4 2.6.16.60-0.42.8 (SLES 10),
5 2.6.27.39-0.3.1 (SLES11),
6 2.6.18-164.11.1.el5 (RHEL 5)
7 2.6.18-164.11.1.0.1.el5 (OEL 5)
8 * Client support for unpatched kernels:
9 (see http://wiki.lustre.org/index.php?title=Patchless_Client)
10 2.6.16 - 2.6.30 vanilla (kernel.org)
11 * Recommended e2fsprogs version: 1.41.10-sun2
12 * The async journal commit feature (bug 19128) and the cancel
13 lock before replay feature (bug 16774) are disabled by default.
15 Severity : enhancement
17 Description: add conf-param -d option to remove permanent settings.
18 Details : Add the ability to remove permanent lctl conf_param settings.
19 (Previously conf_param settings could only be changed, not
20 removed.) This also provides a method to change failover
21 nid locations. Improve lctl man page.
23 Severity : enhancement
25 Description: add list_param to b1_8 and add "-R" option to list params
28 Severity : enhancement
30 Description: lfs quota output is not very convenient for awk/sed-parsing
31 Details : Some positions in lfs quota output table could be empty or
32 non-empty which made it hard to parse it with scripts, now a dash
33 is put instead of space where there is not supposed to be any data.
35 Severity : enhancement
37 Description: fix obdfilter-survey script to work properly with remote oss-s
39 Severity : enhancement
41 Description: add new OBDFILTER_SURVEY test suite
43 Severity : enhancement
45 Description: add new multiple mount protection (MMP) test suite
47 Severity : enhancement
49 Description: add support for async journal commit in echo client
51 Severity : enhancement
53 Description: allow userland programs to include <lustre/lustre_idl.h>
54 from stardard include directories
56 Severity : enhancement
58 Description: The prune-icache-use-trylock is no longer needed now that
59 the patch from bug 20008 is landed.
63 Description: The shrink grant feature is still active on the client although the
64 connect flag is not set.
68 Description: Don't leak grant space if the write failed with quota exceeded.
72 Description: a race condition could lead to SIGBUS being sent to an
73 application using mmap-ped files from Lustre
74 Details : truncate_complete_page implementation for the patchless
75 client could arbitrarily unset PG_Uptodate flag for a
76 page being kicked from the page cache, an uptodate check
77 right after a readpage call in filemap_fault could fail
78 because of that as though the page read had been unsuccessful.
82 Description: dlm lock slab shrinking is not efficient
83 Details : The dlm_locks slab can grow significantly and consumes a lot of
84 memory on the server. Set a hardlimit to grant_plan.
88 Description: Lustre does not do 1MB IOs to HW RAID
89 Details : Bump MAX_PHYS/HW_SEGMENTS and SG_ALL to 256 in the RHEL5 kernel.
90 This is what we do already for SLES kernels.
94 Description: LSI Fusion MPT driver hacks to improve performance
95 Details : Set CONFIG_FUSION_MAX_SGE to 256 for RHEL5
97 Severity : enhancement
99 Description: increase default md stripe_cache_size to 16k
102 Bugzilla : 15587/21439
103 Description: don't handle security.capability xattr
104 Details : CONFIG_SECURITY_FILE_CAPABILITIES is enabled by default on SLES11.
105 This results in additional getxattr calls, causing VBR test
106 failures as well as a preformance drop when writing.
110 Description: obdfilter-survey is no longer working
111 Details : revert patch from bug 20355 to resolve an issue with lctl --threads
112 not working correctly with $(PTHREAD_LIBS) being linked to lctl.
116 Description: ll_shrink_cache does not handle __GFP_FS properly
120 Description: lfs getstripe shows wrong info for directories
121 Details : Set correct LOVEA default values for filesystem-wide.
125 Description: FSX checksum false positves due to mmap IO
126 Details : Use OBD_FL_MMAP flag for IOs on a memory mapped file. Do not print
127 checksum errors, if the flag is set on a request.
131 Description: file operations after eviction have successful return values
132 Details : use vfs ->flush callback to return any pending async errors
137 Description: mdsrate fails to write after 1.3+M files opened
138 Details : decrease memory usage on clients by recycling dentries and
143 Description: obdfilter-survey gives unreasonably high numbers
144 Details : Wait for all threads to complete when running test_brw.
148 Description: do not set lustre read_only device when server umount and keep
149 client records for recoverable ones
153 Description: move sync_on_lock_cancel tunable to the obdfilter layer
154 Details : move the tunable to trigger a journal flush on lock cancel from
155 the ost layer to the obdfilter layer. This tunable is useful
156 when using the async journal commit feature.
160 Description: exp->exp_nid_stats == NULL in filter_tally()
161 Details : fix race with per-nid stats by delaying procfs cleanup until
164 -------------------------------------------------------------------------------
166 2010-04-30 Oracle, Inc.
168 * Support for kernels:
169 2.6.16.60-0.42.8 (SLES 10),
170 2.6.27.39-0.3.1 (SLES11),
171 2.6.18-164.11.1.el5 (RHEL 5)
172 2.6.18-164.11.1.0.1.el5 (OEL 5)
173 * Client support for unpatched kernels:
174 (see http://wiki.lustre.org/index.php?title=Patchless_Client)
175 2.6.16 - 2.6.30 vanilla (kernel.org)
176 * Recommended e2fsprogs version: 1.41.10-sun2
177 * The async journal commit feature (bug 19128) and the cancel
178 lock before replay feature (bug 16774) are disabled by default.
182 Description: open-unlinked directories trigger MDS LBUG
183 Details : Fix regression introduced by the patch from bug 19640.
184 ext3_inc_count() can reset nlink to 1 when the directory
185 is indexed and inode->i_nlink == 2. Work around the problem
186 by incrementing nlink by 2 instead of 1.
190 Description: Reconnects are not throttled
191 Details : Don't wake up pinger on reconnect failures and rely on regular
192 pings to trigger the next reconnection. Please note that the
193 pinger already uses a smaller interval if the import is
197 Frequency : only with NFS export
199 Description: Console flooded with error message from ll_inode_from_lock()
201 Details : in mds_open, initialize the child_res_id before enqueuing
202 the OPEN lock for the child inode, then to avoid senting
203 wrong ldlm_res_id to client.
207 Description: allow multiple instances of the same nid in NID hash
208 Details : Case of multiple separate clients from the same NID (as
209 with liblustre) is legitimate and so we should allow
210 multiple instances of the same NID in nid hash.
214 Description: the readahead code can sleep on a semaphore while holding a
216 Details : in ras_update, "lov_get_info" could be called during increasing
217 readahead windows, which tries to get the mutex lock "lov_lock"
218 while holding the spin_lock "ras_lock", then causes system
223 Description: ASSERTION(cli->cl_avail_grant >= 0) failed
224 Details : fix assertion failure in the grant code.
228 Description: Use CNETERR (which is rate limited) in specific places in
229 the portal's LNET driver to avoid flooding the console.
233 Description: include last created object in precreate slow case
237 Description: don't do rep-ack if not created anything
238 Details : mds_open currently always put a lock into a rep-ack regardless
239 if something was created or not. This is pointless and only
240 creates needless contention. In fact the entire idea was to do
241 this for real creates as a recovery protection.
245 Description: Spurious error messages from smp_processor_id() on preemptible
247 Details : Disable a preemption by grabbing the lock in fs_trace_get_tcd()
248 first. The function fs_trace_get_tcd() was moved up.
252 Description: interval_erase() fix
253 Details : interval_erase() calls update_maxhigh() properly when child
258 Description: "lfs df" does not print stats for all mountpoints
259 Details : Print all mounted lustre filesystems with "lfs df"
263 Description: lfs setstripe -p not longer work with a relative pathname
264 Details : Use realpath() to provide absolute pathname.
268 Description: fix for truncated reply buffer
269 Details : reply buffer could be referred by reply_in_callback after released
274 Description: lustre.lov error when backing up symlinks with extended attributes
275 Details : Improved logic in ll_listxattr()
279 Description: properly handle null value for setattr -n lustre.lov
280 Details : Running "setfattr -n trusted.lov ." causes a NULL dereference
281 in ll_setxattr() due to no checking if "value" is NULL.
282 This command now resets to the default striping when executed
287 Description: stack overflow on lock cancellation due to fsync call
288 Details : sync_on_lock_cancel is needed for recovery when async journal
289 is enabled, but we actually just need to make sure that
290 metadata blocks have hit the journal, so doing a fs sync
291 should be enough and should consume less stack (just create an
292 empty handle and commmit it).
296 Description: using current->journal_info to store per-thread data leads
297 to problem under memory pressure
298 Details : disable the per-thread data (current->journal_info) containing
299 the lock info during I/O to work around the issue for short tem
303 Description: control DCACHE_LUSTRE_INVALID flag with MDS_INODELOCK_LOOKUP lock
304 Details : DCACHE_LUSTRE_INVALID is controlled by MDS_INODELOCK_LOOKUP
305 lock which is corresponding to "IT_LOOKUP", do not skip invalidate
310 Description: Cannot send after transport shutdown
311 Details : Clear imp_vbr_failed flag upon eviction
315 Description: soft lock in request set code during recovery
316 Details : during recovery, uses req->rq_set itself to replay the request
317 instead of ptlrpcd_recovery_pc
321 Description: Use CFS_ALLOC_IO instead of _STD in llap_from_page_with_lockh
322 Details : During an ll_readahead under ll_readpage, we have seen the the
323 OBD_SLAB_ALLOC hang under ldlm_pools_shrink when trying to lock
324 a page that is already locked by the readahead code.
329 Description: stop waitting for next replay transno if shutdown
330 Details : if the system is shutting down, wake up service thread blocked
331 to wait for next replay transno during recovery, then all the
332 references held by queued requests can be dropped and device
337 Description: lov_merge_lvb()) ASSERTION(spin_is_locked(&lsm->lsm_lock)) failed
338 Details : Protect lli->lli_smd pointer updates with lli->lli_lock.
342 Description: per-nid stats should not access lustre-hash internal structures
347 Description: mount.lustre fails to pass some options to mount()
352 Description: ext4 extent allocation is slower than in ext3
353 Details : Increase the default value of MB_DEFAULT_ORDER2_REQS to 8,
354 enlarge ext4 preallocation table for 2048 4K blocks extents
359 Description: incorrect triggering of synchronous IO
360 Details : The OSC can mistakenly fall back to synchronous IO when the
361 max_dirty_mb limit is reached and no write requests have yet
362 been issued. This can occur when the dirty pages are spread
363 over many files all of which are below the optimal request size.
367 Description: Optimize quota_ctl operations by sending requests in parallel
368 Details : Send MDS->OST quota_ctl requests in parallel, do not resend.
369 Compiled from two attachments in the ticket.
373 Description: ordering issue between transaction start & i_mutex
374 Details : start the transaction earlier in llog_lvfs_destroy to get
375 transaction start and inode mutex lock nested properly.
379 Description: lru resize SLV can get stuck
380 Details : calculate SLV with a greater precision to not lose small
381 changes due to interger math truncation; round up SLV only
382 if the amount of granted locks less than the limit to not
383 get stuck with this SLV
387 Description: avoid divide-by-zero in lprocfs_rd_import()
391 Description: lfs quota failed when OSTs are down
392 Details : really return approximate block/inode usage when OSTs are down
396 Description: llobdstat fix and enhancement
397 Details : add a counter to set a limit to how many samples will be returned
398 fix a wildcard in the path to limit to obdfilter stats only
402 Description: recovery-small 51 hang - MDT can not stop, Mount still busy with
404 Details : abort lock enqueue processing sooner when we umount is in progress.
406 Severity : enhancement
408 Description: Update RHEL5.4 kernel to 2.6.18-164.11.1.el5 and OEL5.4 kernel to
409 2.6.18-164.11.1.0.1.el5.
411 Severity : enhancement
413 Description: error message improvements
414 Details : Use INFO/WARN instead of WARN/ERROR for the slow messages.
415 Simplify MDT/OST service start message.
416 Suppress "changing the import ..." warning.
419 Bugzilla : 21961/17914
420 Description: ignore trailing -mdc when determining index number
423 Frequency : only with SLES11
425 Description: fix for a race condition in linux quotas implementation
426 Details : dq_flags(struct dquot) access is not properly locked which could
427 lead to certain inconsistencies when accessing it using non-atomic
428 bit operations like __set_bit in do_set_dqblk.
432 Description: return any pending async errors in close(2) using flush callback
434 -------------------------------------------------------------------------------
436 2010-01-29 Sun Microsystems, Inc.
438 * Support for kernels:
439 2.6.16.60-0.42.8 (SLES 10),
440 2.6.27.39-0.3.1 (SLES11),
441 2.6.18-164.11.1.el5 (RHEL 5)
442 2.6.18-164.6.1.0.1.el5 (OEL 5)
443 * Client support for unpatched kernels:
444 (see http://wiki.lustre.org/index.php?title=Patchless_Client)
445 2.6.16 - 2.6.30 vanilla (kernel.org)
446 * Recommended e2fsprogs version: 1.41.6.sun1
447 * The async journal commit feature (bug 19128) and the cancel
448 lock before replay feature (bug 16774) are disabled by default.
450 Severity : enhancement
452 Description: Update RHEL5.4 kernel to 2.6.18-164.11.1.el5 and
453 OEL5.4 kernel to 2.6.18-164.11.1.0.1.el5.
455 Severity : enhancement
456 Bugzilla : 21511/19848
457 Description: Update SLES11 kernel to 2.6.27.39-0.3.1.
459 Severity : enhancement
461 Description: Update supported SLES10 kernel to 2.6.16.60-0.42.8.
463 Severity : enhancement
465 Description: Update kernel to RHEL5.4 2.6.18-164.6.1.el5 and
466 OEL5 2.6.18-164.6.1.0.1.el5(Both in-kernel OFED enabled).
468 Severity : enhancement
470 Description: Build kernels (RHEL5, OEL5 and SLES10/11) using the vendor's own
473 Severity : enhancement
475 Description: Vanilla kernel 2.6.30 patchless client support.
480 Description: bad entry in directory xxx: inode out of bounds
481 Details : fix locking issue in the rename path which could race with any
482 other operations updating the same directory.
484 Severity : enhancement
486 Description: Make watchdog timer messages to be more clear and descriptive.
490 Description: cp -p command does not preserve the dates and timestamp
491 Details : mtime could be spoiled by a write callback
495 Description: Clear imp_force_reconnect correctly in ptlrpc_connect_interpret()
499 Description: Allow non-root access for "lfs check".
500 Details : Added a check in obd_class_ioctl() for OBD_IOC_PING_TARGET.
502 Severity : enhancement
504 Description: quotacheck performance/scaling issues
505 Details : reduce quotacheck time on empty filesystem by skipping uninit
508 Severity : enhancement
510 Description: Enhancement for lfs(1) command to use numeric uid/gid.
512 Severity : enhancement
514 Description: Adjust locks' extents on their first enqueue, so that at the time
515 they get granted, there is no need for another pass through the
516 queues since they are already shaped into the proper forms.
520 Description: Fix mds_shrink_intent_reply()/mds_intent_policy() to pass correct
521 arguments and prevent LBUG() in lustre_shrink_reply_v2().
525 Description: Change tunefs.lustre and mkfs.lustre --mountfsoptions so that
526 exactly the specified mount options are used. Leaving off
527 any "mandatory" mount options is an error. Leaving off any
528 default mount options causes a warning, but is allowed.
529 Change errors=remount-ro from mandatory to default. Sanitize
530 the mount string before storing it. Update man pages accordingly.
534 Description: mds_getattr() should return 0, even if mds_fid2entry() fails with
535 -ENOENT. Also fix in ptlrpc_expire_one_request() to print signed
538 Severity : enhancement
540 Description: Remove set_info(KEY_UNLINKED) from MDS/OSC
542 Severity : enhancement
544 Description: Clients can replay thousands of unused locks during recovery
545 Details : Don't replay unused locks (only read locks for now) during
546 recovery. This feature is disabled by default and can be
547 enabled by running the following command on the clients:
548 lctl get_param ldlm.cancel_unused_locks_before_replay
552 Description: can't stat file in some situation.
553 Details : improve initialize osc date when target is added to mds and
554 ability to resend too big getattr request is client isn't have info
559 Description: Prevent inconsistences between linux and lustre mount structures.
560 Details : Wait indefinitely in server_wait_finished() until mnt_count drops.
561 Make the sleep interruptible.
563 Severity : enhancement
565 Description: Communicate OST degraded/readonly state via statfs to MDS
566 Details : Flags in the statfs returned from OSTs indicate whether the
567 OST is in a degraded RAID state, or if the filesystem has
568 turned read-only after a filesystem error is detected.
573 Description: don't panic if EPROTO was hit when reading symlink
574 Details : correctly handling request reference in error cases.
579 Description: open sometimes returns ENOENT instead of EACCES
580 Details : checking permission should be part of open part of mds_open, not
581 lookup part. so server should be set DISP_OPEN_OPEN disposition
582 before starting permission check.
583 Also not need revalidate dentry if client already have LOOKUP lock.
587 Description: enable client interface failover
588 Frequency : on servers with multiple network interfaces
589 Details : When a child reconnects from another NID, properly update export
590 nid hash position and ldlm reverse import.
592 Severity : enhancement
594 Description: implemented direct I/O with arbitrary (nonaligned) memory
595 addresses and file offsets.
597 Severity : enhancement
599 Description: added more recovery timeout options.
601 Severity : enhancement
603 Description: added llapi_file_open, llapi_file_create, llapi_file_get_stripe
608 Description: Avoid deadlock for local client writes
609 Frequency : only on systems with clients writing to an OST on the same node
610 Details : Use new OBD_BRW_MEMALLOC flag to notify OST about writes in the
611 memory freeing context. This allows OST threads to set the
612 PF_MEMALLOC flag on task structures in order to allocate memory
613 from reserved pools and complete IO.
614 Use GFP_HIGHUSER for OST allocations for non-local client writes,
615 so that the OST threads generate memory pressure and allow
616 inactive pages to be reclaimed.
621 Descriptoin: lock ordering violation between &cli->cl_sem and _lprocfs_lock
622 Details : move ldlm namespace creation in setup phase to avoid grab
623 _lprocfs_lock with cli_sem held.
627 Frequency : only during format of test systems
628 Description: Unable to run several mkfs.lustre on loop devices at the same time
629 Details : mkfs.lustre returns error 256 on the concurrent loop devices
630 formatting. The solution is to proper handle the error.
632 Severity : enhancement
634 Description: implement async create (obd_async_create) method for osc, to avoid
635 too long waiting new ost objects with holding ldlm lock.
639 Frequency : occasionally during network problems
640 Description: client not allowed to reconnect to OST because of active request
641 Details : abort bulk requests received by the OST once the client has timed
642 out since the client will resend the request anyway. The client
643 also now retries to reconnect to the same server if a connect
644 request failed with EBUSY or -EAGAIN.
647 Frequency : rare, if used wide striped file and one ost in down.
649 Descriptoin: don't return error if we created a subset of objects for file.
650 Details : lov_update_create_set() uses set->set_success as index for created
651 objects, so if some requests failed, they will have hole at end of
652 array and we can use qos_shrink_lsm for allocate correct lsm.
656 Description: Slow stale export processing during normal start up
657 Details : The global mgc lock prevents OST setup to be run in parallel.
658 Replace the global lock with a per-config_llog_data semaphore.
662 Description: Out or order replies might be lost on replay
663 Details : In ptlrpc_retain_replayable_request if we cannot find retained
664 request with tid smaller then one currently being added, add it
665 to the start, not end of the list.
669 Description: BUG: soft lockup - CPU#1 stuck for 10s! [ll_mdt_07:4523]
670 Details : add cond_resched() calls to avoid hogging the cpu for too long
671 in the hash code. Make also lustre_hash_for_each_empty() more
674 Severity : enhancement
676 Description: Performance improvements for debug messages with D_RPCTRACE,
677 D_LDLM, D_QUOTA options.
680 Frequency : only with NFS export
682 Description: (lov_merge.c:74:lov_merge_lvb())
683 ASSERTION(spin_is_locked(&lsm->lsm_lock)) failed (SR 71691004)
684 Details : Fix a race in the nfs export code by populating inode
685 info while the new inode is still locked
687 Severity : enhancement
689 Description: add a new file in procfs called force_lbug. Writting to this
690 file triggers a LBUG. Only for test purpose.
694 Description: OOM killer causes node hang
695 Details : really interrupt the sleep in osc_enter_cache on signals
699 Description: LustreError: 9153:0:(quota_context.c:622:dqacq_completion()) LBUG
700 Details : fix race during quota release on the slave.
702 Severity : enhancement
704 Description: smaller hash bucket sizes, cleanups
705 Details : increase hash table sizes and enabled rehashing for pools, quota,
706 uuid, nid & per-nid stats.
708 Severity : enhancement
710 Description: Add ldiskfs maxdirsize mount option
711 Details : add max_dir size mount option
715 Description: panic in ll_statahead_thread
716 Details : prevent parent thread to be killed before its child
719 Frequency : only with 16TB device
721 Description: unable to perform "mount -t lustre" of 16TB OST device
722 Details : Mounting 16TB LUNs failed due to three bugs in mkfs.lustre.
726 Description: ASSERTION(atomic_read(&imp->imp_inflight) == 0) failed
727 Details : unregistering should be zero if no RPC inflight.
731 Description: hyperion: Oops during metabench
732 Details : Correct the refcount of lov_request_set
734 Severity : enhancement
736 Description: Add mptlinux and nxge drivers to Lustre builds
738 Severity : enhancement
740 Description: Fix watchdog timer message to be more clear
741 Details : Make watchdog timer messages more clear and descriptive.
745 Description: LNET soft lockups in socknal_cd thread
746 Details : don't hog CPU for active-connecting if another connd is
747 accepting connecting-requst from the same peer
751 Description: recovery-small test_17 hang
752 Details : Land several AT improvements & fixes.
756 Description: MDS panic and hanging client processes
757 Details : Replace exp_ops_stats with exp_nid_stats->nid_stats
761 Description: OSS stuck in recovery.
762 Details : fix race during recovery. class_unlink_export,
763 class_set_export_delayed and target_queue_last_replay_reply
764 may race while increasing/decreasing obd_recoverable_clients
765 and obd_delayed_clients, causing recovery to wait forever.
767 Severity : enhancement
769 Description: add cascading_rw.c to lustre/tests
773 Description: filter_last_id() NULL deref
774 Details : lprocfs_filter_rd_last_id() should check for the fully
775 setup obd device, before proceeding further.
777 Severity : enhancement
779 Description: Loadgen improvements
780 Details : stacksize and locking fixes for loadgen
784 Description: Quiet CERROR("dirty %d > system dirty_max %d\n"
785 Details : The atomic_read() allowing the atomic_inc() are not covered
786 by a lock. Thus they may safely race and trip this CERROR()
787 unless we add in a small fudge factor (+1).
789 Severity : enhancement
791 Description: shrink_slab: nr=-9223362083340912175
792 Details : fix spurious message from shrink_slab reporing negative nr
796 Description: Quiet bogus previously committed transno error
797 Details : suppress the "server went back in time" error message which
798 is always printed even in the common case after a client eviction
800 Severity : enhancement
802 Description: Parallel statfs() calls result in client eviction
803 Details : cache statfs data for 1s.
807 Description: parallel-scale test_compilebench: @@@@@@ FAIL: compilebench
809 Details : fix serveral issues in pinger code causing clients not to ping
810 servers for too long, resulting in evictions.
814 Description: e2fsck should warn when MMP update interval is extended
815 Details : print mmp_check_interval and make it possible to abort mount
816 operation in case it takes too long.
820 Description: mdsrate-create-large.sh, BUG: soft lockup - CPU#0 stuck for 10s!
821 Details : fix bug in the RHEL5's jbd2 callback patch.
825 Description: drop number of active requests when queued for recovery
826 Details : Now that we take a reference on the original request instead of
827 making a copy of it for recovery. We need to drop the number of
828 active requests or the queued requests will prevent all request
829 processing when they exceed (srv->srv_threads_running - 1).
831 Severity : enhancement
833 Description: refuse to invalidate operational quota files when they are in use
834 Details : an attempt to invalidate operational quota files on the quota
835 master is not actually permitted by VFS (returning -EPERM), but we
836 should not depend on that and should return the error earlier.
840 Description: Applications stuck in jbd2_log_wait_commit during exit
841 Details : fix deadlock between kjournald2 trying to acquire the page lock
842 owned by an ost_io thread waiting for journal commit.
844 -------------------------------------------------------------------------------
846 2009-10-16 Sun Microsystems, Inc.
848 * Support for kernels:
849 2.6.16.60-0.42.4 (SLES 10),
850 2.6.27.29-0.1 (SLES11, i686 & x84_64 only),
851 2.6.18-128.7.1.el5 (RHEL 5),
852 2.6.18-128.7.1.el5 (OEL 5),
853 * Client support for unpatched kernels:
854 (see http://wiki.lustre.org/index.php?title=Patchless_Client)
855 2.6.16 - 2.6.27 vanilla (kernel.org)
856 * Recommended e2fsprogs version: 1.41.6.sun1
857 * ext4 support for RHEL5 is experimental and thus should not be
860 Severity : enhancement
862 Description: Add OEL5 support.
864 Severity : enhancement
866 Description: Update kernel to SLES11 2.6.27.29-0.1.
870 Description: File checksum failures with OST read cache on
871 Details : Disable page poisoning when the bulk transfer has to be aborted
872 because the client got evicted.
876 Description: Don't allow make backward step on assiging osc next id.
877 Details : race between allocation next id and ll_sync thread can be cause
878 of set wrong osc next id and can be kill valid ost objects.
880 Severity : enhancement
882 Description: Update kernel to RHEL5 2.6.18-128.7.1.el5.
884 Severity : enhancement
886 Description: Update kernel to SLES10 SP2 2.6.16.60-0.42.4.
890 Description: Changes in raid5-large-io-rhel5.patch to calculate sectors properly
894 Description: Increase the default BLK_DEF_MAX_SECTORS value for RHEL5 and SLES11
898 Description: Error handling in osc_statfs_interpret() has been improved.
899 Details : Check in osc_statfs_interpret() for EBADR.
903 Description: Do not send statfs() requests to OSTs disabled by administrator.
904 Details : Check in lov_prep_statfs_set() for non-NULL ltd_exp.
908 Description: Do not update ctime for the deleted inode.
909 Details : Check in mds_reint_unlink() before calling fsfilt_setattr().
913 Description: Increase of the size of the LDLM resource hash.
914 Details : Bump up RES_HASH_BITS=12.
918 Description: correctly send lsm on open replay
919 Details : MDS is trust to LSM size on replay open, but client can set wrong size
924 Description: Deadlock between filter_destroy() and filter_commitrw_write().
925 Details : filter_destroy() does not hold the DLM lock over the whole
926 operation. If the DLM lock is dropped, filter_commitrw() can go
927 through, causing the deadlock between page lock and i_mutex.
928 The i_alloc_sem should also be hold in filter_destroy() while
933 Description: truncate starts GFP_FS allocation under transaction causing deadlock
934 Details : ldiskfs_truncate calls grab_cache_page which may start page
935 allocation under an open transaction. This may lead to
936 calling prune_icache with consequent lustre reentrance.
939 Frequency : only when down/upgrading the MDS to 1.6/1.8 while 1.8 clients are
940 still up and when the OST pool feature is used
942 Description: interop testing got LBUG when run dd with OST pool
943 :LustreError: 30032:0:(llite_lib.c:1913:ll_replace_lsm()) LBUG
944 Details : down/upgrading the MDS to a version that doesn't/does support OST
945 pool can cause clients to crash because the lsm has changed
950 Description: missing tree_status on 1.8.1 RPM build
951 Details : make rpms failed due because the tree_status file is missing.
955 Description: continuing LustreError "mds adjust qunit failed!"
956 Details : don't print message on the console when ->adjust_qunit fails.
960 Description: don't increase ldlm timeout if previous client was evicted
961 Details : if a client doesn't respond to a blocking callback within the
962 adaptive ldlm enqueue timeout, don't adjust the adaptive estimate
963 when the lock is next granted.
967 Description: ost is being unmounted w/o all writes to last_rcvd landing on disk.
968 affects recovery negatively.
969 Details : make sure all exports have been properly destroyed by the zombie
970 thread processed before stopping the target.
974 Description: Performance degradation with O_DIRECT between 1.6 & 1.8.1 b190
975 Details : disable write barrier for ext4/SLES11.
979 Description: Kernel panic - not syncing: Out of memory and no killable
980 processes... on OSS when iozone
981 Details : fix memory leak in the journal checksum patch.
985 Description: group quota "too many blocks" OSS crashes
986 Details : we should keep the same uid/gid for lquota_chkquota() and
987 lquota_pending_commit()
991 Description: LustreError: 9153:0:(quota_context.c:622:dqacq_completion()) LBUG
992 Details : don't LBUG on release quota error. Just a workaround until the
993 problem is understood.
995 ------------------------------------------------------------------------------
996 2009-07-31 Sun Microsystems, Inc.
998 * Support for kernels:
999 2.6.16.60-0.39.3 (SLES 10),
1000 2.6.27.23-0.1 (SLES11, i686 & x84_64 only),
1001 2.6.18-128.1.14.el5 (RHEL 5),
1002 * Client support for unpatched kernels:
1003 (see http://wiki.lustre.org/index.php?title=Patchless_Client)
1004 2.6.16 - 2.6.27 vanilla (kernel.org)
1005 * Recommended e2fsprogs version: 1.41.6.sun1
1006 * File join has been disabled in this release, refer to Bugzilla 16929.
1007 * NFS export disabled when stack size < 8192. Since the NFSv4 export
1008 of Lustre filesystem with 4K stack may cause a stack overflow. For
1009 more information, please refer to bugzilla 17630.
1010 * ext4 support for RHEL5 is experimental and thus should not be
1013 Severity : enhancement
1015 Description: Update kernel to SLES10 SP2 2.6.16.60-0.39.3.
1018 Frequency : with 1.8 server and 1.6 clients
1020 Description: correctly shrink reply for avoid send too big message to client.
1021 Details : 1.8 mds is allocate to big buffer to LOV EA data and this produce
1022 some problems with sending this reply to 1.6 client.
1026 Description: Repeated atomic allocation failures.
1027 Details : Use GFP_HIGHUSER | __GFP_NOMEMALLOC flags for memory allocations
1028 to generate memory pressure and allow reclaiming of inactive pages.
1029 At the same time, do not allow to exhaust emergency pools.
1030 For local clients the use of GFP_NOFS will be introduced in 1.8.2
1032 Severity : enhancement
1033 Bugzilla : 19846, 18289
1034 Description: Update kernel to RHEL5 2.6.18-128.1.14.el5.
1036 Severity : enhancement
1037 Bugzilla : 19625, 16893, 18668, 19848
1038 Description: Add support for SLES11 2.6.27.23-0.1.
1040 Severity : enhancement
1042 Description: Update client support to vanila kernels up to 2.6.27.
1044 Severity : enhancement
1046 Description: Update kernel to SLES10 SP2 2.6.16.60-0.37.
1048 Severity : enhancement
1050 Description: Compile with -Werror by default for i686 and x86_64.
1054 Description: resolve race between obd_disconnect and class_disconnect_exports
1055 Details : if obd_disconnect will be called to already disconnected export he
1056 forget release one reference and osc module can't unloaded.
1058 Severity : enhancement
1060 Description: move AT tunable parameters for more consistent usage
1061 Details : add AT tunables under /proc/sys/lustre, add to conf_param parsing
1065 Descriptoin: correctly skip time estimate if in recovery
1066 Details : rq_send_state insn't bitmask so using bitwise ops is forbid.
1070 Descriptoin: OSS DeadLock
1071 Details : Use trylock to prevent deadlock when shrink icache.
1073 Severity : enhancement
1075 Description: Allow tuning service thread via /proc
1076 Details : For each service a new
1077 /proc/fs/lustre/{service}/*/thread_{min,max,started} entry is
1078 created that can be used to set min/max thread counts, and get the
1079 current number of running threads.
1081 Severity : enhancement
1083 Description: Add state history info file, enhance import info file
1084 Details : Track import connection state changes in a new osc/mdc proc file;
1085 add overview-type data to the osc/mdc import proc file.
1089 Description: Reduce small size read RPC
1090 Details : Set read-ahead limite for every file and only do read-ahead when
1091 available read-ahead pages are bigger than 1M to avoid small size
1096 Description: free_entry erroneously used groups_free instead of put_group_info
1098 Severity : enhancement
1100 Description: Make read-ahead stripe size aligned.
1102 Severity : enhancement
1104 Description: MDS create should not wait for statfs RPC while holding DLM lock.
1107 Frequency : rare, connect and disconnect target at same time
1109 Descriptoin: ASSERTION(atomic_read(&imp->imp_inflight) == 0
1110 Details : don't call obd_disconnect under lov_lock. this long time
1111 operation and can block ptlrpcd which answer to connect request.
1114 Frequency : start MDS on uncleanly shutdowned MDS device
1116 Descriptoin: ll_sync thread stay in waiting mds<>ost recovery finished
1117 Details : stay in waiting mds<>ost recovery finished produce random bugs
1118 due race between two ll_sync thread for one lov target. send
1119 ACTIVATE event only if connect realy finished and import have
1123 Frequency : start MDS on uncleanly shutdowned MDS device
1125 Descriptoin: aborting recovery hang on MDS
1126 Details : don't throttle destroy RPCs for the MDT.
1130 Description: Slow reads beyond 8Tb offsets.
1131 Details : Page index integer overflow in ll_read_ahead_page
1135 Description: Soft lockup on OSS after MDS failover
1136 Details : MSG_CONNECT_INITIAL is not set on the initial MDS->OST connect.
1137 As a conseqence, the patch from bug 18224 is not operational
1138 and the MDS export cannot be reused on the OSTs until it gets
1142 Frequency : rare, only if using MMP with Linux RAID
1144 Description: MMP doesn't work with Linux RAID
1145 Details : While using HA for Lustre servers with Linux RAID, it is possible
1146 that MMP will not detect multiple mounts. To make this work we
1147 need to unplug the device queue in RAID when the MMP block is being
1148 written. Also while reading the MMP block, we should read it from
1149 disk and not the cached one.
1152 Frequency : rare, during recovery
1154 Description: Assertion failure in ldlm_lock_put
1155 Details : Do not put cancelled locks into replay list, hold references on
1156 locks in replay list
1160 Description: 1.6.5 mdsrate performance is slower than 1.4.11/12 (MDS is not cpu bound!)
1161 Details : create_count always drops to the min value (=32) because grow_count
1162 is being changed before the precreate RPC completes.
1165 Frequency : Only in RHEL5 when mounting multiple ext3 filesystems
1168 Description: "kmem_cache_create: duplicate cache jbd_4k" error message
1169 Details : add proper locking for creation of jbd_4k slab cache
1173 Description: MMP check in ext3_remount() fails without displaying any error
1174 Details : When multiple mount protection fails during remount, proper error
1179 Description: Rare Client crash on resend if the file was deleted.
1180 Details : When file is opened, but open reply is lost and file is
1181 subsequently deleted before resend, resend processing logic
1182 breaks trying to open the file again, should not try to open.
1186 Description: add check for >8TB ldiskfs filesystems
1187 Details : ext3-based ldiskfs does not support greater than 8TB LUNs.
1188 Don't allow >8TB ldiskfs filesystems to be mounted without
1189 force_over_8tb mount option
1193 Description: Client locked up when running multiple instances of an app. on
1194 multiple mount points
1195 Details : ll_shrink_cache() can sleep while holding the ll_sb_lock.
1196 Convert ll_sb_lock to a read/write semaphore to fix the problem.
1200 Description: Cannot acces an NFS-mounted Lustre filesystem
1201 Details : An NFS client cannot access the Lustre filesystem NFS-mounted
1202 from a Lustre-client exporting the Lustre filesystem via NFS.
1206 Description: panic in ll_statahead_thread
1207 Details : grab dentry reference in parent process.
1210 -------------------------------------------------------------------------------
1212 tbd Sun Microsystems, Inc.
1214 * Support for kernels:
1215 2.6.16.60-0.31 (SLES 10),
1216 2.6.18-128.1.6.el5 (RHEL 5),
1217 2.6.22.14 vanilla (kernel.org)
1218 * Client support for unpatched kernels:
1219 (see http://wiki.lustre.org/index.php?title=Patchless_Client)
1220 2.6.16 - 2.6.22 vanilla (kernel.org)
1221 * Recommended e2fsprogs version: 1.40.11-sun1
1222 * File join has been disabled, refer to Bugzilla 16929.
1223 * A new Lustre ADIO driver is available for MPICH2-1.0.7.
1224 * NFS export disabled when stack size < 8192. Since the NFSv4 export of
1225 Lustre filesystem with 4K stack may cause a stack overflow. For more
1226 information, please refer to bugzilla 17630.
1228 Severity : enhancement
1230 Description: Update to RHEL5.3 kernel-2.6.18-128.1.6.el5.
1232 Severity : enhancement
1234 Description: Update OFED release to 1.4.1 RC4
1236 Severity : major, only with big OST
1238 Description: Very poor metadata performance on Infiniband lustre configuration
1239 Details : OST object precreation becomes very slow on big OSTs. This is due
1240 to the ialloc patch spending too much time scanning groups.
1243 Frequency : during recovery
1245 Description: don't mix llog inodes with normal.
1246 Details : allocate inodes for log in last inode group
1251 Description: fix lqs' reference which won't be put in some situations
1252 Details : This patch fixes:
1253 1. In quota_check_common(), this function will check quota
1254 for user and group, but only send one return via "pending".
1255 In most cases, the pendings should be same. But that is not
1257 2. If quotaoff runs between lquota_chkquota() and
1258 lquota_pending_commit(), the same thing will happen too.
1259 That is why it comes:
1260 - if (!ll_sb_any_quota_active(qctxt->lqc_sb))
1263 -------------------------------------------------------------------------------
1265 2008-12-31 Sun Microsystems, Inc.
1267 * Support for kernels:
1268 2.6.5-7.314 (SLES 9),
1269 2.6.9-67.0.22.EL (RHEL 4),
1270 2.6.16.60-0.31 (SLES 10),
1271 2.6.18-92.1.17.el5 (RHEL 5),
1272 2.6.22.14 vanilla (kernel.org)
1273 * Client support for unpatched kernels:
1274 (see http://wiki.lustre.org/index.php?title=Patchless_Client)
1275 2.6.16 - 2.6.22 vanilla (kernel.org)
1276 * Client support for unpatched kernels:
1277 we do not recommend using patchless RHEL4 clients with kernels
1278 prior to 2.6.9-55EL (RHEL4U5).
1279 * Recommended e2fsprogs version: 1.40.11-sun1
1280 * Note that reiserfs quotas are disabled on SLES 10 in this kernel.
1281 * RHEL 4 and RHEL 5/SLES 10 clients behaves differently on 'cd' to a
1282 removed cwd "./" (refer to Bugzilla 14399).
1283 * A new quota file format has been introduced in 1.6.5.
1284 The format conversion from prior releases is handled transparently,
1285 but releases older than 1.4.12/1.6.5 don't understand this new
1286 format. The automatic format conversion can be avoided by running
1287 the following command on the MDS:
1288 'tunefs.lustre --param="mdt.quota_type=ug1" $MDTDEV'.
1289 For more information, please refer to bugzilla 13904.
1290 * A new quota file format was introduced in 1.6.6/1.8.0 (kernels 2.6.16+).
1291 The format conversion from prior releases is handled transparently,
1292 but releases older than 1.6.6/1.8.0 don't understand this new
1293 format. The automatic format conversion can be avoided by running
1294 the following commands on the MDS and OSS servers (for
1295 pre 1.4.12-1.6.5 quota files):
1296 'tunefs.lustre --param="mdt.quota_type=ug1" $MDTDEV',
1297 'tunefs.lustre --param="ost.quota_type=ug1" $MDTDEV'
1298 or (for 1.4.12/1.6.5 quota files)
1299 'tunefs.lustre --param="mdt.quota_type=ug2" $MDTDEV',
1300 'tunefs.lustre --param="ost.quota_type=ug2" $MDTDEV'
1301 For more information, please refer to bugzilla 13904.
1302 * Output of lfs quota has been made less detailed by default,
1303 old (verbose) output can be obtained by using -v option.
1304 * File join has been disabled in this release, refer to Bugzilla 16929.
1305 * A new Lustre ADIO driver is available for MPICH2-1.0.7.
1306 * NFS export disabled when stack size < 8192. Since the NFSv4 export of
1307 Lustre filesystem with 4K stack may cause a stack overflow. For more
1308 information, please refer to bugzilla 17630.
1310 Severity : enhancement
1312 Description: Caching OSS
1313 Details : introduce data caching on the OSS. The OSS now relies on the linux
1314 kernel page cache to keep recently accessed data in memory.
1315 It is worth noting that all write requests are still flushed
1316 synchronously as in lustre 1.6.
1318 Severity : enhancement
1320 Description: Version based recovery
1321 Details : introduce finer grained recovery able to detect transaction
1322 dependencies and can deal with transaction gaps caused by clients
1323 failing at the same time as the server.
1325 Severity : enhancement
1327 Description: Enable adaptive timeouts by default
1328 Details : The Lustre timeout value in /proc/sys/lustre/timeout is now
1329 managed dynamically based on server load and should not need
1330 to be tuned manually based on cluster size. This allows Lustre
1331 to work under a wider variety of system sizes and loads, without
1332 unnecessarily causing lengthy recovery times.
1334 Severity : enhancement
1336 Description: Add OST Pools support
1337 Details : File striping can now be set to use an arbitrary pool of OSTs
1339 Severity : enhancement
1341 Description: add lazystatfs mount option to allow statfs(2) to skip down OSTs
1342 Details : allow skip disconnected ost for send statfs request and hide error
1346 Frequency : rare, on llog test 6
1348 Descriptoin: don't allow connect to already connected import
1349 Details : allowing connect to already connected import is hide connecting problem.
1352 Frequency : rare, on failed llog setup
1354 Descriptoin: don't leak obd reference on failed llog setup
1355 Details : for failed llog setup - mgc forget call class_destroy_import for
1356 client import, move destroy import to more generic place.
1361 Descriptoin: allow kill process which wait statahead result
1362 Details : for some reasons 'ls' can stick in waiting result from statahead,
1363 in this case need way for kill this process.
1366 Frequency : rare, at shutdown
1368 Descriptoin: panic at umount
1369 Details : llap_shrinker can be raced with killing super block from list and
1370 this produce panic with access to already freeded pointer
1375 Descriptoin: don't lose wakeup for imp_recovery_waitq
1376 Details : recover_import_no_retry or invalidate_import and import_close can
1377 both sleep on imp_recovery_waitq, but we was send only one wakeup
1383 Descriptoin: panic in mds_open
1384 Details : don't confuse mds_finish_transno() with PTR_ERR(-ENOENT)
1389 Descriptoin: stuck in cache_remove_extent() or panic with accessing to already
1391 Details : release lock refernce only after add page to pages list.
1394 Frequency : always with long access acl
1396 Descriptoin: mds can't pack reply with long acl.
1397 Details : mds don't control size of acl but they limited by reint/getattr
1402 Frequency : on remount
1404 Description: external journal device not working after the remount
1405 Details : clear dev_rdonly flag for external journal devices in
1411 Description: shutdown vs evict race
1412 Details : client_disconnect_export vs connect request race.
1413 if client will evicted at this time - we start invalidate
1414 thread without referece to import and import can be freed
1420 Description: shrink LOV EAs before replying
1421 Details : correctly adjust LOV EA buffer for reply.
1426 Description: don't skip ost target if they assigned to file
1427 Details : Drop slow OSCs if we can, but not for requested start idx.
1428 This means "if OSC is slow and it is not the requested
1429 start OST, then it can be skipped, otherwise skip it only
1430 if it is inactive/recovering/out-of-space.
1432 Severity : enhancement
1434 Description: Update to RHEL5 kernel-2.6.18-92.1.17.el5.
1437 Frequency : rare, need acl's on inode.
1439 Description: client can't handle ost additional correctly
1440 Details : if ost was added after client connected to mds client can have
1441 hit lnet_try_match_md ... to big messages to wide striped files.
1442 in this case need teach client to handle config events about add
1443 lov target and update client max ea size at that event.
1445 Severity : enhancement
1447 Description: Update to sles9 kernel-2.6.5-7.314.
1449 Severity : enhancement
1451 Description: Update to SLES10 SP2 kernel-2.6.16.60-0.31.
1454 Frequency : Create a symlink file with a very long name
1456 Description: ldlm_cancel_pack()) ASSERTION(max >= dlm->lock_count + count)
1457 Details : If there is no extra space in the request for early cancels,
1458 ldlm_req_handles_avail() returns 0 instead of a negative value.
1463 Description: mds is deadlocked
1464 Details : in rare cases, inode in catalog can have i_no less than have parent
1465 i_no, this produce wrong order for locking during open, and
1466 parallel unlink can be lock open. this need teach mds_open to grab
1467 locks in resource id order, not at parent -> child order.
1469 Severity : enhancement
1471 Description: Add /proc entry for import status
1472 Details : The mdc, osc, and mgc import directories now have
1473 an import directory that contains useful import data for debugging
1474 connection problems.
1476 Severity : enhancement
1478 Description: Re-disable certain /proc logging
1479 Details : Enable and disable client's offset_stats, extents_stats and
1480 extents_stats_per_process stats logging on the fly.
1483 Frequency : Only on FC kernels 2.6.22+
1485 Description: oops in statahead
1486 Details : Do not drop reference count for the dentry from VFS when lookup,
1487 VFS will do that by itself.
1489 Severity : enhancement
1491 Description: Generic /proc file permissions
1492 Details : Set /Proc file permissions in a more generic way to enable non-
1493 root users operate on some /proc files.
1497 Description: Hitting mdc_commit_close() ASSERTION
1498 Details : Properly handle request reference release in
1499 ll_release_openhandle().
1503 Frequency : only patchless client
1504 Description: add workaround for race between add/remove dentry from hash
1506 Severity : enhancement
1508 Description: Allow OST glimpses to return PW locks
1512 Description: LBUG when llog conf file is full
1513 Details : When llog bitmap is full, ENOSPC should be returned for plain
1518 Description: Prevent import from entering FULL state when server in recovery
1522 Description: service mount cannot take device name with ":"
1523 Details : Only when device name contains ":/" will mount treat it as
1529 Description: replace ptlrpcd with the statahead thread to interpret the async
1530 statahead RPC callback
1534 Frequency : on recovery
1535 Description: I/O failures after umount during fail back
1536 Details : if client reconnected to restarted server we need join to recovery
1537 instead of find server handler is changed and process self eviction
1538 with cancel all locks.
1540 Severity : enhancement
1542 Description: Update to RHEL5 kernel-2.6.18-92.1.10.el5.
1546 Description: Kernel BUG tries to release flock
1547 Details : Lustre does not destroy flock lock before last reference goes
1548 away. So always drop flock locks when client is evicted and
1549 perform unlock regardless of successfulness of speaking to MDS.
1551 Severity : enhancement
1553 Description: Update to SLES10 SP2 kernel-2.6.16.60-0.27.
1555 Severity : enhancement
1557 Description: Upcall on Lustre log has been dumped
1558 Details : Allow for a user mode script to be called once a Lustre log has
1559 been dumped. It passes the filename of the dumped log to the
1560 script, the location of the script can be specified via
1561 /proc/sys/lnet/debug_log_upcall.
1566 Description: avoid messages about idr_remove called for id that is not allocated
1567 Details : Move assigment s_dev for clustered nfs to end of initialization,
1568 for avoid problem with error handling.
1573 Description: avoid Already found the key in hash [CONN_UNUSED_HASH] messages
1574 Details : When connection is reused this not moved from CONN_UNUSED_HASH into
1575 CONN_USED_HASH and this prodice warning when put connection again
1581 Description: avoid ASSERTION(client_stat->nid_exp_ref_count == 0) failed
1582 Details : release reference to stats when client disconnected, not
1583 when export destroyed for avoid races when client destroyed
1584 after main ost export.
1588 Description: more cleanup in mds_lov
1589 Details : add workaround for get valid ost count for avoid warnings about
1590 drop too big messages, not init llog cat under semphore which
1591 can be blocked on reconnect and break normal replay, fix access
1594 Severity : enhancement
1596 Description: Export bytes_read/bytes_write count on OSC/OST.
1600 Description: Early reply size mismatch, MGC loses connection
1601 Details : Apply the MGS_CONNECT_SUPPORTED mask at reconnect time so
1602 the connect flags are properly negotiated.
1606 Description: Properly propagate oinfo flags from lov to osc for statfs
1607 Details : restore missing copy oi_flags to lov requests.
1611 Description: exports in /proc are broken
1612 Details : recreate /proc entries for clients when they reconnect.
1614 Severity : enhancement
1616 Description: Add man pages for llobdstat(8), llstat(8), plot-llstat(8),
1617 : l_getgroups(8), lst(8), routerstat(8)
1618 Details : included man pages for llobdstat(8), llstat(8),
1619 : plot-llstat(8), l_getgroups(8), lst(8), routerstat(8)
1621 Severity : enhancement
1623 Description: Implement lustre ll_show_options method.
1625 Severity : enhancement
1627 Description: Update to SLES9 kernel-2.6.5-7.312.
1629 Severity : enhancement
1631 Description: Update to RHEL4 kernel-2.6.9-67.0.22.EL.
1635 Description: exports in /proc are broken
1636 Details : recreate /proc entries for clients when they reconnect.
1640 Description: don't fail open with -ERANGE
1641 Details : if client connected until mds will be know about real ost count
1642 get LOV EA can be fail because mds not allocate enougth buffer
1647 Description: Resolve device initialization race
1648 Details : Prevent proc handler from accessing devices added to the
1649 obd_devs array but yet be intialized.
1651 Severity : enhancement
1653 Description: configure's --enable-quota should check the
1654 : kernel .config for CONFIG_QUOTA
1655 Details : configure is terminated if --enable-quota is passed but
1656 : no quota support is in kernel
1658 Severity : enhancement
1660 Description: Update to SLES10 SP2 kernel-2.6.16.60-0.23.
1662 Severity : enhancement
1664 Description: Update to RHEL5 kernel-2.6.18-92.1.6.el5.
1668 Frequency : rare, on PPC clients
1669 Description: don't swab ost objects in response about directory, because
1671 Details : bug similar bug 14856, but in different function.
1673 Severity : enhancement
1675 Description: lfs quota tool enhancement
1676 Details : added units specifiers support for setquota, default to
1677 current uid/gid for quota report, short quota stats by
1678 default, nonpositional parameters for setquota, added
1679 llapi_quotactl manual page.
1681 Severity : enhancement
1683 Description: *optional* service tags registration
1684 Details : if the "service tags" package is installed on a Lustre node
1685 When the filesystem is mounted, a local-node service tag will
1686 be created. See http://inventory.sun.com/ for more information
1687 about the Service Tags asset management system.
1691 Description: Client runs out of low memory
1692 Details : Consider only lowmem when counting initial number of llap pages
1695 Frequency : occasional
1697 Description: add refcount for osc callbacks, so avoid panic on shutdown
1699 Severity : enhancement
1701 Description: Update to RHEL4 kernel-2.6.9-67.0.20.
1704 Frequency : testing only
1706 Description: sanity test 65a fails if stripecount of -1 is set
1707 Details : handle -1 striping on filesystem in ll_dirstripe_verify
1710 Frequency : only in unusual configurations
1712 Description: Kernel panic with find ost index.
1713 Details : lov_obd have panic if some OST's have sparse indexes.
1715 Severity : enhancement
1717 Description: Update to RHEL5 kernel-2.6.18-53.1.21.el5.
1720 Frequency : rarely, if filesystem is mounted with -o flock
1722 Description: do not process already freed flock
1723 Details : flock can possibly be freed by another thread before it reaches
1724 to ldlm_flock_completion_ast.
1727 Frequency : rarely, if filesystem is mounted with -o flock
1729 Description: LBUG during stress test
1730 Details : Need properly lock accesses the flock deadlock detection list.
1733 Frequency : rarely, if binaries are being run from Lustre
1735 Description: oops in page fault handler
1736 Details : kernel page fault handler can return two special 'pages' in
1737 error case, don't try dereference NOPAGE_SIGBUS and NOPAGE_OMM.
1740 Frequency : rarely, during shutdown
1742 Description: timeout with invalidate import.
1743 Details : ptlrpcd_check call obd_zombie_impexp_cull and wait request which
1744 should be handled by ptlrpcd. This produce long age waiting and
1745 -ETIMEOUT ptlrpc_invalidate_import and as result LASSERT.
1751 Description: ASSERTION(CheckWriteback(page,cmd)) failed
1752 Details : badly clear PG_Writeback bit in ll_ap_completion can produce false
1756 Frequency : only with broken builds/installations
1758 Description: no LBUG if lquota.ko and fsfilt_ldiskfs.ko are different versions
1759 Details : just return an error to a user, put a console error message
1761 Severity : enhancement
1763 Description: Update to RHEL5 kernel-2.6.18-53.1.19.el5.
1765 Severity : enhancement
1767 Description: Update to RHEL4 kernel-2.6.9-67.0.15.
1769 Severity : enhancement
1771 Description: enable MGS and MDT services start separately
1772 Details : add a 'nomgs' option in mount.lustre to enable start a MDT with
1773 a co-located MGS without starting the MGS, which is a complement
1774 to 'nosvc' mount option.
1777 Frequency : always, on big-endian systems
1779 Description: cleanup in ptlrpc code, related to PPC platform
1780 Details : store magic in native order avoid panic's in recovery on PPC
1781 node and forbid from this error in future. Also fix posibily
1782 of twice swab data. Fix get lov striping to userpace.
1785 Frequency : rarely, if replay get lost on server
1787 Description: server incorrectly drop resent replays lead to recovery failure.
1788 Details : do not drop replay according to msg flags, instead we check the
1789 per-export recovery request queue for duplication of transno.
1792 Frequency : after recovery
1794 Description: precreate to many object's after del orphan.
1795 Details : del orphan st in oscc last_id == next_id and this triger growing
1796 count of precreated objects. Set flag LOW to skip increase count
1797 of precreated objects.
1800 Frequency : rare, on clear nid stats
1802 Description: ASSERTION(client_stat->nid_exp_ref_count == 0)
1803 Details : when clean nid stats sometimes try destroy live entry,
1804 and this produce panic in free.
1807 Frequency : occasionally since 1.6.4
1809 Description: Stack overflow during MDS log replay
1810 Details : ease stack pressure by using a thread dealing llog_process.
1813 Frequency : very rare
1815 Description: MDT cannot be unmounted, reporting "Mount still busy"
1816 Details : Mountpoint references were being leaked during open reply
1817 reconstruction after an MDS restart. Drop mountpoint reference
1818 in reconstruct_open() and free dentry reference also.
1823 Description: wait until IO finished before start new when do lock cancel.
1824 Details : VM protocol want old IO finished before start new, in this case
1825 need wait until PG_writeback is cleared until check dirty flag
1826 and call writepages in lock cancel callback.
1831 Description: mds_mfd_close() ASSERTION(rc == 0)
1832 Details : In mds_mfd_close(), we need protect inode's writecount change
1833 within its orphan write semaphore to prevent possible races.
1836 Frequency : rare, on shutdown ost
1838 Description: don't hit live lock with umount ost.
1839 Details : shrink_dcache_parent can be in long loop with destroy dentries,
1840 use shrink_dcache_sb instead.
1843 Frequency : only when echo_client is used
1845 Description: don't panic with use echo_client
1846 Details : echo client pass NULL as client nid pointer and this produce NULL
1847 pointer dereference.
1850 Frequency : Always on 32-bit PowerPC systems
1852 Description: fix build on PPC32
1853 Details : compile code with -m64 flag produce wrong object file for PPC32.
1858 Description: MDS LBUG: ASSERTION(!IS_ERR(dchild))
1859 Details : In reconstruct_* functions, LASSERTs on both the data supplied
1860 by a client, and the data on disk are dangerous and incorrect.
1861 Change them with client eviction.
1863 Severity : enhancement
1865 Description: skiplist implementation simplification
1866 Details : skiplists are used to group compatible locks on granted list
1867 that was implemented as tracking first and last lock of each lock
1868 group the patch changes that to using doubly linked lists
1872 Description: delete compatibility for 32bit qdata
1873 Details : as planned, when lustre is beyond b1_8, lquota won't support 32bit
1874 qunit. That means servers of b1_4 and servers of b1_8 can't be
1875 used together if users want to use quota.
1878 Frequency : only with administrator action
1880 Description: mount failure if config log has invalid conf_param setting
1881 Details : If administrator specified an incorrect configuration parameter
1882 with "lctl conf_param" this would cause an error during future
1883 client mounts. Instead, ignore the bad configuration parameter.
1886 Frequency : blocks per group < blocksize*8 and uninit_groups is enabled
1888 Description: ldiskfs error: XXX blocks in bitmap, YYY in gd
1889 Details : If blocks per group is less than blocksize*8, set rest of the
1893 Frequency : Application do stride read on lustre
1895 Description: The read performance will drop a lot if the application does
1897 Details : Because the stride_start_offset are missing in stride read-ahead,
1898 it will cause clients read a lot of unused pages in read-ahead,
1899 then the read-performance drops.
1903 Description: more ldlm soft lockups
1904 Details : In ldlm_resource_add_lock(), call to ldlm_resource_dump()
1905 starve other threads from the resource lock for a long time in
1906 case of long waiting queue, so change the debug level from
1907 D_OTHER to the less frequently used D_INFO.
1909 Severity : enhancement
1911 Description: add -gid, -group, -uid, -user options to lfs find
1913 Severity : enhancement
1915 Description: ll_recover_lost_found_objs - recover objects in lost+found
1916 Details : OST corruption and subsequent e2fsck can leave objects in the
1917 lost+found directory. Using the "ll_recover_lost_found_objs"
1918 tool, these objects can be retrieved and data can be salvaged
1919 by using the object ID saved in the fid EA on each object.
1924 Description: this bug _only_ happens when inode quota limitation is very low
1925 (less than 12), so that inode quota unit is 1 at initialization.
1926 Details : if remaining quota equates 1, it is a sign to demonstate that quota
1927 is effective now. So least quota qunit should be 2.
1931 Description: Hung threads in invalidate_inode_pages2_range
1932 Details : The direct IO path doesn't call check_rpcs to submit a new RPC once
1933 one is completed. As a result, some RPCs are stuck in the queue
1938 Description: Procfs and llog threads access destoryed import sometimes.
1939 Details : Sync the import destoryed process with procfs and llog threads by
1940 the import refcount and semaphore.
1944 Description: mds fails to respond, threads stuck in ldlm_completion_ast
1945 Details : Sort source/child resource pair after updating child resource.
1950 Description: kernel BUG at ldiskfs2_ext_new_extent_cb
1951 Details : If insertion of an extent fails, then discard the inode
1952 preallocation and free data blocks else it can lead to duplicate
1957 Description: don't always update ctime in ext3_xattr_set_handle()
1958 Details : Current xattr code updates inode ctime in ext3_xattr_set_handle()
1959 In some cases the ctime should not be updated, for example for
1960 2.0->1.8 compatibility it is necessary to delete an xattr and it
1961 should not update the ctime.
1965 Description: add quota statistics
1966 Details : 1. sort out quota proc entries and proc code.
1967 2. add quota statistics
1972 Description: quotas are not honored with O_DIRECT
1973 Details : all writes with the flag O_DIRECT will use grants which leads to
1974 this problem. Now using OBD_BRW_SYNC to guard this.
1978 Bugzilla : 15713/16362
1979 Description: Assertion in iopen_connect_dentry in 1.6.3
1980 Details : looking up an inode via iopen with the wrong generation number can
1981 populate the dcache with a disconneced dentry while the inode
1982 number is in the process of being reallocated. This causes an
1983 assertion failure in iopen since the inode's dentry list contains
1984 both a connected and disconnected dentry.
1988 Description: assertion failure in ldlm_handle2lock()
1989 Details : fix a race between class_handle_unhash() and class_handle2object()
1990 introduced in lustre 1.6.5 by bug 13622.
1992 Severity : enhancement
1994 Description: superblock lock contention with many SMP cores on one client
1995 Details : several client filesystem locks were highly contended on SMP
1996 NUMA systems with 8 or more cores. Per-CPU datastructures
1997 and more efficient locking implemented to reduce contention.
2002 Description: Kernel BUG: sd_iostats_bump: unexpected disk index
2003 Details : remove the limit of 256 scsi disks in the sd_iostat patch
2008 Description: oops in sd_iostats_seq_show()
2009 Details : unloading/reloading the scsi low level driver triggers a kernel
2010 bug when trying to access the sd iostat file.
2015 Description: Kernel panics during QLogic driver reload
2016 Details : REQ_BLOCK_PC requests are not handled properly in the sd iostat
2017 patch, causing memory corruption.
2022 Description: journal_dev option does not work in b1_6
2023 Details : pass mount option during pre-mount.
2025 Severity : enhancement
2027 Description: Add a FIEMAP(FIle Extent MAP) ioctl for ldiskfs
2028 Details : FIEMAP ioctl will allow an application to efficiently fetch the
2029 extent information of a file. It can be used to map logical blocks
2030 in a file to physical blocks in the block device.
2033 Frequency : only with adaptive timeout enabled
2035 Description: DEBUG_REQ() bad paging request
2036 Details : ptlrpc_at_recv_early_reply() should not modify req->rq_repmsg
2037 because it can be accessed by reply_in_callback() without the
2041 Frequency : only on Cray X2
2043 Description: X2 build failures
2044 Details : fix build failures on Cray X2.
2048 Description: xid & resent requests
2049 Details : Initialize RPC XID from clock at startup (randomly if clock is
2054 Description: quota recovery deadlock during mds failover
2055 Details : This patch includes att18982, att18236, att18237 in bz14840.
2057 1. fix osts hang when mds does failover with quotaon
2058 2. prevent watchdog storm when osts threads wait for the
2063 Description: kernel panic on racer
2064 Details : Do not access dchild->d_inode when IS_ERR(dchild) is true.
2066 Severity : enhancement
2068 Description: Add lustre_start utility to start or stop multiple Lustre servers
2073 Description: Lustre GPF in {:ptlrpc:ptlrpc_server_free_request+373}
2074 Details : In case of memory pressure, list_del() can be called twice on
2075 req->rq_history_list, causing a kernel oops.
2079 Description: (ptllnd_peer.c:557:kptllnd_peer_check_sends()) ASSERTION(!in_interrupt()) failed
2080 Details : fix stack overflow in the distributed lock manager by defering
2081 export eviction after a failed ast to the elt thread instead of
2082 handling it in the dlm interpret routine.
2084 Severity : enhancement
2086 Description: More exported tunables for mballoc
2087 Details : Add support for tunable preallocation window and new tunables for
2088 large/small requests
2092 Description: Detect corruption of block bitmap and checking for preallocations
2093 Details : Checks validity of on-disk block bitmap. Also it does better
2094 checking of number of applied preallocations. When corruption is
2095 found, it turns filesystem readonly to prevent further corruptions.
2099 Frequency : only for big-endian servers
2100 Description: Check for big-endian system while mounting fs with extents feature
2101 Details : Mounting a filesystem with extents feature will fail on big-endian
2102 systems since ext3-based ldiskfs is not supported on big-endian
2103 systems. Can be over-ridden with "bigendian_extents" mount option.
2107 Description: Excessive recovery window
2108 Details : With AT enabled, the recovery window can be excessively long (6000+
2109 seconds). To address this problem, we no longer use
2110 OBD_RECOVERY_FACTOR when extending the recovery window (the connect
2111 timeout no longer depends on the service time, it is set to
2112 INITIAL_CONNECT_TIMEOUT now) and clients report the old service
2113 time via pb_service_time.
2116 Descriptoin: Don't sync journal after every i/o
2117 Details : Implement write RPC replay to allow server replies for write RPCs
2118 before data is on disk. However, this feature is disabled by
2119 default since some issues leading to data corruptions have been
2120 found during recovery (e.g. bug 19128). This feature can be enabled
2121 by running the following command on the OSSs:
2122 lctl set_param obdfilter.*.sync_journal=0
2126 Description: Watchdog triggered on MDS failover
2127 Details : enable OBD_CONNECT_MDT flag when connecting from the MDS so that
2128 the OSTs know that the MDS "UUID" can be reused for the same export
2129 from a different NID, so we do not need to wait for the export to
2134 Descriptoin: Lustre detected file system corruption with inode out of bounds
2135 Details : don't update i_size on MDS_CLOSE for directories. This causes
2136 directory corruptions on the MDT.
2138 -------------------------------------------------------------------------------
2140 2008-05-26 Sun Microsystems, Inc.
2142 * Support for kernels:
2143 2.6.5-7.311 (SLES 9),
2144 2.6.9-67.0.7.EL (RHEL 4),
2145 2.6.16.54-0.2.5 (SLES 10),
2146 2.6.18-53.1.14.el5 (RHEL 5),
2147 2.6.22.14 vanilla (kernel.org)
2148 * Client support for unpatched kernels:
2149 (see http://wiki.lustre.org/index.php?title=Patchless_Client)
2150 2.6.16 - 2.6.22 vanilla (kernel.org)
2151 * Due to problems with nested symlinks and FMODE_EXEC (bug 12652),
2152 we do not recommend using patchless RHEL4 clients with kernels
2153 prior to 2.6.9-55EL (RHEL4U5).
2154 * Recommended e2fsprogs version: 1.40.7-sun1
2155 * Note that reiserfs quotas are disabled on SLES 10 in this kernel.
2156 * RHEL 4 and RHEL 5/SLES 10 clients behaves differently on 'cd' to a
2157 removed cwd "./" (refer to Bugzilla 14399).
2158 * A new quota file format has been introduced in 1.6.5.
2159 The format conversion from prior releases is handled transparently,
2160 but releases older than 1.4.12/1.6.5 will not understand this new
2161 format. The automatic format conversion can be avoided by running
2162 the following command on the MDS before upgrading:
2163 'tunefs.lustre --param="mdt.quota_type=ug1" $MDTDEV'.
2164 For more information, please refer to bugzilla 13904.
2168 Description: quota performance fix
2169 Details : quota data is written in journalled mode instead of ordered to
2170 increase performance
2174 Description: lfs support for human-readable quota grace time strings
2175 Details : lfs setquota -t and lfs quota -t represent quota grace times
2176 in "XXwXXdXXhXXmXXs" format instead of large values in seconds
2179 Frequency : always with o2ib 1.3 and sles10
2181 Description: fix build with SLES10 and o2ib v3.
2182 Details : sles10 uses diffrent name for Module.symver file but configure
2183 assume this file has same name on RHEL/SLES/vanila kernels.
2186 Frequency : very rare, if additional xattrs are used on kernels >= 2.6.12
2188 Description: MDS may lose file striping (and hence file data) in some cases
2189 Details : If there are additional extended attributes stored on the MDS,
2190 in particular ACLs, SELinux, or user attributes (if user_xattr
2191 is specified for the client mount options) then there is a risk
2192 of attribute loss. Additionally, the Lustre file striping
2193 needs to be larger than default (e.g. striped over all OSTs),
2194 and an additional attribute must be stored initially in the
2195 inode and then increase in size enough to be moved to the
2196 external attribute block (e.g. ACL growing in size) for file
2199 Severity : enhancement
2201 Description: add message levels for liblustreapi
2206 Description: MDT cannot be unmounted, reporting "Mount still busy"
2207 Details : Mountpoint references were being leaked during open reply
2208 reconstruction after an MDS restart. Drop mountpoint reference
2209 in reconstruct_open() and free dentry reference also.
2214 Description: fix for occasional failure case of -ENOSPC in recovery-small tests
2215 Details : Move the 'good_osts' check before the 'total_bavail' check. This
2216 will result in an -EAGAIN and in the exit call path we call
2217 alloc_rr() which will with increasing aggressiveness attempt to
2218 aquire precreated objects on the minimum number of required OSCs.
2222 Description: Use old size assignment to avoid deadlock
2223 Details : This reverts the changes in bugs 2369 and bug 14138 that introduced
2224 the scheduling while holding a spinlock. We do not need locking
2225 for size in ll_update_inode() because size is only updated from
2226 the MDS for directories or files without objects, so there is no
2227 other place to do the update, and concurrent access to such inodes
2228 are protected by the inode lock.
2232 Description: Use __u64 instead of int for valid bits
2236 Description: resolve "_IOWR redefined" build error on SLES10
2240 Description: dump the memory debugging after all modules are unloaded to
2241 suppress false negative in conf_sanity test 39
2245 Description: the recovery timer never expires
2246 Details : for new client connect request, the recovery timer should not be
2247 reset, otherwise recovery timer will never expired, if the old
2248 client never come. Only old client connect and first connection
2249 req should trigger recovery timer reset.
2253 Description: the min numbers of lproc stats are wrong
2254 Details : adding a new constant LC_MIN_INIT and use it for initialization
2258 Frequency : always with interactive lfs
2260 Description: Reinitialize optind to 0 so that interactive lfs works in all cases
2263 Frequency : with multiple concurrent readdir processes in same directory
2264 Bugzilla : 15406, 15169, 15175
2265 Description: misc fixes for directory readahead.
2266 Details : prevent previous statahead async RPC callback from processing the
2267 current "statahead_info", race condition between async RPC callback
2268 add dentry into dentry hash table and "ls" thread revalidate such
2269 dentry, statahead his/miss control for hidden items, and so on.
2271 Severity : enhancement
2273 Description: build kernel-ib packages for OFED 1.3 in our release cycle
2277 Description: incore types cleaning in quota code (with respect to 64-bit limits)
2278 Details : several u32 variables declarations are replaced with u64 declarations
2283 Description: fix SLES kernel versioning
2284 Details : the kernel version for our SLES 10 kernel did not include a "-"
2285 before the "smp" at the end. while this was not a problem in
2286 general, it did mean that software trying to use the kernel
2287 version to try to detect a vendor specific kernel would fail.
2288 this was most evident by the OFED build scripts.
2293 Description: Don't update lov_desc members until making sure they are valid
2294 Details : When updating lov_desc members via proc fs, need fix their
2295 validities before doing the real update.
2298 Frequency : very rare
2300 Description: don't put request into delay list while invalidate in flight.
2301 Details : ptlrpc_delay_request sometimes put in delay list while invalidate
2302 import in flight. this produce timeout for invalidate and sometimes
2303 can cause stale data.
2305 Severity : enhancement
2307 Description: Update kernel to SLES9 2.6.5-7.311.
2309 Severity : enhancement
2311 Description: Update kernel to RHEL4 2.6.9-67.0.7.
2316 Frequency : on PPC only
2317 Description: not convert ost objects for directory because it's not exist.
2318 Details : ll_dir_getstripe assume dirrectory has ost objects but this wrong.
2320 Severity : enhancement
2322 Description: Fix warnings with compile liblustre at sles10/rhel5 which have
2323 __u64 as usingied long long type.
2326 Frequency : rare, on shutdown
2328 Description: race process ast vs remove callback
2329 Details : removing callback before disconnect import open race with
2330 processing callback.
2332 Severity : enhancement
2334 Description: Update kernel to SLES9 2.6.5-7.311.
2336 Severity : enhancement
2338 Description: Files open for execute are not marked busy on SLES10
2339 Details : Add FMODE_EXEC to SLES10 SP1 server kernel series.
2341 Severity : enhancement
2343 Description: Add server support for vanilla-2.6.22.14.
2346 Frequency : occasional
2348 Description: Avoid lov_create() getting stuck in obd_statfs_rqset()
2349 Details : If an OST is down the MDS will hang indefinitely in
2350 obd_statfs_rqset() waiting for the statfs data. While for
2351 MDS QOS usage of statfs, it should not stuck in waiting.
2353 Severity : enhancement
2355 Description: Disable adaptive timeouts by default
2358 Frequency : on network error
2360 Description: panic with double free request if network error
2361 Details : mdc_finish_enqueue is finish request if any network error occuring,
2362 but it's true only for synchronus enqueue, for async enqueue
2363 (via ptlrpcd) this incorrect and ptlrpcd want finish request
2367 Frequency : rare, on recovery
2369 Description: read procfs can produce deadlock in some situation
2370 Details : Holding lprocfs lock with send rpc can produce block for destroy
2371 obd objects and this also block reconnect with -EALREADY.
2372 This isn't fix all lprocfs bugs - but make it rare.
2374 Severity : enhancement
2376 Description: Update kernel to RHEL5 2.6.18-53.1.14.el5.
2379 Frequency : frequent on X2 node
2381 Description: mdc_set_open_replay_data LBUG
2382 Details : Set replay data for requests that are eligible for replay.
2387 Description: lustre_mgs: operation 101 on unconnected MGS
2388 Details : When MGC is disconnected from MGS long enough, MGS will evict the
2389 MGC, and late on MGC cannot successfully connect to MGS and a lot
2390 of the error messages complaining that MGS is not connected.
2393 Frequency : rare, depends on device drivers and load
2395 Description: MDS or OSS nodes crash due to stack overflow
2396 Details : Code changes in 1.6.4 increased the stack usage of some functions.
2397 In some cases, in conjunction with device drivers that use a lot
2398 of stack, the MDS (or possibly OSS) service threads could overflow
2399 the stack. One change which was identified to consume additional
2400 stack has been reworked to avoid the extra stack usage.
2402 Severity : enhancement
2404 Description: Update to RHEL5 latest kernel-2.6.18-53.1.13.el5.
2406 Severity : enhancement
2408 Description: Update to SLES10 SP1 latest kernel-2.6.16.54-0.2.5.
2410 Severity : enhancement
2412 Description: Update to RHEL5 latest kernel-2.6.18-53.1.6.el5.
2414 Serverity : enhancement
2416 Description: Update RHEL4 kernel to 2.6.9-67.0.4.
2419 Frequency : rare on shutdown OST
2421 Description: Don't allow skipping OSTs if index has been specified.
2422 Details : Don't allow skipping OSTs if index has been specified, make locking
2423 in internal create lots better.
2428 Description: ASSERTION(!PageDirty(page)) failed
2429 Details : Wrong check could lead to an assertion failure under specific
2435 Description: LBUG in ptlrpc_check_set() bad phase ebc0de00
2436 Details : access to bitfield in structure is always rounded to long
2437 and this produce problem with not atomic change any bit.
2442 Description: Lustre make rpms failed.
2443 Details : Remove ldiskfs spec file to avoids rpmbuild be confused when
2444 builds Lustre rpms from tarball.
2446 Severity : enhancement
2448 Description: Update to SLES9 SP4 kernel-2.6.5-7.308.
2451 Frequency : rare on shutdown OST
2453 Description: If llog cancel was not send before clean_exports phase, this can
2454 produce deadlock in llog code.
2455 Details : If llog thread has last reference to obd and call class_import_put
2456 this produce deadlock because llog_cleanup_commit_master wait when
2457 last llog_commit_thread exited, but this never success because was
2458 called from llog_commit_thread.
2461 Frequency : only if OST index is skipped
2463 Description: NULL lov_tgts causing MDS oops
2464 Details : more safe checks for NULL lov_tgts for avoid oops.
2466 Severity : enhancement
2468 Description: Update to RHEL4 latest kernel-2.6.9-67.0.1.EL.
2470 Severity : enhancement
2472 Description: Update to RHEL5 latest kernel-2.6.18-53.1.4.el5.
2477 Description: make mgs_setparam() handle fsname containing dash
2478 Details : fsname containing a dash does not work with lctl conf_param
2480 Severity : enhancement
2482 Description: Update to RHEL4 Update-6 kernel-2.6.9-67.EL.
2485 Frequency : rare, in recovery and (or) destroy lovobjid file.
2487 Description: rewrite lov objid code.
2488 Details : Cleanup for lov objid code, remove scability problems and wrong
2489 locking. Fix sending last_id into ost.
2491 Severity : enhancement
2493 Description: Update to SLES10 SP1 latest kernel-2.6.16.54-0.2.3.
2495 Severity : enhancement
2497 Description: Update to RHEL5 Update-1 kernel 2.6.18-53.el5.
2498 Details : Use d_move_locked instead of __d_move.
2501 Frequency : rare, at shutdown
2503 Description: access already free / zero obd_namespace.
2504 Details : if client_disconnect_export was called without force flag set,
2505 and exist connect request in flight, this can produce access to
2506 NULL pointer (or already free pointer) when connect_interpret
2507 store ocd flags in obd_namespace.
2510 Frequency : only at startup
2512 Description: not alloc memory with spinlock held.
2513 Details : allocation memory with GFP_KERNEL can produce sleep deadlock,
2514 if any spinlock held.
2519 Description: lfs find does not continue on file error
2520 Details : Continue other files processing when a file/dir is absent.
2524 Description: Inconsistent usage of lustre_pack_reply()
2525 Details : Standardize the usage of lustre_pack_reply() such that it
2526 always generate a CERROR on failure.
2529 Frequency : very rare
2531 Description: Fix replay if there is an un-replied request and open
2532 Details : In some cases, older replay request will revert the
2533 mcd->mcd_last_xid on MDS which is used to record the client's
2534 latest sent request.
2536 Severity : enhancement
2538 Description: Update to RHEL5 kernel 2.6.18-8.1.15.el5.
2540 Severity : enhancement
2542 Description: Update to SLES10 SP1 kernel 2.6.16.53-0.16
2544 Severity : enhancement
2546 Description: Update to SLES9 kernel-2.6.5-7.287.3.
2548 Severity : enhancement
2550 Description: Update to RHEL4 kernel-2.6.9-55.0.12.EL.
2552 Severity : enhancement
2554 Description: Build SLES10 patchless client fails
2555 Details : The configure was broken by run ./configure with
2556 --with-linux-obj=.... argument for patchless client. When the
2557 configure use --with-linux-obj, the LINUXINCLUDE= -Iinclude
2558 can't search header adequately. Use absolute path such as
2559 -I($LINUX)/include instead.
2561 Severity : enhancement
2563 Description: Lustre Page Accounting
2564 Details : New macros for page alloc and free which enable accounting
2565 of page allocation of Lustre. Use percpu counters to store memory
2566 and page statistics.
2569 Frequency : only if debugging is disabled
2571 Description: LASSERT_{REQ,REP}SWAB macros are buggy
2572 Details : If SWAB_PARANOIA is disabled, the LASSERT_REQSWAB and
2573 LASSERT_REPSWAB macros become no-ops, which is incorrect. Drop
2574 these macros and replace them with their definitions instead.
2579 Description: interrupt oig_wait produce painc on resend.
2580 Details : brw_redo_request can be used for resend requests from ptlrpcd and
2581 private set, and this produce situation when rq_ptlrpcd_data not
2582 copyed to new allocated request and triggered LBUG on assert
2583 req->rq_ptlrpcd_data != NULL. But this member used only for wakeup
2584 ptlrpcd set if request is changed and can be safety changed to use
2587 Severity : enhancement
2589 Description: organize the server-side client stats on per-nid basis
2590 Details : Change the structure of stats under obdfilter and mds to
2598 The "uuid"s file would list the uuids of _active_ exports.
2599 And the clear entry is to clear all stats and stale nids.
2604 Description: Processes looping in ll_readdir() on Lustre clients finally causing
2605 a full node pseudo-hang
2606 Details : Concurrent access to the same directory from multiple clients with
2607 intensive file creation/removal can cause a client node to spin in
2608 ll_readdir(). i_version must be increased every time the lock is
2609 cancelled to ensure a revalidate is done.
2614 Description: touch file failed when fs is not full
2615 Details : OST in recovery should not be discarded by MDS in alloc_qos(),
2616 otherwise we can get ENOSP while fs is not full.
2619 Frequency : only for Cray XT3
2620 Bugzilla : 12829/13455
2621 Description: Changing primary group doesn't change the group lustre assigns to
2623 Details : When CRAY_XT3 is defined, the fsgid supplied by the client is
2624 overridden with the primary group provided by the group upcall,
2625 whereas the supplied fsgid can be trusted if it is in the list of
2626 supplementary groups returned by the group upcall.
2628 Severity : enhancement
2630 Description: Root Squash Functionality
2631 Details : Implementation of NFS-like root squash capability. Specifically,
2632 don't allow someone with root access on a client node to be able
2633 to manipulate files owned by root on a server node.
2635 Severity : enhancement
2637 Description: Slow trucate/writes to huge files at high offsets.
2638 Details : Directly associate cached pages to lock that protect those pages,
2639 this allows us to quickly find what pages to write and remove
2640 once lock callback is received.
2645 Description: Too many locks accumulating on client during NFS usage
2646 Details : mds_open improperly used accmode to find out access mode to a
2647 file. Also mdc_intent_lock logic to find out if we already have
2648 lock similar to just received was flawed since introduction of
2649 skiplists - locks are now added to the front of the granted
2654 Description: Hit ASSERTION(obd->obd_stopping == 1) failed in some setup failed
2656 Details : In obd setup failure handler, obd_stopping will not necessarily to
2657 be 1, and obd_set_up should also be checked to make sure whether
2658 obd is completely setup.
2660 Severity : enhancement
2662 Description: Allow masking D_WARNING, D_ERROR messages from console
2663 Details : Console messages can now be disabled via lnet.printk.
2668 Description: User code with malformed file open parameter crashes client node
2669 Details : Before packing join_file req, all the related reference should be
2670 checked carefully in case some malformed flags cause fake join_file
2676 Description: shrink/enlarge qunit size when needed; fix the problem of coarse
2677 grain of quota doing harm to quota's accuracy
2678 Details : qunit size will be changed when quota limitation is too low/high;
2679 record the pending quota write in order to get more accureate
2680 quota; delete the patch for bug12588, which is unnecessary when
2681 this patch is landed. This bug also contains fixes for bug 14526,
2682 14299, 14601 and 13794.
2686 Description: LDLM_ENQUEUE races with LDLM_CP_CALLBACK
2687 Details : ldlm_completion_ast() assumes that a lock is granted when the req
2688 mode is equal to the granted mode. However, it should also check
2689 that LDLM_FL_CP_REQD is not set.
2693 Description: Heavy nfs access might result in deadlocks
2694 Details : After ELC code landed, it is now improper to enqueue any mds
2695 locks under och_sem, because enqueue might want to decide to
2696 cancel open locks for same inode we are holding och_sem for.
2700 Description: 35% write performance drop with ldiskfs2 when quotas are on
2701 Details : Enable ext3 journalled quota by default to improve performance
2702 when quotas are turned on.
2706 Description: Client eviction while running blogbench
2707 Details : A lot of unlink operations with concurrent I/O can lead to a
2708 deadlock causing evictions. To address the problem, the number of
2709 oustanding OST_DESTROY requests is now throttled to
2710 max_rpcs_in_flight per OSC and LDLM_FL_DISCARD_DATA blocking
2711 callbacks are processed in priority.
2714 Frequency : RHEL4 only
2716 Description: mkfs is very slow on IA64/RHEL4
2717 Details : A performance regression has been discovered in the MPT Fusion
2718 driver between versions 3.02.73rh and 3.02.99.00rh. As a
2719 consequence, we have downgraded the MPT Fusion driver in the RHEL4
2720 kernel from 3.02.99.00 to 3.02.73 until this problem is fixed.
2723 Frequency : PPC/PPC64 only
2725 Description: conflicts between asm-ppc64/types.h and lustre_types.h
2726 Details : fix duplicated definitions between asm-ppc64/types.h and
2727 lustre_types.h on PPC.
2730 Frequency : PPC/PPC64 only
2732 Description: asm-ppc/segment.h does not exist
2733 Details : fix compile issue on PPC.
2737 Description: data checksumming impacts single node performance
2738 Details : add support for several checksum algorithms. Currently, CRC32 and
2739 Adler-32 are supported. The checksum type can be changed on the fly
2740 through /proc/fs/lustre/osc/*/checksum_type.
2744 Description: use adler32 for page checksums
2745 Details : when available, use the Adler-32 algorithm instead of CRC32 for
2750 Description: better handle error messages in extents code
2752 Severity : enhancement
2754 Description: SNMP support enhancement
2755 Details : Adding total number of sampled request for an MDS node in snmp
2758 Severity : enhancement
2760 Description: Optimize ldlm waiting list processing for PR extent locks
2761 Details : When processing waiting list for read extent lock and meeting read
2762 lock that is same or wider to it that is not contended, skip
2763 processing rest of the list and immediatelly return current
2764 status of conflictness, since we are guaranteed there are no
2765 conflicting locks in the rest of the list.
2769 Description: Time out and refuse to reconnect
2770 Details : When the failover node is the primary node, it is possible
2771 to have two identical connections in imp_conn_list. We must
2772 compare not conn's pointers but NIDs, otherwise we
2773 can defeat connection throttling.
2777 Description: Client not clear own cache if answer to reconnect is lost.
2778 Details : Client gets evicted from server. Now client also thinks it is
2779 disconnected (or gets ENOTCONN on its operation) and decides to
2780 reconnect. Server receives reconnect message, but cannot find
2781 export. New export is created that is fully valid (new cookie!),
2782 but reply is lost and not reported to client. Client reconnects
2783 again and gets back a just-created connection, but it is not new
2784 so client thinks it was not evicted and does not do recovery.
2788 Description: Detect stride IO mode in read-ahead
2789 Details : When a client does stride read, read-ahead should detect that and
2790 read-ahead pages according to the detected stride pattern.
2794 Description: build for x2 fails
2795 Details : fix compile issue on Cray systems.
2797 Severity : enhancement
2799 Description: implement readv/writev APIs(aio_read/aio_writes in newer kernels)
2800 Details : This greatly improves speed of NFS writes on 2.6 kernels.
2803 Frequency : only on PPC/SLES10
2805 Description: "BITS_PER_LONG is not 32 or 64" in linux/idr.h
2806 Details : On SLES10/PPC, fs.h includes idr.h which requires BITS_PER_LONG to
2807 be defined. Add a hack in mkfs_lustre.c to work around this compile
2812 Description: LASSERT on MDS when client holding flock lock dies
2813 Details : ldlm pool logic depends on number of granted locks equal to
2814 number of released locks which is not true for flock locks, so
2815 just exclude such locks from consideration.
2819 Description: MDS deadlock with many ll_sync_lov threads and I/O stalled
2820 Details : Use fsfilt_sync() for both the whole filesystem sync and
2821 individual file sync to eliminate dangerous inode locking
2822 with I_LOCK that can lead to a deadlock.
2826 Description: Update an obsolete wirecheck.c generator
2827 Details : Update wirecheck.c/wirehdr.c and regenerate wiretest.c
2831 Description: Client can panic on open sometimes
2832 Details : It is possible that we try to free already freed request in
2833 ll_file_open in some error cases when we send request from
2838 Description: performance in 1.6.3
2839 Details : Force q->max_phys_segments to MAX_PHYS_SEGMENTS on SLES10 to be
2840 sure that 1MB requests are not fragmented by the block layer.
2844 Description: LDLM soft lockups - improvement
2845 Details : It is be possible to send the lock handle along with each read
2846 or write request because the client is already doing a lock match
2847 itself so there isn't any reason the OST should have to re-do that
2853 Description: lfs quota fails with deactivated OSTS
2854 Details : With this patch, three improvements are included:
2855 1. detete the softlimit in mds and osts when use "lfs quota".
2856 2. display the inaccurate data in the output of "lfs quota".
2857 3. try to get quota info when "lfs quota" is executed.
2862 Description: Extent locks not granted with no conflicts sometimes.
2863 Details : When race occurs in glimpse handler and nothing is returned,
2864 we do not reprocess the queue after lock cancel, and that leads
2865 to a stall until next activity on a resource
2868 Frequency : failover with quotaon
2870 Description: during mds failovers with quota on, OSTs got into deadlock state
2871 and causing dumpstack.
2872 Details : for every quota slave, at any time, there is only one quota req
2873 is sent to quota master for every uid/gid. Before that quota req
2874 returns, all the thread relative to the same uid/gid will wait.
2875 So if the quota req is lost because mds failovers or any other
2876 reasons, this bug will be hit. Now, dqacq_interpret() will handle
2877 quota reqs who time out.
2879 Severity : enhancement
2882 Description: when quota slave checks if quota is enough, there is an unnecessary
2884 Details : place this wait on necessary place instead of always waiting.
2886 --------------------------------------------------------------------------------
2888 2007-12-07 Cluster File Systems, Inc. <info@clusterfs.com>
2890 * Support for kernels:
2891 2.6.5-7.286 (SLES 9),
2892 2.6.9-55.0.9.EL (RHEL 4),
2893 2.6.16.53-0.8 (SLES 10),
2894 2.6.18-8.1.14.el5 (RHEL 5),
2895 2.6.18.8 vanilla (kernel.org)
2896 * Client support for unpatched kernels:
2897 (see http://wiki.lustre.org/index.php?title=Patchless_Client)
2898 2.6.16 - 2.6.22 vanilla (kernel.org)
2899 * Due to recently discovered recovery problems, we do not recommend
2900 using patchless RHEL 4 clients with this or any earlier release.
2901 * Recommended e2fsprogs version: 1.40.2-cfs1
2902 * Note that reiserfs quotas are disabled on SLES 10 in this kernel.
2905 Frequency : occasional
2907 Description: MDS hang or stay in waiting lock
2908 Details : If client receive lock with CBPENDING flag ldlm need send lock
2909 cancel as separate rpc, to avoid situation when cancel request
2910 can't processed due all i/o threads stay in wait lock.
2913 Frequency : occasional
2915 Description: Do not fail import if osc_interpret_create gets -EAGAIN
2916 Details : If osc_interpret_create got -EAGAIN it immediately exits and
2917 wakeup oscc_waitq. After wakeup oscc_wait_for_objects call
2918 oscc_has_objects and see OSC has no objests and call
2919 oscc_internal_create to resend create request.
2921 Severity : enhancement
2923 Description: Update kernel patches for SLES10 2.6.16.53-0.8.
2924 Details : Update which_patch & target file for SLES10 latest kernel.
2926 Severity : enhancement
2928 Description: add --type and --size parameters to lfs find
2929 Details : Enhance lfs find by adding filetype and filesize parameters. Also
2930 multiple OBDs can now be specified for the --obd option.
2932 Severity : enhancement
2934 Description: eliminate client locks in face of contention
2935 Details : file contention detection and lockless i/o implementation
2936 for contended files.
2938 Severity : enhancement
2940 Description: Remove client patches from SLES 10 kernel.
2941 Details : This causes SLES 10 clients to behave as patchless clients
2942 even on a Lustre-patched (server) kernel.
2944 Severity : enhancement
2946 Description: use i_size_read and i_size_write in 2.6 port
2947 Details : replace inode->i_size access with i_size_read/write()
2950 Frequency : when removing large files
2952 Description: scheduling issue during removal of large Lustre files
2953 Details : Don't take the BKL in fsfilt_ext3_setattr() for 2.6 kernels.
2954 It causes scheduling issues when removing large files (17TB in the
2960 Description: 1.4.11 Can't handle directories with stripe set and extended ACLs
2961 Details : Impossible (EPROTO is returned) to access a directory that has a
2962 non-default striping and ACLs.
2965 Frequency : only on ppc
2967 Description: /proc/fs/lustre/devices broken on ppc
2968 Details : The patch as applied to 1.6.2 doesn't look correct for all arches.
2969 We should make sure the type of 'index' is loff_t and then cast
2970 explicitly as needed below. Do not assign an explicitly cast
2974 Frequency : only for rhel5
2976 Description: Kernel patches update for RHEL5 2.6.18-8.1.10.el5.
2977 Details : Modify the target file & which_kernel.
2980 Frequency : if the uninit_groups feature is enabled on ldiskfs
2982 Description: e2fsck reports "invalid unused inodes count"
2983 Details : If a new ldiskfs filesystem is created with the "uninit_groups"
2984 feature and only a single inode is created in a group then the
2985 "bg_unused_inodes" count is incorrectly updated. Creating a
2986 second inode in that group would update it correctly.
2991 Description: buffer overruns could theoretically occur
2992 Details : llapi_semantic_traverse() modifies the "path" argument by
2993 appending values to the end of the origin string, and a buffer
2994 overrun may occur. Adding buffer overrun check in liblustreapi.
2996 Severity : enhancement
2998 Description: Add jbd statistics patch for RHEL5 and 2.6.18-vanilla.
3001 Frequency : only if filesystem is inconsistent
3003 Description: handle "serious error: objid * already exists" more gracefully
3004 Details : If LAST_ID value on disk is smaller than the objects existing in
3005 the O/0/d* directories, it indicates disk corruption and causes an
3006 LBUG(). If the object is 0-length, then we should use the existing
3007 object. This will help to avoid a full fsck in most cases.
3009 Severity : enhancement
3011 Description: Kernel patches update for RHEL4 2.6.9-55.0.6.
3012 Details : Modify vm-tunables-rhel4.patch.
3014 Severity : enhancement
3016 Description: Kernel config for 2.6.18-vanilla.
3017 Details : Modify targets/2.6-vanilla.target.in.
3018 Add config file kernel-2.6.18-2.6-vanilla-i686.config.
3019 Add config file kernel-2.6.18-2.6-vanilla-i686-smp.config.
3020 Add config file kernel-2.6.18-2.6-vanilla-x86_64.config.
3021 Add config file kernel-2.6.18-2.6-vanilla-x86_64-smp.config.
3024 Frequency : occasional
3026 Description: improve handling recoverable errors
3027 Details : If request processed with error which can be recoverable on server
3028 request should be resend, otherwise page released from cache and
3034 Description: Kernel patches update for RHEL5 2.6.18-8.1.14.el5.
3035 Details : Modify target file & which_patch.
3036 A flaw was found in the IA32 system call emulation provided
3037 on AMD64 and Intel 64 platforms. An improperly validated 64-bit
3038 value could be stored in the %RAX register, which could trigger an
3039 out-of-bounds system call table access. An untrusted local user
3040 could exploit this flaw to run code in the kernel
3041 (ie a root privilege escalation). (CVE-2007-4573).
3045 Description: change order of libsysio includes
3046 Details : '#include sysio.h' should always come before '#include xtio.h'
3048 Severity : enhancement
3050 Description: adapt the lustre_config script to support the upgrade case
3051 Details : Add "-u" option for lustre_config script to support upgrading 1.4
3052 server targets to 1.6 in parallel.
3057 Description: To avoid grant space > avaible space when the disk is almost
3058 full. Without this patch you might see the error "grant XXXX >
3059 available" or some LBUG about grant, when the disk is almost
3061 Details : In filter_check_grant, for non_grant cache write, we should
3062 check the left space by if (*left > ungranted + bytes), instead
3063 of (*left > ungranted), because only we are sure the left space
3064 is enough for another "bytes", then the ungrant space should be
3065 increase. In client, we should update cl_avail_grant only there is
3066 OBD_MD_FLGRANT in the reply.
3071 Description: Update RHEL 4 kernel to fix local root privilege escalation.
3072 Details : Update to the latest RHEL 4 kernel to fix the vulnerability
3073 described in CVE-2007-4573. This problem could allow untrusted
3074 local users to gain root access.
3077 Frequency : when using O_DIRECT and quotas
3079 Description: Incorrect file ownership on O_DIRECT output files
3080 Details : block usage reported by 'lfs quota' does not take into account
3081 files that have been written with O_DIRECT.
3085 Description: (rw.c:1323:ll_read_ahead_pages()) ASSERTION(page_idx > ria->ria_stoff) failed
3086 Details : Once the unmatched stride IO mode is detected, shrink the stride-ahead
3087 window to 0. If it does hit cache miss, and read-pattern is still
3088 stride-io mode, does not reset the stride window, but also does not
3089 increase the stride window length in this case.
3091 --------------------------------------------------------------------------------
3093 2007-09-27 Cluster File Systems, Inc. <info@clusterfs.com>
3095 * Support for kernels:
3096 2.6.5-7.286 (SLES 9),
3097 2.6.9-55.0.2.EL (RHEL 4),
3098 2.6.16.46-0.14 (SLES 10),
3099 2.6.18-8.1.8.el5 (RHEL 5),
3100 2.6.18.8 vanilla (kernel.org)
3101 * Client support for unpatched kernels:
3102 (see http://wiki.lustre.org/index.php?title=Patchless_Client)
3103 2.6.16 - 2.6.21 vanilla (kernel.org)
3104 * Due to recently discovered recovery problems, we do not recommend
3105 using patchless RHEL 4 clients with this or any earlier release.
3106 * Recommended e2fsprogs version: 1.40.2-cfs1
3107 * Note that reiserfs quotas are disabled on SLES 10 in this kernel.
3111 Description: Fix errors in lfs documentation
3112 Details : Fixes man pages
3114 Severity : enhancement
3116 Description: Adaptive timeouts
3117 Details : RPC timeouts adapt to changing server load and network
3118 conditions to reduce resend attempts and improve recovery time.
3120 Severity : enhancement
3122 Description: llapi_file_create() does not allow some changes
3123 Details : add llapi_file_open() that allows specifying the file creation
3124 mode and open flags, and also returns an open file handle.
3126 Severity : enhancement
3128 Description: df doesn't work properly if diskfs blocksize != 4K
3129 Details : Choose biggest blocksize of OST's as the LOV's blocksize.
3132 Frequency : always on directories with default striping set
3134 Description: lfs find on -1 stripe looping in lsm_lmm_verify_common()
3135 Details : Avoid lov_verify_lmm_common() on directory with -1 stripe count.
3137 Severity : enhancement
3139 Description: merge and cleanup kernel patches.
3140 Details : Remove mnt_lustre_list in vfs_intent-2.6-rhel4.patch.
3142 Severity : enhancement
3144 Description: RedHat Update kernel for RHEL5
3145 Details : Kernel config file for RHEL5.
3147 Severity : enhancement
3149 Description: OSS needs mutliple precreate threads
3150 Details : Add ability to start more than one create thread per OSS.
3153 Frequency : only with quota on the root user
3155 Description: mds_obd_create error creating tmp object
3156 Details : When the user sets quota on root, llog will be affected and can't
3157 create files and write files.
3160 Frequency : Always on ia64 patchless client, and possibly others.
3162 Description: Add EXPORT_SYMBOL check for node_to_cpumask symbol.
3163 Details : This allows the patchless client to be loaded on architectures
3164 without this export.
3169 Description: disorder of journal start and llog_add cause deadlock.
3170 Details : in llog_origin_connect, journal start should happen before
3171 llog_add keep the same order as other functions to avoid
3174 Severity : enhancement
3176 Description: RedHat Update kernel for RHEL5
3177 Details : Modify the kernel config file more closer RHEL5.
3179 Severity : enhancement
3181 Description: Build failure against Centos5 (RHEL5)
3182 Details : Define PAGE_SIZE when it isn't present.
3185 Frequency : occasionally when using NFS
3187 Description: "ll_intent_file_open()) lock enqueue: err: -13" with nfs
3188 Details : with NFS, the anon dentry's parent was set to itself in
3189 d_alloc_anon(), so in MDS, we use rec->ur_fid1 to find the
3190 corresponding dentry other than use rec->ur_name.
3192 Severity : enhancement
3194 Description: enable data checksumming by default
3195 Details : enable checksum by default, allow --disable-checksum
3196 configure option and "-o nochecksum" mount option. Checksums
3197 can also be disabled at runtime via $LPROC/osc/*/checksum_pages.
3200 Frequency : Occasionally with failover
3202 Description: Client eviction due to failover config
3203 Details : after a connection loss, the lustre client should attempt to
3204 reconnect to the last active server first before trying the
3205 other potential connections.
3207 Severity : enhancement
3209 Description: client-side metadata stat-ahead during readdir(directory readahead)
3210 Details : perform client-side metadata stat-ahead when the client detects
3211 readdir and sequential stat of dir entries therein
3216 Description: when mds and osts use different quota unit(32bit and 64bit),
3217 quota will be released repeatly.
3218 Details : void sending multiple quota reqs to mds, which will keep the
3219 status between the reqs.
3222 Frequency : only with liblustre clients on XT3
3224 Description: evictions taking too long
3225 Details : allow llrd to evict clients directly on OSTs
3229 Description: osts not allocated evenly to files
3230 Details : change the condition to increase offset_idx
3232 Severity : enhancement
3234 Description: self-adjustable client's lru lists
3235 Details : use adaptive algorithm for managing client cached locks lru
3236 lists according to current server load, other client's work
3237 pattern, memory activities, etc. Both, server and client
3238 side namespaces provide number of proc tunables for controlling
3243 Description: rewrite ext2-derived code in llite/dir.c and obdclass/uuid.c
3244 Details : rewrite inherited code (uuid parsing code from ext2 utils and
3245 readdir code from ext3) from scratch preserving functionality.
3249 Description: Only those disconnect error should be returned by rq_status.
3250 Details : In open/enqueue processs, Some errors, which will cause client
3251 disconnected, should be returned by rq_status, while other
3252 errors should still be returned by intent, then mdc or llite will
3255 Severity : enhancement
3257 Description: Tune the kernel for good SCSI performance.
3258 Details : Set the value of /sys/block/{dev}/queue/max_sectors_kb
3259 to the value of /sys/block/{dev}/queue/max_hw_sectors_kb
3263 Frequency : Always for filesystems larger than 2TB on 32-bit systems.
3264 Bugzilla : 13547, 13627
3265 Description: Data corruption for OSTs that are formatted larger than 2TB
3267 Details : When generating the bio request for lustre file writes the
3268 sector number would overflow a temporary variable before being
3269 used for the IO. The data reads correctly from Lustre (which
3270 will overflow in a similar manner) but other file data or
3271 filesystem metadata may be corrupted in some cases.
3275 Description: "lfs find -obd UUID" prints directories
3276 Details : "lfs find -obd UUID" will return all directory names instead
3277 of just file names. It is incorrect because the directories
3278 do not reside on the OSTs.
3281 Bugzilla : 13596, 13614
3282 Description: MDS hang after unclean shutdown of lots of clients
3283 Details : Never resend AST requests.
3287 Frequency : Always, for kernels after 2.6.16
3288 Description: Fix warning idr_remove called for id=.. which is not allocated.
3289 Details : Last kernels save old s_dev before kill super and not allow
3290 to restore from callback - restore it before call kill_anon_super.
3294 Description: Fix error on 'ls .' at the top of the Lustre mount.
3295 Details : Don't revalidate dentry if it is a root dentry.
3297 --------------------------------------------------------------------------------
3299 2007-08-27 Cluster File Systems, Inc. <info@clusterfs.com>
3301 * Support for kernels:
3302 2.6.5-7.286 (SLES 9),
3303 2.6.9-55.0.2.EL (RHEL 4),
3304 2.6.16.46-0.14 (SLES 10),
3305 2.6.18-8.1.8.el5 (RHEL 5),
3306 2.6.18.8 vanilla (kernel.org)
3307 * Client support for unpatched kernels:
3308 (see http://wiki.lustre.org/index.php?title=Patchless_Client)
3309 2.6.16 - 2.6.22 vanilla (kernel.org)
3310 * Due to recently discovered recovery problems, we do not recommend
3311 using patchless RHEL 4 clients with this or any earlier release.
3312 * Recommended e2fsprogs version: 1.39.cfs8
3313 * Note that reiserfs quotas are disabled on SLES 10 in this kernel.
3318 Description: block reactivating mgc import until all deactivates complete
3319 Details : Fix race when failing back MDT/MGS to itself (testing)
3321 Severity : enhancement
3323 Description: lfs setstripe enhancement
3324 Details : Make lfs setstripe understand 'k', 'm' and 'g' for stripe size.
3327 Frequency : mds/oss recovery
3329 Description: llog ctxt is refrenced after it has been freed.
3330 Details : llog ctxt refcount was added to avoide the race between ctxt free
3331 and llog recovery process. Each llog user must hold ctxt refcount
3332 before it access the llog. And the llog ctxt can only be freed
3333 when its refcount is zero.
3335 Severity : enhancement
3337 Description: randomly memory allocation failure util
3338 Details : Make lustre randomly failed allocating memory for testing purpose.
3341 Frequency : only for SLES10
3343 Description: Update kernel patch for SLES10 SP1
3344 Details : Add patch blkdev_tunables-2.6-sles10.patch to 2.6-sles10.series.
3346 Severity : enhancement
3348 Description: omit set fsid for export NFS
3349 Details : fix set/restore device id for avoid EMFILE error and mark lustre fs
3350 as FS_REQUIRES_DEV for avoid problems with generate fsid.
3353 Frequency : after network failures
3355 Description: Add sync option to mount_lustre.c
3356 Details : Client loses data written to lustre after a network interruption.
3358 Severity : enhancement
3360 Description: Error message improvement.
3361 Details : Merging of two LCONSOLE_ERROR_MSG into one.
3363 Severity : enhancement
3365 Description: don't use GFP_* in generic Lustre code.
3366 Details : Use cfs_alloc_* functions and CFS_* flags for code portability.
3368 Severity : enhancement
3370 Description: obdclass is limited by single OBD_ALLOC(idarray)
3371 Details : replace OBD_ALLOC/OBD_FREE with OBD_VMALLOC/OBD_VFREE
3373 Severity : enhancement
3375 Description: updated patchess for new RHEL4 kernel
3376 Details : Fixed ext3-unlink-race.patch per Kalpak's comment.
3378 Severity : enhancement
3380 Description: warnings with build patchless client with vanila 2.6.19 and up
3381 Details : change old ctl_table style and replace ctl_table/ctl_table_header
3382 with cfs_sysctl_table_t/cfs_sysctl_table_header_t
3384 Severity : enhancement
3386 Description: O_DIRECT bypasses client statistics.
3387 Details : When running with O_DIRECT I/O, neither the client rpc_stats nor
3388 read_ahead_stats were updated. Copied stats section from
3389 osc_send_oap_rpc() into async_internal().
3392 Frequency : only for Cray XT3
3394 Description: peer credits not enough on many OST per OSS systems.
3395 Details : Use new lnet way to add credits as we need those for pings and ASTs
3398 Frequency : only with liblustre
3400 Description: Liblustre is not releasing flock locks on file close.
3401 Details : Release flock locks on file close.
3404 Frequency : only for RHEL4
3406 Description: Update kernel patches for kernel-2.6.9-55.0.2.EL
3407 Details : Remove inode-nr_unused-2.6.9-rhel4.patch from 2.6-rhel4.series
3408 Update target file and kernel config.
3412 Description: lustre support for RHEL5
3413 Details : Add support for RHEL5.
3418 Description: ASSERTION(export != NULL) failed in target_handle_connect
3419 Details : Assetion hit is result of rare race between disconnect and connet
3420 to same nid. target_handle_connect found old connect cockie and
3421 tried to reconnect, but can't find export for this cockie.
3426 Description: umount blocks forever on error
3427 Details : In result of wrong using obd_no_recov and obd_force flags client
3428 can hand if cancel or some other requests is lost.
3431 Frequency : Only for SLES
3433 Description: sanity_quota fail test_1
3434 Details : There are multiple occurences of $TSTUSR in SLES's /etc/group
3435 file, which makes TSTID[2] inunique.
3437 Severity : enhancement
3439 Description: Kernel patches for SLES9 2.6.5-7.286 kernel
3440 Details : Update target/ChangeLog/which_patch .
3442 Severity : enhancement
3444 Description: jbd statistics
3445 Details : Port older jbd statistics patch for sles10
3450 Description: lvbo_init failed for resource with missing objects.
3451 Details : Fix returning error if we do stat for file with missing/corrupted
3452 objects and i_size set to all sum of size of all avaible objects.
3453 if we to truncate/write to missing object - it is recreated.
3458 Description: Oops in read and write path when failing to allocate lock.
3459 Details : Check if lock allocation failed and return error back.
3462 Frequency : When flocks are used.
3464 Description: assertion failure in ldlm_cli_enquque_fini for non NULL lock.
3465 Details : Flock locks might destroy just granted lock if it could be merged
3466 with another existing flock, this is done in completion handler,
3467 so teach ldlm_cli_enquque_fini that this is a valid case for
3473 Description: reply_lock_interpret crash due to race with it and lock cancel.
3474 Details : Do not replay locks that are being cancelled. Do not reference
3475 locks by their address during replay, just by their handle.
3478 Frequency : only with deactivated OSTs
3480 Description: lstripe command fails for valid OST index
3481 Details : The stripe offset is compared to 'lov->desc.ld_tgt_count'
3482 instead of lov->desc.ld_active_tgt_count.
3484 Severity : enhancement
3486 Description: Build failure against Centos5 (RHEL5)
3487 Details : Use getpagesize() instead of PAGE_SIZE.
3489 --------------------------------------------------------------------------------
3491 2007-07-30 Cluster File Systems, Inc. <info@clusterfs.com>
3493 * Support for kernels:
3494 2.6.5-7.283 (SLES 9),
3495 2.6.9-55.EL (RHEL 4),
3496 2.6.16.46-0.14 (SLES 10),
3497 2.6.18.8 vanilla (kernel.org)
3498 * Client support for unpatched kernels:
3499 (see http://wiki.lustre.org/index.php?title=Patchless_Client)
3500 2.6.16 - 2.6.22 vanilla (kernel.org)
3501 * Due to recently discovered recovery problems, we do not recommend
3502 using patchless RHEL 4 clients with this or any earlier release.
3503 * Recommended e2fsprogs version: 1.39.cfs8
3504 * Note that reiserfs quotas are disabled on SLES 10 in this kernel.
3505 * Starting with this release, the ldiskfs backing filesystem required
3506 by Lustre is now in its own package, lustre-ldiskfs. This package
3507 should be installed. It is versioned separately from Lustre and
3508 may be released separately in future.
3513 Description: server LBUG when shutting down
3514 Details : Block umount forever until the mount refcount is zero rather
3515 than giving up after an arbitrary timeout.
3517 Severity : enhancement
3519 Description: add optional extra BUILD_VERSION info
3520 Details : add a new environment variable (namely LUSTRE_VERS) which allows
3521 to override the lustre version.
3524 Frequency : 2.6.18 servers only
3526 Description: ll_kern_mount() doesn't release the module reference
3527 Details : The ldiskfs module reference count never drops down to 0
3528 because ll_kern_mount() doesn't release the module reference.
3533 Description: server LBUG when using old ost_num_threads parameter
3534 Details : Accept the old ost_num_threads parameter but warn that it
3535 is deprecated, and fix an off-by-one error that caused an LBUG.
3540 Description: Transient SCSI error results in persistent IO issue
3541 Details : iobuf->dr_error is not reinitialized to 0 between two
3545 Frequency : sometimes when underlying device returns I/O errors
3547 Description: OSTs not going read-only during write failures
3548 Details : OSTs are not remounted read-only when the journal commit threads
3549 get I/O errors because fsfilt_ext3 calls journal_start/stop()
3550 instead of the ext3 wrappers.
3553 Frequency : only with 10000 clients or more
3555 Description: poor connect scaling with increasing client count
3556 Details : Don't run filter_grant_sanity_check for more than 100 exports
3557 to improve scaling for large numbers of clients.
3560 Frequency : SLES10 only
3562 Description: sanity-quota.sh quotacheck failed: rc = -22
3563 Details : Quotas cannot be enabled on SLES10.
3566 Frequency : liblustre clients only
3568 Description: getdirentries does not give error when run on compute nodes
3569 Details : getdirentries does not fail when the size specified as an argument
3570 is too small to contain at least one entry
3572 Severity : enhancement
3574 Description: Add LNET router traceability for debug purposes
3575 Details : If a checksum failure occurs with a router as part of the
3576 IO path, the NID of the last router that forwarded the bulk data
3577 is printed so it can be identified.
3582 Description: OST "spontaneously" evicts client; client has imp_pingable == 0
3583 Details : Due to a race condition, liblustre clients were occasionally
3584 evicted incorrectly.
3586 Severity : enhancement
3588 Description: lfs setstripe use optional parameters instead of postional
3591 Severity : enhancement
3593 Description: Nanosecond timestamp support for ldiskfs
3594 Details : The on-disk ldiskfs filesystem has added support for nanosecond
3595 resolution timestamps. There is not yet support for this at
3596 the Lustre filesystem level.
3599 Frequency : during server recovery
3601 Description: MDS failing to send precreate requests due to OSCC_FLAG_RECOVERING
3602 Details : request with rq_no_resend flag not awake l_wait_event if they get a
3606 Frequency : nfs export on patchless client
3608 Description: connectathon hang when test nfs export over patchless client
3609 Details : Disconnected dentry cannot be found with lookup, so we do not need
3610 to unhash it or make it invalid
3613 Description: fix llapi_lov_get_uuids() to allow many OSTs to be returned
3614 Details: : Change llapi_lov_get_uuids() to read the UUIDs from /proc instead
3615 of using an ioctl. This allows lfsck for > 160 OSTs to succeed.
3620 Description: open req refcounting wrong on reconnect
3621 Details : If reconnect happened between getting open reply from server and
3622 call to mdc_set_replay_data in ll_file_open, we will schedule
3623 replay for unreferenced request that we are about to free.
3624 Subsequent close will crash in variety of ways.
3625 Check that request is still eligible for replay in
3626 mdc_set_replay_data().
3631 Description: disable writes to filesystem when reading health_check file
3632 Details : the default for reading the health_check proc file has changed
3633 to NOT do a journal transaction and write to disk, because this
3634 can cause reads of the /proc file to hang and block HA state
3635 checking on a healthy but otherwise heavily loaded system. It
3636 is possible to return to the previous behaviour during configure
3637 with --enable-health-write.
3639 Severity : enhancement
3641 Description: 64-bit inode version
3642 Details: : Add a on-disk 64-bit inode version for ext3 to track changes made
3643 to the inode. This will be required for version-based recovery.
3648 Description: MDS fails to start if a duplicate client export is detected
3649 Details : in some rare cases it was possible for a client to connect to
3650 an MDS multiple times. Upon recovery the MDS would detect this
3651 and fail during startup. Handle this more gracefully.
3653 Severity : enhancement
3655 Description: Add -o localflock option to simulate old noflock behaviour.
3656 Details : This will achieve local-only flock/fcntl locks coherentness.
3661 Description: log_commit_thread vs filter_destroy race leads to crash
3662 Details : Take import reference before releasing llog record semaphore
3667 Description: Wrong request locking in request set processing
3668 Details : ptlrpc_check_set wrongly uses req->rq_lock for proctect add to
3669 imp_delayed_list, in this place should be used imp_lock.
3672 Frequency : when reconnecting
3674 Description: Grant leak when OSC reconnect to OST
3675 Details : When osc reconnect ost, OST(filter) should check whether it
3676 should grant more space to client by comparing fed_grant and
3677 cl_avail_grant, and return the granted space to client instead
3678 of "new granted" space, because client will call osc_init_grant
3679 to update the client grant space info.
3682 Frequency : when client reconnects to OST
3684 Description: Grant leak when OSC does a resend and replays bulk write
3685 Details : When osc reconnect to OST, OST(filter) should clear grant info of
3686 bulk write request, because the grant info while be sync between
3687 OSC and OST when reconnect, and we should ignore the grant info
3688 these of resend/replay write req.
3693 Description: Grant space more than avaiable space sometimes.
3694 Details : When then OST is about to be full, if two bulk writing from
3695 different clients came to OST. Accord the avaliable space of the
3696 OST, the first req should be permitted, and the second one
3697 should be denied by ENOSPC. But if the seconde arrived before
3698 the first one is commited. The OST might wrongly permit second
3699 writing, which will cause grant space > avaiable space.
3702 Frequency : when client is evicted
3704 Description: Grant might be wrongly erased when osc is evicted by OST
3705 Details : when the import is evicted by server, it will fork another
3706 thread ptlrpc_invalidate_import_thread to invalidate the
3707 import, where the grant will be set to 0. While the original
3708 thread will update the grant it got when connecting. So if
3709 the former happened latter, the grant will be wrongly errased
3710 because of this race.
3715 Description: Checking Stale with correct fid
3716 Details : ll_revalidate_it should uses de_inode instead of op_data.fid2
3717 to check whether it is stale, because sometimes, we want the
3718 enqueue happened anyway, and op_data.fid2 will not be initialized.
3720 Severity : enhancement
3722 Description: update patchless client
3723 Details : Add support for patchless client with 2.6.20, 2.6.21 and RHEL 5
3726 Frequency : only with 2.4 kernel
3728 Description: random memory corruption
3729 Details : size of struct ll_inode_info is to big for union inode.u and this
3730 can be cause of random memory corruption.
3735 Description: Memory leak in recovery
3736 Details : Lov_mds_md was not free in an error handler in mds_create_object.
3737 It should also check obd_fail before fsfilt_start, otherwise if
3738 fsfilt_start return -EROFS,(failover mds during mds recovery).
3739 then the req will return with repmsg->transno = 0 and rc = EROFS.
3740 and we met hit the assert LASSERT(req->rq_reqmsg->transno ==
3741 req->rq_repmsg->transno) in ptlrpc_replay_interpret. Fcc should
3742 be freed no matter whether fsfilt_commit success or not.
3745 Frequency : only with large numbers of cores on a single node
3747 Description: Prevents from taking the superblock lock in llap_from_page for
3749 Details : using LL_ORIGIN_REMOVEPAGE origin flag instead of LL_ORIGIN_UNKNOW
3750 for llap_from_page call in ll_removepage() prevents from taking
3751 the superblock lock for a soon died page.
3756 Description: Not check open intent error before release open handle
3757 Details : in some rare cases, the open intent error is not checked before
3758 release open handle, which may cause
3759 ASSERTION(open_req->rq_transno != 0), because it tries to release
3760 the failed open handle.
3765 Description: Set cat log bitmap only after create log success.
3766 Details : in some rare cases, the cat log bitmap is set too early. and it
3767 should be set only after create log success.
3771 Description: Accessing a block bevice can re-enable I/O when Lustre is
3772 tearing down a device.
3773 Details : dev_clear_rdonly(bdev) must be called in kill_bdev() instead of
3778 Description: service threads may hog cpus when there are a lot of requests
3779 Details : Insert cond_resched to give other threads a chance to use some CPU
3784 Description: the cat log was not initialized in recovery
3785 Details : When mds(mgs) do recovery, the tgt_count might be zero, so the
3786 unlink log on mds will not be initialized until mds post
3787 recovery. And also in mds post recovery, the unlink log will
3788 initialization will be done asynchronausly, so there will be race
3789 between add unlink log and unlink log initialization.
3793 Description: brw_stats were being printed incorrectly
3794 Details : brw_stats were being printed as log2 but all of them were not
3795 recorded as log2. Also remove some code duplication arising from
3796 filter_tally_{read,write}.
3800 Frequency : rare, only in recovery.
3801 Description: ASSERTION(req->rq_type != LI_POISON) failed
3802 Details : imp_lock should be held while iterating over imp_sending_list for
3803 prevent destroy request after get timeout in ptlrpc_queue_wait.
3807 Description: replay-single.sh test 52 fails
3808 Details : A lock's skiplist need to be cleanup when it being unlinked
3809 from its resource list.
3814 Description: Short directio read returns full requested size rather than
3816 Details : Direct I/O operations should return actual amount of bytes
3817 transferred rather than requested size.
3819 Severity : enhancement
3821 Description: metadata RPC reduction (e.g. for rm performance)
3822 Details : decrease the amount of synchronous RPC between clients and servers
3823 by canceling conflicing lock before the operation on the client
3824 and packing thier handles into the main operation RPC to server.
3826 Severity : enhancement
3828 Description: add #ifdef HAVE_KERNEL_CONFIG_H
3829 Details : kernels from 2.6.19 not need include linux/config.h, but add
3830 include linux/autoconf.h in commpiler command line.
3832 Severity : enhancement
3834 Description: patchless client support for 2.6.22 kernel
3835 Details : 2.6.22 has only one visble change, SLAB_CTOR_* constants is
3836 removed. In this case we need drop using os depended interface to
3837 kmem_cache and use cfs_mem_cache API.
3842 Description: fix mal-formatted messages
3843 Details : fix some mal-formatted DEBUG_REQ and LCONSOLE_ERROR_MSG messages
3847 Frequency : always in liblustre
3848 Description: wrong IS_ERR implementation in liblustre.h
3849 Details : fix IS_ERR implementation in liblustre.h for right detect errors.
3854 Description: Correct condition for output debug message.
3855 Details : inode i_nlink equal zero is not enough for output message about
3856 disk corruption, i_ctime and i_mode should be also checked.
3860 Frequency : always in patchless client
3861 Description: add configure check for truncate_complete_page
3862 Details : improve checks for exported symbols. This allow run check without
3863 sources, but with Module.symvers shipped with kernel distribution.
3864 add check for truncate_complete_page used by patchless client.
3868 Description: sanity.sh test_77h fails with "test_77h file compare failed"
3869 Details : test_77h uses a file which was messed by other test case.
3873 Description: Not Check whether lov_tgts is NULL in some lov functions
3874 Details : Checking whether lov_tgts is NULL in some functions.
3878 Description: replace obdo_alloc() with OBDO_ALLOC macro
3879 Details : nothing special is done in obdo_alloc() function, and for
3880 debugging purpose, it needs to be replaced with macros.
3884 Description: bad return value and errno from fcntl call
3885 Details : In liblustre API, errno should be a negative value if error
3890 Description: ptlrpc_check_set() LBUG
3891 Details : In case of positive reply from server and failed client bulk
3892 callback after bulk transfer shouldn't LBUG, but process this
3893 request as erroneous.
3895 Severity : enhancement
3897 Description: VFS operations stats tool.
3898 Details : Tool which collects stats by tracking value written in pid,
3899 ppid, gid and uses llstat to generate output to plot graph using
3901 Updated lustre/utils/Makefile.am
3902 Added lustre/utils/ltrack_stats.c
3904 Severity : enhancement
3906 Description: 2.6.18 server support (lustre 1.6.1)
3907 Details : Support for 2.6.18 kernels on the server side.
3912 Description: ASSERTION(imp->imp_conn_current) failed
3913 Details : an assertion failure is hit if a client node boots and attempts to
3914 mount a lustre filesystem faster than RECONNECT_INTERVAL seconds.
3917 Frequency : only for i686
3919 Description: 1.4.11 RC1 build fails for RHEL 4, i686
3920 Details : Fixed config variable for build.
3925 Description: Updated patchess for new RHEL4 kernel
3926 Details : Updated patch inode-nr_unused-2.6.9-rhel4.patch
3927 Updated patch jbd-stats-2.6.9.patch
3928 Updated patch qsnet-rhel4-2.6.patch
3929 Updated patch quota-deadlock-on-pagelock-core.patch
3930 Updated patch vfs_intent-2.6-rhel4.patch
3931 Updated patch vfs_races-2.6-rhel4.patch
3932 Updated series file 2.6-rhel4-titech.series
3933 Updated series file 2.6-rhel4.series
3934 Updated kernel config files
3939 Description: lquota slave complains LBUG when reconnecting with mds
3941 Details : quota slave depends qctxt->lqc_import to send its quota request.
3942 This pointer will be invalid if mds did failover or broke its
3943 connect to osts, which leads to LBUG.
3946 Frequency : when qunit size is too small(less than 20M)
3948 Description: write is stopped by improper -EDQUOT
3949 Details : If the master is busy and qunit size is small enough(let's say 1M),
3950 the slave can not get quota from master on time, which will lead
3951 slave to trigger a -EQUOTA to client.
3956 Description: Deadlock during metadata tests
3957 Details : in prune_dir_dentries(), shrink_dcache_parent() should not be
3958 called with the per-dentry lock held.
3961 Frequency : SLES9 only
3963 Description: Lustre patched kernel for SLES9 SP3 has NR_CPUS set to 8
3964 Details : set CONFIG_NR_CPUS to 128 instead of 8.
3966 Severity : enhancement
3968 Description: remove fs_prep_san_write operation and related patches
3969 Details : remove the ext3-san-jdike patches which are no longer useful.
3974 Description: LDISKFS-fs error (device sdc): ldiskfs_free_blocks
3975 Details : a disk corruption can cause the mballoc code to assert on a
3976 double free or other extent corruptions. Handle these with
3977 ext3_error() instead of assertions.
3980 Frequency : only with mballoc3 code and deep extent trees
3982 Description: ldiskfs_ext_search_right: bad header in inode: unexpected eh_depth
3983 Details : a wrong check of extent headers in ldiskfs_ext_search_right()
3984 can cause the filesystem to be remounted read-only.
3989 Description: lfsck built against 1.4.x cannot run against 1.6.0 lustre
3990 Details : the definition for OBD_IOC_GETNAME changed in 1.6.0. One of the
3991 few external users of this ioctl number is lfsck's call to
3992 llapi_lov_get_uuids() and this caused lfsck to fail at startup.
3993 Add the old ioctl number to the handler so both old and new
3998 Description: parallel lock callbacks
3999 Details : Instead of sending blocking and completion callbacks as separated
4000 requests, adding them to a set and sending in parallel.
4004 Description: Disable most debugging by default
4005 Details : To improve performance, disable most logging (for debug purposes)
4006 by default. VFSTRACE, RPCTRACE, and DLMTRACE are now off by
4007 default, and HA includes fewer messages.
4011 Frequency : only run on patchless client.
4012 Description: use do_facet on sanity.sh for test handling recoverables errors
4013 Details : use do_facet instead of direct use sysctl for set fail_loc on OST
4017 Description: hash tables for lists of nids, connections and uuids
4018 Details : Hash tables noticeably help when a lot of clients connect to a
4019 server, to faster identify duplicate connections or reconnects,
4020 also to faster find export to evict in manual eviction case.
4024 Description: Sometimes, when the server evict a client, and the client will
4025 not be evicted as soon as possible.
4026 Details : In enqueue req, the error was returned by intent, instead of
4027 rq_status which make ptlrpc layer not detect this error, and
4028 does not evict the client. So enqueue error should be returned
4032 Frequency : only at startup
4034 Description: Delay client connections to MDT until fist MDT->OST connect
4035 Details : If a client tried to create a new file before the MDT had
4036 connected to any OSTs, the create would return EIO. Now
4037 the client will simply block until the MDT connects to the
4038 first OST and the create can succeed.
4041 Frequency : only for SLES9
4043 Description: Routinely utilize latest Quadrics drivers in CFS releases
4044 Details : Update patch qsnet-suse-2.6.patch.
4047 Frequency : only for sles10
4049 Description: Update patches for SLES 10 SP1 kernel.
4050 Details : Update the patch vfs_intent-2.6-sles10.patch.
4055 Description: Routinely utilize latest Quadrics drivers in CFS releases
4056 Details : Update patch qsnet-rhel4-2.6.patch.
4059 Frequency : at statup only
4061 Description: mds_lov_synchronize race leads to various problems
4062 Details : simultaneous MDT->OST connections at startup can cause the
4063 sync to abort, leaving the OSC in a bad state.
4068 Description: Using wrong pointer in osc_brw_prep_request
4069 Details : Access to array[-1] can produce panic if kernel compiled with
4070 CONFIG_PAGE_ALLOC enabled
4072 Severity : enhancement
4074 Description: Async OSC create to avoid the blocking unnecessarily.
4075 Details : If a OST has no remain object, system will block on the creating
4076 when need to create a new object on this OST. Now, ways use
4077 pre-created objects when available, instead of blocking on an
4078 empty osc while others are not empty. If we must block, we block
4079 for the shortest possible period of time.
4083 Frequency : only in recovery
4084 Description: Mark OST as early accessible if his start SYNC.
4085 Details : osc_precreate return flag early accessible if oscc marked as
4086 OSCC_FLAG_SYNC_IN_PROGRESS.
4091 Description: Sometimes precreate code can triger create object on wrong ost
4092 Details : Wrong protected or not not restored variables aftre precreate loop
4093 can produce creation object on wrong ost.
4096 Frequency : oss recovery
4098 Description: llog_commit_thread cleanup should sync with llog_commit_thread
4100 Details : llog_commit_thread_count should be synced between llog_commit
4101 start and cleanup, so new llog_commit thread should not be started
4102 when llog_commit threads being stopped to avoid accessing some
4105 Severity : enhancement
4107 Description: Add printing inode info into message about error in writepage.
4109 --------------------------------------------------------------------------------
4111 2007-05-03 Cluster File Systems, Inc. <info@clusterfs.com>
4116 Frequency : on some architectures
4118 Description: 1.6 client sometimes fails to mount from a 1.4 MDT
4119 Details : Uninitialized flags sometimes cause configuration commands to
4123 Frequency : patchless clients only
4125 Description: missing __iget() symbol export
4126 Details : The __iget() symbol export is missing. To avoid the need for
4127 this on patchless clients the deathrow inode reaper is turned
4128 off, and we depend on the VM to clean up old inodes. This
4129 dependency was during via the fix for bug 12181.
4134 Description: sanity.sh fail: test_52b
4135 Details : The ll_inode_to_ext_flags() has a glitch which makes MDS return
4136 incorrect inode's flags to client.
4138 --------------------------------------------------------------------------------
4140 2007-04-19 Cluster File Systems, Inc. <info@clusterfs.com>
4142 * CONFIGURATION CHANGE. This version of Lustre WILL NOT
4143 INTEROPERATE with older versions automatically. In many cases a
4144 special upgrade step is needed. Please read the
4145 user documentation before upgrading any part of a 1.4.x system.
4146 * WARNING: Lustre configuration and startup changes are required with
4147 this release. See https://mail.clusterfs.com/wikis/lustre/MountConf
4149 * Support for kernels:
4150 2.4.21-47.0.1.EL (RHEL 3),
4151 2.6.5-7.283 (SLES 9),
4152 2.6.9-42.0.10.EL (RHEL 4),
4153 2.6.12.6 vanilla (kernel.org),
4154 2.6.16.27-0.9 (SLES10)
4155 * Client support for unpatched kernels:
4156 (see https://mail.clusterfs.com/wikis/lustre/PatchlessClient)
4157 2.6.16 - 2.6.19 vanilla (kernel.org),
4158 2.6.9-42.0.8EL (RHEL 4)
4159 * Recommended e2fsprogs version: 1.39.cfs6
4160 * Note that reiserfs quotas are disabled on SLES 10 in this kernel
4163 Severity : enhancement
4165 Description: Async OSC create to avoid the blocking unnecessarily.
4166 Details : If a OST has no remain object, system will block on the creating
4167 when need to create a new object on this OST. Now, ways use
4168 pre-created objects when available, instead of blocking on an
4169 empty osc while others are not empty. If we must block, we block
4170 for the shortest possible period of time.
4172 Severity : enhancement
4174 Description: MountConf
4175 Details : Lustre configuration is now managed via mkfs and mount
4176 commands instead of lmc and lconf. New obd types (MGS, MGC)
4177 are added for dynamic configuration management. See
4178 https://mail.clusterfs.com/wikis/lustre/MountConf for
4181 Severity : enhancement
4183 Description: dynamic OST addition
4184 Details : OSTs can now be added to a live filesystem
4186 Severity : enhancement
4188 Description: startup order invariance
4189 Details : MDTs and OSTs can be started in any order. Clients only
4190 require the MDT to complete startup.
4192 Severity : enhancement
4194 Description: parallel, asynchronous orphan cleanup
4195 Details : orphan cleanup is now performed in separate threads for each
4196 OST, allowing parallel non-blocking operation.
4198 Severity : enhancement
4200 Description: optimized stripe assignment
4201 Details : stripe assignments are now made based on ost space available,
4202 ost previous usage, and OSS previous usage, in order to try
4203 to optimize storage space and networking resources.
4205 Severity : enhancement
4207 Description: Permanently set tunables
4208 Details : All writable /proc/fs/lustre tunables can now be permanently
4209 set on a per-server basis, at mkfs time or on a live
4212 Severity : enhancement
4214 Description: Lustre message v2
4215 Details : Add lustre message format v2.
4217 Severity : enhancement
4219 Description: client OST exclusion list
4220 Details : Clients can be started with a list of OSTs that should be
4221 declared "inactive" for known non-responsive OSTs.
4225 Description: ENOENT returned for valid filehandle during dbench.
4226 Details : Check if a directory has children when invalidating dentries
4227 associated with an inode during lock cancellation. This fixes
4228 an incorrect ENOENT sometimes seen for valid filehandles during
4229 testing with dbench.
4232 Frequency : SFS test only (otherwise harmless)
4234 Description: SPEC SFS validation failure on NFS v2 over lustre.
4235 Details : Changes the blocksize for regular files to be 2x RPC size,
4236 and not depend on stripe size.
4238 Severity : enhancement
4240 Description: fine-grained SMP locking inside DLM
4241 Details : Improve DLM performance on SMP systems by removing the single
4242 per-namespace lock and replace it with per-resource locks.
4244 Severity : enhancement
4246 Description: don't hold multiple extent locks at one time
4247 Details : To avoid client eviction during large writes, locks are not
4248 held on multiple stripes at one time or for very large writes.
4249 Otherwise, clients can block waiting for a lock on a failed OST
4250 while holding locks on other OSTs and be evicted.
4252 Severity : enhancement
4254 Description: Multiple MD RPCs in flight.
4255 Details : Further unserialise some read-only MDT RPCs - learn about intents.
4256 To avoid overly-overloading MDT, introduce a limit on number of
4257 MDT RPCs in flight for a single client and add /proc controls
4258 to adjust this limit.
4260 Severity : enhancement
4262 Description: client read/write statistics
4263 Details : Add client read/write call usage stats for performance
4264 analysis of user processes.
4265 /proc/fs/lustre/llite/*/offset_stats shows non-sequential
4266 file access. extents_stats shows chunk size distribution.
4267 extents_stats_per_process show chunk size distribution per
4270 Severity : enhancement
4272 Description: per-client statistics on server
4273 Details : Add ldlm and operations statistics for each client in
4274 /proc/fs/lustre/mds|obdfilter/*/exports/
4276 Severity : enhancement
4278 Description: improved MDT statistics
4279 Details : Add detailed MDT operations statistics in
4280 /proc/fs/lustre/mds/*/stats
4282 Severity : enhancement
4284 Description: VFS operations stats
4285 Details : Add client VFS call stats, trackable by pid, ppid, or gid
4286 /proc/fs/lustre/llite/*/stats_track_[pid|ppid|gid]
4291 Description: Fix client-side osc byte counters
4292 Details : The osc read/write byte counters in
4293 /proc/fs/lustre/osc/*/stats are now working
4296 Frequency : always as root on SLES
4298 Description: Failure of copying files with lustre special EAs.
4299 Details : Client side always return success for setxattr call for lustre
4300 special xattr (currently only "trusted.lov").
4305 Description: Refcount LNET uuids
4306 Details : The global LNET uuid list grew linearly with every startup;
4307 refcount repeated list entries instead of always adding to
4310 Severity : enhancement
4312 Description: Dynamic service threads
4313 Details : Within a small range, start extra service threads
4314 automatically when the request queue builds up.
4317 Frequency : mixed-endian client/server environments
4319 Description: mixed-endian crashes
4320 Details : The new msg_v2 system had some failures in mixed-endian
4323 Severity : enhancement
4325 Description: Easy OST removal
4326 Details : OSTs can be permanently deactivated with e.g. 'lctl
4327 conf_param lustre-OST0001.osc.active=0'
4329 Severity : enhancement
4331 Description: MGS proc entries
4332 Details : Added basic proc entries for the MGS showing what filesystems
4335 Severity : enhancement
4337 Description: provide MGS failover
4338 Details : Added config lock reacquisition after MGS server failover.
4340 Severity : enhancement
4342 Description: add Linux 2.4 support
4343 Details : Added support for RHEL 2.4.21 kernel for 1.6 servers and clients
4347 Description: a large application tries to do I/O to the same resource and dies
4348 in the middle of it.
4349 Details : Check the req->rq_arrival time after the call to
4350 ost_brw_lock_get(), but before we do anything about
4351 processing it & sending the BULK transfer request. This
4352 should help move old stale pending locks off the queue as
4353 quickly as obd_timeout.
4356 Frequency : when an incorrect nid is specified during startup
4358 Description: ptlrpc connect to non-existant node causes kernel crash
4359 Details : LNET can't be re-entered from an event callback, which
4360 happened when we expire a message after the export has been
4361 cleaned up. Instead, hand the zombie cleanup off to another
4364 Severity : enhancement
4366 Description: plain/inodebits lock performance improvement
4367 Details : Grouping plain/inodebits in granted list by their request modes
4368 and bits policy, thus improving the performance of search through
4372 Frequency : only if OST filesystem is corrupted
4374 Description: client incorrectly hits assertion in ptlrpc_replay_req()
4375 Details : for a short time RPCs with bulk IO are in the replay list,
4376 but replay of bulk IOs is unimplemented. If the OST filesystem
4377 is corrupted due to disk cache incoherency and then replay is
4378 started it is possible to trip an assertion. Avoid putting
4379 committed RPCs into the replay list at all to avoid this issue.
4382 Frequency : liblustre (e.g. catamount) on a large cluster with >= 8 OSTs/OSS
4384 Description: System hang on startup
4385 Details : This bug allowed the liblustre (e.g. catamount) client to
4386 return to the app before handling all startup RPCs. This
4387 could leave the node unresponsive to lustre network traffic
4388 and manifested as a server ptllnd timeout.
4390 Severity : enhancement
4392 Description: Add "/proc/sys/lustre/debug_peer_on_timeout"
4393 Details : liblustre envirable: LIBLUSTRE_DEBUG_PEER_ON_TIMEOUT
4394 boolean to control whether to print peer debug info when a
4395 client's RPC times out.
4398 Frequency : only for kernels with patches from Lustre below 1.4.3
4400 Description: Remove old rdonly API
4401 Details : Remove old rdonly API which unused from at least lustre 1.4.3
4404 Frequency : only for devices with external journals
4406 Description: Set external device read-only also
4407 Details : During a commanded failover stop, we set the disk device
4408 read-only while the server shuts down. We now also set any
4409 external journal device read-only at the same time.
4412 Frequency : when upgrading from 1.4 while trying to change parameters
4414 Description: The wrong (new) MDC name was used when setting parameters for
4415 upgraded MDT's. Also allows changing of OSC (and MDC)
4416 parameters if --writeconf is specified at tunefs upgrade time.
4419 Frequency : when setting specific ost indicies
4421 Description: QOS code breaks on skipped indicies
4422 Details : Add checks for missing OST indicies in the QOS code, so OSTs
4423 created with --index need not be sequential.
4425 Severity : enhancement
4427 Description: Add uninit_groups feature to ldiskfs2 to speed up e2fsck
4428 Details : The uninit_groups feature works in conjunction with the kernel
4429 filesystem code (ldiskfs2 only) and e2fsprogs-1.39-cfs6 to speed
4430 up the pass1 processing of e2fsck. This is a read-only feature
4431 in ldiskfs2 only, so older kernels and current ldiskfs cannot
4432 mount filesystems that have had this feature enabled.
4434 Severity : enhancement
4436 Description: Improve multi-block allocation algorithm to avoid fragmentation
4437 Details : The mballoc3 code (ldiskfs2 only) adds new mechanisms to improve
4438 allocation locality and avoid filesystem fragmentation.
4440 ------------------------------------------------------------------------------
4442 2007-04-01 Cluster File Systems, Inc. <info@clusterfs.com>
4444 * Support for kernels:
4445 2.4.21-47.0.1.EL (RHEL 3)
4446 2.6.5-7.283 (SLES 9)
4447 2.6.9-42.0.10.EL (RHEL 4)
4448 2.6.12.6 vanilla (kernel.org)
4449 2.6.16.27-0.9 (SLES 10)
4450 * Recommended e2fsprogs version: 1.39.cfs5
4452 * Note that reiserfs quotas are disabled on SLES 10 in this kernel
4456 Frequency : occasional, depends on client load and configuration
4457 Bugzilla : 12181, 12203
4458 Description: data loss for recently-modified files
4460 Details : In some cases it is possible that recently written or created
4461 files may not be written to disk in a timely manner (this should
4462 normally be within 30s unless client IO load is very high).
4463 The problem appears as zero-length files or files that are a
4464 multiple of 1MB in size after a client crash or client eviction
4465 that are missing data at the end of the file.
4467 This problem is more likely to be hit on clients where files are
4468 repeatedly created and unlinked in the same directory, clients
4469 have a large amount of RAM, have many CPUs, the filesystem has
4470 many OSTs, the clients are rebooted frequently, and/or the files
4471 are not accessed by other nodes after being written.
4473 The presence of the problem can be detected by looking at
4474 /proc/sys/fs/inode-state. If the first number (nr_inodes) is
4475 smaller than the second (nr_unused) then dirty files will not
4476 be flushed automatically to disk. "sync; sleep 10" should be
4477 run several times on the node before unmounting it to update
4478 Lustre (this is also safe to run on nodes without this problem).
4480 There is also a related kernel bug in the RHEL4 4 2.6.9 kernel
4481 that can cause this same problem, so customers using that kernel
4482 also need to update the kernel in addition to Lustre. In order
4483 to properly fix this bug, the RHEL3 2.4.21 kernel is also updated.
4485 It is normal that files written just before a client crash (less
4486 than 30s) may not yet have been flushed to disk, even for local
4490 Frequency : frequent on thin XT3 nodes
4492 Description: UUID collision on thin XT3 Linux nodes
4493 Details : UUIDs on Compute Node Linux XT3 nodes were not generated
4494 randomly, since we relied on an insufficiently-seeded PRNG.
4499 Description: OSS hangs after "All ost request buffers busy"
4500 Details : A deadlock between quota and journal operations caused OSS
4501 hangs after printing "All ost request buffers busy."
4504 Frequency : always on liblustre builds
4506 Description: Cleanup compiler warnings on liblustre
4509 Frequency : always on liblustre builds on XT3
4511 Description: LC_CONFIG_CDEBUG don't run while build liblustre on XT3.
4515 Description: Addition of EXT3_FEATURE_RO_COMPAT_DIR_NLINKS flag for
4516 > 32000 subdirectories
4517 Details : Add EXT3_FEATURE_RO_COMPAT_DIR_NLINK flag to
4518 EXT3_FEATURE_RO_COMPAT_SUPP. This flag will be set whenever
4519 subdirectory count crosses 32000. This will aid e2fsck to
4520 correctly handle more than 32000 subdirectories.
4523 Frequency : liblustre (e.g. catamount) on a large cluster with >= 8 OSTs/OSS
4525 Description: System hang on startup
4526 Details : This bug allowed the liblustre (e.g. catamount) client to
4527 return to the app before handling all startup RPCs. This
4528 could leave the node unresponsive to lustre network traffic
4529 and manifested as a server ptllnd timeout.
4531 Severity : enhancement
4533 Description: Add "/proc/sys/lustre/debug_peer_on_timeout"
4534 (liblustre envirable: LIBLUSTRE_DEBUG_PEER_ON_TIMEOUT)
4535 boolean to control whether to print peer debug info when a
4536 client's RPC times out.
4541 Description: make O_SYNC working on 2.6 kernels
4542 Details : 2.6 kernels use different method for mark pages for write,
4543 so need add a code to lustre for O_SYNC work.
4548 Description: Failure to close file and release space on NFS
4549 Details : Put inode details into lock acquired in ll_intent_file_open.
4550 Use mdc_intent_lock in ll_intent_open to properly
4551 detect all kind of errors unhandled by mdc_enqueue.
4556 Description: proc file read during shutdown sometimes raced obd removal,
4558 Details : Add lock to prevent obd access after proc file removal.
4561 Frequency : Only for files larger than 4GB on 32-bit clients.
4563 Description: improperly doing page alignment of locks
4564 Details : Modify lustre core code to use CFS_PAGE_* defines instead of
4565 PAGE_*. Make CFS_PAGE_MASK a 64-bit mask.
4570 Description: RPCs being resent when they shouldn't be
4571 Details : Some RPCs that should not be resent are being resent. This
4572 can cause inconsistencies in the RPC state machine. Do not
4573 resend such requests.
4576 Frequency : rare, only with NFS export
4578 Description: Crash on NFS re-export node
4579 Details : under very unusual load conditions an assertion is hit in
4580 ll_intent_file_open()
4583 Frequency : only if OST filesystem is corrupted
4585 Description: client incorrectly hits assertion in ptlrpc_replay_req()
4586 Details : for a short time RPCs with bulk IO are in the replay list,
4587 but replay of bulk IOs is unimplemented. If the OST filesystem
4588 is corrupted due to disk cache incoherency and then replay is
4589 started it is possible to trip an assertion. Avoid putting
4590 committed RPCs into the replay list at all to avoid this issue.
4595 Description: large O_DIRECT requests fail under memory pressure/fragmentation
4596 Details : Large single O_DIRECT read and write calls can fail to allocate
4597 a sufficiently large buffer to process the request. In case of
4598 allocation failure the allocation is retried with a smaller
4599 buffer and broken into smaller requests.
4601 Severity : enhancement
4603 Description: Add -o localflock option to simulate old noflock behaviour.
4604 Details : This will achieve local-only flock/fcntl locks coherentness.
4609 Description: versioning check is incomplete
4610 Details : Checking the version difference of client vs. server, report
4611 error if the gap is too big.
4616 Description: add support PG_writeback bit
4617 Details : add support for PG_writeback bit for Lustre, for more carefull
4618 work with page cache in 2.6 kernel. This also fix some deadlocks
4619 and remove hack for work O_SYNC with 2.6 kernel.
4621 Severity : enhancement
4623 Description: Add uninit_groups feature to ldiskfs2 to speed up e2fsck
4624 Details : The uninit_groups feature works in conjunction with the kernel
4625 filesystem code (ldiskfs2 only) and e2fsprogs-1.39-cfs6 to speed
4626 up the pass1 processing of e2fsck. This is a read-only feature
4627 in ldiskfs2 only, so older kernels and current ldiskfs cannot
4628 mount filesystems that have had this feature enabled.
4630 Severity : enhancement
4632 Description: Improve multi-block allocation algorithm to avoid fragmentation
4633 Details : The mballoc3 code (ldiskfs2 only) adds new mechanisms to improve
4634 allocation locality and avoid filesystem fragmentation.
4636 ------------------------------------------------------------------------------
4638 2007-02-09 Cluster File Systems, Inc. <info@clusterfs.com>
4640 * Support for kernels:
4641 2.6.9-42.0.3.EL (RHEL 4)
4642 2.6.5-7.276 (SLES 9)
4643 2.4.21-47.0.1.EL (RHEL 3)
4644 2.6.12.6 vanilla (kernel.org)
4645 2.6.16.21-0.8 (SLES10)
4646 * Recommended e2fsprogs version: 1.39.cfs2-0
4648 * The backwards-compatible /proc/sys/portals symlink has been removed
4649 in this release. Before upgrading, please ensure that you change
4650 any configuration scripts or /etc/sysctl.conf files that access
4651 /proc/sys/portals/* or sysctl portals.* to use the corresponding
4652 entry in /proc/sys/lnet or sysctl lnet.*. This change can be made
4653 in advance of the upgrade on any system running Lustre 1.4.6 or
4654 newer, since /proc/sys/lnet was added in that version.
4655 * Note that reiserfs quotas are disabled on SLES 10 in this kernel
4659 Frequency : only when quota is used
4661 Description: avoid scanning export list for quota master
4662 Details : Change the algorithms to avoid scanning export list in order
4663 to improve the efficiency.
4666 Frequency : MDS failover only, very rarely
4668 Description: "went back in time" messages on mds failover
4669 Details : The greatest transno may be lost when the current operation
4670 finishes with an error (transno==0) and the client's last_rcvd
4671 record is over-written. Save the greatest transno in the
4672 mds_last_transno for this case.
4675 Frequency : always for specific kernels and striping counts
4677 Description: client may get "Matching packet too big" without ACL support
4678 Details : Clients compiled without CONFIG_FS_POSIX_ACL get an error message
4679 when trying to access files in certain configurations. The
4680 clients should in fact be denied when mounting because they do
4681 not understand ACLs.
4684 Frequency : Cray XT3 with more than 4000 clients and multiple jobs
4686 Description: many clients connecting with IO in progress causes connect timeouts
4687 Details : Avoid synchronous journal commits to avoid delays caused by many
4688 clients connecting/disconnecting when bulk IO is in progress.
4689 Queue liblustre connect requests on OST_REQUEST_PORTAL instead of
4690 OST_IO_PORTAL to avoid delays behind potentially many pending
4694 Frequency : occasionally with multiple writers to a single file
4696 Description: shared writes to file may result in wrong size reported by stat()
4697 Details : Allow growing of kms when extent lock is cancelled
4700 Frequency : always with random mmap IO to multi-striped file
4702 Description: mmap write might be lost if we are writing to a 'hole' in stripe
4703 Details : Only if the hole is at the end of OST object so that kms is too
4704 small. Fix is to increase kms accordingly in ll_nopage.
4707 Frequency : rare, only if OST filesystem is inconsistent with MDS filesystem
4709 Description: writes to a missing object would leak memory on the OST
4710 Details : If there is an inconsistency between the MDS and OST filesystems,
4711 such that the MDS references an object that doesn't exist, writes
4712 to that object will leak memory due to incorrect cleanup in the
4713 error handling path, eventually running out of memory on the OST.
4718 Description: Creating too long symlink causes lustre errors
4719 Details : Check symlink and name lengths before sending requests to MDS.
4722 Frequency : only if flock is enabled (not on by default)
4724 Description: posix locks not released on fd closure on 2.6.9+
4725 Details : We failed to add posix locks to list of inode locks on 2.6.9+
4726 kernels, this caused such locks not to be released on fd close and
4727 then assertions on fs unmount about still used locks.
4730 Frequency : MDS failover only, very rarely
4732 Description: clients may get ASSERTION(granted_lock != NULL)
4733 Details : When request was taking a long time, and a client was resending
4734 a getattr by name lock request. The were multiple lock requests
4735 with the same client lock handle and
4736 mds_getattr_name->fixup_handle_for_resent_request found one of the
4737 lock handles but later failed with ASSERTION(granted_lock != NULL).
4742 Description: handle->h_buffer_credits > 0, assertion failure
4743 Details : h_buffer_credits is zero after truncate, causing assertion
4744 failure. This patch extends the transaction or creates a new
4748 Frequency : NFS re-export or patchless client
4749 Bugzilla : 11179, 10796
4750 Description: Crash on NFS re-export node (__d_move)
4751 Details : We do not want to hash the dentry if we don't have a lock.
4752 But if this dentry is later used in d_move, we'd hit uninitialised
4753 list head d_hash, so we just do this to init d_hash field but
4754 leave dentry unhashed.
4757 Frequency : NFS re-export or patchless client
4759 Description: NFS exports has problem with symbolic link
4760 Details : lustre client didn't properly install dentry when re-exported
4761 to NFS or running patchless client.
4764 Frequency : NFS re-export or patchless client
4766 Description: Various nfs/patchless fixes.
4767 Details : fixes reuse disconected alias for lookup process - this fixes
4768 warning "find_exported_dentry: npd != pd",
4769 fix permission error with open files at nfs.
4770 fix apply umask when do revalidate.
4773 Frequency : occasional
4775 Description: Crash on NFS re-export node
4776 Details : calling clear_page() on the wrong pointer triggered oops in
4777 generic_mapping_read().
4780 Frequency : rarely, using O_DIRECT IO
4782 Description: unaligned directio crashes client with LASSERT
4783 Details : check for unaligned buffers before trying any requests.
4786 Frequency : rarely, using CFS RAID5 patches in non-standard kernel series
4788 Description: stale data returned from RAID cache
4789 Details : If only a small amount of IO is done to the RAID device before
4790 reading it again it is possible to get stale data from the RAID
4791 cache instead of reading it from disk.
4794 Frequency : always for sles10 kernel
4796 Description: sles10 support
4797 Details : ll_follow_link: compile fixes and using of nd_set_link
4798 under newer kernels.
4801 Frequency : depends on arch, kernel and compiler version, always on sles10
4804 Description: recursive or deep enough symlinks cause stack overflow
4805 Details : getting rid of large stack-allocated variable in
4809 Frequency : depends on hardware
4811 Description: lustre write performance loss in the SLES10 kernel
4812 Details : the performance loss is caused by using of write barriers in the
4813 ext3 code. The SLES10 kernel turns barrier support on by
4814 default. The fix is to undo that change for ldiskfs.
4816 ------------------------------------------------------------------------------
4818 2006-12-09 Cluster File Systems, Inc. <info@clusterfs.com>
4820 * Support for kernels:
4821 2.6.9-42.0.3EL (RHEL 4)
4822 2.6.5-7.276 (SLES 9)
4823 2.4.21-47.0.1.EL (RHEL 3)
4824 2.6.12.6 vanilla (kernel.org)
4828 Frequency : quota enabled and large files being deleted
4830 Description: releasing more than 4GB of quota at once hangs OST
4831 Details : If a user deletes more than 4GB of files on a single OST it
4832 will cause the OST to spin in an infinite loop. Release
4833 quota in < 4GB chunks, or use a 64-bit value for 1.4.7.1+.
4838 Description: statfs data retrieved from /proc may be stale or zero
4839 Details : When reading per-device statfs data from /proc, in the
4840 {kbytes,files}_{total,free,avail} files, it may appear
4841 as zero or be out of date.
4844 Frequency : systems with MD RAID1 external journal devices
4846 Description: lconf's call to blkid is confused by RAID1 journal devices
4847 Details : Use the "blkid -l" flag to locate the MD RAID device instead
4848 of returning all block devices that match the journal UUID.
4851 Frequency : always, for aggregate stripe size over 4GB
4853 Description: "lfs setstripe" fails assertion when setting 4GB+ stripe width
4854 Details : Using "lfs setstripe" to set stripe size * stripe count over 4GB
4855 will fail the kernel with "ASSERTION(lsm->lsm_xfersize != 0)"
4858 Frequency : always if "lfs find" used on a local file/directory
4860 Description: "lfs find" segfaults if used on a local file/directory
4861 Details : The case where a directory component was not specified wasn't
4862 handled correctly. Handle this properly.
4865 Frequency : always on ppc64
4867 Description: the write to an ext3 filesystem mounted with mballoc got stuck
4868 Details : ext3_mb_generate_buddy() uses find_next_bit() which does not
4869 perform endianness conversion.
4872 Frequency : rarely (truncate to non-zero file size after write under load)
4873 Bugzilla : 10730, 10687
4874 Description: Files padded with zeros to next 4K multiple
4875 Details : With filesystems mounted using the "extents" option (2.6 kernels)
4876 it is possible that files that are truncated to a non-zero size
4877 immediately after being written are filled with zero bytes beyond
4878 the truncated size. No file data is lost.
4880 Severity : enhancement
4882 Description: Allow recovery/failover for liblustre clients.
4883 Details : liblustre clients were unaware of failover configurations until
4886 Severity : enhancement
4888 Description: user file locks should fail when not mounting with flock option
4889 Details : Set up an error-returning stub in ll_file_operations.lock field
4890 to prevent incorrect behaviour when client is mounted without
4891 flock option. Also, set up properly f_op->flock field for
4895 Frequency : always on ia64
4897 Description: "lfs df" loops on printing out MDS statfs information
4898 Details : The obd_ioctl_data was not initialized and in some systems
4899 this caused a failure during the ioctl that did not return
4900 an error. Initialize the struct and return an error on failure.
4903 Frequency : SLES 9 only
4905 Description: Error of copying files with lustre special EAs as root
4906 Details : Client side always return success for setxattr call for lustre
4907 special xattr (currently only "trusted.lov").
4910 Frequency : rarely on clusters with both ia64+i386 clients
4912 Description: ia64+i686 clients doing shared IO on the same file may LBUG
4913 Details : In rare cases when both ia64+i686 (or other mixed-PAGE_SIZE)
4914 clients are doing concurrent writes to the same file it is
4915 possible that the ia64 clients may LASSERT because the OST
4916 extent locks are not PAGE_SIZE aligned. Ensure that grown
4917 locks are always aligned on the request boundary.
4920 Frequency : specific use, occasional
4922 Description: Overwriting in use executable truncates on-disk binary image
4923 Details : If one node attempts to overwrite an executable in use by
4924 another node, we now correctly return ETXTBSY instead of
4925 truncating the file.
4927 Severity : enhancement
4929 Description: Async OSC create to avoid the blocking unnecessarily.
4930 Details : If a OST has no remain object, system will block on the creating
4931 when need to create a new object on this OST. Now, ways use
4932 pre-created objects when available, instead of blocking on an
4933 empty osc while others are not empty. If we must block, we block
4934 for the shortest possible period of time.
4939 Description: chmod on Lustre root is propagated to other clients
4940 Details : Re-validate root's dentry in ll_lookup_it to avoid having it
4941 invalid by the follow_mount time.
4944 Frequency : liblustre clients only
4946 Description: Race in 'instant cancel' lock handling could lead to such locks
4947 never to be granted in case of SMP MDS
4948 Details : Do not destroy not yet granted but cbpending locks in
4952 Frequency : replay/resend of open
4954 Description: non null lock assetion failure in mds_intent_policy
4955 Details : Trying to replay/resend lockless open requests resulted in
4956 mds_open() returning 0 with no lock. Now it sets a flag if
4957 a lock is going to be returned.
4959 Severity : enhancement
4961 Description: Checksum enhancements
4962 Details : New checksum enhancements allow for resending RPCs that failed
4965 Severity : enhancement
4967 Description: Tunables on number of dirty pages in cacche
4968 Details : Allow to set limit on number of dirty pages cached.
4973 Description: client crash on unmount - lock still has references
4974 Details : In some error handling cases it was possible to leak a lock
4975 reference on a client while accessing a file. This was not
4976 harmful to the client during operation, but would cause the
4977 client to crash when the filesystem is unmounted.
4980 Frequency : specific case, rare
4982 Description: ETXTBSY on mds though file not in use by client
4983 Details : ETXTBSY is no longer incorrectly returned when attempting to
4984 chmod or chown a directory that the user previously tried to
4985 execute or a currently-executing binary.
4988 Frequency : extremely rare except on liblustre-based clients
4990 Description: Lustre space not freed when files are deleted
4991 Details : Clean up open-unlinked files after client eviction. Previously
4992 the unlink was skipped and the files remained as orphans.
4997 Description: OST failure "would be an LBUG" in waiting_locks_callback()
4998 Details : In some cases it was possible to send a blocking callback to
4999 a client doing a glimpse, even though that client didn't get
5000 a lock granted. When the glimpse lock is cancelled on the OST
5001 the freed lock is left on the waiting list and corrupted the list.
5004 Frequency : all core dumps
5006 Description: Broke core dumps to lustre
5007 Details : Negative dentry may be unhashed if parent does not have UPDATE
5008 lock, but some callers, e.g. do_coredump, expect dentry to be
5009 hashed after successful create, hash it in ll_create_it.
5011 ------------------------------------------------------------------------------
5013 2006-09-13 Cluster File Systems, Inc. <info@clusterfs.com>
5015 * Support for kernels:
5016 2.6.9-42.0.2.EL (RHEL 4)
5017 2.6.5-7.276 (SLES 9)
5018 2.4.21-40.EL (RHEL 3)
5019 2.6.12.6 vanilla (kernel.org)
5023 Frequency : always on RHEL 3
5025 Description: Number of open files grows over time
5026 Details : The number of open files grows over time, whether or not
5027 Lustre is started. This was due to a filp leak introduced
5028 by one of our kernel patches.
5030 ------------------------------------------------------------------------------
5032 2006-08-20 Cluster File Systems, Inc. <info@clusterfs.com>
5034 * Support for kernels:
5035 2.6.9-42.EL (RHEL 4)
5036 2.6.5-7.267 (SLES 9)
5037 2.4.21-40.EL (RHEL 3)
5038 2.6.12.6 vanilla (kernel.org)
5043 Bugzilla : 5719, 9635, 9792, 9684
5044 Description: OST (or MDS) trips assertions in (re)connection under heavy load
5045 Details : If a server is under heavy load and cannot reply to new
5046 connection requests before the client resends the (re)connect,
5047 the connection handling code can behave badly if two service
5048 threads are concurrently handing separate (re)connections from
5049 the same client. Add better locking to the connection handling
5050 code, and ensure that only a single connection will be processed
5051 for a given client UUID, even if the lock is dropped.
5053 Severity : enhancement
5055 Description: add TCP zero-copy support to kernel
5056 Details : Add support to the kernel TCP stack to allow zero-copy bulk
5057 sends if the hardware supports scatter-gather and checksumming.
5058 This allows socklnd to do client-write and server-read more
5059 efficiently and reduce CPU utilization from skbuf copying.
5062 Frequency : only if NFS exporting from client
5064 Description: NULL pointer deref in ll_iocontrol() if chattr mknod file
5065 Details : If setting attributes on a file created under NFS that had
5066 never been opened it would be possible to oops the client
5067 if the file had no objects.
5071 Bugzilla : 9326, 10402, 10897
5072 Description: client crash in ptlrpcd_wake() thread when sending async RPC
5073 Details : It is possible that ptlrpcd_wake() dereferences a freed async
5074 RPC. In rare cases the ptlrpcd thread alread processed the RPC
5075 before ptlrpcd_wake() was called and the request was freed.
5078 Frequency : always for liblustre
5080 Description: liblustre client does MDS+OSTs setattr RPC for each write
5081 Details : When doing a write from a liblustre client, the client
5082 incorrectly issued an RPC to the MDS and each OST the file was
5083 striped over in order to update the timestamps. When writing
5084 with small chunks and many clients this could overwhelm the MDS
5085 with RPCs. In all cases it would slow down the write because
5086 these RPCs are unnecessary.
5088 Severity : enhancement
5090 Description: allow number of MDS service threads to be changed at module load
5091 Details : It is now possible to change the number of MDS service threads
5092 running. Adding "options mds mds_num_threads={N}" to the MDS's
5093 /etc/modprobe.conf will set the number of threads for the next
5094 time Lustre is restarted (assuming the "mds" module is also
5095 reloaded at that time). The default number of threads will
5096 stay the same, 32 for most systems.
5101 Description: OST crash if filesystem is unformatted or corrupt
5102 Details : If an OST is started on a device that has never been formatted
5103 or if the filesystem is corrupt and cannot even mount then the
5104 error handling cleanup routines would dereference a NULL pointer.
5109 Description: NULL pointer deref in llap_from_page.
5110 Details : get_cache_page_nowait can return a page with NULL (or otherwise
5111 incorrect) mapping if the page was truncated/reclaimed while it was
5112 searched for. Check for this condition and skip such pages when
5113 doing readahead. Introduce extra check to llap_from_page() to
5114 verify page->mapping->host is non-NULL (so page is not anonymous).
5117 Frequency : Sometimes when using sys_sendfile
5119 Description: "page not covered by a lock" warnings from ll_readpage
5120 Details : sendfile called ll_readpage without right page locks present.
5121 Now we introduced ll_file_sendfile that does necessary locking
5122 around call to generic_file_sendfile() much like we do in
5126 Frequency : with certain MDS communication failures at client mount time
5128 Description: NULL pointer deref after failed client mount
5129 Details : a client connection request may delayed by the network layer
5130 and not be sent until after the PTLRPC layer has timed out the
5131 request. If the client fails the mount immediately it will try
5132 to clean up before the network times out the request. Add a
5133 reference from the request import to the obd device and delay
5134 the cleanup until the network drops the request.
5137 Frequency : occasionally during client (re)connect
5139 Description: assertion failure during client (re)connect
5140 Details : processing a client connection request may be delayed by the
5141 client or server longer than the client connect timeout. This
5142 causes the client to resend the connection request. If the
5143 original connection request is replied in this interval, the
5144 client may trip an assertion failure in ptlrpc_connect_interpret()
5145 which thought it would be the only running connect process.
5148 Frequency : only with obd_echo servers and clients that are rebooted
5150 Description: kernel BUG accessing uninitialized data structure
5151 Details : When running an obd_echo server it did not start the ping_evictor
5152 thread, and when a client was evicted an uninitialized data
5153 structure was accessed. Start the ping_evictor in the RPC
5154 service startup instead of the OBD startup.
5156 Severity : enhancement
5157 Bugzilla : 10193 (patchless)
5158 Description: Remove dependency on various unexported kernel interfaces.
5159 Details : No longer need reparent_to_init, exit_mm, exit_files,
5160 sock_getsockopt, filemap_populate, FMODE_EXEC, put_filp.
5163 Frequency : rare (only users of deprecated and unsupported LDAP config)
5165 Description: write_conf for zeroconf mount queried LDAP incorrectly for client
5166 Details : LDAP apparently contains 'lustreName' attributes instead of
5167 'name'. A simple remapping of the name is sufficient.
5170 Frequency : rare (only with non-default dump_on_timeout debug enabled)
5172 Description: waiting_locks_callback trips kernel BUG if client is evicted
5173 Details : Running with the dump_on_timeout debug flag turned on makes
5174 it possible that the waiting_locks_callback() can try to dump
5175 the Lustre kernel debug logs from an interrupt handler. Defer
5176 this log dumping to the expired_lock_main() thread.
5178 Severity : enhancement
5180 Description: Support NFS exporting on 2.6 kernels.
5181 Details : Implement non-rawops metadata methods for NFS server to use without
5182 changing NFS server code.
5185 Frequency : very rare (synthetic metadata workload only)
5187 Description: two racing renames might cause an MDS thread to deadlock
5188 Details : Running the "racer" program may cause one MDS thread to rename
5189 a file from being the source of a rename to being the target of
5190 a rename at exactly the same time that another thread is doing
5191 so, and the second thread has already enqueued these locks after
5192 doing a lookup of the target and is trying to relock them in
5193 order. Ensure that we don't try to re-lock the same resource.
5196 Frequency : only very large systems with liblustre clients
5198 Description: slow eviction of liblustre clients with the "evict_by_nid" RPC
5199 Details : Use asynchronous set_info RPCs to send the "evict_by_nid" to
5200 all OSTs in parallel. This allows the eviction of stale liblustre
5201 clients to proceed much faster than if they were done in series,
5202 and also offers similar improvements for other set_info RPCs.
5207 Description: excessive CPU usage during initial read phase on client
5208 Details : During the initial read phase on a client, it would agressively
5209 retry readahead on the file, consuming too much CPU and impacting
5210 performance (since 1.4.5.8). Improve the readahead algorithm
5211 to avoid this, and also improve some other common cases (read
5212 of small files in particular, where "small" is files smaller than
5213 /proc/fs/lustre/llite/*/max_read_ahead_whole_mb, 2MB by default).
5218 Description: MDS crash when receiving packet with unknown intent.
5219 Details : Do not LBUG in unknown intent case, just return -EFAULT
5221 Severity : enhancement
5222 Bugzilla : 9293, 9385
5223 Description: MDS RPCs are serialised on client. This is unnecessary for some.
5224 Details : Do not serialize getattr (non-intent version) and statfs.
5227 Frequency : occasional, when OST network is overloaded/intermittent
5229 Description: client evicted by OST after bulk IO timeout
5230 Details : If a client sends a bulk IO request (read or write) the OST
5231 may evict the client if it is unresposive to its data GET/PUT
5232 request. This is incorrect if the network is overloaded (takes
5233 too long to transfer the RPC data) or dropped the OST GET/PUT
5234 request. There is no need to evict the client at all, since
5235 the pinger and/or lock callbacks will handle this, and the
5236 client can restart the bulk request.
5239 Frequency : Always when mmapping file with no objects
5241 Description: client crashes when mmapping file with no objects
5242 Details : Check that we actually have objects in a file before doing any
5243 operations on objects in ll_vm_open, ll_vm_close and
5249 Description: Request leak when working with deleted CWD
5250 Details : Introduce advanced request refcount tracking for requests
5251 referenced from lustre intent.
5253 Severity : Enhancement
5255 Description: Cache open file handles on client.
5256 Details : MDS now will return special lock along with openhandle, if
5257 requested and client is allowed to hold openhandle, even if unused,
5258 until such a lock is revoked. Helps NFS a lot, since NFS is opening
5259 closing files for every read/write openration.
5261 Severity : Enhancement
5263 Description: Cache open negative dentries on client when possible.
5264 Details : Guard negative dentries with UPDATE lock on parent dir, drop
5265 negative dentries on lock revocation.
5270 Description: Remounting a client read-only wasn't possible with a zconf mount
5271 Details : It wasn't possible to remount a client read-only with llmount.
5273 Severity : enhancement
5274 Description: Include MPICH 1.2.6 Lustre ADIO interface patch
5275 Details : In lustre/contrib/ or /usr/share/lustre in RPM a patch for
5276 MPICH is included to add Lustre-specific ADIO interfaces.
5277 This is based closely on the UFS ADIO layer and only differs
5278 in file creation, in order to allow the OST striping to be set.
5279 This is user-contributed code and not supported by CFS.
5284 Description: extended inode attributes (immutable, append-only) work improperly
5285 when 2.4 and 2.6 kernels are used on client/server or vice versa
5286 Details : Introduce kernel-independent values for these flags.
5288 Severity : enhancement
5291 Description: Allow fractional MB tunings for lustre in /proc/ filesystem.
5292 Details : Many of the /proc/ tunables can only be tuned at a megabyte
5293 granularity. Now, Fractional MB granularity is be supported,
5294 this is very useful for low memory system.
5296 Severity : enhancement
5298 Description: Getattr by fid
5299 Details : Getting a file attributes by its fid, obtaining UPDATE|LOOKUP
5300 locks, avoids extra getattr rpc requests to MDS, allows '/' to
5301 have locks and avoids getattr rpc requests for it on every stat.
5304 Frequency : Always, for filesystems larger than 2TB
5306 Description: ldiskfs crash at mount for filesystem larger than 2TB with mballoc
5307 Details : Kenrel kmalloc limits allocations to 128kB and this prevents
5308 filesystems larger than 2TB to be mounted with mballoc enabled.
5311 Frequency : Always, for 32-bit kernel without CONFIG_LBD and filesystem > 2TB
5313 Description: filesystem corruption for non-standard kernels and very large OSTs
5314 Details : If a 32-bit kernel is compiled without CONFIG_LBD enabled and a
5315 filesystems larger than 2TB is mounted then the kernel will
5316 silently corrupt the start of the filesystem. CONFIG_LBD is
5317 enabled for all CFS-supported kernels, but the possibility of
5318 this happening with a modified kernel config exists.
5320 Severity : enhancement
5322 Description: add client O_DIRECT support for 2.6 kernels
5323 Details : It is now possible to do O_DIRECT reads and writes to files
5324 in the Lustre client mountpoint on 2.6 kernel clients.
5326 Severity : enhancement
5328 Description: parallel glimpse, setattr, statfs, punch, destroy requests
5329 Details : Sends glimpse, setattr, statfs, punch, destroy requests to OSTs in
5330 parallel, not waiting for response from every OST before sending
5331 a rpc to the next OST.
5336 Description: setattr vs write race when updating file timestamps
5337 Details : Client processes that update a file timestamp into the past
5338 right after writing to the file (e.g. tar) it is possible that
5339 the updated file modification time can be reset to the current
5340 time due to a race between processing the setattr and write RPC.
5342 Severity : enhancement
5344 Description: Bring 'lfs find' closer in line with regular Linux find.
5345 Details : lfs find util supports -atime, -mtime, -ctime, -maxdepth, -print,
5346 -print0 options and obtains all the needed info through the lustre
5349 Severity : enhancement
5351 Description: support up to 1024 configured devices on one node
5352 Details : change obd_dev array from statically allocated to dynamically
5353 allocated structs as they are first used to reduce memory usage
5358 Description: Flush dirty partially truncated pages during truncate
5359 Details : Immediatelly flush partially truncated pages in filter_setattr,
5360 this way we completely avoid having any pages in page cache on OST
5361 and can retire ugly workarounds during writes to flush such pages.
5366 Description: i_sem vs transaction deadlock in mds_obd_destroy during unlink.
5367 Details : protect inode from truncation within vfs_unlink() context
5368 just take a reference before calling vfs_unlink() and release it
5369 when parent's i_sem is free.
5372 Frequency : always, if extents are used on OSTs
5374 Description: index ei_leaf_hi (48-bit extension) is not zeroed in extent index
5375 Details : OSTs using the extents format would not zero the high 16 bits of
5376 the index physical block number. This is not a problem for any
5377 OST filesystems smaller than 16TB, and no kernels support ext3
5378 filesystems larger than 16TB yet. This is fixed in 1.4.7 (all
5379 new/modified files) and can be fixed for existing filesystems
5380 with e2fsprogs-1.39-cfs1.
5385 Description: import connection selection may be incorrect if timer wraps
5386 Details : Using a 32-bit jiffies timer with HZ=1000 may cause backup
5387 import connections to be ignored if the 32-bit jiffies counter
5388 wraps. Use a 64-bit jiffies counter.
5391 Frequency : during server recovery
5393 Description: crash after server is denying duplicate export
5394 Details : If clients are resending connect requests to the server, the
5395 server refuses to allow a client to connect multiple times.
5396 Fixed a bug in the handling of this case.
5399 Frequency : very large clusters immediately after boot
5401 Description: LNET request buffers exhausted under heavy short-term load
5402 Details : If a large number of client requests are generated on a service
5403 that has previously never seen so many requests it is possible
5404 that the request buffer growth cannot keep up with the spike in
5405 demand. Instead of dropping incoming requests, they are held in
5406 the LND until the RPC service can accept more requests.
5409 Frequency : Sometimes during replay
5411 Description: Assertion failure in ll_local_open after replay.
5412 Details : If replay happened on an open request reply before we were able
5413 to set replay handler, reply will become not swabbed tripping the
5414 assertion in ll_local_open. Now we set the handler right after
5415 recognising of open request
5418 Frequency : very rare
5420 Description: kernel reports "badness in vsnprintf"
5421 Details : Reading from the "recovery_status" /proc file in small chunks
5422 may cause a negative length in lprocfs_obd_rd_recovery_status()
5423 call to vsnprintf() (which is otherwise harmless). Exit early
5424 if there is no more space in the output buffer.
5426 Severity : enhancement
5428 Description: clear OBD RPC statistics by writing to them
5429 Details : It is now possible to clear the OBD RPC statistics by writing
5430 to the "stats" file.
5435 Description: Client mtime is not the same on different clients after utimes
5436 Details : In some cases, the client was using the utimes() syscall on
5437 a file cached on another node. The clients now validate the
5438 ctime from the MDS + OSTs to determine which one is right.
5443 Description: Inability to activate failout mode
5444 Details : lconf script incorrectly assumed that in python string's numeric
5445 value is used in comparisons.
5448 Frequency : always with multiple stripes per file
5450 Description: Inefficient object allocation for mutli-stripe files
5451 Details : When selecting which OSTs to stripe files over, for files with
5452 a stripe count that divides evenly into the number of OSTs,
5453 the MDS is always picking the same starting OST for each file.
5454 Return the OST selection heuristic to the original design.
5459 Description: mount failures may take full timeout to return an error
5460 Details : Under some heavy load conditions it is possible that a
5461 failed mount can wait for the full obd_timeout interval,
5462 possibly several minutes, before reporting an error.
5463 Instead return an error as soon as the status is known.
5465 ------------------------------------------------------------------------------
5467 2006-02-14 Cluster File Systems, Inc. <info@clusterfs.com>
5469 * WIRE PROTOCOL CHANGE. This version of Lustre networking WILL NOT
5470 INTEROPERATE with older versions automatically. Please read the
5471 user documentation before upgrading any part of a live system.
5472 * WARNING: Lustre networking configuration changes are required with
5473 this release. See https://bugzilla.clusterfs.com/show_bug.cgi?id=10052
5476 * Support for kernels:
5477 2.6.9-22.0.2.EL (RHEL 4)
5478 2.6.5-7.244 (SLES 9)
5479 2.6.12.6 vanilla (kernel.org)
5482 Severity : enhancement
5483 Bugzilla : 7981/8208
5484 Description: Introduced Lustre Networking (LNET)
5485 Details : LNET is new networking infrastructure for Lustre, it includes
5486 a reorganized network configuration mode (see the user
5487 documentation for full details) as well as support for routing
5488 between different network fabrics. Lustre Networking Devices
5489 (LNDs) for the supported network fabrics have also been
5490 created for this new infrastructure.
5492 Severity : enhancement
5493 Description: Introduced Access control lists
5494 Details : clients can set ACLs on files and directories in order to have
5495 more fine-grained permissions than the standard Unix UGO+RWX.
5496 The MDS must be started with the "-o acl" mount option.
5498 Severity : enhancement
5499 Description: Introduced filesystem quotas
5500 Details : Administrators may now establish per-user quotas on the
5503 Severity : enhancement
5505 Description: Configuration change for the XT3
5506 The PTLLND is now used to run Lustre over Portals on the XT3
5507 The configure option(s) --with-cray-portals are no longer used.
5508 Rather --with-portals=<path-to-portals-includes> is used to
5509 enable building on the XT3. In addition to enable XT3 specific
5510 features the option --enable-cray-xt3 must be used.
5515 Description: Running on many-way SMP OSTs can trigger oops in llcd_send()
5516 Details : A race between allocating a new llcd and re-getting the llcd_lock
5517 allowed another thread to grab newly-allocated llcd.
5519 Severity : enhancement
5521 Description: 2.6 OST async journal commit and locking fix to improve performance
5522 Details : The filter_direct_io()+filter_commitrw_write() journal commits for
5523 2.6 kernels are now async as they already were in 2.4 kernels so
5524 that they can commit concurrently with the network bulk transfer.
5525 For block-allocated files the filter allocation semaphore is held
5526 to avoid filesystem fragmentation during allocation. BKL lock
5527 removed for 2.6 xattr operations where it is no longer needed.
5532 Description: lconf incorrectly determined whether two IP networks could talk
5533 Details : In some more complicated routing and multiple-network
5534 configurations, lconf will avoid trying to make a network
5535 connection to a disjoint part of the IP space. It was doing the
5536 math incorrectly for one set of cases.
5541 Description: Fix for potential infinite loop processing records in an llog.
5542 Details : If an llog record is corrupted/zeroed, it is possible to loop
5543 forever in llog_process(). Validate the llog record length
5544 and skip the remainder of the block on error.
5547 Frequency : occasional (liblustre only)
5549 Description: liblustre could not open files whose last component is a symlink
5550 Details : sysio_path_walk() would incorrectly pass the open intent to
5551 intermediate path components.
5554 Frequency : rare (liblustre only with non-standard tuning)
5555 Bugzilla : 7201 (7350)
5556 Description: Tuning the MDC DLM LRU size to zero triggers client LASSERT
5557 Details : llu_lookup_finish_locks() tries to set lock data on a lock
5558 after it has been released, only do this for referenced locks
5560 Severity : enhancement
5562 Description: specifying an (invalid) directory default stripe_size of -1
5563 would reset the directory default striping
5564 Details : stripe_size -1 was used internally to signal directory stripe
5565 removal, now use "all default" to signal dir stripe removal
5566 as a directory striping of "all default" is not useful
5569 Frequency : common for large clusters running liblustre clients
5571 Description: doing an ls when liblustre clients are running is slow
5572 Details : sending a glimpse AST to a liblustre client waits for every AST
5573 to time out, as liblustre clients will not respond. Since they
5574 cannot cache data we refresh the OST lock LVB from disk instead.
5576 Severity : enhancement
5578 Description: doing an ls at the same time as file IO can be slow
5579 Details : enqueue and other "small" requests can be blocked behind many
5580 large IO requests. Create a new OST IO portal for non-IO
5581 requests so they can be processed faster.
5584 Frequency : rare (only HPUX clients mounting unsupported re-exported NFS vol)
5586 Description: an HPUX NFS client would get -EACCESS when ftruncate()ing a newly
5587 created file with mode 000
5588 Details : the Linux NFS server relies on an MDS_OPEN_OWNEROVERRIDE hack to
5589 allow an ftruncate() as a non-root user to a file with mode 000.
5590 Lustre now respects this flag to disable mode checks when
5591 truncating a file owned by the user
5594 Frequency : liblustre-only, when liblustre client dies unexpectedly or becomes
5597 Description: Revoking locks from clients that went dead or catatonic might take
5599 Details : New lock flags FL_CANCEL_ON_BLOCK used by liblustre makes
5600 cancellation of such locks instant on servers without waiting for
5601 any reply from clients. Clients drops these locks when cancel
5602 notification from server is received without replying.
5605 Frequency : liblustre-only, when liblustre client dies or becomes busy
5607 Description: Doing ls on Linux clients can take a long time with active
5609 Details : Liblustre client cannot handle ASTs in timely manner, so avoid
5610 granting such locks to it in the first place if possible. Locks
5611 are taken by proxy on the OST during the read or write and
5612 dropped immediately afterward. Add connect flags handling, do
5613 not grant locks to liblustre clients for glimpse ASTs.
5615 Severity : enhancement
5617 Description: Improve read-ahead algorithm to avoid excessive IO for random reads
5618 Details : Existing read-ahead algorithm is tuned for the case of streamlined
5619 sequential reads and behaves badly with applications doing random
5620 reads. Improve it by reading ahead at least read region, and
5621 avoiding excessive large RPC for small reads.
5623 Severity : enhancement
5625 Description: Creating more than 1000 files for a single job may cause a load
5626 imbalance on the OSTs if there are also a large number of OSTs.
5627 Details : qos_prep_create() uses an OST index reseed value that is an
5628 even multiple of the number of available OSTs so that if the
5629 reseed happens in the middle of the object allocation it will
5630 still utilize the OSTs as uniformly as possible.
5635 Description: OST or MDS may oops in ping_evictor_main()
5636 Details : ping_evictor_main() drops obd_dev_lock if deleting a stale export
5637 but doesn't restart at beginning of obd_exports_timed list
5640 Severity : enhancement
5642 Description: improve by-nid export eviction on the MDS and OST
5643 Details : allow multiple exports with the same NID to be evicted at one
5644 time without re-searching the exports list.
5647 Frequency : rare, only with supplementary groups enabled on SMP 2.6 kernels
5649 Description: MDS may oops in groups_free()
5650 Details : in rare race conditions a newly allocated group_info struct is
5651 freed again, and this can be NULL. The 2.4 compatibility code
5652 for groups_free() checked for a NULL pointer, but 2.6 did not.
5655 Frequency : common for liblustre clients doing little filesystem IO
5656 Bugzilla : 9352, 7313
5657 Description: server may evict liblustre clients accessing contended locks
5658 Details : if a client is granted a lock or receives a completion AST
5659 with a blocking AST already set it would not reply to the AST
5660 for LDLM_FL_CANCEL_ON_BLOCK locks. It now replies to such ASTs.
5663 Frequency : lfs setstripe, only systems with more than 160 OSTs
5665 Description: unable to set striping with a starting offset beyond OST 160
5666 Details : llapi_create_file() incorrectly limited the starting stripe
5667 index to the maximum single-file stripe count.
5670 Frequency : LDAP users only
5672 Description: lconf did not handle in-kernel recovery with LDAP properly
5673 Details : lconf/LustreDB get_refs() is searching the wrong namespace
5675 Severity : enhancement
5677 Description: bind OST threads to NUMA nodes to improve performance
5678 Details : all OST threads are uniformly bound to CPUs on a single NUMA
5679 node and do their allocations there to localize memory access
5681 Severity : enhancement
5683 Description: llmount can determine client NID directly from Myrinet (GM)
5684 Details : the client NID code from gmnalnid was moved directly into
5685 llmount, removing the need to use this or specifying the
5686 client NID explicitly when mounting GM clients with zeroconf
5689 Frequency : if client is started with down MDS
5691 Description: if client is started with down MDS mount hangs in ptlrpc_queue_wait
5692 Details : Having an LWI_INTR() wait event (interruptible, but no timeout)
5693 will wait indefinitely in ptlrpc_queue_wait->l_wait_event() after
5694 ptlrpc_import_delayed_req() because we didn't check if the
5695 request was interrupted, and we also didn't break out of the
5696 event loop if there was no timeout
5701 Description: data loss during non-page-aligned writes to a single file from
5702 both multiple nodes and multiple threads on one node at same time
5703 Details : updates to KMS and lsm weren't protected by common lock. Resulting
5704 inconsistency led to false short-reads, that were cached and later
5705 used by ->prepare_write() to fill in partially written page,
5706 leading to data loss.
5709 Frequency : always, if lconf --abort_recovery used
5711 Description: lconf --abort_recovery fails with 'Operation not supported'
5712 Details : lconf was attempting to abort recovery on the MDT device and not
5715 Severity : enhancement
5717 Description: remove cleanup logs
5718 Details : replace lconf-generated cleanup logs with lustre internal
5719 cleanup routines. Eliminates the need for client-cleanup and
5722 Severity : enhancement
5724 Description: add support for EAs (user and system) on lustre filesystems
5725 Details : it is now possible to store extended attributes in the Lustre
5726 client filesystem, and with the user_xattr mount option it
5727 is possible to allow users to store EAs on their files also
5729 Severity : enhancement
5731 Description: Add possibility (config option) to show minimal available OST free
5733 Details : When compiled with --enable-mindf configure option, statfs(2)
5734 (and so, df) will return least minimal free space available from
5735 all OSTs as amount of free space on FS, instead of summary of
5736 free spaces of all OSTs.
5738 Severity : enhancement
5740 Description: do not expand extent locks acquired on OST-side
5741 Details : Modify ldlm_extent_policy() to not expand local locks, acquired
5742 by server: they are not cached anyway.
5745 Frequency : when mmap is used/binaries executed from Lustre
5747 Description: Unmmap pages before throwing them away from read cache.
5748 Details : llap_shrink cache now attempts to unmap pages before discarding
5749 them (if unmapping failed - do not discard). SLES9 kernel has
5750 extra checks that trigger if this unmapping is not done first.
5755 Description: lconf didn't resolve symlinks before checking to see whether a
5756 given mountpoint was already in use
5759 Frequency : when migrating failover services
5760 Bugzilla : 6395, 9514
5761 Description: When migrating a subset of services from a node (e.g. failback
5762 from a failover service node) the remaining services would
5763 time out and evict clients.
5764 Details : lconf --force (implied by --failover) sets the global obd_timeout
5765 to 5 seconds in order to quickly disconnect, but this caused
5766 other RPCs to time out too quickly. Do not change the global
5767 obd_timeout for force cleanup, only set it for DISCONNECT RPCs.
5769 Severity : enhancement
5770 Frequency : if MDS is started with down OST
5771 Bugzilla : 9439,5706
5772 Description: Allow startup/shutdown of an MDS without depending on the
5773 availability of the OSTs.
5774 Details : Asynchronously call mds_lov_synchronize during MDS startup.
5775 Add appropriate locking and lov-osc refcounts for safe
5776 cleaning. Add osc abort_inflight calls in case the
5777 synchronize never started.
5780 Frequency : occasional (Cray XT3 only)
5782 Description: root not authorized to access files in CRAY_PORTALS environment
5783 Details : The client process capabilities were not honoured on the MDS in
5784 a CRAY_PORTALS/CRAY_XT3 environment. If the file had previously
5785 been accessed by an authorized user then root was able to access
5786 the file on the local client also. The root user capabilities
5787 are now allowed on the MDS, as this environment has secure UID.
5790 Frequency : occasional
5792 Description: ldiskfs "too long searching" message happens too often
5793 Details : A debugging message (otherwise harmless) prints too often on
5794 the OST console. This has been reduced to only happen when
5795 there are fragmentation problems on the filesystem.
5800 Description: Division by zero in statfs when all OSCs are inactive
5801 Details : lov_get_stripecnt() returns zero due to incorrect order of checks,
5802 lov_statfs divides by value returned by lov_get_stripecnt().
5806 Bugzilla : 9489, 3273
5807 Description: First write from each client to each OST was only 4kB in size,
5808 to initialize client writeback cache, which caused sub-optimal
5809 RPCs and poor layout on disk for the first writen file.
5810 Details : Clients now request an initial cache grant at (re)connect time
5811 and so that they can start streaming writes to the cache right
5812 away and always do full-sized RPCs if there is enough data.
5813 If the OST is rebooted the client also re-establishes its grant
5814 so that client cached writes will be honoured under the grant.
5819 Description: Slow ls (and stat(2) syscall) on files residing on IO-loaded OSTs
5820 Details : Now I/O RPCs go to different portal number and (presumably) fast
5821 lock requests (and glimses) and other RPCs get their own service
5822 threads pool that should be able to service those RPCs
5825 Severity : enhancement
5827 Description: Ability to exchange lustre version between client and servers and
5828 issue warnings at client side if client is too old. Also for
5829 liblustre clients there is ability to refuse connection of too old
5831 Details : New 'version' field is added to connect data structure that is
5832 filled with version info. That info is later checked by server and
5836 Frequency : rare, liblustre only.
5837 Bugzilla : 9296, 9581
5838 Description: Two simultaneous writes from liblustre at offset within same page
5839 might proceed at the same time overwriting eachother with stale
5841 Details : I/O lock withing llu_file_prwv was released too early, before data
5842 actually was hitting the wire. Extended lock-holding time until
5843 server acknowledges receiving data.
5846 Frequency : extremely rare. Never observed in practice.
5848 Description: avoid generating lustre_handle cookie of 0.
5849 Details : class_handle_hash() generates handle cookies by incrementing
5850 global counter, and can hit 0 occasionaly (this is unlikely, but
5851 not impossible, because initial value of cookie counter is
5852 selected randonly). Value of 0 is used as a sentinel meaning
5853 "unassigned handle" --- avoid it. Also coalesce two critical
5854 sections in this function into one.
5856 Severity : enhancement
5858 Description: allow liblustre clients to delegate truncate locking to OST
5859 Details : To avoid overhead of locking, liblustre client instructs OST to
5860 take extent lock in ost_punch() on client's behalf. New connection
5861 flag is added to handle backward compatibility.
5863 Severity : enhancement
5864 Bugzilla : 4928, 7341, 9758
5865 Description: allow number of OST service threads to be specified
5866 Details : a module parameter allows the number of OST service threads
5867 to be specified via "options ost ost_num_threads={N}" in the
5868 OSS's /etc/modules.conf or /etc/modprobe.conf.
5872 Bugzilla : 6146, 9635, 9895
5873 Description: servers crash with bad pointer in target_handle_connect()
5874 Details : In rare cases when a client is reconnecting it was possible that
5875 the connection request was the last reference for that export.
5876 We would temporarily drop the export reference and get a new
5877 one, but this may have been the last reference and the export
5878 was just destroyed. Get new reference before dropping old one.
5880 Severity : enhancement
5881 Frequency : if client is started with failover MDS
5883 Description: Allow multiple MDS hostnames in the mount command
5884 Details : Try to read the configuration from all specified MDS
5885 hostnames during a client mount in case the "primary"
5888 Severity : enhancement
5890 Description: Stop sending data to evicted clients as soon as possible.
5891 Details : Check if the client we are about to send or are sending data to
5892 was evicted already. (Check is done every second of waiting,
5893 for which l_wait_event interface was extended to allow checking
5894 of exit condition at specified intervals).
5897 Frequency : rare, normally only when NFS exporting is done from client
5899 Description: 'bad disk LOV MAGIC: 0x00000000' error when chown'ing files
5901 Details : Make mds_get_md() recognise empty md case and set lmm size to 0.
5904 Frequency : always, if srand() is called before liblustre initialization
5906 Description: Liblustre uses system PRNG disturbing its usage by user application
5907 Details : Introduce internal to lustre fast and high-quality PRNG for
5908 lustre usage and make liblustre and some other places in generic
5909 lustre code to use it.
5911 Severity : enhancement
5912 Bugzilla : 9477, 9557, 9870
5913 Description: Verify that the MDS configuration logs are updated when xml is
5914 Details : Check if the .xml configuration logs are newer than the config
5915 logs stored on the MDS and report an error if this is the case.
5916 Request --write-conf, or allow starting with --old_conf.
5918 Severity : enhancement
5920 Description: Handle symlinks in the path when checking if Lustre is mounted.
5921 Details : Resolve intermediate symlinks when checking if a client has
5922 mounted a filesystem to avoid duplicate client mounts.
5927 Description: lconf can hit an error exception but still return success.
5928 Details : The lconf command catches the Command error exception at the top
5929 level script context and will exit with the associated exit
5930 status, but doesn't ensure that this exit status is non-zero.
5935 Description: failure of ptlrpc thread startup can cause oops
5936 Details : Starting a ptlrpc service thread can fail if there are a large
5937 number of threads or the server memory is very fragmented.
5938 Handle this without oopsing.
5941 Frequency : always, only if liblustre and non-default acceptor port was used
5943 Description: liblustre cannot connect to servers with non-default acceptor port
5944 Details : tcpnal_set_default_params() was not called and was therefore
5945 ignoring the environment varaible TCPNAL_PORT, as well as other
5946 TCPNAL_ environment variables
5951 Description: two objects could be created on the same OST for a single file
5952 Details : If an OST is down, in some cases it was possible to create two
5953 objects on a single OST for a single file. No problems other
5954 than potential performance impact and spurious error messages.
5958 Bugzilla : 5681, 9562
5959 Description: Client may oops in ll_unhash_aliases
5960 Details : Client dcache may become inconsistent in race condition.
5961 In some cases "getcwd" can fail if the current directory is
5967 Description: Inode refcounting problems in NFS export code
5968 Details : link_raw functions used to call d_instantiate without obtaining
5969 extra inode reference first.
5973 Bugzilla : 9942, 9903
5974 Description: Referencing freed requests leading to crash, memleaks with NFS.
5975 Details : We used to require that call to ll_revalidate_it was always
5976 followed by ll_lookup_it. Also with revalidate_special() it is
5977 possible to call ll_revalidate_it() twice for the same dentry
5978 even if first occurence returned success. This fix changes semantic
5979 between DISP_ENQ_COMPLETE disposition flag to mean there is extra
5980 reference on a request referred from the intent.
5981 ll_intent_release() then releases such a request.
5984 Frequency : rare, normally benchmark loads only
5986 Description: unlinked inodes were kept in memory on the client
5987 Details : If a client is repeatedly creating and unlinking files it
5988 can accumulate a lot of stale inodes in the inode slab cache.
5989 If there is no other client load running this can cause the
5990 client node to run out of memory. Instead flush old inodes
5991 from client cache that have the same inode number as a new inode.
5994 Frequency : SLES9 2.6.5 kernel and long filenames only
5995 Bugzilla : 9969, 10379
5996 Description: utime reports stale NFS file handle
5997 Details : SLES9 uses out-of-dentry names in some cases, which confused
5998 the lustre dentry revalidation. Change it to always use the
6002 Frequency : rare, unless heavy write-truncate concurrency is continuous
6003 Bugzilla : 4180, 6984, 7171, 9963, 9331
6004 Description: OST becomes very slow and/or deadlocked during object unlink
6005 Details : filter_destroy() was holding onto the parent directory lock
6006 while truncating+unlinking objects. For very large objects this
6007 may block other threads for a long time and slow overall OST
6008 responsiveness. It may also be possible to get a lock ordering
6009 deadlock in this case, or run out of journal credits because of
6010 the combined truncate+unlink. Solution is to do object truncate
6011 first in one transaction without parent lock, and then do the
6012 final unlink in a new transaction with the parent lock. This
6013 reduces the lock hold time dramatically.
6016 Frequency : rare, 2.4 kernels only
6018 Description: MDS or OST cleanup may trip kernel BUG when dropping kernel lock
6019 Details : mds_cleanup() and filter_cleanup() need to drop the kernel lock
6020 before unmounting their filesystem in order to avoid deadlock.
6021 The kernel_locked() function in 2.4 kernels only checks whether
6022 the kernel lock is held, not whether it is this process that is
6023 holding it as 2.6 kernels do.
6028 Description: MDS or OST may oops/LBUG if a client is connecting multiple times
6029 Details : The client ptlrpc code may be trying to reconnect to a down
6030 server before a previous connection attempt has timed out.
6031 Increase the reconnect interval to be longer than the connection
6032 timeout interval to avoid sending duplicate connections to
6036 Frequency : echo_client brw_test command
6038 Description: fix echo_client to work with OST preallocated code
6039 Details : OST preallocation code (5137) didn't take echo_client IO path
6040 into account: echo_client calls filter methods outside of any
6041 OST thread and, hence, there is no per-thread preallocated
6042 pages and buffers to use. Solution: hijack pga pages for IO. As
6043 a byproduct, this avoids unnecessary data copying.
6047 Bugzilla : 3555, 5962, 6025, 6155, 6296, 9574
6048 Description: Client can oops in mdc_commit_close() after open replay
6049 Details : It was possible for the MDS to return an open request with no
6050 transaction number in mds_finish_transno() if the client was
6051 evicted, but without actually returning an error. Clients
6052 would later try to replay that open and may trip an assertion
6053 Simplify the client close codepath, and always return an error
6054 from the MDS in case the open is not successful.
6057 Frequency : rare, 2.6 OSTs only
6059 Description: OST may deadlock under high load on fragmented files
6060 Details : If there was a heavy load and highly-fragmented OST filesystems
6061 it was possible to have all the OST threads deadlock waiting on
6062 allocation of biovecs, because the biovecs were not released
6063 until the entire RPC IO was completed. Instead, release biovecs
6064 as soon as they are complete to ensure forward IO progress.
6066 Severity : enhancement
6068 Description: Support for specifying external journal device at mount
6069 Details : If an OST or MDS device is formatted with an external journal
6070 device, this device major/minor is stored in the ext3 superblock
6071 and may not be valid for failover. Allow detecting and
6072 specifying the external journal at mount time.
6077 Description: Mounting an MDS with pending unlinked files may cause oops
6078 Details : target_finish_recovery() calls mds_postrecov() which returned
6079 the number of orphans unlinked. mds_lov_connect->mds_postsetup()
6080 considers this an error and immediately begins cleaning up the
6081 lov, just after starting the mds_lov process
6083 Severity : enhancement
6085 Description: Implement 'lfs df' to report actual free space on per-OST basis
6086 Details : Add sub-command 'df' on 'lfs' to report the disk space usage of
6087 MDS/OSDs. Usage: lfs df [-i][-h]. Command Options: '-i' to report
6088 usage of objects; '-h' to report in human readable format.
6090 ------------------------------------------------------------------------------
6092 2005-08-26 Cluster File Systems, Inc. <info@clusterfs.com>
6099 Description: Mounting an ldiskfs file system with mballoc may crash OST node.
6100 Details : ldiskfs mballoc code may reference an uninitialized buddy struct
6101 at startup during orphan unlinking. Instead, skip buddy update
6102 before setup, as it will be regenerated after recovery is complete.
6107 Description: If an OST is inactive, its locks might reference stale inodes.
6108 Details : lov_change_cbdata() must iterate over all namespaces, even if
6109 they are inactive to clear inode references from the lock.
6111 Severity : enhancement
6112 Frequency : occasional, if non-standard max_dirty_mb used
6114 Description: Client will block write RPCs if not enough grant
6115 Details : If a client has max_dirty_mb smaller than max_rpcs_in_flight,
6116 then the client will block writes while waiting for another RPC
6117 to complete instead of consuming its dirty limit. With change
6118 we get improved performance when max_dirty_mb is small.
6120 Severity : enhancement
6121 Bugzilla : 3389, 6253
6122 Description: Add support for supplementary groups on the MDS.
6123 Details : The MDS has an upcall /proc/fs/lustre/mds/{mds}/group_upcall
6124 (set to /usr/sbin/l_getgroups if enabled) which will do MDS-side
6125 lookups for user supplementary groups into a cache.
6129 Description: O_CREAT|O_EXCL open flags in liblustre always return -EEXIST
6130 Details : Make libsysio to not enforce O_EXCL by clearing the flag,
6131 for liblustre O_EXCL is enforced by MDS.
6135 Description: readdir never returns NULL in liblustre.
6136 Details : Corrected llu_iop_getdirentries logic, to return offset of next
6137 dentry in struct dirent.
6141 Frequency : liblustre only, depends on application IO pattern
6142 Description: liblustre clients evicted if not contacting servers
6143 Details : Don't put liblustre clients into the ping_evictor list, so
6144 they will not be evicted by the pinger ever.
6146 Severity : enhancement
6148 Description: Add ability to evict clients by NID from MDS.
6149 Details : By echoing "nid:$NID" string into
6150 /proc/fs/lustre/mds/.../evict_client client with nid that equals to
6151 $NID would be instantly evicted from this MDS and from all active
6152 OSTs connected to it.
6156 Description: Do not query file size twice, somewhat slowing stat(2) calls.
6157 Details : lookup_it_finish() used to query file size from OSTs that was not
6162 Description: service threads change working directory to that of init
6163 Details : Starting lustre service threads may pin the working directory
6164 of the parent thread, making that filesystem busy. Threads
6165 now change to the working directory of init to avoid this.
6169 Frequency : during shutdown only
6170 Description: shutdown with a failed MDS or OST can cause unmount to hang
6171 Details : Don't resend DISCONNECT messages in ptlrpc_disconnect_import()
6176 Frequency : 2.6 only
6177 Description: chmod/chown may include an extra supplementary group
6178 Details : ll{,u}_mdc_pack_op_data() does not properly initialize the
6179 supplementary group and if none is specified this is used.
6182 Bugzilla : 5479 (6816)
6184 Description: Racing open + rm can assert client in mdc_set_open_replay_data()
6185 Details : If lookup is in progress on a file that is unlinked we might try
6186 to revalidate the inode and fail in revalidate after lookup is
6187 complete and ll_file_open() enqueues the open again but
6188 it_open_error() was not checking DISP_OPEN_OPEN errors correctly.
6191 Frequency : always, if lconf --abort_recovery used
6193 Description: lconf --abort_recovery fails with 'Operation not supported'
6194 Details : lconf was attempting to abort recovery on the MDT device and not
6197 ------------------------------------------------------------------------------
6199 2005-08-08 Cluster File Systems, Inc. <info@clusterfs.com>
6204 Frequency : rare (only unsupported configurations with a node running as an
6206 Bugzilla : 6514, 5137
6207 Description: Mounting a Lustre file system on a node running as an OST could
6209 Details : OSTs now preallocates memory needed to write out data at
6210 startup, instead of when needed, to avoid having to
6211 allocate memory in possibly low memory situations.
6212 Specifically, if the file system is mounted on on OST,
6213 memory pressure could force it to try to write out data,
6214 which it needed to allocate memory to do. Due to the low
6215 memory, it would be unable to do so and the node would
6216 become unresponsive.
6218 Severity : enhancement
6220 Description: Addition of lconf --service command line option
6221 Details : lconf now accepts a '--service <arg>' option, which is
6222 shorthand for 'lconf --group <arg> --select <arg>=<hostname>'
6224 Severity : enhancement
6226 Description: Failover mode is now the default for OSTs.
6227 Details : By default, OSTs will now run in failover mode. To return to
6228 the old behaviour, add '--failout' to the lmc line for OSTs.
6230 Severity : enhancement
6232 Description: Health checks are now provided for MDS and OSTs
6233 Details : Additional detailed health check information on MSD and OSTs
6234 is now provided through the procfs health_check value.
6237 Frequency : occasional, depends on IO load
6239 Description: Disk fragmentation on the OSTs could eventually cause slowdowns
6240 after numerous create/delete cycles
6241 Details : The ext3 inode allocation policy would not allocate new inodes
6242 very well on the OSTs because there are no new directories
6243 being created. Instead we look for groups with free space if
6244 the parent directories are nearly full.
6249 Description: Network or server problems during mount may cause partially
6250 mounted clients instead of returning an error.
6251 Details : The config llog parsing code may overwrite the error return
6252 code during mount error handling, returning success instead
6258 Description: MDS can fail to allocate large reply buffers
6259 Details : After long uptimes the MDS can fail to allocate large reply
6260 buffers (e.g. zconf client mount config records) due to memory
6261 fragmentation or consumption by the buffer cache. Preallocate
6262 some large reply buffers so that these replies can be sent even
6263 under memory pressure.
6267 Frequency : rare (liblustre)
6268 Description: fsx running with liblustre complained that using truncate() to
6269 extend the file doesn't work. This patch corrects that issue.
6270 Details : This is the liblustre equivalent of the fix for bug 6196. Fixes
6271 ATTR_SIZE and lsm use in llu_setattr_raw.
6275 Frequency : rare, only 2.6 kernels
6276 Description: Unusual file access patterns on the MDS may result in inode
6277 data being lost in very rare circumstances.
6278 Details : Bad interaction between the ea-in-inode patch and the "no-read"
6279 code in the 2.6 kernel caused the inode and/or EA data not to
6280 be read from disk, causing single-file corruption.
6284 Frequency : rare, only 2.6 filesystems using extents
6285 Description: Heavy concurrent write and delete load may cause data corruption.
6286 Details : It was possible under high-load situations to have an extent
6287 metadata block in the block device cache from a just-unlinked
6288 file overwrite a newly-allocated data block. We now unmap any
6289 metadata buffers that alias just-allocated data blocks.
6293 Frequency : filesystems with default stripe_count larger than 77
6294 Description: lconf+mke2fs fail when formatting filesystem with > 77 stripes
6295 Details : lconf specifies an inode size of 4096 bytes when the default
6296 stripe_count is larger than 77. This conflicts with the default
6297 inode density of 1 per 4096 bytes. Allocate smaller inodes in
6298 this case to avoid pinning too much memory for large EAs.
6300 ------------------------------------------------------------------------------
6302 2005-07-07 Cluster File Systems, Inc. <info@clusterfs.com>
6307 Frequency : rare (extremely heavy IO load with hundreds of clients)
6309 Description: Client is evicted, gets IO error writing to file
6310 Details : lock ordering changes for bug 5492 reintroduced bug 3267 and
6311 caused clients to be evicted for AST timeouts. The fixes in
6312 bug 5192 mean we no longer need to have such short AST timeouts
6313 so ldlm_timeout has been increased.
6316 Frequency : occasional during --force or --failover shutdown under load
6317 Bugzilla : 5949, 4834
6318 Description: Server oops/LBUG if stopped with --force or --failover under load
6319 Details : a collection of import/export refcount and cleanup ordering
6320 issues fixed for safer force cleanup
6323 Frequency : only filesystems larger than 120 OSTs
6324 Bugzilla : 5990, 6223
6325 Description: lfs getstripe would oops on a very large filesystem
6326 Details : lov_getconfig used kfree on vmalloc'd memory
6329 Frequency : only filesystems exporting via NFS to Solaris 10 clients
6330 Bugzilla : 6242, 6243
6331 Description: reading from files that had been truncated to a non-zero size
6332 but never opened returned no data
6333 Details : ll_file_read() reads zeros from no-object files to EOF
6338 Description: A bug in MDS/OSS recovery could cause the OSS to fail an assertion
6339 Details : There's little harm in aborting MDS/OSS recovery and letting it
6340 try again, so I removed the LASSERT and return an error instead.
6342 Severity : enhancement
6344 Description: New debugging infrastructure for tracking down data corruption
6345 Details : The I/O checksum code was replaced to: (a) control it at runtime,
6346 (b) cover more of the client-side code path, and (c) try to narrow
6347 down where problems occurred
6351 Bugzilla : 3819, 4364, 4397, 6313
6352 Description: Racing close and eviction MDS could cause assertion in mds_close
6353 Details : It was possible to get multiple mfd references during close and
6354 client eviction, leading to one thread referencing a freed mfd.
6356 Severity: : enhancement
6357 Bugzilla : 3262, 6359
6358 Description: Attempts to reconnect to servers are now more aggressive.
6359 Details : This builds on the enhanced upcall-less recovery that was added
6360 in 1.4.2. When trying to reconnect to servers, clients will
6361 now try each server in the failover group every 10 seconds. By
6362 default, clients would previously try one server every 25 seconds.
6367 Description: After recovery, certain operations trigger a failed
6368 assertion on a client.
6369 Details : Failing over an mds, using lconf -d --failover, while a
6370 client was doing a readdir() call would cause the client to
6371 LBUG after recovery completed and the readdir() was resent.
6373 Severity : enhancement
6375 Description: Default groups are now added by lconf
6376 Details : You can now run lconf --group <servicename> without having to
6377 manually add groups with lmc.
6380 Frequency : occasional
6382 Description: Nodes with an elan id of 0 trigger a failed assertion
6385 Frequency : always when accessing e.g. tty/console device nodes
6387 Description: tty and some other devices nodes cannot be used on lustre
6388 Details : file's private_data field is used by device data and lustre
6389 values in there got lost. New field was added to struct file to
6390 store fs-specific private data.
6393 Frequency : when exporting Lustre via NFS
6395 Description: NFSD failed occasionally when looking up a path component
6396 Details : NFSD is looking up ".." which was broken in ext3 directories
6397 that had grown large enough to become hashed.
6400 Frequency : Clusters with multiple interfaces not on the same subnet
6402 Description: Nodes will repeatedly try to reconnect to an interface which it
6403 cannot reach and report an error to the log.
6404 Details : Extra peer list entries will be created by lconf with some peers
6405 unreachable. lconf now validates the peer before adding it.
6408 Frequency : Only if a default stripe is set on the filesystem root.
6410 Description: Setting a default stripe on the filesystem root prevented the
6411 filesystem from being remounted.
6412 Details : The client was sending extra request flags in the root getattr
6413 request and did not allocate a reply buffer for the dir EA.
6416 Frequency : occasional, higher if lots of files are accessed by one client
6417 Bugzilla : 6159, 6097
6418 Description: Client trips assertion regarding lsm mismatch/magic
6419 Details : While revalidating inodes the VFS looks up inodes with ifind()
6420 and in rare cases can find an inode that is being freed.
6421 The ll_test_inode() code will free the lsm during ifind()
6422 when it finds an existing inode and then the VFS later attaches
6423 this free lsm to a new inode.
6427 Bugzilla : 6422, 7030
6428 Description: MDS deadlock between mkdir and client eviction
6429 Details : Creating a new file via mkdir or mknod (starting a transaction
6430 and getting the ns lock) can deadlock with client eviction
6431 (gets ns lock and trying to finish a synchronous transaction).
6434 Frequency : occasional
6435 Description: While starting a server, the fsfilt_ext3 module could not be
6437 Details : CFS's improved ext3 filesystem is named ldiskfs for 2.6
6438 kernels. Previously, lconf would still use the ext3 name
6439 when trying to load modules. Now, it will correctly use
6440 ext3 on 2.4 and ldiskfs on 2.6.
6442 Severity : enhancement
6443 Description: The default stripe count has been changed to 1
6444 Details : The interpretation of the default stripe count (0, to lfs
6445 or lmc) has been changed to mean striping across a single
6446 OST, rather than all available. For general usage we have
6447 found a stripe count of 1 or 2 works best.
6449 Severity : enhancement
6450 Description: Add support for compiling against Cray portals.
6451 Details : Conditional compiling for some areas that are different
6455 Frequency : occasional
6456 Bugzilla : 6409, 6834
6457 Description: Creating files with an explicit stripe count may lead to
6458 a failed assertion on the MDS
6459 Details : If some OSTs are full or unavailable, creating files may
6460 trigger a failed assertion on the MDS. Now, Lustre will
6461 try to use other servers or return an error to the
6465 Frequency : occasional
6467 Description: Multiple concurrent overlapping read+write on multiple SMP nodes
6468 caused lock timeout during readahead (since 1.4.2).
6469 Details : Processes doing readahead might match a lock that hasn't been
6470 granted yet if there are overlapping and conflicting lock
6471 requests. The readahead process waits on ungranted lock
6472 (original lock is CBPENDING), while OST waits for that process
6473 to cancel CBPENDING read lock and eventually evicts client.
6475 Severity : enhancement
6477 Description: Initial enabling of flock support for clients
6478 Details : Implements fcntl advisory locking and file status functions.
6479 This feature is provided as an optional mount flag (default
6480 off), and is NOT CURRENTLY SUPPORTED. Not all types of record
6481 locking are implemented yet, and those that are are not guaranteed
6482 to be completely correct in production environments.
6483 mount -t lustre -o [flock|noflock] ...
6486 Frequency : occasional
6488 Description: OSTs running 2.4 kernels but with extents enabled might trip an
6489 assertion in the ext3 JBD (journaling) layer.
6490 Details : The b_committed_data struct is protected by the big kernel lock
6491 in 2.4 kernels, serializing journal_commit_transaction() and
6492 ext3_get_block_handle->ext3_new_block->find_next_usable_block()
6493 access to this struct. In 2.6 kernels there is finer grained
6494 locking to improve SMP performance of the JBD layer.
6498 Description: Changes the "SCSI I/O Stats" kernel patch to default to "enabled"
6500 -----------------------------------------------------------------------------
6502 2005-05-05 Cluster File Systems, Inc. <info@clusterfs.com>
6504 NOTE: Lustre 1.4.2 uses an incompatible network protocol than previous
6505 versions of Lustre. Please update all servers and clients to
6506 version 1.4.2 or later at the same time. You must also run
6507 "lconf --write-conf {config}.xml" on the MDS while it is stopped
6508 to update the configuration logs.
6510 - fix for HPUX NFS client breakage when NFS exporting Lustre (5781)
6511 - mdc_enqueue does not need max_mds_easize request buffer on send (5707)
6512 - swab llog records of type '0' so we get proper header size/idx (5861)
6513 - send llog cancel req to DLM cancel portal instead of cb portal (5515)
6514 - fix rename of one directory over another leaking an inode (5953)
6515 - avoid SetPageDirty on 2.6 (5981)
6516 - don't re-add just-being-destroyed locks to the waiting list (5653)
6517 - when creating new directories, inherit the parent's custom
6518 striping settings if present parent (3048)
6519 - flush buffers from cache before direct IO in 2.6 obdfilter (4982)
6520 - don't hold i_size_sem in ll_nopage() and ll_ap_refresh_count (6077)
6521 - don't hold client locks on temporary worklist from l_lru (5666)
6522 - handle IO errors in 2.6 obdfilter bio completion routine (6046)
6523 - automatically evict dead clients (5921)
6524 - Update file size properly in create+truncate+fstat case (6196)
6525 - Do not unhash mountpoint dentries, do not allow removal of
6527 - Avoid lock ordering deadlock issue with write/truncate (6203,5654)
6528 - reserve enough journal credits in fsfilt_start_log for setattr (4554)
6529 - ldlm_enqueue freed-export error path would always LBUG (6149,6184)
6530 - don't reference lr_lvb_data until after we hold lr_lvb_sem (6170)
6531 - don't overwrite last_rcvd if there is a *_client_add() error (6086)
6532 - Correctly handle reads of files with no objects (6243)
6533 - lctl recover will also mark a device active if deactivate used (5933)
6535 - by default create 1 inode per 4kB space on MDS, per 16kB on OSTs
6536 - allow --write-conf on an MDS with different nettype than client (5619)
6537 - don't write config llogs to MDS for mounts not from that MDS (5617)
6538 - lconf should create multiple TCP connections from a client (5201)
6539 - init scripts are now turned off by default; run chkconfig --on
6540 lustre and chkconfig --on lustrefs to use them
6541 - upcalls are no longer needed for clients to recover to failover
6543 - add --abort-recovery option to lconf to abort recovery on device
6545 - add support for an arbitrary number of OSTs (3026)
6546 - Quota support protocol changes.
6547 - forward compatibility changes to wire structs (6007)
6548 - rmmod NALs that might be loaded because of /etc/modules.conf (6133)
6549 - support for mountfsoptions and clientoptions to the Lustre LDAP (5873)
6550 - improved "lustre status" script
6551 - initialize blocksize for non-regular files (6062)
6552 - added --disable-server and --disable-client configure options (5782)
6553 - introduce a lookup cache for lconf to avoid repeated DB scans (6204)
6554 - Vanilla 2.4.29 support
6555 - increase maximum number of obd devices to 520 (6242)
6556 - remove the tcp-zero-copy patch from the suse-2.4 series (5902)
6557 - Quadrics Elan drivers are now included for the RHEL 3 2.4.21 and
6558 SLES 9 2.6.5 kernels
6559 - limit stripes per file to 160 (the maximum EA size) (6093)
6561 2005-03-22 Cluster File Systems, Inc. <info@clusterfs.com>
6564 - don't LASSERT in ll_release on NULL lld with NFS export (4655, 5760)
6565 - hold NS lock when calling handle_ast_error->del_waiting_lock (5746)
6566 - fix setattr mtime regression from lovcleanup merge (4829, 5669)
6567 - workaround for 2.6 crash in ll_unhash_aliases (5687, 5210)
6568 - small ext3 extents cleanups and fixes (5733)
6569 - improved mballoc code, several small races and bugs fixed (5733, 5638)
6570 - kernel version 43 - fix remove_suid bugs in both 2.4 and 2.6 (5695)
6571 - avoid needless client->OST connect, fix handle mismatch (5317)
6572 - fix DLM error path that led to out-of-sync client, long delays (5779)
6573 - support common vfs-enforced mount options (nodev,nosuid,noexec) (5637)
6574 - fix several locking issues related to i_size (5492,5624,5654,5672)
6575 - don't move pending lock onto export if it is already evicted (5683)
6576 - fix kernel oops when creating .foo in unlinked directory (5548)
6577 - fix deadlock in obdfilter statistics vs. object create (5811)
6578 - use time_{before,after} to avoid timer jiffies wrap (5882)
6579 - shutdown --force/--failover stability (3607,3651,4797,5203,4834)
6580 - Do not leak request if server was not able to process it (5154)
6581 - If mds_open unable to find parent dir, make that negative lookup(5154)
6582 - don't create new directories with extent-mapping (5909, 5936)
6584 - fix lustre/lustrefs init scripts for SuSE (patch from Scali, 5702)
6585 - don't hold the pinger_sem in ptlrpc_pinger_sending_on_import
6586 - change obd_increase_kms to obd_adjust_kms (up or down) (5654)
6587 - lconf, lmc search both /usr/lib and /usr/lib64 for Python libs (5800)
6588 - support for RHEL4 kernel on i686 (5773)
6589 - provide error messages when incompatible logs are encountered (5898)
6591 2005-02-18 Cluster File Systems, Inc. <info@clusterfs.com>
6592 * version 1.4.0.10 (1.4.1 release candidate 1)
6594 - don't keep a lock reference when lock is not granted (4238)
6595 - unsafe list practices (rarely) led to infinite eviction loop (4908)
6596 - add per-fs limit of Lustre pages in page cache, avoid OOM (4699)
6597 - drop import inflight refcount on signal_completed_replay error (5255)
6598 - unlock page after async write error during send (3677)
6599 - handle missing objects in filter_preprw_read properly (5265)
6600 - no transno return for symlink open, don't save no-trasno open (3440)
6601 - don't try to complete elan receive that already failed (4012)
6602 - free RPC server reply state on error (5406)
6603 - clean up thread from ptlrpc_start_thread() on error (5160)
6604 - readahead could read extra page into cache that wasn't ejected (5388)
6605 - prevent races in class_attach/setup/cleanup/detach (5260)
6606 - don't dereference de->d_inode after l_dput of de (5458)
6607 - use "int" for stripe value returned from lock_to_stripe (5544)
6608 - mballoc allocation and error-checking fixes in 2.6 (5504)
6609 - block device patches to fix I/O request sizes in 2.6 (5482)
6610 - look up hostnames for IB nals (5602)
6611 - 2.6 changed lock ordering of 2 semaphores, caused deadlock (5654)
6612 - don't start multiple acceptors for the same port (5277)
6613 - fix incorrect LASSERT in mds_getattr_name (5635)
6614 - export a proc file for general "ping" checking (5628)
6615 - fix "lfs check" to not block when the MDS is down (5628)
6617 - service request history (4965)
6618 - put {ll,lov,osc}_async_page structs in a single slab (4699)
6619 - create an "evict_client" /proc entry on OSTs, like the MDS has
6620 - fix mount usage message, return errors per mount(8) (5168)
6621 - change grep [] to grep "[]" in tests so they work in more UMLs
6622 - fix ppc64/x86_64 spec to use %{_libdir} instead of /usr/lib (5389)
6623 - remove ancient LOV_MAGIC_V0 EA support (5047)
6624 - add "disk I/Os in flight" and "I/O req time" stats in obdfilter
6625 - align r/w RPCs to PTLRPC_MAX_BRW_SIZE boundary for performance (3451)
6626 - allow readahead allocations to fail when low on memory (5383)
6627 - mmap locking landed again, after considerable improvement (2828)
6628 - add get_hostaddr() to lustreDB.py for LDAP support (5459)
6630 2004-11-23 Cluster File Systems, Inc. <info@clusterfs.com>
6633 - send OST transaction number in read/write reply to free req (4966)
6634 - don't ASSERT in ptl_send_rpc() if we run out of memory (5119)
6635 - lock /proc/sys/portals/routes internal state, avoiding oops (4827)
6636 - the watchdog thread now runs as interruptible (5246)
6637 - flock/lockf fixes (but it's still disabled, pending 5135)
6638 - don't use EXT3 constants in llite code (5094)
6639 - memory shortage at startup could cause assertion (5176)
6641 - reorganization of lov code
6642 - single portals codebase
6644 - add extents/mballoc support (5025)
6645 - direct I/O reads in the obdfilter (4048)
6646 - kernel patches from LNXI for 2.6 (bluesmoke, perfctr, mtd, kexec)
6648 tbd Cluster File Systems, Inc. <info@clusterfs.com>
6651 - send OST transaction number in read/write reply to free req (4966)
6652 - don't ASSERT in ptl_send_rpc() if we run out of memory (5119)
6653 - lock /proc/sys/portals/routes internal state, avoiding oops (4827)
6654 - the watchdog thread now runs as interruptible (5246)
6655 - handle missing objects in filter_preprw_read properly (5265)
6656 - unsafe list practices (rarely) led to infinite eviction loop (4908)
6657 - drop import inflight refcount on signal_completed_replay error (5255)
6658 - unlock page after async write error during send (3677)
6659 - return original error code on reconstructed replies (3761)
6660 - no transno return for symlink open, don't save no-trasno open (3440)
6662 - add pid to ldlm debugging output (4922)
6663 - bump the watchdog timeouts -- we can't handle 30sec yet
6664 - extra debugging for orphan dentry/inode bug (5259)
6666 2004-11-16 Cluster File Systems, Inc. <info@clusterfs.com>
6669 - fix TCP_NODELAY bug, which caused extreme perf regression (5134)
6670 - allocate qswnal tx descriptors singly to avoid fragmentation (4504)
6671 - don't LBUG on obdo_alloc() failure, use OBD_SLAB_ALLOC() (4800)
6672 - fix NULL dereference in /proc/sys/portals/routes (4827)
6673 - allow failed mdc_close() operations to be interrupted (4561)
6674 - stop precreate on OST before MDS would time out on it (4778)
6675 - don't send partial-page writes before EOF from client (4410)
6676 - discard client grant for sub-page writes on large-page clients (4520)
6677 - don't free dentries not owned by NFS code, check generation (4806)
6678 - fix lsm leak if mds_create_objects() fails (4801)
6679 - limit debug_daemon file size, always print CERROR messages (4789)
6680 - use transno after validating reply (3892)
6681 - process timed out requests if import state changes (3754)
6682 - update mtime on OST during writes, return in glimpse (4829)
6683 - add mkfsoptions to LDAP (4679)
6684 - use ->max_readahead method instead of zapping global ra (5039)
6685 - don't interrupt __l_wait_event() during strace
6687 - add software watchdogs to catch hung threads quickly (4941)
6688 - make lustrefs init script start after nfs is mounted
6689 - fix CWARN/ERROR duplication (4930)
6690 - return async write errors to application if possible (2248)
6691 - add /proc/sys/portal/memused (bytes allocated by PORTALS_ALLOC)
6692 - print NAL number in %x format (4645)
6693 - update barely-supported suse-2.4.21-171 series (4842)
6694 - support for sles 9 %post scripts
6695 - support for building 2.6 kernel-source packages
6696 - support for sles km_* packages
6698 2004-10-07 Cluster File Systems, Inc. <info@clusterfs.com>
6701 - ignore -ENOENT errors in osc_destroy (3639)
6702 - notify osc create thread that OSC is being cleaned up (4600)
6703 - add nettype argument for llmount in #5d in conf-sanity.sh (3936)
6704 - reconstruct ost_handle() like mds_handle() (4657)
6705 - create a new thread to do import eviction to avoid deadlock (3969)
6706 - let lconf resolve symlinked-to devices (4629)
6707 - don't unlink "objects" from directory with default EA (4554)
6708 - hold socknal file ref over connect in case target is down (4394)
6709 - allow more than 32000 subdirectories in a single directory (3244)
6710 - fix blocks count for O_DIRECT writes (3751)
6711 - OST returns ENOSPC from object create when no space left (4539)
6712 - don't send truncate RPC if file size isn't changing (4410)
6713 - limit OSC precreate to 1/2 of value OST considers bogus (4778)
6714 - bind to privileged port in socknal and tcpnal (3689)
6716 - rate limit CERROR/CWARN console message to avoid overload (4519)
6717 - GETFILEINFO dir ioctl returns LOV EA + MDS stat in 1 call (3327)
6718 - basic mmap support (3918)
6719 - kernel patch series update from b1_4 (4711)
6721 2004-09-16 Cluster File Systems, Inc. <info@clusterfs.com>
6724 - avoid crash during MDS cleanup with OST shut down (2775)
6725 - fix loi_list_lock/oig_lock inversion on interrupted IO (4136)
6726 - don't use bad inodes on the MDS (3744)
6727 - dynamic object preallocation to improve recovery speed (4236)
6728 - don't hold spinlock over lock dumping or change debug flags (4401)
6729 - don't zero obd_dev when it is force cleaned (3651)
6730 - print grants to console if they go negative (4431)
6731 - "lctl deactivate" will stop automatic recovery attempts (3406)
6732 - look for existing locks in ldlm_handle_enqueue() (3764)
6733 - don't resolve lock handle twice in recovery avoiding race (4401)
6734 - revalidate should check working dir is a directory (4134)
6736 - don't always mark "slow" obdfilter messages as errors (4418)
6738 2004-08-24 Cluster File Systems, Inc. <info@clusterfs.com>
6741 - don't close LustreDB during write_conf until it is done (3860)
6742 - fix typo in lconf for_each_profile (3821)
6743 - allow dumping logs from multiple threads at one time (3820)
6744 - don't allow multiple threads in OSC recovery (3812)
6745 - fix debug_size parameters (3864)
6746 - fix mds_postrecov to initialize import for llog ctxt (3121)
6747 - replace config semaphore with spinlock (3306)
6748 - be sure to send a reply for a CANCEL rpc with bad export (3863)
6749 - don't allow enqueue to complete on a destroyed export (3822)
6750 - down write_lock before checking llog header bitmap (3825)
6751 - recover from lock replay timeout (3764)
6752 - up llog sem before sending rpc (3652)
6753 - reduce ns lock hold times when setting kms (3267)
6754 - change a dlm LBUG to LASSERTF, to maybe learn something (4228)
6755 - fix NULL deref and obd_dev leak on setup error (3312)
6756 - replace some LBUG about llog ops with error handling (3841)
6757 - don't match INVALID dentries from d_lookup and spin (3784)
6758 - hold dcache_lock while marking dentries INVALID and hashing (4255)
6759 - fix invalid assertion in ptlrpc_set_wait (3880)
6761 - add libwrap support for the TCP acceptor (3996)
6762 - add /proc/sys/portals/routes for non-root route listing (3994)
6763 - allow setting MDS UUID in .xml (2580)
6764 - print the stack of a process that LBUGs (4228)
6766 2004-07-14 Cluster File Systems, Inc. <info@clusterfs.com>
6769 - don't cleanup request in ll_file_open() on failed MDS open (3430)
6770 - make sure to unset replay flag from failed open requests (3440)
6771 - if default stripe count is 0, use OST count for inode size (3636)
6772 - update parent mtime/ctime on client for create/unlink (2611)
6773 - drop dentry ref in ext3_add_link from open_connect_dentry (3266)
6774 - free recovery state on server during a forced cleanup (3571)
6775 - unregister_reply for resent reqs (3063)
6776 - loop back devices mounting and status check on 2.6 (3563)
6777 - fix resource-creation race that can provoke i_size == 0 (3513)
6778 - don't try to use bad inodes returned from MDS/OST fs lookup (3688)
6779 - more debugging for page-accounting assertion (3746)
6780 - return -ENOENT instead of asserting if ost getattr+unlink race (3558)
6781 - avoid deadlock after precreation failure (3758)
6782 - fix race and lock order deadlock in orphan handling (3450, 3750)
6783 - add validity checks when grabbing inodes from l_ast_data (3599)
6785 - add /proc/.../recovery_status to obdfilter (3428)
6786 - lightweight CDEBUG infrastructure, debug daemon (3668)
6787 - change default OSC RPC parameters to be better on small clusters
6788 - turn off OST read cache for files smaller than 32MB
6789 - install man pages and include them in rpms (3100)
6790 - add new init script for (un)mounting lustre filesystems (2593)
6791 - run chkconfig in %post for init scripts (3701)
6792 - drop scimac NAL (unmaintained)
6794 2004-06-17 Cluster File Systems, Inc. <info@clusterfs.com>
6797 - clean kiobufs before and after use (3485)
6798 - strip trailing '/'s before comparing paths with /proc/mounts (3486)
6799 - remove assertions to work around "in-flight rpcs" recovery bug (3063)
6800 - change init script to fail more clearly if not run as root (1528)
6801 - allow clients to reconnect during replay (1742)
6802 - fix ns_lock/i_sem lock ordering deadlock for kms update (3477)
6803 - don't do DNS lookups on NIDs too small for IP addresses (3442)
6804 - re-awaken ptlrpcd if new requests arrive during check_set (3554)
6805 - fix cond_resched (3554)
6806 - only evict unfinished clients after recovery (3515)
6807 - allow bulk resend, prevent data loss (3570)
6808 - dynamic ptlrpc request buffer allocation (2102)
6809 - don't allow unlinking open directory if it isn't empty (2904)
6810 - set MDS/OST threads to umask 0 to not clobber client modes (3359)
6811 - remove extraneous obd dereference causing LASSERT failure (3334)
6812 - don't use get_cycles() when creating temp. files on the mds (3156)
6813 - hold i_sem when setting i_size in ll_extent_lock() (3564)
6814 - handle EEXIST for set-stripe, set proper directory name (3336)
6816 - servers can dump a log evicting a client - lustre.dump_on_timeout=1
6817 - fix ksocknal_fmb_callback() error messages (2918)
6819 2004-05-27 Cluster File Systems, Inc. <info@clusterfs.com>
6822 - don't copy lvb into (possibly NULL) reply on error (2983)
6823 - don't deref dentry after dput, don't free lvb on error (2922)
6824 - use the kms to determine writeback rpc length (2947)
6825 - increment oti_logcookies when osc is inactive (2948)
6826 - update client's i_blocks count via lvb messages (2543)
6827 - handle intent open/close of special files properly (1557)
6828 - mount MDS with errors=remount-ro, like obdfilter (2009)
6829 - initialize lock handle to avoid ASSERT on error cleanup (3057)
6830 - don't use cancelling-locks' kms values (2947)
6831 - use highest lock extent for kms, not last one (2925)
6832 - don't dereference ERR_PTR() dentry in error handling path (3107)
6833 - fix thread race in portals_debug_dumplog() (3122)
6834 - create lprocfs device entries at setup instead of at attach (1519)
6835 - common AST error handler, don't evict client on completion race (3145)
6836 - zero nameidata in detach_mnt in 2.6 (3118)
6837 - verify d_inode after revalidate_special is valid in 2.6 (3116)
6838 - use lustre_put_super() to handle zconf unmounts in 2.6 (3064)
6839 - initialize RPC timeout timer earlier for 2.6 (3219)
6840 - don't dereference NULL reply buffer if mdc_close was never sent (2410)
6841 - print nal/nid for unknown nid (3258)
6842 - additional checks for oscc recovery before doing precreate (3284)
6843 - fix ll_extent_lock() error return code for 64-bit systems (3043)
6844 - don't crash in mdc_close for bad permissions on open (3285)
6845 - zero i_rdev for non-device files (3147)
6846 - clear page->private before handing to FS, better assertion (3119)
6847 - tune the read pipeline (3236)
6848 - fix incorrect decref of invalidated dentry (2350)
6849 - provide read-ahead stats and refine rpc in flight stats (3328)
6850 - don't hold journal transaction open across create RPC (3313)
6851 - update atime on MDS at close time (3265)
6852 - close LDAP connection when recovering to avoid server load (3315)
6853 - update iopen-2.6 patch with fixes from 2399,2517,2904 (3301)
6854 - don't leak open file on MDS after open resend (3325)
6855 - serialize filter_precreate and filter_destroy_precreated (3329)
6856 - loop device shouldn't call sync_dev() for nul device (3092)
6857 - clear page cache after eviction (2766)
6858 - resynchronize MDS->OST in background (2824)
6859 - refuse to mount the same filesystem twice on same mountpoint (3394)
6860 - allow llmount to create routes for mounting behind routers (3320)
6861 - push lock cancellation to blocking thread for glimpse ASTs (3409)
6862 - don't call osc_set_data_with_check() for TEST_LOCK matches (3159)
6863 - fix rare problem with rename on htree directories (3417)
6865 - allow default OST striping configuration per directory (1414)
6866 - fix compilation for qswnal for 2.6 kernels (3125)
6867 - increase maximum number of MDS request buffers for large systems
6868 - change liblustreapi to be useful for external progs like lfsck (3098)
6869 - increase local configuration timeout for slow disks (3353)
6870 - allow configuring ldlm AST timeout - lustre.ldlm_timeout=<seconds>
6872 2004-03-22 Cluster File Systems, Inc. <info@clusterfs.com>
6875 - fixes for glimpse AST timeouts / incorrectly 0-sized files (2818)
6876 - don't overwrite extent policy data in reply if lock was blocked (2901)
6877 - drop filter export grants atomically with removal from device (2663)
6878 - del obd_self_export from work_list in class_disconnect_exports (2908)
6879 - don't LBUG if MDS recovery times out during orphan cleanup (2530)
6880 - swab reply message in mdc_close, other PPC fixes (2464)
6881 - fix destroying of named logs (2325)
6882 - overwrite old logs when running lconf --write_conf (2264)
6883 - bump LLOG_CHUNKSIZE to 8k to allow for larger clusters (2306)
6884 - fix race in target_handle_connect (2898)
6885 - mds_reint_create() should take same inode create lock (2926)
6886 - correct journal credits calculated for CANCEL_UNLINK_LOG (2931)
6887 - don't close files for self_export to avoid uninitialized obd (2936)
6888 - allow MDS with the same name as client node (2939)
6889 - hold dentry reference for closed log files for unlink (2325)
6890 - reserve space for all logs during transactions (2059)
6891 - don't evict page beyond end of stripe extent (2925)
6892 - don't oops on a deleted current working directory (2399)
6893 - handle hard links to targets without a parent properly (2517)
6894 - don't dereference NULL lock when racing during eviction (2867)
6895 - don't grow lock extents when lots of conflicting locks (2919)
6897 2004-03-04 Cluster File Systems, Inc. <info@clusterfs.com>
6900 - account for cache space usage on clients to avoid data loss (974)
6901 - lfsck support in lustre kernel code (2349)
6902 - reduce journal credits needed for BRW writes (2370)
6903 - orphan handling to avoid losing space on client/server crashes
6904 - ptlrpcd can be blocked, stopping ALL progress (2477)
6905 - use lock value blocks to assist in proper KMS, faster stat (1021)
6906 - takes i_sem instead of DLM locks internally on obdfilter (2720)
6907 - recovery for initial connections (2355)
6908 - fixes for mds_cleanup_orphans (1934)
6909 - abort_recovery crashes MDS in b_eq (mds_unlink_orphan) (2584)
6910 - block all file creations until orphan recovery completes (1901)
6911 - client remove rq_connection from request struct (2423)
6912 - conf-sanity test_5, proper cleanup in umount log not availale (2640)
6913 - recovery timer race (2670)
6914 - mdc_close recovey bug (2532)
6915 - ptlrpc cleanup bug (2710)
6916 - mds timeout on local locks (2588)
6917 - namespace lock held during RPCs (2431)
6918 - handle interrupted sync write properly (2503)
6919 - don't try to handle a message that hasn't been replied to (2699)
6920 - client assert failure during cleanup after abort recovery (2701)
6921 - leak mdc device after failed mount (2712)
6922 - ptlrpc_check_set allows timedout requests to complete (2714)
6923 - wait for inflight reqs when ptlrpcd finishes (2710)
6924 - make sure unregistered services are removed from the srv_list
6925 - reset bulk XID's when resending them (caught by 1138 test)
6926 - unregister_bulk after timeout
6927 - fix lconf error (2694)
6928 - handle write after unfinished setstripe, stripe-only getstripe (2388)
6929 - readahead locks pages, leaves pending causing memory pressure (2673)
6930 - increase OST request buffers to 4096 on large machines (2729)
6931 - fix up permission of existing directories in simple_mkdir (2661)
6932 - init deleted item, add assertions ptlrpc_abort_inflight() (2725)
6933 - don't assign transno to errored transactions (2742)
6934 - don't delete objects on OST if given a bogus objid from MDS (2751)
6935 - handle large client PAGE_SIZE readdir on small PAGE_SIZE MDS (2777)
6936 - if rq_no_resend, then timeout request after recovery (2432)
6937 - fix MDS llog_logid record size, 64-bit array alignment (2733)
6938 - don't call usermode_helper from ptlrpcd, DEFAULT upcall (2773)
6939 - put magic in mount.lustre data, check for bad/NULL mount data (2529)
6940 - MDS recovery shouldn't delete objects that it has given out (2730)
6941 - if enqueue arrives after completion, don't clobber LVB (2819)
6942 - don't unlock pages twice when trigger_group_io returns error (2814)
6943 - don't deref NULL rq_repmsg if ldlm_handle_enqueue failed (2822)
6944 - don't write pages to disk if there was an error (1450)
6945 - don't ping imports that have recovery disabled (2676)
6946 - take buffered bytes into account when balancing socknal conn (2817)
6947 - hold a DLM lock over readdir always, use truncate_inode_pages (2706)
6948 - reconnect unlink llog connection after MDS reconnects to OST (2816)
6949 - remove little-endian swabbing of llog records (1987)
6950 - set/limit i_blksize to LL_MAX_BLKSIZE on client (2884)
6951 - retry reposting request buffers if they fail (1191)
6952 - grow extent at grant time to avoid granting a revoked lock (2809)
6953 - lock revoke doesn't evict page if covered by a second lock (2765)
6954 - disable VM readahead to avoid reading outside lock extents (2805)
6956 - return LL_SUPER_MAGIC from statfs for the filesystem type (1972)
6957 - updated kernel patches for hp-2.4.20 kernel (2681)
6959 2004-02-07 Cluster File Systems, Inc. <info@clusterfs.com>
6962 - fix truncated write corruption (2366)
6963 - fix for failed assertion in iopen_connect_dentry (1792,2517)
6965 - don't flag the ptlrpcd thread with PF_MEMALLOC (2636)
6966 - ensure len(uuid) < 37 in lmc (1171)
6967 - fix ia64 OOPS in llog_test (2255)
6968 - zero end of page at obdfilter for partial page writes (2648)
6969 - don't leave stale dentries around after renames (bug 2428)
6970 - fix timeouts when evicting a client with a single lock held (2642)
6971 - set deadline for the initial HELLO message to drain (2634)
6972 - print out dotted-quad IP addresses in the socknal (2302)
6974 - additional debugging for MDS client eviction problem (2443)
6975 - fix mkfsoptions support for osts (2603, 2604)
6977 2004-01-27 Cluster File Systems, Inc. <info@clusterfs.com>
6980 - add series for the vanilla 2.6.0 kernel
6981 - add series for the vanilla 2.4.24 kernel
6982 - add series for a cray x86/64 UL kernel drop
6983 - fix xattr patches for the vanilla 2.4.19 series
6985 - generate true UUIDs in lmc (1171)
6986 - have portals stack dumping break in UML (2466)
6987 - avoid bad dchild deref; avoid inum lock w/o creation (2362)
6988 - allocate with _NOFS in ldlm to avoid deadlock (1933)
6989 - wake callback waiting threads on client eviction (2460)
6990 - Add --ptldebug and --subsystem to lmc (1719)
6991 - update assertion to allow safe interrupt allocation
6992 - set rq_no_resend for cancel requests (2432)
6993 - recalculate ptlrpcd timeout after resend (2494)
6994 - call vfs_rmdir when removing pending directories (2368)
6995 - fix renaming a file to itself (2429)
6996 - lmc creates a default one-stripe lov (2454)
6997 - expand procfs space to handle large clusters (2326)
6998 - increase UML stack to avoid overflow
6999 - update lconf's list of debug and subsystem masks
7000 - fix lfs find --obd (2510)
7001 - /proc tunable for disabling filter read caching (2591)
7002 - stop rpm packages from altering slapd.conf (2301)
7003 - disable nagle in the socknal under 0conf (2578)
7004 - choose mds inode size based on stripe count (2572)
7005 - fix kernel-source rpm problems (2516)
7007 - add --disable-doc to avoid pdf generation (2421)
7008 - update documentation, tests, type-os, comments
7009 - avoid format warnings on ia64
7010 - remove the TOE NAL
7011 - tiny code cleanups by removing unused fields
7013 2004-01-07 Cluster File Systems, Inc. <info@clusterfs.com>
7016 - fix obvious semaphore misuse in as-yet-unused setattr path (2348)
7017 - remove the most blatant lies from BUILDING file (2371)
7018 - change default debug level to reasonable production setting
7019 - reduce client side cache size to reduce cache flush time
7020 - reduce max RPCs in flight to avoid unnecessary file fragmentation
7021 - make TCP zerocopy and pinger support enabled by default (2476)
7022 - sync writes completed after process exits caused crashes (2319)
7023 - maintain correct mount count on the MDS (2356)
7024 - backout 1557, because 2316 wasn't really fixed
7025 - better file I/O statistics gathering in /proc
7026 - don't take unnecessary, deadlock-inducing bug in readpage (2383)
7027 - another kernel patch to fix zero-copy TCP function export
7028 - don't take duplicate lock when processing re-sent getattr (2420)
7029 - lctl uses obd_self_export instead of creating new conn (2353)
7030 - MDS/OST recovery case which requires object creation asserted (2425)
7031 - move lfs from /usr/sbin to /usr/bin in packages
7032 - fix race between mds_client_add and mds_client_free (2417)
7033 - use kmalloc instead of slabs in portals (2430)
7034 - don't create duplicate records when a failover MDS is present (2442)
7035 - remove unnecessary mount age check (2332)
7036 - don't remove directory inodes from locks prematurely (2451)
7037 - don't break if MDS service name is the same as hostname (2103)
7038 - fix races in client write RPC generation when cache full (2482)
7040 2003-12-13 Cluster File Systems, Inc. <info@clusterfs.com>
7043 - remove now-unused request->rq_obd (278)
7044 - if an allocation fails, print out how much memory we've used (1933)
7045 - use PORTAL_SLAB_ALLOC for structures, to get GFP_MEMALLOC (1933)
7046 - add the "configurable stack size" patch to most series files (1256)
7047 - ability to write large log records, for 100+ OST configs (2306)
7048 - fix NULL deref when filter_prep fails (2314)
7049 - fix operator precedence error in filter_sync
7050 - dynamic allocation of socknal TX descriptors (2315)
7051 - fix a missed case in the GFP_MEMALLOC patch, can cause deadlock (2310)
7052 - fix gcc 2.96 compilation problem in xattr kernel patch (2294)
7053 - ensure that CWARN messages in Portals always get to the syslog
7054 - __init/__exit are not for prototype decls (ldlm_init/exit)
7055 - x86-64 compile warning fixes
7056 - fix gateway LMC keyword conflict (2318)
7057 - fix MDS lock inversions in getattr/reint paths (1844)
7058 - fix a rare lock re-ordering bug, which caused deadlock (2322)
7059 - fix i_sem/journal inversion in fsfilt_ext3_write_record (2306)
7060 - DLM race condition prevented some lock evictions (2328)
7061 - ENOMEM detection and retry on socknal sends (2230)
7062 - use GFP_NOFS throughout Lustre, to combat ENOMEM (2230)
7063 - move osc_rpcd into ptlrpc, for use in MDC and others (2329)
7064 - protect MDS inode fsdata with stronger locking; fixes assertion (2313)
7065 - better error messages when a client is rejected during recovery (1505)
7066 - avoid cancelling locks which were never granted, after failure (2330)
7067 - fix i_sem/journal inversion in mds_client_add (2333)
7068 - fix truncate/getattr lock cycle deadlock (2334)
7069 - use rpcd to send close; allows resend after timeout, avoid leak (1897)
7070 - fix two rare exit paths which could leak an l_lock() ref (2321)
7071 - fencepost error in MDS/OST orphan recovery (2226)
7072 - make log record alignment 8 bytes (1988)
7073 - lstripe now fails when requested offset > ost_count (2237)
7074 - ensure that all kernel series have a complete list.h (1607)
7075 - fix crashes in special-file operations (2316)
7076 - lctl create/brw OID mismatch, caused by obsolete filter loop (2339)
7078 - allow configurable automake binary, for testing new versions
7079 - small update to the lfs documentation
7081 2003-12-03 Cluster File Systems, Inc. <info@clusterfs.com>
7083 * fix negative export reference count in fsfilt_sync (2312)
7085 2003-12-01 Cluster File Systems, Inc. <info@clusterfs.com>
7086 * release candidate 0.9.1
7088 - orphans are moved into the PENDING directory for possible recovery
7089 - replayed opens now open by fid for orphan/rename safety (1042)
7090 - last close of an orphan inode generates a transno (683)
7091 - chdir() and mount() now pin the directory entry (1020)
7092 - avoid CERROR in normal ll_setattr_raw() error case (1500)
7093 - discard very old requests without processing them (1502)
7094 - remove some common, well-understood CERRORs (1505)
7095 - require O_DIRECT I/O to be page-sized to workaround IA64 crash (1609)
7096 - clear "grant" flags in OST replies until OST grant code lands (1644)
7097 - fix read performance by not clobbering i_blksize on client (1598)
7098 - fix __ldlm_handle2lock oops by not dereferencing lock after PUT (1625)
7099 - make LRU size a /proc tunable, clears locks when reduced (707)
7100 - fix some lprocfs rot that prevented ptlbd from loading (1732)
7101 - server locks take references on exports now (1558)
7102 - build fixes for 2.4.20-rh trees (1663)
7103 - return an error from lov_create if all OSCs are inactive (1751)
7104 - fix import levels when a reconnect happens without a timeout (1597)
7105 - exit early from mds_open if we get a lookup error (1749)
7106 - partial page read at EOF wouldn't wait for disk before sending (1642)
7107 - avoid NULL deref in obdfilter when reading page past EOF (1592)
7108 - avoid LASSERT in ll_intent_lock if server failed very early (1090)
7109 - fix LBUG in ll_it_open_error with rc = -2 (1861)
7110 - write/truncate lock inversion (1639)
7111 - Don't auto-load obdclass, portals modules during cleanup (1495)
7112 - fix timestamps from jumping to "now" (1763)
7113 - extra journal assertions (1648)
7114 - add an extra multiunlink test (1771)
7115 - fix read_record/write_record API (1776)
7116 - fix leak of offset_extent, possible incorrect i_size later (1772)
7117 - fix lasserts in mis-matched transnos during open-unlink testing (1541)
7118 - Debugging for the kqswnal_get_idle_tx problems (1820)
7119 - Allow recovery to be attempted multiple times (1536)
7120 - Write out MDS last_rcvd file after it is first created (1600)
7121 - Fix tx_descriptor leak in failed transmit situations (1827)
7122 - ext3 journaling fixes for assertion failure after IO error (1871)
7123 - class_export_put() on freed export after completion AST error (1896)
7124 - Fix revalidate looping in VFS (1322)
7125 - Don't access a freed export during MDS_REINT timeout (1521)
7126 - Add open-unlink recovery support on the MDS (1673,1764)
7127 - Return an error if no MDS data was read from last_rcvd (1946)
7128 - Fix for lookup "." or ".." crash on error (1932,1931,1935)
7129 - Don't setup a disk device that doesn't match exported UUID (317)
7130 - Reduce bulk RPC timeout to avoid cascading client/OST failures (1845)
7131 - avoid committing NULL handle in force close
7132 - local.sh is now a one-stripe LOV configuration
7133 - POSIX utime.4 -EPERM on FIFO not owned by user (56)
7134 - fix ext3 htree duplicate directory entry corruption (1516)
7135 - POSIX creat.13, fstat.1, open.18, stat.3 new file atime/mtime (2020)
7136 - update to new LOV EA format (2097)
7137 - interoperability for different PAGE_SIZE/wordsize (686,1821,1343,2042)
7139 2003-06-15 Phil Schwan <phil@clusterfs.com>
7142 - imports and exports cleanup too early, need refcounts (349, 879, 1045)
7143 - per-import/export recovery handling (958, 931, 959)
7144 - multiple last-rcvd slots, for serving multiple FSes (949)
7145 - connections are again shared between multiple imp/exports (963, 964)
7146 - "umount -f" would hang if any requests needed to be sent (393, 978)
7147 - avoid pinning large req buffer by copying for queued messages (989)
7148 - add "uuid" to "lctl device" command to help upcalls (991)
7149 - "open" RPCs with transnos would confuse recovery counters (1037)
7150 - do proper endian conversion of all wire messages (288, 340, 891)
7151 - remove OST bulk get LBUGs, fix ost_brw_write cleanup (1126)
7152 - call waiting locks callback from LDLM recovery thread (1127, 1151)
7153 - fix ptlrpc_connection leak in target_handle_connect (1174)
7154 - fix import refcounting bug in OST and MDS cleanup (1134)
7155 - if an invalid-at-open-time OSC returned before close(), LBUG (1150)
7156 - fix very unlikely obd_types race condition (501)
7157 - remove osc_open hack for echo_client (1187)
7158 - we leaked exports/dlmimps for forcibly disconnected clients (1143)
7159 - a failure in read_inode2 leads to deadlock (1139)
7160 - cancel ack-locks as soon as transaction is committed (1072)
7161 - fix major leaks and crashes in the bulk I/O path (937, 1057)
7162 - make sure to commitrw after any preprw to avoid deadlock (1162)
7163 - failing to execute a file in a lustre FS would lock inode (1203)
7164 - small DEBUG_REQ fix to avoid dereferencing a NULL (1227)
7165 - don't ASSERT while cleaning up an incompletely-setup obd (1248)
7166 - obd_uuid2tgt would walk off the end of the list (1255)
7167 - on IA64 the osc would give portals incorrect bulk size (1258)
7168 - fix debug daemon ioctl interface; allows daemon on ia64 (1274)
7169 - fix lock inversion caused by new llite matching code (1282)
7170 - limit the number of dirty pages on a client to 10MB (1286)
7171 - timed out locks were not being corrected cancelled (1289)
7172 - fix O_DIRECT above 4GB on IA-32 (1292)
7173 * major user-visible changes
7174 - fail out/fail over policy now controlled by the upcall (993)
7176 - add OBD_PING to check server availability and failure (954)
7177 - lustre messages are now sent in sending host order (288, 340, 891)
7178 - add eadatalen to MDS getattr reply (340)
7179 - OST read replies may contain second buffer, with per-page status (593)
7181 2003-03-11 Phil Schwan <phil@clusterfs.com>
7184 - LDLM_DEBUG macro fix, for gcc 3.2 (850)
7185 - failed open()s could cause deadlock; fixed (867, 869)
7186 - stop cancelling OST locks when files are closed (481)
7187 - overlapping XID spaces caused network corruption (851, 853)
7188 - fix unsafe fsfilt counter arithmetic; change to atomic_t
7189 - setattr_raw added, to do single-RPC, server-side setattrs
7190 - lmc/lconf syntax change for OST UUIDs
7191 - fix crashy race condition between ptlrpc_free_req and osc_close
7192 - don't use request in mdc_enqueue if we hit a timeout (889)
7193 - don't set the inode i_size for regular files from the MDS (896)
7194 - handle out of order completion AST (842)
7195 - don't LBUG if a lock request times out after receiving AST (913)
7196 - avoid d_rehash race in ll_find_alias by rehashing inside dcache_lock
7197 - if a bad lock AST arrives, send an error instead of dropping entirely
7198 - return 0 from revalidate2 if ll_intent_lock returns -EINTR (912)
7199 - fix leak in bulk IO when only partially completed (899, 900, 926)
7200 - fix O_DIRECT for ia64 (55)
7201 - (almost) eliminate Lustre-kernel-thread effects on load average (722)
7202 - C-z after timeout could hang a process forever; fixed (977)
7204 - client-side I/O cache (678, 924, 929, 941, 970)
7206 - READPAGE and SETATTRs which don't take server-side locks get
7209 2003-02-11 Phil Schwan <phil@clusterfs.com>
7212 - Fix ldlm_lock_match on the MDS to avoid matching remote locks (592)
7213 - Fix fsfilt_extN_readpage() to read a full page of directory
7214 entries, or fake the remainder if PAGE_SIZE != blocksize (500)
7215 - Avoid extra mdc_getattr() in ll_intent_lock when possible (534, 604)
7216 - Fix imbalanced LOV object allocation and out-of-bound access (469)
7217 - Most intent operations were removed, in favour of a new RPC mode
7218 that does a single RPC to the server and bypasses most of the VFS
7219 - All LDLM resource ID arrays were removed in favour of ldlm_res_id
7220 - Aggressively cancel local locks on DLM servers
7221 - mds_reint_unlink sends EA to the client if it's the last nlink.
7222 client uses that EA to unlink OST objects.
7223 - mds_reint_{rename,unlink,link} were rewritten to take ordered locks
7224 - recursive symlinks were fixed (439)
7225 - fixed NULL deref in DEBUG_REQ
7226 - filter_update_lastobjid no longer calls sync, which annoyed extN
7227 - fixed multi-client small-writes to a single file problem (445)
7228 - fixed mtime updates during file writes (607)
7229 - fixed vector writes on obdfilter causing problems when ENOSPC (670)
7230 - fixed bug in obd_brw_read/write() (under guise of testing 367)
7231 - fixed Linux OST size reporting problem (444, 656)
7232 - OST now updates object mtime with writes or setattr (607, 619)
7233 - client verifies file size before zeroing page past EOF (445)
7234 - OST now writes last allocated objid to disk with allocation (108)
7235 - LOV on echo now works (409)
7237 - mds_reint_unlink sends a new buffer, with the EA included. this
7238 buffer is only valid if body->valid & OBD_MD_FLEASIZE, which is only
7239 set if a regular file was being unlinked, and it was the last link
7240 - use PtlGet from the target for bulk writes (315)
7241 - OST now updates object mtime with writes or setattr (607, 619)
7242 - LDLM now has a grant-time callback to revalidate locked items, if
7244 - Many MDS operations were reorganized to combat race conditions
7246 - Merge b_intel branch (updated lprocfs code) - now at /proc/fs/lustre
7247 - configure check to avoid gcc version 2.96 20000731-2.96-98) (606)
7249 2003-01-06 Andreas Dilger <adilger@clusterfs.com>
7252 - Fully reactivate OST imports after reconnection (512, others)
7253 - Make sure client sees our -ENOTCONN from mds_handle (513 - partial)
7254 - More graceful error handling for truncating on dead OST (515)
7255 - Don't error out unless we're actually accessing dead stripes (474)
7256 - Fix garbage sizes when stripes are missing (410)
7257 - LRU counters were broken, causing constant lock purge (433, 432)
7258 - garbage on read from stripes with failed OSTs (441)
7259 - mark OSCs as active before reconnecting during recovery (438)
7260 - lov_enqueue and lov_cancel need to handle inactive OSTs (403)
7261 - lfind did not preserve OST order in output (443)
7262 - symlinks cause hung clients, incorrect data (439)
7263 - stop dereferencing request after dropping refcount (457)
7264 - don't LASSERT(spin_is_locked) on non-SMP (455)
7265 - fixes for many rename() bugs
7266 - fstat didn't correctly synchronize attributes (399)
7267 - server must handle lock cancellation during blocking AST prep (487)
7268 - bulk descriptors were free()d too soon (511)
7269 - fix paths in lconf, which would load incorrect modules (451, 507)
7270 - fix confusing lconf 'host not found' error message (386)
7271 - fix lock order deadlock on OST (O/R i_sem before journal ops, 478)
7272 - fix race condition in mdc_blocking_ast() for inode access (526)
7273 - fix lov_unpackmd() unpacking wrong number of stripes (537)
7274 - fix lov_set_osc_active() marking wrong OSC inactive (440)
7275 - fix bad lstripe lov_unpackmd() assertion (fix layering too) (527)
7276 - fix multiple writes of stripe MD to MDS (358, maybe 519)
7277 - fix lstripe in several ways (kernel side) (527)
7278 - fix request leak in ldlm_cli_enqueue (262)
7279 - incorrect OSC was marked inactive after OST failure
7280 - call mds_fs_cleanup before unmounting filesystem (524)
7281 - fix races between taking ns_lock and ldlm_lock_change_resource
7282 - fix races updating LOV export open file list
7283 - fix lov_enqueue error path, avoid decref-ing bad lock handle
7284 - fix recovery NULL deref in ldlm_cli_cancel_unused
7285 - fix some DLM races by using new hash table for lock handles (419)
7286 - permit the client to specify desired inodes, at replay
7287 - duplicate requests when we queue them for replay reintegration
7288 - fix last_rcvd offset calculation
7289 - sync after each recovered transaction, so we always make progress
7290 - never, not always, ERESTART requests without transnos
7291 - store the lov_desc in the MDS, so we don't depend on getlovinfo to
7293 - skip replay if the MDS says that the client is already connected
7294 - don't check for a recovery-enabled export to match lctl's UUID
7295 - don't INC_USE_COUNT for phantom exports
7296 - don't crash when cleaning up phantom exports (567)
7297 - don't double-finish or set replay data for errored mdc_open requests
7298 - abort requests when they time out, so we don't get old replies
7299 - send/receive replies for AST messages again
7300 - if the client says that it doesn't have the lock, cancel it on the
7302 - if we timeout during I/O, don't try to cancel an in-use lock; instead
7303 mark it as destroyed, it will all work out when decref is called
7304 - fix module use counts (22, 581)
7306 - ASTs now expect a reply (server cancels lock on error reply)
7308 2002-12-02 Andreas Dilger <adilger@clusterfs.com>
7311 - fix many simultaneous client startup (392)
7312 - fix dentry->d_it clobbering
7313 - credentials weren't being shipped for readdir/getattr operations
7314 - remove invalid assertions triggered during some concurrent MD
7316 - proper Lustre versions added (336, 389)
7317 - fix memory leak for create error case (398)
7318 - fix LOV locking bug that would get cli/srv out of sync
7319 - fix echo client over LOV (409)
7320 - fix dbench 2, extN refcount problem (170, 258, 356, 418)
7321 - fix double-O_EXCL intent crash (424)
7322 - avoid sending multiple lock CANCELs (352)
7324 - MDS can do multi-client recovery (modulo bugs in new code)
7326 - many updates, edits, cleanups
7328 2002-11-18 Phil Schwan <phil@clusterfs.com>
7331 - fix null d_it dereference (346)
7332 - fix full OST/dbench hang (333)
7333 - fix permission problem with file removal (286)
7334 - fix removal of OSCs from LOV when they fail
7335 - fix NULL deref during bulk timeout (214)
7336 - fix problems related to multiple filesystems on one MDS (241)
7337 - fixed serious subtle metadata locking bugs
7338 - free locks on clients when inodes are removed due to memory
7340 - fix inode pointer in lock data (285)
7341 - partial support for multiple MDS on a single host (241)
7342 - data locks weren't cancelled at clear_inode time (290, 311)
7343 - intent locks could lead to unbounded lock growth (205)
7344 - added a maximum lock count, an LRU list, and a flusher
7345 - fix multiple rename (365)
7346 - properly abstracted the echo client
7347 - OSC locked 1 byte too many; fixed
7348 - rewrote brw callback code:
7349 - fixed recovery bugs related to LOVs (306)
7350 - fixed too-many-pages-in-one-write crash (191)
7351 - fixed (again) crash in sync_io_timeout (214)
7352 - probably fixed callback-related race (385)
7354 - Add capability to MDS protocol
7355 - LDLM cancellations and callbacks on different portals
7357 2002-10-28 Andreas Dilger <adilger@clusterfs.com>
7360 - limit client IOV size to PTL_MD_MAX_IOV (611336, 191)
7361 - defer open object destruction to close time (601981, 138)
7362 - open/close OST file handle in obdo (OBD_MD_FLHANDLE) (601981, 138)
7363 - move LDLM_ENQUEUE/CONVERT back to MDS portal (625069)
7364 - abstract ll_lookup2, fix ll_revalidate2 to use abstraction (256)
7365 - don't call obd_setattr in ll_file_release for destroyed objects
7366 * protocol change to lustre_msg: move |version| and add |flags|
7367 * protocol change to osc_punch: "start" in "o_size", "end" in "o_blocks"
7368 * lock replay: for LDLM_FL_REPLAY trust client to do right thing
7369 * added replay of create, unlink, link and rename operations during
7370 MDS failover; recovery should be much more robust now
7371 * remove failed OSCs from LOVs (only lov_create uses this so far)
7372 * the lustre-HOWTO was brought (more) up to date (582544)
7374 2002-10-23 Phil Schwan <phil@clusterfs.com>
7377 - in-use dentries weren't being reused properly (617851)
7378 - prevent multiple LDLM setup (599178)
7379 - fix LOV size calculations for truncate (617853)
7380 - fix client handling of MDS intent errors (POSIX)
7381 - fix permission bug in lovstripe.c test (624321)
7382 - fix MDS thread deadlock - move LDLM handler to DLM portal (625069)
7383 - truncate past end of file could corrupt data
7384 - proper cleanup after timeouts, crashes, etc (592524, 550815)
7385 - a race in recovery could return ETIMEDOUT to apps (623947)
7386 - building outside the source directory was fixed
7387 * the lustre-HOWTO was brought (more) up to date (582544)
7388 * major progress was made on recovery functionality
7390 2002-10-10 Phil Schwan <phil@clusterfs.com>
7393 - recovery deadlock fix
7394 - rm -rf causes LBUG fix (617817)
7395 - file open by multiple tasks fix (618962)
7396 - directory permissions bugs (602707 and 620007)
7397 - journal_stop fixed with locking (611313)
7398 - O_APPEND failures resolved (618273, perhaps 614459)
7399 - lconf PATH fix (619770)
7400 - IA64 build fix (621450)
7401 - RPC buffer sizes scale with amount of memory
7403 2002-10-01 Phil Schwan <phil@clusterfs.com>
7406 - locks would be cancelled without throwing away data pages,
7407 resulting in inconsistent data (605627)
7408 - inode attributes were not always being refreshed (605627, 612449)
7409 - lconf now continues to cleanup after lctl reports an error
7410 - MDS now enforces user permissions (602707)
7411 - lprocfs cleanup fixed, but not yet enabled (614157)
7412 - fixed infinite server hang, should a client not respond to an AST
7413 - avoid going into recovery if user calls readlink() with a buffer
7414 that's too small (613941)
7415 - AST RPCs no longer require replies (614867) -- this may be changed
7416 - don't crash server if client sends an IOV that's too big (611336)
7417 - fixed lock conversion deadlock (611892)
7418 - fixed the following of symlinks (614622)
7419 * recovery: the server can remove locks from a client that dies, other
7420 clients can make progress
7421 * more extN patch fixes
7422 * compile-time configurable ptlrpc buffer allocations
7424 - collaborative read cache document
7425 - Lustre Lite Performance CDR document-in-progress
7427 2002-09-20 Andreas Dilger <adilger@clusterfs.com>
7430 - fix typo in patch-2.4.18
7432 2002-09-20 Andreas Dilger <adilger@clusterfs.com>
7435 - clear ptlrpc request each time in handle_incoming_request()
7436 - unlink of files now destroys the object on the OST
7438 2002-09-19 Peter Braam <braam@clusterfs.com>
7440 * add hard link support
7441 * change obdfile creation method
7442 * kernel patch changed
7444 2002-09-19 Peter Braam <braam@clusterfs.com>
7447 - stack overflow bug in extN fixed
7449 2002-09-18 Andreas Dilger <adilger@clusterfs.com>
7451 * documentation updates
7452 - add man pages for config tools
7453 - update tests/README to describe testing with new config tools
7454 - finish metadata API descriptions
7455 * bug fixes and cleanups
7456 - statfs workaround for 16TB limit
7457 - LOV stripe allocation improved, can stripe on subset of OSTs
7458 - LOV file size/IO offset was wrong for files > 4GB in size
7459 - object EA data was being dropped, caused files to be unreadable
7460 - memory overflow with non-LOV OST caused memory corruption
7461 - fixed regression tests to work with new config tools, obdfilter
7462 - fixed bug when directory size became larger than 1 block
7463 - fixed bug (for single client case) when PWD was deleted
7464 - invalidate local directory pages when doing intent-based ops
7465 - avoid LDLM oops when lock callback contained bad data
7467 2002-09-09 Andreas Dilger <adilger@clusterfs.com>
7469 * documentation updates
7470 * bug fixes and cleanups
7471 - configuration tools
7474 - 64-bit compile warnings
7475 - 64-bit internal statfs data
7477 * test_brw on persistent OST devices
7479 * lprocfs (disabled)
7481 2002-09-04 Andreas Dilger <adilger@clusterfs.com>
7483 * documentation updates
7484 * bug fixes and cleanups
7485 * configuration tools
7487 2002-08-30 Peter J. Braam <braam@clusterfs.com>
7490 * many small fixes to 0_5_4
7491 * io/network handling
7492 * thinkos in MDS operations
7494 2002-08-24 Peter J. Braam <braam@clusterfs.com>
7497 * crucial basic fixes to 0.5.3
7498 * IOR, Iozone work over Elan
7501 2002-08-07 Phil Schwan <phil@clusterfs.com>
7502 * version 0_5_3, our first alpha
7503 * we use the new Portals iovs
7504 * documentation updates
7505 * bug fixes and cleanups
7506 * small changes in the DLM wire protocol
7508 2002-07-25 Peter J. Braam <braam@clusterfs.com>
7509 * version 0_5_1 with some initial stability,
7510 * locking on MD and file I/O.
7511 * documentation updates
7512 * several bug fixes since 0.5.0
7513 * small changes in wire protocol
7515 2002-07-18 Phil Schwan <phil@clusterfs.com>
7517 * delivered as Lustre Light Alpha
7518 * fixed a crash after handling invalid MDS requests
7519 * fixed directory pages for architectures with non-4k pages sizes
7521 2002-07-11 Andreas Dilger <adilger@clusterfs.com>
7522 * release version v0_4_4
7523 * Moves TCP acceptor to be on port 2432 (unused Coda port) instead
7525 * Fixes a number of interruption problems with OST operations.
7526 * Update documentation for portals header changes
7527 * Move all wire protocol structs/defines to lustre_idl.h
7528 * Fixes symlink length bug.
7529 * Add tcpdump to repository.
7531 2002-07-05 Andreas Dilger <adilger@clusterfs.com>
7532 * release version v0_4_3
7533 * Fixes statfs for inodes on extN.
7534 * Fixes bug in runtests which would delete /etc/hosts.
7535 * Use 64-bit object IDs wherever possible (not into VFS though)
7536 Remove ost_get_info, which is unused by lustre, and out of date.
7538 2002-07-03 Peter Braam <braam@clusterfs.com>
7539 * release version v0_4_2 Fixes a lookup error (type not passed)
7540 * move forward to head of Portals
7541 * move forward to latest Lustre kernel
7543 2002-06-25 Peter Braam <braam@clusterfs.com>
7544 * release version v0_4_1. Hopefully stable on single node use.