4 2.6.16.60-0.42.8 (SLES 10),
5 2.6.27.39-0.3.1 (SLES11),
6 2.6.18-194.3.1.el5 (RHEL 5)
7 2.6.18-194.3.1.0.1.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 Severity : enhancement
17 Description: Update RHEL5.5 kernel to 2.6.18-194.3.1.el5 and OEL5.5 kernel to
18 2.6.18-194.3.1.0.1.el5.
22 Description: mds_orphan_add_link()) error linking orphan to PENDING
23 Details : quota limits might disallow linking orphans to PENDING
24 when unlinking a file - temporary raise threads' privileges
25 when processing unlinks.
27 Severity : enhancement
29 Description: add conf-param -d option to remove permanent settings.
30 Details : Add the ability to remove permanent lctl conf_param settings.
31 (Previously conf_param settings could only be changed, not
32 removed.) This also provides a method to change failover
33 nid locations. Improve lctl man page.
35 Severity : enhancement
37 Description: add list_param to b1_8 and add "-R" option to list params
40 Severity : enhancement
42 Description: lfs quota output is not very convenient for awk/sed-parsing
43 Details : Some positions in lfs quota output table could be empty or
44 non-empty which made it hard to parse it with scripts, now a dash
45 is put instead of space where there is not supposed to be any data.
47 Severity : enhancement
49 Description: fix obdfilter-survey script to work properly with remote oss-s
51 Severity : enhancement
53 Description: add new OBDFILTER_SURVEY test suite
55 Severity : enhancement
57 Description: add new multiple mount protection (MMP) test suite
59 Severity : enhancement
61 Description: add support for async journal commit in echo client
63 Severity : enhancement
65 Description: allow userland programs to include <lustre/lustre_idl.h>
66 from stardard include directories
68 Severity : enhancement
70 Description: The prune-icache-use-trylock is no longer needed now that
71 the patch from bug 20008 is landed.
75 Description: The shrink grant feature is still active on the client although the
76 connect flag is not set.
80 Description: Don't leak grant space if the write failed with quota exceeded.
84 Description: a race condition could lead to SIGBUS being sent to an
85 application using mmap-ped files from Lustre
86 Details : truncate_complete_page implementation for the patchless
87 client could arbitrarily unset PG_Uptodate flag for a
88 page being kicked from the page cache, an uptodate check
89 right after a readpage call in filemap_fault could fail
90 because of that as though the page read had been unsuccessful.
94 Description: dlm lock slab shrinking is not efficient
95 Details : The dlm_locks slab can grow significantly and consumes a lot of
96 memory on the server. Set a hardlimit to grant_plan.
100 Description: Lustre does not do 1MB IOs to HW RAID
101 Details : Bump MAX_PHYS/HW_SEGMENTS and SG_ALL to 256 in the RHEL5 kernel.
102 This is what we do already for SLES kernels.
106 Description: LSI Fusion MPT driver hacks to improve performance
107 Details : Set CONFIG_FUSION_MAX_SGE to 256 for RHEL5
109 Severity : enhancement
111 Description: increase default md stripe_cache_size to 16k
114 Bugzilla : 15587/21439
115 Description: don't handle security.capability xattr
116 Details : CONFIG_SECURITY_FILE_CAPABILITIES is enabled by default on SLES11.
117 This results in additional getxattr calls, causing VBR test
118 failures as well as a preformance drop when writing.
122 Description: obdfilter-survey is no longer working
123 Details : revert patch from bug 20355 to resolve an issue with lctl --threads
124 not working correctly with $(PTHREAD_LIBS) being linked to lctl.
128 Description: ll_shrink_cache does not handle __GFP_FS properly
132 Description: lfs getstripe shows wrong info for directories
133 Details : Set correct LOVEA default values for filesystem-wide.
137 Description: FSX checksum false positves due to mmap IO
138 Details : Use OBD_FL_MMAP flag for IOs on a memory mapped file. Do not print
139 checksum errors, if the flag is set on a request.
143 Description: file operations after eviction have successful return values
144 Details : use vfs ->flush callback to return any pending async errors
149 Description: mdsrate fails to write after 1.3+M files opened
150 Details : decrease memory usage on clients by recycling dentries and
155 Description: obdfilter-survey gives unreasonably high numbers
156 Details : Wait for all threads to complete when running test_brw.
160 Description: do not set lustre read_only device when server umount and keep
161 client records for recoverable ones
165 Description: move sync_on_lock_cancel tunable to the obdfilter layer
166 Details : move the tunable to trigger a journal flush on lock cancel from
167 the ost layer to the obdfilter layer. This tunable is useful
168 when using the async journal commit feature.
172 Description: exp->exp_nid_stats == NULL in filter_tally()
173 Details : fix race with per-nid stats by delaying procfs cleanup until
178 Description: extent lock cancellation on client can keep the cpu busy for too
183 Description: Do not fail OST activation when a llog is not found, just
184 issue an error message.
188 Description: Don't enable extents by default for MDT.
192 Description: Protect bitfield access to ptlrpc_request's rq_flags, since
193 the AT code can access it concurrently while sending early
196 -------------------------------------------------------------------------------
198 2010-04-30 Oracle, Inc.
200 * Support for kernels:
201 2.6.16.60-0.42.8 (SLES 10),
202 2.6.27.39-0.3.1 (SLES11),
203 2.6.18-164.11.1.el5 (RHEL 5)
204 2.6.18-164.11.1.0.1.el5 (OEL 5)
205 * Client support for unpatched kernels:
206 (see http://wiki.lustre.org/index.php?title=Patchless_Client)
207 2.6.16 - 2.6.30 vanilla (kernel.org)
208 * Recommended e2fsprogs version: 1.41.10-sun2
209 * The async journal commit feature (bug 19128) and the cancel
210 lock before replay feature (bug 16774) are disabled by default.
214 Description: open-unlinked directories trigger MDS LBUG
215 Details : Fix regression introduced by the patch from bug 19640.
216 ext3_inc_count() can reset nlink to 1 when the directory
217 is indexed and inode->i_nlink == 2. Work around the problem
218 by incrementing nlink by 2 instead of 1.
222 Description: Reconnects are not throttled
223 Details : Don't wake up pinger on reconnect failures and rely on regular
224 pings to trigger the next reconnection. Please note that the
225 pinger already uses a smaller interval if the import is
229 Frequency : only with NFS export
231 Description: Console flooded with error message from ll_inode_from_lock()
233 Details : in mds_open, initialize the child_res_id before enqueuing
234 the OPEN lock for the child inode, then to avoid senting
235 wrong ldlm_res_id to client.
239 Description: allow multiple instances of the same nid in NID hash
240 Details : Case of multiple separate clients from the same NID (as
241 with liblustre) is legitimate and so we should allow
242 multiple instances of the same NID in nid hash.
246 Description: the readahead code can sleep on a semaphore while holding a
248 Details : in ras_update, "lov_get_info" could be called during increasing
249 readahead windows, which tries to get the mutex lock "lov_lock"
250 while holding the spin_lock "ras_lock", then causes system
255 Description: ASSERTION(cli->cl_avail_grant >= 0) failed
256 Details : fix assertion failure in the grant code.
260 Description: Use CNETERR (which is rate limited) in specific places in
261 the portal's LNET driver to avoid flooding the console.
265 Description: include last created object in precreate slow case
269 Description: don't do rep-ack if not created anything
270 Details : mds_open currently always put a lock into a rep-ack regardless
271 if something was created or not. This is pointless and only
272 creates needless contention. In fact the entire idea was to do
273 this for real creates as a recovery protection.
277 Description: Spurious error messages from smp_processor_id() on preemptible
279 Details : Disable a preemption by grabbing the lock in fs_trace_get_tcd()
280 first. The function fs_trace_get_tcd() was moved up.
284 Description: interval_erase() fix
285 Details : interval_erase() calls update_maxhigh() properly when child
290 Description: "lfs df" does not print stats for all mountpoints
291 Details : Print all mounted lustre filesystems with "lfs df"
295 Description: lfs setstripe -p not longer work with a relative pathname
296 Details : Use realpath() to provide absolute pathname.
300 Description: fix for truncated reply buffer
301 Details : reply buffer could be referred by reply_in_callback after released
306 Description: lustre.lov error when backing up symlinks with extended attributes
307 Details : Improved logic in ll_listxattr()
311 Description: properly handle null value for setattr -n lustre.lov
312 Details : Running "setfattr -n trusted.lov ." causes a NULL dereference
313 in ll_setxattr() due to no checking if "value" is NULL.
314 This command now resets to the default striping when executed
319 Description: stack overflow on lock cancellation due to fsync call
320 Details : sync_on_lock_cancel is needed for recovery when async journal
321 is enabled, but we actually just need to make sure that
322 metadata blocks have hit the journal, so doing a fs sync
323 should be enough and should consume less stack (just create an
324 empty handle and commmit it).
328 Description: using current->journal_info to store per-thread data leads
329 to problem under memory pressure
330 Details : disable the per-thread data (current->journal_info) containing
331 the lock info during I/O to work around the issue for short tem
335 Description: control DCACHE_LUSTRE_INVALID flag with MDS_INODELOCK_LOOKUP lock
336 Details : DCACHE_LUSTRE_INVALID is controlled by MDS_INODELOCK_LOOKUP
337 lock which is corresponding to "IT_LOOKUP", do not skip invalidate
342 Description: Cannot send after transport shutdown
343 Details : Clear imp_vbr_failed flag upon eviction
347 Description: soft lock in request set code during recovery
348 Details : during recovery, uses req->rq_set itself to replay the request
349 instead of ptlrpcd_recovery_pc
353 Description: Use CFS_ALLOC_IO instead of _STD in llap_from_page_with_lockh
354 Details : During an ll_readahead under ll_readpage, we have seen the the
355 OBD_SLAB_ALLOC hang under ldlm_pools_shrink when trying to lock
356 a page that is already locked by the readahead code.
361 Description: stop waitting for next replay transno if shutdown
362 Details : if the system is shutting down, wake up service thread blocked
363 to wait for next replay transno during recovery, then all the
364 references held by queued requests can be dropped and device
369 Description: lov_merge_lvb()) ASSERTION(spin_is_locked(&lsm->lsm_lock)) failed
370 Details : Protect lli->lli_smd pointer updates with lli->lli_lock.
374 Description: per-nid stats should not access lustre-hash internal structures
379 Description: mount.lustre fails to pass some options to mount()
384 Description: ext4 extent allocation is slower than in ext3
385 Details : Increase the default value of MB_DEFAULT_ORDER2_REQS to 8,
386 enlarge ext4 preallocation table for 2048 4K blocks extents
391 Description: incorrect triggering of synchronous IO
392 Details : The OSC can mistakenly fall back to synchronous IO when the
393 max_dirty_mb limit is reached and no write requests have yet
394 been issued. This can occur when the dirty pages are spread
395 over many files all of which are below the optimal request size.
399 Description: Optimize quota_ctl operations by sending requests in parallel
400 Details : Send MDS->OST quota_ctl requests in parallel, do not resend.
401 Compiled from two attachments in the ticket.
405 Description: ordering issue between transaction start & i_mutex
406 Details : start the transaction earlier in llog_lvfs_destroy to get
407 transaction start and inode mutex lock nested properly.
411 Description: lru resize SLV can get stuck
412 Details : calculate SLV with a greater precision to not lose small
413 changes due to interger math truncation; round up SLV only
414 if the amount of granted locks less than the limit to not
415 get stuck with this SLV
419 Description: avoid divide-by-zero in lprocfs_rd_import()
423 Description: lfs quota failed when OSTs are down
424 Details : really return approximate block/inode usage when OSTs are down
428 Description: llobdstat fix and enhancement
429 Details : add a counter to set a limit to how many samples will be returned
430 fix a wildcard in the path to limit to obdfilter stats only
434 Description: recovery-small 51 hang - MDT can not stop, Mount still busy with
436 Details : abort lock enqueue processing sooner when we umount is in progress.
438 Severity : enhancement
440 Description: Update RHEL5.4 kernel to 2.6.18-164.11.1.el5 and OEL5.4 kernel to
441 2.6.18-164.11.1.0.1.el5.
443 Severity : enhancement
445 Description: error message improvements
446 Details : Use INFO/WARN instead of WARN/ERROR for the slow messages.
447 Simplify MDT/OST service start message.
448 Suppress "changing the import ..." warning.
451 Bugzilla : 21961/17914
452 Description: ignore trailing -mdc when determining index number
455 Frequency : only with SLES11
457 Description: fix for a race condition in linux quotas implementation
458 Details : dq_flags(struct dquot) access is not properly locked which could
459 lead to certain inconsistencies when accessing it using non-atomic
460 bit operations like __set_bit in do_set_dqblk.
464 Description: return any pending async errors in close(2) using flush callback
466 -------------------------------------------------------------------------------
468 2010-01-29 Sun Microsystems, Inc.
470 * Support for kernels:
471 2.6.16.60-0.42.8 (SLES 10),
472 2.6.27.39-0.3.1 (SLES11),
473 2.6.18-164.11.1.el5 (RHEL 5)
474 2.6.18-164.6.1.0.1.el5 (OEL 5)
475 * Client support for unpatched kernels:
476 (see http://wiki.lustre.org/index.php?title=Patchless_Client)
477 2.6.16 - 2.6.30 vanilla (kernel.org)
478 * Recommended e2fsprogs version: 1.41.6.sun1
479 * The async journal commit feature (bug 19128) and the cancel
480 lock before replay feature (bug 16774) are disabled by default.
482 Severity : enhancement
484 Description: Update RHEL5.4 kernel to 2.6.18-164.11.1.el5 and
485 OEL5.4 kernel to 2.6.18-164.11.1.0.1.el5.
487 Severity : enhancement
488 Bugzilla : 21511/19848
489 Description: Update SLES11 kernel to 2.6.27.39-0.3.1.
491 Severity : enhancement
493 Description: Update supported SLES10 kernel to 2.6.16.60-0.42.8.
495 Severity : enhancement
497 Description: Update kernel to RHEL5.4 2.6.18-164.6.1.el5 and
498 OEL5 2.6.18-164.6.1.0.1.el5(Both in-kernel OFED enabled).
500 Severity : enhancement
502 Description: Build kernels (RHEL5, OEL5 and SLES10/11) using the vendor's own
505 Severity : enhancement
507 Description: Vanilla kernel 2.6.30 patchless client support.
512 Description: bad entry in directory xxx: inode out of bounds
513 Details : fix locking issue in the rename path which could race with any
514 other operations updating the same directory.
516 Severity : enhancement
518 Description: Make watchdog timer messages to be more clear and descriptive.
522 Description: cp -p command does not preserve the dates and timestamp
523 Details : mtime could be spoiled by a write callback
527 Description: Clear imp_force_reconnect correctly in ptlrpc_connect_interpret()
531 Description: Allow non-root access for "lfs check".
532 Details : Added a check in obd_class_ioctl() for OBD_IOC_PING_TARGET.
534 Severity : enhancement
536 Description: quotacheck performance/scaling issues
537 Details : reduce quotacheck time on empty filesystem by skipping uninit
540 Severity : enhancement
542 Description: Enhancement for lfs(1) command to use numeric uid/gid.
544 Severity : enhancement
546 Description: Adjust locks' extents on their first enqueue, so that at the time
547 they get granted, there is no need for another pass through the
548 queues since they are already shaped into the proper forms.
552 Description: Fix mds_shrink_intent_reply()/mds_intent_policy() to pass correct
553 arguments and prevent LBUG() in lustre_shrink_reply_v2().
557 Description: Change tunefs.lustre and mkfs.lustre --mountfsoptions so that
558 exactly the specified mount options are used. Leaving off
559 any "mandatory" mount options is an error. Leaving off any
560 default mount options causes a warning, but is allowed.
561 Change errors=remount-ro from mandatory to default. Sanitize
562 the mount string before storing it. Update man pages accordingly.
566 Description: mds_getattr() should return 0, even if mds_fid2entry() fails with
567 -ENOENT. Also fix in ptlrpc_expire_one_request() to print signed
570 Severity : enhancement
572 Description: Remove set_info(KEY_UNLINKED) from MDS/OSC
574 Severity : enhancement
576 Description: Clients can replay thousands of unused locks during recovery
577 Details : Don't replay unused locks (only read locks for now) during
578 recovery. This feature is disabled by default and can be
579 enabled by running the following command on the clients:
580 lctl get_param ldlm.cancel_unused_locks_before_replay
584 Description: can't stat file in some situation.
585 Details : improve initialize osc date when target is added to mds and
586 ability to resend too big getattr request is client isn't have info
591 Description: Prevent inconsistences between linux and lustre mount structures.
592 Details : Wait indefinitely in server_wait_finished() until mnt_count drops.
593 Make the sleep interruptible.
595 Severity : enhancement
597 Description: Communicate OST degraded/readonly state via statfs to MDS
598 Details : Flags in the statfs returned from OSTs indicate whether the
599 OST is in a degraded RAID state, or if the filesystem has
600 turned read-only after a filesystem error is detected.
605 Description: don't panic if EPROTO was hit when reading symlink
606 Details : correctly handling request reference in error cases.
611 Description: open sometimes returns ENOENT instead of EACCES
612 Details : checking permission should be part of open part of mds_open, not
613 lookup part. so server should be set DISP_OPEN_OPEN disposition
614 before starting permission check.
615 Also not need revalidate dentry if client already have LOOKUP lock.
619 Description: enable client interface failover
620 Frequency : on servers with multiple network interfaces
621 Details : When a child reconnects from another NID, properly update export
622 nid hash position and ldlm reverse import.
624 Severity : enhancement
626 Description: implemented direct I/O with arbitrary (nonaligned) memory
627 addresses and file offsets.
629 Severity : enhancement
631 Description: added more recovery timeout options.
633 Severity : enhancement
635 Description: added llapi_file_open, llapi_file_create, llapi_file_get_stripe
640 Description: Avoid deadlock for local client writes
641 Frequency : only on systems with clients writing to an OST on the same node
642 Details : Use new OBD_BRW_MEMALLOC flag to notify OST about writes in the
643 memory freeing context. This allows OST threads to set the
644 PF_MEMALLOC flag on task structures in order to allocate memory
645 from reserved pools and complete IO.
646 Use GFP_HIGHUSER for OST allocations for non-local client writes,
647 so that the OST threads generate memory pressure and allow
648 inactive pages to be reclaimed.
653 Descriptoin: lock ordering violation between &cli->cl_sem and _lprocfs_lock
654 Details : move ldlm namespace creation in setup phase to avoid grab
655 _lprocfs_lock with cli_sem held.
659 Frequency : only during format of test systems
660 Description: Unable to run several mkfs.lustre on loop devices at the same time
661 Details : mkfs.lustre returns error 256 on the concurrent loop devices
662 formatting. The solution is to proper handle the error.
664 Severity : enhancement
666 Description: implement async create (obd_async_create) method for osc, to avoid
667 too long waiting new ost objects with holding ldlm lock.
671 Frequency : occasionally during network problems
672 Description: client not allowed to reconnect to OST because of active request
673 Details : abort bulk requests received by the OST once the client has timed
674 out since the client will resend the request anyway. The client
675 also now retries to reconnect to the same server if a connect
676 request failed with EBUSY or -EAGAIN.
679 Frequency : rare, if used wide striped file and one ost in down.
681 Descriptoin: don't return error if we created a subset of objects for file.
682 Details : lov_update_create_set() uses set->set_success as index for created
683 objects, so if some requests failed, they will have hole at end of
684 array and we can use qos_shrink_lsm for allocate correct lsm.
688 Description: Slow stale export processing during normal start up
689 Details : The global mgc lock prevents OST setup to be run in parallel.
690 Replace the global lock with a per-config_llog_data semaphore.
694 Description: Out or order replies might be lost on replay
695 Details : In ptlrpc_retain_replayable_request if we cannot find retained
696 request with tid smaller then one currently being added, add it
697 to the start, not end of the list.
701 Description: BUG: soft lockup - CPU#1 stuck for 10s! [ll_mdt_07:4523]
702 Details : add cond_resched() calls to avoid hogging the cpu for too long
703 in the hash code. Make also lustre_hash_for_each_empty() more
706 Severity : enhancement
708 Description: Performance improvements for debug messages with D_RPCTRACE,
709 D_LDLM, D_QUOTA options.
712 Frequency : only with NFS export
714 Description: (lov_merge.c:74:lov_merge_lvb())
715 ASSERTION(spin_is_locked(&lsm->lsm_lock)) failed (SR 71691004)
716 Details : Fix a race in the nfs export code by populating inode
717 info while the new inode is still locked
719 Severity : enhancement
721 Description: add a new file in procfs called force_lbug. Writting to this
722 file triggers a LBUG. Only for test purpose.
726 Description: OOM killer causes node hang
727 Details : really interrupt the sleep in osc_enter_cache on signals
731 Description: LustreError: 9153:0:(quota_context.c:622:dqacq_completion()) LBUG
732 Details : fix race during quota release on the slave.
734 Severity : enhancement
736 Description: smaller hash bucket sizes, cleanups
737 Details : increase hash table sizes and enabled rehashing for pools, quota,
738 uuid, nid & per-nid stats.
740 Severity : enhancement
742 Description: Add ldiskfs maxdirsize mount option
743 Details : add max_dir size mount option
747 Description: panic in ll_statahead_thread
748 Details : prevent parent thread to be killed before its child
751 Frequency : only with 16TB device
753 Description: unable to perform "mount -t lustre" of 16TB OST device
754 Details : Mounting 16TB LUNs failed due to three bugs in mkfs.lustre.
758 Description: ASSERTION(atomic_read(&imp->imp_inflight) == 0) failed
759 Details : unregistering should be zero if no RPC inflight.
763 Description: hyperion: Oops during metabench
764 Details : Correct the refcount of lov_request_set
766 Severity : enhancement
768 Description: Add mptlinux and nxge drivers to Lustre builds
770 Severity : enhancement
772 Description: Fix watchdog timer message to be more clear
773 Details : Make watchdog timer messages more clear and descriptive.
777 Description: LNET soft lockups in socknal_cd thread
778 Details : don't hog CPU for active-connecting if another connd is
779 accepting connecting-requst from the same peer
783 Description: recovery-small test_17 hang
784 Details : Land several AT improvements & fixes.
788 Description: MDS panic and hanging client processes
789 Details : Replace exp_ops_stats with exp_nid_stats->nid_stats
793 Description: OSS stuck in recovery.
794 Details : fix race during recovery. class_unlink_export,
795 class_set_export_delayed and target_queue_last_replay_reply
796 may race while increasing/decreasing obd_recoverable_clients
797 and obd_delayed_clients, causing recovery to wait forever.
799 Severity : enhancement
801 Description: add cascading_rw.c to lustre/tests
805 Description: filter_last_id() NULL deref
806 Details : lprocfs_filter_rd_last_id() should check for the fully
807 setup obd device, before proceeding further.
809 Severity : enhancement
811 Description: Loadgen improvements
812 Details : stacksize and locking fixes for loadgen
816 Description: Quiet CERROR("dirty %d > system dirty_max %d\n"
817 Details : The atomic_read() allowing the atomic_inc() are not covered
818 by a lock. Thus they may safely race and trip this CERROR()
819 unless we add in a small fudge factor (+1).
821 Severity : enhancement
823 Description: shrink_slab: nr=-9223362083340912175
824 Details : fix spurious message from shrink_slab reporing negative nr
828 Description: Quiet bogus previously committed transno error
829 Details : suppress the "server went back in time" error message which
830 is always printed even in the common case after a client eviction
832 Severity : enhancement
834 Description: Parallel statfs() calls result in client eviction
835 Details : cache statfs data for 1s.
839 Description: parallel-scale test_compilebench: @@@@@@ FAIL: compilebench
841 Details : fix serveral issues in pinger code causing clients not to ping
842 servers for too long, resulting in evictions.
846 Description: e2fsck should warn when MMP update interval is extended
847 Details : print mmp_check_interval and make it possible to abort mount
848 operation in case it takes too long.
852 Description: mdsrate-create-large.sh, BUG: soft lockup - CPU#0 stuck for 10s!
853 Details : fix bug in the RHEL5's jbd2 callback patch.
857 Description: drop number of active requests when queued for recovery
858 Details : Now that we take a reference on the original request instead of
859 making a copy of it for recovery. We need to drop the number of
860 active requests or the queued requests will prevent all request
861 processing when they exceed (srv->srv_threads_running - 1).
863 Severity : enhancement
865 Description: refuse to invalidate operational quota files when they are in use
866 Details : an attempt to invalidate operational quota files on the quota
867 master is not actually permitted by VFS (returning -EPERM), but we
868 should not depend on that and should return the error earlier.
872 Description: Applications stuck in jbd2_log_wait_commit during exit
873 Details : fix deadlock between kjournald2 trying to acquire the page lock
874 owned by an ost_io thread waiting for journal commit.
876 -------------------------------------------------------------------------------
878 2009-10-16 Sun Microsystems, Inc.
880 * Support for kernels:
881 2.6.16.60-0.42.4 (SLES 10),
882 2.6.27.29-0.1 (SLES11, i686 & x84_64 only),
883 2.6.18-128.7.1.el5 (RHEL 5),
884 2.6.18-128.7.1.el5 (OEL 5),
885 * Client support for unpatched kernels:
886 (see http://wiki.lustre.org/index.php?title=Patchless_Client)
887 2.6.16 - 2.6.27 vanilla (kernel.org)
888 * Recommended e2fsprogs version: 1.41.6.sun1
889 * ext4 support for RHEL5 is experimental and thus should not be
892 Severity : enhancement
894 Description: Add OEL5 support.
896 Severity : enhancement
898 Description: Update kernel to SLES11 2.6.27.29-0.1.
902 Description: File checksum failures with OST read cache on
903 Details : Disable page poisoning when the bulk transfer has to be aborted
904 because the client got evicted.
908 Description: Don't allow make backward step on assiging osc next id.
909 Details : race between allocation next id and ll_sync thread can be cause
910 of set wrong osc next id and can be kill valid ost objects.
912 Severity : enhancement
914 Description: Update kernel to RHEL5 2.6.18-128.7.1.el5.
916 Severity : enhancement
918 Description: Update kernel to SLES10 SP2 2.6.16.60-0.42.4.
922 Description: Changes in raid5-large-io-rhel5.patch to calculate sectors properly
926 Description: Increase the default BLK_DEF_MAX_SECTORS value for RHEL5 and SLES11
930 Description: Error handling in osc_statfs_interpret() has been improved.
931 Details : Check in osc_statfs_interpret() for EBADR.
935 Description: Do not send statfs() requests to OSTs disabled by administrator.
936 Details : Check in lov_prep_statfs_set() for non-NULL ltd_exp.
940 Description: Do not update ctime for the deleted inode.
941 Details : Check in mds_reint_unlink() before calling fsfilt_setattr().
945 Description: Increase of the size of the LDLM resource hash.
946 Details : Bump up RES_HASH_BITS=12.
950 Description: correctly send lsm on open replay
951 Details : MDS is trust to LSM size on replay open, but client can set wrong size
956 Description: Deadlock between filter_destroy() and filter_commitrw_write().
957 Details : filter_destroy() does not hold the DLM lock over the whole
958 operation. If the DLM lock is dropped, filter_commitrw() can go
959 through, causing the deadlock between page lock and i_mutex.
960 The i_alloc_sem should also be hold in filter_destroy() while
965 Description: truncate starts GFP_FS allocation under transaction causing deadlock
966 Details : ldiskfs_truncate calls grab_cache_page which may start page
967 allocation under an open transaction. This may lead to
968 calling prune_icache with consequent lustre reentrance.
971 Frequency : only when down/upgrading the MDS to 1.6/1.8 while 1.8 clients are
972 still up and when the OST pool feature is used
974 Description: interop testing got LBUG when run dd with OST pool
975 :LustreError: 30032:0:(llite_lib.c:1913:ll_replace_lsm()) LBUG
976 Details : down/upgrading the MDS to a version that doesn't/does support OST
977 pool can cause clients to crash because the lsm has changed
982 Description: missing tree_status on 1.8.1 RPM build
983 Details : make rpms failed due because the tree_status file is missing.
987 Description: continuing LustreError "mds adjust qunit failed!"
988 Details : don't print message on the console when ->adjust_qunit fails.
992 Description: don't increase ldlm timeout if previous client was evicted
993 Details : if a client doesn't respond to a blocking callback within the
994 adaptive ldlm enqueue timeout, don't adjust the adaptive estimate
995 when the lock is next granted.
999 Description: ost is being unmounted w/o all writes to last_rcvd landing on disk.
1000 affects recovery negatively.
1001 Details : make sure all exports have been properly destroyed by the zombie
1002 thread processed before stopping the target.
1006 Description: Performance degradation with O_DIRECT between 1.6 & 1.8.1 b190
1007 Details : disable write barrier for ext4/SLES11.
1011 Description: Kernel panic - not syncing: Out of memory and no killable
1012 processes... on OSS when iozone
1013 Details : fix memory leak in the journal checksum patch.
1017 Description: group quota "too many blocks" OSS crashes
1018 Details : we should keep the same uid/gid for lquota_chkquota() and
1019 lquota_pending_commit()
1023 Description: LustreError: 9153:0:(quota_context.c:622:dqacq_completion()) LBUG
1024 Details : don't LBUG on release quota error. Just a workaround until the
1025 problem is understood.
1027 ------------------------------------------------------------------------------
1028 2009-07-31 Sun Microsystems, Inc.
1030 * Support for kernels:
1031 2.6.16.60-0.39.3 (SLES 10),
1032 2.6.27.23-0.1 (SLES11, i686 & x84_64 only),
1033 2.6.18-128.1.14.el5 (RHEL 5),
1034 * Client support for unpatched kernels:
1035 (see http://wiki.lustre.org/index.php?title=Patchless_Client)
1036 2.6.16 - 2.6.27 vanilla (kernel.org)
1037 * Recommended e2fsprogs version: 1.41.6.sun1
1038 * File join has been disabled in this release, refer to Bugzilla 16929.
1039 * NFS export disabled when stack size < 8192. Since the NFSv4 export
1040 of Lustre filesystem with 4K stack may cause a stack overflow. For
1041 more information, please refer to bugzilla 17630.
1042 * ext4 support for RHEL5 is experimental and thus should not be
1045 Severity : enhancement
1047 Description: Update kernel to SLES10 SP2 2.6.16.60-0.39.3.
1050 Frequency : with 1.8 server and 1.6 clients
1052 Description: correctly shrink reply for avoid send too big message to client.
1053 Details : 1.8 mds is allocate to big buffer to LOV EA data and this produce
1054 some problems with sending this reply to 1.6 client.
1058 Description: Repeated atomic allocation failures.
1059 Details : Use GFP_HIGHUSER | __GFP_NOMEMALLOC flags for memory allocations
1060 to generate memory pressure and allow reclaiming of inactive pages.
1061 At the same time, do not allow to exhaust emergency pools.
1062 For local clients the use of GFP_NOFS will be introduced in 1.8.2
1064 Severity : enhancement
1065 Bugzilla : 19846, 18289
1066 Description: Update kernel to RHEL5 2.6.18-128.1.14.el5.
1068 Severity : enhancement
1069 Bugzilla : 19625, 16893, 18668, 19848
1070 Description: Add support for SLES11 2.6.27.23-0.1.
1072 Severity : enhancement
1074 Description: Update client support to vanila kernels up to 2.6.27.
1076 Severity : enhancement
1078 Description: Update kernel to SLES10 SP2 2.6.16.60-0.37.
1080 Severity : enhancement
1082 Description: Compile with -Werror by default for i686 and x86_64.
1086 Description: resolve race between obd_disconnect and class_disconnect_exports
1087 Details : if obd_disconnect will be called to already disconnected export he
1088 forget release one reference and osc module can't unloaded.
1090 Severity : enhancement
1092 Description: move AT tunable parameters for more consistent usage
1093 Details : add AT tunables under /proc/sys/lustre, add to conf_param parsing
1097 Descriptoin: correctly skip time estimate if in recovery
1098 Details : rq_send_state insn't bitmask so using bitwise ops is forbid.
1102 Descriptoin: OSS DeadLock
1103 Details : Use trylock to prevent deadlock when shrink icache.
1105 Severity : enhancement
1107 Description: Allow tuning service thread via /proc
1108 Details : For each service a new
1109 /proc/fs/lustre/{service}/*/thread_{min,max,started} entry is
1110 created that can be used to set min/max thread counts, and get the
1111 current number of running threads.
1113 Severity : enhancement
1115 Description: Add state history info file, enhance import info file
1116 Details : Track import connection state changes in a new osc/mdc proc file;
1117 add overview-type data to the osc/mdc import proc file.
1121 Description: Reduce small size read RPC
1122 Details : Set read-ahead limite for every file and only do read-ahead when
1123 available read-ahead pages are bigger than 1M to avoid small size
1128 Description: free_entry erroneously used groups_free instead of put_group_info
1130 Severity : enhancement
1132 Description: Make read-ahead stripe size aligned.
1134 Severity : enhancement
1136 Description: MDS create should not wait for statfs RPC while holding DLM lock.
1139 Frequency : rare, connect and disconnect target at same time
1141 Descriptoin: ASSERTION(atomic_read(&imp->imp_inflight) == 0
1142 Details : don't call obd_disconnect under lov_lock. this long time
1143 operation and can block ptlrpcd which answer to connect request.
1146 Frequency : start MDS on uncleanly shutdowned MDS device
1148 Descriptoin: ll_sync thread stay in waiting mds<>ost recovery finished
1149 Details : stay in waiting mds<>ost recovery finished produce random bugs
1150 due race between two ll_sync thread for one lov target. send
1151 ACTIVATE event only if connect realy finished and import have
1155 Frequency : start MDS on uncleanly shutdowned MDS device
1157 Descriptoin: aborting recovery hang on MDS
1158 Details : don't throttle destroy RPCs for the MDT.
1162 Description: Slow reads beyond 8Tb offsets.
1163 Details : Page index integer overflow in ll_read_ahead_page
1167 Description: Soft lockup on OSS after MDS failover
1168 Details : MSG_CONNECT_INITIAL is not set on the initial MDS->OST connect.
1169 As a conseqence, the patch from bug 18224 is not operational
1170 and the MDS export cannot be reused on the OSTs until it gets
1174 Frequency : rare, only if using MMP with Linux RAID
1176 Description: MMP doesn't work with Linux RAID
1177 Details : While using HA for Lustre servers with Linux RAID, it is possible
1178 that MMP will not detect multiple mounts. To make this work we
1179 need to unplug the device queue in RAID when the MMP block is being
1180 written. Also while reading the MMP block, we should read it from
1181 disk and not the cached one.
1184 Frequency : rare, during recovery
1186 Description: Assertion failure in ldlm_lock_put
1187 Details : Do not put cancelled locks into replay list, hold references on
1188 locks in replay list
1192 Description: 1.6.5 mdsrate performance is slower than 1.4.11/12 (MDS is not cpu bound!)
1193 Details : create_count always drops to the min value (=32) because grow_count
1194 is being changed before the precreate RPC completes.
1197 Frequency : Only in RHEL5 when mounting multiple ext3 filesystems
1200 Description: "kmem_cache_create: duplicate cache jbd_4k" error message
1201 Details : add proper locking for creation of jbd_4k slab cache
1205 Description: MMP check in ext3_remount() fails without displaying any error
1206 Details : When multiple mount protection fails during remount, proper error
1211 Description: Rare Client crash on resend if the file was deleted.
1212 Details : When file is opened, but open reply is lost and file is
1213 subsequently deleted before resend, resend processing logic
1214 breaks trying to open the file again, should not try to open.
1218 Description: add check for >8TB ldiskfs filesystems
1219 Details : ext3-based ldiskfs does not support greater than 8TB LUNs.
1220 Don't allow >8TB ldiskfs filesystems to be mounted without
1221 force_over_8tb mount option
1225 Description: Client locked up when running multiple instances of an app. on
1226 multiple mount points
1227 Details : ll_shrink_cache() can sleep while holding the ll_sb_lock.
1228 Convert ll_sb_lock to a read/write semaphore to fix the problem.
1232 Description: Cannot acces an NFS-mounted Lustre filesystem
1233 Details : An NFS client cannot access the Lustre filesystem NFS-mounted
1234 from a Lustre-client exporting the Lustre filesystem via NFS.
1238 Description: panic in ll_statahead_thread
1239 Details : grab dentry reference in parent process.
1242 -------------------------------------------------------------------------------
1244 tbd Sun Microsystems, Inc.
1246 * Support for kernels:
1247 2.6.16.60-0.31 (SLES 10),
1248 2.6.18-128.1.6.el5 (RHEL 5),
1249 2.6.22.14 vanilla (kernel.org)
1250 * Client support for unpatched kernels:
1251 (see http://wiki.lustre.org/index.php?title=Patchless_Client)
1252 2.6.16 - 2.6.22 vanilla (kernel.org)
1253 * Recommended e2fsprogs version: 1.40.11-sun1
1254 * File join has been disabled, refer to Bugzilla 16929.
1255 * A new Lustre ADIO driver is available for MPICH2-1.0.7.
1256 * NFS export disabled when stack size < 8192. Since the NFSv4 export of
1257 Lustre filesystem with 4K stack may cause a stack overflow. For more
1258 information, please refer to bugzilla 17630.
1260 Severity : enhancement
1262 Description: Update to RHEL5.3 kernel-2.6.18-128.1.6.el5.
1264 Severity : enhancement
1266 Description: Update OFED release to 1.4.1 RC4
1268 Severity : major, only with big OST
1270 Description: Very poor metadata performance on Infiniband lustre configuration
1271 Details : OST object precreation becomes very slow on big OSTs. This is due
1272 to the ialloc patch spending too much time scanning groups.
1275 Frequency : during recovery
1277 Description: don't mix llog inodes with normal.
1278 Details : allocate inodes for log in last inode group
1283 Description: fix lqs' reference which won't be put in some situations
1284 Details : This patch fixes:
1285 1. In quota_check_common(), this function will check quota
1286 for user and group, but only send one return via "pending".
1287 In most cases, the pendings should be same. But that is not
1289 2. If quotaoff runs between lquota_chkquota() and
1290 lquota_pending_commit(), the same thing will happen too.
1291 That is why it comes:
1292 - if (!ll_sb_any_quota_active(qctxt->lqc_sb))
1295 -------------------------------------------------------------------------------
1297 2008-12-31 Sun Microsystems, Inc.
1299 * Support for kernels:
1300 2.6.5-7.314 (SLES 9),
1301 2.6.9-67.0.22.EL (RHEL 4),
1302 2.6.16.60-0.31 (SLES 10),
1303 2.6.18-92.1.17.el5 (RHEL 5),
1304 2.6.22.14 vanilla (kernel.org)
1305 * Client support for unpatched kernels:
1306 (see http://wiki.lustre.org/index.php?title=Patchless_Client)
1307 2.6.16 - 2.6.22 vanilla (kernel.org)
1308 * Client support for unpatched kernels:
1309 we do not recommend using patchless RHEL4 clients with kernels
1310 prior to 2.6.9-55EL (RHEL4U5).
1311 * Recommended e2fsprogs version: 1.40.11-sun1
1312 * Note that reiserfs quotas are disabled on SLES 10 in this kernel.
1313 * RHEL 4 and RHEL 5/SLES 10 clients behaves differently on 'cd' to a
1314 removed cwd "./" (refer to Bugzilla 14399).
1315 * A new quota file format has been introduced in 1.6.5.
1316 The format conversion from prior releases is handled transparently,
1317 but releases older than 1.4.12/1.6.5 don't understand this new
1318 format. The automatic format conversion can be avoided by running
1319 the following command on the MDS:
1320 'tunefs.lustre --param="mdt.quota_type=ug1" $MDTDEV'.
1321 For more information, please refer to bugzilla 13904.
1322 * A new quota file format was introduced in 1.6.6/1.8.0 (kernels 2.6.16+).
1323 The format conversion from prior releases is handled transparently,
1324 but releases older than 1.6.6/1.8.0 don't understand this new
1325 format. The automatic format conversion can be avoided by running
1326 the following commands on the MDS and OSS servers (for
1327 pre 1.4.12-1.6.5 quota files):
1328 'tunefs.lustre --param="mdt.quota_type=ug1" $MDTDEV',
1329 'tunefs.lustre --param="ost.quota_type=ug1" $MDTDEV'
1330 or (for 1.4.12/1.6.5 quota files)
1331 'tunefs.lustre --param="mdt.quota_type=ug2" $MDTDEV',
1332 'tunefs.lustre --param="ost.quota_type=ug2" $MDTDEV'
1333 For more information, please refer to bugzilla 13904.
1334 * Output of lfs quota has been made less detailed by default,
1335 old (verbose) output can be obtained by using -v option.
1336 * File join has been disabled in this release, refer to Bugzilla 16929.
1337 * A new Lustre ADIO driver is available for MPICH2-1.0.7.
1338 * NFS export disabled when stack size < 8192. Since the NFSv4 export of
1339 Lustre filesystem with 4K stack may cause a stack overflow. For more
1340 information, please refer to bugzilla 17630.
1342 Severity : enhancement
1344 Description: Caching OSS
1345 Details : introduce data caching on the OSS. The OSS now relies on the linux
1346 kernel page cache to keep recently accessed data in memory.
1347 It is worth noting that all write requests are still flushed
1348 synchronously as in lustre 1.6.
1350 Severity : enhancement
1352 Description: Version based recovery
1353 Details : introduce finer grained recovery able to detect transaction
1354 dependencies and can deal with transaction gaps caused by clients
1355 failing at the same time as the server.
1357 Severity : enhancement
1359 Description: Enable adaptive timeouts by default
1360 Details : The Lustre timeout value in /proc/sys/lustre/timeout is now
1361 managed dynamically based on server load and should not need
1362 to be tuned manually based on cluster size. This allows Lustre
1363 to work under a wider variety of system sizes and loads, without
1364 unnecessarily causing lengthy recovery times.
1366 Severity : enhancement
1368 Description: Add OST Pools support
1369 Details : File striping can now be set to use an arbitrary pool of OSTs
1371 Severity : enhancement
1373 Description: add lazystatfs mount option to allow statfs(2) to skip down OSTs
1374 Details : allow skip disconnected ost for send statfs request and hide error
1378 Frequency : rare, on llog test 6
1380 Descriptoin: don't allow connect to already connected import
1381 Details : allowing connect to already connected import is hide connecting problem.
1384 Frequency : rare, on failed llog setup
1386 Descriptoin: don't leak obd reference on failed llog setup
1387 Details : for failed llog setup - mgc forget call class_destroy_import for
1388 client import, move destroy import to more generic place.
1393 Descriptoin: allow kill process which wait statahead result
1394 Details : for some reasons 'ls' can stick in waiting result from statahead,
1395 in this case need way for kill this process.
1398 Frequency : rare, at shutdown
1400 Descriptoin: panic at umount
1401 Details : llap_shrinker can be raced with killing super block from list and
1402 this produce panic with access to already freeded pointer
1407 Descriptoin: don't lose wakeup for imp_recovery_waitq
1408 Details : recover_import_no_retry or invalidate_import and import_close can
1409 both sleep on imp_recovery_waitq, but we was send only one wakeup
1415 Descriptoin: panic in mds_open
1416 Details : don't confuse mds_finish_transno() with PTR_ERR(-ENOENT)
1421 Descriptoin: stuck in cache_remove_extent() or panic with accessing to already
1423 Details : release lock refernce only after add page to pages list.
1426 Frequency : always with long access acl
1428 Descriptoin: mds can't pack reply with long acl.
1429 Details : mds don't control size of acl but they limited by reint/getattr
1434 Frequency : on remount
1436 Description: external journal device not working after the remount
1437 Details : clear dev_rdonly flag for external journal devices in
1443 Description: shutdown vs evict race
1444 Details : client_disconnect_export vs connect request race.
1445 if client will evicted at this time - we start invalidate
1446 thread without referece to import and import can be freed
1452 Description: shrink LOV EAs before replying
1453 Details : correctly adjust LOV EA buffer for reply.
1458 Description: don't skip ost target if they assigned to file
1459 Details : Drop slow OSCs if we can, but not for requested start idx.
1460 This means "if OSC is slow and it is not the requested
1461 start OST, then it can be skipped, otherwise skip it only
1462 if it is inactive/recovering/out-of-space.
1464 Severity : enhancement
1466 Description: Update to RHEL5 kernel-2.6.18-92.1.17.el5.
1469 Frequency : rare, need acl's on inode.
1471 Description: client can't handle ost additional correctly
1472 Details : if ost was added after client connected to mds client can have
1473 hit lnet_try_match_md ... to big messages to wide striped files.
1474 in this case need teach client to handle config events about add
1475 lov target and update client max ea size at that event.
1477 Severity : enhancement
1479 Description: Update to sles9 kernel-2.6.5-7.314.
1481 Severity : enhancement
1483 Description: Update to SLES10 SP2 kernel-2.6.16.60-0.31.
1486 Frequency : Create a symlink file with a very long name
1488 Description: ldlm_cancel_pack()) ASSERTION(max >= dlm->lock_count + count)
1489 Details : If there is no extra space in the request for early cancels,
1490 ldlm_req_handles_avail() returns 0 instead of a negative value.
1495 Description: mds is deadlocked
1496 Details : in rare cases, inode in catalog can have i_no less than have parent
1497 i_no, this produce wrong order for locking during open, and
1498 parallel unlink can be lock open. this need teach mds_open to grab
1499 locks in resource id order, not at parent -> child order.
1501 Severity : enhancement
1503 Description: Add /proc entry for import status
1504 Details : The mdc, osc, and mgc import directories now have
1505 an import directory that contains useful import data for debugging
1506 connection problems.
1508 Severity : enhancement
1510 Description: Re-disable certain /proc logging
1511 Details : Enable and disable client's offset_stats, extents_stats and
1512 extents_stats_per_process stats logging on the fly.
1515 Frequency : Only on FC kernels 2.6.22+
1517 Description: oops in statahead
1518 Details : Do not drop reference count for the dentry from VFS when lookup,
1519 VFS will do that by itself.
1521 Severity : enhancement
1523 Description: Generic /proc file permissions
1524 Details : Set /Proc file permissions in a more generic way to enable non-
1525 root users operate on some /proc files.
1529 Description: Hitting mdc_commit_close() ASSERTION
1530 Details : Properly handle request reference release in
1531 ll_release_openhandle().
1535 Frequency : only patchless client
1536 Description: add workaround for race between add/remove dentry from hash
1538 Severity : enhancement
1540 Description: Allow OST glimpses to return PW locks
1544 Description: LBUG when llog conf file is full
1545 Details : When llog bitmap is full, ENOSPC should be returned for plain
1550 Description: Prevent import from entering FULL state when server in recovery
1554 Description: service mount cannot take device name with ":"
1555 Details : Only when device name contains ":/" will mount treat it as
1561 Description: replace ptlrpcd with the statahead thread to interpret the async
1562 statahead RPC callback
1566 Frequency : on recovery
1567 Description: I/O failures after umount during fail back
1568 Details : if client reconnected to restarted server we need join to recovery
1569 instead of find server handler is changed and process self eviction
1570 with cancel all locks.
1572 Severity : enhancement
1574 Description: Update to RHEL5 kernel-2.6.18-92.1.10.el5.
1578 Description: Kernel BUG tries to release flock
1579 Details : Lustre does not destroy flock lock before last reference goes
1580 away. So always drop flock locks when client is evicted and
1581 perform unlock regardless of successfulness of speaking to MDS.
1583 Severity : enhancement
1585 Description: Update to SLES10 SP2 kernel-2.6.16.60-0.27.
1587 Severity : enhancement
1589 Description: Upcall on Lustre log has been dumped
1590 Details : Allow for a user mode script to be called once a Lustre log has
1591 been dumped. It passes the filename of the dumped log to the
1592 script, the location of the script can be specified via
1593 /proc/sys/lnet/debug_log_upcall.
1598 Description: avoid messages about idr_remove called for id that is not allocated
1599 Details : Move assigment s_dev for clustered nfs to end of initialization,
1600 for avoid problem with error handling.
1605 Description: avoid Already found the key in hash [CONN_UNUSED_HASH] messages
1606 Details : When connection is reused this not moved from CONN_UNUSED_HASH into
1607 CONN_USED_HASH and this prodice warning when put connection again
1613 Description: avoid ASSERTION(client_stat->nid_exp_ref_count == 0) failed
1614 Details : release reference to stats when client disconnected, not
1615 when export destroyed for avoid races when client destroyed
1616 after main ost export.
1620 Description: more cleanup in mds_lov
1621 Details : add workaround for get valid ost count for avoid warnings about
1622 drop too big messages, not init llog cat under semphore which
1623 can be blocked on reconnect and break normal replay, fix access
1626 Severity : enhancement
1628 Description: Export bytes_read/bytes_write count on OSC/OST.
1632 Description: Early reply size mismatch, MGC loses connection
1633 Details : Apply the MGS_CONNECT_SUPPORTED mask at reconnect time so
1634 the connect flags are properly negotiated.
1638 Description: Properly propagate oinfo flags from lov to osc for statfs
1639 Details : restore missing copy oi_flags to lov requests.
1643 Description: exports in /proc are broken
1644 Details : recreate /proc entries for clients when they reconnect.
1646 Severity : enhancement
1648 Description: Add man pages for llobdstat(8), llstat(8), plot-llstat(8),
1649 : l_getgroups(8), lst(8), routerstat(8)
1650 Details : included man pages for llobdstat(8), llstat(8),
1651 : plot-llstat(8), l_getgroups(8), lst(8), routerstat(8)
1653 Severity : enhancement
1655 Description: Implement lustre ll_show_options method.
1657 Severity : enhancement
1659 Description: Update to SLES9 kernel-2.6.5-7.312.
1661 Severity : enhancement
1663 Description: Update to RHEL4 kernel-2.6.9-67.0.22.EL.
1667 Description: exports in /proc are broken
1668 Details : recreate /proc entries for clients when they reconnect.
1672 Description: don't fail open with -ERANGE
1673 Details : if client connected until mds will be know about real ost count
1674 get LOV EA can be fail because mds not allocate enougth buffer
1679 Description: Resolve device initialization race
1680 Details : Prevent proc handler from accessing devices added to the
1681 obd_devs array but yet be intialized.
1683 Severity : enhancement
1685 Description: configure's --enable-quota should check the
1686 : kernel .config for CONFIG_QUOTA
1687 Details : configure is terminated if --enable-quota is passed but
1688 : no quota support is in kernel
1690 Severity : enhancement
1692 Description: Update to SLES10 SP2 kernel-2.6.16.60-0.23.
1694 Severity : enhancement
1696 Description: Update to RHEL5 kernel-2.6.18-92.1.6.el5.
1700 Frequency : rare, on PPC clients
1701 Description: don't swab ost objects in response about directory, because
1703 Details : bug similar bug 14856, but in different function.
1705 Severity : enhancement
1707 Description: lfs quota tool enhancement
1708 Details : added units specifiers support for setquota, default to
1709 current uid/gid for quota report, short quota stats by
1710 default, nonpositional parameters for setquota, added
1711 llapi_quotactl manual page.
1713 Severity : enhancement
1715 Description: *optional* service tags registration
1716 Details : if the "service tags" package is installed on a Lustre node
1717 When the filesystem is mounted, a local-node service tag will
1718 be created. See http://inventory.sun.com/ for more information
1719 about the Service Tags asset management system.
1723 Description: Client runs out of low memory
1724 Details : Consider only lowmem when counting initial number of llap pages
1727 Frequency : occasional
1729 Description: add refcount for osc callbacks, so avoid panic on shutdown
1731 Severity : enhancement
1733 Description: Update to RHEL4 kernel-2.6.9-67.0.20.
1736 Frequency : testing only
1738 Description: sanity test 65a fails if stripecount of -1 is set
1739 Details : handle -1 striping on filesystem in ll_dirstripe_verify
1742 Frequency : only in unusual configurations
1744 Description: Kernel panic with find ost index.
1745 Details : lov_obd have panic if some OST's have sparse indexes.
1747 Severity : enhancement
1749 Description: Update to RHEL5 kernel-2.6.18-53.1.21.el5.
1752 Frequency : rarely, if filesystem is mounted with -o flock
1754 Description: do not process already freed flock
1755 Details : flock can possibly be freed by another thread before it reaches
1756 to ldlm_flock_completion_ast.
1759 Frequency : rarely, if filesystem is mounted with -o flock
1761 Description: LBUG during stress test
1762 Details : Need properly lock accesses the flock deadlock detection list.
1765 Frequency : rarely, if binaries are being run from Lustre
1767 Description: oops in page fault handler
1768 Details : kernel page fault handler can return two special 'pages' in
1769 error case, don't try dereference NOPAGE_SIGBUS and NOPAGE_OMM.
1772 Frequency : rarely, during shutdown
1774 Description: timeout with invalidate import.
1775 Details : ptlrpcd_check call obd_zombie_impexp_cull and wait request which
1776 should be handled by ptlrpcd. This produce long age waiting and
1777 -ETIMEOUT ptlrpc_invalidate_import and as result LASSERT.
1783 Description: ASSERTION(CheckWriteback(page,cmd)) failed
1784 Details : badly clear PG_Writeback bit in ll_ap_completion can produce false
1788 Frequency : only with broken builds/installations
1790 Description: no LBUG if lquota.ko and fsfilt_ldiskfs.ko are different versions
1791 Details : just return an error to a user, put a console error message
1793 Severity : enhancement
1795 Description: Update to RHEL5 kernel-2.6.18-53.1.19.el5.
1797 Severity : enhancement
1799 Description: Update to RHEL4 kernel-2.6.9-67.0.15.
1801 Severity : enhancement
1803 Description: enable MGS and MDT services start separately
1804 Details : add a 'nomgs' option in mount.lustre to enable start a MDT with
1805 a co-located MGS without starting the MGS, which is a complement
1806 to 'nosvc' mount option.
1809 Frequency : always, on big-endian systems
1811 Description: cleanup in ptlrpc code, related to PPC platform
1812 Details : store magic in native order avoid panic's in recovery on PPC
1813 node and forbid from this error in future. Also fix posibily
1814 of twice swab data. Fix get lov striping to userpace.
1817 Frequency : rarely, if replay get lost on server
1819 Description: server incorrectly drop resent replays lead to recovery failure.
1820 Details : do not drop replay according to msg flags, instead we check the
1821 per-export recovery request queue for duplication of transno.
1824 Frequency : after recovery
1826 Description: precreate to many object's after del orphan.
1827 Details : del orphan st in oscc last_id == next_id and this triger growing
1828 count of precreated objects. Set flag LOW to skip increase count
1829 of precreated objects.
1832 Frequency : rare, on clear nid stats
1834 Description: ASSERTION(client_stat->nid_exp_ref_count == 0)
1835 Details : when clean nid stats sometimes try destroy live entry,
1836 and this produce panic in free.
1839 Frequency : occasionally since 1.6.4
1841 Description: Stack overflow during MDS log replay
1842 Details : ease stack pressure by using a thread dealing llog_process.
1845 Frequency : very rare
1847 Description: MDT cannot be unmounted, reporting "Mount still busy"
1848 Details : Mountpoint references were being leaked during open reply
1849 reconstruction after an MDS restart. Drop mountpoint reference
1850 in reconstruct_open() and free dentry reference also.
1855 Description: wait until IO finished before start new when do lock cancel.
1856 Details : VM protocol want old IO finished before start new, in this case
1857 need wait until PG_writeback is cleared until check dirty flag
1858 and call writepages in lock cancel callback.
1863 Description: mds_mfd_close() ASSERTION(rc == 0)
1864 Details : In mds_mfd_close(), we need protect inode's writecount change
1865 within its orphan write semaphore to prevent possible races.
1868 Frequency : rare, on shutdown ost
1870 Description: don't hit live lock with umount ost.
1871 Details : shrink_dcache_parent can be in long loop with destroy dentries,
1872 use shrink_dcache_sb instead.
1875 Frequency : only when echo_client is used
1877 Description: don't panic with use echo_client
1878 Details : echo client pass NULL as client nid pointer and this produce NULL
1879 pointer dereference.
1882 Frequency : Always on 32-bit PowerPC systems
1884 Description: fix build on PPC32
1885 Details : compile code with -m64 flag produce wrong object file for PPC32.
1890 Description: MDS LBUG: ASSERTION(!IS_ERR(dchild))
1891 Details : In reconstruct_* functions, LASSERTs on both the data supplied
1892 by a client, and the data on disk are dangerous and incorrect.
1893 Change them with client eviction.
1895 Severity : enhancement
1897 Description: skiplist implementation simplification
1898 Details : skiplists are used to group compatible locks on granted list
1899 that was implemented as tracking first and last lock of each lock
1900 group the patch changes that to using doubly linked lists
1904 Description: delete compatibility for 32bit qdata
1905 Details : as planned, when lustre is beyond b1_8, lquota won't support 32bit
1906 qunit. That means servers of b1_4 and servers of b1_8 can't be
1907 used together if users want to use quota.
1910 Frequency : only with administrator action
1912 Description: mount failure if config log has invalid conf_param setting
1913 Details : If administrator specified an incorrect configuration parameter
1914 with "lctl conf_param" this would cause an error during future
1915 client mounts. Instead, ignore the bad configuration parameter.
1918 Frequency : blocks per group < blocksize*8 and uninit_groups is enabled
1920 Description: ldiskfs error: XXX blocks in bitmap, YYY in gd
1921 Details : If blocks per group is less than blocksize*8, set rest of the
1925 Frequency : Application do stride read on lustre
1927 Description: The read performance will drop a lot if the application does
1929 Details : Because the stride_start_offset are missing in stride read-ahead,
1930 it will cause clients read a lot of unused pages in read-ahead,
1931 then the read-performance drops.
1935 Description: more ldlm soft lockups
1936 Details : In ldlm_resource_add_lock(), call to ldlm_resource_dump()
1937 starve other threads from the resource lock for a long time in
1938 case of long waiting queue, so change the debug level from
1939 D_OTHER to the less frequently used D_INFO.
1941 Severity : enhancement
1943 Description: add -gid, -group, -uid, -user options to lfs find
1945 Severity : enhancement
1947 Description: ll_recover_lost_found_objs - recover objects in lost+found
1948 Details : OST corruption and subsequent e2fsck can leave objects in the
1949 lost+found directory. Using the "ll_recover_lost_found_objs"
1950 tool, these objects can be retrieved and data can be salvaged
1951 by using the object ID saved in the fid EA on each object.
1956 Description: this bug _only_ happens when inode quota limitation is very low
1957 (less than 12), so that inode quota unit is 1 at initialization.
1958 Details : if remaining quota equates 1, it is a sign to demonstate that quota
1959 is effective now. So least quota qunit should be 2.
1963 Description: Hung threads in invalidate_inode_pages2_range
1964 Details : The direct IO path doesn't call check_rpcs to submit a new RPC once
1965 one is completed. As a result, some RPCs are stuck in the queue
1970 Description: Procfs and llog threads access destoryed import sometimes.
1971 Details : Sync the import destoryed process with procfs and llog threads by
1972 the import refcount and semaphore.
1976 Description: mds fails to respond, threads stuck in ldlm_completion_ast
1977 Details : Sort source/child resource pair after updating child resource.
1982 Description: kernel BUG at ldiskfs2_ext_new_extent_cb
1983 Details : If insertion of an extent fails, then discard the inode
1984 preallocation and free data blocks else it can lead to duplicate
1989 Description: don't always update ctime in ext3_xattr_set_handle()
1990 Details : Current xattr code updates inode ctime in ext3_xattr_set_handle()
1991 In some cases the ctime should not be updated, for example for
1992 2.0->1.8 compatibility it is necessary to delete an xattr and it
1993 should not update the ctime.
1997 Description: add quota statistics
1998 Details : 1. sort out quota proc entries and proc code.
1999 2. add quota statistics
2004 Description: quotas are not honored with O_DIRECT
2005 Details : all writes with the flag O_DIRECT will use grants which leads to
2006 this problem. Now using OBD_BRW_SYNC to guard this.
2010 Bugzilla : 15713/16362
2011 Description: Assertion in iopen_connect_dentry in 1.6.3
2012 Details : looking up an inode via iopen with the wrong generation number can
2013 populate the dcache with a disconneced dentry while the inode
2014 number is in the process of being reallocated. This causes an
2015 assertion failure in iopen since the inode's dentry list contains
2016 both a connected and disconnected dentry.
2020 Description: assertion failure in ldlm_handle2lock()
2021 Details : fix a race between class_handle_unhash() and class_handle2object()
2022 introduced in lustre 1.6.5 by bug 13622.
2024 Severity : enhancement
2026 Description: superblock lock contention with many SMP cores on one client
2027 Details : several client filesystem locks were highly contended on SMP
2028 NUMA systems with 8 or more cores. Per-CPU datastructures
2029 and more efficient locking implemented to reduce contention.
2034 Description: Kernel BUG: sd_iostats_bump: unexpected disk index
2035 Details : remove the limit of 256 scsi disks in the sd_iostat patch
2040 Description: oops in sd_iostats_seq_show()
2041 Details : unloading/reloading the scsi low level driver triggers a kernel
2042 bug when trying to access the sd iostat file.
2047 Description: Kernel panics during QLogic driver reload
2048 Details : REQ_BLOCK_PC requests are not handled properly in the sd iostat
2049 patch, causing memory corruption.
2054 Description: journal_dev option does not work in b1_6
2055 Details : pass mount option during pre-mount.
2057 Severity : enhancement
2059 Description: Add a FIEMAP(FIle Extent MAP) ioctl for ldiskfs
2060 Details : FIEMAP ioctl will allow an application to efficiently fetch the
2061 extent information of a file. It can be used to map logical blocks
2062 in a file to physical blocks in the block device.
2065 Frequency : only with adaptive timeout enabled
2067 Description: DEBUG_REQ() bad paging request
2068 Details : ptlrpc_at_recv_early_reply() should not modify req->rq_repmsg
2069 because it can be accessed by reply_in_callback() without the
2073 Frequency : only on Cray X2
2075 Description: X2 build failures
2076 Details : fix build failures on Cray X2.
2080 Description: xid & resent requests
2081 Details : Initialize RPC XID from clock at startup (randomly if clock is
2086 Description: quota recovery deadlock during mds failover
2087 Details : This patch includes att18982, att18236, att18237 in bz14840.
2089 1. fix osts hang when mds does failover with quotaon
2090 2. prevent watchdog storm when osts threads wait for the
2095 Description: kernel panic on racer
2096 Details : Do not access dchild->d_inode when IS_ERR(dchild) is true.
2098 Severity : enhancement
2100 Description: Add lustre_start utility to start or stop multiple Lustre servers
2105 Description: Lustre GPF in {:ptlrpc:ptlrpc_server_free_request+373}
2106 Details : In case of memory pressure, list_del() can be called twice on
2107 req->rq_history_list, causing a kernel oops.
2111 Description: (ptllnd_peer.c:557:kptllnd_peer_check_sends()) ASSERTION(!in_interrupt()) failed
2112 Details : fix stack overflow in the distributed lock manager by defering
2113 export eviction after a failed ast to the elt thread instead of
2114 handling it in the dlm interpret routine.
2116 Severity : enhancement
2118 Description: More exported tunables for mballoc
2119 Details : Add support for tunable preallocation window and new tunables for
2120 large/small requests
2124 Description: Detect corruption of block bitmap and checking for preallocations
2125 Details : Checks validity of on-disk block bitmap. Also it does better
2126 checking of number of applied preallocations. When corruption is
2127 found, it turns filesystem readonly to prevent further corruptions.
2131 Frequency : only for big-endian servers
2132 Description: Check for big-endian system while mounting fs with extents feature
2133 Details : Mounting a filesystem with extents feature will fail on big-endian
2134 systems since ext3-based ldiskfs is not supported on big-endian
2135 systems. Can be over-ridden with "bigendian_extents" mount option.
2139 Description: Excessive recovery window
2140 Details : With AT enabled, the recovery window can be excessively long (6000+
2141 seconds). To address this problem, we no longer use
2142 OBD_RECOVERY_FACTOR when extending the recovery window (the connect
2143 timeout no longer depends on the service time, it is set to
2144 INITIAL_CONNECT_TIMEOUT now) and clients report the old service
2145 time via pb_service_time.
2148 Descriptoin: Don't sync journal after every i/o
2149 Details : Implement write RPC replay to allow server replies for write RPCs
2150 before data is on disk. However, this feature is disabled by
2151 default since some issues leading to data corruptions have been
2152 found during recovery (e.g. bug 19128). This feature can be enabled
2153 by running the following command on the OSSs:
2154 lctl set_param obdfilter.*.sync_journal=0
2158 Description: Watchdog triggered on MDS failover
2159 Details : enable OBD_CONNECT_MDT flag when connecting from the MDS so that
2160 the OSTs know that the MDS "UUID" can be reused for the same export
2161 from a different NID, so we do not need to wait for the export to
2166 Descriptoin: Lustre detected file system corruption with inode out of bounds
2167 Details : don't update i_size on MDS_CLOSE for directories. This causes
2168 directory corruptions on the MDT.
2170 -------------------------------------------------------------------------------
2172 2008-05-26 Sun Microsystems, Inc.
2174 * Support for kernels:
2175 2.6.5-7.311 (SLES 9),
2176 2.6.9-67.0.7.EL (RHEL 4),
2177 2.6.16.54-0.2.5 (SLES 10),
2178 2.6.18-53.1.14.el5 (RHEL 5),
2179 2.6.22.14 vanilla (kernel.org)
2180 * Client support for unpatched kernels:
2181 (see http://wiki.lustre.org/index.php?title=Patchless_Client)
2182 2.6.16 - 2.6.22 vanilla (kernel.org)
2183 * Due to problems with nested symlinks and FMODE_EXEC (bug 12652),
2184 we do not recommend using patchless RHEL4 clients with kernels
2185 prior to 2.6.9-55EL (RHEL4U5).
2186 * Recommended e2fsprogs version: 1.40.7-sun1
2187 * Note that reiserfs quotas are disabled on SLES 10 in this kernel.
2188 * RHEL 4 and RHEL 5/SLES 10 clients behaves differently on 'cd' to a
2189 removed cwd "./" (refer to Bugzilla 14399).
2190 * A new quota file format has been introduced in 1.6.5.
2191 The format conversion from prior releases is handled transparently,
2192 but releases older than 1.4.12/1.6.5 will not understand this new
2193 format. The automatic format conversion can be avoided by running
2194 the following command on the MDS before upgrading:
2195 'tunefs.lustre --param="mdt.quota_type=ug1" $MDTDEV'.
2196 For more information, please refer to bugzilla 13904.
2200 Description: quota performance fix
2201 Details : quota data is written in journalled mode instead of ordered to
2202 increase performance
2206 Description: lfs support for human-readable quota grace time strings
2207 Details : lfs setquota -t and lfs quota -t represent quota grace times
2208 in "XXwXXdXXhXXmXXs" format instead of large values in seconds
2211 Frequency : always with o2ib 1.3 and sles10
2213 Description: fix build with SLES10 and o2ib v3.
2214 Details : sles10 uses diffrent name for Module.symver file but configure
2215 assume this file has same name on RHEL/SLES/vanila kernels.
2218 Frequency : very rare, if additional xattrs are used on kernels >= 2.6.12
2220 Description: MDS may lose file striping (and hence file data) in some cases
2221 Details : If there are additional extended attributes stored on the MDS,
2222 in particular ACLs, SELinux, or user attributes (if user_xattr
2223 is specified for the client mount options) then there is a risk
2224 of attribute loss. Additionally, the Lustre file striping
2225 needs to be larger than default (e.g. striped over all OSTs),
2226 and an additional attribute must be stored initially in the
2227 inode and then increase in size enough to be moved to the
2228 external attribute block (e.g. ACL growing in size) for file
2231 Severity : enhancement
2233 Description: add message levels for liblustreapi
2238 Description: MDT cannot be unmounted, reporting "Mount still busy"
2239 Details : Mountpoint references were being leaked during open reply
2240 reconstruction after an MDS restart. Drop mountpoint reference
2241 in reconstruct_open() and free dentry reference also.
2246 Description: fix for occasional failure case of -ENOSPC in recovery-small tests
2247 Details : Move the 'good_osts' check before the 'total_bavail' check. This
2248 will result in an -EAGAIN and in the exit call path we call
2249 alloc_rr() which will with increasing aggressiveness attempt to
2250 aquire precreated objects on the minimum number of required OSCs.
2254 Description: Use old size assignment to avoid deadlock
2255 Details : This reverts the changes in bugs 2369 and bug 14138 that introduced
2256 the scheduling while holding a spinlock. We do not need locking
2257 for size in ll_update_inode() because size is only updated from
2258 the MDS for directories or files without objects, so there is no
2259 other place to do the update, and concurrent access to such inodes
2260 are protected by the inode lock.
2264 Description: Use __u64 instead of int for valid bits
2268 Description: resolve "_IOWR redefined" build error on SLES10
2272 Description: dump the memory debugging after all modules are unloaded to
2273 suppress false negative in conf_sanity test 39
2277 Description: the recovery timer never expires
2278 Details : for new client connect request, the recovery timer should not be
2279 reset, otherwise recovery timer will never expired, if the old
2280 client never come. Only old client connect and first connection
2281 req should trigger recovery timer reset.
2285 Description: the min numbers of lproc stats are wrong
2286 Details : adding a new constant LC_MIN_INIT and use it for initialization
2290 Frequency : always with interactive lfs
2292 Description: Reinitialize optind to 0 so that interactive lfs works in all cases
2295 Frequency : with multiple concurrent readdir processes in same directory
2296 Bugzilla : 15406, 15169, 15175
2297 Description: misc fixes for directory readahead.
2298 Details : prevent previous statahead async RPC callback from processing the
2299 current "statahead_info", race condition between async RPC callback
2300 add dentry into dentry hash table and "ls" thread revalidate such
2301 dentry, statahead his/miss control for hidden items, and so on.
2303 Severity : enhancement
2305 Description: build kernel-ib packages for OFED 1.3 in our release cycle
2309 Description: incore types cleaning in quota code (with respect to 64-bit limits)
2310 Details : several u32 variables declarations are replaced with u64 declarations
2315 Description: fix SLES kernel versioning
2316 Details : the kernel version for our SLES 10 kernel did not include a "-"
2317 before the "smp" at the end. while this was not a problem in
2318 general, it did mean that software trying to use the kernel
2319 version to try to detect a vendor specific kernel would fail.
2320 this was most evident by the OFED build scripts.
2325 Description: Don't update lov_desc members until making sure they are valid
2326 Details : When updating lov_desc members via proc fs, need fix their
2327 validities before doing the real update.
2330 Frequency : very rare
2332 Description: don't put request into delay list while invalidate in flight.
2333 Details : ptlrpc_delay_request sometimes put in delay list while invalidate
2334 import in flight. this produce timeout for invalidate and sometimes
2335 can cause stale data.
2337 Severity : enhancement
2339 Description: Update kernel to SLES9 2.6.5-7.311.
2341 Severity : enhancement
2343 Description: Update kernel to RHEL4 2.6.9-67.0.7.
2348 Frequency : on PPC only
2349 Description: not convert ost objects for directory because it's not exist.
2350 Details : ll_dir_getstripe assume dirrectory has ost objects but this wrong.
2352 Severity : enhancement
2354 Description: Fix warnings with compile liblustre at sles10/rhel5 which have
2355 __u64 as usingied long long type.
2358 Frequency : rare, on shutdown
2360 Description: race process ast vs remove callback
2361 Details : removing callback before disconnect import open race with
2362 processing callback.
2364 Severity : enhancement
2366 Description: Update kernel to SLES9 2.6.5-7.311.
2368 Severity : enhancement
2370 Description: Files open for execute are not marked busy on SLES10
2371 Details : Add FMODE_EXEC to SLES10 SP1 server kernel series.
2373 Severity : enhancement
2375 Description: Add server support for vanilla-2.6.22.14.
2378 Frequency : occasional
2380 Description: Avoid lov_create() getting stuck in obd_statfs_rqset()
2381 Details : If an OST is down the MDS will hang indefinitely in
2382 obd_statfs_rqset() waiting for the statfs data. While for
2383 MDS QOS usage of statfs, it should not stuck in waiting.
2385 Severity : enhancement
2387 Description: Disable adaptive timeouts by default
2390 Frequency : on network error
2392 Description: panic with double free request if network error
2393 Details : mdc_finish_enqueue is finish request if any network error occuring,
2394 but it's true only for synchronus enqueue, for async enqueue
2395 (via ptlrpcd) this incorrect and ptlrpcd want finish request
2399 Frequency : rare, on recovery
2401 Description: read procfs can produce deadlock in some situation
2402 Details : Holding lprocfs lock with send rpc can produce block for destroy
2403 obd objects and this also block reconnect with -EALREADY.
2404 This isn't fix all lprocfs bugs - but make it rare.
2406 Severity : enhancement
2408 Description: Update kernel to RHEL5 2.6.18-53.1.14.el5.
2411 Frequency : frequent on X2 node
2413 Description: mdc_set_open_replay_data LBUG
2414 Details : Set replay data for requests that are eligible for replay.
2419 Description: lustre_mgs: operation 101 on unconnected MGS
2420 Details : When MGC is disconnected from MGS long enough, MGS will evict the
2421 MGC, and late on MGC cannot successfully connect to MGS and a lot
2422 of the error messages complaining that MGS is not connected.
2425 Frequency : rare, depends on device drivers and load
2427 Description: MDS or OSS nodes crash due to stack overflow
2428 Details : Code changes in 1.6.4 increased the stack usage of some functions.
2429 In some cases, in conjunction with device drivers that use a lot
2430 of stack, the MDS (or possibly OSS) service threads could overflow
2431 the stack. One change which was identified to consume additional
2432 stack has been reworked to avoid the extra stack usage.
2434 Severity : enhancement
2436 Description: Update to RHEL5 latest kernel-2.6.18-53.1.13.el5.
2438 Severity : enhancement
2440 Description: Update to SLES10 SP1 latest kernel-2.6.16.54-0.2.5.
2442 Severity : enhancement
2444 Description: Update to RHEL5 latest kernel-2.6.18-53.1.6.el5.
2446 Serverity : enhancement
2448 Description: Update RHEL4 kernel to 2.6.9-67.0.4.
2451 Frequency : rare on shutdown OST
2453 Description: Don't allow skipping OSTs if index has been specified.
2454 Details : Don't allow skipping OSTs if index has been specified, make locking
2455 in internal create lots better.
2460 Description: ASSERTION(!PageDirty(page)) failed
2461 Details : Wrong check could lead to an assertion failure under specific
2467 Description: LBUG in ptlrpc_check_set() bad phase ebc0de00
2468 Details : access to bitfield in structure is always rounded to long
2469 and this produce problem with not atomic change any bit.
2474 Description: Lustre make rpms failed.
2475 Details : Remove ldiskfs spec file to avoids rpmbuild be confused when
2476 builds Lustre rpms from tarball.
2478 Severity : enhancement
2480 Description: Update to SLES9 SP4 kernel-2.6.5-7.308.
2483 Frequency : rare on shutdown OST
2485 Description: If llog cancel was not send before clean_exports phase, this can
2486 produce deadlock in llog code.
2487 Details : If llog thread has last reference to obd and call class_import_put
2488 this produce deadlock because llog_cleanup_commit_master wait when
2489 last llog_commit_thread exited, but this never success because was
2490 called from llog_commit_thread.
2493 Frequency : only if OST index is skipped
2495 Description: NULL lov_tgts causing MDS oops
2496 Details : more safe checks for NULL lov_tgts for avoid oops.
2498 Severity : enhancement
2500 Description: Update to RHEL4 latest kernel-2.6.9-67.0.1.EL.
2502 Severity : enhancement
2504 Description: Update to RHEL5 latest kernel-2.6.18-53.1.4.el5.
2509 Description: make mgs_setparam() handle fsname containing dash
2510 Details : fsname containing a dash does not work with lctl conf_param
2512 Severity : enhancement
2514 Description: Update to RHEL4 Update-6 kernel-2.6.9-67.EL.
2517 Frequency : rare, in recovery and (or) destroy lovobjid file.
2519 Description: rewrite lov objid code.
2520 Details : Cleanup for lov objid code, remove scability problems and wrong
2521 locking. Fix sending last_id into ost.
2523 Severity : enhancement
2525 Description: Update to SLES10 SP1 latest kernel-2.6.16.54-0.2.3.
2527 Severity : enhancement
2529 Description: Update to RHEL5 Update-1 kernel 2.6.18-53.el5.
2530 Details : Use d_move_locked instead of __d_move.
2533 Frequency : rare, at shutdown
2535 Description: access already free / zero obd_namespace.
2536 Details : if client_disconnect_export was called without force flag set,
2537 and exist connect request in flight, this can produce access to
2538 NULL pointer (or already free pointer) when connect_interpret
2539 store ocd flags in obd_namespace.
2542 Frequency : only at startup
2544 Description: not alloc memory with spinlock held.
2545 Details : allocation memory with GFP_KERNEL can produce sleep deadlock,
2546 if any spinlock held.
2551 Description: lfs find does not continue on file error
2552 Details : Continue other files processing when a file/dir is absent.
2556 Description: Inconsistent usage of lustre_pack_reply()
2557 Details : Standardize the usage of lustre_pack_reply() such that it
2558 always generate a CERROR on failure.
2561 Frequency : very rare
2563 Description: Fix replay if there is an un-replied request and open
2564 Details : In some cases, older replay request will revert the
2565 mcd->mcd_last_xid on MDS which is used to record the client's
2566 latest sent request.
2568 Severity : enhancement
2570 Description: Update to RHEL5 kernel 2.6.18-8.1.15.el5.
2572 Severity : enhancement
2574 Description: Update to SLES10 SP1 kernel 2.6.16.53-0.16
2576 Severity : enhancement
2578 Description: Update to SLES9 kernel-2.6.5-7.287.3.
2580 Severity : enhancement
2582 Description: Update to RHEL4 kernel-2.6.9-55.0.12.EL.
2584 Severity : enhancement
2586 Description: Build SLES10 patchless client fails
2587 Details : The configure was broken by run ./configure with
2588 --with-linux-obj=.... argument for patchless client. When the
2589 configure use --with-linux-obj, the LINUXINCLUDE= -Iinclude
2590 can't search header adequately. Use absolute path such as
2591 -I($LINUX)/include instead.
2593 Severity : enhancement
2595 Description: Lustre Page Accounting
2596 Details : New macros for page alloc and free which enable accounting
2597 of page allocation of Lustre. Use percpu counters to store memory
2598 and page statistics.
2601 Frequency : only if debugging is disabled
2603 Description: LASSERT_{REQ,REP}SWAB macros are buggy
2604 Details : If SWAB_PARANOIA is disabled, the LASSERT_REQSWAB and
2605 LASSERT_REPSWAB macros become no-ops, which is incorrect. Drop
2606 these macros and replace them with their definitions instead.
2611 Description: interrupt oig_wait produce painc on resend.
2612 Details : brw_redo_request can be used for resend requests from ptlrpcd and
2613 private set, and this produce situation when rq_ptlrpcd_data not
2614 copyed to new allocated request and triggered LBUG on assert
2615 req->rq_ptlrpcd_data != NULL. But this member used only for wakeup
2616 ptlrpcd set if request is changed and can be safety changed to use
2619 Severity : enhancement
2621 Description: organize the server-side client stats on per-nid basis
2622 Details : Change the structure of stats under obdfilter and mds to
2630 The "uuid"s file would list the uuids of _active_ exports.
2631 And the clear entry is to clear all stats and stale nids.
2636 Description: Processes looping in ll_readdir() on Lustre clients finally causing
2637 a full node pseudo-hang
2638 Details : Concurrent access to the same directory from multiple clients with
2639 intensive file creation/removal can cause a client node to spin in
2640 ll_readdir(). i_version must be increased every time the lock is
2641 cancelled to ensure a revalidate is done.
2646 Description: touch file failed when fs is not full
2647 Details : OST in recovery should not be discarded by MDS in alloc_qos(),
2648 otherwise we can get ENOSP while fs is not full.
2651 Frequency : only for Cray XT3
2652 Bugzilla : 12829/13455
2653 Description: Changing primary group doesn't change the group lustre assigns to
2655 Details : When CRAY_XT3 is defined, the fsgid supplied by the client is
2656 overridden with the primary group provided by the group upcall,
2657 whereas the supplied fsgid can be trusted if it is in the list of
2658 supplementary groups returned by the group upcall.
2660 Severity : enhancement
2662 Description: Root Squash Functionality
2663 Details : Implementation of NFS-like root squash capability. Specifically,
2664 don't allow someone with root access on a client node to be able
2665 to manipulate files owned by root on a server node.
2667 Severity : enhancement
2669 Description: Slow trucate/writes to huge files at high offsets.
2670 Details : Directly associate cached pages to lock that protect those pages,
2671 this allows us to quickly find what pages to write and remove
2672 once lock callback is received.
2677 Description: Too many locks accumulating on client during NFS usage
2678 Details : mds_open improperly used accmode to find out access mode to a
2679 file. Also mdc_intent_lock logic to find out if we already have
2680 lock similar to just received was flawed since introduction of
2681 skiplists - locks are now added to the front of the granted
2686 Description: Hit ASSERTION(obd->obd_stopping == 1) failed in some setup failed
2688 Details : In obd setup failure handler, obd_stopping will not necessarily to
2689 be 1, and obd_set_up should also be checked to make sure whether
2690 obd is completely setup.
2692 Severity : enhancement
2694 Description: Allow masking D_WARNING, D_ERROR messages from console
2695 Details : Console messages can now be disabled via lnet.printk.
2700 Description: User code with malformed file open parameter crashes client node
2701 Details : Before packing join_file req, all the related reference should be
2702 checked carefully in case some malformed flags cause fake join_file
2708 Description: shrink/enlarge qunit size when needed; fix the problem of coarse
2709 grain of quota doing harm to quota's accuracy
2710 Details : qunit size will be changed when quota limitation is too low/high;
2711 record the pending quota write in order to get more accureate
2712 quota; delete the patch for bug12588, which is unnecessary when
2713 this patch is landed. This bug also contains fixes for bug 14526,
2714 14299, 14601 and 13794.
2718 Description: LDLM_ENQUEUE races with LDLM_CP_CALLBACK
2719 Details : ldlm_completion_ast() assumes that a lock is granted when the req
2720 mode is equal to the granted mode. However, it should also check
2721 that LDLM_FL_CP_REQD is not set.
2725 Description: Heavy nfs access might result in deadlocks
2726 Details : After ELC code landed, it is now improper to enqueue any mds
2727 locks under och_sem, because enqueue might want to decide to
2728 cancel open locks for same inode we are holding och_sem for.
2732 Description: 35% write performance drop with ldiskfs2 when quotas are on
2733 Details : Enable ext3 journalled quota by default to improve performance
2734 when quotas are turned on.
2738 Description: Client eviction while running blogbench
2739 Details : A lot of unlink operations with concurrent I/O can lead to a
2740 deadlock causing evictions. To address the problem, the number of
2741 oustanding OST_DESTROY requests is now throttled to
2742 max_rpcs_in_flight per OSC and LDLM_FL_DISCARD_DATA blocking
2743 callbacks are processed in priority.
2746 Frequency : RHEL4 only
2748 Description: mkfs is very slow on IA64/RHEL4
2749 Details : A performance regression has been discovered in the MPT Fusion
2750 driver between versions 3.02.73rh and 3.02.99.00rh. As a
2751 consequence, we have downgraded the MPT Fusion driver in the RHEL4
2752 kernel from 3.02.99.00 to 3.02.73 until this problem is fixed.
2755 Frequency : PPC/PPC64 only
2757 Description: conflicts between asm-ppc64/types.h and lustre_types.h
2758 Details : fix duplicated definitions between asm-ppc64/types.h and
2759 lustre_types.h on PPC.
2762 Frequency : PPC/PPC64 only
2764 Description: asm-ppc/segment.h does not exist
2765 Details : fix compile issue on PPC.
2769 Description: data checksumming impacts single node performance
2770 Details : add support for several checksum algorithms. Currently, CRC32 and
2771 Adler-32 are supported. The checksum type can be changed on the fly
2772 through /proc/fs/lustre/osc/*/checksum_type.
2776 Description: use adler32 for page checksums
2777 Details : when available, use the Adler-32 algorithm instead of CRC32 for
2782 Description: better handle error messages in extents code
2784 Severity : enhancement
2786 Description: SNMP support enhancement
2787 Details : Adding total number of sampled request for an MDS node in snmp
2790 Severity : enhancement
2792 Description: Optimize ldlm waiting list processing for PR extent locks
2793 Details : When processing waiting list for read extent lock and meeting read
2794 lock that is same or wider to it that is not contended, skip
2795 processing rest of the list and immediatelly return current
2796 status of conflictness, since we are guaranteed there are no
2797 conflicting locks in the rest of the list.
2801 Description: Time out and refuse to reconnect
2802 Details : When the failover node is the primary node, it is possible
2803 to have two identical connections in imp_conn_list. We must
2804 compare not conn's pointers but NIDs, otherwise we
2805 can defeat connection throttling.
2809 Description: Client not clear own cache if answer to reconnect is lost.
2810 Details : Client gets evicted from server. Now client also thinks it is
2811 disconnected (or gets ENOTCONN on its operation) and decides to
2812 reconnect. Server receives reconnect message, but cannot find
2813 export. New export is created that is fully valid (new cookie!),
2814 but reply is lost and not reported to client. Client reconnects
2815 again and gets back a just-created connection, but it is not new
2816 so client thinks it was not evicted and does not do recovery.
2820 Description: Detect stride IO mode in read-ahead
2821 Details : When a client does stride read, read-ahead should detect that and
2822 read-ahead pages according to the detected stride pattern.
2826 Description: build for x2 fails
2827 Details : fix compile issue on Cray systems.
2829 Severity : enhancement
2831 Description: implement readv/writev APIs(aio_read/aio_writes in newer kernels)
2832 Details : This greatly improves speed of NFS writes on 2.6 kernels.
2835 Frequency : only on PPC/SLES10
2837 Description: "BITS_PER_LONG is not 32 or 64" in linux/idr.h
2838 Details : On SLES10/PPC, fs.h includes idr.h which requires BITS_PER_LONG to
2839 be defined. Add a hack in mkfs_lustre.c to work around this compile
2844 Description: LASSERT on MDS when client holding flock lock dies
2845 Details : ldlm pool logic depends on number of granted locks equal to
2846 number of released locks which is not true for flock locks, so
2847 just exclude such locks from consideration.
2851 Description: MDS deadlock with many ll_sync_lov threads and I/O stalled
2852 Details : Use fsfilt_sync() for both the whole filesystem sync and
2853 individual file sync to eliminate dangerous inode locking
2854 with I_LOCK that can lead to a deadlock.
2858 Description: Update an obsolete wirecheck.c generator
2859 Details : Update wirecheck.c/wirehdr.c and regenerate wiretest.c
2863 Description: Client can panic on open sometimes
2864 Details : It is possible that we try to free already freed request in
2865 ll_file_open in some error cases when we send request from
2870 Description: performance in 1.6.3
2871 Details : Force q->max_phys_segments to MAX_PHYS_SEGMENTS on SLES10 to be
2872 sure that 1MB requests are not fragmented by the block layer.
2876 Description: LDLM soft lockups - improvement
2877 Details : It is be possible to send the lock handle along with each read
2878 or write request because the client is already doing a lock match
2879 itself so there isn't any reason the OST should have to re-do that
2885 Description: lfs quota fails with deactivated OSTS
2886 Details : With this patch, three improvements are included:
2887 1. detete the softlimit in mds and osts when use "lfs quota".
2888 2. display the inaccurate data in the output of "lfs quota".
2889 3. try to get quota info when "lfs quota" is executed.
2894 Description: Extent locks not granted with no conflicts sometimes.
2895 Details : When race occurs in glimpse handler and nothing is returned,
2896 we do not reprocess the queue after lock cancel, and that leads
2897 to a stall until next activity on a resource
2900 Frequency : failover with quotaon
2902 Description: during mds failovers with quota on, OSTs got into deadlock state
2903 and causing dumpstack.
2904 Details : for every quota slave, at any time, there is only one quota req
2905 is sent to quota master for every uid/gid. Before that quota req
2906 returns, all the thread relative to the same uid/gid will wait.
2907 So if the quota req is lost because mds failovers or any other
2908 reasons, this bug will be hit. Now, dqacq_interpret() will handle
2909 quota reqs who time out.
2911 Severity : enhancement
2914 Description: when quota slave checks if quota is enough, there is an unnecessary
2916 Details : place this wait on necessary place instead of always waiting.
2918 --------------------------------------------------------------------------------
2920 2007-12-07 Cluster File Systems, Inc. <info@clusterfs.com>
2922 * Support for kernels:
2923 2.6.5-7.286 (SLES 9),
2924 2.6.9-55.0.9.EL (RHEL 4),
2925 2.6.16.53-0.8 (SLES 10),
2926 2.6.18-8.1.14.el5 (RHEL 5),
2927 2.6.18.8 vanilla (kernel.org)
2928 * Client support for unpatched kernels:
2929 (see http://wiki.lustre.org/index.php?title=Patchless_Client)
2930 2.6.16 - 2.6.22 vanilla (kernel.org)
2931 * Due to recently discovered recovery problems, we do not recommend
2932 using patchless RHEL 4 clients with this or any earlier release.
2933 * Recommended e2fsprogs version: 1.40.2-cfs1
2934 * Note that reiserfs quotas are disabled on SLES 10 in this kernel.
2937 Frequency : occasional
2939 Description: MDS hang or stay in waiting lock
2940 Details : If client receive lock with CBPENDING flag ldlm need send lock
2941 cancel as separate rpc, to avoid situation when cancel request
2942 can't processed due all i/o threads stay in wait lock.
2945 Frequency : occasional
2947 Description: Do not fail import if osc_interpret_create gets -EAGAIN
2948 Details : If osc_interpret_create got -EAGAIN it immediately exits and
2949 wakeup oscc_waitq. After wakeup oscc_wait_for_objects call
2950 oscc_has_objects and see OSC has no objests and call
2951 oscc_internal_create to resend create request.
2953 Severity : enhancement
2955 Description: Update kernel patches for SLES10 2.6.16.53-0.8.
2956 Details : Update which_patch & target file for SLES10 latest kernel.
2958 Severity : enhancement
2960 Description: add --type and --size parameters to lfs find
2961 Details : Enhance lfs find by adding filetype and filesize parameters. Also
2962 multiple OBDs can now be specified for the --obd option.
2964 Severity : enhancement
2966 Description: eliminate client locks in face of contention
2967 Details : file contention detection and lockless i/o implementation
2968 for contended files.
2970 Severity : enhancement
2972 Description: Remove client patches from SLES 10 kernel.
2973 Details : This causes SLES 10 clients to behave as patchless clients
2974 even on a Lustre-patched (server) kernel.
2976 Severity : enhancement
2978 Description: use i_size_read and i_size_write in 2.6 port
2979 Details : replace inode->i_size access with i_size_read/write()
2982 Frequency : when removing large files
2984 Description: scheduling issue during removal of large Lustre files
2985 Details : Don't take the BKL in fsfilt_ext3_setattr() for 2.6 kernels.
2986 It causes scheduling issues when removing large files (17TB in the
2992 Description: 1.4.11 Can't handle directories with stripe set and extended ACLs
2993 Details : Impossible (EPROTO is returned) to access a directory that has a
2994 non-default striping and ACLs.
2997 Frequency : only on ppc
2999 Description: /proc/fs/lustre/devices broken on ppc
3000 Details : The patch as applied to 1.6.2 doesn't look correct for all arches.
3001 We should make sure the type of 'index' is loff_t and then cast
3002 explicitly as needed below. Do not assign an explicitly cast
3006 Frequency : only for rhel5
3008 Description: Kernel patches update for RHEL5 2.6.18-8.1.10.el5.
3009 Details : Modify the target file & which_kernel.
3012 Frequency : if the uninit_groups feature is enabled on ldiskfs
3014 Description: e2fsck reports "invalid unused inodes count"
3015 Details : If a new ldiskfs filesystem is created with the "uninit_groups"
3016 feature and only a single inode is created in a group then the
3017 "bg_unused_inodes" count is incorrectly updated. Creating a
3018 second inode in that group would update it correctly.
3023 Description: buffer overruns could theoretically occur
3024 Details : llapi_semantic_traverse() modifies the "path" argument by
3025 appending values to the end of the origin string, and a buffer
3026 overrun may occur. Adding buffer overrun check in liblustreapi.
3028 Severity : enhancement
3030 Description: Add jbd statistics patch for RHEL5 and 2.6.18-vanilla.
3033 Frequency : only if filesystem is inconsistent
3035 Description: handle "serious error: objid * already exists" more gracefully
3036 Details : If LAST_ID value on disk is smaller than the objects existing in
3037 the O/0/d* directories, it indicates disk corruption and causes an
3038 LBUG(). If the object is 0-length, then we should use the existing
3039 object. This will help to avoid a full fsck in most cases.
3041 Severity : enhancement
3043 Description: Kernel patches update for RHEL4 2.6.9-55.0.6.
3044 Details : Modify vm-tunables-rhel4.patch.
3046 Severity : enhancement
3048 Description: Kernel config for 2.6.18-vanilla.
3049 Details : Modify targets/2.6-vanilla.target.in.
3050 Add config file kernel-2.6.18-2.6-vanilla-i686.config.
3051 Add config file kernel-2.6.18-2.6-vanilla-i686-smp.config.
3052 Add config file kernel-2.6.18-2.6-vanilla-x86_64.config.
3053 Add config file kernel-2.6.18-2.6-vanilla-x86_64-smp.config.
3056 Frequency : occasional
3058 Description: improve handling recoverable errors
3059 Details : If request processed with error which can be recoverable on server
3060 request should be resend, otherwise page released from cache and
3066 Description: Kernel patches update for RHEL5 2.6.18-8.1.14.el5.
3067 Details : Modify target file & which_patch.
3068 A flaw was found in the IA32 system call emulation provided
3069 on AMD64 and Intel 64 platforms. An improperly validated 64-bit
3070 value could be stored in the %RAX register, which could trigger an
3071 out-of-bounds system call table access. An untrusted local user
3072 could exploit this flaw to run code in the kernel
3073 (ie a root privilege escalation). (CVE-2007-4573).
3077 Description: change order of libsysio includes
3078 Details : '#include sysio.h' should always come before '#include xtio.h'
3080 Severity : enhancement
3082 Description: adapt the lustre_config script to support the upgrade case
3083 Details : Add "-u" option for lustre_config script to support upgrading 1.4
3084 server targets to 1.6 in parallel.
3089 Description: To avoid grant space > avaible space when the disk is almost
3090 full. Without this patch you might see the error "grant XXXX >
3091 available" or some LBUG about grant, when the disk is almost
3093 Details : In filter_check_grant, for non_grant cache write, we should
3094 check the left space by if (*left > ungranted + bytes), instead
3095 of (*left > ungranted), because only we are sure the left space
3096 is enough for another "bytes", then the ungrant space should be
3097 increase. In client, we should update cl_avail_grant only there is
3098 OBD_MD_FLGRANT in the reply.
3103 Description: Update RHEL 4 kernel to fix local root privilege escalation.
3104 Details : Update to the latest RHEL 4 kernel to fix the vulnerability
3105 described in CVE-2007-4573. This problem could allow untrusted
3106 local users to gain root access.
3109 Frequency : when using O_DIRECT and quotas
3111 Description: Incorrect file ownership on O_DIRECT output files
3112 Details : block usage reported by 'lfs quota' does not take into account
3113 files that have been written with O_DIRECT.
3117 Description: (rw.c:1323:ll_read_ahead_pages()) ASSERTION(page_idx > ria->ria_stoff) failed
3118 Details : Once the unmatched stride IO mode is detected, shrink the stride-ahead
3119 window to 0. If it does hit cache miss, and read-pattern is still
3120 stride-io mode, does not reset the stride window, but also does not
3121 increase the stride window length in this case.
3123 --------------------------------------------------------------------------------
3125 2007-09-27 Cluster File Systems, Inc. <info@clusterfs.com>
3127 * Support for kernels:
3128 2.6.5-7.286 (SLES 9),
3129 2.6.9-55.0.2.EL (RHEL 4),
3130 2.6.16.46-0.14 (SLES 10),
3131 2.6.18-8.1.8.el5 (RHEL 5),
3132 2.6.18.8 vanilla (kernel.org)
3133 * Client support for unpatched kernels:
3134 (see http://wiki.lustre.org/index.php?title=Patchless_Client)
3135 2.6.16 - 2.6.21 vanilla (kernel.org)
3136 * Due to recently discovered recovery problems, we do not recommend
3137 using patchless RHEL 4 clients with this or any earlier release.
3138 * Recommended e2fsprogs version: 1.40.2-cfs1
3139 * Note that reiserfs quotas are disabled on SLES 10 in this kernel.
3143 Description: Fix errors in lfs documentation
3144 Details : Fixes man pages
3146 Severity : enhancement
3148 Description: Adaptive timeouts
3149 Details : RPC timeouts adapt to changing server load and network
3150 conditions to reduce resend attempts and improve recovery time.
3152 Severity : enhancement
3154 Description: llapi_file_create() does not allow some changes
3155 Details : add llapi_file_open() that allows specifying the file creation
3156 mode and open flags, and also returns an open file handle.
3158 Severity : enhancement
3160 Description: df doesn't work properly if diskfs blocksize != 4K
3161 Details : Choose biggest blocksize of OST's as the LOV's blocksize.
3164 Frequency : always on directories with default striping set
3166 Description: lfs find on -1 stripe looping in lsm_lmm_verify_common()
3167 Details : Avoid lov_verify_lmm_common() on directory with -1 stripe count.
3169 Severity : enhancement
3171 Description: merge and cleanup kernel patches.
3172 Details : Remove mnt_lustre_list in vfs_intent-2.6-rhel4.patch.
3174 Severity : enhancement
3176 Description: RedHat Update kernel for RHEL5
3177 Details : Kernel config file for RHEL5.
3179 Severity : enhancement
3181 Description: OSS needs mutliple precreate threads
3182 Details : Add ability to start more than one create thread per OSS.
3185 Frequency : only with quota on the root user
3187 Description: mds_obd_create error creating tmp object
3188 Details : When the user sets quota on root, llog will be affected and can't
3189 create files and write files.
3192 Frequency : Always on ia64 patchless client, and possibly others.
3194 Description: Add EXPORT_SYMBOL check for node_to_cpumask symbol.
3195 Details : This allows the patchless client to be loaded on architectures
3196 without this export.
3201 Description: disorder of journal start and llog_add cause deadlock.
3202 Details : in llog_origin_connect, journal start should happen before
3203 llog_add keep the same order as other functions to avoid
3206 Severity : enhancement
3208 Description: RedHat Update kernel for RHEL5
3209 Details : Modify the kernel config file more closer RHEL5.
3211 Severity : enhancement
3213 Description: Build failure against Centos5 (RHEL5)
3214 Details : Define PAGE_SIZE when it isn't present.
3217 Frequency : occasionally when using NFS
3219 Description: "ll_intent_file_open()) lock enqueue: err: -13" with nfs
3220 Details : with NFS, the anon dentry's parent was set to itself in
3221 d_alloc_anon(), so in MDS, we use rec->ur_fid1 to find the
3222 corresponding dentry other than use rec->ur_name.
3224 Severity : enhancement
3226 Description: enable data checksumming by default
3227 Details : enable checksum by default, allow --disable-checksum
3228 configure option and "-o nochecksum" mount option. Checksums
3229 can also be disabled at runtime via $LPROC/osc/*/checksum_pages.
3232 Frequency : Occasionally with failover
3234 Description: Client eviction due to failover config
3235 Details : after a connection loss, the lustre client should attempt to
3236 reconnect to the last active server first before trying the
3237 other potential connections.
3239 Severity : enhancement
3241 Description: client-side metadata stat-ahead during readdir(directory readahead)
3242 Details : perform client-side metadata stat-ahead when the client detects
3243 readdir and sequential stat of dir entries therein
3248 Description: when mds and osts use different quota unit(32bit and 64bit),
3249 quota will be released repeatly.
3250 Details : void sending multiple quota reqs to mds, which will keep the
3251 status between the reqs.
3254 Frequency : only with liblustre clients on XT3
3256 Description: evictions taking too long
3257 Details : allow llrd to evict clients directly on OSTs
3261 Description: osts not allocated evenly to files
3262 Details : change the condition to increase offset_idx
3264 Severity : enhancement
3266 Description: self-adjustable client's lru lists
3267 Details : use adaptive algorithm for managing client cached locks lru
3268 lists according to current server load, other client's work
3269 pattern, memory activities, etc. Both, server and client
3270 side namespaces provide number of proc tunables for controlling
3275 Description: rewrite ext2-derived code in llite/dir.c and obdclass/uuid.c
3276 Details : rewrite inherited code (uuid parsing code from ext2 utils and
3277 readdir code from ext3) from scratch preserving functionality.
3281 Description: Only those disconnect error should be returned by rq_status.
3282 Details : In open/enqueue processs, Some errors, which will cause client
3283 disconnected, should be returned by rq_status, while other
3284 errors should still be returned by intent, then mdc or llite will
3287 Severity : enhancement
3289 Description: Tune the kernel for good SCSI performance.
3290 Details : Set the value of /sys/block/{dev}/queue/max_sectors_kb
3291 to the value of /sys/block/{dev}/queue/max_hw_sectors_kb
3295 Frequency : Always for filesystems larger than 2TB on 32-bit systems.
3296 Bugzilla : 13547, 13627
3297 Description: Data corruption for OSTs that are formatted larger than 2TB
3299 Details : When generating the bio request for lustre file writes the
3300 sector number would overflow a temporary variable before being
3301 used for the IO. The data reads correctly from Lustre (which
3302 will overflow in a similar manner) but other file data or
3303 filesystem metadata may be corrupted in some cases.
3307 Description: "lfs find -obd UUID" prints directories
3308 Details : "lfs find -obd UUID" will return all directory names instead
3309 of just file names. It is incorrect because the directories
3310 do not reside on the OSTs.
3313 Bugzilla : 13596, 13614
3314 Description: MDS hang after unclean shutdown of lots of clients
3315 Details : Never resend AST requests.
3319 Frequency : Always, for kernels after 2.6.16
3320 Description: Fix warning idr_remove called for id=.. which is not allocated.
3321 Details : Last kernels save old s_dev before kill super and not allow
3322 to restore from callback - restore it before call kill_anon_super.
3326 Description: Fix error on 'ls .' at the top of the Lustre mount.
3327 Details : Don't revalidate dentry if it is a root dentry.
3329 --------------------------------------------------------------------------------
3331 2007-08-27 Cluster File Systems, Inc. <info@clusterfs.com>
3333 * Support for kernels:
3334 2.6.5-7.286 (SLES 9),
3335 2.6.9-55.0.2.EL (RHEL 4),
3336 2.6.16.46-0.14 (SLES 10),
3337 2.6.18-8.1.8.el5 (RHEL 5),
3338 2.6.18.8 vanilla (kernel.org)
3339 * Client support for unpatched kernels:
3340 (see http://wiki.lustre.org/index.php?title=Patchless_Client)
3341 2.6.16 - 2.6.22 vanilla (kernel.org)
3342 * Due to recently discovered recovery problems, we do not recommend
3343 using patchless RHEL 4 clients with this or any earlier release.
3344 * Recommended e2fsprogs version: 1.39.cfs8
3345 * Note that reiserfs quotas are disabled on SLES 10 in this kernel.
3350 Description: block reactivating mgc import until all deactivates complete
3351 Details : Fix race when failing back MDT/MGS to itself (testing)
3353 Severity : enhancement
3355 Description: lfs setstripe enhancement
3356 Details : Make lfs setstripe understand 'k', 'm' and 'g' for stripe size.
3359 Frequency : mds/oss recovery
3361 Description: llog ctxt is refrenced after it has been freed.
3362 Details : llog ctxt refcount was added to avoide the race between ctxt free
3363 and llog recovery process. Each llog user must hold ctxt refcount
3364 before it access the llog. And the llog ctxt can only be freed
3365 when its refcount is zero.
3367 Severity : enhancement
3369 Description: randomly memory allocation failure util
3370 Details : Make lustre randomly failed allocating memory for testing purpose.
3373 Frequency : only for SLES10
3375 Description: Update kernel patch for SLES10 SP1
3376 Details : Add patch blkdev_tunables-2.6-sles10.patch to 2.6-sles10.series.
3378 Severity : enhancement
3380 Description: omit set fsid for export NFS
3381 Details : fix set/restore device id for avoid EMFILE error and mark lustre fs
3382 as FS_REQUIRES_DEV for avoid problems with generate fsid.
3385 Frequency : after network failures
3387 Description: Add sync option to mount_lustre.c
3388 Details : Client loses data written to lustre after a network interruption.
3390 Severity : enhancement
3392 Description: Error message improvement.
3393 Details : Merging of two LCONSOLE_ERROR_MSG into one.
3395 Severity : enhancement
3397 Description: don't use GFP_* in generic Lustre code.
3398 Details : Use cfs_alloc_* functions and CFS_* flags for code portability.
3400 Severity : enhancement
3402 Description: obdclass is limited by single OBD_ALLOC(idarray)
3403 Details : replace OBD_ALLOC/OBD_FREE with OBD_VMALLOC/OBD_VFREE
3405 Severity : enhancement
3407 Description: updated patchess for new RHEL4 kernel
3408 Details : Fixed ext3-unlink-race.patch per Kalpak's comment.
3410 Severity : enhancement
3412 Description: warnings with build patchless client with vanila 2.6.19 and up
3413 Details : change old ctl_table style and replace ctl_table/ctl_table_header
3414 with cfs_sysctl_table_t/cfs_sysctl_table_header_t
3416 Severity : enhancement
3418 Description: O_DIRECT bypasses client statistics.
3419 Details : When running with O_DIRECT I/O, neither the client rpc_stats nor
3420 read_ahead_stats were updated. Copied stats section from
3421 osc_send_oap_rpc() into async_internal().
3424 Frequency : only for Cray XT3
3426 Description: peer credits not enough on many OST per OSS systems.
3427 Details : Use new lnet way to add credits as we need those for pings and ASTs
3430 Frequency : only with liblustre
3432 Description: Liblustre is not releasing flock locks on file close.
3433 Details : Release flock locks on file close.
3436 Frequency : only for RHEL4
3438 Description: Update kernel patches for kernel-2.6.9-55.0.2.EL
3439 Details : Remove inode-nr_unused-2.6.9-rhel4.patch from 2.6-rhel4.series
3440 Update target file and kernel config.
3444 Description: lustre support for RHEL5
3445 Details : Add support for RHEL5.
3450 Description: ASSERTION(export != NULL) failed in target_handle_connect
3451 Details : Assetion hit is result of rare race between disconnect and connet
3452 to same nid. target_handle_connect found old connect cockie and
3453 tried to reconnect, but can't find export for this cockie.
3458 Description: umount blocks forever on error
3459 Details : In result of wrong using obd_no_recov and obd_force flags client
3460 can hand if cancel or some other requests is lost.
3463 Frequency : Only for SLES
3465 Description: sanity_quota fail test_1
3466 Details : There are multiple occurences of $TSTUSR in SLES's /etc/group
3467 file, which makes TSTID[2] inunique.
3469 Severity : enhancement
3471 Description: Kernel patches for SLES9 2.6.5-7.286 kernel
3472 Details : Update target/ChangeLog/which_patch .
3474 Severity : enhancement
3476 Description: jbd statistics
3477 Details : Port older jbd statistics patch for sles10
3482 Description: lvbo_init failed for resource with missing objects.
3483 Details : Fix returning error if we do stat for file with missing/corrupted
3484 objects and i_size set to all sum of size of all avaible objects.
3485 if we to truncate/write to missing object - it is recreated.
3490 Description: Oops in read and write path when failing to allocate lock.
3491 Details : Check if lock allocation failed and return error back.
3494 Frequency : When flocks are used.
3496 Description: assertion failure in ldlm_cli_enquque_fini for non NULL lock.
3497 Details : Flock locks might destroy just granted lock if it could be merged
3498 with another existing flock, this is done in completion handler,
3499 so teach ldlm_cli_enquque_fini that this is a valid case for
3505 Description: reply_lock_interpret crash due to race with it and lock cancel.
3506 Details : Do not replay locks that are being cancelled. Do not reference
3507 locks by their address during replay, just by their handle.
3510 Frequency : only with deactivated OSTs
3512 Description: lstripe command fails for valid OST index
3513 Details : The stripe offset is compared to 'lov->desc.ld_tgt_count'
3514 instead of lov->desc.ld_active_tgt_count.
3516 Severity : enhancement
3518 Description: Build failure against Centos5 (RHEL5)
3519 Details : Use getpagesize() instead of PAGE_SIZE.
3521 --------------------------------------------------------------------------------
3523 2007-07-30 Cluster File Systems, Inc. <info@clusterfs.com>
3525 * Support for kernels:
3526 2.6.5-7.283 (SLES 9),
3527 2.6.9-55.EL (RHEL 4),
3528 2.6.16.46-0.14 (SLES 10),
3529 2.6.18.8 vanilla (kernel.org)
3530 * Client support for unpatched kernels:
3531 (see http://wiki.lustre.org/index.php?title=Patchless_Client)
3532 2.6.16 - 2.6.22 vanilla (kernel.org)
3533 * Due to recently discovered recovery problems, we do not recommend
3534 using patchless RHEL 4 clients with this or any earlier release.
3535 * Recommended e2fsprogs version: 1.39.cfs8
3536 * Note that reiserfs quotas are disabled on SLES 10 in this kernel.
3537 * Starting with this release, the ldiskfs backing filesystem required
3538 by Lustre is now in its own package, lustre-ldiskfs. This package
3539 should be installed. It is versioned separately from Lustre and
3540 may be released separately in future.
3545 Description: server LBUG when shutting down
3546 Details : Block umount forever until the mount refcount is zero rather
3547 than giving up after an arbitrary timeout.
3549 Severity : enhancement
3551 Description: add optional extra BUILD_VERSION info
3552 Details : add a new environment variable (namely LUSTRE_VERS) which allows
3553 to override the lustre version.
3556 Frequency : 2.6.18 servers only
3558 Description: ll_kern_mount() doesn't release the module reference
3559 Details : The ldiskfs module reference count never drops down to 0
3560 because ll_kern_mount() doesn't release the module reference.
3565 Description: server LBUG when using old ost_num_threads parameter
3566 Details : Accept the old ost_num_threads parameter but warn that it
3567 is deprecated, and fix an off-by-one error that caused an LBUG.
3572 Description: Transient SCSI error results in persistent IO issue
3573 Details : iobuf->dr_error is not reinitialized to 0 between two
3577 Frequency : sometimes when underlying device returns I/O errors
3579 Description: OSTs not going read-only during write failures
3580 Details : OSTs are not remounted read-only when the journal commit threads
3581 get I/O errors because fsfilt_ext3 calls journal_start/stop()
3582 instead of the ext3 wrappers.
3585 Frequency : only with 10000 clients or more
3587 Description: poor connect scaling with increasing client count
3588 Details : Don't run filter_grant_sanity_check for more than 100 exports
3589 to improve scaling for large numbers of clients.
3592 Frequency : SLES10 only
3594 Description: sanity-quota.sh quotacheck failed: rc = -22
3595 Details : Quotas cannot be enabled on SLES10.
3598 Frequency : liblustre clients only
3600 Description: getdirentries does not give error when run on compute nodes
3601 Details : getdirentries does not fail when the size specified as an argument
3602 is too small to contain at least one entry
3604 Severity : enhancement
3606 Description: Add LNET router traceability for debug purposes
3607 Details : If a checksum failure occurs with a router as part of the
3608 IO path, the NID of the last router that forwarded the bulk data
3609 is printed so it can be identified.
3614 Description: OST "spontaneously" evicts client; client has imp_pingable == 0
3615 Details : Due to a race condition, liblustre clients were occasionally
3616 evicted incorrectly.
3618 Severity : enhancement
3620 Description: lfs setstripe use optional parameters instead of postional
3623 Severity : enhancement
3625 Description: Nanosecond timestamp support for ldiskfs
3626 Details : The on-disk ldiskfs filesystem has added support for nanosecond
3627 resolution timestamps. There is not yet support for this at
3628 the Lustre filesystem level.
3631 Frequency : during server recovery
3633 Description: MDS failing to send precreate requests due to OSCC_FLAG_RECOVERING
3634 Details : request with rq_no_resend flag not awake l_wait_event if they get a
3638 Frequency : nfs export on patchless client
3640 Description: connectathon hang when test nfs export over patchless client
3641 Details : Disconnected dentry cannot be found with lookup, so we do not need
3642 to unhash it or make it invalid
3645 Description: fix llapi_lov_get_uuids() to allow many OSTs to be returned
3646 Details: : Change llapi_lov_get_uuids() to read the UUIDs from /proc instead
3647 of using an ioctl. This allows lfsck for > 160 OSTs to succeed.
3652 Description: open req refcounting wrong on reconnect
3653 Details : If reconnect happened between getting open reply from server and
3654 call to mdc_set_replay_data in ll_file_open, we will schedule
3655 replay for unreferenced request that we are about to free.
3656 Subsequent close will crash in variety of ways.
3657 Check that request is still eligible for replay in
3658 mdc_set_replay_data().
3663 Description: disable writes to filesystem when reading health_check file
3664 Details : the default for reading the health_check proc file has changed
3665 to NOT do a journal transaction and write to disk, because this
3666 can cause reads of the /proc file to hang and block HA state
3667 checking on a healthy but otherwise heavily loaded system. It
3668 is possible to return to the previous behaviour during configure
3669 with --enable-health-write.
3671 Severity : enhancement
3673 Description: 64-bit inode version
3674 Details: : Add a on-disk 64-bit inode version for ext3 to track changes made
3675 to the inode. This will be required for version-based recovery.
3680 Description: MDS fails to start if a duplicate client export is detected
3681 Details : in some rare cases it was possible for a client to connect to
3682 an MDS multiple times. Upon recovery the MDS would detect this
3683 and fail during startup. Handle this more gracefully.
3685 Severity : enhancement
3687 Description: Add -o localflock option to simulate old noflock behaviour.
3688 Details : This will achieve local-only flock/fcntl locks coherentness.
3693 Description: log_commit_thread vs filter_destroy race leads to crash
3694 Details : Take import reference before releasing llog record semaphore
3699 Description: Wrong request locking in request set processing
3700 Details : ptlrpc_check_set wrongly uses req->rq_lock for proctect add to
3701 imp_delayed_list, in this place should be used imp_lock.
3704 Frequency : when reconnecting
3706 Description: Grant leak when OSC reconnect to OST
3707 Details : When osc reconnect ost, OST(filter) should check whether it
3708 should grant more space to client by comparing fed_grant and
3709 cl_avail_grant, and return the granted space to client instead
3710 of "new granted" space, because client will call osc_init_grant
3711 to update the client grant space info.
3714 Frequency : when client reconnects to OST
3716 Description: Grant leak when OSC does a resend and replays bulk write
3717 Details : When osc reconnect to OST, OST(filter) should clear grant info of
3718 bulk write request, because the grant info while be sync between
3719 OSC and OST when reconnect, and we should ignore the grant info
3720 these of resend/replay write req.
3725 Description: Grant space more than avaiable space sometimes.
3726 Details : When then OST is about to be full, if two bulk writing from
3727 different clients came to OST. Accord the avaliable space of the
3728 OST, the first req should be permitted, and the second one
3729 should be denied by ENOSPC. But if the seconde arrived before
3730 the first one is commited. The OST might wrongly permit second
3731 writing, which will cause grant space > avaiable space.
3734 Frequency : when client is evicted
3736 Description: Grant might be wrongly erased when osc is evicted by OST
3737 Details : when the import is evicted by server, it will fork another
3738 thread ptlrpc_invalidate_import_thread to invalidate the
3739 import, where the grant will be set to 0. While the original
3740 thread will update the grant it got when connecting. So if
3741 the former happened latter, the grant will be wrongly errased
3742 because of this race.
3747 Description: Checking Stale with correct fid
3748 Details : ll_revalidate_it should uses de_inode instead of op_data.fid2
3749 to check whether it is stale, because sometimes, we want the
3750 enqueue happened anyway, and op_data.fid2 will not be initialized.
3752 Severity : enhancement
3754 Description: update patchless client
3755 Details : Add support for patchless client with 2.6.20, 2.6.21 and RHEL 5
3758 Frequency : only with 2.4 kernel
3760 Description: random memory corruption
3761 Details : size of struct ll_inode_info is to big for union inode.u and this
3762 can be cause of random memory corruption.
3767 Description: Memory leak in recovery
3768 Details : Lov_mds_md was not free in an error handler in mds_create_object.
3769 It should also check obd_fail before fsfilt_start, otherwise if
3770 fsfilt_start return -EROFS,(failover mds during mds recovery).
3771 then the req will return with repmsg->transno = 0 and rc = EROFS.
3772 and we met hit the assert LASSERT(req->rq_reqmsg->transno ==
3773 req->rq_repmsg->transno) in ptlrpc_replay_interpret. Fcc should
3774 be freed no matter whether fsfilt_commit success or not.
3777 Frequency : only with large numbers of cores on a single node
3779 Description: Prevents from taking the superblock lock in llap_from_page for
3781 Details : using LL_ORIGIN_REMOVEPAGE origin flag instead of LL_ORIGIN_UNKNOW
3782 for llap_from_page call in ll_removepage() prevents from taking
3783 the superblock lock for a soon died page.
3788 Description: Not check open intent error before release open handle
3789 Details : in some rare cases, the open intent error is not checked before
3790 release open handle, which may cause
3791 ASSERTION(open_req->rq_transno != 0), because it tries to release
3792 the failed open handle.
3797 Description: Set cat log bitmap only after create log success.
3798 Details : in some rare cases, the cat log bitmap is set too early. and it
3799 should be set only after create log success.
3803 Description: Accessing a block bevice can re-enable I/O when Lustre is
3804 tearing down a device.
3805 Details : dev_clear_rdonly(bdev) must be called in kill_bdev() instead of
3810 Description: service threads may hog cpus when there are a lot of requests
3811 Details : Insert cond_resched to give other threads a chance to use some CPU
3816 Description: the cat log was not initialized in recovery
3817 Details : When mds(mgs) do recovery, the tgt_count might be zero, so the
3818 unlink log on mds will not be initialized until mds post
3819 recovery. And also in mds post recovery, the unlink log will
3820 initialization will be done asynchronausly, so there will be race
3821 between add unlink log and unlink log initialization.
3825 Description: brw_stats were being printed incorrectly
3826 Details : brw_stats were being printed as log2 but all of them were not
3827 recorded as log2. Also remove some code duplication arising from
3828 filter_tally_{read,write}.
3832 Frequency : rare, only in recovery.
3833 Description: ASSERTION(req->rq_type != LI_POISON) failed
3834 Details : imp_lock should be held while iterating over imp_sending_list for
3835 prevent destroy request after get timeout in ptlrpc_queue_wait.
3839 Description: replay-single.sh test 52 fails
3840 Details : A lock's skiplist need to be cleanup when it being unlinked
3841 from its resource list.
3846 Description: Short directio read returns full requested size rather than
3848 Details : Direct I/O operations should return actual amount of bytes
3849 transferred rather than requested size.
3851 Severity : enhancement
3853 Description: metadata RPC reduction (e.g. for rm performance)
3854 Details : decrease the amount of synchronous RPC between clients and servers
3855 by canceling conflicing lock before the operation on the client
3856 and packing thier handles into the main operation RPC to server.
3858 Severity : enhancement
3860 Description: add #ifdef HAVE_KERNEL_CONFIG_H
3861 Details : kernels from 2.6.19 not need include linux/config.h, but add
3862 include linux/autoconf.h in commpiler command line.
3864 Severity : enhancement
3866 Description: patchless client support for 2.6.22 kernel
3867 Details : 2.6.22 has only one visble change, SLAB_CTOR_* constants is
3868 removed. In this case we need drop using os depended interface to
3869 kmem_cache and use cfs_mem_cache API.
3874 Description: fix mal-formatted messages
3875 Details : fix some mal-formatted DEBUG_REQ and LCONSOLE_ERROR_MSG messages
3879 Frequency : always in liblustre
3880 Description: wrong IS_ERR implementation in liblustre.h
3881 Details : fix IS_ERR implementation in liblustre.h for right detect errors.
3886 Description: Correct condition for output debug message.
3887 Details : inode i_nlink equal zero is not enough for output message about
3888 disk corruption, i_ctime and i_mode should be also checked.
3892 Frequency : always in patchless client
3893 Description: add configure check for truncate_complete_page
3894 Details : improve checks for exported symbols. This allow run check without
3895 sources, but with Module.symvers shipped with kernel distribution.
3896 add check for truncate_complete_page used by patchless client.
3900 Description: sanity.sh test_77h fails with "test_77h file compare failed"
3901 Details : test_77h uses a file which was messed by other test case.
3905 Description: Not Check whether lov_tgts is NULL in some lov functions
3906 Details : Checking whether lov_tgts is NULL in some functions.
3910 Description: replace obdo_alloc() with OBDO_ALLOC macro
3911 Details : nothing special is done in obdo_alloc() function, and for
3912 debugging purpose, it needs to be replaced with macros.
3916 Description: bad return value and errno from fcntl call
3917 Details : In liblustre API, errno should be a negative value if error
3922 Description: ptlrpc_check_set() LBUG
3923 Details : In case of positive reply from server and failed client bulk
3924 callback after bulk transfer shouldn't LBUG, but process this
3925 request as erroneous.
3927 Severity : enhancement
3929 Description: VFS operations stats tool.
3930 Details : Tool which collects stats by tracking value written in pid,
3931 ppid, gid and uses llstat to generate output to plot graph using
3933 Updated lustre/utils/Makefile.am
3934 Added lustre/utils/ltrack_stats.c
3936 Severity : enhancement
3938 Description: 2.6.18 server support (lustre 1.6.1)
3939 Details : Support for 2.6.18 kernels on the server side.
3944 Description: ASSERTION(imp->imp_conn_current) failed
3945 Details : an assertion failure is hit if a client node boots and attempts to
3946 mount a lustre filesystem faster than RECONNECT_INTERVAL seconds.
3949 Frequency : only for i686
3951 Description: 1.4.11 RC1 build fails for RHEL 4, i686
3952 Details : Fixed config variable for build.
3957 Description: Updated patchess for new RHEL4 kernel
3958 Details : Updated patch inode-nr_unused-2.6.9-rhel4.patch
3959 Updated patch jbd-stats-2.6.9.patch
3960 Updated patch qsnet-rhel4-2.6.patch
3961 Updated patch quota-deadlock-on-pagelock-core.patch
3962 Updated patch vfs_intent-2.6-rhel4.patch
3963 Updated patch vfs_races-2.6-rhel4.patch
3964 Updated series file 2.6-rhel4-titech.series
3965 Updated series file 2.6-rhel4.series
3966 Updated kernel config files
3971 Description: lquota slave complains LBUG when reconnecting with mds
3973 Details : quota slave depends qctxt->lqc_import to send its quota request.
3974 This pointer will be invalid if mds did failover or broke its
3975 connect to osts, which leads to LBUG.
3978 Frequency : when qunit size is too small(less than 20M)
3980 Description: write is stopped by improper -EDQUOT
3981 Details : If the master is busy and qunit size is small enough(let's say 1M),
3982 the slave can not get quota from master on time, which will lead
3983 slave to trigger a -EQUOTA to client.
3988 Description: Deadlock during metadata tests
3989 Details : in prune_dir_dentries(), shrink_dcache_parent() should not be
3990 called with the per-dentry lock held.
3993 Frequency : SLES9 only
3995 Description: Lustre patched kernel for SLES9 SP3 has NR_CPUS set to 8
3996 Details : set CONFIG_NR_CPUS to 128 instead of 8.
3998 Severity : enhancement
4000 Description: remove fs_prep_san_write operation and related patches
4001 Details : remove the ext3-san-jdike patches which are no longer useful.
4006 Description: LDISKFS-fs error (device sdc): ldiskfs_free_blocks
4007 Details : a disk corruption can cause the mballoc code to assert on a
4008 double free or other extent corruptions. Handle these with
4009 ext3_error() instead of assertions.
4012 Frequency : only with mballoc3 code and deep extent trees
4014 Description: ldiskfs_ext_search_right: bad header in inode: unexpected eh_depth
4015 Details : a wrong check of extent headers in ldiskfs_ext_search_right()
4016 can cause the filesystem to be remounted read-only.
4021 Description: lfsck built against 1.4.x cannot run against 1.6.0 lustre
4022 Details : the definition for OBD_IOC_GETNAME changed in 1.6.0. One of the
4023 few external users of this ioctl number is lfsck's call to
4024 llapi_lov_get_uuids() and this caused lfsck to fail at startup.
4025 Add the old ioctl number to the handler so both old and new
4030 Description: parallel lock callbacks
4031 Details : Instead of sending blocking and completion callbacks as separated
4032 requests, adding them to a set and sending in parallel.
4036 Description: Disable most debugging by default
4037 Details : To improve performance, disable most logging (for debug purposes)
4038 by default. VFSTRACE, RPCTRACE, and DLMTRACE are now off by
4039 default, and HA includes fewer messages.
4043 Frequency : only run on patchless client.
4044 Description: use do_facet on sanity.sh for test handling recoverables errors
4045 Details : use do_facet instead of direct use sysctl for set fail_loc on OST
4049 Description: hash tables for lists of nids, connections and uuids
4050 Details : Hash tables noticeably help when a lot of clients connect to a
4051 server, to faster identify duplicate connections or reconnects,
4052 also to faster find export to evict in manual eviction case.
4056 Description: Sometimes, when the server evict a client, and the client will
4057 not be evicted as soon as possible.
4058 Details : In enqueue req, the error was returned by intent, instead of
4059 rq_status which make ptlrpc layer not detect this error, and
4060 does not evict the client. So enqueue error should be returned
4064 Frequency : only at startup
4066 Description: Delay client connections to MDT until fist MDT->OST connect
4067 Details : If a client tried to create a new file before the MDT had
4068 connected to any OSTs, the create would return EIO. Now
4069 the client will simply block until the MDT connects to the
4070 first OST and the create can succeed.
4073 Frequency : only for SLES9
4075 Description: Routinely utilize latest Quadrics drivers in CFS releases
4076 Details : Update patch qsnet-suse-2.6.patch.
4079 Frequency : only for sles10
4081 Description: Update patches for SLES 10 SP1 kernel.
4082 Details : Update the patch vfs_intent-2.6-sles10.patch.
4087 Description: Routinely utilize latest Quadrics drivers in CFS releases
4088 Details : Update patch qsnet-rhel4-2.6.patch.
4091 Frequency : at statup only
4093 Description: mds_lov_synchronize race leads to various problems
4094 Details : simultaneous MDT->OST connections at startup can cause the
4095 sync to abort, leaving the OSC in a bad state.
4100 Description: Using wrong pointer in osc_brw_prep_request
4101 Details : Access to array[-1] can produce panic if kernel compiled with
4102 CONFIG_PAGE_ALLOC enabled
4104 Severity : enhancement
4106 Description: Async OSC create to avoid the blocking unnecessarily.
4107 Details : If a OST has no remain object, system will block on the creating
4108 when need to create a new object on this OST. Now, ways use
4109 pre-created objects when available, instead of blocking on an
4110 empty osc while others are not empty. If we must block, we block
4111 for the shortest possible period of time.
4115 Frequency : only in recovery
4116 Description: Mark OST as early accessible if his start SYNC.
4117 Details : osc_precreate return flag early accessible if oscc marked as
4118 OSCC_FLAG_SYNC_IN_PROGRESS.
4123 Description: Sometimes precreate code can triger create object on wrong ost
4124 Details : Wrong protected or not not restored variables aftre precreate loop
4125 can produce creation object on wrong ost.
4128 Frequency : oss recovery
4130 Description: llog_commit_thread cleanup should sync with llog_commit_thread
4132 Details : llog_commit_thread_count should be synced between llog_commit
4133 start and cleanup, so new llog_commit thread should not be started
4134 when llog_commit threads being stopped to avoid accessing some
4137 Severity : enhancement
4139 Description: Add printing inode info into message about error in writepage.
4141 --------------------------------------------------------------------------------
4143 2007-05-03 Cluster File Systems, Inc. <info@clusterfs.com>
4148 Frequency : on some architectures
4150 Description: 1.6 client sometimes fails to mount from a 1.4 MDT
4151 Details : Uninitialized flags sometimes cause configuration commands to
4155 Frequency : patchless clients only
4157 Description: missing __iget() symbol export
4158 Details : The __iget() symbol export is missing. To avoid the need for
4159 this on patchless clients the deathrow inode reaper is turned
4160 off, and we depend on the VM to clean up old inodes. This
4161 dependency was during via the fix for bug 12181.
4166 Description: sanity.sh fail: test_52b
4167 Details : The ll_inode_to_ext_flags() has a glitch which makes MDS return
4168 incorrect inode's flags to client.
4170 --------------------------------------------------------------------------------
4172 2007-04-19 Cluster File Systems, Inc. <info@clusterfs.com>
4174 * CONFIGURATION CHANGE. This version of Lustre WILL NOT
4175 INTEROPERATE with older versions automatically. In many cases a
4176 special upgrade step is needed. Please read the
4177 user documentation before upgrading any part of a 1.4.x system.
4178 * WARNING: Lustre configuration and startup changes are required with
4179 this release. See https://mail.clusterfs.com/wikis/lustre/MountConf
4181 * Support for kernels:
4182 2.4.21-47.0.1.EL (RHEL 3),
4183 2.6.5-7.283 (SLES 9),
4184 2.6.9-42.0.10.EL (RHEL 4),
4185 2.6.12.6 vanilla (kernel.org),
4186 2.6.16.27-0.9 (SLES10)
4187 * Client support for unpatched kernels:
4188 (see https://mail.clusterfs.com/wikis/lustre/PatchlessClient)
4189 2.6.16 - 2.6.19 vanilla (kernel.org),
4190 2.6.9-42.0.8EL (RHEL 4)
4191 * Recommended e2fsprogs version: 1.39.cfs6
4192 * Note that reiserfs quotas are disabled on SLES 10 in this kernel
4195 Severity : enhancement
4197 Description: Async OSC create to avoid the blocking unnecessarily.
4198 Details : If a OST has no remain object, system will block on the creating
4199 when need to create a new object on this OST. Now, ways use
4200 pre-created objects when available, instead of blocking on an
4201 empty osc while others are not empty. If we must block, we block
4202 for the shortest possible period of time.
4204 Severity : enhancement
4206 Description: MountConf
4207 Details : Lustre configuration is now managed via mkfs and mount
4208 commands instead of lmc and lconf. New obd types (MGS, MGC)
4209 are added for dynamic configuration management. See
4210 https://mail.clusterfs.com/wikis/lustre/MountConf for
4213 Severity : enhancement
4215 Description: dynamic OST addition
4216 Details : OSTs can now be added to a live filesystem
4218 Severity : enhancement
4220 Description: startup order invariance
4221 Details : MDTs and OSTs can be started in any order. Clients only
4222 require the MDT to complete startup.
4224 Severity : enhancement
4226 Description: parallel, asynchronous orphan cleanup
4227 Details : orphan cleanup is now performed in separate threads for each
4228 OST, allowing parallel non-blocking operation.
4230 Severity : enhancement
4232 Description: optimized stripe assignment
4233 Details : stripe assignments are now made based on ost space available,
4234 ost previous usage, and OSS previous usage, in order to try
4235 to optimize storage space and networking resources.
4237 Severity : enhancement
4239 Description: Permanently set tunables
4240 Details : All writable /proc/fs/lustre tunables can now be permanently
4241 set on a per-server basis, at mkfs time or on a live
4244 Severity : enhancement
4246 Description: Lustre message v2
4247 Details : Add lustre message format v2.
4249 Severity : enhancement
4251 Description: client OST exclusion list
4252 Details : Clients can be started with a list of OSTs that should be
4253 declared "inactive" for known non-responsive OSTs.
4257 Description: ENOENT returned for valid filehandle during dbench.
4258 Details : Check if a directory has children when invalidating dentries
4259 associated with an inode during lock cancellation. This fixes
4260 an incorrect ENOENT sometimes seen for valid filehandles during
4261 testing with dbench.
4264 Frequency : SFS test only (otherwise harmless)
4266 Description: SPEC SFS validation failure on NFS v2 over lustre.
4267 Details : Changes the blocksize for regular files to be 2x RPC size,
4268 and not depend on stripe size.
4270 Severity : enhancement
4272 Description: fine-grained SMP locking inside DLM
4273 Details : Improve DLM performance on SMP systems by removing the single
4274 per-namespace lock and replace it with per-resource locks.
4276 Severity : enhancement
4278 Description: don't hold multiple extent locks at one time
4279 Details : To avoid client eviction during large writes, locks are not
4280 held on multiple stripes at one time or for very large writes.
4281 Otherwise, clients can block waiting for a lock on a failed OST
4282 while holding locks on other OSTs and be evicted.
4284 Severity : enhancement
4286 Description: Multiple MD RPCs in flight.
4287 Details : Further unserialise some read-only MDT RPCs - learn about intents.
4288 To avoid overly-overloading MDT, introduce a limit on number of
4289 MDT RPCs in flight for a single client and add /proc controls
4290 to adjust this limit.
4292 Severity : enhancement
4294 Description: client read/write statistics
4295 Details : Add client read/write call usage stats for performance
4296 analysis of user processes.
4297 /proc/fs/lustre/llite/*/offset_stats shows non-sequential
4298 file access. extents_stats shows chunk size distribution.
4299 extents_stats_per_process show chunk size distribution per
4302 Severity : enhancement
4304 Description: per-client statistics on server
4305 Details : Add ldlm and operations statistics for each client in
4306 /proc/fs/lustre/mds|obdfilter/*/exports/
4308 Severity : enhancement
4310 Description: improved MDT statistics
4311 Details : Add detailed MDT operations statistics in
4312 /proc/fs/lustre/mds/*/stats
4314 Severity : enhancement
4316 Description: VFS operations stats
4317 Details : Add client VFS call stats, trackable by pid, ppid, or gid
4318 /proc/fs/lustre/llite/*/stats_track_[pid|ppid|gid]
4323 Description: Fix client-side osc byte counters
4324 Details : The osc read/write byte counters in
4325 /proc/fs/lustre/osc/*/stats are now working
4328 Frequency : always as root on SLES
4330 Description: Failure of copying files with lustre special EAs.
4331 Details : Client side always return success for setxattr call for lustre
4332 special xattr (currently only "trusted.lov").
4337 Description: Refcount LNET uuids
4338 Details : The global LNET uuid list grew linearly with every startup;
4339 refcount repeated list entries instead of always adding to
4342 Severity : enhancement
4344 Description: Dynamic service threads
4345 Details : Within a small range, start extra service threads
4346 automatically when the request queue builds up.
4349 Frequency : mixed-endian client/server environments
4351 Description: mixed-endian crashes
4352 Details : The new msg_v2 system had some failures in mixed-endian
4355 Severity : enhancement
4357 Description: Easy OST removal
4358 Details : OSTs can be permanently deactivated with e.g. 'lctl
4359 conf_param lustre-OST0001.osc.active=0'
4361 Severity : enhancement
4363 Description: MGS proc entries
4364 Details : Added basic proc entries for the MGS showing what filesystems
4367 Severity : enhancement
4369 Description: provide MGS failover
4370 Details : Added config lock reacquisition after MGS server failover.
4372 Severity : enhancement
4374 Description: add Linux 2.4 support
4375 Details : Added support for RHEL 2.4.21 kernel for 1.6 servers and clients
4379 Description: a large application tries to do I/O to the same resource and dies
4380 in the middle of it.
4381 Details : Check the req->rq_arrival time after the call to
4382 ost_brw_lock_get(), but before we do anything about
4383 processing it & sending the BULK transfer request. This
4384 should help move old stale pending locks off the queue as
4385 quickly as obd_timeout.
4388 Frequency : when an incorrect nid is specified during startup
4390 Description: ptlrpc connect to non-existant node causes kernel crash
4391 Details : LNET can't be re-entered from an event callback, which
4392 happened when we expire a message after the export has been
4393 cleaned up. Instead, hand the zombie cleanup off to another
4396 Severity : enhancement
4398 Description: plain/inodebits lock performance improvement
4399 Details : Grouping plain/inodebits in granted list by their request modes
4400 and bits policy, thus improving the performance of search through
4404 Frequency : only if OST filesystem is corrupted
4406 Description: client incorrectly hits assertion in ptlrpc_replay_req()
4407 Details : for a short time RPCs with bulk IO are in the replay list,
4408 but replay of bulk IOs is unimplemented. If the OST filesystem
4409 is corrupted due to disk cache incoherency and then replay is
4410 started it is possible to trip an assertion. Avoid putting
4411 committed RPCs into the replay list at all to avoid this issue.
4414 Frequency : liblustre (e.g. catamount) on a large cluster with >= 8 OSTs/OSS
4416 Description: System hang on startup
4417 Details : This bug allowed the liblustre (e.g. catamount) client to
4418 return to the app before handling all startup RPCs. This
4419 could leave the node unresponsive to lustre network traffic
4420 and manifested as a server ptllnd timeout.
4422 Severity : enhancement
4424 Description: Add "/proc/sys/lustre/debug_peer_on_timeout"
4425 Details : liblustre envirable: LIBLUSTRE_DEBUG_PEER_ON_TIMEOUT
4426 boolean to control whether to print peer debug info when a
4427 client's RPC times out.
4430 Frequency : only for kernels with patches from Lustre below 1.4.3
4432 Description: Remove old rdonly API
4433 Details : Remove old rdonly API which unused from at least lustre 1.4.3
4436 Frequency : only for devices with external journals
4438 Description: Set external device read-only also
4439 Details : During a commanded failover stop, we set the disk device
4440 read-only while the server shuts down. We now also set any
4441 external journal device read-only at the same time.
4444 Frequency : when upgrading from 1.4 while trying to change parameters
4446 Description: The wrong (new) MDC name was used when setting parameters for
4447 upgraded MDT's. Also allows changing of OSC (and MDC)
4448 parameters if --writeconf is specified at tunefs upgrade time.
4451 Frequency : when setting specific ost indicies
4453 Description: QOS code breaks on skipped indicies
4454 Details : Add checks for missing OST indicies in the QOS code, so OSTs
4455 created with --index need not be sequential.
4457 Severity : enhancement
4459 Description: Add uninit_groups feature to ldiskfs2 to speed up e2fsck
4460 Details : The uninit_groups feature works in conjunction with the kernel
4461 filesystem code (ldiskfs2 only) and e2fsprogs-1.39-cfs6 to speed
4462 up the pass1 processing of e2fsck. This is a read-only feature
4463 in ldiskfs2 only, so older kernels and current ldiskfs cannot
4464 mount filesystems that have had this feature enabled.
4466 Severity : enhancement
4468 Description: Improve multi-block allocation algorithm to avoid fragmentation
4469 Details : The mballoc3 code (ldiskfs2 only) adds new mechanisms to improve
4470 allocation locality and avoid filesystem fragmentation.
4472 ------------------------------------------------------------------------------
4474 2007-04-01 Cluster File Systems, Inc. <info@clusterfs.com>
4476 * Support for kernels:
4477 2.4.21-47.0.1.EL (RHEL 3)
4478 2.6.5-7.283 (SLES 9)
4479 2.6.9-42.0.10.EL (RHEL 4)
4480 2.6.12.6 vanilla (kernel.org)
4481 2.6.16.27-0.9 (SLES 10)
4482 * Recommended e2fsprogs version: 1.39.cfs5
4484 * Note that reiserfs quotas are disabled on SLES 10 in this kernel
4488 Frequency : occasional, depends on client load and configuration
4489 Bugzilla : 12181, 12203
4490 Description: data loss for recently-modified files
4492 Details : In some cases it is possible that recently written or created
4493 files may not be written to disk in a timely manner (this should
4494 normally be within 30s unless client IO load is very high).
4495 The problem appears as zero-length files or files that are a
4496 multiple of 1MB in size after a client crash or client eviction
4497 that are missing data at the end of the file.
4499 This problem is more likely to be hit on clients where files are
4500 repeatedly created and unlinked in the same directory, clients
4501 have a large amount of RAM, have many CPUs, the filesystem has
4502 many OSTs, the clients are rebooted frequently, and/or the files
4503 are not accessed by other nodes after being written.
4505 The presence of the problem can be detected by looking at
4506 /proc/sys/fs/inode-state. If the first number (nr_inodes) is
4507 smaller than the second (nr_unused) then dirty files will not
4508 be flushed automatically to disk. "sync; sleep 10" should be
4509 run several times on the node before unmounting it to update
4510 Lustre (this is also safe to run on nodes without this problem).
4512 There is also a related kernel bug in the RHEL4 4 2.6.9 kernel
4513 that can cause this same problem, so customers using that kernel
4514 also need to update the kernel in addition to Lustre. In order
4515 to properly fix this bug, the RHEL3 2.4.21 kernel is also updated.
4517 It is normal that files written just before a client crash (less
4518 than 30s) may not yet have been flushed to disk, even for local
4522 Frequency : frequent on thin XT3 nodes
4524 Description: UUID collision on thin XT3 Linux nodes
4525 Details : UUIDs on Compute Node Linux XT3 nodes were not generated
4526 randomly, since we relied on an insufficiently-seeded PRNG.
4531 Description: OSS hangs after "All ost request buffers busy"
4532 Details : A deadlock between quota and journal operations caused OSS
4533 hangs after printing "All ost request buffers busy."
4536 Frequency : always on liblustre builds
4538 Description: Cleanup compiler warnings on liblustre
4541 Frequency : always on liblustre builds on XT3
4543 Description: LC_CONFIG_CDEBUG don't run while build liblustre on XT3.
4547 Description: Addition of EXT3_FEATURE_RO_COMPAT_DIR_NLINKS flag for
4548 > 32000 subdirectories
4549 Details : Add EXT3_FEATURE_RO_COMPAT_DIR_NLINK flag to
4550 EXT3_FEATURE_RO_COMPAT_SUPP. This flag will be set whenever
4551 subdirectory count crosses 32000. This will aid e2fsck to
4552 correctly handle more than 32000 subdirectories.
4555 Frequency : liblustre (e.g. catamount) on a large cluster with >= 8 OSTs/OSS
4557 Description: System hang on startup
4558 Details : This bug allowed the liblustre (e.g. catamount) client to
4559 return to the app before handling all startup RPCs. This
4560 could leave the node unresponsive to lustre network traffic
4561 and manifested as a server ptllnd timeout.
4563 Severity : enhancement
4565 Description: Add "/proc/sys/lustre/debug_peer_on_timeout"
4566 (liblustre envirable: LIBLUSTRE_DEBUG_PEER_ON_TIMEOUT)
4567 boolean to control whether to print peer debug info when a
4568 client's RPC times out.
4573 Description: make O_SYNC working on 2.6 kernels
4574 Details : 2.6 kernels use different method for mark pages for write,
4575 so need add a code to lustre for O_SYNC work.
4580 Description: Failure to close file and release space on NFS
4581 Details : Put inode details into lock acquired in ll_intent_file_open.
4582 Use mdc_intent_lock in ll_intent_open to properly
4583 detect all kind of errors unhandled by mdc_enqueue.
4588 Description: proc file read during shutdown sometimes raced obd removal,
4590 Details : Add lock to prevent obd access after proc file removal.
4593 Frequency : Only for files larger than 4GB on 32-bit clients.
4595 Description: improperly doing page alignment of locks
4596 Details : Modify lustre core code to use CFS_PAGE_* defines instead of
4597 PAGE_*. Make CFS_PAGE_MASK a 64-bit mask.
4602 Description: RPCs being resent when they shouldn't be
4603 Details : Some RPCs that should not be resent are being resent. This
4604 can cause inconsistencies in the RPC state machine. Do not
4605 resend such requests.
4608 Frequency : rare, only with NFS export
4610 Description: Crash on NFS re-export node
4611 Details : under very unusual load conditions an assertion is hit in
4612 ll_intent_file_open()
4615 Frequency : only if OST filesystem is corrupted
4617 Description: client incorrectly hits assertion in ptlrpc_replay_req()
4618 Details : for a short time RPCs with bulk IO are in the replay list,
4619 but replay of bulk IOs is unimplemented. If the OST filesystem
4620 is corrupted due to disk cache incoherency and then replay is
4621 started it is possible to trip an assertion. Avoid putting
4622 committed RPCs into the replay list at all to avoid this issue.
4627 Description: large O_DIRECT requests fail under memory pressure/fragmentation
4628 Details : Large single O_DIRECT read and write calls can fail to allocate
4629 a sufficiently large buffer to process the request. In case of
4630 allocation failure the allocation is retried with a smaller
4631 buffer and broken into smaller requests.
4633 Severity : enhancement
4635 Description: Add -o localflock option to simulate old noflock behaviour.
4636 Details : This will achieve local-only flock/fcntl locks coherentness.
4641 Description: versioning check is incomplete
4642 Details : Checking the version difference of client vs. server, report
4643 error if the gap is too big.
4648 Description: add support PG_writeback bit
4649 Details : add support for PG_writeback bit for Lustre, for more carefull
4650 work with page cache in 2.6 kernel. This also fix some deadlocks
4651 and remove hack for work O_SYNC with 2.6 kernel.
4653 Severity : enhancement
4655 Description: Add uninit_groups feature to ldiskfs2 to speed up e2fsck
4656 Details : The uninit_groups feature works in conjunction with the kernel
4657 filesystem code (ldiskfs2 only) and e2fsprogs-1.39-cfs6 to speed
4658 up the pass1 processing of e2fsck. This is a read-only feature
4659 in ldiskfs2 only, so older kernels and current ldiskfs cannot
4660 mount filesystems that have had this feature enabled.
4662 Severity : enhancement
4664 Description: Improve multi-block allocation algorithm to avoid fragmentation
4665 Details : The mballoc3 code (ldiskfs2 only) adds new mechanisms to improve
4666 allocation locality and avoid filesystem fragmentation.
4668 ------------------------------------------------------------------------------
4670 2007-02-09 Cluster File Systems, Inc. <info@clusterfs.com>
4672 * Support for kernels:
4673 2.6.9-42.0.3.EL (RHEL 4)
4674 2.6.5-7.276 (SLES 9)
4675 2.4.21-47.0.1.EL (RHEL 3)
4676 2.6.12.6 vanilla (kernel.org)
4677 2.6.16.21-0.8 (SLES10)
4678 * Recommended e2fsprogs version: 1.39.cfs2-0
4680 * The backwards-compatible /proc/sys/portals symlink has been removed
4681 in this release. Before upgrading, please ensure that you change
4682 any configuration scripts or /etc/sysctl.conf files that access
4683 /proc/sys/portals/* or sysctl portals.* to use the corresponding
4684 entry in /proc/sys/lnet or sysctl lnet.*. This change can be made
4685 in advance of the upgrade on any system running Lustre 1.4.6 or
4686 newer, since /proc/sys/lnet was added in that version.
4687 * Note that reiserfs quotas are disabled on SLES 10 in this kernel
4691 Frequency : only when quota is used
4693 Description: avoid scanning export list for quota master
4694 Details : Change the algorithms to avoid scanning export list in order
4695 to improve the efficiency.
4698 Frequency : MDS failover only, very rarely
4700 Description: "went back in time" messages on mds failover
4701 Details : The greatest transno may be lost when the current operation
4702 finishes with an error (transno==0) and the client's last_rcvd
4703 record is over-written. Save the greatest transno in the
4704 mds_last_transno for this case.
4707 Frequency : always for specific kernels and striping counts
4709 Description: client may get "Matching packet too big" without ACL support
4710 Details : Clients compiled without CONFIG_FS_POSIX_ACL get an error message
4711 when trying to access files in certain configurations. The
4712 clients should in fact be denied when mounting because they do
4713 not understand ACLs.
4716 Frequency : Cray XT3 with more than 4000 clients and multiple jobs
4718 Description: many clients connecting with IO in progress causes connect timeouts
4719 Details : Avoid synchronous journal commits to avoid delays caused by many
4720 clients connecting/disconnecting when bulk IO is in progress.
4721 Queue liblustre connect requests on OST_REQUEST_PORTAL instead of
4722 OST_IO_PORTAL to avoid delays behind potentially many pending
4726 Frequency : occasionally with multiple writers to a single file
4728 Description: shared writes to file may result in wrong size reported by stat()
4729 Details : Allow growing of kms when extent lock is cancelled
4732 Frequency : always with random mmap IO to multi-striped file
4734 Description: mmap write might be lost if we are writing to a 'hole' in stripe
4735 Details : Only if the hole is at the end of OST object so that kms is too
4736 small. Fix is to increase kms accordingly in ll_nopage.
4739 Frequency : rare, only if OST filesystem is inconsistent with MDS filesystem
4741 Description: writes to a missing object would leak memory on the OST
4742 Details : If there is an inconsistency between the MDS and OST filesystems,
4743 such that the MDS references an object that doesn't exist, writes
4744 to that object will leak memory due to incorrect cleanup in the
4745 error handling path, eventually running out of memory on the OST.
4750 Description: Creating too long symlink causes lustre errors
4751 Details : Check symlink and name lengths before sending requests to MDS.
4754 Frequency : only if flock is enabled (not on by default)
4756 Description: posix locks not released on fd closure on 2.6.9+
4757 Details : We failed to add posix locks to list of inode locks on 2.6.9+
4758 kernels, this caused such locks not to be released on fd close and
4759 then assertions on fs unmount about still used locks.
4762 Frequency : MDS failover only, very rarely
4764 Description: clients may get ASSERTION(granted_lock != NULL)
4765 Details : When request was taking a long time, and a client was resending
4766 a getattr by name lock request. The were multiple lock requests
4767 with the same client lock handle and
4768 mds_getattr_name->fixup_handle_for_resent_request found one of the
4769 lock handles but later failed with ASSERTION(granted_lock != NULL).
4774 Description: handle->h_buffer_credits > 0, assertion failure
4775 Details : h_buffer_credits is zero after truncate, causing assertion
4776 failure. This patch extends the transaction or creates a new
4780 Frequency : NFS re-export or patchless client
4781 Bugzilla : 11179, 10796
4782 Description: Crash on NFS re-export node (__d_move)
4783 Details : We do not want to hash the dentry if we don't have a lock.
4784 But if this dentry is later used in d_move, we'd hit uninitialised
4785 list head d_hash, so we just do this to init d_hash field but
4786 leave dentry unhashed.
4789 Frequency : NFS re-export or patchless client
4791 Description: NFS exports has problem with symbolic link
4792 Details : lustre client didn't properly install dentry when re-exported
4793 to NFS or running patchless client.
4796 Frequency : NFS re-export or patchless client
4798 Description: Various nfs/patchless fixes.
4799 Details : fixes reuse disconected alias for lookup process - this fixes
4800 warning "find_exported_dentry: npd != pd",
4801 fix permission error with open files at nfs.
4802 fix apply umask when do revalidate.
4805 Frequency : occasional
4807 Description: Crash on NFS re-export node
4808 Details : calling clear_page() on the wrong pointer triggered oops in
4809 generic_mapping_read().
4812 Frequency : rarely, using O_DIRECT IO
4814 Description: unaligned directio crashes client with LASSERT
4815 Details : check for unaligned buffers before trying any requests.
4818 Frequency : rarely, using CFS RAID5 patches in non-standard kernel series
4820 Description: stale data returned from RAID cache
4821 Details : If only a small amount of IO is done to the RAID device before
4822 reading it again it is possible to get stale data from the RAID
4823 cache instead of reading it from disk.
4826 Frequency : always for sles10 kernel
4828 Description: sles10 support
4829 Details : ll_follow_link: compile fixes and using of nd_set_link
4830 under newer kernels.
4833 Frequency : depends on arch, kernel and compiler version, always on sles10
4836 Description: recursive or deep enough symlinks cause stack overflow
4837 Details : getting rid of large stack-allocated variable in
4841 Frequency : depends on hardware
4843 Description: lustre write performance loss in the SLES10 kernel
4844 Details : the performance loss is caused by using of write barriers in the
4845 ext3 code. The SLES10 kernel turns barrier support on by
4846 default. The fix is to undo that change for ldiskfs.
4848 ------------------------------------------------------------------------------
4850 2006-12-09 Cluster File Systems, Inc. <info@clusterfs.com>
4852 * Support for kernels:
4853 2.6.9-42.0.3EL (RHEL 4)
4854 2.6.5-7.276 (SLES 9)
4855 2.4.21-47.0.1.EL (RHEL 3)
4856 2.6.12.6 vanilla (kernel.org)
4860 Frequency : quota enabled and large files being deleted
4862 Description: releasing more than 4GB of quota at once hangs OST
4863 Details : If a user deletes more than 4GB of files on a single OST it
4864 will cause the OST to spin in an infinite loop. Release
4865 quota in < 4GB chunks, or use a 64-bit value for 1.4.7.1+.
4870 Description: statfs data retrieved from /proc may be stale or zero
4871 Details : When reading per-device statfs data from /proc, in the
4872 {kbytes,files}_{total,free,avail} files, it may appear
4873 as zero or be out of date.
4876 Frequency : systems with MD RAID1 external journal devices
4878 Description: lconf's call to blkid is confused by RAID1 journal devices
4879 Details : Use the "blkid -l" flag to locate the MD RAID device instead
4880 of returning all block devices that match the journal UUID.
4883 Frequency : always, for aggregate stripe size over 4GB
4885 Description: "lfs setstripe" fails assertion when setting 4GB+ stripe width
4886 Details : Using "lfs setstripe" to set stripe size * stripe count over 4GB
4887 will fail the kernel with "ASSERTION(lsm->lsm_xfersize != 0)"
4890 Frequency : always if "lfs find" used on a local file/directory
4892 Description: "lfs find" segfaults if used on a local file/directory
4893 Details : The case where a directory component was not specified wasn't
4894 handled correctly. Handle this properly.
4897 Frequency : always on ppc64
4899 Description: the write to an ext3 filesystem mounted with mballoc got stuck
4900 Details : ext3_mb_generate_buddy() uses find_next_bit() which does not
4901 perform endianness conversion.
4904 Frequency : rarely (truncate to non-zero file size after write under load)
4905 Bugzilla : 10730, 10687
4906 Description: Files padded with zeros to next 4K multiple
4907 Details : With filesystems mounted using the "extents" option (2.6 kernels)
4908 it is possible that files that are truncated to a non-zero size
4909 immediately after being written are filled with zero bytes beyond
4910 the truncated size. No file data is lost.
4912 Severity : enhancement
4914 Description: Allow recovery/failover for liblustre clients.
4915 Details : liblustre clients were unaware of failover configurations until
4918 Severity : enhancement
4920 Description: user file locks should fail when not mounting with flock option
4921 Details : Set up an error-returning stub in ll_file_operations.lock field
4922 to prevent incorrect behaviour when client is mounted without
4923 flock option. Also, set up properly f_op->flock field for
4927 Frequency : always on ia64
4929 Description: "lfs df" loops on printing out MDS statfs information
4930 Details : The obd_ioctl_data was not initialized and in some systems
4931 this caused a failure during the ioctl that did not return
4932 an error. Initialize the struct and return an error on failure.
4935 Frequency : SLES 9 only
4937 Description: Error of copying files with lustre special EAs as root
4938 Details : Client side always return success for setxattr call for lustre
4939 special xattr (currently only "trusted.lov").
4942 Frequency : rarely on clusters with both ia64+i386 clients
4944 Description: ia64+i686 clients doing shared IO on the same file may LBUG
4945 Details : In rare cases when both ia64+i686 (or other mixed-PAGE_SIZE)
4946 clients are doing concurrent writes to the same file it is
4947 possible that the ia64 clients may LASSERT because the OST
4948 extent locks are not PAGE_SIZE aligned. Ensure that grown
4949 locks are always aligned on the request boundary.
4952 Frequency : specific use, occasional
4954 Description: Overwriting in use executable truncates on-disk binary image
4955 Details : If one node attempts to overwrite an executable in use by
4956 another node, we now correctly return ETXTBSY instead of
4957 truncating the file.
4959 Severity : enhancement
4961 Description: Async OSC create to avoid the blocking unnecessarily.
4962 Details : If a OST has no remain object, system will block on the creating
4963 when need to create a new object on this OST. Now, ways use
4964 pre-created objects when available, instead of blocking on an
4965 empty osc while others are not empty. If we must block, we block
4966 for the shortest possible period of time.
4971 Description: chmod on Lustre root is propagated to other clients
4972 Details : Re-validate root's dentry in ll_lookup_it to avoid having it
4973 invalid by the follow_mount time.
4976 Frequency : liblustre clients only
4978 Description: Race in 'instant cancel' lock handling could lead to such locks
4979 never to be granted in case of SMP MDS
4980 Details : Do not destroy not yet granted but cbpending locks in
4984 Frequency : replay/resend of open
4986 Description: non null lock assetion failure in mds_intent_policy
4987 Details : Trying to replay/resend lockless open requests resulted in
4988 mds_open() returning 0 with no lock. Now it sets a flag if
4989 a lock is going to be returned.
4991 Severity : enhancement
4993 Description: Checksum enhancements
4994 Details : New checksum enhancements allow for resending RPCs that failed
4997 Severity : enhancement
4999 Description: Tunables on number of dirty pages in cacche
5000 Details : Allow to set limit on number of dirty pages cached.
5005 Description: client crash on unmount - lock still has references
5006 Details : In some error handling cases it was possible to leak a lock
5007 reference on a client while accessing a file. This was not
5008 harmful to the client during operation, but would cause the
5009 client to crash when the filesystem is unmounted.
5012 Frequency : specific case, rare
5014 Description: ETXTBSY on mds though file not in use by client
5015 Details : ETXTBSY is no longer incorrectly returned when attempting to
5016 chmod or chown a directory that the user previously tried to
5017 execute or a currently-executing binary.
5020 Frequency : extremely rare except on liblustre-based clients
5022 Description: Lustre space not freed when files are deleted
5023 Details : Clean up open-unlinked files after client eviction. Previously
5024 the unlink was skipped and the files remained as orphans.
5029 Description: OST failure "would be an LBUG" in waiting_locks_callback()
5030 Details : In some cases it was possible to send a blocking callback to
5031 a client doing a glimpse, even though that client didn't get
5032 a lock granted. When the glimpse lock is cancelled on the OST
5033 the freed lock is left on the waiting list and corrupted the list.
5036 Frequency : all core dumps
5038 Description: Broke core dumps to lustre
5039 Details : Negative dentry may be unhashed if parent does not have UPDATE
5040 lock, but some callers, e.g. do_coredump, expect dentry to be
5041 hashed after successful create, hash it in ll_create_it.
5043 ------------------------------------------------------------------------------
5045 2006-09-13 Cluster File Systems, Inc. <info@clusterfs.com>
5047 * Support for kernels:
5048 2.6.9-42.0.2.EL (RHEL 4)
5049 2.6.5-7.276 (SLES 9)
5050 2.4.21-40.EL (RHEL 3)
5051 2.6.12.6 vanilla (kernel.org)
5055 Frequency : always on RHEL 3
5057 Description: Number of open files grows over time
5058 Details : The number of open files grows over time, whether or not
5059 Lustre is started. This was due to a filp leak introduced
5060 by one of our kernel patches.
5062 ------------------------------------------------------------------------------
5064 2006-08-20 Cluster File Systems, Inc. <info@clusterfs.com>
5066 * Support for kernels:
5067 2.6.9-42.EL (RHEL 4)
5068 2.6.5-7.267 (SLES 9)
5069 2.4.21-40.EL (RHEL 3)
5070 2.6.12.6 vanilla (kernel.org)
5075 Bugzilla : 5719, 9635, 9792, 9684
5076 Description: OST (or MDS) trips assertions in (re)connection under heavy load
5077 Details : If a server is under heavy load and cannot reply to new
5078 connection requests before the client resends the (re)connect,
5079 the connection handling code can behave badly if two service
5080 threads are concurrently handing separate (re)connections from
5081 the same client. Add better locking to the connection handling
5082 code, and ensure that only a single connection will be processed
5083 for a given client UUID, even if the lock is dropped.
5085 Severity : enhancement
5087 Description: add TCP zero-copy support to kernel
5088 Details : Add support to the kernel TCP stack to allow zero-copy bulk
5089 sends if the hardware supports scatter-gather and checksumming.
5090 This allows socklnd to do client-write and server-read more
5091 efficiently and reduce CPU utilization from skbuf copying.
5094 Frequency : only if NFS exporting from client
5096 Description: NULL pointer deref in ll_iocontrol() if chattr mknod file
5097 Details : If setting attributes on a file created under NFS that had
5098 never been opened it would be possible to oops the client
5099 if the file had no objects.
5103 Bugzilla : 9326, 10402, 10897
5104 Description: client crash in ptlrpcd_wake() thread when sending async RPC
5105 Details : It is possible that ptlrpcd_wake() dereferences a freed async
5106 RPC. In rare cases the ptlrpcd thread alread processed the RPC
5107 before ptlrpcd_wake() was called and the request was freed.
5110 Frequency : always for liblustre
5112 Description: liblustre client does MDS+OSTs setattr RPC for each write
5113 Details : When doing a write from a liblustre client, the client
5114 incorrectly issued an RPC to the MDS and each OST the file was
5115 striped over in order to update the timestamps. When writing
5116 with small chunks and many clients this could overwhelm the MDS
5117 with RPCs. In all cases it would slow down the write because
5118 these RPCs are unnecessary.
5120 Severity : enhancement
5122 Description: allow number of MDS service threads to be changed at module load
5123 Details : It is now possible to change the number of MDS service threads
5124 running. Adding "options mds mds_num_threads={N}" to the MDS's
5125 /etc/modprobe.conf will set the number of threads for the next
5126 time Lustre is restarted (assuming the "mds" module is also
5127 reloaded at that time). The default number of threads will
5128 stay the same, 32 for most systems.
5133 Description: OST crash if filesystem is unformatted or corrupt
5134 Details : If an OST is started on a device that has never been formatted
5135 or if the filesystem is corrupt and cannot even mount then the
5136 error handling cleanup routines would dereference a NULL pointer.
5141 Description: NULL pointer deref in llap_from_page.
5142 Details : get_cache_page_nowait can return a page with NULL (or otherwise
5143 incorrect) mapping if the page was truncated/reclaimed while it was
5144 searched for. Check for this condition and skip such pages when
5145 doing readahead. Introduce extra check to llap_from_page() to
5146 verify page->mapping->host is non-NULL (so page is not anonymous).
5149 Frequency : Sometimes when using sys_sendfile
5151 Description: "page not covered by a lock" warnings from ll_readpage
5152 Details : sendfile called ll_readpage without right page locks present.
5153 Now we introduced ll_file_sendfile that does necessary locking
5154 around call to generic_file_sendfile() much like we do in
5158 Frequency : with certain MDS communication failures at client mount time
5160 Description: NULL pointer deref after failed client mount
5161 Details : a client connection request may delayed by the network layer
5162 and not be sent until after the PTLRPC layer has timed out the
5163 request. If the client fails the mount immediately it will try
5164 to clean up before the network times out the request. Add a
5165 reference from the request import to the obd device and delay
5166 the cleanup until the network drops the request.
5169 Frequency : occasionally during client (re)connect
5171 Description: assertion failure during client (re)connect
5172 Details : processing a client connection request may be delayed by the
5173 client or server longer than the client connect timeout. This
5174 causes the client to resend the connection request. If the
5175 original connection request is replied in this interval, the
5176 client may trip an assertion failure in ptlrpc_connect_interpret()
5177 which thought it would be the only running connect process.
5180 Frequency : only with obd_echo servers and clients that are rebooted
5182 Description: kernel BUG accessing uninitialized data structure
5183 Details : When running an obd_echo server it did not start the ping_evictor
5184 thread, and when a client was evicted an uninitialized data
5185 structure was accessed. Start the ping_evictor in the RPC
5186 service startup instead of the OBD startup.
5188 Severity : enhancement
5189 Bugzilla : 10193 (patchless)
5190 Description: Remove dependency on various unexported kernel interfaces.
5191 Details : No longer need reparent_to_init, exit_mm, exit_files,
5192 sock_getsockopt, filemap_populate, FMODE_EXEC, put_filp.
5195 Frequency : rare (only users of deprecated and unsupported LDAP config)
5197 Description: write_conf for zeroconf mount queried LDAP incorrectly for client
5198 Details : LDAP apparently contains 'lustreName' attributes instead of
5199 'name'. A simple remapping of the name is sufficient.
5202 Frequency : rare (only with non-default dump_on_timeout debug enabled)
5204 Description: waiting_locks_callback trips kernel BUG if client is evicted
5205 Details : Running with the dump_on_timeout debug flag turned on makes
5206 it possible that the waiting_locks_callback() can try to dump
5207 the Lustre kernel debug logs from an interrupt handler. Defer
5208 this log dumping to the expired_lock_main() thread.
5210 Severity : enhancement
5212 Description: Support NFS exporting on 2.6 kernels.
5213 Details : Implement non-rawops metadata methods for NFS server to use without
5214 changing NFS server code.
5217 Frequency : very rare (synthetic metadata workload only)
5219 Description: two racing renames might cause an MDS thread to deadlock
5220 Details : Running the "racer" program may cause one MDS thread to rename
5221 a file from being the source of a rename to being the target of
5222 a rename at exactly the same time that another thread is doing
5223 so, and the second thread has already enqueued these locks after
5224 doing a lookup of the target and is trying to relock them in
5225 order. Ensure that we don't try to re-lock the same resource.
5228 Frequency : only very large systems with liblustre clients
5230 Description: slow eviction of liblustre clients with the "evict_by_nid" RPC
5231 Details : Use asynchronous set_info RPCs to send the "evict_by_nid" to
5232 all OSTs in parallel. This allows the eviction of stale liblustre
5233 clients to proceed much faster than if they were done in series,
5234 and also offers similar improvements for other set_info RPCs.
5239 Description: excessive CPU usage during initial read phase on client
5240 Details : During the initial read phase on a client, it would agressively
5241 retry readahead on the file, consuming too much CPU and impacting
5242 performance (since 1.4.5.8). Improve the readahead algorithm
5243 to avoid this, and also improve some other common cases (read
5244 of small files in particular, where "small" is files smaller than
5245 /proc/fs/lustre/llite/*/max_read_ahead_whole_mb, 2MB by default).
5250 Description: MDS crash when receiving packet with unknown intent.
5251 Details : Do not LBUG in unknown intent case, just return -EFAULT
5253 Severity : enhancement
5254 Bugzilla : 9293, 9385
5255 Description: MDS RPCs are serialised on client. This is unnecessary for some.
5256 Details : Do not serialize getattr (non-intent version) and statfs.
5259 Frequency : occasional, when OST network is overloaded/intermittent
5261 Description: client evicted by OST after bulk IO timeout
5262 Details : If a client sends a bulk IO request (read or write) the OST
5263 may evict the client if it is unresposive to its data GET/PUT
5264 request. This is incorrect if the network is overloaded (takes
5265 too long to transfer the RPC data) or dropped the OST GET/PUT
5266 request. There is no need to evict the client at all, since
5267 the pinger and/or lock callbacks will handle this, and the
5268 client can restart the bulk request.
5271 Frequency : Always when mmapping file with no objects
5273 Description: client crashes when mmapping file with no objects
5274 Details : Check that we actually have objects in a file before doing any
5275 operations on objects in ll_vm_open, ll_vm_close and
5281 Description: Request leak when working with deleted CWD
5282 Details : Introduce advanced request refcount tracking for requests
5283 referenced from lustre intent.
5285 Severity : Enhancement
5287 Description: Cache open file handles on client.
5288 Details : MDS now will return special lock along with openhandle, if
5289 requested and client is allowed to hold openhandle, even if unused,
5290 until such a lock is revoked. Helps NFS a lot, since NFS is opening
5291 closing files for every read/write openration.
5293 Severity : Enhancement
5295 Description: Cache open negative dentries on client when possible.
5296 Details : Guard negative dentries with UPDATE lock on parent dir, drop
5297 negative dentries on lock revocation.
5302 Description: Remounting a client read-only wasn't possible with a zconf mount
5303 Details : It wasn't possible to remount a client read-only with llmount.
5305 Severity : enhancement
5306 Description: Include MPICH 1.2.6 Lustre ADIO interface patch
5307 Details : In lustre/contrib/ or /usr/share/lustre in RPM a patch for
5308 MPICH is included to add Lustre-specific ADIO interfaces.
5309 This is based closely on the UFS ADIO layer and only differs
5310 in file creation, in order to allow the OST striping to be set.
5311 This is user-contributed code and not supported by CFS.
5316 Description: extended inode attributes (immutable, append-only) work improperly
5317 when 2.4 and 2.6 kernels are used on client/server or vice versa
5318 Details : Introduce kernel-independent values for these flags.
5320 Severity : enhancement
5323 Description: Allow fractional MB tunings for lustre in /proc/ filesystem.
5324 Details : Many of the /proc/ tunables can only be tuned at a megabyte
5325 granularity. Now, Fractional MB granularity is be supported,
5326 this is very useful for low memory system.
5328 Severity : enhancement
5330 Description: Getattr by fid
5331 Details : Getting a file attributes by its fid, obtaining UPDATE|LOOKUP
5332 locks, avoids extra getattr rpc requests to MDS, allows '/' to
5333 have locks and avoids getattr rpc requests for it on every stat.
5336 Frequency : Always, for filesystems larger than 2TB
5338 Description: ldiskfs crash at mount for filesystem larger than 2TB with mballoc
5339 Details : Kenrel kmalloc limits allocations to 128kB and this prevents
5340 filesystems larger than 2TB to be mounted with mballoc enabled.
5343 Frequency : Always, for 32-bit kernel without CONFIG_LBD and filesystem > 2TB
5345 Description: filesystem corruption for non-standard kernels and very large OSTs
5346 Details : If a 32-bit kernel is compiled without CONFIG_LBD enabled and a
5347 filesystems larger than 2TB is mounted then the kernel will
5348 silently corrupt the start of the filesystem. CONFIG_LBD is
5349 enabled for all CFS-supported kernels, but the possibility of
5350 this happening with a modified kernel config exists.
5352 Severity : enhancement
5354 Description: add client O_DIRECT support for 2.6 kernels
5355 Details : It is now possible to do O_DIRECT reads and writes to files
5356 in the Lustre client mountpoint on 2.6 kernel clients.
5358 Severity : enhancement
5360 Description: parallel glimpse, setattr, statfs, punch, destroy requests
5361 Details : Sends glimpse, setattr, statfs, punch, destroy requests to OSTs in
5362 parallel, not waiting for response from every OST before sending
5363 a rpc to the next OST.
5368 Description: setattr vs write race when updating file timestamps
5369 Details : Client processes that update a file timestamp into the past
5370 right after writing to the file (e.g. tar) it is possible that
5371 the updated file modification time can be reset to the current
5372 time due to a race between processing the setattr and write RPC.
5374 Severity : enhancement
5376 Description: Bring 'lfs find' closer in line with regular Linux find.
5377 Details : lfs find util supports -atime, -mtime, -ctime, -maxdepth, -print,
5378 -print0 options and obtains all the needed info through the lustre
5381 Severity : enhancement
5383 Description: support up to 1024 configured devices on one node
5384 Details : change obd_dev array from statically allocated to dynamically
5385 allocated structs as they are first used to reduce memory usage
5390 Description: Flush dirty partially truncated pages during truncate
5391 Details : Immediatelly flush partially truncated pages in filter_setattr,
5392 this way we completely avoid having any pages in page cache on OST
5393 and can retire ugly workarounds during writes to flush such pages.
5398 Description: i_sem vs transaction deadlock in mds_obd_destroy during unlink.
5399 Details : protect inode from truncation within vfs_unlink() context
5400 just take a reference before calling vfs_unlink() and release it
5401 when parent's i_sem is free.
5404 Frequency : always, if extents are used on OSTs
5406 Description: index ei_leaf_hi (48-bit extension) is not zeroed in extent index
5407 Details : OSTs using the extents format would not zero the high 16 bits of
5408 the index physical block number. This is not a problem for any
5409 OST filesystems smaller than 16TB, and no kernels support ext3
5410 filesystems larger than 16TB yet. This is fixed in 1.4.7 (all
5411 new/modified files) and can be fixed for existing filesystems
5412 with e2fsprogs-1.39-cfs1.
5417 Description: import connection selection may be incorrect if timer wraps
5418 Details : Using a 32-bit jiffies timer with HZ=1000 may cause backup
5419 import connections to be ignored if the 32-bit jiffies counter
5420 wraps. Use a 64-bit jiffies counter.
5423 Frequency : during server recovery
5425 Description: crash after server is denying duplicate export
5426 Details : If clients are resending connect requests to the server, the
5427 server refuses to allow a client to connect multiple times.
5428 Fixed a bug in the handling of this case.
5431 Frequency : very large clusters immediately after boot
5433 Description: LNET request buffers exhausted under heavy short-term load
5434 Details : If a large number of client requests are generated on a service
5435 that has previously never seen so many requests it is possible
5436 that the request buffer growth cannot keep up with the spike in
5437 demand. Instead of dropping incoming requests, they are held in
5438 the LND until the RPC service can accept more requests.
5441 Frequency : Sometimes during replay
5443 Description: Assertion failure in ll_local_open after replay.
5444 Details : If replay happened on an open request reply before we were able
5445 to set replay handler, reply will become not swabbed tripping the
5446 assertion in ll_local_open. Now we set the handler right after
5447 recognising of open request
5450 Frequency : very rare
5452 Description: kernel reports "badness in vsnprintf"
5453 Details : Reading from the "recovery_status" /proc file in small chunks
5454 may cause a negative length in lprocfs_obd_rd_recovery_status()
5455 call to vsnprintf() (which is otherwise harmless). Exit early
5456 if there is no more space in the output buffer.
5458 Severity : enhancement
5460 Description: clear OBD RPC statistics by writing to them
5461 Details : It is now possible to clear the OBD RPC statistics by writing
5462 to the "stats" file.
5467 Description: Client mtime is not the same on different clients after utimes
5468 Details : In some cases, the client was using the utimes() syscall on
5469 a file cached on another node. The clients now validate the
5470 ctime from the MDS + OSTs to determine which one is right.
5475 Description: Inability to activate failout mode
5476 Details : lconf script incorrectly assumed that in python string's numeric
5477 value is used in comparisons.
5480 Frequency : always with multiple stripes per file
5482 Description: Inefficient object allocation for mutli-stripe files
5483 Details : When selecting which OSTs to stripe files over, for files with
5484 a stripe count that divides evenly into the number of OSTs,
5485 the MDS is always picking the same starting OST for each file.
5486 Return the OST selection heuristic to the original design.
5491 Description: mount failures may take full timeout to return an error
5492 Details : Under some heavy load conditions it is possible that a
5493 failed mount can wait for the full obd_timeout interval,
5494 possibly several minutes, before reporting an error.
5495 Instead return an error as soon as the status is known.
5497 ------------------------------------------------------------------------------
5499 2006-02-14 Cluster File Systems, Inc. <info@clusterfs.com>
5501 * WIRE PROTOCOL CHANGE. This version of Lustre networking WILL NOT
5502 INTEROPERATE with older versions automatically. Please read the
5503 user documentation before upgrading any part of a live system.
5504 * WARNING: Lustre networking configuration changes are required with
5505 this release. See https://bugzilla.clusterfs.com/show_bug.cgi?id=10052
5508 * Support for kernels:
5509 2.6.9-22.0.2.EL (RHEL 4)
5510 2.6.5-7.244 (SLES 9)
5511 2.6.12.6 vanilla (kernel.org)
5514 Severity : enhancement
5515 Bugzilla : 7981/8208
5516 Description: Introduced Lustre Networking (LNET)
5517 Details : LNET is new networking infrastructure for Lustre, it includes
5518 a reorganized network configuration mode (see the user
5519 documentation for full details) as well as support for routing
5520 between different network fabrics. Lustre Networking Devices
5521 (LNDs) for the supported network fabrics have also been
5522 created for this new infrastructure.
5524 Severity : enhancement
5525 Description: Introduced Access control lists
5526 Details : clients can set ACLs on files and directories in order to have
5527 more fine-grained permissions than the standard Unix UGO+RWX.
5528 The MDS must be started with the "-o acl" mount option.
5530 Severity : enhancement
5531 Description: Introduced filesystem quotas
5532 Details : Administrators may now establish per-user quotas on the
5535 Severity : enhancement
5537 Description: Configuration change for the XT3
5538 The PTLLND is now used to run Lustre over Portals on the XT3
5539 The configure option(s) --with-cray-portals are no longer used.
5540 Rather --with-portals=<path-to-portals-includes> is used to
5541 enable building on the XT3. In addition to enable XT3 specific
5542 features the option --enable-cray-xt3 must be used.
5547 Description: Running on many-way SMP OSTs can trigger oops in llcd_send()
5548 Details : A race between allocating a new llcd and re-getting the llcd_lock
5549 allowed another thread to grab newly-allocated llcd.
5551 Severity : enhancement
5553 Description: 2.6 OST async journal commit and locking fix to improve performance
5554 Details : The filter_direct_io()+filter_commitrw_write() journal commits for
5555 2.6 kernels are now async as they already were in 2.4 kernels so
5556 that they can commit concurrently with the network bulk transfer.
5557 For block-allocated files the filter allocation semaphore is held
5558 to avoid filesystem fragmentation during allocation. BKL lock
5559 removed for 2.6 xattr operations where it is no longer needed.
5564 Description: lconf incorrectly determined whether two IP networks could talk
5565 Details : In some more complicated routing and multiple-network
5566 configurations, lconf will avoid trying to make a network
5567 connection to a disjoint part of the IP space. It was doing the
5568 math incorrectly for one set of cases.
5573 Description: Fix for potential infinite loop processing records in an llog.
5574 Details : If an llog record is corrupted/zeroed, it is possible to loop
5575 forever in llog_process(). Validate the llog record length
5576 and skip the remainder of the block on error.
5579 Frequency : occasional (liblustre only)
5581 Description: liblustre could not open files whose last component is a symlink
5582 Details : sysio_path_walk() would incorrectly pass the open intent to
5583 intermediate path components.
5586 Frequency : rare (liblustre only with non-standard tuning)
5587 Bugzilla : 7201 (7350)
5588 Description: Tuning the MDC DLM LRU size to zero triggers client LASSERT
5589 Details : llu_lookup_finish_locks() tries to set lock data on a lock
5590 after it has been released, only do this for referenced locks
5592 Severity : enhancement
5594 Description: specifying an (invalid) directory default stripe_size of -1
5595 would reset the directory default striping
5596 Details : stripe_size -1 was used internally to signal directory stripe
5597 removal, now use "all default" to signal dir stripe removal
5598 as a directory striping of "all default" is not useful
5601 Frequency : common for large clusters running liblustre clients
5603 Description: doing an ls when liblustre clients are running is slow
5604 Details : sending a glimpse AST to a liblustre client waits for every AST
5605 to time out, as liblustre clients will not respond. Since they
5606 cannot cache data we refresh the OST lock LVB from disk instead.
5608 Severity : enhancement
5610 Description: doing an ls at the same time as file IO can be slow
5611 Details : enqueue and other "small" requests can be blocked behind many
5612 large IO requests. Create a new OST IO portal for non-IO
5613 requests so they can be processed faster.
5616 Frequency : rare (only HPUX clients mounting unsupported re-exported NFS vol)
5618 Description: an HPUX NFS client would get -EACCESS when ftruncate()ing a newly
5619 created file with mode 000
5620 Details : the Linux NFS server relies on an MDS_OPEN_OWNEROVERRIDE hack to
5621 allow an ftruncate() as a non-root user to a file with mode 000.
5622 Lustre now respects this flag to disable mode checks when
5623 truncating a file owned by the user
5626 Frequency : liblustre-only, when liblustre client dies unexpectedly or becomes
5629 Description: Revoking locks from clients that went dead or catatonic might take
5631 Details : New lock flags FL_CANCEL_ON_BLOCK used by liblustre makes
5632 cancellation of such locks instant on servers without waiting for
5633 any reply from clients. Clients drops these locks when cancel
5634 notification from server is received without replying.
5637 Frequency : liblustre-only, when liblustre client dies or becomes busy
5639 Description: Doing ls on Linux clients can take a long time with active
5641 Details : Liblustre client cannot handle ASTs in timely manner, so avoid
5642 granting such locks to it in the first place if possible. Locks
5643 are taken by proxy on the OST during the read or write and
5644 dropped immediately afterward. Add connect flags handling, do
5645 not grant locks to liblustre clients for glimpse ASTs.
5647 Severity : enhancement
5649 Description: Improve read-ahead algorithm to avoid excessive IO for random reads
5650 Details : Existing read-ahead algorithm is tuned for the case of streamlined
5651 sequential reads and behaves badly with applications doing random
5652 reads. Improve it by reading ahead at least read region, and
5653 avoiding excessive large RPC for small reads.
5655 Severity : enhancement
5657 Description: Creating more than 1000 files for a single job may cause a load
5658 imbalance on the OSTs if there are also a large number of OSTs.
5659 Details : qos_prep_create() uses an OST index reseed value that is an
5660 even multiple of the number of available OSTs so that if the
5661 reseed happens in the middle of the object allocation it will
5662 still utilize the OSTs as uniformly as possible.
5667 Description: OST or MDS may oops in ping_evictor_main()
5668 Details : ping_evictor_main() drops obd_dev_lock if deleting a stale export
5669 but doesn't restart at beginning of obd_exports_timed list
5672 Severity : enhancement
5674 Description: improve by-nid export eviction on the MDS and OST
5675 Details : allow multiple exports with the same NID to be evicted at one
5676 time without re-searching the exports list.
5679 Frequency : rare, only with supplementary groups enabled on SMP 2.6 kernels
5681 Description: MDS may oops in groups_free()
5682 Details : in rare race conditions a newly allocated group_info struct is
5683 freed again, and this can be NULL. The 2.4 compatibility code
5684 for groups_free() checked for a NULL pointer, but 2.6 did not.
5687 Frequency : common for liblustre clients doing little filesystem IO
5688 Bugzilla : 9352, 7313
5689 Description: server may evict liblustre clients accessing contended locks
5690 Details : if a client is granted a lock or receives a completion AST
5691 with a blocking AST already set it would not reply to the AST
5692 for LDLM_FL_CANCEL_ON_BLOCK locks. It now replies to such ASTs.
5695 Frequency : lfs setstripe, only systems with more than 160 OSTs
5697 Description: unable to set striping with a starting offset beyond OST 160
5698 Details : llapi_create_file() incorrectly limited the starting stripe
5699 index to the maximum single-file stripe count.
5702 Frequency : LDAP users only
5704 Description: lconf did not handle in-kernel recovery with LDAP properly
5705 Details : lconf/LustreDB get_refs() is searching the wrong namespace
5707 Severity : enhancement
5709 Description: bind OST threads to NUMA nodes to improve performance
5710 Details : all OST threads are uniformly bound to CPUs on a single NUMA
5711 node and do their allocations there to localize memory access
5713 Severity : enhancement
5715 Description: llmount can determine client NID directly from Myrinet (GM)
5716 Details : the client NID code from gmnalnid was moved directly into
5717 llmount, removing the need to use this or specifying the
5718 client NID explicitly when mounting GM clients with zeroconf
5721 Frequency : if client is started with down MDS
5723 Description: if client is started with down MDS mount hangs in ptlrpc_queue_wait
5724 Details : Having an LWI_INTR() wait event (interruptible, but no timeout)
5725 will wait indefinitely in ptlrpc_queue_wait->l_wait_event() after
5726 ptlrpc_import_delayed_req() because we didn't check if the
5727 request was interrupted, and we also didn't break out of the
5728 event loop if there was no timeout
5733 Description: data loss during non-page-aligned writes to a single file from
5734 both multiple nodes and multiple threads on one node at same time
5735 Details : updates to KMS and lsm weren't protected by common lock. Resulting
5736 inconsistency led to false short-reads, that were cached and later
5737 used by ->prepare_write() to fill in partially written page,
5738 leading to data loss.
5741 Frequency : always, if lconf --abort_recovery used
5743 Description: lconf --abort_recovery fails with 'Operation not supported'
5744 Details : lconf was attempting to abort recovery on the MDT device and not
5747 Severity : enhancement
5749 Description: remove cleanup logs
5750 Details : replace lconf-generated cleanup logs with lustre internal
5751 cleanup routines. Eliminates the need for client-cleanup and
5754 Severity : enhancement
5756 Description: add support for EAs (user and system) on lustre filesystems
5757 Details : it is now possible to store extended attributes in the Lustre
5758 client filesystem, and with the user_xattr mount option it
5759 is possible to allow users to store EAs on their files also
5761 Severity : enhancement
5763 Description: Add possibility (config option) to show minimal available OST free
5765 Details : When compiled with --enable-mindf configure option, statfs(2)
5766 (and so, df) will return least minimal free space available from
5767 all OSTs as amount of free space on FS, instead of summary of
5768 free spaces of all OSTs.
5770 Severity : enhancement
5772 Description: do not expand extent locks acquired on OST-side
5773 Details : Modify ldlm_extent_policy() to not expand local locks, acquired
5774 by server: they are not cached anyway.
5777 Frequency : when mmap is used/binaries executed from Lustre
5779 Description: Unmmap pages before throwing them away from read cache.
5780 Details : llap_shrink cache now attempts to unmap pages before discarding
5781 them (if unmapping failed - do not discard). SLES9 kernel has
5782 extra checks that trigger if this unmapping is not done first.
5787 Description: lconf didn't resolve symlinks before checking to see whether a
5788 given mountpoint was already in use
5791 Frequency : when migrating failover services
5792 Bugzilla : 6395, 9514
5793 Description: When migrating a subset of services from a node (e.g. failback
5794 from a failover service node) the remaining services would
5795 time out and evict clients.
5796 Details : lconf --force (implied by --failover) sets the global obd_timeout
5797 to 5 seconds in order to quickly disconnect, but this caused
5798 other RPCs to time out too quickly. Do not change the global
5799 obd_timeout for force cleanup, only set it for DISCONNECT RPCs.
5801 Severity : enhancement
5802 Frequency : if MDS is started with down OST
5803 Bugzilla : 9439,5706
5804 Description: Allow startup/shutdown of an MDS without depending on the
5805 availability of the OSTs.
5806 Details : Asynchronously call mds_lov_synchronize during MDS startup.
5807 Add appropriate locking and lov-osc refcounts for safe
5808 cleaning. Add osc abort_inflight calls in case the
5809 synchronize never started.
5812 Frequency : occasional (Cray XT3 only)
5814 Description: root not authorized to access files in CRAY_PORTALS environment
5815 Details : The client process capabilities were not honoured on the MDS in
5816 a CRAY_PORTALS/CRAY_XT3 environment. If the file had previously
5817 been accessed by an authorized user then root was able to access
5818 the file on the local client also. The root user capabilities
5819 are now allowed on the MDS, as this environment has secure UID.
5822 Frequency : occasional
5824 Description: ldiskfs "too long searching" message happens too often
5825 Details : A debugging message (otherwise harmless) prints too often on
5826 the OST console. This has been reduced to only happen when
5827 there are fragmentation problems on the filesystem.
5832 Description: Division by zero in statfs when all OSCs are inactive
5833 Details : lov_get_stripecnt() returns zero due to incorrect order of checks,
5834 lov_statfs divides by value returned by lov_get_stripecnt().
5838 Bugzilla : 9489, 3273
5839 Description: First write from each client to each OST was only 4kB in size,
5840 to initialize client writeback cache, which caused sub-optimal
5841 RPCs and poor layout on disk for the first writen file.
5842 Details : Clients now request an initial cache grant at (re)connect time
5843 and so that they can start streaming writes to the cache right
5844 away and always do full-sized RPCs if there is enough data.
5845 If the OST is rebooted the client also re-establishes its grant
5846 so that client cached writes will be honoured under the grant.
5851 Description: Slow ls (and stat(2) syscall) on files residing on IO-loaded OSTs
5852 Details : Now I/O RPCs go to different portal number and (presumably) fast
5853 lock requests (and glimses) and other RPCs get their own service
5854 threads pool that should be able to service those RPCs
5857 Severity : enhancement
5859 Description: Ability to exchange lustre version between client and servers and
5860 issue warnings at client side if client is too old. Also for
5861 liblustre clients there is ability to refuse connection of too old
5863 Details : New 'version' field is added to connect data structure that is
5864 filled with version info. That info is later checked by server and
5868 Frequency : rare, liblustre only.
5869 Bugzilla : 9296, 9581
5870 Description: Two simultaneous writes from liblustre at offset within same page
5871 might proceed at the same time overwriting eachother with stale
5873 Details : I/O lock withing llu_file_prwv was released too early, before data
5874 actually was hitting the wire. Extended lock-holding time until
5875 server acknowledges receiving data.
5878 Frequency : extremely rare. Never observed in practice.
5880 Description: avoid generating lustre_handle cookie of 0.
5881 Details : class_handle_hash() generates handle cookies by incrementing
5882 global counter, and can hit 0 occasionaly (this is unlikely, but
5883 not impossible, because initial value of cookie counter is
5884 selected randonly). Value of 0 is used as a sentinel meaning
5885 "unassigned handle" --- avoid it. Also coalesce two critical
5886 sections in this function into one.
5888 Severity : enhancement
5890 Description: allow liblustre clients to delegate truncate locking to OST
5891 Details : To avoid overhead of locking, liblustre client instructs OST to
5892 take extent lock in ost_punch() on client's behalf. New connection
5893 flag is added to handle backward compatibility.
5895 Severity : enhancement
5896 Bugzilla : 4928, 7341, 9758
5897 Description: allow number of OST service threads to be specified
5898 Details : a module parameter allows the number of OST service threads
5899 to be specified via "options ost ost_num_threads={N}" in the
5900 OSS's /etc/modules.conf or /etc/modprobe.conf.
5904 Bugzilla : 6146, 9635, 9895
5905 Description: servers crash with bad pointer in target_handle_connect()
5906 Details : In rare cases when a client is reconnecting it was possible that
5907 the connection request was the last reference for that export.
5908 We would temporarily drop the export reference and get a new
5909 one, but this may have been the last reference and the export
5910 was just destroyed. Get new reference before dropping old one.
5912 Severity : enhancement
5913 Frequency : if client is started with failover MDS
5915 Description: Allow multiple MDS hostnames in the mount command
5916 Details : Try to read the configuration from all specified MDS
5917 hostnames during a client mount in case the "primary"
5920 Severity : enhancement
5922 Description: Stop sending data to evicted clients as soon as possible.
5923 Details : Check if the client we are about to send or are sending data to
5924 was evicted already. (Check is done every second of waiting,
5925 for which l_wait_event interface was extended to allow checking
5926 of exit condition at specified intervals).
5929 Frequency : rare, normally only when NFS exporting is done from client
5931 Description: 'bad disk LOV MAGIC: 0x00000000' error when chown'ing files
5933 Details : Make mds_get_md() recognise empty md case and set lmm size to 0.
5936 Frequency : always, if srand() is called before liblustre initialization
5938 Description: Liblustre uses system PRNG disturbing its usage by user application
5939 Details : Introduce internal to lustre fast and high-quality PRNG for
5940 lustre usage and make liblustre and some other places in generic
5941 lustre code to use it.
5943 Severity : enhancement
5944 Bugzilla : 9477, 9557, 9870
5945 Description: Verify that the MDS configuration logs are updated when xml is
5946 Details : Check if the .xml configuration logs are newer than the config
5947 logs stored on the MDS and report an error if this is the case.
5948 Request --write-conf, or allow starting with --old_conf.
5950 Severity : enhancement
5952 Description: Handle symlinks in the path when checking if Lustre is mounted.
5953 Details : Resolve intermediate symlinks when checking if a client has
5954 mounted a filesystem to avoid duplicate client mounts.
5959 Description: lconf can hit an error exception but still return success.
5960 Details : The lconf command catches the Command error exception at the top
5961 level script context and will exit with the associated exit
5962 status, but doesn't ensure that this exit status is non-zero.
5967 Description: failure of ptlrpc thread startup can cause oops
5968 Details : Starting a ptlrpc service thread can fail if there are a large
5969 number of threads or the server memory is very fragmented.
5970 Handle this without oopsing.
5973 Frequency : always, only if liblustre and non-default acceptor port was used
5975 Description: liblustre cannot connect to servers with non-default acceptor port
5976 Details : tcpnal_set_default_params() was not called and was therefore
5977 ignoring the environment varaible TCPNAL_PORT, as well as other
5978 TCPNAL_ environment variables
5983 Description: two objects could be created on the same OST for a single file
5984 Details : If an OST is down, in some cases it was possible to create two
5985 objects on a single OST for a single file. No problems other
5986 than potential performance impact and spurious error messages.
5990 Bugzilla : 5681, 9562
5991 Description: Client may oops in ll_unhash_aliases
5992 Details : Client dcache may become inconsistent in race condition.
5993 In some cases "getcwd" can fail if the current directory is
5999 Description: Inode refcounting problems in NFS export code
6000 Details : link_raw functions used to call d_instantiate without obtaining
6001 extra inode reference first.
6005 Bugzilla : 9942, 9903
6006 Description: Referencing freed requests leading to crash, memleaks with NFS.
6007 Details : We used to require that call to ll_revalidate_it was always
6008 followed by ll_lookup_it. Also with revalidate_special() it is
6009 possible to call ll_revalidate_it() twice for the same dentry
6010 even if first occurence returned success. This fix changes semantic
6011 between DISP_ENQ_COMPLETE disposition flag to mean there is extra
6012 reference on a request referred from the intent.
6013 ll_intent_release() then releases such a request.
6016 Frequency : rare, normally benchmark loads only
6018 Description: unlinked inodes were kept in memory on the client
6019 Details : If a client is repeatedly creating and unlinking files it
6020 can accumulate a lot of stale inodes in the inode slab cache.
6021 If there is no other client load running this can cause the
6022 client node to run out of memory. Instead flush old inodes
6023 from client cache that have the same inode number as a new inode.
6026 Frequency : SLES9 2.6.5 kernel and long filenames only
6027 Bugzilla : 9969, 10379
6028 Description: utime reports stale NFS file handle
6029 Details : SLES9 uses out-of-dentry names in some cases, which confused
6030 the lustre dentry revalidation. Change it to always use the
6034 Frequency : rare, unless heavy write-truncate concurrency is continuous
6035 Bugzilla : 4180, 6984, 7171, 9963, 9331
6036 Description: OST becomes very slow and/or deadlocked during object unlink
6037 Details : filter_destroy() was holding onto the parent directory lock
6038 while truncating+unlinking objects. For very large objects this
6039 may block other threads for a long time and slow overall OST
6040 responsiveness. It may also be possible to get a lock ordering
6041 deadlock in this case, or run out of journal credits because of
6042 the combined truncate+unlink. Solution is to do object truncate
6043 first in one transaction without parent lock, and then do the
6044 final unlink in a new transaction with the parent lock. This
6045 reduces the lock hold time dramatically.
6048 Frequency : rare, 2.4 kernels only
6050 Description: MDS or OST cleanup may trip kernel BUG when dropping kernel lock
6051 Details : mds_cleanup() and filter_cleanup() need to drop the kernel lock
6052 before unmounting their filesystem in order to avoid deadlock.
6053 The kernel_locked() function in 2.4 kernels only checks whether
6054 the kernel lock is held, not whether it is this process that is
6055 holding it as 2.6 kernels do.
6060 Description: MDS or OST may oops/LBUG if a client is connecting multiple times
6061 Details : The client ptlrpc code may be trying to reconnect to a down
6062 server before a previous connection attempt has timed out.
6063 Increase the reconnect interval to be longer than the connection
6064 timeout interval to avoid sending duplicate connections to
6068 Frequency : echo_client brw_test command
6070 Description: fix echo_client to work with OST preallocated code
6071 Details : OST preallocation code (5137) didn't take echo_client IO path
6072 into account: echo_client calls filter methods outside of any
6073 OST thread and, hence, there is no per-thread preallocated
6074 pages and buffers to use. Solution: hijack pga pages for IO. As
6075 a byproduct, this avoids unnecessary data copying.
6079 Bugzilla : 3555, 5962, 6025, 6155, 6296, 9574
6080 Description: Client can oops in mdc_commit_close() after open replay
6081 Details : It was possible for the MDS to return an open request with no
6082 transaction number in mds_finish_transno() if the client was
6083 evicted, but without actually returning an error. Clients
6084 would later try to replay that open and may trip an assertion
6085 Simplify the client close codepath, and always return an error
6086 from the MDS in case the open is not successful.
6089 Frequency : rare, 2.6 OSTs only
6091 Description: OST may deadlock under high load on fragmented files
6092 Details : If there was a heavy load and highly-fragmented OST filesystems
6093 it was possible to have all the OST threads deadlock waiting on
6094 allocation of biovecs, because the biovecs were not released
6095 until the entire RPC IO was completed. Instead, release biovecs
6096 as soon as they are complete to ensure forward IO progress.
6098 Severity : enhancement
6100 Description: Support for specifying external journal device at mount
6101 Details : If an OST or MDS device is formatted with an external journal
6102 device, this device major/minor is stored in the ext3 superblock
6103 and may not be valid for failover. Allow detecting and
6104 specifying the external journal at mount time.
6109 Description: Mounting an MDS with pending unlinked files may cause oops
6110 Details : target_finish_recovery() calls mds_postrecov() which returned
6111 the number of orphans unlinked. mds_lov_connect->mds_postsetup()
6112 considers this an error and immediately begins cleaning up the
6113 lov, just after starting the mds_lov process
6115 Severity : enhancement
6117 Description: Implement 'lfs df' to report actual free space on per-OST basis
6118 Details : Add sub-command 'df' on 'lfs' to report the disk space usage of
6119 MDS/OSDs. Usage: lfs df [-i][-h]. Command Options: '-i' to report
6120 usage of objects; '-h' to report in human readable format.
6122 ------------------------------------------------------------------------------
6124 2005-08-26 Cluster File Systems, Inc. <info@clusterfs.com>
6131 Description: Mounting an ldiskfs file system with mballoc may crash OST node.
6132 Details : ldiskfs mballoc code may reference an uninitialized buddy struct
6133 at startup during orphan unlinking. Instead, skip buddy update
6134 before setup, as it will be regenerated after recovery is complete.
6139 Description: If an OST is inactive, its locks might reference stale inodes.
6140 Details : lov_change_cbdata() must iterate over all namespaces, even if
6141 they are inactive to clear inode references from the lock.
6143 Severity : enhancement
6144 Frequency : occasional, if non-standard max_dirty_mb used
6146 Description: Client will block write RPCs if not enough grant
6147 Details : If a client has max_dirty_mb smaller than max_rpcs_in_flight,
6148 then the client will block writes while waiting for another RPC
6149 to complete instead of consuming its dirty limit. With change
6150 we get improved performance when max_dirty_mb is small.
6152 Severity : enhancement
6153 Bugzilla : 3389, 6253
6154 Description: Add support for supplementary groups on the MDS.
6155 Details : The MDS has an upcall /proc/fs/lustre/mds/{mds}/group_upcall
6156 (set to /usr/sbin/l_getgroups if enabled) which will do MDS-side
6157 lookups for user supplementary groups into a cache.
6161 Description: O_CREAT|O_EXCL open flags in liblustre always return -EEXIST
6162 Details : Make libsysio to not enforce O_EXCL by clearing the flag,
6163 for liblustre O_EXCL is enforced by MDS.
6167 Description: readdir never returns NULL in liblustre.
6168 Details : Corrected llu_iop_getdirentries logic, to return offset of next
6169 dentry in struct dirent.
6173 Frequency : liblustre only, depends on application IO pattern
6174 Description: liblustre clients evicted if not contacting servers
6175 Details : Don't put liblustre clients into the ping_evictor list, so
6176 they will not be evicted by the pinger ever.
6178 Severity : enhancement
6180 Description: Add ability to evict clients by NID from MDS.
6181 Details : By echoing "nid:$NID" string into
6182 /proc/fs/lustre/mds/.../evict_client client with nid that equals to
6183 $NID would be instantly evicted from this MDS and from all active
6184 OSTs connected to it.
6188 Description: Do not query file size twice, somewhat slowing stat(2) calls.
6189 Details : lookup_it_finish() used to query file size from OSTs that was not
6194 Description: service threads change working directory to that of init
6195 Details : Starting lustre service threads may pin the working directory
6196 of the parent thread, making that filesystem busy. Threads
6197 now change to the working directory of init to avoid this.
6201 Frequency : during shutdown only
6202 Description: shutdown with a failed MDS or OST can cause unmount to hang
6203 Details : Don't resend DISCONNECT messages in ptlrpc_disconnect_import()
6208 Frequency : 2.6 only
6209 Description: chmod/chown may include an extra supplementary group
6210 Details : ll{,u}_mdc_pack_op_data() does not properly initialize the
6211 supplementary group and if none is specified this is used.
6214 Bugzilla : 5479 (6816)
6216 Description: Racing open + rm can assert client in mdc_set_open_replay_data()
6217 Details : If lookup is in progress on a file that is unlinked we might try
6218 to revalidate the inode and fail in revalidate after lookup is
6219 complete and ll_file_open() enqueues the open again but
6220 it_open_error() was not checking DISP_OPEN_OPEN errors correctly.
6223 Frequency : always, if lconf --abort_recovery used
6225 Description: lconf --abort_recovery fails with 'Operation not supported'
6226 Details : lconf was attempting to abort recovery on the MDT device and not
6229 ------------------------------------------------------------------------------
6231 2005-08-08 Cluster File Systems, Inc. <info@clusterfs.com>
6236 Frequency : rare (only unsupported configurations with a node running as an
6238 Bugzilla : 6514, 5137
6239 Description: Mounting a Lustre file system on a node running as an OST could
6241 Details : OSTs now preallocates memory needed to write out data at
6242 startup, instead of when needed, to avoid having to
6243 allocate memory in possibly low memory situations.
6244 Specifically, if the file system is mounted on on OST,
6245 memory pressure could force it to try to write out data,
6246 which it needed to allocate memory to do. Due to the low
6247 memory, it would be unable to do so and the node would
6248 become unresponsive.
6250 Severity : enhancement
6252 Description: Addition of lconf --service command line option
6253 Details : lconf now accepts a '--service <arg>' option, which is
6254 shorthand for 'lconf --group <arg> --select <arg>=<hostname>'
6256 Severity : enhancement
6258 Description: Failover mode is now the default for OSTs.
6259 Details : By default, OSTs will now run in failover mode. To return to
6260 the old behaviour, add '--failout' to the lmc line for OSTs.
6262 Severity : enhancement
6264 Description: Health checks are now provided for MDS and OSTs
6265 Details : Additional detailed health check information on MSD and OSTs
6266 is now provided through the procfs health_check value.
6269 Frequency : occasional, depends on IO load
6271 Description: Disk fragmentation on the OSTs could eventually cause slowdowns
6272 after numerous create/delete cycles
6273 Details : The ext3 inode allocation policy would not allocate new inodes
6274 very well on the OSTs because there are no new directories
6275 being created. Instead we look for groups with free space if
6276 the parent directories are nearly full.
6281 Description: Network or server problems during mount may cause partially
6282 mounted clients instead of returning an error.
6283 Details : The config llog parsing code may overwrite the error return
6284 code during mount error handling, returning success instead
6290 Description: MDS can fail to allocate large reply buffers
6291 Details : After long uptimes the MDS can fail to allocate large reply
6292 buffers (e.g. zconf client mount config records) due to memory
6293 fragmentation or consumption by the buffer cache. Preallocate
6294 some large reply buffers so that these replies can be sent even
6295 under memory pressure.
6299 Frequency : rare (liblustre)
6300 Description: fsx running with liblustre complained that using truncate() to
6301 extend the file doesn't work. This patch corrects that issue.
6302 Details : This is the liblustre equivalent of the fix for bug 6196. Fixes
6303 ATTR_SIZE and lsm use in llu_setattr_raw.
6307 Frequency : rare, only 2.6 kernels
6308 Description: Unusual file access patterns on the MDS may result in inode
6309 data being lost in very rare circumstances.
6310 Details : Bad interaction between the ea-in-inode patch and the "no-read"
6311 code in the 2.6 kernel caused the inode and/or EA data not to
6312 be read from disk, causing single-file corruption.
6316 Frequency : rare, only 2.6 filesystems using extents
6317 Description: Heavy concurrent write and delete load may cause data corruption.
6318 Details : It was possible under high-load situations to have an extent
6319 metadata block in the block device cache from a just-unlinked
6320 file overwrite a newly-allocated data block. We now unmap any
6321 metadata buffers that alias just-allocated data blocks.
6325 Frequency : filesystems with default stripe_count larger than 77
6326 Description: lconf+mke2fs fail when formatting filesystem with > 77 stripes
6327 Details : lconf specifies an inode size of 4096 bytes when the default
6328 stripe_count is larger than 77. This conflicts with the default
6329 inode density of 1 per 4096 bytes. Allocate smaller inodes in
6330 this case to avoid pinning too much memory for large EAs.
6332 ------------------------------------------------------------------------------
6334 2005-07-07 Cluster File Systems, Inc. <info@clusterfs.com>
6339 Frequency : rare (extremely heavy IO load with hundreds of clients)
6341 Description: Client is evicted, gets IO error writing to file
6342 Details : lock ordering changes for bug 5492 reintroduced bug 3267 and
6343 caused clients to be evicted for AST timeouts. The fixes in
6344 bug 5192 mean we no longer need to have such short AST timeouts
6345 so ldlm_timeout has been increased.
6348 Frequency : occasional during --force or --failover shutdown under load
6349 Bugzilla : 5949, 4834
6350 Description: Server oops/LBUG if stopped with --force or --failover under load
6351 Details : a collection of import/export refcount and cleanup ordering
6352 issues fixed for safer force cleanup
6355 Frequency : only filesystems larger than 120 OSTs
6356 Bugzilla : 5990, 6223
6357 Description: lfs getstripe would oops on a very large filesystem
6358 Details : lov_getconfig used kfree on vmalloc'd memory
6361 Frequency : only filesystems exporting via NFS to Solaris 10 clients
6362 Bugzilla : 6242, 6243
6363 Description: reading from files that had been truncated to a non-zero size
6364 but never opened returned no data
6365 Details : ll_file_read() reads zeros from no-object files to EOF
6370 Description: A bug in MDS/OSS recovery could cause the OSS to fail an assertion
6371 Details : There's little harm in aborting MDS/OSS recovery and letting it
6372 try again, so I removed the LASSERT and return an error instead.
6374 Severity : enhancement
6376 Description: New debugging infrastructure for tracking down data corruption
6377 Details : The I/O checksum code was replaced to: (a) control it at runtime,
6378 (b) cover more of the client-side code path, and (c) try to narrow
6379 down where problems occurred
6383 Bugzilla : 3819, 4364, 4397, 6313
6384 Description: Racing close and eviction MDS could cause assertion in mds_close
6385 Details : It was possible to get multiple mfd references during close and
6386 client eviction, leading to one thread referencing a freed mfd.
6388 Severity: : enhancement
6389 Bugzilla : 3262, 6359
6390 Description: Attempts to reconnect to servers are now more aggressive.
6391 Details : This builds on the enhanced upcall-less recovery that was added
6392 in 1.4.2. When trying to reconnect to servers, clients will
6393 now try each server in the failover group every 10 seconds. By
6394 default, clients would previously try one server every 25 seconds.
6399 Description: After recovery, certain operations trigger a failed
6400 assertion on a client.
6401 Details : Failing over an mds, using lconf -d --failover, while a
6402 client was doing a readdir() call would cause the client to
6403 LBUG after recovery completed and the readdir() was resent.
6405 Severity : enhancement
6407 Description: Default groups are now added by lconf
6408 Details : You can now run lconf --group <servicename> without having to
6409 manually add groups with lmc.
6412 Frequency : occasional
6414 Description: Nodes with an elan id of 0 trigger a failed assertion
6417 Frequency : always when accessing e.g. tty/console device nodes
6419 Description: tty and some other devices nodes cannot be used on lustre
6420 Details : file's private_data field is used by device data and lustre
6421 values in there got lost. New field was added to struct file to
6422 store fs-specific private data.
6425 Frequency : when exporting Lustre via NFS
6427 Description: NFSD failed occasionally when looking up a path component
6428 Details : NFSD is looking up ".." which was broken in ext3 directories
6429 that had grown large enough to become hashed.
6432 Frequency : Clusters with multiple interfaces not on the same subnet
6434 Description: Nodes will repeatedly try to reconnect to an interface which it
6435 cannot reach and report an error to the log.
6436 Details : Extra peer list entries will be created by lconf with some peers
6437 unreachable. lconf now validates the peer before adding it.
6440 Frequency : Only if a default stripe is set on the filesystem root.
6442 Description: Setting a default stripe on the filesystem root prevented the
6443 filesystem from being remounted.
6444 Details : The client was sending extra request flags in the root getattr
6445 request and did not allocate a reply buffer for the dir EA.
6448 Frequency : occasional, higher if lots of files are accessed by one client
6449 Bugzilla : 6159, 6097
6450 Description: Client trips assertion regarding lsm mismatch/magic
6451 Details : While revalidating inodes the VFS looks up inodes with ifind()
6452 and in rare cases can find an inode that is being freed.
6453 The ll_test_inode() code will free the lsm during ifind()
6454 when it finds an existing inode and then the VFS later attaches
6455 this free lsm to a new inode.
6459 Bugzilla : 6422, 7030
6460 Description: MDS deadlock between mkdir and client eviction
6461 Details : Creating a new file via mkdir or mknod (starting a transaction
6462 and getting the ns lock) can deadlock with client eviction
6463 (gets ns lock and trying to finish a synchronous transaction).
6466 Frequency : occasional
6467 Description: While starting a server, the fsfilt_ext3 module could not be
6469 Details : CFS's improved ext3 filesystem is named ldiskfs for 2.6
6470 kernels. Previously, lconf would still use the ext3 name
6471 when trying to load modules. Now, it will correctly use
6472 ext3 on 2.4 and ldiskfs on 2.6.
6474 Severity : enhancement
6475 Description: The default stripe count has been changed to 1
6476 Details : The interpretation of the default stripe count (0, to lfs
6477 or lmc) has been changed to mean striping across a single
6478 OST, rather than all available. For general usage we have
6479 found a stripe count of 1 or 2 works best.
6481 Severity : enhancement
6482 Description: Add support for compiling against Cray portals.
6483 Details : Conditional compiling for some areas that are different
6487 Frequency : occasional
6488 Bugzilla : 6409, 6834
6489 Description: Creating files with an explicit stripe count may lead to
6490 a failed assertion on the MDS
6491 Details : If some OSTs are full or unavailable, creating files may
6492 trigger a failed assertion on the MDS. Now, Lustre will
6493 try to use other servers or return an error to the
6497 Frequency : occasional
6499 Description: Multiple concurrent overlapping read+write on multiple SMP nodes
6500 caused lock timeout during readahead (since 1.4.2).
6501 Details : Processes doing readahead might match a lock that hasn't been
6502 granted yet if there are overlapping and conflicting lock
6503 requests. The readahead process waits on ungranted lock
6504 (original lock is CBPENDING), while OST waits for that process
6505 to cancel CBPENDING read lock and eventually evicts client.
6507 Severity : enhancement
6509 Description: Initial enabling of flock support for clients
6510 Details : Implements fcntl advisory locking and file status functions.
6511 This feature is provided as an optional mount flag (default
6512 off), and is NOT CURRENTLY SUPPORTED. Not all types of record
6513 locking are implemented yet, and those that are are not guaranteed
6514 to be completely correct in production environments.
6515 mount -t lustre -o [flock|noflock] ...
6518 Frequency : occasional
6520 Description: OSTs running 2.4 kernels but with extents enabled might trip an
6521 assertion in the ext3 JBD (journaling) layer.
6522 Details : The b_committed_data struct is protected by the big kernel lock
6523 in 2.4 kernels, serializing journal_commit_transaction() and
6524 ext3_get_block_handle->ext3_new_block->find_next_usable_block()
6525 access to this struct. In 2.6 kernels there is finer grained
6526 locking to improve SMP performance of the JBD layer.
6530 Description: Changes the "SCSI I/O Stats" kernel patch to default to "enabled"
6532 -----------------------------------------------------------------------------
6534 2005-05-05 Cluster File Systems, Inc. <info@clusterfs.com>
6536 NOTE: Lustre 1.4.2 uses an incompatible network protocol than previous
6537 versions of Lustre. Please update all servers and clients to
6538 version 1.4.2 or later at the same time. You must also run
6539 "lconf --write-conf {config}.xml" on the MDS while it is stopped
6540 to update the configuration logs.
6542 - fix for HPUX NFS client breakage when NFS exporting Lustre (5781)
6543 - mdc_enqueue does not need max_mds_easize request buffer on send (5707)
6544 - swab llog records of type '0' so we get proper header size/idx (5861)
6545 - send llog cancel req to DLM cancel portal instead of cb portal (5515)
6546 - fix rename of one directory over another leaking an inode (5953)
6547 - avoid SetPageDirty on 2.6 (5981)
6548 - don't re-add just-being-destroyed locks to the waiting list (5653)
6549 - when creating new directories, inherit the parent's custom
6550 striping settings if present parent (3048)
6551 - flush buffers from cache before direct IO in 2.6 obdfilter (4982)
6552 - don't hold i_size_sem in ll_nopage() and ll_ap_refresh_count (6077)
6553 - don't hold client locks on temporary worklist from l_lru (5666)
6554 - handle IO errors in 2.6 obdfilter bio completion routine (6046)
6555 - automatically evict dead clients (5921)
6556 - Update file size properly in create+truncate+fstat case (6196)
6557 - Do not unhash mountpoint dentries, do not allow removal of
6559 - Avoid lock ordering deadlock issue with write/truncate (6203,5654)
6560 - reserve enough journal credits in fsfilt_start_log for setattr (4554)
6561 - ldlm_enqueue freed-export error path would always LBUG (6149,6184)
6562 - don't reference lr_lvb_data until after we hold lr_lvb_sem (6170)
6563 - don't overwrite last_rcvd if there is a *_client_add() error (6086)
6564 - Correctly handle reads of files with no objects (6243)
6565 - lctl recover will also mark a device active if deactivate used (5933)
6567 - by default create 1 inode per 4kB space on MDS, per 16kB on OSTs
6568 - allow --write-conf on an MDS with different nettype than client (5619)
6569 - don't write config llogs to MDS for mounts not from that MDS (5617)
6570 - lconf should create multiple TCP connections from a client (5201)
6571 - init scripts are now turned off by default; run chkconfig --on
6572 lustre and chkconfig --on lustrefs to use them
6573 - upcalls are no longer needed for clients to recover to failover
6575 - add --abort-recovery option to lconf to abort recovery on device
6577 - add support for an arbitrary number of OSTs (3026)
6578 - Quota support protocol changes.
6579 - forward compatibility changes to wire structs (6007)
6580 - rmmod NALs that might be loaded because of /etc/modules.conf (6133)
6581 - support for mountfsoptions and clientoptions to the Lustre LDAP (5873)
6582 - improved "lustre status" script
6583 - initialize blocksize for non-regular files (6062)
6584 - added --disable-server and --disable-client configure options (5782)
6585 - introduce a lookup cache for lconf to avoid repeated DB scans (6204)
6586 - Vanilla 2.4.29 support
6587 - increase maximum number of obd devices to 520 (6242)
6588 - remove the tcp-zero-copy patch from the suse-2.4 series (5902)
6589 - Quadrics Elan drivers are now included for the RHEL 3 2.4.21 and
6590 SLES 9 2.6.5 kernels
6591 - limit stripes per file to 160 (the maximum EA size) (6093)
6593 2005-03-22 Cluster File Systems, Inc. <info@clusterfs.com>
6596 - don't LASSERT in ll_release on NULL lld with NFS export (4655, 5760)
6597 - hold NS lock when calling handle_ast_error->del_waiting_lock (5746)
6598 - fix setattr mtime regression from lovcleanup merge (4829, 5669)
6599 - workaround for 2.6 crash in ll_unhash_aliases (5687, 5210)
6600 - small ext3 extents cleanups and fixes (5733)
6601 - improved mballoc code, several small races and bugs fixed (5733, 5638)
6602 - kernel version 43 - fix remove_suid bugs in both 2.4 and 2.6 (5695)
6603 - avoid needless client->OST connect, fix handle mismatch (5317)
6604 - fix DLM error path that led to out-of-sync client, long delays (5779)
6605 - support common vfs-enforced mount options (nodev,nosuid,noexec) (5637)
6606 - fix several locking issues related to i_size (5492,5624,5654,5672)
6607 - don't move pending lock onto export if it is already evicted (5683)
6608 - fix kernel oops when creating .foo in unlinked directory (5548)
6609 - fix deadlock in obdfilter statistics vs. object create (5811)
6610 - use time_{before,after} to avoid timer jiffies wrap (5882)
6611 - shutdown --force/--failover stability (3607,3651,4797,5203,4834)
6612 - Do not leak request if server was not able to process it (5154)
6613 - If mds_open unable to find parent dir, make that negative lookup(5154)
6614 - don't create new directories with extent-mapping (5909, 5936)
6616 - fix lustre/lustrefs init scripts for SuSE (patch from Scali, 5702)
6617 - don't hold the pinger_sem in ptlrpc_pinger_sending_on_import
6618 - change obd_increase_kms to obd_adjust_kms (up or down) (5654)
6619 - lconf, lmc search both /usr/lib and /usr/lib64 for Python libs (5800)
6620 - support for RHEL4 kernel on i686 (5773)
6621 - provide error messages when incompatible logs are encountered (5898)
6623 2005-02-18 Cluster File Systems, Inc. <info@clusterfs.com>
6624 * version 1.4.0.10 (1.4.1 release candidate 1)
6626 - don't keep a lock reference when lock is not granted (4238)
6627 - unsafe list practices (rarely) led to infinite eviction loop (4908)
6628 - add per-fs limit of Lustre pages in page cache, avoid OOM (4699)
6629 - drop import inflight refcount on signal_completed_replay error (5255)
6630 - unlock page after async write error during send (3677)
6631 - handle missing objects in filter_preprw_read properly (5265)
6632 - no transno return for symlink open, don't save no-trasno open (3440)
6633 - don't try to complete elan receive that already failed (4012)
6634 - free RPC server reply state on error (5406)
6635 - clean up thread from ptlrpc_start_thread() on error (5160)
6636 - readahead could read extra page into cache that wasn't ejected (5388)
6637 - prevent races in class_attach/setup/cleanup/detach (5260)
6638 - don't dereference de->d_inode after l_dput of de (5458)
6639 - use "int" for stripe value returned from lock_to_stripe (5544)
6640 - mballoc allocation and error-checking fixes in 2.6 (5504)
6641 - block device patches to fix I/O request sizes in 2.6 (5482)
6642 - look up hostnames for IB nals (5602)
6643 - 2.6 changed lock ordering of 2 semaphores, caused deadlock (5654)
6644 - don't start multiple acceptors for the same port (5277)
6645 - fix incorrect LASSERT in mds_getattr_name (5635)
6646 - export a proc file for general "ping" checking (5628)
6647 - fix "lfs check" to not block when the MDS is down (5628)
6649 - service request history (4965)
6650 - put {ll,lov,osc}_async_page structs in a single slab (4699)
6651 - create an "evict_client" /proc entry on OSTs, like the MDS has
6652 - fix mount usage message, return errors per mount(8) (5168)
6653 - change grep [] to grep "[]" in tests so they work in more UMLs
6654 - fix ppc64/x86_64 spec to use %{_libdir} instead of /usr/lib (5389)
6655 - remove ancient LOV_MAGIC_V0 EA support (5047)
6656 - add "disk I/Os in flight" and "I/O req time" stats in obdfilter
6657 - align r/w RPCs to PTLRPC_MAX_BRW_SIZE boundary for performance (3451)
6658 - allow readahead allocations to fail when low on memory (5383)
6659 - mmap locking landed again, after considerable improvement (2828)
6660 - add get_hostaddr() to lustreDB.py for LDAP support (5459)
6662 2004-11-23 Cluster File Systems, Inc. <info@clusterfs.com>
6665 - send OST transaction number in read/write reply to free req (4966)
6666 - don't ASSERT in ptl_send_rpc() if we run out of memory (5119)
6667 - lock /proc/sys/portals/routes internal state, avoiding oops (4827)
6668 - the watchdog thread now runs as interruptible (5246)
6669 - flock/lockf fixes (but it's still disabled, pending 5135)
6670 - don't use EXT3 constants in llite code (5094)
6671 - memory shortage at startup could cause assertion (5176)
6673 - reorganization of lov code
6674 - single portals codebase
6676 - add extents/mballoc support (5025)
6677 - direct I/O reads in the obdfilter (4048)
6678 - kernel patches from LNXI for 2.6 (bluesmoke, perfctr, mtd, kexec)
6680 tbd Cluster File Systems, Inc. <info@clusterfs.com>
6683 - send OST transaction number in read/write reply to free req (4966)
6684 - don't ASSERT in ptl_send_rpc() if we run out of memory (5119)
6685 - lock /proc/sys/portals/routes internal state, avoiding oops (4827)
6686 - the watchdog thread now runs as interruptible (5246)
6687 - handle missing objects in filter_preprw_read properly (5265)
6688 - unsafe list practices (rarely) led to infinite eviction loop (4908)
6689 - drop import inflight refcount on signal_completed_replay error (5255)
6690 - unlock page after async write error during send (3677)
6691 - return original error code on reconstructed replies (3761)
6692 - no transno return for symlink open, don't save no-trasno open (3440)
6694 - add pid to ldlm debugging output (4922)
6695 - bump the watchdog timeouts -- we can't handle 30sec yet
6696 - extra debugging for orphan dentry/inode bug (5259)
6698 2004-11-16 Cluster File Systems, Inc. <info@clusterfs.com>
6701 - fix TCP_NODELAY bug, which caused extreme perf regression (5134)
6702 - allocate qswnal tx descriptors singly to avoid fragmentation (4504)
6703 - don't LBUG on obdo_alloc() failure, use OBD_SLAB_ALLOC() (4800)
6704 - fix NULL dereference in /proc/sys/portals/routes (4827)
6705 - allow failed mdc_close() operations to be interrupted (4561)
6706 - stop precreate on OST before MDS would time out on it (4778)
6707 - don't send partial-page writes before EOF from client (4410)
6708 - discard client grant for sub-page writes on large-page clients (4520)
6709 - don't free dentries not owned by NFS code, check generation (4806)
6710 - fix lsm leak if mds_create_objects() fails (4801)
6711 - limit debug_daemon file size, always print CERROR messages (4789)
6712 - use transno after validating reply (3892)
6713 - process timed out requests if import state changes (3754)
6714 - update mtime on OST during writes, return in glimpse (4829)
6715 - add mkfsoptions to LDAP (4679)
6716 - use ->max_readahead method instead of zapping global ra (5039)
6717 - don't interrupt __l_wait_event() during strace
6719 - add software watchdogs to catch hung threads quickly (4941)
6720 - make lustrefs init script start after nfs is mounted
6721 - fix CWARN/ERROR duplication (4930)
6722 - return async write errors to application if possible (2248)
6723 - add /proc/sys/portal/memused (bytes allocated by PORTALS_ALLOC)
6724 - print NAL number in %x format (4645)
6725 - update barely-supported suse-2.4.21-171 series (4842)
6726 - support for sles 9 %post scripts
6727 - support for building 2.6 kernel-source packages
6728 - support for sles km_* packages
6730 2004-10-07 Cluster File Systems, Inc. <info@clusterfs.com>
6733 - ignore -ENOENT errors in osc_destroy (3639)
6734 - notify osc create thread that OSC is being cleaned up (4600)
6735 - add nettype argument for llmount in #5d in conf-sanity.sh (3936)
6736 - reconstruct ost_handle() like mds_handle() (4657)
6737 - create a new thread to do import eviction to avoid deadlock (3969)
6738 - let lconf resolve symlinked-to devices (4629)
6739 - don't unlink "objects" from directory with default EA (4554)
6740 - hold socknal file ref over connect in case target is down (4394)
6741 - allow more than 32000 subdirectories in a single directory (3244)
6742 - fix blocks count for O_DIRECT writes (3751)
6743 - OST returns ENOSPC from object create when no space left (4539)
6744 - don't send truncate RPC if file size isn't changing (4410)
6745 - limit OSC precreate to 1/2 of value OST considers bogus (4778)
6746 - bind to privileged port in socknal and tcpnal (3689)
6748 - rate limit CERROR/CWARN console message to avoid overload (4519)
6749 - GETFILEINFO dir ioctl returns LOV EA + MDS stat in 1 call (3327)
6750 - basic mmap support (3918)
6751 - kernel patch series update from b1_4 (4711)
6753 2004-09-16 Cluster File Systems, Inc. <info@clusterfs.com>
6756 - avoid crash during MDS cleanup with OST shut down (2775)
6757 - fix loi_list_lock/oig_lock inversion on interrupted IO (4136)
6758 - don't use bad inodes on the MDS (3744)
6759 - dynamic object preallocation to improve recovery speed (4236)
6760 - don't hold spinlock over lock dumping or change debug flags (4401)
6761 - don't zero obd_dev when it is force cleaned (3651)
6762 - print grants to console if they go negative (4431)
6763 - "lctl deactivate" will stop automatic recovery attempts (3406)
6764 - look for existing locks in ldlm_handle_enqueue() (3764)
6765 - don't resolve lock handle twice in recovery avoiding race (4401)
6766 - revalidate should check working dir is a directory (4134)
6768 - don't always mark "slow" obdfilter messages as errors (4418)
6770 2004-08-24 Cluster File Systems, Inc. <info@clusterfs.com>
6773 - don't close LustreDB during write_conf until it is done (3860)
6774 - fix typo in lconf for_each_profile (3821)
6775 - allow dumping logs from multiple threads at one time (3820)
6776 - don't allow multiple threads in OSC recovery (3812)
6777 - fix debug_size parameters (3864)
6778 - fix mds_postrecov to initialize import for llog ctxt (3121)
6779 - replace config semaphore with spinlock (3306)
6780 - be sure to send a reply for a CANCEL rpc with bad export (3863)
6781 - don't allow enqueue to complete on a destroyed export (3822)
6782 - down write_lock before checking llog header bitmap (3825)
6783 - recover from lock replay timeout (3764)
6784 - up llog sem before sending rpc (3652)
6785 - reduce ns lock hold times when setting kms (3267)
6786 - change a dlm LBUG to LASSERTF, to maybe learn something (4228)
6787 - fix NULL deref and obd_dev leak on setup error (3312)
6788 - replace some LBUG about llog ops with error handling (3841)
6789 - don't match INVALID dentries from d_lookup and spin (3784)
6790 - hold dcache_lock while marking dentries INVALID and hashing (4255)
6791 - fix invalid assertion in ptlrpc_set_wait (3880)
6793 - add libwrap support for the TCP acceptor (3996)
6794 - add /proc/sys/portals/routes for non-root route listing (3994)
6795 - allow setting MDS UUID in .xml (2580)
6796 - print the stack of a process that LBUGs (4228)
6798 2004-07-14 Cluster File Systems, Inc. <info@clusterfs.com>
6801 - don't cleanup request in ll_file_open() on failed MDS open (3430)
6802 - make sure to unset replay flag from failed open requests (3440)
6803 - if default stripe count is 0, use OST count for inode size (3636)
6804 - update parent mtime/ctime on client for create/unlink (2611)
6805 - drop dentry ref in ext3_add_link from open_connect_dentry (3266)
6806 - free recovery state on server during a forced cleanup (3571)
6807 - unregister_reply for resent reqs (3063)
6808 - loop back devices mounting and status check on 2.6 (3563)
6809 - fix resource-creation race that can provoke i_size == 0 (3513)
6810 - don't try to use bad inodes returned from MDS/OST fs lookup (3688)
6811 - more debugging for page-accounting assertion (3746)
6812 - return -ENOENT instead of asserting if ost getattr+unlink race (3558)
6813 - avoid deadlock after precreation failure (3758)
6814 - fix race and lock order deadlock in orphan handling (3450, 3750)
6815 - add validity checks when grabbing inodes from l_ast_data (3599)
6817 - add /proc/.../recovery_status to obdfilter (3428)
6818 - lightweight CDEBUG infrastructure, debug daemon (3668)
6819 - change default OSC RPC parameters to be better on small clusters
6820 - turn off OST read cache for files smaller than 32MB
6821 - install man pages and include them in rpms (3100)
6822 - add new init script for (un)mounting lustre filesystems (2593)
6823 - run chkconfig in %post for init scripts (3701)
6824 - drop scimac NAL (unmaintained)
6826 2004-06-17 Cluster File Systems, Inc. <info@clusterfs.com>
6829 - clean kiobufs before and after use (3485)
6830 - strip trailing '/'s before comparing paths with /proc/mounts (3486)
6831 - remove assertions to work around "in-flight rpcs" recovery bug (3063)
6832 - change init script to fail more clearly if not run as root (1528)
6833 - allow clients to reconnect during replay (1742)
6834 - fix ns_lock/i_sem lock ordering deadlock for kms update (3477)
6835 - don't do DNS lookups on NIDs too small for IP addresses (3442)
6836 - re-awaken ptlrpcd if new requests arrive during check_set (3554)
6837 - fix cond_resched (3554)
6838 - only evict unfinished clients after recovery (3515)
6839 - allow bulk resend, prevent data loss (3570)
6840 - dynamic ptlrpc request buffer allocation (2102)
6841 - don't allow unlinking open directory if it isn't empty (2904)
6842 - set MDS/OST threads to umask 0 to not clobber client modes (3359)
6843 - remove extraneous obd dereference causing LASSERT failure (3334)
6844 - don't use get_cycles() when creating temp. files on the mds (3156)
6845 - hold i_sem when setting i_size in ll_extent_lock() (3564)
6846 - handle EEXIST for set-stripe, set proper directory name (3336)
6848 - servers can dump a log evicting a client - lustre.dump_on_timeout=1
6849 - fix ksocknal_fmb_callback() error messages (2918)
6851 2004-05-27 Cluster File Systems, Inc. <info@clusterfs.com>
6854 - don't copy lvb into (possibly NULL) reply on error (2983)
6855 - don't deref dentry after dput, don't free lvb on error (2922)
6856 - use the kms to determine writeback rpc length (2947)
6857 - increment oti_logcookies when osc is inactive (2948)
6858 - update client's i_blocks count via lvb messages (2543)
6859 - handle intent open/close of special files properly (1557)
6860 - mount MDS with errors=remount-ro, like obdfilter (2009)
6861 - initialize lock handle to avoid ASSERT on error cleanup (3057)
6862 - don't use cancelling-locks' kms values (2947)
6863 - use highest lock extent for kms, not last one (2925)
6864 - don't dereference ERR_PTR() dentry in error handling path (3107)
6865 - fix thread race in portals_debug_dumplog() (3122)
6866 - create lprocfs device entries at setup instead of at attach (1519)
6867 - common AST error handler, don't evict client on completion race (3145)
6868 - zero nameidata in detach_mnt in 2.6 (3118)
6869 - verify d_inode after revalidate_special is valid in 2.6 (3116)
6870 - use lustre_put_super() to handle zconf unmounts in 2.6 (3064)
6871 - initialize RPC timeout timer earlier for 2.6 (3219)
6872 - don't dereference NULL reply buffer if mdc_close was never sent (2410)
6873 - print nal/nid for unknown nid (3258)
6874 - additional checks for oscc recovery before doing precreate (3284)
6875 - fix ll_extent_lock() error return code for 64-bit systems (3043)
6876 - don't crash in mdc_close for bad permissions on open (3285)
6877 - zero i_rdev for non-device files (3147)
6878 - clear page->private before handing to FS, better assertion (3119)
6879 - tune the read pipeline (3236)
6880 - fix incorrect decref of invalidated dentry (2350)
6881 - provide read-ahead stats and refine rpc in flight stats (3328)
6882 - don't hold journal transaction open across create RPC (3313)
6883 - update atime on MDS at close time (3265)
6884 - close LDAP connection when recovering to avoid server load (3315)
6885 - update iopen-2.6 patch with fixes from 2399,2517,2904 (3301)
6886 - don't leak open file on MDS after open resend (3325)
6887 - serialize filter_precreate and filter_destroy_precreated (3329)
6888 - loop device shouldn't call sync_dev() for nul device (3092)
6889 - clear page cache after eviction (2766)
6890 - resynchronize MDS->OST in background (2824)
6891 - refuse to mount the same filesystem twice on same mountpoint (3394)
6892 - allow llmount to create routes for mounting behind routers (3320)
6893 - push lock cancellation to blocking thread for glimpse ASTs (3409)
6894 - don't call osc_set_data_with_check() for TEST_LOCK matches (3159)
6895 - fix rare problem with rename on htree directories (3417)
6897 - allow default OST striping configuration per directory (1414)
6898 - fix compilation for qswnal for 2.6 kernels (3125)
6899 - increase maximum number of MDS request buffers for large systems
6900 - change liblustreapi to be useful for external progs like lfsck (3098)
6901 - increase local configuration timeout for slow disks (3353)
6902 - allow configuring ldlm AST timeout - lustre.ldlm_timeout=<seconds>
6904 2004-03-22 Cluster File Systems, Inc. <info@clusterfs.com>
6907 - fixes for glimpse AST timeouts / incorrectly 0-sized files (2818)
6908 - don't overwrite extent policy data in reply if lock was blocked (2901)
6909 - drop filter export grants atomically with removal from device (2663)
6910 - del obd_self_export from work_list in class_disconnect_exports (2908)
6911 - don't LBUG if MDS recovery times out during orphan cleanup (2530)
6912 - swab reply message in mdc_close, other PPC fixes (2464)
6913 - fix destroying of named logs (2325)
6914 - overwrite old logs when running lconf --write_conf (2264)
6915 - bump LLOG_CHUNKSIZE to 8k to allow for larger clusters (2306)
6916 - fix race in target_handle_connect (2898)
6917 - mds_reint_create() should take same inode create lock (2926)
6918 - correct journal credits calculated for CANCEL_UNLINK_LOG (2931)
6919 - don't close files for self_export to avoid uninitialized obd (2936)
6920 - allow MDS with the same name as client node (2939)
6921 - hold dentry reference for closed log files for unlink (2325)
6922 - reserve space for all logs during transactions (2059)
6923 - don't evict page beyond end of stripe extent (2925)
6924 - don't oops on a deleted current working directory (2399)
6925 - handle hard links to targets without a parent properly (2517)
6926 - don't dereference NULL lock when racing during eviction (2867)
6927 - don't grow lock extents when lots of conflicting locks (2919)
6929 2004-03-04 Cluster File Systems, Inc. <info@clusterfs.com>
6932 - account for cache space usage on clients to avoid data loss (974)
6933 - lfsck support in lustre kernel code (2349)
6934 - reduce journal credits needed for BRW writes (2370)
6935 - orphan handling to avoid losing space on client/server crashes
6936 - ptlrpcd can be blocked, stopping ALL progress (2477)
6937 - use lock value blocks to assist in proper KMS, faster stat (1021)
6938 - takes i_sem instead of DLM locks internally on obdfilter (2720)
6939 - recovery for initial connections (2355)
6940 - fixes for mds_cleanup_orphans (1934)
6941 - abort_recovery crashes MDS in b_eq (mds_unlink_orphan) (2584)
6942 - block all file creations until orphan recovery completes (1901)
6943 - client remove rq_connection from request struct (2423)
6944 - conf-sanity test_5, proper cleanup in umount log not availale (2640)
6945 - recovery timer race (2670)
6946 - mdc_close recovey bug (2532)
6947 - ptlrpc cleanup bug (2710)
6948 - mds timeout on local locks (2588)
6949 - namespace lock held during RPCs (2431)
6950 - handle interrupted sync write properly (2503)
6951 - don't try to handle a message that hasn't been replied to (2699)
6952 - client assert failure during cleanup after abort recovery (2701)
6953 - leak mdc device after failed mount (2712)
6954 - ptlrpc_check_set allows timedout requests to complete (2714)
6955 - wait for inflight reqs when ptlrpcd finishes (2710)
6956 - make sure unregistered services are removed from the srv_list
6957 - reset bulk XID's when resending them (caught by 1138 test)
6958 - unregister_bulk after timeout
6959 - fix lconf error (2694)
6960 - handle write after unfinished setstripe, stripe-only getstripe (2388)
6961 - readahead locks pages, leaves pending causing memory pressure (2673)
6962 - increase OST request buffers to 4096 on large machines (2729)
6963 - fix up permission of existing directories in simple_mkdir (2661)
6964 - init deleted item, add assertions ptlrpc_abort_inflight() (2725)
6965 - don't assign transno to errored transactions (2742)
6966 - don't delete objects on OST if given a bogus objid from MDS (2751)
6967 - handle large client PAGE_SIZE readdir on small PAGE_SIZE MDS (2777)
6968 - if rq_no_resend, then timeout request after recovery (2432)
6969 - fix MDS llog_logid record size, 64-bit array alignment (2733)
6970 - don't call usermode_helper from ptlrpcd, DEFAULT upcall (2773)
6971 - put magic in mount.lustre data, check for bad/NULL mount data (2529)
6972 - MDS recovery shouldn't delete objects that it has given out (2730)
6973 - if enqueue arrives after completion, don't clobber LVB (2819)
6974 - don't unlock pages twice when trigger_group_io returns error (2814)
6975 - don't deref NULL rq_repmsg if ldlm_handle_enqueue failed (2822)
6976 - don't write pages to disk if there was an error (1450)
6977 - don't ping imports that have recovery disabled (2676)
6978 - take buffered bytes into account when balancing socknal conn (2817)
6979 - hold a DLM lock over readdir always, use truncate_inode_pages (2706)
6980 - reconnect unlink llog connection after MDS reconnects to OST (2816)
6981 - remove little-endian swabbing of llog records (1987)
6982 - set/limit i_blksize to LL_MAX_BLKSIZE on client (2884)
6983 - retry reposting request buffers if they fail (1191)
6984 - grow extent at grant time to avoid granting a revoked lock (2809)
6985 - lock revoke doesn't evict page if covered by a second lock (2765)
6986 - disable VM readahead to avoid reading outside lock extents (2805)
6988 - return LL_SUPER_MAGIC from statfs for the filesystem type (1972)
6989 - updated kernel patches for hp-2.4.20 kernel (2681)
6991 2004-02-07 Cluster File Systems, Inc. <info@clusterfs.com>
6994 - fix truncated write corruption (2366)
6995 - fix for failed assertion in iopen_connect_dentry (1792,2517)
6997 - don't flag the ptlrpcd thread with PF_MEMALLOC (2636)
6998 - ensure len(uuid) < 37 in lmc (1171)
6999 - fix ia64 OOPS in llog_test (2255)
7000 - zero end of page at obdfilter for partial page writes (2648)
7001 - don't leave stale dentries around after renames (bug 2428)
7002 - fix timeouts when evicting a client with a single lock held (2642)
7003 - set deadline for the initial HELLO message to drain (2634)
7004 - print out dotted-quad IP addresses in the socknal (2302)
7006 - additional debugging for MDS client eviction problem (2443)
7007 - fix mkfsoptions support for osts (2603, 2604)
7009 2004-01-27 Cluster File Systems, Inc. <info@clusterfs.com>
7012 - add series for the vanilla 2.6.0 kernel
7013 - add series for the vanilla 2.4.24 kernel
7014 - add series for a cray x86/64 UL kernel drop
7015 - fix xattr patches for the vanilla 2.4.19 series
7017 - generate true UUIDs in lmc (1171)
7018 - have portals stack dumping break in UML (2466)
7019 - avoid bad dchild deref; avoid inum lock w/o creation (2362)
7020 - allocate with _NOFS in ldlm to avoid deadlock (1933)
7021 - wake callback waiting threads on client eviction (2460)
7022 - Add --ptldebug and --subsystem to lmc (1719)
7023 - update assertion to allow safe interrupt allocation
7024 - set rq_no_resend for cancel requests (2432)
7025 - recalculate ptlrpcd timeout after resend (2494)
7026 - call vfs_rmdir when removing pending directories (2368)
7027 - fix renaming a file to itself (2429)
7028 - lmc creates a default one-stripe lov (2454)
7029 - expand procfs space to handle large clusters (2326)
7030 - increase UML stack to avoid overflow
7031 - update lconf's list of debug and subsystem masks
7032 - fix lfs find --obd (2510)
7033 - /proc tunable for disabling filter read caching (2591)
7034 - stop rpm packages from altering slapd.conf (2301)
7035 - disable nagle in the socknal under 0conf (2578)
7036 - choose mds inode size based on stripe count (2572)
7037 - fix kernel-source rpm problems (2516)
7039 - add --disable-doc to avoid pdf generation (2421)
7040 - update documentation, tests, type-os, comments
7041 - avoid format warnings on ia64
7042 - remove the TOE NAL
7043 - tiny code cleanups by removing unused fields
7045 2004-01-07 Cluster File Systems, Inc. <info@clusterfs.com>
7048 - fix obvious semaphore misuse in as-yet-unused setattr path (2348)
7049 - remove the most blatant lies from BUILDING file (2371)
7050 - change default debug level to reasonable production setting
7051 - reduce client side cache size to reduce cache flush time
7052 - reduce max RPCs in flight to avoid unnecessary file fragmentation
7053 - make TCP zerocopy and pinger support enabled by default (2476)
7054 - sync writes completed after process exits caused crashes (2319)
7055 - maintain correct mount count on the MDS (2356)
7056 - backout 1557, because 2316 wasn't really fixed
7057 - better file I/O statistics gathering in /proc
7058 - don't take unnecessary, deadlock-inducing bug in readpage (2383)
7059 - another kernel patch to fix zero-copy TCP function export
7060 - don't take duplicate lock when processing re-sent getattr (2420)
7061 - lctl uses obd_self_export instead of creating new conn (2353)
7062 - MDS/OST recovery case which requires object creation asserted (2425)
7063 - move lfs from /usr/sbin to /usr/bin in packages
7064 - fix race between mds_client_add and mds_client_free (2417)
7065 - use kmalloc instead of slabs in portals (2430)
7066 - don't create duplicate records when a failover MDS is present (2442)
7067 - remove unnecessary mount age check (2332)
7068 - don't remove directory inodes from locks prematurely (2451)
7069 - don't break if MDS service name is the same as hostname (2103)
7070 - fix races in client write RPC generation when cache full (2482)
7072 2003-12-13 Cluster File Systems, Inc. <info@clusterfs.com>
7075 - remove now-unused request->rq_obd (278)
7076 - if an allocation fails, print out how much memory we've used (1933)
7077 - use PORTAL_SLAB_ALLOC for structures, to get GFP_MEMALLOC (1933)
7078 - add the "configurable stack size" patch to most series files (1256)
7079 - ability to write large log records, for 100+ OST configs (2306)
7080 - fix NULL deref when filter_prep fails (2314)
7081 - fix operator precedence error in filter_sync
7082 - dynamic allocation of socknal TX descriptors (2315)
7083 - fix a missed case in the GFP_MEMALLOC patch, can cause deadlock (2310)
7084 - fix gcc 2.96 compilation problem in xattr kernel patch (2294)
7085 - ensure that CWARN messages in Portals always get to the syslog
7086 - __init/__exit are not for prototype decls (ldlm_init/exit)
7087 - x86-64 compile warning fixes
7088 - fix gateway LMC keyword conflict (2318)
7089 - fix MDS lock inversions in getattr/reint paths (1844)
7090 - fix a rare lock re-ordering bug, which caused deadlock (2322)
7091 - fix i_sem/journal inversion in fsfilt_ext3_write_record (2306)
7092 - DLM race condition prevented some lock evictions (2328)
7093 - ENOMEM detection and retry on socknal sends (2230)
7094 - use GFP_NOFS throughout Lustre, to combat ENOMEM (2230)
7095 - move osc_rpcd into ptlrpc, for use in MDC and others (2329)
7096 - protect MDS inode fsdata with stronger locking; fixes assertion (2313)
7097 - better error messages when a client is rejected during recovery (1505)
7098 - avoid cancelling locks which were never granted, after failure (2330)
7099 - fix i_sem/journal inversion in mds_client_add (2333)
7100 - fix truncate/getattr lock cycle deadlock (2334)
7101 - use rpcd to send close; allows resend after timeout, avoid leak (1897)
7102 - fix two rare exit paths which could leak an l_lock() ref (2321)
7103 - fencepost error in MDS/OST orphan recovery (2226)
7104 - make log record alignment 8 bytes (1988)
7105 - lstripe now fails when requested offset > ost_count (2237)
7106 - ensure that all kernel series have a complete list.h (1607)
7107 - fix crashes in special-file operations (2316)
7108 - lctl create/brw OID mismatch, caused by obsolete filter loop (2339)
7110 - allow configurable automake binary, for testing new versions
7111 - small update to the lfs documentation
7113 2003-12-03 Cluster File Systems, Inc. <info@clusterfs.com>
7115 * fix negative export reference count in fsfilt_sync (2312)
7117 2003-12-01 Cluster File Systems, Inc. <info@clusterfs.com>
7118 * release candidate 0.9.1
7120 - orphans are moved into the PENDING directory for possible recovery
7121 - replayed opens now open by fid for orphan/rename safety (1042)
7122 - last close of an orphan inode generates a transno (683)
7123 - chdir() and mount() now pin the directory entry (1020)
7124 - avoid CERROR in normal ll_setattr_raw() error case (1500)
7125 - discard very old requests without processing them (1502)
7126 - remove some common, well-understood CERRORs (1505)
7127 - require O_DIRECT I/O to be page-sized to workaround IA64 crash (1609)
7128 - clear "grant" flags in OST replies until OST grant code lands (1644)
7129 - fix read performance by not clobbering i_blksize on client (1598)
7130 - fix __ldlm_handle2lock oops by not dereferencing lock after PUT (1625)
7131 - make LRU size a /proc tunable, clears locks when reduced (707)
7132 - fix some lprocfs rot that prevented ptlbd from loading (1732)
7133 - server locks take references on exports now (1558)
7134 - build fixes for 2.4.20-rh trees (1663)
7135 - return an error from lov_create if all OSCs are inactive (1751)
7136 - fix import levels when a reconnect happens without a timeout (1597)
7137 - exit early from mds_open if we get a lookup error (1749)
7138 - partial page read at EOF wouldn't wait for disk before sending (1642)
7139 - avoid NULL deref in obdfilter when reading page past EOF (1592)
7140 - avoid LASSERT in ll_intent_lock if server failed very early (1090)
7141 - fix LBUG in ll_it_open_error with rc = -2 (1861)
7142 - write/truncate lock inversion (1639)
7143 - Don't auto-load obdclass, portals modules during cleanup (1495)
7144 - fix timestamps from jumping to "now" (1763)
7145 - extra journal assertions (1648)
7146 - add an extra multiunlink test (1771)
7147 - fix read_record/write_record API (1776)
7148 - fix leak of offset_extent, possible incorrect i_size later (1772)
7149 - fix lasserts in mis-matched transnos during open-unlink testing (1541)
7150 - Debugging for the kqswnal_get_idle_tx problems (1820)
7151 - Allow recovery to be attempted multiple times (1536)
7152 - Write out MDS last_rcvd file after it is first created (1600)
7153 - Fix tx_descriptor leak in failed transmit situations (1827)
7154 - ext3 journaling fixes for assertion failure after IO error (1871)
7155 - class_export_put() on freed export after completion AST error (1896)
7156 - Fix revalidate looping in VFS (1322)
7157 - Don't access a freed export during MDS_REINT timeout (1521)
7158 - Add open-unlink recovery support on the MDS (1673,1764)
7159 - Return an error if no MDS data was read from last_rcvd (1946)
7160 - Fix for lookup "." or ".." crash on error (1932,1931,1935)
7161 - Don't setup a disk device that doesn't match exported UUID (317)
7162 - Reduce bulk RPC timeout to avoid cascading client/OST failures (1845)
7163 - avoid committing NULL handle in force close
7164 - local.sh is now a one-stripe LOV configuration
7165 - POSIX utime.4 -EPERM on FIFO not owned by user (56)
7166 - fix ext3 htree duplicate directory entry corruption (1516)
7167 - POSIX creat.13, fstat.1, open.18, stat.3 new file atime/mtime (2020)
7168 - update to new LOV EA format (2097)
7169 - interoperability for different PAGE_SIZE/wordsize (686,1821,1343,2042)
7171 2003-06-15 Phil Schwan <phil@clusterfs.com>
7174 - imports and exports cleanup too early, need refcounts (349, 879, 1045)
7175 - per-import/export recovery handling (958, 931, 959)
7176 - multiple last-rcvd slots, for serving multiple FSes (949)
7177 - connections are again shared between multiple imp/exports (963, 964)
7178 - "umount -f" would hang if any requests needed to be sent (393, 978)
7179 - avoid pinning large req buffer by copying for queued messages (989)
7180 - add "uuid" to "lctl device" command to help upcalls (991)
7181 - "open" RPCs with transnos would confuse recovery counters (1037)
7182 - do proper endian conversion of all wire messages (288, 340, 891)
7183 - remove OST bulk get LBUGs, fix ost_brw_write cleanup (1126)
7184 - call waiting locks callback from LDLM recovery thread (1127, 1151)
7185 - fix ptlrpc_connection leak in target_handle_connect (1174)
7186 - fix import refcounting bug in OST and MDS cleanup (1134)
7187 - if an invalid-at-open-time OSC returned before close(), LBUG (1150)
7188 - fix very unlikely obd_types race condition (501)
7189 - remove osc_open hack for echo_client (1187)
7190 - we leaked exports/dlmimps for forcibly disconnected clients (1143)
7191 - a failure in read_inode2 leads to deadlock (1139)
7192 - cancel ack-locks as soon as transaction is committed (1072)
7193 - fix major leaks and crashes in the bulk I/O path (937, 1057)
7194 - make sure to commitrw after any preprw to avoid deadlock (1162)
7195 - failing to execute a file in a lustre FS would lock inode (1203)
7196 - small DEBUG_REQ fix to avoid dereferencing a NULL (1227)
7197 - don't ASSERT while cleaning up an incompletely-setup obd (1248)
7198 - obd_uuid2tgt would walk off the end of the list (1255)
7199 - on IA64 the osc would give portals incorrect bulk size (1258)
7200 - fix debug daemon ioctl interface; allows daemon on ia64 (1274)
7201 - fix lock inversion caused by new llite matching code (1282)
7202 - limit the number of dirty pages on a client to 10MB (1286)
7203 - timed out locks were not being corrected cancelled (1289)
7204 - fix O_DIRECT above 4GB on IA-32 (1292)
7205 * major user-visible changes
7206 - fail out/fail over policy now controlled by the upcall (993)
7208 - add OBD_PING to check server availability and failure (954)
7209 - lustre messages are now sent in sending host order (288, 340, 891)
7210 - add eadatalen to MDS getattr reply (340)
7211 - OST read replies may contain second buffer, with per-page status (593)
7213 2003-03-11 Phil Schwan <phil@clusterfs.com>
7216 - LDLM_DEBUG macro fix, for gcc 3.2 (850)
7217 - failed open()s could cause deadlock; fixed (867, 869)
7218 - stop cancelling OST locks when files are closed (481)
7219 - overlapping XID spaces caused network corruption (851, 853)
7220 - fix unsafe fsfilt counter arithmetic; change to atomic_t
7221 - setattr_raw added, to do single-RPC, server-side setattrs
7222 - lmc/lconf syntax change for OST UUIDs
7223 - fix crashy race condition between ptlrpc_free_req and osc_close
7224 - don't use request in mdc_enqueue if we hit a timeout (889)
7225 - don't set the inode i_size for regular files from the MDS (896)
7226 - handle out of order completion AST (842)
7227 - don't LBUG if a lock request times out after receiving AST (913)
7228 - avoid d_rehash race in ll_find_alias by rehashing inside dcache_lock
7229 - if a bad lock AST arrives, send an error instead of dropping entirely
7230 - return 0 from revalidate2 if ll_intent_lock returns -EINTR (912)
7231 - fix leak in bulk IO when only partially completed (899, 900, 926)
7232 - fix O_DIRECT for ia64 (55)
7233 - (almost) eliminate Lustre-kernel-thread effects on load average (722)
7234 - C-z after timeout could hang a process forever; fixed (977)
7236 - client-side I/O cache (678, 924, 929, 941, 970)
7238 - READPAGE and SETATTRs which don't take server-side locks get
7241 2003-02-11 Phil Schwan <phil@clusterfs.com>
7244 - Fix ldlm_lock_match on the MDS to avoid matching remote locks (592)
7245 - Fix fsfilt_extN_readpage() to read a full page of directory
7246 entries, or fake the remainder if PAGE_SIZE != blocksize (500)
7247 - Avoid extra mdc_getattr() in ll_intent_lock when possible (534, 604)
7248 - Fix imbalanced LOV object allocation and out-of-bound access (469)
7249 - Most intent operations were removed, in favour of a new RPC mode
7250 that does a single RPC to the server and bypasses most of the VFS
7251 - All LDLM resource ID arrays were removed in favour of ldlm_res_id
7252 - Aggressively cancel local locks on DLM servers
7253 - mds_reint_unlink sends EA to the client if it's the last nlink.
7254 client uses that EA to unlink OST objects.
7255 - mds_reint_{rename,unlink,link} were rewritten to take ordered locks
7256 - recursive symlinks were fixed (439)
7257 - fixed NULL deref in DEBUG_REQ
7258 - filter_update_lastobjid no longer calls sync, which annoyed extN
7259 - fixed multi-client small-writes to a single file problem (445)
7260 - fixed mtime updates during file writes (607)
7261 - fixed vector writes on obdfilter causing problems when ENOSPC (670)
7262 - fixed bug in obd_brw_read/write() (under guise of testing 367)
7263 - fixed Linux OST size reporting problem (444, 656)
7264 - OST now updates object mtime with writes or setattr (607, 619)
7265 - client verifies file size before zeroing page past EOF (445)
7266 - OST now writes last allocated objid to disk with allocation (108)
7267 - LOV on echo now works (409)
7269 - mds_reint_unlink sends a new buffer, with the EA included. this
7270 buffer is only valid if body->valid & OBD_MD_FLEASIZE, which is only
7271 set if a regular file was being unlinked, and it was the last link
7272 - use PtlGet from the target for bulk writes (315)
7273 - OST now updates object mtime with writes or setattr (607, 619)
7274 - LDLM now has a grant-time callback to revalidate locked items, if
7276 - Many MDS operations were reorganized to combat race conditions
7278 - Merge b_intel branch (updated lprocfs code) - now at /proc/fs/lustre
7279 - configure check to avoid gcc version 2.96 20000731-2.96-98) (606)
7281 2003-01-06 Andreas Dilger <adilger@clusterfs.com>
7284 - Fully reactivate OST imports after reconnection (512, others)
7285 - Make sure client sees our -ENOTCONN from mds_handle (513 - partial)
7286 - More graceful error handling for truncating on dead OST (515)
7287 - Don't error out unless we're actually accessing dead stripes (474)
7288 - Fix garbage sizes when stripes are missing (410)
7289 - LRU counters were broken, causing constant lock purge (433, 432)
7290 - garbage on read from stripes with failed OSTs (441)
7291 - mark OSCs as active before reconnecting during recovery (438)
7292 - lov_enqueue and lov_cancel need to handle inactive OSTs (403)
7293 - lfind did not preserve OST order in output (443)
7294 - symlinks cause hung clients, incorrect data (439)
7295 - stop dereferencing request after dropping refcount (457)
7296 - don't LASSERT(spin_is_locked) on non-SMP (455)
7297 - fixes for many rename() bugs
7298 - fstat didn't correctly synchronize attributes (399)
7299 - server must handle lock cancellation during blocking AST prep (487)
7300 - bulk descriptors were free()d too soon (511)
7301 - fix paths in lconf, which would load incorrect modules (451, 507)
7302 - fix confusing lconf 'host not found' error message (386)
7303 - fix lock order deadlock on OST (O/R i_sem before journal ops, 478)
7304 - fix race condition in mdc_blocking_ast() for inode access (526)
7305 - fix lov_unpackmd() unpacking wrong number of stripes (537)
7306 - fix lov_set_osc_active() marking wrong OSC inactive (440)
7307 - fix bad lstripe lov_unpackmd() assertion (fix layering too) (527)
7308 - fix multiple writes of stripe MD to MDS (358, maybe 519)
7309 - fix lstripe in several ways (kernel side) (527)
7310 - fix request leak in ldlm_cli_enqueue (262)
7311 - incorrect OSC was marked inactive after OST failure
7312 - call mds_fs_cleanup before unmounting filesystem (524)
7313 - fix races between taking ns_lock and ldlm_lock_change_resource
7314 - fix races updating LOV export open file list
7315 - fix lov_enqueue error path, avoid decref-ing bad lock handle
7316 - fix recovery NULL deref in ldlm_cli_cancel_unused
7317 - fix some DLM races by using new hash table for lock handles (419)
7318 - permit the client to specify desired inodes, at replay
7319 - duplicate requests when we queue them for replay reintegration
7320 - fix last_rcvd offset calculation
7321 - sync after each recovered transaction, so we always make progress
7322 - never, not always, ERESTART requests without transnos
7323 - store the lov_desc in the MDS, so we don't depend on getlovinfo to
7325 - skip replay if the MDS says that the client is already connected
7326 - don't check for a recovery-enabled export to match lctl's UUID
7327 - don't INC_USE_COUNT for phantom exports
7328 - don't crash when cleaning up phantom exports (567)
7329 - don't double-finish or set replay data for errored mdc_open requests
7330 - abort requests when they time out, so we don't get old replies
7331 - send/receive replies for AST messages again
7332 - if the client says that it doesn't have the lock, cancel it on the
7334 - if we timeout during I/O, don't try to cancel an in-use lock; instead
7335 mark it as destroyed, it will all work out when decref is called
7336 - fix module use counts (22, 581)
7338 - ASTs now expect a reply (server cancels lock on error reply)
7340 2002-12-02 Andreas Dilger <adilger@clusterfs.com>
7343 - fix many simultaneous client startup (392)
7344 - fix dentry->d_it clobbering
7345 - credentials weren't being shipped for readdir/getattr operations
7346 - remove invalid assertions triggered during some concurrent MD
7348 - proper Lustre versions added (336, 389)
7349 - fix memory leak for create error case (398)
7350 - fix LOV locking bug that would get cli/srv out of sync
7351 - fix echo client over LOV (409)
7352 - fix dbench 2, extN refcount problem (170, 258, 356, 418)
7353 - fix double-O_EXCL intent crash (424)
7354 - avoid sending multiple lock CANCELs (352)
7356 - MDS can do multi-client recovery (modulo bugs in new code)
7358 - many updates, edits, cleanups
7360 2002-11-18 Phil Schwan <phil@clusterfs.com>
7363 - fix null d_it dereference (346)
7364 - fix full OST/dbench hang (333)
7365 - fix permission problem with file removal (286)
7366 - fix removal of OSCs from LOV when they fail
7367 - fix NULL deref during bulk timeout (214)
7368 - fix problems related to multiple filesystems on one MDS (241)
7369 - fixed serious subtle metadata locking bugs
7370 - free locks on clients when inodes are removed due to memory
7372 - fix inode pointer in lock data (285)
7373 - partial support for multiple MDS on a single host (241)
7374 - data locks weren't cancelled at clear_inode time (290, 311)
7375 - intent locks could lead to unbounded lock growth (205)
7376 - added a maximum lock count, an LRU list, and a flusher
7377 - fix multiple rename (365)
7378 - properly abstracted the echo client
7379 - OSC locked 1 byte too many; fixed
7380 - rewrote brw callback code:
7381 - fixed recovery bugs related to LOVs (306)
7382 - fixed too-many-pages-in-one-write crash (191)
7383 - fixed (again) crash in sync_io_timeout (214)
7384 - probably fixed callback-related race (385)
7386 - Add capability to MDS protocol
7387 - LDLM cancellations and callbacks on different portals
7389 2002-10-28 Andreas Dilger <adilger@clusterfs.com>
7392 - limit client IOV size to PTL_MD_MAX_IOV (611336, 191)
7393 - defer open object destruction to close time (601981, 138)
7394 - open/close OST file handle in obdo (OBD_MD_FLHANDLE) (601981, 138)
7395 - move LDLM_ENQUEUE/CONVERT back to MDS portal (625069)
7396 - abstract ll_lookup2, fix ll_revalidate2 to use abstraction (256)
7397 - don't call obd_setattr in ll_file_release for destroyed objects
7398 * protocol change to lustre_msg: move |version| and add |flags|
7399 * protocol change to osc_punch: "start" in "o_size", "end" in "o_blocks"
7400 * lock replay: for LDLM_FL_REPLAY trust client to do right thing
7401 * added replay of create, unlink, link and rename operations during
7402 MDS failover; recovery should be much more robust now
7403 * remove failed OSCs from LOVs (only lov_create uses this so far)
7404 * the lustre-HOWTO was brought (more) up to date (582544)
7406 2002-10-23 Phil Schwan <phil@clusterfs.com>
7409 - in-use dentries weren't being reused properly (617851)
7410 - prevent multiple LDLM setup (599178)
7411 - fix LOV size calculations for truncate (617853)
7412 - fix client handling of MDS intent errors (POSIX)
7413 - fix permission bug in lovstripe.c test (624321)
7414 - fix MDS thread deadlock - move LDLM handler to DLM portal (625069)
7415 - truncate past end of file could corrupt data
7416 - proper cleanup after timeouts, crashes, etc (592524, 550815)
7417 - a race in recovery could return ETIMEDOUT to apps (623947)
7418 - building outside the source directory was fixed
7419 * the lustre-HOWTO was brought (more) up to date (582544)
7420 * major progress was made on recovery functionality
7422 2002-10-10 Phil Schwan <phil@clusterfs.com>
7425 - recovery deadlock fix
7426 - rm -rf causes LBUG fix (617817)
7427 - file open by multiple tasks fix (618962)
7428 - directory permissions bugs (602707 and 620007)
7429 - journal_stop fixed with locking (611313)
7430 - O_APPEND failures resolved (618273, perhaps 614459)
7431 - lconf PATH fix (619770)
7432 - IA64 build fix (621450)
7433 - RPC buffer sizes scale with amount of memory
7435 2002-10-01 Phil Schwan <phil@clusterfs.com>
7438 - locks would be cancelled without throwing away data pages,
7439 resulting in inconsistent data (605627)
7440 - inode attributes were not always being refreshed (605627, 612449)
7441 - lconf now continues to cleanup after lctl reports an error
7442 - MDS now enforces user permissions (602707)
7443 - lprocfs cleanup fixed, but not yet enabled (614157)
7444 - fixed infinite server hang, should a client not respond to an AST
7445 - avoid going into recovery if user calls readlink() with a buffer
7446 that's too small (613941)
7447 - AST RPCs no longer require replies (614867) -- this may be changed
7448 - don't crash server if client sends an IOV that's too big (611336)
7449 - fixed lock conversion deadlock (611892)
7450 - fixed the following of symlinks (614622)
7451 * recovery: the server can remove locks from a client that dies, other
7452 clients can make progress
7453 * more extN patch fixes
7454 * compile-time configurable ptlrpc buffer allocations
7456 - collaborative read cache document
7457 - Lustre Lite Performance CDR document-in-progress
7459 2002-09-20 Andreas Dilger <adilger@clusterfs.com>
7462 - fix typo in patch-2.4.18
7464 2002-09-20 Andreas Dilger <adilger@clusterfs.com>
7467 - clear ptlrpc request each time in handle_incoming_request()
7468 - unlink of files now destroys the object on the OST
7470 2002-09-19 Peter Braam <braam@clusterfs.com>
7472 * add hard link support
7473 * change obdfile creation method
7474 * kernel patch changed
7476 2002-09-19 Peter Braam <braam@clusterfs.com>
7479 - stack overflow bug in extN fixed
7481 2002-09-18 Andreas Dilger <adilger@clusterfs.com>
7483 * documentation updates
7484 - add man pages for config tools
7485 - update tests/README to describe testing with new config tools
7486 - finish metadata API descriptions
7487 * bug fixes and cleanups
7488 - statfs workaround for 16TB limit
7489 - LOV stripe allocation improved, can stripe on subset of OSTs
7490 - LOV file size/IO offset was wrong for files > 4GB in size
7491 - object EA data was being dropped, caused files to be unreadable
7492 - memory overflow with non-LOV OST caused memory corruption
7493 - fixed regression tests to work with new config tools, obdfilter
7494 - fixed bug when directory size became larger than 1 block
7495 - fixed bug (for single client case) when PWD was deleted
7496 - invalidate local directory pages when doing intent-based ops
7497 - avoid LDLM oops when lock callback contained bad data
7499 2002-09-09 Andreas Dilger <adilger@clusterfs.com>
7501 * documentation updates
7502 * bug fixes and cleanups
7503 - configuration tools
7506 - 64-bit compile warnings
7507 - 64-bit internal statfs data
7509 * test_brw on persistent OST devices
7511 * lprocfs (disabled)
7513 2002-09-04 Andreas Dilger <adilger@clusterfs.com>
7515 * documentation updates
7516 * bug fixes and cleanups
7517 * configuration tools
7519 2002-08-30 Peter J. Braam <braam@clusterfs.com>
7522 * many small fixes to 0_5_4
7523 * io/network handling
7524 * thinkos in MDS operations
7526 2002-08-24 Peter J. Braam <braam@clusterfs.com>
7529 * crucial basic fixes to 0.5.3
7530 * IOR, Iozone work over Elan
7533 2002-08-07 Phil Schwan <phil@clusterfs.com>
7534 * version 0_5_3, our first alpha
7535 * we use the new Portals iovs
7536 * documentation updates
7537 * bug fixes and cleanups
7538 * small changes in the DLM wire protocol
7540 2002-07-25 Peter J. Braam <braam@clusterfs.com>
7541 * version 0_5_1 with some initial stability,
7542 * locking on MD and file I/O.
7543 * documentation updates
7544 * several bug fixes since 0.5.0
7545 * small changes in wire protocol
7547 2002-07-18 Phil Schwan <phil@clusterfs.com>
7549 * delivered as Lustre Light Alpha
7550 * fixed a crash after handling invalid MDS requests
7551 * fixed directory pages for architectures with non-4k pages sizes
7553 2002-07-11 Andreas Dilger <adilger@clusterfs.com>
7554 * release version v0_4_4
7555 * Moves TCP acceptor to be on port 2432 (unused Coda port) instead
7557 * Fixes a number of interruption problems with OST operations.
7558 * Update documentation for portals header changes
7559 * Move all wire protocol structs/defines to lustre_idl.h
7560 * Fixes symlink length bug.
7561 * Add tcpdump to repository.
7563 2002-07-05 Andreas Dilger <adilger@clusterfs.com>
7564 * release version v0_4_3
7565 * Fixes statfs for inodes on extN.
7566 * Fixes bug in runtests which would delete /etc/hosts.
7567 * Use 64-bit object IDs wherever possible (not into VFS though)
7568 Remove ost_get_info, which is unused by lustre, and out of date.
7570 2002-07-03 Peter Braam <braam@clusterfs.com>
7571 * release version v0_4_2 Fixes a lookup error (type not passed)
7572 * move forward to head of Portals
7573 * move forward to latest Lustre kernel
7575 2002-06-25 Peter Braam <braam@clusterfs.com>
7576 * release version v0_4_1. Hopefully stable on single node use.
git://git.whamcloud.com / fs / lustre-release.git / blob