Whamcloud - gitweb
ericm [Thu, 15 Jan 2004 10:26:14 +0000 (10:26 +0000)]
merge from HEAD:
20030115
tianying [Thu, 15 Jan 2004 08:27:42 +0000 (08:27 +0000)]
Update from HEAD
cvs2svn [Sat, 10 Jan 2004 21:06:53 +0000 (21:06 +0000)]
This commit was manufactured by cvs2svn to create branch 'unlabeled-1.1.8'.
alex [Sat, 10 Jan 2004 21:06:46 +0000 (21:06 +0000)]
- series against second suse 2.4.21 kernel
b2505
alex [Sat, 10 Jan 2004 21:01:15 +0000 (21:01 +0000)]
- vanilla-2.4.24 series and needed patches
NOTE: do not use it for a while, please. ext3-delete-thread must
be reworked first.
alex [Fri, 9 Jan 2004 22:31:18 +0000 (22:31 +0000)]
file kgdb_2.6.0 was initially added on branch b_v26.
alex [Fri, 9 Jan 2004 22:31:17 +0000 (22:31 +0000)]
file vfs_intent_2.6.0.patch was initially added on branch b_v26.
alex [Fri, 9 Jan 2004 22:31:16 +0000 (22:31 +0000)]
file kernel_text_address-2.6.0.patch was initially added on branch b_v26.
alex [Fri, 9 Jan 2004 22:31:15 +0000 (22:31 +0000)]
file invalidate_show-2.6.0.patch was initially added on branch b_v26.
alex [Fri, 9 Jan 2004 22:31:14 +0000 (22:31 +0000)]
file ext3-start_this_handle-must-return-error.patch was initially added on branch b_v26.
alex [Fri, 9 Jan 2004 22:31:13 +0000 (22:31 +0000)]
file ext3-init-generation-2.6.0.patch was initially added on branch b_v26.
alex [Fri, 9 Jan 2004 22:31:12 +0000 (22:31 +0000)]
file export_symbols-2.6.0.patch was initially added on branch b_v26.
alex [Fri, 9 Jan 2004 22:31:10 +0000 (22:31 +0000)]
file 2.6.0-mm2.patch was initially added on branch b_v26.
jacob [Fri, 9 Jan 2004 21:10:11 +0000 (21:10 +0000)]
b=2429
r=phil,shaver
fix renaming a file to itself, and add a couple of tests for this
case.
also, make rename.pl not spin waiting for children.
jacob [Fri, 9 Jan 2004 20:27:17 +0000 (20:27 +0000)]
b=2502
fix a bug in --disable-doc handling, and add snapfs to DISTDIRS
wangdi [Fri, 9 Jan 2004 16:26:23 +0000 (16:26 +0000)]
fix some bugs in snapfs
wangdi [Fri, 9 Jan 2004 16:19:28 +0000 (16:19 +0000)]
add loop_device_get_info.patch in snapfs series
wangdi [Fri, 9 Jan 2004 16:10:43 +0000 (16:10 +0000)]
fix some bug in snap-core.patch 1) add some item in ext3_sb 2)some error handler
niu [Fri, 9 Jan 2004 07:02:26 +0000 (07:02 +0000)]
file localrpc_test.sh was initially added on branch b_localprc.
niu [Fri, 9 Jan 2004 07:02:23 +0000 (07:02 +0000)]
file ptlrpc_local.c was initially added on branch b_localprc.
youfeng [Fri, 9 Jan 2004 05:07:13 +0000 (05:07 +0000)]
b 2198
r braam
perl llanalyze dk.str -dlm MDC_uml_mds1_MNT_localhost
perl llanalyze dk.str \
-extract_start="004000:000400:0:
1073288749.491802:1780:898:6772" \
-extract_end="000040:080000:0:
1073288750.493802:1652:974:7014"
jacob [Thu, 8 Jan 2004 23:13:30 +0000 (23:13 +0000)]
b=2368
r=phil
- call vfs_rmdir() on directory orphans when removing them from
PENDING
- add a test for this case to replay-single
- add open(O_DIRECTORY) to multiop
rread [Thu, 8 Jan 2004 22:49:58 +0000 (22:49 +0000)]
Break, instead of panic, in the UML version of portals_debug_dumpstack.
rread [Thu, 8 Jan 2004 21:17:56 +0000 (21:17 +0000)]
b=2432,2494
test cases for these bugs.
Also some test enhancments
- add more lctl marks
- add SETUP CLEANUP options to replay-single
rread [Thu, 8 Jan 2004 21:07:23 +0000 (21:07 +0000)]
b=2494
Make sure we reset the timer after resending a request. Renamed
sending_error to force_timer_recalc, because that's what it does.
rread [Thu, 8 Jan 2004 20:21:38 +0000 (20:21 +0000)]
b=2432
Merging fix from b_llp2. Set the rq_no_resend for cancel requests so
ldlm_cli_cancel() can reconstruct the cancel with the correct
cookie.
adilger [Thu, 8 Jan 2004 08:36:26 +0000 (08:36 +0000)]
Fix typo in patch.
adilger [Thu, 8 Jan 2004 07:34:32 +0000 (07:34 +0000)]
Add missing ChangeLog entry.
zab [Thu, 8 Jan 2004 07:24:03 +0000 (07:24 +0000)]
Fix races in client write RPC generation when cache full
b=2482
(tested in buffalo against b1_0)
zab [Thu, 8 Jan 2004 06:37:29 +0000 (06:37 +0000)]
- fix up eeb's assertion fix. __GFP_WAIT isn't sufficient, it's GFP_ATOMIC that
callers use.
eeb [Wed, 7 Jan 2004 20:07:55 +0000 (20:07 +0000)]
. Added an important comment in parse_get()
adilger [Wed, 7 Jan 2004 19:47:11 +0000 (19:47 +0000)]
At least make some effort to detect errors in llog testing.
adilger [Wed, 7 Jan 2004 17:46:34 +0000 (17:46 +0000)]
Quiet error if run without args.
adilger [Wed, 7 Jan 2004 17:20:17 +0000 (17:20 +0000)]
Use munlink and not unlink (unlink doesn't exist in non-RH9 setups).
shaver [Wed, 7 Jan 2004 15:47:41 +0000 (15:47 +0000)]
Reverting 64K-stack change; r=phil.
wangdi [Wed, 7 Jan 2004 14:26:40 +0000 (14:26 +0000)]
update snapfs, include super.c filter.c snapfs_internal.h
wangdi [Wed, 7 Jan 2004 14:24:22 +0000 (14:24 +0000)]
add options.c
eeb [Wed, 7 Jan 2004 13:28:03 +0000 (13:28 +0000)]
. fixed ia64 format warnings
eeb [Wed, 7 Jan 2004 13:26:29 +0000 (13:26 +0000)]
. fixed assertion when called from interrupt handler (allocating lib_msg_t)
jerrifer [Wed, 7 Jan 2004 09:39:58 +0000 (09:39 +0000)]
set stacksize to 64K
wangchao [Wed, 7 Jan 2004 08:20:56 +0000 (08:20 +0000)]
fix a typo.
jerrifer [Wed, 7 Jan 2004 06:43:10 +0000 (06:43 +0000)]
add conf-sanity.sh to acceptance-small.sh
jerrifer [Wed, 7 Jan 2004 06:26:25 +0000 (06:26 +0000)]
update config file
wangchao [Wed, 7 Jan 2004 05:52:06 +0000 (05:52 +0000)]
b=2103
r=shaver
add tag to name2uuid() to deal with the broken of lmc when mds name and hostname are the same.
wangchao [Wed, 7 Jan 2004 05:07:18 +0000 (05:07 +0000)]
b=1719
r=shaver
Add --ptldebug and --subsystem to lmc.
adilger [Tue, 6 Jan 2004 18:34:09 +0000 (18:34 +0000)]
Make TCP zerocopy and pinger support enabled by default (affinity already was)
b=2476
r=phil
wangdi [Tue, 6 Jan 2004 14:33:36 +0000 (14:33 +0000)]
update snapfs
rread [Tue, 6 Jan 2004 00:58:08 +0000 (00:58 +0000)]
b=2460
Add a test to recovery-small to trigger this bug. Also fix some
typos/brainos in recovery-small and test-framework, and set the debug
level to the default for local configs,
jacob [Mon, 5 Jan 2004 20:05:51 +0000 (20:05 +0000)]
b=2421
r=phil
(i'm just committing jerrifer's work + a couple of typo fixes)
add a --disable-doc option to lustre's configure which skips .pdf doc
building (to ease rpm creation).
rread [Mon, 5 Jan 2004 18:57:17 +0000 (18:57 +0000)]
b=2460
r=phil
Make sure threads waiting for a completion callback are awoken when
the client is evicted. Also set LDLM_FL_CANCEL explicitely in
cleanup_resouces, so callers of ldmm_namespace_cleanup no
longer need to specify it.
adilger [Mon, 5 Jan 2004 18:52:14 +0000 (18:52 +0000)]
Don't use SLAB_KERNEL (GFP_KERNEL) allocations for locks and resources.
This is the bug that started it all, and was fixed only in b_llnl_prod...
b=1933
r=zab
adilger [Mon, 5 Jan 2004 18:07:18 +0000 (18:07 +0000)]
We haven't opened objects on OSTs in a long time. Remove unused field.
adilger [Mon, 5 Jan 2004 18:06:19 +0000 (18:06 +0000)]
Don't dereference a bad dchild if we had a lookup error (Alex found this).
Don't get inum lock if we didn't create an inode.
b=2362
wangdi [Mon, 5 Jan 2004 14:39:49 +0000 (14:39 +0000)]
fix some bugs of snapfs
jerrifer [Mon, 5 Jan 2004 06:03:49 +0000 (06:03 +0000)]
change tag from b_devel to HEAD
wangdi [Mon, 5 Jan 2004 03:09:46 +0000 (03:09 +0000)]
reomve i_filterdata in this patch, for vfs_intent patch have this item already. fix this temporary
wangdi [Sun, 4 Jan 2004 07:48:23 +0000 (07:48 +0000)]
add snapfs building support in lustre
wangdi [Sun, 4 Jan 2004 07:39:51 +0000 (07:39 +0000)]
add snapfs-2.4.20 kernel series
wangdi [Sun, 4 Jan 2004 07:37:43 +0000 (07:37 +0000)]
oops make a mistake, change snapfs_core-2.4.20.path to snapfs_core-2.4.20.patch
wangdi [Sun, 4 Jan 2004 07:32:14 +0000 (07:32 +0000)]
add snapfs_core-2.4.20.path
wangdi [Sun, 4 Jan 2004 07:29:43 +0000 (07:29 +0000)]
add snapconf utils
wangdi [Sun, 4 Jan 2004 07:28:26 +0000 (07:28 +0000)]
add snapfs_internal.h snapfs_support.h Makefile.am
wangdi [Sun, 4 Jan 2004 07:24:39 +0000 (07:24 +0000)]
port snapfs to 2.4
wangdi [Sun, 4 Jan 2004 07:22:40 +0000 (07:22 +0000)]
add snapfs to cvs
adilger [Sat, 3 Jan 2004 00:46:33 +0000 (00:46 +0000)]
Make "error" always report the test that failed, for SANITYLOG runs.
Make a bit more effort to clean up test cruft.
adilger [Sat, 3 Jan 2004 00:02:31 +0000 (00:02 +0000)]
Don't remove directory inodes from locks prematurely. Causes pages not to
be evicted if lock is revoked.
b=2451
r=phik
adilger [Fri, 2 Jan 2004 23:57:30 +0000 (23:57 +0000)]
Update ChangeLog with already-fixed bugs.
ericm [Fri, 2 Jan 2004 13:58:39 +0000 (13:58 +0000)]
file replay_ost_single.c was initially added on branch b_eq.
youfeng [Fri, 2 Jan 2004 04:04:29 +0000 (04:04 +0000)]
b_2466
r = braam
make portals_debug_dumpstack(void) calling panic("LBUG") in uml
eeb [Wed, 31 Dec 2003 14:58:38 +0000 (14:58 +0000)]
* Removed toenal
wangchao [Wed, 31 Dec 2003 07:42:11 +0000 (07:42 +0000)]
b=1171
r=shaver
create true uuid in lmc
ccooper [Wed, 31 Dec 2003 03:34:44 +0000 (03:34 +0000)]
- new test: run create and rename in same directory on multiple nodes in parallel
braam [Wed, 31 Dec 2003 02:35:58 +0000 (02:35 +0000)]
file filter_lvb.c was initially added on branch b_size.
adilger [Tue, 30 Dec 2003 20:56:18 +0000 (20:56 +0000)]
Reduce per-OSC client side cache to avoid pathalogical cache-flush times
with lots of clients. Tested at LLNL for a long time already.
Reduce max per-OSC RPCs in flight to reduce OST file fragmentation with
out-of-order writes to the filesystem.
r=Zach
ericm [Tue, 30 Dec 2003 08:24:46 +0000 (08:24 +0000)]
merge HEAD to b_eq: tag
20031230
alex [Tue, 30 Dec 2003 05:44:45 +0000 (05:44 +0000)]
- patch from Oleg to export get_kmem_end in UML kernel
tianying [Tue, 30 Dec 2003 02:37:44 +0000 (02:37 +0000)]
remove mount age check
b: 2332
r: Shaver
rread [Mon, 29 Dec 2003 20:36:00 +0000 (20:36 +0000)]
b=2442
Only create config logs for the MDS on the local node.
zab [Sun, 28 Dec 2003 21:33:36 +0000 (21:33 +0000)]
- remove more forgotten relics
zab [Fri, 26 Dec 2003 20:20:02 +0000 (20:20 +0000)]
b=2430
Use kmalloc() instead of slab to allocate the MD and MSG portal descriptors.
On x86_64 the MD and MSG structs got quite large, into the 2k range, though
pages are still 4k there. This tricked slab into trying to back their slabs
with multi-page allocations. These allocations failed under high load which
lead to LBUG()s and timeouts.
While we're in here, properly initialize the in_use counts and stop using
sleeping allocations while holding the state lock.
ccooper [Wed, 24 Dec 2003 19:18:29 +0000 (19:18 +0000)]
- use munlink instead of unlink
ccooper [Wed, 24 Dec 2003 16:15:27 +0000 (16:15 +0000)]
- ability to skip O_DIRECT tests on O_DIRECT-deficient systems (LLNL)
rread [Wed, 24 Dec 2003 01:38:20 +0000 (01:38 +0000)]
- test-frameworkize recovery-small.sh, and recovery-small.sh will
now run in a local, single node mode by default.
- Add some tests for old bugs to recovery-small.sh. These tests were
in the ltest/recovery scripts, but that is now unused.
- Move the OBD_FAIL funcs to test-framework.
- add the -v command line option for test scripts; it will print out
the commands its executing on the nodes.
rread [Tue, 23 Dec 2003 20:29:09 +0000 (20:29 +0000)]
b=2417
r=shaver
Prevent a race between mds_client_add and mds_client_free by clearing the client
index bit after the data has been zeroed. Also make config messages consistent
(with each other and with the obdfilter messages).
jerrifer [Tue, 23 Dec 2003 10:21:35 +0000 (10:21 +0000)]
change lfs from /usr/sbin to /usr/bin
adilger [Tue, 23 Dec 2003 08:49:40 +0000 (08:49 +0000)]
Fix valid uninitialized variable gcc warning.
youfeng [Mon, 22 Dec 2003 03:31:52 +0000 (03:31 +0000)]
b=2074
r=Adilger
Log:
remove the '\n's not at the end of log lines
green [Sun, 21 Dec 2003 10:26:13 +0000 (10:26 +0000)]
Reverted #974 for now as it causes problems for people.
Approved by Andreas Dilger.
phil [Sun, 21 Dec 2003 07:51:42 +0000 (07:51 +0000)]
Make the namespace/resource/lock dumping somewhat more compact, so
that less log space is wasted, and it's easier to visually scan.
phil [Sun, 21 Dec 2003 07:46:39 +0000 (07:46 +0000)]
b=2425
Jacob reported that when MDS/OST recovery requires new objects to be
created, the OST throws an assertion.
Bug 2425 remains open to track the creation of many more tests for
missing MDS/OST recovery cases.
phil [Sun, 21 Dec 2003 07:41:47 +0000 (07:41 +0000)]
Remove pesky $Id tag which only causes conflicts
rread [Fri, 19 Dec 2003 19:45:29 +0000 (19:45 +0000)]
b=2353
r=shaver
Delete IOC_CONNECT,DISCONNECT and use obd_self_export instead
of creating connections for lctl. Also delete the IOC_DEVICE comamnd
and make the ioctl interface stateless. The lctl probe command is now
a noop, and lctl device is still used to set the device, although the
current device state is only saved in lctl now, and not the kernel.
shaver [Fri, 19 Dec 2003 14:17:00 +0000 (14:17 +0000)]
b=2420: don't acquire a duplicate lock when processing a resent GETATTR, just
grab the dchild directly and sample the data. Fixes recovery-small.sh.
r=phik,buffalo
eeb [Fri, 19 Dec 2003 13:58:10 +0000 (13:58 +0000)]
* PtlMDUnlink() can no longer return PTL_MD_INUSE, since it commits the MD
for destruction. If no network I/O is current at the time, a
PTL_EVENT_UNLINK event is created.
* The 'unlinked_me' field of an event has been replaced by a simple flag
'unlinked' that is set if the event signals the destruction of the MD.
* Events have a new 'status' field. This is PTL_OK on successful
completion, and any other portals errno on completion with failure.
CWARN() messages in these callbacks log abnormal completion.
* All event callbacks changed to handle the UNLINK event, completion
status and unlinked flag.
* All abnormal completions changed to work with PltMDUnlink and the new
callbacks.
* Removed bd_complete from ptlrpc_bulk_desc and added bd_success.
Communications have completed when bd_network_rw gets cleared. If
bd_success is set, then bd_nob_transferred tells you how much data
was sent/received.
* Changed MDS and OST bulk completion to deal with failed bulk transfers.
The PtlBD server just LASSERTS things went OK, so we can be reminded to
implement better error handling there too.
* ptlrpc_wake_client_req() inline helper.
* Changed the lib/NAL interface as follows....
. cb_callback() is optional and defaults to calling the event queue's
callback if it is left NULL.
. cb_read(), cb_write(), cb_map(), cb_map_pages(), return PTL_OK on
success and another portals errno on failure.
. cb_send(), cb_send_pages(), cb_recv(), cb_recv_pages() return PTL_OK
if and only if they can commit to calling lib_finalize() when the
relevent message completes (possibly with error).
. cb_send(), cb_send_pages(), cb_recv(), cb_recv_pages() may not modify
the iovec/ptl_kiov_t they are passed, and must do I/O on the
subsection of this scatter/gather buffer starting at 'offset' for
'mlen' bytes. This greatly simplifies portals lib level descriptor
management at minimal expense to the NAL.
. portals lib now exports lib_extract_iov(), lib_extract_kiov() and the
other iov helpers take an additional 'offset' parameter, to simplify
offset buffer coding in the NAL.
. lib_parse() is void (i.e. returns no value).
. lib_finalize() takes an addition ptl_errno_t completion status.
...note that NALs other than qswnal and socknal need to have these
changes implemented properly and tested.
* Swapped some loose fprintf()s for CERROR()
* Dropped PORTAL_SLAB_ALLOC(); portals just uses PORTAL_ALLOC() now.
Since there are no slabs now, I also changed #ifdef PTL_USE_SLAB_CACHE
to #ifndef PTL_USE_LIB_FREELIST
* Changed lib_msg_alloc() so it is _never_ called with the statelock held,
just like all the other allocators.
* Changed dynamic MD allocation to size the MD by the number of fragments.
* Dropped a bunch of dross, plus the iovs from lib_msg_t so they become
tiny again.
alex [Fri, 19 Dec 2003 11:16:11 +0000 (11:16 +0000)]
- tcp_sendpage_zccd() must be exported always
phil [Thu, 18 Dec 2003 10:21:23 +0000 (10:21 +0000)]
b=2383
Stop taking a PR lock in mds_readpage; a PR is already held by the
client, so if there is a PW in the queue, deadlock will result. Just
assume that the client has a lock.
phil [Thu, 18 Dec 2003 09:45:44 +0000 (09:45 +0000)]
Print the service name in the mds RECOVERY: message
zab [Thu, 18 Dec 2003 04:13:42 +0000 (04:13 +0000)]
b=2252
r=adilger
(didn't see regressions in buffalo, confirmed read throughput increases
with sf and fpp multi-node IOR)
This cleans up llite's readpage path and implements our own read-ahead window
that hangs off of ll_file_data. The broad goal is to keep a fair amount of
read-ahead pages issued and queued which can be fired off into read rpcs as
read-ahead rpcs are completed.
zab [Thu, 18 Dec 2003 03:59:08 +0000 (03:59 +0000)]
- put llite page cache pages in a list_head for the duration
of their stay in the page cache. This lets us display the contents
of the page cache via llite/*/dump_pgcache file. This was done as part
of b=2252 and is being committed seperately from the read-ahead work.
adilger [Wed, 17 Dec 2003 19:49:18 +0000 (19:49 +0000)]
Silence bogus compiler warning.