serialization support for extents/mballoc case
alex [Thu, 1 Sep 2005 21:53:36 +0000 (21:53 +0000)]
- add fo_alloc_lock serialization support for extents/mballoc case

18 years ago- bump OST_MAX_THREADS to 48 for benchmarking
alex [Thu, 1 Sep 2005 19:59:45 +0000 (19:59 +0000)]
- bump OST_MAX_THREADS to 48 for benchmarking

18 years ago- serialize all allocations on OSS by fo_alloc_lock. for benchmarking
alex [Thu, 1 Sep 2005 18:14:13 +0000 (18:14 +0000)]
- serialize all allocations on OSS by fo_alloc_lock.  for benchmarking

18 years agoBranch: HEAD
wangdi [Thu, 1 Sep 2005 07:58:17 +0000 (07:58 +0000)]
Branch: HEAD
lproc for update config should return count, not 0

18 years agoBranch: HEAD
wangdi [Thu, 1 Sep 2005 06:01:38 +0000 (06:01 +0000)]
Branch: HEAD
lli_posix_acl and lli_key_info is not reliable on checking whether the acl and key_info
is valid on the client

18 years agoAdd support for b_lnet branch of the ./build tree to cleanly
pjkirner [Thu, 1 Sep 2005 04:02:09 +0000 (04:02 +0000)]
Add support for b_lnet branch of the ./build tree to cleanly
seperate the changes related to LNET reorg that is going on on the
b1_4_newconfig branch(s).

(Consulted with Jacob and Andreas on this change)

18 years agoBranch HEAD
adilger [Wed, 31 Aug 2005 20:00:53 +0000 (20:00 +0000)]
Branch HEAD
Check out the tagged version of the lustre-build scripts, so we are sure
to use the same ones as when the packages were originally built.

18 years agochange error msg to warning.
lsy [Wed, 31 Aug 2005 16:05:47 +0000 (16:05 +0000)]
change error msg to warning.
some cleanup.

18 years ago- as we don't set PG_writeback on pages being flushed, we need another
alex [Wed, 31 Aug 2005 15:35:05 +0000 (15:35 +0000)]
- as we don't set PG_writeback on pages being flushed, we need another
  way to support fsync(2). the simplest one is to wait for empty list
  lli_pending_write_llaps

18 years agoFix build breakage introduced by lnet reorg.
pjkirner [Wed, 31 Aug 2005 02:20:33 +0000 (02:20 +0000)]
Fix build breakage introduced by lnet reorg.
Completely removed changes until something more compatable can be designed.

Backed out changes to build tree:
autogen.sh                      reverted back to rev 1.4
autoMakefile.am.toplevel        reverted back to rev 1.8
buildcvs                        reverted back to rev 1.44
Makefile.in.toplevel            reverted back to rev 1.1
autoconf/lustre-build-darwin.m4 reverted back to rev 1.8
autoconf/lustre-build-linux.m4  reverted back to rev 1.15
autoconf/lustre-build.m4        reverted back to rev 1.28

18 years agob=7390, 7409
tappro [Wed, 31 Aug 2005 02:11:23 +0000 (02:11 +0000)]
b=7390, 7409
changes in auditing of failed operations when audit is set on directory

18 years agoPhase 1 of LNET Reorg
pjkirner [Tue, 30 Aug 2005 20:09:53 +0000 (20:09 +0000)]
Phase 1 of LNET Reorg
r=eeb
r=pjk

18 years agoBranch: HEAD
wangdi [Tue, 30 Aug 2005 14:14:40 +0000 (14:14 +0000)]
Branch: HEAD
forget lprof->lp_gkc in profile delete

18 years ago- give concurrent allocation one more chance
alex [Tue, 30 Aug 2005 11:51:19 +0000 (11:51 +0000)]
- give concurrent allocation one more chance

18 years agohack to always use root credential between client and oss.
ericm [Mon, 29 Aug 2005 20:59:19 +0000 (20:59 +0000)]
hack to always use root credential between client and oss.

18 years agob=7268
alex [Mon, 29 Aug 2005 19:42:30 +0000 (19:42 +0000)]
b=7268

 - dump time ost_brw_write() spent waiting for bulk completion

18 years ago- replay i/o epochs along with other attributes to improve isize consistency
alex [Mon, 29 Aug 2005 12:28:58 +0000 (12:28 +0000)]
- replay i/o epochs along with other attributes to improve isize consistency
  without involving OST

18 years agominor fix:
lsy [Mon, 29 Aug 2005 10:33:29 +0000 (10:33 +0000)]
minor fix:
* for partial write only search capability once.

18 years ago- ll_writepages() waits for I/O completion if filesystem is being umounted
alex [Mon, 29 Aug 2005 10:30:37 +0000 (10:30 +0000)]
- ll_writepages() waits for I/O completion if filesystem is being umounted

18 years ago- dump log if OSC recovery takes too long
alex [Mon, 29 Aug 2005 10:28:16 +0000 (10:28 +0000)]
- dump log if OSC recovery takes too long

18 years agor=tappro
alex [Mon, 29 Aug 2005 09:49:28 +0000 (09:49 +0000)]
r=tappro

 - couple unused vars have been removed

18 years agoBranch: HEAD
wangdi [Mon, 29 Aug 2005 07:22:19 +0000 (07:22 +0000)]
Branch: HEAD
revert back the fix for b7379, it is not right.

18 years ago- fill io epoch in MDS_CLOSE only if file is writtable
alex [Sun, 28 Aug 2005 17:57:15 +0000 (17:57 +0000)]
- fill io epoch in MDS_CLOSE only if file is writtable

18 years ago- set and check io epoch only for writtable files
alex [Sun, 28 Aug 2005 13:16:48 +0000 (13:16 +0000)]
- set and check io epoch only for writtable files

18 years agob=7412
alex [Sun, 28 Aug 2005 13:14:33 +0000 (13:14 +0000)]
b=7412
r=ericm

 - keep FLSIZE for non-regular files

18 years agoBranch: HEAD
wangdi [Sun, 28 Aug 2005 02:56:22 +0000 (02:56 +0000)]
Branch: HEAD
a tmp fix for b7379.

18 years agoa minor reorganization of capability code.
lsy [Sat, 27 Aug 2005 10:15:23 +0000 (10:15 +0000)]
a minor reorganization of capability code.
some cleanup.

18 years agob=7390
tappro [Fri, 26 Aug 2005 12:59:06 +0000 (12:59 +0000)]
b=7390
previous fix was wrong

18 years agob=7390
tappro [Fri, 26 Aug 2005 11:48:04 +0000 (11:48 +0000)]
b=7390

- enable audit on every child in dir while setting audit on that dir.

18 years ago- fix case when inode can be NULL.
tappro [Fri, 26 Aug 2005 11:44:26 +0000 (11:44 +0000)]
- fix case when inode can be NULL.

18 years agob=7390
tappro [Fri, 26 Aug 2005 11:42:29 +0000 (11:42 +0000)]
b=7390
- add special handler for auditing access failures

18 years ago- all stats gatherd in OSC debugging time (cache loading, etc.) moved to proc. Added...
yury [Fri, 26 Aug 2005 08:53:03 +0000 (08:53 +0000)]
- all stats gatherd in OSC debugging time (cache loading, etc.) moved to proc. Added new proc file ./cache_stats showing them.
- some RPCs related new stats added to ./rpc_stats

18 years agobranch: HEAD
wangdi [Fri, 26 Aug 2005 03:45:05 +0000 (03:45 +0000)]
branch: HEAD
some updates in sanity-crypto.sh

18 years agoshouldn't touch this.
lsy [Thu, 25 Aug 2005 14:55:33 +0000 (14:55 +0000)]
shouldn't touch this.

18 years agofixes:
lsy [Thu, 25 Aug 2005 14:52:52 +0000 (14:52 +0000)]
fixes:
* previous implementation doesn't support multi-open/truncate by different
  users on the same inode.
* update DLD.
* cleanup.

18 years agofix for namelen used for audit purposes.
tappro [Thu, 25 Aug 2005 12:58:02 +0000 (12:58 +0000)]
fix for namelen used for audit purposes.

18 years agob=7390
tappro [Thu, 25 Aug 2005 09:41:12 +0000 (09:41 +0000)]
b=7390
- move reint auditing into mds_audit
- auth audit should use child inode if it exists to recognize correct audit setting

18 years agob=7390
tappro [Thu, 25 Aug 2005 09:36:28 +0000 (09:36 +0000)]
b=7390
audit on file didn't work on unlink due to loss of audit setting. Now audit mask is saved in smfs_inode_info also.

18 years agoBranch: HEAD
wangdi [Thu, 25 Aug 2005 07:07:27 +0000 (07:07 +0000)]
Branch: HEAD
b7148 typo fix

18 years ago- returned back grant related changes moved from 1_4 as they seem cause issues with #45
yury [Thu, 25 Aug 2005 06:34:25 +0000 (06:34 +0000)]
- returned back grant related changes moved from 1_4 as they seem cause issues with #45

18 years agoBranch: HEAD
wangdi [Thu, 25 Aug 2005 06:28:16 +0000 (06:28 +0000)]
Branch: HEAD
tmp hack fix for b7148, now we use confobd to write config log, so mds_config_version
is useless in config log name defination. Remove it temporarily.
b7148

18 years agob=7178
alex [Wed, 24 Aug 2005 21:49:24 +0000 (21:49 +0000)]
b=7178
r=tappro

 - avoid possibility to FLSIZE being sent back to client w/o checking

18 years agob=7178
alex [Wed, 24 Aug 2005 21:31:01 +0000 (21:31 +0000)]
b=7178

 - let's try to catch the case when late MDS_CLOSE updates attrs. from
   out-of-date i/o epoch

18 years agofile random-reads.c was initially added on branch b1_4.
nikita [Wed, 24 Aug 2005 19:27:03 +0000 (19:27 +0000)]
file random-reads.c was initially added on branch b1_4.

18 years ago- returned back OSC dirty size.
yury [Wed, 24 Aug 2005 14:05:51 +0000 (14:05 +0000)]
- returned back OSC dirty size.
- enabled filter_tally_write() to have brw_stats on OST

18 years agoBranch: HEAD
wangdi [Wed, 24 Aug 2005 09:45:12 +0000 (09:45 +0000)]
Branch: HEAD
b7375
when decrypt page, should decrypt the whole page array, not only the first
page

18 years agob=7356
alex [Wed, 24 Aug 2005 04:36:29 +0000 (04:36 +0000)]
b=7356

 - we should allow processing for request being resent during request
   replay stage. but we may do that only when all clients are connected
   (or after evicting clients that haven't connect in time)

18 years agoBranch: HEAD
wangdi [Wed, 24 Aug 2005 03:33:33 +0000 (03:33 +0000)]
Branch: HEAD
add lmv_precleanup for cleanup llog for lmv

18 years agob=7391
tappro [Tue, 23 Aug 2005 15:11:58 +0000 (15:11 +0000)]
b=7391
lctl fix for setattr
remove unused symlink opcode

18 years agoremove unused variable
tappro [Tue, 23 Aug 2005 15:08:25 +0000 (15:08 +0000)]
remove unused variable

18 years agob=7214
tappro [Tue, 23 Aug 2005 08:40:34 +0000 (08:40 +0000)]
b=7214
- fix duplicate name in open/stat auditing. Name is needed only if lustre id cannot be saved.

18 years agoset fsuid upon readpage.
lsy [Tue, 23 Aug 2005 06:26:36 +0000 (06:26 +0000)]
set fsuid upon readpage.

18 years agoadd "blacklist" support into OSS capabilities.
ericm [Tue, 23 Aug 2005 03:48:48 +0000 (03:48 +0000)]
add "blacklist" support into OSS capabilities.

18 years agoAdd LNET CVS Repository Support
pjkirner [Mon, 22 Aug 2005 23:47:06 +0000 (23:47 +0000)]
Add LNET CVS Repository Support

18 years agoremove some temporary debug msgs.
ericm [Mon, 22 Aug 2005 21:13:47 +0000 (21:13 +0000)]
remove some temporary debug msgs.

18 years agofew changes for debug purposes in OSC:
yury [Mon, 22 Aug 2005 16:54:08 +0000 (16:54 +0000)]
few changes for debug purposes in OSC:
- increased number in flight requests by factor 2. Thus twise is increased cl_dirty_max (in fact cache size)
- disabled check for in flight requests in osc_check_rpcs() - number of in flight RPCs is limited anyway by ->cl_dirty_max

18 years ago- added CERROR LBUG in filter_grant_check() from 1_4
yury [Mon, 22 Aug 2005 14:46:58 +0000 (14:46 +0000)]
- added CERROR LBUG in filter_grant_check() from 1_4

18 years ago- added missed fix to osc_exit_cache() from 1_4. It updates lost_grant mor carefuly...
yury [Mon, 22 Aug 2005 14:28:25 +0000 (14:28 +0000)]
- added missed fix to osc_exit_cache() from 1_4. It updates lost_grant mor carefuly what is important for cases when blocksize is not equal to page size.
- fixed warning about init of var in llog test stuff.

18 years ago- added more debug in osc. Currently two ideas are working for checks why osc does...
yury [Mon, 22 Aug 2005 11:57:40 +0000 (11:57 +0000)]
- added more debug in osc. Currently two ideas are working for checks why osc does not give desired band:

(1) cache size is choosen kind of cumbersome way and not enough, what makes writing thread waiting for cache in osc_enter_cache(). Cache size curremtly does not depend on how good wire is able to send RPCs to remote node what seems to be an artificial limit in some cases. If this is so, the fix would be to adjust cache size depending on much it is needed. This however does not explain why second subtest suffers from slowdown whereas first one does not.

(2) due to some reasons (limited cache or limited grant from OST) many async RPCs are falled back to synchronous IO what makes substential slowdown.

Added debug on real cluster should show what is going on there. Idea that issue on server has a little chance to be true, as server uses direct IO and there is almost not ability to be slowed down due to some cache issues, etc.

18 years ago* simplified gmnal thread startup/shutdown
eeb [Mon, 22 Aug 2005 11:42:01 +0000 (11:42 +0000)]
*   simplified gmnal thread startup/shutdown

18 years agowhile sanity test60 uninitialized llog_cookie in llog_test_4() can invoke assertion...
tappro [Mon, 22 Aug 2005 11:21:59 +0000 (11:21 +0000)]
while sanity test60 uninitialized llog_cookie in llog_test_4() can invoke assertion in llog_cat_current_log()

18 years ago* gmnal: removed spinlock round GM API calls that don't need to be serialised
eeb [Mon, 22 Aug 2005 10:58:22 +0000 (10:58 +0000)]
*   gmnal: removed spinlock round GM API calls that don't need to be serialised
    (they'd been added "blindly" previously)

*   gmnal netregression ran for 36 hours.

18 years ago- fixes about cobd_switch() which caused issues with cleanup:
yury [Mon, 22 Aug 2005 09:48:30 +0000 (09:48 +0000)]
- fixes about cobd_switch() which caused issues with cleanup:

  - avoid double connections to choosen export due to using lctl on umounted client (1st connection) and after that mount the client (2nd connection). Avoid double EA init along with double connection.

  - disconnect correct export on cobd_disconnect()

- avoid possible oops on double cobd_cleanup() due to freed cmobd->write_srv, but not zerod out after that.
- in cobd_switch() set ->cache_on field into new value only if COBD_CONNECT or COBD_SWITCH bias is specified.

- in LMV and LOV in lconf cleanup LMV and LOV first and after that cleanup all their targets. This is because LMV, etc. cleanup needs targets OBD attached, with correct obd_name, etc. This fixes prevents oops in disconnect time whrn using test45 configuration.

18 years agofixes:
lsy [Mon, 22 Aug 2005 08:31:00 +0000 (08:31 +0000)]
fixes:
* pack fid for direct io.
* seperate verify_fid from verify_capa to fully support truncate.
* debug message cleanup.

18 years agosanity test58 was failed due to structures changing. Setting proper values did fix...
tappro [Mon, 22 Aug 2005 07:41:59 +0000 (07:41 +0000)]
sanity test58 was failed due to structures changing. Setting proper values did fix that.

18 years agob=7354
tappro [Sun, 21 Aug 2005 22:56:12 +0000 (22:56 +0000)]
b=7354
 smfs didn't update backfs file pos with smfs file pos value before operation. This were critical for llog processing in some situations.

18 years agoBranch: HEAD
wangdi [Sun, 21 Aug 2005 06:32:18 +0000 (06:32 +0000)]
Branch: HEAD
comment out llog_init in lmv_add_mdc, because we already do llog_init in mdc_setup

18 years agoBranch HEAD
wangdi [Sat, 20 Aug 2005 19:37:38 +0000 (19:37 +0000)]
Branch HEAD
land add mds code to HEAD

18 years agodon't let ctx obj share storage with request buffer, which might be
ericm [Sat, 20 Aug 2005 19:08:25 +0000 (19:08 +0000)]
don't let ctx obj share storage with request buffer, which might be
released in case of recovery, request be saved and delayed handed.

18 years ago- more debug stuff in osc. Added av. pipe loading, max pipe loading, etc.
yury [Sat, 20 Aug 2005 16:46:00 +0000 (16:46 +0000)]
- more debug stuff in osc. Added av. pipe loading, max pipe loading, etc.
- some names cleanup in gns and capa threads related stuff to make all llite
  threads (gns, capa, close) coherent in names to each other.

18 years agoadd some debug message.
ericm [Fri, 19 Aug 2005 19:57:22 +0000 (19:57 +0000)]
add some debug message.

18 years agoBranch: HEAD
wangdi [Fri, 19 Aug 2005 17:55:54 +0000 (17:55 +0000)]
Branch: HEAD
b7374,7375 offset should be mapped inside page. when crypt and decrypt

18 years ago* cleaned up startup/shutdown handling
eeb [Fri, 19 Aug 2005 17:11:42 +0000 (17:11 +0000)]
*    cleaned up startup/shutdown handling
*    queue rx descriptors for handling by thread directly

18 years ago- temporary disable type check
tappro [Fri, 19 Aug 2005 10:31:51 +0000 (10:31 +0000)]
- temporary disable type check

18 years agouse inode->i_mode instead of fid type, and comment out the assertions, because
niu [Fri, 19 Aug 2005 09:36:15 +0000 (09:36 +0000)]
use inode->i_mode instead of fid type, and comment out the assertions, because
type in lustre id are not always correct.(client don't update type in lli_id;
id for cross-ref inode never store mode)

18 years agopack lustre id to oss in setattr on open
niu [Fri, 19 Aug 2005 08:50:23 +0000 (08:50 +0000)]
pack lustre id to oss in setattr on open