Whamcloud - gitweb
fs/lustre-release.git
7 years agoLU-2722 clio: directIO thread races with completion thread
Vitaly Fertman [Thu, 31 Jan 2013 12:12:42 +0000 (16:12 +0400)]
LU-2722 clio: directIO thread races with completion thread

ll_direct_IO_26()) ASSERTION( obj->cob_transient_pages == 0 ) failed

The directio thread puts pages to the transfer in osc_io_submit(), and
waits for its completion. Upon completion, osc_completion() is called,
which notifies waiters IO is completed and only after that put the
page. the original directIO thread wakes up and may succeed to assert
on the last page which is not yet put by osc_completion().

Signed-off-by: Vitaly Fertman <vitaly_fertman@xyratex.com>
Change-Id: Ib912e913885a9ff2cb1e9c865ae2003a676c2b5b
Xyratex-bug-id: MRP-838
Reviewed-on: http://review.whamcloud.com/5223
Reviewed-by: Jinshan Xiong <jinshan.xiong@intel.com>
Tested-by: Hudson
Tested-by: Maloo <whamcloud.maloo@gmail.com>
Reviewed-by: Niu Yawei <yawei.niu@intel.com>
Reviewed-by: Oleg Drokin <oleg.drokin@intel.com>
7 years agoLU-2467 ptlrpc: Allow OBD_PINGs to be suppressed
Li Wei [Wed, 30 Jan 2013 12:38:59 +0000 (20:38 +0800)]
LU-2467 ptlrpc: Allow OBD_PINGs to be suppressed

This patch introduces a new ptlrpc module parameter, "suppress_pings",
to provide an option for reducing excessive OBD_PINGs in large
clusters.  The parameter affects all MDTs and OSTs on a node.  It is
off (zero) by default, giving a behavior identical to current
implementation.  If it is on (non-zero), all clients of the affected
targets who understand OBD_CONNECT_PINGLESS will know, at connect
time, that pings are not required.  When suppressing pings, there must
be an external mechanism to notify the targets of client deaths, via
the targets' "evict_client" procfs entries.  In addition, a highly
available standalone MGS is also recommended when suppressing pings,
so that clients are notified (through Imperative Recovery) of target
recoveries.

The changes do two basically independent things.  One is initializing
import and export states (i.e., imp_connect_data and
exp_obd_chain_timed) according to "suppress_pings", since whether to
ping or not is a property of each import-export pair.  MGC pings can
not be suppressed, because maintaining MGS connections is dictated by
the reliance on Imperative Recovery.

The other thing is changing pinger and import routines to respect the
import property set earlier.  (The export side does not need any
change at all.)  Pings are still needed to query last committed
transactions if there are uncommitted requests on an import, so that
resources pinned for replays can be released even when applications
become idle.  An early version of this patch removes imports that do
not need to be pinged from pinger_imports and add them back when last
committed transactions are needed or recoveries must be initiated.
This version does not do that because a) the overheads of iterating
through 10,000 imports are not prohibitively large---around 10 ms, b)
adding imports back to pinger_imports requires the global pinger_mutex
to be held, and c) the imp_lock contention added on each import is
small.

Change-Id: Iabc84d395c978c3f156c52aebfad83621facb4fe
Signed-off-by: Li Wei <wei.g.li@intel.com>
Reviewed-on: http://review.whamcloud.com/5009
Reviewed-by: Lai Siyao <lai.siyao@intel.com>
Tested-by: Hudson
Reviewed-by: Emoly Liu <emoly.liu@intel.com>
Tested-by: Maloo <whamcloud.maloo@gmail.com>
Reviewed-by: Andreas Dilger <andreas.dilger@intel.com>
Reviewed-by: Oleg Drokin <oleg.drokin@intel.com>
7 years agoLU-2467 protocol: Add OBD_CONNECT_PINGLESS
Li Wei [Thu, 31 Jan 2013 09:48:23 +0000 (17:48 +0800)]
LU-2467 protocol: Add OBD_CONNECT_PINGLESS

Reserve a bit for OBD_CONNECT_PINGLESS, which indicates a client is
capable of suppressing keep-alive OBD_PINGs.  If granted by a server,
it means the server does not require (but still allows) pings.

Signed-off-by: Li Wei <wei.g.li@intel.com>
Change-Id: I7575082244a84f2e633a1936fcf894da3c6ee7dd
Reviewed-on: http://review.whamcloud.com/5229
Reviewed-by: Andreas Dilger <andreas.dilger@intel.com>
Reviewed-by: Oleg Drokin <oleg.drokin@intel.com>
Tested-by: Hudson
Tested-by: Maloo <whamcloud.maloo@gmail.com>
7 years agoLU-2723 clio: a client hangs on osc_extent_wait in dio
Vitaly Fertman [Thu, 31 Jan 2013 12:36:19 +0000 (16:36 +0400)]
LU-2723 clio: a client hangs on osc_extent_wait in dio

If dio overlaps with mapped region, first overlapped pages are sent
by generic_file_direct_write->filemap_write_and_wait_range->writepages
After that the dio pages are sent by write_begin/write_end, writepages
However, write_end->osc_page_cache_add does not release extent because
the IO is not considered as synchronous. as the result, the following
writepages->osc_cache_writeback_range() does not send pages as the
extent is still ACTIVE and osc_cache_wait_range() hangs forever.

The patch make directIO to be considered as sync by lustre so that
extent was released at write_end.

Signed-off-by: Vitaly Fertman <vitaly_fertman@xyratex.com>
Xyratex-bug-id: MRP-839
Change-Id: I0c0e1599ee180f467826e4b6715c3cbfd0c6eb5c
Reviewed-on: http://review.whamcloud.com/5224
Reviewed-by: Jinshan Xiong <jinshan.xiong@intel.com>
Tested-by: Hudson
Tested-by: Maloo <whamcloud.maloo@gmail.com>
Reviewed-by: Oleg Drokin <oleg.drokin@intel.com>
7 years agoLU-2194 tests: add debugging to recovery-small/19
Nathaniel Clark [Mon, 21 Jan 2013 20:15:20 +0000 (15:15 -0500)]
LU-2194 tests: add debugging to recovery-small/19

Add error messages to 19a and 19b

Signed-off-by: Nathaniel Clark <nathaniel.l.clark@intel.com>
Change-Id: Ief35c063734654c5a6f3dab0d283c4b7a32706e8
Reviewed-on: http://review.whamcloud.com/5141
Reviewed-by: Li Wei <wei.g.li@intel.com>
Reviewed-by: Jian Yu <jian.yu@intel.com>
Tested-by: Hudson
Tested-by: Maloo <whamcloud.maloo@gmail.com>
Reviewed-by: Andreas Dilger <andreas.dilger@intel.com>
7 years agoLU-1187 mdt: directory remote open fix
Wang Di [Tue, 2 Oct 2012 13:54:53 +0000 (06:54 -0700)]
LU-1187 mdt: directory remote open fix

Directory remote open continue as normal open, i.e. return the
remote fid to client, and client will send the open request to
the right MDT.

Set correct it disposition in mdt_open_by_fid.

Signed-off-by: wang di <di.wang@intel.com>
Change-Id: I32d5dee9cd1eac47787e2d85a5db1e164d2846d3
Reviewed-on: http://review.whamcloud.com/4934
Reviewed-by: Andreas Dilger <andreas.dilger@intel.com>
Tested-by: Hudson
Reviewed-by: Alex Zhuravlev <alexey.zhuravlev@intel.com>
Tested-by: Maloo <whamcloud.maloo@gmail.com>
Reviewed-by: Mike Pershin <mike.pershin@intel.com>
7 years agoLU-1187 mdd: a few missing stuff in MD stack for DNE.
wangdi [Tue, 1 Oct 2013 11:25:49 +0000 (04:25 -0700)]
LU-1187 mdd: a few missing stuff in MD stack for DNE.

1. Assign the index operation for the directory create
in declare phase, which is needed for creating the
object in OUT.

2. Declare dotdot insertion for remote directory creation,
so the insert update can be packed into RPC and send to
the remote MDT (out) to be executed.

3. Add hint to the object declare phase, by which OSP can
get some info during update RPC packing, like parent fid.

4. Separate lu_object_exists into lu_object_remote and
lu_object_exists, so it can check whether remote object
exists as well.

Signed-off-by: wang di <di.wang@intel.com>
Change-Id: Icb359443d9982ee8567af933c5def42dc51a3a7a
Reviewed-on: http://review.whamcloud.com/4930
Tested-by: Hudson
Reviewed-by: Andreas Dilger <andreas.dilger@intel.com>
Reviewed-by: Alex Zhuravlev <alexey.zhuravlev@intel.com>
Tested-by: Maloo <whamcloud.maloo@gmail.com>
Reviewed-by: Mike Pershin <mike.pershin@intel.com>
7 years agoLU-2654 ldlm: ldlm_resource_dump should be called with resource locked
Oleg Drokin [Sat, 2 Feb 2013 04:08:59 +0000 (23:08 -0500)]
LU-2654 ldlm: ldlm_resource_dump should be called with resource locked

Fix ldlm_resource_complain to lock resource before calling
ldlm_resource_dump

Change-Id: I9dfbe84bd63de3093c99e7379daeca30d02fc875
Signed-off-by: Oleg Drokin <green@whamcloud.com>
Reviewed-on: http://review.whamcloud.com/5254
Tested-by: Hudson
Tested-by: Maloo <whamcloud.maloo@gmail.com>
Reviewed-by: Alex Zhuravlev <alexey.zhuravlev@intel.com>
Reviewed-by: Liang Zhen <liang.zhen@intel.com>
Reviewed-by: Oleg Drokin <oleg.drokin@intel.com>
Tested-by: Oleg Drokin <oleg.drokin@intel.com>
7 years agoLU-1187 mdt: add sanity check for rename and link
wangdi [Sun, 21 Oct 2012 10:49:02 +0000 (03:49 -0700)]
LU-1187 mdt: add sanity check for rename and link

Add sanity check for rename/link, so remote rename/link
will return EXDEV.

Signed-off-by: wang di <di.wang@whamcloud.com>
Change-Id: I841c8ffaa74b162fcca5cd1661c383215350cccc
Reviewed-on: http://review.whamcloud.com/4348
Tested-by: Hudson
Reviewed-by: Andreas Dilger <andreas.dilger@intel.com>
Tested-by: Maloo <whamcloud.maloo@gmail.com>
Reviewed-by: Alex Zhuravlev <alexey.zhuravlev@intel.com>
Reviewed-by: Oleg Drokin <oleg.drokin@intel.com>
7 years agoLU-1187 mdt: unlink remote directory
wangdi [Tue, 19 Nov 2013 14:34:21 +0000 (06:34 -0800)]
LU-1187 mdt: unlink remote directory

Send unlink req to the slave MDT, so both open/close
and unlink will send request to the slave MDT, where
the remote object resides. Then it would be able to
check orphan (unlink open file) locally.

Add lock_object api for enqueue remote object.

Change-Id: I7483b0f023e4e3de6597da58d3d9f3e96c0d53b7
Signed-off-by: Wang Di <di.wang@intel.com>
Reviewed-on: http://review.whamcloud.com/4339
Reviewed-by: Andreas Dilger <andreas.dilger@intel.com>
Reviewed-by: Alex Zhuravlev <alexey.zhuravlev@intel.com>
Tested-by: Hudson
Tested-by: Maloo <whamcloud.maloo@gmail.com>
7 years agoLU-1187 utils: add lfs setdirstripe/getdirstripe
wangdi [Sat, 16 Nov 2013 09:19:02 +0000 (01:19 -0800)]
LU-1187 utils: add lfs setdirstripe/getdirstripe

1.Add lfs setdirstripe/mkdir/getdirstripe to create
remote directory.

2.Add lfs rm_entry to just remove the entry of remote
directory.

Change-Id: Idaecfbb9dd97e3ae9fbc70e5b902a91a7b2ed18e
Signed-off-by: Wang Di <di.wang@intel.com>
Reviewed-on: http://review.whamcloud.com/4341
Reviewed-by: Andreas Dilger <andreas.dilger@intel.com>
Tested-by: Hudson
Tested-by: Maloo <whamcloud.maloo@gmail.com>
Reviewed-by: Alex Zhuravlev <alexey.zhuravlev@intel.com>
7 years agoLU-1866 lfsck: enhance otable-based iteration
Fan Yong [Sat, 12 Jan 2013 11:47:12 +0000 (19:47 +0800)]
LU-1866 lfsck: enhance otable-based iteration

1) Use ::load() method to unplug OSD layer iteration.

2) Use ::put() method to wakeup the LFSCK main engine
   if it is blocked at low layer for ::next() call.

Test-Parameters: envdefinitions=ENABLE_QUOTA=yes testlist=sanity-scrub
Signed-off-by: Fan Yong <fan.yong@intel.com>
Change-Id: I28e5689ad3a4ee96a3f17a64e8d6dca553a86d77
Reviewed-on: http://review.whamcloud.com/4906
Reviewed-by: Andreas Dilger <andreas.dilger@intel.com>
Tested-by: Hudson
Tested-by: Maloo <whamcloud.maloo@gmail.com>
Tested-by: Andreas Dilger <andreas.dilger@intel.com>
Reviewed-by: Alex Zhuravlev <alexey.zhuravlev@intel.com>
7 years agoLU-1866 osd: FID-in-LMA and OI files
Fan Yong [Sat, 12 Jan 2013 11:08:35 +0000 (19:08 +0800)]
LU-1866 osd: FID-in-LMA and OI files

Generate FID-in-LMA not only for normal objects, but also for
IGIF objects, server-side local files, and other non-OST/LLOG
objects.

LAST_ID has FID-in-LMA but no related mappings in the OI file.

At the same time, OI mapping usage is also extended to above
objects. Only the objects marked as "LDISKFS_STATE_LUSTRE_NO_OI"
or OST objects have no OI mappings the OI files.

It is part of 1.8 IGIF objects upgrage handling. OI scrub will
automatically generate IGIF-in-LMA and add related OI mappings
in the OI files.

Test-Parameters: envdefinitions=ENABLE_QUOTA=yes testlist=sanity-scrub

Signed-off-by: Fan Yong <fan.yong@intel.com>
Change-Id: If1e1434c238f2ab737e35433bfad1abbb00fcb00
Reviewed-on: http://review.whamcloud.com/4904
Reviewed-by: Andreas Dilger <andreas.dilger@intel.com>
Tested-by: Hudson
Tested-by: Maloo <whamcloud.maloo@gmail.com>
Reviewed-by: Alex Zhuravlev <alexey.zhuravlev@intel.com>
7 years agoLU-2733 osd: not return "-ENOENT" for zfs osd_object_init
Fan Yong [Sat, 12 Jan 2013 23:02:35 +0000 (07:02 +0800)]
LU-2733 osd: not return "-ENOENT" for zfs osd_object_init

For zfs backend, the osd_object_init() should not return
"-ENOENT" if osd_fid_lookup() cannot find the object,
which is normal for new created cases.

Signed-off-by: Fan Yong <fan.yong@intel.com>
Change-Id: I76e5d802d0caf8ac3a9992278ec2f7c6c6e2cf07
Reviewed-on: http://review.whamcloud.com/5242
Reviewed-by: Alex Zhuravlev <alexey.zhuravlev@intel.com>
Tested-by: Hudson
Reviewed-by: Andreas Dilger <andreas.dilger@intel.com>
Tested-by: Maloo <whamcloud.maloo@gmail.com>
7 years agoLU-1187 osd: add agent/local inode for remote directory
wangdi [Sat, 16 Nov 2013 08:41:27 +0000 (00:41 -0800)]
LU-1187 osd: add agent/local inode for remote directory

A local inode will be created for remote directory entry, so
to make e2fsck check pass.

The remote directory on the remote MDT will be put under the
agent directory.

Remove LA_TYPE checking during osd_inode_set, because when
initialize the remote object, it needs to get TYPE information
from remote attr get, where we need type information anyway,
(osp_md_attr_get ->out_attr_get->osd_attr_get | LA_TYPE).
So osd_attr_get will always return LA_TYPE valid information
here, then if it do attr_get/modify/attr_set might hit this
ASSERT, So during osd_attr_set, we can just ignore type.

Signed-off-by: wang di <di.wang@intel.com>
Change-Id: I7f177fa8ae2ab450cde78a8c2da9a2770ec6b0cb
Reviewed-on: http://review.whamcloud.com/4931
Reviewed-by: Alex Zhuravlev <alexey.zhuravlev@intel.com>
Reviewed-by: Andreas Dilger <andreas.dilger@intel.com>
Tested-by: Hudson
Tested-by: Maloo <whamcloud.maloo@gmail.com>
7 years agoLU-1866 scrub: initial OI scrub
Fan Yong [Sat, 12 Jan 2013 08:03:35 +0000 (16:03 +0800)]
LU-1866 scrub: initial OI scrub

We will extend OI mapping usage (in OI directories/files) to
all the objects, except for the objects which are marked as
"LDISKFS_STATE_LUSTRE_NO_OI" or OST object.

The OI mappings for the server-side local files may become
invalid because of server-side file-level back/restore, or
server crash. Those files are necessary for server mount,
so related OI mappings must be checked and rebuilt before
they are really used. That is the initial OI scrub work.

Initial OI scrub is used for quickly checking and rebuilding
OI mappings for server-side local files during server mount.

Another change is about MDT_LAST_RECV_OID/OFD_LAST_RECV_OID.
Currently, we can just use the same OID for the "last_rcvd"
files on different targets. So merge the two OIDs as single
OID: LAST_RECV_OID, which also simplifies OI scrub work.

Signed-off-by: Fan Yong <fan.yong@intel.com>
Change-Id: Iddca4812d3c82ca43f1e852fd74475b11b6e6dea
Reviewed-on: http://review.whamcloud.com/4903
Reviewed-by: Alex Zhuravlev <alexey.zhuravlev@intel.com>
Tested-by: Hudson
Reviewed-by: Andreas Dilger <andreas.dilger@intel.com>
Tested-by: Maloo <whamcloud.maloo@gmail.com>
7 years agoLU-1866 osd: ancillary work for initial OI scrub
Fan Yong [Sat, 12 Jan 2013 00:41:01 +0000 (08:41 +0800)]
LU-1866 osd: ancillary work for initial OI scrub

1) Abstract some general interfaces that can be shared by common
   OSD operations, by normal OI scrub, and by initial OI scrub.

2) Mark the object as "LDISKFS_STATE_LUSTRE_NO_OI" if it is not
   unnecessary to add OI mapping for the object, such as backend
   local root, OI directories/files, and others created before OI
   files created.

3) Do not check fld in OI scurb, which may be not ready yet.

4) Reorganize OI scrub code to be more readable.

Signed-off-by: Fan Yong <fan.yong@intel.com>
Change-Id: Ife8168df2454437ad18550b73d26f7372c21776f
Reviewed-on: http://review.whamcloud.com/4902
Reviewed-by: Alex Zhuravlev <alexey.zhuravlev@intel.com>
Tested-by: Hudson
Reviewed-by: Andreas Dilger <andreas.dilger@intel.com>
Tested-by: Maloo <whamcloud.maloo@gmail.com>
7 years agoLU-2725 ost: ost_rw_hpreq_check should return 0 or 1
Oleg Drokin [Thu, 31 Jan 2013 19:10:08 +0000 (14:10 -0500)]
LU-2725 ost: ost_rw_hpreq_check should return 0 or 1

ost_rw_hpreq_check was returning number of matched locks which is wrong.

Change-Id: Ic4d3dd4a186b616c13ffd9f897254da3c19f84f8
Signed-off-by: Oleg Drokin <green@whamcloud.com>
Reviewed-on: http://review.whamcloud.com/5233
Reviewed-by: Andreas Dilger <andreas.dilger@intel.com>
Tested-by: Oleg Drokin <oleg.drokin@intel.com>
Reviewed-by: Mike Pershin <mike.pershin@intel.com>
Tested-by: Hudson
Tested-by: Maloo <whamcloud.maloo@gmail.com>
7 years agoLU-1431 ptlrpc: PTLRPC_BRW_MAX_SIZE usage cleanup.
Sergii Glushchenko [Tue, 29 Jan 2013 14:19:40 +0000 (16:19 +0200)]
LU-1431 ptlrpc: PTLRPC_BRW_MAX_SIZE usage cleanup.

Clean-up the layering in current code by eliminating direct
usage of PTLRPC_BRW_MAX_SIZE macro outside of the ptlrpc
module. This should help us acheive "floating" max brw size
value across the cluster, which in turn should help with
4MB IO task.

Signed-off-by: Sergii Glushchenko <sergii_glushchenko@xyratex.com>
Reviewed-by: Alexey Lyashkov <alexey_lyashkov@xyratex.com>
Reviewed-by: Alexander Zarochentsev <Alexander_Zarochentsev@xyratex.com>
Change-Id: Ib9f6265bd28f25c34b27eb6ad1a56a752b71a4ac
Xyratex-bug-id: MRP-687
Reviewed-on: http://review.whamcloud.com/4876
Reviewed-by: Andreas Dilger <andreas.dilger@intel.com>
Tested-by: Hudson
Tested-by: Maloo <whamcloud.maloo@gmail.com>
Reviewed-by: Oleg Drokin <oleg.drokin@intel.com>
7 years agoLU-1866 fid: cleanup object visibility definition and check
Fan Yong [Sat, 12 Jan 2013 00:40:24 +0000 (08:40 +0800)]
LU-1866 fid: cleanup object visibility definition and check

1) The normal fid, IGIF fid, "ROOT" fid, ".lustre" fid are
   used for client-mdt visible objects.

2) Remove redundan osd_igif.{c,h}

3) Hide server-side local root for up layer dir readpage.

4) It is unnecessary to append FID-in-dient for dot name entry.
   Because we can get it from the directory object directly.

Signed-off-by: Fan Yong <fan.yong@intel.com>
Change-Id: If98de0c15c0cd3a0e93b666a8d48366b0d8a2cb0
Reviewed-on: http://review.whamcloud.com/4901
Reviewed-by: Mike Pershin <mike.pershin@intel.com>
Reviewed-by: Alex Zhuravlev <alexey.zhuravlev@intel.com>
Tested-by: Hudson
Reviewed-by: Andreas Dilger <andreas.dilger@intel.com>
Tested-by: Maloo <whamcloud.maloo@gmail.com>
7 years agoLU-1866 misc: fix some issues found during LFSCK
Fan Yong [Sat, 12 Jan 2013 00:30:37 +0000 (08:30 +0800)]
LU-1866 misc: fix some issues found during LFSCK

1) Test scripts issues for sanity.sh and sanity-scrub.sh

2) Incorrectly use open flag 'MDS_OPEN_OWNEROVERRIDE" for
   permission check for set_attr.

3) Other code clean, such as mdt_body, md_op_spec, wiretest.

Signed-off-by: Fan Yong <fan.yong@intel.com>
Change-Id: I52fbfe8481e35edba75e8fe7e0ab2fec094eabae
Reviewed-on: http://review.whamcloud.com/5046
Tested-by: Hudson
Tested-by: Maloo <whamcloud.maloo@gmail.com>
Reviewed-by: Andreas Dilger <andreas.dilger@intel.com>
Reviewed-by: wangdi <di.wang@intel.com>
7 years agoLU-1187 mdt: Only create remote dir on MDT0
Wang Di [Thu, 26 Jul 2012 10:36:45 +0000 (03:36 -0700)]
LU-1187 mdt: Only create remote dir on MDT0

1.Add checking for only creating remote dir on MDT0.
2.Add enable_remote_dir parameter to enalbe remote dir creation
on other MDTs.

Change-Id: I272d4763f07115c5aa0432eb4bbdda5349711d25
Signed-off-by: Wang Di <di.wang@whamcloud.com>
Reviewed-on: http://review.whamcloud.com/4336
Tested-by: Hudson
Tested-by: Maloo <whamcloud.maloo@gmail.com>
Reviewed-by: Alex Zhuravlev <alexey.zhuravlev@intel.com>
Reviewed-by: Mike Pershin <mike.pershin@intel.com>
Reviewed-by: Oleg Drokin <green@whamcloud.com>
7 years agoLU-1187 lod: prepare default stripe EA for remote directory
Wang Di [Tue, 3 Apr 2012 06:11:09 +0000 (23:11 -0700)]
LU-1187 lod: prepare default stripe EA for remote directory

During remote directory create, it should prepare the default
stripe EA in declare, so the default EA can be set correctly
on the remote directory.

Change-Id: Ief93e72cc033f3553865f2abba53e81df0f3fa89
Signed-off-by: Wang Di <di.wang@whamcloud.com>
Reviewed-on: http://review.whamcloud.com/4335
Tested-by: Hudson
Tested-by: Maloo <whamcloud.maloo@gmail.com>
Reviewed-by: Alex Zhuravlev <alexey.zhuravlev@intel.com>
Reviewed-by: Mike Pershin <mike.pershin@intel.com>
Reviewed-by: Oleg Drokin <green@whamcloud.com>
7 years agoLU-1187 mdt: Add pre_cleanup phase in MDT stack cleanup
wangdi [Wed, 2 Oct 2013 15:37:40 +0000 (08:37 -0700)]
LU-1187 mdt: Add pre_cleanup phase in MDT stack cleanup

Add pre_cleanup phase in MDT stack cleanup, so MD OSP can be
disconnected before cleanup MDT stack.

Signed-off-by: wang di <di.wang@whamcloud.com>
Change-Id: Id1a56c393c22b4fc526c81c05b11bfb656a80559
Reviewed-on: http://review.whamcloud.com/4929
Tested-by: Hudson
Tested-by: Maloo <whamcloud.maloo@gmail.com>
Reviewed-by: Alex Zhuravlev <alexey.zhuravlev@intel.com>
Reviewed-by: Mike Pershin <mike.pershin@intel.com>
Reviewed-by: Oleg Drokin <green@whamcloud.com>
7 years agoLU-1187 mdt: add out handler for object update
wangdi [Sat, 16 Nov 2013 18:53:15 +0000 (10:53 -0800)]
LU-1187 mdt: add out handler for object update

Add object update handler on remote MDT. Because the RPC
only includes object updates, (no metadata operation), the
out_handler will skip MDD layer and call osd layer directly.

Signed-off-by: wang di <di.wang@intel.com>
Change-Id: I9d553145cedfffcbb26d839ed8c7d9cc0887ed86
Reviewed-on: http://review.whamcloud.com/4928
Reviewed-by: Mike Pershin <mike.pershin@intel.com>
Reviewed-by: Alex Zhuravlev <alexey.zhuravlev@intel.com>
Tested-by: Hudson
Tested-by: Maloo <whamcloud.maloo@gmail.com>
Reviewed-by: Oleg Drokin <green@whamcloud.com>
7 years agoLU-1187 osp: add osp_md_object for remote directory.
wangdi [Sat, 16 Nov 2013 18:52:14 +0000 (10:52 -0800)]
LU-1187 osp: add osp_md_object for remote directory.

In declare phase, the update for remote object will be added
in the remote list of the transaction, which will then being
sent to the remote MDT during trans_start.

Add osp_md_object for remote directory, any updates for
remote operation will be packed into RPC by api in
the osp_md_object

Signed-off-by: wang di <di.wang@intel.com>
Change-Id: I5fac24e21e6663dca0e2d2fa5d90d4a07a002136
Reviewed-on: http://review.whamcloud.com/4927
Reviewed-by: Mike Pershin <mike.pershin@intel.com>
Reviewed-by: Alex Zhuravlev <alexey.zhuravlev@intel.com>
Tested-by: Hudson
Tested-by: Maloo <whamcloud.maloo@gmail.com>
Reviewed-by: Oleg Drokin <green@whamcloud.com>
7 years agoLU-2466 lnet: Use Hash Table for Remote Route List
Doug Oucharek [Sat, 12 Jan 2013 01:09:15 +0000 (17:09 -0800)]
LU-2466 lnet: Use Hash Table for Remote Route List

This change updates the remote route list (the_lnet.ln_remote_nets)
to be a hash table (the_lnet.ln_remote_nets_hash) to speed up
access when dealing with a large number of routes (over 1000).

Signed-off-by: Doug Oucharek <doug.s.oucharek@intel.com>
Change-Id: I2dac8075ff38d29bd38905b5b1b002603c5a56b6
Reviewed-on: http://review.whamcloud.com/5023
Reviewed-by: Liang Zhen <liang.zhen@intel.com>
Tested-by: Hudson
Reviewed-by: John Hammond <john.hammond@intel.com>
Tested-by: Maloo <whamcloud.maloo@gmail.com>
Reviewed-by: Oleg Drokin <green@whamcloud.com>
7 years agoLU-2700 mdt: serializing the rename
wangdi [Sat, 16 Nov 2013 03:05:28 +0000 (19:05 -0800)]
LU-2700 mdt: serializing the rename

BFL is needed in the rename process even though the source and
target are under the same directory, because we still need
serialize the locking of these two objects.

Signed-off-by: wang di <di.wang@intel.com>
Change-Id: I61593e3902f1f81ac282c020d912fdf28ce2cc2c
Reviewed-on: http://review.whamcloud.com/5203
Tested-by: Hudson
Tested-by: Maloo <whamcloud.maloo@gmail.com>
Tested-by: Oleg Drokin <green@whamcloud.com>
Reviewed-by: Andreas Dilger <andreas.dilger@intel.com>
Reviewed-by: Alex Zhuravlev <alexey.zhuravlev@intel.com>
7 years agoLU-1876 hsm: Avoid deadlock on i_mutex and layout change
Jinshan Xiong [Thu, 24 Jan 2013 21:28:21 +0000 (14:28 -0700)]
LU-1876 hsm: Avoid deadlock on i_mutex and layout change

The deadlock is as follows:
Proc 1 (write): start an IO -> try to acquire i_mutex in
            generic_file_aio_write()
Proc 2 (setattr): Acquire i_mutex in ll_setattr_raw -> initialise IO
            -> ll_layout_refresh() -> wait for IO above.

It's actually not necessary to acquire i_mutex in ll_setattr_raw()
until setattr really starts.

Signed-off-by: Jinshan Xiong <jinshan.xiong@intel.com>
Change-Id: I1988d67eab8bcac2ee63b9ec0ecab06d0b7cfc66
Reviewed-on: http://review.whamcloud.com/5159
Tested-by: Hudson
Reviewed-by: jacques-Charles Lafoucriere <jacques-charles.lafoucriere@cea.fr>
Tested-by: Maloo <whamcloud.maloo@gmail.com>
Reviewed-by: Johann Lombardi <johann.lombardi@intel.com>
7 years agoLU-1876 hsm: bugfix for deadlock in glimpse and layout change
Jinshan Xiong [Thu, 24 Jan 2013 00:44:34 +0000 (16:44 -0800)]
LU-1876 hsm: bugfix for deadlock in glimpse and layout change

This dead lock is as follows: glimpse calls ccc_inode_lsm_get()
inside glimpse IO, it will be blocked at grabbing lo_type_guard
sem, because another thread is changing layout which holds
lo_type_guard but waiting for all active IO to complete.

The basic rule: any attempt to grab lo_type_guard inside an IO is
generically risky.

Signed-off-by: Jinshan Xiong <jinshan.xiong@intel.com>
Change-Id: Ib41c03f0a73e2ef7cd37e7689c2d55f7fd0484bf
Reviewed-on: http://review.whamcloud.com/5158
Reviewed-by: jacques-Charles Lafoucriere <jacques-charles.lafoucriere@cea.fr>
Tested-by: Hudson
Tested-by: Maloo <whamcloud.maloo@gmail.com>
Reviewed-by: Johann Lombardi <johann.lombardi@intel.com>
7 years agoLU-1282 lprocfs: reduce lprocfs stats memory use
Bobi Jam [Fri, 18 Jan 2013 16:54:32 +0000 (00:54 +0800)]
LU-1282 lprocfs: reduce lprocfs stats memory use

* Move percpu counter common data out, do not need to store them
  redundantly in percpu counter.
* LPROCFS_STATS_FLAG_IRQ_SAFE flag implies three things:
  1. lprocfs_counter needs lc_sum_irq.
  2. when a stats is a percpu counter stat, all its percpu counter
     gets allocated along with the stats itself.
  3. when a stats is a non-percpu stat, lprocfs_stats_lock() needs
     disable irq.
* change lprocfs_counter to make non-irq-safe stats counter do not
  use lc_sum_irq counter, which can save memory.
* Right now, only obd_memory stats use LPROCFS_STATS_FLAG_IRQ_SAFE
  flag.

Signed-off-by: Bobi Jam <bobijam.xu@intel.com>
Change-Id: I2eb037f9dcda983844857fc068c428c4fa387e7a
Reviewed-on: http://review.whamcloud.com/3246
Reviewed-by: Jinshan Xiong <jinshan.xiong@intel.com>
Tested-by: Hudson
Tested-by: Maloo <whamcloud.maloo@gmail.com>
Reviewed-by: Andreas Dilger <andreas.dilger@intel.com>
7 years agoLU-2221 ptlrpc: kerberos support for kernel>=2.6.24
Fan Yong [Sun, 6 Jan 2013 23:42:27 +0000 (07:42 +0800)]
LU-2221 ptlrpc: kerberos support for kernel>=2.6.24

Since kernel 2.6.24 the scatterlist struct has no field "page".
Then related Lustre kerberos code cannot work anymore.

So do not access scatterlist::page directly, instead, use the
scatterlist functions sg_set_page and sg_assign_page.

Signed-off-by: Fan Yong <fan.yong@intel.com>
Signed-off-by: Thomas Stibor <thomas@stibor.net>
Change-Id: I446925bb42c1e018a55a69948383c8f71976f1fa
Reviewed-on: http://review.whamcloud.com/4394
Tested-by: Hudson
Tested-by: Maloo <whamcloud.maloo@gmail.com>
Reviewed-by: Andreas Dilger <andreas.dilger@intel.com>
Reviewed-by: Nathaniel Clark <nathaniel.l.clark@intel.com>
7 years agoLU-2592 tests: do not remove $TMP/*active in rpc.sh
Jian Yu [Fri, 18 Jan 2013 09:26:49 +0000 (17:26 +0800)]
LU-2592 tests: do not remove $TMP/*active in rpc.sh

The $TMP/*active files record the current active facets info
under failover test configuration. They are removed from
init_test_env() initially. If failover tests use do_rpc_nodes()
which performs rpc.sh and in which init_test_env() is used,
then the active facets info will be lost during the testing.

This patch fixes the above issue by introducing an RPC_MODE
variable which controls that the $TMP/*active files will not
be removed from init_test_env() in rpc.sh.

Test-Parameters: envdefinitions=SLOW=yes,ENABLE_QUOTA=yes clientcount=4 osscount=2 mdscount=2 austeroptions=-R failover=true useiscsi=true testlist=replay-vbr
Signed-off-by: Jian Yu <jian.yu@intel.com>
Change-Id: Iab464449e7c41ffd76256e9ec60b48b326f3ea9e
Reviewed-on: http://review.whamcloud.com/5127
Tested-by: Hudson
Tested-by: Maloo <whamcloud.maloo@gmail.com>
Reviewed-by: Keith Mannthey <keith.mannthey@intel.com>
Reviewed-by: Wei Liu <wei3.liu@intel.com>
Reviewed-by: Oleg Drokin <green@whamcloud.com>
7 years agoLU-2693 tests: check Lustre version in conf-sanity test 66
Jian Yu [Tue, 29 Jan 2013 04:12:59 +0000 (12:12 +0800)]
LU-2693 tests: check Lustre version in conf-sanity test 66

Lustre starts supporting the "replace nids" feature
(commit 6aa30ff) since version 2.3.59. We need add
a Lustre version check into conf-sanity test 66 to
make sure the test does not run on unsupported
Lustre version.

Test-Parameters: envdefinitions=SLOW=yes \
serverjob=lustre-b2_1 serverbuildno=164 \
testlist=conf-sanity

Signed-off-by: Jian Yu <jian.yu@intel.com>
Change-Id: I7eee38664e66059bbc028c3eb7d9741b1351b883
Reviewed-on: http://review.whamcloud.com/5198
Tested-by: Hudson
Tested-by: Maloo <whamcloud.maloo@gmail.com>
Reviewed-by: Keith Mannthey <keith.mannthey@intel.com>
Reviewed-by: Wei Liu <wei3.liu@intel.com>
Reviewed-by: Oleg Drokin <green@whamcloud.com>
7 years agoLU-2697 osp: remove trailing newline on thread names
chas williams - CONTRACTOR [Mon, 28 Jan 2013 14:58:35 +0000 (09:58 -0500)]
LU-2697 osp: remove trailing newline on thread names

The trailing newline on the thread name causes confusion with some
utilities that examine this string.

Signed-off-by: chas williams - CONTRACTOR <chas@cmf.nrl.navy.mil>
Change-Id: I031aafd2541b1829bc721ebcb9f52cb5e73101b3
Reviewed-on: http://review.whamcloud.com/5189
Reviewed-by: Bob Glossman <bob.glossman@intel.com>
Reviewed-by: Nathaniel Clark <nathaniel.l.clark@intel.com>
Tested-by: Hudson
Tested-by: Maloo <whamcloud.maloo@gmail.com>
Reviewed-by: Oleg Drokin <green@whamcloud.com>
7 years agoLU-2709 build: fix 'memory corruption' errors
Sebastien Buisson [Wed, 30 Jan 2013 14:55:47 +0000 (15:55 +0100)]
LU-2709 build: fix 'memory corruption' errors

Fix 'memory corruption' defects found by Coverity version 6.5.0:
Out-of-bounds access (OVERRUN_STATIC)
Overrunning static array by passing it to a
function which indexes it at a higher byte position.
Unbounded source buffer (STRING_SIZE)
Passing string of unknown size to a function that expects
a string of a particular size.

Signed-off-by: Sebastien Buisson <sebastien.buisson@bull.net>
Change-Id: Iabc9bd7609d285a3709737748f2100b90769d795
Reviewed-on: http://review.whamcloud.com/5210
Tested-by: Hudson
Reviewed-by: Keith Mannthey <keith.mannthey@intel.com>
Reviewed-by: Bob Glossman <bob.glossman@intel.com>
Tested-by: Maloo <whamcloud.maloo@gmail.com>
Reviewed-by: Oleg Drokin <green@whamcloud.com>
7 years agoLU-1897 test: replay-single test_70b dbench not found
Keith Mannthey [Mon, 21 Jan 2013 18:02:07 +0000 (10:02 -0800)]
LU-1897 test: replay-single test_70b dbench not found

dbench is exiting before test_70b completes is work.

dbench is being used as a background load in this test.
In one observed case test was taking a little under 90s but
dbench was exiting at 62s. Duration was set at 90.

Presently dbench starts for 60s then the system sleeps
for 12s, then we start a new 60s timer in which we expect
dbench to be running.  There is a large window in which
this test will fail.

The main issue is the use of Duration for use as both the
timer of the real test and the timer of the background load.

I raised the real Duration and started the timer for the real
test before starting dbench such that the real test should
finish before dbench does or about the same time.

Signed-off-by: Keith Mannthey <keith.mannthey@intel.com>
Change-Id: Iba243fe3132a5fa677e5a7bcd09e491727ba092a
Reviewed-on: http://review.whamcloud.com/4973
Reviewed-by: Bob Glossman <bob.glossman@intel.com>
Tested-by: Hudson
Reviewed-by: Jian Yu <jian.yu@intel.com>
Tested-by: Maloo <whamcloud.maloo@gmail.com>
Reviewed-by: Oleg Drokin <green@whamcloud.com>
7 years agoLU-2101 clio: cl_sync_io_wait() need a barrier
Liang Zhen [Sun, 7 Oct 2012 17:05:44 +0000 (01:05 +0800)]
LU-2101 clio: cl_sync_io_wait() need a barrier

cl_sync_io_note() has a small window between cfs_atomic_dec_and_test
and cfs_waitq_broadcast, another thread might have destroyed anchor
between this window, then cl_sync_io_note() is calling wakeup on a
poisoned waitq and crash.

Signed-off-by: Liang Zhen <liang.zhen@intel.com>
Change-Id: I04bfc738c8c5bec3a1e8b51131fae4199ef5ec9a
Reviewed-on: http://review.whamcloud.com/5199
Reviewed-by: Johann Lombardi <johann.lombardi@intel.com>
Tested-by: Hudson
Reviewed-by: Jinshan Xiong <jinshan.xiong@intel.com>
Tested-by: Maloo <whamcloud.maloo@gmail.com>
7 years agoLU-2010 tests: replay-single/44a,b,45 find only tested mdc
Nathaniel Clark [Fri, 30 Nov 2012 16:50:55 +0000 (11:50 -0500)]
LU-2010 tests: replay-single/44a,b,45 find only tested mdc

If multiple filesystems are mounted, then test_45 would have found
mdcdevs for each filesystem.  This corrects that to only look for the
filesystem that is being tested against.

Also correctly delete old test files.

Signed-off-by: Nathaniel Clark <nathaniel.l.clark@intel.com>
Change-Id: I0297cf9b21e9c97ee468cd81d5d4ac409c59d1e0
Reviewed-on: http://review.whamcloud.com/4724
Tested-by: Hudson
Tested-by: Maloo <whamcloud.maloo@gmail.com>
Reviewed-by: Jian Yu <jian.yu@intel.com>
Reviewed-by: Oleg Drokin <green@whamcloud.com>
7 years agoLU-2576 osc: various fix in osc_enter_cache()
Niu Yawei [Sun, 6 Jan 2013 08:21:03 +0000 (03:21 -0500)]
LU-2576 osc: various fix in osc_enter_cache()

- It should not go to sleep if there isn't any inflight write
  RPC, otherwise, it could not be waked up for a long time,
  until dirty flush triggered on other OSCs or other objects.

- If the page can't be granted due to too many dirty pages
  (> obd_max_dirty_pages), osc_enter_cache() should return
  -EDQUOT.

- It should pass NULL osc object to  osc_io_unplug(), otherwise
  dirty flush can't be triggered if the passed object is
  clean.

- It should not wait in a loop of "while (cli->cl_dirty > 0)",
  trigger the dirty flush once is enough. If there are too
  many threads that consumed grants immediately, then we
  should turn to sync write, but not trigger flush and wait
  for grant again (that'll block io process on grant).

- Refuse setting max_dirty_mb as 0 via proc file.

Signed-off-by: Niu Yawei <yawei.niu@intel.com>
Change-Id: I449cba07bd427749ab023c249d9e200aba1b406a
Reviewed-on: http://review.whamcloud.com/4963
Tested-by: Hudson
Tested-by: Maloo <whamcloud.maloo@gmail.com>
Reviewed-by: Prakash Surya <surya1@llnl.gov>
Reviewed-by: Jinshan Xiong <jinshan.xiong@intel.com>
Reviewed-by: Oleg Drokin <green@whamcloud.com>
7 years agoLU-2663 tests: don't remove objects by debugfs
Niu Yawei [Mon, 28 Jan 2013 09:19:57 +0000 (04:19 -0500)]
LU-2663 tests: don't remove objects by debugfs

In lfsck.sh, we shouldn't use debugfs to remove objects, that'll
cause quota usage inconsistence warning in e2fsck.

Fix improper LBUG() in mdd_create_data(), MDS_OPEN_HAS_OBJS is
still used by lfsck.

Signed-off-by: Niu Yawei <yawei.niu@intel.com>
Change-Id: I93c6f5c40e359d0d8402868be548088d6e2ac5da
Reviewed-on: http://review.whamcloud.com/5186
Tested-by: Hudson
Reviewed-by: Andreas Dilger <andreas.dilger@intel.com>
Reviewed-by: Emoly Liu <emoly.liu@intel.com>
Tested-by: Andreas Dilger <andreas.dilger@intel.com>
7 years agoLU-2640 osd: enhance transaction debug to trace rollback
Fan Yong [Thu, 10 Jan 2013 15:17:38 +0000 (23:17 +0800)]
LU-2640 osd: enhance transaction debug to trace rollback

Some modifications may need to rollback because of failures during
the transactions. The original transaction trace mechanism did not
consider rollback cases, so cannot correctly record sub-operations
and the rollback cases.

In this patch, not only the specified sub-operation will be traced,
but also its rollback operation (if may rollback) will be recorded
also. For example:

When declare for ref_add, it will record the ref_add sub-operation
by "osd_thandle::ot_declare_ops[OSD_OT_REF_ADD]"; when consume the
ref_add credit, the possible rollback will be recorded through the
"osd_thandle::ot_declare_ops_rb[OSD_OT_REF_ADD]"; and when need to
rollback the ref_add (by ref_del) in subsequent processing, it can
be traced by the "osd_thandle::ot_declare_ops_rb[OSD_OT_REF_ADD]";
and once rollback starts, only "osd_thandle::ot_declare_ops_rb[X]"
can be consumed.

Signed-off-by: Fan Yong <fan.yong@intel.com>
Change-Id: Iade214eff468c2298f84920b997040e06f1a204f
Reviewed-on: http://review.whamcloud.com/5138
Tested-by: Hudson
Tested-by: Maloo <whamcloud.maloo@gmail.com>
Reviewed-by: Andreas Dilger <andreas.dilger@intel.com>
7 years agoLU-2061 hsm: send HSM resquest to CDT
jcl [Thu, 10 Jan 2013 17:42:12 +0000 (18:42 +0100)]
LU-2061 hsm: send HSM resquest to CDT

This patch implements the user interface to ask for hsm requests

It implements:
- lfs hsm_archive
- lfs hsm_restore
- lfs hsm_cancel
- lfs hsm_remove
- lfs hsm_release

The matching llapi call:
- llapi_hsm_user_request_alloc()
- llapi_hsm_request()

related MDS functions:
- mdt_hsm_request()

and the MDS RPC:
- MDS_HSM_REQUEST

Signed-off-by: JC Lafoucriere <jacques-charles.lafoucriere@cea.fr>
Change-Id: I635f6e0518ec12c359dbb528c3a61216650d952d
Reviewed-on: http://review.whamcloud.com/5030
Reviewed-by: Johann Lombardi <johann.lombardi@intel.com>
Tested-by: Hudson
Tested-by: Maloo <whamcloud.maloo@gmail.com>
Reviewed-by: Andreas Dilger <andreas.dilger@intel.com>
7 years agoLU-2061 hsm: get list of HSM actions
Aurelien Degremont [Wed, 18 Apr 2012 15:05:59 +0000 (17:05 +0200)]
LU-2061 hsm: get list of HSM actions

This patch implements a MDS RPC to get
the list of registered and running hsm actions on files

It implements:
- lfs hsm_action

The matching llapi call:
- llapi_hsm_current_action()

The matching MDS function:
- mdt_hsm_action()

and MDS RPC:
- MDS_HSM_ACTION

Signed-off-by: JC Lafoucriere <jacques-charles.lafoucriere@cea.fr>
Change-Id: If4b00949ca8c94c8572cc81a7f7b8db32914542c
Reviewed-on: http://review.whamcloud.com/5029
Reviewed-by: Johann Lombardi <johann.lombardi@intel.com>
Tested-by: Hudson
Tested-by: Maloo <whamcloud.maloo@gmail.com>
Reviewed-by: Andreas Dilger <andreas.dilger@intel.com>
7 years agoLU-1338 hsm: HSM flags feature
Aurelien Degremont [Mon, 14 Jan 2013 22:43:54 +0000 (23:43 +0100)]
LU-1338 hsm: HSM flags feature

This extends lustre_hsm_attrs struct in MDD to store HSM per-file
information. It adds infrastructure (function, RPCs, ...) to read and
modify it, from MDT code or from user-space on clients through llapi
and lfs command.

This patch implements:
- lfs hsm_state
- lfs hsm_set
- lfs hsm_clear

matching llapi functions:
- llapi_hsm_state_get()
- llapi_hsm_state_set()

matching mdt functions:
- mdt_hsm_state_get()
- mdt_hsm_state_set()

and MDS RPC:
- MDS_HSM_STATE_GET
- MDS_HSM_STATE_SET

Signed-off-by: Aurelien Degremont <aurelien.degremont@cea.fr>
Signed-off-by: JC Lafoucriere <jacques-charles.lafoucriere@cea.fr>
Change-Id: Ic8f82ddc9a56206307c2e5be2523fb7ce42b8638
Reviewed-on: http://review.whamcloud.com/3035
Tested-by: Hudson
Reviewed-by: Johann Lombardi <johann.lombardi@intel.com>
Tested-by: Maloo <whamcloud.maloo@gmail.com>
Reviewed-by: Andreas Dilger <andreas.dilger@intel.com>
7 years agoLU-2638 ldiskfs: update dotdot entry for rename
Fan Yong [Thu, 10 Jan 2013 13:32:57 +0000 (21:32 +0800)]
LU-2638 ldiskfs: update dotdot entry for rename

On master, when rename a directory, its old dotdot entry will
be removed firstly, then insert the new dotdot entry, and try
to append FID-in-dirent. But the space for dotdot entry maybe
not enough to hold the new dotdot with FID-in-dirent, such as
the MDT device restored from file-level backup, or the device
is upgraded from 1.8. If we do not move data in the directory
block, the FID-in-dirent will overwrite the dx_root and cause
the directory crashed. Currently, we do not want to introduce
complex logic to handle directory data moving, instead, under
such case, ignore the FID-in-dirent for the new dotdot entry,
and just insert the new dotdot entry.

Signed-off-by: Fan Yong <fan.yong@intel.com>
Change-Id: Id38f5136fb2bc607344c37ca10a32e63ab427e3d
Reviewed-on: http://review.whamcloud.com/5179
Tested-by: Hudson
Tested-by: Maloo <whamcloud.maloo@gmail.com>
Reviewed-by: Andreas Dilger <andreas.dilger@intel.com>
Reviewed-by: Alex Zhuravlev <alexey.zhuravlev@intel.com>
7 years agoLU-1187 lod: move seq client init from MDT to LOD
wangdi [Wed, 2 Oct 2013 11:40:56 +0000 (04:40 -0700)]
LU-1187 lod: move seq client init from MDT to LOD

Move seq client init from MDT to LOD, so to unify
config log for adding OSP and MD OSP.

Change-Id: I93f91c0dd6f3568196a99eecf422e0b409abda31
Signed-off-by: Wang Di <di.wang@whamcloud.com>
Reviewed-on: http://review.whamcloud.com/4337
Tested-by: Hudson
Tested-by: Maloo <whamcloud.maloo@gmail.com>
Reviewed-by: Alex Zhuravlev <alexey.zhuravlev@intel.com>
Reviewed-by: Mike Pershin <mike.pershin@intel.com>
Reviewed-by: Andreas Dilger <andreas.dilger@intel.com>
7 years agoLU-1187 lod: Add remote object for DNE
wangdi [Tue, 1 Oct 2013 11:13:00 +0000 (04:13 -0700)]
LU-1187 lod: Add remote object for DNE

LOD will do FLD lookup to check whether the object
is for OST or remote MDT object.

Because there are remote object invlove, it might
return some error other than ENOMEM for object allocation.

Signed-off-by: wang di <di.wang@intel.com>
Change-Id: Ie226b3c475524fc7e705665c6200e4c66d50acf0
Reviewed-on: http://review.whamcloud.com/4924
Tested-by: Hudson
Reviewed-by: Alex Zhuravlev <alexey.zhuravlev@intel.com>
Reviewed-by: Mike Pershin <mike.pershin@intel.com>
Tested-by: Maloo <whamcloud.maloo@gmail.com>
Reviewed-by: Oleg Drokin <green@whamcloud.com>
7 years agoLU-2062 hsm: llapi hsm calls for CopyTool
jcl [Mon, 3 Dec 2012 20:13:13 +0000 (21:13 +0100)]
LU-2062 hsm: llapi hsm calls for CopyTool

llapi hsm calls for CopyTool
- llapi_hsm_copytool_start()
- llapi_hsm_copytool_fini()
- llapi_hsm_copytool_recv()
- llapi_hsm_copytool_free()
- llapi_hsm_copy_start()
- llapi_hsm_copy_end()
- llapi_hsm_progress()
- llapi_hsm_import()

and related MDT functions
- mdt_hsm_ct_register()
- mdt_hsm_ct_unregister()
- mdt_hsm_progress()

The new RPC are:
- MDS_HSM_CT_REGISTER
- MDS_HSM_CT_UNREGISTER
- MDS_HSM_PROGRESS

Signed-off-by: JC Lafoucriere <jacques-charles.lafoucriere@cea.fr>
Change-Id: I93617ac6e66a3324344b57f1a22d7c851d715262
Reviewed-on: http://review.whamcloud.com/4736
Tested-by: Hudson
Reviewed-by: Johann Lombardi <johann.lombardi@intel.com>
Reviewed-by: Andreas Dilger <andreas.dilger@intel.com>
Tested-by: Maloo <whamcloud.maloo@gmail.com>
7 years agoNew tag 2.3.60 2.3.60 v2_3_60 v2_3_60_0
Oleg Drokin [Wed, 30 Jan 2013 18:53:09 +0000 (13:53 -0500)]
New tag 2.3.60

Also another can kicking for LU-2449, hopefully last this time.

Change-Id: I1455e48f1b557bc8787ef234e1894b0f9d7f4843
Signed-off-by: Oleg Drokin <green@whamcloud.com>
7 years agoLU-2183 osp: cleanup osp-on-ost
wangdi [Wed, 9 Oct 2013 02:56:07 +0000 (19:56 -0700)]
LU-2183 osp: cleanup osp-on-ost

With new quota design and FIDonOST introduced, a new connection
from OST/MDTn to MDT0 is required for sending quota and sequence
RPCs. This connection was implemented by adding a new special
osp device named osp-on-ost, which introduced complexity to osp
code, and many hacks of checking the osp device name in both
client and server side code.

In this patch, we removed the osp-on-ost, and introduced a new
device type lwp (Light Weight Proxy) to manage the light weight
connection established from OSTs/MDTs to MDT0.

Signed-off-by: Niu Yawei <niu@whamcloud.com>
Signed-off-by: Wang Di <di.wang@whamcloud.com>
Change-Id: Ic1ba25f941a4f971a542dc4be8a81e28ee0f477b
Reviewed-on: http://review.whamcloud.com/4923
Tested-by: Hudson
Reviewed-by: Andreas Dilger <andreas.dilger@intel.com>
Reviewed-by: Alex Zhuravlev <alexey.zhuravlev@intel.com>
Reviewed-by: Mike Pershin <mike.pershin@intel.com>
Tested-by: Maloo <whamcloud.maloo@gmail.com>
7 years agoLU-398 ptlrpc: Add the NRS framework and FIFO policy
Nikitas Angelinas [Wed, 20 Jun 2012 09:31:16 +0000 (10:31 +0100)]
LU-398 ptlrpc: Add the NRS framework and FIFO policy

The Network Request Scheduler (NRS) allows a user to control the way
in which RPCs are dispatched from PTLRPC services. This can be used to
various effects, most importantly improved fairness across the
cluster, better I/O performance under certain workloads, and allowing
for QoS semantics to be used at the filesystem level. Central to the
operation of NRS are NRS policies, each of which implements handling
of RPCs in a different way, aiming to achieve a particular goal.

This patch adds the core NRS framework and the NRS FIFO policy, which
is a logical wrapper around previous, non-NRS functionality, and is
used as the default policy for handling all types of RPCs.

Signed-off-by: Nikitas Angelinas <nikitas_angelinas@xyratex.com>
Co-authored-by: Liang Zhen <liang@whamcloud.com>
Change-Id: I7fcd7885cc89b653d9bc482da533aae2f129bdf9
Oracle-bug-id: b=13634
Xyratex-bug-id: MRP-73
Reviewed-on: http://review.whamcloud.com/4411
Tested-by: Hudson
Tested-by: Maloo <whamcloud.maloo@gmail.com>
Reviewed-by: Mike Pershin <mike.pershin@intel.com>
Reviewed-by: Liang Zhen <liang.zhen@intel.com>
Reviewed-by: Oleg Drokin <green@whamcloud.com>
7 years agoLU-2441 mdd: Implement volatile file
jcl [Mon, 17 Dec 2012 00:55:30 +0000 (01:55 +0100)]
LU-2441 mdd: Implement volatile file

Add a way to create files which are created unlinked, so
they are automatically removed at last close

User interface is:
- llapi_create_volatile_idx()
- llapi_create_volatile()

Signed-off-by: JC Lafoucriere <jacques-charles.lafoucriere@cea.fr>
Change-Id: Ia33a6832d31b83191e4d364da4350cedbe6eef39
Reviewed-on: http://review.whamcloud.com/4836
Tested-by: Hudson
Reviewed-by: Andreas Dilger <andreas.dilger@intel.com>
Tested-by: Maloo <whamcloud.maloo@gmail.com>
Reviewed-by: Johann Lombardi <johann.lombardi@intel.com>
Reviewed-by: Jinshan Xiong <jinshan.xiong@intel.com>
7 years agoLU-1187 tests: chmod +x for dir_remote.sh
wangdi [Mon, 28 Oct 2013 08:30:12 +0000 (01:30 -0700)]
LU-1187 tests: chmod +x for dir_remote.sh

chmod +x for dir_remote.sh

Signed-off-by: wang di <di.wang@whamcloud.com>
Change-Id: I77d670594c128d4f18903bd247a49df7f8ead970
Reviewed-on: http://review.whamcloud.com/4997
Tested-by: Hudson
Reviewed-by: Jian Yu <jian.yu@intel.com>
Tested-by: Maloo <whamcloud.maloo@gmail.com>
Reviewed-by: Oleg Drokin <green@whamcloud.com>
7 years agoLU-1876 hsm: account for active IOs correctly
Jinshan Xiong [Wed, 23 Jan 2013 07:32:24 +0000 (23:32 -0800)]
LU-1876 hsm: account for active IOs correctly

We used to use refcount of lsm to account for active IOs but this
turns out wrong because empty layout files don't have a lsm.

In this patch, lov_object::lo_active_ios is invented to account for
active IOs for both raid0 and empty layout;

Signed-off-by: Jinshan Xiong <jinshan.xiong@intel.com>
Change-Id: I1b7aa16367af1d0a54f6c70bb4155b20071af225
Reviewed-on: http://review.whamcloud.com/5157
Tested-by: Hudson
Reviewed-by: Johann Lombardi <johann.lombardi@intel.com>
Tested-by: Maloo <whamcloud.maloo@gmail.com>
Reviewed-by: Oleg Drokin <green@whamcloud.com>
7 years agoLU-1187 tests: add cleanup for sanity 24q.
wangdi [Mon, 12 Nov 2012 22:30:58 +0000 (14:30 -0800)]
LU-1187 tests: add cleanup for sanity 24q.

Add trap cleanup for sanity 24q.

Signed-off-by: wang di <di.wang@intel.com>
Change-Id: Ie8b096cd9906d07812d743cdccdf619b977aaed6
Reviewed-on: http://review.whamcloud.com/4521
Tested-by: Hudson
Reviewed-by: Jian Yu <jian.yu@intel.com>
Tested-by: Maloo <whamcloud.maloo@gmail.com>
Reviewed-by: Andreas Dilger <andreas.dilger@intel.com>
7 years agoLU-1187 mdt: return lock to client for remote dir.
wangdi [Fri, 1 Nov 2013 12:20:10 +0000 (05:20 -0700)]
LU-1187 mdt: return lock to client for remote dir.

In mdt_reint_open, MDT should return LOOKUP inodebit lock for
remote directory.

Signed-off-by: wang di <di.wang@intel.com>
Change-Id: Iaa7c6a13b71b627c5792f4b28bb3c5520b2132e5
Reviewed-on: http://review.whamcloud.com/5026
Tested-by: Hudson
Reviewed-by: Jinshan Xiong <jinshan.xiong@intel.com>
Tested-by: Maloo <whamcloud.maloo@gmail.com>
Reviewed-by: Andreas Dilger <andreas.dilger@intel.com>
Reviewed-by: Oleg Drokin <green@whamcloud.com>
7 years agoLU-1445 ost: enable fid on OST support on OST
wangdi [Thu, 26 Sep 2013 11:57:13 +0000 (04:57 -0700)]
LU-1445 ost: enable fid on OST support on OST

Enable fid on OST support on OST.

Signed-off-by: Wang Di <di.wang@intel.com>
Change-Id: I51c242e7630d52c578f4ddfe1149b43eb9d5f7c5
Reviewed-on: http://review.whamcloud.com/4791
Reviewed-by: Mike Pershin <mike.pershin@intel.com>
Tested-by: Hudson
Tested-by: Maloo <whamcloud.maloo@gmail.com>
Reviewed-by: Andreas Dilger <andreas.dilger@intel.com>
Reviewed-by: Alex Zhuravlev <alexey.zhuravlev@intel.com>
7 years agoLU-1506 ldlm: correct lock res build for FID on OST
wangdi [Thu, 26 Sep 2013 13:16:26 +0000 (06:16 -0700)]
LU-1506 ldlm: correct lock res build for FID on OST

Because 1.8/2.1/2.2 client always put oid to res[0] and
oseq to res[1], OFD will do the same to avoid incompatible
problems.

Signed-off-by: wang di <di.wang@intel.com>
Change-Id: Ic4cf66b1bc23631af42530787796510653f82a61
Reviewed-on: http://review.whamcloud.com/3081
Tested-by: Hudson
Reviewed-by: Mike Pershin <mike.pershin@intel.com>
Tested-by: Maloo <whamcloud.maloo@gmail.com>
Reviewed-by: Andreas Dilger <andreas.dilger@intel.com>
7 years agoLU-1445 mdt: allow lightweight connection even if no OST
wangdi [Sat, 5 Oct 2013 14:33:55 +0000 (07:33 -0700)]
LU-1445 mdt: allow lightweight connection even if no OST

MDT should allow lightweight connection from OST even if there
is no OST being activated, because OST might need do FLD lookup
during recovery phase.

Signed-off-by: wang di <di.wang@intel.com>
Change-Id: Ica44aafdbcbf29f10d314bc65de8c1f5ff5d2940
Reviewed-on: http://review.whamcloud.com/4920
Reviewed-by: Andreas Dilger <andreas.dilger@intel.com>
Tested-by: Hudson
Tested-by: Maloo <whamcloud.maloo@gmail.com>
Reviewed-by: Alex Zhuravlev <alexey.zhuravlev@intel.com>
Reviewed-by: Mike Pershin <mike.pershin@intel.com>
7 years agoLU-1445 fld: allow fld lookup during recovery
wangdi [Fri, 4 Oct 2013 11:47:25 +0000 (04:47 -0700)]
LU-1445 fld: allow fld lookup during recovery

Allow fld lookup during recovery process, so OST can send fld
lookup req to MDT0 during recovery phase, instead of waiting
the recovery is done.

Signed-off-by: wang di <di.wang@intel.com>
Change-Id: I69990eb4225d98f20ce2dc197cd3c16b5c6f9617
Reviewed-on: http://review.whamcloud.com/4919
Reviewed-by: Mike Pershin <mike.pershin@intel.com>
Tested-by: Hudson
Reviewed-by: Andreas Dilger <andreas.dilger@intel.com>
Tested-by: Maloo <whamcloud.maloo@gmail.com>
Reviewed-by: Alex Zhuravlev <alexey.zhuravlev@intel.com>
7 years agoLU-1445 osd: Add multiple sequence support for osd-zfs
wangdi [Tue, 1 Oct 2013 03:16:27 +0000 (20:16 -0700)]
LU-1445 osd: Add multiple sequence support for osd-zfs

Add osd_seq_list for osd-zfs, then it can allocate new sequence
after FID on OST is landed.

Test-Parameters: mdtfilesystemtype=zfs ostfilesystemtype=zfs
Signed-off-by: wang di <di.wang@intel.com>
Change-Id: I66bae3987a1ae167aa48a3a1d4b6e50f300f1dd8
Reviewed-on: http://review.whamcloud.com/4837
Reviewed-by: Andreas Dilger <andreas.dilger@intel.com>
Tested-by: Hudson
Tested-by: Maloo <whamcloud.maloo@gmail.com>
Reviewed-by: Alex Zhuravlev <alexey.zhuravlev@intel.com>
Reviewed-by: Mike Pershin <mike.pershin@intel.com>
7 years agoLU-1445 osd: add fld look up in osd.
wangdi [Thu, 26 Sep 2013 11:48:51 +0000 (04:48 -0700)]
LU-1445 osd: add fld look up in osd.

Lookup FLDB during oi_insert, so we know it is MDT or
OST objects. And it also needs to init FLD service
on OFD as well, so object on OST can lookup FLDB
as well.

Change fci_lock to read/write lock, since fld_cache_lookup
will become the hot path after FLD lookup is added in OSD
layer.

Change-Id: I79ef723de603214b7136a51c36f0cb63513b5640
Signed-off-by: Wang Di <di.wang@intel.com>
Reviewed-on: http://review.whamcloud.com/4330
Reviewed-by: Mike Pershin <mike.pershin@intel.com>
Tested-by: Hudson
Tested-by: Maloo <whamcloud.maloo@gmail.com>
Reviewed-by: Andreas Dilger <andreas.dilger@intel.com>
Reviewed-by: Alex Zhuravlev <alexey.zhuravlev@intel.com>
7 years agoLU-1445 osp: rollover to the new seq synchronously
wangdi [Thu, 26 Sep 2013 09:57:55 +0000 (02:57 -0700)]
LU-1445 osp: rollover to the new seq synchronously

For simplicity, OSP will track only one sequence for precreation,
so it will need

1. Stop precreation when the current sequence is about to used up,
until all of the precreation objects has been used up.

2. ofd will precreate objects synchronously when the current sequence
is about to used up.

Signed-off-by: wang di <di.wang@intel.com>
Change-Id: I4c0de8f2869c9a8e7a4b0d2abe06041e84885b2e
Reviewed-on: http://review.whamcloud.com/4790
Reviewed-by: Mike Pershin <mike.pershin@intel.com>
Reviewed-by: Andreas Dilger <andreas.dilger@intel.com>
Tested-by: Hudson
Tested-by: Maloo <whamcloud.maloo@gmail.com>
Reviewed-by: Alex Zhuravlev <alexey.zhuravlev@intel.com>
7 years agoLU-1445 osp: Use FID to track precreate cache.
wangdi [Thu, 26 Sep 2013 02:24:17 +0000 (19:24 -0700)]
LU-1445 osp: Use FID to track precreate cache.

Change last_used/pre_used/pre_created_id to real fid, so
to track precreate cache with real fid.

For simplicity, precreate/orphan_cleanup between osp and
OFD must be within the same sequence.

Precreate process will be:

1. OSP will request one sequence from OST initially.

2. During precreation, osp will allocate a group of fids
for the precreation request within the current sequence,
then send the request to OST(OFD) with these fids.

3. When getting the precreate request, OST will precreate
the objects by those fids in the request.

Signed-off-by: wang di <di.wang@intel.com>
Change-Id: Ib7f1016321c064656158c54f4475922a199e19b3
Reviewed-on: http://review.whamcloud.com/4789
Reviewed-by: Mike Pershin <mike.pershin@intel.com>
Reviewed-by: Andreas Dilger <andreas.dilger@intel.com>
Tested-by: Hudson
Tested-by: Maloo <whamcloud.maloo@gmail.com>
Reviewed-by: Alex Zhuravlev <alexey.zhuravlev@intel.com>
7 years agoLU-2674 osp: set proper connection flag
Niu Yawei [Fri, 25 Jan 2013 04:05:30 +0000 (23:05 -0500)]
LU-2674 osp: set proper connection flag

Set proper connection flag for the osp-on-ost.

Signed-off-by: Niu Yawei <yawei.niu@intel.com>
Change-Id: Idf80df593f22309565c92630f5c04fc3781e3d34
Reviewed-on: http://review.whamcloud.com/5167
Reviewed-by: wangdi <di.wang@intel.com>
Reviewed-by: Doug Oucharek <doug.s.oucharek@intel.com>
Tested-by: Hudson
Reviewed-by: Andreas Dilger <andreas.dilger@intel.com>
Tested-by: Maloo <whamcloud.maloo@gmail.com>
7 years agoLU-2561 mdd: create decares attr set with parent attr
Lai Siyao [Wed, 12 Dec 2012 05:20:33 +0000 (13:20 +0800)]
LU-2561 mdd: create decares attr set with parent attr

mdd_declare_create() mistakenly declares attr set for child with
parent attr, as will cause truncate operation.

Signed-off-by: Lai Siyao <lai.siyao@intel.com>
Change-Id: I79dbbb2b52e1d41aaa70c36ce6c197ab2cd684d8
Reviewed-on: http://review.whamcloud.com/4956
Reviewed-by: Prakash Surya <surya1@llnl.gov>
Tested-by: Hudson
Reviewed-by: Bobi Jam <bobijam@gmail.com>
Tested-by: Maloo <whamcloud.maloo@gmail.com>
Reviewed-by: Oleg Drokin <green@whamcloud.com>
7 years agoLU-2555 ldiskfs: fix ext4_mb_add_n_trim()
Niu Yawei [Sun, 6 Jan 2013 05:16:36 +0000 (00:16 -0500)]
LU-2555 ldiskfs: fix ext4_mb_add_n_trim()

In ext4_mb_add_n_trim(), lg_prealloc_lock should be taken
when changing the lg_prealloc_list.

Signed-off-by: Niu Yawei <yawei.niu@intel.com>
Change-Id: Icc62d87a682b0ebf70ad12e148a52b864bf1fdb8
Reviewed-on: http://review.whamcloud.com/4961
Reviewed-by: Bobi Jam <bobijam@gmail.com>
Tested-by: Hudson
Reviewed-by: Alex Zhuravlev <alexey.zhuravlev@intel.com>
Tested-by: Maloo <whamcloud.maloo@gmail.com>
Reviewed-by: Oleg Drokin <green@whamcloud.com>
7 years agoLU-2550 osc: set resend count properly
Niu Yawei [Mon, 14 Jan 2013 04:01:30 +0000 (23:01 -0500)]
LU-2550 osc: set resend count properly

The resend count of new io request should be set properly
in osc_brw_redo_request().

Signed-off-by: Niu Yawei <yawei.niu@intel.com>
Change-Id: Ifce58cf45ae3bfa3b73427f732e281b468cf982b
Reviewed-on: http://review.whamcloud.com/5012
Tested-by: Hudson
Reviewed-by: Bobi Jam <bobijam@gmail.com>
Reviewed-by: Johann Lombardi <johann.lombardi@intel.com>
Tested-by: Maloo <whamcloud.maloo@gmail.com>
7 years agoLU-2517 kernel: Kernel update [RHEL6.3 2.6.32-279.19.1.el6]
yangsheng [Wed, 9 Jan 2013 08:32:14 +0000 (16:32 +0800)]
LU-2517 kernel: Kernel update [RHEL6.3 2.6.32-279.19.1.el6]

Update RHEL6.3 kernel to 2.6.32-279.19.1.el6.

Signed-off-by: yang sheng <ys@whamcloud.com>
Change-Id: I1e89c76d4ec0cc736a0e479a465e59e2262522c0
Reviewed-on: http://review.whamcloud.com/4980
Tested-by: Hudson
Tested-by: Maloo <whamcloud.maloo@gmail.com>
Reviewed-by: Bobi Jam <bobijam@gmail.com>
Reviewed-by: Lai Siyao <lai.siyao@intel.com>
Reviewed-by: Oleg Drokin <green@whamcloud.com>
7 years agoLU-1486 tests: fix sanityn test_45g() to add error check
Yu Jian [Thu, 30 Aug 2012 06:50:10 +0000 (14:50 +0800)]
LU-1486 tests: fix sanityn test_45g() to add error check

This patch fixes sanityn test_45g() to add error check for
"stat must fail".

Signed-off-by: Yu Jian <yujian@whamcloud.com>
Change-Id: Iff65f2684244ab58e0d0cab68db0795f5604187e
Reviewed-on: http://review.whamcloud.com/3826
Tested-by: Hudson
Reviewed-by: Wei Liu <wei3.liu@intel.com>
Tested-by: Maloo <whamcloud.maloo@gmail.com>
Reviewed-by: Minh Diep <minh.diep@intel.com>
Reviewed-by: Oleg Drokin <green@whamcloud.com>
7 years agoLU-661 tests: don't check file-2 in replay-dual 0b
Mikhail Pershin [Tue, 22 Nov 2011 06:59:26 +0000 (10:59 +0400)]
LU-661 tests: don't check file-2 in replay-dual 0b

It is uncertain was it replayed on not yet, so it might exist
and test will fail. Remove wrong check from test

Signed-off-by: Mikhail Pershin <mike.pershin@intel.com>
Change-Id: I724f389f0504f3c8c388b2b271b3a64cb61339ab
Reviewed-on: http://review.whamcloud.com/4999
Tested-by: Hudson
Reviewed-by: Andreas Dilger <andreas.dilger@intel.com>
Tested-by: Maloo <whamcloud.maloo@gmail.com>
Reviewed-by: Fan Yong <fan.yong@intel.com>
7 years agoLU-2416 test: Some test using wrong lctl params
James Simmons [Mon, 7 Jan 2013 17:55:42 +0000 (12:55 -0500)]
LU-2416 test: Some test using wrong lctl params

With the move of obdfilter to ofd some of the lctl
parameters got moved around. For the test suite a function
{get,set}_obdfilter_param was written to handle this change so
this patch updates a few more test to use this api.

Signed-off-by: James Simmons <uja.ornl@gmail.com>
Change-Id: I05ca4711b198061de1378b79d2f7f35d5a1c75af
Reviewed-on: http://review.whamcloud.com/4726
Reviewed-by: Li Wei <wei.g.li@intel.com>
Reviewed-by: Jian Yu <jian.yu@intel.com>
Tested-by: Hudson
Tested-by: Maloo <whamcloud.maloo@gmail.com>
Reviewed-by: Oleg Drokin <green@whamcloud.com>
7 years agoLU-2158 fsfilt: Remove bits from fsfilt_ext3.c/lustre_fsfilt.h
Prakash Surya [Tue, 18 Dec 2012 22:15:38 +0000 (14:15 -0800)]
LU-2158 fsfilt: Remove bits from fsfilt_ext3.c/lustre_fsfilt.h

The only remaining dependency needed from the fsfilt_ext3.c for use in
osd-ldiskfs is the fsfilt_ext3_map_inode_pages function. Thus much of
the fsfilt_ext3.c contents have been removed by this commit because
they are no longer used.

Also, since much of the fsfilt_operations structure is not used anymore,
that structure was trimmed to the bare essentials needed by the
osd-ldiskfs layer. The related inline helper functions were also removed.

Change-Id: Ie0f7d428d15862b1b17df140b7152cb0b56ce3be
Signed-off-by: Prakash Surya <surya1@llnl.gov>
Signed-off-by: Christopher J. Morrone <morrone2@llnl.gov>
Reviewed-on: http://review.whamcloud.com/4863
Tested-by: Hudson
Reviewed-by: Andreas Dilger <andreas.dilger@intel.com>
Reviewed-by: James Simmons <uja.ornl@gmail.com>
Tested-by: Maloo <whamcloud.maloo@gmail.com>
7 years agoLU-2158 lvfs: Remove cruft from lvfs_linux.h
Prakash Surya [Wed, 19 Oct 2011 19:01:08 +0000 (12:01 -0700)]
LU-2158 lvfs: Remove cruft from lvfs_linux.h

Various bits from the lvfs_linux.h file were removed as they are no
longer used. Specifically, the following was removed:

 * l_inode
 * lvfs_sbdev_sync
 * lvfs_set_rdonly
 * lvfs_clear_rdonly
 * lvfs_dentry_params

Change-Id: I993a1416bb548746ca717904a483c0cd6245768e
Signed-off-by: Prakash Surya <surya1@llnl.gov>
Reviewed-on: http://review.whamcloud.com/4862
Reviewed-by: Andreas Dilger <andreas.dilger@intel.com>
Tested-by: Hudson
Tested-by: Maloo <whamcloud.maloo@gmail.com>
Reviewed-by: James Simmons <uja.ornl@gmail.com>
7 years agoLU-2158 lvfs: Remove ll_sleep from linux/lvfs.h
Prakash Surya [Wed, 19 Oct 2011 17:55:37 +0000 (10:55 -0700)]
LU-2158 lvfs: Remove ll_sleep from linux/lvfs.h

The ll_sleep function was removed because it is not used anywhere.

Change-Id: Ic22552c69635f3da5b2cdd3dd67b9930d63a48d1
Signed-off-by: Prakash Surya <surya1@llnl.gov>
Reviewed-on: http://review.whamcloud.com/4861
Reviewed-by: Andreas Dilger <andreas.dilger@intel.com>
Tested-by: Hudson
Tested-by: Maloo <whamcloud.maloo@gmail.com>
Reviewed-by: James Simmons <uja.ornl@gmail.com>
7 years agoLU-2158 lvfs: Remove ll_fid2str from lvfs.h
Prakash Surya [Wed, 19 Oct 2011 17:47:41 +0000 (10:47 -0700)]
LU-2158 lvfs: Remove ll_fid2str from lvfs.h

The ll_fid2str function was removed because it is not used anywhere.

Change-Id: I7e16b3d5e2fa10af0d5763909a42871a8c86a3ff
Signed-off-by: Prakash Surya <surya1@llnl.gov>
Reviewed-on: http://review.whamcloud.com/4860
Reviewed-by: Andreas Dilger <andreas.dilger@intel.com>
Tested-by: Hudson
Tested-by: Maloo <whamcloud.maloo@gmail.com>
Reviewed-by: James Simmons <uja.ornl@gmail.com>
7 years agoLU-2618 test: recovery-small test_24b dumps dmesg
Kyrylo Shatskyy [Tue, 15 Jan 2013 22:22:45 +0000 (00:22 +0200)]
LU-2618 test: recovery-small test_24b dumps dmesg

Redirected output of 'dmesg -c' to null device

Signed-off-by: Kyrylo Shatskyy <kyrylo_shatskyy@xyratex.com>
Change-Id: I6b06122e4c9600aac6da476f7083c1e051d92c98
Reviewed-on: http://review.whamcloud.com/5033
Tested-by: Hudson
Reviewed-by: Andreas Dilger <andreas.dilger@intel.com>
Reviewed-by: Keith Mannthey <keith.mannthey@intel.com>
Tested-by: Maloo <whamcloud.maloo@gmail.com>
7 years agoLU-2590 lod: magic changed after swab
Bobi Jam [Fri, 11 Jan 2013 04:08:30 +0000 (12:08 +0800)]
LU-2590 lod: magic changed after swab

In lod_qos_parse_config(), the stack magic variable should be changed
to native endianness as well.

Signed-off-by: Bobi Jam <bobijam.xu@intel.com>
Change-Id: I940c40f20f1ec55d7b55eeffe8d61404a1badfdb
Reviewed-on: http://review.whamcloud.com/4996
Tested-by: Hudson
Reviewed-by: Andreas Dilger <andreas.dilger@intel.com>
Tested-by: Maloo <whamcloud.maloo@gmail.com>
Reviewed-by: Alex Zhuravlev <alexey.zhuravlev@intel.com>
7 years agoLU-2558 test: recovery-small 19a FAIL no eviction
yangsheng [Thu, 10 Jan 2013 13:15:39 +0000 (21:15 +0800)]
LU-2558 test: recovery-small 19a FAIL no eviction

Use stat instead of mcreate since latter cannot
ensure enqueue a lock.

Signed-off-by: yang sheng <ys@whamcloud.com>
Change-Id: Ifb63f679305e707df64457c4de599ca1b5fb0f00
Reviewed-on: http://review.whamcloud.com/4987
Reviewed-by: Fan Yong <fan.yong@intel.com>
Reviewed-by: Jian Yu <jian.yu@intel.com>
Tested-by: Hudson
Reviewed-by: Lai Siyao <lai.siyao@intel.com>
Tested-by: Maloo <whamcloud.maloo@gmail.com>
Reviewed-by: Oleg Drokin <green@whamcloud.com>
7 years agoLU-1270 utils: properly initialize traverse param
James Simmons [Wed, 12 Dec 2012 15:57:41 +0000 (10:57 -0500)]
LU-1270 utils: properly initialize traverse param

The find_param::mdtindex intialization was neglected.

Signed-off-by: James Simmons <uja.ornl@gmail.com>
Signed-off-by: Bobi Jam <bobijam@whamcloud.com>
Change-Id: I3018c346a4bbab0d7da7c7efdeeb201098ac1e53
Reviewed-on: http://review.whamcloud.com/2411
Tested-by: Hudson
Tested-by: Maloo <whamcloud.maloo@gmail.com>
Reviewed-by: Andreas Dilger <andreas.dilger@intel.com>
7 years agoLU-2309 config: ignore unknown configuration param
Jian Yu [Wed, 16 Jan 2013 06:37:35 +0000 (14:37 +0800)]
LU-2309 config: ignore unknown configuration param

Client or server should not fail to mount if it hits
a configuration parameter that it doesn't understand.
This patch fixes class_process_config() to meet
the above requirement.

The patch also improves conf-sanity test 42 to verify
that invalid sys config param should not prevent client
or server from mounting.

Signed-off-by: Jian Yu <jian.yu@intel.com>
Change-Id: I33a6eb3c2beb3dd7425a097e8e7b5a2766852a96
Reviewed-on: http://review.whamcloud.com/5037
Tested-by: Hudson
Reviewed-by: Andreas Dilger <andreas.dilger@intel.com>
Reviewed-by: Lai Siyao <lai.siyao@intel.com>
Tested-by: Maloo <whamcloud.maloo@gmail.com>
7 years agoLU-2336 llog: hide the harmless log message
Hongchao Zhang [Sun, 13 Jan 2013 13:15:49 +0000 (21:15 +0800)]
LU-2336 llog: hide the harmless log message

in llog_cat_add_rec, the llog_write_rec could return -ENOSPC
if the current log is full, and it will move to the next log
file, then it should not be treated as an error.

Change-Id: Ie2e203dc01572e170e9a383c63dcc670db093355
Signed-off-by: Hongchao Zhang <hongchao.zhang@whamcloud.com>
Reviewed-on: http://review.whamcloud.com/5146
Tested-by: Hudson
Reviewed-by: Prakash Surya <surya1@llnl.gov>
Tested-by: Maloo <whamcloud.maloo@gmail.com>
Reviewed-by: Andreas Dilger <andreas.dilger@intel.com>
7 years agoLU-2158 lvfs: Remove functions from lvfs_linux.c
Prakash Surya [Tue, 18 Dec 2012 21:19:02 +0000 (13:19 -0800)]
LU-2158 lvfs: Remove functions from lvfs_linux.c

The following functions were removed from the lvfs_linux.c file as they
are no longer used:

 * simple_mknod
 * lustre_fread
 * lustre_fwrite
 * lustre_fsync
 * l_filldir
 * l_readdir
 * l_notify_change
 * simple_truncate
 * lvfs_check_io_health

Change-Id: I6684815c85e87974277295e6906fd6d7c8dfe1be
Signed-off-by: Prakash Surya <surya1@llnl.gov>
Signed-off-by: Christopher J. Morrone <morrone2@llnl.gov>
Reviewed-on: http://review.whamcloud.com/4859
Reviewed-by: James Simmons <uja.ornl@gmail.com>
Tested-by: Hudson
Tested-by: Maloo <whamcloud.maloo@gmail.com>
Reviewed-by: Andreas Dilger <andreas.dilger@intel.com>
7 years agoLU-2644 build: fix 'resource leak' errors
Sebastien Buisson [Fri, 18 Jan 2013 15:00:40 +0000 (16:00 +0100)]
LU-2644 build: fix 'resource leak' errors

Fix 'resource leak' defects found by Coverity version 6.0.3:
Resource leak (RESOURCE_LEAK)
Variable going out of scope leaks the storage it points to.

Signed-off-by: Sebastien Buisson <sebastien.buisson@bull.net>
Change-Id: I9fae1905848a83393b7258449bfafd1fc0a342a7
Reviewed-on: http://review.whamcloud.com/5131
Reviewed-by: Bob Glossman <bob.glossman@intel.com>
Reviewed-by: Keith Mannthey <keith.mannthey@intel.com>
Tested-by: Hudson
Reviewed-by: Andreas Dilger <andreas.dilger@intel.com>
Tested-by: Maloo <whamcloud.maloo@gmail.com>
Reviewed-by: Oleg Drokin <green@whamcloud.com>
7 years agoLU-744 clio: save memory allocations for cl_page
Jinshan Xiong [Thu, 20 Dec 2012 02:07:45 +0000 (18:07 -0800)]
LU-744 clio: save memory allocations for cl_page

It used to allocate 6 piece of memory to create a page:
 - top cl_page, llite & lov page slice;
 - subpage, lovsub_page & osc_page.

This patch will allocate top cl_page, llite and lov page slice in
the same piece of memory, similar to sub pages, so that only 2
memory allocations will be needed for each page.

Signed-off-by: Jinshan Xiong <jinshan.xiong@intel.com>
Change-Id: I1c4dd0541bda23e7d739131f73bcd895b8dd2c13
Reviewed-on: http://review.whamcloud.com/4943
Tested-by: Hudson
Tested-by: Maloo <whamcloud.maloo@gmail.com>
Reviewed-by: Fan Yong <fan.yong@intel.com>
Reviewed-by: Andreas Dilger <andreas.dilger@intel.com>
7 years agoLU-1199 build: Assume that ldiskfs PDO is now standard
Christopher J. Morrone [Fri, 11 Jan 2013 23:35:00 +0000 (15:35 -0800)]
LU-1199 build: Assume that ldiskfs PDO is now standard

PDO (Parallel Directory Operations) is now a standard ldiskfs feature.
The old patch series files that lacked PDO support have been removed
from the tree, so now we can drop all of the places that conditionally
use PDO, and just assume that PDO is standard.

Also remove support for rhel5 series file in ldiskfs, since that
no longer exists in the tree.

Change-Id: I76b5b31a4614c1ade3b4b5890a111cdabc4afa63
Signed-off-by: Christopher J. Morrone <morrone2@llnl.gov>
Reviewed-on: http://review.whamcloud.com/5004
Tested-by: Hudson
Reviewed-by: Andreas Dilger <andreas.dilger@intel.com>
Tested-by: Maloo <whamcloud.maloo@gmail.com>
Reviewed-by: Liang Zhen <liang.zhen@intel.com>
Reviewed-by: James Simmons <uja.ornl@gmail.com>
7 years agoLU-1095 debug: Quiet/cleanup various common console messages
Brian Behlendorf [Fri, 19 Feb 2010 22:39:40 +0000 (14:39 -0800)]
LU-1095 debug: Quiet/cleanup various common console messages

Turn off several common message we always observe when restarting
servers assuming they are for the expected return code.  Simply
move others which are not helpful to an administrator to the
internal debug kernel log.  Less noise means you are more likely
to spot the important error messages.

Signed-off-by: Christopher J. Morrone <morrone2@llnl.gov>
Signed-off-by: Prakash Surya <surya1@llnl.gov>
Change-Id: I523eacfe3c72480d62c915537427aa6c99428647
Reviewed-on: http://review.whamcloud.com/2201
Tested-by: Hudson
Reviewed-by: Mike Pershin <mike.pershin@intel.com>
Tested-by: Maloo <whamcloud.maloo@gmail.com>
Reviewed-by: Oleg Drokin <green@whamcloud.com>
7 years agoLU-1199 build: Remove autoMakefile.am.toplevel
Christopher J. Morrone [Tue, 18 Dec 2012 19:31:28 +0000 (11:31 -0800)]
LU-1199 build: Remove autoMakefile.am.toplevel

Now that ldiskfs no longer shares lustre's build system,
there is no need for the added complication of
the autoMakefile.am.toplevel file.  We move its contents
into the actual top level autoMakefile.am file.

Change-Id: Ie2787465f1b60146dd44c2e4564e236782de1afa
Signed-off-by: Christopher J. Morrone <morrone2@llnl.gov>
Reviewed-on: http://review.whamcloud.com/4975
Tested-by: Hudson
Reviewed-by: James Simmons <uja.ornl@gmail.com>
Reviewed-by: Andreas Dilger <andreas.dilger@intel.com>
Tested-by: Maloo <whamcloud.maloo@gmail.com>
Reviewed-by: Minh Diep <minh.diep@intel.com>
7 years agoLU-2361 quota: quiet warning when acct isn't enabled
Johann Lombardi [Tue, 8 Jan 2013 09:34:32 +0000 (10:34 +0100)]
LU-2361 quota: quiet warning when acct isn't enabled

Don't print a warning to the console on every mount when space
accounting is disabled.

Signed-off-by: Johann Lombardi <johann.lombardi@intel.com>
Change-Id: I7e1855c21f509b755c0452824eefe1084a1c33f1
Reviewed-on: http://review.whamcloud.com/4968
Tested-by: Hudson
Reviewed-by: Niu Yawei <yawei.niu@intel.com>
Reviewed-by: Andreas Dilger <andreas.dilger@intel.com>
Tested-by: Maloo <whamcloud.maloo@gmail.com>
7 years agoLU-1199 build: Simplify autotools version checks
Christopher J. Morrone [Wed, 28 Nov 2012 04:55:57 +0000 (20:55 -0800)]
LU-1199 build: Simplify autotools version checks

Simplify the check for minumum autoconf and automake
versions by using standard macros in configure.ac
rather than using shell functions in autogen.sh.

Note that I am bumping the minimum automake version
from 1.9 to 1.10.

Change-Id: I009c316a54d8d517abea0d1808dc244e2353b2aa
Signed-off-by: Christopher J. Morrone <morrone2@llnl.gov>
Reviewed-on: http://review.whamcloud.com/4693
Reviewed-by: Andreas Dilger <andreas.dilger@intel.com>
Tested-by: Hudson
Tested-by: Maloo <whamcloud.maloo@gmail.com>
Reviewed-by: Brian J. Murrell <brian.murrell@intel.com>
Reviewed-by: Minh Diep <minh.diep@intel.com>
7 years agoLU-2602 tests: add missing bug number to error_ignore call
Bob Glossman [Fri, 11 Jan 2013 01:07:40 +0000 (17:07 -0800)]
LU-2602 tests: add missing bug number to error_ignore call

error_ignore called with the wrong number of arguments
led to a corrupted and hard to interpret test log

Signed-off-by: Bob Glossman <bob.glossman@intel.com>
Change-Id: I6d2150eaa3c1b8cf044ddafd40d66cd64bb25314
Reviewed-on: http://review.whamcloud.com/4995
Tested-by: Hudson
Reviewed-by: Andreas Dilger <andreas.dilger@intel.com>
Tested-by: Maloo <whamcloud.maloo@gmail.com>
Reviewed-by: Minh Diep <minh.diep@intel.com>
Reviewed-by: Keith Mannthey <keith.mannthey@intel.com>
7 years agoLU-1812 mgs: assign PTR_ERR() to prevent warnings
chas williams - CONTRACTOR [Tue, 8 Jan 2013 15:06:18 +0000 (10:06 -0500)]
LU-1812 mgs: assign PTR_ERR() to prevent warnings

Some newer kernels fail with the "error: ignoring return value of
'PTR_ERR', declared with attribute warn_unused_result".

Signed-off-by: chas williams - CONTRACTOR <chas@cmf.nrl.navy.mil>
Change-Id: I94ff7d8371e2daf7440dd5fb295c209ed671bc74
Reviewed-on: http://review.whamcloud.com/4971
Reviewed-by: James Simmons <uja.ornl@gmail.com>
Tested-by: Hudson
Tested-by: Maloo <whamcloud.maloo@gmail.com>
Reviewed-by: Peng Tao <bergwolf@gmail.com>
Reviewed-by: Andreas Dilger <andreas.dilger@intel.com>
7 years agoLU-2607 mdd: preserve input arguments in mdd_create()
Alex Zhuravlev [Thu, 17 Jan 2013 07:43:08 +0000 (11:43 +0400)]
LU-2607 mdd: preserve input arguments in mdd_create()

all input arguments are supposed to stay as is so that
the caller can re-use them again.

Signed-off-by: Alex Zhuravlev <alexey.zhuravlev@intel.com>
Change-Id: Ie2be78e702bfa9b7febafe7d106f09beccbf48d3
Reviewed-on: http://review.whamcloud.com/5048
Reviewed-by: wangdi <di.wang@intel.com>
Reviewed-by: Mike Pershin <mike.pershin@intel.com>
Tested-by: Hudson
Tested-by: Maloo <whamcloud.maloo@gmail.com>
Reviewed-by: Oleg Drokin <green@whamcloud.com>
7 years agoLU-2563 tests: Print correct names in mgsdevname()
Li Wei [Fri, 4 Jan 2013 03:55:18 +0000 (11:55 +0800)]
LU-2563 tests: Print correct names in mgsdevname()

Current mgsdevname() prints names of block devices (or files for loop
devices) regardless of MGS back end types.  This is plainly a
violation of what the function is expected to do.  If an MGS is
ZFS-based, mgsdevname() should print the dataset name instead of the
VDev names.  This patch fixes the regression and makes sure that
mgsdevname() and mgsvdevname() work for both standalone and
MDT-combined MGSs.

Signed-off-by: Li Wei <wei.g.li@intel.com>
Change-Id: I5885af48e811fcd98aa00c5d901db30e5837b505
Signed-off-by: Li Wei <wei.g.li@intel.com>
Reviewed-on: http://review.whamcloud.com/4950
Tested-by: Hudson
Reviewed-by: Jian Yu <jian.yu@intel.com>
Tested-by: Maloo <whamcloud.maloo@gmail.com>
Reviewed-by: wangdi <di.wang@intel.com>
Reviewed-by: Oleg Drokin <green@whamcloud.com>
7 years agoLU-2587 quota: bump version after migration
Niu Yawei [Tue, 8 Jan 2013 10:39:56 +0000 (05:39 -0500)]
LU-2587 quota: bump version after migration

The global index version should be bumped to 2 if migration
happened, otherwise, the global index copy version on slave will
be same as the global index's (both equal to 1 after creation),
and the migrated limits will fail to be reintegrated.

Signed-off-by: Niu Yawei <yawei.niu@intel.com>
Change-Id: I2d596a33e2eb4495dfbb7bd63d37199297a9b214
Reviewed-on: http://review.whamcloud.com/4969
Reviewed-by: Johann Lombardi <johann.lombardi@intel.com>
Tested-by: Hudson
Reviewed-by: Fan Yong <fan.yong@intel.com>
Tested-by: Maloo <whamcloud.maloo@gmail.com>
7 years agoLU-2072 ldiskfs: Leak of s_mb_prealloc_table in ldiskfs
Bruno Faccini [Thu, 10 Jan 2013 14:27:25 +0000 (15:27 +0100)]
LU-2072 ldiskfs: Leak of s_mb_prealloc_table in ldiskfs

ext4-prealloc-rhel[5,6] patches adds kmalloc'ed
sbi->s_mb_prealloc_table field which is never freed
in [ldiskfs,ext4]_mb_release()/ext4-vmalloc-rhel[5,6].patch

Signed-off-by: Bruno Faccini <bruno.faccini@intel.com>
Change-Id: I0ca3ff1e7ae8daa0260cd070c5ac52c843248a1b
Reviewed-on: http://review.whamcloud.com/4683
Tested-by: Hudson
Reviewed-by: Andreas Dilger <andreas.dilger@intel.com>
Reviewed-by: Johann Lombardi <johann.lombardi@intel.com>
Tested-by: Maloo <whamcloud.maloo@gmail.com>
7 years agoLU-2321 debug: Quiet uninitialized target connects
Prakash Surya [Tue, 20 Nov 2012 21:04:03 +0000 (13:04 -0800)]
LU-2321 debug: Quiet uninitialized target connects

When a target is not fully initialized and receives a connection
attempt, it will refuse the connection and print a message to the
console stating this fact. This is "normal operation", so rather
than using LCONSOLE_WARN to convey this information, this patch
modifies it to use CDEBUG.

Signed-off-by: Prakash Surya <surya1@llnl.gov>
Change-Id: Ia5d098fcfe2ad9ae2a1406166087db8181496a7d
Reviewed-on: http://review.whamcloud.com/4640
Tested-by: Hudson
Reviewed-by: Andreas Dilger <andreas.dilger@intel.com>
Tested-by: Maloo <whamcloud.maloo@gmail.com>
Reviewed-by: Oleg Drokin <green@whamcloud.com>
7 years agoLU-1665 libcfs: alignment of hsb_head[0] can be suboptimal
chas williams - CONTRACTOR [Fri, 10 Aug 2012 19:21:10 +0000 (15:21 -0400)]
LU-1665 libcfs: alignment of hsb_head[0] can be suboptimal

On ia64, hsb_head[0] of struct cfs_hash_bucket could be aligned to a
4-byte boundary.  However, after allocating the hash bucket and
casting, the first member of the struct pointed to by hsb_head[0]
might prefer an 8-byte alignment.  This causes a constant stream
of unaligned access warnings from the kernel.

For other cpus this might waste some small amount of space, but
potentially produces more natural alignments (8 byte objects on 8
byte boundaries).

Signed-off-by: chas williams - CONTRACTOR <chas@cmf.nrl.navy.mil>
Change-Id: I4561ab3e6553b5dcef840b8e4e9aa3023b7c3586
Reviewed-on: http://review.whamcloud.com/4965
Reviewed-by: Andreas Dilger <andreas.dilger@intel.com>
Tested-by: Hudson
Reviewed-by: Keith Mannthey <keith.mannthey@intel.com>
Tested-by: Maloo <whamcloud.maloo@gmail.com>
7 years agoLU-1961 build: fix 'dereference before null check' errors
Sebastien Buisson [Tue, 18 Sep 2012 07:13:41 +0000 (09:13 +0200)]
LU-1961 build: fix 'dereference before null check' errors

Fix 'dereference before null check' defects found by Coverity version
6.0.3:
Dereference before null check (REVERSE_INULL)
Null-checking variable suggests that it may be null, but it has
already been dereferenced on all paths leading to the check.

Signed-off-by: Sebastien Buisson <sebastien.buisson@bull.net>
Change-Id: I0c1be27ea443865fcb7640976839d95d0dc07b9a
Reviewed-on: http://review.whamcloud.com/4023
Tested-by: Hudson
Reviewed-by: Bob Glossman <bob.glossman@intel.com>
Tested-by: Maloo <whamcloud.maloo@gmail.com>
Reviewed-by: Keith Mannthey <keith.mannthey@intel.com>
Reviewed-by: Oleg Drokin <green@whamcloud.com>