Whamcloud - gitweb
fs/lustre-release.git
9 years agob=23289 disable osd iam debug by default
Jian Yu [Tue, 7 Sep 2010 10:01:05 +0000 (14:01 +0400)]
b=23289 disable osd iam debug by default

Disable debug counter in iam.

o=zhen.liang
i=alex.zhuravlev
i=di.wang

9 years agob=23289 disable RS_DEBUG by default
Jian Yu [Tue, 7 Sep 2010 10:08:25 +0000 (14:08 +0400)]
b=23289 disable RS_DEBUG by default

Disable RS_DEBUG by default.

o=zhen.liang
i=andreas.dilger
i=Mikhail.Pershin

9 years agob=23316 BUG: soft lockup - CPU#2 stuck for 10s! [ll_cfg_requeue:2851]
Dmitry Zogin [Tue, 7 Sep 2010 14:05:16 +0000 (18:05 +0400)]
b=23316 BUG: soft lockup - CPU#2 stuck for 10s! [ll_cfg_requeue:2851]

 test-framework.sh cleanup patch

9 years agoMerge branch 'master' of git.lustre.org:prime/lustre
Mikhail Pershin [Tue, 7 Sep 2010 20:08:00 +0000 (00:08 +0400)]
Merge branch 'master' of git.lustre.org:prime/lustre

9 years agob=23728 Doxygen style comment - sptlrpc api.
Eric Mei [Tue, 7 Sep 2010 13:40:53 +0000 (17:40 +0400)]
b=23728 Doxygen style comment - sptlrpc api.

9 years agoUpdated for build 03. 2.0.51.0 v2_0_51_0
unknown [Tue, 7 Sep 2010 14:36:41 +0000 (07:36 -0700)]
Updated for build 03.

9 years agobz=23673 add "updates" dir to lustre-modules location.
Vitaly Fertman [Tue, 7 Sep 2010 12:46:01 +0000 (16:46 +0400)]
bz=23673 add "updates" dir to lustre-modules location.

o=grev
i=vitaly

9 years agob=23402 mmp_init() fix
Elena Gryaznova [Mon, 30 Aug 2010 10:02:56 +0000 (14:02 +0400)]
b=23402 mmp_init() fix

i=jian.yu

9 years agob=17748 fix sanityn test 34 to support multiple OSS nodes
Jian Yu [Sat, 4 Sep 2010 14:29:50 +0000 (18:29 +0400)]
b=17748 fix sanityn test 34 to support multiple OSS nodes

i=vitaly.fertman
i=grev

9 years agob=23588 conf-sanity.sh:test_50g - wait for new OST addition propagation to a client
Vladimir Saveliev [Sat, 28 Aug 2010 02:39:55 +0000 (06:39 +0400)]
b=23588 conf-sanity.sh:test_50g - wait for new OST addition propagation to a client

+ fix for memory leak in osc_enqueue_base()

i=nathan.rutman

9 years agob=23014 sgpdd-survey acc-sm integration
Elena Gryaznova [Mon, 6 Sep 2010 10:56:23 +0000 (14:56 +0400)]
b=23014 sgpdd-survey acc-sm integration

i=Andreas.Dilger
i=Manoj.P.Joseph

9 years agob=23515 recovery_*_scale tests need more than 2 clients
Elena Gryaznova [Mon, 6 Sep 2010 11:07:42 +0000 (15:07 +0400)]
b=23515 recovery_*_scale tests need more than 2 clients

i=Mikhail.Pershin

9 years agob=23363 obdfilter-survey: no echo client and ost_name not set
Elena Gryaznova [Mon, 6 Sep 2010 11:58:31 +0000 (15:58 +0400)]
b=23363 obdfilter-survey: no echo client and ost_name not set

i=Andrew.Perepechko

libecho fix: use NETTYPE instead of tcp only

9 years agob=11678 add check to void unnecessary warning
Landen [Mon, 6 Sep 2010 14:31:24 +0000 (18:31 +0400)]
b=11678 add check to void unnecessary warning

i=johann
i=andrew.perepechko

9 years agob21012 conf-sanity test_52: @@@@@@ FAIL: Unable to tar files
Rahul Deshmukh [Mon, 6 Sep 2010 09:22:10 +0000 (13:22 +0400)]
b21012 conf-sanity test_52: @@@@@@ FAIL: Unable to tar files

b21012 conf-sanity test_52: @@@@@@ FAIL: Unable to tar files

Fixed timing problem of test_52
i=grev
i=vladimir.saveliev

9 years agob=23663 fix buffer overflow in liblustre
Brian J. Murrell [Tue, 31 Aug 2010 23:47:27 +0000 (03:47 +0400)]
b=23663 fix buffer overflow in liblustre

i=dmitry
i=andrew

The space allocated to ll_instance to store the "%p" notation for a
pointer was failing to allocate enough space for the leading "0x" in
addition to the space for the actual address bytes.

9 years agob=22978 better conformation to coding standards
Brian J. Murrell [Tue, 31 Aug 2010 19:13:00 +0000 (23:13 +0400)]
b=22978 better conformation to coding standards

Per adilger:
- the preferred Lustre coding style is to not use assignments inside
  conditionals
- Lustre coding style recommends against using negation on non-boolean
  values
- while fixing this code, we may as well print the flag in octal, which
is how it is normally defined.

i=whitebear
i=panda

9 years agob=19700 remove obsolete IB LNDs from HEAD
Maxim Patlasov [Mon, 30 Aug 2010 11:12:56 +0000 (15:12 +0400)]
b=19700 remove obsolete IB LNDs from HEAD

9 years agoRevert "b20288 fix a deadlock in kiblnd_check_conns i=isaac i=maxim"
Mikhail Pershin [Thu, 2 Sep 2010 13:36:52 +0000 (17:36 +0400)]
Revert "b20288 fix a deadlock in kiblnd_check_conns i=isaac i=maxim"

This reverts commit ac54ab5516bc3c1fb995f8189234123cbbac61e7.

9 years agob=23636 extend exp_lock protected area in target_process_req_flags
Mikhail Pershin [Tue, 31 Aug 2010 12:21:19 +0000 (16:21 +0400)]
b=23636 extend exp_lock protected area in target_process_req_flags

Issue was introduced by bug 17660, previously this was protected by obd lock

i=zam
i=vitaly

9 years agoRevert "b=17471 change conf_param syntax to match set_param"
Mikhail Pershin [Tue, 31 Aug 2010 11:17:20 +0000 (15:17 +0400)]
Revert "b=17471 change conf_param syntax to match set_param"

This reverts commit 9fe6b20f03f57eb3a224dd203f74fb8fd8411c36.

9 years agob=23234 use a regular expression to parse ip_addr from ret_str in lc_net
Elena Gryaznova [Mon, 30 Aug 2010 17:12:50 +0000 (21:12 +0400)]
b=23234 use a regular expression to parse ip_addr from ret_str in lc_net

a=Chris Horn (CRAY)
i=Elena Gryaznova
i=Andrew Perepechko

9 years agob=16909 CERROR to LCONSOLE_WARN for lnet_send errors
Brian Behlendorf [Mon, 30 Aug 2010 20:20:29 +0000 (00:20 +0400)]
b=16909 CERROR to LCONSOLE_WARN for lnet_send errors

These errors are not uncommon when restarting services and there is
no need to include the additional lustre debug noise in them.

  LustreError: 8831:0:(lib-move.c:1427:lnet_send()) No route to
  12345-192.168.65.112@o2ib6 via 172.16.2.201@tcp (all routers down)

Changed to:

  Lustre: No route to 12345-192.168.65.112@o2ib6 via 172.16.2.201@tcp
  (all routers down)

9 years agoRevert "b=22423 Reconnects are not throttled"
Mikhail Pershin [Mon, 30 Aug 2010 21:17:09 +0000 (01:17 +0400)]
Revert "b=22423 Reconnects are not throttled"

This reverts commit 52a7f6e56654faa1fa2ed6d9a745cc018de6ff86.

9 years agob=23500 Send quota broadcast requests in parallel
Landen [Mon, 23 Aug 2010 14:48:13 +0000 (18:48 +0400)]
b=23500 Send quota broadcast requests in parallel

Quota broadcast requests are sent sequentially, while it would be
more efficient to send all the requests simultaneously.

i=johann
i=panda

9 years agob=22476 shrink dlm slab on servers even if __GFP_FS is not set.
Vitaly Fertman [Thu, 26 Aug 2010 13:31:44 +0000 (17:31 +0400)]
b=22476 shrink dlm slab on servers even if __GFP_FS is not set.

o=johann
i=vitaly
i=green

9 years agob=22476 grant_plan upper hard limit.
Vitaly Fertman [Thu, 26 Aug 2010 13:31:45 +0000 (17:31 +0400)]
b=22476 grant_plan upper hard limit.

i=johann
i=dmitry

9 years agob=22415 cut off the "*" mark at the end of digital string for quota usage when out...
Fan Yong [Sun, 29 Aug 2010 16:28:44 +0000 (20:28 +0400)]
b=22415 cut off the "*" mark at the end of digital string for quota usage when out of quota

cut off the "*" mark at the end of digital string for quota usage when out of quota.

i=grev
i=landen

9 years agob=21871 Just cleanup export nid_stat while exp_refcount == 0.
yangsheng [Fri, 20 Aug 2010 00:55:43 +0000 (04:55 +0400)]
b=21871 Just cleanup export nid_stat while exp_refcount == 0.

i=nathan
i=johann

9 years agob=22430 parse names with hiphen properly
Maxim Patlasov [Sun, 22 Aug 2010 14:42:10 +0000 (18:42 +0400)]
b=22430 parse names with hiphen properly

i=liang
The names like sfire13-ib@o2ib confuse lctl ping because it assumes
"[PID-]NID[@network]" format. The patch fixes the problem by trying
"PID-NID" format first and falling back to "NID" format if the attempt
failed.

9 years agob=22884 File to OST allocation on an empty file system varies greatly.
Dmitry Zogin [Thu, 19 Aug 2010 14:49:57 +0000 (18:49 +0400)]
b=22884 File to OST allocation on an empty file system varies greatly.

 osc_precreate() should return 0, if there are enough objects left.

 i=andreas.dilger
 i=andrew.perepechko

9 years agob=22407 substitute %i with %d
Maxim Patlasov [Mon, 23 Aug 2010 09:52:06 +0000 (13:52 +0400)]
b=22407 substitute %i with %d

i=alex
Common in-kernel code should use %d intstead of %i in print format
strings because solaris native prims (like cmn_err()) don't support
%i.

9 years agob=13585 Remove i_filterdata patches.
yangsheng [Fri, 20 Aug 2010 00:37:50 +0000 (04:37 +0400)]
b=13585 Remove i_filterdata patches.

i=adilger

9 years agob=23368 MMP: Watchdog detected LOCKUP on CPU 1 RIP: __list_add+0x6/0x68
Rahul Deshmukh [Thu, 19 Aug 2010 17:02:56 +0000 (21:02 +0400)]
b=23368 MMP: Watchdog detected LOCKUP on CPU 1 RIP: __list_add+0x6/0x68

b=23368  MMP: Watchdog detected LOCKUP on CPU 1 RIP: __list_add+0x6/0x68

i=sheng.yang

Ported the ext4 patches for master (rhel5) disable delalloc,disable mb-cache
and remove extents warning.

9 years agob=22967 s/AC_LANG_CONFTEST/LB_LINUX_CONFTEST/
Brian J. Murrell [Thu, 19 Aug 2010 21:39:53 +0000 (01:39 +0400)]
b=22967 s/AC_LANG_CONFTEST/LB_LINUX_CONFTEST/

There was a macro update in more recent releases of autoconf that requires
us to use AC_LANG_CONFTEST instead of the LB_LINUX_CONFTEST that we currently
use.  The problem using LB_LINUX_CONFTEST causes is that as configure is
determining capabilities and setting capability flags, these are not being
used when compiling further conftest.c programs.

So for example if a macro determines if foo is available and then sets FOO
if it is, and then a test in a subsequent macro tries to use FOO, it will
find it undefined.

i=mjmac
i=adilger

9 years agob=22846 debuginfo packages for lustre and ldiskfs
Brian J. Murrell [Thu, 19 Aug 2010 21:34:13 +0000 (01:34 +0400)]
b=22846 debuginfo packages for lustre and ldiskfs

With our current .spec files for lustre and ldiskfs, debuginfo packages
are not produced.  It seems that while other distros do this implicitly,
SLES needs to be told to do so explicitly.

i=mjmac
i=wangyb

9 years agob=21452 support for weak-modules
Brian J. Murrell [Thu, 19 Aug 2010 20:56:30 +0000 (00:56 +0400)]
b=21452 support for weak-modules

Add support to our RPM SPEC for the weak-modules script.

This requires that we install our modules under
/lib/modules/$(uname -r)/updates/kernel.  I think this is the correct
location for us in any case given that we are a kernel "addon" package.
Relax the kernel Requires: to work better with weak-modules.
Use the external dependency generator as the internal one cannot deal
with kernel modules.

i=mjmac
i=wangyb

9 years agob=22632 also build mptlinux on SLES11
Johann Lombardi [Fri, 13 Aug 2010 14:00:55 +0000 (18:00 +0400)]
b=22632 also build mptlinux on SLES11

i=johann

Add SLES11 to the list of platforms we build mptlinux on.
FWIW, RDAC fails to build on SLES11 so it has not been added here.

9 years agob=22632 update mptsas driver to PH16-4.18.20.04
Brian J. Murrell [Fri, 13 Aug 2010 14:00:54 +0000 (18:00 +0400)]
b=22632 update mptsas driver to PH16-4.18.20.04

i=wangyb
i=yangsheng

9 years agob=23035 split changelog time record into ctime,mtime,atime
Nathan Rutman [Fri, 13 Aug 2010 17:36:41 +0000 (21:36 +0400)]
b=23035 split changelog time record into ctime,mtime,atime

9 years agob=23298 backport some llapi_changelog/copytool fixes from HSM branch
Nathan Rutman [Fri, 13 Aug 2010 17:06:24 +0000 (21:06 +0400)]
b=23298 backport some llapi_changelog/copytool fixes from HSM branch

455f63ad2b92cea37793a3bce5d412b30ea144c5 zero kuc private memory
d1949858dabc7de7738e2179cfc8e4b3d6fa799f make llapi_copytool_recv thread safe
9582d218eb6abef52814b364e16e5b3b24c29658 make llapi_changelog_recv thread safe

9 years agob=20407 TF: "HARD" failovers with multiple targets per server
Elena Gryaznova [Tue, 24 Aug 2010 19:52:29 +0000 (23:52 +0400)]
b=20407 TF: "HARD" failovers with multiple targets per server

i=Brian.Murrell
i=Li.Wei

9 years agob=23531 fix wrong variable use cases in "lustre_acl_xattr_merge2ext()"
Fan Yong [Sat, 21 Aug 2010 14:08:14 +0000 (18:08 +0400)]
b=23531 fix wrong variable use cases in "lustre_acl_xattr_merge2ext()"

fix wrong variable use cases in "lustre_acl_xattr_merge2ext()"

i=Mikhail.Pershin
i=di.wang

9 years agob=23595 return registration errors
Nathan Rutman [Mon, 23 Aug 2010 17:57:28 +0000 (21:57 +0400)]
b=23595 return registration errors

9 years agob=16909 add CNETERR() cleanup console messages
Cliff White [Fri, 20 Aug 2010 21:08:24 +0000 (01:08 +0400)]
b=16909 add CNETERR() cleanup console messages

i=johann.lombardi
i=maxim.patlasov

Change LNET GET/PUT console errors to CNETERR, add CNETERR macro

9 years agob=23634 sanity-sec RUNAS fix
Mikhail Pershin [Tue, 24 Aug 2010 19:13:57 +0000 (23:13 +0400)]
b=23634 sanity-sec RUNAS fix

fix regression from bug 14242 on master

i=tappro
i=vitaly

9 years agob=17485 Don't reset the lov_objid to last_id sent by OST
Vladimir Saveliev [Fri, 20 Aug 2010 15:03:09 +0000 (19:03 +0400)]
b=17485 Don't reset the lov_objid to last_id sent by OST

https://bugzilla.lustre.org/attachment.cgi?id=30559 for head

i=Hongchao.zhang
i=andrew.perepechko

9 years agob=17485 testcase to verify MDS not to allocate same objid to different files
Vladimir Saveliev [Fri, 20 Aug 2010 15:02:36 +0000 (19:02 +0400)]
b=17485 testcase to verify MDS not to allocate same objid to different files

i=johann.lombardi
i=grev

9 years agob=17760 Separate locking for obd bitfield and recovery
Mikhail Pershin [Sat, 21 Aug 2010 10:01:48 +0000 (14:01 +0400)]
b=17760 Separate locking for obd bitfield and recovery

i=adilger
i=zam

9 years agob=23267 Make lustre_idl.h include needed definitions from libcfs.
Ricardo M. Correia [Tue, 17 Aug 2010 17:09:54 +0000 (21:09 +0400)]
b=23267 Make lustre_idl.h include needed definitions from libcfs.

i=nathan.rutman
i=mikhail.pershin

It seems that lustre_idl.h uses symbols such as LASSERT, CLASSERT, LPUX64, etc,
but it doesn't include libcfs.h. This was causing a compilation failure in
b_hd_kdmu. The fix is to include libcfs.h.

9 years agob=22934 fix writeconf testing: require modules for tunefs, don't use SINGLEMDS
Nathan Rutman [Thu, 19 Aug 2010 21:40:05 +0000 (01:40 +0400)]
b=22934 fix writeconf testing: require modules for tunefs, don't use SINGLEMDS

i=johann

9 years agob=17682 Improve debug capability for production systems
Dmitry Zogin [Fri, 20 Aug 2010 14:50:07 +0000 (18:50 +0400)]
b=17682 Improve debug capability for production systems

 Add a warning, when debug/subsystem_debug masks are set by
 numerical values, rather than by symbolic names.

9 years agob=17682 Improve debug capability for production systems
Dmitry Zogin [Fri, 20 Aug 2010 14:44:44 +0000 (18:44 +0400)]
b=17682 Improve debug capability for production systems

 Improve performance of D_RPCTRACE debugging.

 i=andreas.dilger
 i=andrew.perepechko

9 years agob=22610 a truncate_complete_page fix
Mikhail Pershin [Fri, 20 Aug 2010 19:25:26 +0000 (23:25 +0400)]
b=22610 a truncate_complete_page fix

truncate_complete_page implementation for the patchless
client could arbitrarily unset PG_Uptodate flag for a
page being kicked from the page cache, an uptodate check
right after a readpage call in filemap_fault could fail
because of that as though the page read had been unsuccessful.

i=Oleg Drokin
i=Johann Lombardi

9 years agob=23595 fix conf-sanity 57 for remote ost
Nathan Rutman [Fri, 20 Aug 2010 19:00:55 +0000 (23:00 +0400)]
b=23595 fix conf-sanity 57 for remote ost

9 years agob=14242 test_6g fails when b_release_1_6_4 is run on Cray XT3
Elena Gryaznova [Fri, 20 Aug 2010 16:47:22 +0000 (20:47 +0400)]
b=14242 test_6g fails when b_release_1_6_4 is run on Cray XT3

i=Andreas.Dilger

9 years agob=22364 sanity-quota: remote obsolete test_10, test_16
Elena Gryaznova [Fri, 20 Aug 2010 16:59:36 +0000 (20:59 +0400)]
b=22364 sanity-quota: remote obsolete test_10, test_16

i=Yong.Fan

9 years agob=18016 fix index type in ll_read_ahead_page
Oleg Drokin [Fri, 20 Aug 2010 16:00:39 +0000 (20:00 +0400)]
b=18016 fix index type in ll_read_ahead_page

The page index type in ll_read_ahead_page was int which is wrong
and breaks after we cross ~10T barrier breaking read ahead beyod
that offset. Changred it to pgoff_t

i=wangdi
i=panda

9 years agob=22418 2.6.32 compat synchronize_rcu already defined
Mikhail Pershin [Fri, 20 Aug 2010 16:29:06 +0000 (20:29 +0400)]
b=22418 2.6.32 compat synchronize_rcu already defined

i=adilger
i=kalpak

9 years agob=21980 cache `ll_obdo_cache': Can't free all objects
Dmitry Zogin [Thu, 19 Aug 2010 14:53:31 +0000 (18:53 +0400)]
b=21980 cache `ll_obdo_cache': Can't free all objects

 Always use OBDO_ALLOC/FREE for obdo allocations to prevent slab fragmentation.
 Also removing OBD_FL_TEMPORARY flag relicts.

 o=johann
 i=di.wang
 i=dmitry.zoguine

9 years agob=23179 MDS node unresponsive
Isaac Huang [Thu, 19 Aug 2010 17:11:46 +0000 (21:11 +0400)]
b=23179 MDS node unresponsive

Better hash distribution, doubled hash size for the lnet cookie hash.

i=zhen.liang
i=andreas.dilger

9 years agob=23120 eliminate defunct changelog_send_thread processes
Nathan Rutman [Thu, 19 Aug 2010 21:36:44 +0000 (01:36 +0400)]
b=23120 eliminate defunct changelog_send_thread processes

o=thomas.leibovici@cea.fr
i=nathan

9 years agob=16909 Console message cleanup
Cliff White [Thu, 19 Aug 2010 20:28:13 +0000 (00:28 +0400)]
b=16909 Console message cleanup

i=maxim.patlasov
i=johann.lombardi

Use D_NET for DEBUG_REQ

9 years agob=22797 notify users that recovery-[random,double]-scale tests are not functional...
Elena Gryaznova [Thu, 19 Aug 2010 11:23:08 +0000 (15:23 +0400)]
b=22797 notify users that recovery-[random,double]-scale tests are not functional for SOFT failure mode

i=Brian.Murrell

9 years agob=22906 limit filesystem size for LUN size close to 16TB
Mikhail Pershin [Thu, 19 Aug 2010 16:36:07 +0000 (20:36 +0400)]
b=22906 limit filesystem size for LUN size close to 16TB

a=adilger
i=zhiqi.tao
i=girish.shilamkar

9 years agob=21825 doxygen comments for CMM
Mikhail Pershin [Thu, 19 Aug 2010 15:22:42 +0000 (19:22 +0400)]
b=21825 doxygen comments for CMM

i=di.wang

9 years agob=20334 fix conflicting ioctl numbers
Mikhail Pershin [Thu, 19 Aug 2010 15:01:54 +0000 (19:01 +0400)]
b=20334 fix conflicting ioctl numbers

i=nathan
i=tappro

9 years agob=20973 Doxygen comments for LNet API
Isaac Huang [Wed, 18 Aug 2010 05:40:56 +0000 (09:40 +0400)]
b=20973 Doxygen comments for LNet API

Doxygen comments for LNet API. This is a comment-only change.

i=maxim

9 years agob21619 hash ME on RDMA portal i=isaac i=maxim
Liang Zhen [Tue, 17 Aug 2010 07:06:13 +0000 (11:06 +0400)]
b21619 hash ME on RDMA portal i=isaac i=maxim

9 years agob=22882 MMP might sleep negative time
Rahul Deshmukh [Thu, 19 Aug 2010 06:29:34 +0000 (10:29 +0400)]
b=22882 MMP might sleep negative time

b=22882 MMP might sleep negative time

i=andreas
i=girish

Ported MMP might sleep negative time patch.

9 years agobz=21869 improve filter_fid2dentry() error message
Vitaly Fertman [Wed, 18 Aug 2010 20:30:04 +0000 (00:30 +0400)]
bz=21869 improve filter_fid2dentry() error message

o=adilger
i=girish
i=landen

9 years agob=23196 take lov reference in lov_quota_adjust_qunit so as to avoid races
Andrew Perepechko [Wed, 18 Aug 2010 18:12:07 +0000 (22:12 +0400)]
b=23196 take lov reference in lov_quota_adjust_qunit so as to avoid races

a=Johann Lombardi
i=ZhiYong Tian
i=Andrew Perepechko

9 years agob=22360 return -EIO in close() if there was any async I/O error
Andrew Perepechko [Tue, 17 Aug 2010 09:31:39 +0000 (13:31 +0400)]
b=22360 return -EIO in close() if there was any async I/O error

i=Andreas Dilger
i=Johann Lombardi
i=ZhiYong Tian

9 years agob=22923 increase show_error timeout from 5 to 10
Elena Gryaznova [Tue, 17 Aug 2010 19:12:57 +0000 (23:12 +0400)]
b=22923 increase show_error timeout from 5 to 10

i=Maxim.Patlasov

9 years agob=22923 increase lst stat timeout
Elena Gryaznova [Tue, 17 Aug 2010 19:08:20 +0000 (23:08 +0400)]
b=22923 increase lst stat timeout

i=Maxim.Patlasov

9 years agob=22107 pin object's inode in memory to avoid certain timeouts
Andrew Perepechko [Mon, 16 Aug 2010 21:42:10 +0000 (01:42 +0400)]
b=22107 pin object's inode in memory to avoid certain timeouts

i=Andreas Dilger
i=Johann Lombardi

9 years agob=21259 udev rule to set /dev/obd perms 666
Dmitry Zogin [Fri, 13 Aug 2010 14:58:43 +0000 (18:58 +0400)]
b=21259 udev rule to set /dev/obd perms 666

 i=christopher.morrone
 i=dmitry.zoguine

9 years agob=22423 Reconnects are not throttled
Mikhail Pershin [Thu, 19 Aug 2010 07:02:05 +0000 (11:02 +0400)]
b=22423 Reconnects are not throttled

 Rely on pings to issue reconnects. Conf-sanity.sh test_35 modified.

 i=nathan.ruthman
 i=dmitry.zoguine

9 years agob=19102 set default values of filesystem-wide for LOVEA.
yangsheng [Fri, 13 Aug 2010 13:19:26 +0000 (17:19 +0400)]
b=19102 set default values of filesystem-wide for LOVEA.

i=adilger
i=nathan

9 years agob=17471 change conf_param syntax to match set_param
Nathan Rutman [Fri, 13 Aug 2010 22:11:07 +0000 (02:11 +0400)]
b=17471 change conf_param syntax to match set_param

i=adilger
i=emoly.liu

9 years agob=22656 Prevent failover nids from registering with MGS first.
Nathan Rutman [Fri, 13 Aug 2010 20:29:51 +0000 (00:29 +0400)]
b=22656 Prevent failover nids from registering with MGS first.

Make the check in mgs_handle_target_reg()

o=Joseph Herring
i=nathan.rutman
i=andreas.dilger

9 years agob=22464 add writeconf as mount option (includes bz 23228)
Nathan Rutman [Fri, 13 Aug 2010 16:47:47 +0000 (20:47 +0400)]
b=22464 add writeconf as mount option (includes bz 23228)

i=alexg
i=adilger

9 years agob=22137 kernel oops at replay-single test_61d.
Dmitry Zogin [Fri, 13 Aug 2010 13:39:38 +0000 (17:39 +0400)]
b=22137 kernel oops at replay-single test_61d.

 replay-single.sh test_61d was modified to operate with MGS in case of
 the different MGS and MDS.

 i=grev

9 years agob20288 fix a deadlock in kiblnd_check_conns i=isaac i=maxim
Liang Zhen [Tue, 17 Aug 2010 06:26:17 +0000 (10:26 +0400)]
b20288 fix a deadlock in kiblnd_check_conns i=isaac i=maxim

9 years agob=23185 check for both arches
Brian J. Murrell [Fri, 13 Aug 2010 15:03:10 +0000 (19:03 +0400)]
b=23185 check for both arches

When we build our version of the SLES kernel, we optimize it for i686
whereas the SUSE kernel is i386.  The actual arch makes a difference in
where the Module.symvers can be found, so just look in both locations
to cover both the upstream vendor kernel as well as our patched kernel.

9 years agob=23185 use resolve_arch() in the test too
Brian J. Murrell [Fri, 13 Aug 2010 15:03:09 +0000 (19:03 +0400)]
b=23185 use resolve_arch() in the test too

Missed a usage need for resolve_arch() in the test for the file we
extracted, using resolve_arch().

9 years agob=23185 properly support sles10 ppc64
Brian J. Murrell [Fri, 13 Aug 2010 15:03:08 +0000 (19:03 +0400)]
b=23185 properly support sles10 ppc64

We never really have properly supported ppc64 on sles10:
- RPMSMPTYPE for ppc64 on SLES10 is ppc64, not smp
- add a function to resolve an arch to their canonical value

Comment out some code to align the sles distro lbuild with rhel5.

9 years agob=21452 kABI tracking
Brian J. Murrell [Fri, 13 Aug 2010 14:17:01 +0000 (18:17 +0400)]
b=21452 kABI tracking

Changes required to build packages that properly interact with
the kABI tracking included in SLES and RHEL kernels:
- take local copies of the find_requires generators and make modif-
  ications to them so that they can find the kernel artifacts installed
  in our build tree
  + thusly, override the __find_requires macro in all of the RPM
    build operations

Fix package versioning of patchless packages to also include the
timestamp, just like patched packages do, for non-release builds.
Also, remove the _lustre from the kernel version in the lustre
package names for patchless packages.

Note that changes needed to enable the external dependency checker
and thus include kABI dependencies in the RPM are not included here
for RHEL5.  Once RH have updated their kABI whitelist to include the
symbols we need, the patches in bz attachments 30546-7 need to land
for RHEL5 weak modules to work and a bug needs to be filed with OFA
to land the change in bz attachment 30547.

i=wangyb
i=mjmac

9 years agob=17166 remove <2.6.9 compatibility code from HEAD
LiuYing [Tue, 17 Aug 2010 03:40:49 +0000 (07:40 +0400)]
b=17166 remove <2.6.9 compatibility code from HEAD

HEAD requires 2.6.9 and above, so we remove older code.
The majority of the remaining compatibility code is in libcfs and lnet.

i=andreas
i=isaac

9 years agob=13752 Remove 2.4 kernel definition of cfs_waitq_wait_event_interruptible_timeout().
yangsheng [Fri, 13 Aug 2010 13:12:44 +0000 (17:12 +0400)]
b=13752 Remove 2.4 kernel definition of cfs_waitq_wait_event_interruptible_timeout().

i=issac

9 years agob=23122 Get ia64/rhel5 build successful.
yangsheng [Fri, 13 Aug 2010 13:09:41 +0000 (17:09 +0400)]
b=23122 Get ia64/rhel5 build successful.

9 years agoMerge branch 'master' of git.lustre.org:prime/lustre
Mikhail Pershin [Tue, 17 Aug 2010 05:18:37 +0000 (09:18 +0400)]
Merge branch 'master' of git.lustre.org:prime/lustre

9 years agob=19625 ext4 support for SLES11
Rahul Deshmukh [Sat, 14 Aug 2010 08:09:50 +0000 (12:09 +0400)]
b=19625 ext4 support for SLES11

Bug 19625 ext4 support for SLES11

Ported the ext4 support for SLES11 related changes for master.

9 years agoUpdated for the next bi-weekly build. 2.0.50.0 v2_0_50_0
unknown [Mon, 16 Aug 2010 22:22:34 +0000 (15:22 -0700)]
Updated for the next bi-weekly build.

9 years agob=23221 merge lock descripters if overlapped
Vitaly Fertman [Fri, 13 Aug 2010 08:55:00 +0000 (12:55 +0400)]
b=23221 merge lock descripters if overlapped

i=green
i=ericm

lock descriptors prepared by vvp_mmap_locks() and ccc_io_one_lock_index()
may overlap and must be merged

9 years agob=23196 take lov reference in lov_quota_check
Andrew Perepechko [Mon, 16 Aug 2010 11:58:06 +0000 (15:58 +0400)]
b=23196 take lov reference in lov_quota_check

i=Johann Lombardi
i=ZhiYong Tian

9 years agob23289 smp improvement for watchdog i=eric.mei i=maxim
Liang Zhen [Sun, 15 Aug 2010 16:46:52 +0000 (20:46 +0400)]
b23289 smp improvement for watchdog i=eric.mei i=maxim

9 years agob22078 wrong value is used by parse_net
Liang Zhen [Sun, 15 Aug 2010 16:46:00 +0000 (20:46 +0400)]
b22078 wrong value is used by parse_net

i=nathan i=robert i=qiang.wu

9 years agob=17630 Don't export nfs when work with 4k stack.
yangsheng [Fri, 13 Aug 2010 13:05:28 +0000 (17:05 +0400)]
b=17630 Don't export nfs when work with 4k stack.

i=adilger

9 years agob=20563
pravin [Fri, 13 Aug 2010 11:19:11 +0000 (15:19 +0400)]
b=20563

i=vitaly
i=andreas
patch adds mount option to generate 32bit ino, this can be used
for 32bit application compatibility.