do not drop replay according to msg flags, instead we check the
per-export recovery request queue for duplication of transno.
b=15756
r=adilger
r=rread

i=Adilger

config for several remote clients; replay-single test_70a, test_70b

i=Adilger

skip test_32a, test_32b for NETTYPE != tcp

b=15759
r=rread, green

Don't consider -EMFILE (too many threads) a fatal startup error

r=Andreas

to remove instances of /proc/sys in sanity.sh by using either sysctl or lctl {get,set}_param

Quiet build warnings when building MPI test programs.

b=13285
r=adilger, deen

1. updated ext3-mballoc3-core.patch and added ext3-max-dir-size.patch to limit the maximum size of a directory;
2. updated other ldiskfs patches to resolve patch conflicts caused by the above two new patches;
3. added a test case in sanity.sh.

r=andrew.perepechko, johann

Patch by Adilger to replace /proc/fs/lustre access with lctl set_param in sanity-quota.sh

r=Adilger

Changed replay-single.sh to use 'lctl {get,set}_param'

i=adilger

disable LIBLUSTRE test for NETTYPE != tcp && NETTYPE != ptl

Ignore compiled file.

Description: MDS may lose file striping (and hence file data) in some cases
Details    : If there are additional extended attributes stored on the MDS,
	     in particular ACLs, SELinux, or user attributes (if user_xattr
	     is specified for the client mount options) then there is a risk
	     of attribute loss.  Additionally, the Lustre file striping
	     needs to be larger than default (e.g. striped over all OSTs),
	     and an additional attribute must be stored initially in the
	     inode and then increase in size enough to be moved to the
	     external attribute block (e.g. ACL growing in size) for file
	     data to be lost.
i=kalpak.shah
i=alex.zhuravlev
b=15777

Sync ChangeLog from b_release_1_6_5.

b=12375
i=deen
i=grev

i=Adilger

sanity test_118* fixes; set fail loc on all oss nodes

i=Adilger

the minor cleanup of the previous fix:
-- after we get rid of $tdir we do not need to create $tmpdir, because
   unzip -d <dir> does not need <dir> exists, it needs the parent directory
   of <dir> exists; unzip creates <dir> if it does not exists.

b=14790
i=Johann
i=ZhiYong

additional test. check if setting limits twice doesn't break oplimits.

Reinitialize optind to 0 on each lfs_* call, in case lfs is being used interactively.  Most lfs_* functions do this already, but a few (find, setstripe, quotachown) do not.
 b=15212
 i=rupesh
 i=johann

Fix typo in error message.

b=15521
r=adilger, grev

the test case for bug 15521

i=Adilger

conf-sanity test_32a, test_32b: create tmp dir correctly

i=Adilger

disable LIBLUSTRE test for NETTYPE != tcp

i=vitaly.vertman
i=oleg.drokin

Lockless i/o code fixes and improvements:
 (1) lockless truncate checks for OBD_CONNECT_TRUNCLOCK and ll_file_punch
     sets OBD_FL_TRUNCLOCK correctly.
 (2) an lproc control for lockless truncate,
     lproc statistics for lockless truncate.
 (3) sanityN tests for lockless code

b=15517
i=johann
i=umka

Fix compile warnings on i686.

b=15691
i=huanghua
i=tappro

1) Prevent "ll_sa_wrong" from being increased repeatedly when hit ratio is low.
2) Remount client before sanity 123a to eliminate previous test effect.

b=15643
i=johann

Prolong the waiting time in test_12 of sanity-quota.sh

i=grev

fix for test_16 failure due to trying to create mntpt dir on Read-Only fs.

b=14909
i=tianzy

use cancel_lru_locks instead of "sync; sleep 1; sync" in sanity-quota.sh.

b=14929
i=zam

b=15406
i=huanghua
i=tappro
i=vitaly

Back port dir_SA fixes from HEAD (b1_8_dir_ra) to b1_6.

i=Adilger
skip test_37 if CLIENTMODSONLY is set

change at_max according to AT status.
b=13659
r=nathan
r=rread
r=grev

i=Adilger
force test_20a and test_20b to do write on the correct ost

i=Nathan.Rutman
skip test_37 for client only testing

i=Adilger
allow both "cfs" and "sun" e2fsck

i=Yury.Umanets
i=Alexey.Lyashkov
force test_69 to fail if directio write failed

r=grev
- fixes using check_runas_id(). Adds check for used runas id in test_72.

i=Yury.Umanets
force test_66b to use FSNAME

i=Oleg.Drokin
load crc16 module if it exists

Add a new item(lq_lock) in struct lustre_qunit which protects every single qunit
and qunit_hash_lock only protects qunit hash. It will improve parellelism.
i=johann
i=andrew

i=Adilger
load crc16