git://git.whamcloud.com - fs/lustre-release.git/log

LU-1396 osc: control the RPC rate between MDS and OST

1. Limit the RPC rate with setting 50 maxim rpc in flight between
MDS and OST.

2. Add specify flag in oa to tell whether the destory from echo_md
destory or MDT close/unlink orphan, where we can not throttle the
destory RPC, since it might block ptlrpcd thread. See Bug 16006.

Signed-off-by: Wang Di <di.wang@whamcloud.com>
Change-Id: Ie3b555eecd2716a82b058f645d002f46d8c4dd36
Reviewed-on: http://review.whamcloud.com/2899
Tested-by: Hudson
Tested-by: Maloo <whamcloud.maloo@gmail.com>
Reviewed-by: Bobi Jam <bobijam@whamcloud.com>
Reviewed-by: Andreas Dilger <adilger@whamcloud.com>
Reviewed-by: Oleg Drokin <green@whamcloud.com>

LU-1305 osd: body operations for osd-zfs

read/write, get/put bufs, punch

Signed-off-by: Alex Zhuravlev <bzzz@whamcloud.com>
Change-Id: Ie2ccd127b52b00904ec0ce5ba143466ca0328f95
Reviewed-on: http://review.whamcloud.com/2968
Reviewed-by: Andreas Dilger <adilger@whamcloud.com>
Tested-by: Hudson
Reviewed-by: Johann Lombardi <johann@whamcloud.com>

LU-1305 osd: index operations for osd-zfs

index insert/delete/lookup, iterator support

Signed-off-by: Alex Zhuravlev <bzzz@whamcloud.com>
Change-Id: I46819a5dd4da601716e516fb54942aa611497aef
Reviewed-on: http://review.whamcloud.com/2967
Reviewed-by: Andreas Dilger <adilger@whamcloud.com>
Tested-by: Hudson
Reviewed-by: Johann Lombardi <johann@whamcloud.com>

LU-56 libcfs: CPT affinity workitem scheduler

this patch covered multipled changes:
- flexible APIs for creating WI schedulers
  a) therioticall user can create any number of WI schedulers, each
     scheduler can have its own threads pool
  b) user can create CPT affinity WI schedulers for each CPT, it's
     reserved for LNet selftest.
- rehashing and LNet selftest will not share WI schedulers anymore
- libcfs will only start a WI scheduler with small number of threads
  for cfs_hash rehashing
- LNet selftest will create its own schedulers on starting of module,
  and destroy schedulers on shutting down of module

Signed-off-by: Liang Zhen <liang@whamcloud.com>
Change-Id: Idf66a83817fe847ed29e052e0ddc2a4fed498f1a
Reviewed-on: http://review.whamcloud.com/2729
Tested-by: Hudson
Tested-by: Maloo <whamcloud.maloo@gmail.com>
Reviewed-by: Bobi Jam <bobijam@whamcloud.com>
Reviewed-by: Doug Oucharek <doug@whamcloud.com>
Reviewed-by: Oleg Drokin <green@whamcloud.com>

LU-56 obdclass: SMP improvement for lu_key

ptlrpc service threads need to call lu_context_init/fini in each
loop (for each RPC), this could be a big performance issue on
fat SMP machine if we always add lu_context to remember list
because operations on lu_context with LCT_REMEMBER are serialized
by one spinlock and we need to lock/unlock it for multiple times
for each RPC.

But we found LCT_REMEMBER is abused at here, it's impossible that
server stack is being unloaded but service threads are still
running, so we can simply remove LCT_REMEMBER flag from ->rq_session,
and made some small changes to bypass global lock in lu_context_init
and lu_context_fini if w/o LCT_REMEMBER.

Signed-off-by: Liang Zhen <liang@whamcloud.com>
Change-Id: I5875a90365a103707526483047ec7628f6964a56
Reviewed-on: http://review.whamcloud.com/2824
Tested-by: Hudson
Tested-by: Maloo <whamcloud.maloo@gmail.com>
Reviewed-by: Jinshan Xiong <jinshan.xiong@whamcloud.com>
Reviewed-by: Fan Yong <yong.fan@whamcloud.com>
Reviewed-by: Oleg Drokin <green@whamcloud.com>

LU-56 lnet: multiple cleanups for inspection

This patch covered multiple cleanups for previous patches:
- user "features" to replace "versions" of router checker ping
- code cleanup for lnet_ni_alloc
- fix a loading issue on 32-bit system
- comments cleanup and some small changes
- coding style cleanup

Signed-off-by: Liang Zhen <liang@whamcloud.com>
Change-Id: I96e5fa260d93082851c5146883df1a5b8a96ef42
Reviewed-on: http://review.whamcloud.com/3180
Reviewed-by: Bobi Jam <bobijam@whamcloud.com>
Tested-by: Hudson
Tested-by: Maloo <whamcloud.maloo@gmail.com>
Reviewed-by: Andreas Dilger <adilger@whamcloud.com>

new tag 2.2.58

Change-Id: I22a3a84ce6e2ed4260c3a8652131afc43df22e46
Signed-off-by: Oleg Drokin <green@whamcloud.com>

LU-56 lnet: allow user to bind NI on CPTs

By default, NI will be bond on all CPTs, which means messages for a
NI could be handled by LND threads on any CPT (hashed by NID).
This patch add a new parameter for NI configuration, it allows user
to bind NI on specified CPT(s):

- tcp0(eth1)[0,1]
bind NI (tcp0) on CPT0 and CPT1
- o2ib(ib0)[2-5]
bind NI (o2ib) on CPT2,3,4,5

Expression between square brackets are CPTs that user wants this NI
to bind, if user provided this expression, messages for the NI
will only be handled by LND threads running on specified CPTs.

This is an intermediate patch, to get this feature we also need
upcoming LND patches.

Signed-off-by: Liang Zhen <liang@whamcloud.com>
Change-Id: I706a92c6da181ed0fec857cc25b5ae27a7a7c36b
Reviewed-on: http://review.whamcloud.com/3114
Tested-by: Hudson
Tested-by: Maloo <whamcloud.maloo@gmail.com>
Reviewed-by: Doug Oucharek <doug@whamcloud.com>
Reviewed-by: Bobi Jam <bobijam@whamcloud.com>
Reviewed-by: Oleg Drokin <green@whamcloud.com>

LU-56 lnet: Partitioned LNet networks

We have implemented partitioned LNet sources (MD/ME/EQ),
This patch created partitioned data for other LNet objects:
- Peer-tables
Peers are hashed into peer-table on different partitions by NID
- NI refcount and message queue
NI will have refcount and message queue for each partition
- counters for each partition

These objects are protected by percpt lock lnet_t::ln_net_lock,
which replaced the original LNET_LOCK

Signed-off-by: Liang Zhen <liang@whamcloud.com>
Change-Id: I7c8c1359aca04a7f859672ccd3268f0282505dd5
Reviewed-on: http://review.whamcloud.com/3113
Tested-by: Hudson
Tested-by: Maloo <whamcloud.maloo@gmail.com>
Reviewed-by: Doug Oucharek <doug@whamcloud.com>
Reviewed-by: Bobi Jam <bobijam@whamcloud.com>
Reviewed-by: Oleg Drokin <green@whamcloud.com>

LU-1341 tests: recovery-small test_19* fixes

- test 19a uses ELC, so no eviction happens;
- test 19b cancels conflicting lock in advance due to CLIO logic,
no eviction again;

fix tests to check what is expected - eviction must take place.

Change-Id: Ib5365c9f90d1fc388fe81196661b262da5e7ad78
Xyratex-bug-id: MRP-482
Reviewed-by: Alexander Zarochentsev <alexander_zarochentsev@xyratex.com>
Reviewed-by: Alexander Lezhoev <Alexander_Lezhoev@xyratex.com>
Signed-off-by: Vitaly Fertman <vitaly_fertman@xyratex.com>
Reviewed-on: http://review.whamcloud.com/2592
Tested-by: Hudson
Reviewed-by: Jinshan Xiong <jinshan.xiong@whamcloud.com>
Reviewed-by: Yu Jian <yujian@whamcloud.com>
Tested-by: Maloo <whamcloud.maloo@gmail.com>
Reviewed-by: Oleg Drokin <green@whamcloud.com>

LU-1157 ldlm: replace waiting flock lists by hashes

replace per-export list by per-export hash to locate a lock with
blocking export & owner.

Change-Id: I9c4089579bbf126781e232ea7021317fd10223e9
Xyratex-Bug-ID: MRP-385
Reviewed-by: Andriy Skulysh <andriy_skulysh@xyratex.com>
Signed-off-by: Vitaly Fertman <vitaly_fertman@xyratex.com>
Reviewed-on: http://review.whamcloud.com/2240
Tested-by: Hudson
Tested-by: Maloo <whamcloud.maloo@gmail.com>
Reviewed-by: Oleg Drokin <green@whamcloud.com>

LU-56 lnet: cleanup for rtrpool and LNet counter

This patch covered a few of things:
- code cleanup for router buffer pools
- code cleanup for error handling in lnet_prepare()
- code cleanup for LNet counters

This is an intermediate patch for LNet SMP improvements.

Signed-off-by: Liang Zhen <liang@whamcloud.com>
Change-Id: I554d6acb79a55dd77f709d3b6633f157f50a8cee
Reviewed-on: http://review.whamcloud.com/3091
Reviewed-by: Doug Oucharek <doug@whamcloud.com>
Tested-by: Hudson
Tested-by: Maloo <whamcloud.maloo@gmail.com>
Reviewed-by: Bobi Jam <bobijam@whamcloud.com>
Reviewed-by: Oleg Drokin <green@whamcloud.com>

LU-56 lnet: Partitioned LNet resources (ME/MD/EQ)

We already have a new lock lnet_res_lock to protect LNet resources,
but it's still a global lock and could have performance issue.
This patch created partitioned data for LNet, resources are
spreaded into different partitions. Also, lnet_res_lock is not
a single spinlock anymore, it's a percpt lock now, which means
LNet only needs to lock one partition at a time while operating
MD/ME belonging to that partition.

There are a few things are still serialized by exclusive lock:
- EQ allocation/free
- LNetEQPoll (non-zero size EQ)
- delay message on lazy portal
- Steaing MD between partitions.

There operations are either rare or deprecated so they shouldn't
become performance problem.

Signed-off-by: Liang Zhen <liang@whamcloud.com>
Change-Id: If5e88b92dd508b84c0fd91725b3aaed424dd3108
Reviewed-on: http://review.whamcloud.com/3078
Reviewed-by: Bobi Jam <bobijam@whamcloud.com>
Tested-by: Hudson
Reviewed-by: Doug Oucharek <doug@whamcloud.com>
Tested-by: Maloo <whamcloud.maloo@gmail.com>
Reviewed-by: Oleg Drokin <green@whamcloud.com>

LU-56 ptlrpc: cleanup of ptlrpc_unregister_service

This patch is only a cleanup of ptlrpc_unregister_service(),
it's an intermediate patch for partitioned ptlrpc service.

Signed-off-by: Liang Zhen <liang@whamcloud.com>
Change-Id: I334fa3813f77711defd598bbb7f688d3ea6026be
Reviewed-on: http://review.whamcloud.com/2917
Tested-by: Hudson
Tested-by: Maloo <whamcloud.maloo@gmail.com>
Reviewed-by: wangdi <di.wang@whamcloud.com>
Reviewed-by: Andreas Dilger <adilger@whamcloud.com>

LU-56 ptlrpc: svc thread starting/stopping cleanup

This patch covered two things:
- serialize creation of ptlrpc service thread
  In current version we can parallel create service threads, so there
  could be "hole" of thread ID if one creation failed, it could be
  problemaic because some modules require thread ID to be strictly
  contiguous and unique. Serializing thread creation can resolve this
  issue.
- code cleanup for for stopping of ptlrpc servcie threads, this is
  just for the next step work of partitioned ptlrpc service.

Signed-off-by: Liang Zhen <liang@whamcloud.com>
Change-Id: Ied8ad89003aa9d53fa73a4e5166a2c8d07a1aae9
Reviewed-on: http://review.whamcloud.com/2912
Tested-by: Hudson
Tested-by: Maloo <whamcloud.maloo@gmail.com>
Reviewed-by: wangdi <di.wang@whamcloud.com>
Reviewed-by: Andreas Dilger <adilger@whamcloud.com>

LU-506 kernel: 2.6.38 kernel macro check cleanup

* HAVE_DCACHE_LOCK is used to mark kernel >= 2.6.38, instead of other
kernel macros, this makes it easier to maintain kernel support in
the future.
* minor cleanups.

Signed-off-by: Lai Siyao <laisiyao@whamcloud.com>
Change-Id: Ic9a6a8c339a10b34e2b1b06b47fd9e252cd3ec2c
Reviewed-on: http://review.whamcloud.com/3159
Reviewed-by: Yang Sheng <ys@whamcloud.com>
Tested-by: Hudson
Tested-by: Maloo <whamcloud.maloo@gmail.com>
Reviewed-by: Andreas Dilger <adilger@whamcloud.com>
Reviewed-by: Fan Yong <yong.fan@whamcloud.com>

LU-723 ldiskfs: Drop support for ext3 based ldiskfs

Building ldiskfs from ext3 is no longer supported, thus this change
attempts to clean things up and only support building ldiskfs from
ext4 sources. The main changes made, involved replacing the @BACKFS@
autoconf variable with 'ext4', and removing the HAVE_EXT4_LDISKFS
macro. It is now safe to assume that ldiskfs is build based entirely
on ext4.

Signed-off-by: Prakash Surya <surya1@llnl.gov>
Change-Id: Ia0eff6e4c1b755806ccfbf554e4a36201f9f1a64
Reviewed-on: http://review.whamcloud.com/1643
Tested-by: Hudson
Reviewed-by: Andreas Dilger <adilger@whamcloud.com>
Tested-by: Maloo <whamcloud.maloo@gmail.com>
Reviewed-by: Oleg Drokin <green@whamcloud.com>

LU-1194 llog: fix for not sync llcd at thread stop

If llog_obd_repl_cancel() happend between llog_sync() and
class_import_put() at filter_llog_finish(), llog_recov_thread_stop()
throw LBUG. This patch fix this issue by adding new flags to llog_ctxt.

Signed-off-by: Alexander Boyko <alexander_boyko@xyratex.com>
Reviewed-by: Andriy Skulysh <andriy_skulysh@xyratex.com>
Reviewed-by: Alexander Zarochentsev <alexander_zarochentsev@xyratex.com>
Xyratex-bug-id: MRP-456
Change-Id: Ife79adfe6cde0f2090776cd27cd87f65c1e988e2
Reviewed-on: http://review.whamcloud.com/2789
Reviewed-by: Andriy Skulysh <Andriy_Skulysh@xyratex.com>
Tested-by: Hudson
Tested-by: Maloo <whamcloud.maloo@gmail.com>
Reviewed-by: Mike Pershin <tappro@whamcloud.com>
Reviewed-by: Oleg Drokin <green@whamcloud.com>

LU-56 lnet: reduce stack usage of "match" functions

Use new structure lnet_match_info to transfer the parameters
of LNet "match" functions and reduce stack usage.

This is an intermediate patch for LNet SMP improvements.

Signed-off-by: Liang Zhen <liang@whamcloud.com>
Change-Id: I710a78c58add8609606f5d6de1f975ffc5200439
Reviewed-on: http://review.whamcloud.com/3070
Reviewed-by: Doug Oucharek <doug@whamcloud.com>
Tested-by: Hudson
Tested-by: Maloo <whamcloud.maloo@gmail.com>
Reviewed-by: Bobi Jam <bobijam@whamcloud.com>
Reviewed-by: Oleg Drokin <green@whamcloud.com>

LU-56 lnet: Granulate LNet lock

LNet is using a global lock LNET_LOCK to serialize all operations
and event callbacks of LNet, it's a big performance issue on fat
SMP machines because of high lock contention.

We have submitted many changes to separate critical logic of LNet
and this patch is the key step for finer-grained LNet locking.
This patch add a new lock "lnet_res_lock", all operations on LNet
resources (ME, MD, EQ) are under protection of this lock, we still
keep LNET_LOCK so far, but it's only called for serializing
operations on NI, peer, credits and routers.

This is still an intermediate patch for LNet SMP improvements, both
LNET_LOCK and lnet_res_lock are just spinlock now, they will be
replaced by percpt lock in upcoming patches.

Signed-off-by: Liang Zhen <liang@whamcloud.com>
Change-Id: I313caffd21776ee3474c2a1391ea78f002b47790
Reviewed-on: http://review.whamcloud.com/3056
Reviewed-by: Doug Oucharek <doug@whamcloud.com>
Reviewed-by: Bobi Jam <bobijam@whamcloud.com>
Tested-by: Hudson
Tested-by: Maloo <whamcloud.maloo@gmail.com>
Reviewed-by: Lai Siyao <laisiyao@whamcloud.com>
Reviewed-by: Oleg Drokin <green@whamcloud.com>

LU-1305 osd: object operations for osd-zfs

allocate/release, create/destroy, attr set/get, etc

Signed-off-by: Alex Zhuravlev <bzzz@whamcloud.com>
Change-Id: I4e56a175b6fd5008d89cdd509b47f8bc00ce5552
Reviewed-on: http://review.whamcloud.com/2966
Tested-by: Hudson
Tested-by: Maloo <whamcloud.maloo@gmail.com>
Reviewed-by: Fan Yong <yong.fan@whamcloud.com>
Reviewed-by: Andreas Dilger <adilger@whamcloud.com>

LU-1305 osd: object index for osd-zfs

functionality to map fids to dnodes

Signed-off-by: Alex Zhuravlev <bzzz@whamcloud.com>
Change-Id: If8153907342a8c0ac3fe8f657f7dc740db6b2753
Reviewed-on: http://review.whamcloud.com/2964
Tested-by: Hudson
Tested-by: Maloo <whamcloud.maloo@gmail.com>
Reviewed-by: Fan Yong <yong.fan@whamcloud.com>
Reviewed-by: Andreas Dilger <adilger@whamcloud.com>

LU-1305 osd: osd_handler.c for osd-zfs

functions to setup/mount zfs backend

Signed-off-by: Alex Zhuravlev <bzzz@whamcloud.com>
Change-Id: I200709e0c86e4f7aae35528cbae7f7b08e094f47
Reviewed-on: http://review.whamcloud.com/2963
Tested-by: Hudson
Reviewed-by: Andreas Dilger <adilger@whamcloud.com>

LU-56 ptlrpc: partition data for ptlrpc service

We will have multiple partition data & threads for ptlrpc service,
this patch is the first step work, we moved quite a lot members
of ptlrpc_service to a new structure ptlrpc_service_part.
Now we only create one instance of ptlrpc_service_part for each
service, but we will have multiple instances for each service
very soon (instance per CPT, CPU ParTion).

Signed-off-by: Liang Zhen <liang@whamcloud.com>
Change-Id: I63d816bdf44a22528c6097fe348060f57d862df3
Reviewed-on: http://review.whamcloud.com/2895
Tested-by: Hudson
Reviewed-by: wangdi <di.wang@whamcloud.com>
Tested-by: Maloo <whamcloud.maloo@gmail.com>
Reviewed-by: Andreas Dilger <adilger@whamcloud.com>

LU-957 scrub: trigger OI scrub if found bad OI entry

If some RPC involves OI lookup and finds inconsistent OI mapping,
it should trigger OI scrub to check and repair the inconsistency.

Known issues:
When the fid is returned to client, the OI mapping corresponding
to such fid may be not updated, or not committed to disk yet. If
server crashed before OI scrub completed, then recovery with the
fid corresponding to inconsistent OI mapping may fail or blocked.

Signed-off-by: Fan Yong <yong.fan@whamcloud.com>
Change-Id: I9709386aa6d42954b619f6b1342adae59a2ec5a9
Reviewed-on: http://review.whamcloud.com/2554
Tested-by: Hudson
Tested-by: Maloo <whamcloud.maloo@gmail.com>
Reviewed-by: Alex Zhuravlev <bzzz@whamcloud.com>
Reviewed-by: Andreas Dilger <adilger@whamcloud.com>

LU-1459 llite: Don't LBUG when import has LUSTRE_IMP_NEW state

When a disabled OSC/OST is configured in the system at mount
time, a client will LBUG if calling "lfs check servers".
Disabling the LBUG causes client to return -EIO instead.

Signed-off-by: Jeremy Filizetti <jeremy.filizetti@gmail.com>
Signed-off-by: Bobi Jam <bobijam@whamcloud.com>
Change-Id: I1844b66e56259da28129df2c60d2542e9c95aeee
Reviewed-on: http://review.whamcloud.com/2998
Tested-by: Hudson
Tested-by: Maloo <whamcloud.maloo@gmail.com>
Reviewed-by: Fan Yong <yong.fan@whamcloud.com>
Reviewed-by: Andreas Dilger <adilger@whamcloud.com>

LU-1493 quota: extra release caused by race

There is a race between the check_cur_qunit() and the
dqacq_completion(): check_cur_qunit() read hardlimit
and calculate how much quota need be acquired/released
based on the hardlimit, however, the hardlimit can be
changed by the dqacq_completion() at anytime. So that
could result in extra quota acquire/release when there
is inflight dqacq.

In general, such extra dqacq dosen't bring fatal error,
unless an extra release is going to release more than
'hardlimit' quota.

To minimize the code changes (anyway, it'll be totally
rewritten in the new quota design), we just do one more
check here to avoid the extra release which could bring
fatal error. A better solution could be calculating the
qd_count here and removing the lqs_blk/ino_rec stuff.

Signed-off-by: Niu Yawei <niu@whamcloud.com>
Change-Id: I0ad5ff0f32e39f32872c201ad1d545fbd9d1a57d
Reviewed-on: http://review.whamcloud.com/3074
Tested-by: Hudson
Tested-by: Maloo <whamcloud.maloo@gmail.com>
Reviewed-by: Johann Lombardi <johann@whamcloud.com>
Reviewed-by: Fan Yong <yong.fan@whamcloud.com>

LU-1438 quota: quota active checking is missed on slave

On quota slave, we missed checking if quota is enabled in the
quota_check_common() and several other places. Which could cause
slave retry acquire quota in quota_chk_acq_common() infinitely
when the quota is already turned off on master.

Signed-off-by: Niu Yawei <niu@whamcloud.com>
Change-Id: Iaa48c7cca05daf595b6d3b7e4025c7650e460918
Reviewed-on: http://review.whamcloud.com/3097
Tested-by: Hudson
Reviewed-by: Fan Yong <yong.fan@whamcloud.com>
Tested-by: Maloo <whamcloud.maloo@gmail.com>
Reviewed-by: Johann Lombardi <johann@whamcloud.com>

LU-957 scrub: osd-ldiskfs itable based iteration

Implement inode table based object iteration in osd-ldiskfs.
It is implemented as DT iteration APIs, which can be used by
up layer LFSCK to scan the whole device sequentially.

Signed-off-by: Fan Yong <yong.fan@whamcloud.com>
Change-Id: I3d80bb4a174d47429764e5cca35e4f07be52d50b
Reviewed-on: http://review.whamcloud.com/2553
Tested-by: Hudson
Tested-by: Maloo <whamcloud.maloo@gmail.com>
Reviewed-by: Alex Zhuravlev <bzzz@whamcloud.com>
Reviewed-by: Andreas Dilger <adilger@whamcloud.com>

LU-56 lnet: code cleanup for lib-move.c

Most changes in this patch are just code cleanup:
- remove one unnecessary lock dance for message forwarding
- move some code blocks to make functions cleaner
- rename lnet_ni_peer_alive to lnet_ni_query_locked

It's an intermediate patch LNet SMP improvements.

Signed-off-by: Liang Zhen <liang@whamcloud.com>
Change-Id: Ia7cfe37a5a4f896be4fcc7f6c5cf9c27268de9ba
Reviewed-on: http://review.whamcloud.com/3048
Tested-by: Hudson
Tested-by: Maloo <whamcloud.maloo@gmail.com>
Reviewed-by: Doug Oucharek <doug@whamcloud.com>
Reviewed-by: Bobi Jam <bobijam@whamcloud.com>
Reviewed-by: Andreas Dilger <adilger@whamcloud.com>

LU-56 lnet: match-table for Portals

Create sub-object named as "match-table" for each Portal, MEs will
be attached match-table instead of Portal.
Although we only have one match-table for each Portal in this patch,
but in upcoming changes, we will create multiple match-tables
for each Portal:
- unique-match Portal
  MEs will be scattered to different match-tables by match info
- wildcard Portal
  LND threads just grab ME/MD from match-table corresponding to
  current CPT (CPU partition).

We also did some code cleanup for delayed message in this patch.

Signed-off-by: Liang Zhen <liang@whamcloud.com>
Change-Id: I2b24723c3bd2a6664f2b241840de19d5f43be11f
Reviewed-on: http://review.whamcloud.com/3043
Reviewed-by: Doug Oucharek <doug@whamcloud.com>
Tested-by: Hudson
Tested-by: Maloo <whamcloud.maloo@gmail.com>
Reviewed-by: Bobi Jam <bobijam@whamcloud.com>
Reviewed-by: Andreas Dilger <adilger@whamcloud.com>

LU-56 lnet: code cleanup for lib-md.c

This patch just did some code cleanup for lib-md.c, it's an
intermediate patch for LNet SMP improvements.

Signed-off-by: Liang Zhen <liang@whamcloud.com>
Change-Id: I5d162852008284ae08f3b9e7935c9251b245fb5c
Reviewed-on: http://review.whamcloud.com/3010
Tested-by: Hudson
Reviewed-by: Doug Oucharek <doug@whamcloud.com>
Tested-by: Maloo <whamcloud.maloo@gmail.com>
Reviewed-by: Bobi Jam <bobijam@whamcloud.com>
Reviewed-by: Andreas Dilger <adilger@whamcloud.com>

LU-56 lnet: split lnet_commit_md and cleanup

Briefly, LNet has two types of operations: 1) resource operations
on EQ, MD and ME; 2) network operations on credits, peer, NI etc.
Most of them are in different logic blocks, but there are still
a few functions have mixed operations, lnet_commit_md() is one of
these functions.

This patch is a key step for LNet SMP improvements, lnet_commit_md()
of current LNet did two things: 1) consume MD threshold and commit
it to network; 2) attach the message on active message list to show
it might have taken LNet message credits. This patch split it into
two sets of functions, lnet_msg_attach/detach_md and
lnet_msg_commit/decommit,

We need to call lnet_msg_attach_md and lnet_msg_comit before deliver
a message to network, call lnet_msg_detach_md and lnet_msg_decommit
while finalizing a message. These two sets of functions will be
protected by different locks in upcoming patches.

Another change in this patch is, we moved most counters operations
into lnet_msg_decommit, this will make code cleaner and easier for
the next step work. The reason that we put counters operations in
lnet_msg_decommit instead of lnet_msg_commit is because we can't
guarantee lnet_msg_commit is always called after lnet_msg_attach_md,
and counters need information filled by lnet_msg_attach_md, which
means lnet_msg_commit might not have enough information for counters.
The thing we know for sure is lnet_msg_decommit is always called
after lnet_msg_attach_md, so it's safe to put counter operations
inside lnet_msg_decommit.

Signed-off-by: Liang Zhen <liang@whamcloud.com>
Change-Id: Iee961fbff6e6e39ed76c14c0bd5b6560a7730af9
Reviewed-on: http://review.whamcloud.com/2997
Tested-by: Hudson
Reviewed-by: Doug Oucharek <doug@whamcloud.com>
Reviewed-by: Bobi Jam <bobijam@whamcloud.com>
Tested-by: Maloo <whamcloud.maloo@gmail.com>
Reviewed-by: Andreas Dilger <adilger@whamcloud.com>

LU-1095 debug: fix missing CDEBUG() newline

The console message cleanup 389fde827be2ee6fb4ee08e955d773a2a16e70c6
lost a newline in one of the messages, which itself generates a
warning on the console.

Signed-off-by: Andreas Dilger <adilger@whamcloud.com>
Change-Id: Ie5eed2977e496e082d5b8e62bfc39e0df93fcab0
Reviewed-on: http://review.whamcloud.com/3134
Reviewed-by: Li Wei <liwei@whamcloud.com>
Tested-by: Hudson
Tested-by: Maloo <whamcloud.maloo@gmail.com>

LU-1030 test: delete test file at the end of test script

At sanity.sh:test_49, we need to delete the test file after test is
finished otherwise the following test cases may fail due to lack of
storage capacity.

Signed-off-by: Jinshan Xiong <jinshan.xiong@whamcloud.com>
Change-Id: Iec9e93f1a4844156345608bf9f7a9bf679b36253
Reviewed-on: http://review.whamcloud.com/3158
Reviewed-by: Yu Jian <yujian@whamcloud.com>
Tested-by: Hudson
Reviewed-by: Andreas Dilger <adilger@whamcloud.com>
Tested-by: Maloo <whamcloud.maloo@gmail.com>

LU-1305 osd: procfs support for osd-zfs

procfs support

Signed-off-by: Alex Zhuravlev <bzzz@whamcloud.com>
Change-Id: I599d1317f762b50965f760a590f2f99ffd86bef5
Reviewed-on: http://review.whamcloud.com/2965
Tested-by: Hudson
Tested-by: Maloo <whamcloud.maloo@gmail.com>
Reviewed-by: Fan Yong <yong.fan@whamcloud.com>
Reviewed-by: Andreas Dilger <adilger@whamcloud.com>

LU-1305 obdclass: lprocfs helpers

few lprocfs helpers to dump statfs data via OSD API

Signed-off-by: Alex Zhuravlev <bzzz@whamcloud.com>
Change-Id: If90c410c9155f480baa8d344a670cc1a48077f07
Reviewed-on: http://review.whamcloud.com/3089
Tested-by: Hudson
Tested-by: Maloo <whamcloud.maloo@gmail.com>
Reviewed-by: Andreas Dilger <adilger@whamcloud.com>

LU-56 lnet: LNet message event cleanup

In current version, LNet message events are assigned at many places,
it's not easy to maintain, also, assignments of some fields of event
need to be protected by lock and some don't need, this will be a
problem for upcoming LNet SMP improvements.
This patch created a new function lnet_build_msg_event(), all events
are assigned inside this function and it can be called w/o lock
protection.
This is an intermediate patch for LNet SMP improvements.

Signed-off-by: Liang Zhen <liang@whamcloud.com>
Change-Id: Ief8722a15d17909e3844149c55d24fb22ed0c04c
Reviewed-on: http://review.whamcloud.com/2983
Reviewed-by: Bobi Jam <bobijam@whamcloud.com>
Reviewed-by: Doug Oucharek <doug@whamcloud.com>
Tested-by: Hudson
Tested-by: Maloo <whamcloud.maloo@gmail.com>
Reviewed-by: Andreas Dilger <adilger@whamcloud.com>

LU-56 lnet: eliminate a few locking dance in LNet

This patch gets rid of a few unnecessary locking dance from
LNet, also changed type of lnet_ni_t::ni_last_alive from jiffy to
second because per-second checking should be good enough.

Signed-off-by: Liang Zhen <liang@whamcloud.com>
Change-Id: I0e3269f0a5ae0cfecd9611584d5e213a670e38f6
Reviewed-on: http://review.whamcloud.com/2933
Tested-by: Hudson
Reviewed-by: Doug Oucharek <doug@whamcloud.com>
Tested-by: Maloo <whamcloud.maloo@gmail.com>
Reviewed-by: Bobi Jam <bobijam@whamcloud.com>
Reviewed-by: Andreas Dilger <adilger@whamcloud.com>

LU-56 lnet: parse RC ping in event callback

In current LNet, we don't parse router-checker ping in EQ callback,
NI information from remote side is attached on peer and it's parsed
for each sending. This is OK for now but with upcoming SMP
improvements, the global LNET_LOCK is removed and EQ callback and
peer/router are proteced by different locks, which means sink buffer
of RC ping could be in progressing of change when other thread is
reading it in lnet_send(), and get wrong information.

This patch will parse ping information in EQ callback, and store it
in a new member lnet_route_t::lr_downis which is just a integer,
although reader in lnet_send() still don't have lock while reading
it, but it's OK because it's just an integer and reader can
atomically read reliable result even w/o any lock.

It should also be a little faster than original implememtation
because we don't have to parse RC ping, which should be updated
once for many seconds, for each sending.

Signed-off-by: Liang Zhen <liang@whamcloud.com>
Change-Id: I5b89fabfd22f22ba171ef63d30da2d32e2788b16
Reviewed-on: http://review.whamcloud.com/2932
Reviewed-by: Doug Oucharek <doug@whamcloud.com>
Tested-by: Hudson
Tested-by: Maloo <whamcloud.maloo@gmail.com>
Reviewed-by: Bobi Jam <bobijam@whamcloud.com>
Reviewed-by: Andreas Dilger <adilger@whamcloud.com>

LU-56 lnet: router-checker (RC) cleanup

This patched covered a couple of things:
- remove global router-checker data
  In current LNet we can have both global RC data and per-peer
  RC data, the later is only for asymmetrical router failure so far,
  however we can use per-peer RC data even asym_router_failure is
  disabled, it's just increase a little memory overhead but will
  simplify router checker.
  Also, I'm thinking that asym_router_failure is a nice feature and
  it should be enabled by default in the future.

- Code cleanup for router checker, again, it's still an intermediate
  patch for LNet SMP improvements, we need many code cleanup before
  we change locking mechanism of LNet.

Signed-off-by: Liang Zhen <liang@whamcloud.com>
Change-Id: I830e706ce5099f15f122b521f74b6dfcde443592
Reviewed-on: http://review.whamcloud.com/2930
Tested-by: Hudson
Reviewed-by: Doug Oucharek <doug@whamcloud.com>
Tested-by: Maloo <whamcloud.maloo@gmail.com>
Reviewed-by: Bobi Jam <bobijam@whamcloud.com>
Reviewed-by: Andreas Dilger <adilger@whamcloud.com>

LU-1534 osd: Fix LBUGs when destroying IGIF objects

When destroying MDT objects with IGIF FIDs, the following assertion
may occur:

  LustreError: 28147:0:(osd_internal.h:665:osd_fid2oi()) ASSERTION(
  !fid_is_igif(fid) ) failed:
  LustreError: 28147:0:(osd_internal.h:665:osd_fid2oi()) LBUG
  Pid: 28147, comm: mdt_00

With call traces like this:

  osd_oi_delete()
  osd_object_destroy()
  mdd_object_kill()
  mdd_finish_unlink()
  mdd_unlink()

This patch fixes osd_oi_delete() to avoid trying to delete IGIF FIDs
from OI.  Also, the assertion is enhanced to print more debugging
information.

Signed-off-by: Li Wei <liwei@whamcloud.com>
Change-Id: I95fe9db8aec53a496a8dc84c87594ed81f3fe106
Signed-off-by: Li Wei <liwei@whamcloud.com>
Reviewed-on: http://review.whamcloud.com/3120
Reviewed-by: Fan Yong <yong.fan@whamcloud.com>
Reviewed-by: Andreas Dilger <adilger@whamcloud.com>
Tested-by: Hudson

2.2.57 tag

Change-Id: I75ef5cf27c4cd7c0c4dd4b1885f4838b5da39002

LU-1428 ldlm: fix a race in ldlm_lock_destroy_internal

ldlm_lock::l_exp_hash should be protected by internal lock of
cfs_hash, but we called cfs_hlist_unhashed(lock::l_exp_hash)
w/o holding cfs_hash lock in ldlm_lock_destroy_internal,
which means if someone called ldlm_lock_cancel on a lock while
export::exp_lock_hash is in progress of rehashing (thread context of
cfs_workitem), there could be tiny window between deleting this lock
from bucket[A] and re-adding it to bucket[B] of l_exp_hash, and
cfs_hlist_unhashed(lock::l_exp_hash) will return 1 in this window,
then we destroyed a lock but left it on l_exp_hash forever because
lock::l_destroyed has been set to 1 and ldlm_lock_destroy_internal()
wouldn't be able to remove the lock from l_exp_hash even it's called
infinite times in ldlm_cancel_locks_for_export_cb().

This patch also added some debug information to
ldlm_cancel_locks_for_export_cb in case this patch can't fix this
problem.

Signed-off-by: Liang Zhen <liang@whamcloud.com>
Change-Id: Ia0932658b3f085a55535e36bee4fb833e74fa242
Reviewed-on: http://review.whamcloud.com/3028
Tested-by: Hudson
Reviewed-by: Fan Yong <yong.fan@whamcloud.com>
Tested-by: Maloo <whamcloud.maloo@gmail.com>
Reviewed-by: Lai Siyao <laisiyao@whamcloud.com>
Reviewed-by: Andreas Dilger <adilger@whamcloud.com>

LU-1518 mdd: Revert "Fixup mdd_{obf,dot_lustre}_obj_ops."

This reverts commit cf0d73004a20923c690591c1f3f61a8c872ad86e

Reverting this patch due to build failure (moo_version_get not defined).

Change-Id: I7cac5c6d6de2924d46c3a74ed4aa138f50501a16
Reviewed-on: http://review.whamcloud.com/3137
Reviewed-by: Johann Lombardi <johann@whamcloud.com>
Tested-by: Johann Lombardi <johann@whamcloud.com>

LU-1371 ldiskfs: fix osd_it_iam_rec() to support binary keys

osd_it_iam_rec() assumes a string key and does not support binary key.

Signed-off-by: Johann Lombardi <johann@whamcloud.com>
Change-Id: I1a09c9b9ad53e593cad642b6ae8a7d417ea83c9d
Reviewed-on: http://review.whamcloud.com/2647
Tested-by: Hudson
Tested-by: Maloo <whamcloud.maloo@gmail.com>
Reviewed-by: Liang Zhen <liang@whamcloud.com>
Reviewed-by: wangdi <di.wang@whamcloud.com>
Reviewed-by: Andreas Dilger <adilger@whamcloud.com>

LU-1518 mdd: Fixup mdd_{obf,dot_lustre}_obj_ops.

Define several missing md_object ops for .lustre/fid. Unify
attribute handling for .lustre with that of normal md_objects.

Change-Id: I892904af4ef01b3687e60a41250b018d67c86734
Signed-off-by: John L. Hammond <jhammond@tacc.utexas.edu>
Reviewed-on: http://review.whamcloud.com/3103
Tested-by: Hudson
Reviewed-by: Andreas Dilger <adilger@whamcloud.com>
Reviewed-by: wangdi <di.wang@whamcloud.com>
Reviewed-by: Lai Siyao <laisiyao@whamcloud.com>
Reviewed-by: Alex Zhuravlev <bzzz@whamcloud.com>
Tested-by: Maloo <whamcloud.maloo@gmail.com>

LU-988 ost: add debugging for inode timestamps

Intermittent test failures for sanity.sh test_39j are showing the
inode timestamps going backward. Add internal debugging to track the
LVB timestamps to see where the old timestamps are coming from, to
determine if this is a problem with the VM environment being used for
testing, or if it is a genuine problem in the Lustre code.

Signed-off-by: Bobi Jam <bobijam@whamcloud.com>
Change-Id: I590d8b28b7a37010af6a43d329917022b840d074
Reviewed-on: http://review.whamcloud.com/3069
Tested-by: Hudson
Tested-by: Maloo <whamcloud.maloo@gmail.com>
Reviewed-by: Lai Siyao <laisiyao@whamcloud.com>
Reviewed-by: Fan Yong <yong.fan@whamcloud.com>
Reviewed-by: Andreas Dilger <adilger@whamcloud.com>

LU-1305 libcfs: S_OSD subsystem

new definition for osd components

Signed-off-by: Alex Zhuravlev <bzzz@whamcloud.com>
Change-Id: Ia224dbd09caecb5cc31488627bb6a958b680377a
Reviewed-on: http://review.whamcloud.com/3088
Tested-by: Hudson
Reviewed-by: Andreas Dilger <adilger@whamcloud.com>
Tested-by: Maloo <whamcloud.maloo@gmail.com>

LU-1030 osc: new IO engine implementation

New IO engine to manage dirty pages with osc_extent.

Osc_extent is a data structure to manage a series of contiguous
blocks; however, the pages in an extent is not required to be
contiguous. An extent must be written out in one RPC.

The purpose of introducing extents are:
1. make grants work for extent-based OSD;
2. form better IO by picking up contiguous pages to compose RPC;
3. reimplement ll_writepages() with CIT_FSYNC.

Signed-off-by: Jinshan Xiong <jinshan.xiong@whamcloud.com>
Change-Id: I3ef619c1f07eefd201236ab55e5fd858791d41e0
Reviewed-on: http://review.whamcloud.com/2270
Reviewed-by: Johann Lombardi <johann@whamcloud.com>
Tested-by: Hudson
Tested-by: Maloo <whamcloud.maloo@gmail.com>
Reviewed-by: Andreas Dilger <adilger@whamcloud.com>

LU-1014 mountconf: Revert "MGS should process parameter config"

This reverts commit 6cde8692384a1d3a6d0e8114629fa943c951fe40.

Previous commit is wrong, will fix it in another patch.

Signed-off-by: Lai Siyao <laisiyao@whamcloud.com>
Change-Id: I5868cb1908f9b123db5b8bd7be9e364ae9a51d6a
Reviewed-on: http://review.whamcloud.com/2843
Tested-by: Hudson
Reviewed-by: Andreas Dilger <adilger@whamcloud.com>
Reviewed-by: James Simmons <uja.ornl@gmail.com>
Tested-by: Maloo <whamcloud.maloo@gmail.com>

LU-1305 osd: quota support for osd-zfs

export of accouting information and enforcement

Signed-off-by: Alex Zhuravlev <bzzz@whamcloud.com>
Change-Id: Iac5100960a5107277ccd628ba3a682cb9b03c0ed
Reviewed-on: http://review.whamcloud.com/2970
Reviewed-by: Johann Lombardi <johann@whamcloud.com>
Tested-by: Hudson
Tested-by: Maloo <whamcloud.maloo@gmail.com>
Reviewed-by: Andreas Dilger <adilger@whamcloud.com>

LU-1305 defs: zfs definitions

couple definitions required for osd-zfs

Signed-off-by: Alex Zhuravlev <bzzz@whamcloud.com>
Change-Id: I994ab7353edef7f4ccc39707534e7c262e635af3
Reviewed-on: http://review.whamcloud.com/2962
Tested-by: Hudson
Tested-by: Maloo <whamcloud.maloo@gmail.com>
Reviewed-by: Fan Yong <yong.fan@whamcloud.com>
Reviewed-by: Andreas Dilger <adilger@whamcloud.com>

LU-1406 ofd: add OBD methods to handle OST requests

Finally add OBD handlers to process incoming requests.
Add stub for quota functions, quota is non-functional in OFD.
Fix couple minor issues:
- slot checking in ofd_last_rcvd_update() must check lr_idx
- remove osd_write_locked() assertion from __osd_xattr_set(),
it is too excessive there

Signed-off-by: Mikhail Pershin <tappro@whamcloud.com>
Change-Id: Iab270cdf5f2ad41df86845190c5c6db3436279c8
Reviewed-on: http://review.whamcloud.com/2894
Tested-by: Hudson
Reviewed-by: Andreas Dilger <adilger@whamcloud.com>
Tested-by: Maloo <whamcloud.maloo@gmail.com>
Reviewed-by: Alex Zhuravlev <bzzz@whamcloud.com>

LU-1406 ofd: IO operations

add IO functions to OFD

Signed-off-by: Mikhail Pershin <tappro@whamcloud.com>
Change-Id: Ie78e92f5f770ff07c94bf87b2d5ee8300673d271
Reviewed-on: http://review.whamcloud.com/2893
Reviewed-by: Andreas Dilger <adilger@whamcloud.com>
Tested-by: Hudson
Tested-by: Maloo <whamcloud.maloo@gmail.com>
Reviewed-by: Alex Zhuravlev <bzzz@whamcloud.com>

LU-1406 ofd: add DLM and LVB code

OFD dlm and lvb functions

Signed-off-by: Mikhail Pershin <tappro@whamcloud.com>
Change-Id: Ieddd2b16129f62c6e380d007aaa1f2de70e8a9ca
Reviewed-on: http://review.whamcloud.com/2892
Tested-by: Hudson
Reviewed-by: Andreas Dilger <adilger@whamcloud.com>
Tested-by: Maloo <whamcloud.maloo@gmail.com>
Reviewed-by: Alex Zhuravlev <bzzz@whamcloud.com>

LU-1305 osd: lnb_file_offset in struct niobuf_local

introduced in orion so that every buffer can be marked
with a specific offset.

Signed-off-by: Alex Zhuravlev <bzzz@whamcloud.com>
Change-Id: Ic8c69fd401300107d589a093023ffe41c0f0eddd
Reviewed-on: http://review.whamcloud.com/2884
Tested-by: Hudson
Reviewed-by: Andreas Dilger <adilger@whamcloud.com>
Tested-by: Maloo <whamcloud.maloo@gmail.com>

LU-1305 osd: new index flag DT_IND_RANGE

introduced in orion to recognize whether the index support
range lookups (used by fldb)

Signed-off-by: Alex Zhuravlev <bzzz@whamcloud.com>
Change-Id: I40efa613697ebe8394557d5f687132ca9f542859
Reviewed-on: http://review.whamcloud.com/2883
Tested-by: Hudson
Tested-by: Maloo <whamcloud.maloo@gmail.com>
Reviewed-by: Andreas Dilger <adilger@whamcloud.com>

2.3.56 tag

Change-Id: Ia76ec8d6266348c1135292d6de4ded530e6a3889

LU-1031: add grouplock test to sanity.sh

add a simple grouplock test to sanity.sh (test_82)
replace hardcoded use of multiop by a $MULTIOP in
test scripts.

Change-Id: I94ba2248b7efafe258873c468e6d0cabdfcb7d09
Signed-off-by: jcl <jacques-charles.lafoucriere@cea.fr>
Reviewed-on: http://review.whamcloud.com/2098
Tested-by: Hudson
Tested-by: Maloo <whamcloud.maloo@gmail.com>
Reviewed-by: Oleg Drokin <green@whamcloud.com>

LU-1513 libcfs: call cfs_alloc with proper masks

Do not use kernel GFP masks directly when call cfs_alloc,
instead, Lustre special flags can be used to control the
allocator's behavior.

cfs_alloc will convert these Lustre special flags to some
kernle GFP masks, then call kmalloc with these GFP masks.

Signed-off-by: Fan Yong <yong.fan@whamcloud.com>
Change-Id: I3189e143ff0cf65a08a1bdf2b8476ab151dc308e
Reviewed-on: http://review.whamcloud.com/3118
Reviewed-by: Oleg Drokin <green@whamcloud.com>
Tested-by: Hudson
Tested-by: Maloo <whamcloud.maloo@gmail.com>
Reviewed-by: Andreas Dilger <adilger@whamcloud.com>

LU-56 ptlrpc: common code to validate nthreads

Add a common function to validate ptlrpc service threads number,
instead of checking it before registration of each service.
This function (ptlrpc_server_nthreads_check) will be enhanced
soon for CPT affinity ptlrpc service.

Signed-off-by: Liang Zhen <liang@whamcloud.com>
Change-Id: I5db09396f4a95e36d8f5d38777c51eafb7dea5b1
Reviewed-on: http://review.whamcloud.com/2879
Tested-by: Hudson
Reviewed-by: Andreas Dilger <adilger@whamcloud.com>
Tested-by: Maloo <whamcloud.maloo@gmail.com>
Reviewed-by: wangdi <di.wang@whamcloud.com>
Reviewed-by: Oleg Drokin <green@whamcloud.com>

LU-482 tests: cleanup clients/OSTs after MDS error

If the MDS fails to mount due to LU-482, ensure that the clients
and OSTs are unmounted cleanly. Otherwise, the normal unmount
will hang on the clients as they wait for the MDS to be recovered.

Signed-off-by: Andreas Dilger <adilger@whamcloud.com>
Change-Id: Iccd8f81f714974063eaeec45ffbd80f9cef3242b
Reviewed-on: http://review.whamcloud.com/3019
Tested-by: Hudson
Tested-by: Maloo <whamcloud.maloo@gmail.com>
Reviewed-by: Oleg Drokin <green@whamcloud.com>

LU-827 tests: fix user ID for sanityn test_60

If sanityn.sh test_60() is run on a test system before any of the
quota users are configured, the test will fail. The test does not
require that any specific user ID is used, so instead use RUNAS_ID
(which is a numeric value and does not even need to be defined in
/etc/passwd for this test to work).

Signed-off-by: Andreas Dilger <adilger@whamcloud.com>
Change-Id: Iaca88344e0dc615f44417e4175e5cddd146c2a7a
Reviewed-on: http://review.whamcloud.com/3021
Tested-by: Hudson
Tested-by: Maloo <whamcloud.maloo@gmail.com>
Reviewed-by: Johann Lombardi <johann@whamcloud.com>
Reviewed-by: Yu Jian <yujian@whamcloud.com>

New tag 2.2.55

Change-Id: I61c9699e1a557d3d5c0848c5e8628f70d256c280
Signed-off-by: Oleg Drokin <green@whamcloud.com>

LU-1286 obdclass: oops in lprocfs_remove_nolock

Move lov /proc entry cleanup from lov_precleanup to
lov_cleanup, so the lov proc entry will be removed after
the pool has been removed, otherwise it will
cause oops during pool proc entry deletion.

Signed-off-by: Di Wang <di.wang@whamcloud.com>
Change-Id: I1d1b03d02348caaa1e421aa1de1535c8cf0ba3f4
Reviewed-on: http://review.whamcloud.com/2450
Tested-by: Hudson
Reviewed-by: Andreas Dilger <adilger@whamcloud.com>
Tested-by: Maloo <whamcloud.maloo@gmail.com>
Reviewed-by: Oleg Drokin <green@whamcloud.com>

LU-1421 clio: check cp_owner of cl_page after the page is owned

After page writeback is supported in clio, there exists a short
period that page is unlocked but cp_owner is set. This confuses
the assertion of cl_page_assume(). This problem can be fixed by
checking cp_owner after the page is owned.

Signed-off-by: Jinshan Xiong <jinshan.xiong@whamcloud.com>
Change-Id: Iafcdbb9b0dbb5f081a84c1da1b7ad8051d406988
Reviewed-on: http://review.whamcloud.com/3027
Tested-by: Hudson
Tested-by: Maloo <whamcloud.maloo@gmail.com>
Reviewed-by: Niu Yawei <niu@whamcloud.com>
Reviewed-by: Oleg Drokin <green@whamcloud.com>

LU-506 kernel: some other minor cleanup of BKL

Remove lock_kernel/unlock_kernel from cfs_daemonize.
Remove HAVE_UNLOCKED_IOCTL check as it is defined after 2.6.12.

Signed-off-by: Liu Xuezhao <xuezhao.liu@emc.com>
Change-Id: I6e41c0c3fa76c33dae4f820a63bd3ba62ff704ec
Reviewed-on: http://review.whamcloud.com/2950
Tested-by: Hudson
Reviewed-by: Lai Siyao <laisiyao@whamcloud.com>
Reviewed-by: Andreas Dilger <adilger@whamcloud.com>
Tested-by: Maloo <whamcloud.maloo@gmail.com>
Reviewed-by: Oleg Drokin <green@whamcloud.com>

LU-506 kernel: FC15 - support dcache scalability changes.

* dcache supports rcu-walk, but lustre doesn't, so
revalidate/check_acl/permission fall back to ref-walk.
* dcache_lock is completely removed in latest kernel, many
fields are protected by inode->i_lock now.
* remove super hack d_rehash_cond(), and treats
DCACHE_LUSTRE_INVALID similar to DCACHE_DISCONNECTED, therefore
dentry doesn't need to be dropped and rehashed frequently.
* .lookup(LOOKUP_CREATE) calls d_add() dentry directly, and
.create only needs to d_instantiate() this dentry.
* set sb->s_d_op to ll_d_ops, and d_alloc() will initialize
dentry with it, therefore it's no need to set dentry operations
explicitely for new kernel.
* .d_delete is called before dentry refcount decrease for new
kernel.
* other cleanups.

Change-Id: Ib08bd4a45dba8d3d5b6150e9a4df95728fdf33e2
Signed-off-by: Lai Siyao <laisiyao@whamcloud.com>
Signed-off-by: yang sheng <ys@whamcloud.com>
Reviewed-on: http://review.whamcloud.com/1865
Reviewed-by: Fan Yong <yong.fan@whamcloud.com>
Tested-by: Hudson
Tested-by: Maloo <whamcloud.maloo@gmail.com>
Reviewed-by: Andreas Dilger <adilger@whamcloud.com>
Reviewed-by: Oleg Drokin <green@whamcloud.com>

Revert "LU-1201 checksum: add libcfs crypto hash"

This reverts commit 84a3fd67356c8073a917ea6abd63928055e38156
that broke build for external ofed on rhel5 kernels.

Signed-off-by: Oleg Drokin <green@whamcloud.com>

LU-1373 ptlrpc: add flow control extension to ptlrpc req set

This patch allows new requests to be added to a request set while
this latter has already requests in flight. This is done by adding a
callback function invoked by the request set to generate RPCs.
The request set will fire a new RPC each time one completes, keeping
the number of RPCs in flight equals to set->set_max_inflight.

Lock callbacks can thus be sent by the service thread again. This
avoids doing disk I/O from the ptlrpcd context.

Signed-off-by: Johann Lombardi <johann@whamcloud.com>
Change-Id: If95922fa8da1dfa7ce7c98d6bfe35e9a5f5bb34f
Reviewed-on: http://review.whamcloud.com/2650
Tested-by: Hudson
Reviewed-by: Fan Yong <yong.fan@whamcloud.com>
Reviewed-by: Jinshan Xiong <jinshan.xiong@whamcloud.com>
Tested-by: Maloo <whamcloud.maloo@gmail.com>
Reviewed-by: Oleg Drokin <green@whamcloud.com>

LU-1461 vmalloc usage needs linux/vmalloc.h include

It seems the ext4-vmalloc patch forgot to include linux/vmalloc.h
in mballoc.c

Signed-off-by: Oleg Drokin <green@whamcloud.com>
Test Requirements: fortestonly
Test-Parameters: ostcount=2

Change-Id: I994b75353ce13bac41f45e6bf494191b43304e3c
Reviewed-on: http://review.whamcloud.com/3036
Tested-by: Hudson
Tested-by: Oleg Drokin <green@whamcloud.com>
Reviewed-by: Yu Jian <yujian@whamcloud.com>
Reviewed-by: Alex Zhuravlev <bzzz@whamcloud.com>
Tested-by: Maloo <whamcloud.maloo@gmail.com>

LU-56 lnet: move "match" functions to lib-ptl.c

This is still an intermediate patch for LNet SMP improvements,
it covered a few things:
- create a new file lib-ptl.c, all functions about portals
  are moved into this file.
- always precreate match hash-table for all portals, it's a little
  wasty for unique-portal, but it will save a lot of problems for
  upcoming patches
- instead of storing all portals in contiguous buffer, we allocate
  memory for each portal in turn, this change is also for upcoming
  patches.

Signed-off-by: Liang Zhen <liang@whamcloud.com>
Change-Id: I19e61c7f3a01f1c90a9f3f78d48d81dc00cd037d
Reviewed-on: http://review.whamcloud.com/2926
Tested-by: Hudson
Reviewed-by: Doug Oucharek <doug@whamcloud.com>
Tested-by: Maloo <whamcloud.maloo@gmail.com>
Reviewed-by: Bobi Jam <bobijam@whamcloud.com>
Reviewed-by: Oleg Drokin <green@whamcloud.com>

LU-933 mdc: allow disabling the mdc_rpc_lock

It is desirable to allow disabling the client mdc_{get,put}_rpc_lock()
in order to allow clients to send multiple filesystem-modifying RPCs
at the same time. While this would break MDS recovery (due to
insufficient transaction slots in the MDS last_rcvd file) it would
allow a smaller number of clients to generate a much higher RPC load
on the MDS. This is ideal for MDS/RPC load testing purposes, and can
also be used to help evaluate the potential benefits of implementing
the multi-slot last_rcvd feature.

Signed-off-by: Liang Zhen <liang@whamcloud.com>
Change-Id: I39be390ed37f0327bf7eb8015d528195b8568e0b
Reviewed-on: http://review.whamcloud.com/2084
Tested-by: Hudson
Reviewed-by: Fan Yong <yong.fan@whamcloud.com>
Reviewed-by: Andreas Dilger <adilger@whamcloud.com>
Tested-by: Maloo <whamcloud.maloo@gmail.com>
Reviewed-by: Oleg Drokin <green@whamcloud.com>

LU-56 lnet: allow to create EQ with zero eq_size

Current LNet doesn't allow to create EQ with zero eq_size, it's
an unnecessary limit if the EQ has callback, because EQ callback
is guaranteed to get every event. EQ with no-zero eq_size is only
useful for LNetEQPoll, there is no Lustre use-case that is using
both EQ callback and LNetEQPoll on a same EQ.

Also, even with upcoming LNet SMP improvements, there has to be
a lock to protect enqueue/dequeue operations of event, which might
hurt performance because of lock contention.

This patch will allow user to create EQ with eq_size being zero,
which means events will only be delivered by callback and user
will get nothing by calling LNetEQPoll.

In multiple thread environment, we would suggest user to create EQ
only with callback (eq_size=0) if the EQ might get high rate events,
LNetEQPoll can still work fine if eq_size is non-zero and nothing
will be changed at all. User is still able to create EQ with both
callback and non-zero eq_size, and call LNetEQPoll to get event,
although it's deprecated.

Summary: still can support all current use-cases, although set
eq_size to zero and only rely on callback will be the better way.

Signed-off-by: Liang Zhen <liang@whamcloud.com>
Change-Id: Ia984f2f65eb5fd064a36dfd7b399e46378013c57
Reviewed-on: http://review.whamcloud.com/2925
Reviewed-by: Doug Oucharek <doug@whamcloud.com>
Tested-by: Hudson
Tested-by: Maloo <whamcloud.maloo@gmail.com>
Reviewed-by: Bobi Jam <bobijam@whamcloud.com>
Reviewed-by: Oleg Drokin <green@whamcloud.com>

LU-56 lnet: cleanup for LNet Event Queue

This patch is just a cleanup patch for LNet Event Queue (EQ),
it's an intermediate patch for upcoming LNet SMP improvement
for EQ.

Signed-off-by: Liang Zhen <liang@whamcloud.com>
Change-Id: I2761c0c4118fa709a921b4060a7bd5ec6700a3f0
Reviewed-on: http://review.whamcloud.com/2924
Reviewed-by: Bobi Jam <bobijam@whamcloud.com>
Reviewed-by: Doug Oucharek <doug@whamcloud.com>
Tested-by: Hudson
Tested-by: Maloo <whamcloud.maloo@gmail.com>
Reviewed-by: Oleg Drokin <green@whamcloud.com>

LU-56 lnet: new internal object lnet_peer_table

Instead of attaching all peers under the big global data of LNet, we
created lnet_peer_table which is just a simple container of peers.

We also made another small change in this patch, when the last
refcount of a peer is released, the dead peer wouldn't be released
anymore, instead it will be attached on deathrow list of peer-table
and released while destroying the peer-table. Peers on deathrow list
can be re-used if we need create a new peer.

Both changes in this patch are just preparations for other LNet SMP
improvements. For example, we will create instances of
lnet_peer_table for each CPT in following patches.

Signed-off-by: Liang Zhen <liang@whamcloud.com>
Change-Id: I60e6f2e564f9ab9c86278c9a08e24aaac114e89c
Reviewed-on: http://review.whamcloud.com/2923
Reviewed-by: Bobi Jam <bobijam@whamcloud.com>
Reviewed-by: Doug Oucharek <doug@whamcloud.com>
Tested-by: Hudson
Tested-by: Maloo <whamcloud.maloo@gmail.com>
Reviewed-by: Oleg Drokin <green@whamcloud.com>

LU-56 ptlrpc: clean up ptlrpc svc initializing APIs

Both ptlrpc_init_svc and ptlrpc_init_svc_conf have long parameter list
and difficult for user, also, it doesn't make too much sense to keep
both of them. This patch will merge them to a new one
ptlrpc_register_service, and improve the structured parameter
ptlrpc_service_conf and make it more clear for user.
This is the first patch for ptlrpc service SMP improvements, more
patches depending on this will come soon.

Signed-off-by: Liang Zhen <liang@whamcloud.com>
Change-Id: I7b2932d4235fb418bee8e9d65f8b3e4f62d2849e
Reviewed-on: http://review.whamcloud.com/2878
Tested-by: Hudson
Tested-by: Maloo <whamcloud.maloo@gmail.com>
Reviewed-by: wangdi <di.wang@whamcloud.com>
Reviewed-by: Andreas Dilger <adilger@whamcloud.com>
Reviewed-by: Oleg Drokin <green@whamcloud.com>

LU-1406 ofd: add object methods

OFD objects methods except IO

Signed-off-by: Mikhail Pershin <tappro@whamcloud.com>
Change-Id: I17d1e9fd462f8022b516f16df71e40be0879efea
Reviewed-on: http://review.whamcloud.com/2887
Reviewed-by: Alex Zhuravlev <bzzz@whamcloud.com>
Tested-by: Hudson
Tested-by: Maloo <whamcloud.maloo@gmail.com>
Reviewed-by: Andreas Dilger <adilger@whamcloud.com>

LU-1406 ofd: transaction and last_rcvd functions

Add ofd transaction handlers and last_rcvd update function

Signed-off-by: Mikhail Pershin <tappro@whamcloud.com>
Change-Id: I14467658ea36d5b71eec7ce183190509a049eb09
Reviewed-on: http://review.whamcloud.com/2886
Tested-by: Hudson
Tested-by: Maloo <whamcloud.maloo@gmail.com>
Reviewed-by: Andreas Dilger <adilger@whamcloud.com>
Reviewed-by: Alex Zhuravlev <bzzz@whamcloud.com>
Reviewed-by: Oleg Drokin <green@whamcloud.com>

LU-1406 ofd: grant support

Add grants functionality

Signed-off-by: Mikhail Pershin <tappro@whamcloud.com>
Change-Id: I660508681e1e33b01dfbdb44ac32705575ee85e1
Reviewed-on: http://review.whamcloud.com/2871
Tested-by: Hudson
Reviewed-by: Andreas Dilger <adilger@whamcloud.com>
Tested-by: Maloo <whamcloud.maloo@gmail.com>
Reviewed-by: Johann Lombardi <johann@whamcloud.com>
Reviewed-by: Oleg Drokin <green@whamcloud.com>

LU-1201 checksum: add libcfs crypto hash

Add libcfs crypto hash and cleanup all lustre hash checksumming.
Now lustre hash calculations base on linux kernel crypto api for kernel,
and base on libcfs crypto implementation for userlevel. So any improvement
at linux kernel for hashes would improve lustre.

Signed-off-by: Alexander Boyko <alexander_boyko@xyratex.com>
Reviewed-by: Alexey Lyashkov <alexey_lyashkov@xyratex.com>
Reviewed-by: Alexander Zarochentsev <alexander_zarochentsev@xyratex.com>
Xyratex-bug-id: MRP-337, MRP-471
Change-Id: I71e7d4fbf45ee40c0c15e235eb71fb4aa86c17d1
Reviewed-on: http://review.whamcloud.com/2276
Tested-by: Hudson
Tested-by: Maloo <whamcloud.maloo@gmail.com>
Reviewed-by: Andreas Dilger <adilger@whamcloud.com>
Reviewed-by: Oleg Drokin <green@whamcloud.com>

LU-1402 libcfs: CFS_ALLOC_HIGH is __GFP_HIGHMEM

According to the comment above CFS_ALLOC_HIGH, the flag
allows allocator to return memory in the highmen zone, but
cfs_alloc_flags_to_gfp() incorrectly converts it to
__GFP_HIGH, which enables the use of emergency pools.

__GFP_HIGHMEM should be used instead. Also it makes sense to rename
CFS_ALLOC_HIGH to CFS_ALLOC_HIGHMEM to avoid further confusion.

Signed-off-by: Isaac Huang <Isaac_Huang@xyratex.com>
Signed-off-by: Andreas Dilger <adilger@whamcloud.com>
Change-Id: I088f36faec1fe22902065d680730731f8deab5d2
Reviewed-on: http://review.whamcloud.com/2732
Tested-by: Hudson
Tested-by: Maloo <whamcloud.maloo@gmail.com>
Reviewed-by: wangdi <di.wang@whamcloud.com>
Reviewed-by: Liang Zhen <liang@whamcloud.com>

LU-1467 ost: ASSERTION(lock->l_req_mode == lock->l_granted_mode)

The lock may be cancel while ost_prolong_lock_one invoked, so
just return in this case.

Signed-off-by: yang sheng <ys@whamcloud.com>
Change-Id: Ica6ad9199e4b210145e99d2420925803b18a7edd
Reviewed-on: http://review.whamcloud.com/3042
Tested-by: Hudson
Reviewed-by: Jinshan Xiong <jinshan.xiong@whamcloud.com>
Reviewed-by: Niu Yawei <niu@whamcloud.com>
Tested-by: Maloo <whamcloud.maloo@gmail.com>
Reviewed-by: Oleg Drokin <green@whamcloud.com>

LU-1406 ofd: add FMD support

Add FMD code in OFD

Signed-off-by: Mikhail Pershin <tappro@whamcloud.com>
Change-Id: Iec1cd619cd99b49f5a40c21fc7e8815b5ae392e0
Reviewed-on: http://review.whamcloud.com/2863
Tested-by: Hudson
Tested-by: Maloo <whamcloud.maloo@gmail.com>
Reviewed-by: Andreas Dilger <adilger@whamcloud.com>
Reviewed-by: Johann Lombardi <johann@whamcloud.com>
Reviewed-by: Alex Zhuravlev <bzzz@whamcloud.com>

LU-1406 ofd: add procfs values

Add procfs functionality to OFD

Signed-off-by: Mikhail Pershin <tappro@whamcloud.com>
Change-Id: If2c0a66479d9df8d585cde651244b86bf78636ce
Reviewed-on: http://review.whamcloud.com/2862
Tested-by: Hudson
Tested-by: Maloo <whamcloud.maloo@gmail.com>
Reviewed-by: Andreas Dilger <adilger@whamcloud.com>
Reviewed-by: Alex Zhuravlev <bzzz@whamcloud.com>

LU-1406 ofd: OFD device methods

Add capability code and basic device functions: statfs, set_info,
get_info, iocontrol, ping, sync, health_check

Signed-off-by: Mikhail Pershin <tappro@whamcloud.com>
Change-Id: I7a0a671ef45e8f4860344bd8c9a9a306d60bd780
Reviewed-on: http://review.whamcloud.com/2856
Tested-by: Hudson
Tested-by: Maloo <whamcloud.maloo@gmail.com>
Reviewed-by: Andreas Dilger <adilger@whamcloud.com>
Reviewed-by: Alex Zhuravlev <bzzz@whamcloud.com>
Reviewed-by: Oleg Drokin <green@whamcloud.com>

LU-1422 cray: Cray platforms set the initial connection value to high

While testing imperative recovery it was observed that cray systems
had a noticable delay that did not happen on other systems. It was
discovered that the INITIAL_CONNECTION_TIMEOUT has set to a value
for the old Cray catamount systems which is not long valid and
now impacts recovery performace for both the IR and SR case. Also
this patch removes some catamount hacks from liblustre.

Signed-off-by: James Simmons <uja.ornl@gmail.com>
Change-Id: Ia6ff6e22886c49c861ad9469e675f38fe42d6fb4
Reviewed-on: http://review.whamcloud.com/2835
Reviewed-by: Andreas Dilger <adilger@whamcloud.com>
Reviewed-by: Cory Spitz <spitzcor@cray.com>
Tested-by: Hudson
Tested-by: Maloo <whamcloud.maloo@gmail.com>
Reviewed-by: Oleg Drokin <green@whamcloud.com>

LU-1378 fid: Add console info for super seq allocation

Add console information for super sequence allocation. Because
one super sequence will include 1 billion sequences, it rarely
happens in reality, so it will not cause the flood of console
msg.

Signed-off-by: Di Wang <di.wang@whamcloud.com>
Change-Id: I5154ee2a03006680b6a08d588287bf3941149457
Reviewed-on: http://review.whamcloud.com/2701
Tested-by: Hudson
Reviewed-by: Andreas Dilger <adilger@whamcloud.com>
Tested-by: Maloo <whamcloud.maloo@gmail.com>
Reviewed-by: Fan Yong <yong.fan@whamcloud.com>
Reviewed-by: Oleg Drokin <green@whamcloud.com>

LU-1282 lprocfs: free stat memory as it is cleared

Clear stats also free the stat memory occupation.

Signed-off-by: Bobi Jam <bobijam@whamcloud.com>
Change-Id: Ie0050eb716883aaeca08f4b3bacc772e83b87679
Reviewed-on: http://review.whamcloud.com/3026
Tested-by: Hudson
Tested-by: Maloo <whamcloud.maloo@gmail.com>
Reviewed-by: Andreas Dilger <adilger@whamcloud.com>
Reviewed-by: Oleg Drokin <green@whamcloud.com>

LU-80 tests: make lfs_migrate script executable

When running sanity.sh test_56w from the build tree, the test fails
because lfs_migrate does not have execute permission set. Set the
execute permissions on this script in the build tree.

Signed-off-by: Andreas Dilger <adilger@whamcloud.com>
Change-Id: I4af5d7ff4b41353d2e73937292020505ef8de0d9
Reviewed-on: http://review.whamcloud.com/3020
Tested-by: Hudson
Tested-by: Maloo <whamcloud.maloo@gmail.com>
Reviewed-by: Yu Jian <yujian@whamcloud.com>
Reviewed-by: Oleg Drokin <green@whamcloud.com>

LU-56 lnet: container for LNet message

Adding a simple message container to LNet, and use it to replace
global message queues and message finalizers.
It's a work step of LNet SMP improvements, we will create instances
for each CPT in following patches.

Signed-off-by: Liang Zhen <liang@whamcloud.com>
Change-Id: I00f79d7598c4cd1274d7b72f7c4b482d9df75b7a
Reviewed-on: http://review.whamcloud.com/2922
Reviewed-by: Bobi Jam <bobijam@whamcloud.com>
Reviewed-by: Doug Oucharek <doug@whamcloud.com>
Tested-by: Hudson
Tested-by: Maloo <whamcloud.maloo@gmail.com>
Reviewed-by: Oleg Drokin <green@whamcloud.com>

LU-56 lnet: abstract container for EQ/ME/MD

This patch implemented a simple resource container for LNet,
it also created several instances of it to manage different LNet
objects (EQ/ME/MD), instead of putting resources in global
table/list.
It's a work step of LNet SMP improvements, we will create instances
for each CPT in following patches.

Signed-off-by: Liang Zhen <liang@whamcloud.com>
Change-Id: I5e03d618a5c932f032b45b4acdbd4aec546cf9a9
Reviewed-on: http://review.whamcloud.com/2921
Tested-by: Hudson
Tested-by: Maloo <whamcloud.maloo@gmail.com>
Reviewed-by: Bobi Jam <bobijam@whamcloud.com>
Reviewed-by: Lai Siyao <laisiyao@whamcloud.com>
Reviewed-by: Doug Oucharek <doug@whamcloud.com>
Reviewed-by: Andreas Dilger <adilger@whamcloud.com>

LU-56 lnet: add lnet_*_free_locked for LNet

LNet has a set of internal functions lnet_*_free to release LNet
message or LNet descriptors (EQ, MD, ME), they require caller to
hold LNET_LOCK on calling, this is somehow confusing because
lnet_*_alloc should be called w/o any lock.
This patch add another set of functions named as lnet_*_free_locked
to replace original lnet_*_free which require caller to hold lock,
and still provide lnet_*_free which can be called w/o any lock.

This patch is just the first step work for LNet SMP improvements.

Signed-off-by: Liang Zhen <liang@whamcloud.com>
Change-Id: I4686d83fc08695f117bf6b3b53d5ad56fb03bcd1
Reviewed-on: http://review.whamcloud.com/2919
Reviewed-by: Andreas Dilger <adilger@whamcloud.com>
Reviewed-by: Doug Oucharek <doug@whamcloud.com>
Tested-by: Hudson
Tested-by: Maloo <whamcloud.maloo@gmail.com>

LU-1408 debug: initialize debug_msg_data if needed

initialize the libcfs_debug_msg_data only if needed

Change-Id: Id80ea290b982153c86dae2f71c3d42753fa8848f
Signed-off-by: Hongchao Zhang <hongchao.zhang@whamcloud.com>
Reviewed-on: http://review.whamcloud.com/2889
Reviewed-by: Oleg Drokin <green@whamcloud.com>
Tested-by: Hudson
Reviewed-by: James Simmons <uja.ornl@gmail.com>
Tested-by: Maloo <whamcloud.maloo@gmail.com>
Reviewed-by: Shuichi Ihara <sihara@ddn.com>

LU-1406 ofd: add connect functions

Patch adds connect functions and export init/fini

Signed-off-by: Mikhail Pershin <tappro@whamcloud.com>
Change-Id: I6535d37845863c269dc279c6acc8be8bb8c54432
Reviewed-on: http://review.whamcloud.com/2855
Tested-by: Hudson
Tested-by: Maloo <whamcloud.maloo@gmail.com>
Reviewed-by: Andreas Dilger <adilger@whamcloud.com>
Reviewed-by: Oleg Drokin <green@whamcloud.com>

LU-957 scrub: OI scrub against ldiskfs

The OI scrub is totally implemented inside osd-ldiskfs,
and driven by the inode iterator. For each object with
normal fid, its fid in the LMA and related OI entry in
the OI file will be compared, if they are inconsistent,
then the OI entry will be updated.

Signed-off-by: Fan Yong <yong.fan@whamcloud.com>
Change-Id: I04d25b3433336fb63c312f795a8631328beb3aa7
Reviewed-on: http://review.whamcloud.com/2552
Tested-by: Hudson
Reviewed-by: Andreas Dilger <adilger@whamcloud.com>
Tested-by: Maloo <whamcloud.maloo@gmail.com>
Reviewed-by: Alex Zhuravlev <bzzz@whamcloud.com>
Reviewed-by: Oleg Drokin <green@whamcloud.com>

LU-957 scrub: OSD layer cleanup for OI scrub

1) Abstract some common functions in OSD layer.
2) Call lu_object_put() in osd_object_find() under failure case.
3) Do not release buffer_head until all the needed contents have
been accessed in osd_oi_index_create_one().

Signed-off-by: Fan Yong <yong.fan@whamcloud.com>
Change-Id: I3e6880271ee73ec2c4638b02c7e9c55fc827bdf3
Reviewed-on: http://review.whamcloud.com/2551
Reviewed-by: Andreas Dilger <adilger@whamcloud.com>
Tested-by: Hudson
Tested-by: Maloo <whamcloud.maloo@gmail.com>
Reviewed-by: Oleg Drokin <green@whamcloud.com>

LU-957 scrub: Ancillary work for LFSCK/OI scrub

1) New debug sub-system - 'D_LFSCK'
For Lustre fsck/scrub running trace.

2) New MDT mount option - 'noscrub'
To disable auto triggering OI scrub when MDT mounts up
or by RPC which accesses inconsistent OI mapping entry.

3) Any object which will be used when Lustre server mounts
should not be added into OI files to guarantee that the
Lustre server can start up even though OI files corrupt.

Signed-off-by: Fan Yong <yong.fan@whamcloud.com>
Change-Id: I8b1f1f1160db25ce05bd2d5688f9ff1507e3841b
Reviewed-on: http://review.whamcloud.com/2897
Tested-by: Hudson
Reviewed-by: Andreas Dilger <adilger@whamcloud.com>
Tested-by: Maloo <whamcloud.maloo@gmail.com>
Reviewed-by: Oleg Drokin <green@whamcloud.com>