and exist connect request in flight, this can produce access to
  NULL pointer (or already free pointer) when connect_interpret
  store ocd flags in obd_namespace.

  b=14260
  i=adilger
  i=johann

b=14418
i=green
i=johann

b=12211
origianl patch producer=adilger
i=green, bobijam

move obd_fail_check() from being a static inline to being a function.
We already protect this function from being called needlessly by use of
unlikely() and checking obd_fail_loc != 0 before calling it.

Having such a large function inline bloats the code and likely reduces
performance by putting extra (though unlikely) code in every function.

b=14215
i=johann, adilger

user pointer parameter (struct ldlm_res_id*) in ldlm_cli_enqueue_local()
to reduce the stack usage.

i=Adilger

remove useless "echo"

i=Adilger

report the slow and excluded tests separately

fix a race condition of early reply release.
b=14274
r=nathan, andreas

r=alex

i=Adilger

force test_5c to use mount_client

i=Adilger

typo fixed (found by Huang Hua)

r=adilger,shadow,deen

- enable OBD_CONNECT_LRU_RESIZE only when lru resize is enabled in configure time.

i=Adilger

test_29 cleanup

b=13563
i=green
i=shadow

1) cancel lru locks in ldlm_cli_cancel() and send 1 batched cancel RPC;
2) pass the whole list of canceled locks to the async blocking thread, not
lock-by-lock, and send cancels in batched cancel RPCs.

r=adilger,nikita,shadow

- lots of fixes in ldlm pools code for smoother work and correct memory pressure handling.

r=johann,adilger

Commit missing parts of bug 10718 to correctly clear lock cache on liblustre.

b=13843
i=adilger
i=nathan

Fix coding style.
(Dynamically create DLM blocking callback threads)

b=13843
i=adilger
i=nathan

Dynamically create DLM blocking callback threads
+ fix an issue with MacOS implementation of atomic_sub_and_test().

b=12646
i=johann, shadow

Add opc parameter to ost_checksum_bulk().

b=13715
i=johann, nathan

Don't let out-dated timedout request invalidate the import.

b=14208
i=zhenyu.xu, johann

Replace macro __LINUX__ with __linux__.

r=vitaly,shadow,adilger(some arly versions)

Attach cached pagers to locks that protect those pages. This helps to quickly
find those pages for flush and removal purposes on AST reception and
significantly speeds up lock revocation on big files with pages at
high offsets.

b=12749
i=adilger
i=johann

b=12829
i=adilger
i=bobijam

Fix tiny nit in the changelog entry.

b=12829
i=adilger
i=bobijam

When CRAY_XT3 is defined, the fsgid supplied by the client is
overridden with the primary group provided by the group upcall,
whereas the supplied fsgid can be trust if it is in the list of
supplementary groups returned by the group upcall.

b=11301

i=nathan.rutman

cleanup: remove tdir creation

i=Adilger

get rid of relative path in setup_test102

b=14326
r=johann,bwzhou

Use old size assignment to avoid deadlock caused by ll_inode_size_lock calls down semaphore under spinlock held.

b=14135
i=nathan
i=huanghua

Original "mgc_logname2resid" process both "logname" (which consists of fsname-nodetype)
and "fsname", it can not distinguish "-" contained fsname and the logname.
Split it into two functions: "mgc_logname2resid" and "mgs_fsname2resid" for that.

b=14283
i=zhenyu.xu, oleg.drokin

show "used" as only the used space, to match "df".

i=Adilger

acc-sm "SLOW" list (all scripts)

i=grev
i=adilger
branch=b1_6

change rundbench script to use contents of DBENCH_LIB variable to
find the location of the dbench executable (like acceptance-small.sh does)
so that we use the same environment variables to specify the location
of the dbench executable in both acceptance-small.sh and rundbench
instead of using different environment variables to configure the
same thing in the two different scripts.

b=14270
i=johann, kalpak.shah

Description: lfs find does not continue on file error
Details    : Continue other files processing when a file/dir is absent.

b=14016
Fix BUGON when do per-client proc cleanup. Patch from shadow
i=Andreas,Nathan

i=Adilger

get rid of not default stripe size; use new-style setstripe args

typo fixed

r=andrew.perepechko
r=johann

Branch b1_6
With this patch, three improvements are included:
1. detete the softlimit in mds and osts when use "lfs quota".
2. display the inaccurate data in the output of "lfs quota".
3. try to get quota info when "lfs quota" is executed.