4 2.6.16.60-0.42.8 (SLES 10),
5 2.6.27.39-0.3.1 (SLES11),
6 2.6.18-164.11.1.el5 (RHEL 5)
7 2.6.18-164.11.1.0.1.el5 (OEL 5)
8 * Client support for unpatched kernels:
9 (see http://wiki.lustre.org/index.php?title=Patchless_Client)
10 2.6.16 - 2.6.30 vanilla (kernel.org)
11 * Recommended e2fsprogs version: 1.41.10-sun2
12 * The async journal commit feature (bug 19128) and the cancel
13 lock before replay feature (bug 16774) are disabled by default.
15 -------------------------------------------------------------------------------
17 2010-04-30 Oracle, Inc.
19 * Support for kernels:
20 2.6.16.60-0.42.8 (SLES 10),
21 2.6.27.39-0.3.1 (SLES11),
22 2.6.18-164.11.1.el5 (RHEL 5)
23 2.6.18-164.11.1.0.1.el5 (OEL 5)
24 * Client support for unpatched kernels:
25 (see http://wiki.lustre.org/index.php?title=Patchless_Client)
26 2.6.16 - 2.6.30 vanilla (kernel.org)
27 * Recommended e2fsprogs version: 1.41.10-sun2
28 * The async journal commit feature (bug 19128) and the cancel
29 lock before replay feature (bug 16774) are disabled by default.
31 Severity : enhancement
33 Description: lfs quota output is not very convenient for awk/sed-parsing
34 Details : Some positions if lfs quota output table could be empty or
35 non-empty which made it hard to parse it with scripts, now
36 a dash is put instead of space where there is not supposed
41 Description: open-unlinked directories trigger MDS LBUG
42 Details : Fix regression introduced by the patch from bug 19640.
43 ext3_inc_count() can reset nlink to 1 when the directory
44 is indexed and inode->i_nlink == 2. Work around the problem
45 by incrementing nlink by 2 instead of 1.
49 Description: Reconnects are not throttled
50 Details : Don't wake up pinger on reconnect failures and rely on regular
51 pings to trigger the next reconnection. Please note that the
52 pinger already uses a smaller interval if the import is
56 Frequency : only with NFS export
58 Description: Console flooded with error message from ll_inode_from_lock()
60 Details : in mds_open, initialize the child_res_id before enqueuing
61 the OPEN lock for the child inode, then to avoid senting
62 wrong ldlm_res_id to client.
66 Description: allow multiple instances of the same nid in NID hash
67 Details : Case of multiple separate clients from the same NID (as
68 with liblustre) is legitimate and so we should allow
69 multiple instances of the same NID in nid hash.
73 Description: the readahead code can sleep on a semaphore while holding a
75 Details : in ras_update, "lov_get_info" could be called during increasing
76 readahead windows, which tries to get the mutex lock "lov_lock"
77 while holding the spin_lock "ras_lock", then causes system
82 Description: ASSERTION(cli->cl_avail_grant >= 0) failed
83 Details : fix assertion failure in the grant code.
87 Description: Use CNETERR (which is rate limited) in specific places in
88 the portal's LNET driver to avoid flooding the console.
92 Description: include last created object in precreate slow case
96 Description: don't do rep-ack if not created anything
97 Details : mds_open currently always put a lock into a rep-ack regardless
98 if something was created or not. This is pointless and only
99 creates needless contention. In fact the entire idea was to do
100 this for real creates as a recovery protection.
104 Description: Spurious error messages from smp_processor_id() on preemptible
106 Details : Disable a preemption by grabbing the lock in fs_trace_get_tcd()
107 first. The function fs_trace_get_tcd() was moved up.
111 Description: interval_erase() fix
112 Details : interval_erase() calls update_maxhigh() properly when child
117 Description: "lfs df" does not print stats for all mountpoints
118 Details : Print all mounted lustre filesystems with "lfs df"
122 Description: lfs setstripe -p not longer work with a relative pathname
123 Details : Use realpath() to provide absolute pathname.
127 Description: fix for truncated reply buffer
128 Details : reply buffer could be referred by reply_in_callback after released
133 Description: lustre.lov error when backing up symlinks with extended attributes
134 Details : Improved logic in ll_listxattr()
138 Description: properly handle null value for setattr -n lustre.lov
139 Details : Running "setfattr -n trusted.lov ." causes a NULL dereference
140 in ll_setxattr() due to no checking if "value" is NULL.
141 This command now resets to the default striping when executed
146 Description: stack overflow on lock cancellation due to fsync call
147 Details : sync_on_lock_cancel is needed for recovery when async journal
148 is enabled, but we actually just need to make sure that
149 metadata blocks have hit the journal, so doing a fs sync
150 should be enough and should consume less stack (just create an
151 empty handle and commmit it).
155 Description: using current->journal_info to store per-thread data leads
156 to problem under memory pressure
157 Details : disable the per-thread data (current->journal_info) containing
158 the lock info during I/O to work around the issue for short tem
162 Description: control DCACHE_LUSTRE_INVALID flag with MDS_INODELOCK_LOOKUP lock
163 Details : DCACHE_LUSTRE_INVALID is controlled by MDS_INODELOCK_LOOKUP
164 lock which is corresponding to "IT_LOOKUP", do not skip invalidate
169 Description: Cannot send after transport shutdown
170 Details : Clear imp_vbr_failed flag upon eviction
174 Description: soft lock in request set code during recovery
175 Details : during recovery, uses req->rq_set itself to replay the request
176 instead of ptlrpcd_recovery_pc
180 Description: Use CFS_ALLOC_IO instead of _STD in llap_from_page_with_lockh
181 Details : During an ll_readahead under ll_readpage, we have seen the the
182 OBD_SLAB_ALLOC hang under ldlm_pools_shrink when trying to lock
183 a page that is already locked by the readahead code.
188 Description: stop waitting for next replay transno if shutdown
189 Details : if the system is shutting down, wake up service thread blocked
190 to wait for next replay transno during recovery, then all the
191 references held by queued requests can be dropped and device
196 Description: lov_merge_lvb()) ASSERTION(spin_is_locked(&lsm->lsm_lock)) failed
197 Details : Protect lli->lli_smd pointer updates with lli->lli_lock.
201 Description: per-nid stats should not access lustre-hash internal structures
206 Description: mount.lustre fails to pass some options to mount()
211 Description: ext4 extent allocation is slower than in ext3
212 Details : Increase the default value of MB_DEFAULT_ORDER2_REQS to 8,
213 enlarge ext4 preallocation table for 2048 4K blocks extents
218 Description: incorrect triggering of synchronous IO
219 Details : The OSC can mistakenly fall back to synchronous IO when the
220 max_dirty_mb limit is reached and no write requests have yet
221 been issued. This can occur when the dirty pages are spread
222 over many files all of which are below the optimal request size.
226 Description: Optimize quota_ctl operations by sending requests in parallel
227 Details : Send MDS->OST quota_ctl requests in parallel, do not resend.
228 Compiled from two attachments in the ticket.
232 Description: ordering issue between transaction start & i_mutex
233 Details : start the transaction earlier in llog_lvfs_destroy to get
234 transaction start and inode mutex lock nested properly.
238 Description: lru resize SLV can get stuck
239 Details : calculate SLV with a greater precision to not lose small
240 changes due to interger math truncation; round up SLV only
241 if the amount of granted locks less than the limit to not
242 get stuck with this SLV
246 Description: avoid divide-by-zero in lprocfs_rd_import()
250 Description: lfs quota failed when OSTs are down
251 Details : really return approximate block/inode usage when OSTs are down
255 Description: llobdstat fix and enhancement
256 Details : add a counter to set a limit to how many samples will be returned
257 fix a wildcard in the path to limit to obdfilter stats only
261 Description: recovery-small 51 hang - MDT can not stop, Mount still busy with
263 Details : abort lock enqueue processing sooner when we umount is in progress.
265 Severity : enhancement
267 Description: Update RHEL5.4 kernel to 2.6.18-164.11.1.el5 and OEL5.4 kernel to
268 2.6.18-164.11.1.0.1.el5.
270 Severity : enhancement
272 Description: error message improvements
273 Details : Use INFO/WARN instead of WARN/ERROR for the slow messages.
274 Simplify MDT/OST service start message.
275 Suppress "changing the import ..." warning.
278 Bugzilla : 21961/17914
279 Description: ignore trailing -mdc when determining index number
282 Frequency : only with SLES11
284 Description: fix for a race condition in linux quotas implementation
285 Details : dq_flags(struct dquot) access is not properly locked which could
286 lead to certain inconsistencies when accessing it using non-atomic
287 bit operations like __set_bit in do_set_dqblk.
291 Description: return any pending async errors in close(2) using flush callback
293 -------------------------------------------------------------------------------
295 2010-01-29 Sun Microsystems, Inc.
297 * Support for kernels:
298 2.6.16.60-0.42.8 (SLES 10),
299 2.6.27.39-0.3.1 (SLES11),
300 2.6.18-164.11.1.el5 (RHEL 5)
301 2.6.18-164.6.1.0.1.el5 (OEL 5)
302 * Client support for unpatched kernels:
303 (see http://wiki.lustre.org/index.php?title=Patchless_Client)
304 2.6.16 - 2.6.30 vanilla (kernel.org)
305 * Recommended e2fsprogs version: 1.41.6.sun1
306 * The async journal commit feature (bug 19128) and the cancel
307 lock before replay feature (bug 16774) are disabled by default.
309 Severity : enhancement
311 Description: Update RHEL5.4 kernel to 2.6.18-164.11.1.el5 and
312 OEL5.4 kernel to 2.6.18-164.11.1.0.1.el5.
314 Severity : enhancement
315 Bugzilla : 21511/19848
316 Description: Update SLES11 kernel to 2.6.27.39-0.3.1.
318 Severity : enhancement
320 Description: Update supported SLES10 kernel to 2.6.16.60-0.42.8.
322 Severity : enhancement
324 Description: Update kernel to RHEL5.4 2.6.18-164.6.1.el5 and
325 OEL5 2.6.18-164.6.1.0.1.el5(Both in-kernel OFED enabled).
327 Severity : enhancement
329 Description: Build kernels (RHEL5, OEL5 and SLES10/11) using the vendor's own
332 Severity : enhancement
334 Description: Vanilla kernel 2.6.30 patchless client support.
339 Description: bad entry in directory xxx: inode out of bounds
340 Details : fix locking issue in the rename path which could race with any
341 other operations updating the same directory.
343 Severity : enhancement
345 Description: Make watchdog timer messages to be more clear and descriptive.
349 Description: cp -p command does not preserve the dates and timestamp
350 Details : mtime could be spoiled by a write callback
354 Description: Clear imp_force_reconnect correctly in ptlrpc_connect_interpret()
358 Description: Allow non-root access for "lfs check".
359 Details : Added a check in obd_class_ioctl() for OBD_IOC_PING_TARGET.
361 Severity : enhancement
363 Description: quotacheck performance/scaling issues
364 Details : reduce quotacheck time on empty filesystem by skipping uninit
367 Severity : enhancement
369 Description: Enhancement for lfs(1) command to use numeric uid/gid.
371 Severity : enhancement
373 Description: Adjust locks' extents on their first enqueue, so that at the time
374 they get granted, there is no need for another pass through the
375 queues since they are already shaped into the proper forms.
379 Description: Fix mds_shrink_intent_reply()/mds_intent_policy() to pass correct
380 arguments and prevent LBUG() in lustre_shrink_reply_v2().
384 Description: Change tunefs.lustre and mkfs.lustre --mountfsoptions so that
385 exactly the specified mount options are used. Leaving off
386 any "mandatory" mount options is an error. Leaving off any
387 default mount options causes a warning, but is allowed.
388 Change errors=remount-ro from mandatory to default. Sanitize
389 the mount string before storing it. Update man pages accordingly.
393 Description: mds_getattr() should return 0, even if mds_fid2entry() fails with
394 -ENOENT. Also fix in ptlrpc_expire_one_request() to print signed
397 Severity : enhancement
399 Description: Remove set_info(KEY_UNLINKED) from MDS/OSC
401 Severity : enhancement
403 Description: Clients can replay thousands of unused locks during recovery
404 Details : Don't replay unused locks (only read locks for now) during
405 recovery. This feature is disabled by default and can be
406 enabled by running the following command on the clients:
407 lctl get_param ldlm.cancel_unused_locks_before_replay
411 Description: can't stat file in some situation.
412 Details : improve initialize osc date when target is added to mds and
413 ability to resend too big getattr request is client isn't have info
418 Description: Prevent inconsistences between linux and lustre mount structures.
419 Details : Wait indefinitely in server_wait_finished() until mnt_count drops.
420 Make the sleep interruptible.
422 Severity : enhancement
424 Description: Communicate OST degraded/readonly state via statfs to MDS
425 Details : Flags in the statfs returned from OSTs indicate whether the
426 OST is in a degraded RAID state, or if the filesystem has
427 turned read-only after a filesystem error is detected.
432 Description: don't panic if EPROTO was hit when reading symlink
433 Details : correctly handling request reference in error cases.
438 Description: open sometimes returns ENOENT instead of EACCES
439 Details : checking permission should be part of open part of mds_open, not
440 lookup part. so server should be set DISP_OPEN_OPEN disposition
441 before starting permission check.
442 Also not need revalidate dentry if client already have LOOKUP lock.
446 Description: enable client interface failover
447 Frequency : on servers with multiple network interfaces
448 Details : When a child reconnects from another NID, properly update export
449 nid hash position and ldlm reverse import.
451 Severity : enhancement
453 Description: implemented direct I/O with arbitrary (nonaligned) memory
454 addresses and file offsets.
456 Severity : enhancement
458 Description: added more recovery timeout options.
460 Severity : enhancement
462 Description: added llapi_file_open, llapi_file_create, llapi_file_get_stripe
467 Description: Avoid deadlock for local client writes
468 Frequency : only on systems with clients writing to an OST on the same node
469 Details : Use new OBD_BRW_MEMALLOC flag to notify OST about writes in the
470 memory freeing context. This allows OST threads to set the
471 PF_MEMALLOC flag on task structures in order to allocate memory
472 from reserved pools and complete IO.
473 Use GFP_HIGHUSER for OST allocations for non-local client writes,
474 so that the OST threads generate memory pressure and allow
475 inactive pages to be reclaimed.
480 Descriptoin: lock ordering violation between &cli->cl_sem and _lprocfs_lock
481 Details : move ldlm namespace creation in setup phase to avoid grab
482 _lprocfs_lock with cli_sem held.
486 Frequency : only during format of test systems
487 Description: Unable to run several mkfs.lustre on loop devices at the same time
488 Details : mkfs.lustre returns error 256 on the concurrent loop devices
489 formatting. The solution is to proper handle the error.
491 Severity : enhancement
493 Description: implement async create (obd_async_create) method for osc, to avoid
494 too long waiting new ost objects with holding ldlm lock.
498 Frequency : occasionally during network problems
499 Description: client not allowed to reconnect to OST because of active request
500 Details : abort bulk requests received by the OST once the client has timed
501 out since the client will resend the request anyway. The client
502 also now retries to reconnect to the same server if a connect
503 request failed with EBUSY or -EAGAIN.
506 Frequency : rare, if used wide striped file and one ost in down.
508 Descriptoin: don't return error if we created a subset of objects for file.
509 Details : lov_update_create_set() uses set->set_success as index for created
510 objects, so if some requests failed, they will have hole at end of
511 array and we can use qos_shrink_lsm for allocate correct lsm.
515 Description: Slow stale export processing during normal start up
516 Details : The global mgc lock prevents OST setup to be run in parallel.
517 Replace the global lock with a per-config_llog_data semaphore.
521 Description: Out or order replies might be lost on replay
522 Details : In ptlrpc_retain_replayable_request if we cannot find retained
523 request with tid smaller then one currently being added, add it
524 to the start, not end of the list.
528 Description: BUG: soft lockup - CPU#1 stuck for 10s! [ll_mdt_07:4523]
529 Details : add cond_resched() calls to avoid hogging the cpu for too long
530 in the hash code. Make also lustre_hash_for_each_empty() more
533 Severity : enhancement
535 Description: Performance improvements for debug messages with D_RPCTRACE,
536 D_LDLM, D_QUOTA options.
539 Frequency : only with NFS export
541 Description: (lov_merge.c:74:lov_merge_lvb())
542 ASSERTION(spin_is_locked(&lsm->lsm_lock)) failed (SR 71691004)
543 Details : Fix a race in the nfs export code by populating inode
544 info while the new inode is still locked
546 Severity : enhancement
548 Description: add a new file in procfs called force_lbug. Writting to this
549 file triggers a LBUG. Only for test purpose.
553 Description: OOM killer causes node hang
554 Details : really interrupt the sleep in osc_enter_cache on signals
558 Description: LustreError: 9153:0:(quota_context.c:622:dqacq_completion()) LBUG
559 Details : fix race during quota release on the slave.
561 Severity : enhancement
563 Description: smaller hash bucket sizes, cleanups
564 Details : increase hash table sizes and enabled rehashing for pools, uuid,
567 Severity : enhancement
569 Description: Add ldiskfs maxdirsize mount option
570 Details : add max_dir size mount option
574 Description: panic in ll_statahead_thread
575 Details : prevent parent thread to be killed before its child
578 Frequency : only with 16TB device
580 Description: unable to perform "mount -t lustre" of 16TB OST device
581 Details : Mounting 16TB LUNs failed due to three bugs in mkfs.lustre.
585 Description: ASSERTION(atomic_read(&imp->imp_inflight) == 0) failed
586 Details : unregistering should be zero if no RPC inflight.
590 Description: hyperion: Oops during metabench
591 Details : Correct the refcount of lov_request_set
593 Severity : enhancement
595 Description: Add mptlinux and nxge drivers to Lustre builds
597 Severity : enhancement
599 Description: Fix watchdog timer message to be more clear
600 Details : Make watchdog timer messages more clear and descriptive.
604 Description: LNET soft lockups in socknal_cd thread
605 Details : don't hog CPU for active-connecting if another connd is
606 accepting connecting-requst from the same peer
610 Description: recovery-small test_17 hang
611 Details : Land several AT improvements & fixes.
615 Description: MDS panic and hanging client processes
616 Details : Replace exp_ops_stats with exp_nid_stats->nid_stats
620 Description: OSS stuck in recovery.
621 Details : fix race during recovery. class_unlink_export,
622 class_set_export_delayed and target_queue_last_replay_reply
623 may race while increasing/decreasing obd_recoverable_clients
624 and obd_delayed_clients, causing recovery to wait forever.
626 Severity : enhancement
628 Description: add cascading_rw.c to lustre/tests
632 Description: filter_last_id() NULL deref
633 Details : lprocfs_filter_rd_last_id() should check for the fully
634 setup obd device, before proceeding further.
636 Severity : enhancement
638 Description: Loadgen improvements
639 Details : stacksize and locking fixes for loadgen
643 Description: Quiet CERROR("dirty %d > system dirty_max %d\n"
644 Details : The atomic_read() allowing the atomic_inc() are not covered
645 by a lock. Thus they may safely race and trip this CERROR()
646 unless we add in a small fudge factor (+1).
648 Severity : enhancement
650 Description: shrink_slab: nr=-9223362083340912175
651 Details : fix spurious message from shrink_slab reporing negative nr
655 Description: Quiet bogus previously committed transno error
656 Details : suppress the "server went back in time" error message which
657 is always printed even in the common case after a client eviction
659 Severity : enhancement
661 Description: Parallel statfs() calls result in client eviction
662 Details : cache statfs data for 1s.
666 Description: parallel-scale test_compilebench: @@@@@@ FAIL: compilebench
668 Details : fix serveral issues in pinger code causing clients not to ping
669 servers for too long, resulting in evictions.
673 Description: e2fsck should warn when MMP update interval is extended
674 Details : print mmp_check_interval and make it possible to abort mount
675 operation in case it takes too long.
679 Description: mdsrate-create-large.sh, BUG: soft lockup - CPU#0 stuck for 10s!
680 Details : fix bug in the RHEL5's jbd2 callback patch.
684 Description: drop number of active requests when queued for recovery
685 Details : Now that we take a reference on the original request instead of
686 making a copy of it for recovery. We need to drop the number of
687 active requests or the queued requests will prevent all request
688 processing when they exceed (srv->srv_threads_running - 1).
690 Severity : enhancement
692 Description: refuse to invalidate operational quota files when they are in use
693 Details : an attempt to invalidate operational quota files on the quota
694 master is not actually permitted by VFS (returning -EPERM), but we
695 should not depend on that and should return the error earlier.
699 Description: Applications stuck in jbd2_log_wait_commit during exit
700 Details : fix deadlock between kjournald2 trying to acquire the page lock
701 owned by an ost_io thread waiting for journal commit.
703 -------------------------------------------------------------------------------
705 2009-10-16 Sun Microsystems, Inc.
707 * Support for kernels:
708 2.6.16.60-0.42.4 (SLES 10),
709 2.6.27.29-0.1 (SLES11, i686 & x84_64 only),
710 2.6.18-128.7.1.el5 (RHEL 5),
711 2.6.18-128.7.1.el5 (OEL 5),
712 * Client support for unpatched kernels:
713 (see http://wiki.lustre.org/index.php?title=Patchless_Client)
714 2.6.16 - 2.6.27 vanilla (kernel.org)
715 * Recommended e2fsprogs version: 1.41.6.sun1
716 * ext4 support for RHEL5 is experimental and thus should not be
719 Severity : enhancement
721 Description: Add OEL5 support.
723 Severity : enhancement
725 Description: Update kernel to SLES11 2.6.27.29-0.1.
729 Description: File checksum failures with OST read cache on
730 Details : Disable page poisoning when the bulk transfer has to be aborted
731 because the client got evicted.
735 Description: Don't allow make backward step on assiging osc next id.
736 Details : race between allocation next id and ll_sync thread can be cause
737 of set wrong osc next id and can be kill valid ost objects.
739 Severity : enhancement
741 Description: Update kernel to RHEL5 2.6.18-128.7.1.el5.
743 Severity : enhancement
745 Description: Update kernel to SLES10 SP2 2.6.16.60-0.42.4.
749 Description: Changes in raid5-large-io-rhel5.patch to calculate sectors properly
753 Description: Increase the default BLK_DEF_MAX_SECTORS value for RHEL5 and SLES11
757 Description: Error handling in osc_statfs_interpret() has been improved.
758 Details : Check in osc_statfs_interpret() for EBADR.
762 Description: Do not send statfs() requests to OSTs disabled by administrator.
763 Details : Check in lov_prep_statfs_set() for non-NULL ltd_exp.
767 Description: Do not update ctime for the deleted inode.
768 Details : Check in mds_reint_unlink() before calling fsfilt_setattr().
772 Description: Increase of the size of the LDLM resource hash.
773 Details : Bump up RES_HASH_BITS=12.
777 Description: correctly send lsm on open replay
778 Details : MDS is trust to LSM size on replay open, but client can set wrong size
783 Description: Deadlock between filter_destroy() and filter_commitrw_write().
784 Details : filter_destroy() does not hold the DLM lock over the whole
785 operation. If the DLM lock is dropped, filter_commitrw() can go
786 through, causing the deadlock between page lock and i_mutex.
787 The i_alloc_sem should also be hold in filter_destroy() while
792 Description: truncate starts GFP_FS allocation under transaction causing deadlock
793 Details : ldiskfs_truncate calls grab_cache_page which may start page
794 allocation under an open transaction. This may lead to
795 calling prune_icache with consequent lustre reentrance.
798 Frequency : only when down/upgrading the MDS to 1.6/1.8 while 1.8 clients are
799 still up and when the OST pool feature is used
801 Description: interop testing got LBUG when run dd with OST pool
802 :LustreError: 30032:0:(llite_lib.c:1913:ll_replace_lsm()) LBUG
803 Details : down/upgrading the MDS to a version that doesn't/does support OST
804 pool can cause clients to crash because the lsm has changed
809 Description: missing tree_status on 1.8.1 RPM build
810 Details : make rpms failed due because the tree_status file is missing.
814 Description: continuing LustreError "mds adjust qunit failed!"
815 Details : don't print message on the console when ->adjust_qunit fails.
819 Description: don't increase ldlm timeout if previous client was evicted
820 Details : if a client doesn't respond to a blocking callback within the
821 adaptive ldlm enqueue timeout, don't adjust the adaptive estimate
822 when the lock is next granted.
826 Description: ost is being unmounted w/o all writes to last_rcvd landing on disk.
827 affects recovery negatively.
828 Details : make sure all exports have been properly destroyed by the zombie
829 thread processed before stopping the target.
833 Description: Performance degradation with O_DIRECT between 1.6 & 1.8.1 b190
834 Details : disable write barrier for ext4/SLES11.
838 Description: Kernel panic - not syncing: Out of memory and no killable
839 processes... on OSS when iozone
840 Details : fix memory leak in the journal checksum patch.
844 Description: group quota "too many blocks" OSS crashes
845 Details : we should keep the same uid/gid for lquota_chkquota() and
846 lquota_pending_commit()
850 Description: LustreError: 9153:0:(quota_context.c:622:dqacq_completion()) LBUG
851 Details : don't LBUG on release quota error. Just a workaround until the
852 problem is understood.
854 ------------------------------------------------------------------------------
855 2009-07-31 Sun Microsystems, Inc.
857 * Support for kernels:
858 2.6.16.60-0.39.3 (SLES 10),
859 2.6.27.23-0.1 (SLES11, i686 & x84_64 only),
860 2.6.18-128.1.14.el5 (RHEL 5),
861 * Client support for unpatched kernels:
862 (see http://wiki.lustre.org/index.php?title=Patchless_Client)
863 2.6.16 - 2.6.27 vanilla (kernel.org)
864 * Recommended e2fsprogs version: 1.41.6.sun1
865 * File join has been disabled in this release, refer to Bugzilla 16929.
866 * NFS export disabled when stack size < 8192. Since the NFSv4 export
867 of Lustre filesystem with 4K stack may cause a stack overflow. For
868 more information, please refer to bugzilla 17630.
869 * ext4 support for RHEL5 is experimental and thus should not be
872 Severity : enhancement
874 Description: Update kernel to SLES10 SP2 2.6.16.60-0.39.3.
877 Frequency : with 1.8 server and 1.6 clients
879 Description: correctly shrink reply for avoid send too big message to client.
880 Details : 1.8 mds is allocate to big buffer to LOV EA data and this produce
881 some problems with sending this reply to 1.6 client.
885 Description: Repeated atomic allocation failures.
886 Details : Use GFP_HIGHUSER | __GFP_NOMEMALLOC flags for memory allocations
887 to generate memory pressure and allow reclaiming of inactive pages.
888 At the same time, do not allow to exhaust emergency pools.
889 For local clients the use of GFP_NOFS will be introduced in 1.8.2
891 Severity : enhancement
892 Bugzilla : 19846, 18289
893 Description: Update kernel to RHEL5 2.6.18-128.1.14.el5.
895 Severity : enhancement
896 Bugzilla : 19625, 16893, 18668, 19848
897 Description: Add support for SLES11 2.6.27.23-0.1.
899 Severity : enhancement
901 Description: Update client support to vanila kernels up to 2.6.27.
903 Severity : enhancement
905 Description: Update kernel to SLES10 SP2 2.6.16.60-0.37.
907 Severity : enhancement
909 Description: Compile with -Werror by default for i686 and x86_64.
913 Description: resolve race between obd_disconnect and class_disconnect_exports
914 Details : if obd_disconnect will be called to already disconnected export he
915 forget release one reference and osc module can't unloaded.
917 Severity : enhancement
919 Description: move AT tunable parameters for more consistent usage
920 Details : add AT tunables under /proc/sys/lustre, add to conf_param parsing
924 Descriptoin: correctly skip time estimate if in recovery
925 Details : rq_send_state insn't bitmask so using bitwise ops is forbid.
929 Descriptoin: OSS DeadLock
930 Details : Use trylock to prevent deadlock when shrink icache.
932 Severity : enhancement
934 Description: Allow tuning service thread via /proc
935 Details : For each service a new
936 /proc/fs/lustre/{service}/*/thread_{min,max,started} entry is
937 created that can be used to set min/max thread counts, and get the
938 current number of running threads.
940 Severity : enhancement
942 Description: Add state history info file, enhance import info file
943 Details : Track import connection state changes in a new osc/mdc proc file;
944 add overview-type data to the osc/mdc import proc file.
948 Description: Reduce small size read RPC
949 Details : Set read-ahead limite for every file and only do read-ahead when
950 available read-ahead pages are bigger than 1M to avoid small size
955 Description: free_entry erroneously used groups_free instead of put_group_info
957 Severity : enhancement
959 Description: Make read-ahead stripe size aligned.
961 Severity : enhancement
963 Description: MDS create should not wait for statfs RPC while holding DLM lock.
966 Frequency : rare, connect and disconnect target at same time
968 Descriptoin: ASSERTION(atomic_read(&imp->imp_inflight) == 0
969 Details : don't call obd_disconnect under lov_lock. this long time
970 operation and can block ptlrpcd which answer to connect request.
973 Frequency : start MDS on uncleanly shutdowned MDS device
975 Descriptoin: ll_sync thread stay in waiting mds<>ost recovery finished
976 Details : stay in waiting mds<>ost recovery finished produce random bugs
977 due race between two ll_sync thread for one lov target. send
978 ACTIVATE event only if connect realy finished and import have
982 Frequency : start MDS on uncleanly shutdowned MDS device
984 Descriptoin: aborting recovery hang on MDS
985 Details : don't throttle destroy RPCs for the MDT.
989 Description: Slow reads beyond 8Tb offsets.
990 Details : Page index integer overflow in ll_read_ahead_page
994 Description: Soft lockup on OSS after MDS failover
995 Details : MSG_CONNECT_INITIAL is not set on the initial MDS->OST connect.
996 As a conseqence, the patch from bug 18224 is not operational
997 and the MDS export cannot be reused on the OSTs until it gets
1001 Frequency : rare, only if using MMP with Linux RAID
1003 Description: MMP doesn't work with Linux RAID
1004 Details : While using HA for Lustre servers with Linux RAID, it is possible
1005 that MMP will not detect multiple mounts. To make this work we
1006 need to unplug the device queue in RAID when the MMP block is being
1007 written. Also while reading the MMP block, we should read it from
1008 disk and not the cached one.
1011 Frequency : rare, during recovery
1013 Description: Assertion failure in ldlm_lock_put
1014 Details : Do not put cancelled locks into replay list, hold references on
1015 locks in replay list
1019 Description: 1.6.5 mdsrate performance is slower than 1.4.11/12 (MDS is not cpu bound!)
1020 Details : create_count always drops to the min value (=32) because grow_count
1021 is being changed before the precreate RPC completes.
1024 Frequency : Only in RHEL5 when mounting multiple ext3 filesystems
1027 Description: "kmem_cache_create: duplicate cache jbd_4k" error message
1028 Details : add proper locking for creation of jbd_4k slab cache
1032 Description: MMP check in ext3_remount() fails without displaying any error
1033 Details : When multiple mount protection fails during remount, proper error
1038 Description: Rare Client crash on resend if the file was deleted.
1039 Details : When file is opened, but open reply is lost and file is
1040 subsequently deleted before resend, resend processing logic
1041 breaks trying to open the file again, should not try to open.
1045 Description: add check for >8TB ldiskfs filesystems
1046 Details : ext3-based ldiskfs does not support greater than 8TB LUNs.
1047 Don't allow >8TB ldiskfs filesystems to be mounted without
1048 force_over_8tb mount option
1052 Description: Client locked up when running multiple instances of an app. on
1053 multiple mount points
1054 Details : ll_shrink_cache() can sleep while holding the ll_sb_lock.
1055 Convert ll_sb_lock to a read/write semaphore to fix the problem.
1059 Description: Cannot acces an NFS-mounted Lustre filesystem
1060 Details : An NFS client cannot access the Lustre filesystem NFS-mounted
1061 from a Lustre-client exporting the Lustre filesystem via NFS.
1065 Description: panic in ll_statahead_thread
1066 Details : grab dentry reference in parent process.
1069 -------------------------------------------------------------------------------
1071 tbd Sun Microsystems, Inc.
1073 * Support for kernels:
1074 2.6.16.60-0.31 (SLES 10),
1075 2.6.18-128.1.6.el5 (RHEL 5),
1076 2.6.22.14 vanilla (kernel.org)
1077 * Client support for unpatched kernels:
1078 (see http://wiki.lustre.org/index.php?title=Patchless_Client)
1079 2.6.16 - 2.6.22 vanilla (kernel.org)
1080 * Recommended e2fsprogs version: 1.40.11-sun1
1081 * File join has been disabled, refer to Bugzilla 16929.
1082 * A new Lustre ADIO driver is available for MPICH2-1.0.7.
1083 * NFS export disabled when stack size < 8192. Since the NFSv4 export of
1084 Lustre filesystem with 4K stack may cause a stack overflow. For more
1085 information, please refer to bugzilla 17630.
1087 Severity : enhancement
1089 Description: Update to RHEL5.3 kernel-2.6.18-128.1.6.el5.
1091 Severity : enhancement
1093 Description: Update OFED release to 1.4.1 RC4
1095 Severity : major, only with big OST
1097 Description: Very poor metadata performance on Infiniband lustre configuration
1098 Details : OST object precreation becomes very slow on big OSTs. This is due
1099 to the ialloc patch spending too much time scanning groups.
1102 Frequency : during recovery
1104 Description: don't mix llog inodes with normal.
1105 Details : allocate inodes for log in last inode group
1110 Description: fix lqs' reference which won't be put in some situations
1111 Details : This patch fixes:
1112 1. In quota_check_common(), this function will check quota
1113 for user and group, but only send one return via "pending".
1114 In most cases, the pendings should be same. But that is not
1116 2. If quotaoff runs between lquota_chkquota() and
1117 lquota_pending_commit(), the same thing will happen too.
1118 That is why it comes:
1119 - if (!ll_sb_any_quota_active(qctxt->lqc_sb))
1122 -------------------------------------------------------------------------------
1124 2008-12-31 Sun Microsystems, Inc.
1126 * Support for kernels:
1127 2.6.5-7.314 (SLES 9),
1128 2.6.9-67.0.22.EL (RHEL 4),
1129 2.6.16.60-0.31 (SLES 10),
1130 2.6.18-92.1.17.el5 (RHEL 5),
1131 2.6.22.14 vanilla (kernel.org)
1132 * Client support for unpatched kernels:
1133 (see http://wiki.lustre.org/index.php?title=Patchless_Client)
1134 2.6.16 - 2.6.22 vanilla (kernel.org)
1135 * Client support for unpatched kernels:
1136 we do not recommend using patchless RHEL4 clients with kernels
1137 prior to 2.6.9-55EL (RHEL4U5).
1138 * Recommended e2fsprogs version: 1.40.11-sun1
1139 * Note that reiserfs quotas are disabled on SLES 10 in this kernel.
1140 * RHEL 4 and RHEL 5/SLES 10 clients behaves differently on 'cd' to a
1141 removed cwd "./" (refer to Bugzilla 14399).
1142 * A new quota file format has been introduced in 1.6.5.
1143 The format conversion from prior releases is handled transparently,
1144 but releases older than 1.4.12/1.6.5 don't understand this new
1145 format. The automatic format conversion can be avoided by running
1146 the following command on the MDS:
1147 'tunefs.lustre --param="mdt.quota_type=ug1" $MDTDEV'.
1148 For more information, please refer to bugzilla 13904.
1149 * A new quota file format was introduced in 1.6.6/1.8.0 (kernels 2.6.16+).
1150 The format conversion from prior releases is handled transparently,
1151 but releases older than 1.6.6/1.8.0 don't understand this new
1152 format. The automatic format conversion can be avoided by running
1153 the following commands on the MDS and OSS servers (for
1154 pre 1.4.12-1.6.5 quota files):
1155 'tunefs.lustre --param="mdt.quota_type=ug1" $MDTDEV',
1156 'tunefs.lustre --param="ost.quota_type=ug1" $MDTDEV'
1157 or (for 1.4.12/1.6.5 quota files)
1158 'tunefs.lustre --param="mdt.quota_type=ug2" $MDTDEV',
1159 'tunefs.lustre --param="ost.quota_type=ug2" $MDTDEV'
1160 For more information, please refer to bugzilla 13904.
1161 * Output of lfs quota has been made less detailed by default,
1162 old (verbose) output can be obtained by using -v option.
1163 * File join has been disabled in this release, refer to Bugzilla 16929.
1164 * A new Lustre ADIO driver is available for MPICH2-1.0.7.
1165 * NFS export disabled when stack size < 8192. Since the NFSv4 export of
1166 Lustre filesystem with 4K stack may cause a stack overflow. For more
1167 information, please refer to bugzilla 17630.
1169 Severity : enhancement
1171 Description: Caching OSS
1172 Details : introduce data caching on the OSS. The OSS now relies on the linux
1173 kernel page cache to keep recently accessed data in memory.
1174 It is worth noting that all write requests are still flushed
1175 synchronously as in lustre 1.6.
1177 Severity : enhancement
1179 Description: Version based recovery
1180 Details : introduce finer grained recovery able to detect transaction
1181 dependencies and can deal with transaction gaps caused by clients
1182 failing at the same time as the server.
1184 Severity : enhancement
1186 Description: Enable adaptive timeouts by default
1187 Details : The Lustre timeout value in /proc/sys/lustre/timeout is now
1188 managed dynamically based on server load and should not need
1189 to be tuned manually based on cluster size. This allows Lustre
1190 to work under a wider variety of system sizes and loads, without
1191 unnecessarily causing lengthy recovery times.
1193 Severity : enhancement
1195 Description: Add OST Pools support
1196 Details : File striping can now be set to use an arbitrary pool of OSTs
1198 Severity : enhancement
1200 Description: add lazystatfs mount option to allow statfs(2) to skip down OSTs
1201 Details : allow skip disconnected ost for send statfs request and hide error
1205 Frequency : rare, on llog test 6
1207 Descriptoin: don't allow connect to already connected import
1208 Details : allowing connect to already connected import is hide connecting problem.
1211 Frequency : rare, on failed llog setup
1213 Descriptoin: don't leak obd reference on failed llog setup
1214 Details : for failed llog setup - mgc forget call class_destroy_import for
1215 client import, move destroy import to more generic place.
1220 Descriptoin: allow kill process which wait statahead result
1221 Details : for some reasons 'ls' can stick in waiting result from statahead,
1222 in this case need way for kill this process.
1225 Frequency : rare, at shutdown
1227 Descriptoin: panic at umount
1228 Details : llap_shrinker can be raced with killing super block from list and
1229 this produce panic with access to already freeded pointer
1234 Descriptoin: don't lose wakeup for imp_recovery_waitq
1235 Details : recover_import_no_retry or invalidate_import and import_close can
1236 both sleep on imp_recovery_waitq, but we was send only one wakeup
1242 Descriptoin: panic in mds_open
1243 Details : don't confuse mds_finish_transno() with PTR_ERR(-ENOENT)
1248 Descriptoin: stuck in cache_remove_extent() or panic with accessing to already
1250 Details : release lock refernce only after add page to pages list.
1253 Frequency : always with long access acl
1255 Descriptoin: mds can't pack reply with long acl.
1256 Details : mds don't control size of acl but they limited by reint/getattr
1261 Frequency : on remount
1263 Description: external journal device not working after the remount
1264 Details : clear dev_rdonly flag for external journal devices in
1270 Description: shutdown vs evict race
1271 Details : client_disconnect_export vs connect request race.
1272 if client will evicted at this time - we start invalidate
1273 thread without referece to import and import can be freed
1279 Description: shrink LOV EAs before replying
1280 Details : correctly adjust LOV EA buffer for reply.
1285 Description: don't skip ost target if they assigned to file
1286 Details : Drop slow OSCs if we can, but not for requested start idx.
1287 This means "if OSC is slow and it is not the requested
1288 start OST, then it can be skipped, otherwise skip it only
1289 if it is inactive/recovering/out-of-space.
1291 Severity : enhancement
1293 Description: Update to RHEL5 kernel-2.6.18-92.1.17.el5.
1296 Frequency : rare, need acl's on inode.
1298 Description: client can't handle ost additional correctly
1299 Details : if ost was added after client connected to mds client can have
1300 hit lnet_try_match_md ... to big messages to wide striped files.
1301 in this case need teach client to handle config events about add
1302 lov target and update client max ea size at that event.
1304 Severity : enhancement
1306 Description: Update to sles9 kernel-2.6.5-7.314.
1308 Severity : enhancement
1310 Description: Update to SLES10 SP2 kernel-2.6.16.60-0.31.
1313 Frequency : Create a symlink file with a very long name
1315 Description: ldlm_cancel_pack()) ASSERTION(max >= dlm->lock_count + count)
1316 Details : If there is no extra space in the request for early cancels,
1317 ldlm_req_handles_avail() returns 0 instead of a negative value.
1322 Description: mds is deadlocked
1323 Details : in rare cases, inode in catalog can have i_no less than have parent
1324 i_no, this produce wrong order for locking during open, and
1325 parallel unlink can be lock open. this need teach mds_open to grab
1326 locks in resource id order, not at parent -> child order.
1328 Severity : enhancement
1330 Description: Add /proc entry for import status
1331 Details : The mdc, osc, and mgc import directories now have
1332 an import directory that contains useful import data for debugging
1333 connection problems.
1335 Severity : enhancement
1337 Description: Re-disable certain /proc logging
1338 Details : Enable and disable client's offset_stats, extents_stats and
1339 extents_stats_per_process stats logging on the fly.
1342 Frequency : Only on FC kernels 2.6.22+
1344 Description: oops in statahead
1345 Details : Do not drop reference count for the dentry from VFS when lookup,
1346 VFS will do that by itself.
1348 Severity : enhancement
1350 Description: Generic /proc file permissions
1351 Details : Set /Proc file permissions in a more generic way to enable non-
1352 root users operate on some /proc files.
1356 Description: Hitting mdc_commit_close() ASSERTION
1357 Details : Properly handle request reference release in
1358 ll_release_openhandle().
1362 Frequency : only patchless client
1363 Description: add workaround for race between add/remove dentry from hash
1365 Severity : enhancement
1367 Description: Allow OST glimpses to return PW locks
1371 Description: LBUG when llog conf file is full
1372 Details : When llog bitmap is full, ENOSPC should be returned for plain
1377 Description: Prevent import from entering FULL state when server in recovery
1381 Description: service mount cannot take device name with ":"
1382 Details : Only when device name contains ":/" will mount treat it as
1388 Description: replace ptlrpcd with the statahead thread to interpret the async
1389 statahead RPC callback
1393 Frequency : on recovery
1394 Description: I/O failures after umount during fail back
1395 Details : if client reconnected to restarted server we need join to recovery
1396 instead of find server handler is changed and process self eviction
1397 with cancel all locks.
1399 Severity : enhancement
1401 Description: Update to RHEL5 kernel-2.6.18-92.1.10.el5.
1405 Description: Kernel BUG tries to release flock
1406 Details : Lustre does not destroy flock lock before last reference goes
1407 away. So always drop flock locks when client is evicted and
1408 perform unlock regardless of successfulness of speaking to MDS.
1410 Severity : enhancement
1412 Description: Update to SLES10 SP2 kernel-2.6.16.60-0.27.
1414 Severity : enhancement
1416 Description: Upcall on Lustre log has been dumped
1417 Details : Allow for a user mode script to be called once a Lustre log has
1418 been dumped. It passes the filename of the dumped log to the
1419 script, the location of the script can be specified via
1420 /proc/sys/lnet/debug_log_upcall.
1425 Description: avoid messages about idr_remove called for id that is not allocated
1426 Details : Move assigment s_dev for clustered nfs to end of initialization,
1427 for avoid problem with error handling.
1432 Description: avoid Already found the key in hash [CONN_UNUSED_HASH] messages
1433 Details : When connection is reused this not moved from CONN_UNUSED_HASH into
1434 CONN_USED_HASH and this prodice warning when put connection again
1440 Description: avoid ASSERTION(client_stat->nid_exp_ref_count == 0) failed
1441 Details : release reference to stats when client disconnected, not
1442 when export destroyed for avoid races when client destroyed
1443 after main ost export.
1447 Description: more cleanup in mds_lov
1448 Details : add workaround for get valid ost count for avoid warnings about
1449 drop too big messages, not init llog cat under semphore which
1450 can be blocked on reconnect and break normal replay, fix access
1453 Severity : enhancement
1455 Description: Export bytes_read/bytes_write count on OSC/OST.
1459 Description: Early reply size mismatch, MGC loses connection
1460 Details : Apply the MGS_CONNECT_SUPPORTED mask at reconnect time so
1461 the connect flags are properly negotiated.
1465 Description: Properly propagate oinfo flags from lov to osc for statfs
1466 Details : restore missing copy oi_flags to lov requests.
1470 Description: exports in /proc are broken
1471 Details : recreate /proc entries for clients when they reconnect.
1473 Severity : enhancement
1475 Description: Add man pages for llobdstat(8), llstat(8), plot-llstat(8),
1476 : l_getgroups(8), lst(8), routerstat(8)
1477 Details : included man pages for llobdstat(8), llstat(8),
1478 : plot-llstat(8), l_getgroups(8), lst(8), routerstat(8)
1480 Severity : enhancement
1482 Description: Implement lustre ll_show_options method.
1484 Severity : enhancement
1486 Description: Update to SLES9 kernel-2.6.5-7.312.
1488 Severity : enhancement
1490 Description: Update to RHEL4 kernel-2.6.9-67.0.22.EL.
1494 Description: exports in /proc are broken
1495 Details : recreate /proc entries for clients when they reconnect.
1499 Description: don't fail open with -ERANGE
1500 Details : if client connected until mds will be know about real ost count
1501 get LOV EA can be fail because mds not allocate enougth buffer
1506 Description: Resolve device initialization race
1507 Details : Prevent proc handler from accessing devices added to the
1508 obd_devs array but yet be intialized.
1510 Severity : enhancement
1512 Description: configure's --enable-quota should check the
1513 : kernel .config for CONFIG_QUOTA
1514 Details : configure is terminated if --enable-quota is passed but
1515 : no quota support is in kernel
1517 Severity : enhancement
1519 Description: Update to SLES10 SP2 kernel-2.6.16.60-0.23.
1521 Severity : enhancement
1523 Description: Update to RHEL5 kernel-2.6.18-92.1.6.el5.
1527 Frequency : rare, on PPC clients
1528 Description: don't swab ost objects in response about directory, because
1530 Details : bug similar bug 14856, but in different function.
1532 Severity : enhancement
1534 Description: lfs quota tool enhancement
1535 Details : added units specifiers support for setquota, default to
1536 current uid/gid for quota report, short quota stats by
1537 default, nonpositional parameters for setquota, added
1538 llapi_quotactl manual page.
1540 Severity : enhancement
1542 Description: *optional* service tags registration
1543 Details : if the "service tags" package is installed on a Lustre node
1544 When the filesystem is mounted, a local-node service tag will
1545 be created. See http://inventory.sun.com/ for more information
1546 about the Service Tags asset management system.
1550 Description: Client runs out of low memory
1551 Details : Consider only lowmem when counting initial number of llap pages
1554 Frequency : occasional
1556 Description: add refcount for osc callbacks, so avoid panic on shutdown
1558 Severity : enhancement
1560 Description: Update to RHEL4 kernel-2.6.9-67.0.20.
1563 Frequency : testing only
1565 Description: sanity test 65a fails if stripecount of -1 is set
1566 Details : handle -1 striping on filesystem in ll_dirstripe_verify
1569 Frequency : only in unusual configurations
1571 Description: Kernel panic with find ost index.
1572 Details : lov_obd have panic if some OST's have sparse indexes.
1574 Severity : enhancement
1576 Description: Update to RHEL5 kernel-2.6.18-53.1.21.el5.
1579 Frequency : rarely, if filesystem is mounted with -o flock
1581 Description: do not process already freed flock
1582 Details : flock can possibly be freed by another thread before it reaches
1583 to ldlm_flock_completion_ast.
1586 Frequency : rarely, if filesystem is mounted with -o flock
1588 Description: LBUG during stress test
1589 Details : Need properly lock accesses the flock deadlock detection list.
1592 Frequency : rarely, if binaries are being run from Lustre
1594 Description: oops in page fault handler
1595 Details : kernel page fault handler can return two special 'pages' in
1596 error case, don't try dereference NOPAGE_SIGBUS and NOPAGE_OMM.
1599 Frequency : rarely, during shutdown
1601 Description: timeout with invalidate import.
1602 Details : ptlrpcd_check call obd_zombie_impexp_cull and wait request which
1603 should be handled by ptlrpcd. This produce long age waiting and
1604 -ETIMEOUT ptlrpc_invalidate_import and as result LASSERT.
1610 Description: ASSERTION(CheckWriteback(page,cmd)) failed
1611 Details : badly clear PG_Writeback bit in ll_ap_completion can produce false
1615 Frequency : only with broken builds/installations
1617 Description: no LBUG if lquota.ko and fsfilt_ldiskfs.ko are different versions
1618 Details : just return an error to a user, put a console error message
1620 Severity : enhancement
1622 Description: Update to RHEL5 kernel-2.6.18-53.1.19.el5.
1624 Severity : enhancement
1626 Description: Update to RHEL4 kernel-2.6.9-67.0.15.
1628 Severity : enhancement
1630 Description: enable MGS and MDT services start separately
1631 Details : add a 'nomgs' option in mount.lustre to enable start a MDT with
1632 a co-located MGS without starting the MGS, which is a complement
1633 to 'nosvc' mount option.
1636 Frequency : always, on big-endian systems
1638 Description: cleanup in ptlrpc code, related to PPC platform
1639 Details : store magic in native order avoid panic's in recovery on PPC
1640 node and forbid from this error in future. Also fix posibily
1641 of twice swab data. Fix get lov striping to userpace.
1644 Frequency : rarely, if replay get lost on server
1646 Description: server incorrectly drop resent replays lead to recovery failure.
1647 Details : do not drop replay according to msg flags, instead we check the
1648 per-export recovery request queue for duplication of transno.
1651 Frequency : after recovery
1653 Description: precreate to many object's after del orphan.
1654 Details : del orphan st in oscc last_id == next_id and this triger growing
1655 count of precreated objects. Set flag LOW to skip increase count
1656 of precreated objects.
1659 Frequency : rare, on clear nid stats
1661 Description: ASSERTION(client_stat->nid_exp_ref_count == 0)
1662 Details : when clean nid stats sometimes try destroy live entry,
1663 and this produce panic in free.
1666 Frequency : occasionally since 1.6.4
1668 Description: Stack overflow during MDS log replay
1669 Details : ease stack pressure by using a thread dealing llog_process.
1672 Frequency : very rare
1674 Description: MDT cannot be unmounted, reporting "Mount still busy"
1675 Details : Mountpoint references were being leaked during open reply
1676 reconstruction after an MDS restart. Drop mountpoint reference
1677 in reconstruct_open() and free dentry reference also.
1682 Description: wait until IO finished before start new when do lock cancel.
1683 Details : VM protocol want old IO finished before start new, in this case
1684 need wait until PG_writeback is cleared until check dirty flag
1685 and call writepages in lock cancel callback.
1690 Description: mds_mfd_close() ASSERTION(rc == 0)
1691 Details : In mds_mfd_close(), we need protect inode's writecount change
1692 within its orphan write semaphore to prevent possible races.
1695 Frequency : rare, on shutdown ost
1697 Description: don't hit live lock with umount ost.
1698 Details : shrink_dcache_parent can be in long loop with destroy dentries,
1699 use shrink_dcache_sb instead.
1702 Frequency : only when echo_client is used
1704 Description: don't panic with use echo_client
1705 Details : echo client pass NULL as client nid pointer and this produce NULL
1706 pointer dereference.
1709 Frequency : Always on 32-bit PowerPC systems
1711 Description: fix build on PPC32
1712 Details : compile code with -m64 flag produce wrong object file for PPC32.
1717 Description: MDS LBUG: ASSERTION(!IS_ERR(dchild))
1718 Details : In reconstruct_* functions, LASSERTs on both the data supplied
1719 by a client, and the data on disk are dangerous and incorrect.
1720 Change them with client eviction.
1722 Severity : enhancement
1724 Description: skiplist implementation simplification
1725 Details : skiplists are used to group compatible locks on granted list
1726 that was implemented as tracking first and last lock of each lock
1727 group the patch changes that to using doubly linked lists
1731 Description: delete compatibility for 32bit qdata
1732 Details : as planned, when lustre is beyond b1_8, lquota won't support 32bit
1733 qunit. That means servers of b1_4 and servers of b1_8 can't be
1734 used together if users want to use quota.
1737 Frequency : only with administrator action
1739 Description: mount failure if config log has invalid conf_param setting
1740 Details : If administrator specified an incorrect configuration parameter
1741 with "lctl conf_param" this would cause an error during future
1742 client mounts. Instead, ignore the bad configuration parameter.
1745 Frequency : blocks per group < blocksize*8 and uninit_groups is enabled
1747 Description: ldiskfs error: XXX blocks in bitmap, YYY in gd
1748 Details : If blocks per group is less than blocksize*8, set rest of the
1752 Frequency : Application do stride read on lustre
1754 Description: The read performance will drop a lot if the application does
1756 Details : Because the stride_start_offset are missing in stride read-ahead,
1757 it will cause clients read a lot of unused pages in read-ahead,
1758 then the read-performance drops.
1762 Description: more ldlm soft lockups
1763 Details : In ldlm_resource_add_lock(), call to ldlm_resource_dump()
1764 starve other threads from the resource lock for a long time in
1765 case of long waiting queue, so change the debug level from
1766 D_OTHER to the less frequently used D_INFO.
1768 Severity : enhancement
1770 Description: add -gid, -group, -uid, -user options to lfs find
1772 Severity : enhancement
1774 Description: ll_recover_lost_found_objs - recover objects in lost+found
1775 Details : OST corruption and subsequent e2fsck can leave objects in the
1776 lost+found directory. Using the "ll_recover_lost_found_objs"
1777 tool, these objects can be retrieved and data can be salvaged
1778 by using the object ID saved in the fid EA on each object.
1783 Description: this bug _only_ happens when inode quota limitation is very low
1784 (less than 12), so that inode quota unit is 1 at initialization.
1785 Details : if remaining quota equates 1, it is a sign to demonstate that quota
1786 is effective now. So least quota qunit should be 2.
1790 Description: Hung threads in invalidate_inode_pages2_range
1791 Details : The direct IO path doesn't call check_rpcs to submit a new RPC once
1792 one is completed. As a result, some RPCs are stuck in the queue
1797 Description: Procfs and llog threads access destoryed import sometimes.
1798 Details : Sync the import destoryed process with procfs and llog threads by
1799 the import refcount and semaphore.
1803 Description: mds fails to respond, threads stuck in ldlm_completion_ast
1804 Details : Sort source/child resource pair after updating child resource.
1809 Description: kernel BUG at ldiskfs2_ext_new_extent_cb
1810 Details : If insertion of an extent fails, then discard the inode
1811 preallocation and free data blocks else it can lead to duplicate
1816 Description: don't always update ctime in ext3_xattr_set_handle()
1817 Details : Current xattr code updates inode ctime in ext3_xattr_set_handle()
1818 In some cases the ctime should not be updated, for example for
1819 2.0->1.8 compatibility it is necessary to delete an xattr and it
1820 should not update the ctime.
1824 Description: add quota statistics
1825 Details : 1. sort out quota proc entries and proc code.
1826 2. add quota statistics
1831 Description: quotas are not honored with O_DIRECT
1832 Details : all writes with the flag O_DIRECT will use grants which leads to
1833 this problem. Now using OBD_BRW_SYNC to guard this.
1837 Bugzilla : 15713/16362
1838 Description: Assertion in iopen_connect_dentry in 1.6.3
1839 Details : looking up an inode via iopen with the wrong generation number can
1840 populate the dcache with a disconneced dentry while the inode
1841 number is in the process of being reallocated. This causes an
1842 assertion failure in iopen since the inode's dentry list contains
1843 both a connected and disconnected dentry.
1847 Description: assertion failure in ldlm_handle2lock()
1848 Details : fix a race between class_handle_unhash() and class_handle2object()
1849 introduced in lustre 1.6.5 by bug 13622.
1851 Severity : enhancement
1853 Description: superblock lock contention with many SMP cores on one client
1854 Details : several client filesystem locks were highly contended on SMP
1855 NUMA systems with 8 or more cores. Per-CPU datastructures
1856 and more efficient locking implemented to reduce contention.
1861 Description: Kernel BUG: sd_iostats_bump: unexpected disk index
1862 Details : remove the limit of 256 scsi disks in the sd_iostat patch
1867 Description: oops in sd_iostats_seq_show()
1868 Details : unloading/reloading the scsi low level driver triggers a kernel
1869 bug when trying to access the sd iostat file.
1874 Description: Kernel panics during QLogic driver reload
1875 Details : REQ_BLOCK_PC requests are not handled properly in the sd iostat
1876 patch, causing memory corruption.
1881 Description: journal_dev option does not work in b1_6
1882 Details : pass mount option during pre-mount.
1884 Severity : enhancement
1886 Description: Add a FIEMAP(FIle Extent MAP) ioctl for ldiskfs
1887 Details : FIEMAP ioctl will allow an application to efficiently fetch the
1888 extent information of a file. It can be used to map logical blocks
1889 in a file to physical blocks in the block device.
1892 Frequency : only with adaptive timeout enabled
1894 Description: DEBUG_REQ() bad paging request
1895 Details : ptlrpc_at_recv_early_reply() should not modify req->rq_repmsg
1896 because it can be accessed by reply_in_callback() without the
1900 Frequency : only on Cray X2
1902 Description: X2 build failures
1903 Details : fix build failures on Cray X2.
1907 Description: xid & resent requests
1908 Details : Initialize RPC XID from clock at startup (randomly if clock is
1913 Description: quota recovery deadlock during mds failover
1914 Details : This patch includes att18982, att18236, att18237 in bz14840.
1916 1. fix osts hang when mds does failover with quotaon
1917 2. prevent watchdog storm when osts threads wait for the
1922 Description: kernel panic on racer
1923 Details : Do not access dchild->d_inode when IS_ERR(dchild) is true.
1925 Severity : enhancement
1927 Description: Add lustre_start utility to start or stop multiple Lustre servers
1932 Description: Lustre GPF in {:ptlrpc:ptlrpc_server_free_request+373}
1933 Details : In case of memory pressure, list_del() can be called twice on
1934 req->rq_history_list, causing a kernel oops.
1938 Description: (ptllnd_peer.c:557:kptllnd_peer_check_sends()) ASSERTION(!in_interrupt()) failed
1939 Details : fix stack overflow in the distributed lock manager by defering
1940 export eviction after a failed ast to the elt thread instead of
1941 handling it in the dlm interpret routine.
1943 Severity : enhancement
1945 Description: More exported tunables for mballoc
1946 Details : Add support for tunable preallocation window and new tunables for
1947 large/small requests
1951 Description: Detect corruption of block bitmap and checking for preallocations
1952 Details : Checks validity of on-disk block bitmap. Also it does better
1953 checking of number of applied preallocations. When corruption is
1954 found, it turns filesystem readonly to prevent further corruptions.
1958 Frequency : only for big-endian servers
1959 Description: Check for big-endian system while mounting fs with extents feature
1960 Details : Mounting a filesystem with extents feature will fail on big-endian
1961 systems since ext3-based ldiskfs is not supported on big-endian
1962 systems. Can be over-ridden with "bigendian_extents" mount option.
1966 Description: Excessive recovery window
1967 Details : With AT enabled, the recovery window can be excessively long (6000+
1968 seconds). To address this problem, we no longer use
1969 OBD_RECOVERY_FACTOR when extending the recovery window (the connect
1970 timeout no longer depends on the service time, it is set to
1971 INITIAL_CONNECT_TIMEOUT now) and clients report the old service
1972 time via pb_service_time.
1975 Descriptoin: Don't sync journal after every i/o
1976 Details : Implement write RPC replay to allow server replies for write RPCs
1977 before data is on disk. However, this feature is disabled by
1978 default since some issues leading to data corruptions have been
1979 found during recovery (e.g. bug 19128). This feature can be enabled
1980 by running the following command on the OSSs:
1981 lctl set_param obdfilter.*.sync_journal=0
1985 Description: Watchdog triggered on MDS failover
1986 Details : enable OBD_CONNECT_MDT flag when connecting from the MDS so that
1987 the OSTs know that the MDS "UUID" can be reused for the same export
1988 from a different NID, so we do not need to wait for the export to
1993 Descriptoin: Lustre detected file system corruption with inode out of bounds
1994 Details : don't update i_size on MDS_CLOSE for directories. This causes
1995 directory corruptions on the MDT.
1997 -------------------------------------------------------------------------------
1999 2008-05-26 Sun Microsystems, Inc.
2001 * Support for kernels:
2002 2.6.5-7.311 (SLES 9),
2003 2.6.9-67.0.7.EL (RHEL 4),
2004 2.6.16.54-0.2.5 (SLES 10),
2005 2.6.18-53.1.14.el5 (RHEL 5),
2006 2.6.22.14 vanilla (kernel.org)
2007 * Client support for unpatched kernels:
2008 (see http://wiki.lustre.org/index.php?title=Patchless_Client)
2009 2.6.16 - 2.6.22 vanilla (kernel.org)
2010 * Due to problems with nested symlinks and FMODE_EXEC (bug 12652),
2011 we do not recommend using patchless RHEL4 clients with kernels
2012 prior to 2.6.9-55EL (RHEL4U5).
2013 * Recommended e2fsprogs version: 1.40.7-sun1
2014 * Note that reiserfs quotas are disabled on SLES 10 in this kernel.
2015 * RHEL 4 and RHEL 5/SLES 10 clients behaves differently on 'cd' to a
2016 removed cwd "./" (refer to Bugzilla 14399).
2017 * A new quota file format has been introduced in 1.6.5.
2018 The format conversion from prior releases is handled transparently,
2019 but releases older than 1.4.12/1.6.5 will not understand this new
2020 format. The automatic format conversion can be avoided by running
2021 the following command on the MDS before upgrading:
2022 'tunefs.lustre --param="mdt.quota_type=ug1" $MDTDEV'.
2023 For more information, please refer to bugzilla 13904.
2027 Description: quota performance fix
2028 Details : quota data is written in journalled mode instead of ordered to
2029 increase performance
2033 Description: lfs support for human-readable quota grace time strings
2034 Details : lfs setquota -t and lfs quota -t represent quota grace times
2035 in "XXwXXdXXhXXmXXs" format instead of large values in seconds
2038 Frequency : always with o2ib 1.3 and sles10
2040 Description: fix build with SLES10 and o2ib v3.
2041 Details : sles10 uses diffrent name for Module.symver file but configure
2042 assume this file has same name on RHEL/SLES/vanila kernels.
2045 Frequency : very rare, if additional xattrs are used on kernels >= 2.6.12
2047 Description: MDS may lose file striping (and hence file data) in some cases
2048 Details : If there are additional extended attributes stored on the MDS,
2049 in particular ACLs, SELinux, or user attributes (if user_xattr
2050 is specified for the client mount options) then there is a risk
2051 of attribute loss. Additionally, the Lustre file striping
2052 needs to be larger than default (e.g. striped over all OSTs),
2053 and an additional attribute must be stored initially in the
2054 inode and then increase in size enough to be moved to the
2055 external attribute block (e.g. ACL growing in size) for file
2058 Severity : enhancement
2060 Description: add message levels for liblustreapi
2065 Description: MDT cannot be unmounted, reporting "Mount still busy"
2066 Details : Mountpoint references were being leaked during open reply
2067 reconstruction after an MDS restart. Drop mountpoint reference
2068 in reconstruct_open() and free dentry reference also.
2073 Description: fix for occasional failure case of -ENOSPC in recovery-small tests
2074 Details : Move the 'good_osts' check before the 'total_bavail' check. This
2075 will result in an -EAGAIN and in the exit call path we call
2076 alloc_rr() which will with increasing aggressiveness attempt to
2077 aquire precreated objects on the minimum number of required OSCs.
2081 Description: Use old size assignment to avoid deadlock
2082 Details : This reverts the changes in bugs 2369 and bug 14138 that introduced
2083 the scheduling while holding a spinlock. We do not need locking
2084 for size in ll_update_inode() because size is only updated from
2085 the MDS for directories or files without objects, so there is no
2086 other place to do the update, and concurrent access to such inodes
2087 are protected by the inode lock.
2091 Description: Use __u64 instead of int for valid bits
2095 Description: resolve "_IOWR redefined" build error on SLES10
2099 Description: dump the memory debugging after all modules are unloaded to
2100 suppress false negative in conf_sanity test 39
2104 Description: the recovery timer never expires
2105 Details : for new client connect request, the recovery timer should not be
2106 reset, otherwise recovery timer will never expired, if the old
2107 client never come. Only old client connect and first connection
2108 req should trigger recovery timer reset.
2112 Description: the min numbers of lproc stats are wrong
2113 Details : adding a new constant LC_MIN_INIT and use it for initialization
2117 Frequency : always with interactive lfs
2119 Description: Reinitialize optind to 0 so that interactive lfs works in all cases
2122 Frequency : with multiple concurrent readdir processes in same directory
2123 Bugzilla : 15406, 15169, 15175
2124 Description: misc fixes for directory readahead.
2125 Details : prevent previous statahead async RPC callback from processing the
2126 current "statahead_info", race condition between async RPC callback
2127 add dentry into dentry hash table and "ls" thread revalidate such
2128 dentry, statahead his/miss control for hidden items, and so on.
2130 Severity : enhancement
2132 Description: build kernel-ib packages for OFED 1.3 in our release cycle
2136 Description: incore types cleaning in quota code (with respect to 64-bit limits)
2137 Details : several u32 variables declarations are replaced with u64 declarations
2142 Description: fix SLES kernel versioning
2143 Details : the kernel version for our SLES 10 kernel did not include a "-"
2144 before the "smp" at the end. while this was not a problem in
2145 general, it did mean that software trying to use the kernel
2146 version to try to detect a vendor specific kernel would fail.
2147 this was most evident by the OFED build scripts.
2152 Description: Don't update lov_desc members until making sure they are valid
2153 Details : When updating lov_desc members via proc fs, need fix their
2154 validities before doing the real update.
2157 Frequency : very rare
2159 Description: don't put request into delay list while invalidate in flight.
2160 Details : ptlrpc_delay_request sometimes put in delay list while invalidate
2161 import in flight. this produce timeout for invalidate and sometimes
2162 can cause stale data.
2164 Severity : enhancement
2166 Description: Update kernel to SLES9 2.6.5-7.311.
2168 Severity : enhancement
2170 Description: Update kernel to RHEL4 2.6.9-67.0.7.
2175 Frequency : on PPC only
2176 Description: not convert ost objects for directory because it's not exist.
2177 Details : ll_dir_getstripe assume dirrectory has ost objects but this wrong.
2179 Severity : enhancement
2181 Description: Fix warnings with compile liblustre at sles10/rhel5 which have
2182 __u64 as usingied long long type.
2185 Frequency : rare, on shutdown
2187 Description: race process ast vs remove callback
2188 Details : removing callback before disconnect import open race with
2189 processing callback.
2191 Severity : enhancement
2193 Description: Update kernel to SLES9 2.6.5-7.311.
2195 Severity : enhancement
2197 Description: Files open for execute are not marked busy on SLES10
2198 Details : Add FMODE_EXEC to SLES10 SP1 server kernel series.
2200 Severity : enhancement
2202 Description: Add server support for vanilla-2.6.22.14.
2205 Frequency : occasional
2207 Description: Avoid lov_create() getting stuck in obd_statfs_rqset()
2208 Details : If an OST is down the MDS will hang indefinitely in
2209 obd_statfs_rqset() waiting for the statfs data. While for
2210 MDS QOS usage of statfs, it should not stuck in waiting.
2212 Severity : enhancement
2214 Description: Disable adaptive timeouts by default
2217 Frequency : on network error
2219 Description: panic with double free request if network error
2220 Details : mdc_finish_enqueue is finish request if any network error occuring,
2221 but it's true only for synchronus enqueue, for async enqueue
2222 (via ptlrpcd) this incorrect and ptlrpcd want finish request
2226 Frequency : rare, on recovery
2228 Description: read procfs can produce deadlock in some situation
2229 Details : Holding lprocfs lock with send rpc can produce block for destroy
2230 obd objects and this also block reconnect with -EALREADY.
2231 This isn't fix all lprocfs bugs - but make it rare.
2233 Severity : enhancement
2235 Description: Update kernel to RHEL5 2.6.18-53.1.14.el5.
2238 Frequency : frequent on X2 node
2240 Description: mdc_set_open_replay_data LBUG
2241 Details : Set replay data for requests that are eligible for replay.
2246 Description: lustre_mgs: operation 101 on unconnected MGS
2247 Details : When MGC is disconnected from MGS long enough, MGS will evict the
2248 MGC, and late on MGC cannot successfully connect to MGS and a lot
2249 of the error messages complaining that MGS is not connected.
2252 Frequency : rare, depends on device drivers and load
2254 Description: MDS or OSS nodes crash due to stack overflow
2255 Details : Code changes in 1.6.4 increased the stack usage of some functions.
2256 In some cases, in conjunction with device drivers that use a lot
2257 of stack, the MDS (or possibly OSS) service threads could overflow
2258 the stack. One change which was identified to consume additional
2259 stack has been reworked to avoid the extra stack usage.
2261 Severity : enhancement
2263 Description: Update to RHEL5 latest kernel-2.6.18-53.1.13.el5.
2265 Severity : enhancement
2267 Description: Update to SLES10 SP1 latest kernel-2.6.16.54-0.2.5.
2269 Severity : enhancement
2271 Description: Update to RHEL5 latest kernel-2.6.18-53.1.6.el5.
2273 Serverity : enhancement
2275 Description: Update RHEL4 kernel to 2.6.9-67.0.4.
2278 Frequency : rare on shutdown OST
2280 Description: Don't allow skipping OSTs if index has been specified.
2281 Details : Don't allow skipping OSTs if index has been specified, make locking
2282 in internal create lots better.
2287 Description: ASSERTION(!PageDirty(page)) failed
2288 Details : Wrong check could lead to an assertion failure under specific
2294 Description: LBUG in ptlrpc_check_set() bad phase ebc0de00
2295 Details : access to bitfield in structure is always rounded to long
2296 and this produce problem with not atomic change any bit.
2301 Description: Lustre make rpms failed.
2302 Details : Remove ldiskfs spec file to avoids rpmbuild be confused when
2303 builds Lustre rpms from tarball.
2305 Severity : enhancement
2307 Description: Update to SLES9 SP4 kernel-2.6.5-7.308.
2310 Frequency : rare on shutdown OST
2312 Description: If llog cancel was not send before clean_exports phase, this can
2313 produce deadlock in llog code.
2314 Details : If llog thread has last reference to obd and call class_import_put
2315 this produce deadlock because llog_cleanup_commit_master wait when
2316 last llog_commit_thread exited, but this never success because was
2317 called from llog_commit_thread.
2320 Frequency : only if OST index is skipped
2322 Description: NULL lov_tgts causing MDS oops
2323 Details : more safe checks for NULL lov_tgts for avoid oops.
2325 Severity : enhancement
2327 Description: Update to RHEL4 latest kernel-2.6.9-67.0.1.EL.
2329 Severity : enhancement
2331 Description: Update to RHEL5 latest kernel-2.6.18-53.1.4.el5.
2336 Description: make mgs_setparam() handle fsname containing dash
2337 Details : fsname containing a dash does not work with lctl conf_param
2339 Severity : enhancement
2341 Description: Update to RHEL4 Update-6 kernel-2.6.9-67.EL.
2344 Frequency : rare, in recovery and (or) destroy lovobjid file.
2346 Description: rewrite lov objid code.
2347 Details : Cleanup for lov objid code, remove scability problems and wrong
2348 locking. Fix sending last_id into ost.
2350 Severity : enhancement
2352 Description: Update to SLES10 SP1 latest kernel-2.6.16.54-0.2.3.
2354 Severity : enhancement
2356 Description: Update to RHEL5 Update-1 kernel 2.6.18-53.el5.
2357 Details : Use d_move_locked instead of __d_move.
2360 Frequency : rare, at shutdown
2362 Description: access already free / zero obd_namespace.
2363 Details : if client_disconnect_export was called without force flag set,
2364 and exist connect request in flight, this can produce access to
2365 NULL pointer (or already free pointer) when connect_interpret
2366 store ocd flags in obd_namespace.
2369 Frequency : only at startup
2371 Description: not alloc memory with spinlock held.
2372 Details : allocation memory with GFP_KERNEL can produce sleep deadlock,
2373 if any spinlock held.
2378 Description: lfs find does not continue on file error
2379 Details : Continue other files processing when a file/dir is absent.
2383 Description: Inconsistent usage of lustre_pack_reply()
2384 Details : Standardize the usage of lustre_pack_reply() such that it
2385 always generate a CERROR on failure.
2388 Frequency : very rare
2390 Description: Fix replay if there is an un-replied request and open
2391 Details : In some cases, older replay request will revert the
2392 mcd->mcd_last_xid on MDS which is used to record the client's
2393 latest sent request.
2395 Severity : enhancement
2397 Description: Update to RHEL5 kernel 2.6.18-8.1.15.el5.
2399 Severity : enhancement
2401 Description: Update to SLES10 SP1 kernel 2.6.16.53-0.16
2403 Severity : enhancement
2405 Description: Update to SLES9 kernel-2.6.5-7.287.3.
2407 Severity : enhancement
2409 Description: Update to RHEL4 kernel-2.6.9-55.0.12.EL.
2411 Severity : enhancement
2413 Description: Build SLES10 patchless client fails
2414 Details : The configure was broken by run ./configure with
2415 --with-linux-obj=.... argument for patchless client. When the
2416 configure use --with-linux-obj, the LINUXINCLUDE= -Iinclude
2417 can't search header adequately. Use absolute path such as
2418 -I($LINUX)/include instead.
2420 Severity : enhancement
2422 Description: Lustre Page Accounting
2423 Details : New macros for page alloc and free which enable accounting
2424 of page allocation of Lustre. Use percpu counters to store memory
2425 and page statistics.
2428 Frequency : only if debugging is disabled
2430 Description: LASSERT_{REQ,REP}SWAB macros are buggy
2431 Details : If SWAB_PARANOIA is disabled, the LASSERT_REQSWAB and
2432 LASSERT_REPSWAB macros become no-ops, which is incorrect. Drop
2433 these macros and replace them with their definitions instead.
2438 Description: interrupt oig_wait produce painc on resend.
2439 Details : brw_redo_request can be used for resend requests from ptlrpcd and
2440 private set, and this produce situation when rq_ptlrpcd_data not
2441 copyed to new allocated request and triggered LBUG on assert
2442 req->rq_ptlrpcd_data != NULL. But this member used only for wakeup
2443 ptlrpcd set if request is changed and can be safety changed to use
2446 Severity : enhancement
2448 Description: organize the server-side client stats on per-nid basis
2449 Details : Change the structure of stats under obdfilter and mds to
2457 The "uuid"s file would list the uuids of _active_ exports.
2458 And the clear entry is to clear all stats and stale nids.
2463 Description: Processes looping in ll_readdir() on Lustre clients finally causing
2464 a full node pseudo-hang
2465 Details : Concurrent access to the same directory from multiple clients with
2466 intensive file creation/removal can cause a client node to spin in
2467 ll_readdir(). i_version must be increased every time the lock is
2468 cancelled to ensure a revalidate is done.
2473 Description: touch file failed when fs is not full
2474 Details : OST in recovery should not be discarded by MDS in alloc_qos(),
2475 otherwise we can get ENOSP while fs is not full.
2478 Frequency : only for Cray XT3
2479 Bugzilla : 12829/13455
2480 Description: Changing primary group doesn't change the group lustre assigns to
2482 Details : When CRAY_XT3 is defined, the fsgid supplied by the client is
2483 overridden with the primary group provided by the group upcall,
2484 whereas the supplied fsgid can be trusted if it is in the list of
2485 supplementary groups returned by the group upcall.
2487 Severity : enhancement
2489 Description: Root Squash Functionality
2490 Details : Implementation of NFS-like root squash capability. Specifically,
2491 don't allow someone with root access on a client node to be able
2492 to manipulate files owned by root on a server node.
2494 Severity : enhancement
2496 Description: Slow trucate/writes to huge files at high offsets.
2497 Details : Directly associate cached pages to lock that protect those pages,
2498 this allows us to quickly find what pages to write and remove
2499 once lock callback is received.
2504 Description: Too many locks accumulating on client during NFS usage
2505 Details : mds_open improperly used accmode to find out access mode to a
2506 file. Also mdc_intent_lock logic to find out if we already have
2507 lock similar to just received was flawed since introduction of
2508 skiplists - locks are now added to the front of the granted
2513 Description: Hit ASSERTION(obd->obd_stopping == 1) failed in some setup failed
2515 Details : In obd setup failure handler, obd_stopping will not necessarily to
2516 be 1, and obd_set_up should also be checked to make sure whether
2517 obd is completely setup.
2519 Severity : enhancement
2521 Description: Allow masking D_WARNING, D_ERROR messages from console
2522 Details : Console messages can now be disabled via lnet.printk.
2527 Description: User code with malformed file open parameter crashes client node
2528 Details : Before packing join_file req, all the related reference should be
2529 checked carefully in case some malformed flags cause fake join_file
2535 Description: shrink/enlarge qunit size when needed; fix the problem of coarse
2536 grain of quota doing harm to quota's accuracy
2537 Details : qunit size will be changed when quota limitation is too low/high;
2538 record the pending quota write in order to get more accureate
2539 quota; delete the patch for bug12588, which is unnecessary when
2540 this patch is landed. This bug also contains fixes for bug 14526,
2541 14299, 14601 and 13794.
2545 Description: LDLM_ENQUEUE races with LDLM_CP_CALLBACK
2546 Details : ldlm_completion_ast() assumes that a lock is granted when the req
2547 mode is equal to the granted mode. However, it should also check
2548 that LDLM_FL_CP_REQD is not set.
2552 Description: Heavy nfs access might result in deadlocks
2553 Details : After ELC code landed, it is now improper to enqueue any mds
2554 locks under och_sem, because enqueue might want to decide to
2555 cancel open locks for same inode we are holding och_sem for.
2559 Description: 35% write performance drop with ldiskfs2 when quotas are on
2560 Details : Enable ext3 journalled quota by default to improve performance
2561 when quotas are turned on.
2565 Description: Client eviction while running blogbench
2566 Details : A lot of unlink operations with concurrent I/O can lead to a
2567 deadlock causing evictions. To address the problem, the number of
2568 oustanding OST_DESTROY requests is now throttled to
2569 max_rpcs_in_flight per OSC and LDLM_FL_DISCARD_DATA blocking
2570 callbacks are processed in priority.
2573 Frequency : RHEL4 only
2575 Description: mkfs is very slow on IA64/RHEL4
2576 Details : A performance regression has been discovered in the MPT Fusion
2577 driver between versions 3.02.73rh and 3.02.99.00rh. As a
2578 consequence, we have downgraded the MPT Fusion driver in the RHEL4
2579 kernel from 3.02.99.00 to 3.02.73 until this problem is fixed.
2582 Frequency : PPC/PPC64 only
2584 Description: conflicts between asm-ppc64/types.h and lustre_types.h
2585 Details : fix duplicated definitions between asm-ppc64/types.h and
2586 lustre_types.h on PPC.
2589 Frequency : PPC/PPC64 only
2591 Description: asm-ppc/segment.h does not exist
2592 Details : fix compile issue on PPC.
2596 Description: data checksumming impacts single node performance
2597 Details : add support for several checksum algorithms. Currently, CRC32 and
2598 Adler-32 are supported. The checksum type can be changed on the fly
2599 through /proc/fs/lustre/osc/*/checksum_type.
2603 Description: use adler32 for page checksums
2604 Details : when available, use the Adler-32 algorithm instead of CRC32 for
2609 Description: better handle error messages in extents code
2611 Severity : enhancement
2613 Description: SNMP support enhancement
2614 Details : Adding total number of sampled request for an MDS node in snmp
2617 Severity : enhancement
2619 Description: Optimize ldlm waiting list processing for PR extent locks
2620 Details : When processing waiting list for read extent lock and meeting read
2621 lock that is same or wider to it that is not contended, skip
2622 processing rest of the list and immediatelly return current
2623 status of conflictness, since we are guaranteed there are no
2624 conflicting locks in the rest of the list.
2628 Description: Time out and refuse to reconnect
2629 Details : When the failover node is the primary node, it is possible
2630 to have two identical connections in imp_conn_list. We must
2631 compare not conn's pointers but NIDs, otherwise we
2632 can defeat connection throttling.
2636 Description: Client not clear own cache if answer to reconnect is lost.
2637 Details : Client gets evicted from server. Now client also thinks it is
2638 disconnected (or gets ENOTCONN on its operation) and decides to
2639 reconnect. Server receives reconnect message, but cannot find
2640 export. New export is created that is fully valid (new cookie!),
2641 but reply is lost and not reported to client. Client reconnects
2642 again and gets back a just-created connection, but it is not new
2643 so client thinks it was not evicted and does not do recovery.
2647 Description: Detect stride IO mode in read-ahead
2648 Details : When a client does stride read, read-ahead should detect that and
2649 read-ahead pages according to the detected stride pattern.
2653 Description: build for x2 fails
2654 Details : fix compile issue on Cray systems.
2656 Severity : enhancement
2658 Description: implement readv/writev APIs(aio_read/aio_writes in newer kernels)
2659 Details : This greatly improves speed of NFS writes on 2.6 kernels.
2662 Frequency : only on PPC/SLES10
2664 Description: "BITS_PER_LONG is not 32 or 64" in linux/idr.h
2665 Details : On SLES10/PPC, fs.h includes idr.h which requires BITS_PER_LONG to
2666 be defined. Add a hack in mkfs_lustre.c to work around this compile
2671 Description: LASSERT on MDS when client holding flock lock dies
2672 Details : ldlm pool logic depends on number of granted locks equal to
2673 number of released locks which is not true for flock locks, so
2674 just exclude such locks from consideration.
2678 Description: MDS deadlock with many ll_sync_lov threads and I/O stalled
2679 Details : Use fsfilt_sync() for both the whole filesystem sync and
2680 individual file sync to eliminate dangerous inode locking
2681 with I_LOCK that can lead to a deadlock.
2685 Description: Update an obsolete wirecheck.c generator
2686 Details : Update wirecheck.c/wirehdr.c and regenerate wiretest.c
2690 Description: Client can panic on open sometimes
2691 Details : It is possible that we try to free already freed request in
2692 ll_file_open in some error cases when we send request from
2697 Description: performance in 1.6.3
2698 Details : Force q->max_phys_segments to MAX_PHYS_SEGMENTS on SLES10 to be
2699 sure that 1MB requests are not fragmented by the block layer.
2703 Description: LDLM soft lockups - improvement
2704 Details : It is be possible to send the lock handle along with each read
2705 or write request because the client is already doing a lock match
2706 itself so there isn't any reason the OST should have to re-do that
2712 Description: lfs quota fails with deactivated OSTS
2713 Details : With this patch, three improvements are included:
2714 1. detete the softlimit in mds and osts when use "lfs quota".
2715 2. display the inaccurate data in the output of "lfs quota".
2716 3. try to get quota info when "lfs quota" is executed.
2721 Description: Extent locks not granted with no conflicts sometimes.
2722 Details : When race occurs in glimpse handler and nothing is returned,
2723 we do not reprocess the queue after lock cancel, and that leads
2724 to a stall until next activity on a resource
2727 Frequency : failover with quotaon
2729 Description: during mds failovers with quota on, OSTs got into deadlock state
2730 and causing dumpstack.
2731 Details : for every quota slave, at any time, there is only one quota req
2732 is sent to quota master for every uid/gid. Before that quota req
2733 returns, all the thread relative to the same uid/gid will wait.
2734 So if the quota req is lost because mds failovers or any other
2735 reasons, this bug will be hit. Now, dqacq_interpret() will handle
2736 quota reqs who time out.
2738 Severity : enhancement
2741 Description: when quota slave checks if quota is enough, there is an unnecessary
2743 Details : place this wait on necessary place instead of always waiting.
2745 --------------------------------------------------------------------------------
2747 2007-12-07 Cluster File Systems, Inc. <info@clusterfs.com>
2749 * Support for kernels:
2750 2.6.5-7.286 (SLES 9),
2751 2.6.9-55.0.9.EL (RHEL 4),
2752 2.6.16.53-0.8 (SLES 10),
2753 2.6.18-8.1.14.el5 (RHEL 5),
2754 2.6.18.8 vanilla (kernel.org)
2755 * Client support for unpatched kernels:
2756 (see http://wiki.lustre.org/index.php?title=Patchless_Client)
2757 2.6.16 - 2.6.22 vanilla (kernel.org)
2758 * Due to recently discovered recovery problems, we do not recommend
2759 using patchless RHEL 4 clients with this or any earlier release.
2760 * Recommended e2fsprogs version: 1.40.2-cfs1
2761 * Note that reiserfs quotas are disabled on SLES 10 in this kernel.
2764 Frequency : occasional
2766 Description: MDS hang or stay in waiting lock
2767 Details : If client receive lock with CBPENDING flag ldlm need send lock
2768 cancel as separate rpc, to avoid situation when cancel request
2769 can't processed due all i/o threads stay in wait lock.
2772 Frequency : occasional
2774 Description: Do not fail import if osc_interpret_create gets -EAGAIN
2775 Details : If osc_interpret_create got -EAGAIN it immediately exits and
2776 wakeup oscc_waitq. After wakeup oscc_wait_for_objects call
2777 oscc_has_objects and see OSC has no objests and call
2778 oscc_internal_create to resend create request.
2780 Severity : enhancement
2782 Description: Update kernel patches for SLES10 2.6.16.53-0.8.
2783 Details : Update which_patch & target file for SLES10 latest kernel.
2785 Severity : enhancement
2787 Description: add --type and --size parameters to lfs find
2788 Details : Enhance lfs find by adding filetype and filesize parameters. Also
2789 multiple OBDs can now be specified for the --obd option.
2791 Severity : enhancement
2793 Description: eliminate client locks in face of contention
2794 Details : file contention detection and lockless i/o implementation
2795 for contended files.
2797 Severity : enhancement
2799 Description: Remove client patches from SLES 10 kernel.
2800 Details : This causes SLES 10 clients to behave as patchless clients
2801 even on a Lustre-patched (server) kernel.
2803 Severity : enhancement
2805 Description: use i_size_read and i_size_write in 2.6 port
2806 Details : replace inode->i_size access with i_size_read/write()
2809 Frequency : when removing large files
2811 Description: scheduling issue during removal of large Lustre files
2812 Details : Don't take the BKL in fsfilt_ext3_setattr() for 2.6 kernels.
2813 It causes scheduling issues when removing large files (17TB in the
2819 Description: 1.4.11 Can't handle directories with stripe set and extended ACLs
2820 Details : Impossible (EPROTO is returned) to access a directory that has a
2821 non-default striping and ACLs.
2824 Frequency : only on ppc
2826 Description: /proc/fs/lustre/devices broken on ppc
2827 Details : The patch as applied to 1.6.2 doesn't look correct for all arches.
2828 We should make sure the type of 'index' is loff_t and then cast
2829 explicitly as needed below. Do not assign an explicitly cast
2833 Frequency : only for rhel5
2835 Description: Kernel patches update for RHEL5 2.6.18-8.1.10.el5.
2836 Details : Modify the target file & which_kernel.
2839 Frequency : if the uninit_groups feature is enabled on ldiskfs
2841 Description: e2fsck reports "invalid unused inodes count"
2842 Details : If a new ldiskfs filesystem is created with the "uninit_groups"
2843 feature and only a single inode is created in a group then the
2844 "bg_unused_inodes" count is incorrectly updated. Creating a
2845 second inode in that group would update it correctly.
2850 Description: buffer overruns could theoretically occur
2851 Details : llapi_semantic_traverse() modifies the "path" argument by
2852 appending values to the end of the origin string, and a buffer
2853 overrun may occur. Adding buffer overrun check in liblustreapi.
2855 Severity : enhancement
2857 Description: Add jbd statistics patch for RHEL5 and 2.6.18-vanilla.
2860 Frequency : only if filesystem is inconsistent
2862 Description: handle "serious error: objid * already exists" more gracefully
2863 Details : If LAST_ID value on disk is smaller than the objects existing in
2864 the O/0/d* directories, it indicates disk corruption and causes an
2865 LBUG(). If the object is 0-length, then we should use the existing
2866 object. This will help to avoid a full fsck in most cases.
2868 Severity : enhancement
2870 Description: Kernel patches update for RHEL4 2.6.9-55.0.6.
2871 Details : Modify vm-tunables-rhel4.patch.
2873 Severity : enhancement
2875 Description: Kernel config for 2.6.18-vanilla.
2876 Details : Modify targets/2.6-vanilla.target.in.
2877 Add config file kernel-2.6.18-2.6-vanilla-i686.config.
2878 Add config file kernel-2.6.18-2.6-vanilla-i686-smp.config.
2879 Add config file kernel-2.6.18-2.6-vanilla-x86_64.config.
2880 Add config file kernel-2.6.18-2.6-vanilla-x86_64-smp.config.
2883 Frequency : occasional
2885 Description: improve handling recoverable errors
2886 Details : If request processed with error which can be recoverable on server
2887 request should be resend, otherwise page released from cache and
2893 Description: Kernel patches update for RHEL5 2.6.18-8.1.14.el5.
2894 Details : Modify target file & which_patch.
2895 A flaw was found in the IA32 system call emulation provided
2896 on AMD64 and Intel 64 platforms. An improperly validated 64-bit
2897 value could be stored in the %RAX register, which could trigger an
2898 out-of-bounds system call table access. An untrusted local user
2899 could exploit this flaw to run code in the kernel
2900 (ie a root privilege escalation). (CVE-2007-4573).
2904 Description: change order of libsysio includes
2905 Details : '#include sysio.h' should always come before '#include xtio.h'
2907 Severity : enhancement
2909 Description: adapt the lustre_config script to support the upgrade case
2910 Details : Add "-u" option for lustre_config script to support upgrading 1.4
2911 server targets to 1.6 in parallel.
2916 Description: To avoid grant space > avaible space when the disk is almost
2917 full. Without this patch you might see the error "grant XXXX >
2918 available" or some LBUG about grant, when the disk is almost
2920 Details : In filter_check_grant, for non_grant cache write, we should
2921 check the left space by if (*left > ungranted + bytes), instead
2922 of (*left > ungranted), because only we are sure the left space
2923 is enough for another "bytes", then the ungrant space should be
2924 increase. In client, we should update cl_avail_grant only there is
2925 OBD_MD_FLGRANT in the reply.
2930 Description: Update RHEL 4 kernel to fix local root privilege escalation.
2931 Details : Update to the latest RHEL 4 kernel to fix the vulnerability
2932 described in CVE-2007-4573. This problem could allow untrusted
2933 local users to gain root access.
2936 Frequency : when using O_DIRECT and quotas
2938 Description: Incorrect file ownership on O_DIRECT output files
2939 Details : block usage reported by 'lfs quota' does not take into account
2940 files that have been written with O_DIRECT.
2944 Description: (rw.c:1323:ll_read_ahead_pages()) ASSERTION(page_idx > ria->ria_stoff) failed
2945 Details : Once the unmatched stride IO mode is detected, shrink the stride-ahead
2946 window to 0. If it does hit cache miss, and read-pattern is still
2947 stride-io mode, does not reset the stride window, but also does not
2948 increase the stride window length in this case.
2950 --------------------------------------------------------------------------------
2952 2007-09-27 Cluster File Systems, Inc. <info@clusterfs.com>
2954 * Support for kernels:
2955 2.6.5-7.286 (SLES 9),
2956 2.6.9-55.0.2.EL (RHEL 4),
2957 2.6.16.46-0.14 (SLES 10),
2958 2.6.18-8.1.8.el5 (RHEL 5),
2959 2.6.18.8 vanilla (kernel.org)
2960 * Client support for unpatched kernels:
2961 (see http://wiki.lustre.org/index.php?title=Patchless_Client)
2962 2.6.16 - 2.6.21 vanilla (kernel.org)
2963 * Due to recently discovered recovery problems, we do not recommend
2964 using patchless RHEL 4 clients with this or any earlier release.
2965 * Recommended e2fsprogs version: 1.40.2-cfs1
2966 * Note that reiserfs quotas are disabled on SLES 10 in this kernel.
2970 Description: Fix errors in lfs documentation
2971 Details : Fixes man pages
2973 Severity : enhancement
2975 Description: Adaptive timeouts
2976 Details : RPC timeouts adapt to changing server load and network
2977 conditions to reduce resend attempts and improve recovery time.
2979 Severity : enhancement
2981 Description: llapi_file_create() does not allow some changes
2982 Details : add llapi_file_open() that allows specifying the file creation
2983 mode and open flags, and also returns an open file handle.
2985 Severity : enhancement
2987 Description: df doesn't work properly if diskfs blocksize != 4K
2988 Details : Choose biggest blocksize of OST's as the LOV's blocksize.
2991 Frequency : always on directories with default striping set
2993 Description: lfs find on -1 stripe looping in lsm_lmm_verify_common()
2994 Details : Avoid lov_verify_lmm_common() on directory with -1 stripe count.
2996 Severity : enhancement
2998 Description: merge and cleanup kernel patches.
2999 Details : Remove mnt_lustre_list in vfs_intent-2.6-rhel4.patch.
3001 Severity : enhancement
3003 Description: RedHat Update kernel for RHEL5
3004 Details : Kernel config file for RHEL5.
3006 Severity : enhancement
3008 Description: OSS needs mutliple precreate threads
3009 Details : Add ability to start more than one create thread per OSS.
3012 Frequency : only with quota on the root user
3014 Description: mds_obd_create error creating tmp object
3015 Details : When the user sets quota on root, llog will be affected and can't
3016 create files and write files.
3019 Frequency : Always on ia64 patchless client, and possibly others.
3021 Description: Add EXPORT_SYMBOL check for node_to_cpumask symbol.
3022 Details : This allows the patchless client to be loaded on architectures
3023 without this export.
3028 Description: disorder of journal start and llog_add cause deadlock.
3029 Details : in llog_origin_connect, journal start should happen before
3030 llog_add keep the same order as other functions to avoid
3033 Severity : enhancement
3035 Description: RedHat Update kernel for RHEL5
3036 Details : Modify the kernel config file more closer RHEL5.
3038 Severity : enhancement
3040 Description: Build failure against Centos5 (RHEL5)
3041 Details : Define PAGE_SIZE when it isn't present.
3044 Frequency : occasionally when using NFS
3046 Description: "ll_intent_file_open()) lock enqueue: err: -13" with nfs
3047 Details : with NFS, the anon dentry's parent was set to itself in
3048 d_alloc_anon(), so in MDS, we use rec->ur_fid1 to find the
3049 corresponding dentry other than use rec->ur_name.
3051 Severity : enhancement
3053 Description: enable data checksumming by default
3054 Details : enable checksum by default, allow --disable-checksum
3055 configure option and "-o nochecksum" mount option. Checksums
3056 can also be disabled at runtime via $LPROC/osc/*/checksum_pages.
3059 Frequency : Occasionally with failover
3061 Description: Client eviction due to failover config
3062 Details : after a connection loss, the lustre client should attempt to
3063 reconnect to the last active server first before trying the
3064 other potential connections.
3066 Severity : enhancement
3068 Description: client-side metadata stat-ahead during readdir(directory readahead)
3069 Details : perform client-side metadata stat-ahead when the client detects
3070 readdir and sequential stat of dir entries therein
3075 Description: when mds and osts use different quota unit(32bit and 64bit),
3076 quota will be released repeatly.
3077 Details : void sending multiple quota reqs to mds, which will keep the
3078 status between the reqs.
3081 Frequency : only with liblustre clients on XT3
3083 Description: evictions taking too long
3084 Details : allow llrd to evict clients directly on OSTs
3088 Description: osts not allocated evenly to files
3089 Details : change the condition to increase offset_idx
3091 Severity : enhancement
3093 Description: self-adjustable client's lru lists
3094 Details : use adaptive algorithm for managing client cached locks lru
3095 lists according to current server load, other client's work
3096 pattern, memory activities, etc. Both, server and client
3097 side namespaces provide number of proc tunables for controlling
3102 Description: rewrite ext2-derived code in llite/dir.c and obdclass/uuid.c
3103 Details : rewrite inherited code (uuid parsing code from ext2 utils and
3104 readdir code from ext3) from scratch preserving functionality.
3108 Description: Only those disconnect error should be returned by rq_status.
3109 Details : In open/enqueue processs, Some errors, which will cause client
3110 disconnected, should be returned by rq_status, while other
3111 errors should still be returned by intent, then mdc or llite will
3114 Severity : enhancement
3116 Description: Tune the kernel for good SCSI performance.
3117 Details : Set the value of /sys/block/{dev}/queue/max_sectors_kb
3118 to the value of /sys/block/{dev}/queue/max_hw_sectors_kb
3122 Frequency : Always for filesystems larger than 2TB on 32-bit systems.
3123 Bugzilla : 13547, 13627
3124 Description: Data corruption for OSTs that are formatted larger than 2TB
3126 Details : When generating the bio request for lustre file writes the
3127 sector number would overflow a temporary variable before being
3128 used for the IO. The data reads correctly from Lustre (which
3129 will overflow in a similar manner) but other file data or
3130 filesystem metadata may be corrupted in some cases.
3134 Description: "lfs find -obd UUID" prints directories
3135 Details : "lfs find -obd UUID" will return all directory names instead
3136 of just file names. It is incorrect because the directories
3137 do not reside on the OSTs.
3140 Bugzilla : 13596, 13614
3141 Description: MDS hang after unclean shutdown of lots of clients
3142 Details : Never resend AST requests.
3146 Frequency : Always, for kernels after 2.6.16
3147 Description: Fix warning idr_remove called for id=.. which is not allocated.
3148 Details : Last kernels save old s_dev before kill super and not allow
3149 to restore from callback - restore it before call kill_anon_super.
3153 Description: Fix error on 'ls .' at the top of the Lustre mount.
3154 Details : Don't revalidate dentry if it is a root dentry.
3156 --------------------------------------------------------------------------------
3158 2007-08-27 Cluster File Systems, Inc. <info@clusterfs.com>
3160 * Support for kernels:
3161 2.6.5-7.286 (SLES 9),
3162 2.6.9-55.0.2.EL (RHEL 4),
3163 2.6.16.46-0.14 (SLES 10),
3164 2.6.18-8.1.8.el5 (RHEL 5),
3165 2.6.18.8 vanilla (kernel.org)
3166 * Client support for unpatched kernels:
3167 (see http://wiki.lustre.org/index.php?title=Patchless_Client)
3168 2.6.16 - 2.6.22 vanilla (kernel.org)
3169 * Due to recently discovered recovery problems, we do not recommend
3170 using patchless RHEL 4 clients with this or any earlier release.
3171 * Recommended e2fsprogs version: 1.39.cfs8
3172 * Note that reiserfs quotas are disabled on SLES 10 in this kernel.
3177 Description: block reactivating mgc import until all deactivates complete
3178 Details : Fix race when failing back MDT/MGS to itself (testing)
3180 Severity : enhancement
3182 Description: lfs setstripe enhancement
3183 Details : Make lfs setstripe understand 'k', 'm' and 'g' for stripe size.
3186 Frequency : mds/oss recovery
3188 Description: llog ctxt is refrenced after it has been freed.
3189 Details : llog ctxt refcount was added to avoide the race between ctxt free
3190 and llog recovery process. Each llog user must hold ctxt refcount
3191 before it access the llog. And the llog ctxt can only be freed
3192 when its refcount is zero.
3194 Severity : enhancement
3196 Description: randomly memory allocation failure util
3197 Details : Make lustre randomly failed allocating memory for testing purpose.
3200 Frequency : only for SLES10
3202 Description: Update kernel patch for SLES10 SP1
3203 Details : Add patch blkdev_tunables-2.6-sles10.patch to 2.6-sles10.series.
3205 Severity : enhancement
3207 Description: omit set fsid for export NFS
3208 Details : fix set/restore device id for avoid EMFILE error and mark lustre fs
3209 as FS_REQUIRES_DEV for avoid problems with generate fsid.
3212 Frequency : after network failures
3214 Description: Add sync option to mount_lustre.c
3215 Details : Client loses data written to lustre after a network interruption.
3217 Severity : enhancement
3219 Description: Error message improvement.
3220 Details : Merging of two LCONSOLE_ERROR_MSG into one.
3222 Severity : enhancement
3224 Description: don't use GFP_* in generic Lustre code.
3225 Details : Use cfs_alloc_* functions and CFS_* flags for code portability.
3227 Severity : enhancement
3229 Description: obdclass is limited by single OBD_ALLOC(idarray)
3230 Details : replace OBD_ALLOC/OBD_FREE with OBD_VMALLOC/OBD_VFREE
3232 Severity : enhancement
3234 Description: updated patchess for new RHEL4 kernel
3235 Details : Fixed ext3-unlink-race.patch per Kalpak's comment.
3237 Severity : enhancement
3239 Description: warnings with build patchless client with vanila 2.6.19 and up
3240 Details : change old ctl_table style and replace ctl_table/ctl_table_header
3241 with cfs_sysctl_table_t/cfs_sysctl_table_header_t
3243 Severity : enhancement
3245 Description: O_DIRECT bypasses client statistics.
3246 Details : When running with O_DIRECT I/O, neither the client rpc_stats nor
3247 read_ahead_stats were updated. Copied stats section from
3248 osc_send_oap_rpc() into async_internal().
3251 Frequency : only for Cray XT3
3253 Description: peer credits not enough on many OST per OSS systems.
3254 Details : Use new lnet way to add credits as we need those for pings and ASTs
3257 Frequency : only with liblustre
3259 Description: Liblustre is not releasing flock locks on file close.
3260 Details : Release flock locks on file close.
3263 Frequency : only for RHEL4
3265 Description: Update kernel patches for kernel-2.6.9-55.0.2.EL
3266 Details : Remove inode-nr_unused-2.6.9-rhel4.patch from 2.6-rhel4.series
3267 Update target file and kernel config.
3271 Description: lustre support for RHEL5
3272 Details : Add support for RHEL5.
3277 Description: ASSERTION(export != NULL) failed in target_handle_connect
3278 Details : Assetion hit is result of rare race between disconnect and connet
3279 to same nid. target_handle_connect found old connect cockie and
3280 tried to reconnect, but can't find export for this cockie.
3285 Description: umount blocks forever on error
3286 Details : In result of wrong using obd_no_recov and obd_force flags client
3287 can hand if cancel or some other requests is lost.
3290 Frequency : Only for SLES
3292 Description: sanity_quota fail test_1
3293 Details : There are multiple occurences of $TSTUSR in SLES's /etc/group
3294 file, which makes TSTID[2] inunique.
3296 Severity : enhancement
3298 Description: Kernel patches for SLES9 2.6.5-7.286 kernel
3299 Details : Update target/ChangeLog/which_patch .
3301 Severity : enhancement
3303 Description: jbd statistics
3304 Details : Port older jbd statistics patch for sles10
3309 Description: lvbo_init failed for resource with missing objects.
3310 Details : Fix returning error if we do stat for file with missing/corrupted
3311 objects and i_size set to all sum of size of all avaible objects.
3312 if we to truncate/write to missing object - it is recreated.
3317 Description: Oops in read and write path when failing to allocate lock.
3318 Details : Check if lock allocation failed and return error back.
3321 Frequency : When flocks are used.
3323 Description: assertion failure in ldlm_cli_enquque_fini for non NULL lock.
3324 Details : Flock locks might destroy just granted lock if it could be merged
3325 with another existing flock, this is done in completion handler,
3326 so teach ldlm_cli_enquque_fini that this is a valid case for
3332 Description: reply_lock_interpret crash due to race with it and lock cancel.
3333 Details : Do not replay locks that are being cancelled. Do not reference
3334 locks by their address during replay, just by their handle.
3337 Frequency : only with deactivated OSTs
3339 Description: lstripe command fails for valid OST index
3340 Details : The stripe offset is compared to 'lov->desc.ld_tgt_count'
3341 instead of lov->desc.ld_active_tgt_count.
3343 Severity : enhancement
3345 Description: Build failure against Centos5 (RHEL5)
3346 Details : Use getpagesize() instead of PAGE_SIZE.
3348 --------------------------------------------------------------------------------
3350 2007-07-30 Cluster File Systems, Inc. <info@clusterfs.com>
3352 * Support for kernels:
3353 2.6.5-7.283 (SLES 9),
3354 2.6.9-55.EL (RHEL 4),
3355 2.6.16.46-0.14 (SLES 10),
3356 2.6.18.8 vanilla (kernel.org)
3357 * Client support for unpatched kernels:
3358 (see http://wiki.lustre.org/index.php?title=Patchless_Client)
3359 2.6.16 - 2.6.22 vanilla (kernel.org)
3360 * Due to recently discovered recovery problems, we do not recommend
3361 using patchless RHEL 4 clients with this or any earlier release.
3362 * Recommended e2fsprogs version: 1.39.cfs8
3363 * Note that reiserfs quotas are disabled on SLES 10 in this kernel.
3364 * Starting with this release, the ldiskfs backing filesystem required
3365 by Lustre is now in its own package, lustre-ldiskfs. This package
3366 should be installed. It is versioned separately from Lustre and
3367 may be released separately in future.
3372 Description: server LBUG when shutting down
3373 Details : Block umount forever until the mount refcount is zero rather
3374 than giving up after an arbitrary timeout.
3376 Severity : enhancement
3378 Description: add optional extra BUILD_VERSION info
3379 Details : add a new environment variable (namely LUSTRE_VERS) which allows
3380 to override the lustre version.
3383 Frequency : 2.6.18 servers only
3385 Description: ll_kern_mount() doesn't release the module reference
3386 Details : The ldiskfs module reference count never drops down to 0
3387 because ll_kern_mount() doesn't release the module reference.
3392 Description: server LBUG when using old ost_num_threads parameter
3393 Details : Accept the old ost_num_threads parameter but warn that it
3394 is deprecated, and fix an off-by-one error that caused an LBUG.
3399 Description: Transient SCSI error results in persistent IO issue
3400 Details : iobuf->dr_error is not reinitialized to 0 between two
3404 Frequency : sometimes when underlying device returns I/O errors
3406 Description: OSTs not going read-only during write failures
3407 Details : OSTs are not remounted read-only when the journal commit threads
3408 get I/O errors because fsfilt_ext3 calls journal_start/stop()
3409 instead of the ext3 wrappers.
3412 Frequency : only with 10000 clients or more
3414 Description: poor connect scaling with increasing client count
3415 Details : Don't run filter_grant_sanity_check for more than 100 exports
3416 to improve scaling for large numbers of clients.
3419 Frequency : SLES10 only
3421 Description: sanity-quota.sh quotacheck failed: rc = -22
3422 Details : Quotas cannot be enabled on SLES10.
3425 Frequency : liblustre clients only
3427 Description: getdirentries does not give error when run on compute nodes
3428 Details : getdirentries does not fail when the size specified as an argument
3429 is too small to contain at least one entry
3431 Severity : enhancement
3433 Description: Add LNET router traceability for debug purposes
3434 Details : If a checksum failure occurs with a router as part of the
3435 IO path, the NID of the last router that forwarded the bulk data
3436 is printed so it can be identified.
3441 Description: OST "spontaneously" evicts client; client has imp_pingable == 0
3442 Details : Due to a race condition, liblustre clients were occasionally
3443 evicted incorrectly.
3445 Severity : enhancement
3447 Description: lfs setstripe use optional parameters instead of postional
3450 Severity : enhancement
3452 Description: Nanosecond timestamp support for ldiskfs
3453 Details : The on-disk ldiskfs filesystem has added support for nanosecond
3454 resolution timestamps. There is not yet support for this at
3455 the Lustre filesystem level.
3458 Frequency : during server recovery
3460 Description: MDS failing to send precreate requests due to OSCC_FLAG_RECOVERING
3461 Details : request with rq_no_resend flag not awake l_wait_event if they get a
3465 Frequency : nfs export on patchless client
3467 Description: connectathon hang when test nfs export over patchless client
3468 Details : Disconnected dentry cannot be found with lookup, so we do not need
3469 to unhash it or make it invalid
3472 Description: fix llapi_lov_get_uuids() to allow many OSTs to be returned
3473 Details: : Change llapi_lov_get_uuids() to read the UUIDs from /proc instead
3474 of using an ioctl. This allows lfsck for > 160 OSTs to succeed.
3479 Description: open req refcounting wrong on reconnect
3480 Details : If reconnect happened between getting open reply from server and
3481 call to mdc_set_replay_data in ll_file_open, we will schedule
3482 replay for unreferenced request that we are about to free.
3483 Subsequent close will crash in variety of ways.
3484 Check that request is still eligible for replay in
3485 mdc_set_replay_data().
3490 Description: disable writes to filesystem when reading health_check file
3491 Details : the default for reading the health_check proc file has changed
3492 to NOT do a journal transaction and write to disk, because this
3493 can cause reads of the /proc file to hang and block HA state
3494 checking on a healthy but otherwise heavily loaded system. It
3495 is possible to return to the previous behaviour during configure
3496 with --enable-health-write.
3498 Severity : enhancement
3500 Description: 64-bit inode version
3501 Details: : Add a on-disk 64-bit inode version for ext3 to track changes made
3502 to the inode. This will be required for version-based recovery.
3507 Description: MDS fails to start if a duplicate client export is detected
3508 Details : in some rare cases it was possible for a client to connect to
3509 an MDS multiple times. Upon recovery the MDS would detect this
3510 and fail during startup. Handle this more gracefully.
3512 Severity : enhancement
3514 Description: Add -o localflock option to simulate old noflock behaviour.
3515 Details : This will achieve local-only flock/fcntl locks coherentness.
3520 Description: log_commit_thread vs filter_destroy race leads to crash
3521 Details : Take import reference before releasing llog record semaphore
3526 Description: Wrong request locking in request set processing
3527 Details : ptlrpc_check_set wrongly uses req->rq_lock for proctect add to
3528 imp_delayed_list, in this place should be used imp_lock.
3531 Frequency : when reconnecting
3533 Description: Grant leak when OSC reconnect to OST
3534 Details : When osc reconnect ost, OST(filter) should check whether it
3535 should grant more space to client by comparing fed_grant and
3536 cl_avail_grant, and return the granted space to client instead
3537 of "new granted" space, because client will call osc_init_grant
3538 to update the client grant space info.
3541 Frequency : when client reconnects to OST
3543 Description: Grant leak when OSC does a resend and replays bulk write
3544 Details : When osc reconnect to OST, OST(filter) should clear grant info of
3545 bulk write request, because the grant info while be sync between
3546 OSC and OST when reconnect, and we should ignore the grant info
3547 these of resend/replay write req.
3552 Description: Grant space more than avaiable space sometimes.
3553 Details : When then OST is about to be full, if two bulk writing from
3554 different clients came to OST. Accord the avaliable space of the
3555 OST, the first req should be permitted, and the second one
3556 should be denied by ENOSPC. But if the seconde arrived before
3557 the first one is commited. The OST might wrongly permit second
3558 writing, which will cause grant space > avaiable space.
3561 Frequency : when client is evicted
3563 Description: Grant might be wrongly erased when osc is evicted by OST
3564 Details : when the import is evicted by server, it will fork another
3565 thread ptlrpc_invalidate_import_thread to invalidate the
3566 import, where the grant will be set to 0. While the original
3567 thread will update the grant it got when connecting. So if
3568 the former happened latter, the grant will be wrongly errased
3569 because of this race.
3574 Description: Checking Stale with correct fid
3575 Details : ll_revalidate_it should uses de_inode instead of op_data.fid2
3576 to check whether it is stale, because sometimes, we want the
3577 enqueue happened anyway, and op_data.fid2 will not be initialized.
3579 Severity : enhancement
3581 Description: update patchless client
3582 Details : Add support for patchless client with 2.6.20, 2.6.21 and RHEL 5
3585 Frequency : only with 2.4 kernel
3587 Description: random memory corruption
3588 Details : size of struct ll_inode_info is to big for union inode.u and this
3589 can be cause of random memory corruption.
3594 Description: Memory leak in recovery
3595 Details : Lov_mds_md was not free in an error handler in mds_create_object.
3596 It should also check obd_fail before fsfilt_start, otherwise if
3597 fsfilt_start return -EROFS,(failover mds during mds recovery).
3598 then the req will return with repmsg->transno = 0 and rc = EROFS.
3599 and we met hit the assert LASSERT(req->rq_reqmsg->transno ==
3600 req->rq_repmsg->transno) in ptlrpc_replay_interpret. Fcc should
3601 be freed no matter whether fsfilt_commit success or not.
3604 Frequency : only with large numbers of cores on a single node
3606 Description: Prevents from taking the superblock lock in llap_from_page for
3608 Details : using LL_ORIGIN_REMOVEPAGE origin flag instead of LL_ORIGIN_UNKNOW
3609 for llap_from_page call in ll_removepage() prevents from taking
3610 the superblock lock for a soon died page.
3615 Description: Not check open intent error before release open handle
3616 Details : in some rare cases, the open intent error is not checked before
3617 release open handle, which may cause
3618 ASSERTION(open_req->rq_transno != 0), because it tries to release
3619 the failed open handle.
3624 Description: Set cat log bitmap only after create log success.
3625 Details : in some rare cases, the cat log bitmap is set too early. and it
3626 should be set only after create log success.
3630 Description: Accessing a block bevice can re-enable I/O when Lustre is
3631 tearing down a device.
3632 Details : dev_clear_rdonly(bdev) must be called in kill_bdev() instead of
3637 Description: service threads may hog cpus when there are a lot of requests
3638 Details : Insert cond_resched to give other threads a chance to use some CPU
3643 Description: the cat log was not initialized in recovery
3644 Details : When mds(mgs) do recovery, the tgt_count might be zero, so the
3645 unlink log on mds will not be initialized until mds post
3646 recovery. And also in mds post recovery, the unlink log will
3647 initialization will be done asynchronausly, so there will be race
3648 between add unlink log and unlink log initialization.
3652 Description: brw_stats were being printed incorrectly
3653 Details : brw_stats were being printed as log2 but all of them were not
3654 recorded as log2. Also remove some code duplication arising from
3655 filter_tally_{read,write}.
3659 Frequency : rare, only in recovery.
3660 Description: ASSERTION(req->rq_type != LI_POISON) failed
3661 Details : imp_lock should be held while iterating over imp_sending_list for
3662 prevent destroy request after get timeout in ptlrpc_queue_wait.
3666 Description: replay-single.sh test 52 fails
3667 Details : A lock's skiplist need to be cleanup when it being unlinked
3668 from its resource list.
3673 Description: Short directio read returns full requested size rather than
3675 Details : Direct I/O operations should return actual amount of bytes
3676 transferred rather than requested size.
3678 Severity : enhancement
3680 Description: metadata RPC reduction (e.g. for rm performance)
3681 Details : decrease the amount of synchronous RPC between clients and servers
3682 by canceling conflicing lock before the operation on the client
3683 and packing thier handles into the main operation RPC to server.
3685 Severity : enhancement
3687 Description: add #ifdef HAVE_KERNEL_CONFIG_H
3688 Details : kernels from 2.6.19 not need include linux/config.h, but add
3689 include linux/autoconf.h in commpiler command line.
3691 Severity : enhancement
3693 Description: patchless client support for 2.6.22 kernel
3694 Details : 2.6.22 has only one visble change, SLAB_CTOR_* constants is
3695 removed. In this case we need drop using os depended interface to
3696 kmem_cache and use cfs_mem_cache API.
3701 Description: fix mal-formatted messages
3702 Details : fix some mal-formatted DEBUG_REQ and LCONSOLE_ERROR_MSG messages
3706 Frequency : always in liblustre
3707 Description: wrong IS_ERR implementation in liblustre.h
3708 Details : fix IS_ERR implementation in liblustre.h for right detect errors.
3713 Description: Correct condition for output debug message.
3714 Details : inode i_nlink equal zero is not enough for output message about
3715 disk corruption, i_ctime and i_mode should be also checked.
3719 Frequency : always in patchless client
3720 Description: add configure check for truncate_complete_page
3721 Details : improve checks for exported symbols. This allow run check without
3722 sources, but with Module.symvers shipped with kernel distribution.
3723 add check for truncate_complete_page used by patchless client.
3727 Description: sanity.sh test_77h fails with "test_77h file compare failed"
3728 Details : test_77h uses a file which was messed by other test case.
3732 Description: Not Check whether lov_tgts is NULL in some lov functions
3733 Details : Checking whether lov_tgts is NULL in some functions.
3737 Description: replace obdo_alloc() with OBDO_ALLOC macro
3738 Details : nothing special is done in obdo_alloc() function, and for
3739 debugging purpose, it needs to be replaced with macros.
3743 Description: bad return value and errno from fcntl call
3744 Details : In liblustre API, errno should be a negative value if error
3749 Description: ptlrpc_check_set() LBUG
3750 Details : In case of positive reply from server and failed client bulk
3751 callback after bulk transfer shouldn't LBUG, but process this
3752 request as erroneous.
3754 Severity : enhancement
3756 Description: VFS operations stats tool.
3757 Details : Tool which collects stats by tracking value written in pid,
3758 ppid, gid and uses llstat to generate output to plot graph using
3760 Updated lustre/utils/Makefile.am
3761 Added lustre/utils/ltrack_stats.c
3763 Severity : enhancement
3765 Description: 2.6.18 server support (lustre 1.6.1)
3766 Details : Support for 2.6.18 kernels on the server side.
3771 Description: ASSERTION(imp->imp_conn_current) failed
3772 Details : an assertion failure is hit if a client node boots and attempts to
3773 mount a lustre filesystem faster than RECONNECT_INTERVAL seconds.
3776 Frequency : only for i686
3778 Description: 1.4.11 RC1 build fails for RHEL 4, i686
3779 Details : Fixed config variable for build.
3784 Description: Updated patchess for new RHEL4 kernel
3785 Details : Updated patch inode-nr_unused-2.6.9-rhel4.patch
3786 Updated patch jbd-stats-2.6.9.patch
3787 Updated patch qsnet-rhel4-2.6.patch
3788 Updated patch quota-deadlock-on-pagelock-core.patch
3789 Updated patch vfs_intent-2.6-rhel4.patch
3790 Updated patch vfs_races-2.6-rhel4.patch
3791 Updated series file 2.6-rhel4-titech.series
3792 Updated series file 2.6-rhel4.series
3793 Updated kernel config files
3798 Description: lquota slave complains LBUG when reconnecting with mds
3800 Details : quota slave depends qctxt->lqc_import to send its quota request.
3801 This pointer will be invalid if mds did failover or broke its
3802 connect to osts, which leads to LBUG.
3805 Frequency : when qunit size is too small(less than 20M)
3807 Description: write is stopped by improper -EDQUOT
3808 Details : If the master is busy and qunit size is small enough(let's say 1M),
3809 the slave can not get quota from master on time, which will lead
3810 slave to trigger a -EQUOTA to client.
3815 Description: Deadlock during metadata tests
3816 Details : in prune_dir_dentries(), shrink_dcache_parent() should not be
3817 called with the per-dentry lock held.
3820 Frequency : SLES9 only
3822 Description: Lustre patched kernel for SLES9 SP3 has NR_CPUS set to 8
3823 Details : set CONFIG_NR_CPUS to 128 instead of 8.
3825 Severity : enhancement
3827 Description: remove fs_prep_san_write operation and related patches
3828 Details : remove the ext3-san-jdike patches which are no longer useful.
3833 Description: LDISKFS-fs error (device sdc): ldiskfs_free_blocks
3834 Details : a disk corruption can cause the mballoc code to assert on a
3835 double free or other extent corruptions. Handle these with
3836 ext3_error() instead of assertions.
3839 Frequency : only with mballoc3 code and deep extent trees
3841 Description: ldiskfs_ext_search_right: bad header in inode: unexpected eh_depth
3842 Details : a wrong check of extent headers in ldiskfs_ext_search_right()
3843 can cause the filesystem to be remounted read-only.
3848 Description: lfsck built against 1.4.x cannot run against 1.6.0 lustre
3849 Details : the definition for OBD_IOC_GETNAME changed in 1.6.0. One of the
3850 few external users of this ioctl number is lfsck's call to
3851 llapi_lov_get_uuids() and this caused lfsck to fail at startup.
3852 Add the old ioctl number to the handler so both old and new
3857 Description: parallel lock callbacks
3858 Details : Instead of sending blocking and completion callbacks as separated
3859 requests, adding them to a set and sending in parallel.
3863 Description: Disable most debugging by default
3864 Details : To improve performance, disable most logging (for debug purposes)
3865 by default. VFSTRACE, RPCTRACE, and DLMTRACE are now off by
3866 default, and HA includes fewer messages.
3870 Frequency : only run on patchless client.
3871 Description: use do_facet on sanity.sh for test handling recoverables errors
3872 Details : use do_facet instead of direct use sysctl for set fail_loc on OST
3876 Description: hash tables for lists of nids, connections and uuids
3877 Details : Hash tables noticeably help when a lot of clients connect to a
3878 server, to faster identify duplicate connections or reconnects,
3879 also to faster find export to evict in manual eviction case.
3883 Description: Sometimes, when the server evict a client, and the client will
3884 not be evicted as soon as possible.
3885 Details : In enqueue req, the error was returned by intent, instead of
3886 rq_status which make ptlrpc layer not detect this error, and
3887 does not evict the client. So enqueue error should be returned
3891 Frequency : only at startup
3893 Description: Delay client connections to MDT until fist MDT->OST connect
3894 Details : If a client tried to create a new file before the MDT had
3895 connected to any OSTs, the create would return EIO. Now
3896 the client will simply block until the MDT connects to the
3897 first OST and the create can succeed.
3900 Frequency : only for SLES9
3902 Description: Routinely utilize latest Quadrics drivers in CFS releases
3903 Details : Update patch qsnet-suse-2.6.patch.
3906 Frequency : only for sles10
3908 Description: Update patches for SLES 10 SP1 kernel.
3909 Details : Update the patch vfs_intent-2.6-sles10.patch.
3914 Description: Routinely utilize latest Quadrics drivers in CFS releases
3915 Details : Update patch qsnet-rhel4-2.6.patch.
3918 Frequency : at statup only
3920 Description: mds_lov_synchronize race leads to various problems
3921 Details : simultaneous MDT->OST connections at startup can cause the
3922 sync to abort, leaving the OSC in a bad state.
3927 Description: Using wrong pointer in osc_brw_prep_request
3928 Details : Access to array[-1] can produce panic if kernel compiled with
3929 CONFIG_PAGE_ALLOC enabled
3931 Severity : enhancement
3933 Description: Async OSC create to avoid the blocking unnecessarily.
3934 Details : If a OST has no remain object, system will block on the creating
3935 when need to create a new object on this OST. Now, ways use
3936 pre-created objects when available, instead of blocking on an
3937 empty osc while others are not empty. If we must block, we block
3938 for the shortest possible period of time.
3942 Frequency : only in recovery
3943 Description: Mark OST as early accessible if his start SYNC.
3944 Details : osc_precreate return flag early accessible if oscc marked as
3945 OSCC_FLAG_SYNC_IN_PROGRESS.
3950 Description: Sometimes precreate code can triger create object on wrong ost
3951 Details : Wrong protected or not not restored variables aftre precreate loop
3952 can produce creation object on wrong ost.
3955 Frequency : oss recovery
3957 Description: llog_commit_thread cleanup should sync with llog_commit_thread
3959 Details : llog_commit_thread_count should be synced between llog_commit
3960 start and cleanup, so new llog_commit thread should not be started
3961 when llog_commit threads being stopped to avoid accessing some
3964 Severity : enhancement
3966 Description: Add printing inode info into message about error in writepage.
3968 --------------------------------------------------------------------------------
3970 2007-05-03 Cluster File Systems, Inc. <info@clusterfs.com>
3975 Frequency : on some architectures
3977 Description: 1.6 client sometimes fails to mount from a 1.4 MDT
3978 Details : Uninitialized flags sometimes cause configuration commands to
3982 Frequency : patchless clients only
3984 Description: missing __iget() symbol export
3985 Details : The __iget() symbol export is missing. To avoid the need for
3986 this on patchless clients the deathrow inode reaper is turned
3987 off, and we depend on the VM to clean up old inodes. This
3988 dependency was during via the fix for bug 12181.
3993 Description: sanity.sh fail: test_52b
3994 Details : The ll_inode_to_ext_flags() has a glitch which makes MDS return
3995 incorrect inode's flags to client.
3997 --------------------------------------------------------------------------------
3999 2007-04-19 Cluster File Systems, Inc. <info@clusterfs.com>
4001 * CONFIGURATION CHANGE. This version of Lustre WILL NOT
4002 INTEROPERATE with older versions automatically. In many cases a
4003 special upgrade step is needed. Please read the
4004 user documentation before upgrading any part of a 1.4.x system.
4005 * WARNING: Lustre configuration and startup changes are required with
4006 this release. See https://mail.clusterfs.com/wikis/lustre/MountConf
4008 * Support for kernels:
4009 2.4.21-47.0.1.EL (RHEL 3),
4010 2.6.5-7.283 (SLES 9),
4011 2.6.9-42.0.10.EL (RHEL 4),
4012 2.6.12.6 vanilla (kernel.org),
4013 2.6.16.27-0.9 (SLES10)
4014 * Client support for unpatched kernels:
4015 (see https://mail.clusterfs.com/wikis/lustre/PatchlessClient)
4016 2.6.16 - 2.6.19 vanilla (kernel.org),
4017 2.6.9-42.0.8EL (RHEL 4)
4018 * Recommended e2fsprogs version: 1.39.cfs6
4019 * Note that reiserfs quotas are disabled on SLES 10 in this kernel
4022 Severity : enhancement
4024 Description: Async OSC create to avoid the blocking unnecessarily.
4025 Details : If a OST has no remain object, system will block on the creating
4026 when need to create a new object on this OST. Now, ways use
4027 pre-created objects when available, instead of blocking on an
4028 empty osc while others are not empty. If we must block, we block
4029 for the shortest possible period of time.
4031 Severity : enhancement
4033 Description: MountConf
4034 Details : Lustre configuration is now managed via mkfs and mount
4035 commands instead of lmc and lconf. New obd types (MGS, MGC)
4036 are added for dynamic configuration management. See
4037 https://mail.clusterfs.com/wikis/lustre/MountConf for
4040 Severity : enhancement
4042 Description: dynamic OST addition
4043 Details : OSTs can now be added to a live filesystem
4045 Severity : enhancement
4047 Description: startup order invariance
4048 Details : MDTs and OSTs can be started in any order. Clients only
4049 require the MDT to complete startup.
4051 Severity : enhancement
4053 Description: parallel, asynchronous orphan cleanup
4054 Details : orphan cleanup is now performed in separate threads for each
4055 OST, allowing parallel non-blocking operation.
4057 Severity : enhancement
4059 Description: optimized stripe assignment
4060 Details : stripe assignments are now made based on ost space available,
4061 ost previous usage, and OSS previous usage, in order to try
4062 to optimize storage space and networking resources.
4064 Severity : enhancement
4066 Description: Permanently set tunables
4067 Details : All writable /proc/fs/lustre tunables can now be permanently
4068 set on a per-server basis, at mkfs time or on a live
4071 Severity : enhancement
4073 Description: Lustre message v2
4074 Details : Add lustre message format v2.
4076 Severity : enhancement
4078 Description: client OST exclusion list
4079 Details : Clients can be started with a list of OSTs that should be
4080 declared "inactive" for known non-responsive OSTs.
4084 Description: ENOENT returned for valid filehandle during dbench.
4085 Details : Check if a directory has children when invalidating dentries
4086 associated with an inode during lock cancellation. This fixes
4087 an incorrect ENOENT sometimes seen for valid filehandles during
4088 testing with dbench.
4091 Frequency : SFS test only (otherwise harmless)
4093 Description: SPEC SFS validation failure on NFS v2 over lustre.
4094 Details : Changes the blocksize for regular files to be 2x RPC size,
4095 and not depend on stripe size.
4097 Severity : enhancement
4099 Description: fine-grained SMP locking inside DLM
4100 Details : Improve DLM performance on SMP systems by removing the single
4101 per-namespace lock and replace it with per-resource locks.
4103 Severity : enhancement
4105 Description: don't hold multiple extent locks at one time
4106 Details : To avoid client eviction during large writes, locks are not
4107 held on multiple stripes at one time or for very large writes.
4108 Otherwise, clients can block waiting for a lock on a failed OST
4109 while holding locks on other OSTs and be evicted.
4111 Severity : enhancement
4113 Description: Multiple MD RPCs in flight.
4114 Details : Further unserialise some read-only MDT RPCs - learn about intents.
4115 To avoid overly-overloading MDT, introduce a limit on number of
4116 MDT RPCs in flight for a single client and add /proc controls
4117 to adjust this limit.
4119 Severity : enhancement
4121 Description: client read/write statistics
4122 Details : Add client read/write call usage stats for performance
4123 analysis of user processes.
4124 /proc/fs/lustre/llite/*/offset_stats shows non-sequential
4125 file access. extents_stats shows chunk size distribution.
4126 extents_stats_per_process show chunk size distribution per
4129 Severity : enhancement
4131 Description: per-client statistics on server
4132 Details : Add ldlm and operations statistics for each client in
4133 /proc/fs/lustre/mds|obdfilter/*/exports/
4135 Severity : enhancement
4137 Description: improved MDT statistics
4138 Details : Add detailed MDT operations statistics in
4139 /proc/fs/lustre/mds/*/stats
4141 Severity : enhancement
4143 Description: VFS operations stats
4144 Details : Add client VFS call stats, trackable by pid, ppid, or gid
4145 /proc/fs/lustre/llite/*/stats_track_[pid|ppid|gid]
4150 Description: Fix client-side osc byte counters
4151 Details : The osc read/write byte counters in
4152 /proc/fs/lustre/osc/*/stats are now working
4155 Frequency : always as root on SLES
4157 Description: Failure of copying files with lustre special EAs.
4158 Details : Client side always return success for setxattr call for lustre
4159 special xattr (currently only "trusted.lov").
4164 Description: Refcount LNET uuids
4165 Details : The global LNET uuid list grew linearly with every startup;
4166 refcount repeated list entries instead of always adding to
4169 Severity : enhancement
4171 Description: Dynamic service threads
4172 Details : Within a small range, start extra service threads
4173 automatically when the request queue builds up.
4176 Frequency : mixed-endian client/server environments
4178 Description: mixed-endian crashes
4179 Details : The new msg_v2 system had some failures in mixed-endian
4182 Severity : enhancement
4184 Description: Easy OST removal
4185 Details : OSTs can be permanently deactivated with e.g. 'lctl
4186 conf_param lustre-OST0001.osc.active=0'
4188 Severity : enhancement
4190 Description: MGS proc entries
4191 Details : Added basic proc entries for the MGS showing what filesystems
4194 Severity : enhancement
4196 Description: provide MGS failover
4197 Details : Added config lock reacquisition after MGS server failover.
4199 Severity : enhancement
4201 Description: add Linux 2.4 support
4202 Details : Added support for RHEL 2.4.21 kernel for 1.6 servers and clients
4206 Description: a large application tries to do I/O to the same resource and dies
4207 in the middle of it.
4208 Details : Check the req->rq_arrival time after the call to
4209 ost_brw_lock_get(), but before we do anything about
4210 processing it & sending the BULK transfer request. This
4211 should help move old stale pending locks off the queue as
4212 quickly as obd_timeout.
4215 Frequency : when an incorrect nid is specified during startup
4217 Description: ptlrpc connect to non-existant node causes kernel crash
4218 Details : LNET can't be re-entered from an event callback, which
4219 happened when we expire a message after the export has been
4220 cleaned up. Instead, hand the zombie cleanup off to another
4223 Severity : enhancement
4225 Description: plain/inodebits lock performance improvement
4226 Details : Grouping plain/inodebits in granted list by their request modes
4227 and bits policy, thus improving the performance of search through
4231 Frequency : only if OST filesystem is corrupted
4233 Description: client incorrectly hits assertion in ptlrpc_replay_req()
4234 Details : for a short time RPCs with bulk IO are in the replay list,
4235 but replay of bulk IOs is unimplemented. If the OST filesystem
4236 is corrupted due to disk cache incoherency and then replay is
4237 started it is possible to trip an assertion. Avoid putting
4238 committed RPCs into the replay list at all to avoid this issue.
4241 Frequency : liblustre (e.g. catamount) on a large cluster with >= 8 OSTs/OSS
4243 Description: System hang on startup
4244 Details : This bug allowed the liblustre (e.g. catamount) client to
4245 return to the app before handling all startup RPCs. This
4246 could leave the node unresponsive to lustre network traffic
4247 and manifested as a server ptllnd timeout.
4249 Severity : enhancement
4251 Description: Add "/proc/sys/lustre/debug_peer_on_timeout"
4252 Details : liblustre envirable: LIBLUSTRE_DEBUG_PEER_ON_TIMEOUT
4253 boolean to control whether to print peer debug info when a
4254 client's RPC times out.
4257 Frequency : only for kernels with patches from Lustre below 1.4.3
4259 Description: Remove old rdonly API
4260 Details : Remove old rdonly API which unused from at least lustre 1.4.3
4263 Frequency : only for devices with external journals
4265 Description: Set external device read-only also
4266 Details : During a commanded failover stop, we set the disk device
4267 read-only while the server shuts down. We now also set any
4268 external journal device read-only at the same time.
4271 Frequency : when upgrading from 1.4 while trying to change parameters
4273 Description: The wrong (new) MDC name was used when setting parameters for
4274 upgraded MDT's. Also allows changing of OSC (and MDC)
4275 parameters if --writeconf is specified at tunefs upgrade time.
4278 Frequency : when setting specific ost indicies
4280 Description: QOS code breaks on skipped indicies
4281 Details : Add checks for missing OST indicies in the QOS code, so OSTs
4282 created with --index need not be sequential.
4284 Severity : enhancement
4286 Description: Add uninit_groups feature to ldiskfs2 to speed up e2fsck
4287 Details : The uninit_groups feature works in conjunction with the kernel
4288 filesystem code (ldiskfs2 only) and e2fsprogs-1.39-cfs6 to speed
4289 up the pass1 processing of e2fsck. This is a read-only feature
4290 in ldiskfs2 only, so older kernels and current ldiskfs cannot
4291 mount filesystems that have had this feature enabled.
4293 Severity : enhancement
4295 Description: Improve multi-block allocation algorithm to avoid fragmentation
4296 Details : The mballoc3 code (ldiskfs2 only) adds new mechanisms to improve
4297 allocation locality and avoid filesystem fragmentation.
4299 ------------------------------------------------------------------------------
4301 2007-04-01 Cluster File Systems, Inc. <info@clusterfs.com>
4303 * Support for kernels:
4304 2.4.21-47.0.1.EL (RHEL 3)
4305 2.6.5-7.283 (SLES 9)
4306 2.6.9-42.0.10.EL (RHEL 4)
4307 2.6.12.6 vanilla (kernel.org)
4308 2.6.16.27-0.9 (SLES 10)
4309 * Recommended e2fsprogs version: 1.39.cfs5
4311 * Note that reiserfs quotas are disabled on SLES 10 in this kernel
4315 Frequency : occasional, depends on client load and configuration
4316 Bugzilla : 12181, 12203
4317 Description: data loss for recently-modified files
4319 Details : In some cases it is possible that recently written or created
4320 files may not be written to disk in a timely manner (this should
4321 normally be within 30s unless client IO load is very high).
4322 The problem appears as zero-length files or files that are a
4323 multiple of 1MB in size after a client crash or client eviction
4324 that are missing data at the end of the file.
4326 This problem is more likely to be hit on clients where files are
4327 repeatedly created and unlinked in the same directory, clients
4328 have a large amount of RAM, have many CPUs, the filesystem has
4329 many OSTs, the clients are rebooted frequently, and/or the files
4330 are not accessed by other nodes after being written.
4332 The presence of the problem can be detected by looking at
4333 /proc/sys/fs/inode-state. If the first number (nr_inodes) is
4334 smaller than the second (nr_unused) then dirty files will not
4335 be flushed automatically to disk. "sync; sleep 10" should be
4336 run several times on the node before unmounting it to update
4337 Lustre (this is also safe to run on nodes without this problem).
4339 There is also a related kernel bug in the RHEL4 4 2.6.9 kernel
4340 that can cause this same problem, so customers using that kernel
4341 also need to update the kernel in addition to Lustre. In order
4342 to properly fix this bug, the RHEL3 2.4.21 kernel is also updated.
4344 It is normal that files written just before a client crash (less
4345 than 30s) may not yet have been flushed to disk, even for local
4349 Frequency : frequent on thin XT3 nodes
4351 Description: UUID collision on thin XT3 Linux nodes
4352 Details : UUIDs on Compute Node Linux XT3 nodes were not generated
4353 randomly, since we relied on an insufficiently-seeded PRNG.
4358 Description: OSS hangs after "All ost request buffers busy"
4359 Details : A deadlock between quota and journal operations caused OSS
4360 hangs after printing "All ost request buffers busy."
4363 Frequency : always on liblustre builds
4365 Description: Cleanup compiler warnings on liblustre
4368 Frequency : always on liblustre builds on XT3
4370 Description: LC_CONFIG_CDEBUG don't run while build liblustre on XT3.
4374 Description: Addition of EXT3_FEATURE_RO_COMPAT_DIR_NLINKS flag for
4375 > 32000 subdirectories
4376 Details : Add EXT3_FEATURE_RO_COMPAT_DIR_NLINK flag to
4377 EXT3_FEATURE_RO_COMPAT_SUPP. This flag will be set whenever
4378 subdirectory count crosses 32000. This will aid e2fsck to
4379 correctly handle more than 32000 subdirectories.
4382 Frequency : liblustre (e.g. catamount) on a large cluster with >= 8 OSTs/OSS
4384 Description: System hang on startup
4385 Details : This bug allowed the liblustre (e.g. catamount) client to
4386 return to the app before handling all startup RPCs. This
4387 could leave the node unresponsive to lustre network traffic
4388 and manifested as a server ptllnd timeout.
4390 Severity : enhancement
4392 Description: Add "/proc/sys/lustre/debug_peer_on_timeout"
4393 (liblustre envirable: LIBLUSTRE_DEBUG_PEER_ON_TIMEOUT)
4394 boolean to control whether to print peer debug info when a
4395 client's RPC times out.
4400 Description: make O_SYNC working on 2.6 kernels
4401 Details : 2.6 kernels use different method for mark pages for write,
4402 so need add a code to lustre for O_SYNC work.
4407 Description: Failure to close file and release space on NFS
4408 Details : Put inode details into lock acquired in ll_intent_file_open.
4409 Use mdc_intent_lock in ll_intent_open to properly
4410 detect all kind of errors unhandled by mdc_enqueue.
4415 Description: proc file read during shutdown sometimes raced obd removal,
4417 Details : Add lock to prevent obd access after proc file removal.
4420 Frequency : Only for files larger than 4GB on 32-bit clients.
4422 Description: improperly doing page alignment of locks
4423 Details : Modify lustre core code to use CFS_PAGE_* defines instead of
4424 PAGE_*. Make CFS_PAGE_MASK a 64-bit mask.
4429 Description: RPCs being resent when they shouldn't be
4430 Details : Some RPCs that should not be resent are being resent. This
4431 can cause inconsistencies in the RPC state machine. Do not
4432 resend such requests.
4435 Frequency : rare, only with NFS export
4437 Description: Crash on NFS re-export node
4438 Details : under very unusual load conditions an assertion is hit in
4439 ll_intent_file_open()
4442 Frequency : only if OST filesystem is corrupted
4444 Description: client incorrectly hits assertion in ptlrpc_replay_req()
4445 Details : for a short time RPCs with bulk IO are in the replay list,
4446 but replay of bulk IOs is unimplemented. If the OST filesystem
4447 is corrupted due to disk cache incoherency and then replay is
4448 started it is possible to trip an assertion. Avoid putting
4449 committed RPCs into the replay list at all to avoid this issue.
4454 Description: large O_DIRECT requests fail under memory pressure/fragmentation
4455 Details : Large single O_DIRECT read and write calls can fail to allocate
4456 a sufficiently large buffer to process the request. In case of
4457 allocation failure the allocation is retried with a smaller
4458 buffer and broken into smaller requests.
4460 Severity : enhancement
4462 Description: Add -o localflock option to simulate old noflock behaviour.
4463 Details : This will achieve local-only flock/fcntl locks coherentness.
4468 Description: versioning check is incomplete
4469 Details : Checking the version difference of client vs. server, report
4470 error if the gap is too big.
4475 Description: add support PG_writeback bit
4476 Details : add support for PG_writeback bit for Lustre, for more carefull
4477 work with page cache in 2.6 kernel. This also fix some deadlocks
4478 and remove hack for work O_SYNC with 2.6 kernel.
4480 Severity : enhancement
4482 Description: Add uninit_groups feature to ldiskfs2 to speed up e2fsck
4483 Details : The uninit_groups feature works in conjunction with the kernel
4484 filesystem code (ldiskfs2 only) and e2fsprogs-1.39-cfs6 to speed
4485 up the pass1 processing of e2fsck. This is a read-only feature
4486 in ldiskfs2 only, so older kernels and current ldiskfs cannot
4487 mount filesystems that have had this feature enabled.
4489 Severity : enhancement
4491 Description: Improve multi-block allocation algorithm to avoid fragmentation
4492 Details : The mballoc3 code (ldiskfs2 only) adds new mechanisms to improve
4493 allocation locality and avoid filesystem fragmentation.
4495 ------------------------------------------------------------------------------
4497 2007-02-09 Cluster File Systems, Inc. <info@clusterfs.com>
4499 * Support for kernels:
4500 2.6.9-42.0.3.EL (RHEL 4)
4501 2.6.5-7.276 (SLES 9)
4502 2.4.21-47.0.1.EL (RHEL 3)
4503 2.6.12.6 vanilla (kernel.org)
4504 2.6.16.21-0.8 (SLES10)
4505 * Recommended e2fsprogs version: 1.39.cfs2-0
4507 * The backwards-compatible /proc/sys/portals symlink has been removed
4508 in this release. Before upgrading, please ensure that you change
4509 any configuration scripts or /etc/sysctl.conf files that access
4510 /proc/sys/portals/* or sysctl portals.* to use the corresponding
4511 entry in /proc/sys/lnet or sysctl lnet.*. This change can be made
4512 in advance of the upgrade on any system running Lustre 1.4.6 or
4513 newer, since /proc/sys/lnet was added in that version.
4514 * Note that reiserfs quotas are disabled on SLES 10 in this kernel
4518 Frequency : only when quota is used
4520 Description: avoid scanning export list for quota master
4521 Details : Change the algorithms to avoid scanning export list in order
4522 to improve the efficiency.
4525 Frequency : MDS failover only, very rarely
4527 Description: "went back in time" messages on mds failover
4528 Details : The greatest transno may be lost when the current operation
4529 finishes with an error (transno==0) and the client's last_rcvd
4530 record is over-written. Save the greatest transno in the
4531 mds_last_transno for this case.
4534 Frequency : always for specific kernels and striping counts
4536 Description: client may get "Matching packet too big" without ACL support
4537 Details : Clients compiled without CONFIG_FS_POSIX_ACL get an error message
4538 when trying to access files in certain configurations. The
4539 clients should in fact be denied when mounting because they do
4540 not understand ACLs.
4543 Frequency : Cray XT3 with more than 4000 clients and multiple jobs
4545 Description: many clients connecting with IO in progress causes connect timeouts
4546 Details : Avoid synchronous journal commits to avoid delays caused by many
4547 clients connecting/disconnecting when bulk IO is in progress.
4548 Queue liblustre connect requests on OST_REQUEST_PORTAL instead of
4549 OST_IO_PORTAL to avoid delays behind potentially many pending
4553 Frequency : occasionally with multiple writers to a single file
4555 Description: shared writes to file may result in wrong size reported by stat()
4556 Details : Allow growing of kms when extent lock is cancelled
4559 Frequency : always with random mmap IO to multi-striped file
4561 Description: mmap write might be lost if we are writing to a 'hole' in stripe
4562 Details : Only if the hole is at the end of OST object so that kms is too
4563 small. Fix is to increase kms accordingly in ll_nopage.
4566 Frequency : rare, only if OST filesystem is inconsistent with MDS filesystem
4568 Description: writes to a missing object would leak memory on the OST
4569 Details : If there is an inconsistency between the MDS and OST filesystems,
4570 such that the MDS references an object that doesn't exist, writes
4571 to that object will leak memory due to incorrect cleanup in the
4572 error handling path, eventually running out of memory on the OST.
4577 Description: Creating too long symlink causes lustre errors
4578 Details : Check symlink and name lengths before sending requests to MDS.
4581 Frequency : only if flock is enabled (not on by default)
4583 Description: posix locks not released on fd closure on 2.6.9+
4584 Details : We failed to add posix locks to list of inode locks on 2.6.9+
4585 kernels, this caused such locks not to be released on fd close and
4586 then assertions on fs unmount about still used locks.
4589 Frequency : MDS failover only, very rarely
4591 Description: clients may get ASSERTION(granted_lock != NULL)
4592 Details : When request was taking a long time, and a client was resending
4593 a getattr by name lock request. The were multiple lock requests
4594 with the same client lock handle and
4595 mds_getattr_name->fixup_handle_for_resent_request found one of the
4596 lock handles but later failed with ASSERTION(granted_lock != NULL).
4601 Description: handle->h_buffer_credits > 0, assertion failure
4602 Details : h_buffer_credits is zero after truncate, causing assertion
4603 failure. This patch extends the transaction or creates a new
4607 Frequency : NFS re-export or patchless client
4608 Bugzilla : 11179, 10796
4609 Description: Crash on NFS re-export node (__d_move)
4610 Details : We do not want to hash the dentry if we don't have a lock.
4611 But if this dentry is later used in d_move, we'd hit uninitialised
4612 list head d_hash, so we just do this to init d_hash field but
4613 leave dentry unhashed.
4616 Frequency : NFS re-export or patchless client
4618 Description: NFS exports has problem with symbolic link
4619 Details : lustre client didn't properly install dentry when re-exported
4620 to NFS or running patchless client.
4623 Frequency : NFS re-export or patchless client
4625 Description: Various nfs/patchless fixes.
4626 Details : fixes reuse disconected alias for lookup process - this fixes
4627 warning "find_exported_dentry: npd != pd",
4628 fix permission error with open files at nfs.
4629 fix apply umask when do revalidate.
4632 Frequency : occasional
4634 Description: Crash on NFS re-export node
4635 Details : calling clear_page() on the wrong pointer triggered oops in
4636 generic_mapping_read().
4639 Frequency : rarely, using O_DIRECT IO
4641 Description: unaligned directio crashes client with LASSERT
4642 Details : check for unaligned buffers before trying any requests.
4645 Frequency : rarely, using CFS RAID5 patches in non-standard kernel series
4647 Description: stale data returned from RAID cache
4648 Details : If only a small amount of IO is done to the RAID device before
4649 reading it again it is possible to get stale data from the RAID
4650 cache instead of reading it from disk.
4653 Frequency : always for sles10 kernel
4655 Description: sles10 support
4656 Details : ll_follow_link: compile fixes and using of nd_set_link
4657 under newer kernels.
4660 Frequency : depends on arch, kernel and compiler version, always on sles10
4663 Description: recursive or deep enough symlinks cause stack overflow
4664 Details : getting rid of large stack-allocated variable in
4668 Frequency : depends on hardware
4670 Description: lustre write performance loss in the SLES10 kernel
4671 Details : the performance loss is caused by using of write barriers in the
4672 ext3 code. The SLES10 kernel turns barrier support on by
4673 default. The fix is to undo that change for ldiskfs.
4675 ------------------------------------------------------------------------------
4677 2006-12-09 Cluster File Systems, Inc. <info@clusterfs.com>
4679 * Support for kernels:
4680 2.6.9-42.0.3EL (RHEL 4)
4681 2.6.5-7.276 (SLES 9)
4682 2.4.21-47.0.1.EL (RHEL 3)
4683 2.6.12.6 vanilla (kernel.org)
4687 Frequency : quota enabled and large files being deleted
4689 Description: releasing more than 4GB of quota at once hangs OST
4690 Details : If a user deletes more than 4GB of files on a single OST it
4691 will cause the OST to spin in an infinite loop. Release
4692 quota in < 4GB chunks, or use a 64-bit value for 1.4.7.1+.
4697 Description: statfs data retrieved from /proc may be stale or zero
4698 Details : When reading per-device statfs data from /proc, in the
4699 {kbytes,files}_{total,free,avail} files, it may appear
4700 as zero or be out of date.
4703 Frequency : systems with MD RAID1 external journal devices
4705 Description: lconf's call to blkid is confused by RAID1 journal devices
4706 Details : Use the "blkid -l" flag to locate the MD RAID device instead
4707 of returning all block devices that match the journal UUID.
4710 Frequency : always, for aggregate stripe size over 4GB
4712 Description: "lfs setstripe" fails assertion when setting 4GB+ stripe width
4713 Details : Using "lfs setstripe" to set stripe size * stripe count over 4GB
4714 will fail the kernel with "ASSERTION(lsm->lsm_xfersize != 0)"
4717 Frequency : always if "lfs find" used on a local file/directory
4719 Description: "lfs find" segfaults if used on a local file/directory
4720 Details : The case where a directory component was not specified wasn't
4721 handled correctly. Handle this properly.
4724 Frequency : always on ppc64
4726 Description: the write to an ext3 filesystem mounted with mballoc got stuck
4727 Details : ext3_mb_generate_buddy() uses find_next_bit() which does not
4728 perform endianness conversion.
4731 Frequency : rarely (truncate to non-zero file size after write under load)
4732 Bugzilla : 10730, 10687
4733 Description: Files padded with zeros to next 4K multiple
4734 Details : With filesystems mounted using the "extents" option (2.6 kernels)
4735 it is possible that files that are truncated to a non-zero size
4736 immediately after being written are filled with zero bytes beyond
4737 the truncated size. No file data is lost.
4739 Severity : enhancement
4741 Description: Allow recovery/failover for liblustre clients.
4742 Details : liblustre clients were unaware of failover configurations until
4745 Severity : enhancement
4747 Description: user file locks should fail when not mounting with flock option
4748 Details : Set up an error-returning stub in ll_file_operations.lock field
4749 to prevent incorrect behaviour when client is mounted without
4750 flock option. Also, set up properly f_op->flock field for
4754 Frequency : always on ia64
4756 Description: "lfs df" loops on printing out MDS statfs information
4757 Details : The obd_ioctl_data was not initialized and in some systems
4758 this caused a failure during the ioctl that did not return
4759 an error. Initialize the struct and return an error on failure.
4762 Frequency : SLES 9 only
4764 Description: Error of copying files with lustre special EAs as root
4765 Details : Client side always return success for setxattr call for lustre
4766 special xattr (currently only "trusted.lov").
4769 Frequency : rarely on clusters with both ia64+i386 clients
4771 Description: ia64+i686 clients doing shared IO on the same file may LBUG
4772 Details : In rare cases when both ia64+i686 (or other mixed-PAGE_SIZE)
4773 clients are doing concurrent writes to the same file it is
4774 possible that the ia64 clients may LASSERT because the OST
4775 extent locks are not PAGE_SIZE aligned. Ensure that grown
4776 locks are always aligned on the request boundary.
4779 Frequency : specific use, occasional
4781 Description: Overwriting in use executable truncates on-disk binary image
4782 Details : If one node attempts to overwrite an executable in use by
4783 another node, we now correctly return ETXTBSY instead of
4784 truncating the file.
4786 Severity : enhancement
4788 Description: Async OSC create to avoid the blocking unnecessarily.
4789 Details : If a OST has no remain object, system will block on the creating
4790 when need to create a new object on this OST. Now, ways use
4791 pre-created objects when available, instead of blocking on an
4792 empty osc while others are not empty. If we must block, we block
4793 for the shortest possible period of time.
4798 Description: chmod on Lustre root is propagated to other clients
4799 Details : Re-validate root's dentry in ll_lookup_it to avoid having it
4800 invalid by the follow_mount time.
4803 Frequency : liblustre clients only
4805 Description: Race in 'instant cancel' lock handling could lead to such locks
4806 never to be granted in case of SMP MDS
4807 Details : Do not destroy not yet granted but cbpending locks in
4811 Frequency : replay/resend of open
4813 Description: non null lock assetion failure in mds_intent_policy
4814 Details : Trying to replay/resend lockless open requests resulted in
4815 mds_open() returning 0 with no lock. Now it sets a flag if
4816 a lock is going to be returned.
4818 Severity : enhancement
4820 Description: Checksum enhancements
4821 Details : New checksum enhancements allow for resending RPCs that failed
4824 Severity : enhancement
4826 Description: Tunables on number of dirty pages in cacche
4827 Details : Allow to set limit on number of dirty pages cached.
4832 Description: client crash on unmount - lock still has references
4833 Details : In some error handling cases it was possible to leak a lock
4834 reference on a client while accessing a file. This was not
4835 harmful to the client during operation, but would cause the
4836 client to crash when the filesystem is unmounted.
4839 Frequency : specific case, rare
4841 Description: ETXTBSY on mds though file not in use by client
4842 Details : ETXTBSY is no longer incorrectly returned when attempting to
4843 chmod or chown a directory that the user previously tried to
4844 execute or a currently-executing binary.
4847 Frequency : extremely rare except on liblustre-based clients
4849 Description: Lustre space not freed when files are deleted
4850 Details : Clean up open-unlinked files after client eviction. Previously
4851 the unlink was skipped and the files remained as orphans.
4856 Description: OST failure "would be an LBUG" in waiting_locks_callback()
4857 Details : In some cases it was possible to send a blocking callback to
4858 a client doing a glimpse, even though that client didn't get
4859 a lock granted. When the glimpse lock is cancelled on the OST
4860 the freed lock is left on the waiting list and corrupted the list.
4863 Frequency : all core dumps
4865 Description: Broke core dumps to lustre
4866 Details : Negative dentry may be unhashed if parent does not have UPDATE
4867 lock, but some callers, e.g. do_coredump, expect dentry to be
4868 hashed after successful create, hash it in ll_create_it.
4870 ------------------------------------------------------------------------------
4872 2006-09-13 Cluster File Systems, Inc. <info@clusterfs.com>
4874 * Support for kernels:
4875 2.6.9-42.0.2.EL (RHEL 4)
4876 2.6.5-7.276 (SLES 9)
4877 2.4.21-40.EL (RHEL 3)
4878 2.6.12.6 vanilla (kernel.org)
4882 Frequency : always on RHEL 3
4884 Description: Number of open files grows over time
4885 Details : The number of open files grows over time, whether or not
4886 Lustre is started. This was due to a filp leak introduced
4887 by one of our kernel patches.
4889 ------------------------------------------------------------------------------
4891 2006-08-20 Cluster File Systems, Inc. <info@clusterfs.com>
4893 * Support for kernels:
4894 2.6.9-42.EL (RHEL 4)
4895 2.6.5-7.267 (SLES 9)
4896 2.4.21-40.EL (RHEL 3)
4897 2.6.12.6 vanilla (kernel.org)
4902 Bugzilla : 5719, 9635, 9792, 9684
4903 Description: OST (or MDS) trips assertions in (re)connection under heavy load
4904 Details : If a server is under heavy load and cannot reply to new
4905 connection requests before the client resends the (re)connect,
4906 the connection handling code can behave badly if two service
4907 threads are concurrently handing separate (re)connections from
4908 the same client. Add better locking to the connection handling
4909 code, and ensure that only a single connection will be processed
4910 for a given client UUID, even if the lock is dropped.
4912 Severity : enhancement
4914 Description: add TCP zero-copy support to kernel
4915 Details : Add support to the kernel TCP stack to allow zero-copy bulk
4916 sends if the hardware supports scatter-gather and checksumming.
4917 This allows socklnd to do client-write and server-read more
4918 efficiently and reduce CPU utilization from skbuf copying.
4921 Frequency : only if NFS exporting from client
4923 Description: NULL pointer deref in ll_iocontrol() if chattr mknod file
4924 Details : If setting attributes on a file created under NFS that had
4925 never been opened it would be possible to oops the client
4926 if the file had no objects.
4930 Bugzilla : 9326, 10402, 10897
4931 Description: client crash in ptlrpcd_wake() thread when sending async RPC
4932 Details : It is possible that ptlrpcd_wake() dereferences a freed async
4933 RPC. In rare cases the ptlrpcd thread alread processed the RPC
4934 before ptlrpcd_wake() was called and the request was freed.
4937 Frequency : always for liblustre
4939 Description: liblustre client does MDS+OSTs setattr RPC for each write
4940 Details : When doing a write from a liblustre client, the client
4941 incorrectly issued an RPC to the MDS and each OST the file was
4942 striped over in order to update the timestamps. When writing
4943 with small chunks and many clients this could overwhelm the MDS
4944 with RPCs. In all cases it would slow down the write because
4945 these RPCs are unnecessary.
4947 Severity : enhancement
4949 Description: allow number of MDS service threads to be changed at module load
4950 Details : It is now possible to change the number of MDS service threads
4951 running. Adding "options mds mds_num_threads={N}" to the MDS's
4952 /etc/modprobe.conf will set the number of threads for the next
4953 time Lustre is restarted (assuming the "mds" module is also
4954 reloaded at that time). The default number of threads will
4955 stay the same, 32 for most systems.
4960 Description: OST crash if filesystem is unformatted or corrupt
4961 Details : If an OST is started on a device that has never been formatted
4962 or if the filesystem is corrupt and cannot even mount then the
4963 error handling cleanup routines would dereference a NULL pointer.
4968 Description: NULL pointer deref in llap_from_page.
4969 Details : get_cache_page_nowait can return a page with NULL (or otherwise
4970 incorrect) mapping if the page was truncated/reclaimed while it was
4971 searched for. Check for this condition and skip such pages when
4972 doing readahead. Introduce extra check to llap_from_page() to
4973 verify page->mapping->host is non-NULL (so page is not anonymous).
4976 Frequency : Sometimes when using sys_sendfile
4978 Description: "page not covered by a lock" warnings from ll_readpage
4979 Details : sendfile called ll_readpage without right page locks present.
4980 Now we introduced ll_file_sendfile that does necessary locking
4981 around call to generic_file_sendfile() much like we do in
4985 Frequency : with certain MDS communication failures at client mount time
4987 Description: NULL pointer deref after failed client mount
4988 Details : a client connection request may delayed by the network layer
4989 and not be sent until after the PTLRPC layer has timed out the
4990 request. If the client fails the mount immediately it will try
4991 to clean up before the network times out the request. Add a
4992 reference from the request import to the obd device and delay
4993 the cleanup until the network drops the request.
4996 Frequency : occasionally during client (re)connect
4998 Description: assertion failure during client (re)connect
4999 Details : processing a client connection request may be delayed by the
5000 client or server longer than the client connect timeout. This
5001 causes the client to resend the connection request. If the
5002 original connection request is replied in this interval, the
5003 client may trip an assertion failure in ptlrpc_connect_interpret()
5004 which thought it would be the only running connect process.
5007 Frequency : only with obd_echo servers and clients that are rebooted
5009 Description: kernel BUG accessing uninitialized data structure
5010 Details : When running an obd_echo server it did not start the ping_evictor
5011 thread, and when a client was evicted an uninitialized data
5012 structure was accessed. Start the ping_evictor in the RPC
5013 service startup instead of the OBD startup.
5015 Severity : enhancement
5016 Bugzilla : 10193 (patchless)
5017 Description: Remove dependency on various unexported kernel interfaces.
5018 Details : No longer need reparent_to_init, exit_mm, exit_files,
5019 sock_getsockopt, filemap_populate, FMODE_EXEC, put_filp.
5022 Frequency : rare (only users of deprecated and unsupported LDAP config)
5024 Description: write_conf for zeroconf mount queried LDAP incorrectly for client
5025 Details : LDAP apparently contains 'lustreName' attributes instead of
5026 'name'. A simple remapping of the name is sufficient.
5029 Frequency : rare (only with non-default dump_on_timeout debug enabled)
5031 Description: waiting_locks_callback trips kernel BUG if client is evicted
5032 Details : Running with the dump_on_timeout debug flag turned on makes
5033 it possible that the waiting_locks_callback() can try to dump
5034 the Lustre kernel debug logs from an interrupt handler. Defer
5035 this log dumping to the expired_lock_main() thread.
5037 Severity : enhancement
5039 Description: Support NFS exporting on 2.6 kernels.
5040 Details : Implement non-rawops metadata methods for NFS server to use without
5041 changing NFS server code.
5044 Frequency : very rare (synthetic metadata workload only)
5046 Description: two racing renames might cause an MDS thread to deadlock
5047 Details : Running the "racer" program may cause one MDS thread to rename
5048 a file from being the source of a rename to being the target of
5049 a rename at exactly the same time that another thread is doing
5050 so, and the second thread has already enqueued these locks after
5051 doing a lookup of the target and is trying to relock them in
5052 order. Ensure that we don't try to re-lock the same resource.
5055 Frequency : only very large systems with liblustre clients
5057 Description: slow eviction of liblustre clients with the "evict_by_nid" RPC
5058 Details : Use asynchronous set_info RPCs to send the "evict_by_nid" to
5059 all OSTs in parallel. This allows the eviction of stale liblustre
5060 clients to proceed much faster than if they were done in series,
5061 and also offers similar improvements for other set_info RPCs.
5066 Description: excessive CPU usage during initial read phase on client
5067 Details : During the initial read phase on a client, it would agressively
5068 retry readahead on the file, consuming too much CPU and impacting
5069 performance (since 1.4.5.8). Improve the readahead algorithm
5070 to avoid this, and also improve some other common cases (read
5071 of small files in particular, where "small" is files smaller than
5072 /proc/fs/lustre/llite/*/max_read_ahead_whole_mb, 2MB by default).
5077 Description: MDS crash when receiving packet with unknown intent.
5078 Details : Do not LBUG in unknown intent case, just return -EFAULT
5080 Severity : enhancement
5081 Bugzilla : 9293, 9385
5082 Description: MDS RPCs are serialised on client. This is unnecessary for some.
5083 Details : Do not serialize getattr (non-intent version) and statfs.
5086 Frequency : occasional, when OST network is overloaded/intermittent
5088 Description: client evicted by OST after bulk IO timeout
5089 Details : If a client sends a bulk IO request (read or write) the OST
5090 may evict the client if it is unresposive to its data GET/PUT
5091 request. This is incorrect if the network is overloaded (takes
5092 too long to transfer the RPC data) or dropped the OST GET/PUT
5093 request. There is no need to evict the client at all, since
5094 the pinger and/or lock callbacks will handle this, and the
5095 client can restart the bulk request.
5098 Frequency : Always when mmapping file with no objects
5100 Description: client crashes when mmapping file with no objects
5101 Details : Check that we actually have objects in a file before doing any
5102 operations on objects in ll_vm_open, ll_vm_close and
5108 Description: Request leak when working with deleted CWD
5109 Details : Introduce advanced request refcount tracking for requests
5110 referenced from lustre intent.
5112 Severity : Enhancement
5114 Description: Cache open file handles on client.
5115 Details : MDS now will return special lock along with openhandle, if
5116 requested and client is allowed to hold openhandle, even if unused,
5117 until such a lock is revoked. Helps NFS a lot, since NFS is opening
5118 closing files for every read/write openration.
5120 Severity : Enhancement
5122 Description: Cache open negative dentries on client when possible.
5123 Details : Guard negative dentries with UPDATE lock on parent dir, drop
5124 negative dentries on lock revocation.
5129 Description: Remounting a client read-only wasn't possible with a zconf mount
5130 Details : It wasn't possible to remount a client read-only with llmount.
5132 Severity : enhancement
5133 Description: Include MPICH 1.2.6 Lustre ADIO interface patch
5134 Details : In lustre/contrib/ or /usr/share/lustre in RPM a patch for
5135 MPICH is included to add Lustre-specific ADIO interfaces.
5136 This is based closely on the UFS ADIO layer and only differs
5137 in file creation, in order to allow the OST striping to be set.
5138 This is user-contributed code and not supported by CFS.
5143 Description: extended inode attributes (immutable, append-only) work improperly
5144 when 2.4 and 2.6 kernels are used on client/server or vice versa
5145 Details : Introduce kernel-independent values for these flags.
5147 Severity : enhancement
5150 Description: Allow fractional MB tunings for lustre in /proc/ filesystem.
5151 Details : Many of the /proc/ tunables can only be tuned at a megabyte
5152 granularity. Now, Fractional MB granularity is be supported,
5153 this is very useful for low memory system.
5155 Severity : enhancement
5157 Description: Getattr by fid
5158 Details : Getting a file attributes by its fid, obtaining UPDATE|LOOKUP
5159 locks, avoids extra getattr rpc requests to MDS, allows '/' to
5160 have locks and avoids getattr rpc requests for it on every stat.
5163 Frequency : Always, for filesystems larger than 2TB
5165 Description: ldiskfs crash at mount for filesystem larger than 2TB with mballoc
5166 Details : Kenrel kmalloc limits allocations to 128kB and this prevents
5167 filesystems larger than 2TB to be mounted with mballoc enabled.
5170 Frequency : Always, for 32-bit kernel without CONFIG_LBD and filesystem > 2TB
5172 Description: filesystem corruption for non-standard kernels and very large OSTs
5173 Details : If a 32-bit kernel is compiled without CONFIG_LBD enabled and a
5174 filesystems larger than 2TB is mounted then the kernel will
5175 silently corrupt the start of the filesystem. CONFIG_LBD is
5176 enabled for all CFS-supported kernels, but the possibility of
5177 this happening with a modified kernel config exists.
5179 Severity : enhancement
5181 Description: add client O_DIRECT support for 2.6 kernels
5182 Details : It is now possible to do O_DIRECT reads and writes to files
5183 in the Lustre client mountpoint on 2.6 kernel clients.
5185 Severity : enhancement
5187 Description: parallel glimpse, setattr, statfs, punch, destroy requests
5188 Details : Sends glimpse, setattr, statfs, punch, destroy requests to OSTs in
5189 parallel, not waiting for response from every OST before sending
5190 a rpc to the next OST.
5195 Description: setattr vs write race when updating file timestamps
5196 Details : Client processes that update a file timestamp into the past
5197 right after writing to the file (e.g. tar) it is possible that
5198 the updated file modification time can be reset to the current
5199 time due to a race between processing the setattr and write RPC.
5201 Severity : enhancement
5203 Description: Bring 'lfs find' closer in line with regular Linux find.
5204 Details : lfs find util supports -atime, -mtime, -ctime, -maxdepth, -print,
5205 -print0 options and obtains all the needed info through the lustre
5208 Severity : enhancement
5210 Description: support up to 1024 configured devices on one node
5211 Details : change obd_dev array from statically allocated to dynamically
5212 allocated structs as they are first used to reduce memory usage
5217 Description: Flush dirty partially truncated pages during truncate
5218 Details : Immediatelly flush partially truncated pages in filter_setattr,
5219 this way we completely avoid having any pages in page cache on OST
5220 and can retire ugly workarounds during writes to flush such pages.
5225 Description: i_sem vs transaction deadlock in mds_obd_destroy during unlink.
5226 Details : protect inode from truncation within vfs_unlink() context
5227 just take a reference before calling vfs_unlink() and release it
5228 when parent's i_sem is free.
5231 Frequency : always, if extents are used on OSTs
5233 Description: index ei_leaf_hi (48-bit extension) is not zeroed in extent index
5234 Details : OSTs using the extents format would not zero the high 16 bits of
5235 the index physical block number. This is not a problem for any
5236 OST filesystems smaller than 16TB, and no kernels support ext3
5237 filesystems larger than 16TB yet. This is fixed in 1.4.7 (all
5238 new/modified files) and can be fixed for existing filesystems
5239 with e2fsprogs-1.39-cfs1.
5244 Description: import connection selection may be incorrect if timer wraps
5245 Details : Using a 32-bit jiffies timer with HZ=1000 may cause backup
5246 import connections to be ignored if the 32-bit jiffies counter
5247 wraps. Use a 64-bit jiffies counter.
5250 Frequency : during server recovery
5252 Description: crash after server is denying duplicate export
5253 Details : If clients are resending connect requests to the server, the
5254 server refuses to allow a client to connect multiple times.
5255 Fixed a bug in the handling of this case.
5258 Frequency : very large clusters immediately after boot
5260 Description: LNET request buffers exhausted under heavy short-term load
5261 Details : If a large number of client requests are generated on a service
5262 that has previously never seen so many requests it is possible
5263 that the request buffer growth cannot keep up with the spike in
5264 demand. Instead of dropping incoming requests, they are held in
5265 the LND until the RPC service can accept more requests.
5268 Frequency : Sometimes during replay
5270 Description: Assertion failure in ll_local_open after replay.
5271 Details : If replay happened on an open request reply before we were able
5272 to set replay handler, reply will become not swabbed tripping the
5273 assertion in ll_local_open. Now we set the handler right after
5274 recognising of open request
5277 Frequency : very rare
5279 Description: kernel reports "badness in vsnprintf"
5280 Details : Reading from the "recovery_status" /proc file in small chunks
5281 may cause a negative length in lprocfs_obd_rd_recovery_status()
5282 call to vsnprintf() (which is otherwise harmless). Exit early
5283 if there is no more space in the output buffer.
5285 Severity : enhancement
5287 Description: clear OBD RPC statistics by writing to them
5288 Details : It is now possible to clear the OBD RPC statistics by writing
5289 to the "stats" file.
5294 Description: Client mtime is not the same on different clients after utimes
5295 Details : In some cases, the client was using the utimes() syscall on
5296 a file cached on another node. The clients now validate the
5297 ctime from the MDS + OSTs to determine which one is right.
5302 Description: Inability to activate failout mode
5303 Details : lconf script incorrectly assumed that in python string's numeric
5304 value is used in comparisons.
5307 Frequency : always with multiple stripes per file
5309 Description: Inefficient object allocation for mutli-stripe files
5310 Details : When selecting which OSTs to stripe files over, for files with
5311 a stripe count that divides evenly into the number of OSTs,
5312 the MDS is always picking the same starting OST for each file.
5313 Return the OST selection heuristic to the original design.
5318 Description: mount failures may take full timeout to return an error
5319 Details : Under some heavy load conditions it is possible that a
5320 failed mount can wait for the full obd_timeout interval,
5321 possibly several minutes, before reporting an error.
5322 Instead return an error as soon as the status is known.
5324 ------------------------------------------------------------------------------
5326 2006-02-14 Cluster File Systems, Inc. <info@clusterfs.com>
5328 * WIRE PROTOCOL CHANGE. This version of Lustre networking WILL NOT
5329 INTEROPERATE with older versions automatically. Please read the
5330 user documentation before upgrading any part of a live system.
5331 * WARNING: Lustre networking configuration changes are required with
5332 this release. See https://bugzilla.clusterfs.com/show_bug.cgi?id=10052
5335 * Support for kernels:
5336 2.6.9-22.0.2.EL (RHEL 4)
5337 2.6.5-7.244 (SLES 9)
5338 2.6.12.6 vanilla (kernel.org)
5341 Severity : enhancement
5342 Bugzilla : 7981/8208
5343 Description: Introduced Lustre Networking (LNET)
5344 Details : LNET is new networking infrastructure for Lustre, it includes
5345 a reorganized network configuration mode (see the user
5346 documentation for full details) as well as support for routing
5347 between different network fabrics. Lustre Networking Devices
5348 (LNDs) for the supported network fabrics have also been
5349 created for this new infrastructure.
5351 Severity : enhancement
5352 Description: Introduced Access control lists
5353 Details : clients can set ACLs on files and directories in order to have
5354 more fine-grained permissions than the standard Unix UGO+RWX.
5355 The MDS must be started with the "-o acl" mount option.
5357 Severity : enhancement
5358 Description: Introduced filesystem quotas
5359 Details : Administrators may now establish per-user quotas on the
5362 Severity : enhancement
5364 Description: Configuration change for the XT3
5365 The PTLLND is now used to run Lustre over Portals on the XT3
5366 The configure option(s) --with-cray-portals are no longer used.
5367 Rather --with-portals=<path-to-portals-includes> is used to
5368 enable building on the XT3. In addition to enable XT3 specific
5369 features the option --enable-cray-xt3 must be used.
5374 Description: Running on many-way SMP OSTs can trigger oops in llcd_send()
5375 Details : A race between allocating a new llcd and re-getting the llcd_lock
5376 allowed another thread to grab newly-allocated llcd.
5378 Severity : enhancement
5380 Description: 2.6 OST async journal commit and locking fix to improve performance
5381 Details : The filter_direct_io()+filter_commitrw_write() journal commits for
5382 2.6 kernels are now async as they already were in 2.4 kernels so
5383 that they can commit concurrently with the network bulk transfer.
5384 For block-allocated files the filter allocation semaphore is held
5385 to avoid filesystem fragmentation during allocation. BKL lock
5386 removed for 2.6 xattr operations where it is no longer needed.
5391 Description: lconf incorrectly determined whether two IP networks could talk
5392 Details : In some more complicated routing and multiple-network
5393 configurations, lconf will avoid trying to make a network
5394 connection to a disjoint part of the IP space. It was doing the
5395 math incorrectly for one set of cases.
5400 Description: Fix for potential infinite loop processing records in an llog.
5401 Details : If an llog record is corrupted/zeroed, it is possible to loop
5402 forever in llog_process(). Validate the llog record length
5403 and skip the remainder of the block on error.
5406 Frequency : occasional (liblustre only)
5408 Description: liblustre could not open files whose last component is a symlink
5409 Details : sysio_path_walk() would incorrectly pass the open intent to
5410 intermediate path components.
5413 Frequency : rare (liblustre only with non-standard tuning)
5414 Bugzilla : 7201 (7350)
5415 Description: Tuning the MDC DLM LRU size to zero triggers client LASSERT
5416 Details : llu_lookup_finish_locks() tries to set lock data on a lock
5417 after it has been released, only do this for referenced locks
5419 Severity : enhancement
5421 Description: specifying an (invalid) directory default stripe_size of -1
5422 would reset the directory default striping
5423 Details : stripe_size -1 was used internally to signal directory stripe
5424 removal, now use "all default" to signal dir stripe removal
5425 as a directory striping of "all default" is not useful
5428 Frequency : common for large clusters running liblustre clients
5430 Description: doing an ls when liblustre clients are running is slow
5431 Details : sending a glimpse AST to a liblustre client waits for every AST
5432 to time out, as liblustre clients will not respond. Since they
5433 cannot cache data we refresh the OST lock LVB from disk instead.
5435 Severity : enhancement
5437 Description: doing an ls at the same time as file IO can be slow
5438 Details : enqueue and other "small" requests can be blocked behind many
5439 large IO requests. Create a new OST IO portal for non-IO
5440 requests so they can be processed faster.
5443 Frequency : rare (only HPUX clients mounting unsupported re-exported NFS vol)
5445 Description: an HPUX NFS client would get -EACCESS when ftruncate()ing a newly
5446 created file with mode 000
5447 Details : the Linux NFS server relies on an MDS_OPEN_OWNEROVERRIDE hack to
5448 allow an ftruncate() as a non-root user to a file with mode 000.
5449 Lustre now respects this flag to disable mode checks when
5450 truncating a file owned by the user
5453 Frequency : liblustre-only, when liblustre client dies unexpectedly or becomes
5456 Description: Revoking locks from clients that went dead or catatonic might take
5458 Details : New lock flags FL_CANCEL_ON_BLOCK used by liblustre makes
5459 cancellation of such locks instant on servers without waiting for
5460 any reply from clients. Clients drops these locks when cancel
5461 notification from server is received without replying.
5464 Frequency : liblustre-only, when liblustre client dies or becomes busy
5466 Description: Doing ls on Linux clients can take a long time with active
5468 Details : Liblustre client cannot handle ASTs in timely manner, so avoid
5469 granting such locks to it in the first place if possible. Locks
5470 are taken by proxy on the OST during the read or write and
5471 dropped immediately afterward. Add connect flags handling, do
5472 not grant locks to liblustre clients for glimpse ASTs.
5474 Severity : enhancement
5476 Description: Improve read-ahead algorithm to avoid excessive IO for random reads
5477 Details : Existing read-ahead algorithm is tuned for the case of streamlined
5478 sequential reads and behaves badly with applications doing random
5479 reads. Improve it by reading ahead at least read region, and
5480 avoiding excessive large RPC for small reads.
5482 Severity : enhancement
5484 Description: Creating more than 1000 files for a single job may cause a load
5485 imbalance on the OSTs if there are also a large number of OSTs.
5486 Details : qos_prep_create() uses an OST index reseed value that is an
5487 even multiple of the number of available OSTs so that if the
5488 reseed happens in the middle of the object allocation it will
5489 still utilize the OSTs as uniformly as possible.
5494 Description: OST or MDS may oops in ping_evictor_main()
5495 Details : ping_evictor_main() drops obd_dev_lock if deleting a stale export
5496 but doesn't restart at beginning of obd_exports_timed list
5499 Severity : enhancement
5501 Description: improve by-nid export eviction on the MDS and OST
5502 Details : allow multiple exports with the same NID to be evicted at one
5503 time without re-searching the exports list.
5506 Frequency : rare, only with supplementary groups enabled on SMP 2.6 kernels
5508 Description: MDS may oops in groups_free()
5509 Details : in rare race conditions a newly allocated group_info struct is
5510 freed again, and this can be NULL. The 2.4 compatibility code
5511 for groups_free() checked for a NULL pointer, but 2.6 did not.
5514 Frequency : common for liblustre clients doing little filesystem IO
5515 Bugzilla : 9352, 7313
5516 Description: server may evict liblustre clients accessing contended locks
5517 Details : if a client is granted a lock or receives a completion AST
5518 with a blocking AST already set it would not reply to the AST
5519 for LDLM_FL_CANCEL_ON_BLOCK locks. It now replies to such ASTs.
5522 Frequency : lfs setstripe, only systems with more than 160 OSTs
5524 Description: unable to set striping with a starting offset beyond OST 160
5525 Details : llapi_create_file() incorrectly limited the starting stripe
5526 index to the maximum single-file stripe count.
5529 Frequency : LDAP users only
5531 Description: lconf did not handle in-kernel recovery with LDAP properly
5532 Details : lconf/LustreDB get_refs() is searching the wrong namespace
5534 Severity : enhancement
5536 Description: bind OST threads to NUMA nodes to improve performance
5537 Details : all OST threads are uniformly bound to CPUs on a single NUMA
5538 node and do their allocations there to localize memory access
5540 Severity : enhancement
5542 Description: llmount can determine client NID directly from Myrinet (GM)
5543 Details : the client NID code from gmnalnid was moved directly into
5544 llmount, removing the need to use this or specifying the
5545 client NID explicitly when mounting GM clients with zeroconf
5548 Frequency : if client is started with down MDS
5550 Description: if client is started with down MDS mount hangs in ptlrpc_queue_wait
5551 Details : Having an LWI_INTR() wait event (interruptible, but no timeout)
5552 will wait indefinitely in ptlrpc_queue_wait->l_wait_event() after
5553 ptlrpc_import_delayed_req() because we didn't check if the
5554 request was interrupted, and we also didn't break out of the
5555 event loop if there was no timeout
5560 Description: data loss during non-page-aligned writes to a single file from
5561 both multiple nodes and multiple threads on one node at same time
5562 Details : updates to KMS and lsm weren't protected by common lock. Resulting
5563 inconsistency led to false short-reads, that were cached and later
5564 used by ->prepare_write() to fill in partially written page,
5565 leading to data loss.
5568 Frequency : always, if lconf --abort_recovery used
5570 Description: lconf --abort_recovery fails with 'Operation not supported'
5571 Details : lconf was attempting to abort recovery on the MDT device and not
5574 Severity : enhancement
5576 Description: remove cleanup logs
5577 Details : replace lconf-generated cleanup logs with lustre internal
5578 cleanup routines. Eliminates the need for client-cleanup and
5581 Severity : enhancement
5583 Description: add support for EAs (user and system) on lustre filesystems
5584 Details : it is now possible to store extended attributes in the Lustre
5585 client filesystem, and with the user_xattr mount option it
5586 is possible to allow users to store EAs on their files also
5588 Severity : enhancement
5590 Description: Add possibility (config option) to show minimal available OST free
5592 Details : When compiled with --enable-mindf configure option, statfs(2)
5593 (and so, df) will return least minimal free space available from
5594 all OSTs as amount of free space on FS, instead of summary of
5595 free spaces of all OSTs.
5597 Severity : enhancement
5599 Description: do not expand extent locks acquired on OST-side
5600 Details : Modify ldlm_extent_policy() to not expand local locks, acquired
5601 by server: they are not cached anyway.
5604 Frequency : when mmap is used/binaries executed from Lustre
5606 Description: Unmmap pages before throwing them away from read cache.
5607 Details : llap_shrink cache now attempts to unmap pages before discarding
5608 them (if unmapping failed - do not discard). SLES9 kernel has
5609 extra checks that trigger if this unmapping is not done first.
5614 Description: lconf didn't resolve symlinks before checking to see whether a
5615 given mountpoint was already in use
5618 Frequency : when migrating failover services
5619 Bugzilla : 6395, 9514
5620 Description: When migrating a subset of services from a node (e.g. failback
5621 from a failover service node) the remaining services would
5622 time out and evict clients.
5623 Details : lconf --force (implied by --failover) sets the global obd_timeout
5624 to 5 seconds in order to quickly disconnect, but this caused
5625 other RPCs to time out too quickly. Do not change the global
5626 obd_timeout for force cleanup, only set it for DISCONNECT RPCs.
5628 Severity : enhancement
5629 Frequency : if MDS is started with down OST
5630 Bugzilla : 9439,5706
5631 Description: Allow startup/shutdown of an MDS without depending on the
5632 availability of the OSTs.
5633 Details : Asynchronously call mds_lov_synchronize during MDS startup.
5634 Add appropriate locking and lov-osc refcounts for safe
5635 cleaning. Add osc abort_inflight calls in case the
5636 synchronize never started.
5639 Frequency : occasional (Cray XT3 only)
5641 Description: root not authorized to access files in CRAY_PORTALS environment
5642 Details : The client process capabilities were not honoured on the MDS in
5643 a CRAY_PORTALS/CRAY_XT3 environment. If the file had previously
5644 been accessed by an authorized user then root was able to access
5645 the file on the local client also. The root user capabilities
5646 are now allowed on the MDS, as this environment has secure UID.
5649 Frequency : occasional
5651 Description: ldiskfs "too long searching" message happens too often
5652 Details : A debugging message (otherwise harmless) prints too often on
5653 the OST console. This has been reduced to only happen when
5654 there are fragmentation problems on the filesystem.
5659 Description: Division by zero in statfs when all OSCs are inactive
5660 Details : lov_get_stripecnt() returns zero due to incorrect order of checks,
5661 lov_statfs divides by value returned by lov_get_stripecnt().
5665 Bugzilla : 9489, 3273
5666 Description: First write from each client to each OST was only 4kB in size,
5667 to initialize client writeback cache, which caused sub-optimal
5668 RPCs and poor layout on disk for the first writen file.
5669 Details : Clients now request an initial cache grant at (re)connect time
5670 and so that they can start streaming writes to the cache right
5671 away and always do full-sized RPCs if there is enough data.
5672 If the OST is rebooted the client also re-establishes its grant
5673 so that client cached writes will be honoured under the grant.
5678 Description: Slow ls (and stat(2) syscall) on files residing on IO-loaded OSTs
5679 Details : Now I/O RPCs go to different portal number and (presumably) fast
5680 lock requests (and glimses) and other RPCs get their own service
5681 threads pool that should be able to service those RPCs
5684 Severity : enhancement
5686 Description: Ability to exchange lustre version between client and servers and
5687 issue warnings at client side if client is too old. Also for
5688 liblustre clients there is ability to refuse connection of too old
5690 Details : New 'version' field is added to connect data structure that is
5691 filled with version info. That info is later checked by server and
5695 Frequency : rare, liblustre only.
5696 Bugzilla : 9296, 9581
5697 Description: Two simultaneous writes from liblustre at offset within same page
5698 might proceed at the same time overwriting eachother with stale
5700 Details : I/O lock withing llu_file_prwv was released too early, before data
5701 actually was hitting the wire. Extended lock-holding time until
5702 server acknowledges receiving data.
5705 Frequency : extremely rare. Never observed in practice.
5707 Description: avoid generating lustre_handle cookie of 0.
5708 Details : class_handle_hash() generates handle cookies by incrementing
5709 global counter, and can hit 0 occasionaly (this is unlikely, but
5710 not impossible, because initial value of cookie counter is
5711 selected randonly). Value of 0 is used as a sentinel meaning
5712 "unassigned handle" --- avoid it. Also coalesce two critical
5713 sections in this function into one.
5715 Severity : enhancement
5717 Description: allow liblustre clients to delegate truncate locking to OST
5718 Details : To avoid overhead of locking, liblustre client instructs OST to
5719 take extent lock in ost_punch() on client's behalf. New connection
5720 flag is added to handle backward compatibility.
5722 Severity : enhancement
5723 Bugzilla : 4928, 7341, 9758
5724 Description: allow number of OST service threads to be specified
5725 Details : a module parameter allows the number of OST service threads
5726 to be specified via "options ost ost_num_threads={N}" in the
5727 OSS's /etc/modules.conf or /etc/modprobe.conf.
5731 Bugzilla : 6146, 9635, 9895
5732 Description: servers crash with bad pointer in target_handle_connect()
5733 Details : In rare cases when a client is reconnecting it was possible that
5734 the connection request was the last reference for that export.
5735 We would temporarily drop the export reference and get a new
5736 one, but this may have been the last reference and the export
5737 was just destroyed. Get new reference before dropping old one.
5739 Severity : enhancement
5740 Frequency : if client is started with failover MDS
5742 Description: Allow multiple MDS hostnames in the mount command
5743 Details : Try to read the configuration from all specified MDS
5744 hostnames during a client mount in case the "primary"
5747 Severity : enhancement
5749 Description: Stop sending data to evicted clients as soon as possible.
5750 Details : Check if the client we are about to send or are sending data to
5751 was evicted already. (Check is done every second of waiting,
5752 for which l_wait_event interface was extended to allow checking
5753 of exit condition at specified intervals).
5756 Frequency : rare, normally only when NFS exporting is done from client
5758 Description: 'bad disk LOV MAGIC: 0x00000000' error when chown'ing files
5760 Details : Make mds_get_md() recognise empty md case and set lmm size to 0.
5763 Frequency : always, if srand() is called before liblustre initialization
5765 Description: Liblustre uses system PRNG disturbing its usage by user application
5766 Details : Introduce internal to lustre fast and high-quality PRNG for
5767 lustre usage and make liblustre and some other places in generic
5768 lustre code to use it.
5770 Severity : enhancement
5771 Bugzilla : 9477, 9557, 9870
5772 Description: Verify that the MDS configuration logs are updated when xml is
5773 Details : Check if the .xml configuration logs are newer than the config
5774 logs stored on the MDS and report an error if this is the case.
5775 Request --write-conf, or allow starting with --old_conf.
5777 Severity : enhancement
5779 Description: Handle symlinks in the path when checking if Lustre is mounted.
5780 Details : Resolve intermediate symlinks when checking if a client has
5781 mounted a filesystem to avoid duplicate client mounts.
5786 Description: lconf can hit an error exception but still return success.
5787 Details : The lconf command catches the Command error exception at the top
5788 level script context and will exit with the associated exit
5789 status, but doesn't ensure that this exit status is non-zero.
5794 Description: failure of ptlrpc thread startup can cause oops
5795 Details : Starting a ptlrpc service thread can fail if there are a large
5796 number of threads or the server memory is very fragmented.
5797 Handle this without oopsing.
5800 Frequency : always, only if liblustre and non-default acceptor port was used
5802 Description: liblustre cannot connect to servers with non-default acceptor port
5803 Details : tcpnal_set_default_params() was not called and was therefore
5804 ignoring the environment varaible TCPNAL_PORT, as well as other
5805 TCPNAL_ environment variables
5810 Description: two objects could be created on the same OST for a single file
5811 Details : If an OST is down, in some cases it was possible to create two
5812 objects on a single OST for a single file. No problems other
5813 than potential performance impact and spurious error messages.
5817 Bugzilla : 5681, 9562
5818 Description: Client may oops in ll_unhash_aliases
5819 Details : Client dcache may become inconsistent in race condition.
5820 In some cases "getcwd" can fail if the current directory is
5826 Description: Inode refcounting problems in NFS export code
5827 Details : link_raw functions used to call d_instantiate without obtaining
5828 extra inode reference first.
5832 Bugzilla : 9942, 9903
5833 Description: Referencing freed requests leading to crash, memleaks with NFS.
5834 Details : We used to require that call to ll_revalidate_it was always
5835 followed by ll_lookup_it. Also with revalidate_special() it is
5836 possible to call ll_revalidate_it() twice for the same dentry
5837 even if first occurence returned success. This fix changes semantic
5838 between DISP_ENQ_COMPLETE disposition flag to mean there is extra
5839 reference on a request referred from the intent.
5840 ll_intent_release() then releases such a request.
5843 Frequency : rare, normally benchmark loads only
5845 Description: unlinked inodes were kept in memory on the client
5846 Details : If a client is repeatedly creating and unlinking files it
5847 can accumulate a lot of stale inodes in the inode slab cache.
5848 If there is no other client load running this can cause the
5849 client node to run out of memory. Instead flush old inodes
5850 from client cache that have the same inode number as a new inode.
5853 Frequency : SLES9 2.6.5 kernel and long filenames only
5854 Bugzilla : 9969, 10379
5855 Description: utime reports stale NFS file handle
5856 Details : SLES9 uses out-of-dentry names in some cases, which confused
5857 the lustre dentry revalidation. Change it to always use the
5861 Frequency : rare, unless heavy write-truncate concurrency is continuous
5862 Bugzilla : 4180, 6984, 7171, 9963, 9331
5863 Description: OST becomes very slow and/or deadlocked during object unlink
5864 Details : filter_destroy() was holding onto the parent directory lock
5865 while truncating+unlinking objects. For very large objects this
5866 may block other threads for a long time and slow overall OST
5867 responsiveness. It may also be possible to get a lock ordering
5868 deadlock in this case, or run out of journal credits because of
5869 the combined truncate+unlink. Solution is to do object truncate
5870 first in one transaction without parent lock, and then do the
5871 final unlink in a new transaction with the parent lock. This
5872 reduces the lock hold time dramatically.
5875 Frequency : rare, 2.4 kernels only
5877 Description: MDS or OST cleanup may trip kernel BUG when dropping kernel lock
5878 Details : mds_cleanup() and filter_cleanup() need to drop the kernel lock
5879 before unmounting their filesystem in order to avoid deadlock.
5880 The kernel_locked() function in 2.4 kernels only checks whether
5881 the kernel lock is held, not whether it is this process that is
5882 holding it as 2.6 kernels do.
5887 Description: MDS or OST may oops/LBUG if a client is connecting multiple times
5888 Details : The client ptlrpc code may be trying to reconnect to a down
5889 server before a previous connection attempt has timed out.
5890 Increase the reconnect interval to be longer than the connection
5891 timeout interval to avoid sending duplicate connections to
5895 Frequency : echo_client brw_test command
5897 Description: fix echo_client to work with OST preallocated code
5898 Details : OST preallocation code (5137) didn't take echo_client IO path
5899 into account: echo_client calls filter methods outside of any
5900 OST thread and, hence, there is no per-thread preallocated
5901 pages and buffers to use. Solution: hijack pga pages for IO. As
5902 a byproduct, this avoids unnecessary data copying.
5906 Bugzilla : 3555, 5962, 6025, 6155, 6296, 9574
5907 Description: Client can oops in mdc_commit_close() after open replay
5908 Details : It was possible for the MDS to return an open request with no
5909 transaction number in mds_finish_transno() if the client was
5910 evicted, but without actually returning an error. Clients
5911 would later try to replay that open and may trip an assertion
5912 Simplify the client close codepath, and always return an error
5913 from the MDS in case the open is not successful.
5916 Frequency : rare, 2.6 OSTs only
5918 Description: OST may deadlock under high load on fragmented files
5919 Details : If there was a heavy load and highly-fragmented OST filesystems
5920 it was possible to have all the OST threads deadlock waiting on
5921 allocation of biovecs, because the biovecs were not released
5922 until the entire RPC IO was completed. Instead, release biovecs
5923 as soon as they are complete to ensure forward IO progress.
5925 Severity : enhancement
5927 Description: Support for specifying external journal device at mount
5928 Details : If an OST or MDS device is formatted with an external journal
5929 device, this device major/minor is stored in the ext3 superblock
5930 and may not be valid for failover. Allow detecting and
5931 specifying the external journal at mount time.
5936 Description: Mounting an MDS with pending unlinked files may cause oops
5937 Details : target_finish_recovery() calls mds_postrecov() which returned
5938 the number of orphans unlinked. mds_lov_connect->mds_postsetup()
5939 considers this an error and immediately begins cleaning up the
5940 lov, just after starting the mds_lov process
5942 Severity : enhancement
5944 Description: Implement 'lfs df' to report actual free space on per-OST basis
5945 Details : Add sub-command 'df' on 'lfs' to report the disk space usage of
5946 MDS/OSDs. Usage: lfs df [-i][-h]. Command Options: '-i' to report
5947 usage of objects; '-h' to report in human readable format.
5949 ------------------------------------------------------------------------------
5951 2005-08-26 Cluster File Systems, Inc. <info@clusterfs.com>
5958 Description: Mounting an ldiskfs file system with mballoc may crash OST node.
5959 Details : ldiskfs mballoc code may reference an uninitialized buddy struct
5960 at startup during orphan unlinking. Instead, skip buddy update
5961 before setup, as it will be regenerated after recovery is complete.
5966 Description: If an OST is inactive, its locks might reference stale inodes.
5967 Details : lov_change_cbdata() must iterate over all namespaces, even if
5968 they are inactive to clear inode references from the lock.
5970 Severity : enhancement
5971 Frequency : occasional, if non-standard max_dirty_mb used
5973 Description: Client will block write RPCs if not enough grant
5974 Details : If a client has max_dirty_mb smaller than max_rpcs_in_flight,
5975 then the client will block writes while waiting for another RPC
5976 to complete instead of consuming its dirty limit. With change
5977 we get improved performance when max_dirty_mb is small.
5979 Severity : enhancement
5980 Bugzilla : 3389, 6253
5981 Description: Add support for supplementary groups on the MDS.
5982 Details : The MDS has an upcall /proc/fs/lustre/mds/{mds}/group_upcall
5983 (set to /usr/sbin/l_getgroups if enabled) which will do MDS-side
5984 lookups for user supplementary groups into a cache.
5988 Description: O_CREAT|O_EXCL open flags in liblustre always return -EEXIST
5989 Details : Make libsysio to not enforce O_EXCL by clearing the flag,
5990 for liblustre O_EXCL is enforced by MDS.
5994 Description: readdir never returns NULL in liblustre.
5995 Details : Corrected llu_iop_getdirentries logic, to return offset of next
5996 dentry in struct dirent.
6000 Frequency : liblustre only, depends on application IO pattern
6001 Description: liblustre clients evicted if not contacting servers
6002 Details : Don't put liblustre clients into the ping_evictor list, so
6003 they will not be evicted by the pinger ever.
6005 Severity : enhancement
6007 Description: Add ability to evict clients by NID from MDS.
6008 Details : By echoing "nid:$NID" string into
6009 /proc/fs/lustre/mds/.../evict_client client with nid that equals to
6010 $NID would be instantly evicted from this MDS and from all active
6011 OSTs connected to it.
6015 Description: Do not query file size twice, somewhat slowing stat(2) calls.
6016 Details : lookup_it_finish() used to query file size from OSTs that was not
6021 Description: service threads change working directory to that of init
6022 Details : Starting lustre service threads may pin the working directory
6023 of the parent thread, making that filesystem busy. Threads
6024 now change to the working directory of init to avoid this.
6028 Frequency : during shutdown only
6029 Description: shutdown with a failed MDS or OST can cause unmount to hang
6030 Details : Don't resend DISCONNECT messages in ptlrpc_disconnect_import()
6035 Frequency : 2.6 only
6036 Description: chmod/chown may include an extra supplementary group
6037 Details : ll{,u}_mdc_pack_op_data() does not properly initialize the
6038 supplementary group and if none is specified this is used.
6041 Bugzilla : 5479 (6816)
6043 Description: Racing open + rm can assert client in mdc_set_open_replay_data()
6044 Details : If lookup is in progress on a file that is unlinked we might try
6045 to revalidate the inode and fail in revalidate after lookup is
6046 complete and ll_file_open() enqueues the open again but
6047 it_open_error() was not checking DISP_OPEN_OPEN errors correctly.
6050 Frequency : always, if lconf --abort_recovery used
6052 Description: lconf --abort_recovery fails with 'Operation not supported'
6053 Details : lconf was attempting to abort recovery on the MDT device and not
6056 ------------------------------------------------------------------------------
6058 2005-08-08 Cluster File Systems, Inc. <info@clusterfs.com>
6063 Frequency : rare (only unsupported configurations with a node running as an
6065 Bugzilla : 6514, 5137
6066 Description: Mounting a Lustre file system on a node running as an OST could
6068 Details : OSTs now preallocates memory needed to write out data at
6069 startup, instead of when needed, to avoid having to
6070 allocate memory in possibly low memory situations.
6071 Specifically, if the file system is mounted on on OST,
6072 memory pressure could force it to try to write out data,
6073 which it needed to allocate memory to do. Due to the low
6074 memory, it would be unable to do so and the node would
6075 become unresponsive.
6077 Severity : enhancement
6079 Description: Addition of lconf --service command line option
6080 Details : lconf now accepts a '--service <arg>' option, which is
6081 shorthand for 'lconf --group <arg> --select <arg>=<hostname>'
6083 Severity : enhancement
6085 Description: Failover mode is now the default for OSTs.
6086 Details : By default, OSTs will now run in failover mode. To return to
6087 the old behaviour, add '--failout' to the lmc line for OSTs.
6089 Severity : enhancement
6091 Description: Health checks are now provided for MDS and OSTs
6092 Details : Additional detailed health check information on MSD and OSTs
6093 is now provided through the procfs health_check value.
6096 Frequency : occasional, depends on IO load
6098 Description: Disk fragmentation on the OSTs could eventually cause slowdowns
6099 after numerous create/delete cycles
6100 Details : The ext3 inode allocation policy would not allocate new inodes
6101 very well on the OSTs because there are no new directories
6102 being created. Instead we look for groups with free space if
6103 the parent directories are nearly full.
6108 Description: Network or server problems during mount may cause partially
6109 mounted clients instead of returning an error.
6110 Details : The config llog parsing code may overwrite the error return
6111 code during mount error handling, returning success instead
6117 Description: MDS can fail to allocate large reply buffers
6118 Details : After long uptimes the MDS can fail to allocate large reply
6119 buffers (e.g. zconf client mount config records) due to memory
6120 fragmentation or consumption by the buffer cache. Preallocate
6121 some large reply buffers so that these replies can be sent even
6122 under memory pressure.
6126 Frequency : rare (liblustre)
6127 Description: fsx running with liblustre complained that using truncate() to
6128 extend the file doesn't work. This patch corrects that issue.
6129 Details : This is the liblustre equivalent of the fix for bug 6196. Fixes
6130 ATTR_SIZE and lsm use in llu_setattr_raw.
6134 Frequency : rare, only 2.6 kernels
6135 Description: Unusual file access patterns on the MDS may result in inode
6136 data being lost in very rare circumstances.
6137 Details : Bad interaction between the ea-in-inode patch and the "no-read"
6138 code in the 2.6 kernel caused the inode and/or EA data not to
6139 be read from disk, causing single-file corruption.
6143 Frequency : rare, only 2.6 filesystems using extents
6144 Description: Heavy concurrent write and delete load may cause data corruption.
6145 Details : It was possible under high-load situations to have an extent
6146 metadata block in the block device cache from a just-unlinked
6147 file overwrite a newly-allocated data block. We now unmap any
6148 metadata buffers that alias just-allocated data blocks.
6152 Frequency : filesystems with default stripe_count larger than 77
6153 Description: lconf+mke2fs fail when formatting filesystem with > 77 stripes
6154 Details : lconf specifies an inode size of 4096 bytes when the default
6155 stripe_count is larger than 77. This conflicts with the default
6156 inode density of 1 per 4096 bytes. Allocate smaller inodes in
6157 this case to avoid pinning too much memory for large EAs.
6159 ------------------------------------------------------------------------------
6161 2005-07-07 Cluster File Systems, Inc. <info@clusterfs.com>
6166 Frequency : rare (extremely heavy IO load with hundreds of clients)
6168 Description: Client is evicted, gets IO error writing to file
6169 Details : lock ordering changes for bug 5492 reintroduced bug 3267 and
6170 caused clients to be evicted for AST timeouts. The fixes in
6171 bug 5192 mean we no longer need to have such short AST timeouts
6172 so ldlm_timeout has been increased.
6175 Frequency : occasional during --force or --failover shutdown under load
6176 Bugzilla : 5949, 4834
6177 Description: Server oops/LBUG if stopped with --force or --failover under load
6178 Details : a collection of import/export refcount and cleanup ordering
6179 issues fixed for safer force cleanup
6182 Frequency : only filesystems larger than 120 OSTs
6183 Bugzilla : 5990, 6223
6184 Description: lfs getstripe would oops on a very large filesystem
6185 Details : lov_getconfig used kfree on vmalloc'd memory
6188 Frequency : only filesystems exporting via NFS to Solaris 10 clients
6189 Bugzilla : 6242, 6243
6190 Description: reading from files that had been truncated to a non-zero size
6191 but never opened returned no data
6192 Details : ll_file_read() reads zeros from no-object files to EOF
6197 Description: A bug in MDS/OSS recovery could cause the OSS to fail an assertion
6198 Details : There's little harm in aborting MDS/OSS recovery and letting it
6199 try again, so I removed the LASSERT and return an error instead.
6201 Severity : enhancement
6203 Description: New debugging infrastructure for tracking down data corruption
6204 Details : The I/O checksum code was replaced to: (a) control it at runtime,
6205 (b) cover more of the client-side code path, and (c) try to narrow
6206 down where problems occurred
6210 Bugzilla : 3819, 4364, 4397, 6313
6211 Description: Racing close and eviction MDS could cause assertion in mds_close
6212 Details : It was possible to get multiple mfd references during close and
6213 client eviction, leading to one thread referencing a freed mfd.
6215 Severity: : enhancement
6216 Bugzilla : 3262, 6359
6217 Description: Attempts to reconnect to servers are now more aggressive.
6218 Details : This builds on the enhanced upcall-less recovery that was added
6219 in 1.4.2. When trying to reconnect to servers, clients will
6220 now try each server in the failover group every 10 seconds. By
6221 default, clients would previously try one server every 25 seconds.
6226 Description: After recovery, certain operations trigger a failed
6227 assertion on a client.
6228 Details : Failing over an mds, using lconf -d --failover, while a
6229 client was doing a readdir() call would cause the client to
6230 LBUG after recovery completed and the readdir() was resent.
6232 Severity : enhancement
6234 Description: Default groups are now added by lconf
6235 Details : You can now run lconf --group <servicename> without having to
6236 manually add groups with lmc.
6239 Frequency : occasional
6241 Description: Nodes with an elan id of 0 trigger a failed assertion
6244 Frequency : always when accessing e.g. tty/console device nodes
6246 Description: tty and some other devices nodes cannot be used on lustre
6247 Details : file's private_data field is used by device data and lustre
6248 values in there got lost. New field was added to struct file to
6249 store fs-specific private data.
6252 Frequency : when exporting Lustre via NFS
6254 Description: NFSD failed occasionally when looking up a path component
6255 Details : NFSD is looking up ".." which was broken in ext3 directories
6256 that had grown large enough to become hashed.
6259 Frequency : Clusters with multiple interfaces not on the same subnet
6261 Description: Nodes will repeatedly try to reconnect to an interface which it
6262 cannot reach and report an error to the log.
6263 Details : Extra peer list entries will be created by lconf with some peers
6264 unreachable. lconf now validates the peer before adding it.
6267 Frequency : Only if a default stripe is set on the filesystem root.
6269 Description: Setting a default stripe on the filesystem root prevented the
6270 filesystem from being remounted.
6271 Details : The client was sending extra request flags in the root getattr
6272 request and did not allocate a reply buffer for the dir EA.
6275 Frequency : occasional, higher if lots of files are accessed by one client
6276 Bugzilla : 6159, 6097
6277 Description: Client trips assertion regarding lsm mismatch/magic
6278 Details : While revalidating inodes the VFS looks up inodes with ifind()
6279 and in rare cases can find an inode that is being freed.
6280 The ll_test_inode() code will free the lsm during ifind()
6281 when it finds an existing inode and then the VFS later attaches
6282 this free lsm to a new inode.
6286 Bugzilla : 6422, 7030
6287 Description: MDS deadlock between mkdir and client eviction
6288 Details : Creating a new file via mkdir or mknod (starting a transaction
6289 and getting the ns lock) can deadlock with client eviction
6290 (gets ns lock and trying to finish a synchronous transaction).
6293 Frequency : occasional
6294 Description: While starting a server, the fsfilt_ext3 module could not be
6296 Details : CFS's improved ext3 filesystem is named ldiskfs for 2.6
6297 kernels. Previously, lconf would still use the ext3 name
6298 when trying to load modules. Now, it will correctly use
6299 ext3 on 2.4 and ldiskfs on 2.6.
6301 Severity : enhancement
6302 Description: The default stripe count has been changed to 1
6303 Details : The interpretation of the default stripe count (0, to lfs
6304 or lmc) has been changed to mean striping across a single
6305 OST, rather than all available. For general usage we have
6306 found a stripe count of 1 or 2 works best.
6308 Severity : enhancement
6309 Description: Add support for compiling against Cray portals.
6310 Details : Conditional compiling for some areas that are different
6314 Frequency : occasional
6315 Bugzilla : 6409, 6834
6316 Description: Creating files with an explicit stripe count may lead to
6317 a failed assertion on the MDS
6318 Details : If some OSTs are full or unavailable, creating files may
6319 trigger a failed assertion on the MDS. Now, Lustre will
6320 try to use other servers or return an error to the
6324 Frequency : occasional
6326 Description: Multiple concurrent overlapping read+write on multiple SMP nodes
6327 caused lock timeout during readahead (since 1.4.2).
6328 Details : Processes doing readahead might match a lock that hasn't been
6329 granted yet if there are overlapping and conflicting lock
6330 requests. The readahead process waits on ungranted lock
6331 (original lock is CBPENDING), while OST waits for that process
6332 to cancel CBPENDING read lock and eventually evicts client.
6334 Severity : enhancement
6336 Description: Initial enabling of flock support for clients
6337 Details : Implements fcntl advisory locking and file status functions.
6338 This feature is provided as an optional mount flag (default
6339 off), and is NOT CURRENTLY SUPPORTED. Not all types of record
6340 locking are implemented yet, and those that are are not guaranteed
6341 to be completely correct in production environments.
6342 mount -t lustre -o [flock|noflock] ...
6345 Frequency : occasional
6347 Description: OSTs running 2.4 kernels but with extents enabled might trip an
6348 assertion in the ext3 JBD (journaling) layer.
6349 Details : The b_committed_data struct is protected by the big kernel lock
6350 in 2.4 kernels, serializing journal_commit_transaction() and
6351 ext3_get_block_handle->ext3_new_block->find_next_usable_block()
6352 access to this struct. In 2.6 kernels there is finer grained
6353 locking to improve SMP performance of the JBD layer.
6357 Description: Changes the "SCSI I/O Stats" kernel patch to default to "enabled"
6359 -----------------------------------------------------------------------------
6361 2005-05-05 Cluster File Systems, Inc. <info@clusterfs.com>
6363 NOTE: Lustre 1.4.2 uses an incompatible network protocol than previous
6364 versions of Lustre. Please update all servers and clients to
6365 version 1.4.2 or later at the same time. You must also run
6366 "lconf --write-conf {config}.xml" on the MDS while it is stopped
6367 to update the configuration logs.
6369 - fix for HPUX NFS client breakage when NFS exporting Lustre (5781)
6370 - mdc_enqueue does not need max_mds_easize request buffer on send (5707)
6371 - swab llog records of type '0' so we get proper header size/idx (5861)
6372 - send llog cancel req to DLM cancel portal instead of cb portal (5515)
6373 - fix rename of one directory over another leaking an inode (5953)
6374 - avoid SetPageDirty on 2.6 (5981)
6375 - don't re-add just-being-destroyed locks to the waiting list (5653)
6376 - when creating new directories, inherit the parent's custom
6377 striping settings if present parent (3048)
6378 - flush buffers from cache before direct IO in 2.6 obdfilter (4982)
6379 - don't hold i_size_sem in ll_nopage() and ll_ap_refresh_count (6077)
6380 - don't hold client locks on temporary worklist from l_lru (5666)
6381 - handle IO errors in 2.6 obdfilter bio completion routine (6046)
6382 - automatically evict dead clients (5921)
6383 - Update file size properly in create+truncate+fstat case (6196)
6384 - Do not unhash mountpoint dentries, do not allow removal of
6386 - Avoid lock ordering deadlock issue with write/truncate (6203,5654)
6387 - reserve enough journal credits in fsfilt_start_log for setattr (4554)
6388 - ldlm_enqueue freed-export error path would always LBUG (6149,6184)
6389 - don't reference lr_lvb_data until after we hold lr_lvb_sem (6170)
6390 - don't overwrite last_rcvd if there is a *_client_add() error (6086)
6391 - Correctly handle reads of files with no objects (6243)
6392 - lctl recover will also mark a device active if deactivate used (5933)
6394 - by default create 1 inode per 4kB space on MDS, per 16kB on OSTs
6395 - allow --write-conf on an MDS with different nettype than client (5619)
6396 - don't write config llogs to MDS for mounts not from that MDS (5617)
6397 - lconf should create multiple TCP connections from a client (5201)
6398 - init scripts are now turned off by default; run chkconfig --on
6399 lustre and chkconfig --on lustrefs to use them
6400 - upcalls are no longer needed for clients to recover to failover
6402 - add --abort-recovery option to lconf to abort recovery on device
6404 - add support for an arbitrary number of OSTs (3026)
6405 - Quota support protocol changes.
6406 - forward compatibility changes to wire structs (6007)
6407 - rmmod NALs that might be loaded because of /etc/modules.conf (6133)
6408 - support for mountfsoptions and clientoptions to the Lustre LDAP (5873)
6409 - improved "lustre status" script
6410 - initialize blocksize for non-regular files (6062)
6411 - added --disable-server and --disable-client configure options (5782)
6412 - introduce a lookup cache for lconf to avoid repeated DB scans (6204)
6413 - Vanilla 2.4.29 support
6414 - increase maximum number of obd devices to 520 (6242)
6415 - remove the tcp-zero-copy patch from the suse-2.4 series (5902)
6416 - Quadrics Elan drivers are now included for the RHEL 3 2.4.21 and
6417 SLES 9 2.6.5 kernels
6418 - limit stripes per file to 160 (the maximum EA size) (6093)
6420 2005-03-22 Cluster File Systems, Inc. <info@clusterfs.com>
6423 - don't LASSERT in ll_release on NULL lld with NFS export (4655, 5760)
6424 - hold NS lock when calling handle_ast_error->del_waiting_lock (5746)
6425 - fix setattr mtime regression from lovcleanup merge (4829, 5669)
6426 - workaround for 2.6 crash in ll_unhash_aliases (5687, 5210)
6427 - small ext3 extents cleanups and fixes (5733)
6428 - improved mballoc code, several small races and bugs fixed (5733, 5638)
6429 - kernel version 43 - fix remove_suid bugs in both 2.4 and 2.6 (5695)
6430 - avoid needless client->OST connect, fix handle mismatch (5317)
6431 - fix DLM error path that led to out-of-sync client, long delays (5779)
6432 - support common vfs-enforced mount options (nodev,nosuid,noexec) (5637)
6433 - fix several locking issues related to i_size (5492,5624,5654,5672)
6434 - don't move pending lock onto export if it is already evicted (5683)
6435 - fix kernel oops when creating .foo in unlinked directory (5548)
6436 - fix deadlock in obdfilter statistics vs. object create (5811)
6437 - use time_{before,after} to avoid timer jiffies wrap (5882)
6438 - shutdown --force/--failover stability (3607,3651,4797,5203,4834)
6439 - Do not leak request if server was not able to process it (5154)
6440 - If mds_open unable to find parent dir, make that negative lookup(5154)
6441 - don't create new directories with extent-mapping (5909, 5936)
6443 - fix lustre/lustrefs init scripts for SuSE (patch from Scali, 5702)
6444 - don't hold the pinger_sem in ptlrpc_pinger_sending_on_import
6445 - change obd_increase_kms to obd_adjust_kms (up or down) (5654)
6446 - lconf, lmc search both /usr/lib and /usr/lib64 for Python libs (5800)
6447 - support for RHEL4 kernel on i686 (5773)
6448 - provide error messages when incompatible logs are encountered (5898)
6450 2005-02-18 Cluster File Systems, Inc. <info@clusterfs.com>
6451 * version 1.4.0.10 (1.4.1 release candidate 1)
6453 - don't keep a lock reference when lock is not granted (4238)
6454 - unsafe list practices (rarely) led to infinite eviction loop (4908)
6455 - add per-fs limit of Lustre pages in page cache, avoid OOM (4699)
6456 - drop import inflight refcount on signal_completed_replay error (5255)
6457 - unlock page after async write error during send (3677)
6458 - handle missing objects in filter_preprw_read properly (5265)
6459 - no transno return for symlink open, don't save no-trasno open (3440)
6460 - don't try to complete elan receive that already failed (4012)
6461 - free RPC server reply state on error (5406)
6462 - clean up thread from ptlrpc_start_thread() on error (5160)
6463 - readahead could read extra page into cache that wasn't ejected (5388)
6464 - prevent races in class_attach/setup/cleanup/detach (5260)
6465 - don't dereference de->d_inode after l_dput of de (5458)
6466 - use "int" for stripe value returned from lock_to_stripe (5544)
6467 - mballoc allocation and error-checking fixes in 2.6 (5504)
6468 - block device patches to fix I/O request sizes in 2.6 (5482)
6469 - look up hostnames for IB nals (5602)
6470 - 2.6 changed lock ordering of 2 semaphores, caused deadlock (5654)
6471 - don't start multiple acceptors for the same port (5277)
6472 - fix incorrect LASSERT in mds_getattr_name (5635)
6473 - export a proc file for general "ping" checking (5628)
6474 - fix "lfs check" to not block when the MDS is down (5628)
6476 - service request history (4965)
6477 - put {ll,lov,osc}_async_page structs in a single slab (4699)
6478 - create an "evict_client" /proc entry on OSTs, like the MDS has
6479 - fix mount usage message, return errors per mount(8) (5168)
6480 - change grep [] to grep "[]" in tests so they work in more UMLs
6481 - fix ppc64/x86_64 spec to use %{_libdir} instead of /usr/lib (5389)
6482 - remove ancient LOV_MAGIC_V0 EA support (5047)
6483 - add "disk I/Os in flight" and "I/O req time" stats in obdfilter
6484 - align r/w RPCs to PTLRPC_MAX_BRW_SIZE boundary for performance (3451)
6485 - allow readahead allocations to fail when low on memory (5383)
6486 - mmap locking landed again, after considerable improvement (2828)
6487 - add get_hostaddr() to lustreDB.py for LDAP support (5459)
6489 2004-11-23 Cluster File Systems, Inc. <info@clusterfs.com>
6492 - send OST transaction number in read/write reply to free req (4966)
6493 - don't ASSERT in ptl_send_rpc() if we run out of memory (5119)
6494 - lock /proc/sys/portals/routes internal state, avoiding oops (4827)
6495 - the watchdog thread now runs as interruptible (5246)
6496 - flock/lockf fixes (but it's still disabled, pending 5135)
6497 - don't use EXT3 constants in llite code (5094)
6498 - memory shortage at startup could cause assertion (5176)
6500 - reorganization of lov code
6501 - single portals codebase
6503 - add extents/mballoc support (5025)
6504 - direct I/O reads in the obdfilter (4048)
6505 - kernel patches from LNXI for 2.6 (bluesmoke, perfctr, mtd, kexec)
6507 tbd Cluster File Systems, Inc. <info@clusterfs.com>
6510 - send OST transaction number in read/write reply to free req (4966)
6511 - don't ASSERT in ptl_send_rpc() if we run out of memory (5119)
6512 - lock /proc/sys/portals/routes internal state, avoiding oops (4827)
6513 - the watchdog thread now runs as interruptible (5246)
6514 - handle missing objects in filter_preprw_read properly (5265)
6515 - unsafe list practices (rarely) led to infinite eviction loop (4908)
6516 - drop import inflight refcount on signal_completed_replay error (5255)
6517 - unlock page after async write error during send (3677)
6518 - return original error code on reconstructed replies (3761)
6519 - no transno return for symlink open, don't save no-trasno open (3440)
6521 - add pid to ldlm debugging output (4922)
6522 - bump the watchdog timeouts -- we can't handle 30sec yet
6523 - extra debugging for orphan dentry/inode bug (5259)
6525 2004-11-16 Cluster File Systems, Inc. <info@clusterfs.com>
6528 - fix TCP_NODELAY bug, which caused extreme perf regression (5134)
6529 - allocate qswnal tx descriptors singly to avoid fragmentation (4504)
6530 - don't LBUG on obdo_alloc() failure, use OBD_SLAB_ALLOC() (4800)
6531 - fix NULL dereference in /proc/sys/portals/routes (4827)
6532 - allow failed mdc_close() operations to be interrupted (4561)
6533 - stop precreate on OST before MDS would time out on it (4778)
6534 - don't send partial-page writes before EOF from client (4410)
6535 - discard client grant for sub-page writes on large-page clients (4520)
6536 - don't free dentries not owned by NFS code, check generation (4806)
6537 - fix lsm leak if mds_create_objects() fails (4801)
6538 - limit debug_daemon file size, always print CERROR messages (4789)
6539 - use transno after validating reply (3892)
6540 - process timed out requests if import state changes (3754)
6541 - update mtime on OST during writes, return in glimpse (4829)
6542 - add mkfsoptions to LDAP (4679)
6543 - use ->max_readahead method instead of zapping global ra (5039)
6544 - don't interrupt __l_wait_event() during strace
6546 - add software watchdogs to catch hung threads quickly (4941)
6547 - make lustrefs init script start after nfs is mounted
6548 - fix CWARN/ERROR duplication (4930)
6549 - return async write errors to application if possible (2248)
6550 - add /proc/sys/portal/memused (bytes allocated by PORTALS_ALLOC)
6551 - print NAL number in %x format (4645)
6552 - update barely-supported suse-2.4.21-171 series (4842)
6553 - support for sles 9 %post scripts
6554 - support for building 2.6 kernel-source packages
6555 - support for sles km_* packages
6557 2004-10-07 Cluster File Systems, Inc. <info@clusterfs.com>
6560 - ignore -ENOENT errors in osc_destroy (3639)
6561 - notify osc create thread that OSC is being cleaned up (4600)
6562 - add nettype argument for llmount in #5d in conf-sanity.sh (3936)
6563 - reconstruct ost_handle() like mds_handle() (4657)
6564 - create a new thread to do import eviction to avoid deadlock (3969)
6565 - let lconf resolve symlinked-to devices (4629)
6566 - don't unlink "objects" from directory with default EA (4554)
6567 - hold socknal file ref over connect in case target is down (4394)
6568 - allow more than 32000 subdirectories in a single directory (3244)
6569 - fix blocks count for O_DIRECT writes (3751)
6570 - OST returns ENOSPC from object create when no space left (4539)
6571 - don't send truncate RPC if file size isn't changing (4410)
6572 - limit OSC precreate to 1/2 of value OST considers bogus (4778)
6573 - bind to privileged port in socknal and tcpnal (3689)
6575 - rate limit CERROR/CWARN console message to avoid overload (4519)
6576 - GETFILEINFO dir ioctl returns LOV EA + MDS stat in 1 call (3327)
6577 - basic mmap support (3918)
6578 - kernel patch series update from b1_4 (4711)
6580 2004-09-16 Cluster File Systems, Inc. <info@clusterfs.com>
6583 - avoid crash during MDS cleanup with OST shut down (2775)
6584 - fix loi_list_lock/oig_lock inversion on interrupted IO (4136)
6585 - don't use bad inodes on the MDS (3744)
6586 - dynamic object preallocation to improve recovery speed (4236)
6587 - don't hold spinlock over lock dumping or change debug flags (4401)
6588 - don't zero obd_dev when it is force cleaned (3651)
6589 - print grants to console if they go negative (4431)
6590 - "lctl deactivate" will stop automatic recovery attempts (3406)
6591 - look for existing locks in ldlm_handle_enqueue() (3764)
6592 - don't resolve lock handle twice in recovery avoiding race (4401)
6593 - revalidate should check working dir is a directory (4134)
6595 - don't always mark "slow" obdfilter messages as errors (4418)
6597 2004-08-24 Cluster File Systems, Inc. <info@clusterfs.com>
6600 - don't close LustreDB during write_conf until it is done (3860)
6601 - fix typo in lconf for_each_profile (3821)
6602 - allow dumping logs from multiple threads at one time (3820)
6603 - don't allow multiple threads in OSC recovery (3812)
6604 - fix debug_size parameters (3864)
6605 - fix mds_postrecov to initialize import for llog ctxt (3121)
6606 - replace config semaphore with spinlock (3306)
6607 - be sure to send a reply for a CANCEL rpc with bad export (3863)
6608 - don't allow enqueue to complete on a destroyed export (3822)
6609 - down write_lock before checking llog header bitmap (3825)
6610 - recover from lock replay timeout (3764)
6611 - up llog sem before sending rpc (3652)
6612 - reduce ns lock hold times when setting kms (3267)
6613 - change a dlm LBUG to LASSERTF, to maybe learn something (4228)
6614 - fix NULL deref and obd_dev leak on setup error (3312)
6615 - replace some LBUG about llog ops with error handling (3841)
6616 - don't match INVALID dentries from d_lookup and spin (3784)
6617 - hold dcache_lock while marking dentries INVALID and hashing (4255)
6618 - fix invalid assertion in ptlrpc_set_wait (3880)
6620 - add libwrap support for the TCP acceptor (3996)
6621 - add /proc/sys/portals/routes for non-root route listing (3994)
6622 - allow setting MDS UUID in .xml (2580)
6623 - print the stack of a process that LBUGs (4228)
6625 2004-07-14 Cluster File Systems, Inc. <info@clusterfs.com>
6628 - don't cleanup request in ll_file_open() on failed MDS open (3430)
6629 - make sure to unset replay flag from failed open requests (3440)
6630 - if default stripe count is 0, use OST count for inode size (3636)
6631 - update parent mtime/ctime on client for create/unlink (2611)
6632 - drop dentry ref in ext3_add_link from open_connect_dentry (3266)
6633 - free recovery state on server during a forced cleanup (3571)
6634 - unregister_reply for resent reqs (3063)
6635 - loop back devices mounting and status check on 2.6 (3563)
6636 - fix resource-creation race that can provoke i_size == 0 (3513)
6637 - don't try to use bad inodes returned from MDS/OST fs lookup (3688)
6638 - more debugging for page-accounting assertion (3746)
6639 - return -ENOENT instead of asserting if ost getattr+unlink race (3558)
6640 - avoid deadlock after precreation failure (3758)
6641 - fix race and lock order deadlock in orphan handling (3450, 3750)
6642 - add validity checks when grabbing inodes from l_ast_data (3599)
6644 - add /proc/.../recovery_status to obdfilter (3428)
6645 - lightweight CDEBUG infrastructure, debug daemon (3668)
6646 - change default OSC RPC parameters to be better on small clusters
6647 - turn off OST read cache for files smaller than 32MB
6648 - install man pages and include them in rpms (3100)
6649 - add new init script for (un)mounting lustre filesystems (2593)
6650 - run chkconfig in %post for init scripts (3701)
6651 - drop scimac NAL (unmaintained)
6653 2004-06-17 Cluster File Systems, Inc. <info@clusterfs.com>
6656 - clean kiobufs before and after use (3485)
6657 - strip trailing '/'s before comparing paths with /proc/mounts (3486)
6658 - remove assertions to work around "in-flight rpcs" recovery bug (3063)
6659 - change init script to fail more clearly if not run as root (1528)
6660 - allow clients to reconnect during replay (1742)
6661 - fix ns_lock/i_sem lock ordering deadlock for kms update (3477)
6662 - don't do DNS lookups on NIDs too small for IP addresses (3442)
6663 - re-awaken ptlrpcd if new requests arrive during check_set (3554)
6664 - fix cond_resched (3554)
6665 - only evict unfinished clients after recovery (3515)
6666 - allow bulk resend, prevent data loss (3570)
6667 - dynamic ptlrpc request buffer allocation (2102)
6668 - don't allow unlinking open directory if it isn't empty (2904)
6669 - set MDS/OST threads to umask 0 to not clobber client modes (3359)
6670 - remove extraneous obd dereference causing LASSERT failure (3334)
6671 - don't use get_cycles() when creating temp. files on the mds (3156)
6672 - hold i_sem when setting i_size in ll_extent_lock() (3564)
6673 - handle EEXIST for set-stripe, set proper directory name (3336)
6675 - servers can dump a log evicting a client - lustre.dump_on_timeout=1
6676 - fix ksocknal_fmb_callback() error messages (2918)
6678 2004-05-27 Cluster File Systems, Inc. <info@clusterfs.com>
6681 - don't copy lvb into (possibly NULL) reply on error (2983)
6682 - don't deref dentry after dput, don't free lvb on error (2922)
6683 - use the kms to determine writeback rpc length (2947)
6684 - increment oti_logcookies when osc is inactive (2948)
6685 - update client's i_blocks count via lvb messages (2543)
6686 - handle intent open/close of special files properly (1557)
6687 - mount MDS with errors=remount-ro, like obdfilter (2009)
6688 - initialize lock handle to avoid ASSERT on error cleanup (3057)
6689 - don't use cancelling-locks' kms values (2947)
6690 - use highest lock extent for kms, not last one (2925)
6691 - don't dereference ERR_PTR() dentry in error handling path (3107)
6692 - fix thread race in portals_debug_dumplog() (3122)
6693 - create lprocfs device entries at setup instead of at attach (1519)
6694 - common AST error handler, don't evict client on completion race (3145)
6695 - zero nameidata in detach_mnt in 2.6 (3118)
6696 - verify d_inode after revalidate_special is valid in 2.6 (3116)
6697 - use lustre_put_super() to handle zconf unmounts in 2.6 (3064)
6698 - initialize RPC timeout timer earlier for 2.6 (3219)
6699 - don't dereference NULL reply buffer if mdc_close was never sent (2410)
6700 - print nal/nid for unknown nid (3258)
6701 - additional checks for oscc recovery before doing precreate (3284)
6702 - fix ll_extent_lock() error return code for 64-bit systems (3043)
6703 - don't crash in mdc_close for bad permissions on open (3285)
6704 - zero i_rdev for non-device files (3147)
6705 - clear page->private before handing to FS, better assertion (3119)
6706 - tune the read pipeline (3236)
6707 - fix incorrect decref of invalidated dentry (2350)
6708 - provide read-ahead stats and refine rpc in flight stats (3328)
6709 - don't hold journal transaction open across create RPC (3313)
6710 - update atime on MDS at close time (3265)
6711 - close LDAP connection when recovering to avoid server load (3315)
6712 - update iopen-2.6 patch with fixes from 2399,2517,2904 (3301)
6713 - don't leak open file on MDS after open resend (3325)
6714 - serialize filter_precreate and filter_destroy_precreated (3329)
6715 - loop device shouldn't call sync_dev() for nul device (3092)
6716 - clear page cache after eviction (2766)
6717 - resynchronize MDS->OST in background (2824)
6718 - refuse to mount the same filesystem twice on same mountpoint (3394)
6719 - allow llmount to create routes for mounting behind routers (3320)
6720 - push lock cancellation to blocking thread for glimpse ASTs (3409)
6721 - don't call osc_set_data_with_check() for TEST_LOCK matches (3159)
6722 - fix rare problem with rename on htree directories (3417)
6724 - allow default OST striping configuration per directory (1414)
6725 - fix compilation for qswnal for 2.6 kernels (3125)
6726 - increase maximum number of MDS request buffers for large systems
6727 - change liblustreapi to be useful for external progs like lfsck (3098)
6728 - increase local configuration timeout for slow disks (3353)
6729 - allow configuring ldlm AST timeout - lustre.ldlm_timeout=<seconds>
6731 2004-03-22 Cluster File Systems, Inc. <info@clusterfs.com>
6734 - fixes for glimpse AST timeouts / incorrectly 0-sized files (2818)
6735 - don't overwrite extent policy data in reply if lock was blocked (2901)
6736 - drop filter export grants atomically with removal from device (2663)
6737 - del obd_self_export from work_list in class_disconnect_exports (2908)
6738 - don't LBUG if MDS recovery times out during orphan cleanup (2530)
6739 - swab reply message in mdc_close, other PPC fixes (2464)
6740 - fix destroying of named logs (2325)
6741 - overwrite old logs when running lconf --write_conf (2264)
6742 - bump LLOG_CHUNKSIZE to 8k to allow for larger clusters (2306)
6743 - fix race in target_handle_connect (2898)
6744 - mds_reint_create() should take same inode create lock (2926)
6745 - correct journal credits calculated for CANCEL_UNLINK_LOG (2931)
6746 - don't close files for self_export to avoid uninitialized obd (2936)
6747 - allow MDS with the same name as client node (2939)
6748 - hold dentry reference for closed log files for unlink (2325)
6749 - reserve space for all logs during transactions (2059)
6750 - don't evict page beyond end of stripe extent (2925)
6751 - don't oops on a deleted current working directory (2399)
6752 - handle hard links to targets without a parent properly (2517)
6753 - don't dereference NULL lock when racing during eviction (2867)
6754 - don't grow lock extents when lots of conflicting locks (2919)
6756 2004-03-04 Cluster File Systems, Inc. <info@clusterfs.com>
6759 - account for cache space usage on clients to avoid data loss (974)
6760 - lfsck support in lustre kernel code (2349)
6761 - reduce journal credits needed for BRW writes (2370)
6762 - orphan handling to avoid losing space on client/server crashes
6763 - ptlrpcd can be blocked, stopping ALL progress (2477)
6764 - use lock value blocks to assist in proper KMS, faster stat (1021)
6765 - takes i_sem instead of DLM locks internally on obdfilter (2720)
6766 - recovery for initial connections (2355)
6767 - fixes for mds_cleanup_orphans (1934)
6768 - abort_recovery crashes MDS in b_eq (mds_unlink_orphan) (2584)
6769 - block all file creations until orphan recovery completes (1901)
6770 - client remove rq_connection from request struct (2423)
6771 - conf-sanity test_5, proper cleanup in umount log not availale (2640)
6772 - recovery timer race (2670)
6773 - mdc_close recovey bug (2532)
6774 - ptlrpc cleanup bug (2710)
6775 - mds timeout on local locks (2588)
6776 - namespace lock held during RPCs (2431)
6777 - handle interrupted sync write properly (2503)
6778 - don't try to handle a message that hasn't been replied to (2699)
6779 - client assert failure during cleanup after abort recovery (2701)
6780 - leak mdc device after failed mount (2712)
6781 - ptlrpc_check_set allows timedout requests to complete (2714)
6782 - wait for inflight reqs when ptlrpcd finishes (2710)
6783 - make sure unregistered services are removed from the srv_list
6784 - reset bulk XID's when resending them (caught by 1138 test)
6785 - unregister_bulk after timeout
6786 - fix lconf error (2694)
6787 - handle write after unfinished setstripe, stripe-only getstripe (2388)
6788 - readahead locks pages, leaves pending causing memory pressure (2673)
6789 - increase OST request buffers to 4096 on large machines (2729)
6790 - fix up permission of existing directories in simple_mkdir (2661)
6791 - init deleted item, add assertions ptlrpc_abort_inflight() (2725)
6792 - don't assign transno to errored transactions (2742)
6793 - don't delete objects on OST if given a bogus objid from MDS (2751)
6794 - handle large client PAGE_SIZE readdir on small PAGE_SIZE MDS (2777)
6795 - if rq_no_resend, then timeout request after recovery (2432)
6796 - fix MDS llog_logid record size, 64-bit array alignment (2733)
6797 - don't call usermode_helper from ptlrpcd, DEFAULT upcall (2773)
6798 - put magic in mount.lustre data, check for bad/NULL mount data (2529)
6799 - MDS recovery shouldn't delete objects that it has given out (2730)
6800 - if enqueue arrives after completion, don't clobber LVB (2819)
6801 - don't unlock pages twice when trigger_group_io returns error (2814)
6802 - don't deref NULL rq_repmsg if ldlm_handle_enqueue failed (2822)
6803 - don't write pages to disk if there was an error (1450)
6804 - don't ping imports that have recovery disabled (2676)
6805 - take buffered bytes into account when balancing socknal conn (2817)
6806 - hold a DLM lock over readdir always, use truncate_inode_pages (2706)
6807 - reconnect unlink llog connection after MDS reconnects to OST (2816)
6808 - remove little-endian swabbing of llog records (1987)
6809 - set/limit i_blksize to LL_MAX_BLKSIZE on client (2884)
6810 - retry reposting request buffers if they fail (1191)
6811 - grow extent at grant time to avoid granting a revoked lock (2809)
6812 - lock revoke doesn't evict page if covered by a second lock (2765)
6813 - disable VM readahead to avoid reading outside lock extents (2805)
6815 - return LL_SUPER_MAGIC from statfs for the filesystem type (1972)
6816 - updated kernel patches for hp-2.4.20 kernel (2681)
6818 2004-02-07 Cluster File Systems, Inc. <info@clusterfs.com>
6821 - fix truncated write corruption (2366)
6822 - fix for failed assertion in iopen_connect_dentry (1792,2517)
6824 - don't flag the ptlrpcd thread with PF_MEMALLOC (2636)
6825 - ensure len(uuid) < 37 in lmc (1171)
6826 - fix ia64 OOPS in llog_test (2255)
6827 - zero end of page at obdfilter for partial page writes (2648)
6828 - don't leave stale dentries around after renames (bug 2428)
6829 - fix timeouts when evicting a client with a single lock held (2642)
6830 - set deadline for the initial HELLO message to drain (2634)
6831 - print out dotted-quad IP addresses in the socknal (2302)
6833 - additional debugging for MDS client eviction problem (2443)
6834 - fix mkfsoptions support for osts (2603, 2604)
6836 2004-01-27 Cluster File Systems, Inc. <info@clusterfs.com>
6839 - add series for the vanilla 2.6.0 kernel
6840 - add series for the vanilla 2.4.24 kernel
6841 - add series for a cray x86/64 UL kernel drop
6842 - fix xattr patches for the vanilla 2.4.19 series
6844 - generate true UUIDs in lmc (1171)
6845 - have portals stack dumping break in UML (2466)
6846 - avoid bad dchild deref; avoid inum lock w/o creation (2362)
6847 - allocate with _NOFS in ldlm to avoid deadlock (1933)
6848 - wake callback waiting threads on client eviction (2460)
6849 - Add --ptldebug and --subsystem to lmc (1719)
6850 - update assertion to allow safe interrupt allocation
6851 - set rq_no_resend for cancel requests (2432)
6852 - recalculate ptlrpcd timeout after resend (2494)
6853 - call vfs_rmdir when removing pending directories (2368)
6854 - fix renaming a file to itself (2429)
6855 - lmc creates a default one-stripe lov (2454)
6856 - expand procfs space to handle large clusters (2326)
6857 - increase UML stack to avoid overflow
6858 - update lconf's list of debug and subsystem masks
6859 - fix lfs find --obd (2510)
6860 - /proc tunable for disabling filter read caching (2591)
6861 - stop rpm packages from altering slapd.conf (2301)
6862 - disable nagle in the socknal under 0conf (2578)
6863 - choose mds inode size based on stripe count (2572)
6864 - fix kernel-source rpm problems (2516)
6866 - add --disable-doc to avoid pdf generation (2421)
6867 - update documentation, tests, type-os, comments
6868 - avoid format warnings on ia64
6869 - remove the TOE NAL
6870 - tiny code cleanups by removing unused fields
6872 2004-01-07 Cluster File Systems, Inc. <info@clusterfs.com>
6875 - fix obvious semaphore misuse in as-yet-unused setattr path (2348)
6876 - remove the most blatant lies from BUILDING file (2371)
6877 - change default debug level to reasonable production setting
6878 - reduce client side cache size to reduce cache flush time
6879 - reduce max RPCs in flight to avoid unnecessary file fragmentation
6880 - make TCP zerocopy and pinger support enabled by default (2476)
6881 - sync writes completed after process exits caused crashes (2319)
6882 - maintain correct mount count on the MDS (2356)
6883 - backout 1557, because 2316 wasn't really fixed
6884 - better file I/O statistics gathering in /proc
6885 - don't take unnecessary, deadlock-inducing bug in readpage (2383)
6886 - another kernel patch to fix zero-copy TCP function export
6887 - don't take duplicate lock when processing re-sent getattr (2420)
6888 - lctl uses obd_self_export instead of creating new conn (2353)
6889 - MDS/OST recovery case which requires object creation asserted (2425)
6890 - move lfs from /usr/sbin to /usr/bin in packages
6891 - fix race between mds_client_add and mds_client_free (2417)
6892 - use kmalloc instead of slabs in portals (2430)
6893 - don't create duplicate records when a failover MDS is present (2442)
6894 - remove unnecessary mount age check (2332)
6895 - don't remove directory inodes from locks prematurely (2451)
6896 - don't break if MDS service name is the same as hostname (2103)
6897 - fix races in client write RPC generation when cache full (2482)
6899 2003-12-13 Cluster File Systems, Inc. <info@clusterfs.com>
6902 - remove now-unused request->rq_obd (278)
6903 - if an allocation fails, print out how much memory we've used (1933)
6904 - use PORTAL_SLAB_ALLOC for structures, to get GFP_MEMALLOC (1933)
6905 - add the "configurable stack size" patch to most series files (1256)
6906 - ability to write large log records, for 100+ OST configs (2306)
6907 - fix NULL deref when filter_prep fails (2314)
6908 - fix operator precedence error in filter_sync
6909 - dynamic allocation of socknal TX descriptors (2315)
6910 - fix a missed case in the GFP_MEMALLOC patch, can cause deadlock (2310)
6911 - fix gcc 2.96 compilation problem in xattr kernel patch (2294)
6912 - ensure that CWARN messages in Portals always get to the syslog
6913 - __init/__exit are not for prototype decls (ldlm_init/exit)
6914 - x86-64 compile warning fixes
6915 - fix gateway LMC keyword conflict (2318)
6916 - fix MDS lock inversions in getattr/reint paths (1844)
6917 - fix a rare lock re-ordering bug, which caused deadlock (2322)
6918 - fix i_sem/journal inversion in fsfilt_ext3_write_record (2306)
6919 - DLM race condition prevented some lock evictions (2328)
6920 - ENOMEM detection and retry on socknal sends (2230)
6921 - use GFP_NOFS throughout Lustre, to combat ENOMEM (2230)
6922 - move osc_rpcd into ptlrpc, for use in MDC and others (2329)
6923 - protect MDS inode fsdata with stronger locking; fixes assertion (2313)
6924 - better error messages when a client is rejected during recovery (1505)
6925 - avoid cancelling locks which were never granted, after failure (2330)
6926 - fix i_sem/journal inversion in mds_client_add (2333)
6927 - fix truncate/getattr lock cycle deadlock (2334)
6928 - use rpcd to send close; allows resend after timeout, avoid leak (1897)
6929 - fix two rare exit paths which could leak an l_lock() ref (2321)
6930 - fencepost error in MDS/OST orphan recovery (2226)
6931 - make log record alignment 8 bytes (1988)
6932 - lstripe now fails when requested offset > ost_count (2237)
6933 - ensure that all kernel series have a complete list.h (1607)
6934 - fix crashes in special-file operations (2316)
6935 - lctl create/brw OID mismatch, caused by obsolete filter loop (2339)
6937 - allow configurable automake binary, for testing new versions
6938 - small update to the lfs documentation
6940 2003-12-03 Cluster File Systems, Inc. <info@clusterfs.com>
6942 * fix negative export reference count in fsfilt_sync (2312)
6944 2003-12-01 Cluster File Systems, Inc. <info@clusterfs.com>
6945 * release candidate 0.9.1
6947 - orphans are moved into the PENDING directory for possible recovery
6948 - replayed opens now open by fid for orphan/rename safety (1042)
6949 - last close of an orphan inode generates a transno (683)
6950 - chdir() and mount() now pin the directory entry (1020)
6951 - avoid CERROR in normal ll_setattr_raw() error case (1500)
6952 - discard very old requests without processing them (1502)
6953 - remove some common, well-understood CERRORs (1505)
6954 - require O_DIRECT I/O to be page-sized to workaround IA64 crash (1609)
6955 - clear "grant" flags in OST replies until OST grant code lands (1644)
6956 - fix read performance by not clobbering i_blksize on client (1598)
6957 - fix __ldlm_handle2lock oops by not dereferencing lock after PUT (1625)
6958 - make LRU size a /proc tunable, clears locks when reduced (707)
6959 - fix some lprocfs rot that prevented ptlbd from loading (1732)
6960 - server locks take references on exports now (1558)
6961 - build fixes for 2.4.20-rh trees (1663)
6962 - return an error from lov_create if all OSCs are inactive (1751)
6963 - fix import levels when a reconnect happens without a timeout (1597)
6964 - exit early from mds_open if we get a lookup error (1749)
6965 - partial page read at EOF wouldn't wait for disk before sending (1642)
6966 - avoid NULL deref in obdfilter when reading page past EOF (1592)
6967 - avoid LASSERT in ll_intent_lock if server failed very early (1090)
6968 - fix LBUG in ll_it_open_error with rc = -2 (1861)
6969 - write/truncate lock inversion (1639)
6970 - Don't auto-load obdclass, portals modules during cleanup (1495)
6971 - fix timestamps from jumping to "now" (1763)
6972 - extra journal assertions (1648)
6973 - add an extra multiunlink test (1771)
6974 - fix read_record/write_record API (1776)
6975 - fix leak of offset_extent, possible incorrect i_size later (1772)
6976 - fix lasserts in mis-matched transnos during open-unlink testing (1541)
6977 - Debugging for the kqswnal_get_idle_tx problems (1820)
6978 - Allow recovery to be attempted multiple times (1536)
6979 - Write out MDS last_rcvd file after it is first created (1600)
6980 - Fix tx_descriptor leak in failed transmit situations (1827)
6981 - ext3 journaling fixes for assertion failure after IO error (1871)
6982 - class_export_put() on freed export after completion AST error (1896)
6983 - Fix revalidate looping in VFS (1322)
6984 - Don't access a freed export during MDS_REINT timeout (1521)
6985 - Add open-unlink recovery support on the MDS (1673,1764)
6986 - Return an error if no MDS data was read from last_rcvd (1946)
6987 - Fix for lookup "." or ".." crash on error (1932,1931,1935)
6988 - Don't setup a disk device that doesn't match exported UUID (317)
6989 - Reduce bulk RPC timeout to avoid cascading client/OST failures (1845)
6990 - avoid committing NULL handle in force close
6991 - local.sh is now a one-stripe LOV configuration
6992 - POSIX utime.4 -EPERM on FIFO not owned by user (56)
6993 - fix ext3 htree duplicate directory entry corruption (1516)
6994 - POSIX creat.13, fstat.1, open.18, stat.3 new file atime/mtime (2020)
6995 - update to new LOV EA format (2097)
6996 - interoperability for different PAGE_SIZE/wordsize (686,1821,1343,2042)
6998 2003-06-15 Phil Schwan <phil@clusterfs.com>
7001 - imports and exports cleanup too early, need refcounts (349, 879, 1045)
7002 - per-import/export recovery handling (958, 931, 959)
7003 - multiple last-rcvd slots, for serving multiple FSes (949)
7004 - connections are again shared between multiple imp/exports (963, 964)
7005 - "umount -f" would hang if any requests needed to be sent (393, 978)
7006 - avoid pinning large req buffer by copying for queued messages (989)
7007 - add "uuid" to "lctl device" command to help upcalls (991)
7008 - "open" RPCs with transnos would confuse recovery counters (1037)
7009 - do proper endian conversion of all wire messages (288, 340, 891)
7010 - remove OST bulk get LBUGs, fix ost_brw_write cleanup (1126)
7011 - call waiting locks callback from LDLM recovery thread (1127, 1151)
7012 - fix ptlrpc_connection leak in target_handle_connect (1174)
7013 - fix import refcounting bug in OST and MDS cleanup (1134)
7014 - if an invalid-at-open-time OSC returned before close(), LBUG (1150)
7015 - fix very unlikely obd_types race condition (501)
7016 - remove osc_open hack for echo_client (1187)
7017 - we leaked exports/dlmimps for forcibly disconnected clients (1143)
7018 - a failure in read_inode2 leads to deadlock (1139)
7019 - cancel ack-locks as soon as transaction is committed (1072)
7020 - fix major leaks and crashes in the bulk I/O path (937, 1057)
7021 - make sure to commitrw after any preprw to avoid deadlock (1162)
7022 - failing to execute a file in a lustre FS would lock inode (1203)
7023 - small DEBUG_REQ fix to avoid dereferencing a NULL (1227)
7024 - don't ASSERT while cleaning up an incompletely-setup obd (1248)
7025 - obd_uuid2tgt would walk off the end of the list (1255)
7026 - on IA64 the osc would give portals incorrect bulk size (1258)
7027 - fix debug daemon ioctl interface; allows daemon on ia64 (1274)
7028 - fix lock inversion caused by new llite matching code (1282)
7029 - limit the number of dirty pages on a client to 10MB (1286)
7030 - timed out locks were not being corrected cancelled (1289)
7031 - fix O_DIRECT above 4GB on IA-32 (1292)
7032 * major user-visible changes
7033 - fail out/fail over policy now controlled by the upcall (993)
7035 - add OBD_PING to check server availability and failure (954)
7036 - lustre messages are now sent in sending host order (288, 340, 891)
7037 - add eadatalen to MDS getattr reply (340)
7038 - OST read replies may contain second buffer, with per-page status (593)
7040 2003-03-11 Phil Schwan <phil@clusterfs.com>
7043 - LDLM_DEBUG macro fix, for gcc 3.2 (850)
7044 - failed open()s could cause deadlock; fixed (867, 869)
7045 - stop cancelling OST locks when files are closed (481)
7046 - overlapping XID spaces caused network corruption (851, 853)
7047 - fix unsafe fsfilt counter arithmetic; change to atomic_t
7048 - setattr_raw added, to do single-RPC, server-side setattrs
7049 - lmc/lconf syntax change for OST UUIDs
7050 - fix crashy race condition between ptlrpc_free_req and osc_close
7051 - don't use request in mdc_enqueue if we hit a timeout (889)
7052 - don't set the inode i_size for regular files from the MDS (896)
7053 - handle out of order completion AST (842)
7054 - don't LBUG if a lock request times out after receiving AST (913)
7055 - avoid d_rehash race in ll_find_alias by rehashing inside dcache_lock
7056 - if a bad lock AST arrives, send an error instead of dropping entirely
7057 - return 0 from revalidate2 if ll_intent_lock returns -EINTR (912)
7058 - fix leak in bulk IO when only partially completed (899, 900, 926)
7059 - fix O_DIRECT for ia64 (55)
7060 - (almost) eliminate Lustre-kernel-thread effects on load average (722)
7061 - C-z after timeout could hang a process forever; fixed (977)
7063 - client-side I/O cache (678, 924, 929, 941, 970)
7065 - READPAGE and SETATTRs which don't take server-side locks get
7068 2003-02-11 Phil Schwan <phil@clusterfs.com>
7071 - Fix ldlm_lock_match on the MDS to avoid matching remote locks (592)
7072 - Fix fsfilt_extN_readpage() to read a full page of directory
7073 entries, or fake the remainder if PAGE_SIZE != blocksize (500)
7074 - Avoid extra mdc_getattr() in ll_intent_lock when possible (534, 604)
7075 - Fix imbalanced LOV object allocation and out-of-bound access (469)
7076 - Most intent operations were removed, in favour of a new RPC mode
7077 that does a single RPC to the server and bypasses most of the VFS
7078 - All LDLM resource ID arrays were removed in favour of ldlm_res_id
7079 - Aggressively cancel local locks on DLM servers
7080 - mds_reint_unlink sends EA to the client if it's the last nlink.
7081 client uses that EA to unlink OST objects.
7082 - mds_reint_{rename,unlink,link} were rewritten to take ordered locks
7083 - recursive symlinks were fixed (439)
7084 - fixed NULL deref in DEBUG_REQ
7085 - filter_update_lastobjid no longer calls sync, which annoyed extN
7086 - fixed multi-client small-writes to a single file problem (445)
7087 - fixed mtime updates during file writes (607)
7088 - fixed vector writes on obdfilter causing problems when ENOSPC (670)
7089 - fixed bug in obd_brw_read/write() (under guise of testing 367)
7090 - fixed Linux OST size reporting problem (444, 656)
7091 - OST now updates object mtime with writes or setattr (607, 619)
7092 - client verifies file size before zeroing page past EOF (445)
7093 - OST now writes last allocated objid to disk with allocation (108)
7094 - LOV on echo now works (409)
7096 - mds_reint_unlink sends a new buffer, with the EA included. this
7097 buffer is only valid if body->valid & OBD_MD_FLEASIZE, which is only
7098 set if a regular file was being unlinked, and it was the last link
7099 - use PtlGet from the target for bulk writes (315)
7100 - OST now updates object mtime with writes or setattr (607, 619)
7101 - LDLM now has a grant-time callback to revalidate locked items, if
7103 - Many MDS operations were reorganized to combat race conditions
7105 - Merge b_intel branch (updated lprocfs code) - now at /proc/fs/lustre
7106 - configure check to avoid gcc version 2.96 20000731-2.96-98) (606)
7108 2003-01-06 Andreas Dilger <adilger@clusterfs.com>
7111 - Fully reactivate OST imports after reconnection (512, others)
7112 - Make sure client sees our -ENOTCONN from mds_handle (513 - partial)
7113 - More graceful error handling for truncating on dead OST (515)
7114 - Don't error out unless we're actually accessing dead stripes (474)
7115 - Fix garbage sizes when stripes are missing (410)
7116 - LRU counters were broken, causing constant lock purge (433, 432)
7117 - garbage on read from stripes with failed OSTs (441)
7118 - mark OSCs as active before reconnecting during recovery (438)
7119 - lov_enqueue and lov_cancel need to handle inactive OSTs (403)
7120 - lfind did not preserve OST order in output (443)
7121 - symlinks cause hung clients, incorrect data (439)
7122 - stop dereferencing request after dropping refcount (457)
7123 - don't LASSERT(spin_is_locked) on non-SMP (455)
7124 - fixes for many rename() bugs
7125 - fstat didn't correctly synchronize attributes (399)
7126 - server must handle lock cancellation during blocking AST prep (487)
7127 - bulk descriptors were free()d too soon (511)
7128 - fix paths in lconf, which would load incorrect modules (451, 507)
7129 - fix confusing lconf 'host not found' error message (386)
7130 - fix lock order deadlock on OST (O/R i_sem before journal ops, 478)
7131 - fix race condition in mdc_blocking_ast() for inode access (526)
7132 - fix lov_unpackmd() unpacking wrong number of stripes (537)
7133 - fix lov_set_osc_active() marking wrong OSC inactive (440)
7134 - fix bad lstripe lov_unpackmd() assertion (fix layering too) (527)
7135 - fix multiple writes of stripe MD to MDS (358, maybe 519)
7136 - fix lstripe in several ways (kernel side) (527)
7137 - fix request leak in ldlm_cli_enqueue (262)
7138 - incorrect OSC was marked inactive after OST failure
7139 - call mds_fs_cleanup before unmounting filesystem (524)
7140 - fix races between taking ns_lock and ldlm_lock_change_resource
7141 - fix races updating LOV export open file list
7142 - fix lov_enqueue error path, avoid decref-ing bad lock handle
7143 - fix recovery NULL deref in ldlm_cli_cancel_unused
7144 - fix some DLM races by using new hash table for lock handles (419)
7145 - permit the client to specify desired inodes, at replay
7146 - duplicate requests when we queue them for replay reintegration
7147 - fix last_rcvd offset calculation
7148 - sync after each recovered transaction, so we always make progress
7149 - never, not always, ERESTART requests without transnos
7150 - store the lov_desc in the MDS, so we don't depend on getlovinfo to
7152 - skip replay if the MDS says that the client is already connected
7153 - don't check for a recovery-enabled export to match lctl's UUID
7154 - don't INC_USE_COUNT for phantom exports
7155 - don't crash when cleaning up phantom exports (567)
7156 - don't double-finish or set replay data for errored mdc_open requests
7157 - abort requests when they time out, so we don't get old replies
7158 - send/receive replies for AST messages again
7159 - if the client says that it doesn't have the lock, cancel it on the
7161 - if we timeout during I/O, don't try to cancel an in-use lock; instead
7162 mark it as destroyed, it will all work out when decref is called
7163 - fix module use counts (22, 581)
7165 - ASTs now expect a reply (server cancels lock on error reply)
7167 2002-12-02 Andreas Dilger <adilger@clusterfs.com>
7170 - fix many simultaneous client startup (392)
7171 - fix dentry->d_it clobbering
7172 - credentials weren't being shipped for readdir/getattr operations
7173 - remove invalid assertions triggered during some concurrent MD
7175 - proper Lustre versions added (336, 389)
7176 - fix memory leak for create error case (398)
7177 - fix LOV locking bug that would get cli/srv out of sync
7178 - fix echo client over LOV (409)
7179 - fix dbench 2, extN refcount problem (170, 258, 356, 418)
7180 - fix double-O_EXCL intent crash (424)
7181 - avoid sending multiple lock CANCELs (352)
7183 - MDS can do multi-client recovery (modulo bugs in new code)
7185 - many updates, edits, cleanups
7187 2002-11-18 Phil Schwan <phil@clusterfs.com>
7190 - fix null d_it dereference (346)
7191 - fix full OST/dbench hang (333)
7192 - fix permission problem with file removal (286)
7193 - fix removal of OSCs from LOV when they fail
7194 - fix NULL deref during bulk timeout (214)
7195 - fix problems related to multiple filesystems on one MDS (241)
7196 - fixed serious subtle metadata locking bugs
7197 - free locks on clients when inodes are removed due to memory
7199 - fix inode pointer in lock data (285)
7200 - partial support for multiple MDS on a single host (241)
7201 - data locks weren't cancelled at clear_inode time (290, 311)
7202 - intent locks could lead to unbounded lock growth (205)
7203 - added a maximum lock count, an LRU list, and a flusher
7204 - fix multiple rename (365)
7205 - properly abstracted the echo client
7206 - OSC locked 1 byte too many; fixed
7207 - rewrote brw callback code:
7208 - fixed recovery bugs related to LOVs (306)
7209 - fixed too-many-pages-in-one-write crash (191)
7210 - fixed (again) crash in sync_io_timeout (214)
7211 - probably fixed callback-related race (385)
7213 - Add capability to MDS protocol
7214 - LDLM cancellations and callbacks on different portals
7216 2002-10-28 Andreas Dilger <adilger@clusterfs.com>
7219 - limit client IOV size to PTL_MD_MAX_IOV (611336, 191)
7220 - defer open object destruction to close time (601981, 138)
7221 - open/close OST file handle in obdo (OBD_MD_FLHANDLE) (601981, 138)
7222 - move LDLM_ENQUEUE/CONVERT back to MDS portal (625069)
7223 - abstract ll_lookup2, fix ll_revalidate2 to use abstraction (256)
7224 - don't call obd_setattr in ll_file_release for destroyed objects
7225 * protocol change to lustre_msg: move |version| and add |flags|
7226 * protocol change to osc_punch: "start" in "o_size", "end" in "o_blocks"
7227 * lock replay: for LDLM_FL_REPLAY trust client to do right thing
7228 * added replay of create, unlink, link and rename operations during
7229 MDS failover; recovery should be much more robust now
7230 * remove failed OSCs from LOVs (only lov_create uses this so far)
7231 * the lustre-HOWTO was brought (more) up to date (582544)
7233 2002-10-23 Phil Schwan <phil@clusterfs.com>
7236 - in-use dentries weren't being reused properly (617851)
7237 - prevent multiple LDLM setup (599178)
7238 - fix LOV size calculations for truncate (617853)
7239 - fix client handling of MDS intent errors (POSIX)
7240 - fix permission bug in lovstripe.c test (624321)
7241 - fix MDS thread deadlock - move LDLM handler to DLM portal (625069)
7242 - truncate past end of file could corrupt data
7243 - proper cleanup after timeouts, crashes, etc (592524, 550815)
7244 - a race in recovery could return ETIMEDOUT to apps (623947)
7245 - building outside the source directory was fixed
7246 * the lustre-HOWTO was brought (more) up to date (582544)
7247 * major progress was made on recovery functionality
7249 2002-10-10 Phil Schwan <phil@clusterfs.com>
7252 - recovery deadlock fix
7253 - rm -rf causes LBUG fix (617817)
7254 - file open by multiple tasks fix (618962)
7255 - directory permissions bugs (602707 and 620007)
7256 - journal_stop fixed with locking (611313)
7257 - O_APPEND failures resolved (618273, perhaps 614459)
7258 - lconf PATH fix (619770)
7259 - IA64 build fix (621450)
7260 - RPC buffer sizes scale with amount of memory
7262 2002-10-01 Phil Schwan <phil@clusterfs.com>
7265 - locks would be cancelled without throwing away data pages,
7266 resulting in inconsistent data (605627)
7267 - inode attributes were not always being refreshed (605627, 612449)
7268 - lconf now continues to cleanup after lctl reports an error
7269 - MDS now enforces user permissions (602707)
7270 - lprocfs cleanup fixed, but not yet enabled (614157)
7271 - fixed infinite server hang, should a client not respond to an AST
7272 - avoid going into recovery if user calls readlink() with a buffer
7273 that's too small (613941)
7274 - AST RPCs no longer require replies (614867) -- this may be changed
7275 - don't crash server if client sends an IOV that's too big (611336)
7276 - fixed lock conversion deadlock (611892)
7277 - fixed the following of symlinks (614622)
7278 * recovery: the server can remove locks from a client that dies, other
7279 clients can make progress
7280 * more extN patch fixes
7281 * compile-time configurable ptlrpc buffer allocations
7283 - collaborative read cache document
7284 - Lustre Lite Performance CDR document-in-progress
7286 2002-09-20 Andreas Dilger <adilger@clusterfs.com>
7289 - fix typo in patch-2.4.18
7291 2002-09-20 Andreas Dilger <adilger@clusterfs.com>
7294 - clear ptlrpc request each time in handle_incoming_request()
7295 - unlink of files now destroys the object on the OST
7297 2002-09-19 Peter Braam <braam@clusterfs.com>
7299 * add hard link support
7300 * change obdfile creation method
7301 * kernel patch changed
7303 2002-09-19 Peter Braam <braam@clusterfs.com>
7306 - stack overflow bug in extN fixed
7308 2002-09-18 Andreas Dilger <adilger@clusterfs.com>
7310 * documentation updates
7311 - add man pages for config tools
7312 - update tests/README to describe testing with new config tools
7313 - finish metadata API descriptions
7314 * bug fixes and cleanups
7315 - statfs workaround for 16TB limit
7316 - LOV stripe allocation improved, can stripe on subset of OSTs
7317 - LOV file size/IO offset was wrong for files > 4GB in size
7318 - object EA data was being dropped, caused files to be unreadable
7319 - memory overflow with non-LOV OST caused memory corruption
7320 - fixed regression tests to work with new config tools, obdfilter
7321 - fixed bug when directory size became larger than 1 block
7322 - fixed bug (for single client case) when PWD was deleted
7323 - invalidate local directory pages when doing intent-based ops
7324 - avoid LDLM oops when lock callback contained bad data
7326 2002-09-09 Andreas Dilger <adilger@clusterfs.com>
7328 * documentation updates
7329 * bug fixes and cleanups
7330 - configuration tools
7333 - 64-bit compile warnings
7334 - 64-bit internal statfs data
7336 * test_brw on persistent OST devices
7338 * lprocfs (disabled)
7340 2002-09-04 Andreas Dilger <adilger@clusterfs.com>
7342 * documentation updates
7343 * bug fixes and cleanups
7344 * configuration tools
7346 2002-08-30 Peter J. Braam <braam@clusterfs.com>
7349 * many small fixes to 0_5_4
7350 * io/network handling
7351 * thinkos in MDS operations
7353 2002-08-24 Peter J. Braam <braam@clusterfs.com>
7356 * crucial basic fixes to 0.5.3
7357 * IOR, Iozone work over Elan
7360 2002-08-07 Phil Schwan <phil@clusterfs.com>
7361 * version 0_5_3, our first alpha
7362 * we use the new Portals iovs
7363 * documentation updates
7364 * bug fixes and cleanups
7365 * small changes in the DLM wire protocol
7367 2002-07-25 Peter J. Braam <braam@clusterfs.com>
7368 * version 0_5_1 with some initial stability,
7369 * locking on MD and file I/O.
7370 * documentation updates
7371 * several bug fixes since 0.5.0
7372 * small changes in wire protocol
7374 2002-07-18 Phil Schwan <phil@clusterfs.com>
7376 * delivered as Lustre Light Alpha
7377 * fixed a crash after handling invalid MDS requests
7378 * fixed directory pages for architectures with non-4k pages sizes
7380 2002-07-11 Andreas Dilger <adilger@clusterfs.com>
7381 * release version v0_4_4
7382 * Moves TCP acceptor to be on port 2432 (unused Coda port) instead
7384 * Fixes a number of interruption problems with OST operations.
7385 * Update documentation for portals header changes
7386 * Move all wire protocol structs/defines to lustre_idl.h
7387 * Fixes symlink length bug.
7388 * Add tcpdump to repository.
7390 2002-07-05 Andreas Dilger <adilger@clusterfs.com>
7391 * release version v0_4_3
7392 * Fixes statfs for inodes on extN.
7393 * Fixes bug in runtests which would delete /etc/hosts.
7394 * Use 64-bit object IDs wherever possible (not into VFS though)
7395 Remove ost_get_info, which is unused by lustre, and out of date.
7397 2002-07-03 Peter Braam <braam@clusterfs.com>
7398 * release version v0_4_2 Fixes a lookup error (type not passed)
7399 * move forward to head of Portals
7400 * move forward to latest Lustre kernel
7402 2002-06-25 Peter Braam <braam@clusterfs.com>
7403 * release version v0_4_1. Hopefully stable on single node use.
git://git.whamcloud.com / fs / lustre-release.git / blob