Whamcloud - gitweb
braam [Tue, 15 Jul 2003 08:33:13 +0000 (08:33 +0000)]
- missing patches for b_unify from the 2.5 series
braam [Tue, 15 Jul 2003 08:27:57 +0000 (08:27 +0000)]
- add the unhash method to the kernel
braam [Tue, 15 Jul 2003 03:51:12 +0000 (03:51 +0000)]
- 2.5 kernel patches from ad for unify
mmex [Tue, 15 Jul 2003 02:42:06 +0000 (02:42 +0000)]
- add ext3-map_inode_page.patch to all series
- add ext3-error-export.patch to all series and add new export symbols that
will be needed in last_rcvd_records_via_metadata patch.
braam [Tue, 15 Jul 2003 00:01:21 +0000 (00:01 +0000)]
- remaining b_unify ---> b_ad merge
adilger [Mon, 14 Jul 2003 22:37:20 +0000 (22:37 +0000)]
Instead of setting all of the valid flags, and then clearing them for
regular files if an EA is found, we instead leave them unset for regular
files and only set them if we find we do not find an EA. This avoids
sending back bad data to the client if we abort early because of error
and do not clear the valid flags (POSIX.os/files/open/T.open 31).
shaver [Mon, 14 Jul 2003 22:32:59 +0000 (22:32 +0000)]
revert mmex's accidental checkin
mmex [Mon, 14 Jul 2003 22:22:28 +0000 (22:22 +0000)]
- cleanup come of the configuration files to be more compatible and
define some new things to help building rpms.
braam [Sat, 12 Jul 2003 21:41:29 +0000 (21:41 +0000)]
- add a chaos 2.4.20 patch series to b_intent
mmex [Sat, 12 Jul 2003 21:21:46 +0000 (21:21 +0000)]
- fix my typo
- include string.h so we don't get a warning everytime dcache.h is included,
which is often.
mmex [Sat, 12 Jul 2003 20:51:27 +0000 (20:51 +0000)]
- resolve conflict between 2.4.20 vanilla kernel and hp kernel. The XFS
patches cause us problems.
alex_tomas [Sat, 12 Jul 2003 18:30:39 +0000 (18:30 +0000)]
- ll_fill_super() calls lprocfs_register_mountpoint() _before_
*iget
- lprocfs_counter_add() shows backtrace in counter var isn't initialized yet
alex_tomas [Sat, 12 Jul 2003 09:51:50 +0000 (09:51 +0000)]
- inode-protection-from-pdflush.patch: an inode may have I_SKIP_PDFLUSH
flag which protects it from to be processed by pdflush's writeback
mechanism. this allows us to work locking problem with lustre_fwrite()
around
- MDS sets I_SKIP_PDFLUSH flag on 'last record' file
- filter_prep() sets I_SKIP_PDFLUSH flag
- inode-protection-from-pdflush.patch has been included in kgdb-2.5.73
series
alex_tomas [Sat, 12 Jul 2003 09:36:56 +0000 (09:36 +0000)]
- ext3-noread-inode.patch added
braam [Sat, 12 Jul 2003 08:28:53 +0000 (08:28 +0000)]
- cleanup in super and llite_lib to make this closer to b_unify.
morrone [Sat, 12 Jul 2003 00:49:01 +0000 (00:49 +0000)]
Stop routerstat from resetting the router's statistics after every read, which
will allow multiple users to run routerstat at the same time. Since the
stats aren't reset, it also needs to handle rollover.
shaver [Fri, 11 Jul 2003 08:36:06 +0000 (08:36 +0000)]
Substantive:
- don't resend pinger pings; there's really no point, and it does so terribly
confuse the mighty set_wait state machine.
- teach the mighty set_wait state machine about rq_no_resend.
Less substantive:
- defense of the sacred 80th column
adilger [Fri, 11 Jul 2003 07:48:04 +0000 (07:48 +0000)]
Truncate thread for 2.4.18-18 kernel.
b=1436
braam [Fri, 11 Jul 2003 05:52:49 +0000 (05:52 +0000)]
- test and cleanup alex work on super/super25 and llite_lib
- fix the follow link issue introduced during the merge
mmex [Fri, 11 Jul 2003 01:37:28 +0000 (01:37 +0000)]
- fix buffer overrun in openfile
- add test 39 which tests for mtime change on file creat
alex_tomas [Thu, 10 Jul 2003 20:43:06 +0000 (20:43 +0000)]
- almost all routines from super.c and super25.c extracted into llite_lib.c
alex_tomas [Thu, 10 Jul 2003 20:17:47 +0000 (20:17 +0000)]
- llite_lib.c is collection of routines beind used by super.c and super25.c
shaver [Thu, 10 Jul 2003 19:07:37 +0000 (19:07 +0000)]
Fix logging and silly list-math bug. (Some day, some happy day, I will remember
the order of the arguments to list_add.)
Half-hearted attempt at disconnecting the mgmt_cli at cleanup time instead of
last-deregister, to work around the lconf-socket-closure Russian Roulette
situation. Will get better (reverted, in fact) when eeb lands b_multinet.
adilger [Thu, 10 Jul 2003 18:57:31 +0000 (18:57 +0000)]
Landing utime et. al. fixes:
- bug 1014, 36 - utime fixes
- bug 1365 - crash on ops with no objects (truncate, utime)
- bug 318 - readahead prep work - OST gets client file handle on reads
braam [Thu, 10 Jul 2003 18:38:17 +0000 (18:38 +0000)]
- fix up the filter Makefile to pick up Phil's new files
braam [Wed, 9 Jul 2003 23:52:26 +0000 (23:52 +0000)]
- kernel patches for 2.4.18
adilger [Wed, 9 Jul 2003 22:03:30 +0000 (22:03 +0000)]
Add fix-tabs header + copyright.
braam [Wed, 9 Jul 2003 18:39:30 +0000 (18:39 +0000)]
- another merge of b_devel into b_ad
pschwan [Wed, 9 Jul 2003 18:25:29 +0000 (18:25 +0000)]
- osc_mark_page_dirty was always returning -EDQUOT because no grant
had been received
- add the ext3 map_inode_page patch
- call write_inode_now(), instead of filemap_fdata{sync,wait}, which
correctly takes inodes off of the superblock dirty list
- add an O_DIRECT-style filter_commitrw_write
devendragarg [Wed, 9 Jul 2003 12:20:36 +0000 (12:20 +0000)]
Updated with review comments.
- replace multiple print functions with one display method.
adilger [Wed, 9 Jul 2003 02:06:28 +0000 (02:06 +0000)]
Have a return value to quiet compiler warning.
adilger [Wed, 9 Jul 2003 01:47:42 +0000 (01:47 +0000)]
Move some common definitions into filter_internal.h
shaver [Wed, 9 Jul 2003 01:00:40 +0000 (01:00 +0000)]
teach jt_dbg_modules about mgmt_svc and mgmt_cli
shaver [Wed, 9 Jul 2003 00:35:38 +0000 (00:35 +0000)]
lconf support for client-side loading of management client as required.
Bump config version to account for new elements, just to be safe.
adilger [Tue, 8 Jul 2003 23:34:04 +0000 (23:34 +0000)]
Whitespace only.
braam [Tue, 8 Jul 2003 22:52:34 +0000 (22:52 +0000)]
- minor fork of chaos vfs intent patch
braam [Tue, 8 Jul 2003 22:40:55 +0000 (22:40 +0000)]
- we could call this "merely a merge from b_devel". Notes:
- lprocfs is not managable
- this oopses on mount, will be fixed soon.
adilger [Tue, 8 Jul 2003 21:08:31 +0000 (21:08 +0000)]
Minor cleanups.
adilger [Tue, 8 Jul 2003 19:34:04 +0000 (19:34 +0000)]
Grab i_sem on the parent for simple_mknod() and simple_mkdir() functions.
With b_orphan these are used during real operations and not just setup.
adilger [Tue, 8 Jul 2003 17:58:46 +0000 (17:58 +0000)]
Move one more inline prototype to a header.
adilger [Tue, 8 Jul 2003 17:41:49 +0000 (17:41 +0000)]
Add lov_internal.h for LOV function definitions/structs.
shaver [Tue, 8 Jul 2003 17:28:38 +0000 (17:28 +0000)]
Clean up opcode/portal selection a bit, and add support for the management
client as a client type. Add optional (default off) connection from client
OBDs to the management client for event registration/deregistration.
shaver [Tue, 8 Jul 2003 17:11:35 +0000 (17:11 +0000)]
add S_MGMT for management-service logging
shaver [Tue, 8 Jul 2003 17:09:33 +0000 (17:09 +0000)]
Initial implementation of management-ping client.
adilger [Mon, 7 Jul 2003 21:32:46 +0000 (21:32 +0000)]
Move some function declarations into llite_internal.h.
pschwan [Mon, 7 Jul 2003 19:07:26 +0000 (19:07 +0000)]
b=1141
Back out my obdfilter transaction changes; there are at least two
serious lock inversion bugs, and these changes will need to be rethought.
pschwan [Mon, 7 Jul 2003 05:55:29 +0000 (05:55 +0000)]
b=1141
This patch breaks the previous single brw disk transaction into two:
one opened and closed in each of preprw and commitrw. It also avoids
holding the i_sem across the brw. No stunning results yet, but I expect
that our behaviour in the face of client bulk timeout is substantially
improved. Some more analysis will be required before we turn this
into a performance win, I think.
pschwan [Mon, 7 Jul 2003 00:06:46 +0000 (00:06 +0000)]
Remove two printks from startup; every line that we don't print is
1100 fewer lines in the MCR /var/log/messages
pschwan [Sun, 6 Jul 2003 22:55:42 +0000 (22:55 +0000)]
Some kernels apparently need an explicit include for timeval here
pschwan [Sun, 6 Jul 2003 20:07:45 +0000 (20:07 +0000)]
b=1141
Another mostly-reorganization commit, to reduce the size of my working
patch set.
- remove gratuitous lustre_put_page() inline
- split filter_preprw() into filter_preprw_read() and filter_preprw_write()
- reverse the 2.5 waitfor_one_page macro into a 2.4 wait_on_page_locked macro
- unimportant changes galore
pschwan [Sun, 6 Jul 2003 18:49:43 +0000 (18:49 +0000)]
b=1141
A reorganization of obdfilter/ before I launch into the real work.
There should be no functional changes.
- removes unused 'niobuf_remote' argument from fsfilt_brw_start
- splits filter.c into filter.c, filter_io.c, filter_log.c, and filter_san.c
More of the same is on the way, but this is a good start.
pschwan [Sun, 6 Jul 2003 01:16:48 +0000 (01:16 +0000)]
b=1502
- Discard incredibly old requests (older than "obd_timeout" seconds) in
ptlrpc_main without even trying to process them, because the client
has already given up on us. See the bug text for more details on why
this is the right thing to do.
- Clean up deeply-nested ptlrpc_main()
- On Zach's good advice, I also fixed Portals to record when a packet
arrives in useful microseconds instead of relatively useless cycles
pschwan [Sat, 5 Jul 2003 22:47:31 +0000 (22:47 +0000)]
land b_merge on b_devel. Includes b_mount and b_orphan, which includes:
- orphans are moved into the PENDING directory for possible recovery
- replayed opens now open by fid for orphan/rename safety (1042)
- last close of an orphan inode generates a transno (683)
- avoid CERROR in normal ll_setattr_raw() error case (1500)
Added code which is currently disabled:
- dentry pinning from b_mount for cwd and mountpoints (1020)
- logging code for orphans (and many other things)
shaver [Sat, 5 Jul 2003 14:50:51 +0000 (14:50 +0000)]
small mgmt_svc tweaks
adilger [Sat, 5 Jul 2003 07:03:38 +0000 (07:03 +0000)]
Allow some "slop" in the timestamps, because UML isn't very good at keeping
accurate times if it is busy running.
sravi [Fri, 4 Jul 2003 07:17:50 +0000 (07:17 +0000)]
new ioctl statfs has been added as part of the globalhealth. Now the lctl have
command statfs, which is used for stat on MDS or OST.
This is a utility get the MDS/OST status and generate the snmp trap.
rread [Thu, 3 Jul 2003 23:05:04 +0000 (23:05 +0000)]
b=1469
Create a simple wrapper around gethostbyname, so we can print a sane
error message when the host is not found, instead of "No such file or
directory."
zab [Thu, 3 Jul 2003 22:36:21 +0000 (22:36 +0000)]
ok, no, really, commit the qla2xxx driver.
shaver [Thu, 3 Jul 2003 19:18:33 +0000 (19:18 +0000)]
initial mgmt/monitoring service skeleton
sravi [Thu, 3 Jul 2003 17:16:38 +0000 (17:16 +0000)]
This file is to check the health of the OST as part of the global-health.
This does not modify any other files lustre/portals.
May need to optimize this a bit.
eeb [Thu, 3 Jul 2003 15:40:30 +0000 (15:40 +0000)]
* Added missing kportal_put_ni() when forcing portals failures
pschwan [Thu, 3 Jul 2003 05:16:48 +0000 (05:16 +0000)]
- the filter doesn't have a cli_obd, so pass an import instead of a
conn in the llcd
- avoid double obd_transno_commit_cb()
- initialize the fo_sem for great justice
pschwan [Wed, 2 Jul 2003 23:26:55 +0000 (23:26 +0000)]
b=1501
- return -ENOTDIR from mds_open() if we're trying to open a
non-directory with O_DIRECTORY
- add tests/o_directory.c and make it in sanity test 38
zab [Wed, 2 Jul 2003 21:36:16 +0000 (21:36 +0000)]
- add the b1 qla2xxx 2.5 driver that we're using. b3 panics.
pschwan [Wed, 2 Jul 2003 18:40:59 +0000 (18:40 +0000)]
- add set_info calls, to inform the filter when the MDS connects
- make OST_LOG_CANCEL into a more generic OBD_LOG_CANCEL packet
- the filter can't use the normal obd methods unless we also give it an MDC
- lov_connect now initializes the export lists unconditionally
- the filter needs to register for the commit callback if there are
cookies to be cancelled, regardless of whether this is a failover OST
- hacked around in the lprocfs sewer until it built again
shaver [Wed, 2 Jul 2003 13:32:34 +0000 (13:32 +0000)]
Client-side portions of bug 974: push cache and dirty-page management down into
OSC, in preparation for pre-allocation grants. Also fixes 975 by providing
/proc/fs/lustre/llite/fs0/max_dirty_pages to limit client-side cache.
wangdi [Tue, 1 Jul 2003 11:04:09 +0000 (11:04 +0000)]
add tg3_netconsole.pc
wangdi [Tue, 1 Jul 2003 11:02:45 +0000 (11:02 +0000)]
add netconsole_sysrq
wangdi [Tue, 1 Jul 2003 11:01:41 +0000 (11:01 +0000)]
add tg3_netconsole.patch
rread [Tue, 1 Jul 2003 07:35:18 +0000 (07:35 +0000)]
b=1424
r=shaver,others
- cleanup import level and generation
- fixes several bugs:
b=1420 ost and client both timeout, client never reconnects
b=1440 very small change to call the lov set_active in recovery
b=1448 another small change to send DISCONNECT at a normal level
zab [Mon, 30 Jun 2003 23:24:37 +0000 (23:24 +0000)]
- move e1000_Poll() into a less broken place
- s/static// in front of stop_this_cpu so lkcd compiles under smp
+ mask these changes with tons of line numbering and diff -p noise
braam [Mon, 30 Jun 2003 21:30:46 +0000 (21:30 +0000)]
- add netdump patches to 2.5.73 kgdb series.
braam [Mon, 30 Jun 2003 21:26:49 +0000 (21:26 +0000)]
- add the kexec patch to 2.5.73
braam [Mon, 30 Jun 2003 20:52:50 +0000 (20:52 +0000)]
- move things forward to 2.5.73, include lkcd patches
shaver [Mon, 30 Jun 2003 20:11:27 +0000 (20:11 +0000)]
b=957: add pinger for recoverable clients, in order to promptly detect server
failure and join recovery. Currently DISABLED (#ifdef ENABLE_PINGER)
pending some discussion about deployment.
Also: use the --config (or file.xml) instead of 'none' in the mount line, so
that mounted filesystems have a little more context.
braam [Mon, 30 Jun 2003 17:55:07 +0000 (17:55 +0000)]
- major cleanup of the 2.5 kernel patch
- we indicate now with two flags when the last path name component
is being looked up, ending almost all confusion about the code
path.
- many small fixes.
- sanity finishes without failures, until at cleanup something hangs.
adilger [Fri, 27 Jun 2003 23:56:27 +0000 (23:56 +0000)]
Fix utime.c so it calls utime() on the right file and print valid errors.
Fix mkdirdeep.c so it is generally useful
- don't dump/trace by default
- exit on errors
- allow deeper than 100 tree (up to PATH_MAX).
- don't clobber stack generating stat pathnames.
braam [Fri, 27 Jun 2003 20:42:20 +0000 (20:42 +0000)]
- fix up super.c
- minor change to llite_lib.c to avoid a warning.
shaver [Fri, 27 Jun 2003 17:25:31 +0000 (17:25 +0000)]
Fix lock-match fallout from 1228: use the inode against which the intent lock
was granted, often not the one against which the lock was requested, when
matching l_data.
Also, match locks in ll_dir_readpage.
Remove some deadwood fops-functions.
meijia [Fri, 27 Jun 2003 04:23:47 +0000 (04:23 +0000)]
[liblustre]: fix an segfault when running iozone. don't more checking
before remove an inode.
shaver [Thu, 26 Jun 2003 20:20:01 +0000 (20:20 +0000)]
adding otree files (not yet used)
mmex [Thu, 26 Jun 2003 19:37:36 +0000 (19:37 +0000)]
Script to test ls speed on a single host using two mount points.
Not a great test but can give basic information on ls speed issues.
pschwan [Thu, 26 Jun 2003 17:49:35 +0000 (17:49 +0000)]
- Many unused variable warning fixes (several of which were not really a
problem, I think)
- list_splice_init() isn't defined in 2.4.18
- FSTYPE change for lov.sh
pschwan [Wed, 25 Jun 2003 22:38:32 +0000 (22:38 +0000)]
- fixes the maximum OST packet size, a lot of list management, and a
couple other bugs
- starts threads at OST and MDS startup
- able to complete a round of create/setattr/cancel if SENDNOW is set
- some corruption happens when I ask it to build up and cancel whole
pages; I need to run on multiple systems to start narrowing down where.
meijia [Wed, 25 Jun 2003 05:09:24 +0000 (05:09 +0000)]
[liblustre]: add back the script
adilger [Tue, 24 Jun 2003 23:34:21 +0000 (23:34 +0000)]
Fix for bug 1466, delete thread locking up machine during orphan recovery.
adilger [Tue, 24 Jun 2003 00:04:20 +0000 (00:04 +0000)]
Simple utime testing script, as I can't concentrate on anything more complex.
adilger [Mon, 23 Jun 2003 18:20:54 +0000 (18:20 +0000)]
Use a better variable name than "i" for the open file descriptor.
meijia [Mon, 23 Jun 2003 11:56:54 +0000 (11:56 +0000)]
[liblustre]:
- treat an closed inode as invalid.
- reap inode right after unlink/rmdir and when found stale ones.
braam [Sun, 22 Jun 2003 21:12:01 +0000 (21:12 +0000)]
add patches for 2.5.72. Minor Portals miscompilation still needs a
fix, but I cannot do that now I think.
braam [Sat, 21 Jun 2003 20:34:46 +0000 (20:34 +0000)]
many changes to the intent patch. Most of sanity runs now.
behlendo [Fri, 20 Jun 2003 23:57:46 +0000 (23:57 +0000)]
- Ensure stdout is flushed after each new line, this makes it possible to
redirect the resulting output to a file.
- %Lu doesn't format properly, changed to %lu
girish_c [Fri, 20 Jun 2003 22:43:18 +0000 (22:43 +0000)]
Fixes Connectathon bigfile(30 MB)test bug
pschwan [Fri, 20 Jun 2003 16:54:24 +0000 (16:54 +0000)]
b_orphan builds again; now for a some testing
eeb [Wed, 18 Jun 2003 22:08:38 +0000 (22:08 +0000)]
* 1439 fix
adilger [Wed, 18 Jun 2003 18:10:06 +0000 (18:10 +0000)]
Fix patches that were eaten by patch scripts.
girish_c [Wed, 18 Jun 2003 16:39:43 +0000 (16:39 +0000)]
NFS export patch for lustre
1) Patch is updated with review comments
2) All Connectathon tests passes except bigfile(30 MB)test
adilger [Wed, 18 Jun 2003 09:07:19 +0000 (09:07 +0000)]
Fix delete thread so that it doesn't sleep uninterruptibly and raise load avg.
Do not start delete thread if asyncdel mount option is not given.
pschwan [Tue, 17 Jun 2003 09:58:09 +0000 (09:58 +0000)]
haven't even tried to build, but getting closer:
- move the recovd_thread to ptlrpc, because it uses ptlrpc functions
- have one global llog_commit_master, in ptlrpc
meijia [Tue, 17 Jun 2003 03:12:53 +0000 (03:12 +0000)]
[liblustre]: add script to generate libraries