4 2.6.16.60-0.42.8 (SLES 10),
5 2.6.27.39-0.3.1 (SLES11),
6 2.6.18-194.3.1.el5 (RHEL 5)
7 2.6.18-194.3.1.0.1.el5 (OEL 5)
8 * Client support for unpatched kernels:
9 (see http://wiki.lustre.org/index.php?title=Patchless_Client)
10 2.6.16 - 2.6.30 vanilla (kernel.org)
11 * Recommended e2fsprogs version: 1.41.10-sun2
12 * The async journal commit feature (bug 19128) and the cancel
13 lock before replay feature (bug 16774) are disabled by default.
15 Severity : enhancement
17 Description: Update RHEL5.5 kernel to 2.6.18-194.3.1.el5 and OEL5.5 kernel to
18 2.6.18-194.3.1.0.1.el5.
22 Description: mds_orphan_add_link()) error linking orphan to PENDING
23 Details : quota limits might disallow linking orphans to PENDING
24 when unlinking a file - temporary raise threads' privileges
25 when processing unlinks.
27 Severity : enhancement
29 Description: add conf-param -d option to remove permanent settings.
30 Details : Add the ability to remove permanent lctl conf_param settings.
31 (Previously conf_param settings could only be changed, not
32 removed.) This also provides a method to change failover
33 nid locations. Improve lctl man page.
35 Severity : enhancement
37 Description: add list_param to b1_8 and add "-R" option to list params
40 Severity : enhancement
42 Description: lfs quota output is not very convenient for awk/sed-parsing
43 Details : Some positions in lfs quota output table could be empty or
44 non-empty which made it hard to parse it with scripts, now a dash
45 is put instead of space where there is not supposed to be any data.
47 Severity : enhancement
49 Description: fix obdfilter-survey script to work properly with remote oss-s
51 Severity : enhancement
53 Description: add new OBDFILTER_SURVEY test suite
55 Severity : enhancement
57 Description: add new multiple mount protection (MMP) test suite
59 Severity : enhancement
61 Description: add support for async journal commit in echo client
63 Severity : enhancement
65 Description: allow userland programs to include <lustre/lustre_idl.h>
66 from stardard include directories
68 Severity : enhancement
70 Description: The prune-icache-use-trylock is no longer needed now that
71 the patch from bug 20008 is landed.
75 Description: The shrink grant feature is still active on the client although the
76 connect flag is not set.
80 Description: Don't leak grant space if the write failed with quota exceeded.
84 Description: a race condition could lead to SIGBUS being sent to an
85 application using mmap-ped files from Lustre
86 Details : truncate_complete_page implementation for the patchless
87 client could arbitrarily unset PG_Uptodate flag for a
88 page being kicked from the page cache, an uptodate check
89 right after a readpage call in filemap_fault could fail
90 because of that as though the page read had been unsuccessful.
94 Description: dlm lock slab shrinking is not efficient
95 Details : The dlm_locks slab can grow significantly and consumes a lot of
96 memory on the server. Set a hardlimit to grant_plan.
100 Description: Lustre does not do 1MB IOs to HW RAID
101 Details : Bump MAX_PHYS/HW_SEGMENTS and SG_ALL to 256 in the RHEL5 kernel.
102 This is what we do already for SLES kernels.
106 Description: LSI Fusion MPT driver hacks to improve performance
107 Details : Set CONFIG_FUSION_MAX_SGE to 256 for RHEL5
109 Severity : enhancement
111 Description: increase default md stripe_cache_size to 16k
114 Bugzilla : 15587/21439
115 Description: don't handle security.capability xattr
116 Details : CONFIG_SECURITY_FILE_CAPABILITIES is enabled by default on SLES11.
117 This results in additional getxattr calls, causing VBR test
118 failures as well as a preformance drop when writing.
122 Description: obdfilter-survey is no longer working
123 Details : revert patch from bug 20355 to resolve an issue with lctl --threads
124 not working correctly with $(PTHREAD_LIBS) being linked to lctl.
128 Description: ll_shrink_cache does not handle __GFP_FS properly
132 Description: lfs getstripe shows wrong info for directories
133 Details : Set correct LOVEA default values for filesystem-wide.
137 Description: FSX checksum false positves due to mmap IO
138 Details : Use OBD_FL_MMAP flag for IOs on a memory mapped file. Do not print
139 checksum errors, if the flag is set on a request.
143 Description: file operations after eviction have successful return values
144 Details : use vfs ->flush callback to return any pending async errors
149 Description: mdsrate fails to write after 1.3+M files opened
150 Details : decrease memory usage on clients by recycling dentries and
155 Description: obdfilter-survey gives unreasonably high numbers
156 Details : Wait for all threads to complete when running test_brw.
160 Description: do not set lustre read_only device when server umount and keep
161 client records for recoverable ones
165 Description: move sync_on_lock_cancel tunable to the obdfilter layer
166 Details : move the tunable to trigger a journal flush on lock cancel from
167 the ost layer to the obdfilter layer. This tunable is useful
168 when using the async journal commit feature.
172 Description: exp->exp_nid_stats == NULL in filter_tally()
173 Details : fix race with per-nid stats by delaying procfs cleanup until
178 Description: extent lock cancellation on client can keep the cpu busy for too
183 Description: Do not fail OST activation when a llog is not found, just
184 issue an error message.
188 Description: Don't enable extents by default for MDT.
192 Description: Protect bitfield access to ptlrpc_request's rq_flags, since
193 the AT code can access it concurrently while sending early
196 -------------------------------------------------------------------------------
198 2010-04-30 Oracle, Inc.
200 * Support for kernels:
201 2.6.16.60-0.42.8 (SLES 10),
202 2.6.27.39-0.3.1 (SLES11),
203 2.6.18-164.11.1.el5 (RHEL 5)
204 2.6.18-164.11.1.0.1.el5 (OEL 5)
205 * Client support for unpatched kernels:
206 (see http://wiki.lustre.org/index.php?title=Patchless_Client)
207 2.6.16 - 2.6.30 vanilla (kernel.org)
208 * Recommended e2fsprogs version: 1.41.10-sun2
209 * The async journal commit feature (bug 19128) and the cancel
210 lock before replay feature (bug 16774) are disabled by default.
214 Description: open-unlinked directories trigger MDS LBUG
215 Details : Fix regression introduced by the patch from bug 19640.
216 ext3_inc_count() can reset nlink to 1 when the directory
217 is indexed and inode->i_nlink == 2. Work around the problem
218 by incrementing nlink by 2 instead of 1.
222 Description: Reconnects are not throttled
223 Details : Don't wake up pinger on reconnect failures and rely on regular
224 pings to trigger the next reconnection. Please note that the
225 pinger already uses a smaller interval if the import is
229 Frequency : only with NFS export
231 Description: Console flooded with error message from ll_inode_from_lock()
233 Details : in mds_open, initialize the child_res_id before enqueuing
234 the OPEN lock for the child inode, then to avoid senting
235 wrong ldlm_res_id to client.
239 Description: allow multiple instances of the same nid in NID hash
240 Details : Case of multiple separate clients from the same NID (as
241 with liblustre) is legitimate and so we should allow
242 multiple instances of the same NID in nid hash.
246 Description: the readahead code can sleep on a semaphore while holding a
248 Details : in ras_update, "lov_get_info" could be called during increasing
249 readahead windows, which tries to get the mutex lock "lov_lock"
250 while holding the spin_lock "ras_lock", then causes system
255 Description: ASSERTION(cli->cl_avail_grant >= 0) failed
256 Details : fix assertion failure in the grant code.
260 Description: Use CNETERR (which is rate limited) in specific places in
261 the portal's LNET driver to avoid flooding the console.
265 Description: include last created object in precreate slow case
269 Description: don't do rep-ack if not created anything
270 Details : mds_open currently always put a lock into a rep-ack regardless
271 if something was created or not. This is pointless and only
272 creates needless contention. In fact the entire idea was to do
273 this for real creates as a recovery protection.
277 Description: Spurious error messages from smp_processor_id() on preemptible
279 Details : Disable a preemption by grabbing the lock in fs_trace_get_tcd()
280 first. The function fs_trace_get_tcd() was moved up.
284 Description: interval_erase() fix
285 Details : interval_erase() calls update_maxhigh() properly when child
290 Description: "lfs df" does not print stats for all mountpoints
291 Details : Print all mounted lustre filesystems with "lfs df"
295 Description: lfs setstripe -p not longer work with a relative pathname
296 Details : Use realpath() to provide absolute pathname.
300 Description: fix for truncated reply buffer
301 Details : reply buffer could be referred by reply_in_callback after released
306 Description: lustre.lov error when backing up symlinks with extended attributes
307 Details : Improved logic in ll_listxattr()
311 Description: properly handle null value for setattr -n lustre.lov
312 Details : Running "setfattr -n trusted.lov ." causes a NULL dereference
313 in ll_setxattr() due to no checking if "value" is NULL.
314 This command now resets to the default striping when executed
319 Description: stack overflow on lock cancellation due to fsync call
320 Details : sync_on_lock_cancel is needed for recovery when async journal
321 is enabled, but we actually just need to make sure that
322 metadata blocks have hit the journal, so doing a fs sync
323 should be enough and should consume less stack (just create an
324 empty handle and commmit it).
328 Description: using current->journal_info to store per-thread data leads
329 to problem under memory pressure
330 Details : disable the per-thread data (current->journal_info) containing
331 the lock info during I/O to work around the issue for short tem
335 Description: control DCACHE_LUSTRE_INVALID flag with MDS_INODELOCK_LOOKUP lock
336 Details : DCACHE_LUSTRE_INVALID is controlled by MDS_INODELOCK_LOOKUP
337 lock which is corresponding to "IT_LOOKUP", do not skip invalidate
342 Description: Cannot send after transport shutdown
343 Details : Clear imp_vbr_failed flag upon eviction
347 Description: soft lock in request set code during recovery
348 Details : during recovery, uses req->rq_set itself to replay the request
349 instead of ptlrpcd_recovery_pc
353 Description: Use CFS_ALLOC_IO instead of _STD in llap_from_page_with_lockh
354 Details : During an ll_readahead under ll_readpage, we have seen the the
355 OBD_SLAB_ALLOC hang under ldlm_pools_shrink when trying to lock
356 a page that is already locked by the readahead code.
361 Description: stop waitting for next replay transno if shutdown
362 Details : if the system is shutting down, wake up service thread blocked
363 to wait for next replay transno during recovery, then all the
364 references held by queued requests can be dropped and device
369 Description: lov_merge_lvb()) ASSERTION(spin_is_locked(&lsm->lsm_lock)) failed
370 Details : Protect lli->lli_smd pointer updates with lli->lli_lock.
374 Description: per-nid stats should not access lustre-hash internal structures
379 Description: mount.lustre fails to pass some options to mount()
384 Description: ext4 extent allocation is slower than in ext3
385 Details : Increase the default value of MB_DEFAULT_ORDER2_REQS to 8,
386 enlarge ext4 preallocation table for 2048 4K blocks extents
391 Description: incorrect triggering of synchronous IO
392 Details : The OSC can mistakenly fall back to synchronous IO when the
393 max_dirty_mb limit is reached and no write requests have yet
394 been issued. This can occur when the dirty pages are spread
395 over many files all of which are below the optimal request size.
399 Description: Optimize quota_ctl operations by sending requests in parallel
400 Details : Send MDS->OST quota_ctl requests in parallel, do not resend.
401 Compiled from two attachments in the ticket.
405 Description: ordering issue between transaction start & i_mutex
406 Details : start the transaction earlier in llog_lvfs_destroy to get
407 transaction start and inode mutex lock nested properly.
411 Description: lru resize SLV can get stuck
412 Details : calculate SLV with a greater precision to not lose small
413 changes due to interger math truncation; round up SLV only
414 if the amount of granted locks less than the limit to not
415 get stuck with this SLV
419 Description: avoid divide-by-zero in lprocfs_rd_import()
421 Severity : enhancement
423 Description: "weak-modules" support
424 Details : Implement "weak-modules" support which enables kernel modules
425 to be used with any kernel that implements the same kABI. In
426 order to achieve this modules are now installed in
427 /lib/modules/$(uname -r)/updates/kernel on all distributions.
431 Description: lfs quota failed when OSTs are down
432 Details : really return approximate block/inode usage when OSTs are down
436 Description: llobdstat fix and enhancement
437 Details : add a counter to set a limit to how many samples will be returned
438 fix a wildcard in the path to limit to obdfilter stats only
442 Description: recovery-small 51 hang - MDT can not stop, Mount still busy with
444 Details : abort lock enqueue processing sooner when we umount is in progress.
446 Severity : enhancement
448 Description: Update RHEL5.4 kernel to 2.6.18-164.11.1.el5 and OEL5.4 kernel to
449 2.6.18-164.11.1.0.1.el5.
451 Severity : enhancement
453 Description: error message improvements
454 Details : Use INFO/WARN instead of WARN/ERROR for the slow messages.
455 Simplify MDT/OST service start message.
456 Suppress "changing the import ..." warning.
459 Bugzilla : 21961/17914
460 Description: ignore trailing -mdc when determining index number
463 Frequency : only with SLES11
465 Description: fix for a race condition in linux quotas implementation
466 Details : dq_flags(struct dquot) access is not properly locked which could
467 lead to certain inconsistencies when accessing it using non-atomic
468 bit operations like __set_bit in do_set_dqblk.
472 Description: return any pending async errors in close(2) using flush callback
474 -------------------------------------------------------------------------------
476 2010-01-29 Sun Microsystems, Inc.
478 * Support for kernels:
479 2.6.16.60-0.42.8 (SLES 10),
480 2.6.27.39-0.3.1 (SLES11),
481 2.6.18-164.11.1.el5 (RHEL 5)
482 2.6.18-164.6.1.0.1.el5 (OEL 5)
483 * Client support for unpatched kernels:
484 (see http://wiki.lustre.org/index.php?title=Patchless_Client)
485 2.6.16 - 2.6.30 vanilla (kernel.org)
486 * Recommended e2fsprogs version: 1.41.6.sun1
487 * The async journal commit feature (bug 19128) and the cancel
488 lock before replay feature (bug 16774) are disabled by default.
490 Severity : enhancement
492 Description: Update RHEL5.4 kernel to 2.6.18-164.11.1.el5 and
493 OEL5.4 kernel to 2.6.18-164.11.1.0.1.el5.
495 Severity : enhancement
496 Bugzilla : 21511/19848
497 Description: Update SLES11 kernel to 2.6.27.39-0.3.1.
499 Severity : enhancement
501 Description: Update supported SLES10 kernel to 2.6.16.60-0.42.8.
503 Severity : enhancement
505 Description: Update kernel to RHEL5.4 2.6.18-164.6.1.el5 and
506 OEL5 2.6.18-164.6.1.0.1.el5(Both in-kernel OFED enabled).
508 Severity : enhancement
510 Description: Build kernels (RHEL5, OEL5 and SLES10/11) using the vendor's own
513 Severity : enhancement
515 Description: Vanilla kernel 2.6.30 patchless client support.
520 Description: bad entry in directory xxx: inode out of bounds
521 Details : fix locking issue in the rename path which could race with any
522 other operations updating the same directory.
524 Severity : enhancement
526 Description: Make watchdog timer messages to be more clear and descriptive.
530 Description: cp -p command does not preserve the dates and timestamp
531 Details : mtime could be spoiled by a write callback
535 Description: Clear imp_force_reconnect correctly in ptlrpc_connect_interpret()
539 Description: Allow non-root access for "lfs check".
540 Details : Added a check in obd_class_ioctl() for OBD_IOC_PING_TARGET.
542 Severity : enhancement
544 Description: quotacheck performance/scaling issues
545 Details : reduce quotacheck time on empty filesystem by skipping uninit
548 Severity : enhancement
550 Description: Enhancement for lfs(1) command to use numeric uid/gid.
552 Severity : enhancement
554 Description: Adjust locks' extents on their first enqueue, so that at the time
555 they get granted, there is no need for another pass through the
556 queues since they are already shaped into the proper forms.
560 Description: Fix mds_shrink_intent_reply()/mds_intent_policy() to pass correct
561 arguments and prevent LBUG() in lustre_shrink_reply_v2().
565 Description: Change tunefs.lustre and mkfs.lustre --mountfsoptions so that
566 exactly the specified mount options are used. Leaving off
567 any "mandatory" mount options is an error. Leaving off any
568 default mount options causes a warning, but is allowed.
569 Change errors=remount-ro from mandatory to default. Sanitize
570 the mount string before storing it. Update man pages accordingly.
574 Description: mds_getattr() should return 0, even if mds_fid2entry() fails with
575 -ENOENT. Also fix in ptlrpc_expire_one_request() to print signed
578 Severity : enhancement
580 Description: Remove set_info(KEY_UNLINKED) from MDS/OSC
582 Severity : enhancement
584 Description: Clients can replay thousands of unused locks during recovery
585 Details : Don't replay unused locks (only read locks for now) during
586 recovery. This feature is disabled by default and can be
587 enabled by running the following command on the clients:
588 lctl get_param ldlm.cancel_unused_locks_before_replay
592 Description: can't stat file in some situation.
593 Details : improve initialize osc date when target is added to mds and
594 ability to resend too big getattr request is client isn't have info
599 Description: Prevent inconsistences between linux and lustre mount structures.
600 Details : Wait indefinitely in server_wait_finished() until mnt_count drops.
601 Make the sleep interruptible.
603 Severity : enhancement
605 Description: Communicate OST degraded/readonly state via statfs to MDS
606 Details : Flags in the statfs returned from OSTs indicate whether the
607 OST is in a degraded RAID state, or if the filesystem has
608 turned read-only after a filesystem error is detected.
613 Description: don't panic if EPROTO was hit when reading symlink
614 Details : correctly handling request reference in error cases.
619 Description: open sometimes returns ENOENT instead of EACCES
620 Details : checking permission should be part of open part of mds_open, not
621 lookup part. so server should be set DISP_OPEN_OPEN disposition
622 before starting permission check.
623 Also not need revalidate dentry if client already have LOOKUP lock.
627 Description: enable client interface failover
628 Frequency : on servers with multiple network interfaces
629 Details : When a child reconnects from another NID, properly update export
630 nid hash position and ldlm reverse import.
632 Severity : enhancement
634 Description: implemented direct I/O with arbitrary (nonaligned) memory
635 addresses and file offsets.
637 Severity : enhancement
639 Description: added more recovery timeout options.
641 Severity : enhancement
643 Description: added llapi_file_open, llapi_file_create, llapi_file_get_stripe
648 Description: Avoid deadlock for local client writes
649 Frequency : only on systems with clients writing to an OST on the same node
650 Details : Use new OBD_BRW_MEMALLOC flag to notify OST about writes in the
651 memory freeing context. This allows OST threads to set the
652 PF_MEMALLOC flag on task structures in order to allocate memory
653 from reserved pools and complete IO.
654 Use GFP_HIGHUSER for OST allocations for non-local client writes,
655 so that the OST threads generate memory pressure and allow
656 inactive pages to be reclaimed.
661 Descriptoin: lock ordering violation between &cli->cl_sem and _lprocfs_lock
662 Details : move ldlm namespace creation in setup phase to avoid grab
663 _lprocfs_lock with cli_sem held.
667 Frequency : only during format of test systems
668 Description: Unable to run several mkfs.lustre on loop devices at the same time
669 Details : mkfs.lustre returns error 256 on the concurrent loop devices
670 formatting. The solution is to proper handle the error.
672 Severity : enhancement
674 Description: implement async create (obd_async_create) method for osc, to avoid
675 too long waiting new ost objects with holding ldlm lock.
679 Frequency : occasionally during network problems
680 Description: client not allowed to reconnect to OST because of active request
681 Details : abort bulk requests received by the OST once the client has timed
682 out since the client will resend the request anyway. The client
683 also now retries to reconnect to the same server if a connect
684 request failed with EBUSY or -EAGAIN.
687 Frequency : rare, if used wide striped file and one ost in down.
689 Descriptoin: don't return error if we created a subset of objects for file.
690 Details : lov_update_create_set() uses set->set_success as index for created
691 objects, so if some requests failed, they will have hole at end of
692 array and we can use qos_shrink_lsm for allocate correct lsm.
696 Description: Slow stale export processing during normal start up
697 Details : The global mgc lock prevents OST setup to be run in parallel.
698 Replace the global lock with a per-config_llog_data semaphore.
702 Description: Out or order replies might be lost on replay
703 Details : In ptlrpc_retain_replayable_request if we cannot find retained
704 request with tid smaller then one currently being added, add it
705 to the start, not end of the list.
709 Description: BUG: soft lockup - CPU#1 stuck for 10s! [ll_mdt_07:4523]
710 Details : add cond_resched() calls to avoid hogging the cpu for too long
711 in the hash code. Make also lustre_hash_for_each_empty() more
714 Severity : enhancement
716 Description: Performance improvements for debug messages with D_RPCTRACE,
717 D_LDLM, D_QUOTA options.
720 Frequency : only with NFS export
722 Description: (lov_merge.c:74:lov_merge_lvb())
723 ASSERTION(spin_is_locked(&lsm->lsm_lock)) failed (SR 71691004)
724 Details : Fix a race in the nfs export code by populating inode
725 info while the new inode is still locked
727 Severity : enhancement
729 Description: add a new file in procfs called force_lbug. Writting to this
730 file triggers a LBUG. Only for test purpose.
734 Description: OOM killer causes node hang
735 Details : really interrupt the sleep in osc_enter_cache on signals
739 Description: LustreError: 9153:0:(quota_context.c:622:dqacq_completion()) LBUG
740 Details : fix race during quota release on the slave.
742 Severity : enhancement
744 Description: smaller hash bucket sizes, cleanups
745 Details : increase hash table sizes and enabled rehashing for pools, quota,
746 uuid, nid & per-nid stats.
748 Severity : enhancement
750 Description: Add ldiskfs maxdirsize mount option
751 Details : add max_dir size mount option
755 Description: panic in ll_statahead_thread
756 Details : prevent parent thread to be killed before its child
759 Frequency : only with 16TB device
761 Description: unable to perform "mount -t lustre" of 16TB OST device
762 Details : Mounting 16TB LUNs failed due to three bugs in mkfs.lustre.
766 Description: ASSERTION(atomic_read(&imp->imp_inflight) == 0) failed
767 Details : unregistering should be zero if no RPC inflight.
771 Description: hyperion: Oops during metabench
772 Details : Correct the refcount of lov_request_set
774 Severity : enhancement
776 Description: Add mptlinux and nxge drivers to Lustre builds
778 Severity : enhancement
780 Description: Fix watchdog timer message to be more clear
781 Details : Make watchdog timer messages more clear and descriptive.
785 Description: LNET soft lockups in socknal_cd thread
786 Details : don't hog CPU for active-connecting if another connd is
787 accepting connecting-requst from the same peer
791 Description: recovery-small test_17 hang
792 Details : Land several AT improvements & fixes.
796 Description: MDS panic and hanging client processes
797 Details : Replace exp_ops_stats with exp_nid_stats->nid_stats
801 Description: OSS stuck in recovery.
802 Details : fix race during recovery. class_unlink_export,
803 class_set_export_delayed and target_queue_last_replay_reply
804 may race while increasing/decreasing obd_recoverable_clients
805 and obd_delayed_clients, causing recovery to wait forever.
807 Severity : enhancement
809 Description: add cascading_rw.c to lustre/tests
813 Description: filter_last_id() NULL deref
814 Details : lprocfs_filter_rd_last_id() should check for the fully
815 setup obd device, before proceeding further.
817 Severity : enhancement
819 Description: Loadgen improvements
820 Details : stacksize and locking fixes for loadgen
824 Description: Quiet CERROR("dirty %d > system dirty_max %d\n"
825 Details : The atomic_read() allowing the atomic_inc() are not covered
826 by a lock. Thus they may safely race and trip this CERROR()
827 unless we add in a small fudge factor (+1).
829 Severity : enhancement
831 Description: shrink_slab: nr=-9223362083340912175
832 Details : fix spurious message from shrink_slab reporing negative nr
836 Description: Quiet bogus previously committed transno error
837 Details : suppress the "server went back in time" error message which
838 is always printed even in the common case after a client eviction
840 Severity : enhancement
842 Description: Parallel statfs() calls result in client eviction
843 Details : cache statfs data for 1s.
847 Description: parallel-scale test_compilebench: @@@@@@ FAIL: compilebench
849 Details : fix serveral issues in pinger code causing clients not to ping
850 servers for too long, resulting in evictions.
854 Description: e2fsck should warn when MMP update interval is extended
855 Details : print mmp_check_interval and make it possible to abort mount
856 operation in case it takes too long.
860 Description: mdsrate-create-large.sh, BUG: soft lockup - CPU#0 stuck for 10s!
861 Details : fix bug in the RHEL5's jbd2 callback patch.
865 Description: drop number of active requests when queued for recovery
866 Details : Now that we take a reference on the original request instead of
867 making a copy of it for recovery. We need to drop the number of
868 active requests or the queued requests will prevent all request
869 processing when they exceed (srv->srv_threads_running - 1).
871 Severity : enhancement
873 Description: refuse to invalidate operational quota files when they are in use
874 Details : an attempt to invalidate operational quota files on the quota
875 master is not actually permitted by VFS (returning -EPERM), but we
876 should not depend on that and should return the error earlier.
880 Description: Applications stuck in jbd2_log_wait_commit during exit
881 Details : fix deadlock between kjournald2 trying to acquire the page lock
882 owned by an ost_io thread waiting for journal commit.
884 -------------------------------------------------------------------------------
886 2009-10-16 Sun Microsystems, Inc.
888 * Support for kernels:
889 2.6.16.60-0.42.4 (SLES 10),
890 2.6.27.29-0.1 (SLES11, i686 & x84_64 only),
891 2.6.18-128.7.1.el5 (RHEL 5),
892 2.6.18-128.7.1.el5 (OEL 5),
893 * Client support for unpatched kernels:
894 (see http://wiki.lustre.org/index.php?title=Patchless_Client)
895 2.6.16 - 2.6.27 vanilla (kernel.org)
896 * Recommended e2fsprogs version: 1.41.6.sun1
897 * ext4 support for RHEL5 is experimental and thus should not be
900 Severity : enhancement
902 Description: Add OEL5 support.
904 Severity : enhancement
906 Description: Update kernel to SLES11 2.6.27.29-0.1.
910 Description: File checksum failures with OST read cache on
911 Details : Disable page poisoning when the bulk transfer has to be aborted
912 because the client got evicted.
916 Description: Don't allow make backward step on assiging osc next id.
917 Details : race between allocation next id and ll_sync thread can be cause
918 of set wrong osc next id and can be kill valid ost objects.
920 Severity : enhancement
922 Description: Update kernel to RHEL5 2.6.18-128.7.1.el5.
924 Severity : enhancement
926 Description: Update kernel to SLES10 SP2 2.6.16.60-0.42.4.
930 Description: Changes in raid5-large-io-rhel5.patch to calculate sectors properly
934 Description: Increase the default BLK_DEF_MAX_SECTORS value for RHEL5 and SLES11
938 Description: Error handling in osc_statfs_interpret() has been improved.
939 Details : Check in osc_statfs_interpret() for EBADR.
943 Description: Do not send statfs() requests to OSTs disabled by administrator.
944 Details : Check in lov_prep_statfs_set() for non-NULL ltd_exp.
948 Description: Do not update ctime for the deleted inode.
949 Details : Check in mds_reint_unlink() before calling fsfilt_setattr().
953 Description: Increase of the size of the LDLM resource hash.
954 Details : Bump up RES_HASH_BITS=12.
958 Description: correctly send lsm on open replay
959 Details : MDS is trust to LSM size on replay open, but client can set wrong size
964 Description: Deadlock between filter_destroy() and filter_commitrw_write().
965 Details : filter_destroy() does not hold the DLM lock over the whole
966 operation. If the DLM lock is dropped, filter_commitrw() can go
967 through, causing the deadlock between page lock and i_mutex.
968 The i_alloc_sem should also be hold in filter_destroy() while
973 Description: truncate starts GFP_FS allocation under transaction causing deadlock
974 Details : ldiskfs_truncate calls grab_cache_page which may start page
975 allocation under an open transaction. This may lead to
976 calling prune_icache with consequent lustre reentrance.
979 Frequency : only when down/upgrading the MDS to 1.6/1.8 while 1.8 clients are
980 still up and when the OST pool feature is used
982 Description: interop testing got LBUG when run dd with OST pool
983 :LustreError: 30032:0:(llite_lib.c:1913:ll_replace_lsm()) LBUG
984 Details : down/upgrading the MDS to a version that doesn't/does support OST
985 pool can cause clients to crash because the lsm has changed
990 Description: missing tree_status on 1.8.1 RPM build
991 Details : make rpms failed due because the tree_status file is missing.
995 Description: continuing LustreError "mds adjust qunit failed!"
996 Details : don't print message on the console when ->adjust_qunit fails.
1000 Description: don't increase ldlm timeout if previous client was evicted
1001 Details : if a client doesn't respond to a blocking callback within the
1002 adaptive ldlm enqueue timeout, don't adjust the adaptive estimate
1003 when the lock is next granted.
1007 Description: ost is being unmounted w/o all writes to last_rcvd landing on disk.
1008 affects recovery negatively.
1009 Details : make sure all exports have been properly destroyed by the zombie
1010 thread processed before stopping the target.
1014 Description: Performance degradation with O_DIRECT between 1.6 & 1.8.1 b190
1015 Details : disable write barrier for ext4/SLES11.
1019 Description: Kernel panic - not syncing: Out of memory and no killable
1020 processes... on OSS when iozone
1021 Details : fix memory leak in the journal checksum patch.
1025 Description: group quota "too many blocks" OSS crashes
1026 Details : we should keep the same uid/gid for lquota_chkquota() and
1027 lquota_pending_commit()
1031 Description: LustreError: 9153:0:(quota_context.c:622:dqacq_completion()) LBUG
1032 Details : don't LBUG on release quota error. Just a workaround until the
1033 problem is understood.
1035 ------------------------------------------------------------------------------
1036 2009-07-31 Sun Microsystems, Inc.
1038 * Support for kernels:
1039 2.6.16.60-0.39.3 (SLES 10),
1040 2.6.27.23-0.1 (SLES11, i686 & x84_64 only),
1041 2.6.18-128.1.14.el5 (RHEL 5),
1042 * Client support for unpatched kernels:
1043 (see http://wiki.lustre.org/index.php?title=Patchless_Client)
1044 2.6.16 - 2.6.27 vanilla (kernel.org)
1045 * Recommended e2fsprogs version: 1.41.6.sun1
1046 * File join has been disabled in this release, refer to Bugzilla 16929.
1047 * NFS export disabled when stack size < 8192. Since the NFSv4 export
1048 of Lustre filesystem with 4K stack may cause a stack overflow. For
1049 more information, please refer to bugzilla 17630.
1050 * ext4 support for RHEL5 is experimental and thus should not be
1053 Severity : enhancement
1055 Description: Update kernel to SLES10 SP2 2.6.16.60-0.39.3.
1058 Frequency : with 1.8 server and 1.6 clients
1060 Description: correctly shrink reply for avoid send too big message to client.
1061 Details : 1.8 mds is allocate to big buffer to LOV EA data and this produce
1062 some problems with sending this reply to 1.6 client.
1066 Description: Repeated atomic allocation failures.
1067 Details : Use GFP_HIGHUSER | __GFP_NOMEMALLOC flags for memory allocations
1068 to generate memory pressure and allow reclaiming of inactive pages.
1069 At the same time, do not allow to exhaust emergency pools.
1070 For local clients the use of GFP_NOFS will be introduced in 1.8.2
1072 Severity : enhancement
1073 Bugzilla : 19846, 18289
1074 Description: Update kernel to RHEL5 2.6.18-128.1.14.el5.
1076 Severity : enhancement
1077 Bugzilla : 19625, 16893, 18668, 19848
1078 Description: Add support for SLES11 2.6.27.23-0.1.
1080 Severity : enhancement
1082 Description: Update client support to vanila kernels up to 2.6.27.
1084 Severity : enhancement
1086 Description: Update kernel to SLES10 SP2 2.6.16.60-0.37.
1088 Severity : enhancement
1090 Description: Compile with -Werror by default for i686 and x86_64.
1094 Description: resolve race between obd_disconnect and class_disconnect_exports
1095 Details : if obd_disconnect will be called to already disconnected export he
1096 forget release one reference and osc module can't unloaded.
1098 Severity : enhancement
1100 Description: move AT tunable parameters for more consistent usage
1101 Details : add AT tunables under /proc/sys/lustre, add to conf_param parsing
1105 Descriptoin: correctly skip time estimate if in recovery
1106 Details : rq_send_state insn't bitmask so using bitwise ops is forbid.
1110 Descriptoin: OSS DeadLock
1111 Details : Use trylock to prevent deadlock when shrink icache.
1113 Severity : enhancement
1115 Description: Allow tuning service thread via /proc
1116 Details : For each service a new
1117 /proc/fs/lustre/{service}/*/thread_{min,max,started} entry is
1118 created that can be used to set min/max thread counts, and get the
1119 current number of running threads.
1121 Severity : enhancement
1123 Description: Add state history info file, enhance import info file
1124 Details : Track import connection state changes in a new osc/mdc proc file;
1125 add overview-type data to the osc/mdc import proc file.
1129 Description: Reduce small size read RPC
1130 Details : Set read-ahead limite for every file and only do read-ahead when
1131 available read-ahead pages are bigger than 1M to avoid small size
1136 Description: free_entry erroneously used groups_free instead of put_group_info
1138 Severity : enhancement
1140 Description: Make read-ahead stripe size aligned.
1142 Severity : enhancement
1144 Description: MDS create should not wait for statfs RPC while holding DLM lock.
1147 Frequency : rare, connect and disconnect target at same time
1149 Descriptoin: ASSERTION(atomic_read(&imp->imp_inflight) == 0
1150 Details : don't call obd_disconnect under lov_lock. this long time
1151 operation and can block ptlrpcd which answer to connect request.
1154 Frequency : start MDS on uncleanly shutdowned MDS device
1156 Descriptoin: ll_sync thread stay in waiting mds<>ost recovery finished
1157 Details : stay in waiting mds<>ost recovery finished produce random bugs
1158 due race between two ll_sync thread for one lov target. send
1159 ACTIVATE event only if connect realy finished and import have
1163 Frequency : start MDS on uncleanly shutdowned MDS device
1165 Descriptoin: aborting recovery hang on MDS
1166 Details : don't throttle destroy RPCs for the MDT.
1170 Description: Slow reads beyond 8Tb offsets.
1171 Details : Page index integer overflow in ll_read_ahead_page
1175 Description: Soft lockup on OSS after MDS failover
1176 Details : MSG_CONNECT_INITIAL is not set on the initial MDS->OST connect.
1177 As a conseqence, the patch from bug 18224 is not operational
1178 and the MDS export cannot be reused on the OSTs until it gets
1182 Frequency : rare, only if using MMP with Linux RAID
1184 Description: MMP doesn't work with Linux RAID
1185 Details : While using HA for Lustre servers with Linux RAID, it is possible
1186 that MMP will not detect multiple mounts. To make this work we
1187 need to unplug the device queue in RAID when the MMP block is being
1188 written. Also while reading the MMP block, we should read it from
1189 disk and not the cached one.
1192 Frequency : rare, during recovery
1194 Description: Assertion failure in ldlm_lock_put
1195 Details : Do not put cancelled locks into replay list, hold references on
1196 locks in replay list
1200 Description: 1.6.5 mdsrate performance is slower than 1.4.11/12 (MDS is not cpu bound!)
1201 Details : create_count always drops to the min value (=32) because grow_count
1202 is being changed before the precreate RPC completes.
1205 Frequency : Only in RHEL5 when mounting multiple ext3 filesystems
1208 Description: "kmem_cache_create: duplicate cache jbd_4k" error message
1209 Details : add proper locking for creation of jbd_4k slab cache
1213 Description: MMP check in ext3_remount() fails without displaying any error
1214 Details : When multiple mount protection fails during remount, proper error
1219 Description: Rare Client crash on resend if the file was deleted.
1220 Details : When file is opened, but open reply is lost and file is
1221 subsequently deleted before resend, resend processing logic
1222 breaks trying to open the file again, should not try to open.
1226 Description: add check for >8TB ldiskfs filesystems
1227 Details : ext3-based ldiskfs does not support greater than 8TB LUNs.
1228 Don't allow >8TB ldiskfs filesystems to be mounted without
1229 force_over_8tb mount option
1233 Description: Client locked up when running multiple instances of an app. on
1234 multiple mount points
1235 Details : ll_shrink_cache() can sleep while holding the ll_sb_lock.
1236 Convert ll_sb_lock to a read/write semaphore to fix the problem.
1240 Description: Cannot acces an NFS-mounted Lustre filesystem
1241 Details : An NFS client cannot access the Lustre filesystem NFS-mounted
1242 from a Lustre-client exporting the Lustre filesystem via NFS.
1246 Description: panic in ll_statahead_thread
1247 Details : grab dentry reference in parent process.
1250 -------------------------------------------------------------------------------
1252 tbd Sun Microsystems, Inc.
1254 * Support for kernels:
1255 2.6.16.60-0.31 (SLES 10),
1256 2.6.18-128.1.6.el5 (RHEL 5),
1257 2.6.22.14 vanilla (kernel.org)
1258 * Client support for unpatched kernels:
1259 (see http://wiki.lustre.org/index.php?title=Patchless_Client)
1260 2.6.16 - 2.6.22 vanilla (kernel.org)
1261 * Recommended e2fsprogs version: 1.40.11-sun1
1262 * File join has been disabled, refer to Bugzilla 16929.
1263 * A new Lustre ADIO driver is available for MPICH2-1.0.7.
1264 * NFS export disabled when stack size < 8192. Since the NFSv4 export of
1265 Lustre filesystem with 4K stack may cause a stack overflow. For more
1266 information, please refer to bugzilla 17630.
1268 Severity : enhancement
1270 Description: Update to RHEL5.3 kernel-2.6.18-128.1.6.el5.
1272 Severity : enhancement
1274 Description: Update OFED release to 1.4.1 RC4
1276 Severity : major, only with big OST
1278 Description: Very poor metadata performance on Infiniband lustre configuration
1279 Details : OST object precreation becomes very slow on big OSTs. This is due
1280 to the ialloc patch spending too much time scanning groups.
1283 Frequency : during recovery
1285 Description: don't mix llog inodes with normal.
1286 Details : allocate inodes for log in last inode group
1291 Description: fix lqs' reference which won't be put in some situations
1292 Details : This patch fixes:
1293 1. In quota_check_common(), this function will check quota
1294 for user and group, but only send one return via "pending".
1295 In most cases, the pendings should be same. But that is not
1297 2. If quotaoff runs between lquota_chkquota() and
1298 lquota_pending_commit(), the same thing will happen too.
1299 That is why it comes:
1300 - if (!ll_sb_any_quota_active(qctxt->lqc_sb))
1303 -------------------------------------------------------------------------------
1305 2008-12-31 Sun Microsystems, Inc.
1307 * Support for kernels:
1308 2.6.5-7.314 (SLES 9),
1309 2.6.9-67.0.22.EL (RHEL 4),
1310 2.6.16.60-0.31 (SLES 10),
1311 2.6.18-92.1.17.el5 (RHEL 5),
1312 2.6.22.14 vanilla (kernel.org)
1313 * Client support for unpatched kernels:
1314 (see http://wiki.lustre.org/index.php?title=Patchless_Client)
1315 2.6.16 - 2.6.22 vanilla (kernel.org)
1316 * Client support for unpatched kernels:
1317 we do not recommend using patchless RHEL4 clients with kernels
1318 prior to 2.6.9-55EL (RHEL4U5).
1319 * Recommended e2fsprogs version: 1.40.11-sun1
1320 * Note that reiserfs quotas are disabled on SLES 10 in this kernel.
1321 * RHEL 4 and RHEL 5/SLES 10 clients behaves differently on 'cd' to a
1322 removed cwd "./" (refer to Bugzilla 14399).
1323 * A new quota file format has been introduced in 1.6.5.
1324 The format conversion from prior releases is handled transparently,
1325 but releases older than 1.4.12/1.6.5 don't understand this new
1326 format. The automatic format conversion can be avoided by running
1327 the following command on the MDS:
1328 'tunefs.lustre --param="mdt.quota_type=ug1" $MDTDEV'.
1329 For more information, please refer to bugzilla 13904.
1330 * A new quota file format was introduced in 1.6.6/1.8.0 (kernels 2.6.16+).
1331 The format conversion from prior releases is handled transparently,
1332 but releases older than 1.6.6/1.8.0 don't understand this new
1333 format. The automatic format conversion can be avoided by running
1334 the following commands on the MDS and OSS servers (for
1335 pre 1.4.12-1.6.5 quota files):
1336 'tunefs.lustre --param="mdt.quota_type=ug1" $MDTDEV',
1337 'tunefs.lustre --param="ost.quota_type=ug1" $MDTDEV'
1338 or (for 1.4.12/1.6.5 quota files)
1339 'tunefs.lustre --param="mdt.quota_type=ug2" $MDTDEV',
1340 'tunefs.lustre --param="ost.quota_type=ug2" $MDTDEV'
1341 For more information, please refer to bugzilla 13904.
1342 * Output of lfs quota has been made less detailed by default,
1343 old (verbose) output can be obtained by using -v option.
1344 * File join has been disabled in this release, refer to Bugzilla 16929.
1345 * A new Lustre ADIO driver is available for MPICH2-1.0.7.
1346 * NFS export disabled when stack size < 8192. Since the NFSv4 export of
1347 Lustre filesystem with 4K stack may cause a stack overflow. For more
1348 information, please refer to bugzilla 17630.
1350 Severity : enhancement
1352 Description: Caching OSS
1353 Details : introduce data caching on the OSS. The OSS now relies on the linux
1354 kernel page cache to keep recently accessed data in memory.
1355 It is worth noting that all write requests are still flushed
1356 synchronously as in lustre 1.6.
1358 Severity : enhancement
1360 Description: Version based recovery
1361 Details : introduce finer grained recovery able to detect transaction
1362 dependencies and can deal with transaction gaps caused by clients
1363 failing at the same time as the server.
1365 Severity : enhancement
1367 Description: Enable adaptive timeouts by default
1368 Details : The Lustre timeout value in /proc/sys/lustre/timeout is now
1369 managed dynamically based on server load and should not need
1370 to be tuned manually based on cluster size. This allows Lustre
1371 to work under a wider variety of system sizes and loads, without
1372 unnecessarily causing lengthy recovery times.
1374 Severity : enhancement
1376 Description: Add OST Pools support
1377 Details : File striping can now be set to use an arbitrary pool of OSTs
1379 Severity : enhancement
1381 Description: add lazystatfs mount option to allow statfs(2) to skip down OSTs
1382 Details : allow skip disconnected ost for send statfs request and hide error
1386 Frequency : rare, on llog test 6
1388 Descriptoin: don't allow connect to already connected import
1389 Details : allowing connect to already connected import is hide connecting problem.
1392 Frequency : rare, on failed llog setup
1394 Descriptoin: don't leak obd reference on failed llog setup
1395 Details : for failed llog setup - mgc forget call class_destroy_import for
1396 client import, move destroy import to more generic place.
1401 Descriptoin: allow kill process which wait statahead result
1402 Details : for some reasons 'ls' can stick in waiting result from statahead,
1403 in this case need way for kill this process.
1406 Frequency : rare, at shutdown
1408 Descriptoin: panic at umount
1409 Details : llap_shrinker can be raced with killing super block from list and
1410 this produce panic with access to already freeded pointer
1415 Descriptoin: don't lose wakeup for imp_recovery_waitq
1416 Details : recover_import_no_retry or invalidate_import and import_close can
1417 both sleep on imp_recovery_waitq, but we was send only one wakeup
1423 Descriptoin: panic in mds_open
1424 Details : don't confuse mds_finish_transno() with PTR_ERR(-ENOENT)
1429 Descriptoin: stuck in cache_remove_extent() or panic with accessing to already
1431 Details : release lock refernce only after add page to pages list.
1434 Frequency : always with long access acl
1436 Descriptoin: mds can't pack reply with long acl.
1437 Details : mds don't control size of acl but they limited by reint/getattr
1442 Frequency : on remount
1444 Description: external journal device not working after the remount
1445 Details : clear dev_rdonly flag for external journal devices in
1451 Description: shutdown vs evict race
1452 Details : client_disconnect_export vs connect request race.
1453 if client will evicted at this time - we start invalidate
1454 thread without referece to import and import can be freed
1460 Description: shrink LOV EAs before replying
1461 Details : correctly adjust LOV EA buffer for reply.
1466 Description: don't skip ost target if they assigned to file
1467 Details : Drop slow OSCs if we can, but not for requested start idx.
1468 This means "if OSC is slow and it is not the requested
1469 start OST, then it can be skipped, otherwise skip it only
1470 if it is inactive/recovering/out-of-space.
1472 Severity : enhancement
1474 Description: Update to RHEL5 kernel-2.6.18-92.1.17.el5.
1477 Frequency : rare, need acl's on inode.
1479 Description: client can't handle ost additional correctly
1480 Details : if ost was added after client connected to mds client can have
1481 hit lnet_try_match_md ... to big messages to wide striped files.
1482 in this case need teach client to handle config events about add
1483 lov target and update client max ea size at that event.
1485 Severity : enhancement
1487 Description: Update to sles9 kernel-2.6.5-7.314.
1489 Severity : enhancement
1491 Description: Update to SLES10 SP2 kernel-2.6.16.60-0.31.
1494 Frequency : Create a symlink file with a very long name
1496 Description: ldlm_cancel_pack()) ASSERTION(max >= dlm->lock_count + count)
1497 Details : If there is no extra space in the request for early cancels,
1498 ldlm_req_handles_avail() returns 0 instead of a negative value.
1503 Description: mds is deadlocked
1504 Details : in rare cases, inode in catalog can have i_no less than have parent
1505 i_no, this produce wrong order for locking during open, and
1506 parallel unlink can be lock open. this need teach mds_open to grab
1507 locks in resource id order, not at parent -> child order.
1509 Severity : enhancement
1511 Description: Add /proc entry for import status
1512 Details : The mdc, osc, and mgc import directories now have
1513 an import directory that contains useful import data for debugging
1514 connection problems.
1516 Severity : enhancement
1518 Description: Re-disable certain /proc logging
1519 Details : Enable and disable client's offset_stats, extents_stats and
1520 extents_stats_per_process stats logging on the fly.
1523 Frequency : Only on FC kernels 2.6.22+
1525 Description: oops in statahead
1526 Details : Do not drop reference count for the dentry from VFS when lookup,
1527 VFS will do that by itself.
1529 Severity : enhancement
1531 Description: Generic /proc file permissions
1532 Details : Set /Proc file permissions in a more generic way to enable non-
1533 root users operate on some /proc files.
1537 Description: Hitting mdc_commit_close() ASSERTION
1538 Details : Properly handle request reference release in
1539 ll_release_openhandle().
1543 Frequency : only patchless client
1544 Description: add workaround for race between add/remove dentry from hash
1546 Severity : enhancement
1548 Description: Allow OST glimpses to return PW locks
1552 Description: LBUG when llog conf file is full
1553 Details : When llog bitmap is full, ENOSPC should be returned for plain
1558 Description: Prevent import from entering FULL state when server in recovery
1562 Description: service mount cannot take device name with ":"
1563 Details : Only when device name contains ":/" will mount treat it as
1569 Description: replace ptlrpcd with the statahead thread to interpret the async
1570 statahead RPC callback
1574 Frequency : on recovery
1575 Description: I/O failures after umount during fail back
1576 Details : if client reconnected to restarted server we need join to recovery
1577 instead of find server handler is changed and process self eviction
1578 with cancel all locks.
1580 Severity : enhancement
1582 Description: Update to RHEL5 kernel-2.6.18-92.1.10.el5.
1586 Description: Kernel BUG tries to release flock
1587 Details : Lustre does not destroy flock lock before last reference goes
1588 away. So always drop flock locks when client is evicted and
1589 perform unlock regardless of successfulness of speaking to MDS.
1591 Severity : enhancement
1593 Description: Update to SLES10 SP2 kernel-2.6.16.60-0.27.
1595 Severity : enhancement
1597 Description: Upcall on Lustre log has been dumped
1598 Details : Allow for a user mode script to be called once a Lustre log has
1599 been dumped. It passes the filename of the dumped log to the
1600 script, the location of the script can be specified via
1601 /proc/sys/lnet/debug_log_upcall.
1606 Description: avoid messages about idr_remove called for id that is not allocated
1607 Details : Move assigment s_dev for clustered nfs to end of initialization,
1608 for avoid problem with error handling.
1613 Description: avoid Already found the key in hash [CONN_UNUSED_HASH] messages
1614 Details : When connection is reused this not moved from CONN_UNUSED_HASH into
1615 CONN_USED_HASH and this prodice warning when put connection again
1621 Description: avoid ASSERTION(client_stat->nid_exp_ref_count == 0) failed
1622 Details : release reference to stats when client disconnected, not
1623 when export destroyed for avoid races when client destroyed
1624 after main ost export.
1628 Description: more cleanup in mds_lov
1629 Details : add workaround for get valid ost count for avoid warnings about
1630 drop too big messages, not init llog cat under semphore which
1631 can be blocked on reconnect and break normal replay, fix access
1634 Severity : enhancement
1636 Description: Export bytes_read/bytes_write count on OSC/OST.
1640 Description: Early reply size mismatch, MGC loses connection
1641 Details : Apply the MGS_CONNECT_SUPPORTED mask at reconnect time so
1642 the connect flags are properly negotiated.
1646 Description: Properly propagate oinfo flags from lov to osc for statfs
1647 Details : restore missing copy oi_flags to lov requests.
1651 Description: exports in /proc are broken
1652 Details : recreate /proc entries for clients when they reconnect.
1654 Severity : enhancement
1656 Description: Add man pages for llobdstat(8), llstat(8), plot-llstat(8),
1657 : l_getgroups(8), lst(8), routerstat(8)
1658 Details : included man pages for llobdstat(8), llstat(8),
1659 : plot-llstat(8), l_getgroups(8), lst(8), routerstat(8)
1661 Severity : enhancement
1663 Description: Implement lustre ll_show_options method.
1665 Severity : enhancement
1667 Description: Update to SLES9 kernel-2.6.5-7.312.
1669 Severity : enhancement
1671 Description: Update to RHEL4 kernel-2.6.9-67.0.22.EL.
1675 Description: exports in /proc are broken
1676 Details : recreate /proc entries for clients when they reconnect.
1680 Description: don't fail open with -ERANGE
1681 Details : if client connected until mds will be know about real ost count
1682 get LOV EA can be fail because mds not allocate enougth buffer
1687 Description: Resolve device initialization race
1688 Details : Prevent proc handler from accessing devices added to the
1689 obd_devs array but yet be intialized.
1691 Severity : enhancement
1693 Description: configure's --enable-quota should check the
1694 : kernel .config for CONFIG_QUOTA
1695 Details : configure is terminated if --enable-quota is passed but
1696 : no quota support is in kernel
1698 Severity : enhancement
1700 Description: Update to SLES10 SP2 kernel-2.6.16.60-0.23.
1702 Severity : enhancement
1704 Description: Update to RHEL5 kernel-2.6.18-92.1.6.el5.
1708 Frequency : rare, on PPC clients
1709 Description: don't swab ost objects in response about directory, because
1711 Details : bug similar bug 14856, but in different function.
1713 Severity : enhancement
1715 Description: lfs quota tool enhancement
1716 Details : added units specifiers support for setquota, default to
1717 current uid/gid for quota report, short quota stats by
1718 default, nonpositional parameters for setquota, added
1719 llapi_quotactl manual page.
1721 Severity : enhancement
1723 Description: *optional* service tags registration
1724 Details : if the "service tags" package is installed on a Lustre node
1725 When the filesystem is mounted, a local-node service tag will
1726 be created. See http://inventory.sun.com/ for more information
1727 about the Service Tags asset management system.
1731 Description: Client runs out of low memory
1732 Details : Consider only lowmem when counting initial number of llap pages
1735 Frequency : occasional
1737 Description: add refcount for osc callbacks, so avoid panic on shutdown
1739 Severity : enhancement
1741 Description: Update to RHEL4 kernel-2.6.9-67.0.20.
1744 Frequency : testing only
1746 Description: sanity test 65a fails if stripecount of -1 is set
1747 Details : handle -1 striping on filesystem in ll_dirstripe_verify
1750 Frequency : only in unusual configurations
1752 Description: Kernel panic with find ost index.
1753 Details : lov_obd have panic if some OST's have sparse indexes.
1755 Severity : enhancement
1757 Description: Update to RHEL5 kernel-2.6.18-53.1.21.el5.
1760 Frequency : rarely, if filesystem is mounted with -o flock
1762 Description: do not process already freed flock
1763 Details : flock can possibly be freed by another thread before it reaches
1764 to ldlm_flock_completion_ast.
1767 Frequency : rarely, if filesystem is mounted with -o flock
1769 Description: LBUG during stress test
1770 Details : Need properly lock accesses the flock deadlock detection list.
1773 Frequency : rarely, if binaries are being run from Lustre
1775 Description: oops in page fault handler
1776 Details : kernel page fault handler can return two special 'pages' in
1777 error case, don't try dereference NOPAGE_SIGBUS and NOPAGE_OMM.
1780 Frequency : rarely, during shutdown
1782 Description: timeout with invalidate import.
1783 Details : ptlrpcd_check call obd_zombie_impexp_cull and wait request which
1784 should be handled by ptlrpcd. This produce long age waiting and
1785 -ETIMEOUT ptlrpc_invalidate_import and as result LASSERT.
1791 Description: ASSERTION(CheckWriteback(page,cmd)) failed
1792 Details : badly clear PG_Writeback bit in ll_ap_completion can produce false
1796 Frequency : only with broken builds/installations
1798 Description: no LBUG if lquota.ko and fsfilt_ldiskfs.ko are different versions
1799 Details : just return an error to a user, put a console error message
1801 Severity : enhancement
1803 Description: Update to RHEL5 kernel-2.6.18-53.1.19.el5.
1805 Severity : enhancement
1807 Description: Update to RHEL4 kernel-2.6.9-67.0.15.
1809 Severity : enhancement
1811 Description: enable MGS and MDT services start separately
1812 Details : add a 'nomgs' option in mount.lustre to enable start a MDT with
1813 a co-located MGS without starting the MGS, which is a complement
1814 to 'nosvc' mount option.
1817 Frequency : always, on big-endian systems
1819 Description: cleanup in ptlrpc code, related to PPC platform
1820 Details : store magic in native order avoid panic's in recovery on PPC
1821 node and forbid from this error in future. Also fix posibily
1822 of twice swab data. Fix get lov striping to userpace.
1825 Frequency : rarely, if replay get lost on server
1827 Description: server incorrectly drop resent replays lead to recovery failure.
1828 Details : do not drop replay according to msg flags, instead we check the
1829 per-export recovery request queue for duplication of transno.
1832 Frequency : after recovery
1834 Description: precreate to many object's after del orphan.
1835 Details : del orphan st in oscc last_id == next_id and this triger growing
1836 count of precreated objects. Set flag LOW to skip increase count
1837 of precreated objects.
1840 Frequency : rare, on clear nid stats
1842 Description: ASSERTION(client_stat->nid_exp_ref_count == 0)
1843 Details : when clean nid stats sometimes try destroy live entry,
1844 and this produce panic in free.
1847 Frequency : occasionally since 1.6.4
1849 Description: Stack overflow during MDS log replay
1850 Details : ease stack pressure by using a thread dealing llog_process.
1853 Frequency : very rare
1855 Description: MDT cannot be unmounted, reporting "Mount still busy"
1856 Details : Mountpoint references were being leaked during open reply
1857 reconstruction after an MDS restart. Drop mountpoint reference
1858 in reconstruct_open() and free dentry reference also.
1863 Description: wait until IO finished before start new when do lock cancel.
1864 Details : VM protocol want old IO finished before start new, in this case
1865 need wait until PG_writeback is cleared until check dirty flag
1866 and call writepages in lock cancel callback.
1871 Description: mds_mfd_close() ASSERTION(rc == 0)
1872 Details : In mds_mfd_close(), we need protect inode's writecount change
1873 within its orphan write semaphore to prevent possible races.
1876 Frequency : rare, on shutdown ost
1878 Description: don't hit live lock with umount ost.
1879 Details : shrink_dcache_parent can be in long loop with destroy dentries,
1880 use shrink_dcache_sb instead.
1883 Frequency : only when echo_client is used
1885 Description: don't panic with use echo_client
1886 Details : echo client pass NULL as client nid pointer and this produce NULL
1887 pointer dereference.
1890 Frequency : Always on 32-bit PowerPC systems
1892 Description: fix build on PPC32
1893 Details : compile code with -m64 flag produce wrong object file for PPC32.
1898 Description: MDS LBUG: ASSERTION(!IS_ERR(dchild))
1899 Details : In reconstruct_* functions, LASSERTs on both the data supplied
1900 by a client, and the data on disk are dangerous and incorrect.
1901 Change them with client eviction.
1903 Severity : enhancement
1905 Description: skiplist implementation simplification
1906 Details : skiplists are used to group compatible locks on granted list
1907 that was implemented as tracking first and last lock of each lock
1908 group the patch changes that to using doubly linked lists
1912 Description: delete compatibility for 32bit qdata
1913 Details : as planned, when lustre is beyond b1_8, lquota won't support 32bit
1914 qunit. That means servers of b1_4 and servers of b1_8 can't be
1915 used together if users want to use quota.
1918 Frequency : only with administrator action
1920 Description: mount failure if config log has invalid conf_param setting
1921 Details : If administrator specified an incorrect configuration parameter
1922 with "lctl conf_param" this would cause an error during future
1923 client mounts. Instead, ignore the bad configuration parameter.
1926 Frequency : blocks per group < blocksize*8 and uninit_groups is enabled
1928 Description: ldiskfs error: XXX blocks in bitmap, YYY in gd
1929 Details : If blocks per group is less than blocksize*8, set rest of the
1933 Frequency : Application do stride read on lustre
1935 Description: The read performance will drop a lot if the application does
1937 Details : Because the stride_start_offset are missing in stride read-ahead,
1938 it will cause clients read a lot of unused pages in read-ahead,
1939 then the read-performance drops.
1943 Description: more ldlm soft lockups
1944 Details : In ldlm_resource_add_lock(), call to ldlm_resource_dump()
1945 starve other threads from the resource lock for a long time in
1946 case of long waiting queue, so change the debug level from
1947 D_OTHER to the less frequently used D_INFO.
1949 Severity : enhancement
1951 Description: add -gid, -group, -uid, -user options to lfs find
1953 Severity : enhancement
1955 Description: ll_recover_lost_found_objs - recover objects in lost+found
1956 Details : OST corruption and subsequent e2fsck can leave objects in the
1957 lost+found directory. Using the "ll_recover_lost_found_objs"
1958 tool, these objects can be retrieved and data can be salvaged
1959 by using the object ID saved in the fid EA on each object.
1964 Description: this bug _only_ happens when inode quota limitation is very low
1965 (less than 12), so that inode quota unit is 1 at initialization.
1966 Details : if remaining quota equates 1, it is a sign to demonstate that quota
1967 is effective now. So least quota qunit should be 2.
1971 Description: Hung threads in invalidate_inode_pages2_range
1972 Details : The direct IO path doesn't call check_rpcs to submit a new RPC once
1973 one is completed. As a result, some RPCs are stuck in the queue
1978 Description: Procfs and llog threads access destoryed import sometimes.
1979 Details : Sync the import destoryed process with procfs and llog threads by
1980 the import refcount and semaphore.
1984 Description: mds fails to respond, threads stuck in ldlm_completion_ast
1985 Details : Sort source/child resource pair after updating child resource.
1990 Description: kernel BUG at ldiskfs2_ext_new_extent_cb
1991 Details : If insertion of an extent fails, then discard the inode
1992 preallocation and free data blocks else it can lead to duplicate
1997 Description: don't always update ctime in ext3_xattr_set_handle()
1998 Details : Current xattr code updates inode ctime in ext3_xattr_set_handle()
1999 In some cases the ctime should not be updated, for example for
2000 2.0->1.8 compatibility it is necessary to delete an xattr and it
2001 should not update the ctime.
2005 Description: add quota statistics
2006 Details : 1. sort out quota proc entries and proc code.
2007 2. add quota statistics
2012 Description: quotas are not honored with O_DIRECT
2013 Details : all writes with the flag O_DIRECT will use grants which leads to
2014 this problem. Now using OBD_BRW_SYNC to guard this.
2018 Bugzilla : 15713/16362
2019 Description: Assertion in iopen_connect_dentry in 1.6.3
2020 Details : looking up an inode via iopen with the wrong generation number can
2021 populate the dcache with a disconneced dentry while the inode
2022 number is in the process of being reallocated. This causes an
2023 assertion failure in iopen since the inode's dentry list contains
2024 both a connected and disconnected dentry.
2028 Description: assertion failure in ldlm_handle2lock()
2029 Details : fix a race between class_handle_unhash() and class_handle2object()
2030 introduced in lustre 1.6.5 by bug 13622.
2032 Severity : enhancement
2034 Description: superblock lock contention with many SMP cores on one client
2035 Details : several client filesystem locks were highly contended on SMP
2036 NUMA systems with 8 or more cores. Per-CPU datastructures
2037 and more efficient locking implemented to reduce contention.
2042 Description: Kernel BUG: sd_iostats_bump: unexpected disk index
2043 Details : remove the limit of 256 scsi disks in the sd_iostat patch
2048 Description: oops in sd_iostats_seq_show()
2049 Details : unloading/reloading the scsi low level driver triggers a kernel
2050 bug when trying to access the sd iostat file.
2055 Description: Kernel panics during QLogic driver reload
2056 Details : REQ_BLOCK_PC requests are not handled properly in the sd iostat
2057 patch, causing memory corruption.
2062 Description: journal_dev option does not work in b1_6
2063 Details : pass mount option during pre-mount.
2065 Severity : enhancement
2067 Description: Add a FIEMAP(FIle Extent MAP) ioctl for ldiskfs
2068 Details : FIEMAP ioctl will allow an application to efficiently fetch the
2069 extent information of a file. It can be used to map logical blocks
2070 in a file to physical blocks in the block device.
2073 Frequency : only with adaptive timeout enabled
2075 Description: DEBUG_REQ() bad paging request
2076 Details : ptlrpc_at_recv_early_reply() should not modify req->rq_repmsg
2077 because it can be accessed by reply_in_callback() without the
2081 Frequency : only on Cray X2
2083 Description: X2 build failures
2084 Details : fix build failures on Cray X2.
2088 Description: xid & resent requests
2089 Details : Initialize RPC XID from clock at startup (randomly if clock is
2094 Description: quota recovery deadlock during mds failover
2095 Details : This patch includes att18982, att18236, att18237 in bz14840.
2097 1. fix osts hang when mds does failover with quotaon
2098 2. prevent watchdog storm when osts threads wait for the
2103 Description: kernel panic on racer
2104 Details : Do not access dchild->d_inode when IS_ERR(dchild) is true.
2106 Severity : enhancement
2108 Description: Add lustre_start utility to start or stop multiple Lustre servers
2113 Description: Lustre GPF in {:ptlrpc:ptlrpc_server_free_request+373}
2114 Details : In case of memory pressure, list_del() can be called twice on
2115 req->rq_history_list, causing a kernel oops.
2119 Description: (ptllnd_peer.c:557:kptllnd_peer_check_sends()) ASSERTION(!in_interrupt()) failed
2120 Details : fix stack overflow in the distributed lock manager by defering
2121 export eviction after a failed ast to the elt thread instead of
2122 handling it in the dlm interpret routine.
2124 Severity : enhancement
2126 Description: More exported tunables for mballoc
2127 Details : Add support for tunable preallocation window and new tunables for
2128 large/small requests
2132 Description: Detect corruption of block bitmap and checking for preallocations
2133 Details : Checks validity of on-disk block bitmap. Also it does better
2134 checking of number of applied preallocations. When corruption is
2135 found, it turns filesystem readonly to prevent further corruptions.
2139 Frequency : only for big-endian servers
2140 Description: Check for big-endian system while mounting fs with extents feature
2141 Details : Mounting a filesystem with extents feature will fail on big-endian
2142 systems since ext3-based ldiskfs is not supported on big-endian
2143 systems. Can be over-ridden with "bigendian_extents" mount option.
2147 Description: Excessive recovery window
2148 Details : With AT enabled, the recovery window can be excessively long (6000+
2149 seconds). To address this problem, we no longer use
2150 OBD_RECOVERY_FACTOR when extending the recovery window (the connect
2151 timeout no longer depends on the service time, it is set to
2152 INITIAL_CONNECT_TIMEOUT now) and clients report the old service
2153 time via pb_service_time.
2156 Descriptoin: Don't sync journal after every i/o
2157 Details : Implement write RPC replay to allow server replies for write RPCs
2158 before data is on disk. However, this feature is disabled by
2159 default since some issues leading to data corruptions have been
2160 found during recovery (e.g. bug 19128). This feature can be enabled
2161 by running the following command on the OSSs:
2162 lctl set_param obdfilter.*.sync_journal=0
2166 Description: Watchdog triggered on MDS failover
2167 Details : enable OBD_CONNECT_MDT flag when connecting from the MDS so that
2168 the OSTs know that the MDS "UUID" can be reused for the same export
2169 from a different NID, so we do not need to wait for the export to
2174 Descriptoin: Lustre detected file system corruption with inode out of bounds
2175 Details : don't update i_size on MDS_CLOSE for directories. This causes
2176 directory corruptions on the MDT.
2178 -------------------------------------------------------------------------------
2180 2008-05-26 Sun Microsystems, Inc.
2182 * Support for kernels:
2183 2.6.5-7.311 (SLES 9),
2184 2.6.9-67.0.7.EL (RHEL 4),
2185 2.6.16.54-0.2.5 (SLES 10),
2186 2.6.18-53.1.14.el5 (RHEL 5),
2187 2.6.22.14 vanilla (kernel.org)
2188 * Client support for unpatched kernels:
2189 (see http://wiki.lustre.org/index.php?title=Patchless_Client)
2190 2.6.16 - 2.6.22 vanilla (kernel.org)
2191 * Due to problems with nested symlinks and FMODE_EXEC (bug 12652),
2192 we do not recommend using patchless RHEL4 clients with kernels
2193 prior to 2.6.9-55EL (RHEL4U5).
2194 * Recommended e2fsprogs version: 1.40.7-sun1
2195 * Note that reiserfs quotas are disabled on SLES 10 in this kernel.
2196 * RHEL 4 and RHEL 5/SLES 10 clients behaves differently on 'cd' to a
2197 removed cwd "./" (refer to Bugzilla 14399).
2198 * A new quota file format has been introduced in 1.6.5.
2199 The format conversion from prior releases is handled transparently,
2200 but releases older than 1.4.12/1.6.5 will not understand this new
2201 format. The automatic format conversion can be avoided by running
2202 the following command on the MDS before upgrading:
2203 'tunefs.lustre --param="mdt.quota_type=ug1" $MDTDEV'.
2204 For more information, please refer to bugzilla 13904.
2208 Description: quota performance fix
2209 Details : quota data is written in journalled mode instead of ordered to
2210 increase performance
2214 Description: lfs support for human-readable quota grace time strings
2215 Details : lfs setquota -t and lfs quota -t represent quota grace times
2216 in "XXwXXdXXhXXmXXs" format instead of large values in seconds
2219 Frequency : always with o2ib 1.3 and sles10
2221 Description: fix build with SLES10 and o2ib v3.
2222 Details : sles10 uses diffrent name for Module.symver file but configure
2223 assume this file has same name on RHEL/SLES/vanila kernels.
2226 Frequency : very rare, if additional xattrs are used on kernels >= 2.6.12
2228 Description: MDS may lose file striping (and hence file data) in some cases
2229 Details : If there are additional extended attributes stored on the MDS,
2230 in particular ACLs, SELinux, or user attributes (if user_xattr
2231 is specified for the client mount options) then there is a risk
2232 of attribute loss. Additionally, the Lustre file striping
2233 needs to be larger than default (e.g. striped over all OSTs),
2234 and an additional attribute must be stored initially in the
2235 inode and then increase in size enough to be moved to the
2236 external attribute block (e.g. ACL growing in size) for file
2239 Severity : enhancement
2241 Description: add message levels for liblustreapi
2246 Description: MDT cannot be unmounted, reporting "Mount still busy"
2247 Details : Mountpoint references were being leaked during open reply
2248 reconstruction after an MDS restart. Drop mountpoint reference
2249 in reconstruct_open() and free dentry reference also.
2254 Description: fix for occasional failure case of -ENOSPC in recovery-small tests
2255 Details : Move the 'good_osts' check before the 'total_bavail' check. This
2256 will result in an -EAGAIN and in the exit call path we call
2257 alloc_rr() which will with increasing aggressiveness attempt to
2258 aquire precreated objects on the minimum number of required OSCs.
2262 Description: Use old size assignment to avoid deadlock
2263 Details : This reverts the changes in bugs 2369 and bug 14138 that introduced
2264 the scheduling while holding a spinlock. We do not need locking
2265 for size in ll_update_inode() because size is only updated from
2266 the MDS for directories or files without objects, so there is no
2267 other place to do the update, and concurrent access to such inodes
2268 are protected by the inode lock.
2272 Description: Use __u64 instead of int for valid bits
2276 Description: resolve "_IOWR redefined" build error on SLES10
2280 Description: dump the memory debugging after all modules are unloaded to
2281 suppress false negative in conf_sanity test 39
2285 Description: the recovery timer never expires
2286 Details : for new client connect request, the recovery timer should not be
2287 reset, otherwise recovery timer will never expired, if the old
2288 client never come. Only old client connect and first connection
2289 req should trigger recovery timer reset.
2293 Description: the min numbers of lproc stats are wrong
2294 Details : adding a new constant LC_MIN_INIT and use it for initialization
2298 Frequency : always with interactive lfs
2300 Description: Reinitialize optind to 0 so that interactive lfs works in all cases
2303 Frequency : with multiple concurrent readdir processes in same directory
2304 Bugzilla : 15406, 15169, 15175
2305 Description: misc fixes for directory readahead.
2306 Details : prevent previous statahead async RPC callback from processing the
2307 current "statahead_info", race condition between async RPC callback
2308 add dentry into dentry hash table and "ls" thread revalidate such
2309 dentry, statahead his/miss control for hidden items, and so on.
2311 Severity : enhancement
2313 Description: build kernel-ib packages for OFED 1.3 in our release cycle
2317 Description: incore types cleaning in quota code (with respect to 64-bit limits)
2318 Details : several u32 variables declarations are replaced with u64 declarations
2323 Description: fix SLES kernel versioning
2324 Details : the kernel version for our SLES 10 kernel did not include a "-"
2325 before the "smp" at the end. while this was not a problem in
2326 general, it did mean that software trying to use the kernel
2327 version to try to detect a vendor specific kernel would fail.
2328 this was most evident by the OFED build scripts.
2333 Description: Don't update lov_desc members until making sure they are valid
2334 Details : When updating lov_desc members via proc fs, need fix their
2335 validities before doing the real update.
2338 Frequency : very rare
2340 Description: don't put request into delay list while invalidate in flight.
2341 Details : ptlrpc_delay_request sometimes put in delay list while invalidate
2342 import in flight. this produce timeout for invalidate and sometimes
2343 can cause stale data.
2345 Severity : enhancement
2347 Description: Update kernel to SLES9 2.6.5-7.311.
2349 Severity : enhancement
2351 Description: Update kernel to RHEL4 2.6.9-67.0.7.
2356 Frequency : on PPC only
2357 Description: not convert ost objects for directory because it's not exist.
2358 Details : ll_dir_getstripe assume dirrectory has ost objects but this wrong.
2360 Severity : enhancement
2362 Description: Fix warnings with compile liblustre at sles10/rhel5 which have
2363 __u64 as usingied long long type.
2366 Frequency : rare, on shutdown
2368 Description: race process ast vs remove callback
2369 Details : removing callback before disconnect import open race with
2370 processing callback.
2372 Severity : enhancement
2374 Description: Update kernel to SLES9 2.6.5-7.311.
2376 Severity : enhancement
2378 Description: Files open for execute are not marked busy on SLES10
2379 Details : Add FMODE_EXEC to SLES10 SP1 server kernel series.
2381 Severity : enhancement
2383 Description: Add server support for vanilla-2.6.22.14.
2386 Frequency : occasional
2388 Description: Avoid lov_create() getting stuck in obd_statfs_rqset()
2389 Details : If an OST is down the MDS will hang indefinitely in
2390 obd_statfs_rqset() waiting for the statfs data. While for
2391 MDS QOS usage of statfs, it should not stuck in waiting.
2393 Severity : enhancement
2395 Description: Disable adaptive timeouts by default
2398 Frequency : on network error
2400 Description: panic with double free request if network error
2401 Details : mdc_finish_enqueue is finish request if any network error occuring,
2402 but it's true only for synchronus enqueue, for async enqueue
2403 (via ptlrpcd) this incorrect and ptlrpcd want finish request
2407 Frequency : rare, on recovery
2409 Description: read procfs can produce deadlock in some situation
2410 Details : Holding lprocfs lock with send rpc can produce block for destroy
2411 obd objects and this also block reconnect with -EALREADY.
2412 This isn't fix all lprocfs bugs - but make it rare.
2414 Severity : enhancement
2416 Description: Update kernel to RHEL5 2.6.18-53.1.14.el5.
2419 Frequency : frequent on X2 node
2421 Description: mdc_set_open_replay_data LBUG
2422 Details : Set replay data for requests that are eligible for replay.
2427 Description: lustre_mgs: operation 101 on unconnected MGS
2428 Details : When MGC is disconnected from MGS long enough, MGS will evict the
2429 MGC, and late on MGC cannot successfully connect to MGS and a lot
2430 of the error messages complaining that MGS is not connected.
2433 Frequency : rare, depends on device drivers and load
2435 Description: MDS or OSS nodes crash due to stack overflow
2436 Details : Code changes in 1.6.4 increased the stack usage of some functions.
2437 In some cases, in conjunction with device drivers that use a lot
2438 of stack, the MDS (or possibly OSS) service threads could overflow
2439 the stack. One change which was identified to consume additional
2440 stack has been reworked to avoid the extra stack usage.
2442 Severity : enhancement
2444 Description: Update to RHEL5 latest kernel-2.6.18-53.1.13.el5.
2446 Severity : enhancement
2448 Description: Update to SLES10 SP1 latest kernel-2.6.16.54-0.2.5.
2450 Severity : enhancement
2452 Description: Update to RHEL5 latest kernel-2.6.18-53.1.6.el5.
2454 Serverity : enhancement
2456 Description: Update RHEL4 kernel to 2.6.9-67.0.4.
2459 Frequency : rare on shutdown OST
2461 Description: Don't allow skipping OSTs if index has been specified.
2462 Details : Don't allow skipping OSTs if index has been specified, make locking
2463 in internal create lots better.
2468 Description: ASSERTION(!PageDirty(page)) failed
2469 Details : Wrong check could lead to an assertion failure under specific
2475 Description: LBUG in ptlrpc_check_set() bad phase ebc0de00
2476 Details : access to bitfield in structure is always rounded to long
2477 and this produce problem with not atomic change any bit.
2482 Description: Lustre make rpms failed.
2483 Details : Remove ldiskfs spec file to avoids rpmbuild be confused when
2484 builds Lustre rpms from tarball.
2486 Severity : enhancement
2488 Description: Update to SLES9 SP4 kernel-2.6.5-7.308.
2491 Frequency : rare on shutdown OST
2493 Description: If llog cancel was not send before clean_exports phase, this can
2494 produce deadlock in llog code.
2495 Details : If llog thread has last reference to obd and call class_import_put
2496 this produce deadlock because llog_cleanup_commit_master wait when
2497 last llog_commit_thread exited, but this never success because was
2498 called from llog_commit_thread.
2501 Frequency : only if OST index is skipped
2503 Description: NULL lov_tgts causing MDS oops
2504 Details : more safe checks for NULL lov_tgts for avoid oops.
2506 Severity : enhancement
2508 Description: Update to RHEL4 latest kernel-2.6.9-67.0.1.EL.
2510 Severity : enhancement
2512 Description: Update to RHEL5 latest kernel-2.6.18-53.1.4.el5.
2517 Description: make mgs_setparam() handle fsname containing dash
2518 Details : fsname containing a dash does not work with lctl conf_param
2520 Severity : enhancement
2522 Description: Update to RHEL4 Update-6 kernel-2.6.9-67.EL.
2525 Frequency : rare, in recovery and (or) destroy lovobjid file.
2527 Description: rewrite lov objid code.
2528 Details : Cleanup for lov objid code, remove scability problems and wrong
2529 locking. Fix sending last_id into ost.
2531 Severity : enhancement
2533 Description: Update to SLES10 SP1 latest kernel-2.6.16.54-0.2.3.
2535 Severity : enhancement
2537 Description: Update to RHEL5 Update-1 kernel 2.6.18-53.el5.
2538 Details : Use d_move_locked instead of __d_move.
2541 Frequency : rare, at shutdown
2543 Description: access already free / zero obd_namespace.
2544 Details : if client_disconnect_export was called without force flag set,
2545 and exist connect request in flight, this can produce access to
2546 NULL pointer (or already free pointer) when connect_interpret
2547 store ocd flags in obd_namespace.
2550 Frequency : only at startup
2552 Description: not alloc memory with spinlock held.
2553 Details : allocation memory with GFP_KERNEL can produce sleep deadlock,
2554 if any spinlock held.
2559 Description: lfs find does not continue on file error
2560 Details : Continue other files processing when a file/dir is absent.
2564 Description: Inconsistent usage of lustre_pack_reply()
2565 Details : Standardize the usage of lustre_pack_reply() such that it
2566 always generate a CERROR on failure.
2569 Frequency : very rare
2571 Description: Fix replay if there is an un-replied request and open
2572 Details : In some cases, older replay request will revert the
2573 mcd->mcd_last_xid on MDS which is used to record the client's
2574 latest sent request.
2576 Severity : enhancement
2578 Description: Update to RHEL5 kernel 2.6.18-8.1.15.el5.
2580 Severity : enhancement
2582 Description: Update to SLES10 SP1 kernel 2.6.16.53-0.16
2584 Severity : enhancement
2586 Description: Update to SLES9 kernel-2.6.5-7.287.3.
2588 Severity : enhancement
2590 Description: Update to RHEL4 kernel-2.6.9-55.0.12.EL.
2592 Severity : enhancement
2594 Description: Build SLES10 patchless client fails
2595 Details : The configure was broken by run ./configure with
2596 --with-linux-obj=.... argument for patchless client. When the
2597 configure use --with-linux-obj, the LINUXINCLUDE= -Iinclude
2598 can't search header adequately. Use absolute path such as
2599 -I($LINUX)/include instead.
2601 Severity : enhancement
2603 Description: Lustre Page Accounting
2604 Details : New macros for page alloc and free which enable accounting
2605 of page allocation of Lustre. Use percpu counters to store memory
2606 and page statistics.
2609 Frequency : only if debugging is disabled
2611 Description: LASSERT_{REQ,REP}SWAB macros are buggy
2612 Details : If SWAB_PARANOIA is disabled, the LASSERT_REQSWAB and
2613 LASSERT_REPSWAB macros become no-ops, which is incorrect. Drop
2614 these macros and replace them with their definitions instead.
2619 Description: interrupt oig_wait produce painc on resend.
2620 Details : brw_redo_request can be used for resend requests from ptlrpcd and
2621 private set, and this produce situation when rq_ptlrpcd_data not
2622 copyed to new allocated request and triggered LBUG on assert
2623 req->rq_ptlrpcd_data != NULL. But this member used only for wakeup
2624 ptlrpcd set if request is changed and can be safety changed to use
2627 Severity : enhancement
2629 Description: organize the server-side client stats on per-nid basis
2630 Details : Change the structure of stats under obdfilter and mds to
2638 The "uuid"s file would list the uuids of _active_ exports.
2639 And the clear entry is to clear all stats and stale nids.
2644 Description: Processes looping in ll_readdir() on Lustre clients finally causing
2645 a full node pseudo-hang
2646 Details : Concurrent access to the same directory from multiple clients with
2647 intensive file creation/removal can cause a client node to spin in
2648 ll_readdir(). i_version must be increased every time the lock is
2649 cancelled to ensure a revalidate is done.
2654 Description: touch file failed when fs is not full
2655 Details : OST in recovery should not be discarded by MDS in alloc_qos(),
2656 otherwise we can get ENOSP while fs is not full.
2659 Frequency : only for Cray XT3
2660 Bugzilla : 12829/13455
2661 Description: Changing primary group doesn't change the group lustre assigns to
2663 Details : When CRAY_XT3 is defined, the fsgid supplied by the client is
2664 overridden with the primary group provided by the group upcall,
2665 whereas the supplied fsgid can be trusted if it is in the list of
2666 supplementary groups returned by the group upcall.
2668 Severity : enhancement
2670 Description: Root Squash Functionality
2671 Details : Implementation of NFS-like root squash capability. Specifically,
2672 don't allow someone with root access on a client node to be able
2673 to manipulate files owned by root on a server node.
2675 Severity : enhancement
2677 Description: Slow trucate/writes to huge files at high offsets.
2678 Details : Directly associate cached pages to lock that protect those pages,
2679 this allows us to quickly find what pages to write and remove
2680 once lock callback is received.
2685 Description: Too many locks accumulating on client during NFS usage
2686 Details : mds_open improperly used accmode to find out access mode to a
2687 file. Also mdc_intent_lock logic to find out if we already have
2688 lock similar to just received was flawed since introduction of
2689 skiplists - locks are now added to the front of the granted
2694 Description: Hit ASSERTION(obd->obd_stopping == 1) failed in some setup failed
2696 Details : In obd setup failure handler, obd_stopping will not necessarily to
2697 be 1, and obd_set_up should also be checked to make sure whether
2698 obd is completely setup.
2700 Severity : enhancement
2702 Description: Allow masking D_WARNING, D_ERROR messages from console
2703 Details : Console messages can now be disabled via lnet.printk.
2708 Description: User code with malformed file open parameter crashes client node
2709 Details : Before packing join_file req, all the related reference should be
2710 checked carefully in case some malformed flags cause fake join_file
2716 Description: shrink/enlarge qunit size when needed; fix the problem of coarse
2717 grain of quota doing harm to quota's accuracy
2718 Details : qunit size will be changed when quota limitation is too low/high;
2719 record the pending quota write in order to get more accureate
2720 quota; delete the patch for bug12588, which is unnecessary when
2721 this patch is landed. This bug also contains fixes for bug 14526,
2722 14299, 14601 and 13794.
2726 Description: LDLM_ENQUEUE races with LDLM_CP_CALLBACK
2727 Details : ldlm_completion_ast() assumes that a lock is granted when the req
2728 mode is equal to the granted mode. However, it should also check
2729 that LDLM_FL_CP_REQD is not set.
2733 Description: Heavy nfs access might result in deadlocks
2734 Details : After ELC code landed, it is now improper to enqueue any mds
2735 locks under och_sem, because enqueue might want to decide to
2736 cancel open locks for same inode we are holding och_sem for.
2740 Description: 35% write performance drop with ldiskfs2 when quotas are on
2741 Details : Enable ext3 journalled quota by default to improve performance
2742 when quotas are turned on.
2746 Description: Client eviction while running blogbench
2747 Details : A lot of unlink operations with concurrent I/O can lead to a
2748 deadlock causing evictions. To address the problem, the number of
2749 oustanding OST_DESTROY requests is now throttled to
2750 max_rpcs_in_flight per OSC and LDLM_FL_DISCARD_DATA blocking
2751 callbacks are processed in priority.
2754 Frequency : RHEL4 only
2756 Description: mkfs is very slow on IA64/RHEL4
2757 Details : A performance regression has been discovered in the MPT Fusion
2758 driver between versions 3.02.73rh and 3.02.99.00rh. As a
2759 consequence, we have downgraded the MPT Fusion driver in the RHEL4
2760 kernel from 3.02.99.00 to 3.02.73 until this problem is fixed.
2763 Frequency : PPC/PPC64 only
2765 Description: conflicts between asm-ppc64/types.h and lustre_types.h
2766 Details : fix duplicated definitions between asm-ppc64/types.h and
2767 lustre_types.h on PPC.
2770 Frequency : PPC/PPC64 only
2772 Description: asm-ppc/segment.h does not exist
2773 Details : fix compile issue on PPC.
2777 Description: data checksumming impacts single node performance
2778 Details : add support for several checksum algorithms. Currently, CRC32 and
2779 Adler-32 are supported. The checksum type can be changed on the fly
2780 through /proc/fs/lustre/osc/*/checksum_type.
2784 Description: use adler32 for page checksums
2785 Details : when available, use the Adler-32 algorithm instead of CRC32 for
2790 Description: better handle error messages in extents code
2792 Severity : enhancement
2794 Description: SNMP support enhancement
2795 Details : Adding total number of sampled request for an MDS node in snmp
2798 Severity : enhancement
2800 Description: Optimize ldlm waiting list processing for PR extent locks
2801 Details : When processing waiting list for read extent lock and meeting read
2802 lock that is same or wider to it that is not contended, skip
2803 processing rest of the list and immediatelly return current
2804 status of conflictness, since we are guaranteed there are no
2805 conflicting locks in the rest of the list.
2809 Description: Time out and refuse to reconnect
2810 Details : When the failover node is the primary node, it is possible
2811 to have two identical connections in imp_conn_list. We must
2812 compare not conn's pointers but NIDs, otherwise we
2813 can defeat connection throttling.
2817 Description: Client not clear own cache if answer to reconnect is lost.
2818 Details : Client gets evicted from server. Now client also thinks it is
2819 disconnected (or gets ENOTCONN on its operation) and decides to
2820 reconnect. Server receives reconnect message, but cannot find
2821 export. New export is created that is fully valid (new cookie!),
2822 but reply is lost and not reported to client. Client reconnects
2823 again and gets back a just-created connection, but it is not new
2824 so client thinks it was not evicted and does not do recovery.
2828 Description: Detect stride IO mode in read-ahead
2829 Details : When a client does stride read, read-ahead should detect that and
2830 read-ahead pages according to the detected stride pattern.
2834 Description: build for x2 fails
2835 Details : fix compile issue on Cray systems.
2837 Severity : enhancement
2839 Description: implement readv/writev APIs(aio_read/aio_writes in newer kernels)
2840 Details : This greatly improves speed of NFS writes on 2.6 kernels.
2843 Frequency : only on PPC/SLES10
2845 Description: "BITS_PER_LONG is not 32 or 64" in linux/idr.h
2846 Details : On SLES10/PPC, fs.h includes idr.h which requires BITS_PER_LONG to
2847 be defined. Add a hack in mkfs_lustre.c to work around this compile
2852 Description: LASSERT on MDS when client holding flock lock dies
2853 Details : ldlm pool logic depends on number of granted locks equal to
2854 number of released locks which is not true for flock locks, so
2855 just exclude such locks from consideration.
2859 Description: MDS deadlock with many ll_sync_lov threads and I/O stalled
2860 Details : Use fsfilt_sync() for both the whole filesystem sync and
2861 individual file sync to eliminate dangerous inode locking
2862 with I_LOCK that can lead to a deadlock.
2866 Description: Update an obsolete wirecheck.c generator
2867 Details : Update wirecheck.c/wirehdr.c and regenerate wiretest.c
2871 Description: Client can panic on open sometimes
2872 Details : It is possible that we try to free already freed request in
2873 ll_file_open in some error cases when we send request from
2878 Description: performance in 1.6.3
2879 Details : Force q->max_phys_segments to MAX_PHYS_SEGMENTS on SLES10 to be
2880 sure that 1MB requests are not fragmented by the block layer.
2884 Description: LDLM soft lockups - improvement
2885 Details : It is be possible to send the lock handle along with each read
2886 or write request because the client is already doing a lock match
2887 itself so there isn't any reason the OST should have to re-do that
2893 Description: lfs quota fails with deactivated OSTS
2894 Details : With this patch, three improvements are included:
2895 1. detete the softlimit in mds and osts when use "lfs quota".
2896 2. display the inaccurate data in the output of "lfs quota".
2897 3. try to get quota info when "lfs quota" is executed.
2902 Description: Extent locks not granted with no conflicts sometimes.
2903 Details : When race occurs in glimpse handler and nothing is returned,
2904 we do not reprocess the queue after lock cancel, and that leads
2905 to a stall until next activity on a resource
2908 Frequency : failover with quotaon
2910 Description: during mds failovers with quota on, OSTs got into deadlock state
2911 and causing dumpstack.
2912 Details : for every quota slave, at any time, there is only one quota req
2913 is sent to quota master for every uid/gid. Before that quota req
2914 returns, all the thread relative to the same uid/gid will wait.
2915 So if the quota req is lost because mds failovers or any other
2916 reasons, this bug will be hit. Now, dqacq_interpret() will handle
2917 quota reqs who time out.
2919 Severity : enhancement
2922 Description: when quota slave checks if quota is enough, there is an unnecessary
2924 Details : place this wait on necessary place instead of always waiting.
2926 --------------------------------------------------------------------------------
2928 2007-12-07 Cluster File Systems, Inc. <info@clusterfs.com>
2930 * Support for kernels:
2931 2.6.5-7.286 (SLES 9),
2932 2.6.9-55.0.9.EL (RHEL 4),
2933 2.6.16.53-0.8 (SLES 10),
2934 2.6.18-8.1.14.el5 (RHEL 5),
2935 2.6.18.8 vanilla (kernel.org)
2936 * Client support for unpatched kernels:
2937 (see http://wiki.lustre.org/index.php?title=Patchless_Client)
2938 2.6.16 - 2.6.22 vanilla (kernel.org)
2939 * Due to recently discovered recovery problems, we do not recommend
2940 using patchless RHEL 4 clients with this or any earlier release.
2941 * Recommended e2fsprogs version: 1.40.2-cfs1
2942 * Note that reiserfs quotas are disabled on SLES 10 in this kernel.
2945 Frequency : occasional
2947 Description: MDS hang or stay in waiting lock
2948 Details : If client receive lock with CBPENDING flag ldlm need send lock
2949 cancel as separate rpc, to avoid situation when cancel request
2950 can't processed due all i/o threads stay in wait lock.
2953 Frequency : occasional
2955 Description: Do not fail import if osc_interpret_create gets -EAGAIN
2956 Details : If osc_interpret_create got -EAGAIN it immediately exits and
2957 wakeup oscc_waitq. After wakeup oscc_wait_for_objects call
2958 oscc_has_objects and see OSC has no objests and call
2959 oscc_internal_create to resend create request.
2961 Severity : enhancement
2963 Description: Update kernel patches for SLES10 2.6.16.53-0.8.
2964 Details : Update which_patch & target file for SLES10 latest kernel.
2966 Severity : enhancement
2968 Description: add --type and --size parameters to lfs find
2969 Details : Enhance lfs find by adding filetype and filesize parameters. Also
2970 multiple OBDs can now be specified for the --obd option.
2972 Severity : enhancement
2974 Description: eliminate client locks in face of contention
2975 Details : file contention detection and lockless i/o implementation
2976 for contended files.
2978 Severity : enhancement
2980 Description: Remove client patches from SLES 10 kernel.
2981 Details : This causes SLES 10 clients to behave as patchless clients
2982 even on a Lustre-patched (server) kernel.
2984 Severity : enhancement
2986 Description: use i_size_read and i_size_write in 2.6 port
2987 Details : replace inode->i_size access with i_size_read/write()
2990 Frequency : when removing large files
2992 Description: scheduling issue during removal of large Lustre files
2993 Details : Don't take the BKL in fsfilt_ext3_setattr() for 2.6 kernels.
2994 It causes scheduling issues when removing large files (17TB in the
3000 Description: 1.4.11 Can't handle directories with stripe set and extended ACLs
3001 Details : Impossible (EPROTO is returned) to access a directory that has a
3002 non-default striping and ACLs.
3005 Frequency : only on ppc
3007 Description: /proc/fs/lustre/devices broken on ppc
3008 Details : The patch as applied to 1.6.2 doesn't look correct for all arches.
3009 We should make sure the type of 'index' is loff_t and then cast
3010 explicitly as needed below. Do not assign an explicitly cast
3014 Frequency : only for rhel5
3016 Description: Kernel patches update for RHEL5 2.6.18-8.1.10.el5.
3017 Details : Modify the target file & which_kernel.
3020 Frequency : if the uninit_groups feature is enabled on ldiskfs
3022 Description: e2fsck reports "invalid unused inodes count"
3023 Details : If a new ldiskfs filesystem is created with the "uninit_groups"
3024 feature and only a single inode is created in a group then the
3025 "bg_unused_inodes" count is incorrectly updated. Creating a
3026 second inode in that group would update it correctly.
3031 Description: buffer overruns could theoretically occur
3032 Details : llapi_semantic_traverse() modifies the "path" argument by
3033 appending values to the end of the origin string, and a buffer
3034 overrun may occur. Adding buffer overrun check in liblustreapi.
3036 Severity : enhancement
3038 Description: Add jbd statistics patch for RHEL5 and 2.6.18-vanilla.
3041 Frequency : only if filesystem is inconsistent
3043 Description: handle "serious error: objid * already exists" more gracefully
3044 Details : If LAST_ID value on disk is smaller than the objects existing in
3045 the O/0/d* directories, it indicates disk corruption and causes an
3046 LBUG(). If the object is 0-length, then we should use the existing
3047 object. This will help to avoid a full fsck in most cases.
3049 Severity : enhancement
3051 Description: Kernel patches update for RHEL4 2.6.9-55.0.6.
3052 Details : Modify vm-tunables-rhel4.patch.
3054 Severity : enhancement
3056 Description: Kernel config for 2.6.18-vanilla.
3057 Details : Modify targets/2.6-vanilla.target.in.
3058 Add config file kernel-2.6.18-2.6-vanilla-i686.config.
3059 Add config file kernel-2.6.18-2.6-vanilla-i686-smp.config.
3060 Add config file kernel-2.6.18-2.6-vanilla-x86_64.config.
3061 Add config file kernel-2.6.18-2.6-vanilla-x86_64-smp.config.
3064 Frequency : occasional
3066 Description: improve handling recoverable errors
3067 Details : If request processed with error which can be recoverable on server
3068 request should be resend, otherwise page released from cache and
3074 Description: Kernel patches update for RHEL5 2.6.18-8.1.14.el5.
3075 Details : Modify target file & which_patch.
3076 A flaw was found in the IA32 system call emulation provided
3077 on AMD64 and Intel 64 platforms. An improperly validated 64-bit
3078 value could be stored in the %RAX register, which could trigger an
3079 out-of-bounds system call table access. An untrusted local user
3080 could exploit this flaw to run code in the kernel
3081 (ie a root privilege escalation). (CVE-2007-4573).
3085 Description: change order of libsysio includes
3086 Details : '#include sysio.h' should always come before '#include xtio.h'
3088 Severity : enhancement
3090 Description: adapt the lustre_config script to support the upgrade case
3091 Details : Add "-u" option for lustre_config script to support upgrading 1.4
3092 server targets to 1.6 in parallel.
3097 Description: To avoid grant space > avaible space when the disk is almost
3098 full. Without this patch you might see the error "grant XXXX >
3099 available" or some LBUG about grant, when the disk is almost
3101 Details : In filter_check_grant, for non_grant cache write, we should
3102 check the left space by if (*left > ungranted + bytes), instead
3103 of (*left > ungranted), because only we are sure the left space
3104 is enough for another "bytes", then the ungrant space should be
3105 increase. In client, we should update cl_avail_grant only there is
3106 OBD_MD_FLGRANT in the reply.
3111 Description: Update RHEL 4 kernel to fix local root privilege escalation.
3112 Details : Update to the latest RHEL 4 kernel to fix the vulnerability
3113 described in CVE-2007-4573. This problem could allow untrusted
3114 local users to gain root access.
3117 Frequency : when using O_DIRECT and quotas
3119 Description: Incorrect file ownership on O_DIRECT output files
3120 Details : block usage reported by 'lfs quota' does not take into account
3121 files that have been written with O_DIRECT.
3125 Description: (rw.c:1323:ll_read_ahead_pages()) ASSERTION(page_idx > ria->ria_stoff) failed
3126 Details : Once the unmatched stride IO mode is detected, shrink the stride-ahead
3127 window to 0. If it does hit cache miss, and read-pattern is still
3128 stride-io mode, does not reset the stride window, but also does not
3129 increase the stride window length in this case.
3131 --------------------------------------------------------------------------------
3133 2007-09-27 Cluster File Systems, Inc. <info@clusterfs.com>
3135 * Support for kernels:
3136 2.6.5-7.286 (SLES 9),
3137 2.6.9-55.0.2.EL (RHEL 4),
3138 2.6.16.46-0.14 (SLES 10),
3139 2.6.18-8.1.8.el5 (RHEL 5),
3140 2.6.18.8 vanilla (kernel.org)
3141 * Client support for unpatched kernels:
3142 (see http://wiki.lustre.org/index.php?title=Patchless_Client)
3143 2.6.16 - 2.6.21 vanilla (kernel.org)
3144 * Due to recently discovered recovery problems, we do not recommend
3145 using patchless RHEL 4 clients with this or any earlier release.
3146 * Recommended e2fsprogs version: 1.40.2-cfs1
3147 * Note that reiserfs quotas are disabled on SLES 10 in this kernel.
3151 Description: Fix errors in lfs documentation
3152 Details : Fixes man pages
3154 Severity : enhancement
3156 Description: Adaptive timeouts
3157 Details : RPC timeouts adapt to changing server load and network
3158 conditions to reduce resend attempts and improve recovery time.
3160 Severity : enhancement
3162 Description: llapi_file_create() does not allow some changes
3163 Details : add llapi_file_open() that allows specifying the file creation
3164 mode and open flags, and also returns an open file handle.
3166 Severity : enhancement
3168 Description: df doesn't work properly if diskfs blocksize != 4K
3169 Details : Choose biggest blocksize of OST's as the LOV's blocksize.
3172 Frequency : always on directories with default striping set
3174 Description: lfs find on -1 stripe looping in lsm_lmm_verify_common()
3175 Details : Avoid lov_verify_lmm_common() on directory with -1 stripe count.
3177 Severity : enhancement
3179 Description: merge and cleanup kernel patches.
3180 Details : Remove mnt_lustre_list in vfs_intent-2.6-rhel4.patch.
3182 Severity : enhancement
3184 Description: RedHat Update kernel for RHEL5
3185 Details : Kernel config file for RHEL5.
3187 Severity : enhancement
3189 Description: OSS needs mutliple precreate threads
3190 Details : Add ability to start more than one create thread per OSS.
3193 Frequency : only with quota on the root user
3195 Description: mds_obd_create error creating tmp object
3196 Details : When the user sets quota on root, llog will be affected and can't
3197 create files and write files.
3200 Frequency : Always on ia64 patchless client, and possibly others.
3202 Description: Add EXPORT_SYMBOL check for node_to_cpumask symbol.
3203 Details : This allows the patchless client to be loaded on architectures
3204 without this export.
3209 Description: disorder of journal start and llog_add cause deadlock.
3210 Details : in llog_origin_connect, journal start should happen before
3211 llog_add keep the same order as other functions to avoid
3214 Severity : enhancement
3216 Description: RedHat Update kernel for RHEL5
3217 Details : Modify the kernel config file more closer RHEL5.
3219 Severity : enhancement
3221 Description: Build failure against Centos5 (RHEL5)
3222 Details : Define PAGE_SIZE when it isn't present.
3225 Frequency : occasionally when using NFS
3227 Description: "ll_intent_file_open()) lock enqueue: err: -13" with nfs
3228 Details : with NFS, the anon dentry's parent was set to itself in
3229 d_alloc_anon(), so in MDS, we use rec->ur_fid1 to find the
3230 corresponding dentry other than use rec->ur_name.
3232 Severity : enhancement
3234 Description: enable data checksumming by default
3235 Details : enable checksum by default, allow --disable-checksum
3236 configure option and "-o nochecksum" mount option. Checksums
3237 can also be disabled at runtime via $LPROC/osc/*/checksum_pages.
3240 Frequency : Occasionally with failover
3242 Description: Client eviction due to failover config
3243 Details : after a connection loss, the lustre client should attempt to
3244 reconnect to the last active server first before trying the
3245 other potential connections.
3247 Severity : enhancement
3249 Description: client-side metadata stat-ahead during readdir(directory readahead)
3250 Details : perform client-side metadata stat-ahead when the client detects
3251 readdir and sequential stat of dir entries therein
3256 Description: when mds and osts use different quota unit(32bit and 64bit),
3257 quota will be released repeatly.
3258 Details : void sending multiple quota reqs to mds, which will keep the
3259 status between the reqs.
3262 Frequency : only with liblustre clients on XT3
3264 Description: evictions taking too long
3265 Details : allow llrd to evict clients directly on OSTs
3269 Description: osts not allocated evenly to files
3270 Details : change the condition to increase offset_idx
3272 Severity : enhancement
3274 Description: self-adjustable client's lru lists
3275 Details : use adaptive algorithm for managing client cached locks lru
3276 lists according to current server load, other client's work
3277 pattern, memory activities, etc. Both, server and client
3278 side namespaces provide number of proc tunables for controlling
3283 Description: rewrite ext2-derived code in llite/dir.c and obdclass/uuid.c
3284 Details : rewrite inherited code (uuid parsing code from ext2 utils and
3285 readdir code from ext3) from scratch preserving functionality.
3289 Description: Only those disconnect error should be returned by rq_status.
3290 Details : In open/enqueue processs, Some errors, which will cause client
3291 disconnected, should be returned by rq_status, while other
3292 errors should still be returned by intent, then mdc or llite will
3295 Severity : enhancement
3297 Description: Tune the kernel for good SCSI performance.
3298 Details : Set the value of /sys/block/{dev}/queue/max_sectors_kb
3299 to the value of /sys/block/{dev}/queue/max_hw_sectors_kb
3303 Frequency : Always for filesystems larger than 2TB on 32-bit systems.
3304 Bugzilla : 13547, 13627
3305 Description: Data corruption for OSTs that are formatted larger than 2TB
3307 Details : When generating the bio request for lustre file writes the
3308 sector number would overflow a temporary variable before being
3309 used for the IO. The data reads correctly from Lustre (which
3310 will overflow in a similar manner) but other file data or
3311 filesystem metadata may be corrupted in some cases.
3315 Description: "lfs find -obd UUID" prints directories
3316 Details : "lfs find -obd UUID" will return all directory names instead
3317 of just file names. It is incorrect because the directories
3318 do not reside on the OSTs.
3321 Bugzilla : 13596, 13614
3322 Description: MDS hang after unclean shutdown of lots of clients
3323 Details : Never resend AST requests.
3327 Frequency : Always, for kernels after 2.6.16
3328 Description: Fix warning idr_remove called for id=.. which is not allocated.
3329 Details : Last kernels save old s_dev before kill super and not allow
3330 to restore from callback - restore it before call kill_anon_super.
3334 Description: Fix error on 'ls .' at the top of the Lustre mount.
3335 Details : Don't revalidate dentry if it is a root dentry.
3337 --------------------------------------------------------------------------------
3339 2007-08-27 Cluster File Systems, Inc. <info@clusterfs.com>
3341 * Support for kernels:
3342 2.6.5-7.286 (SLES 9),
3343 2.6.9-55.0.2.EL (RHEL 4),
3344 2.6.16.46-0.14 (SLES 10),
3345 2.6.18-8.1.8.el5 (RHEL 5),
3346 2.6.18.8 vanilla (kernel.org)
3347 * Client support for unpatched kernels:
3348 (see http://wiki.lustre.org/index.php?title=Patchless_Client)
3349 2.6.16 - 2.6.22 vanilla (kernel.org)
3350 * Due to recently discovered recovery problems, we do not recommend
3351 using patchless RHEL 4 clients with this or any earlier release.
3352 * Recommended e2fsprogs version: 1.39.cfs8
3353 * Note that reiserfs quotas are disabled on SLES 10 in this kernel.
3358 Description: block reactivating mgc import until all deactivates complete
3359 Details : Fix race when failing back MDT/MGS to itself (testing)
3361 Severity : enhancement
3363 Description: lfs setstripe enhancement
3364 Details : Make lfs setstripe understand 'k', 'm' and 'g' for stripe size.
3367 Frequency : mds/oss recovery
3369 Description: llog ctxt is refrenced after it has been freed.
3370 Details : llog ctxt refcount was added to avoide the race between ctxt free
3371 and llog recovery process. Each llog user must hold ctxt refcount
3372 before it access the llog. And the llog ctxt can only be freed
3373 when its refcount is zero.
3375 Severity : enhancement
3377 Description: randomly memory allocation failure util
3378 Details : Make lustre randomly failed allocating memory for testing purpose.
3381 Frequency : only for SLES10
3383 Description: Update kernel patch for SLES10 SP1
3384 Details : Add patch blkdev_tunables-2.6-sles10.patch to 2.6-sles10.series.
3386 Severity : enhancement
3388 Description: omit set fsid for export NFS
3389 Details : fix set/restore device id for avoid EMFILE error and mark lustre fs
3390 as FS_REQUIRES_DEV for avoid problems with generate fsid.
3393 Frequency : after network failures
3395 Description: Add sync option to mount_lustre.c
3396 Details : Client loses data written to lustre after a network interruption.
3398 Severity : enhancement
3400 Description: Error message improvement.
3401 Details : Merging of two LCONSOLE_ERROR_MSG into one.
3403 Severity : enhancement
3405 Description: don't use GFP_* in generic Lustre code.
3406 Details : Use cfs_alloc_* functions and CFS_* flags for code portability.
3408 Severity : enhancement
3410 Description: obdclass is limited by single OBD_ALLOC(idarray)
3411 Details : replace OBD_ALLOC/OBD_FREE with OBD_VMALLOC/OBD_VFREE
3413 Severity : enhancement
3415 Description: updated patchess for new RHEL4 kernel
3416 Details : Fixed ext3-unlink-race.patch per Kalpak's comment.
3418 Severity : enhancement
3420 Description: warnings with build patchless client with vanila 2.6.19 and up
3421 Details : change old ctl_table style and replace ctl_table/ctl_table_header
3422 with cfs_sysctl_table_t/cfs_sysctl_table_header_t
3424 Severity : enhancement
3426 Description: O_DIRECT bypasses client statistics.
3427 Details : When running with O_DIRECT I/O, neither the client rpc_stats nor
3428 read_ahead_stats were updated. Copied stats section from
3429 osc_send_oap_rpc() into async_internal().
3432 Frequency : only for Cray XT3
3434 Description: peer credits not enough on many OST per OSS systems.
3435 Details : Use new lnet way to add credits as we need those for pings and ASTs
3438 Frequency : only with liblustre
3440 Description: Liblustre is not releasing flock locks on file close.
3441 Details : Release flock locks on file close.
3444 Frequency : only for RHEL4
3446 Description: Update kernel patches for kernel-2.6.9-55.0.2.EL
3447 Details : Remove inode-nr_unused-2.6.9-rhel4.patch from 2.6-rhel4.series
3448 Update target file and kernel config.
3452 Description: lustre support for RHEL5
3453 Details : Add support for RHEL5.
3458 Description: ASSERTION(export != NULL) failed in target_handle_connect
3459 Details : Assetion hit is result of rare race between disconnect and connet
3460 to same nid. target_handle_connect found old connect cockie and
3461 tried to reconnect, but can't find export for this cockie.
3466 Description: umount blocks forever on error
3467 Details : In result of wrong using obd_no_recov and obd_force flags client
3468 can hand if cancel or some other requests is lost.
3471 Frequency : Only for SLES
3473 Description: sanity_quota fail test_1
3474 Details : There are multiple occurences of $TSTUSR in SLES's /etc/group
3475 file, which makes TSTID[2] inunique.
3477 Severity : enhancement
3479 Description: Kernel patches for SLES9 2.6.5-7.286 kernel
3480 Details : Update target/ChangeLog/which_patch .
3482 Severity : enhancement
3484 Description: jbd statistics
3485 Details : Port older jbd statistics patch for sles10
3490 Description: lvbo_init failed for resource with missing objects.
3491 Details : Fix returning error if we do stat for file with missing/corrupted
3492 objects and i_size set to all sum of size of all avaible objects.
3493 if we to truncate/write to missing object - it is recreated.
3498 Description: Oops in read and write path when failing to allocate lock.
3499 Details : Check if lock allocation failed and return error back.
3502 Frequency : When flocks are used.
3504 Description: assertion failure in ldlm_cli_enquque_fini for non NULL lock.
3505 Details : Flock locks might destroy just granted lock if it could be merged
3506 with another existing flock, this is done in completion handler,
3507 so teach ldlm_cli_enquque_fini that this is a valid case for
3513 Description: reply_lock_interpret crash due to race with it and lock cancel.
3514 Details : Do not replay locks that are being cancelled. Do not reference
3515 locks by their address during replay, just by their handle.
3518 Frequency : only with deactivated OSTs
3520 Description: lstripe command fails for valid OST index
3521 Details : The stripe offset is compared to 'lov->desc.ld_tgt_count'
3522 instead of lov->desc.ld_active_tgt_count.
3524 Severity : enhancement
3526 Description: Build failure against Centos5 (RHEL5)
3527 Details : Use getpagesize() instead of PAGE_SIZE.
3529 --------------------------------------------------------------------------------
3531 2007-07-30 Cluster File Systems, Inc. <info@clusterfs.com>
3533 * Support for kernels:
3534 2.6.5-7.283 (SLES 9),
3535 2.6.9-55.EL (RHEL 4),
3536 2.6.16.46-0.14 (SLES 10),
3537 2.6.18.8 vanilla (kernel.org)
3538 * Client support for unpatched kernels:
3539 (see http://wiki.lustre.org/index.php?title=Patchless_Client)
3540 2.6.16 - 2.6.22 vanilla (kernel.org)
3541 * Due to recently discovered recovery problems, we do not recommend
3542 using patchless RHEL 4 clients with this or any earlier release.
3543 * Recommended e2fsprogs version: 1.39.cfs8
3544 * Note that reiserfs quotas are disabled on SLES 10 in this kernel.
3545 * Starting with this release, the ldiskfs backing filesystem required
3546 by Lustre is now in its own package, lustre-ldiskfs. This package
3547 should be installed. It is versioned separately from Lustre and
3548 may be released separately in future.
3553 Description: server LBUG when shutting down
3554 Details : Block umount forever until the mount refcount is zero rather
3555 than giving up after an arbitrary timeout.
3557 Severity : enhancement
3559 Description: add optional extra BUILD_VERSION info
3560 Details : add a new environment variable (namely LUSTRE_VERS) which allows
3561 to override the lustre version.
3564 Frequency : 2.6.18 servers only
3566 Description: ll_kern_mount() doesn't release the module reference
3567 Details : The ldiskfs module reference count never drops down to 0
3568 because ll_kern_mount() doesn't release the module reference.
3573 Description: server LBUG when using old ost_num_threads parameter
3574 Details : Accept the old ost_num_threads parameter but warn that it
3575 is deprecated, and fix an off-by-one error that caused an LBUG.
3580 Description: Transient SCSI error results in persistent IO issue
3581 Details : iobuf->dr_error is not reinitialized to 0 between two
3585 Frequency : sometimes when underlying device returns I/O errors
3587 Description: OSTs not going read-only during write failures
3588 Details : OSTs are not remounted read-only when the journal commit threads
3589 get I/O errors because fsfilt_ext3 calls journal_start/stop()
3590 instead of the ext3 wrappers.
3593 Frequency : only with 10000 clients or more
3595 Description: poor connect scaling with increasing client count
3596 Details : Don't run filter_grant_sanity_check for more than 100 exports
3597 to improve scaling for large numbers of clients.
3600 Frequency : SLES10 only
3602 Description: sanity-quota.sh quotacheck failed: rc = -22
3603 Details : Quotas cannot be enabled on SLES10.
3606 Frequency : liblustre clients only
3608 Description: getdirentries does not give error when run on compute nodes
3609 Details : getdirentries does not fail when the size specified as an argument
3610 is too small to contain at least one entry
3612 Severity : enhancement
3614 Description: Add LNET router traceability for debug purposes
3615 Details : If a checksum failure occurs with a router as part of the
3616 IO path, the NID of the last router that forwarded the bulk data
3617 is printed so it can be identified.
3622 Description: OST "spontaneously" evicts client; client has imp_pingable == 0
3623 Details : Due to a race condition, liblustre clients were occasionally
3624 evicted incorrectly.
3626 Severity : enhancement
3628 Description: lfs setstripe use optional parameters instead of postional
3631 Severity : enhancement
3633 Description: Nanosecond timestamp support for ldiskfs
3634 Details : The on-disk ldiskfs filesystem has added support for nanosecond
3635 resolution timestamps. There is not yet support for this at
3636 the Lustre filesystem level.
3639 Frequency : during server recovery
3641 Description: MDS failing to send precreate requests due to OSCC_FLAG_RECOVERING
3642 Details : request with rq_no_resend flag not awake l_wait_event if they get a
3646 Frequency : nfs export on patchless client
3648 Description: connectathon hang when test nfs export over patchless client
3649 Details : Disconnected dentry cannot be found with lookup, so we do not need
3650 to unhash it or make it invalid
3653 Description: fix llapi_lov_get_uuids() to allow many OSTs to be returned
3654 Details: : Change llapi_lov_get_uuids() to read the UUIDs from /proc instead
3655 of using an ioctl. This allows lfsck for > 160 OSTs to succeed.
3660 Description: open req refcounting wrong on reconnect
3661 Details : If reconnect happened between getting open reply from server and
3662 call to mdc_set_replay_data in ll_file_open, we will schedule
3663 replay for unreferenced request that we are about to free.
3664 Subsequent close will crash in variety of ways.
3665 Check that request is still eligible for replay in
3666 mdc_set_replay_data().
3671 Description: disable writes to filesystem when reading health_check file
3672 Details : the default for reading the health_check proc file has changed
3673 to NOT do a journal transaction and write to disk, because this
3674 can cause reads of the /proc file to hang and block HA state
3675 checking on a healthy but otherwise heavily loaded system. It
3676 is possible to return to the previous behaviour during configure
3677 with --enable-health-write.
3679 Severity : enhancement
3681 Description: 64-bit inode version
3682 Details: : Add a on-disk 64-bit inode version for ext3 to track changes made
3683 to the inode. This will be required for version-based recovery.
3688 Description: MDS fails to start if a duplicate client export is detected
3689 Details : in some rare cases it was possible for a client to connect to
3690 an MDS multiple times. Upon recovery the MDS would detect this
3691 and fail during startup. Handle this more gracefully.
3693 Severity : enhancement
3695 Description: Add -o localflock option to simulate old noflock behaviour.
3696 Details : This will achieve local-only flock/fcntl locks coherentness.
3701 Description: log_commit_thread vs filter_destroy race leads to crash
3702 Details : Take import reference before releasing llog record semaphore
3707 Description: Wrong request locking in request set processing
3708 Details : ptlrpc_check_set wrongly uses req->rq_lock for proctect add to
3709 imp_delayed_list, in this place should be used imp_lock.
3712 Frequency : when reconnecting
3714 Description: Grant leak when OSC reconnect to OST
3715 Details : When osc reconnect ost, OST(filter) should check whether it
3716 should grant more space to client by comparing fed_grant and
3717 cl_avail_grant, and return the granted space to client instead
3718 of "new granted" space, because client will call osc_init_grant
3719 to update the client grant space info.
3722 Frequency : when client reconnects to OST
3724 Description: Grant leak when OSC does a resend and replays bulk write
3725 Details : When osc reconnect to OST, OST(filter) should clear grant info of
3726 bulk write request, because the grant info while be sync between
3727 OSC and OST when reconnect, and we should ignore the grant info
3728 these of resend/replay write req.
3733 Description: Grant space more than avaiable space sometimes.
3734 Details : When then OST is about to be full, if two bulk writing from
3735 different clients came to OST. Accord the avaliable space of the
3736 OST, the first req should be permitted, and the second one
3737 should be denied by ENOSPC. But if the seconde arrived before
3738 the first one is commited. The OST might wrongly permit second
3739 writing, which will cause grant space > avaiable space.
3742 Frequency : when client is evicted
3744 Description: Grant might be wrongly erased when osc is evicted by OST
3745 Details : when the import is evicted by server, it will fork another
3746 thread ptlrpc_invalidate_import_thread to invalidate the
3747 import, where the grant will be set to 0. While the original
3748 thread will update the grant it got when connecting. So if
3749 the former happened latter, the grant will be wrongly errased
3750 because of this race.
3755 Description: Checking Stale with correct fid
3756 Details : ll_revalidate_it should uses de_inode instead of op_data.fid2
3757 to check whether it is stale, because sometimes, we want the
3758 enqueue happened anyway, and op_data.fid2 will not be initialized.
3760 Severity : enhancement
3762 Description: update patchless client
3763 Details : Add support for patchless client with 2.6.20, 2.6.21 and RHEL 5
3766 Frequency : only with 2.4 kernel
3768 Description: random memory corruption
3769 Details : size of struct ll_inode_info is to big for union inode.u and this
3770 can be cause of random memory corruption.
3775 Description: Memory leak in recovery
3776 Details : Lov_mds_md was not free in an error handler in mds_create_object.
3777 It should also check obd_fail before fsfilt_start, otherwise if
3778 fsfilt_start return -EROFS,(failover mds during mds recovery).
3779 then the req will return with repmsg->transno = 0 and rc = EROFS.
3780 and we met hit the assert LASSERT(req->rq_reqmsg->transno ==
3781 req->rq_repmsg->transno) in ptlrpc_replay_interpret. Fcc should
3782 be freed no matter whether fsfilt_commit success or not.
3785 Frequency : only with large numbers of cores on a single node
3787 Description: Prevents from taking the superblock lock in llap_from_page for
3789 Details : using LL_ORIGIN_REMOVEPAGE origin flag instead of LL_ORIGIN_UNKNOW
3790 for llap_from_page call in ll_removepage() prevents from taking
3791 the superblock lock for a soon died page.
3796 Description: Not check open intent error before release open handle
3797 Details : in some rare cases, the open intent error is not checked before
3798 release open handle, which may cause
3799 ASSERTION(open_req->rq_transno != 0), because it tries to release
3800 the failed open handle.
3805 Description: Set cat log bitmap only after create log success.
3806 Details : in some rare cases, the cat log bitmap is set too early. and it
3807 should be set only after create log success.
3811 Description: Accessing a block bevice can re-enable I/O when Lustre is
3812 tearing down a device.
3813 Details : dev_clear_rdonly(bdev) must be called in kill_bdev() instead of
3818 Description: service threads may hog cpus when there are a lot of requests
3819 Details : Insert cond_resched to give other threads a chance to use some CPU
3824 Description: the cat log was not initialized in recovery
3825 Details : When mds(mgs) do recovery, the tgt_count might be zero, so the
3826 unlink log on mds will not be initialized until mds post
3827 recovery. And also in mds post recovery, the unlink log will
3828 initialization will be done asynchronausly, so there will be race
3829 between add unlink log and unlink log initialization.
3833 Description: brw_stats were being printed incorrectly
3834 Details : brw_stats were being printed as log2 but all of them were not
3835 recorded as log2. Also remove some code duplication arising from
3836 filter_tally_{read,write}.
3840 Frequency : rare, only in recovery.
3841 Description: ASSERTION(req->rq_type != LI_POISON) failed
3842 Details : imp_lock should be held while iterating over imp_sending_list for
3843 prevent destroy request after get timeout in ptlrpc_queue_wait.
3847 Description: replay-single.sh test 52 fails
3848 Details : A lock's skiplist need to be cleanup when it being unlinked
3849 from its resource list.
3854 Description: Short directio read returns full requested size rather than
3856 Details : Direct I/O operations should return actual amount of bytes
3857 transferred rather than requested size.
3859 Severity : enhancement
3861 Description: metadata RPC reduction (e.g. for rm performance)
3862 Details : decrease the amount of synchronous RPC between clients and servers
3863 by canceling conflicing lock before the operation on the client
3864 and packing thier handles into the main operation RPC to server.
3866 Severity : enhancement
3868 Description: add #ifdef HAVE_KERNEL_CONFIG_H
3869 Details : kernels from 2.6.19 not need include linux/config.h, but add
3870 include linux/autoconf.h in commpiler command line.
3872 Severity : enhancement
3874 Description: patchless client support for 2.6.22 kernel
3875 Details : 2.6.22 has only one visble change, SLAB_CTOR_* constants is
3876 removed. In this case we need drop using os depended interface to
3877 kmem_cache and use cfs_mem_cache API.
3882 Description: fix mal-formatted messages
3883 Details : fix some mal-formatted DEBUG_REQ and LCONSOLE_ERROR_MSG messages
3887 Frequency : always in liblustre
3888 Description: wrong IS_ERR implementation in liblustre.h
3889 Details : fix IS_ERR implementation in liblustre.h for right detect errors.
3894 Description: Correct condition for output debug message.
3895 Details : inode i_nlink equal zero is not enough for output message about
3896 disk corruption, i_ctime and i_mode should be also checked.
3900 Frequency : always in patchless client
3901 Description: add configure check for truncate_complete_page
3902 Details : improve checks for exported symbols. This allow run check without
3903 sources, but with Module.symvers shipped with kernel distribution.
3904 add check for truncate_complete_page used by patchless client.
3908 Description: sanity.sh test_77h fails with "test_77h file compare failed"
3909 Details : test_77h uses a file which was messed by other test case.
3913 Description: Not Check whether lov_tgts is NULL in some lov functions
3914 Details : Checking whether lov_tgts is NULL in some functions.
3918 Description: replace obdo_alloc() with OBDO_ALLOC macro
3919 Details : nothing special is done in obdo_alloc() function, and for
3920 debugging purpose, it needs to be replaced with macros.
3924 Description: bad return value and errno from fcntl call
3925 Details : In liblustre API, errno should be a negative value if error
3930 Description: ptlrpc_check_set() LBUG
3931 Details : In case of positive reply from server and failed client bulk
3932 callback after bulk transfer shouldn't LBUG, but process this
3933 request as erroneous.
3935 Severity : enhancement
3937 Description: VFS operations stats tool.
3938 Details : Tool which collects stats by tracking value written in pid,
3939 ppid, gid and uses llstat to generate output to plot graph using
3941 Updated lustre/utils/Makefile.am
3942 Added lustre/utils/ltrack_stats.c
3944 Severity : enhancement
3946 Description: 2.6.18 server support (lustre 1.6.1)
3947 Details : Support for 2.6.18 kernels on the server side.
3952 Description: ASSERTION(imp->imp_conn_current) failed
3953 Details : an assertion failure is hit if a client node boots and attempts to
3954 mount a lustre filesystem faster than RECONNECT_INTERVAL seconds.
3957 Frequency : only for i686
3959 Description: 1.4.11 RC1 build fails for RHEL 4, i686
3960 Details : Fixed config variable for build.
3965 Description: Updated patchess for new RHEL4 kernel
3966 Details : Updated patch inode-nr_unused-2.6.9-rhel4.patch
3967 Updated patch jbd-stats-2.6.9.patch
3968 Updated patch qsnet-rhel4-2.6.patch
3969 Updated patch quota-deadlock-on-pagelock-core.patch
3970 Updated patch vfs_intent-2.6-rhel4.patch
3971 Updated patch vfs_races-2.6-rhel4.patch
3972 Updated series file 2.6-rhel4-titech.series
3973 Updated series file 2.6-rhel4.series
3974 Updated kernel config files
3979 Description: lquota slave complains LBUG when reconnecting with mds
3981 Details : quota slave depends qctxt->lqc_import to send its quota request.
3982 This pointer will be invalid if mds did failover or broke its
3983 connect to osts, which leads to LBUG.
3986 Frequency : when qunit size is too small(less than 20M)
3988 Description: write is stopped by improper -EDQUOT
3989 Details : If the master is busy and qunit size is small enough(let's say 1M),
3990 the slave can not get quota from master on time, which will lead
3991 slave to trigger a -EQUOTA to client.
3996 Description: Deadlock during metadata tests
3997 Details : in prune_dir_dentries(), shrink_dcache_parent() should not be
3998 called with the per-dentry lock held.
4001 Frequency : SLES9 only
4003 Description: Lustre patched kernel for SLES9 SP3 has NR_CPUS set to 8
4004 Details : set CONFIG_NR_CPUS to 128 instead of 8.
4006 Severity : enhancement
4008 Description: remove fs_prep_san_write operation and related patches
4009 Details : remove the ext3-san-jdike patches which are no longer useful.
4014 Description: LDISKFS-fs error (device sdc): ldiskfs_free_blocks
4015 Details : a disk corruption can cause the mballoc code to assert on a
4016 double free or other extent corruptions. Handle these with
4017 ext3_error() instead of assertions.
4020 Frequency : only with mballoc3 code and deep extent trees
4022 Description: ldiskfs_ext_search_right: bad header in inode: unexpected eh_depth
4023 Details : a wrong check of extent headers in ldiskfs_ext_search_right()
4024 can cause the filesystem to be remounted read-only.
4029 Description: lfsck built against 1.4.x cannot run against 1.6.0 lustre
4030 Details : the definition for OBD_IOC_GETNAME changed in 1.6.0. One of the
4031 few external users of this ioctl number is lfsck's call to
4032 llapi_lov_get_uuids() and this caused lfsck to fail at startup.
4033 Add the old ioctl number to the handler so both old and new
4038 Description: parallel lock callbacks
4039 Details : Instead of sending blocking and completion callbacks as separated
4040 requests, adding them to a set and sending in parallel.
4044 Description: Disable most debugging by default
4045 Details : To improve performance, disable most logging (for debug purposes)
4046 by default. VFSTRACE, RPCTRACE, and DLMTRACE are now off by
4047 default, and HA includes fewer messages.
4051 Frequency : only run on patchless client.
4052 Description: use do_facet on sanity.sh for test handling recoverables errors
4053 Details : use do_facet instead of direct use sysctl for set fail_loc on OST
4057 Description: hash tables for lists of nids, connections and uuids
4058 Details : Hash tables noticeably help when a lot of clients connect to a
4059 server, to faster identify duplicate connections or reconnects,
4060 also to faster find export to evict in manual eviction case.
4064 Description: Sometimes, when the server evict a client, and the client will
4065 not be evicted as soon as possible.
4066 Details : In enqueue req, the error was returned by intent, instead of
4067 rq_status which make ptlrpc layer not detect this error, and
4068 does not evict the client. So enqueue error should be returned
4072 Frequency : only at startup
4074 Description: Delay client connections to MDT until fist MDT->OST connect
4075 Details : If a client tried to create a new file before the MDT had
4076 connected to any OSTs, the create would return EIO. Now
4077 the client will simply block until the MDT connects to the
4078 first OST and the create can succeed.
4081 Frequency : only for SLES9
4083 Description: Routinely utilize latest Quadrics drivers in CFS releases
4084 Details : Update patch qsnet-suse-2.6.patch.
4087 Frequency : only for sles10
4089 Description: Update patches for SLES 10 SP1 kernel.
4090 Details : Update the patch vfs_intent-2.6-sles10.patch.
4095 Description: Routinely utilize latest Quadrics drivers in CFS releases
4096 Details : Update patch qsnet-rhel4-2.6.patch.
4099 Frequency : at statup only
4101 Description: mds_lov_synchronize race leads to various problems
4102 Details : simultaneous MDT->OST connections at startup can cause the
4103 sync to abort, leaving the OSC in a bad state.
4108 Description: Using wrong pointer in osc_brw_prep_request
4109 Details : Access to array[-1] can produce panic if kernel compiled with
4110 CONFIG_PAGE_ALLOC enabled
4112 Severity : enhancement
4114 Description: Async OSC create to avoid the blocking unnecessarily.
4115 Details : If a OST has no remain object, system will block on the creating
4116 when need to create a new object on this OST. Now, ways use
4117 pre-created objects when available, instead of blocking on an
4118 empty osc while others are not empty. If we must block, we block
4119 for the shortest possible period of time.
4123 Frequency : only in recovery
4124 Description: Mark OST as early accessible if his start SYNC.
4125 Details : osc_precreate return flag early accessible if oscc marked as
4126 OSCC_FLAG_SYNC_IN_PROGRESS.
4131 Description: Sometimes precreate code can triger create object on wrong ost
4132 Details : Wrong protected or not not restored variables aftre precreate loop
4133 can produce creation object on wrong ost.
4136 Frequency : oss recovery
4138 Description: llog_commit_thread cleanup should sync with llog_commit_thread
4140 Details : llog_commit_thread_count should be synced between llog_commit
4141 start and cleanup, so new llog_commit thread should not be started
4142 when llog_commit threads being stopped to avoid accessing some
4145 Severity : enhancement
4147 Description: Add printing inode info into message about error in writepage.
4149 --------------------------------------------------------------------------------
4151 2007-05-03 Cluster File Systems, Inc. <info@clusterfs.com>
4156 Frequency : on some architectures
4158 Description: 1.6 client sometimes fails to mount from a 1.4 MDT
4159 Details : Uninitialized flags sometimes cause configuration commands to
4163 Frequency : patchless clients only
4165 Description: missing __iget() symbol export
4166 Details : The __iget() symbol export is missing. To avoid the need for
4167 this on patchless clients the deathrow inode reaper is turned
4168 off, and we depend on the VM to clean up old inodes. This
4169 dependency was during via the fix for bug 12181.
4174 Description: sanity.sh fail: test_52b
4175 Details : The ll_inode_to_ext_flags() has a glitch which makes MDS return
4176 incorrect inode's flags to client.
4178 --------------------------------------------------------------------------------
4180 2007-04-19 Cluster File Systems, Inc. <info@clusterfs.com>
4182 * CONFIGURATION CHANGE. This version of Lustre WILL NOT
4183 INTEROPERATE with older versions automatically. In many cases a
4184 special upgrade step is needed. Please read the
4185 user documentation before upgrading any part of a 1.4.x system.
4186 * WARNING: Lustre configuration and startup changes are required with
4187 this release. See https://mail.clusterfs.com/wikis/lustre/MountConf
4189 * Support for kernels:
4190 2.4.21-47.0.1.EL (RHEL 3),
4191 2.6.5-7.283 (SLES 9),
4192 2.6.9-42.0.10.EL (RHEL 4),
4193 2.6.12.6 vanilla (kernel.org),
4194 2.6.16.27-0.9 (SLES10)
4195 * Client support for unpatched kernels:
4196 (see https://mail.clusterfs.com/wikis/lustre/PatchlessClient)
4197 2.6.16 - 2.6.19 vanilla (kernel.org),
4198 2.6.9-42.0.8EL (RHEL 4)
4199 * Recommended e2fsprogs version: 1.39.cfs6
4200 * Note that reiserfs quotas are disabled on SLES 10 in this kernel
4203 Severity : enhancement
4205 Description: Async OSC create to avoid the blocking unnecessarily.
4206 Details : If a OST has no remain object, system will block on the creating
4207 when need to create a new object on this OST. Now, ways use
4208 pre-created objects when available, instead of blocking on an
4209 empty osc while others are not empty. If we must block, we block
4210 for the shortest possible period of time.
4212 Severity : enhancement
4214 Description: MountConf
4215 Details : Lustre configuration is now managed via mkfs and mount
4216 commands instead of lmc and lconf. New obd types (MGS, MGC)
4217 are added for dynamic configuration management. See
4218 https://mail.clusterfs.com/wikis/lustre/MountConf for
4221 Severity : enhancement
4223 Description: dynamic OST addition
4224 Details : OSTs can now be added to a live filesystem
4226 Severity : enhancement
4228 Description: startup order invariance
4229 Details : MDTs and OSTs can be started in any order. Clients only
4230 require the MDT to complete startup.
4232 Severity : enhancement
4234 Description: parallel, asynchronous orphan cleanup
4235 Details : orphan cleanup is now performed in separate threads for each
4236 OST, allowing parallel non-blocking operation.
4238 Severity : enhancement
4240 Description: optimized stripe assignment
4241 Details : stripe assignments are now made based on ost space available,
4242 ost previous usage, and OSS previous usage, in order to try
4243 to optimize storage space and networking resources.
4245 Severity : enhancement
4247 Description: Permanently set tunables
4248 Details : All writable /proc/fs/lustre tunables can now be permanently
4249 set on a per-server basis, at mkfs time or on a live
4252 Severity : enhancement
4254 Description: Lustre message v2
4255 Details : Add lustre message format v2.
4257 Severity : enhancement
4259 Description: client OST exclusion list
4260 Details : Clients can be started with a list of OSTs that should be
4261 declared "inactive" for known non-responsive OSTs.
4265 Description: ENOENT returned for valid filehandle during dbench.
4266 Details : Check if a directory has children when invalidating dentries
4267 associated with an inode during lock cancellation. This fixes
4268 an incorrect ENOENT sometimes seen for valid filehandles during
4269 testing with dbench.
4272 Frequency : SFS test only (otherwise harmless)
4274 Description: SPEC SFS validation failure on NFS v2 over lustre.
4275 Details : Changes the blocksize for regular files to be 2x RPC size,
4276 and not depend on stripe size.
4278 Severity : enhancement
4280 Description: fine-grained SMP locking inside DLM
4281 Details : Improve DLM performance on SMP systems by removing the single
4282 per-namespace lock and replace it with per-resource locks.
4284 Severity : enhancement
4286 Description: don't hold multiple extent locks at one time
4287 Details : To avoid client eviction during large writes, locks are not
4288 held on multiple stripes at one time or for very large writes.
4289 Otherwise, clients can block waiting for a lock on a failed OST
4290 while holding locks on other OSTs and be evicted.
4292 Severity : enhancement
4294 Description: Multiple MD RPCs in flight.
4295 Details : Further unserialise some read-only MDT RPCs - learn about intents.
4296 To avoid overly-overloading MDT, introduce a limit on number of
4297 MDT RPCs in flight for a single client and add /proc controls
4298 to adjust this limit.
4300 Severity : enhancement
4302 Description: client read/write statistics
4303 Details : Add client read/write call usage stats for performance
4304 analysis of user processes.
4305 /proc/fs/lustre/llite/*/offset_stats shows non-sequential
4306 file access. extents_stats shows chunk size distribution.
4307 extents_stats_per_process show chunk size distribution per
4310 Severity : enhancement
4312 Description: per-client statistics on server
4313 Details : Add ldlm and operations statistics for each client in
4314 /proc/fs/lustre/mds|obdfilter/*/exports/
4316 Severity : enhancement
4318 Description: improved MDT statistics
4319 Details : Add detailed MDT operations statistics in
4320 /proc/fs/lustre/mds/*/stats
4322 Severity : enhancement
4324 Description: VFS operations stats
4325 Details : Add client VFS call stats, trackable by pid, ppid, or gid
4326 /proc/fs/lustre/llite/*/stats_track_[pid|ppid|gid]
4331 Description: Fix client-side osc byte counters
4332 Details : The osc read/write byte counters in
4333 /proc/fs/lustre/osc/*/stats are now working
4336 Frequency : always as root on SLES
4338 Description: Failure of copying files with lustre special EAs.
4339 Details : Client side always return success for setxattr call for lustre
4340 special xattr (currently only "trusted.lov").
4345 Description: Refcount LNET uuids
4346 Details : The global LNET uuid list grew linearly with every startup;
4347 refcount repeated list entries instead of always adding to
4350 Severity : enhancement
4352 Description: Dynamic service threads
4353 Details : Within a small range, start extra service threads
4354 automatically when the request queue builds up.
4357 Frequency : mixed-endian client/server environments
4359 Description: mixed-endian crashes
4360 Details : The new msg_v2 system had some failures in mixed-endian
4363 Severity : enhancement
4365 Description: Easy OST removal
4366 Details : OSTs can be permanently deactivated with e.g. 'lctl
4367 conf_param lustre-OST0001.osc.active=0'
4369 Severity : enhancement
4371 Description: MGS proc entries
4372 Details : Added basic proc entries for the MGS showing what filesystems
4375 Severity : enhancement
4377 Description: provide MGS failover
4378 Details : Added config lock reacquisition after MGS server failover.
4380 Severity : enhancement
4382 Description: add Linux 2.4 support
4383 Details : Added support for RHEL 2.4.21 kernel for 1.6 servers and clients
4387 Description: a large application tries to do I/O to the same resource and dies
4388 in the middle of it.
4389 Details : Check the req->rq_arrival time after the call to
4390 ost_brw_lock_get(), but before we do anything about
4391 processing it & sending the BULK transfer request. This
4392 should help move old stale pending locks off the queue as
4393 quickly as obd_timeout.
4396 Frequency : when an incorrect nid is specified during startup
4398 Description: ptlrpc connect to non-existant node causes kernel crash
4399 Details : LNET can't be re-entered from an event callback, which
4400 happened when we expire a message after the export has been
4401 cleaned up. Instead, hand the zombie cleanup off to another
4404 Severity : enhancement
4406 Description: plain/inodebits lock performance improvement
4407 Details : Grouping plain/inodebits in granted list by their request modes
4408 and bits policy, thus improving the performance of search through
4412 Frequency : only if OST filesystem is corrupted
4414 Description: client incorrectly hits assertion in ptlrpc_replay_req()
4415 Details : for a short time RPCs with bulk IO are in the replay list,
4416 but replay of bulk IOs is unimplemented. If the OST filesystem
4417 is corrupted due to disk cache incoherency and then replay is
4418 started it is possible to trip an assertion. Avoid putting
4419 committed RPCs into the replay list at all to avoid this issue.
4422 Frequency : liblustre (e.g. catamount) on a large cluster with >= 8 OSTs/OSS
4424 Description: System hang on startup
4425 Details : This bug allowed the liblustre (e.g. catamount) client to
4426 return to the app before handling all startup RPCs. This
4427 could leave the node unresponsive to lustre network traffic
4428 and manifested as a server ptllnd timeout.
4430 Severity : enhancement
4432 Description: Add "/proc/sys/lustre/debug_peer_on_timeout"
4433 Details : liblustre envirable: LIBLUSTRE_DEBUG_PEER_ON_TIMEOUT
4434 boolean to control whether to print peer debug info when a
4435 client's RPC times out.
4438 Frequency : only for kernels with patches from Lustre below 1.4.3
4440 Description: Remove old rdonly API
4441 Details : Remove old rdonly API which unused from at least lustre 1.4.3
4444 Frequency : only for devices with external journals
4446 Description: Set external device read-only also
4447 Details : During a commanded failover stop, we set the disk device
4448 read-only while the server shuts down. We now also set any
4449 external journal device read-only at the same time.
4452 Frequency : when upgrading from 1.4 while trying to change parameters
4454 Description: The wrong (new) MDC name was used when setting parameters for
4455 upgraded MDT's. Also allows changing of OSC (and MDC)
4456 parameters if --writeconf is specified at tunefs upgrade time.
4459 Frequency : when setting specific ost indicies
4461 Description: QOS code breaks on skipped indicies
4462 Details : Add checks for missing OST indicies in the QOS code, so OSTs
4463 created with --index need not be sequential.
4465 Severity : enhancement
4467 Description: Add uninit_groups feature to ldiskfs2 to speed up e2fsck
4468 Details : The uninit_groups feature works in conjunction with the kernel
4469 filesystem code (ldiskfs2 only) and e2fsprogs-1.39-cfs6 to speed
4470 up the pass1 processing of e2fsck. This is a read-only feature
4471 in ldiskfs2 only, so older kernels and current ldiskfs cannot
4472 mount filesystems that have had this feature enabled.
4474 Severity : enhancement
4476 Description: Improve multi-block allocation algorithm to avoid fragmentation
4477 Details : The mballoc3 code (ldiskfs2 only) adds new mechanisms to improve
4478 allocation locality and avoid filesystem fragmentation.
4480 ------------------------------------------------------------------------------
4482 2007-04-01 Cluster File Systems, Inc. <info@clusterfs.com>
4484 * Support for kernels:
4485 2.4.21-47.0.1.EL (RHEL 3)
4486 2.6.5-7.283 (SLES 9)
4487 2.6.9-42.0.10.EL (RHEL 4)
4488 2.6.12.6 vanilla (kernel.org)
4489 2.6.16.27-0.9 (SLES 10)
4490 * Recommended e2fsprogs version: 1.39.cfs5
4492 * Note that reiserfs quotas are disabled on SLES 10 in this kernel
4496 Frequency : occasional, depends on client load and configuration
4497 Bugzilla : 12181, 12203
4498 Description: data loss for recently-modified files
4500 Details : In some cases it is possible that recently written or created
4501 files may not be written to disk in a timely manner (this should
4502 normally be within 30s unless client IO load is very high).
4503 The problem appears as zero-length files or files that are a
4504 multiple of 1MB in size after a client crash or client eviction
4505 that are missing data at the end of the file.
4507 This problem is more likely to be hit on clients where files are
4508 repeatedly created and unlinked in the same directory, clients
4509 have a large amount of RAM, have many CPUs, the filesystem has
4510 many OSTs, the clients are rebooted frequently, and/or the files
4511 are not accessed by other nodes after being written.
4513 The presence of the problem can be detected by looking at
4514 /proc/sys/fs/inode-state. If the first number (nr_inodes) is
4515 smaller than the second (nr_unused) then dirty files will not
4516 be flushed automatically to disk. "sync; sleep 10" should be
4517 run several times on the node before unmounting it to update
4518 Lustre (this is also safe to run on nodes without this problem).
4520 There is also a related kernel bug in the RHEL4 4 2.6.9 kernel
4521 that can cause this same problem, so customers using that kernel
4522 also need to update the kernel in addition to Lustre. In order
4523 to properly fix this bug, the RHEL3 2.4.21 kernel is also updated.
4525 It is normal that files written just before a client crash (less
4526 than 30s) may not yet have been flushed to disk, even for local
4530 Frequency : frequent on thin XT3 nodes
4532 Description: UUID collision on thin XT3 Linux nodes
4533 Details : UUIDs on Compute Node Linux XT3 nodes were not generated
4534 randomly, since we relied on an insufficiently-seeded PRNG.
4539 Description: OSS hangs after "All ost request buffers busy"
4540 Details : A deadlock between quota and journal operations caused OSS
4541 hangs after printing "All ost request buffers busy."
4544 Frequency : always on liblustre builds
4546 Description: Cleanup compiler warnings on liblustre
4549 Frequency : always on liblustre builds on XT3
4551 Description: LC_CONFIG_CDEBUG don't run while build liblustre on XT3.
4555 Description: Addition of EXT3_FEATURE_RO_COMPAT_DIR_NLINKS flag for
4556 > 32000 subdirectories
4557 Details : Add EXT3_FEATURE_RO_COMPAT_DIR_NLINK flag to
4558 EXT3_FEATURE_RO_COMPAT_SUPP. This flag will be set whenever
4559 subdirectory count crosses 32000. This will aid e2fsck to
4560 correctly handle more than 32000 subdirectories.
4563 Frequency : liblustre (e.g. catamount) on a large cluster with >= 8 OSTs/OSS
4565 Description: System hang on startup
4566 Details : This bug allowed the liblustre (e.g. catamount) client to
4567 return to the app before handling all startup RPCs. This
4568 could leave the node unresponsive to lustre network traffic
4569 and manifested as a server ptllnd timeout.
4571 Severity : enhancement
4573 Description: Add "/proc/sys/lustre/debug_peer_on_timeout"
4574 (liblustre envirable: LIBLUSTRE_DEBUG_PEER_ON_TIMEOUT)
4575 boolean to control whether to print peer debug info when a
4576 client's RPC times out.
4581 Description: make O_SYNC working on 2.6 kernels
4582 Details : 2.6 kernels use different method for mark pages for write,
4583 so need add a code to lustre for O_SYNC work.
4588 Description: Failure to close file and release space on NFS
4589 Details : Put inode details into lock acquired in ll_intent_file_open.
4590 Use mdc_intent_lock in ll_intent_open to properly
4591 detect all kind of errors unhandled by mdc_enqueue.
4596 Description: proc file read during shutdown sometimes raced obd removal,
4598 Details : Add lock to prevent obd access after proc file removal.
4601 Frequency : Only for files larger than 4GB on 32-bit clients.
4603 Description: improperly doing page alignment of locks
4604 Details : Modify lustre core code to use CFS_PAGE_* defines instead of
4605 PAGE_*. Make CFS_PAGE_MASK a 64-bit mask.
4610 Description: RPCs being resent when they shouldn't be
4611 Details : Some RPCs that should not be resent are being resent. This
4612 can cause inconsistencies in the RPC state machine. Do not
4613 resend such requests.
4616 Frequency : rare, only with NFS export
4618 Description: Crash on NFS re-export node
4619 Details : under very unusual load conditions an assertion is hit in
4620 ll_intent_file_open()
4623 Frequency : only if OST filesystem is corrupted
4625 Description: client incorrectly hits assertion in ptlrpc_replay_req()
4626 Details : for a short time RPCs with bulk IO are in the replay list,
4627 but replay of bulk IOs is unimplemented. If the OST filesystem
4628 is corrupted due to disk cache incoherency and then replay is
4629 started it is possible to trip an assertion. Avoid putting
4630 committed RPCs into the replay list at all to avoid this issue.
4635 Description: large O_DIRECT requests fail under memory pressure/fragmentation
4636 Details : Large single O_DIRECT read and write calls can fail to allocate
4637 a sufficiently large buffer to process the request. In case of
4638 allocation failure the allocation is retried with a smaller
4639 buffer and broken into smaller requests.
4641 Severity : enhancement
4643 Description: Add -o localflock option to simulate old noflock behaviour.
4644 Details : This will achieve local-only flock/fcntl locks coherentness.
4649 Description: versioning check is incomplete
4650 Details : Checking the version difference of client vs. server, report
4651 error if the gap is too big.
4656 Description: add support PG_writeback bit
4657 Details : add support for PG_writeback bit for Lustre, for more carefull
4658 work with page cache in 2.6 kernel. This also fix some deadlocks
4659 and remove hack for work O_SYNC with 2.6 kernel.
4661 Severity : enhancement
4663 Description: Add uninit_groups feature to ldiskfs2 to speed up e2fsck
4664 Details : The uninit_groups feature works in conjunction with the kernel
4665 filesystem code (ldiskfs2 only) and e2fsprogs-1.39-cfs6 to speed
4666 up the pass1 processing of e2fsck. This is a read-only feature
4667 in ldiskfs2 only, so older kernels and current ldiskfs cannot
4668 mount filesystems that have had this feature enabled.
4670 Severity : enhancement
4672 Description: Improve multi-block allocation algorithm to avoid fragmentation
4673 Details : The mballoc3 code (ldiskfs2 only) adds new mechanisms to improve
4674 allocation locality and avoid filesystem fragmentation.
4676 ------------------------------------------------------------------------------
4678 2007-02-09 Cluster File Systems, Inc. <info@clusterfs.com>
4680 * Support for kernels:
4681 2.6.9-42.0.3.EL (RHEL 4)
4682 2.6.5-7.276 (SLES 9)
4683 2.4.21-47.0.1.EL (RHEL 3)
4684 2.6.12.6 vanilla (kernel.org)
4685 2.6.16.21-0.8 (SLES10)
4686 * Recommended e2fsprogs version: 1.39.cfs2-0
4688 * The backwards-compatible /proc/sys/portals symlink has been removed
4689 in this release. Before upgrading, please ensure that you change
4690 any configuration scripts or /etc/sysctl.conf files that access
4691 /proc/sys/portals/* or sysctl portals.* to use the corresponding
4692 entry in /proc/sys/lnet or sysctl lnet.*. This change can be made
4693 in advance of the upgrade on any system running Lustre 1.4.6 or
4694 newer, since /proc/sys/lnet was added in that version.
4695 * Note that reiserfs quotas are disabled on SLES 10 in this kernel
4699 Frequency : only when quota is used
4701 Description: avoid scanning export list for quota master
4702 Details : Change the algorithms to avoid scanning export list in order
4703 to improve the efficiency.
4706 Frequency : MDS failover only, very rarely
4708 Description: "went back in time" messages on mds failover
4709 Details : The greatest transno may be lost when the current operation
4710 finishes with an error (transno==0) and the client's last_rcvd
4711 record is over-written. Save the greatest transno in the
4712 mds_last_transno for this case.
4715 Frequency : always for specific kernels and striping counts
4717 Description: client may get "Matching packet too big" without ACL support
4718 Details : Clients compiled without CONFIG_FS_POSIX_ACL get an error message
4719 when trying to access files in certain configurations. The
4720 clients should in fact be denied when mounting because they do
4721 not understand ACLs.
4724 Frequency : Cray XT3 with more than 4000 clients and multiple jobs
4726 Description: many clients connecting with IO in progress causes connect timeouts
4727 Details : Avoid synchronous journal commits to avoid delays caused by many
4728 clients connecting/disconnecting when bulk IO is in progress.
4729 Queue liblustre connect requests on OST_REQUEST_PORTAL instead of
4730 OST_IO_PORTAL to avoid delays behind potentially many pending
4734 Frequency : occasionally with multiple writers to a single file
4736 Description: shared writes to file may result in wrong size reported by stat()
4737 Details : Allow growing of kms when extent lock is cancelled
4740 Frequency : always with random mmap IO to multi-striped file
4742 Description: mmap write might be lost if we are writing to a 'hole' in stripe
4743 Details : Only if the hole is at the end of OST object so that kms is too
4744 small. Fix is to increase kms accordingly in ll_nopage.
4747 Frequency : rare, only if OST filesystem is inconsistent with MDS filesystem
4749 Description: writes to a missing object would leak memory on the OST
4750 Details : If there is an inconsistency between the MDS and OST filesystems,
4751 such that the MDS references an object that doesn't exist, writes
4752 to that object will leak memory due to incorrect cleanup in the
4753 error handling path, eventually running out of memory on the OST.
4758 Description: Creating too long symlink causes lustre errors
4759 Details : Check symlink and name lengths before sending requests to MDS.
4762 Frequency : only if flock is enabled (not on by default)
4764 Description: posix locks not released on fd closure on 2.6.9+
4765 Details : We failed to add posix locks to list of inode locks on 2.6.9+
4766 kernels, this caused such locks not to be released on fd close and
4767 then assertions on fs unmount about still used locks.
4770 Frequency : MDS failover only, very rarely
4772 Description: clients may get ASSERTION(granted_lock != NULL)
4773 Details : When request was taking a long time, and a client was resending
4774 a getattr by name lock request. The were multiple lock requests
4775 with the same client lock handle and
4776 mds_getattr_name->fixup_handle_for_resent_request found one of the
4777 lock handles but later failed with ASSERTION(granted_lock != NULL).
4782 Description: handle->h_buffer_credits > 0, assertion failure
4783 Details : h_buffer_credits is zero after truncate, causing assertion
4784 failure. This patch extends the transaction or creates a new
4788 Frequency : NFS re-export or patchless client
4789 Bugzilla : 11179, 10796
4790 Description: Crash on NFS re-export node (__d_move)
4791 Details : We do not want to hash the dentry if we don't have a lock.
4792 But if this dentry is later used in d_move, we'd hit uninitialised
4793 list head d_hash, so we just do this to init d_hash field but
4794 leave dentry unhashed.
4797 Frequency : NFS re-export or patchless client
4799 Description: NFS exports has problem with symbolic link
4800 Details : lustre client didn't properly install dentry when re-exported
4801 to NFS or running patchless client.
4804 Frequency : NFS re-export or patchless client
4806 Description: Various nfs/patchless fixes.
4807 Details : fixes reuse disconected alias for lookup process - this fixes
4808 warning "find_exported_dentry: npd != pd",
4809 fix permission error with open files at nfs.
4810 fix apply umask when do revalidate.
4813 Frequency : occasional
4815 Description: Crash on NFS re-export node
4816 Details : calling clear_page() on the wrong pointer triggered oops in
4817 generic_mapping_read().
4820 Frequency : rarely, using O_DIRECT IO
4822 Description: unaligned directio crashes client with LASSERT
4823 Details : check for unaligned buffers before trying any requests.
4826 Frequency : rarely, using CFS RAID5 patches in non-standard kernel series
4828 Description: stale data returned from RAID cache
4829 Details : If only a small amount of IO is done to the RAID device before
4830 reading it again it is possible to get stale data from the RAID
4831 cache instead of reading it from disk.
4834 Frequency : always for sles10 kernel
4836 Description: sles10 support
4837 Details : ll_follow_link: compile fixes and using of nd_set_link
4838 under newer kernels.
4841 Frequency : depends on arch, kernel and compiler version, always on sles10
4844 Description: recursive or deep enough symlinks cause stack overflow
4845 Details : getting rid of large stack-allocated variable in
4849 Frequency : depends on hardware
4851 Description: lustre write performance loss in the SLES10 kernel
4852 Details : the performance loss is caused by using of write barriers in the
4853 ext3 code. The SLES10 kernel turns barrier support on by
4854 default. The fix is to undo that change for ldiskfs.
4856 ------------------------------------------------------------------------------
4858 2006-12-09 Cluster File Systems, Inc. <info@clusterfs.com>
4860 * Support for kernels:
4861 2.6.9-42.0.3EL (RHEL 4)
4862 2.6.5-7.276 (SLES 9)
4863 2.4.21-47.0.1.EL (RHEL 3)
4864 2.6.12.6 vanilla (kernel.org)
4868 Frequency : quota enabled and large files being deleted
4870 Description: releasing more than 4GB of quota at once hangs OST
4871 Details : If a user deletes more than 4GB of files on a single OST it
4872 will cause the OST to spin in an infinite loop. Release
4873 quota in < 4GB chunks, or use a 64-bit value for 1.4.7.1+.
4878 Description: statfs data retrieved from /proc may be stale or zero
4879 Details : When reading per-device statfs data from /proc, in the
4880 {kbytes,files}_{total,free,avail} files, it may appear
4881 as zero or be out of date.
4884 Frequency : systems with MD RAID1 external journal devices
4886 Description: lconf's call to blkid is confused by RAID1 journal devices
4887 Details : Use the "blkid -l" flag to locate the MD RAID device instead
4888 of returning all block devices that match the journal UUID.
4891 Frequency : always, for aggregate stripe size over 4GB
4893 Description: "lfs setstripe" fails assertion when setting 4GB+ stripe width
4894 Details : Using "lfs setstripe" to set stripe size * stripe count over 4GB
4895 will fail the kernel with "ASSERTION(lsm->lsm_xfersize != 0)"
4898 Frequency : always if "lfs find" used on a local file/directory
4900 Description: "lfs find" segfaults if used on a local file/directory
4901 Details : The case where a directory component was not specified wasn't
4902 handled correctly. Handle this properly.
4905 Frequency : always on ppc64
4907 Description: the write to an ext3 filesystem mounted with mballoc got stuck
4908 Details : ext3_mb_generate_buddy() uses find_next_bit() which does not
4909 perform endianness conversion.
4912 Frequency : rarely (truncate to non-zero file size after write under load)
4913 Bugzilla : 10730, 10687
4914 Description: Files padded with zeros to next 4K multiple
4915 Details : With filesystems mounted using the "extents" option (2.6 kernels)
4916 it is possible that files that are truncated to a non-zero size
4917 immediately after being written are filled with zero bytes beyond
4918 the truncated size. No file data is lost.
4920 Severity : enhancement
4922 Description: Allow recovery/failover for liblustre clients.
4923 Details : liblustre clients were unaware of failover configurations until
4926 Severity : enhancement
4928 Description: user file locks should fail when not mounting with flock option
4929 Details : Set up an error-returning stub in ll_file_operations.lock field
4930 to prevent incorrect behaviour when client is mounted without
4931 flock option. Also, set up properly f_op->flock field for
4935 Frequency : always on ia64
4937 Description: "lfs df" loops on printing out MDS statfs information
4938 Details : The obd_ioctl_data was not initialized and in some systems
4939 this caused a failure during the ioctl that did not return
4940 an error. Initialize the struct and return an error on failure.
4943 Frequency : SLES 9 only
4945 Description: Error of copying files with lustre special EAs as root
4946 Details : Client side always return success for setxattr call for lustre
4947 special xattr (currently only "trusted.lov").
4950 Frequency : rarely on clusters with both ia64+i386 clients
4952 Description: ia64+i686 clients doing shared IO on the same file may LBUG
4953 Details : In rare cases when both ia64+i686 (or other mixed-PAGE_SIZE)
4954 clients are doing concurrent writes to the same file it is
4955 possible that the ia64 clients may LASSERT because the OST
4956 extent locks are not PAGE_SIZE aligned. Ensure that grown
4957 locks are always aligned on the request boundary.
4960 Frequency : specific use, occasional
4962 Description: Overwriting in use executable truncates on-disk binary image
4963 Details : If one node attempts to overwrite an executable in use by
4964 another node, we now correctly return ETXTBSY instead of
4965 truncating the file.
4967 Severity : enhancement
4969 Description: Async OSC create to avoid the blocking unnecessarily.
4970 Details : If a OST has no remain object, system will block on the creating
4971 when need to create a new object on this OST. Now, ways use
4972 pre-created objects when available, instead of blocking on an
4973 empty osc while others are not empty. If we must block, we block
4974 for the shortest possible period of time.
4979 Description: chmod on Lustre root is propagated to other clients
4980 Details : Re-validate root's dentry in ll_lookup_it to avoid having it
4981 invalid by the follow_mount time.
4984 Frequency : liblustre clients only
4986 Description: Race in 'instant cancel' lock handling could lead to such locks
4987 never to be granted in case of SMP MDS
4988 Details : Do not destroy not yet granted but cbpending locks in
4992 Frequency : replay/resend of open
4994 Description: non null lock assetion failure in mds_intent_policy
4995 Details : Trying to replay/resend lockless open requests resulted in
4996 mds_open() returning 0 with no lock. Now it sets a flag if
4997 a lock is going to be returned.
4999 Severity : enhancement
5001 Description: Checksum enhancements
5002 Details : New checksum enhancements allow for resending RPCs that failed
5005 Severity : enhancement
5007 Description: Tunables on number of dirty pages in cacche
5008 Details : Allow to set limit on number of dirty pages cached.
5013 Description: client crash on unmount - lock still has references
5014 Details : In some error handling cases it was possible to leak a lock
5015 reference on a client while accessing a file. This was not
5016 harmful to the client during operation, but would cause the
5017 client to crash when the filesystem is unmounted.
5020 Frequency : specific case, rare
5022 Description: ETXTBSY on mds though file not in use by client
5023 Details : ETXTBSY is no longer incorrectly returned when attempting to
5024 chmod or chown a directory that the user previously tried to
5025 execute or a currently-executing binary.
5028 Frequency : extremely rare except on liblustre-based clients
5030 Description: Lustre space not freed when files are deleted
5031 Details : Clean up open-unlinked files after client eviction. Previously
5032 the unlink was skipped and the files remained as orphans.
5037 Description: OST failure "would be an LBUG" in waiting_locks_callback()
5038 Details : In some cases it was possible to send a blocking callback to
5039 a client doing a glimpse, even though that client didn't get
5040 a lock granted. When the glimpse lock is cancelled on the OST
5041 the freed lock is left on the waiting list and corrupted the list.
5044 Frequency : all core dumps
5046 Description: Broke core dumps to lustre
5047 Details : Negative dentry may be unhashed if parent does not have UPDATE
5048 lock, but some callers, e.g. do_coredump, expect dentry to be
5049 hashed after successful create, hash it in ll_create_it.
5051 ------------------------------------------------------------------------------
5053 2006-09-13 Cluster File Systems, Inc. <info@clusterfs.com>
5055 * Support for kernels:
5056 2.6.9-42.0.2.EL (RHEL 4)
5057 2.6.5-7.276 (SLES 9)
5058 2.4.21-40.EL (RHEL 3)
5059 2.6.12.6 vanilla (kernel.org)
5063 Frequency : always on RHEL 3
5065 Description: Number of open files grows over time
5066 Details : The number of open files grows over time, whether or not
5067 Lustre is started. This was due to a filp leak introduced
5068 by one of our kernel patches.
5070 ------------------------------------------------------------------------------
5072 2006-08-20 Cluster File Systems, Inc. <info@clusterfs.com>
5074 * Support for kernels:
5075 2.6.9-42.EL (RHEL 4)
5076 2.6.5-7.267 (SLES 9)
5077 2.4.21-40.EL (RHEL 3)
5078 2.6.12.6 vanilla (kernel.org)
5083 Bugzilla : 5719, 9635, 9792, 9684
5084 Description: OST (or MDS) trips assertions in (re)connection under heavy load
5085 Details : If a server is under heavy load and cannot reply to new
5086 connection requests before the client resends the (re)connect,
5087 the connection handling code can behave badly if two service
5088 threads are concurrently handing separate (re)connections from
5089 the same client. Add better locking to the connection handling
5090 code, and ensure that only a single connection will be processed
5091 for a given client UUID, even if the lock is dropped.
5093 Severity : enhancement
5095 Description: add TCP zero-copy support to kernel
5096 Details : Add support to the kernel TCP stack to allow zero-copy bulk
5097 sends if the hardware supports scatter-gather and checksumming.
5098 This allows socklnd to do client-write and server-read more
5099 efficiently and reduce CPU utilization from skbuf copying.
5102 Frequency : only if NFS exporting from client
5104 Description: NULL pointer deref in ll_iocontrol() if chattr mknod file
5105 Details : If setting attributes on a file created under NFS that had
5106 never been opened it would be possible to oops the client
5107 if the file had no objects.
5111 Bugzilla : 9326, 10402, 10897
5112 Description: client crash in ptlrpcd_wake() thread when sending async RPC
5113 Details : It is possible that ptlrpcd_wake() dereferences a freed async
5114 RPC. In rare cases the ptlrpcd thread alread processed the RPC
5115 before ptlrpcd_wake() was called and the request was freed.
5118 Frequency : always for liblustre
5120 Description: liblustre client does MDS+OSTs setattr RPC for each write
5121 Details : When doing a write from a liblustre client, the client
5122 incorrectly issued an RPC to the MDS and each OST the file was
5123 striped over in order to update the timestamps. When writing
5124 with small chunks and many clients this could overwhelm the MDS
5125 with RPCs. In all cases it would slow down the write because
5126 these RPCs are unnecessary.
5128 Severity : enhancement
5130 Description: allow number of MDS service threads to be changed at module load
5131 Details : It is now possible to change the number of MDS service threads
5132 running. Adding "options mds mds_num_threads={N}" to the MDS's
5133 /etc/modprobe.conf will set the number of threads for the next
5134 time Lustre is restarted (assuming the "mds" module is also
5135 reloaded at that time). The default number of threads will
5136 stay the same, 32 for most systems.
5141 Description: OST crash if filesystem is unformatted or corrupt
5142 Details : If an OST is started on a device that has never been formatted
5143 or if the filesystem is corrupt and cannot even mount then the
5144 error handling cleanup routines would dereference a NULL pointer.
5149 Description: NULL pointer deref in llap_from_page.
5150 Details : get_cache_page_nowait can return a page with NULL (or otherwise
5151 incorrect) mapping if the page was truncated/reclaimed while it was
5152 searched for. Check for this condition and skip such pages when
5153 doing readahead. Introduce extra check to llap_from_page() to
5154 verify page->mapping->host is non-NULL (so page is not anonymous).
5157 Frequency : Sometimes when using sys_sendfile
5159 Description: "page not covered by a lock" warnings from ll_readpage
5160 Details : sendfile called ll_readpage without right page locks present.
5161 Now we introduced ll_file_sendfile that does necessary locking
5162 around call to generic_file_sendfile() much like we do in
5166 Frequency : with certain MDS communication failures at client mount time
5168 Description: NULL pointer deref after failed client mount
5169 Details : a client connection request may delayed by the network layer
5170 and not be sent until after the PTLRPC layer has timed out the
5171 request. If the client fails the mount immediately it will try
5172 to clean up before the network times out the request. Add a
5173 reference from the request import to the obd device and delay
5174 the cleanup until the network drops the request.
5177 Frequency : occasionally during client (re)connect
5179 Description: assertion failure during client (re)connect
5180 Details : processing a client connection request may be delayed by the
5181 client or server longer than the client connect timeout. This
5182 causes the client to resend the connection request. If the
5183 original connection request is replied in this interval, the
5184 client may trip an assertion failure in ptlrpc_connect_interpret()
5185 which thought it would be the only running connect process.
5188 Frequency : only with obd_echo servers and clients that are rebooted
5190 Description: kernel BUG accessing uninitialized data structure
5191 Details : When running an obd_echo server it did not start the ping_evictor
5192 thread, and when a client was evicted an uninitialized data
5193 structure was accessed. Start the ping_evictor in the RPC
5194 service startup instead of the OBD startup.
5196 Severity : enhancement
5197 Bugzilla : 10193 (patchless)
5198 Description: Remove dependency on various unexported kernel interfaces.
5199 Details : No longer need reparent_to_init, exit_mm, exit_files,
5200 sock_getsockopt, filemap_populate, FMODE_EXEC, put_filp.
5203 Frequency : rare (only users of deprecated and unsupported LDAP config)
5205 Description: write_conf for zeroconf mount queried LDAP incorrectly for client
5206 Details : LDAP apparently contains 'lustreName' attributes instead of
5207 'name'. A simple remapping of the name is sufficient.
5210 Frequency : rare (only with non-default dump_on_timeout debug enabled)
5212 Description: waiting_locks_callback trips kernel BUG if client is evicted
5213 Details : Running with the dump_on_timeout debug flag turned on makes
5214 it possible that the waiting_locks_callback() can try to dump
5215 the Lustre kernel debug logs from an interrupt handler. Defer
5216 this log dumping to the expired_lock_main() thread.
5218 Severity : enhancement
5220 Description: Support NFS exporting on 2.6 kernels.
5221 Details : Implement non-rawops metadata methods for NFS server to use without
5222 changing NFS server code.
5225 Frequency : very rare (synthetic metadata workload only)
5227 Description: two racing renames might cause an MDS thread to deadlock
5228 Details : Running the "racer" program may cause one MDS thread to rename
5229 a file from being the source of a rename to being the target of
5230 a rename at exactly the same time that another thread is doing
5231 so, and the second thread has already enqueued these locks after
5232 doing a lookup of the target and is trying to relock them in
5233 order. Ensure that we don't try to re-lock the same resource.
5236 Frequency : only very large systems with liblustre clients
5238 Description: slow eviction of liblustre clients with the "evict_by_nid" RPC
5239 Details : Use asynchronous set_info RPCs to send the "evict_by_nid" to
5240 all OSTs in parallel. This allows the eviction of stale liblustre
5241 clients to proceed much faster than if they were done in series,
5242 and also offers similar improvements for other set_info RPCs.
5247 Description: excessive CPU usage during initial read phase on client
5248 Details : During the initial read phase on a client, it would agressively
5249 retry readahead on the file, consuming too much CPU and impacting
5250 performance (since 1.4.5.8). Improve the readahead algorithm
5251 to avoid this, and also improve some other common cases (read
5252 of small files in particular, where "small" is files smaller than
5253 /proc/fs/lustre/llite/*/max_read_ahead_whole_mb, 2MB by default).
5258 Description: MDS crash when receiving packet with unknown intent.
5259 Details : Do not LBUG in unknown intent case, just return -EFAULT
5261 Severity : enhancement
5262 Bugzilla : 9293, 9385
5263 Description: MDS RPCs are serialised on client. This is unnecessary for some.
5264 Details : Do not serialize getattr (non-intent version) and statfs.
5267 Frequency : occasional, when OST network is overloaded/intermittent
5269 Description: client evicted by OST after bulk IO timeout
5270 Details : If a client sends a bulk IO request (read or write) the OST
5271 may evict the client if it is unresposive to its data GET/PUT
5272 request. This is incorrect if the network is overloaded (takes
5273 too long to transfer the RPC data) or dropped the OST GET/PUT
5274 request. There is no need to evict the client at all, since
5275 the pinger and/or lock callbacks will handle this, and the
5276 client can restart the bulk request.
5279 Frequency : Always when mmapping file with no objects
5281 Description: client crashes when mmapping file with no objects
5282 Details : Check that we actually have objects in a file before doing any
5283 operations on objects in ll_vm_open, ll_vm_close and
5289 Description: Request leak when working with deleted CWD
5290 Details : Introduce advanced request refcount tracking for requests
5291 referenced from lustre intent.
5293 Severity : Enhancement
5295 Description: Cache open file handles on client.
5296 Details : MDS now will return special lock along with openhandle, if
5297 requested and client is allowed to hold openhandle, even if unused,
5298 until such a lock is revoked. Helps NFS a lot, since NFS is opening
5299 closing files for every read/write openration.
5301 Severity : Enhancement
5303 Description: Cache open negative dentries on client when possible.
5304 Details : Guard negative dentries with UPDATE lock on parent dir, drop
5305 negative dentries on lock revocation.
5310 Description: Remounting a client read-only wasn't possible with a zconf mount
5311 Details : It wasn't possible to remount a client read-only with llmount.
5313 Severity : enhancement
5314 Description: Include MPICH 1.2.6 Lustre ADIO interface patch
5315 Details : In lustre/contrib/ or /usr/share/lustre in RPM a patch for
5316 MPICH is included to add Lustre-specific ADIO interfaces.
5317 This is based closely on the UFS ADIO layer and only differs
5318 in file creation, in order to allow the OST striping to be set.
5319 This is user-contributed code and not supported by CFS.
5324 Description: extended inode attributes (immutable, append-only) work improperly
5325 when 2.4 and 2.6 kernels are used on client/server or vice versa
5326 Details : Introduce kernel-independent values for these flags.
5328 Severity : enhancement
5331 Description: Allow fractional MB tunings for lustre in /proc/ filesystem.
5332 Details : Many of the /proc/ tunables can only be tuned at a megabyte
5333 granularity. Now, Fractional MB granularity is be supported,
5334 this is very useful for low memory system.
5336 Severity : enhancement
5338 Description: Getattr by fid
5339 Details : Getting a file attributes by its fid, obtaining UPDATE|LOOKUP
5340 locks, avoids extra getattr rpc requests to MDS, allows '/' to
5341 have locks and avoids getattr rpc requests for it on every stat.
5344 Frequency : Always, for filesystems larger than 2TB
5346 Description: ldiskfs crash at mount for filesystem larger than 2TB with mballoc
5347 Details : Kenrel kmalloc limits allocations to 128kB and this prevents
5348 filesystems larger than 2TB to be mounted with mballoc enabled.
5351 Frequency : Always, for 32-bit kernel without CONFIG_LBD and filesystem > 2TB
5353 Description: filesystem corruption for non-standard kernels and very large OSTs
5354 Details : If a 32-bit kernel is compiled without CONFIG_LBD enabled and a
5355 filesystems larger than 2TB is mounted then the kernel will
5356 silently corrupt the start of the filesystem. CONFIG_LBD is
5357 enabled for all CFS-supported kernels, but the possibility of
5358 this happening with a modified kernel config exists.
5360 Severity : enhancement
5362 Description: add client O_DIRECT support for 2.6 kernels
5363 Details : It is now possible to do O_DIRECT reads and writes to files
5364 in the Lustre client mountpoint on 2.6 kernel clients.
5366 Severity : enhancement
5368 Description: parallel glimpse, setattr, statfs, punch, destroy requests
5369 Details : Sends glimpse, setattr, statfs, punch, destroy requests to OSTs in
5370 parallel, not waiting for response from every OST before sending
5371 a rpc to the next OST.
5376 Description: setattr vs write race when updating file timestamps
5377 Details : Client processes that update a file timestamp into the past
5378 right after writing to the file (e.g. tar) it is possible that
5379 the updated file modification time can be reset to the current
5380 time due to a race between processing the setattr and write RPC.
5382 Severity : enhancement
5384 Description: Bring 'lfs find' closer in line with regular Linux find.
5385 Details : lfs find util supports -atime, -mtime, -ctime, -maxdepth, -print,
5386 -print0 options and obtains all the needed info through the lustre
5389 Severity : enhancement
5391 Description: support up to 1024 configured devices on one node
5392 Details : change obd_dev array from statically allocated to dynamically
5393 allocated structs as they are first used to reduce memory usage
5398 Description: Flush dirty partially truncated pages during truncate
5399 Details : Immediatelly flush partially truncated pages in filter_setattr,
5400 this way we completely avoid having any pages in page cache on OST
5401 and can retire ugly workarounds during writes to flush such pages.
5406 Description: i_sem vs transaction deadlock in mds_obd_destroy during unlink.
5407 Details : protect inode from truncation within vfs_unlink() context
5408 just take a reference before calling vfs_unlink() and release it
5409 when parent's i_sem is free.
5412 Frequency : always, if extents are used on OSTs
5414 Description: index ei_leaf_hi (48-bit extension) is not zeroed in extent index
5415 Details : OSTs using the extents format would not zero the high 16 bits of
5416 the index physical block number. This is not a problem for any
5417 OST filesystems smaller than 16TB, and no kernels support ext3
5418 filesystems larger than 16TB yet. This is fixed in 1.4.7 (all
5419 new/modified files) and can be fixed for existing filesystems
5420 with e2fsprogs-1.39-cfs1.
5425 Description: import connection selection may be incorrect if timer wraps
5426 Details : Using a 32-bit jiffies timer with HZ=1000 may cause backup
5427 import connections to be ignored if the 32-bit jiffies counter
5428 wraps. Use a 64-bit jiffies counter.
5431 Frequency : during server recovery
5433 Description: crash after server is denying duplicate export
5434 Details : If clients are resending connect requests to the server, the
5435 server refuses to allow a client to connect multiple times.
5436 Fixed a bug in the handling of this case.
5439 Frequency : very large clusters immediately after boot
5441 Description: LNET request buffers exhausted under heavy short-term load
5442 Details : If a large number of client requests are generated on a service
5443 that has previously never seen so many requests it is possible
5444 that the request buffer growth cannot keep up with the spike in
5445 demand. Instead of dropping incoming requests, they are held in
5446 the LND until the RPC service can accept more requests.
5449 Frequency : Sometimes during replay
5451 Description: Assertion failure in ll_local_open after replay.
5452 Details : If replay happened on an open request reply before we were able
5453 to set replay handler, reply will become not swabbed tripping the
5454 assertion in ll_local_open. Now we set the handler right after
5455 recognising of open request
5458 Frequency : very rare
5460 Description: kernel reports "badness in vsnprintf"
5461 Details : Reading from the "recovery_status" /proc file in small chunks
5462 may cause a negative length in lprocfs_obd_rd_recovery_status()
5463 call to vsnprintf() (which is otherwise harmless). Exit early
5464 if there is no more space in the output buffer.
5466 Severity : enhancement
5468 Description: clear OBD RPC statistics by writing to them
5469 Details : It is now possible to clear the OBD RPC statistics by writing
5470 to the "stats" file.
5475 Description: Client mtime is not the same on different clients after utimes
5476 Details : In some cases, the client was using the utimes() syscall on
5477 a file cached on another node. The clients now validate the
5478 ctime from the MDS + OSTs to determine which one is right.
5483 Description: Inability to activate failout mode
5484 Details : lconf script incorrectly assumed that in python string's numeric
5485 value is used in comparisons.
5488 Frequency : always with multiple stripes per file
5490 Description: Inefficient object allocation for mutli-stripe files
5491 Details : When selecting which OSTs to stripe files over, for files with
5492 a stripe count that divides evenly into the number of OSTs,
5493 the MDS is always picking the same starting OST for each file.
5494 Return the OST selection heuristic to the original design.
5499 Description: mount failures may take full timeout to return an error
5500 Details : Under some heavy load conditions it is possible that a
5501 failed mount can wait for the full obd_timeout interval,
5502 possibly several minutes, before reporting an error.
5503 Instead return an error as soon as the status is known.
5505 ------------------------------------------------------------------------------
5507 2006-02-14 Cluster File Systems, Inc. <info@clusterfs.com>
5509 * WIRE PROTOCOL CHANGE. This version of Lustre networking WILL NOT
5510 INTEROPERATE with older versions automatically. Please read the
5511 user documentation before upgrading any part of a live system.
5512 * WARNING: Lustre networking configuration changes are required with
5513 this release. See https://bugzilla.clusterfs.com/show_bug.cgi?id=10052
5516 * Support for kernels:
5517 2.6.9-22.0.2.EL (RHEL 4)
5518 2.6.5-7.244 (SLES 9)
5519 2.6.12.6 vanilla (kernel.org)
5522 Severity : enhancement
5523 Bugzilla : 7981/8208
5524 Description: Introduced Lustre Networking (LNET)
5525 Details : LNET is new networking infrastructure for Lustre, it includes
5526 a reorganized network configuration mode (see the user
5527 documentation for full details) as well as support for routing
5528 between different network fabrics. Lustre Networking Devices
5529 (LNDs) for the supported network fabrics have also been
5530 created for this new infrastructure.
5532 Severity : enhancement
5533 Description: Introduced Access control lists
5534 Details : clients can set ACLs on files and directories in order to have
5535 more fine-grained permissions than the standard Unix UGO+RWX.
5536 The MDS must be started with the "-o acl" mount option.
5538 Severity : enhancement
5539 Description: Introduced filesystem quotas
5540 Details : Administrators may now establish per-user quotas on the
5543 Severity : enhancement
5545 Description: Configuration change for the XT3
5546 The PTLLND is now used to run Lustre over Portals on the XT3
5547 The configure option(s) --with-cray-portals are no longer used.
5548 Rather --with-portals=<path-to-portals-includes> is used to
5549 enable building on the XT3. In addition to enable XT3 specific
5550 features the option --enable-cray-xt3 must be used.
5555 Description: Running on many-way SMP OSTs can trigger oops in llcd_send()
5556 Details : A race between allocating a new llcd and re-getting the llcd_lock
5557 allowed another thread to grab newly-allocated llcd.
5559 Severity : enhancement
5561 Description: 2.6 OST async journal commit and locking fix to improve performance
5562 Details : The filter_direct_io()+filter_commitrw_write() journal commits for
5563 2.6 kernels are now async as they already were in 2.4 kernels so
5564 that they can commit concurrently with the network bulk transfer.
5565 For block-allocated files the filter allocation semaphore is held
5566 to avoid filesystem fragmentation during allocation. BKL lock
5567 removed for 2.6 xattr operations where it is no longer needed.
5572 Description: lconf incorrectly determined whether two IP networks could talk
5573 Details : In some more complicated routing and multiple-network
5574 configurations, lconf will avoid trying to make a network
5575 connection to a disjoint part of the IP space. It was doing the
5576 math incorrectly for one set of cases.
5581 Description: Fix for potential infinite loop processing records in an llog.
5582 Details : If an llog record is corrupted/zeroed, it is possible to loop
5583 forever in llog_process(). Validate the llog record length
5584 and skip the remainder of the block on error.
5587 Frequency : occasional (liblustre only)
5589 Description: liblustre could not open files whose last component is a symlink
5590 Details : sysio_path_walk() would incorrectly pass the open intent to
5591 intermediate path components.
5594 Frequency : rare (liblustre only with non-standard tuning)
5595 Bugzilla : 7201 (7350)
5596 Description: Tuning the MDC DLM LRU size to zero triggers client LASSERT
5597 Details : llu_lookup_finish_locks() tries to set lock data on a lock
5598 after it has been released, only do this for referenced locks
5600 Severity : enhancement
5602 Description: specifying an (invalid) directory default stripe_size of -1
5603 would reset the directory default striping
5604 Details : stripe_size -1 was used internally to signal directory stripe
5605 removal, now use "all default" to signal dir stripe removal
5606 as a directory striping of "all default" is not useful
5609 Frequency : common for large clusters running liblustre clients
5611 Description: doing an ls when liblustre clients are running is slow
5612 Details : sending a glimpse AST to a liblustre client waits for every AST
5613 to time out, as liblustre clients will not respond. Since they
5614 cannot cache data we refresh the OST lock LVB from disk instead.
5616 Severity : enhancement
5618 Description: doing an ls at the same time as file IO can be slow
5619 Details : enqueue and other "small" requests can be blocked behind many
5620 large IO requests. Create a new OST IO portal for non-IO
5621 requests so they can be processed faster.
5624 Frequency : rare (only HPUX clients mounting unsupported re-exported NFS vol)
5626 Description: an HPUX NFS client would get -EACCESS when ftruncate()ing a newly
5627 created file with mode 000
5628 Details : the Linux NFS server relies on an MDS_OPEN_OWNEROVERRIDE hack to
5629 allow an ftruncate() as a non-root user to a file with mode 000.
5630 Lustre now respects this flag to disable mode checks when
5631 truncating a file owned by the user
5634 Frequency : liblustre-only, when liblustre client dies unexpectedly or becomes
5637 Description: Revoking locks from clients that went dead or catatonic might take
5639 Details : New lock flags FL_CANCEL_ON_BLOCK used by liblustre makes
5640 cancellation of such locks instant on servers without waiting for
5641 any reply from clients. Clients drops these locks when cancel
5642 notification from server is received without replying.
5645 Frequency : liblustre-only, when liblustre client dies or becomes busy
5647 Description: Doing ls on Linux clients can take a long time with active
5649 Details : Liblustre client cannot handle ASTs in timely manner, so avoid
5650 granting such locks to it in the first place if possible. Locks
5651 are taken by proxy on the OST during the read or write and
5652 dropped immediately afterward. Add connect flags handling, do
5653 not grant locks to liblustre clients for glimpse ASTs.
5655 Severity : enhancement
5657 Description: Improve read-ahead algorithm to avoid excessive IO for random reads
5658 Details : Existing read-ahead algorithm is tuned for the case of streamlined
5659 sequential reads and behaves badly with applications doing random
5660 reads. Improve it by reading ahead at least read region, and
5661 avoiding excessive large RPC for small reads.
5663 Severity : enhancement
5665 Description: Creating more than 1000 files for a single job may cause a load
5666 imbalance on the OSTs if there are also a large number of OSTs.
5667 Details : qos_prep_create() uses an OST index reseed value that is an
5668 even multiple of the number of available OSTs so that if the
5669 reseed happens in the middle of the object allocation it will
5670 still utilize the OSTs as uniformly as possible.
5675 Description: OST or MDS may oops in ping_evictor_main()
5676 Details : ping_evictor_main() drops obd_dev_lock if deleting a stale export
5677 but doesn't restart at beginning of obd_exports_timed list
5680 Severity : enhancement
5682 Description: improve by-nid export eviction on the MDS and OST
5683 Details : allow multiple exports with the same NID to be evicted at one
5684 time without re-searching the exports list.
5687 Frequency : rare, only with supplementary groups enabled on SMP 2.6 kernels
5689 Description: MDS may oops in groups_free()
5690 Details : in rare race conditions a newly allocated group_info struct is
5691 freed again, and this can be NULL. The 2.4 compatibility code
5692 for groups_free() checked for a NULL pointer, but 2.6 did not.
5695 Frequency : common for liblustre clients doing little filesystem IO
5696 Bugzilla : 9352, 7313
5697 Description: server may evict liblustre clients accessing contended locks
5698 Details : if a client is granted a lock or receives a completion AST
5699 with a blocking AST already set it would not reply to the AST
5700 for LDLM_FL_CANCEL_ON_BLOCK locks. It now replies to such ASTs.
5703 Frequency : lfs setstripe, only systems with more than 160 OSTs
5705 Description: unable to set striping with a starting offset beyond OST 160
5706 Details : llapi_create_file() incorrectly limited the starting stripe
5707 index to the maximum single-file stripe count.
5710 Frequency : LDAP users only
5712 Description: lconf did not handle in-kernel recovery with LDAP properly
5713 Details : lconf/LustreDB get_refs() is searching the wrong namespace
5715 Severity : enhancement
5717 Description: bind OST threads to NUMA nodes to improve performance
5718 Details : all OST threads are uniformly bound to CPUs on a single NUMA
5719 node and do their allocations there to localize memory access
5721 Severity : enhancement
5723 Description: llmount can determine client NID directly from Myrinet (GM)
5724 Details : the client NID code from gmnalnid was moved directly into
5725 llmount, removing the need to use this or specifying the
5726 client NID explicitly when mounting GM clients with zeroconf
5729 Frequency : if client is started with down MDS
5731 Description: if client is started with down MDS mount hangs in ptlrpc_queue_wait
5732 Details : Having an LWI_INTR() wait event (interruptible, but no timeout)
5733 will wait indefinitely in ptlrpc_queue_wait->l_wait_event() after
5734 ptlrpc_import_delayed_req() because we didn't check if the
5735 request was interrupted, and we also didn't break out of the
5736 event loop if there was no timeout
5741 Description: data loss during non-page-aligned writes to a single file from
5742 both multiple nodes and multiple threads on one node at same time
5743 Details : updates to KMS and lsm weren't protected by common lock. Resulting
5744 inconsistency led to false short-reads, that were cached and later
5745 used by ->prepare_write() to fill in partially written page,
5746 leading to data loss.
5749 Frequency : always, if lconf --abort_recovery used
5751 Description: lconf --abort_recovery fails with 'Operation not supported'
5752 Details : lconf was attempting to abort recovery on the MDT device and not
5755 Severity : enhancement
5757 Description: remove cleanup logs
5758 Details : replace lconf-generated cleanup logs with lustre internal
5759 cleanup routines. Eliminates the need for client-cleanup and
5762 Severity : enhancement
5764 Description: add support for EAs (user and system) on lustre filesystems
5765 Details : it is now possible to store extended attributes in the Lustre
5766 client filesystem, and with the user_xattr mount option it
5767 is possible to allow users to store EAs on their files also
5769 Severity : enhancement
5771 Description: Add possibility (config option) to show minimal available OST free
5773 Details : When compiled with --enable-mindf configure option, statfs(2)
5774 (and so, df) will return least minimal free space available from
5775 all OSTs as amount of free space on FS, instead of summary of
5776 free spaces of all OSTs.
5778 Severity : enhancement
5780 Description: do not expand extent locks acquired on OST-side
5781 Details : Modify ldlm_extent_policy() to not expand local locks, acquired
5782 by server: they are not cached anyway.
5785 Frequency : when mmap is used/binaries executed from Lustre
5787 Description: Unmmap pages before throwing them away from read cache.
5788 Details : llap_shrink cache now attempts to unmap pages before discarding
5789 them (if unmapping failed - do not discard). SLES9 kernel has
5790 extra checks that trigger if this unmapping is not done first.
5795 Description: lconf didn't resolve symlinks before checking to see whether a
5796 given mountpoint was already in use
5799 Frequency : when migrating failover services
5800 Bugzilla : 6395, 9514
5801 Description: When migrating a subset of services from a node (e.g. failback
5802 from a failover service node) the remaining services would
5803 time out and evict clients.
5804 Details : lconf --force (implied by --failover) sets the global obd_timeout
5805 to 5 seconds in order to quickly disconnect, but this caused
5806 other RPCs to time out too quickly. Do not change the global
5807 obd_timeout for force cleanup, only set it for DISCONNECT RPCs.
5809 Severity : enhancement
5810 Frequency : if MDS is started with down OST
5811 Bugzilla : 9439,5706
5812 Description: Allow startup/shutdown of an MDS without depending on the
5813 availability of the OSTs.
5814 Details : Asynchronously call mds_lov_synchronize during MDS startup.
5815 Add appropriate locking and lov-osc refcounts for safe
5816 cleaning. Add osc abort_inflight calls in case the
5817 synchronize never started.
5820 Frequency : occasional (Cray XT3 only)
5822 Description: root not authorized to access files in CRAY_PORTALS environment
5823 Details : The client process capabilities were not honoured on the MDS in
5824 a CRAY_PORTALS/CRAY_XT3 environment. If the file had previously
5825 been accessed by an authorized user then root was able to access
5826 the file on the local client also. The root user capabilities
5827 are now allowed on the MDS, as this environment has secure UID.
5830 Frequency : occasional
5832 Description: ldiskfs "too long searching" message happens too often
5833 Details : A debugging message (otherwise harmless) prints too often on
5834 the OST console. This has been reduced to only happen when
5835 there are fragmentation problems on the filesystem.
5840 Description: Division by zero in statfs when all OSCs are inactive
5841 Details : lov_get_stripecnt() returns zero due to incorrect order of checks,
5842 lov_statfs divides by value returned by lov_get_stripecnt().
5846 Bugzilla : 9489, 3273
5847 Description: First write from each client to each OST was only 4kB in size,
5848 to initialize client writeback cache, which caused sub-optimal
5849 RPCs and poor layout on disk for the first writen file.
5850 Details : Clients now request an initial cache grant at (re)connect time
5851 and so that they can start streaming writes to the cache right
5852 away and always do full-sized RPCs if there is enough data.
5853 If the OST is rebooted the client also re-establishes its grant
5854 so that client cached writes will be honoured under the grant.
5859 Description: Slow ls (and stat(2) syscall) on files residing on IO-loaded OSTs
5860 Details : Now I/O RPCs go to different portal number and (presumably) fast
5861 lock requests (and glimses) and other RPCs get their own service
5862 threads pool that should be able to service those RPCs
5865 Severity : enhancement
5867 Description: Ability to exchange lustre version between client and servers and
5868 issue warnings at client side if client is too old. Also for
5869 liblustre clients there is ability to refuse connection of too old
5871 Details : New 'version' field is added to connect data structure that is
5872 filled with version info. That info is later checked by server and
5876 Frequency : rare, liblustre only.
5877 Bugzilla : 9296, 9581
5878 Description: Two simultaneous writes from liblustre at offset within same page
5879 might proceed at the same time overwriting eachother with stale
5881 Details : I/O lock withing llu_file_prwv was released too early, before data
5882 actually was hitting the wire. Extended lock-holding time until
5883 server acknowledges receiving data.
5886 Frequency : extremely rare. Never observed in practice.
5888 Description: avoid generating lustre_handle cookie of 0.
5889 Details : class_handle_hash() generates handle cookies by incrementing
5890 global counter, and can hit 0 occasionaly (this is unlikely, but
5891 not impossible, because initial value of cookie counter is
5892 selected randonly). Value of 0 is used as a sentinel meaning
5893 "unassigned handle" --- avoid it. Also coalesce two critical
5894 sections in this function into one.
5896 Severity : enhancement
5898 Description: allow liblustre clients to delegate truncate locking to OST
5899 Details : To avoid overhead of locking, liblustre client instructs OST to
5900 take extent lock in ost_punch() on client's behalf. New connection
5901 flag is added to handle backward compatibility.
5903 Severity : enhancement
5904 Bugzilla : 4928, 7341, 9758
5905 Description: allow number of OST service threads to be specified
5906 Details : a module parameter allows the number of OST service threads
5907 to be specified via "options ost ost_num_threads={N}" in the
5908 OSS's /etc/modules.conf or /etc/modprobe.conf.
5912 Bugzilla : 6146, 9635, 9895
5913 Description: servers crash with bad pointer in target_handle_connect()
5914 Details : In rare cases when a client is reconnecting it was possible that
5915 the connection request was the last reference for that export.
5916 We would temporarily drop the export reference and get a new
5917 one, but this may have been the last reference and the export
5918 was just destroyed. Get new reference before dropping old one.
5920 Severity : enhancement
5921 Frequency : if client is started with failover MDS
5923 Description: Allow multiple MDS hostnames in the mount command
5924 Details : Try to read the configuration from all specified MDS
5925 hostnames during a client mount in case the "primary"
5928 Severity : enhancement
5930 Description: Stop sending data to evicted clients as soon as possible.
5931 Details : Check if the client we are about to send or are sending data to
5932 was evicted already. (Check is done every second of waiting,
5933 for which l_wait_event interface was extended to allow checking
5934 of exit condition at specified intervals).
5937 Frequency : rare, normally only when NFS exporting is done from client
5939 Description: 'bad disk LOV MAGIC: 0x00000000' error when chown'ing files
5941 Details : Make mds_get_md() recognise empty md case and set lmm size to 0.
5944 Frequency : always, if srand() is called before liblustre initialization
5946 Description: Liblustre uses system PRNG disturbing its usage by user application
5947 Details : Introduce internal to lustre fast and high-quality PRNG for
5948 lustre usage and make liblustre and some other places in generic
5949 lustre code to use it.
5951 Severity : enhancement
5952 Bugzilla : 9477, 9557, 9870
5953 Description: Verify that the MDS configuration logs are updated when xml is
5954 Details : Check if the .xml configuration logs are newer than the config
5955 logs stored on the MDS and report an error if this is the case.
5956 Request --write-conf, or allow starting with --old_conf.
5958 Severity : enhancement
5960 Description: Handle symlinks in the path when checking if Lustre is mounted.
5961 Details : Resolve intermediate symlinks when checking if a client has
5962 mounted a filesystem to avoid duplicate client mounts.
5967 Description: lconf can hit an error exception but still return success.
5968 Details : The lconf command catches the Command error exception at the top
5969 level script context and will exit with the associated exit
5970 status, but doesn't ensure that this exit status is non-zero.
5975 Description: failure of ptlrpc thread startup can cause oops
5976 Details : Starting a ptlrpc service thread can fail if there are a large
5977 number of threads or the server memory is very fragmented.
5978 Handle this without oopsing.
5981 Frequency : always, only if liblustre and non-default acceptor port was used
5983 Description: liblustre cannot connect to servers with non-default acceptor port
5984 Details : tcpnal_set_default_params() was not called and was therefore
5985 ignoring the environment varaible TCPNAL_PORT, as well as other
5986 TCPNAL_ environment variables
5991 Description: two objects could be created on the same OST for a single file
5992 Details : If an OST is down, in some cases it was possible to create two
5993 objects on a single OST for a single file. No problems other
5994 than potential performance impact and spurious error messages.
5998 Bugzilla : 5681, 9562
5999 Description: Client may oops in ll_unhash_aliases
6000 Details : Client dcache may become inconsistent in race condition.
6001 In some cases "getcwd" can fail if the current directory is
6007 Description: Inode refcounting problems in NFS export code
6008 Details : link_raw functions used to call d_instantiate without obtaining
6009 extra inode reference first.
6013 Bugzilla : 9942, 9903
6014 Description: Referencing freed requests leading to crash, memleaks with NFS.
6015 Details : We used to require that call to ll_revalidate_it was always
6016 followed by ll_lookup_it. Also with revalidate_special() it is
6017 possible to call ll_revalidate_it() twice for the same dentry
6018 even if first occurence returned success. This fix changes semantic
6019 between DISP_ENQ_COMPLETE disposition flag to mean there is extra
6020 reference on a request referred from the intent.
6021 ll_intent_release() then releases such a request.
6024 Frequency : rare, normally benchmark loads only
6026 Description: unlinked inodes were kept in memory on the client
6027 Details : If a client is repeatedly creating and unlinking files it
6028 can accumulate a lot of stale inodes in the inode slab cache.
6029 If there is no other client load running this can cause the
6030 client node to run out of memory. Instead flush old inodes
6031 from client cache that have the same inode number as a new inode.
6034 Frequency : SLES9 2.6.5 kernel and long filenames only
6035 Bugzilla : 9969, 10379
6036 Description: utime reports stale NFS file handle
6037 Details : SLES9 uses out-of-dentry names in some cases, which confused
6038 the lustre dentry revalidation. Change it to always use the
6042 Frequency : rare, unless heavy write-truncate concurrency is continuous
6043 Bugzilla : 4180, 6984, 7171, 9963, 9331
6044 Description: OST becomes very slow and/or deadlocked during object unlink
6045 Details : filter_destroy() was holding onto the parent directory lock
6046 while truncating+unlinking objects. For very large objects this
6047 may block other threads for a long time and slow overall OST
6048 responsiveness. It may also be possible to get a lock ordering
6049 deadlock in this case, or run out of journal credits because of
6050 the combined truncate+unlink. Solution is to do object truncate
6051 first in one transaction without parent lock, and then do the
6052 final unlink in a new transaction with the parent lock. This
6053 reduces the lock hold time dramatically.
6056 Frequency : rare, 2.4 kernels only
6058 Description: MDS or OST cleanup may trip kernel BUG when dropping kernel lock
6059 Details : mds_cleanup() and filter_cleanup() need to drop the kernel lock
6060 before unmounting their filesystem in order to avoid deadlock.
6061 The kernel_locked() function in 2.4 kernels only checks whether
6062 the kernel lock is held, not whether it is this process that is
6063 holding it as 2.6 kernels do.
6068 Description: MDS or OST may oops/LBUG if a client is connecting multiple times
6069 Details : The client ptlrpc code may be trying to reconnect to a down
6070 server before a previous connection attempt has timed out.
6071 Increase the reconnect interval to be longer than the connection
6072 timeout interval to avoid sending duplicate connections to
6076 Frequency : echo_client brw_test command
6078 Description: fix echo_client to work with OST preallocated code
6079 Details : OST preallocation code (5137) didn't take echo_client IO path
6080 into account: echo_client calls filter methods outside of any
6081 OST thread and, hence, there is no per-thread preallocated
6082 pages and buffers to use. Solution: hijack pga pages for IO. As
6083 a byproduct, this avoids unnecessary data copying.
6087 Bugzilla : 3555, 5962, 6025, 6155, 6296, 9574
6088 Description: Client can oops in mdc_commit_close() after open replay
6089 Details : It was possible for the MDS to return an open request with no
6090 transaction number in mds_finish_transno() if the client was
6091 evicted, but without actually returning an error. Clients
6092 would later try to replay that open and may trip an assertion
6093 Simplify the client close codepath, and always return an error
6094 from the MDS in case the open is not successful.
6097 Frequency : rare, 2.6 OSTs only
6099 Description: OST may deadlock under high load on fragmented files
6100 Details : If there was a heavy load and highly-fragmented OST filesystems
6101 it was possible to have all the OST threads deadlock waiting on
6102 allocation of biovecs, because the biovecs were not released
6103 until the entire RPC IO was completed. Instead, release biovecs
6104 as soon as they are complete to ensure forward IO progress.
6106 Severity : enhancement
6108 Description: Support for specifying external journal device at mount
6109 Details : If an OST or MDS device is formatted with an external journal
6110 device, this device major/minor is stored in the ext3 superblock
6111 and may not be valid for failover. Allow detecting and
6112 specifying the external journal at mount time.
6117 Description: Mounting an MDS with pending unlinked files may cause oops
6118 Details : target_finish_recovery() calls mds_postrecov() which returned
6119 the number of orphans unlinked. mds_lov_connect->mds_postsetup()
6120 considers this an error and immediately begins cleaning up the
6121 lov, just after starting the mds_lov process
6123 Severity : enhancement
6125 Description: Implement 'lfs df' to report actual free space on per-OST basis
6126 Details : Add sub-command 'df' on 'lfs' to report the disk space usage of
6127 MDS/OSDs. Usage: lfs df [-i][-h]. Command Options: '-i' to report
6128 usage of objects; '-h' to report in human readable format.
6130 ------------------------------------------------------------------------------
6132 2005-08-26 Cluster File Systems, Inc. <info@clusterfs.com>
6139 Description: Mounting an ldiskfs file system with mballoc may crash OST node.
6140 Details : ldiskfs mballoc code may reference an uninitialized buddy struct
6141 at startup during orphan unlinking. Instead, skip buddy update
6142 before setup, as it will be regenerated after recovery is complete.
6147 Description: If an OST is inactive, its locks might reference stale inodes.
6148 Details : lov_change_cbdata() must iterate over all namespaces, even if
6149 they are inactive to clear inode references from the lock.
6151 Severity : enhancement
6152 Frequency : occasional, if non-standard max_dirty_mb used
6154 Description: Client will block write RPCs if not enough grant
6155 Details : If a client has max_dirty_mb smaller than max_rpcs_in_flight,
6156 then the client will block writes while waiting for another RPC
6157 to complete instead of consuming its dirty limit. With change
6158 we get improved performance when max_dirty_mb is small.
6160 Severity : enhancement
6161 Bugzilla : 3389, 6253
6162 Description: Add support for supplementary groups on the MDS.
6163 Details : The MDS has an upcall /proc/fs/lustre/mds/{mds}/group_upcall
6164 (set to /usr/sbin/l_getgroups if enabled) which will do MDS-side
6165 lookups for user supplementary groups into a cache.
6169 Description: O_CREAT|O_EXCL open flags in liblustre always return -EEXIST
6170 Details : Make libsysio to not enforce O_EXCL by clearing the flag,
6171 for liblustre O_EXCL is enforced by MDS.
6175 Description: readdir never returns NULL in liblustre.
6176 Details : Corrected llu_iop_getdirentries logic, to return offset of next
6177 dentry in struct dirent.
6181 Frequency : liblustre only, depends on application IO pattern
6182 Description: liblustre clients evicted if not contacting servers
6183 Details : Don't put liblustre clients into the ping_evictor list, so
6184 they will not be evicted by the pinger ever.
6186 Severity : enhancement
6188 Description: Add ability to evict clients by NID from MDS.
6189 Details : By echoing "nid:$NID" string into
6190 /proc/fs/lustre/mds/.../evict_client client with nid that equals to
6191 $NID would be instantly evicted from this MDS and from all active
6192 OSTs connected to it.
6196 Description: Do not query file size twice, somewhat slowing stat(2) calls.
6197 Details : lookup_it_finish() used to query file size from OSTs that was not
6202 Description: service threads change working directory to that of init
6203 Details : Starting lustre service threads may pin the working directory
6204 of the parent thread, making that filesystem busy. Threads
6205 now change to the working directory of init to avoid this.
6209 Frequency : during shutdown only
6210 Description: shutdown with a failed MDS or OST can cause unmount to hang
6211 Details : Don't resend DISCONNECT messages in ptlrpc_disconnect_import()
6216 Frequency : 2.6 only
6217 Description: chmod/chown may include an extra supplementary group
6218 Details : ll{,u}_mdc_pack_op_data() does not properly initialize the
6219 supplementary group and if none is specified this is used.
6222 Bugzilla : 5479 (6816)
6224 Description: Racing open + rm can assert client in mdc_set_open_replay_data()
6225 Details : If lookup is in progress on a file that is unlinked we might try
6226 to revalidate the inode and fail in revalidate after lookup is
6227 complete and ll_file_open() enqueues the open again but
6228 it_open_error() was not checking DISP_OPEN_OPEN errors correctly.
6231 Frequency : always, if lconf --abort_recovery used
6233 Description: lconf --abort_recovery fails with 'Operation not supported'
6234 Details : lconf was attempting to abort recovery on the MDT device and not
6237 ------------------------------------------------------------------------------
6239 2005-08-08 Cluster File Systems, Inc. <info@clusterfs.com>
6244 Frequency : rare (only unsupported configurations with a node running as an
6246 Bugzilla : 6514, 5137
6247 Description: Mounting a Lustre file system on a node running as an OST could
6249 Details : OSTs now preallocates memory needed to write out data at
6250 startup, instead of when needed, to avoid having to
6251 allocate memory in possibly low memory situations.
6252 Specifically, if the file system is mounted on on OST,
6253 memory pressure could force it to try to write out data,
6254 which it needed to allocate memory to do. Due to the low
6255 memory, it would be unable to do so and the node would
6256 become unresponsive.
6258 Severity : enhancement
6260 Description: Addition of lconf --service command line option
6261 Details : lconf now accepts a '--service <arg>' option, which is
6262 shorthand for 'lconf --group <arg> --select <arg>=<hostname>'
6264 Severity : enhancement
6266 Description: Failover mode is now the default for OSTs.
6267 Details : By default, OSTs will now run in failover mode. To return to
6268 the old behaviour, add '--failout' to the lmc line for OSTs.
6270 Severity : enhancement
6272 Description: Health checks are now provided for MDS and OSTs
6273 Details : Additional detailed health check information on MSD and OSTs
6274 is now provided through the procfs health_check value.
6277 Frequency : occasional, depends on IO load
6279 Description: Disk fragmentation on the OSTs could eventually cause slowdowns
6280 after numerous create/delete cycles
6281 Details : The ext3 inode allocation policy would not allocate new inodes
6282 very well on the OSTs because there are no new directories
6283 being created. Instead we look for groups with free space if
6284 the parent directories are nearly full.
6289 Description: Network or server problems during mount may cause partially
6290 mounted clients instead of returning an error.
6291 Details : The config llog parsing code may overwrite the error return
6292 code during mount error handling, returning success instead
6298 Description: MDS can fail to allocate large reply buffers
6299 Details : After long uptimes the MDS can fail to allocate large reply
6300 buffers (e.g. zconf client mount config records) due to memory
6301 fragmentation or consumption by the buffer cache. Preallocate
6302 some large reply buffers so that these replies can be sent even
6303 under memory pressure.
6307 Frequency : rare (liblustre)
6308 Description: fsx running with liblustre complained that using truncate() to
6309 extend the file doesn't work. This patch corrects that issue.
6310 Details : This is the liblustre equivalent of the fix for bug 6196. Fixes
6311 ATTR_SIZE and lsm use in llu_setattr_raw.
6315 Frequency : rare, only 2.6 kernels
6316 Description: Unusual file access patterns on the MDS may result in inode
6317 data being lost in very rare circumstances.
6318 Details : Bad interaction between the ea-in-inode patch and the "no-read"
6319 code in the 2.6 kernel caused the inode and/or EA data not to
6320 be read from disk, causing single-file corruption.
6324 Frequency : rare, only 2.6 filesystems using extents
6325 Description: Heavy concurrent write and delete load may cause data corruption.
6326 Details : It was possible under high-load situations to have an extent
6327 metadata block in the block device cache from a just-unlinked
6328 file overwrite a newly-allocated data block. We now unmap any
6329 metadata buffers that alias just-allocated data blocks.
6333 Frequency : filesystems with default stripe_count larger than 77
6334 Description: lconf+mke2fs fail when formatting filesystem with > 77 stripes
6335 Details : lconf specifies an inode size of 4096 bytes when the default
6336 stripe_count is larger than 77. This conflicts with the default
6337 inode density of 1 per 4096 bytes. Allocate smaller inodes in
6338 this case to avoid pinning too much memory for large EAs.
6340 ------------------------------------------------------------------------------
6342 2005-07-07 Cluster File Systems, Inc. <info@clusterfs.com>
6347 Frequency : rare (extremely heavy IO load with hundreds of clients)
6349 Description: Client is evicted, gets IO error writing to file
6350 Details : lock ordering changes for bug 5492 reintroduced bug 3267 and
6351 caused clients to be evicted for AST timeouts. The fixes in
6352 bug 5192 mean we no longer need to have such short AST timeouts
6353 so ldlm_timeout has been increased.
6356 Frequency : occasional during --force or --failover shutdown under load
6357 Bugzilla : 5949, 4834
6358 Description: Server oops/LBUG if stopped with --force or --failover under load
6359 Details : a collection of import/export refcount and cleanup ordering
6360 issues fixed for safer force cleanup
6363 Frequency : only filesystems larger than 120 OSTs
6364 Bugzilla : 5990, 6223
6365 Description: lfs getstripe would oops on a very large filesystem
6366 Details : lov_getconfig used kfree on vmalloc'd memory
6369 Frequency : only filesystems exporting via NFS to Solaris 10 clients
6370 Bugzilla : 6242, 6243
6371 Description: reading from files that had been truncated to a non-zero size
6372 but never opened returned no data
6373 Details : ll_file_read() reads zeros from no-object files to EOF
6378 Description: A bug in MDS/OSS recovery could cause the OSS to fail an assertion
6379 Details : There's little harm in aborting MDS/OSS recovery and letting it
6380 try again, so I removed the LASSERT and return an error instead.
6382 Severity : enhancement
6384 Description: New debugging infrastructure for tracking down data corruption
6385 Details : The I/O checksum code was replaced to: (a) control it at runtime,
6386 (b) cover more of the client-side code path, and (c) try to narrow
6387 down where problems occurred
6391 Bugzilla : 3819, 4364, 4397, 6313
6392 Description: Racing close and eviction MDS could cause assertion in mds_close
6393 Details : It was possible to get multiple mfd references during close and
6394 client eviction, leading to one thread referencing a freed mfd.
6396 Severity: : enhancement
6397 Bugzilla : 3262, 6359
6398 Description: Attempts to reconnect to servers are now more aggressive.
6399 Details : This builds on the enhanced upcall-less recovery that was added
6400 in 1.4.2. When trying to reconnect to servers, clients will
6401 now try each server in the failover group every 10 seconds. By
6402 default, clients would previously try one server every 25 seconds.
6407 Description: After recovery, certain operations trigger a failed
6408 assertion on a client.
6409 Details : Failing over an mds, using lconf -d --failover, while a
6410 client was doing a readdir() call would cause the client to
6411 LBUG after recovery completed and the readdir() was resent.
6413 Severity : enhancement
6415 Description: Default groups are now added by lconf
6416 Details : You can now run lconf --group <servicename> without having to
6417 manually add groups with lmc.
6420 Frequency : occasional
6422 Description: Nodes with an elan id of 0 trigger a failed assertion
6425 Frequency : always when accessing e.g. tty/console device nodes
6427 Description: tty and some other devices nodes cannot be used on lustre
6428 Details : file's private_data field is used by device data and lustre
6429 values in there got lost. New field was added to struct file to
6430 store fs-specific private data.
6433 Frequency : when exporting Lustre via NFS
6435 Description: NFSD failed occasionally when looking up a path component
6436 Details : NFSD is looking up ".." which was broken in ext3 directories
6437 that had grown large enough to become hashed.
6440 Frequency : Clusters with multiple interfaces not on the same subnet
6442 Description: Nodes will repeatedly try to reconnect to an interface which it
6443 cannot reach and report an error to the log.
6444 Details : Extra peer list entries will be created by lconf with some peers
6445 unreachable. lconf now validates the peer before adding it.
6448 Frequency : Only if a default stripe is set on the filesystem root.
6450 Description: Setting a default stripe on the filesystem root prevented the
6451 filesystem from being remounted.
6452 Details : The client was sending extra request flags in the root getattr
6453 request and did not allocate a reply buffer for the dir EA.
6456 Frequency : occasional, higher if lots of files are accessed by one client
6457 Bugzilla : 6159, 6097
6458 Description: Client trips assertion regarding lsm mismatch/magic
6459 Details : While revalidating inodes the VFS looks up inodes with ifind()
6460 and in rare cases can find an inode that is being freed.
6461 The ll_test_inode() code will free the lsm during ifind()
6462 when it finds an existing inode and then the VFS later attaches
6463 this free lsm to a new inode.
6467 Bugzilla : 6422, 7030
6468 Description: MDS deadlock between mkdir and client eviction
6469 Details : Creating a new file via mkdir or mknod (starting a transaction
6470 and getting the ns lock) can deadlock with client eviction
6471 (gets ns lock and trying to finish a synchronous transaction).
6474 Frequency : occasional
6475 Description: While starting a server, the fsfilt_ext3 module could not be
6477 Details : CFS's improved ext3 filesystem is named ldiskfs for 2.6
6478 kernels. Previously, lconf would still use the ext3 name
6479 when trying to load modules. Now, it will correctly use
6480 ext3 on 2.4 and ldiskfs on 2.6.
6482 Severity : enhancement
6483 Description: The default stripe count has been changed to 1
6484 Details : The interpretation of the default stripe count (0, to lfs
6485 or lmc) has been changed to mean striping across a single
6486 OST, rather than all available. For general usage we have
6487 found a stripe count of 1 or 2 works best.
6489 Severity : enhancement
6490 Description: Add support for compiling against Cray portals.
6491 Details : Conditional compiling for some areas that are different
6495 Frequency : occasional
6496 Bugzilla : 6409, 6834
6497 Description: Creating files with an explicit stripe count may lead to
6498 a failed assertion on the MDS
6499 Details : If some OSTs are full or unavailable, creating files may
6500 trigger a failed assertion on the MDS. Now, Lustre will
6501 try to use other servers or return an error to the
6505 Frequency : occasional
6507 Description: Multiple concurrent overlapping read+write on multiple SMP nodes
6508 caused lock timeout during readahead (since 1.4.2).
6509 Details : Processes doing readahead might match a lock that hasn't been
6510 granted yet if there are overlapping and conflicting lock
6511 requests. The readahead process waits on ungranted lock
6512 (original lock is CBPENDING), while OST waits for that process
6513 to cancel CBPENDING read lock and eventually evicts client.
6515 Severity : enhancement
6517 Description: Initial enabling of flock support for clients
6518 Details : Implements fcntl advisory locking and file status functions.
6519 This feature is provided as an optional mount flag (default
6520 off), and is NOT CURRENTLY SUPPORTED. Not all types of record
6521 locking are implemented yet, and those that are are not guaranteed
6522 to be completely correct in production environments.
6523 mount -t lustre -o [flock|noflock] ...
6526 Frequency : occasional
6528 Description: OSTs running 2.4 kernels but with extents enabled might trip an
6529 assertion in the ext3 JBD (journaling) layer.
6530 Details : The b_committed_data struct is protected by the big kernel lock
6531 in 2.4 kernels, serializing journal_commit_transaction() and
6532 ext3_get_block_handle->ext3_new_block->find_next_usable_block()
6533 access to this struct. In 2.6 kernels there is finer grained
6534 locking to improve SMP performance of the JBD layer.
6538 Description: Changes the "SCSI I/O Stats" kernel patch to default to "enabled"
6540 -----------------------------------------------------------------------------
6542 2005-05-05 Cluster File Systems, Inc. <info@clusterfs.com>
6544 NOTE: Lustre 1.4.2 uses an incompatible network protocol than previous
6545 versions of Lustre. Please update all servers and clients to
6546 version 1.4.2 or later at the same time. You must also run
6547 "lconf --write-conf {config}.xml" on the MDS while it is stopped
6548 to update the configuration logs.
6550 - fix for HPUX NFS client breakage when NFS exporting Lustre (5781)
6551 - mdc_enqueue does not need max_mds_easize request buffer on send (5707)
6552 - swab llog records of type '0' so we get proper header size/idx (5861)
6553 - send llog cancel req to DLM cancel portal instead of cb portal (5515)
6554 - fix rename of one directory over another leaking an inode (5953)
6555 - avoid SetPageDirty on 2.6 (5981)
6556 - don't re-add just-being-destroyed locks to the waiting list (5653)
6557 - when creating new directories, inherit the parent's custom
6558 striping settings if present parent (3048)
6559 - flush buffers from cache before direct IO in 2.6 obdfilter (4982)
6560 - don't hold i_size_sem in ll_nopage() and ll_ap_refresh_count (6077)
6561 - don't hold client locks on temporary worklist from l_lru (5666)
6562 - handle IO errors in 2.6 obdfilter bio completion routine (6046)
6563 - automatically evict dead clients (5921)
6564 - Update file size properly in create+truncate+fstat case (6196)
6565 - Do not unhash mountpoint dentries, do not allow removal of
6567 - Avoid lock ordering deadlock issue with write/truncate (6203,5654)
6568 - reserve enough journal credits in fsfilt_start_log for setattr (4554)
6569 - ldlm_enqueue freed-export error path would always LBUG (6149,6184)
6570 - don't reference lr_lvb_data until after we hold lr_lvb_sem (6170)
6571 - don't overwrite last_rcvd if there is a *_client_add() error (6086)
6572 - Correctly handle reads of files with no objects (6243)
6573 - lctl recover will also mark a device active if deactivate used (5933)
6575 - by default create 1 inode per 4kB space on MDS, per 16kB on OSTs
6576 - allow --write-conf on an MDS with different nettype than client (5619)
6577 - don't write config llogs to MDS for mounts not from that MDS (5617)
6578 - lconf should create multiple TCP connections from a client (5201)
6579 - init scripts are now turned off by default; run chkconfig --on
6580 lustre and chkconfig --on lustrefs to use them
6581 - upcalls are no longer needed for clients to recover to failover
6583 - add --abort-recovery option to lconf to abort recovery on device
6585 - add support for an arbitrary number of OSTs (3026)
6586 - Quota support protocol changes.
6587 - forward compatibility changes to wire structs (6007)
6588 - rmmod NALs that might be loaded because of /etc/modules.conf (6133)
6589 - support for mountfsoptions and clientoptions to the Lustre LDAP (5873)
6590 - improved "lustre status" script
6591 - initialize blocksize for non-regular files (6062)
6592 - added --disable-server and --disable-client configure options (5782)
6593 - introduce a lookup cache for lconf to avoid repeated DB scans (6204)
6594 - Vanilla 2.4.29 support
6595 - increase maximum number of obd devices to 520 (6242)
6596 - remove the tcp-zero-copy patch from the suse-2.4 series (5902)
6597 - Quadrics Elan drivers are now included for the RHEL 3 2.4.21 and
6598 SLES 9 2.6.5 kernels
6599 - limit stripes per file to 160 (the maximum EA size) (6093)
6601 2005-03-22 Cluster File Systems, Inc. <info@clusterfs.com>
6604 - don't LASSERT in ll_release on NULL lld with NFS export (4655, 5760)
6605 - hold NS lock when calling handle_ast_error->del_waiting_lock (5746)
6606 - fix setattr mtime regression from lovcleanup merge (4829, 5669)
6607 - workaround for 2.6 crash in ll_unhash_aliases (5687, 5210)
6608 - small ext3 extents cleanups and fixes (5733)
6609 - improved mballoc code, several small races and bugs fixed (5733, 5638)
6610 - kernel version 43 - fix remove_suid bugs in both 2.4 and 2.6 (5695)
6611 - avoid needless client->OST connect, fix handle mismatch (5317)
6612 - fix DLM error path that led to out-of-sync client, long delays (5779)
6613 - support common vfs-enforced mount options (nodev,nosuid,noexec) (5637)
6614 - fix several locking issues related to i_size (5492,5624,5654,5672)
6615 - don't move pending lock onto export if it is already evicted (5683)
6616 - fix kernel oops when creating .foo in unlinked directory (5548)
6617 - fix deadlock in obdfilter statistics vs. object create (5811)
6618 - use time_{before,after} to avoid timer jiffies wrap (5882)
6619 - shutdown --force/--failover stability (3607,3651,4797,5203,4834)
6620 - Do not leak request if server was not able to process it (5154)
6621 - If mds_open unable to find parent dir, make that negative lookup(5154)
6622 - don't create new directories with extent-mapping (5909, 5936)
6624 - fix lustre/lustrefs init scripts for SuSE (patch from Scali, 5702)
6625 - don't hold the pinger_sem in ptlrpc_pinger_sending_on_import
6626 - change obd_increase_kms to obd_adjust_kms (up or down) (5654)
6627 - lconf, lmc search both /usr/lib and /usr/lib64 for Python libs (5800)
6628 - support for RHEL4 kernel on i686 (5773)
6629 - provide error messages when incompatible logs are encountered (5898)
6631 2005-02-18 Cluster File Systems, Inc. <info@clusterfs.com>
6632 * version 1.4.0.10 (1.4.1 release candidate 1)
6634 - don't keep a lock reference when lock is not granted (4238)
6635 - unsafe list practices (rarely) led to infinite eviction loop (4908)
6636 - add per-fs limit of Lustre pages in page cache, avoid OOM (4699)
6637 - drop import inflight refcount on signal_completed_replay error (5255)
6638 - unlock page after async write error during send (3677)
6639 - handle missing objects in filter_preprw_read properly (5265)
6640 - no transno return for symlink open, don't save no-trasno open (3440)
6641 - don't try to complete elan receive that already failed (4012)
6642 - free RPC server reply state on error (5406)
6643 - clean up thread from ptlrpc_start_thread() on error (5160)
6644 - readahead could read extra page into cache that wasn't ejected (5388)
6645 - prevent races in class_attach/setup/cleanup/detach (5260)
6646 - don't dereference de->d_inode after l_dput of de (5458)
6647 - use "int" for stripe value returned from lock_to_stripe (5544)
6648 - mballoc allocation and error-checking fixes in 2.6 (5504)
6649 - block device patches to fix I/O request sizes in 2.6 (5482)
6650 - look up hostnames for IB nals (5602)
6651 - 2.6 changed lock ordering of 2 semaphores, caused deadlock (5654)
6652 - don't start multiple acceptors for the same port (5277)
6653 - fix incorrect LASSERT in mds_getattr_name (5635)
6654 - export a proc file for general "ping" checking (5628)
6655 - fix "lfs check" to not block when the MDS is down (5628)
6657 - service request history (4965)
6658 - put {ll,lov,osc}_async_page structs in a single slab (4699)
6659 - create an "evict_client" /proc entry on OSTs, like the MDS has
6660 - fix mount usage message, return errors per mount(8) (5168)
6661 - change grep [] to grep "[]" in tests so they work in more UMLs
6662 - fix ppc64/x86_64 spec to use %{_libdir} instead of /usr/lib (5389)
6663 - remove ancient LOV_MAGIC_V0 EA support (5047)
6664 - add "disk I/Os in flight" and "I/O req time" stats in obdfilter
6665 - align r/w RPCs to PTLRPC_MAX_BRW_SIZE boundary for performance (3451)
6666 - allow readahead allocations to fail when low on memory (5383)
6667 - mmap locking landed again, after considerable improvement (2828)
6668 - add get_hostaddr() to lustreDB.py for LDAP support (5459)
6670 2004-11-23 Cluster File Systems, Inc. <info@clusterfs.com>
6673 - send OST transaction number in read/write reply to free req (4966)
6674 - don't ASSERT in ptl_send_rpc() if we run out of memory (5119)
6675 - lock /proc/sys/portals/routes internal state, avoiding oops (4827)
6676 - the watchdog thread now runs as interruptible (5246)
6677 - flock/lockf fixes (but it's still disabled, pending 5135)
6678 - don't use EXT3 constants in llite code (5094)
6679 - memory shortage at startup could cause assertion (5176)
6681 - reorganization of lov code
6682 - single portals codebase
6684 - add extents/mballoc support (5025)
6685 - direct I/O reads in the obdfilter (4048)
6686 - kernel patches from LNXI for 2.6 (bluesmoke, perfctr, mtd, kexec)
6688 tbd Cluster File Systems, Inc. <info@clusterfs.com>
6691 - send OST transaction number in read/write reply to free req (4966)
6692 - don't ASSERT in ptl_send_rpc() if we run out of memory (5119)
6693 - lock /proc/sys/portals/routes internal state, avoiding oops (4827)
6694 - the watchdog thread now runs as interruptible (5246)
6695 - handle missing objects in filter_preprw_read properly (5265)
6696 - unsafe list practices (rarely) led to infinite eviction loop (4908)
6697 - drop import inflight refcount on signal_completed_replay error (5255)
6698 - unlock page after async write error during send (3677)
6699 - return original error code on reconstructed replies (3761)
6700 - no transno return for symlink open, don't save no-trasno open (3440)
6702 - add pid to ldlm debugging output (4922)
6703 - bump the watchdog timeouts -- we can't handle 30sec yet
6704 - extra debugging for orphan dentry/inode bug (5259)
6706 2004-11-16 Cluster File Systems, Inc. <info@clusterfs.com>
6709 - fix TCP_NODELAY bug, which caused extreme perf regression (5134)
6710 - allocate qswnal tx descriptors singly to avoid fragmentation (4504)
6711 - don't LBUG on obdo_alloc() failure, use OBD_SLAB_ALLOC() (4800)
6712 - fix NULL dereference in /proc/sys/portals/routes (4827)
6713 - allow failed mdc_close() operations to be interrupted (4561)
6714 - stop precreate on OST before MDS would time out on it (4778)
6715 - don't send partial-page writes before EOF from client (4410)
6716 - discard client grant for sub-page writes on large-page clients (4520)
6717 - don't free dentries not owned by NFS code, check generation (4806)
6718 - fix lsm leak if mds_create_objects() fails (4801)
6719 - limit debug_daemon file size, always print CERROR messages (4789)
6720 - use transno after validating reply (3892)
6721 - process timed out requests if import state changes (3754)
6722 - update mtime on OST during writes, return in glimpse (4829)
6723 - add mkfsoptions to LDAP (4679)
6724 - use ->max_readahead method instead of zapping global ra (5039)
6725 - don't interrupt __l_wait_event() during strace
6727 - add software watchdogs to catch hung threads quickly (4941)
6728 - make lustrefs init script start after nfs is mounted
6729 - fix CWARN/ERROR duplication (4930)
6730 - return async write errors to application if possible (2248)
6731 - add /proc/sys/portal/memused (bytes allocated by PORTALS_ALLOC)
6732 - print NAL number in %x format (4645)
6733 - update barely-supported suse-2.4.21-171 series (4842)
6734 - support for sles 9 %post scripts
6735 - support for building 2.6 kernel-source packages
6736 - support for sles km_* packages
6738 2004-10-07 Cluster File Systems, Inc. <info@clusterfs.com>
6741 - ignore -ENOENT errors in osc_destroy (3639)
6742 - notify osc create thread that OSC is being cleaned up (4600)
6743 - add nettype argument for llmount in #5d in conf-sanity.sh (3936)
6744 - reconstruct ost_handle() like mds_handle() (4657)
6745 - create a new thread to do import eviction to avoid deadlock (3969)
6746 - let lconf resolve symlinked-to devices (4629)
6747 - don't unlink "objects" from directory with default EA (4554)
6748 - hold socknal file ref over connect in case target is down (4394)
6749 - allow more than 32000 subdirectories in a single directory (3244)
6750 - fix blocks count for O_DIRECT writes (3751)
6751 - OST returns ENOSPC from object create when no space left (4539)
6752 - don't send truncate RPC if file size isn't changing (4410)
6753 - limit OSC precreate to 1/2 of value OST considers bogus (4778)
6754 - bind to privileged port in socknal and tcpnal (3689)
6756 - rate limit CERROR/CWARN console message to avoid overload (4519)
6757 - GETFILEINFO dir ioctl returns LOV EA + MDS stat in 1 call (3327)
6758 - basic mmap support (3918)
6759 - kernel patch series update from b1_4 (4711)
6761 2004-09-16 Cluster File Systems, Inc. <info@clusterfs.com>
6764 - avoid crash during MDS cleanup with OST shut down (2775)
6765 - fix loi_list_lock/oig_lock inversion on interrupted IO (4136)
6766 - don't use bad inodes on the MDS (3744)
6767 - dynamic object preallocation to improve recovery speed (4236)
6768 - don't hold spinlock over lock dumping or change debug flags (4401)
6769 - don't zero obd_dev when it is force cleaned (3651)
6770 - print grants to console if they go negative (4431)
6771 - "lctl deactivate" will stop automatic recovery attempts (3406)
6772 - look for existing locks in ldlm_handle_enqueue() (3764)
6773 - don't resolve lock handle twice in recovery avoiding race (4401)
6774 - revalidate should check working dir is a directory (4134)
6776 - don't always mark "slow" obdfilter messages as errors (4418)
6778 2004-08-24 Cluster File Systems, Inc. <info@clusterfs.com>
6781 - don't close LustreDB during write_conf until it is done (3860)
6782 - fix typo in lconf for_each_profile (3821)
6783 - allow dumping logs from multiple threads at one time (3820)
6784 - don't allow multiple threads in OSC recovery (3812)
6785 - fix debug_size parameters (3864)
6786 - fix mds_postrecov to initialize import for llog ctxt (3121)
6787 - replace config semaphore with spinlock (3306)
6788 - be sure to send a reply for a CANCEL rpc with bad export (3863)
6789 - don't allow enqueue to complete on a destroyed export (3822)
6790 - down write_lock before checking llog header bitmap (3825)
6791 - recover from lock replay timeout (3764)
6792 - up llog sem before sending rpc (3652)
6793 - reduce ns lock hold times when setting kms (3267)
6794 - change a dlm LBUG to LASSERTF, to maybe learn something (4228)
6795 - fix NULL deref and obd_dev leak on setup error (3312)
6796 - replace some LBUG about llog ops with error handling (3841)
6797 - don't match INVALID dentries from d_lookup and spin (3784)
6798 - hold dcache_lock while marking dentries INVALID and hashing (4255)
6799 - fix invalid assertion in ptlrpc_set_wait (3880)
6801 - add libwrap support for the TCP acceptor (3996)
6802 - add /proc/sys/portals/routes for non-root route listing (3994)
6803 - allow setting MDS UUID in .xml (2580)
6804 - print the stack of a process that LBUGs (4228)
6806 2004-07-14 Cluster File Systems, Inc. <info@clusterfs.com>
6809 - don't cleanup request in ll_file_open() on failed MDS open (3430)
6810 - make sure to unset replay flag from failed open requests (3440)
6811 - if default stripe count is 0, use OST count for inode size (3636)
6812 - update parent mtime/ctime on client for create/unlink (2611)
6813 - drop dentry ref in ext3_add_link from open_connect_dentry (3266)
6814 - free recovery state on server during a forced cleanup (3571)
6815 - unregister_reply for resent reqs (3063)
6816 - loop back devices mounting and status check on 2.6 (3563)
6817 - fix resource-creation race that can provoke i_size == 0 (3513)
6818 - don't try to use bad inodes returned from MDS/OST fs lookup (3688)
6819 - more debugging for page-accounting assertion (3746)
6820 - return -ENOENT instead of asserting if ost getattr+unlink race (3558)
6821 - avoid deadlock after precreation failure (3758)
6822 - fix race and lock order deadlock in orphan handling (3450, 3750)
6823 - add validity checks when grabbing inodes from l_ast_data (3599)
6825 - add /proc/.../recovery_status to obdfilter (3428)
6826 - lightweight CDEBUG infrastructure, debug daemon (3668)
6827 - change default OSC RPC parameters to be better on small clusters
6828 - turn off OST read cache for files smaller than 32MB
6829 - install man pages and include them in rpms (3100)
6830 - add new init script for (un)mounting lustre filesystems (2593)
6831 - run chkconfig in %post for init scripts (3701)
6832 - drop scimac NAL (unmaintained)
6834 2004-06-17 Cluster File Systems, Inc. <info@clusterfs.com>
6837 - clean kiobufs before and after use (3485)
6838 - strip trailing '/'s before comparing paths with /proc/mounts (3486)
6839 - remove assertions to work around "in-flight rpcs" recovery bug (3063)
6840 - change init script to fail more clearly if not run as root (1528)
6841 - allow clients to reconnect during replay (1742)
6842 - fix ns_lock/i_sem lock ordering deadlock for kms update (3477)
6843 - don't do DNS lookups on NIDs too small for IP addresses (3442)
6844 - re-awaken ptlrpcd if new requests arrive during check_set (3554)
6845 - fix cond_resched (3554)
6846 - only evict unfinished clients after recovery (3515)
6847 - allow bulk resend, prevent data loss (3570)
6848 - dynamic ptlrpc request buffer allocation (2102)
6849 - don't allow unlinking open directory if it isn't empty (2904)
6850 - set MDS/OST threads to umask 0 to not clobber client modes (3359)
6851 - remove extraneous obd dereference causing LASSERT failure (3334)
6852 - don't use get_cycles() when creating temp. files on the mds (3156)
6853 - hold i_sem when setting i_size in ll_extent_lock() (3564)
6854 - handle EEXIST for set-stripe, set proper directory name (3336)
6856 - servers can dump a log evicting a client - lustre.dump_on_timeout=1
6857 - fix ksocknal_fmb_callback() error messages (2918)
6859 2004-05-27 Cluster File Systems, Inc. <info@clusterfs.com>
6862 - don't copy lvb into (possibly NULL) reply on error (2983)
6863 - don't deref dentry after dput, don't free lvb on error (2922)
6864 - use the kms to determine writeback rpc length (2947)
6865 - increment oti_logcookies when osc is inactive (2948)
6866 - update client's i_blocks count via lvb messages (2543)
6867 - handle intent open/close of special files properly (1557)
6868 - mount MDS with errors=remount-ro, like obdfilter (2009)
6869 - initialize lock handle to avoid ASSERT on error cleanup (3057)
6870 - don't use cancelling-locks' kms values (2947)
6871 - use highest lock extent for kms, not last one (2925)
6872 - don't dereference ERR_PTR() dentry in error handling path (3107)
6873 - fix thread race in portals_debug_dumplog() (3122)
6874 - create lprocfs device entries at setup instead of at attach (1519)
6875 - common AST error handler, don't evict client on completion race (3145)
6876 - zero nameidata in detach_mnt in 2.6 (3118)
6877 - verify d_inode after revalidate_special is valid in 2.6 (3116)
6878 - use lustre_put_super() to handle zconf unmounts in 2.6 (3064)
6879 - initialize RPC timeout timer earlier for 2.6 (3219)
6880 - don't dereference NULL reply buffer if mdc_close was never sent (2410)
6881 - print nal/nid for unknown nid (3258)
6882 - additional checks for oscc recovery before doing precreate (3284)
6883 - fix ll_extent_lock() error return code for 64-bit systems (3043)
6884 - don't crash in mdc_close for bad permissions on open (3285)
6885 - zero i_rdev for non-device files (3147)
6886 - clear page->private before handing to FS, better assertion (3119)
6887 - tune the read pipeline (3236)
6888 - fix incorrect decref of invalidated dentry (2350)
6889 - provide read-ahead stats and refine rpc in flight stats (3328)
6890 - don't hold journal transaction open across create RPC (3313)
6891 - update atime on MDS at close time (3265)
6892 - close LDAP connection when recovering to avoid server load (3315)
6893 - update iopen-2.6 patch with fixes from 2399,2517,2904 (3301)
6894 - don't leak open file on MDS after open resend (3325)
6895 - serialize filter_precreate and filter_destroy_precreated (3329)
6896 - loop device shouldn't call sync_dev() for nul device (3092)
6897 - clear page cache after eviction (2766)
6898 - resynchronize MDS->OST in background (2824)
6899 - refuse to mount the same filesystem twice on same mountpoint (3394)
6900 - allow llmount to create routes for mounting behind routers (3320)
6901 - push lock cancellation to blocking thread for glimpse ASTs (3409)
6902 - don't call osc_set_data_with_check() for TEST_LOCK matches (3159)
6903 - fix rare problem with rename on htree directories (3417)
6905 - allow default OST striping configuration per directory (1414)
6906 - fix compilation for qswnal for 2.6 kernels (3125)
6907 - increase maximum number of MDS request buffers for large systems
6908 - change liblustreapi to be useful for external progs like lfsck (3098)
6909 - increase local configuration timeout for slow disks (3353)
6910 - allow configuring ldlm AST timeout - lustre.ldlm_timeout=<seconds>
6912 2004-03-22 Cluster File Systems, Inc. <info@clusterfs.com>
6915 - fixes for glimpse AST timeouts / incorrectly 0-sized files (2818)
6916 - don't overwrite extent policy data in reply if lock was blocked (2901)
6917 - drop filter export grants atomically with removal from device (2663)
6918 - del obd_self_export from work_list in class_disconnect_exports (2908)
6919 - don't LBUG if MDS recovery times out during orphan cleanup (2530)
6920 - swab reply message in mdc_close, other PPC fixes (2464)
6921 - fix destroying of named logs (2325)
6922 - overwrite old logs when running lconf --write_conf (2264)
6923 - bump LLOG_CHUNKSIZE to 8k to allow for larger clusters (2306)
6924 - fix race in target_handle_connect (2898)
6925 - mds_reint_create() should take same inode create lock (2926)
6926 - correct journal credits calculated for CANCEL_UNLINK_LOG (2931)
6927 - don't close files for self_export to avoid uninitialized obd (2936)
6928 - allow MDS with the same name as client node (2939)
6929 - hold dentry reference for closed log files for unlink (2325)
6930 - reserve space for all logs during transactions (2059)
6931 - don't evict page beyond end of stripe extent (2925)
6932 - don't oops on a deleted current working directory (2399)
6933 - handle hard links to targets without a parent properly (2517)
6934 - don't dereference NULL lock when racing during eviction (2867)
6935 - don't grow lock extents when lots of conflicting locks (2919)
6937 2004-03-04 Cluster File Systems, Inc. <info@clusterfs.com>
6940 - account for cache space usage on clients to avoid data loss (974)
6941 - lfsck support in lustre kernel code (2349)
6942 - reduce journal credits needed for BRW writes (2370)
6943 - orphan handling to avoid losing space on client/server crashes
6944 - ptlrpcd can be blocked, stopping ALL progress (2477)
6945 - use lock value blocks to assist in proper KMS, faster stat (1021)
6946 - takes i_sem instead of DLM locks internally on obdfilter (2720)
6947 - recovery for initial connections (2355)
6948 - fixes for mds_cleanup_orphans (1934)
6949 - abort_recovery crashes MDS in b_eq (mds_unlink_orphan) (2584)
6950 - block all file creations until orphan recovery completes (1901)
6951 - client remove rq_connection from request struct (2423)
6952 - conf-sanity test_5, proper cleanup in umount log not availale (2640)
6953 - recovery timer race (2670)
6954 - mdc_close recovey bug (2532)
6955 - ptlrpc cleanup bug (2710)
6956 - mds timeout on local locks (2588)
6957 - namespace lock held during RPCs (2431)
6958 - handle interrupted sync write properly (2503)
6959 - don't try to handle a message that hasn't been replied to (2699)
6960 - client assert failure during cleanup after abort recovery (2701)
6961 - leak mdc device after failed mount (2712)
6962 - ptlrpc_check_set allows timedout requests to complete (2714)
6963 - wait for inflight reqs when ptlrpcd finishes (2710)
6964 - make sure unregistered services are removed from the srv_list
6965 - reset bulk XID's when resending them (caught by 1138 test)
6966 - unregister_bulk after timeout
6967 - fix lconf error (2694)
6968 - handle write after unfinished setstripe, stripe-only getstripe (2388)
6969 - readahead locks pages, leaves pending causing memory pressure (2673)
6970 - increase OST request buffers to 4096 on large machines (2729)
6971 - fix up permission of existing directories in simple_mkdir (2661)
6972 - init deleted item, add assertions ptlrpc_abort_inflight() (2725)
6973 - don't assign transno to errored transactions (2742)
6974 - don't delete objects on OST if given a bogus objid from MDS (2751)
6975 - handle large client PAGE_SIZE readdir on small PAGE_SIZE MDS (2777)
6976 - if rq_no_resend, then timeout request after recovery (2432)
6977 - fix MDS llog_logid record size, 64-bit array alignment (2733)
6978 - don't call usermode_helper from ptlrpcd, DEFAULT upcall (2773)
6979 - put magic in mount.lustre data, check for bad/NULL mount data (2529)
6980 - MDS recovery shouldn't delete objects that it has given out (2730)
6981 - if enqueue arrives after completion, don't clobber LVB (2819)
6982 - don't unlock pages twice when trigger_group_io returns error (2814)
6983 - don't deref NULL rq_repmsg if ldlm_handle_enqueue failed (2822)
6984 - don't write pages to disk if there was an error (1450)
6985 - don't ping imports that have recovery disabled (2676)
6986 - take buffered bytes into account when balancing socknal conn (2817)
6987 - hold a DLM lock over readdir always, use truncate_inode_pages (2706)
6988 - reconnect unlink llog connection after MDS reconnects to OST (2816)
6989 - remove little-endian swabbing of llog records (1987)
6990 - set/limit i_blksize to LL_MAX_BLKSIZE on client (2884)
6991 - retry reposting request buffers if they fail (1191)
6992 - grow extent at grant time to avoid granting a revoked lock (2809)
6993 - lock revoke doesn't evict page if covered by a second lock (2765)
6994 - disable VM readahead to avoid reading outside lock extents (2805)
6996 - return LL_SUPER_MAGIC from statfs for the filesystem type (1972)
6997 - updated kernel patches for hp-2.4.20 kernel (2681)
6999 2004-02-07 Cluster File Systems, Inc. <info@clusterfs.com>
7002 - fix truncated write corruption (2366)
7003 - fix for failed assertion in iopen_connect_dentry (1792,2517)
7005 - don't flag the ptlrpcd thread with PF_MEMALLOC (2636)
7006 - ensure len(uuid) < 37 in lmc (1171)
7007 - fix ia64 OOPS in llog_test (2255)
7008 - zero end of page at obdfilter for partial page writes (2648)
7009 - don't leave stale dentries around after renames (bug 2428)
7010 - fix timeouts when evicting a client with a single lock held (2642)
7011 - set deadline for the initial HELLO message to drain (2634)
7012 - print out dotted-quad IP addresses in the socknal (2302)
7014 - additional debugging for MDS client eviction problem (2443)
7015 - fix mkfsoptions support for osts (2603, 2604)
7017 2004-01-27 Cluster File Systems, Inc. <info@clusterfs.com>
7020 - add series for the vanilla 2.6.0 kernel
7021 - add series for the vanilla 2.4.24 kernel
7022 - add series for a cray x86/64 UL kernel drop
7023 - fix xattr patches for the vanilla 2.4.19 series
7025 - generate true UUIDs in lmc (1171)
7026 - have portals stack dumping break in UML (2466)
7027 - avoid bad dchild deref; avoid inum lock w/o creation (2362)
7028 - allocate with _NOFS in ldlm to avoid deadlock (1933)
7029 - wake callback waiting threads on client eviction (2460)
7030 - Add --ptldebug and --subsystem to lmc (1719)
7031 - update assertion to allow safe interrupt allocation
7032 - set rq_no_resend for cancel requests (2432)
7033 - recalculate ptlrpcd timeout after resend (2494)
7034 - call vfs_rmdir when removing pending directories (2368)
7035 - fix renaming a file to itself (2429)
7036 - lmc creates a default one-stripe lov (2454)
7037 - expand procfs space to handle large clusters (2326)
7038 - increase UML stack to avoid overflow
7039 - update lconf's list of debug and subsystem masks
7040 - fix lfs find --obd (2510)
7041 - /proc tunable for disabling filter read caching (2591)
7042 - stop rpm packages from altering slapd.conf (2301)
7043 - disable nagle in the socknal under 0conf (2578)
7044 - choose mds inode size based on stripe count (2572)
7045 - fix kernel-source rpm problems (2516)
7047 - add --disable-doc to avoid pdf generation (2421)
7048 - update documentation, tests, type-os, comments
7049 - avoid format warnings on ia64
7050 - remove the TOE NAL
7051 - tiny code cleanups by removing unused fields
7053 2004-01-07 Cluster File Systems, Inc. <info@clusterfs.com>
7056 - fix obvious semaphore misuse in as-yet-unused setattr path (2348)
7057 - remove the most blatant lies from BUILDING file (2371)
7058 - change default debug level to reasonable production setting
7059 - reduce client side cache size to reduce cache flush time
7060 - reduce max RPCs in flight to avoid unnecessary file fragmentation
7061 - make TCP zerocopy and pinger support enabled by default (2476)
7062 - sync writes completed after process exits caused crashes (2319)
7063 - maintain correct mount count on the MDS (2356)
7064 - backout 1557, because 2316 wasn't really fixed
7065 - better file I/O statistics gathering in /proc
7066 - don't take unnecessary, deadlock-inducing bug in readpage (2383)
7067 - another kernel patch to fix zero-copy TCP function export
7068 - don't take duplicate lock when processing re-sent getattr (2420)
7069 - lctl uses obd_self_export instead of creating new conn (2353)
7070 - MDS/OST recovery case which requires object creation asserted (2425)
7071 - move lfs from /usr/sbin to /usr/bin in packages
7072 - fix race between mds_client_add and mds_client_free (2417)
7073 - use kmalloc instead of slabs in portals (2430)
7074 - don't create duplicate records when a failover MDS is present (2442)
7075 - remove unnecessary mount age check (2332)
7076 - don't remove directory inodes from locks prematurely (2451)
7077 - don't break if MDS service name is the same as hostname (2103)
7078 - fix races in client write RPC generation when cache full (2482)
7080 2003-12-13 Cluster File Systems, Inc. <info@clusterfs.com>
7083 - remove now-unused request->rq_obd (278)
7084 - if an allocation fails, print out how much memory we've used (1933)
7085 - use PORTAL_SLAB_ALLOC for structures, to get GFP_MEMALLOC (1933)
7086 - add the "configurable stack size" patch to most series files (1256)
7087 - ability to write large log records, for 100+ OST configs (2306)
7088 - fix NULL deref when filter_prep fails (2314)
7089 - fix operator precedence error in filter_sync
7090 - dynamic allocation of socknal TX descriptors (2315)
7091 - fix a missed case in the GFP_MEMALLOC patch, can cause deadlock (2310)
7092 - fix gcc 2.96 compilation problem in xattr kernel patch (2294)
7093 - ensure that CWARN messages in Portals always get to the syslog
7094 - __init/__exit are not for prototype decls (ldlm_init/exit)
7095 - x86-64 compile warning fixes
7096 - fix gateway LMC keyword conflict (2318)
7097 - fix MDS lock inversions in getattr/reint paths (1844)
7098 - fix a rare lock re-ordering bug, which caused deadlock (2322)
7099 - fix i_sem/journal inversion in fsfilt_ext3_write_record (2306)
7100 - DLM race condition prevented some lock evictions (2328)
7101 - ENOMEM detection and retry on socknal sends (2230)
7102 - use GFP_NOFS throughout Lustre, to combat ENOMEM (2230)
7103 - move osc_rpcd into ptlrpc, for use in MDC and others (2329)
7104 - protect MDS inode fsdata with stronger locking; fixes assertion (2313)
7105 - better error messages when a client is rejected during recovery (1505)
7106 - avoid cancelling locks which were never granted, after failure (2330)
7107 - fix i_sem/journal inversion in mds_client_add (2333)
7108 - fix truncate/getattr lock cycle deadlock (2334)
7109 - use rpcd to send close; allows resend after timeout, avoid leak (1897)
7110 - fix two rare exit paths which could leak an l_lock() ref (2321)
7111 - fencepost error in MDS/OST orphan recovery (2226)
7112 - make log record alignment 8 bytes (1988)
7113 - lstripe now fails when requested offset > ost_count (2237)
7114 - ensure that all kernel series have a complete list.h (1607)
7115 - fix crashes in special-file operations (2316)
7116 - lctl create/brw OID mismatch, caused by obsolete filter loop (2339)
7118 - allow configurable automake binary, for testing new versions
7119 - small update to the lfs documentation
7121 2003-12-03 Cluster File Systems, Inc. <info@clusterfs.com>
7123 * fix negative export reference count in fsfilt_sync (2312)
7125 2003-12-01 Cluster File Systems, Inc. <info@clusterfs.com>
7126 * release candidate 0.9.1
7128 - orphans are moved into the PENDING directory for possible recovery
7129 - replayed opens now open by fid for orphan/rename safety (1042)
7130 - last close of an orphan inode generates a transno (683)
7131 - chdir() and mount() now pin the directory entry (1020)
7132 - avoid CERROR in normal ll_setattr_raw() error case (1500)
7133 - discard very old requests without processing them (1502)
7134 - remove some common, well-understood CERRORs (1505)
7135 - require O_DIRECT I/O to be page-sized to workaround IA64 crash (1609)
7136 - clear "grant" flags in OST replies until OST grant code lands (1644)
7137 - fix read performance by not clobbering i_blksize on client (1598)
7138 - fix __ldlm_handle2lock oops by not dereferencing lock after PUT (1625)
7139 - make LRU size a /proc tunable, clears locks when reduced (707)
7140 - fix some lprocfs rot that prevented ptlbd from loading (1732)
7141 - server locks take references on exports now (1558)
7142 - build fixes for 2.4.20-rh trees (1663)
7143 - return an error from lov_create if all OSCs are inactive (1751)
7144 - fix import levels when a reconnect happens without a timeout (1597)
7145 - exit early from mds_open if we get a lookup error (1749)
7146 - partial page read at EOF wouldn't wait for disk before sending (1642)
7147 - avoid NULL deref in obdfilter when reading page past EOF (1592)
7148 - avoid LASSERT in ll_intent_lock if server failed very early (1090)
7149 - fix LBUG in ll_it_open_error with rc = -2 (1861)
7150 - write/truncate lock inversion (1639)
7151 - Don't auto-load obdclass, portals modules during cleanup (1495)
7152 - fix timestamps from jumping to "now" (1763)
7153 - extra journal assertions (1648)
7154 - add an extra multiunlink test (1771)
7155 - fix read_record/write_record API (1776)
7156 - fix leak of offset_extent, possible incorrect i_size later (1772)
7157 - fix lasserts in mis-matched transnos during open-unlink testing (1541)
7158 - Debugging for the kqswnal_get_idle_tx problems (1820)
7159 - Allow recovery to be attempted multiple times (1536)
7160 - Write out MDS last_rcvd file after it is first created (1600)
7161 - Fix tx_descriptor leak in failed transmit situations (1827)
7162 - ext3 journaling fixes for assertion failure after IO error (1871)
7163 - class_export_put() on freed export after completion AST error (1896)
7164 - Fix revalidate looping in VFS (1322)
7165 - Don't access a freed export during MDS_REINT timeout (1521)
7166 - Add open-unlink recovery support on the MDS (1673,1764)
7167 - Return an error if no MDS data was read from last_rcvd (1946)
7168 - Fix for lookup "." or ".." crash on error (1932,1931,1935)
7169 - Don't setup a disk device that doesn't match exported UUID (317)
7170 - Reduce bulk RPC timeout to avoid cascading client/OST failures (1845)
7171 - avoid committing NULL handle in force close
7172 - local.sh is now a one-stripe LOV configuration
7173 - POSIX utime.4 -EPERM on FIFO not owned by user (56)
7174 - fix ext3 htree duplicate directory entry corruption (1516)
7175 - POSIX creat.13, fstat.1, open.18, stat.3 new file atime/mtime (2020)
7176 - update to new LOV EA format (2097)
7177 - interoperability for different PAGE_SIZE/wordsize (686,1821,1343,2042)
7179 2003-06-15 Phil Schwan <phil@clusterfs.com>
7182 - imports and exports cleanup too early, need refcounts (349, 879, 1045)
7183 - per-import/export recovery handling (958, 931, 959)
7184 - multiple last-rcvd slots, for serving multiple FSes (949)
7185 - connections are again shared between multiple imp/exports (963, 964)
7186 - "umount -f" would hang if any requests needed to be sent (393, 978)
7187 - avoid pinning large req buffer by copying for queued messages (989)
7188 - add "uuid" to "lctl device" command to help upcalls (991)
7189 - "open" RPCs with transnos would confuse recovery counters (1037)
7190 - do proper endian conversion of all wire messages (288, 340, 891)
7191 - remove OST bulk get LBUGs, fix ost_brw_write cleanup (1126)
7192 - call waiting locks callback from LDLM recovery thread (1127, 1151)
7193 - fix ptlrpc_connection leak in target_handle_connect (1174)
7194 - fix import refcounting bug in OST and MDS cleanup (1134)
7195 - if an invalid-at-open-time OSC returned before close(), LBUG (1150)
7196 - fix very unlikely obd_types race condition (501)
7197 - remove osc_open hack for echo_client (1187)
7198 - we leaked exports/dlmimps for forcibly disconnected clients (1143)
7199 - a failure in read_inode2 leads to deadlock (1139)
7200 - cancel ack-locks as soon as transaction is committed (1072)
7201 - fix major leaks and crashes in the bulk I/O path (937, 1057)
7202 - make sure to commitrw after any preprw to avoid deadlock (1162)
7203 - failing to execute a file in a lustre FS would lock inode (1203)
7204 - small DEBUG_REQ fix to avoid dereferencing a NULL (1227)
7205 - don't ASSERT while cleaning up an incompletely-setup obd (1248)
7206 - obd_uuid2tgt would walk off the end of the list (1255)
7207 - on IA64 the osc would give portals incorrect bulk size (1258)
7208 - fix debug daemon ioctl interface; allows daemon on ia64 (1274)
7209 - fix lock inversion caused by new llite matching code (1282)
7210 - limit the number of dirty pages on a client to 10MB (1286)
7211 - timed out locks were not being corrected cancelled (1289)
7212 - fix O_DIRECT above 4GB on IA-32 (1292)
7213 * major user-visible changes
7214 - fail out/fail over policy now controlled by the upcall (993)
7216 - add OBD_PING to check server availability and failure (954)
7217 - lustre messages are now sent in sending host order (288, 340, 891)
7218 - add eadatalen to MDS getattr reply (340)
7219 - OST read replies may contain second buffer, with per-page status (593)
7221 2003-03-11 Phil Schwan <phil@clusterfs.com>
7224 - LDLM_DEBUG macro fix, for gcc 3.2 (850)
7225 - failed open()s could cause deadlock; fixed (867, 869)
7226 - stop cancelling OST locks when files are closed (481)
7227 - overlapping XID spaces caused network corruption (851, 853)
7228 - fix unsafe fsfilt counter arithmetic; change to atomic_t
7229 - setattr_raw added, to do single-RPC, server-side setattrs
7230 - lmc/lconf syntax change for OST UUIDs
7231 - fix crashy race condition between ptlrpc_free_req and osc_close
7232 - don't use request in mdc_enqueue if we hit a timeout (889)
7233 - don't set the inode i_size for regular files from the MDS (896)
7234 - handle out of order completion AST (842)
7235 - don't LBUG if a lock request times out after receiving AST (913)
7236 - avoid d_rehash race in ll_find_alias by rehashing inside dcache_lock
7237 - if a bad lock AST arrives, send an error instead of dropping entirely
7238 - return 0 from revalidate2 if ll_intent_lock returns -EINTR (912)
7239 - fix leak in bulk IO when only partially completed (899, 900, 926)
7240 - fix O_DIRECT for ia64 (55)
7241 - (almost) eliminate Lustre-kernel-thread effects on load average (722)
7242 - C-z after timeout could hang a process forever; fixed (977)
7244 - client-side I/O cache (678, 924, 929, 941, 970)
7246 - READPAGE and SETATTRs which don't take server-side locks get
7249 2003-02-11 Phil Schwan <phil@clusterfs.com>
7252 - Fix ldlm_lock_match on the MDS to avoid matching remote locks (592)
7253 - Fix fsfilt_extN_readpage() to read a full page of directory
7254 entries, or fake the remainder if PAGE_SIZE != blocksize (500)
7255 - Avoid extra mdc_getattr() in ll_intent_lock when possible (534, 604)
7256 - Fix imbalanced LOV object allocation and out-of-bound access (469)
7257 - Most intent operations were removed, in favour of a new RPC mode
7258 that does a single RPC to the server and bypasses most of the VFS
7259 - All LDLM resource ID arrays were removed in favour of ldlm_res_id
7260 - Aggressively cancel local locks on DLM servers
7261 - mds_reint_unlink sends EA to the client if it's the last nlink.
7262 client uses that EA to unlink OST objects.
7263 - mds_reint_{rename,unlink,link} were rewritten to take ordered locks
7264 - recursive symlinks were fixed (439)
7265 - fixed NULL deref in DEBUG_REQ
7266 - filter_update_lastobjid no longer calls sync, which annoyed extN
7267 - fixed multi-client small-writes to a single file problem (445)
7268 - fixed mtime updates during file writes (607)
7269 - fixed vector writes on obdfilter causing problems when ENOSPC (670)
7270 - fixed bug in obd_brw_read/write() (under guise of testing 367)
7271 - fixed Linux OST size reporting problem (444, 656)
7272 - OST now updates object mtime with writes or setattr (607, 619)
7273 - client verifies file size before zeroing page past EOF (445)
7274 - OST now writes last allocated objid to disk with allocation (108)
7275 - LOV on echo now works (409)
7277 - mds_reint_unlink sends a new buffer, with the EA included. this
7278 buffer is only valid if body->valid & OBD_MD_FLEASIZE, which is only
7279 set if a regular file was being unlinked, and it was the last link
7280 - use PtlGet from the target for bulk writes (315)
7281 - OST now updates object mtime with writes or setattr (607, 619)
7282 - LDLM now has a grant-time callback to revalidate locked items, if
7284 - Many MDS operations were reorganized to combat race conditions
7286 - Merge b_intel branch (updated lprocfs code) - now at /proc/fs/lustre
7287 - configure check to avoid gcc version 2.96 20000731-2.96-98) (606)
7289 2003-01-06 Andreas Dilger <adilger@clusterfs.com>
7292 - Fully reactivate OST imports after reconnection (512, others)
7293 - Make sure client sees our -ENOTCONN from mds_handle (513 - partial)
7294 - More graceful error handling for truncating on dead OST (515)
7295 - Don't error out unless we're actually accessing dead stripes (474)
7296 - Fix garbage sizes when stripes are missing (410)
7297 - LRU counters were broken, causing constant lock purge (433, 432)
7298 - garbage on read from stripes with failed OSTs (441)
7299 - mark OSCs as active before reconnecting during recovery (438)
7300 - lov_enqueue and lov_cancel need to handle inactive OSTs (403)
7301 - lfind did not preserve OST order in output (443)
7302 - symlinks cause hung clients, incorrect data (439)
7303 - stop dereferencing request after dropping refcount (457)
7304 - don't LASSERT(spin_is_locked) on non-SMP (455)
7305 - fixes for many rename() bugs
7306 - fstat didn't correctly synchronize attributes (399)
7307 - server must handle lock cancellation during blocking AST prep (487)
7308 - bulk descriptors were free()d too soon (511)
7309 - fix paths in lconf, which would load incorrect modules (451, 507)
7310 - fix confusing lconf 'host not found' error message (386)
7311 - fix lock order deadlock on OST (O/R i_sem before journal ops, 478)
7312 - fix race condition in mdc_blocking_ast() for inode access (526)
7313 - fix lov_unpackmd() unpacking wrong number of stripes (537)
7314 - fix lov_set_osc_active() marking wrong OSC inactive (440)
7315 - fix bad lstripe lov_unpackmd() assertion (fix layering too) (527)
7316 - fix multiple writes of stripe MD to MDS (358, maybe 519)
7317 - fix lstripe in several ways (kernel side) (527)
7318 - fix request leak in ldlm_cli_enqueue (262)
7319 - incorrect OSC was marked inactive after OST failure
7320 - call mds_fs_cleanup before unmounting filesystem (524)
7321 - fix races between taking ns_lock and ldlm_lock_change_resource
7322 - fix races updating LOV export open file list
7323 - fix lov_enqueue error path, avoid decref-ing bad lock handle
7324 - fix recovery NULL deref in ldlm_cli_cancel_unused
7325 - fix some DLM races by using new hash table for lock handles (419)
7326 - permit the client to specify desired inodes, at replay
7327 - duplicate requests when we queue them for replay reintegration
7328 - fix last_rcvd offset calculation
7329 - sync after each recovered transaction, so we always make progress
7330 - never, not always, ERESTART requests without transnos
7331 - store the lov_desc in the MDS, so we don't depend on getlovinfo to
7333 - skip replay if the MDS says that the client is already connected
7334 - don't check for a recovery-enabled export to match lctl's UUID
7335 - don't INC_USE_COUNT for phantom exports
7336 - don't crash when cleaning up phantom exports (567)
7337 - don't double-finish or set replay data for errored mdc_open requests
7338 - abort requests when they time out, so we don't get old replies
7339 - send/receive replies for AST messages again
7340 - if the client says that it doesn't have the lock, cancel it on the
7342 - if we timeout during I/O, don't try to cancel an in-use lock; instead
7343 mark it as destroyed, it will all work out when decref is called
7344 - fix module use counts (22, 581)
7346 - ASTs now expect a reply (server cancels lock on error reply)
7348 2002-12-02 Andreas Dilger <adilger@clusterfs.com>
7351 - fix many simultaneous client startup (392)
7352 - fix dentry->d_it clobbering
7353 - credentials weren't being shipped for readdir/getattr operations
7354 - remove invalid assertions triggered during some concurrent MD
7356 - proper Lustre versions added (336, 389)
7357 - fix memory leak for create error case (398)
7358 - fix LOV locking bug that would get cli/srv out of sync
7359 - fix echo client over LOV (409)
7360 - fix dbench 2, extN refcount problem (170, 258, 356, 418)
7361 - fix double-O_EXCL intent crash (424)
7362 - avoid sending multiple lock CANCELs (352)
7364 - MDS can do multi-client recovery (modulo bugs in new code)
7366 - many updates, edits, cleanups
7368 2002-11-18 Phil Schwan <phil@clusterfs.com>
7371 - fix null d_it dereference (346)
7372 - fix full OST/dbench hang (333)
7373 - fix permission problem with file removal (286)
7374 - fix removal of OSCs from LOV when they fail
7375 - fix NULL deref during bulk timeout (214)
7376 - fix problems related to multiple filesystems on one MDS (241)
7377 - fixed serious subtle metadata locking bugs
7378 - free locks on clients when inodes are removed due to memory
7380 - fix inode pointer in lock data (285)
7381 - partial support for multiple MDS on a single host (241)
7382 - data locks weren't cancelled at clear_inode time (290, 311)
7383 - intent locks could lead to unbounded lock growth (205)
7384 - added a maximum lock count, an LRU list, and a flusher
7385 - fix multiple rename (365)
7386 - properly abstracted the echo client
7387 - OSC locked 1 byte too many; fixed
7388 - rewrote brw callback code:
7389 - fixed recovery bugs related to LOVs (306)
7390 - fixed too-many-pages-in-one-write crash (191)
7391 - fixed (again) crash in sync_io_timeout (214)
7392 - probably fixed callback-related race (385)
7394 - Add capability to MDS protocol
7395 - LDLM cancellations and callbacks on different portals
7397 2002-10-28 Andreas Dilger <adilger@clusterfs.com>
7400 - limit client IOV size to PTL_MD_MAX_IOV (611336, 191)
7401 - defer open object destruction to close time (601981, 138)
7402 - open/close OST file handle in obdo (OBD_MD_FLHANDLE) (601981, 138)
7403 - move LDLM_ENQUEUE/CONVERT back to MDS portal (625069)
7404 - abstract ll_lookup2, fix ll_revalidate2 to use abstraction (256)
7405 - don't call obd_setattr in ll_file_release for destroyed objects
7406 * protocol change to lustre_msg: move |version| and add |flags|
7407 * protocol change to osc_punch: "start" in "o_size", "end" in "o_blocks"
7408 * lock replay: for LDLM_FL_REPLAY trust client to do right thing
7409 * added replay of create, unlink, link and rename operations during
7410 MDS failover; recovery should be much more robust now
7411 * remove failed OSCs from LOVs (only lov_create uses this so far)
7412 * the lustre-HOWTO was brought (more) up to date (582544)
7414 2002-10-23 Phil Schwan <phil@clusterfs.com>
7417 - in-use dentries weren't being reused properly (617851)
7418 - prevent multiple LDLM setup (599178)
7419 - fix LOV size calculations for truncate (617853)
7420 - fix client handling of MDS intent errors (POSIX)
7421 - fix permission bug in lovstripe.c test (624321)
7422 - fix MDS thread deadlock - move LDLM handler to DLM portal (625069)
7423 - truncate past end of file could corrupt data
7424 - proper cleanup after timeouts, crashes, etc (592524, 550815)
7425 - a race in recovery could return ETIMEDOUT to apps (623947)
7426 - building outside the source directory was fixed
7427 * the lustre-HOWTO was brought (more) up to date (582544)
7428 * major progress was made on recovery functionality
7430 2002-10-10 Phil Schwan <phil@clusterfs.com>
7433 - recovery deadlock fix
7434 - rm -rf causes LBUG fix (617817)
7435 - file open by multiple tasks fix (618962)
7436 - directory permissions bugs (602707 and 620007)
7437 - journal_stop fixed with locking (611313)
7438 - O_APPEND failures resolved (618273, perhaps 614459)
7439 - lconf PATH fix (619770)
7440 - IA64 build fix (621450)
7441 - RPC buffer sizes scale with amount of memory
7443 2002-10-01 Phil Schwan <phil@clusterfs.com>
7446 - locks would be cancelled without throwing away data pages,
7447 resulting in inconsistent data (605627)
7448 - inode attributes were not always being refreshed (605627, 612449)
7449 - lconf now continues to cleanup after lctl reports an error
7450 - MDS now enforces user permissions (602707)
7451 - lprocfs cleanup fixed, but not yet enabled (614157)
7452 - fixed infinite server hang, should a client not respond to an AST
7453 - avoid going into recovery if user calls readlink() with a buffer
7454 that's too small (613941)
7455 - AST RPCs no longer require replies (614867) -- this may be changed
7456 - don't crash server if client sends an IOV that's too big (611336)
7457 - fixed lock conversion deadlock (611892)
7458 - fixed the following of symlinks (614622)
7459 * recovery: the server can remove locks from a client that dies, other
7460 clients can make progress
7461 * more extN patch fixes
7462 * compile-time configurable ptlrpc buffer allocations
7464 - collaborative read cache document
7465 - Lustre Lite Performance CDR document-in-progress
7467 2002-09-20 Andreas Dilger <adilger@clusterfs.com>
7470 - fix typo in patch-2.4.18
7472 2002-09-20 Andreas Dilger <adilger@clusterfs.com>
7475 - clear ptlrpc request each time in handle_incoming_request()
7476 - unlink of files now destroys the object on the OST
7478 2002-09-19 Peter Braam <braam@clusterfs.com>
7480 * add hard link support
7481 * change obdfile creation method
7482 * kernel patch changed
7484 2002-09-19 Peter Braam <braam@clusterfs.com>
7487 - stack overflow bug in extN fixed
7489 2002-09-18 Andreas Dilger <adilger@clusterfs.com>
7491 * documentation updates
7492 - add man pages for config tools
7493 - update tests/README to describe testing with new config tools
7494 - finish metadata API descriptions
7495 * bug fixes and cleanups
7496 - statfs workaround for 16TB limit
7497 - LOV stripe allocation improved, can stripe on subset of OSTs
7498 - LOV file size/IO offset was wrong for files > 4GB in size
7499 - object EA data was being dropped, caused files to be unreadable
7500 - memory overflow with non-LOV OST caused memory corruption
7501 - fixed regression tests to work with new config tools, obdfilter
7502 - fixed bug when directory size became larger than 1 block
7503 - fixed bug (for single client case) when PWD was deleted
7504 - invalidate local directory pages when doing intent-based ops
7505 - avoid LDLM oops when lock callback contained bad data
7507 2002-09-09 Andreas Dilger <adilger@clusterfs.com>
7509 * documentation updates
7510 * bug fixes and cleanups
7511 - configuration tools
7514 - 64-bit compile warnings
7515 - 64-bit internal statfs data
7517 * test_brw on persistent OST devices
7519 * lprocfs (disabled)
7521 2002-09-04 Andreas Dilger <adilger@clusterfs.com>
7523 * documentation updates
7524 * bug fixes and cleanups
7525 * configuration tools
7527 2002-08-30 Peter J. Braam <braam@clusterfs.com>
7530 * many small fixes to 0_5_4
7531 * io/network handling
7532 * thinkos in MDS operations
7534 2002-08-24 Peter J. Braam <braam@clusterfs.com>
7537 * crucial basic fixes to 0.5.3
7538 * IOR, Iozone work over Elan
7541 2002-08-07 Phil Schwan <phil@clusterfs.com>
7542 * version 0_5_3, our first alpha
7543 * we use the new Portals iovs
7544 * documentation updates
7545 * bug fixes and cleanups
7546 * small changes in the DLM wire protocol
7548 2002-07-25 Peter J. Braam <braam@clusterfs.com>
7549 * version 0_5_1 with some initial stability,
7550 * locking on MD and file I/O.
7551 * documentation updates
7552 * several bug fixes since 0.5.0
7553 * small changes in wire protocol
7555 2002-07-18 Phil Schwan <phil@clusterfs.com>
7557 * delivered as Lustre Light Alpha
7558 * fixed a crash after handling invalid MDS requests
7559 * fixed directory pages for architectures with non-4k pages sizes
7561 2002-07-11 Andreas Dilger <adilger@clusterfs.com>
7562 * release version v0_4_4
7563 * Moves TCP acceptor to be on port 2432 (unused Coda port) instead
7565 * Fixes a number of interruption problems with OST operations.
7566 * Update documentation for portals header changes
7567 * Move all wire protocol structs/defines to lustre_idl.h
7568 * Fixes symlink length bug.
7569 * Add tcpdump to repository.
7571 2002-07-05 Andreas Dilger <adilger@clusterfs.com>
7572 * release version v0_4_3
7573 * Fixes statfs for inodes on extN.
7574 * Fixes bug in runtests which would delete /etc/hosts.
7575 * Use 64-bit object IDs wherever possible (not into VFS though)
7576 Remove ost_get_info, which is unused by lustre, and out of date.
7578 2002-07-03 Peter Braam <braam@clusterfs.com>
7579 * release version v0_4_2 Fixes a lookup error (type not passed)
7580 * move forward to head of Portals
7581 * move forward to latest Lustre kernel
7583 2002-06-25 Peter Braam <braam@clusterfs.com>
7584 * release version v0_4_1. Hopefully stable on single node use.