i=Adilger
replace cleanup_and_setup_lustre fn by check_and_setup_lustre fn

r=adilger,johann

- removes deadlock possibility by disabling rehash in hash_del() operations and moving hash_add()
out of spin_locks when calling. Hash table has own mechanisms for protecting its structures and it
also has hash_add_unique() method for using in concurrent run contexts;

- fixed missed lh_put() in hash_add_unique() which led to extra refs in some cases (extra ref to
export) and inability to cleanup;

- fixed __lustre_hash_set_theta() which set @max theta into ->lh_min_theta;

- in lustre_hash_rehash_size() disable rehash also for the case when new and old hash sizes equal
in corner cases (max_size or min_size). Before this fix it could be possible to do needless
rehashes when size is actually did not change but we do this expensive operation;

- disable rehash in hash_add_unique() if no actual add happened since entry with the same key is
already found in the table;

- some cleanups in hash table code;

i=Adilger
check config if lustre is mounted before acc-sm run

i=Brian
assert_DIR cleanup

r=tappro,johann

- implements proper locking for rq pool freeing

b=16860
i=nathan
i=rread

Description: Excessive recovery window
Details    : With AT enabled, the recovery window can be excessively long (6000+
	     seconds). To address this problem, we no longer use
	     OBD_RECOVERY_FACTOR when extending the recovery window (the connect
	     timeout no longer depends on the service time, it is set to
	     INITIAL_CONNECT_TIMEOUT now) and clients report the old service
	     time via pb_service_time.

b=16578
o=adilger

A faster way to get long string.

- make sure that rpcs in RQ_PHASE_UNREGISTERING phase can be marked expired and interrupted.

r=johann,shadow

- fixes ptlrpcd blocking on very long reply unlink waiting. To do so new rpc phase introduced
RQ_PHASE_UNREGISTERING in which request stay until we have reply_in_callback() called by lnet
signaling that reply is unlinked. All requests in this state are skipped in processing by prlrcd
instead of waiting n * 300s on each of them. This allows ptlrpcd to process other rpcs in the set;

- make sure that inflight count is coherent with being present on sending or delay list. That is,
if we see inflight != 0, rpc must be on one of these lists. This is very helpful in
ptlrpc_invalidate_import() to show all rpcs still waiting after invalidating import;

- in ptlrpc_invalidate_import() wait maximal rq_deadline - now from all inflight rpcs instead of
obd_timeout which may be much longer. If calculated timeout is 0, obd_timeout is used. This fixes
the issue that rq_deadline - now > obd_timeout (very easy to see in logs) which led to inflight !=
0 assert because inflight rpcs timed out later than our wait period is finished;

- in ptlrpc_invalidate_import() wait forever for rpcs in UNREGISTERING phase. Check in assert for
inflight == 0 for wait timed out case if no rpcs in UNREGISTERING phase. Only those in
UNREGISTERING phase are allowed to stay longer than obd_timeout;

- added ptlrpc_move_rqphase() function. All phase changes go through it. Add debug_req() there to
track down all phase changes;

- conf_sanity.sh test_45 added to emulate very long reply unlink and also situation when
rq_deadline - now > obd_timeout;

- do not wait forever in ptlrpc_unregister_reply() for async case (using it from sets). sync case
left unchanged;

- make sure that ptlrpc_set_next_timeout() yields 1s timeout (instead of 0s) for the set with rpcs
in "unregistering" stage to prevent ptlrpcd from sleeping forever and hanging in test_45;

- in ptlrpcd() make sure that we do not sleep on 0 timeout.

b=17371
i=Johann Lombardi
i=Oleg Drokin

fix a race between requeue thread processing and umount

i\Adilger
correct remote_[mds|ost] fn to work correctly on configuration
with several MDS/OSS nodes

b=16438
i=adilger
i=girish

Mounting a filesystem with extents feature will fail on big-endian systems since ext3-based ldiskfs is not supported on big-endian systems. This can be over-riden with "bigendian_extents" mount option.

b=15715
r=adilger,green

Fixed the race of destroying and enqueuing a ldlm lock at OST side.

b=16578
i=adilger

Description: ldlm_cancel_pack()) ASSERTION(max >= dlm->lock_count + count)
Details    : If there is no extra space in the request for early cancels,
             ldlm_req_handles_avail() returns 0 instead of a negative value.

b=17534
i=alilger, yangsheng

Fixed for client crash by old-style mount command.

Replace LBUG with RETURN(-EINVAL) to avoid crashing
b=5135
i=adilger
i=johann

  b:12512
  i:grev, adilger

b=17493
i=Andreas Dilger
i=Johann Lombardi
handling of a broken readonly key

fix an error in the test_18 of sanity-quota.sh
b=17523
i=johann
i=panda

Quiet compiler warning about unused label.
Conditional check will be optimized away by compiler.

Fix 80-column line wrapping.

i=Nikita
sanity test_100 fix

i=Nikita
test_53 fix

Remove trailing whitespace.

o=Robert.Read
i=grev
test_27u fix

b=13904
i=Johann Lombardi
i=ZhiYong Tian

64-bit quota support for kernel

b=17379
i=adilger, johann

Test case for recursive symlink.

b=17379
i=Brian(LLNL), johann

Set recursive symlink depth to 5 when kernel has 4K stack.

fix a possible NULL pointer in client_quota_ctl()
b=17486
i=johann
i=panda

i=johann
i=adilger
b=17448

Branch b1_6
b=16887
i=pravin.shelar
i=adilger

Address LBUG, ASSERTION(client_stat->nid_exp_ref_count == 0) failed:count -1
* add client stat on obd_nid_stat after client stat is ready.
* properly decrease exp_nid_stats' nid_exp_ref_count in lprocfs_exp_cleanup().

remove unecessary return.

Once the unmatched stride IO mode is detected, shrink the stride-ahead window to 0.
if it does hit cache miss, and read-pattern is still stride-io mode,
does not reset the stride window, but also does not increase the stride
window length in this case.
b=17197
i=Nikita
i=Andreas

i=johann
i=adilger
coverity fix.  thought I landed this awhile ago, but apparently not...

b=17491
i=nathan
i=rread

Quick fix patch from behlendorf1@llnl.gov.

b=16680
i=adilger, kalpak (o=bzzz)

Detect on-disk corruption of block bitmap and better checking of preallocated blocks.

b=17089
i=wangdi

fix mistake made when the patch was landed.