Whamcloud - gitweb
fs/lustre-release.git
20 years agofix formatting
jacob [Fri, 20 May 2005 23:26:29 +0000 (23:26 +0000)]
fix formatting

20 years agob=6359
jacob [Fri, 20 May 2005 23:25:36 +0000 (23:25 +0000)]
b=6359
r=adilger, phil

Severity:  : enhancement
Bugzilla   : 3262, 6359
Description: Attempts to reconnect to servers are now more aggressive.
Details    : This builds on the enhanced upcall-less recovery that was added
             in 1.4.2.  When trying to reconnect to servers,  clients will
             now try each server in the failover group every 10 seconds.  By
             default, clients would previously try one server every 25 seconds.

20 years agoBranch b1_4
adilger [Fri, 20 May 2005 23:07:42 +0000 (23:07 +0000)]
Branch b1_4
Do it right.

20 years agoBranch b1_4
adilger [Fri, 20 May 2005 23:06:35 +0000 (23:06 +0000)]
Branch b1_4
Increase fsx size for acceptance-small.sh to something reasonable.

20 years agoadd the now-required -DLUSTRE_UTILS=1 to the Lustre build as well
phil [Fri, 20 May 2005 22:39:47 +0000 (22:39 +0000)]
add the now-required -DLUSTRE_UTILS=1 to the Lustre build as well

20 years agoprint more useful things
jacob [Fri, 20 May 2005 20:47:48 +0000 (20:47 +0000)]
print more useful things

20 years agoBranch b1_4
adilger [Fri, 20 May 2005 20:16:27 +0000 (20:16 +0000)]
Branch b1_4
Print a useful error on failure.

20 years agoBranch b1_4
adilger [Fri, 20 May 2005 17:38:57 +0000 (17:38 +0000)]
Branch b1_4
Remove some extraneous logic in mdc_rename_pack().  We already check this
same thing in ll_i2uctxt() as called by ll{,u}_prepare_mdc_op_data() from
ll{,u}_rename_raw().  Verified with POSIX this has no effect.

20 years agoBranch b1_4
adilger [Fri, 20 May 2005 17:17:53 +0000 (17:17 +0000)]
Branch b1_4
Remove unused and duplicating-function file simple.c.  All of this has
been moved into lvfs*.c

20 years agob=4855
jacob [Thu, 19 May 2005 20:56:34 +0000 (20:56 +0000)]
b=4855
r=phil (original patch)

handle crc32_le being a module or disabled.

20 years agob=6349
jacob [Thu, 19 May 2005 17:35:56 +0000 (17:35 +0000)]
b=6349
r=adilger

print the nid instead of the export uuid (if available) when
ping-evicting clients

20 years agoBranch b1_4
adilger [Wed, 18 May 2005 17:11:14 +0000 (17:11 +0000)]
Branch b1_4
It was possible to get multiple mfd references during close and client
eviction, leading to either mds_close() or mds_mfd_close() referencing
a freed mfd.

Now we remove the mfd from the mfd_list and the handle hash under lock so
that once it starts on the road to destruction it is not possible to get a
new reference to it.  Also add comments on usage of mfd-related functions!
b=3819, b=4364, b=4397, b=6313
r=green

20 years agor=phil
jacob [Wed, 18 May 2005 00:51:54 +0000 (00:51 +0000)]
r=phil

stop lying just to be popular.

20 years agolsy reported that LL_CDEBUG_PAGE in osc_checksum_bulk wouldn't compile,
phil [Tue, 17 May 2005 16:25:58 +0000 (16:25 +0000)]
lsy reported that LL_CDEBUG_PAGE in osc_checksum_bulk wouldn't compile,
because the outer scope defines a variable "page_count", and LL_CDEBUG_PAGE
wants to call the VM function page_count().

I still can't figure out why it works fine for me, but I've stopped trying.
Changing the name of the variable is easier.

20 years agofix up for readline (the old one can not handle long input)
lwang [Tue, 17 May 2005 08:43:37 +0000 (08:43 +0000)]
fix up for readline (the old one can not handle long input)

20 years agob=5902
phil [Tue, 17 May 2005 04:04:09 +0000 (04:04 +0000)]
b=5902
r=adilger

Many improvements to the code that checksums I/O, among them:
- add a switch to enable or disable them at runtime
- check the pages while under llite's control, and the buffers while under
  the osc/ost's
- if the server and client see different checksums, try to figure out where
  it went wrong
- use the kernel crc32 routine

20 years agob=6200
phil [Tue, 17 May 2005 03:41:19 +0000 (03:41 +0000)]
b=6200
A bug in MDS/OSS recovery could cause the OSS to fail a (ctxt == NULL)
assertion.

There's little harm in just aborting MDS/OSS recovery and letting
it try again next time, so I removed the LASSERT and return an error instead.
We can revisit it when we put the knife in the llog code for real.

20 years agoBranch b1_4
adilger [Mon, 16 May 2005 22:57:48 +0000 (22:57 +0000)]
Branch b1_4
Disable fsync-under-recovery test, which is preventing remaining tests
from running.
b=5494
r=nathan

20 years agoBranch b1_4
adilger [Mon, 16 May 2005 18:22:42 +0000 (18:22 +0000)]
Branch b1_4
Decrease the ldlm_timeout if obd_timeout is dropped.  It doesn't really make
sense to have AST timeouts longer than the RPC timeouts, but is mostly only
noticable in our regression tests (recovery-small.sh test 10 in particular).
b=6172

20 years agoBranch b1_4
adilger [Mon, 16 May 2005 18:01:46 +0000 (18:01 +0000)]
Branch b1_4
Use a timeout instead of killing writemany - the OST failover may not have
happened yet.
r=nathan

20 years agoBranch b1_4
adilger [Mon, 16 May 2005 17:54:59 +0000 (17:54 +0000)]
Branch b1_4
Make sure that test 52 writemany will exit at some point.
Put writemany usage message at top of code so it is easily found.
r=nathan

20 years ago*** empty log message ***
larry [Sun, 15 May 2005 23:21:51 +0000 (23:21 +0000)]
*** empty log message ***

20 years agoBranch b1_4
adilger [Sat, 14 May 2005 10:32:37 +0000 (10:32 +0000)]
Branch b1_4
Add regression test for reading from a file with no objects (reads zeros,
stops at EOF).
b=6242, b=6243

20 years agoupdate to new errata
jacob [Fri, 13 May 2005 18:54:53 +0000 (18:54 +0000)]
update to new errata

20 years agoBranch b1_4
nathan [Fri, 13 May 2005 18:11:51 +0000 (18:11 +0000)]
Branch b1_4
b=6301
r=adilger
Prevent damage due to racing export disconnects

20 years agoland b_release_1_4_2:
jacob [Fri, 13 May 2005 18:04:05 +0000 (18:04 +0000)]
land b_release_1_4_2:

 - qsnet patches for sles 2.6 and rhel 2.4

20 years agoBranch b1_4
adilger [Thu, 12 May 2005 23:13:24 +0000 (23:13 +0000)]
Branch b1_4
Move all pagecache vs. directio handling into filter_clear_page_cache()
and add comments about what is going on here and why.

Use the fs-specific ->invalidatepage called from truncate_complete_page()
instead of calling block_invalidatepage() directly.

Wait on page for the 2.6 code, and use 2.6 functions and have compat
macros for 2.4 instead.
b=4892
r=green

20 years agoBranch b1_4
adilger [Thu, 12 May 2005 21:28:49 +0000 (21:28 +0000)]
Branch b1_4
Update ChangeLog to new format.

20 years agoBranch b1_4
adilger [Thu, 12 May 2005 20:42:39 +0000 (20:42 +0000)]
Branch b1_4
Add obd_name to message.
b=6301

20 years agoBranch b1_4
adilger [Wed, 11 May 2005 22:04:03 +0000 (22:04 +0000)]
Branch b1_4
Ensure the PING_INTERVAL is at least 1s, otherwise obd_timeout < 4 can
cause severe heartburn.

20 years agoBranch: b1_4
green [Wed, 11 May 2005 21:16:30 +0000 (21:16 +0000)]
Branch: b1_4
b=6243
r=adilger

Properly return 0 for reads that are totally after EOF.

20 years agoBranch b1_4
adilger [Wed, 11 May 2005 19:23:49 +0000 (19:23 +0000)]
Branch b1_4
Revert previous commit, it was confusingly correct.
b=6289

20 years agoBranch b1_4
adilger [Wed, 11 May 2005 18:28:46 +0000 (18:28 +0000)]
Branch b1_4
Pass OBD_BRW_WRITE to filter_alloc_iobuf() from preprw_write().  This
parameter is actually unused at this time, so no real effect.
b=6289
r=fergal (original change)

20 years agoBranch b1_4
adilger [Wed, 11 May 2005 18:03:07 +0000 (18:03 +0000)]
Branch b1_4
Update build version to 1.4.2.2bgl

20 years agoBranch b1_4
adilger [Wed, 11 May 2005 17:51:37 +0000 (17:51 +0000)]
Branch b1_4
Fix typo in bug number.

20 years agoBranch b1_4
nathan [Wed, 11 May 2005 00:07:36 +0000 (00:07 +0000)]
Branch b1_4
b=5493
add asserts to help track this down

20 years agoBranch b1_4
nathan [Tue, 10 May 2005 21:28:44 +0000 (21:28 +0000)]
Branch b1_4
b=5949

20 years agoBranch b1_4
adilger [Tue, 10 May 2005 20:32:23 +0000 (20:32 +0000)]
Branch b1_4
Increase the LDLM AST timeout.  With the lock reordering in bug 5492 et.al.
we now hold ns_lock over vmtruncate.
b=6172
r=phil

20 years agoBranch b1_4
nathan [Tue, 10 May 2005 15:25:39 +0000 (15:25 +0000)]
Branch b1_4
b=6223
r=adilger
Fix lfs find to handle arbitrary number of targets

20 years agoBranch b1_4
adilger [Mon, 9 May 2005 06:05:24 +0000 (06:05 +0000)]
Branch b1_4
Fix compile warning.

20 years ago2.4.29 does not need smp_num_siblings export patch as this is already exported
green [Sun, 8 May 2005 20:49:11 +0000 (20:49 +0000)]
2.4.29 does not need smp_num_siblings export patch as this is already exported

20 years agoMove CONFIG_KALLSYMS question from ISDN menu to kernel hacking menu
green [Sun, 8 May 2005 20:33:40 +0000 (20:33 +0000)]
Move CONFIG_KALLSYMS question from ISDN menu to kernel hacking menu
in 2.4.29 patch

20 years agokallsyms patch for 2.4.29
green [Sun, 8 May 2005 20:22:48 +0000 (20:22 +0000)]
kallsyms patch for 2.4.29
Separated uml & non-uml 2.4.29 series

20 years agob=5907
green [Fri, 6 May 2005 20:22:39 +0000 (20:22 +0000)]
b=5907
r=adilger

Testcase for the bug.

20 years agofix subsystem type
adilger [Fri, 6 May 2005 14:55:49 +0000 (14:55 +0000)]
fix subsystem type

20 years agoBranch: b1_4
adilger [Thu, 5 May 2005 22:28:36 +0000 (22:28 +0000)]
Branch: b1_4
Add bug reference for workaround.
b=6155

20 years agoBranch b1_4
nathan [Thu, 5 May 2005 21:52:47 +0000 (21:52 +0000)]
Branch b1_4
b=5949
r=adilger
Various fixes to make --failover under load safer

20 years agob=6243
green [Thu, 5 May 2005 20:54:22 +0000 (20:54 +0000)]
b=6243
Export clear_user_skas as well (uml-only again)

20 years ago- I committed the ext3-patch-fuzz patch, but forgot to add it to the series
phil [Thu, 5 May 2005 20:50:41 +0000 (20:50 +0000)]
- I committed the ext3-patch-fuzz patch, but forgot to add it to the series
- add the recent clear_user export patch to my uml series

20 years agoBranch: b1_4
adilger [Thu, 5 May 2005 20:16:49 +0000 (20:16 +0000)]
Branch: b1_4
Remove unused function.  class_find_client_obd_notype() does the same and
is more generic.
b=6242

20 years agoBranch: b1_4
adilger [Thu, 5 May 2005 19:41:27 +0000 (19:41 +0000)]
Branch: b1_4
Reorder ll_async_page slightly so the ints are together (slight savings on
64-bit arches).

20 years agoBranch: b1_4
adilger [Thu, 5 May 2005 18:51:17 +0000 (18:51 +0000)]
Branch: b1_4
Limit the number of stripes in a file to the maximum that will fit into
a 4kB EA.  This is needed until we can store larger EAs in ext3 (bug 4424).
b=6093

20 years agoBranch: b1_4
adilger [Thu, 5 May 2005 18:36:35 +0000 (18:36 +0000)]
Branch: b1_4
Update build version to 1.4.1.14.

20 years agoBranch: b1_4
adilger [Thu, 5 May 2005 18:25:02 +0000 (18:25 +0000)]
Branch: b1_4
Make it easier to see dev_read_only set on device.

20 years agoBranch: b1_4
adilger [Thu, 5 May 2005 18:12:01 +0000 (18:12 +0000)]
Branch: b1_4
Placeholder from b_cray so it doesn't get reused.

20 years agob=5902
phil [Thu, 5 May 2005 18:07:11 +0000 (18:07 +0000)]
b=5902
Remove the tcp-zero-copy patch from the 2.4.21-suse-171 series.  We are
virtually certain that the corruption is caused in the TCP stack, and we are
completely unable to reproduce it with zero-copy disabled.  Because this series
is of almost zero interest, just remove the patch and move on.

20 years agoBranch: b1_4
adilger [Thu, 5 May 2005 18:01:27 +0000 (18:01 +0000)]
Branch: b1_4
Remove hopelessly obsolete test program.

20 years agoturn quadrics drivers on
jacob [Thu, 5 May 2005 17:47:32 +0000 (17:47 +0000)]
turn quadrics drivers on

20 years agoBranch: b1_4
adilger [Thu, 5 May 2005 17:44:17 +0000 (17:44 +0000)]
Branch: b1_4
Using 'lctl recover' should mark a device active if it was previously
deactivated.
Add some debugging to ptlrpc_pinger_main() to help analyse bug 5933.
b=5933
r=nathan

20 years agob=6233
jacob [Thu, 5 May 2005 17:41:08 +0000 (17:41 +0000)]
b=6233
originally from oleg, r=jacob

Build quota_context as part of ptlrpc, so it can be shared by the mds
and obdfilter without being linked into both, which sometimes causes
build failures on lester.

20 years agoBranch: b1_4
adilger [Thu, 5 May 2005 16:44:59 +0000 (16:44 +0000)]
Branch: b1_4
Bump the maximum number of devices as a short-term solution.  The long-term
solution to allocate these structs is under implementation but not ready
for a 1.4.2 release.
b=6242

20 years agob=6243
green [Thu, 5 May 2005 05:55:08 +0000 (05:55 +0000)]
b=6243
r=adilger

Correctly return zero-filled regions when doing reads of files with no objects.
Export clear_user_tt symbol in UML as it is now needed by lustre.

20 years agoadd qsnet patch for rhel-2.4.
jacob [Wed, 4 May 2005 23:04:04 +0000 (23:04 +0000)]
add qsnet patch for rhel-2.4.

20 years agoBranch: b1_4
adilger [Wed, 4 May 2005 21:28:01 +0000 (21:28 +0000)]
Branch: b1_4
Commit the routed.sh script to CVS for reference, remove old configs.

20 years agoupdate to latest errata; patches apply and no config changes needed
jacob [Wed, 4 May 2005 19:49:57 +0000 (19:49 +0000)]
update to latest errata; patches apply and no config changes needed

20 years agoBranch: b1_4
adilger [Wed, 4 May 2005 08:30:17 +0000 (08:30 +0000)]
Branch: b1_4
Don't clobber last_rcvd upon mds_client_add() failure.  This was caused by
bad error handling in mds_connect() not setting med_mcd = NULL after freeing
it, and then zeroing the "client data" at med_off = 0.  Includes test.
This commit adds some safety checks to {mds,filter}_finish_transno()
and reconciles the differences between the mds and obdfilter last_rcvd
functions.
b=6086
r=nathan

20 years agob=3031
jacob [Tue, 3 May 2005 22:16:07 +0000 (22:16 +0000)]
b=3031

 - this is what i should have done originally

20 years agoBranch b1_4
nathan [Tue, 3 May 2005 22:13:12 +0000 (22:13 +0000)]
Branch b1_4
b=6223
Fix this related one too.

20 years agob=6223
nathan [Tue, 3 May 2005 21:54:49 +0000 (21:54 +0000)]
b=6223
This should fix the crash (need to use vfree, not kfree).  But it won't
allow lfs getstripe to work for 420 OSTs -- OBD_MAX_IOCTL_BUFFER of 8k allows a
max of around 180 OBDs.

20 years agob=3031
jacob [Tue, 3 May 2005 21:46:41 +0000 (21:46 +0000)]
b=3031

 - remove .flags files on make clean

20 years agob=3418
jacob [Tue, 3 May 2005 21:25:15 +0000 (21:25 +0000)]
b=3418

 - add dependencies on libs

20 years agovanilla-2.4.29 support.
green [Tue, 3 May 2005 21:02:00 +0000 (21:02 +0000)]
vanilla-2.4.29 support.

20 years agoBranch: b1_4
adilger [Fri, 29 Apr 2005 22:28:50 +0000 (22:28 +0000)]
Branch: b1_4
Add a similar fix to the mds_connect() to handle NULL connect data, in
case we ever use echo_client against the MDS.

Print out the request for bad connect requests so we can get the source NID.
This was happening at NCSA for a long time and no way to find which client
was trying to conenct.

20 years agor=phil
nic [Fri, 29 Apr 2005 22:00:13 +0000 (22:00 +0000)]
r=phil

only set data elements with a non-NULL data. this fixes echo_client connects

20 years agoBranch: b1_4
adilger [Fri, 29 Apr 2005 20:24:51 +0000 (20:24 +0000)]
Branch: b1_4
- don't reference lr_lvb_data until after we hold lr_lvb_sem
b=6170

20 years agoBranch: b1_4
adilger [Thu, 28 Apr 2005 23:25:40 +0000 (23:25 +0000)]
Branch: b1_4
Fix compile warnings on 64-bit.

20 years agoBranch b1_4
nathan [Thu, 28 Apr 2005 22:02:49 +0000 (22:02 +0000)]
Branch b1_4
b=6151
utilities all want /dev/obd

20 years agoBranch: b1_4
adilger [Thu, 28 Apr 2005 20:00:21 +0000 (20:00 +0000)]
Branch: b1_4
Placeholders for constants already defined in b_cray.

20 years agoBranch: b1_4
adilger [Thu, 28 Apr 2005 19:50:10 +0000 (19:50 +0000)]
Branch: b1_4
On Altix PAGE_SIZE is #defined as getpagesize() in userspace and CPP
can't deal with that.  Instead, assume we are limited by the RPC size
and not the number of pages (which is currently true for any arch)
and don't use PAGE_SIZE in any CPP conditionals.

20 years agoBranch: b1_4
adilger [Thu, 28 Apr 2005 18:36:04 +0000 (18:36 +0000)]
Branch: b1_4
Fix quota test valid compile warnings.

20 years agoBranch: b1_4
adilger [Thu, 28 Apr 2005 00:22:59 +0000 (00:22 +0000)]
Branch: b1_4
Free lsm if there was an error allocating lsr.

20 years agoBranch: b1_4
adilger [Thu, 28 Apr 2005 00:15:20 +0000 (00:15 +0000)]
Branch: b1_4
Define the quota constants for

20 years agoBranch: b1_4
adilger [Thu, 28 Apr 2005 00:03:30 +0000 (00:03 +0000)]
Branch: b1_4
Don't leave lock on resource list in ldlm_handle_enqueue() error path, as
that will LBUG when lock is destroyed.  Regression test for same.
b=6149, b=6184
r=phil (original patch)

20 years agoBranch: b1_4
adilger [Wed, 27 Apr 2005 21:19:05 +0000 (21:19 +0000)]
Branch: b1_4
Restore lov.sh testing to 5 OSTs.
b=4554

20 years agoChangelog update for commits on 6203, 5007
green [Wed, 27 Apr 2005 20:01:28 +0000 (20:01 +0000)]
Changelog update for commits on 6203, 5007

20 years agoBranch: b1_4
adilger [Wed, 27 Apr 2005 19:04:13 +0000 (19:04 +0000)]
Branch: b1_4
- reserve enough journal credits in fsfilt_start_log for setattr
b=4554

20 years agoBranch: b1_4
adilger [Wed, 27 Apr 2005 18:11:06 +0000 (18:11 +0000)]
Branch: b1_4
Improve lconf performance for large configurations by adding a cache
for lookups instead of searching the DB for each entry.
This sped up a 420-OST config --write-conf from 10h to 6 minutes...
b=6204
r=fergal (original patch)

20 years agob=6203,5654
green [Wed, 27 Apr 2005 17:35:37 +0000 (17:35 +0000)]
b=6203,5654
r=adilger
Release i_sem and i_alloc_sem prior to getting DLM lock in setattr_raw to
avoid lock inversion issues.

20 years agob=5907
green [Wed, 27 Apr 2005 08:22:11 +0000 (08:22 +0000)]
b=5907
r=adilger
Do not allow to remove mountpoints on lustre

20 years agobump version
jacob [Wed, 27 Apr 2005 04:27:28 +0000 (04:27 +0000)]
bump version

20 years agob=5782
jacob [Tue, 26 Apr 2005 23:53:29 +0000 (23:53 +0000)]
b=5782

 - added --disable-server and --disable-client configure options (5782)

20 years agobe more quiet
jacob [Tue, 26 Apr 2005 23:47:20 +0000 (23:47 +0000)]
be more quiet

20 years agob=6205
jacob [Tue, 26 Apr 2005 22:03:51 +0000 (22:03 +0000)]
b=6205
r=adilger

 - move the lustre kernel modules to their own package
 - lbuild now does 'make rpms' to build lustre modules, more or less
 - pass --with-linux to lbuild to build only lustre rpms against an
   existing kernel source tree
 - drop the lustre-ldap rpm

20 years agob=5907
green [Tue, 26 Apr 2005 20:49:30 +0000 (20:49 +0000)]
b=5907
r=adilger
Do not unhash mountpoint dentries and silence error message

20 years agoBranch: b1_4
adilger [Tue, 26 Apr 2005 17:38:08 +0000 (17:38 +0000)]
Branch: b1_4
Quiet compile warning about unused lustre_delete_dquot().

20 years agoUpdate changelog with 6196 bugfix info
green [Mon, 25 Apr 2005 16:30:47 +0000 (16:30 +0000)]
Update changelog with 6196 bugfix info

20 years agob=6196
green [Mon, 25 Apr 2005 08:40:36 +0000 (08:40 +0000)]
b=6196
r=adilger
For inodes with no ost objects call vmtruncate from inode_setattr() to update
inode->i_size, becasue size is only stored on MDS.

20 years agoBranch: b1_4
adilger [Sat, 23 Apr 2005 08:37:55 +0000 (08:37 +0000)]
Branch: b1_4
Put test58 files into a subdirectory.

20 years agoBranch: b1_4
adilger [Sat, 23 Apr 2005 08:08:06 +0000 (08:08 +0000)]
Branch: b1_4
Reduce number of files created because small test setups don't have so many
files and it doesn't really improve the test.  Also, checkstat will fail
if there are too many files on the command line.

20 years agoBranch: b1_4
adilger [Sat, 23 Apr 2005 00:15:10 +0000 (00:15 +0000)]
Branch: b1_4
Proper fix for hokey error message quieting hack, per Nathan's feedback.