i=Alexander Zarochentsev (alexander.zarochentsev@sun.com)
i=ZhiYong Tian (tianzy@sun.com)

fix HAVE_QUOTA_SUPPORT usage

i=Adilger
i=grev
replace proc with lctl [set|get]_param

i=Adilger
build mpitests only if --disable-mpitests is not set

i=Nathan.Rutman
ignore test_116 failure

b=16208
i=johann, adilger

Bring down the stack usage

b=16570
i=robert.read, h.huang

Make MDS_SETXATTR request replayable.

b=16698
i=oleg.drokin
i=robert.read

Branch b1_6

b=16678
i=nathan

i=Nathan.Rutman
restore the original at_max on all nodes

invert printk to CERROR/CWARN in lustre_quota_fmt
b=16851
i=johann
i=panda

b=16676
i=Johann Lombardi (johann@sun.com)
i=Zhenyu Xu (zhenyu.xu@sun.com)

add inode locking around vfs-calling critical sections

b=16646
i=ZhiYong Tian(tianzy@sun.com)
i=Sergey Glushchenko(deen@sun.com)
i=Johann Lombardi(johann@sun.com)

lfs to put * when a limit is reached

b=16758
i=shadow, adilger, johann

Fixed the endian problem for PPC64 patchless client.

i=Adilger
i=Robert.Read
test_14 fix: use multiop instead of exec

i=Robert.Read
o=Brian.Behlendorf
test_27b fix: check mount

i=Adilger
increase MINRES to 2^32 -1

i=shadow
i=johann
b=15198

i=Adilger
runbench fix: skip if dbench or client are not found
t-f skip fn fix: do not fail if TESTSUITELOG is not found;
log fn: escape /

b=16789
i=adilger
i=johann

correction in fiemap macros in wirecheck and other fixes to wirecheck

request will resend before finished ost llog init and this produce
access to non init llog context.

Branch b1_6
b=16839
i=adilger
i=rread

tests 20c, 32, 37 fixes: let pid to be killed

remote command fix for previous commit

b=16098
i=hua.huang

add a special LICENSE file for Cray

i=Adilger
test_101 fix

i=Adilger
check CATASTROPHE on all nodes

i=Adilger
make oos.sh to be verbose; get rid of proc

index of lov target as data parameter.

Branch b1_6
b=16826
i=adilger
i=green

r=shadow,wangdi

- new recov_thread.c code;
- cleanups, fixes

Ignore generated file.

Don't print console message the first time the import is set.

ost count is different between original processing and replay.
not count ost twice in case ost connected before recovery finished.

Branch b_1_6
b=16736
i=tappro
i=green

i=Tappro
test_34 fix: let pid to be killed

b=16156
i=shadow, zhenyu.xu

Use patchless_client instead patchless to avoid misunderstanding.

i=Tappro
test_24 and t-f client_reconnect fn fix

b=15927
i=johann
i=huanghua

Add ChangeLog entry.

avoid I/O failures after umount during fail back
Branch b1_6
b=16611
i=green
i=rread

b=16359
i=shadow, panda

make sure _debug_req don't retrive diffdrent endian lustre_msg.

i=ericm
i=rread
increase fudge for Cray