Whamcloud - gitweb
alex [Wed, 1 Jun 2005 10:26:37 +0000 (10:26 +0000)]
b=6427
- pass OBD_MD_FLID for raw lookup
adilger [Wed, 1 Jun 2005 07:42:11 +0000 (07:42 +0000)]
Branch HEAD
Build fix for HEAD.
r=oleg (original fix)
yury [Wed, 1 Jun 2005 07:12:10 +0000 (07:12 +0000)]
- one more fix about jumping to wrong label in mds_create_objects()
yury [Wed, 1 Jun 2005 07:06:45 +0000 (07:06 +0000)]
- fixed jump to wrong label in mds_create_objects() in error path.
alex [Tue, 31 May 2005 21:39:59 +0000 (21:39 +0000)]
- minor fixes in smfs to support raw lookups
green [Tue, 31 May 2005 20:25:50 +0000 (20:25 +0000)]
Branch: b_cray-1_4_2
If liblustre was built, add liblustre.{a,so} to the lustre package
nic [Tue, 31 May 2005 20:04:25 +0000 (20:04 +0000)]
add larry's changes that allow using mpirun or yod for running IOR.
alex [Tue, 31 May 2005 19:52:48 +0000 (19:52 +0000)]
b=5843
- don't postpone orphan unlinks because
ericm [Tue, 31 May 2005 18:28:51 +0000 (18:28 +0000)]
quiet some verbose gss messages.
adilger [Tue, 31 May 2005 18:15:47 +0000 (18:15 +0000)]
Branch b1_4
On x86_64 the kernel declares __u64 as long long, but userspace uses long.
Fix LP*64 printf format macros to handle this difference to quiet warnings.
wangdi [Tue, 31 May 2005 14:52:12 +0000 (14:52 +0000)]
Brach HEAD
r:alex
add lookup-raw patches for raw lookup acceptance tests
alex [Tue, 31 May 2005 12:41:17 +0000 (12:41 +0000)]
- don't try to show nid for non-connected exports
tappro [Tue, 31 May 2005 11:31:54 +0000 (11:31 +0000)]
- add debug message
alex [Tue, 31 May 2005 11:20:31 +0000 (11:20 +0000)]
b=6379
- even a simpler fix: just skip removal if orphan is already unlinked
semaphore on pending dir should be enough
tappro [Tue, 31 May 2005 11:02:50 +0000 (11:02 +0000)]
b=6414,6366
r=alex
it seems that this ASSERT is not needed here really. Such case is allowed
eeb [Tue, 31 May 2005 10:24:25 +0000 (10:24 +0000)]
* Fix for 6357: handle RapidArray memory registration failures.
alex [Tue, 31 May 2005 10:10:25 +0000 (10:10 +0000)]
- show nid for stale clients
alex [Tue, 31 May 2005 07:58:02 +0000 (07:58 +0000)]
- export show_task()
alex [Tue, 31 May 2005 07:16:22 +0000 (07:16 +0000)]
b=6379
- check for open count only in mds_cleanup_orphans()
- drop orphan flag to avoid simultaneous unlinking from PENDING/
alex [Tue, 31 May 2005 06:41:53 +0000 (06:41 +0000)]
b=6403
- revert wrong fix back
adilger [Mon, 30 May 2005 20:25:33 +0000 (20:25 +0000)]
HEAD
Fix error messages.
alex [Mon, 30 May 2005 13:50:05 +0000 (13:50 +0000)]
- minor debug message fix
alex [Mon, 30 May 2005 11:30:07 +0000 (11:30 +0000)]
b=6403
- remove ugly debug bits
- dump FL_REPLAY reqs once after connect to avoid flooding
yury [Mon, 30 May 2005 07:42:57 +0000 (07:42 +0000)]
- more compilation fixes after discussing them with Andreas
alex [Sun, 29 May 2005 22:43:10 +0000 (22:43 +0000)]
b=6413
- wake recovery thread up on any {REQ|LOCK}_REPLAY_DONE
alex [Sun, 29 May 2005 21:03:52 +0000 (21:03 +0000)]
b=6379
- drop orphan flag to prevent a race between mds_cleanup_orphans() and
mds_mfd_close() when the both try to unlink the same orphan
yury [Sun, 29 May 2005 16:17:17 +0000 (16:17 +0000)]
- many gcc4 compilation fixes (warnings)
alex [Sun, 29 May 2005 09:34:30 +0000 (09:34 +0000)]
b=6403
- ll_dentry_iput() clears all l_ast_data and this makes impossible to
close a file upon openlock cancel
alex [Sat, 28 May 2005 11:16:28 +0000 (11:16 +0000)]
- minor debug messages fixes
adilger [Fri, 27 May 2005 22:45:28 +0000 (22:45 +0000)]
Branch b1_4
Re-enable rate-limiting of CDEBUG messages. The lack of which is causing
grief on large clusters during recovery.
b=4311
lincent [Fri, 27 May 2005 21:31:04 +0000 (21:31 +0000)]
file llog_reader.c was initially added on branch b_cray_newconfig.
alex [Fri, 27 May 2005 21:02:00 +0000 (21:02 +0000)]
b=6403
- additional check because a req with FL_REPLAY can be w/o repmsg
yury [Fri, 27 May 2005 18:19:41 +0000 (18:19 +0000)]
- landing MOD (MDS Originated destroy) plus few ldlm and MDS fixes.
alex [Fri, 27 May 2005 13:38:44 +0000 (13:38 +0000)]
b=6403
- additional debug mess(age?) to learn 6K depth open req queue on cnbuild
tappro [Fri, 27 May 2005 12:51:54 +0000 (12:51 +0000)]
b=6366
r=alex
- delete_inode() for backfs inode should be called synchronous with smfs. To do this smfs_clear_inode_info() is called from smfs_delete_inode().
wangdi [Fri, 27 May 2005 09:50:41 +0000 (09:50 +0000)]
Branch HEAD
add types.h in Makefile.am for some tests build
alex [Thu, 26 May 2005 22:55:27 +0000 (22:55 +0000)]
b=6403
- the situation looks valid with opencache locks, but it floods consoles
over. make it silent for a while
ericm [Thu, 26 May 2005 19:19:56 +0000 (19:19 +0000)]
update to krb5 script: check krytab for root.
ericm [Thu, 26 May 2005 19:14:59 +0000 (19:14 +0000)]
b=6394
* don't touch base rpcpipe chain.
* invalidate svc gss cache item in case of init error.
alex [Thu, 26 May 2005 19:02:36 +0000 (19:02 +0000)]
- minor log format fix
ericm [Thu, 26 May 2005 17:13:15 +0000 (17:13 +0000)]
quite some unnecessary messages and a compiler warning.
alex [Thu, 26 May 2005 17:05:54 +0000 (17:05 +0000)]
b=6398
- additional assert
alex [Thu, 26 May 2005 16:59:58 +0000 (16:59 +0000)]
- don't log open replays
alex [Thu, 26 May 2005 11:08:39 +0000 (11:08 +0000)]
b=6398
- target_start_recovery_thread() starts a thread and must wait for
completion because that thread sets recovering flag. but routine
kernel_thread() returns pid > 0 in successful case, while the
target_start_recover_thread() expects 0. so, first routine
target_start_recovery_thread() returns immediately and chain
mds_postsetup() -> mds_dt_connect() is called with obd_recovering==0.
so, mds_postrecov_common() is called and it calls mds_cleanup_orphans()
alex [Wed, 25 May 2005 21:35:00 +0000 (21:35 +0000)]
b=6368
- use time_after_eq() instead of simple comparing
ericm [Wed, 25 May 2005 18:08:29 +0000 (18:08 +0000)]
* add a bit in ptlrpc_request to indicate ptlrpcs error, gss recovery
fixes, need new nfsutils patches version 8.
* more precise error notify between kernel and lgssd.
* minor gss api param fix. and various minor gss fixes.
adilger [Wed, 25 May 2005 15:33:20 +0000 (15:33 +0000)]
file upcall_cache.c was initially added on branch b1_4_bug3389.
wangdi [Wed, 25 May 2005 07:54:27 +0000 (07:54 +0000)]
Branch: HEAD
remove unsupported target on HEAD
wangdi [Wed, 25 May 2005 07:52:32 +0000 (07:52 +0000)]
Branch: HEAD
remove SMFS options in autoMakefile.am because we should build smfs by default
on HEAD
wangdi [Tue, 24 May 2005 10:11:43 +0000 (10:11 +0000)]
Branch HEAD
1) remove unsupported target in lustre-core.m4 according to boc's suggest
2) add some args on 2.6-fc3 target
alex [Tue, 24 May 2005 10:01:06 +0000 (10:01 +0000)]
- enable tcp keepalive back
alex [Tue, 24 May 2005 09:54:48 +0000 (09:54 +0000)]
- disable tcp keepalive for debugging purpose
jacob [Tue, 24 May 2005 09:52:53 +0000 (09:52 +0000)]
revert unapproved change (lustre-build HEAD is shared by all branches; the proper place for this change is in HEAD's .target.in files)
wangdi [Tue, 24 May 2005 09:08:07 +0000 (09:08 +0000)]
Brach: HEAD
add kernel-2.6.10-2.6-fc3-i686-smp.config for lbuild on HEAD
wangdi [Tue, 24 May 2005 08:56:07 +0000 (08:56 +0000)]
Branch HEAD
change LINUX26 from 0 to 1 for HEAD in lbuild
wangdi [Tue, 24 May 2005 07:47:16 +0000 (07:47 +0000)]
Branch: HEAD
some fix for fc3 lbuild on HEAD
alex [Mon, 23 May 2005 15:38:39 +0000 (15:38 +0000)]
- grrr, they are unsigned ...
tappro [Mon, 23 May 2005 11:29:34 +0000 (11:29 +0000)]
- changes for KML setting and variables in accordance with smfs plugin API
tappro [Mon, 23 May 2005 11:25:57 +0000 (11:25 +0000)]
- KML fixes and updates due to SMFS plugin API.
alex [Mon, 23 May 2005 10:27:32 +0000 (10:27 +0000)]
- brown paper bug (negative value is good)
alex [Sun, 22 May 2005 12:10:37 +0000 (12:10 +0000)]
- take HZ into account
yury [Sun, 22 May 2005 09:44:39 +0000 (09:44 +0000)]
- fixed CDEBUG message
alex [Sat, 21 May 2005 19:00:19 +0000 (19:00 +0000)]
- on cnbuild ping time is set to some strange value, the patch is intended
to catch and fix the problem
yury [Sat, 21 May 2005 13:54:26 +0000 (13:54 +0000)]
- handling out possible signals to @current while waiting for mount.
- checking for possible dentry mounting even if wait function returned failure - this helps to avpid post mounts.
- many changes in sanity-gns.sh
wangdi [Sat, 21 May 2005 12:51:11 +0000 (12:51 +0000)]
Branch HEAD
In smfs_helper, replace list_for_each_entry_safe with list_for_each_entry,
because we may unplugin the plg list
yury [Sat, 21 May 2005 11:14:10 +0000 (11:14 +0000)]
- fixes in sanity-gns.sh
alex [Sat, 21 May 2005 08:07:24 +0000 (08:07 +0000)]
- bug fixed: wrong lock type was passed to ldlm_lock_decref()
phil [Fri, 20 May 2005 23:03:16 +0000 (23:03 +0000)]
r=jacob
check out libsysio HEAD for all lustre builds
phil [Fri, 20 May 2005 22:55:11 +0000 (22:55 +0000)]
b=3359
r=lee@sandia.gov
libsysio fixes which always apply the umask before calling liblustre.
With these changes, liblustre can stop worrying about the umask entirely.
phil [Fri, 20 May 2005 22:53:54 +0000 (22:53 +0000)]
As long as we have lustre hack flags, let's at least compile with -O0,
which makes debugging at least theoretically possible.
phil [Fri, 20 May 2005 22:44:08 +0000 (22:44 +0000)]
add -DLUSTRE_UTILS=1 to the utils build
phil [Fri, 20 May 2005 22:35:35 +0000 (22:35 +0000)]
- Fixes a catstrophic bug introduced in the merge, wherein liblustre OBD_ALLOC
wasn't zeroing memory
- eliminate CDEBUG helper dupes; everything references CDEBUG now
- I wanted to keep obd_memory, portals_debug, et al in liblustre, but I can't
have those bits compiled into userspace utils. Hence the LUSTRE_UTILS
preprocessor definition to distinguish them.
brian [Fri, 20 May 2005 19:03:42 +0000 (19:03 +0000)]
mountain need Qlogic 2300 driver as well.
brian [Fri, 20 May 2005 16:43:32 +0000 (16:43 +0000)]
Quadrics' elan config variables from mountain.
wangdi [Fri, 20 May 2005 15:52:00 +0000 (15:52 +0000)]
Branch: HEAD
some fixes about write back cache.
1)cleanup smfs_init_inode_info
2)miss spin_unlock in class_find_client_obd
3)remove PLG_ALL flags of OBJECTS and LOGS dir on MDS
4)some minor fix in cmobd.sh for WB cache
alex [Fri, 20 May 2005 11:35:45 +0000 (11:35 +0000)]
b=6360
- couple asserts and debug info
phil [Thu, 19 May 2005 23:22:02 +0000 (23:22 +0000)]
add some real userspace definitions for CDEBUG et al
alex [Thu, 19 May 2005 21:18:22 +0000 (21:18 +0000)]
b=6316
- all clients report their minimal transno to be replayed
- server waits for all clients to connect in order to collect
minimal transno's from clients and replay all requests
in right order
alex [Thu, 19 May 2005 21:14:53 +0000 (21:14 +0000)]
b=6358
- simplest mkdir placement to test the idea on mountain
jacob [Thu, 19 May 2005 20:35:56 +0000 (20:35 +0000)]
b=6336
only disable qswnal when building against qsnet inside the kernel.
jacob [Thu, 19 May 2005 20:33:32 +0000 (20:33 +0000)]
b=6336
disable qswnal if neither CONFIG_QSNET nor CONFIG_QSNET_MODULES is
defined (ie on ppc64).
jacob [Thu, 19 May 2005 19:58:38 +0000 (19:58 +0000)]
b=6265
remove .pyc files when uninstalling
wangdi [Thu, 19 May 2005 16:45:33 +0000 (16:45 +0000)]
Branch: HEAD
1.port new lustre_lcfg structure from b1_4 to HEAD
2.some fix on test scripts for WB cache
eeb [Thu, 19 May 2005 14:04:32 +0000 (14:04 +0000)]
* changed router lookups to always return the local destination
socknal + qswnal changes to match
simplified "will this be forwarded" predicate in ptl_send()
* fixed bug in multi-instance socknal to pass the correct NI to the tx
teardown function to allow for the case where the tx hasn't even had a conn
assigned yet.
* fixed routerstat stats formatting
* removed osc::check_elan_limit() and replaced with max_unfragmented_pages()
which actually works. NB it depend on the assumption that file and page
offsets are aligned!
* fixed lconf to stay away from portals, except to 'modprobe' for it first in
'load_module()', and to rmmod all the NALs + portals when ptlrpc gets
unloaded.
alex [Thu, 19 May 2005 11:35:16 +0000 (11:35 +0000)]
- wrong command lctl dk /r/tmp/debug in test 55
ericm [Wed, 18 May 2005 23:26:15 +0000 (23:26 +0000)]
b=6332
create dir across MDS's, subdir need default ACL of its parent to determine
its own default ACL and access ACL. API obd_create() is added 2 params
to hold acl data.
alex [Wed, 18 May 2005 21:29:12 +0000 (21:29 +0000)]
- minor fixes in the script
alex [Wed, 18 May 2005 20:44:06 +0000 (20:44 +0000)]
- ptlrpc_ping_interpret is needless:
general timeout infrastructure is enough for cascading timeouts
alex [Wed, 18 May 2005 20:41:05 +0000 (20:41 +0000)]
- ptlrpc_request_handle_notconn() shouldn't reconnect if recovery on
import is disabled adminitratively
tappro [Wed, 18 May 2005 11:52:27 +0000 (11:52 +0000)]
b=6344
r=alex
- fix bug with flag DCACHE_CROSS_REF that can be unsetted sometimes.
wangdi [Wed, 18 May 2005 09:02:56 +0000 (09:02 +0000)]
Branch: HEAD
some fix about x86_64 lkcd patch
wangdi [Tue, 17 May 2005 10:26:35 +0000 (10:26 +0000)]
Branch HEAD
add some config options for CGG
tappro [Tue, 17 May 2005 09:02:45 +0000 (09:02 +0000)]
b=6274
r=alex
-fix nasty bug with using function in stack
tappro [Mon, 16 May 2005 18:33:01 +0000 (18:33 +0000)]
b=5881, 6274
r=alex
-fix bug with setting i_ino (thanks alex)
-fix issue with non-freed dentry after error
-cleanup smfs_do_lookup() code
liangzhen [Mon, 16 May 2005 18:33:00 +0000 (18:33 +0000)]
Small fix for OS X uninstall script
liangzhen [Mon, 16 May 2005 18:07:50 +0000 (18:07 +0000)]
Files needed by OS X package
yury [Mon, 16 May 2005 16:22:41 +0000 (16:22 +0000)]
- many changes in sanity-gns.sh. Enabled last test (3b).
yury [Mon, 16 May 2005 08:04:51 +0000 (08:04 +0000)]
- fixed umounting all GNS entrie son lustre umount (added umount_lustre patch)
- some typos fixed.
yury [Mon, 16 May 2005 06:48:21 +0000 (06:48 +0000)]
- fixes and improvements in GNS:
- making it work in some stat cases (stat target is behind mount object)
- making it work in sys_mount() control path (is mount point is behind mount object)
- to avoid deadlock is mount case is_subdir() is used for detecting is current dentry is
parent of dentry being in progress of mounting.
- do not mount on revalidate if there is not valid @nd applied.
- using nd->flags if possible for both lookup and revalidate cases.
- added debug info printing in mounting cases.
- totaly removed -ERESTARTSYS approach stuff.
- fixed bug with many waiters on ->ll_gns_mount_finished - only first of them will find GNS
in IDLE state and will start mounting and the rest will assert on wrong GNS state (MOUNTING)
- reinit ->ll_gns_mount_finished after all waiters are awakaned, this fixes the bug when
after first using this completion it does not functions, because complete_all() has
slighly diff. semanthics then complete() and namely does not just decrease ->done but
sets it to MAX_INT/2.
- fully rewritten sanity-gns.sh to make it more flexible and easy to add new sanity tests.
wangdi [Sun, 15 May 2005 17:07:11 +0000 (17:07 +0000)]
Branch: HEAD
Should return 0, if the return value of mds_postrecov_common > 0