4 2.6.16.60-0.42.8 (SLES 10),
5 2.6.27.39-0.3.1 (SLES11),
6 2.6.18-164.11.1.el5 (RHEL 5)
7 2.6.18-164.11.1.0.1.el5 (OEL 5)
8 * Client support for unpatched kernels:
9 (see http://wiki.lustre.org/index.php?title=Patchless_Client)
10 2.6.16 - 2.6.30 vanilla (kernel.org)
11 * Recommended e2fsprogs version: 1.41.6.sun1
12 * The async journal commit feature (bug 19128) and the cancel
13 lock before replay feature (bug 16774) are disabled by default.
17 Description: lfs quota failed when OSTs are down
18 Details : really return approximate block/inode usage when OSTs are down
22 Description: llobdstat fix and enhancement
23 Details : add a counter to set a limit to how many samples will be returned
24 fix a wildcard in the path to limit to obdfilter stats only
28 Description: recovery-small 51 hang - MDT can not stop, Mount still busy with
30 Details : abort lock enqueue processing sooner when we umount is in progress.
32 Severity : enhancement
34 Description: Update RHEL5.4 kernel to 2.6.18-164.11.1.el5 and
35 OEL5.4 kernel to 2.6.18-164.11.1.0.1.el5.
37 Severity : enhancement
39 Description: Simplify MDT/OST service start message.
42 Bugzilla : 21961/17914
43 Description: ignore trailing -mdc when determining index number
47 Description: decrease the usage of memory on clients.
48 Details : 1. On clients, recycle dentries and inodes unused.
49 2. Delete the code related to ll_deathrow(att 6215 in bug 1443). It
51 -------------------------------------------------------------------------------
53 2010-01-29 Sun Microsystems, Inc.
55 * Support for kernels:
56 2.6.16.60-0.42.8 (SLES 10),
57 2.6.27.39-0.3.1 (SLES11),
58 2.6.18-164.11.1.el5 (RHEL 5)
59 2.6.18-164.6.1.0.1.el5 (OEL 5)
60 * Client support for unpatched kernels:
61 (see http://wiki.lustre.org/index.php?title=Patchless_Client)
62 2.6.16 - 2.6.30 vanilla (kernel.org)
63 * Recommended e2fsprogs version: 1.41.6.sun1
64 * The async journal commit feature (bug 19128) and the cancel
65 lock before replay feature (bug 16774) are disabled by default.
67 Severity : enhancement
69 Description: Update RHEL5.4 kernel to 2.6.18-164.11.1.el5 and
70 OEL5.4 kernel to 2.6.18-164.11.1.0.1.el5.
72 Severity : enhancement
73 Bugzilla : 21511/19848
74 Description: Update SLES11 kernel to 2.6.27.39-0.3.1.
76 Severity : enhancement
78 Description: Update supported SLES10 kernel to 2.6.16.60-0.42.8.
80 Severity : enhancement
82 Description: Update kernel to RHEL5.4 2.6.18-164.6.1.el5 and
83 OEL5 2.6.18-164.6.1.0.1.el5(Both in-kernel OFED enabled).
85 Severity : enhancement
87 Description: Build kernels (RHEL5, OEL5 and SLES10/11) using the vendor's own
90 Severity : enhancement
92 Description: Vanilla kernel 2.6.30 patchless client support.
97 Description: bad entry in directory xxx: inode out of bounds
98 Details : fix locking issue in the rename path which could race with any
99 other operations updating the same directory.
101 Severity : enhancement
103 Description: Make watchdog timer messages to be more clear and descriptive.
107 Description: cp -p command does not preserve the dates and timestamp
108 Details : mtime could be spoiled by a write callback
112 Description: Clear imp_force_reconnect correctly in ptlrpc_connect_interpret()
116 Description: Allow non-root access for "lfs check".
117 Details : Added a check in obd_class_ioctl() for OBD_IOC_PING_TARGET.
119 Severity : enhancement
121 Description: quotacheck performance/scaling issues
122 Details : reduce quotacheck time on empty filesystem by skipping uninit
125 Severity : enhancement
127 Description: Enhancement for lfs(1) command to use numeric uid/gid.
129 Severity : enhancement
131 Description: Adjust locks' extents on their first enqueue, so that at the time
132 they get granted, there is no need for another pass through the
133 queues since they are already shaped into the proper forms.
137 Description: Fix mds_shrink_intent_reply()/mds_intent_policy() to pass correct
138 arguments and prevent LBUG() in lustre_shrink_reply_v2().
142 Description: Change tunefs.lustre and mkfs.lustre --mountfsoptions so that
143 exactly the specified mount options are used. Leaving off
144 any "mandatory" mount options is an error. Leaving off any
145 default mount options causes a warning, but is allowed.
146 Change errors=remount-ro from mandatory to default. Sanitize
147 the mount string before storing it. Update man pages accordingly.
151 Description: mds_getattr() should return 0, even if mds_fid2entry() fails with
152 -ENOENT. Also fix in ptlrpc_expire_one_request() to print signed
155 Severity : enhancement
157 Description: Remove set_info(KEY_UNLINKED) from MDS/OSC
159 Severity : enhancement
161 Description: Clients can replay thousands of unused locks during recovery
162 Details : Don't replay unused locks (only read locks for now) during
163 recovery. This feature is disabled by default and can be
164 enabled by running the following command on the clients:
165 lctl get_param ldlm.cancel_unused_locks_before_replay
169 Description: can't stat file in some situation.
170 Details : improve initialize osc date when target is added to mds and
171 ability to resend too big getattr request is client isn't have info
176 Description: Prevent inconsistences between linux and lustre mount structures.
177 Details : Wait indefinitely in server_wait_finished() until mnt_count drops.
178 Make the sleep interruptible.
180 Severity : enhancement
182 Description: Communicate OST degraded/readonly state via statfs to MDS
183 Details : Flags in the statfs returned from OSTs indicate whether the
184 OST is in a degraded RAID state, or if the filesystem has
185 turned read-only after a filesystem error is detected.
190 Description: don't panic if EPROTO was hit when reading symlink
191 Details : correctly handling request reference in error cases.
196 Description: open sometimes returns ENOENT instead of EACCES
197 Details : checking permission should be part of open part of mds_open, not
198 lookup part. so server should be set DISP_OPEN_OPEN disposition
199 before starting permission check.
200 Also not need revalidate dentry if client already have LOOKUP lock.
204 Description: enable client interface failover
205 Frequency : on servers with multiple network interfaces
206 Details : When a child reconnects from another NID, properly update export
207 nid hash position and ldlm reverse import.
209 Severity : enhancement
211 Description: implemented direct I/O with arbitrary (nonaligned) memory
212 addresses and file offsets.
214 Severity : enhancement
216 Description: added more recovery timeout options.
218 Severity : enhancement
220 Description: added llapi_file_open, llapi_file_create, llapi_file_get_stripe
225 Description: Avoid deadlock for local client writes
226 Frequency : only on systems with clients writing to an OST on the same node
227 Details : Use new OBD_BRW_MEMALLOC flag to notify OST about writes in the
228 memory freeing context. This allows OST threads to set the
229 PF_MEMALLOC flag on task structures in order to allocate memory
230 from reserved pools and complete IO.
231 Use GFP_HIGHUSER for OST allocations for non-local client writes,
232 so that the OST threads generate memory pressure and allow
233 inactive pages to be reclaimed.
238 Descriptoin: lock ordering violation between &cli->cl_sem and _lprocfs_lock
239 Details : move ldlm namespace creation in setup phase to avoid grab
240 _lprocfs_lock with cli_sem held.
244 Frequency : only during format of test systems
245 Description: Unable to run several mkfs.lustre on loop devices at the same time
246 Details : mkfs.lustre returns error 256 on the concurrent loop devices
247 formatting. The solution is to proper handle the error.
249 Severity : enhancement
251 Description: implement async create (obd_async_create) method for osc, to avoid
252 too long waiting new ost objects with holding ldlm lock.
256 Frequency : occasionally during network problems
257 Description: client not allowed to reconnect to OST because of active request
258 Details : abort bulk requests received by the OST once the client has timed
259 out since the client will resend the request anyway. The client
260 also now retries to reconnect to the same server if a connect
261 request failed with EBUSY or -EAGAIN.
264 Frequency : rare, if used wide striped file and one ost in down.
266 Descriptoin: don't return error if we created a subset of objects for file.
267 Details : lov_update_create_set() uses set->set_success as index for created
268 objects, so if some requests failed, they will have hole at end of
269 array and we can use qos_shrink_lsm for allocate correct lsm.
273 Description: Slow stale export processing during normal start up
274 Details : The global mgc lock prevents OST setup to be run in parallel.
275 Replace the global lock with a per-config_llog_data semaphore.
279 Description: Out or order replies might be lost on replay
280 Details : In ptlrpc_retain_replayable_request if we cannot find retained
281 request with tid smaller then one currently being added, add it
282 to the start, not end of the list.
286 Description: BUG: soft lockup - CPU#1 stuck for 10s! [ll_mdt_07:4523]
287 Details : add cond_resched() calls to avoid hogging the cpu for too long
288 in the hash code. Make also lustre_hash_for_each_empty() more
291 Severity : enhancement
293 Description: Performance improvements for debug messages with D_RPCTRACE,
294 D_LDLM, D_QUOTA options.
297 Frequency : only with NFS export
299 Description: (lov_merge.c:74:lov_merge_lvb())
300 ASSERTION(spin_is_locked(&lsm->lsm_lock)) failed (SR 71691004)
301 Details : Fix a race in the nfs export code by populating inode
302 info while the new inode is still locked
304 Severity : enhancement
306 Description: add a new file in procfs called force_lbug. Writting to this
307 file triggers a LBUG. Only for test purpose.
311 Description: OOM killer causes node hang
312 Details : really interrupt the sleep in osc_enter_cache on signals
316 Description: LustreError: 9153:0:(quota_context.c:622:dqacq_completion()) LBUG
317 Details : fix race during quota release on the slave.
319 Severity : enhancement
321 Description: smaller hash bucket sizes, cleanups
322 Details : increase hash table sizes and enabled rehashing for pools, uuid,
325 Severity : enhancement
327 Description: Add ldiskfs maxdirsize mount option
328 Details : add max_dir size mount option
332 Description: panic in ll_statahead_thread
333 Details : prevent parent thread to be killed before its child
336 Frequency : only with 16TB device
338 Description: unable to perform "mount -t lustre" of 16TB OST device
339 Details : Mounting 16TB LUNs failed due to three bugs in mkfs.lustre.
343 Description: ASSERTION(atomic_read(&imp->imp_inflight) == 0) failed
344 Details : unregistering should be zero if no RPC inflight.
348 Description: hyperion: Oops during metabench
349 Details : Correct the refcount of lov_request_set
351 Severity : enhancement
353 Description: Add mptlinux and nxge drivers to Lustre builds
355 Severity : enhancement
357 Description: Fix watchdog timer message to be more clear
358 Details : Make watchdog timer messages more clear and descriptive.
362 Description: LNET soft lockups in socknal_cd thread
363 Details : don't hog CPU for active-connecting if another connd is
364 accepting connecting-requst from the same peer
368 Description: recovery-small test_17 hang
369 Details : Land several AT improvements & fixes.
373 Description: MDS panic and hanging client processes
374 Details : Replace exp_ops_stats with exp_nid_stats->nid_stats
378 Description: OSS stuck in recovery.
379 Details : fix race during recovery. class_unlink_export,
380 class_set_export_delayed and target_queue_last_replay_reply
381 may race while increasing/decreasing obd_recoverable_clients
382 and obd_delayed_clients, causing recovery to wait forever.
384 Severity : enhancement
386 Description: add cascading_rw.c to lustre/tests
390 Description: filter_last_id() NULL deref
391 Details : lprocfs_filter_rd_last_id() should check for the fully
392 setup obd device, before proceeding further.
394 Severity : enhancement
396 Description: Loadgen improvements
397 Details : stacksize and locking fixes for loadgen
401 Description: Quiet CERROR("dirty %d > system dirty_max %d\n"
402 Details : The atomic_read() allowing the atomic_inc() are not covered
403 by a lock. Thus they may safely race and trip this CERROR()
404 unless we add in a small fudge factor (+1).
406 Severity : enhancement
408 Description: shrink_slab: nr=-9223362083340912175
409 Details : fix spurious message from shrink_slab reporing negative nr
413 Description: Quiet bogus previously committed transno error
414 Details : suppress the "server went back in time" error message which
415 is always printed even in the common case after a client eviction
417 Severity : enhancement
419 Description: Parallel statfs() calls result in client eviction
420 Details : cache statfs data for 1s.
424 Description: parallel-scale test_compilebench: @@@@@@ FAIL: compilebench
426 Details : fix serveral issues in pinger code causing clients not to ping
427 servers for too long, resulting in evictions.
431 Description: e2fsck should warn when MMP update interval is extended
432 Details : print mmp_check_interval and make it possible to abort mount
433 operation in case it takes too long.
437 Description: mdsrate-create-large.sh, BUG: soft lockup - CPU#0 stuck for 10s!
438 Details : fix bug in the RHEL5's jbd2 callback patch.
442 Description: drop number of active requests when queued for recovery
443 Details : Now that we take a reference on the original request instead of
444 making a copy of it for recovery. We need to drop the number of
445 active requests or the queued requests will prevent all request
446 processing when they exceed (srv->srv_threads_running - 1).
448 Severity : enhancement
450 Description: refuse to invalidate operational quota files when they are in use
451 Details : an attempt to invalidate operational quota files on the quota
452 master is not actually permitted by VFS (returning -EPERM), but we
453 should not depend on that and should return the error earlier.
457 Description: Applications stuck in jbd2_log_wait_commit during exit
458 Details : fix deadlock between kjournald2 trying to acquire the page lock
459 owned by an ost_io thread waiting for journal commit.
461 -------------------------------------------------------------------------------
463 2009-10-16 Sun Microsystems, Inc.
465 * Support for kernels:
466 2.6.16.60-0.42.4 (SLES 10),
467 2.6.27.29-0.1 (SLES11, i686 & x84_64 only),
468 2.6.18-128.7.1.el5 (RHEL 5),
469 2.6.18-128.7.1.el5 (OEL 5),
470 * Client support for unpatched kernels:
471 (see http://wiki.lustre.org/index.php?title=Patchless_Client)
472 2.6.16 - 2.6.27 vanilla (kernel.org)
473 * Recommended e2fsprogs version: 1.41.6.sun1
474 * ext4 support for RHEL5 is experimental and thus should not be
477 Severity : enhancement
479 Description: Add OEL5 support.
481 Severity : enhancement
483 Description: Update kernel to SLES11 2.6.27.29-0.1.
487 Description: File checksum failures with OST read cache on
488 Details : Disable page poisoning when the bulk transfer has to be aborted
489 because the client got evicted.
493 Description: Don't allow make backward step on assiging osc next id.
494 Details : race between allocation next id and ll_sync thread can be cause
495 of set wrong osc next id and can be kill valid ost objects.
497 Severity : enhancement
499 Description: Update kernel to RHEL5 2.6.18-128.7.1.el5.
501 Severity : enhancement
503 Description: Update kernel to SLES10 SP2 2.6.16.60-0.42.4.
507 Description: Changes in raid5-large-io-rhel5.patch to calculate sectors properly
511 Description: Increase the default BLK_DEF_MAX_SECTORS value for RHEL5 and SLES11
515 Description: Error handling in osc_statfs_interpret() has been improved.
516 Details : Check in osc_statfs_interpret() for EBADR.
520 Description: Do not send statfs() requests to OSTs disabled by administrator.
521 Details : Check in lov_prep_statfs_set() for non-NULL ltd_exp.
525 Description: Do not update ctime for the deleted inode.
526 Details : Check in mds_reint_unlink() before calling fsfilt_setattr().
530 Description: Increase of the size of the LDLM resource hash.
531 Details : Bump up RES_HASH_BITS=12.
535 Description: correctly send lsm on open replay
536 Details : MDS is trust to LSM size on replay open, but client can set wrong size
541 Description: Deadlock between filter_destroy() and filter_commitrw_write().
542 Details : filter_destroy() does not hold the DLM lock over the whole
543 operation. If the DLM lock is dropped, filter_commitrw() can go
544 through, causing the deadlock between page lock and i_mutex.
545 The i_alloc_sem should also be hold in filter_destroy() while
550 Description: truncate starts GFP_FS allocation under transaction causing deadlock
551 Details : ldiskfs_truncate calls grab_cache_page which may start page
552 allocation under an open transaction. This may lead to
553 calling prune_icache with consequent lustre reentrance.
556 Frequency : only when down/upgrading the MDS to 1.6/1.8 while 1.8 clients are
557 still up and when the OST pool feature is used
559 Description: interop testing got LBUG when run dd with OST pool
560 :LustreError: 30032:0:(llite_lib.c:1913:ll_replace_lsm()) LBUG
561 Details : down/upgrading the MDS to a version that doesn't/does support OST
562 pool can cause clients to crash because the lsm has changed
567 Description: missing tree_status on 1.8.1 RPM build
568 Details : make rpms failed due because the tree_status file is missing.
572 Description: continuing LustreError "mds adjust qunit failed!"
573 Details : don't print message on the console when ->adjust_qunit fails.
577 Description: don't increase ldlm timeout if previous client was evicted
578 Details : if a client doesn't respond to a blocking callback within the
579 adaptive ldlm enqueue timeout, don't adjust the adaptive estimate
580 when the lock is next granted.
584 Description: ost is being unmounted w/o all writes to last_rcvd landing on disk.
585 affects recovery negatively.
586 Details : make sure all exports have been properly destroyed by the zombie
587 thread processed before stopping the target.
591 Description: Performance degradation with O_DIRECT between 1.6 & 1.8.1 b190
592 Details : disable write barrier for ext4/SLES11.
596 Description: Kernel panic - not syncing: Out of memory and no killable
597 processes... on OSS when iozone
598 Details : fix memory leak in the journal checksum patch.
602 Description: group quota "too many blocks" OSS crashes
603 Details : we should keep the same uid/gid for lquota_chkquota() and
604 lquota_pending_commit()
608 Description: LustreError: 9153:0:(quota_context.c:622:dqacq_completion()) LBUG
609 Details : don't LBUG on release quota error. Just a workaround until the
610 problem is understood.
612 ------------------------------------------------------------------------------
613 2009-07-31 Sun Microsystems, Inc.
615 * Support for kernels:
616 2.6.16.60-0.39.3 (SLES 10),
617 2.6.27.23-0.1 (SLES11, i686 & x84_64 only),
618 2.6.18-128.1.14.el5 (RHEL 5),
619 * Client support for unpatched kernels:
620 (see http://wiki.lustre.org/index.php?title=Patchless_Client)
621 2.6.16 - 2.6.27 vanilla (kernel.org)
622 * Recommended e2fsprogs version: 1.41.6.sun1
623 * File join has been disabled in this release, refer to Bugzilla 16929.
624 * NFS export disabled when stack size < 8192. Since the NFSv4 export
625 of Lustre filesystem with 4K stack may cause a stack overflow. For
626 more information, please refer to bugzilla 17630.
627 * ext4 support for RHEL5 is experimental and thus should not be
630 Severity : enhancement
632 Description: Update kernel to SLES10 SP2 2.6.16.60-0.39.3.
635 Frequency : with 1.8 server and 1.6 clients
637 Description: correctly shrink reply for avoid send too big message to client.
638 Details : 1.8 mds is allocate to big buffer to LOV EA data and this produce
639 some problems with sending this reply to 1.6 client.
643 Description: Repeated atomic allocation failures.
644 Details : Use GFP_HIGHUSER | __GFP_NOMEMALLOC flags for memory allocations
645 to generate memory pressure and allow reclaiming of inactive pages.
646 At the same time, do not allow to exhaust emergency pools.
647 For local clients the use of GFP_NOFS will be introduced in 1.8.2
649 Severity : enhancement
650 Bugzilla : 19846, 18289
651 Description: Update kernel to RHEL5 2.6.18-128.1.14.el5.
653 Severity : enhancement
654 Bugzilla : 19625, 16893, 18668, 19848
655 Description: Add support for SLES11 2.6.27.23-0.1.
657 Severity : enhancement
659 Description: Update client support to vanila kernels up to 2.6.27.
661 Severity : enhancement
663 Description: Update kernel to SLES10 SP2 2.6.16.60-0.37.
665 Severity : enhancement
667 Description: Compile with -Werror by default for i686 and x86_64.
671 Description: resolve race between obd_disconnect and class_disconnect_exports
672 Details : if obd_disconnect will be called to already disconnected export he
673 forget release one reference and osc module can't unloaded.
675 Severity : enhancement
677 Description: move AT tunable parameters for more consistent usage
678 Details : add AT tunables under /proc/sys/lustre, add to conf_param parsing
682 Descriptoin: correctly skip time estimate if in recovery
683 Details : rq_send_state insn't bitmask so using bitwise ops is forbid.
687 Descriptoin: OSS DeadLock
688 Details : Use trylock to prevent deadlock when shrink icache.
690 Severity : enhancement
692 Description: Allow tuning service thread via /proc
693 Details : For each service a new
694 /proc/fs/lustre/{service}/*/thread_{min,max,started} entry is
695 created that can be used to set min/max thread counts, and get the
696 current number of running threads.
698 Severity : enhancement
700 Description: Add state history info file, enhance import info file
701 Details : Track import connection state changes in a new osc/mdc proc file;
702 add overview-type data to the osc/mdc import proc file.
706 Description: Reduce small size read RPC
707 Details : Set read-ahead limite for every file and only do read-ahead when
708 available read-ahead pages are bigger than 1M to avoid small size
713 Description: free_entry erroneously used groups_free instead of put_group_info
715 Severity : enhancement
717 Description: Make read-ahead stripe size aligned.
719 Severity : enhancement
721 Description: MDS create should not wait for statfs RPC while holding DLM lock.
724 Frequency : rare, connect and disconnect target at same time
726 Descriptoin: ASSERTION(atomic_read(&imp->imp_inflight) == 0
727 Details : don't call obd_disconnect under lov_lock. this long time
728 operation and can block ptlrpcd which answer to connect request.
731 Frequency : start MDS on uncleanly shutdowned MDS device
733 Descriptoin: ll_sync thread stay in waiting mds<>ost recovery finished
734 Details : stay in waiting mds<>ost recovery finished produce random bugs
735 due race between two ll_sync thread for one lov target. send
736 ACTIVATE event only if connect realy finished and import have
740 Frequency : start MDS on uncleanly shutdowned MDS device
742 Descriptoin: aborting recovery hang on MDS
743 Details : don't throttle destroy RPCs for the MDT.
747 Description: Slow reads beyond 8Tb offsets.
748 Details : Page index integer overflow in ll_read_ahead_page
752 Description: Soft lockup on OSS after MDS failover
753 Details : MSG_CONNECT_INITIAL is not set on the initial MDS->OST connect.
754 As a conseqence, the patch from bug 18224 is not operational
755 and the MDS export cannot be reused on the OSTs until it gets
759 Frequency : rare, only if using MMP with Linux RAID
761 Description: MMP doesn't work with Linux RAID
762 Details : While using HA for Lustre servers with Linux RAID, it is possible
763 that MMP will not detect multiple mounts. To make this work we
764 need to unplug the device queue in RAID when the MMP block is being
765 written. Also while reading the MMP block, we should read it from
766 disk and not the cached one.
769 Frequency : rare, during recovery
771 Description: Assertion failure in ldlm_lock_put
772 Details : Do not put cancelled locks into replay list, hold references on
777 Description: 1.6.5 mdsrate performance is slower than 1.4.11/12 (MDS is not cpu bound!)
778 Details : create_count always drops to the min value (=32) because grow_count
779 is being changed before the precreate RPC completes.
782 Frequency : Only in RHEL5 when mounting multiple ext3 filesystems
785 Description: "kmem_cache_create: duplicate cache jbd_4k" error message
786 Details : add proper locking for creation of jbd_4k slab cache
790 Description: MMP check in ext3_remount() fails without displaying any error
791 Details : When multiple mount protection fails during remount, proper error
796 Description: Rare Client crash on resend if the file was deleted.
797 Details : When file is opened, but open reply is lost and file is
798 subsequently deleted before resend, resend processing logic
799 breaks trying to open the file again, should not try to open.
803 Description: add check for >8TB ldiskfs filesystems
804 Details : ext3-based ldiskfs does not support greater than 8TB LUNs.
805 Don't allow >8TB ldiskfs filesystems to be mounted without
806 force_over_8tb mount option
810 Description: Client locked up when running multiple instances of an app. on
811 multiple mount points
812 Details : ll_shrink_cache() can sleep while holding the ll_sb_lock.
813 Convert ll_sb_lock to a read/write semaphore to fix the problem.
817 Description: Cannot acces an NFS-mounted Lustre filesystem
818 Details : An NFS client cannot access the Lustre filesystem NFS-mounted
819 from a Lustre-client exporting the Lustre filesystem via NFS.
823 Description: panic in ll_statahead_thread
824 Details : grab dentry reference in parent process.
827 -------------------------------------------------------------------------------
829 tbd Sun Microsystems, Inc.
831 * Support for kernels:
832 2.6.16.60-0.31 (SLES 10),
833 2.6.18-128.1.6.el5 (RHEL 5),
834 2.6.22.14 vanilla (kernel.org)
835 * Client support for unpatched kernels:
836 (see http://wiki.lustre.org/index.php?title=Patchless_Client)
837 2.6.16 - 2.6.22 vanilla (kernel.org)
838 * Recommended e2fsprogs version: 1.40.11-sun1
839 * File join has been disabled, refer to Bugzilla 16929.
840 * A new Lustre ADIO driver is available for MPICH2-1.0.7.
841 * NFS export disabled when stack size < 8192. Since the NFSv4 export of
842 Lustre filesystem with 4K stack may cause a stack overflow. For more
843 information, please refer to bugzilla 17630.
845 Severity : enhancement
847 Description: Update to RHEL5.3 kernel-2.6.18-128.1.6.el5.
849 Severity : enhancement
851 Description: Update OFED release to 1.4.1 RC4
853 Severity : major, only with big OST
855 Description: Very poor metadata performance on Infiniband lustre configuration
856 Details : OST object precreation becomes very slow on big OSTs. This is due
857 to the ialloc patch spending too much time scanning groups.
860 Frequency : during recovery
862 Description: don't mix llog inodes with normal.
863 Details : allocate inodes for log in last inode group
868 Description: fix lqs' reference which won't be put in some situations
869 Details : This patch fixes:
870 1. In quota_check_common(), this function will check quota
871 for user and group, but only send one return via "pending".
872 In most cases, the pendings should be same. But that is not
874 2. If quotaoff runs between lquota_chkquota() and
875 lquota_pending_commit(), the same thing will happen too.
876 That is why it comes:
877 - if (!ll_sb_any_quota_active(qctxt->lqc_sb))
880 -------------------------------------------------------------------------------
882 2008-12-31 Sun Microsystems, Inc.
884 * Support for kernels:
885 2.6.5-7.314 (SLES 9),
886 2.6.9-67.0.22.EL (RHEL 4),
887 2.6.16.60-0.31 (SLES 10),
888 2.6.18-92.1.17.el5 (RHEL 5),
889 2.6.22.14 vanilla (kernel.org)
890 * Client support for unpatched kernels:
891 (see http://wiki.lustre.org/index.php?title=Patchless_Client)
892 2.6.16 - 2.6.22 vanilla (kernel.org)
893 * Client support for unpatched kernels:
894 we do not recommend using patchless RHEL4 clients with kernels
895 prior to 2.6.9-55EL (RHEL4U5).
896 * Recommended e2fsprogs version: 1.40.11-sun1
897 * Note that reiserfs quotas are disabled on SLES 10 in this kernel.
898 * RHEL 4 and RHEL 5/SLES 10 clients behaves differently on 'cd' to a
899 removed cwd "./" (refer to Bugzilla 14399).
900 * A new quota file format has been introduced in 1.6.5.
901 The format conversion from prior releases is handled transparently,
902 but releases older than 1.4.12/1.6.5 don't understand this new
903 format. The automatic format conversion can be avoided by running
904 the following command on the MDS:
905 'tunefs.lustre --param="mdt.quota_type=ug1" $MDTDEV'.
906 For more information, please refer to bugzilla 13904.
907 * A new quota file format was introduced in 1.6.6/1.8.0 (kernels 2.6.16+).
908 The format conversion from prior releases is handled transparently,
909 but releases older than 1.6.6/1.8.0 don't understand this new
910 format. The automatic format conversion can be avoided by running
911 the following commands on the MDS and OSS servers (for
912 pre 1.4.12-1.6.5 quota files):
913 'tunefs.lustre --param="mdt.quota_type=ug1" $MDTDEV',
914 'tunefs.lustre --param="ost.quota_type=ug1" $MDTDEV'
915 or (for 1.4.12/1.6.5 quota files)
916 'tunefs.lustre --param="mdt.quota_type=ug2" $MDTDEV',
917 'tunefs.lustre --param="ost.quota_type=ug2" $MDTDEV'
918 For more information, please refer to bugzilla 13904.
919 * Output of lfs quota has been made less detailed by default,
920 old (verbose) output can be obtained by using -v option.
921 * File join has been disabled in this release, refer to Bugzilla 16929.
922 * A new Lustre ADIO driver is available for MPICH2-1.0.7.
923 * NFS export disabled when stack size < 8192. Since the NFSv4 export of
924 Lustre filesystem with 4K stack may cause a stack overflow. For more
925 information, please refer to bugzilla 17630.
927 Severity : enhancement
929 Description: Caching OSS
930 Details : introduce data caching on the OSS. The OSS now relies on the linux
931 kernel page cache to keep recently accessed data in memory.
932 It is worth noting that all write requests are still flushed
933 synchronously as in lustre 1.6.
935 Severity : enhancement
937 Description: Version based recovery
938 Details : introduce finer grained recovery able to detect transaction
939 dependencies and can deal with transaction gaps caused by clients
940 failing at the same time as the server.
942 Severity : enhancement
944 Description: Enable adaptive timeouts by default
945 Details : The Lustre timeout value in /proc/sys/lustre/timeout is now
946 managed dynamically based on server load and should not need
947 to be tuned manually based on cluster size. This allows Lustre
948 to work under a wider variety of system sizes and loads, without
949 unnecessarily causing lengthy recovery times.
951 Severity : enhancement
953 Description: Add OST Pools support
954 Details : File striping can now be set to use an arbitrary pool of OSTs
956 Severity : enhancement
958 Description: add lazystatfs mount option to allow statfs(2) to skip down OSTs
959 Details : allow skip disconnected ost for send statfs request and hide error
963 Frequency : rare, on llog test 6
965 Descriptoin: don't allow connect to already connected import
966 Details : allowing connect to already connected import is hide connecting problem.
969 Frequency : rare, on failed llog setup
971 Descriptoin: don't leak obd reference on failed llog setup
972 Details : for failed llog setup - mgc forget call class_destroy_import for
973 client import, move destroy import to more generic place.
978 Descriptoin: allow kill process which wait statahead result
979 Details : for some reasons 'ls' can stick in waiting result from statahead,
980 in this case need way for kill this process.
983 Frequency : rare, at shutdown
985 Descriptoin: panic at umount
986 Details : llap_shrinker can be raced with killing super block from list and
987 this produce panic with access to already freeded pointer
992 Descriptoin: don't lose wakeup for imp_recovery_waitq
993 Details : recover_import_no_retry or invalidate_import and import_close can
994 both sleep on imp_recovery_waitq, but we was send only one wakeup
1000 Descriptoin: panic in mds_open
1001 Details : don't confuse mds_finish_transno() with PTR_ERR(-ENOENT)
1006 Descriptoin: stuck in cache_remove_extent() or panic with accessing to already
1008 Details : release lock refernce only after add page to pages list.
1011 Frequency : always with long access acl
1013 Descriptoin: mds can't pack reply with long acl.
1014 Details : mds don't control size of acl but they limited by reint/getattr
1019 Frequency : on remount
1021 Description: external journal device not working after the remount
1022 Details : clear dev_rdonly flag for external journal devices in
1028 Description: shutdown vs evict race
1029 Details : client_disconnect_export vs connect request race.
1030 if client will evicted at this time - we start invalidate
1031 thread without referece to import and import can be freed
1037 Description: shrink LOV EAs before replying
1038 Details : correctly adjust LOV EA buffer for reply.
1043 Description: don't skip ost target if they assigned to file
1044 Details : Drop slow OSCs if we can, but not for requested start idx.
1045 This means "if OSC is slow and it is not the requested
1046 start OST, then it can be skipped, otherwise skip it only
1047 if it is inactive/recovering/out-of-space.
1049 Severity : enhancement
1051 Description: Update to RHEL5 kernel-2.6.18-92.1.17.el5.
1054 Frequency : rare, need acl's on inode.
1056 Description: client can't handle ost additional correctly
1057 Details : if ost was added after client connected to mds client can have
1058 hit lnet_try_match_md ... to big messages to wide striped files.
1059 in this case need teach client to handle config events about add
1060 lov target and update client max ea size at that event.
1062 Severity : enhancement
1064 Description: Update to sles9 kernel-2.6.5-7.314.
1066 Severity : enhancement
1068 Description: Update to SLES10 SP2 kernel-2.6.16.60-0.31.
1071 Frequency : Create a symlink file with a very long name
1073 Description: ldlm_cancel_pack()) ASSERTION(max >= dlm->lock_count + count)
1074 Details : If there is no extra space in the request for early cancels,
1075 ldlm_req_handles_avail() returns 0 instead of a negative value.
1080 Description: mds is deadlocked
1081 Details : in rare cases, inode in catalog can have i_no less than have parent
1082 i_no, this produce wrong order for locking during open, and
1083 parallel unlink can be lock open. this need teach mds_open to grab
1084 locks in resource id order, not at parent -> child order.
1086 Severity : enhancement
1088 Description: Add /proc entry for import status
1089 Details : The mdc, osc, and mgc import directories now have
1090 an import directory that contains useful import data for debugging
1091 connection problems.
1093 Severity : enhancement
1095 Description: Re-disable certain /proc logging
1096 Details : Enable and disable client's offset_stats, extents_stats and
1097 extents_stats_per_process stats logging on the fly.
1100 Frequency : Only on FC kernels 2.6.22+
1102 Description: oops in statahead
1103 Details : Do not drop reference count for the dentry from VFS when lookup,
1104 VFS will do that by itself.
1106 Severity : enhancement
1108 Description: Generic /proc file permissions
1109 Details : Set /Proc file permissions in a more generic way to enable non-
1110 root users operate on some /proc files.
1114 Description: Hitting mdc_commit_close() ASSERTION
1115 Details : Properly handle request reference release in
1116 ll_release_openhandle().
1120 Frequency : only patchless client
1121 Description: add workaround for race between add/remove dentry from hash
1123 Severity : enhancement
1125 Description: Allow OST glimpses to return PW locks
1129 Description: LBUG when llog conf file is full
1130 Details : When llog bitmap is full, ENOSPC should be returned for plain
1135 Description: Prevent import from entering FULL state when server in recovery
1139 Description: service mount cannot take device name with ":"
1140 Details : Only when device name contains ":/" will mount treat it as
1146 Description: replace ptlrpcd with the statahead thread to interpret the async
1147 statahead RPC callback
1151 Frequency : on recovery
1152 Description: I/O failures after umount during fail back
1153 Details : if client reconnected to restarted server we need join to recovery
1154 instead of find server handler is changed and process self eviction
1155 with cancel all locks.
1157 Severity : enhancement
1159 Description: Update to RHEL5 kernel-2.6.18-92.1.10.el5.
1163 Description: Kernel BUG tries to release flock
1164 Details : Lustre does not destroy flock lock before last reference goes
1165 away. So always drop flock locks when client is evicted and
1166 perform unlock regardless of successfulness of speaking to MDS.
1168 Severity : enhancement
1170 Description: Update to SLES10 SP2 kernel-2.6.16.60-0.27.
1172 Severity : enhancement
1174 Description: Upcall on Lustre log has been dumped
1175 Details : Allow for a user mode script to be called once a Lustre log has
1176 been dumped. It passes the filename of the dumped log to the
1177 script, the location of the script can be specified via
1178 /proc/sys/lnet/debug_log_upcall.
1183 Description: avoid messages about idr_remove called for id that is not allocated
1184 Details : Move assigment s_dev for clustered nfs to end of initialization,
1185 for avoid problem with error handling.
1190 Description: avoid Already found the key in hash [CONN_UNUSED_HASH] messages
1191 Details : When connection is reused this not moved from CONN_UNUSED_HASH into
1192 CONN_USED_HASH and this prodice warning when put connection again
1198 Description: avoid ASSERTION(client_stat->nid_exp_ref_count == 0) failed
1199 Details : release reference to stats when client disconnected, not
1200 when export destroyed for avoid races when client destroyed
1201 after main ost export.
1205 Description: more cleanup in mds_lov
1206 Details : add workaround for get valid ost count for avoid warnings about
1207 drop too big messages, not init llog cat under semphore which
1208 can be blocked on reconnect and break normal replay, fix access
1211 Severity : enhancement
1213 Description: Export bytes_read/bytes_write count on OSC/OST.
1217 Description: Early reply size mismatch, MGC loses connection
1218 Details : Apply the MGS_CONNECT_SUPPORTED mask at reconnect time so
1219 the connect flags are properly negotiated.
1223 Description: Properly propagate oinfo flags from lov to osc for statfs
1224 Details : restore missing copy oi_flags to lov requests.
1228 Description: exports in /proc are broken
1229 Details : recreate /proc entries for clients when they reconnect.
1231 Severity : enhancement
1233 Description: Add man pages for llobdstat(8), llstat(8), plot-llstat(8),
1234 : l_getgroups(8), lst(8), routerstat(8)
1235 Details : included man pages for llobdstat(8), llstat(8),
1236 : plot-llstat(8), l_getgroups(8), lst(8), routerstat(8)
1238 Severity : enhancement
1240 Description: Implement lustre ll_show_options method.
1242 Severity : enhancement
1244 Description: Update to SLES9 kernel-2.6.5-7.312.
1246 Severity : enhancement
1248 Description: Update to RHEL4 kernel-2.6.9-67.0.22.EL.
1252 Description: exports in /proc are broken
1253 Details : recreate /proc entries for clients when they reconnect.
1257 Description: don't fail open with -ERANGE
1258 Details : if client connected until mds will be know about real ost count
1259 get LOV EA can be fail because mds not allocate enougth buffer
1264 Description: Resolve device initialization race
1265 Details : Prevent proc handler from accessing devices added to the
1266 obd_devs array but yet be intialized.
1268 Severity : enhancement
1270 Description: configure's --enable-quota should check the
1271 : kernel .config for CONFIG_QUOTA
1272 Details : configure is terminated if --enable-quota is passed but
1273 : no quota support is in kernel
1275 Severity : enhancement
1277 Description: Update to SLES10 SP2 kernel-2.6.16.60-0.23.
1279 Severity : enhancement
1281 Description: Update to RHEL5 kernel-2.6.18-92.1.6.el5.
1285 Frequency : rare, on PPC clients
1286 Description: don't swab ost objects in response about directory, because
1288 Details : bug similar bug 14856, but in different function.
1290 Severity : enhancement
1292 Description: lfs quota tool enhancement
1293 Details : added units specifiers support for setquota, default to
1294 current uid/gid for quota report, short quota stats by
1295 default, nonpositional parameters for setquota, added
1296 llapi_quotactl manual page.
1298 Severity : enhancement
1300 Description: *optional* service tags registration
1301 Details : if the "service tags" package is installed on a Lustre node
1302 When the filesystem is mounted, a local-node service tag will
1303 be created. See http://inventory.sun.com/ for more information
1304 about the Service Tags asset management system.
1308 Description: Client runs out of low memory
1309 Details : Consider only lowmem when counting initial number of llap pages
1312 Frequency : occasional
1314 Description: add refcount for osc callbacks, so avoid panic on shutdown
1316 Severity : enhancement
1318 Description: Update to RHEL4 kernel-2.6.9-67.0.20.
1321 Frequency : testing only
1323 Description: sanity test 65a fails if stripecount of -1 is set
1324 Details : handle -1 striping on filesystem in ll_dirstripe_verify
1327 Frequency : only in unusual configurations
1329 Description: Kernel panic with find ost index.
1330 Details : lov_obd have panic if some OST's have sparse indexes.
1332 Severity : enhancement
1334 Description: Update to RHEL5 kernel-2.6.18-53.1.21.el5.
1337 Frequency : rarely, if filesystem is mounted with -o flock
1339 Description: do not process already freed flock
1340 Details : flock can possibly be freed by another thread before it reaches
1341 to ldlm_flock_completion_ast.
1344 Frequency : rarely, if filesystem is mounted with -o flock
1346 Description: LBUG during stress test
1347 Details : Need properly lock accesses the flock deadlock detection list.
1350 Frequency : rarely, if binaries are being run from Lustre
1352 Description: oops in page fault handler
1353 Details : kernel page fault handler can return two special 'pages' in
1354 error case, don't try dereference NOPAGE_SIGBUS and NOPAGE_OMM.
1357 Frequency : rarely, during shutdown
1359 Description: timeout with invalidate import.
1360 Details : ptlrpcd_check call obd_zombie_impexp_cull and wait request which
1361 should be handled by ptlrpcd. This produce long age waiting and
1362 -ETIMEOUT ptlrpc_invalidate_import and as result LASSERT.
1368 Description: ASSERTION(CheckWriteback(page,cmd)) failed
1369 Details : badly clear PG_Writeback bit in ll_ap_completion can produce false
1373 Frequency : only with broken builds/installations
1375 Description: no LBUG if lquota.ko and fsfilt_ldiskfs.ko are different versions
1376 Details : just return an error to a user, put a console error message
1378 Severity : enhancement
1380 Description: Update to RHEL5 kernel-2.6.18-53.1.19.el5.
1382 Severity : enhancement
1384 Description: Update to RHEL4 kernel-2.6.9-67.0.15.
1386 Severity : enhancement
1388 Description: enable MGS and MDT services start separately
1389 Details : add a 'nomgs' option in mount.lustre to enable start a MDT with
1390 a co-located MGS without starting the MGS, which is a complement
1391 to 'nosvc' mount option.
1394 Frequency : always, on big-endian systems
1396 Description: cleanup in ptlrpc code, related to PPC platform
1397 Details : store magic in native order avoid panic's in recovery on PPC
1398 node and forbid from this error in future. Also fix posibily
1399 of twice swab data. Fix get lov striping to userpace.
1402 Frequency : rarely, if replay get lost on server
1404 Description: server incorrectly drop resent replays lead to recovery failure.
1405 Details : do not drop replay according to msg flags, instead we check the
1406 per-export recovery request queue for duplication of transno.
1409 Frequency : after recovery
1411 Description: precreate to many object's after del orphan.
1412 Details : del orphan st in oscc last_id == next_id and this triger growing
1413 count of precreated objects. Set flag LOW to skip increase count
1414 of precreated objects.
1417 Frequency : rare, on clear nid stats
1419 Description: ASSERTION(client_stat->nid_exp_ref_count == 0)
1420 Details : when clean nid stats sometimes try destroy live entry,
1421 and this produce panic in free.
1424 Frequency : occasionally since 1.6.4
1426 Description: Stack overflow during MDS log replay
1427 Details : ease stack pressure by using a thread dealing llog_process.
1430 Frequency : very rare
1432 Description: MDT cannot be unmounted, reporting "Mount still busy"
1433 Details : Mountpoint references were being leaked during open reply
1434 reconstruction after an MDS restart. Drop mountpoint reference
1435 in reconstruct_open() and free dentry reference also.
1440 Description: wait until IO finished before start new when do lock cancel.
1441 Details : VM protocol want old IO finished before start new, in this case
1442 need wait until PG_writeback is cleared until check dirty flag
1443 and call writepages in lock cancel callback.
1448 Description: mds_mfd_close() ASSERTION(rc == 0)
1449 Details : In mds_mfd_close(), we need protect inode's writecount change
1450 within its orphan write semaphore to prevent possible races.
1453 Frequency : rare, on shutdown ost
1455 Description: don't hit live lock with umount ost.
1456 Details : shrink_dcache_parent can be in long loop with destroy dentries,
1457 use shrink_dcache_sb instead.
1460 Frequency : only when echo_client is used
1462 Description: don't panic with use echo_client
1463 Details : echo client pass NULL as client nid pointer and this produce NULL
1464 pointer dereference.
1467 Frequency : Always on 32-bit PowerPC systems
1469 Description: fix build on PPC32
1470 Details : compile code with -m64 flag produce wrong object file for PPC32.
1475 Description: MDS LBUG: ASSERTION(!IS_ERR(dchild))
1476 Details : In reconstruct_* functions, LASSERTs on both the data supplied
1477 by a client, and the data on disk are dangerous and incorrect.
1478 Change them with client eviction.
1480 Severity : enhancement
1482 Description: skiplist implementation simplification
1483 Details : skiplists are used to group compatible locks on granted list
1484 that was implemented as tracking first and last lock of each lock
1485 group the patch changes that to using doubly linked lists
1489 Description: delete compatibility for 32bit qdata
1490 Details : as planned, when lustre is beyond b1_8, lquota won't support 32bit
1491 qunit. That means servers of b1_4 and servers of b1_8 can't be
1492 used together if users want to use quota.
1495 Frequency : only with administrator action
1497 Description: mount failure if config log has invalid conf_param setting
1498 Details : If administrator specified an incorrect configuration parameter
1499 with "lctl conf_param" this would cause an error during future
1500 client mounts. Instead, ignore the bad configuration parameter.
1503 Frequency : blocks per group < blocksize*8 and uninit_groups is enabled
1505 Description: ldiskfs error: XXX blocks in bitmap, YYY in gd
1506 Details : If blocks per group is less than blocksize*8, set rest of the
1510 Frequency : Application do stride read on lustre
1512 Description: The read performance will drop a lot if the application does
1514 Details : Because the stride_start_offset are missing in stride read-ahead,
1515 it will cause clients read a lot of unused pages in read-ahead,
1516 then the read-performance drops.
1520 Description: more ldlm soft lockups
1521 Details : In ldlm_resource_add_lock(), call to ldlm_resource_dump()
1522 starve other threads from the resource lock for a long time in
1523 case of long waiting queue, so change the debug level from
1524 D_OTHER to the less frequently used D_INFO.
1526 Severity : enhancement
1528 Description: add -gid, -group, -uid, -user options to lfs find
1530 Severity : enhancement
1532 Description: ll_recover_lost_found_objs - recover objects in lost+found
1533 Details : OST corruption and subsequent e2fsck can leave objects in the
1534 lost+found directory. Using the "ll_recover_lost_found_objs"
1535 tool, these objects can be retrieved and data can be salvaged
1536 by using the object ID saved in the fid EA on each object.
1541 Description: this bug _only_ happens when inode quota limitation is very low
1542 (less than 12), so that inode quota unit is 1 at initialization.
1543 Details : if remaining quota equates 1, it is a sign to demonstate that quota
1544 is effective now. So least quota qunit should be 2.
1548 Description: Hung threads in invalidate_inode_pages2_range
1549 Details : The direct IO path doesn't call check_rpcs to submit a new RPC once
1550 one is completed. As a result, some RPCs are stuck in the queue
1555 Description: Procfs and llog threads access destoryed import sometimes.
1556 Details : Sync the import destoryed process with procfs and llog threads by
1557 the import refcount and semaphore.
1561 Description: mds fails to respond, threads stuck in ldlm_completion_ast
1562 Details : Sort source/child resource pair after updating child resource.
1567 Description: kernel BUG at ldiskfs2_ext_new_extent_cb
1568 Details : If insertion of an extent fails, then discard the inode
1569 preallocation and free data blocks else it can lead to duplicate
1574 Description: don't always update ctime in ext3_xattr_set_handle()
1575 Details : Current xattr code updates inode ctime in ext3_xattr_set_handle()
1576 In some cases the ctime should not be updated, for example for
1577 2.0->1.8 compatibility it is necessary to delete an xattr and it
1578 should not update the ctime.
1582 Description: add quota statistics
1583 Details : 1. sort out quota proc entries and proc code.
1584 2. add quota statistics
1589 Description: quotas are not honored with O_DIRECT
1590 Details : all writes with the flag O_DIRECT will use grants which leads to
1591 this problem. Now using OBD_BRW_SYNC to guard this.
1595 Bugzilla : 15713/16362
1596 Description: Assertion in iopen_connect_dentry in 1.6.3
1597 Details : looking up an inode via iopen with the wrong generation number can
1598 populate the dcache with a disconneced dentry while the inode
1599 number is in the process of being reallocated. This causes an
1600 assertion failure in iopen since the inode's dentry list contains
1601 both a connected and disconnected dentry.
1605 Description: assertion failure in ldlm_handle2lock()
1606 Details : fix a race between class_handle_unhash() and class_handle2object()
1607 introduced in lustre 1.6.5 by bug 13622.
1609 Severity : enhancement
1611 Description: superblock lock contention with many SMP cores on one client
1612 Details : several client filesystem locks were highly contended on SMP
1613 NUMA systems with 8 or more cores. Per-CPU datastructures
1614 and more efficient locking implemented to reduce contention.
1619 Description: Kernel BUG: sd_iostats_bump: unexpected disk index
1620 Details : remove the limit of 256 scsi disks in the sd_iostat patch
1625 Description: oops in sd_iostats_seq_show()
1626 Details : unloading/reloading the scsi low level driver triggers a kernel
1627 bug when trying to access the sd iostat file.
1632 Description: Kernel panics during QLogic driver reload
1633 Details : REQ_BLOCK_PC requests are not handled properly in the sd iostat
1634 patch, causing memory corruption.
1639 Description: journal_dev option does not work in b1_6
1640 Details : pass mount option during pre-mount.
1642 Severity : enhancement
1644 Description: Add a FIEMAP(FIle Extent MAP) ioctl for ldiskfs
1645 Details : FIEMAP ioctl will allow an application to efficiently fetch the
1646 extent information of a file. It can be used to map logical blocks
1647 in a file to physical blocks in the block device.
1650 Frequency : only with adaptive timeout enabled
1652 Description: DEBUG_REQ() bad paging request
1653 Details : ptlrpc_at_recv_early_reply() should not modify req->rq_repmsg
1654 because it can be accessed by reply_in_callback() without the
1658 Frequency : only on Cray X2
1660 Description: X2 build failures
1661 Details : fix build failures on Cray X2.
1665 Description: xid & resent requests
1666 Details : Initialize RPC XID from clock at startup (randomly if clock is
1671 Description: quota recovery deadlock during mds failover
1672 Details : This patch includes att18982, att18236, att18237 in bz14840.
1674 1. fix osts hang when mds does failover with quotaon
1675 2. prevent watchdog storm when osts threads wait for the
1680 Description: kernel panic on racer
1681 Details : Do not access dchild->d_inode when IS_ERR(dchild) is true.
1683 Severity : enhancement
1685 Description: Add lustre_start utility to start or stop multiple Lustre servers
1690 Description: Lustre GPF in {:ptlrpc:ptlrpc_server_free_request+373}
1691 Details : In case of memory pressure, list_del() can be called twice on
1692 req->rq_history_list, causing a kernel oops.
1696 Description: (ptllnd_peer.c:557:kptllnd_peer_check_sends()) ASSERTION(!in_interrupt()) failed
1697 Details : fix stack overflow in the distributed lock manager by defering
1698 export eviction after a failed ast to the elt thread instead of
1699 handling it in the dlm interpret routine.
1701 Severity : enhancement
1703 Description: More exported tunables for mballoc
1704 Details : Add support for tunable preallocation window and new tunables for
1705 large/small requests
1709 Description: Detect corruption of block bitmap and checking for preallocations
1710 Details : Checks validity of on-disk block bitmap. Also it does better
1711 checking of number of applied preallocations. When corruption is
1712 found, it turns filesystem readonly to prevent further corruptions.
1716 Frequency : only for big-endian servers
1717 Description: Check for big-endian system while mounting fs with extents feature
1718 Details : Mounting a filesystem with extents feature will fail on big-endian
1719 systems since ext3-based ldiskfs is not supported on big-endian
1720 systems. Can be over-ridden with "bigendian_extents" mount option.
1724 Description: Excessive recovery window
1725 Details : With AT enabled, the recovery window can be excessively long (6000+
1726 seconds). To address this problem, we no longer use
1727 OBD_RECOVERY_FACTOR when extending the recovery window (the connect
1728 timeout no longer depends on the service time, it is set to
1729 INITIAL_CONNECT_TIMEOUT now) and clients report the old service
1730 time via pb_service_time.
1733 Descriptoin: Don't sync journal after every i/o
1734 Details : Implement write RPC replay to allow server replies for write RPCs
1735 before data is on disk. However, this feature is disabled by
1736 default since some issues leading to data corruptions have been
1737 found during recovery (e.g. bug 19128). This feature can be enabled
1738 by running the following command on the OSSs:
1739 lctl set_param obdfilter.*.sync_journal=0
1743 Description: Watchdog triggered on MDS failover
1744 Details : enable OBD_CONNECT_MDT flag when connecting from the MDS so that
1745 the OSTs know that the MDS "UUID" can be reused for the same export
1746 from a different NID, so we do not need to wait for the export to
1751 Descriptoin: Lustre detected file system corruption with inode out of bounds
1752 Details : don't update i_size on MDS_CLOSE for directories. This causes
1753 directory corruptions on the MDT.
1755 -------------------------------------------------------------------------------
1757 2008-05-26 Sun Microsystems, Inc.
1759 * Support for kernels:
1760 2.6.5-7.311 (SLES 9),
1761 2.6.9-67.0.7.EL (RHEL 4),
1762 2.6.16.54-0.2.5 (SLES 10),
1763 2.6.18-53.1.14.el5 (RHEL 5),
1764 2.6.22.14 vanilla (kernel.org)
1765 * Client support for unpatched kernels:
1766 (see http://wiki.lustre.org/index.php?title=Patchless_Client)
1767 2.6.16 - 2.6.22 vanilla (kernel.org)
1768 * Due to problems with nested symlinks and FMODE_EXEC (bug 12652),
1769 we do not recommend using patchless RHEL4 clients with kernels
1770 prior to 2.6.9-55EL (RHEL4U5).
1771 * Recommended e2fsprogs version: 1.40.7-sun1
1772 * Note that reiserfs quotas are disabled on SLES 10 in this kernel.
1773 * RHEL 4 and RHEL 5/SLES 10 clients behaves differently on 'cd' to a
1774 removed cwd "./" (refer to Bugzilla 14399).
1775 * A new quota file format has been introduced in 1.6.5.
1776 The format conversion from prior releases is handled transparently,
1777 but releases older than 1.4.12/1.6.5 will not understand this new
1778 format. The automatic format conversion can be avoided by running
1779 the following command on the MDS before upgrading:
1780 'tunefs.lustre --param="mdt.quota_type=ug1" $MDTDEV'.
1781 For more information, please refer to bugzilla 13904.
1785 Description: quota performance fix
1786 Details : quota data is written in journalled mode instead of ordered to
1787 increase performance
1791 Description: lfs support for human-readable quota grace time strings
1792 Details : lfs setquota -t and lfs quota -t represent quota grace times
1793 in "XXwXXdXXhXXmXXs" format instead of large values in seconds
1796 Frequency : always with o2ib 1.3 and sles10
1798 Description: fix build with SLES10 and o2ib v3.
1799 Details : sles10 uses diffrent name for Module.symver file but configure
1800 assume this file has same name on RHEL/SLES/vanila kernels.
1803 Frequency : very rare, if additional xattrs are used on kernels >= 2.6.12
1805 Description: MDS may lose file striping (and hence file data) in some cases
1806 Details : If there are additional extended attributes stored on the MDS,
1807 in particular ACLs, SELinux, or user attributes (if user_xattr
1808 is specified for the client mount options) then there is a risk
1809 of attribute loss. Additionally, the Lustre file striping
1810 needs to be larger than default (e.g. striped over all OSTs),
1811 and an additional attribute must be stored initially in the
1812 inode and then increase in size enough to be moved to the
1813 external attribute block (e.g. ACL growing in size) for file
1816 Severity : enhancement
1818 Description: add message levels for liblustreapi
1823 Description: MDT cannot be unmounted, reporting "Mount still busy"
1824 Details : Mountpoint references were being leaked during open reply
1825 reconstruction after an MDS restart. Drop mountpoint reference
1826 in reconstruct_open() and free dentry reference also.
1831 Description: fix for occasional failure case of -ENOSPC in recovery-small tests
1832 Details : Move the 'good_osts' check before the 'total_bavail' check. This
1833 will result in an -EAGAIN and in the exit call path we call
1834 alloc_rr() which will with increasing aggressiveness attempt to
1835 aquire precreated objects on the minimum number of required OSCs.
1839 Description: Use old size assignment to avoid deadlock
1840 Details : This reverts the changes in bugs 2369 and bug 14138 that introduced
1841 the scheduling while holding a spinlock. We do not need locking
1842 for size in ll_update_inode() because size is only updated from
1843 the MDS for directories or files without objects, so there is no
1844 other place to do the update, and concurrent access to such inodes
1845 are protected by the inode lock.
1849 Description: Use __u64 instead of int for valid bits
1853 Description: resolve "_IOWR redefined" build error on SLES10
1857 Description: dump the memory debugging after all modules are unloaded to
1858 suppress false negative in conf_sanity test 39
1862 Description: the recovery timer never expires
1863 Details : for new client connect request, the recovery timer should not be
1864 reset, otherwise recovery timer will never expired, if the old
1865 client never come. Only old client connect and first connection
1866 req should trigger recovery timer reset.
1870 Description: the min numbers of lproc stats are wrong
1871 Details : adding a new constant LC_MIN_INIT and use it for initialization
1875 Frequency : always with interactive lfs
1877 Description: Reinitialize optind to 0 so that interactive lfs works in all cases
1880 Frequency : with multiple concurrent readdir processes in same directory
1881 Bugzilla : 15406, 15169, 15175
1882 Description: misc fixes for directory readahead.
1883 Details : prevent previous statahead async RPC callback from processing the
1884 current "statahead_info", race condition between async RPC callback
1885 add dentry into dentry hash table and "ls" thread revalidate such
1886 dentry, statahead his/miss control for hidden items, and so on.
1888 Severity : enhancement
1890 Description: build kernel-ib packages for OFED 1.3 in our release cycle
1894 Description: incore types cleaning in quota code (with respect to 64-bit limits)
1895 Details : several u32 variables declarations are replaced with u64 declarations
1900 Description: fix SLES kernel versioning
1901 Details : the kernel version for our SLES 10 kernel did not include a "-"
1902 before the "smp" at the end. while this was not a problem in
1903 general, it did mean that software trying to use the kernel
1904 version to try to detect a vendor specific kernel would fail.
1905 this was most evident by the OFED build scripts.
1910 Description: Don't update lov_desc members until making sure they are valid
1911 Details : When updating lov_desc members via proc fs, need fix their
1912 validities before doing the real update.
1915 Frequency : very rare
1917 Description: don't put request into delay list while invalidate in flight.
1918 Details : ptlrpc_delay_request sometimes put in delay list while invalidate
1919 import in flight. this produce timeout for invalidate and sometimes
1920 can cause stale data.
1922 Severity : enhancement
1924 Description: Update kernel to SLES9 2.6.5-7.311.
1926 Severity : enhancement
1928 Description: Update kernel to RHEL4 2.6.9-67.0.7.
1933 Frequency : on PPC only
1934 Description: not convert ost objects for directory because it's not exist.
1935 Details : ll_dir_getstripe assume dirrectory has ost objects but this wrong.
1937 Severity : enhancement
1939 Description: Fix warnings with compile liblustre at sles10/rhel5 which have
1940 __u64 as usingied long long type.
1943 Frequency : rare, on shutdown
1945 Description: race process ast vs remove callback
1946 Details : removing callback before disconnect import open race with
1947 processing callback.
1949 Severity : enhancement
1951 Description: Update kernel to SLES9 2.6.5-7.311.
1953 Severity : enhancement
1955 Description: Files open for execute are not marked busy on SLES10
1956 Details : Add FMODE_EXEC to SLES10 SP1 server kernel series.
1958 Severity : enhancement
1960 Description: Add server support for vanilla-2.6.22.14.
1963 Frequency : occasional
1965 Description: Avoid lov_create() getting stuck in obd_statfs_rqset()
1966 Details : If an OST is down the MDS will hang indefinitely in
1967 obd_statfs_rqset() waiting for the statfs data. While for
1968 MDS QOS usage of statfs, it should not stuck in waiting.
1970 Severity : enhancement
1972 Description: Disable adaptive timeouts by default
1975 Frequency : on network error
1977 Description: panic with double free request if network error
1978 Details : mdc_finish_enqueue is finish request if any network error occuring,
1979 but it's true only for synchronus enqueue, for async enqueue
1980 (via ptlrpcd) this incorrect and ptlrpcd want finish request
1984 Frequency : rare, on recovery
1986 Description: read procfs can produce deadlock in some situation
1987 Details : Holding lprocfs lock with send rpc can produce block for destroy
1988 obd objects and this also block reconnect with -EALREADY.
1989 This isn't fix all lprocfs bugs - but make it rare.
1991 Severity : enhancement
1993 Description: Update kernel to RHEL5 2.6.18-53.1.14.el5.
1996 Frequency : frequent on X2 node
1998 Description: mdc_set_open_replay_data LBUG
1999 Details : Set replay data for requests that are eligible for replay.
2004 Description: lustre_mgs: operation 101 on unconnected MGS
2005 Details : When MGC is disconnected from MGS long enough, MGS will evict the
2006 MGC, and late on MGC cannot successfully connect to MGS and a lot
2007 of the error messages complaining that MGS is not connected.
2010 Frequency : rare, depends on device drivers and load
2012 Description: MDS or OSS nodes crash due to stack overflow
2013 Details : Code changes in 1.6.4 increased the stack usage of some functions.
2014 In some cases, in conjunction with device drivers that use a lot
2015 of stack, the MDS (or possibly OSS) service threads could overflow
2016 the stack. One change which was identified to consume additional
2017 stack has been reworked to avoid the extra stack usage.
2019 Severity : enhancement
2021 Description: Update to RHEL5 latest kernel-2.6.18-53.1.13.el5.
2023 Severity : enhancement
2025 Description: Update to SLES10 SP1 latest kernel-2.6.16.54-0.2.5.
2027 Severity : enhancement
2029 Description: Update to RHEL5 latest kernel-2.6.18-53.1.6.el5.
2031 Serverity : enhancement
2033 Description: Update RHEL4 kernel to 2.6.9-67.0.4.
2036 Frequency : rare on shutdown OST
2038 Description: Don't allow skipping OSTs if index has been specified.
2039 Details : Don't allow skipping OSTs if index has been specified, make locking
2040 in internal create lots better.
2045 Description: ASSERTION(!PageDirty(page)) failed
2046 Details : Wrong check could lead to an assertion failure under specific
2052 Description: LBUG in ptlrpc_check_set() bad phase ebc0de00
2053 Details : access to bitfield in structure is always rounded to long
2054 and this produce problem with not atomic change any bit.
2059 Description: Lustre make rpms failed.
2060 Details : Remove ldiskfs spec file to avoids rpmbuild be confused when
2061 builds Lustre rpms from tarball.
2063 Severity : enhancement
2065 Description: Update to SLES9 SP4 kernel-2.6.5-7.308.
2068 Frequency : rare on shutdown OST
2070 Description: If llog cancel was not send before clean_exports phase, this can
2071 produce deadlock in llog code.
2072 Details : If llog thread has last reference to obd and call class_import_put
2073 this produce deadlock because llog_cleanup_commit_master wait when
2074 last llog_commit_thread exited, but this never success because was
2075 called from llog_commit_thread.
2078 Frequency : only if OST index is skipped
2080 Description: NULL lov_tgts causing MDS oops
2081 Details : more safe checks for NULL lov_tgts for avoid oops.
2083 Severity : enhancement
2085 Description: Update to RHEL4 latest kernel-2.6.9-67.0.1.EL.
2087 Severity : enhancement
2089 Description: Update to RHEL5 latest kernel-2.6.18-53.1.4.el5.
2094 Description: make mgs_setparam() handle fsname containing dash
2095 Details : fsname containing a dash does not work with lctl conf_param
2097 Severity : enhancement
2099 Description: Update to RHEL4 Update-6 kernel-2.6.9-67.EL.
2102 Frequency : rare, in recovery and (or) destroy lovobjid file.
2104 Description: rewrite lov objid code.
2105 Details : Cleanup for lov objid code, remove scability problems and wrong
2106 locking. Fix sending last_id into ost.
2108 Severity : enhancement
2110 Description: Update to SLES10 SP1 latest kernel-2.6.16.54-0.2.3.
2112 Severity : enhancement
2114 Description: Update to RHEL5 Update-1 kernel 2.6.18-53.el5.
2115 Details : Use d_move_locked instead of __d_move.
2118 Frequency : rare, at shutdown
2120 Description: access already free / zero obd_namespace.
2121 Details : if client_disconnect_export was called without force flag set,
2122 and exist connect request in flight, this can produce access to
2123 NULL pointer (or already free pointer) when connect_interpret
2124 store ocd flags in obd_namespace.
2127 Frequency : only at startup
2129 Description: not alloc memory with spinlock held.
2130 Details : allocation memory with GFP_KERNEL can produce sleep deadlock,
2131 if any spinlock held.
2136 Description: lfs find does not continue on file error
2137 Details : Continue other files processing when a file/dir is absent.
2141 Description: Inconsistent usage of lustre_pack_reply()
2142 Details : Standardize the usage of lustre_pack_reply() such that it
2143 always generate a CERROR on failure.
2146 Frequency : very rare
2148 Description: Fix replay if there is an un-replied request and open
2149 Details : In some cases, older replay request will revert the
2150 mcd->mcd_last_xid on MDS which is used to record the client's
2151 latest sent request.
2153 Severity : enhancement
2155 Description: Update to RHEL5 kernel 2.6.18-8.1.15.el5.
2157 Severity : enhancement
2159 Description: Update to SLES10 SP1 kernel 2.6.16.53-0.16
2161 Severity : enhancement
2163 Description: Update to SLES9 kernel-2.6.5-7.287.3.
2165 Severity : enhancement
2167 Description: Update to RHEL4 kernel-2.6.9-55.0.12.EL.
2169 Severity : enhancement
2171 Description: Build SLES10 patchless client fails
2172 Details : The configure was broken by run ./configure with
2173 --with-linux-obj=.... argument for patchless client. When the
2174 configure use --with-linux-obj, the LINUXINCLUDE= -Iinclude
2175 can't search header adequately. Use absolute path such as
2176 -I($LINUX)/include instead.
2178 Severity : enhancement
2180 Description: Lustre Page Accounting
2181 Details : New macros for page alloc and free which enable accounting
2182 of page allocation of Lustre. Use percpu counters to store memory
2183 and page statistics.
2186 Frequency : only if debugging is disabled
2188 Description: LASSERT_{REQ,REP}SWAB macros are buggy
2189 Details : If SWAB_PARANOIA is disabled, the LASSERT_REQSWAB and
2190 LASSERT_REPSWAB macros become no-ops, which is incorrect. Drop
2191 these macros and replace them with their definitions instead.
2196 Description: interrupt oig_wait produce painc on resend.
2197 Details : brw_redo_request can be used for resend requests from ptlrpcd and
2198 private set, and this produce situation when rq_ptlrpcd_data not
2199 copyed to new allocated request and triggered LBUG on assert
2200 req->rq_ptlrpcd_data != NULL. But this member used only for wakeup
2201 ptlrpcd set if request is changed and can be safety changed to use
2204 Severity : enhancement
2206 Description: organize the server-side client stats on per-nid basis
2207 Details : Change the structure of stats under obdfilter and mds to
2215 The "uuid"s file would list the uuids of _active_ exports.
2216 And the clear entry is to clear all stats and stale nids.
2221 Description: Processes looping in ll_readdir() on Lustre clients finally causing
2222 a full node pseudo-hang
2223 Details : Concurrent access to the same directory from multiple clients with
2224 intensive file creation/removal can cause a client node to spin in
2225 ll_readdir(). i_version must be increased every time the lock is
2226 cancelled to ensure a revalidate is done.
2231 Description: touch file failed when fs is not full
2232 Details : OST in recovery should not be discarded by MDS in alloc_qos(),
2233 otherwise we can get ENOSP while fs is not full.
2236 Frequency : only for Cray XT3
2237 Bugzilla : 12829/13455
2238 Description: Changing primary group doesn't change the group lustre assigns to
2240 Details : When CRAY_XT3 is defined, the fsgid supplied by the client is
2241 overridden with the primary group provided by the group upcall,
2242 whereas the supplied fsgid can be trusted if it is in the list of
2243 supplementary groups returned by the group upcall.
2245 Severity : enhancement
2247 Description: Root Squash Functionality
2248 Details : Implementation of NFS-like root squash capability. Specifically,
2249 don't allow someone with root access on a client node to be able
2250 to manipulate files owned by root on a server node.
2252 Severity : enhancement
2254 Description: Slow trucate/writes to huge files at high offsets.
2255 Details : Directly associate cached pages to lock that protect those pages,
2256 this allows us to quickly find what pages to write and remove
2257 once lock callback is received.
2262 Description: Too many locks accumulating on client during NFS usage
2263 Details : mds_open improperly used accmode to find out access mode to a
2264 file. Also mdc_intent_lock logic to find out if we already have
2265 lock similar to just received was flawed since introduction of
2266 skiplists - locks are now added to the front of the granted
2271 Description: Hit ASSERTION(obd->obd_stopping == 1) failed in some setup failed
2273 Details : In obd setup failure handler, obd_stopping will not necessarily to
2274 be 1, and obd_set_up should also be checked to make sure whether
2275 obd is completely setup.
2277 Severity : enhancement
2279 Description: Allow masking D_WARNING, D_ERROR messages from console
2280 Details : Console messages can now be disabled via lnet.printk.
2285 Description: User code with malformed file open parameter crashes client node
2286 Details : Before packing join_file req, all the related reference should be
2287 checked carefully in case some malformed flags cause fake join_file
2293 Description: shrink/enlarge qunit size when needed; fix the problem of coarse
2294 grain of quota doing harm to quota's accuracy
2295 Details : qunit size will be changed when quota limitation is too low/high;
2296 record the pending quota write in order to get more accureate
2297 quota; delete the patch for bug12588, which is unnecessary when
2298 this patch is landed. This bug also contains fixes for bug 14526,
2299 14299, 14601 and 13794.
2303 Description: LDLM_ENQUEUE races with LDLM_CP_CALLBACK
2304 Details : ldlm_completion_ast() assumes that a lock is granted when the req
2305 mode is equal to the granted mode. However, it should also check
2306 that LDLM_FL_CP_REQD is not set.
2310 Description: Heavy nfs access might result in deadlocks
2311 Details : After ELC code landed, it is now improper to enqueue any mds
2312 locks under och_sem, because enqueue might want to decide to
2313 cancel open locks for same inode we are holding och_sem for.
2317 Description: 35% write performance drop with ldiskfs2 when quotas are on
2318 Details : Enable ext3 journalled quota by default to improve performance
2319 when quotas are turned on.
2323 Description: Client eviction while running blogbench
2324 Details : A lot of unlink operations with concurrent I/O can lead to a
2325 deadlock causing evictions. To address the problem, the number of
2326 oustanding OST_DESTROY requests is now throttled to
2327 max_rpcs_in_flight per OSC and LDLM_FL_DISCARD_DATA blocking
2328 callbacks are processed in priority.
2331 Frequency : RHEL4 only
2333 Description: mkfs is very slow on IA64/RHEL4
2334 Details : A performance regression has been discovered in the MPT Fusion
2335 driver between versions 3.02.73rh and 3.02.99.00rh. As a
2336 consequence, we have downgraded the MPT Fusion driver in the RHEL4
2337 kernel from 3.02.99.00 to 3.02.73 until this problem is fixed.
2340 Frequency : PPC/PPC64 only
2342 Description: conflicts between asm-ppc64/types.h and lustre_types.h
2343 Details : fix duplicated definitions between asm-ppc64/types.h and
2344 lustre_types.h on PPC.
2347 Frequency : PPC/PPC64 only
2349 Description: asm-ppc/segment.h does not exist
2350 Details : fix compile issue on PPC.
2354 Description: data checksumming impacts single node performance
2355 Details : add support for several checksum algorithms. Currently, CRC32 and
2356 Adler-32 are supported. The checksum type can be changed on the fly
2357 through /proc/fs/lustre/osc/*/checksum_type.
2361 Description: use adler32 for page checksums
2362 Details : when available, use the Adler-32 algorithm instead of CRC32 for
2367 Description: better handle error messages in extents code
2369 Severity : enhancement
2371 Description: SNMP support enhancement
2372 Details : Adding total number of sampled request for an MDS node in snmp
2375 Severity : enhancement
2377 Description: Optimize ldlm waiting list processing for PR extent locks
2378 Details : When processing waiting list for read extent lock and meeting read
2379 lock that is same or wider to it that is not contended, skip
2380 processing rest of the list and immediatelly return current
2381 status of conflictness, since we are guaranteed there are no
2382 conflicting locks in the rest of the list.
2386 Description: Time out and refuse to reconnect
2387 Details : When the failover node is the primary node, it is possible
2388 to have two identical connections in imp_conn_list. We must
2389 compare not conn's pointers but NIDs, otherwise we
2390 can defeat connection throttling.
2394 Description: Client not clear own cache if answer to reconnect is lost.
2395 Details : Client gets evicted from server. Now client also thinks it is
2396 disconnected (or gets ENOTCONN on its operation) and decides to
2397 reconnect. Server receives reconnect message, but cannot find
2398 export. New export is created that is fully valid (new cookie!),
2399 but reply is lost and not reported to client. Client reconnects
2400 again and gets back a just-created connection, but it is not new
2401 so client thinks it was not evicted and does not do recovery.
2405 Description: Detect stride IO mode in read-ahead
2406 Details : When a client does stride read, read-ahead should detect that and
2407 read-ahead pages according to the detected stride pattern.
2411 Description: build for x2 fails
2412 Details : fix compile issue on Cray systems.
2414 Severity : enhancement
2416 Description: implement readv/writev APIs(aio_read/aio_writes in newer kernels)
2417 Details : This greatly improves speed of NFS writes on 2.6 kernels.
2420 Frequency : only on PPC/SLES10
2422 Description: "BITS_PER_LONG is not 32 or 64" in linux/idr.h
2423 Details : On SLES10/PPC, fs.h includes idr.h which requires BITS_PER_LONG to
2424 be defined. Add a hack in mkfs_lustre.c to work around this compile
2429 Description: LASSERT on MDS when client holding flock lock dies
2430 Details : ldlm pool logic depends on number of granted locks equal to
2431 number of released locks which is not true for flock locks, so
2432 just exclude such locks from consideration.
2436 Description: MDS deadlock with many ll_sync_lov threads and I/O stalled
2437 Details : Use fsfilt_sync() for both the whole filesystem sync and
2438 individual file sync to eliminate dangerous inode locking
2439 with I_LOCK that can lead to a deadlock.
2443 Description: Update an obsolete wirecheck.c generator
2444 Details : Update wirecheck.c/wirehdr.c and regenerate wiretest.c
2448 Description: Client can panic on open sometimes
2449 Details : It is possible that we try to free already freed request in
2450 ll_file_open in some error cases when we send request from
2455 Description: performance in 1.6.3
2456 Details : Force q->max_phys_segments to MAX_PHYS_SEGMENTS on SLES10 to be
2457 sure that 1MB requests are not fragmented by the block layer.
2461 Description: LDLM soft lockups - improvement
2462 Details : It is be possible to send the lock handle along with each read
2463 or write request because the client is already doing a lock match
2464 itself so there isn't any reason the OST should have to re-do that
2470 Description: lfs quota fails with deactivated OSTS
2471 Details : With this patch, three improvements are included:
2472 1. detete the softlimit in mds and osts when use "lfs quota".
2473 2. display the inaccurate data in the output of "lfs quota".
2474 3. try to get quota info when "lfs quota" is executed.
2479 Description: Extent locks not granted with no conflicts sometimes.
2480 Details : When race occurs in glimpse handler and nothing is returned,
2481 we do not reprocess the queue after lock cancel, and that leads
2482 to a stall until next activity on a resource
2485 Frequency : failover with quotaon
2487 Description: during mds failovers with quota on, OSTs got into deadlock state
2488 and causing dumpstack.
2489 Details : for every quota slave, at any time, there is only one quota req
2490 is sent to quota master for every uid/gid. Before that quota req
2491 returns, all the thread relative to the same uid/gid will wait.
2492 So if the quota req is lost because mds failovers or any other
2493 reasons, this bug will be hit. Now, dqacq_interpret() will handle
2494 quota reqs who time out.
2496 Severity : enhancement
2499 Description: when quota slave checks if quota is enough, there is an unnecessary
2501 Details : place this wait on necessary place instead of always waiting.
2503 --------------------------------------------------------------------------------
2505 2007-12-07 Cluster File Systems, Inc. <info@clusterfs.com>
2507 * Support for kernels:
2508 2.6.5-7.286 (SLES 9),
2509 2.6.9-55.0.9.EL (RHEL 4),
2510 2.6.16.53-0.8 (SLES 10),
2511 2.6.18-8.1.14.el5 (RHEL 5),
2512 2.6.18.8 vanilla (kernel.org)
2513 * Client support for unpatched kernels:
2514 (see http://wiki.lustre.org/index.php?title=Patchless_Client)
2515 2.6.16 - 2.6.22 vanilla (kernel.org)
2516 * Due to recently discovered recovery problems, we do not recommend
2517 using patchless RHEL 4 clients with this or any earlier release.
2518 * Recommended e2fsprogs version: 1.40.2-cfs1
2519 * Note that reiserfs quotas are disabled on SLES 10 in this kernel.
2522 Frequency : occasional
2524 Description: MDS hang or stay in waiting lock
2525 Details : If client receive lock with CBPENDING flag ldlm need send lock
2526 cancel as separate rpc, to avoid situation when cancel request
2527 can't processed due all i/o threads stay in wait lock.
2530 Frequency : occasional
2532 Description: Do not fail import if osc_interpret_create gets -EAGAIN
2533 Details : If osc_interpret_create got -EAGAIN it immediately exits and
2534 wakeup oscc_waitq. After wakeup oscc_wait_for_objects call
2535 oscc_has_objects and see OSC has no objests and call
2536 oscc_internal_create to resend create request.
2538 Severity : enhancement
2540 Description: Update kernel patches for SLES10 2.6.16.53-0.8.
2541 Details : Update which_patch & target file for SLES10 latest kernel.
2543 Severity : enhancement
2545 Description: add --type and --size parameters to lfs find
2546 Details : Enhance lfs find by adding filetype and filesize parameters. Also
2547 multiple OBDs can now be specified for the --obd option.
2549 Severity : enhancement
2551 Description: eliminate client locks in face of contention
2552 Details : file contention detection and lockless i/o implementation
2553 for contended files.
2555 Severity : enhancement
2557 Description: Remove client patches from SLES 10 kernel.
2558 Details : This causes SLES 10 clients to behave as patchless clients
2559 even on a Lustre-patched (server) kernel.
2561 Severity : enhancement
2563 Description: use i_size_read and i_size_write in 2.6 port
2564 Details : replace inode->i_size access with i_size_read/write()
2567 Frequency : when removing large files
2569 Description: scheduling issue during removal of large Lustre files
2570 Details : Don't take the BKL in fsfilt_ext3_setattr() for 2.6 kernels.
2571 It causes scheduling issues when removing large files (17TB in the
2577 Description: 1.4.11 Can't handle directories with stripe set and extended ACLs
2578 Details : Impossible (EPROTO is returned) to access a directory that has a
2579 non-default striping and ACLs.
2582 Frequency : only on ppc
2584 Description: /proc/fs/lustre/devices broken on ppc
2585 Details : The patch as applied to 1.6.2 doesn't look correct for all arches.
2586 We should make sure the type of 'index' is loff_t and then cast
2587 explicitly as needed below. Do not assign an explicitly cast
2591 Frequency : only for rhel5
2593 Description: Kernel patches update for RHEL5 2.6.18-8.1.10.el5.
2594 Details : Modify the target file & which_kernel.
2597 Frequency : if the uninit_groups feature is enabled on ldiskfs
2599 Description: e2fsck reports "invalid unused inodes count"
2600 Details : If a new ldiskfs filesystem is created with the "uninit_groups"
2601 feature and only a single inode is created in a group then the
2602 "bg_unused_inodes" count is incorrectly updated. Creating a
2603 second inode in that group would update it correctly.
2608 Description: buffer overruns could theoretically occur
2609 Details : llapi_semantic_traverse() modifies the "path" argument by
2610 appending values to the end of the origin string, and a buffer
2611 overrun may occur. Adding buffer overrun check in liblustreapi.
2613 Severity : enhancement
2615 Description: Add jbd statistics patch for RHEL5 and 2.6.18-vanilla.
2618 Frequency : only if filesystem is inconsistent
2620 Description: handle "serious error: objid * already exists" more gracefully
2621 Details : If LAST_ID value on disk is smaller than the objects existing in
2622 the O/0/d* directories, it indicates disk corruption and causes an
2623 LBUG(). If the object is 0-length, then we should use the existing
2624 object. This will help to avoid a full fsck in most cases.
2626 Severity : enhancement
2628 Description: Kernel patches update for RHEL4 2.6.9-55.0.6.
2629 Details : Modify vm-tunables-rhel4.patch.
2631 Severity : enhancement
2633 Description: Kernel config for 2.6.18-vanilla.
2634 Details : Modify targets/2.6-vanilla.target.in.
2635 Add config file kernel-2.6.18-2.6-vanilla-i686.config.
2636 Add config file kernel-2.6.18-2.6-vanilla-i686-smp.config.
2637 Add config file kernel-2.6.18-2.6-vanilla-x86_64.config.
2638 Add config file kernel-2.6.18-2.6-vanilla-x86_64-smp.config.
2641 Frequency : occasional
2643 Description: improve handling recoverable errors
2644 Details : If request processed with error which can be recoverable on server
2645 request should be resend, otherwise page released from cache and
2651 Description: Kernel patches update for RHEL5 2.6.18-8.1.14.el5.
2652 Details : Modify target file & which_patch.
2653 A flaw was found in the IA32 system call emulation provided
2654 on AMD64 and Intel 64 platforms. An improperly validated 64-bit
2655 value could be stored in the %RAX register, which could trigger an
2656 out-of-bounds system call table access. An untrusted local user
2657 could exploit this flaw to run code in the kernel
2658 (ie a root privilege escalation). (CVE-2007-4573).
2662 Description: change order of libsysio includes
2663 Details : '#include sysio.h' should always come before '#include xtio.h'
2665 Severity : enhancement
2667 Description: adapt the lustre_config script to support the upgrade case
2668 Details : Add "-u" option for lustre_config script to support upgrading 1.4
2669 server targets to 1.6 in parallel.
2674 Description: To avoid grant space > avaible space when the disk is almost
2675 full. Without this patch you might see the error "grant XXXX >
2676 available" or some LBUG about grant, when the disk is almost
2678 Details : In filter_check_grant, for non_grant cache write, we should
2679 check the left space by if (*left > ungranted + bytes), instead
2680 of (*left > ungranted), because only we are sure the left space
2681 is enough for another "bytes", then the ungrant space should be
2682 increase. In client, we should update cl_avail_grant only there is
2683 OBD_MD_FLGRANT in the reply.
2688 Description: Update RHEL 4 kernel to fix local root privilege escalation.
2689 Details : Update to the latest RHEL 4 kernel to fix the vulnerability
2690 described in CVE-2007-4573. This problem could allow untrusted
2691 local users to gain root access.
2694 Frequency : when using O_DIRECT and quotas
2696 Description: Incorrect file ownership on O_DIRECT output files
2697 Details : block usage reported by 'lfs quota' does not take into account
2698 files that have been written with O_DIRECT.
2702 Description: (rw.c:1323:ll_read_ahead_pages()) ASSERTION(page_idx > ria->ria_stoff) failed
2703 Details : Once the unmatched stride IO mode is detected, shrink the stride-ahead
2704 window to 0. If it does hit cache miss, and read-pattern is still
2705 stride-io mode, does not reset the stride window, but also does not
2706 increase the stride window length in this case.
2708 --------------------------------------------------------------------------------
2710 2007-09-27 Cluster File Systems, Inc. <info@clusterfs.com>
2712 * Support for kernels:
2713 2.6.5-7.286 (SLES 9),
2714 2.6.9-55.0.2.EL (RHEL 4),
2715 2.6.16.46-0.14 (SLES 10),
2716 2.6.18-8.1.8.el5 (RHEL 5),
2717 2.6.18.8 vanilla (kernel.org)
2718 * Client support for unpatched kernels:
2719 (see http://wiki.lustre.org/index.php?title=Patchless_Client)
2720 2.6.16 - 2.6.21 vanilla (kernel.org)
2721 * Due to recently discovered recovery problems, we do not recommend
2722 using patchless RHEL 4 clients with this or any earlier release.
2723 * Recommended e2fsprogs version: 1.40.2-cfs1
2724 * Note that reiserfs quotas are disabled on SLES 10 in this kernel.
2728 Description: Fix errors in lfs documentation
2729 Details : Fixes man pages
2731 Severity : enhancement
2733 Description: Adaptive timeouts
2734 Details : RPC timeouts adapt to changing server load and network
2735 conditions to reduce resend attempts and improve recovery time.
2737 Severity : enhancement
2739 Description: llapi_file_create() does not allow some changes
2740 Details : add llapi_file_open() that allows specifying the file creation
2741 mode and open flags, and also returns an open file handle.
2743 Severity : enhancement
2745 Description: df doesn't work properly if diskfs blocksize != 4K
2746 Details : Choose biggest blocksize of OST's as the LOV's blocksize.
2749 Frequency : always on directories with default striping set
2751 Description: lfs find on -1 stripe looping in lsm_lmm_verify_common()
2752 Details : Avoid lov_verify_lmm_common() on directory with -1 stripe count.
2754 Severity : enhancement
2756 Description: merge and cleanup kernel patches.
2757 Details : Remove mnt_lustre_list in vfs_intent-2.6-rhel4.patch.
2759 Severity : enhancement
2761 Description: RedHat Update kernel for RHEL5
2762 Details : Kernel config file for RHEL5.
2764 Severity : enhancement
2766 Description: OSS needs mutliple precreate threads
2767 Details : Add ability to start more than one create thread per OSS.
2770 Frequency : only with quota on the root user
2772 Description: mds_obd_create error creating tmp object
2773 Details : When the user sets quota on root, llog will be affected and can't
2774 create files and write files.
2777 Frequency : Always on ia64 patchless client, and possibly others.
2779 Description: Add EXPORT_SYMBOL check for node_to_cpumask symbol.
2780 Details : This allows the patchless client to be loaded on architectures
2781 without this export.
2786 Description: disorder of journal start and llog_add cause deadlock.
2787 Details : in llog_origin_connect, journal start should happen before
2788 llog_add keep the same order as other functions to avoid
2791 Severity : enhancement
2793 Description: RedHat Update kernel for RHEL5
2794 Details : Modify the kernel config file more closer RHEL5.
2796 Severity : enhancement
2798 Description: Build failure against Centos5 (RHEL5)
2799 Details : Define PAGE_SIZE when it isn't present.
2802 Frequency : occasionally when using NFS
2804 Description: "ll_intent_file_open()) lock enqueue: err: -13" with nfs
2805 Details : with NFS, the anon dentry's parent was set to itself in
2806 d_alloc_anon(), so in MDS, we use rec->ur_fid1 to find the
2807 corresponding dentry other than use rec->ur_name.
2809 Severity : enhancement
2811 Description: enable data checksumming by default
2812 Details : enable checksum by default, allow --disable-checksum
2813 configure option and "-o nochecksum" mount option. Checksums
2814 can also be disabled at runtime via $LPROC/osc/*/checksum_pages.
2817 Frequency : Occasionally with failover
2819 Description: Client eviction due to failover config
2820 Details : after a connection loss, the lustre client should attempt to
2821 reconnect to the last active server first before trying the
2822 other potential connections.
2824 Severity : enhancement
2826 Description: client-side metadata stat-ahead during readdir(directory readahead)
2827 Details : perform client-side metadata stat-ahead when the client detects
2828 readdir and sequential stat of dir entries therein
2833 Description: when mds and osts use different quota unit(32bit and 64bit),
2834 quota will be released repeatly.
2835 Details : void sending multiple quota reqs to mds, which will keep the
2836 status between the reqs.
2839 Frequency : only with liblustre clients on XT3
2841 Description: evictions taking too long
2842 Details : allow llrd to evict clients directly on OSTs
2846 Description: osts not allocated evenly to files
2847 Details : change the condition to increase offset_idx
2849 Severity : enhancement
2851 Description: self-adjustable client's lru lists
2852 Details : use adaptive algorithm for managing client cached locks lru
2853 lists according to current server load, other client's work
2854 pattern, memory activities, etc. Both, server and client
2855 side namespaces provide number of proc tunables for controlling
2860 Description: rewrite ext2-derived code in llite/dir.c and obdclass/uuid.c
2861 Details : rewrite inherited code (uuid parsing code from ext2 utils and
2862 readdir code from ext3) from scratch preserving functionality.
2866 Description: Only those disconnect error should be returned by rq_status.
2867 Details : In open/enqueue processs, Some errors, which will cause client
2868 disconnected, should be returned by rq_status, while other
2869 errors should still be returned by intent, then mdc or llite will
2872 Severity : enhancement
2874 Description: Tune the kernel for good SCSI performance.
2875 Details : Set the value of /sys/block/{dev}/queue/max_sectors_kb
2876 to the value of /sys/block/{dev}/queue/max_hw_sectors_kb
2880 Frequency : Always for filesystems larger than 2TB on 32-bit systems.
2881 Bugzilla : 13547, 13627
2882 Description: Data corruption for OSTs that are formatted larger than 2TB
2884 Details : When generating the bio request for lustre file writes the
2885 sector number would overflow a temporary variable before being
2886 used for the IO. The data reads correctly from Lustre (which
2887 will overflow in a similar manner) but other file data or
2888 filesystem metadata may be corrupted in some cases.
2892 Description: "lfs find -obd UUID" prints directories
2893 Details : "lfs find -obd UUID" will return all directory names instead
2894 of just file names. It is incorrect because the directories
2895 do not reside on the OSTs.
2898 Bugzilla : 13596, 13614
2899 Description: MDS hang after unclean shutdown of lots of clients
2900 Details : Never resend AST requests.
2904 Frequency : Always, for kernels after 2.6.16
2905 Description: Fix warning idr_remove called for id=.. which is not allocated.
2906 Details : Last kernels save old s_dev before kill super and not allow
2907 to restore from callback - restore it before call kill_anon_super.
2911 Description: Fix error on 'ls .' at the top of the Lustre mount.
2912 Details : Don't revalidate dentry if it is a root dentry.
2914 --------------------------------------------------------------------------------
2916 2007-08-27 Cluster File Systems, Inc. <info@clusterfs.com>
2918 * Support for kernels:
2919 2.6.5-7.286 (SLES 9),
2920 2.6.9-55.0.2.EL (RHEL 4),
2921 2.6.16.46-0.14 (SLES 10),
2922 2.6.18-8.1.8.el5 (RHEL 5),
2923 2.6.18.8 vanilla (kernel.org)
2924 * Client support for unpatched kernels:
2925 (see http://wiki.lustre.org/index.php?title=Patchless_Client)
2926 2.6.16 - 2.6.22 vanilla (kernel.org)
2927 * Due to recently discovered recovery problems, we do not recommend
2928 using patchless RHEL 4 clients with this or any earlier release.
2929 * Recommended e2fsprogs version: 1.39.cfs8
2930 * Note that reiserfs quotas are disabled on SLES 10 in this kernel.
2935 Description: block reactivating mgc import until all deactivates complete
2936 Details : Fix race when failing back MDT/MGS to itself (testing)
2938 Severity : enhancement
2940 Description: lfs setstripe enhancement
2941 Details : Make lfs setstripe understand 'k', 'm' and 'g' for stripe size.
2944 Frequency : mds/oss recovery
2946 Description: llog ctxt is refrenced after it has been freed.
2947 Details : llog ctxt refcount was added to avoide the race between ctxt free
2948 and llog recovery process. Each llog user must hold ctxt refcount
2949 before it access the llog. And the llog ctxt can only be freed
2950 when its refcount is zero.
2952 Severity : enhancement
2954 Description: randomly memory allocation failure util
2955 Details : Make lustre randomly failed allocating memory for testing purpose.
2958 Frequency : only for SLES10
2960 Description: Update kernel patch for SLES10 SP1
2961 Details : Add patch blkdev_tunables-2.6-sles10.patch to 2.6-sles10.series.
2963 Severity : enhancement
2965 Description: omit set fsid for export NFS
2966 Details : fix set/restore device id for avoid EMFILE error and mark lustre fs
2967 as FS_REQUIRES_DEV for avoid problems with generate fsid.
2970 Frequency : after network failures
2972 Description: Add sync option to mount_lustre.c
2973 Details : Client loses data written to lustre after a network interruption.
2975 Severity : enhancement
2977 Description: Error message improvement.
2978 Details : Merging of two LCONSOLE_ERROR_MSG into one.
2980 Severity : enhancement
2982 Description: don't use GFP_* in generic Lustre code.
2983 Details : Use cfs_alloc_* functions and CFS_* flags for code portability.
2985 Severity : enhancement
2987 Description: obdclass is limited by single OBD_ALLOC(idarray)
2988 Details : replace OBD_ALLOC/OBD_FREE with OBD_VMALLOC/OBD_VFREE
2990 Severity : enhancement
2992 Description: updated patchess for new RHEL4 kernel
2993 Details : Fixed ext3-unlink-race.patch per Kalpak's comment.
2995 Severity : enhancement
2997 Description: warnings with build patchless client with vanila 2.6.19 and up
2998 Details : change old ctl_table style and replace ctl_table/ctl_table_header
2999 with cfs_sysctl_table_t/cfs_sysctl_table_header_t
3001 Severity : enhancement
3003 Description: O_DIRECT bypasses client statistics.
3004 Details : When running with O_DIRECT I/O, neither the client rpc_stats nor
3005 read_ahead_stats were updated. Copied stats section from
3006 osc_send_oap_rpc() into async_internal().
3009 Frequency : only for Cray XT3
3011 Description: peer credits not enough on many OST per OSS systems.
3012 Details : Use new lnet way to add credits as we need those for pings and ASTs
3015 Frequency : only with liblustre
3017 Description: Liblustre is not releasing flock locks on file close.
3018 Details : Release flock locks on file close.
3021 Frequency : only for RHEL4
3023 Description: Update kernel patches for kernel-2.6.9-55.0.2.EL
3024 Details : Remove inode-nr_unused-2.6.9-rhel4.patch from 2.6-rhel4.series
3025 Update target file and kernel config.
3029 Description: lustre support for RHEL5
3030 Details : Add support for RHEL5.
3035 Description: ASSERTION(export != NULL) failed in target_handle_connect
3036 Details : Assetion hit is result of rare race between disconnect and connet
3037 to same nid. target_handle_connect found old connect cockie and
3038 tried to reconnect, but can't find export for this cockie.
3043 Description: umount blocks forever on error
3044 Details : In result of wrong using obd_no_recov and obd_force flags client
3045 can hand if cancel or some other requests is lost.
3048 Frequency : Only for SLES
3050 Description: sanity_quota fail test_1
3051 Details : There are multiple occurences of $TSTUSR in SLES's /etc/group
3052 file, which makes TSTID[2] inunique.
3054 Severity : enhancement
3056 Description: Kernel patches for SLES9 2.6.5-7.286 kernel
3057 Details : Update target/ChangeLog/which_patch .
3059 Severity : enhancement
3061 Description: jbd statistics
3062 Details : Port older jbd statistics patch for sles10
3067 Description: lvbo_init failed for resource with missing objects.
3068 Details : Fix returning error if we do stat for file with missing/corrupted
3069 objects and i_size set to all sum of size of all avaible objects.
3070 if we to truncate/write to missing object - it is recreated.
3075 Description: Oops in read and write path when failing to allocate lock.
3076 Details : Check if lock allocation failed and return error back.
3079 Frequency : When flocks are used.
3081 Description: assertion failure in ldlm_cli_enquque_fini for non NULL lock.
3082 Details : Flock locks might destroy just granted lock if it could be merged
3083 with another existing flock, this is done in completion handler,
3084 so teach ldlm_cli_enquque_fini that this is a valid case for
3090 Description: reply_lock_interpret crash due to race with it and lock cancel.
3091 Details : Do not replay locks that are being cancelled. Do not reference
3092 locks by their address during replay, just by their handle.
3095 Frequency : only with deactivated OSTs
3097 Description: lstripe command fails for valid OST index
3098 Details : The stripe offset is compared to 'lov->desc.ld_tgt_count'
3099 instead of lov->desc.ld_active_tgt_count.
3101 Severity : enhancement
3103 Description: Build failure against Centos5 (RHEL5)
3104 Details : Use getpagesize() instead of PAGE_SIZE.
3106 --------------------------------------------------------------------------------
3108 2007-07-30 Cluster File Systems, Inc. <info@clusterfs.com>
3110 * Support for kernels:
3111 2.6.5-7.283 (SLES 9),
3112 2.6.9-55.EL (RHEL 4),
3113 2.6.16.46-0.14 (SLES 10),
3114 2.6.18.8 vanilla (kernel.org)
3115 * Client support for unpatched kernels:
3116 (see http://wiki.lustre.org/index.php?title=Patchless_Client)
3117 2.6.16 - 2.6.22 vanilla (kernel.org)
3118 * Due to recently discovered recovery problems, we do not recommend
3119 using patchless RHEL 4 clients with this or any earlier release.
3120 * Recommended e2fsprogs version: 1.39.cfs8
3121 * Note that reiserfs quotas are disabled on SLES 10 in this kernel.
3122 * Starting with this release, the ldiskfs backing filesystem required
3123 by Lustre is now in its own package, lustre-ldiskfs. This package
3124 should be installed. It is versioned separately from Lustre and
3125 may be released separately in future.
3130 Description: server LBUG when shutting down
3131 Details : Block umount forever until the mount refcount is zero rather
3132 than giving up after an arbitrary timeout.
3134 Severity : enhancement
3136 Description: add optional extra BUILD_VERSION info
3137 Details : add a new environment variable (namely LUSTRE_VERS) which allows
3138 to override the lustre version.
3141 Frequency : 2.6.18 servers only
3143 Description: ll_kern_mount() doesn't release the module reference
3144 Details : The ldiskfs module reference count never drops down to 0
3145 because ll_kern_mount() doesn't release the module reference.
3150 Description: server LBUG when using old ost_num_threads parameter
3151 Details : Accept the old ost_num_threads parameter but warn that it
3152 is deprecated, and fix an off-by-one error that caused an LBUG.
3157 Description: Transient SCSI error results in persistent IO issue
3158 Details : iobuf->dr_error is not reinitialized to 0 between two
3162 Frequency : sometimes when underlying device returns I/O errors
3164 Description: OSTs not going read-only during write failures
3165 Details : OSTs are not remounted read-only when the journal commit threads
3166 get I/O errors because fsfilt_ext3 calls journal_start/stop()
3167 instead of the ext3 wrappers.
3170 Frequency : only with 10000 clients or more
3172 Description: poor connect scaling with increasing client count
3173 Details : Don't run filter_grant_sanity_check for more than 100 exports
3174 to improve scaling for large numbers of clients.
3177 Frequency : SLES10 only
3179 Description: sanity-quota.sh quotacheck failed: rc = -22
3180 Details : Quotas cannot be enabled on SLES10.
3183 Frequency : liblustre clients only
3185 Description: getdirentries does not give error when run on compute nodes
3186 Details : getdirentries does not fail when the size specified as an argument
3187 is too small to contain at least one entry
3189 Severity : enhancement
3191 Description: Add LNET router traceability for debug purposes
3192 Details : If a checksum failure occurs with a router as part of the
3193 IO path, the NID of the last router that forwarded the bulk data
3194 is printed so it can be identified.
3199 Description: OST "spontaneously" evicts client; client has imp_pingable == 0
3200 Details : Due to a race condition, liblustre clients were occasionally
3201 evicted incorrectly.
3203 Severity : enhancement
3205 Description: lfs setstripe use optional parameters instead of postional
3208 Severity : enhancement
3210 Description: Nanosecond timestamp support for ldiskfs
3211 Details : The on-disk ldiskfs filesystem has added support for nanosecond
3212 resolution timestamps. There is not yet support for this at
3213 the Lustre filesystem level.
3216 Frequency : during server recovery
3218 Description: MDS failing to send precreate requests due to OSCC_FLAG_RECOVERING
3219 Details : request with rq_no_resend flag not awake l_wait_event if they get a
3223 Frequency : nfs export on patchless client
3225 Description: connectathon hang when test nfs export over patchless client
3226 Details : Disconnected dentry cannot be found with lookup, so we do not need
3227 to unhash it or make it invalid
3230 Description: fix llapi_lov_get_uuids() to allow many OSTs to be returned
3231 Details: : Change llapi_lov_get_uuids() to read the UUIDs from /proc instead
3232 of using an ioctl. This allows lfsck for > 160 OSTs to succeed.
3237 Description: open req refcounting wrong on reconnect
3238 Details : If reconnect happened between getting open reply from server and
3239 call to mdc_set_replay_data in ll_file_open, we will schedule
3240 replay for unreferenced request that we are about to free.
3241 Subsequent close will crash in variety of ways.
3242 Check that request is still eligible for replay in
3243 mdc_set_replay_data().
3248 Description: disable writes to filesystem when reading health_check file
3249 Details : the default for reading the health_check proc file has changed
3250 to NOT do a journal transaction and write to disk, because this
3251 can cause reads of the /proc file to hang and block HA state
3252 checking on a healthy but otherwise heavily loaded system. It
3253 is possible to return to the previous behaviour during configure
3254 with --enable-health-write.
3256 Severity : enhancement
3258 Description: 64-bit inode version
3259 Details: : Add a on-disk 64-bit inode version for ext3 to track changes made
3260 to the inode. This will be required for version-based recovery.
3265 Description: MDS fails to start if a duplicate client export is detected
3266 Details : in some rare cases it was possible for a client to connect to
3267 an MDS multiple times. Upon recovery the MDS would detect this
3268 and fail during startup. Handle this more gracefully.
3270 Severity : enhancement
3272 Description: Add -o localflock option to simulate old noflock behaviour.
3273 Details : This will achieve local-only flock/fcntl locks coherentness.
3278 Description: log_commit_thread vs filter_destroy race leads to crash
3279 Details : Take import reference before releasing llog record semaphore
3284 Description: Wrong request locking in request set processing
3285 Details : ptlrpc_check_set wrongly uses req->rq_lock for proctect add to
3286 imp_delayed_list, in this place should be used imp_lock.
3289 Frequency : when reconnecting
3291 Description: Grant leak when OSC reconnect to OST
3292 Details : When osc reconnect ost, OST(filter) should check whether it
3293 should grant more space to client by comparing fed_grant and
3294 cl_avail_grant, and return the granted space to client instead
3295 of "new granted" space, because client will call osc_init_grant
3296 to update the client grant space info.
3299 Frequency : when client reconnects to OST
3301 Description: Grant leak when OSC does a resend and replays bulk write
3302 Details : When osc reconnect to OST, OST(filter) should clear grant info of
3303 bulk write request, because the grant info while be sync between
3304 OSC and OST when reconnect, and we should ignore the grant info
3305 these of resend/replay write req.
3310 Description: Grant space more than avaiable space sometimes.
3311 Details : When then OST is about to be full, if two bulk writing from
3312 different clients came to OST. Accord the avaliable space of the
3313 OST, the first req should be permitted, and the second one
3314 should be denied by ENOSPC. But if the seconde arrived before
3315 the first one is commited. The OST might wrongly permit second
3316 writing, which will cause grant space > avaiable space.
3319 Frequency : when client is evicted
3321 Description: Grant might be wrongly erased when osc is evicted by OST
3322 Details : when the import is evicted by server, it will fork another
3323 thread ptlrpc_invalidate_import_thread to invalidate the
3324 import, where the grant will be set to 0. While the original
3325 thread will update the grant it got when connecting. So if
3326 the former happened latter, the grant will be wrongly errased
3327 because of this race.
3332 Description: Checking Stale with correct fid
3333 Details : ll_revalidate_it should uses de_inode instead of op_data.fid2
3334 to check whether it is stale, because sometimes, we want the
3335 enqueue happened anyway, and op_data.fid2 will not be initialized.
3337 Severity : enhancement
3339 Description: update patchless client
3340 Details : Add support for patchless client with 2.6.20, 2.6.21 and RHEL 5
3343 Frequency : only with 2.4 kernel
3345 Description: random memory corruption
3346 Details : size of struct ll_inode_info is to big for union inode.u and this
3347 can be cause of random memory corruption.
3352 Description: Memory leak in recovery
3353 Details : Lov_mds_md was not free in an error handler in mds_create_object.
3354 It should also check obd_fail before fsfilt_start, otherwise if
3355 fsfilt_start return -EROFS,(failover mds during mds recovery).
3356 then the req will return with repmsg->transno = 0 and rc = EROFS.
3357 and we met hit the assert LASSERT(req->rq_reqmsg->transno ==
3358 req->rq_repmsg->transno) in ptlrpc_replay_interpret. Fcc should
3359 be freed no matter whether fsfilt_commit success or not.
3362 Frequency : only with large numbers of cores on a single node
3364 Description: Prevents from taking the superblock lock in llap_from_page for
3366 Details : using LL_ORIGIN_REMOVEPAGE origin flag instead of LL_ORIGIN_UNKNOW
3367 for llap_from_page call in ll_removepage() prevents from taking
3368 the superblock lock for a soon died page.
3373 Description: Not check open intent error before release open handle
3374 Details : in some rare cases, the open intent error is not checked before
3375 release open handle, which may cause
3376 ASSERTION(open_req->rq_transno != 0), because it tries to release
3377 the failed open handle.
3382 Description: Set cat log bitmap only after create log success.
3383 Details : in some rare cases, the cat log bitmap is set too early. and it
3384 should be set only after create log success.
3388 Description: Accessing a block bevice can re-enable I/O when Lustre is
3389 tearing down a device.
3390 Details : dev_clear_rdonly(bdev) must be called in kill_bdev() instead of
3395 Description: service threads may hog cpus when there are a lot of requests
3396 Details : Insert cond_resched to give other threads a chance to use some CPU
3401 Description: the cat log was not initialized in recovery
3402 Details : When mds(mgs) do recovery, the tgt_count might be zero, so the
3403 unlink log on mds will not be initialized until mds post
3404 recovery. And also in mds post recovery, the unlink log will
3405 initialization will be done asynchronausly, so there will be race
3406 between add unlink log and unlink log initialization.
3410 Description: brw_stats were being printed incorrectly
3411 Details : brw_stats were being printed as log2 but all of them were not
3412 recorded as log2. Also remove some code duplication arising from
3413 filter_tally_{read,write}.
3417 Frequency : rare, only in recovery.
3418 Description: ASSERTION(req->rq_type != LI_POISON) failed
3419 Details : imp_lock should be held while iterating over imp_sending_list for
3420 prevent destroy request after get timeout in ptlrpc_queue_wait.
3424 Description: replay-single.sh test 52 fails
3425 Details : A lock's skiplist need to be cleanup when it being unlinked
3426 from its resource list.
3431 Description: Short directio read returns full requested size rather than
3433 Details : Direct I/O operations should return actual amount of bytes
3434 transferred rather than requested size.
3436 Severity : enhancement
3438 Description: metadata RPC reduction (e.g. for rm performance)
3439 Details : decrease the amount of synchronous RPC between clients and servers
3440 by canceling conflicing lock before the operation on the client
3441 and packing thier handles into the main operation RPC to server.
3443 Severity : enhancement
3445 Description: add #ifdef HAVE_KERNEL_CONFIG_H
3446 Details : kernels from 2.6.19 not need include linux/config.h, but add
3447 include linux/autoconf.h in commpiler command line.
3449 Severity : enhancement
3451 Description: patchless client support for 2.6.22 kernel
3452 Details : 2.6.22 has only one visble change, SLAB_CTOR_* constants is
3453 removed. In this case we need drop using os depended interface to
3454 kmem_cache and use cfs_mem_cache API.
3459 Description: fix mal-formatted messages
3460 Details : fix some mal-formatted DEBUG_REQ and LCONSOLE_ERROR_MSG messages
3464 Frequency : always in liblustre
3465 Description: wrong IS_ERR implementation in liblustre.h
3466 Details : fix IS_ERR implementation in liblustre.h for right detect errors.
3471 Description: Correct condition for output debug message.
3472 Details : inode i_nlink equal zero is not enough for output message about
3473 disk corruption, i_ctime and i_mode should be also checked.
3477 Frequency : always in patchless client
3478 Description: add configure check for truncate_complete_page
3479 Details : improve checks for exported symbols. This allow run check without
3480 sources, but with Module.symvers shipped with kernel distribution.
3481 add check for truncate_complete_page used by patchless client.
3485 Description: sanity.sh test_77h fails with "test_77h file compare failed"
3486 Details : test_77h uses a file which was messed by other test case.
3490 Description: Not Check whether lov_tgts is NULL in some lov functions
3491 Details : Checking whether lov_tgts is NULL in some functions.
3495 Description: replace obdo_alloc() with OBDO_ALLOC macro
3496 Details : nothing special is done in obdo_alloc() function, and for
3497 debugging purpose, it needs to be replaced with macros.
3501 Description: bad return value and errno from fcntl call
3502 Details : In liblustre API, errno should be a negative value if error
3507 Description: ptlrpc_check_set() LBUG
3508 Details : In case of positive reply from server and failed client bulk
3509 callback after bulk transfer shouldn't LBUG, but process this
3510 request as erroneous.
3512 Severity : enhancement
3514 Description: VFS operations stats tool.
3515 Details : Tool which collects stats by tracking value written in pid,
3516 ppid, gid and uses llstat to generate output to plot graph using
3518 Updated lustre/utils/Makefile.am
3519 Added lustre/utils/ltrack_stats.c
3521 Severity : enhancement
3523 Description: 2.6.18 server support (lustre 1.6.1)
3524 Details : Support for 2.6.18 kernels on the server side.
3529 Description: ASSERTION(imp->imp_conn_current) failed
3530 Details : an assertion failure is hit if a client node boots and attempts to
3531 mount a lustre filesystem faster than RECONNECT_INTERVAL seconds.
3534 Frequency : only for i686
3536 Description: 1.4.11 RC1 build fails for RHEL 4, i686
3537 Details : Fixed config variable for build.
3542 Description: Updated patchess for new RHEL4 kernel
3543 Details : Updated patch inode-nr_unused-2.6.9-rhel4.patch
3544 Updated patch jbd-stats-2.6.9.patch
3545 Updated patch qsnet-rhel4-2.6.patch
3546 Updated patch quota-deadlock-on-pagelock-core.patch
3547 Updated patch vfs_intent-2.6-rhel4.patch
3548 Updated patch vfs_races-2.6-rhel4.patch
3549 Updated series file 2.6-rhel4-titech.series
3550 Updated series file 2.6-rhel4.series
3551 Updated kernel config files
3556 Description: lquota slave complains LBUG when reconnecting with mds
3558 Details : quota slave depends qctxt->lqc_import to send its quota request.
3559 This pointer will be invalid if mds did failover or broke its
3560 connect to osts, which leads to LBUG.
3563 Frequency : when qunit size is too small(less than 20M)
3565 Description: write is stopped by improper -EDQUOT
3566 Details : If the master is busy and qunit size is small enough(let's say 1M),
3567 the slave can not get quota from master on time, which will lead
3568 slave to trigger a -EQUOTA to client.
3573 Description: Deadlock during metadata tests
3574 Details : in prune_dir_dentries(), shrink_dcache_parent() should not be
3575 called with the per-dentry lock held.
3578 Frequency : SLES9 only
3580 Description: Lustre patched kernel for SLES9 SP3 has NR_CPUS set to 8
3581 Details : set CONFIG_NR_CPUS to 128 instead of 8.
3583 Severity : enhancement
3585 Description: remove fs_prep_san_write operation and related patches
3586 Details : remove the ext3-san-jdike patches which are no longer useful.
3591 Description: LDISKFS-fs error (device sdc): ldiskfs_free_blocks
3592 Details : a disk corruption can cause the mballoc code to assert on a
3593 double free or other extent corruptions. Handle these with
3594 ext3_error() instead of assertions.
3597 Frequency : only with mballoc3 code and deep extent trees
3599 Description: ldiskfs_ext_search_right: bad header in inode: unexpected eh_depth
3600 Details : a wrong check of extent headers in ldiskfs_ext_search_right()
3601 can cause the filesystem to be remounted read-only.
3606 Description: lfsck built against 1.4.x cannot run against 1.6.0 lustre
3607 Details : the definition for OBD_IOC_GETNAME changed in 1.6.0. One of the
3608 few external users of this ioctl number is lfsck's call to
3609 llapi_lov_get_uuids() and this caused lfsck to fail at startup.
3610 Add the old ioctl number to the handler so both old and new
3615 Description: parallel lock callbacks
3616 Details : Instead of sending blocking and completion callbacks as separated
3617 requests, adding them to a set and sending in parallel.
3621 Description: Disable most debugging by default
3622 Details : To improve performance, disable most logging (for debug purposes)
3623 by default. VFSTRACE, RPCTRACE, and DLMTRACE are now off by
3624 default, and HA includes fewer messages.
3628 Frequency : only run on patchless client.
3629 Description: use do_facet on sanity.sh for test handling recoverables errors
3630 Details : use do_facet instead of direct use sysctl for set fail_loc on OST
3634 Description: hash tables for lists of nids, connections and uuids
3635 Details : Hash tables noticeably help when a lot of clients connect to a
3636 server, to faster identify duplicate connections or reconnects,
3637 also to faster find export to evict in manual eviction case.
3641 Description: Sometimes, when the server evict a client, and the client will
3642 not be evicted as soon as possible.
3643 Details : In enqueue req, the error was returned by intent, instead of
3644 rq_status which make ptlrpc layer not detect this error, and
3645 does not evict the client. So enqueue error should be returned
3649 Frequency : only at startup
3651 Description: Delay client connections to MDT until fist MDT->OST connect
3652 Details : If a client tried to create a new file before the MDT had
3653 connected to any OSTs, the create would return EIO. Now
3654 the client will simply block until the MDT connects to the
3655 first OST and the create can succeed.
3658 Frequency : only for SLES9
3660 Description: Routinely utilize latest Quadrics drivers in CFS releases
3661 Details : Update patch qsnet-suse-2.6.patch.
3664 Frequency : only for sles10
3666 Description: Update patches for SLES 10 SP1 kernel.
3667 Details : Update the patch vfs_intent-2.6-sles10.patch.
3672 Description: Routinely utilize latest Quadrics drivers in CFS releases
3673 Details : Update patch qsnet-rhel4-2.6.patch.
3676 Frequency : at statup only
3678 Description: mds_lov_synchronize race leads to various problems
3679 Details : simultaneous MDT->OST connections at startup can cause the
3680 sync to abort, leaving the OSC in a bad state.
3685 Description: Using wrong pointer in osc_brw_prep_request
3686 Details : Access to array[-1] can produce panic if kernel compiled with
3687 CONFIG_PAGE_ALLOC enabled
3689 Severity : enhancement
3691 Description: Async OSC create to avoid the blocking unnecessarily.
3692 Details : If a OST has no remain object, system will block on the creating
3693 when need to create a new object on this OST. Now, ways use
3694 pre-created objects when available, instead of blocking on an
3695 empty osc while others are not empty. If we must block, we block
3696 for the shortest possible period of time.
3700 Frequency : only in recovery
3701 Description: Mark OST as early accessible if his start SYNC.
3702 Details : osc_precreate return flag early accessible if oscc marked as
3703 OSCC_FLAG_SYNC_IN_PROGRESS.
3708 Description: Sometimes precreate code can triger create object on wrong ost
3709 Details : Wrong protected or not not restored variables aftre precreate loop
3710 can produce creation object on wrong ost.
3713 Frequency : oss recovery
3715 Description: llog_commit_thread cleanup should sync with llog_commit_thread
3717 Details : llog_commit_thread_count should be synced between llog_commit
3718 start and cleanup, so new llog_commit thread should not be started
3719 when llog_commit threads being stopped to avoid accessing some
3722 Severity : enhancement
3724 Description: Add printing inode info into message about error in writepage.
3726 --------------------------------------------------------------------------------
3728 2007-05-03 Cluster File Systems, Inc. <info@clusterfs.com>
3733 Frequency : on some architectures
3735 Description: 1.6 client sometimes fails to mount from a 1.4 MDT
3736 Details : Uninitialized flags sometimes cause configuration commands to
3740 Frequency : patchless clients only
3742 Description: missing __iget() symbol export
3743 Details : The __iget() symbol export is missing. To avoid the need for
3744 this on patchless clients the deathrow inode reaper is turned
3745 off, and we depend on the VM to clean up old inodes. This
3746 dependency was during via the fix for bug 12181.
3751 Description: sanity.sh fail: test_52b
3752 Details : The ll_inode_to_ext_flags() has a glitch which makes MDS return
3753 incorrect inode's flags to client.
3755 --------------------------------------------------------------------------------
3757 2007-04-19 Cluster File Systems, Inc. <info@clusterfs.com>
3759 * CONFIGURATION CHANGE. This version of Lustre WILL NOT
3760 INTEROPERATE with older versions automatically. In many cases a
3761 special upgrade step is needed. Please read the
3762 user documentation before upgrading any part of a 1.4.x system.
3763 * WARNING: Lustre configuration and startup changes are required with
3764 this release. See https://mail.clusterfs.com/wikis/lustre/MountConf
3766 * Support for kernels:
3767 2.4.21-47.0.1.EL (RHEL 3),
3768 2.6.5-7.283 (SLES 9),
3769 2.6.9-42.0.10.EL (RHEL 4),
3770 2.6.12.6 vanilla (kernel.org),
3771 2.6.16.27-0.9 (SLES10)
3772 * Client support for unpatched kernels:
3773 (see https://mail.clusterfs.com/wikis/lustre/PatchlessClient)
3774 2.6.16 - 2.6.19 vanilla (kernel.org),
3775 2.6.9-42.0.8EL (RHEL 4)
3776 * Recommended e2fsprogs version: 1.39.cfs6
3777 * Note that reiserfs quotas are disabled on SLES 10 in this kernel
3780 Severity : enhancement
3782 Description: Async OSC create to avoid the blocking unnecessarily.
3783 Details : If a OST has no remain object, system will block on the creating
3784 when need to create a new object on this OST. Now, ways use
3785 pre-created objects when available, instead of blocking on an
3786 empty osc while others are not empty. If we must block, we block
3787 for the shortest possible period of time.
3789 Severity : enhancement
3791 Description: MountConf
3792 Details : Lustre configuration is now managed via mkfs and mount
3793 commands instead of lmc and lconf. New obd types (MGS, MGC)
3794 are added for dynamic configuration management. See
3795 https://mail.clusterfs.com/wikis/lustre/MountConf for
3798 Severity : enhancement
3800 Description: dynamic OST addition
3801 Details : OSTs can now be added to a live filesystem
3803 Severity : enhancement
3805 Description: startup order invariance
3806 Details : MDTs and OSTs can be started in any order. Clients only
3807 require the MDT to complete startup.
3809 Severity : enhancement
3811 Description: parallel, asynchronous orphan cleanup
3812 Details : orphan cleanup is now performed in separate threads for each
3813 OST, allowing parallel non-blocking operation.
3815 Severity : enhancement
3817 Description: optimized stripe assignment
3818 Details : stripe assignments are now made based on ost space available,
3819 ost previous usage, and OSS previous usage, in order to try
3820 to optimize storage space and networking resources.
3822 Severity : enhancement
3824 Description: Permanently set tunables
3825 Details : All writable /proc/fs/lustre tunables can now be permanently
3826 set on a per-server basis, at mkfs time or on a live
3829 Severity : enhancement
3831 Description: Lustre message v2
3832 Details : Add lustre message format v2.
3834 Severity : enhancement
3836 Description: client OST exclusion list
3837 Details : Clients can be started with a list of OSTs that should be
3838 declared "inactive" for known non-responsive OSTs.
3842 Description: ENOENT returned for valid filehandle during dbench.
3843 Details : Check if a directory has children when invalidating dentries
3844 associated with an inode during lock cancellation. This fixes
3845 an incorrect ENOENT sometimes seen for valid filehandles during
3846 testing with dbench.
3849 Frequency : SFS test only (otherwise harmless)
3851 Description: SPEC SFS validation failure on NFS v2 over lustre.
3852 Details : Changes the blocksize for regular files to be 2x RPC size,
3853 and not depend on stripe size.
3855 Severity : enhancement
3857 Description: fine-grained SMP locking inside DLM
3858 Details : Improve DLM performance on SMP systems by removing the single
3859 per-namespace lock and replace it with per-resource locks.
3861 Severity : enhancement
3863 Description: don't hold multiple extent locks at one time
3864 Details : To avoid client eviction during large writes, locks are not
3865 held on multiple stripes at one time or for very large writes.
3866 Otherwise, clients can block waiting for a lock on a failed OST
3867 while holding locks on other OSTs and be evicted.
3869 Severity : enhancement
3871 Description: Multiple MD RPCs in flight.
3872 Details : Further unserialise some read-only MDT RPCs - learn about intents.
3873 To avoid overly-overloading MDT, introduce a limit on number of
3874 MDT RPCs in flight for a single client and add /proc controls
3875 to adjust this limit.
3877 Severity : enhancement
3879 Description: client read/write statistics
3880 Details : Add client read/write call usage stats for performance
3881 analysis of user processes.
3882 /proc/fs/lustre/llite/*/offset_stats shows non-sequential
3883 file access. extents_stats shows chunk size distribution.
3884 extents_stats_per_process show chunk size distribution per
3887 Severity : enhancement
3889 Description: per-client statistics on server
3890 Details : Add ldlm and operations statistics for each client in
3891 /proc/fs/lustre/mds|obdfilter/*/exports/
3893 Severity : enhancement
3895 Description: improved MDT statistics
3896 Details : Add detailed MDT operations statistics in
3897 /proc/fs/lustre/mds/*/stats
3899 Severity : enhancement
3901 Description: VFS operations stats
3902 Details : Add client VFS call stats, trackable by pid, ppid, or gid
3903 /proc/fs/lustre/llite/*/stats_track_[pid|ppid|gid]
3908 Description: Fix client-side osc byte counters
3909 Details : The osc read/write byte counters in
3910 /proc/fs/lustre/osc/*/stats are now working
3913 Frequency : always as root on SLES
3915 Description: Failure of copying files with lustre special EAs.
3916 Details : Client side always return success for setxattr call for lustre
3917 special xattr (currently only "trusted.lov").
3922 Description: Refcount LNET uuids
3923 Details : The global LNET uuid list grew linearly with every startup;
3924 refcount repeated list entries instead of always adding to
3927 Severity : enhancement
3929 Description: Dynamic service threads
3930 Details : Within a small range, start extra service threads
3931 automatically when the request queue builds up.
3934 Frequency : mixed-endian client/server environments
3936 Description: mixed-endian crashes
3937 Details : The new msg_v2 system had some failures in mixed-endian
3940 Severity : enhancement
3942 Description: Easy OST removal
3943 Details : OSTs can be permanently deactivated with e.g. 'lctl
3944 conf_param lustre-OST0001.osc.active=0'
3946 Severity : enhancement
3948 Description: MGS proc entries
3949 Details : Added basic proc entries for the MGS showing what filesystems
3952 Severity : enhancement
3954 Description: provide MGS failover
3955 Details : Added config lock reacquisition after MGS server failover.
3957 Severity : enhancement
3959 Description: add Linux 2.4 support
3960 Details : Added support for RHEL 2.4.21 kernel for 1.6 servers and clients
3964 Description: a large application tries to do I/O to the same resource and dies
3965 in the middle of it.
3966 Details : Check the req->rq_arrival time after the call to
3967 ost_brw_lock_get(), but before we do anything about
3968 processing it & sending the BULK transfer request. This
3969 should help move old stale pending locks off the queue as
3970 quickly as obd_timeout.
3973 Frequency : when an incorrect nid is specified during startup
3975 Description: ptlrpc connect to non-existant node causes kernel crash
3976 Details : LNET can't be re-entered from an event callback, which
3977 happened when we expire a message after the export has been
3978 cleaned up. Instead, hand the zombie cleanup off to another
3981 Severity : enhancement
3983 Description: plain/inodebits lock performance improvement
3984 Details : Grouping plain/inodebits in granted list by their request modes
3985 and bits policy, thus improving the performance of search through
3989 Frequency : only if OST filesystem is corrupted
3991 Description: client incorrectly hits assertion in ptlrpc_replay_req()
3992 Details : for a short time RPCs with bulk IO are in the replay list,
3993 but replay of bulk IOs is unimplemented. If the OST filesystem
3994 is corrupted due to disk cache incoherency and then replay is
3995 started it is possible to trip an assertion. Avoid putting
3996 committed RPCs into the replay list at all to avoid this issue.
3999 Frequency : liblustre (e.g. catamount) on a large cluster with >= 8 OSTs/OSS
4001 Description: System hang on startup
4002 Details : This bug allowed the liblustre (e.g. catamount) client to
4003 return to the app before handling all startup RPCs. This
4004 could leave the node unresponsive to lustre network traffic
4005 and manifested as a server ptllnd timeout.
4007 Severity : enhancement
4009 Description: Add "/proc/sys/lustre/debug_peer_on_timeout"
4010 Details : liblustre envirable: LIBLUSTRE_DEBUG_PEER_ON_TIMEOUT
4011 boolean to control whether to print peer debug info when a
4012 client's RPC times out.
4015 Frequency : only for kernels with patches from Lustre below 1.4.3
4017 Description: Remove old rdonly API
4018 Details : Remove old rdonly API which unused from at least lustre 1.4.3
4021 Frequency : only for devices with external journals
4023 Description: Set external device read-only also
4024 Details : During a commanded failover stop, we set the disk device
4025 read-only while the server shuts down. We now also set any
4026 external journal device read-only at the same time.
4029 Frequency : when upgrading from 1.4 while trying to change parameters
4031 Description: The wrong (new) MDC name was used when setting parameters for
4032 upgraded MDT's. Also allows changing of OSC (and MDC)
4033 parameters if --writeconf is specified at tunefs upgrade time.
4036 Frequency : when setting specific ost indicies
4038 Description: QOS code breaks on skipped indicies
4039 Details : Add checks for missing OST indicies in the QOS code, so OSTs
4040 created with --index need not be sequential.
4042 Severity : enhancement
4044 Description: Add uninit_groups feature to ldiskfs2 to speed up e2fsck
4045 Details : The uninit_groups feature works in conjunction with the kernel
4046 filesystem code (ldiskfs2 only) and e2fsprogs-1.39-cfs6 to speed
4047 up the pass1 processing of e2fsck. This is a read-only feature
4048 in ldiskfs2 only, so older kernels and current ldiskfs cannot
4049 mount filesystems that have had this feature enabled.
4051 Severity : enhancement
4053 Description: Improve multi-block allocation algorithm to avoid fragmentation
4054 Details : The mballoc3 code (ldiskfs2 only) adds new mechanisms to improve
4055 allocation locality and avoid filesystem fragmentation.
4057 ------------------------------------------------------------------------------
4059 2007-04-01 Cluster File Systems, Inc. <info@clusterfs.com>
4061 * Support for kernels:
4062 2.4.21-47.0.1.EL (RHEL 3)
4063 2.6.5-7.283 (SLES 9)
4064 2.6.9-42.0.10.EL (RHEL 4)
4065 2.6.12.6 vanilla (kernel.org)
4066 2.6.16.27-0.9 (SLES 10)
4067 * Recommended e2fsprogs version: 1.39.cfs5
4069 * Note that reiserfs quotas are disabled on SLES 10 in this kernel
4073 Frequency : occasional, depends on client load and configuration
4074 Bugzilla : 12181, 12203
4075 Description: data loss for recently-modified files
4077 Details : In some cases it is possible that recently written or created
4078 files may not be written to disk in a timely manner (this should
4079 normally be within 30s unless client IO load is very high).
4080 The problem appears as zero-length files or files that are a
4081 multiple of 1MB in size after a client crash or client eviction
4082 that are missing data at the end of the file.
4084 This problem is more likely to be hit on clients where files are
4085 repeatedly created and unlinked in the same directory, clients
4086 have a large amount of RAM, have many CPUs, the filesystem has
4087 many OSTs, the clients are rebooted frequently, and/or the files
4088 are not accessed by other nodes after being written.
4090 The presence of the problem can be detected by looking at
4091 /proc/sys/fs/inode-state. If the first number (nr_inodes) is
4092 smaller than the second (nr_unused) then dirty files will not
4093 be flushed automatically to disk. "sync; sleep 10" should be
4094 run several times on the node before unmounting it to update
4095 Lustre (this is also safe to run on nodes without this problem).
4097 There is also a related kernel bug in the RHEL4 4 2.6.9 kernel
4098 that can cause this same problem, so customers using that kernel
4099 also need to update the kernel in addition to Lustre. In order
4100 to properly fix this bug, the RHEL3 2.4.21 kernel is also updated.
4102 It is normal that files written just before a client crash (less
4103 than 30s) may not yet have been flushed to disk, even for local
4107 Frequency : frequent on thin XT3 nodes
4109 Description: UUID collision on thin XT3 Linux nodes
4110 Details : UUIDs on Compute Node Linux XT3 nodes were not generated
4111 randomly, since we relied on an insufficiently-seeded PRNG.
4116 Description: OSS hangs after "All ost request buffers busy"
4117 Details : A deadlock between quota and journal operations caused OSS
4118 hangs after printing "All ost request buffers busy."
4121 Frequency : always on liblustre builds
4123 Description: Cleanup compiler warnings on liblustre
4126 Frequency : always on liblustre builds on XT3
4128 Description: LC_CONFIG_CDEBUG don't run while build liblustre on XT3.
4132 Description: Addition of EXT3_FEATURE_RO_COMPAT_DIR_NLINKS flag for
4133 > 32000 subdirectories
4134 Details : Add EXT3_FEATURE_RO_COMPAT_DIR_NLINK flag to
4135 EXT3_FEATURE_RO_COMPAT_SUPP. This flag will be set whenever
4136 subdirectory count crosses 32000. This will aid e2fsck to
4137 correctly handle more than 32000 subdirectories.
4140 Frequency : liblustre (e.g. catamount) on a large cluster with >= 8 OSTs/OSS
4142 Description: System hang on startup
4143 Details : This bug allowed the liblustre (e.g. catamount) client to
4144 return to the app before handling all startup RPCs. This
4145 could leave the node unresponsive to lustre network traffic
4146 and manifested as a server ptllnd timeout.
4148 Severity : enhancement
4150 Description: Add "/proc/sys/lustre/debug_peer_on_timeout"
4151 (liblustre envirable: LIBLUSTRE_DEBUG_PEER_ON_TIMEOUT)
4152 boolean to control whether to print peer debug info when a
4153 client's RPC times out.
4158 Description: make O_SYNC working on 2.6 kernels
4159 Details : 2.6 kernels use different method for mark pages for write,
4160 so need add a code to lustre for O_SYNC work.
4165 Description: Failure to close file and release space on NFS
4166 Details : Put inode details into lock acquired in ll_intent_file_open.
4167 Use mdc_intent_lock in ll_intent_open to properly
4168 detect all kind of errors unhandled by mdc_enqueue.
4173 Description: proc file read during shutdown sometimes raced obd removal,
4175 Details : Add lock to prevent obd access after proc file removal.
4178 Frequency : Only for files larger than 4GB on 32-bit clients.
4180 Description: improperly doing page alignment of locks
4181 Details : Modify lustre core code to use CFS_PAGE_* defines instead of
4182 PAGE_*. Make CFS_PAGE_MASK a 64-bit mask.
4187 Description: RPCs being resent when they shouldn't be
4188 Details : Some RPCs that should not be resent are being resent. This
4189 can cause inconsistencies in the RPC state machine. Do not
4190 resend such requests.
4193 Frequency : rare, only with NFS export
4195 Description: Crash on NFS re-export node
4196 Details : under very unusual load conditions an assertion is hit in
4197 ll_intent_file_open()
4200 Frequency : only if OST filesystem is corrupted
4202 Description: client incorrectly hits assertion in ptlrpc_replay_req()
4203 Details : for a short time RPCs with bulk IO are in the replay list,
4204 but replay of bulk IOs is unimplemented. If the OST filesystem
4205 is corrupted due to disk cache incoherency and then replay is
4206 started it is possible to trip an assertion. Avoid putting
4207 committed RPCs into the replay list at all to avoid this issue.
4212 Description: large O_DIRECT requests fail under memory pressure/fragmentation
4213 Details : Large single O_DIRECT read and write calls can fail to allocate
4214 a sufficiently large buffer to process the request. In case of
4215 allocation failure the allocation is retried with a smaller
4216 buffer and broken into smaller requests.
4218 Severity : enhancement
4220 Description: Add -o localflock option to simulate old noflock behaviour.
4221 Details : This will achieve local-only flock/fcntl locks coherentness.
4226 Description: versioning check is incomplete
4227 Details : Checking the version difference of client vs. server, report
4228 error if the gap is too big.
4233 Description: add support PG_writeback bit
4234 Details : add support for PG_writeback bit for Lustre, for more carefull
4235 work with page cache in 2.6 kernel. This also fix some deadlocks
4236 and remove hack for work O_SYNC with 2.6 kernel.
4238 Severity : enhancement
4240 Description: Add uninit_groups feature to ldiskfs2 to speed up e2fsck
4241 Details : The uninit_groups feature works in conjunction with the kernel
4242 filesystem code (ldiskfs2 only) and e2fsprogs-1.39-cfs6 to speed
4243 up the pass1 processing of e2fsck. This is a read-only feature
4244 in ldiskfs2 only, so older kernels and current ldiskfs cannot
4245 mount filesystems that have had this feature enabled.
4247 Severity : enhancement
4249 Description: Improve multi-block allocation algorithm to avoid fragmentation
4250 Details : The mballoc3 code (ldiskfs2 only) adds new mechanisms to improve
4251 allocation locality and avoid filesystem fragmentation.
4253 ------------------------------------------------------------------------------
4255 2007-02-09 Cluster File Systems, Inc. <info@clusterfs.com>
4257 * Support for kernels:
4258 2.6.9-42.0.3.EL (RHEL 4)
4259 2.6.5-7.276 (SLES 9)
4260 2.4.21-47.0.1.EL (RHEL 3)
4261 2.6.12.6 vanilla (kernel.org)
4262 2.6.16.21-0.8 (SLES10)
4263 * Recommended e2fsprogs version: 1.39.cfs2-0
4265 * The backwards-compatible /proc/sys/portals symlink has been removed
4266 in this release. Before upgrading, please ensure that you change
4267 any configuration scripts or /etc/sysctl.conf files that access
4268 /proc/sys/portals/* or sysctl portals.* to use the corresponding
4269 entry in /proc/sys/lnet or sysctl lnet.*. This change can be made
4270 in advance of the upgrade on any system running Lustre 1.4.6 or
4271 newer, since /proc/sys/lnet was added in that version.
4272 * Note that reiserfs quotas are disabled on SLES 10 in this kernel
4276 Frequency : only when quota is used
4278 Description: avoid scanning export list for quota master
4279 Details : Change the algorithms to avoid scanning export list in order
4280 to improve the efficiency.
4283 Frequency : MDS failover only, very rarely
4285 Description: "went back in time" messages on mds failover
4286 Details : The greatest transno may be lost when the current operation
4287 finishes with an error (transno==0) and the client's last_rcvd
4288 record is over-written. Save the greatest transno in the
4289 mds_last_transno for this case.
4292 Frequency : always for specific kernels and striping counts
4294 Description: client may get "Matching packet too big" without ACL support
4295 Details : Clients compiled without CONFIG_FS_POSIX_ACL get an error message
4296 when trying to access files in certain configurations. The
4297 clients should in fact be denied when mounting because they do
4298 not understand ACLs.
4301 Frequency : Cray XT3 with more than 4000 clients and multiple jobs
4303 Description: many clients connecting with IO in progress causes connect timeouts
4304 Details : Avoid synchronous journal commits to avoid delays caused by many
4305 clients connecting/disconnecting when bulk IO is in progress.
4306 Queue liblustre connect requests on OST_REQUEST_PORTAL instead of
4307 OST_IO_PORTAL to avoid delays behind potentially many pending
4311 Frequency : occasionally with multiple writers to a single file
4313 Description: shared writes to file may result in wrong size reported by stat()
4314 Details : Allow growing of kms when extent lock is cancelled
4317 Frequency : always with random mmap IO to multi-striped file
4319 Description: mmap write might be lost if we are writing to a 'hole' in stripe
4320 Details : Only if the hole is at the end of OST object so that kms is too
4321 small. Fix is to increase kms accordingly in ll_nopage.
4324 Frequency : rare, only if OST filesystem is inconsistent with MDS filesystem
4326 Description: writes to a missing object would leak memory on the OST
4327 Details : If there is an inconsistency between the MDS and OST filesystems,
4328 such that the MDS references an object that doesn't exist, writes
4329 to that object will leak memory due to incorrect cleanup in the
4330 error handling path, eventually running out of memory on the OST.
4335 Description: Creating too long symlink causes lustre errors
4336 Details : Check symlink and name lengths before sending requests to MDS.
4339 Frequency : only if flock is enabled (not on by default)
4341 Description: posix locks not released on fd closure on 2.6.9+
4342 Details : We failed to add posix locks to list of inode locks on 2.6.9+
4343 kernels, this caused such locks not to be released on fd close and
4344 then assertions on fs unmount about still used locks.
4347 Frequency : MDS failover only, very rarely
4349 Description: clients may get ASSERTION(granted_lock != NULL)
4350 Details : When request was taking a long time, and a client was resending
4351 a getattr by name lock request. The were multiple lock requests
4352 with the same client lock handle and
4353 mds_getattr_name->fixup_handle_for_resent_request found one of the
4354 lock handles but later failed with ASSERTION(granted_lock != NULL).
4359 Description: handle->h_buffer_credits > 0, assertion failure
4360 Details : h_buffer_credits is zero after truncate, causing assertion
4361 failure. This patch extends the transaction or creates a new
4365 Frequency : NFS re-export or patchless client
4366 Bugzilla : 11179, 10796
4367 Description: Crash on NFS re-export node (__d_move)
4368 Details : We do not want to hash the dentry if we don't have a lock.
4369 But if this dentry is later used in d_move, we'd hit uninitialised
4370 list head d_hash, so we just do this to init d_hash field but
4371 leave dentry unhashed.
4374 Frequency : NFS re-export or patchless client
4376 Description: NFS exports has problem with symbolic link
4377 Details : lustre client didn't properly install dentry when re-exported
4378 to NFS or running patchless client.
4381 Frequency : NFS re-export or patchless client
4383 Description: Various nfs/patchless fixes.
4384 Details : fixes reuse disconected alias for lookup process - this fixes
4385 warning "find_exported_dentry: npd != pd",
4386 fix permission error with open files at nfs.
4387 fix apply umask when do revalidate.
4390 Frequency : occasional
4392 Description: Crash on NFS re-export node
4393 Details : calling clear_page() on the wrong pointer triggered oops in
4394 generic_mapping_read().
4397 Frequency : rarely, using O_DIRECT IO
4399 Description: unaligned directio crashes client with LASSERT
4400 Details : check for unaligned buffers before trying any requests.
4403 Frequency : rarely, using CFS RAID5 patches in non-standard kernel series
4405 Description: stale data returned from RAID cache
4406 Details : If only a small amount of IO is done to the RAID device before
4407 reading it again it is possible to get stale data from the RAID
4408 cache instead of reading it from disk.
4411 Frequency : always for sles10 kernel
4413 Description: sles10 support
4414 Details : ll_follow_link: compile fixes and using of nd_set_link
4415 under newer kernels.
4418 Frequency : depends on arch, kernel and compiler version, always on sles10
4421 Description: recursive or deep enough symlinks cause stack overflow
4422 Details : getting rid of large stack-allocated variable in
4426 Frequency : depends on hardware
4428 Description: lustre write performance loss in the SLES10 kernel
4429 Details : the performance loss is caused by using of write barriers in the
4430 ext3 code. The SLES10 kernel turns barrier support on by
4431 default. The fix is to undo that change for ldiskfs.
4433 ------------------------------------------------------------------------------
4435 2006-12-09 Cluster File Systems, Inc. <info@clusterfs.com>
4437 * Support for kernels:
4438 2.6.9-42.0.3EL (RHEL 4)
4439 2.6.5-7.276 (SLES 9)
4440 2.4.21-47.0.1.EL (RHEL 3)
4441 2.6.12.6 vanilla (kernel.org)
4445 Frequency : quota enabled and large files being deleted
4447 Description: releasing more than 4GB of quota at once hangs OST
4448 Details : If a user deletes more than 4GB of files on a single OST it
4449 will cause the OST to spin in an infinite loop. Release
4450 quota in < 4GB chunks, or use a 64-bit value for 1.4.7.1+.
4455 Description: statfs data retrieved from /proc may be stale or zero
4456 Details : When reading per-device statfs data from /proc, in the
4457 {kbytes,files}_{total,free,avail} files, it may appear
4458 as zero or be out of date.
4461 Frequency : systems with MD RAID1 external journal devices
4463 Description: lconf's call to blkid is confused by RAID1 journal devices
4464 Details : Use the "blkid -l" flag to locate the MD RAID device instead
4465 of returning all block devices that match the journal UUID.
4468 Frequency : always, for aggregate stripe size over 4GB
4470 Description: "lfs setstripe" fails assertion when setting 4GB+ stripe width
4471 Details : Using "lfs setstripe" to set stripe size * stripe count over 4GB
4472 will fail the kernel with "ASSERTION(lsm->lsm_xfersize != 0)"
4475 Frequency : always if "lfs find" used on a local file/directory
4477 Description: "lfs find" segfaults if used on a local file/directory
4478 Details : The case where a directory component was not specified wasn't
4479 handled correctly. Handle this properly.
4482 Frequency : always on ppc64
4484 Description: the write to an ext3 filesystem mounted with mballoc got stuck
4485 Details : ext3_mb_generate_buddy() uses find_next_bit() which does not
4486 perform endianness conversion.
4489 Frequency : rarely (truncate to non-zero file size after write under load)
4490 Bugzilla : 10730, 10687
4491 Description: Files padded with zeros to next 4K multiple
4492 Details : With filesystems mounted using the "extents" option (2.6 kernels)
4493 it is possible that files that are truncated to a non-zero size
4494 immediately after being written are filled with zero bytes beyond
4495 the truncated size. No file data is lost.
4497 Severity : enhancement
4499 Description: Allow recovery/failover for liblustre clients.
4500 Details : liblustre clients were unaware of failover configurations until
4503 Severity : enhancement
4505 Description: user file locks should fail when not mounting with flock option
4506 Details : Set up an error-returning stub in ll_file_operations.lock field
4507 to prevent incorrect behaviour when client is mounted without
4508 flock option. Also, set up properly f_op->flock field for
4512 Frequency : always on ia64
4514 Description: "lfs df" loops on printing out MDS statfs information
4515 Details : The obd_ioctl_data was not initialized and in some systems
4516 this caused a failure during the ioctl that did not return
4517 an error. Initialize the struct and return an error on failure.
4520 Frequency : SLES 9 only
4522 Description: Error of copying files with lustre special EAs as root
4523 Details : Client side always return success for setxattr call for lustre
4524 special xattr (currently only "trusted.lov").
4527 Frequency : rarely on clusters with both ia64+i386 clients
4529 Description: ia64+i686 clients doing shared IO on the same file may LBUG
4530 Details : In rare cases when both ia64+i686 (or other mixed-PAGE_SIZE)
4531 clients are doing concurrent writes to the same file it is
4532 possible that the ia64 clients may LASSERT because the OST
4533 extent locks are not PAGE_SIZE aligned. Ensure that grown
4534 locks are always aligned on the request boundary.
4537 Frequency : specific use, occasional
4539 Description: Overwriting in use executable truncates on-disk binary image
4540 Details : If one node attempts to overwrite an executable in use by
4541 another node, we now correctly return ETXTBSY instead of
4542 truncating the file.
4544 Severity : enhancement
4546 Description: Async OSC create to avoid the blocking unnecessarily.
4547 Details : If a OST has no remain object, system will block on the creating
4548 when need to create a new object on this OST. Now, ways use
4549 pre-created objects when available, instead of blocking on an
4550 empty osc while others are not empty. If we must block, we block
4551 for the shortest possible period of time.
4556 Description: chmod on Lustre root is propagated to other clients
4557 Details : Re-validate root's dentry in ll_lookup_it to avoid having it
4558 invalid by the follow_mount time.
4561 Frequency : liblustre clients only
4563 Description: Race in 'instant cancel' lock handling could lead to such locks
4564 never to be granted in case of SMP MDS
4565 Details : Do not destroy not yet granted but cbpending locks in
4569 Frequency : replay/resend of open
4571 Description: non null lock assetion failure in mds_intent_policy
4572 Details : Trying to replay/resend lockless open requests resulted in
4573 mds_open() returning 0 with no lock. Now it sets a flag if
4574 a lock is going to be returned.
4576 Severity : enhancement
4578 Description: Checksum enhancements
4579 Details : New checksum enhancements allow for resending RPCs that failed
4582 Severity : enhancement
4584 Description: Tunables on number of dirty pages in cacche
4585 Details : Allow to set limit on number of dirty pages cached.
4590 Description: client crash on unmount - lock still has references
4591 Details : In some error handling cases it was possible to leak a lock
4592 reference on a client while accessing a file. This was not
4593 harmful to the client during operation, but would cause the
4594 client to crash when the filesystem is unmounted.
4597 Frequency : specific case, rare
4599 Description: ETXTBSY on mds though file not in use by client
4600 Details : ETXTBSY is no longer incorrectly returned when attempting to
4601 chmod or chown a directory that the user previously tried to
4602 execute or a currently-executing binary.
4605 Frequency : extremely rare except on liblustre-based clients
4607 Description: Lustre space not freed when files are deleted
4608 Details : Clean up open-unlinked files after client eviction. Previously
4609 the unlink was skipped and the files remained as orphans.
4614 Description: OST failure "would be an LBUG" in waiting_locks_callback()
4615 Details : In some cases it was possible to send a blocking callback to
4616 a client doing a glimpse, even though that client didn't get
4617 a lock granted. When the glimpse lock is cancelled on the OST
4618 the freed lock is left on the waiting list and corrupted the list.
4621 Frequency : all core dumps
4623 Description: Broke core dumps to lustre
4624 Details : Negative dentry may be unhashed if parent does not have UPDATE
4625 lock, but some callers, e.g. do_coredump, expect dentry to be
4626 hashed after successful create, hash it in ll_create_it.
4628 ------------------------------------------------------------------------------
4630 2006-09-13 Cluster File Systems, Inc. <info@clusterfs.com>
4632 * Support for kernels:
4633 2.6.9-42.0.2.EL (RHEL 4)
4634 2.6.5-7.276 (SLES 9)
4635 2.4.21-40.EL (RHEL 3)
4636 2.6.12.6 vanilla (kernel.org)
4640 Frequency : always on RHEL 3
4642 Description: Number of open files grows over time
4643 Details : The number of open files grows over time, whether or not
4644 Lustre is started. This was due to a filp leak introduced
4645 by one of our kernel patches.
4647 ------------------------------------------------------------------------------
4649 2006-08-20 Cluster File Systems, Inc. <info@clusterfs.com>
4651 * Support for kernels:
4652 2.6.9-42.EL (RHEL 4)
4653 2.6.5-7.267 (SLES 9)
4654 2.4.21-40.EL (RHEL 3)
4655 2.6.12.6 vanilla (kernel.org)
4660 Bugzilla : 5719, 9635, 9792, 9684
4661 Description: OST (or MDS) trips assertions in (re)connection under heavy load
4662 Details : If a server is under heavy load and cannot reply to new
4663 connection requests before the client resends the (re)connect,
4664 the connection handling code can behave badly if two service
4665 threads are concurrently handing separate (re)connections from
4666 the same client. Add better locking to the connection handling
4667 code, and ensure that only a single connection will be processed
4668 for a given client UUID, even if the lock is dropped.
4670 Severity : enhancement
4672 Description: add TCP zero-copy support to kernel
4673 Details : Add support to the kernel TCP stack to allow zero-copy bulk
4674 sends if the hardware supports scatter-gather and checksumming.
4675 This allows socklnd to do client-write and server-read more
4676 efficiently and reduce CPU utilization from skbuf copying.
4679 Frequency : only if NFS exporting from client
4681 Description: NULL pointer deref in ll_iocontrol() if chattr mknod file
4682 Details : If setting attributes on a file created under NFS that had
4683 never been opened it would be possible to oops the client
4684 if the file had no objects.
4688 Bugzilla : 9326, 10402, 10897
4689 Description: client crash in ptlrpcd_wake() thread when sending async RPC
4690 Details : It is possible that ptlrpcd_wake() dereferences a freed async
4691 RPC. In rare cases the ptlrpcd thread alread processed the RPC
4692 before ptlrpcd_wake() was called and the request was freed.
4695 Frequency : always for liblustre
4697 Description: liblustre client does MDS+OSTs setattr RPC for each write
4698 Details : When doing a write from a liblustre client, the client
4699 incorrectly issued an RPC to the MDS and each OST the file was
4700 striped over in order to update the timestamps. When writing
4701 with small chunks and many clients this could overwhelm the MDS
4702 with RPCs. In all cases it would slow down the write because
4703 these RPCs are unnecessary.
4705 Severity : enhancement
4707 Description: allow number of MDS service threads to be changed at module load
4708 Details : It is now possible to change the number of MDS service threads
4709 running. Adding "options mds mds_num_threads={N}" to the MDS's
4710 /etc/modprobe.conf will set the number of threads for the next
4711 time Lustre is restarted (assuming the "mds" module is also
4712 reloaded at that time). The default number of threads will
4713 stay the same, 32 for most systems.
4718 Description: OST crash if filesystem is unformatted or corrupt
4719 Details : If an OST is started on a device that has never been formatted
4720 or if the filesystem is corrupt and cannot even mount then the
4721 error handling cleanup routines would dereference a NULL pointer.
4726 Description: NULL pointer deref in llap_from_page.
4727 Details : get_cache_page_nowait can return a page with NULL (or otherwise
4728 incorrect) mapping if the page was truncated/reclaimed while it was
4729 searched for. Check for this condition and skip such pages when
4730 doing readahead. Introduce extra check to llap_from_page() to
4731 verify page->mapping->host is non-NULL (so page is not anonymous).
4734 Frequency : Sometimes when using sys_sendfile
4736 Description: "page not covered by a lock" warnings from ll_readpage
4737 Details : sendfile called ll_readpage without right page locks present.
4738 Now we introduced ll_file_sendfile that does necessary locking
4739 around call to generic_file_sendfile() much like we do in
4743 Frequency : with certain MDS communication failures at client mount time
4745 Description: NULL pointer deref after failed client mount
4746 Details : a client connection request may delayed by the network layer
4747 and not be sent until after the PTLRPC layer has timed out the
4748 request. If the client fails the mount immediately it will try
4749 to clean up before the network times out the request. Add a
4750 reference from the request import to the obd device and delay
4751 the cleanup until the network drops the request.
4754 Frequency : occasionally during client (re)connect
4756 Description: assertion failure during client (re)connect
4757 Details : processing a client connection request may be delayed by the
4758 client or server longer than the client connect timeout. This
4759 causes the client to resend the connection request. If the
4760 original connection request is replied in this interval, the
4761 client may trip an assertion failure in ptlrpc_connect_interpret()
4762 which thought it would be the only running connect process.
4765 Frequency : only with obd_echo servers and clients that are rebooted
4767 Description: kernel BUG accessing uninitialized data structure
4768 Details : When running an obd_echo server it did not start the ping_evictor
4769 thread, and when a client was evicted an uninitialized data
4770 structure was accessed. Start the ping_evictor in the RPC
4771 service startup instead of the OBD startup.
4773 Severity : enhancement
4774 Bugzilla : 10193 (patchless)
4775 Description: Remove dependency on various unexported kernel interfaces.
4776 Details : No longer need reparent_to_init, exit_mm, exit_files,
4777 sock_getsockopt, filemap_populate, FMODE_EXEC, put_filp.
4780 Frequency : rare (only users of deprecated and unsupported LDAP config)
4782 Description: write_conf for zeroconf mount queried LDAP incorrectly for client
4783 Details : LDAP apparently contains 'lustreName' attributes instead of
4784 'name'. A simple remapping of the name is sufficient.
4787 Frequency : rare (only with non-default dump_on_timeout debug enabled)
4789 Description: waiting_locks_callback trips kernel BUG if client is evicted
4790 Details : Running with the dump_on_timeout debug flag turned on makes
4791 it possible that the waiting_locks_callback() can try to dump
4792 the Lustre kernel debug logs from an interrupt handler. Defer
4793 this log dumping to the expired_lock_main() thread.
4795 Severity : enhancement
4797 Description: Support NFS exporting on 2.6 kernels.
4798 Details : Implement non-rawops metadata methods for NFS server to use without
4799 changing NFS server code.
4802 Frequency : very rare (synthetic metadata workload only)
4804 Description: two racing renames might cause an MDS thread to deadlock
4805 Details : Running the "racer" program may cause one MDS thread to rename
4806 a file from being the source of a rename to being the target of
4807 a rename at exactly the same time that another thread is doing
4808 so, and the second thread has already enqueued these locks after
4809 doing a lookup of the target and is trying to relock them in
4810 order. Ensure that we don't try to re-lock the same resource.
4813 Frequency : only very large systems with liblustre clients
4815 Description: slow eviction of liblustre clients with the "evict_by_nid" RPC
4816 Details : Use asynchronous set_info RPCs to send the "evict_by_nid" to
4817 all OSTs in parallel. This allows the eviction of stale liblustre
4818 clients to proceed much faster than if they were done in series,
4819 and also offers similar improvements for other set_info RPCs.
4824 Description: excessive CPU usage during initial read phase on client
4825 Details : During the initial read phase on a client, it would agressively
4826 retry readahead on the file, consuming too much CPU and impacting
4827 performance (since 1.4.5.8). Improve the readahead algorithm
4828 to avoid this, and also improve some other common cases (read
4829 of small files in particular, where "small" is files smaller than
4830 /proc/fs/lustre/llite/*/max_read_ahead_whole_mb, 2MB by default).
4835 Description: MDS crash when receiving packet with unknown intent.
4836 Details : Do not LBUG in unknown intent case, just return -EFAULT
4838 Severity : enhancement
4839 Bugzilla : 9293, 9385
4840 Description: MDS RPCs are serialised on client. This is unnecessary for some.
4841 Details : Do not serialize getattr (non-intent version) and statfs.
4844 Frequency : occasional, when OST network is overloaded/intermittent
4846 Description: client evicted by OST after bulk IO timeout
4847 Details : If a client sends a bulk IO request (read or write) the OST
4848 may evict the client if it is unresposive to its data GET/PUT
4849 request. This is incorrect if the network is overloaded (takes
4850 too long to transfer the RPC data) or dropped the OST GET/PUT
4851 request. There is no need to evict the client at all, since
4852 the pinger and/or lock callbacks will handle this, and the
4853 client can restart the bulk request.
4856 Frequency : Always when mmapping file with no objects
4858 Description: client crashes when mmapping file with no objects
4859 Details : Check that we actually have objects in a file before doing any
4860 operations on objects in ll_vm_open, ll_vm_close and
4866 Description: Request leak when working with deleted CWD
4867 Details : Introduce advanced request refcount tracking for requests
4868 referenced from lustre intent.
4870 Severity : Enhancement
4872 Description: Cache open file handles on client.
4873 Details : MDS now will return special lock along with openhandle, if
4874 requested and client is allowed to hold openhandle, even if unused,
4875 until such a lock is revoked. Helps NFS a lot, since NFS is opening
4876 closing files for every read/write openration.
4878 Severity : Enhancement
4880 Description: Cache open negative dentries on client when possible.
4881 Details : Guard negative dentries with UPDATE lock on parent dir, drop
4882 negative dentries on lock revocation.
4887 Description: Remounting a client read-only wasn't possible with a zconf mount
4888 Details : It wasn't possible to remount a client read-only with llmount.
4890 Severity : enhancement
4891 Description: Include MPICH 1.2.6 Lustre ADIO interface patch
4892 Details : In lustre/contrib/ or /usr/share/lustre in RPM a patch for
4893 MPICH is included to add Lustre-specific ADIO interfaces.
4894 This is based closely on the UFS ADIO layer and only differs
4895 in file creation, in order to allow the OST striping to be set.
4896 This is user-contributed code and not supported by CFS.
4901 Description: extended inode attributes (immutable, append-only) work improperly
4902 when 2.4 and 2.6 kernels are used on client/server or vice versa
4903 Details : Introduce kernel-independent values for these flags.
4905 Severity : enhancement
4908 Description: Allow fractional MB tunings for lustre in /proc/ filesystem.
4909 Details : Many of the /proc/ tunables can only be tuned at a megabyte
4910 granularity. Now, Fractional MB granularity is be supported,
4911 this is very useful for low memory system.
4913 Severity : enhancement
4915 Description: Getattr by fid
4916 Details : Getting a file attributes by its fid, obtaining UPDATE|LOOKUP
4917 locks, avoids extra getattr rpc requests to MDS, allows '/' to
4918 have locks and avoids getattr rpc requests for it on every stat.
4921 Frequency : Always, for filesystems larger than 2TB
4923 Description: ldiskfs crash at mount for filesystem larger than 2TB with mballoc
4924 Details : Kenrel kmalloc limits allocations to 128kB and this prevents
4925 filesystems larger than 2TB to be mounted with mballoc enabled.
4928 Frequency : Always, for 32-bit kernel without CONFIG_LBD and filesystem > 2TB
4930 Description: filesystem corruption for non-standard kernels and very large OSTs
4931 Details : If a 32-bit kernel is compiled without CONFIG_LBD enabled and a
4932 filesystems larger than 2TB is mounted then the kernel will
4933 silently corrupt the start of the filesystem. CONFIG_LBD is
4934 enabled for all CFS-supported kernels, but the possibility of
4935 this happening with a modified kernel config exists.
4937 Severity : enhancement
4939 Description: add client O_DIRECT support for 2.6 kernels
4940 Details : It is now possible to do O_DIRECT reads and writes to files
4941 in the Lustre client mountpoint on 2.6 kernel clients.
4943 Severity : enhancement
4945 Description: parallel glimpse, setattr, statfs, punch, destroy requests
4946 Details : Sends glimpse, setattr, statfs, punch, destroy requests to OSTs in
4947 parallel, not waiting for response from every OST before sending
4948 a rpc to the next OST.
4953 Description: setattr vs write race when updating file timestamps
4954 Details : Client processes that update a file timestamp into the past
4955 right after writing to the file (e.g. tar) it is possible that
4956 the updated file modification time can be reset to the current
4957 time due to a race between processing the setattr and write RPC.
4959 Severity : enhancement
4961 Description: Bring 'lfs find' closer in line with regular Linux find.
4962 Details : lfs find util supports -atime, -mtime, -ctime, -maxdepth, -print,
4963 -print0 options and obtains all the needed info through the lustre
4966 Severity : enhancement
4968 Description: support up to 1024 configured devices on one node
4969 Details : change obd_dev array from statically allocated to dynamically
4970 allocated structs as they are first used to reduce memory usage
4975 Description: Flush dirty partially truncated pages during truncate
4976 Details : Immediatelly flush partially truncated pages in filter_setattr,
4977 this way we completely avoid having any pages in page cache on OST
4978 and can retire ugly workarounds during writes to flush such pages.
4983 Description: i_sem vs transaction deadlock in mds_obd_destroy during unlink.
4984 Details : protect inode from truncation within vfs_unlink() context
4985 just take a reference before calling vfs_unlink() and release it
4986 when parent's i_sem is free.
4989 Frequency : always, if extents are used on OSTs
4991 Description: index ei_leaf_hi (48-bit extension) is not zeroed in extent index
4992 Details : OSTs using the extents format would not zero the high 16 bits of
4993 the index physical block number. This is not a problem for any
4994 OST filesystems smaller than 16TB, and no kernels support ext3
4995 filesystems larger than 16TB yet. This is fixed in 1.4.7 (all
4996 new/modified files) and can be fixed for existing filesystems
4997 with e2fsprogs-1.39-cfs1.
5002 Description: import connection selection may be incorrect if timer wraps
5003 Details : Using a 32-bit jiffies timer with HZ=1000 may cause backup
5004 import connections to be ignored if the 32-bit jiffies counter
5005 wraps. Use a 64-bit jiffies counter.
5008 Frequency : during server recovery
5010 Description: crash after server is denying duplicate export
5011 Details : If clients are resending connect requests to the server, the
5012 server refuses to allow a client to connect multiple times.
5013 Fixed a bug in the handling of this case.
5016 Frequency : very large clusters immediately after boot
5018 Description: LNET request buffers exhausted under heavy short-term load
5019 Details : If a large number of client requests are generated on a service
5020 that has previously never seen so many requests it is possible
5021 that the request buffer growth cannot keep up with the spike in
5022 demand. Instead of dropping incoming requests, they are held in
5023 the LND until the RPC service can accept more requests.
5026 Frequency : Sometimes during replay
5028 Description: Assertion failure in ll_local_open after replay.
5029 Details : If replay happened on an open request reply before we were able
5030 to set replay handler, reply will become not swabbed tripping the
5031 assertion in ll_local_open. Now we set the handler right after
5032 recognising of open request
5035 Frequency : very rare
5037 Description: kernel reports "badness in vsnprintf"
5038 Details : Reading from the "recovery_status" /proc file in small chunks
5039 may cause a negative length in lprocfs_obd_rd_recovery_status()
5040 call to vsnprintf() (which is otherwise harmless). Exit early
5041 if there is no more space in the output buffer.
5043 Severity : enhancement
5045 Description: clear OBD RPC statistics by writing to them
5046 Details : It is now possible to clear the OBD RPC statistics by writing
5047 to the "stats" file.
5052 Description: Client mtime is not the same on different clients after utimes
5053 Details : In some cases, the client was using the utimes() syscall on
5054 a file cached on another node. The clients now validate the
5055 ctime from the MDS + OSTs to determine which one is right.
5060 Description: Inability to activate failout mode
5061 Details : lconf script incorrectly assumed that in python string's numeric
5062 value is used in comparisons.
5065 Frequency : always with multiple stripes per file
5067 Description: Inefficient object allocation for mutli-stripe files
5068 Details : When selecting which OSTs to stripe files over, for files with
5069 a stripe count that divides evenly into the number of OSTs,
5070 the MDS is always picking the same starting OST for each file.
5071 Return the OST selection heuristic to the original design.
5076 Description: mount failures may take full timeout to return an error
5077 Details : Under some heavy load conditions it is possible that a
5078 failed mount can wait for the full obd_timeout interval,
5079 possibly several minutes, before reporting an error.
5080 Instead return an error as soon as the status is known.
5082 ------------------------------------------------------------------------------
5084 2006-02-14 Cluster File Systems, Inc. <info@clusterfs.com>
5086 * WIRE PROTOCOL CHANGE. This version of Lustre networking WILL NOT
5087 INTEROPERATE with older versions automatically. Please read the
5088 user documentation before upgrading any part of a live system.
5089 * WARNING: Lustre networking configuration changes are required with
5090 this release. See https://bugzilla.clusterfs.com/show_bug.cgi?id=10052
5093 * Support for kernels:
5094 2.6.9-22.0.2.EL (RHEL 4)
5095 2.6.5-7.244 (SLES 9)
5096 2.6.12.6 vanilla (kernel.org)
5099 Severity : enhancement
5100 Bugzilla : 7981/8208
5101 Description: Introduced Lustre Networking (LNET)
5102 Details : LNET is new networking infrastructure for Lustre, it includes
5103 a reorganized network configuration mode (see the user
5104 documentation for full details) as well as support for routing
5105 between different network fabrics. Lustre Networking Devices
5106 (LNDs) for the supported network fabrics have also been
5107 created for this new infrastructure.
5109 Severity : enhancement
5110 Description: Introduced Access control lists
5111 Details : clients can set ACLs on files and directories in order to have
5112 more fine-grained permissions than the standard Unix UGO+RWX.
5113 The MDS must be started with the "-o acl" mount option.
5115 Severity : enhancement
5116 Description: Introduced filesystem quotas
5117 Details : Administrators may now establish per-user quotas on the
5120 Severity : enhancement
5122 Description: Configuration change for the XT3
5123 The PTLLND is now used to run Lustre over Portals on the XT3
5124 The configure option(s) --with-cray-portals are no longer used.
5125 Rather --with-portals=<path-to-portals-includes> is used to
5126 enable building on the XT3. In addition to enable XT3 specific
5127 features the option --enable-cray-xt3 must be used.
5132 Description: Running on many-way SMP OSTs can trigger oops in llcd_send()
5133 Details : A race between allocating a new llcd and re-getting the llcd_lock
5134 allowed another thread to grab newly-allocated llcd.
5136 Severity : enhancement
5138 Description: 2.6 OST async journal commit and locking fix to improve performance
5139 Details : The filter_direct_io()+filter_commitrw_write() journal commits for
5140 2.6 kernels are now async as they already were in 2.4 kernels so
5141 that they can commit concurrently with the network bulk transfer.
5142 For block-allocated files the filter allocation semaphore is held
5143 to avoid filesystem fragmentation during allocation. BKL lock
5144 removed for 2.6 xattr operations where it is no longer needed.
5149 Description: lconf incorrectly determined whether two IP networks could talk
5150 Details : In some more complicated routing and multiple-network
5151 configurations, lconf will avoid trying to make a network
5152 connection to a disjoint part of the IP space. It was doing the
5153 math incorrectly for one set of cases.
5158 Description: Fix for potential infinite loop processing records in an llog.
5159 Details : If an llog record is corrupted/zeroed, it is possible to loop
5160 forever in llog_process(). Validate the llog record length
5161 and skip the remainder of the block on error.
5164 Frequency : occasional (liblustre only)
5166 Description: liblustre could not open files whose last component is a symlink
5167 Details : sysio_path_walk() would incorrectly pass the open intent to
5168 intermediate path components.
5171 Frequency : rare (liblustre only with non-standard tuning)
5172 Bugzilla : 7201 (7350)
5173 Description: Tuning the MDC DLM LRU size to zero triggers client LASSERT
5174 Details : llu_lookup_finish_locks() tries to set lock data on a lock
5175 after it has been released, only do this for referenced locks
5177 Severity : enhancement
5179 Description: specifying an (invalid) directory default stripe_size of -1
5180 would reset the directory default striping
5181 Details : stripe_size -1 was used internally to signal directory stripe
5182 removal, now use "all default" to signal dir stripe removal
5183 as a directory striping of "all default" is not useful
5186 Frequency : common for large clusters running liblustre clients
5188 Description: doing an ls when liblustre clients are running is slow
5189 Details : sending a glimpse AST to a liblustre client waits for every AST
5190 to time out, as liblustre clients will not respond. Since they
5191 cannot cache data we refresh the OST lock LVB from disk instead.
5193 Severity : enhancement
5195 Description: doing an ls at the same time as file IO can be slow
5196 Details : enqueue and other "small" requests can be blocked behind many
5197 large IO requests. Create a new OST IO portal for non-IO
5198 requests so they can be processed faster.
5201 Frequency : rare (only HPUX clients mounting unsupported re-exported NFS vol)
5203 Description: an HPUX NFS client would get -EACCESS when ftruncate()ing a newly
5204 created file with mode 000
5205 Details : the Linux NFS server relies on an MDS_OPEN_OWNEROVERRIDE hack to
5206 allow an ftruncate() as a non-root user to a file with mode 000.
5207 Lustre now respects this flag to disable mode checks when
5208 truncating a file owned by the user
5211 Frequency : liblustre-only, when liblustre client dies unexpectedly or becomes
5214 Description: Revoking locks from clients that went dead or catatonic might take
5216 Details : New lock flags FL_CANCEL_ON_BLOCK used by liblustre makes
5217 cancellation of such locks instant on servers without waiting for
5218 any reply from clients. Clients drops these locks when cancel
5219 notification from server is received without replying.
5222 Frequency : liblustre-only, when liblustre client dies or becomes busy
5224 Description: Doing ls on Linux clients can take a long time with active
5226 Details : Liblustre client cannot handle ASTs in timely manner, so avoid
5227 granting such locks to it in the first place if possible. Locks
5228 are taken by proxy on the OST during the read or write and
5229 dropped immediately afterward. Add connect flags handling, do
5230 not grant locks to liblustre clients for glimpse ASTs.
5232 Severity : enhancement
5234 Description: Improve read-ahead algorithm to avoid excessive IO for random reads
5235 Details : Existing read-ahead algorithm is tuned for the case of streamlined
5236 sequential reads and behaves badly with applications doing random
5237 reads. Improve it by reading ahead at least read region, and
5238 avoiding excessive large RPC for small reads.
5240 Severity : enhancement
5242 Description: Creating more than 1000 files for a single job may cause a load
5243 imbalance on the OSTs if there are also a large number of OSTs.
5244 Details : qos_prep_create() uses an OST index reseed value that is an
5245 even multiple of the number of available OSTs so that if the
5246 reseed happens in the middle of the object allocation it will
5247 still utilize the OSTs as uniformly as possible.
5252 Description: OST or MDS may oops in ping_evictor_main()
5253 Details : ping_evictor_main() drops obd_dev_lock if deleting a stale export
5254 but doesn't restart at beginning of obd_exports_timed list
5257 Severity : enhancement
5259 Description: improve by-nid export eviction on the MDS and OST
5260 Details : allow multiple exports with the same NID to be evicted at one
5261 time without re-searching the exports list.
5264 Frequency : rare, only with supplementary groups enabled on SMP 2.6 kernels
5266 Description: MDS may oops in groups_free()
5267 Details : in rare race conditions a newly allocated group_info struct is
5268 freed again, and this can be NULL. The 2.4 compatibility code
5269 for groups_free() checked for a NULL pointer, but 2.6 did not.
5272 Frequency : common for liblustre clients doing little filesystem IO
5273 Bugzilla : 9352, 7313
5274 Description: server may evict liblustre clients accessing contended locks
5275 Details : if a client is granted a lock or receives a completion AST
5276 with a blocking AST already set it would not reply to the AST
5277 for LDLM_FL_CANCEL_ON_BLOCK locks. It now replies to such ASTs.
5280 Frequency : lfs setstripe, only systems with more than 160 OSTs
5282 Description: unable to set striping with a starting offset beyond OST 160
5283 Details : llapi_create_file() incorrectly limited the starting stripe
5284 index to the maximum single-file stripe count.
5287 Frequency : LDAP users only
5289 Description: lconf did not handle in-kernel recovery with LDAP properly
5290 Details : lconf/LustreDB get_refs() is searching the wrong namespace
5292 Severity : enhancement
5294 Description: bind OST threads to NUMA nodes to improve performance
5295 Details : all OST threads are uniformly bound to CPUs on a single NUMA
5296 node and do their allocations there to localize memory access
5298 Severity : enhancement
5300 Description: llmount can determine client NID directly from Myrinet (GM)
5301 Details : the client NID code from gmnalnid was moved directly into
5302 llmount, removing the need to use this or specifying the
5303 client NID explicitly when mounting GM clients with zeroconf
5306 Frequency : if client is started with down MDS
5308 Description: if client is started with down MDS mount hangs in ptlrpc_queue_wait
5309 Details : Having an LWI_INTR() wait event (interruptible, but no timeout)
5310 will wait indefinitely in ptlrpc_queue_wait->l_wait_event() after
5311 ptlrpc_import_delayed_req() because we didn't check if the
5312 request was interrupted, and we also didn't break out of the
5313 event loop if there was no timeout
5318 Description: data loss during non-page-aligned writes to a single file from
5319 both multiple nodes and multiple threads on one node at same time
5320 Details : updates to KMS and lsm weren't protected by common lock. Resulting
5321 inconsistency led to false short-reads, that were cached and later
5322 used by ->prepare_write() to fill in partially written page,
5323 leading to data loss.
5326 Frequency : always, if lconf --abort_recovery used
5328 Description: lconf --abort_recovery fails with 'Operation not supported'
5329 Details : lconf was attempting to abort recovery on the MDT device and not
5332 Severity : enhancement
5334 Description: remove cleanup logs
5335 Details : replace lconf-generated cleanup logs with lustre internal
5336 cleanup routines. Eliminates the need for client-cleanup and
5339 Severity : enhancement
5341 Description: add support for EAs (user and system) on lustre filesystems
5342 Details : it is now possible to store extended attributes in the Lustre
5343 client filesystem, and with the user_xattr mount option it
5344 is possible to allow users to store EAs on their files also
5346 Severity : enhancement
5348 Description: Add possibility (config option) to show minimal available OST free
5350 Details : When compiled with --enable-mindf configure option, statfs(2)
5351 (and so, df) will return least minimal free space available from
5352 all OSTs as amount of free space on FS, instead of summary of
5353 free spaces of all OSTs.
5355 Severity : enhancement
5357 Description: do not expand extent locks acquired on OST-side
5358 Details : Modify ldlm_extent_policy() to not expand local locks, acquired
5359 by server: they are not cached anyway.
5362 Frequency : when mmap is used/binaries executed from Lustre
5364 Description: Unmmap pages before throwing them away from read cache.
5365 Details : llap_shrink cache now attempts to unmap pages before discarding
5366 them (if unmapping failed - do not discard). SLES9 kernel has
5367 extra checks that trigger if this unmapping is not done first.
5372 Description: lconf didn't resolve symlinks before checking to see whether a
5373 given mountpoint was already in use
5376 Frequency : when migrating failover services
5377 Bugzilla : 6395, 9514
5378 Description: When migrating a subset of services from a node (e.g. failback
5379 from a failover service node) the remaining services would
5380 time out and evict clients.
5381 Details : lconf --force (implied by --failover) sets the global obd_timeout
5382 to 5 seconds in order to quickly disconnect, but this caused
5383 other RPCs to time out too quickly. Do not change the global
5384 obd_timeout for force cleanup, only set it for DISCONNECT RPCs.
5386 Severity : enhancement
5387 Frequency : if MDS is started with down OST
5388 Bugzilla : 9439,5706
5389 Description: Allow startup/shutdown of an MDS without depending on the
5390 availability of the OSTs.
5391 Details : Asynchronously call mds_lov_synchronize during MDS startup.
5392 Add appropriate locking and lov-osc refcounts for safe
5393 cleaning. Add osc abort_inflight calls in case the
5394 synchronize never started.
5397 Frequency : occasional (Cray XT3 only)
5399 Description: root not authorized to access files in CRAY_PORTALS environment
5400 Details : The client process capabilities were not honoured on the MDS in
5401 a CRAY_PORTALS/CRAY_XT3 environment. If the file had previously
5402 been accessed by an authorized user then root was able to access
5403 the file on the local client also. The root user capabilities
5404 are now allowed on the MDS, as this environment has secure UID.
5407 Frequency : occasional
5409 Description: ldiskfs "too long searching" message happens too often
5410 Details : A debugging message (otherwise harmless) prints too often on
5411 the OST console. This has been reduced to only happen when
5412 there are fragmentation problems on the filesystem.
5417 Description: Division by zero in statfs when all OSCs are inactive
5418 Details : lov_get_stripecnt() returns zero due to incorrect order of checks,
5419 lov_statfs divides by value returned by lov_get_stripecnt().
5423 Bugzilla : 9489, 3273
5424 Description: First write from each client to each OST was only 4kB in size,
5425 to initialize client writeback cache, which caused sub-optimal
5426 RPCs and poor layout on disk for the first writen file.
5427 Details : Clients now request an initial cache grant at (re)connect time
5428 and so that they can start streaming writes to the cache right
5429 away and always do full-sized RPCs if there is enough data.
5430 If the OST is rebooted the client also re-establishes its grant
5431 so that client cached writes will be honoured under the grant.
5436 Description: Slow ls (and stat(2) syscall) on files residing on IO-loaded OSTs
5437 Details : Now I/O RPCs go to different portal number and (presumably) fast
5438 lock requests (and glimses) and other RPCs get their own service
5439 threads pool that should be able to service those RPCs
5442 Severity : enhancement
5444 Description: Ability to exchange lustre version between client and servers and
5445 issue warnings at client side if client is too old. Also for
5446 liblustre clients there is ability to refuse connection of too old
5448 Details : New 'version' field is added to connect data structure that is
5449 filled with version info. That info is later checked by server and
5453 Frequency : rare, liblustre only.
5454 Bugzilla : 9296, 9581
5455 Description: Two simultaneous writes from liblustre at offset within same page
5456 might proceed at the same time overwriting eachother with stale
5458 Details : I/O lock withing llu_file_prwv was released too early, before data
5459 actually was hitting the wire. Extended lock-holding time until
5460 server acknowledges receiving data.
5463 Frequency : extremely rare. Never observed in practice.
5465 Description: avoid generating lustre_handle cookie of 0.
5466 Details : class_handle_hash() generates handle cookies by incrementing
5467 global counter, and can hit 0 occasionaly (this is unlikely, but
5468 not impossible, because initial value of cookie counter is
5469 selected randonly). Value of 0 is used as a sentinel meaning
5470 "unassigned handle" --- avoid it. Also coalesce two critical
5471 sections in this function into one.
5473 Severity : enhancement
5475 Description: allow liblustre clients to delegate truncate locking to OST
5476 Details : To avoid overhead of locking, liblustre client instructs OST to
5477 take extent lock in ost_punch() on client's behalf. New connection
5478 flag is added to handle backward compatibility.
5480 Severity : enhancement
5481 Bugzilla : 4928, 7341, 9758
5482 Description: allow number of OST service threads to be specified
5483 Details : a module parameter allows the number of OST service threads
5484 to be specified via "options ost ost_num_threads={N}" in the
5485 OSS's /etc/modules.conf or /etc/modprobe.conf.
5489 Bugzilla : 6146, 9635, 9895
5490 Description: servers crash with bad pointer in target_handle_connect()
5491 Details : In rare cases when a client is reconnecting it was possible that
5492 the connection request was the last reference for that export.
5493 We would temporarily drop the export reference and get a new
5494 one, but this may have been the last reference and the export
5495 was just destroyed. Get new reference before dropping old one.
5497 Severity : enhancement
5498 Frequency : if client is started with failover MDS
5500 Description: Allow multiple MDS hostnames in the mount command
5501 Details : Try to read the configuration from all specified MDS
5502 hostnames during a client mount in case the "primary"
5505 Severity : enhancement
5507 Description: Stop sending data to evicted clients as soon as possible.
5508 Details : Check if the client we are about to send or are sending data to
5509 was evicted already. (Check is done every second of waiting,
5510 for which l_wait_event interface was extended to allow checking
5511 of exit condition at specified intervals).
5514 Frequency : rare, normally only when NFS exporting is done from client
5516 Description: 'bad disk LOV MAGIC: 0x00000000' error when chown'ing files
5518 Details : Make mds_get_md() recognise empty md case and set lmm size to 0.
5521 Frequency : always, if srand() is called before liblustre initialization
5523 Description: Liblustre uses system PRNG disturbing its usage by user application
5524 Details : Introduce internal to lustre fast and high-quality PRNG for
5525 lustre usage and make liblustre and some other places in generic
5526 lustre code to use it.
5528 Severity : enhancement
5529 Bugzilla : 9477, 9557, 9870
5530 Description: Verify that the MDS configuration logs are updated when xml is
5531 Details : Check if the .xml configuration logs are newer than the config
5532 logs stored on the MDS and report an error if this is the case.
5533 Request --write-conf, or allow starting with --old_conf.
5535 Severity : enhancement
5537 Description: Handle symlinks in the path when checking if Lustre is mounted.
5538 Details : Resolve intermediate symlinks when checking if a client has
5539 mounted a filesystem to avoid duplicate client mounts.
5544 Description: lconf can hit an error exception but still return success.
5545 Details : The lconf command catches the Command error exception at the top
5546 level script context and will exit with the associated exit
5547 status, but doesn't ensure that this exit status is non-zero.
5552 Description: failure of ptlrpc thread startup can cause oops
5553 Details : Starting a ptlrpc service thread can fail if there are a large
5554 number of threads or the server memory is very fragmented.
5555 Handle this without oopsing.
5558 Frequency : always, only if liblustre and non-default acceptor port was used
5560 Description: liblustre cannot connect to servers with non-default acceptor port
5561 Details : tcpnal_set_default_params() was not called and was therefore
5562 ignoring the environment varaible TCPNAL_PORT, as well as other
5563 TCPNAL_ environment variables
5568 Description: two objects could be created on the same OST for a single file
5569 Details : If an OST is down, in some cases it was possible to create two
5570 objects on a single OST for a single file. No problems other
5571 than potential performance impact and spurious error messages.
5575 Bugzilla : 5681, 9562
5576 Description: Client may oops in ll_unhash_aliases
5577 Details : Client dcache may become inconsistent in race condition.
5578 In some cases "getcwd" can fail if the current directory is
5584 Description: Inode refcounting problems in NFS export code
5585 Details : link_raw functions used to call d_instantiate without obtaining
5586 extra inode reference first.
5590 Bugzilla : 9942, 9903
5591 Description: Referencing freed requests leading to crash, memleaks with NFS.
5592 Details : We used to require that call to ll_revalidate_it was always
5593 followed by ll_lookup_it. Also with revalidate_special() it is
5594 possible to call ll_revalidate_it() twice for the same dentry
5595 even if first occurence returned success. This fix changes semantic
5596 between DISP_ENQ_COMPLETE disposition flag to mean there is extra
5597 reference on a request referred from the intent.
5598 ll_intent_release() then releases such a request.
5601 Frequency : rare, normally benchmark loads only
5603 Description: unlinked inodes were kept in memory on the client
5604 Details : If a client is repeatedly creating and unlinking files it
5605 can accumulate a lot of stale inodes in the inode slab cache.
5606 If there is no other client load running this can cause the
5607 client node to run out of memory. Instead flush old inodes
5608 from client cache that have the same inode number as a new inode.
5611 Frequency : SLES9 2.6.5 kernel and long filenames only
5612 Bugzilla : 9969, 10379
5613 Description: utime reports stale NFS file handle
5614 Details : SLES9 uses out-of-dentry names in some cases, which confused
5615 the lustre dentry revalidation. Change it to always use the
5619 Frequency : rare, unless heavy write-truncate concurrency is continuous
5620 Bugzilla : 4180, 6984, 7171, 9963, 9331
5621 Description: OST becomes very slow and/or deadlocked during object unlink
5622 Details : filter_destroy() was holding onto the parent directory lock
5623 while truncating+unlinking objects. For very large objects this
5624 may block other threads for a long time and slow overall OST
5625 responsiveness. It may also be possible to get a lock ordering
5626 deadlock in this case, or run out of journal credits because of
5627 the combined truncate+unlink. Solution is to do object truncate
5628 first in one transaction without parent lock, and then do the
5629 final unlink in a new transaction with the parent lock. This
5630 reduces the lock hold time dramatically.
5633 Frequency : rare, 2.4 kernels only
5635 Description: MDS or OST cleanup may trip kernel BUG when dropping kernel lock
5636 Details : mds_cleanup() and filter_cleanup() need to drop the kernel lock
5637 before unmounting their filesystem in order to avoid deadlock.
5638 The kernel_locked() function in 2.4 kernels only checks whether
5639 the kernel lock is held, not whether it is this process that is
5640 holding it as 2.6 kernels do.
5645 Description: MDS or OST may oops/LBUG if a client is connecting multiple times
5646 Details : The client ptlrpc code may be trying to reconnect to a down
5647 server before a previous connection attempt has timed out.
5648 Increase the reconnect interval to be longer than the connection
5649 timeout interval to avoid sending duplicate connections to
5653 Frequency : echo_client brw_test command
5655 Description: fix echo_client to work with OST preallocated code
5656 Details : OST preallocation code (5137) didn't take echo_client IO path
5657 into account: echo_client calls filter methods outside of any
5658 OST thread and, hence, there is no per-thread preallocated
5659 pages and buffers to use. Solution: hijack pga pages for IO. As
5660 a byproduct, this avoids unnecessary data copying.
5664 Bugzilla : 3555, 5962, 6025, 6155, 6296, 9574
5665 Description: Client can oops in mdc_commit_close() after open replay
5666 Details : It was possible for the MDS to return an open request with no
5667 transaction number in mds_finish_transno() if the client was
5668 evicted, but without actually returning an error. Clients
5669 would later try to replay that open and may trip an assertion
5670 Simplify the client close codepath, and always return an error
5671 from the MDS in case the open is not successful.
5674 Frequency : rare, 2.6 OSTs only
5676 Description: OST may deadlock under high load on fragmented files
5677 Details : If there was a heavy load and highly-fragmented OST filesystems
5678 it was possible to have all the OST threads deadlock waiting on
5679 allocation of biovecs, because the biovecs were not released
5680 until the entire RPC IO was completed. Instead, release biovecs
5681 as soon as they are complete to ensure forward IO progress.
5683 Severity : enhancement
5685 Description: Support for specifying external journal device at mount
5686 Details : If an OST or MDS device is formatted with an external journal
5687 device, this device major/minor is stored in the ext3 superblock
5688 and may not be valid for failover. Allow detecting and
5689 specifying the external journal at mount time.
5694 Description: Mounting an MDS with pending unlinked files may cause oops
5695 Details : target_finish_recovery() calls mds_postrecov() which returned
5696 the number of orphans unlinked. mds_lov_connect->mds_postsetup()
5697 considers this an error and immediately begins cleaning up the
5698 lov, just after starting the mds_lov process
5700 Severity : enhancement
5702 Description: Implement 'lfs df' to report actual free space on per-OST basis
5703 Details : Add sub-command 'df' on 'lfs' to report the disk space usage of
5704 MDS/OSDs. Usage: lfs df [-i][-h]. Command Options: '-i' to report
5705 usage of objects; '-h' to report in human readable format.
5707 ------------------------------------------------------------------------------
5709 2005-08-26 Cluster File Systems, Inc. <info@clusterfs.com>
5716 Description: Mounting an ldiskfs file system with mballoc may crash OST node.
5717 Details : ldiskfs mballoc code may reference an uninitialized buddy struct
5718 at startup during orphan unlinking. Instead, skip buddy update
5719 before setup, as it will be regenerated after recovery is complete.
5724 Description: If an OST is inactive, its locks might reference stale inodes.
5725 Details : lov_change_cbdata() must iterate over all namespaces, even if
5726 they are inactive to clear inode references from the lock.
5728 Severity : enhancement
5729 Frequency : occasional, if non-standard max_dirty_mb used
5731 Description: Client will block write RPCs if not enough grant
5732 Details : If a client has max_dirty_mb smaller than max_rpcs_in_flight,
5733 then the client will block writes while waiting for another RPC
5734 to complete instead of consuming its dirty limit. With change
5735 we get improved performance when max_dirty_mb is small.
5737 Severity : enhancement
5738 Bugzilla : 3389, 6253
5739 Description: Add support for supplementary groups on the MDS.
5740 Details : The MDS has an upcall /proc/fs/lustre/mds/{mds}/group_upcall
5741 (set to /usr/sbin/l_getgroups if enabled) which will do MDS-side
5742 lookups for user supplementary groups into a cache.
5746 Description: O_CREAT|O_EXCL open flags in liblustre always return -EEXIST
5747 Details : Make libsysio to not enforce O_EXCL by clearing the flag,
5748 for liblustre O_EXCL is enforced by MDS.
5752 Description: readdir never returns NULL in liblustre.
5753 Details : Corrected llu_iop_getdirentries logic, to return offset of next
5754 dentry in struct dirent.
5758 Frequency : liblustre only, depends on application IO pattern
5759 Description: liblustre clients evicted if not contacting servers
5760 Details : Don't put liblustre clients into the ping_evictor list, so
5761 they will not be evicted by the pinger ever.
5763 Severity : enhancement
5765 Description: Add ability to evict clients by NID from MDS.
5766 Details : By echoing "nid:$NID" string into
5767 /proc/fs/lustre/mds/.../evict_client client with nid that equals to
5768 $NID would be instantly evicted from this MDS and from all active
5769 OSTs connected to it.
5773 Description: Do not query file size twice, somewhat slowing stat(2) calls.
5774 Details : lookup_it_finish() used to query file size from OSTs that was not
5779 Description: service threads change working directory to that of init
5780 Details : Starting lustre service threads may pin the working directory
5781 of the parent thread, making that filesystem busy. Threads
5782 now change to the working directory of init to avoid this.
5786 Frequency : during shutdown only
5787 Description: shutdown with a failed MDS or OST can cause unmount to hang
5788 Details : Don't resend DISCONNECT messages in ptlrpc_disconnect_import()
5793 Frequency : 2.6 only
5794 Description: chmod/chown may include an extra supplementary group
5795 Details : ll{,u}_mdc_pack_op_data() does not properly initialize the
5796 supplementary group and if none is specified this is used.
5799 Bugzilla : 5479 (6816)
5801 Description: Racing open + rm can assert client in mdc_set_open_replay_data()
5802 Details : If lookup is in progress on a file that is unlinked we might try
5803 to revalidate the inode and fail in revalidate after lookup is
5804 complete and ll_file_open() enqueues the open again but
5805 it_open_error() was not checking DISP_OPEN_OPEN errors correctly.
5808 Frequency : always, if lconf --abort_recovery used
5810 Description: lconf --abort_recovery fails with 'Operation not supported'
5811 Details : lconf was attempting to abort recovery on the MDT device and not
5814 ------------------------------------------------------------------------------
5816 2005-08-08 Cluster File Systems, Inc. <info@clusterfs.com>
5821 Frequency : rare (only unsupported configurations with a node running as an
5823 Bugzilla : 6514, 5137
5824 Description: Mounting a Lustre file system on a node running as an OST could
5826 Details : OSTs now preallocates memory needed to write out data at
5827 startup, instead of when needed, to avoid having to
5828 allocate memory in possibly low memory situations.
5829 Specifically, if the file system is mounted on on OST,
5830 memory pressure could force it to try to write out data,
5831 which it needed to allocate memory to do. Due to the low
5832 memory, it would be unable to do so and the node would
5833 become unresponsive.
5835 Severity : enhancement
5837 Description: Addition of lconf --service command line option
5838 Details : lconf now accepts a '--service <arg>' option, which is
5839 shorthand for 'lconf --group <arg> --select <arg>=<hostname>'
5841 Severity : enhancement
5843 Description: Failover mode is now the default for OSTs.
5844 Details : By default, OSTs will now run in failover mode. To return to
5845 the old behaviour, add '--failout' to the lmc line for OSTs.
5847 Severity : enhancement
5849 Description: Health checks are now provided for MDS and OSTs
5850 Details : Additional detailed health check information on MSD and OSTs
5851 is now provided through the procfs health_check value.
5854 Frequency : occasional, depends on IO load
5856 Description: Disk fragmentation on the OSTs could eventually cause slowdowns
5857 after numerous create/delete cycles
5858 Details : The ext3 inode allocation policy would not allocate new inodes
5859 very well on the OSTs because there are no new directories
5860 being created. Instead we look for groups with free space if
5861 the parent directories are nearly full.
5866 Description: Network or server problems during mount may cause partially
5867 mounted clients instead of returning an error.
5868 Details : The config llog parsing code may overwrite the error return
5869 code during mount error handling, returning success instead
5875 Description: MDS can fail to allocate large reply buffers
5876 Details : After long uptimes the MDS can fail to allocate large reply
5877 buffers (e.g. zconf client mount config records) due to memory
5878 fragmentation or consumption by the buffer cache. Preallocate
5879 some large reply buffers so that these replies can be sent even
5880 under memory pressure.
5884 Frequency : rare (liblustre)
5885 Description: fsx running with liblustre complained that using truncate() to
5886 extend the file doesn't work. This patch corrects that issue.
5887 Details : This is the liblustre equivalent of the fix for bug 6196. Fixes
5888 ATTR_SIZE and lsm use in llu_setattr_raw.
5892 Frequency : rare, only 2.6 kernels
5893 Description: Unusual file access patterns on the MDS may result in inode
5894 data being lost in very rare circumstances.
5895 Details : Bad interaction between the ea-in-inode patch and the "no-read"
5896 code in the 2.6 kernel caused the inode and/or EA data not to
5897 be read from disk, causing single-file corruption.
5901 Frequency : rare, only 2.6 filesystems using extents
5902 Description: Heavy concurrent write and delete load may cause data corruption.
5903 Details : It was possible under high-load situations to have an extent
5904 metadata block in the block device cache from a just-unlinked
5905 file overwrite a newly-allocated data block. We now unmap any
5906 metadata buffers that alias just-allocated data blocks.
5910 Frequency : filesystems with default stripe_count larger than 77
5911 Description: lconf+mke2fs fail when formatting filesystem with > 77 stripes
5912 Details : lconf specifies an inode size of 4096 bytes when the default
5913 stripe_count is larger than 77. This conflicts with the default
5914 inode density of 1 per 4096 bytes. Allocate smaller inodes in
5915 this case to avoid pinning too much memory for large EAs.
5917 ------------------------------------------------------------------------------
5919 2005-07-07 Cluster File Systems, Inc. <info@clusterfs.com>
5924 Frequency : rare (extremely heavy IO load with hundreds of clients)
5926 Description: Client is evicted, gets IO error writing to file
5927 Details : lock ordering changes for bug 5492 reintroduced bug 3267 and
5928 caused clients to be evicted for AST timeouts. The fixes in
5929 bug 5192 mean we no longer need to have such short AST timeouts
5930 so ldlm_timeout has been increased.
5933 Frequency : occasional during --force or --failover shutdown under load
5934 Bugzilla : 5949, 4834
5935 Description: Server oops/LBUG if stopped with --force or --failover under load
5936 Details : a collection of import/export refcount and cleanup ordering
5937 issues fixed for safer force cleanup
5940 Frequency : only filesystems larger than 120 OSTs
5941 Bugzilla : 5990, 6223
5942 Description: lfs getstripe would oops on a very large filesystem
5943 Details : lov_getconfig used kfree on vmalloc'd memory
5946 Frequency : only filesystems exporting via NFS to Solaris 10 clients
5947 Bugzilla : 6242, 6243
5948 Description: reading from files that had been truncated to a non-zero size
5949 but never opened returned no data
5950 Details : ll_file_read() reads zeros from no-object files to EOF
5955 Description: A bug in MDS/OSS recovery could cause the OSS to fail an assertion
5956 Details : There's little harm in aborting MDS/OSS recovery and letting it
5957 try again, so I removed the LASSERT and return an error instead.
5959 Severity : enhancement
5961 Description: New debugging infrastructure for tracking down data corruption
5962 Details : The I/O checksum code was replaced to: (a) control it at runtime,
5963 (b) cover more of the client-side code path, and (c) try to narrow
5964 down where problems occurred
5968 Bugzilla : 3819, 4364, 4397, 6313
5969 Description: Racing close and eviction MDS could cause assertion in mds_close
5970 Details : It was possible to get multiple mfd references during close and
5971 client eviction, leading to one thread referencing a freed mfd.
5973 Severity: : enhancement
5974 Bugzilla : 3262, 6359
5975 Description: Attempts to reconnect to servers are now more aggressive.
5976 Details : This builds on the enhanced upcall-less recovery that was added
5977 in 1.4.2. When trying to reconnect to servers, clients will
5978 now try each server in the failover group every 10 seconds. By
5979 default, clients would previously try one server every 25 seconds.
5984 Description: After recovery, certain operations trigger a failed
5985 assertion on a client.
5986 Details : Failing over an mds, using lconf -d --failover, while a
5987 client was doing a readdir() call would cause the client to
5988 LBUG after recovery completed and the readdir() was resent.
5990 Severity : enhancement
5992 Description: Default groups are now added by lconf
5993 Details : You can now run lconf --group <servicename> without having to
5994 manually add groups with lmc.
5997 Frequency : occasional
5999 Description: Nodes with an elan id of 0 trigger a failed assertion
6002 Frequency : always when accessing e.g. tty/console device nodes
6004 Description: tty and some other devices nodes cannot be used on lustre
6005 Details : file's private_data field is used by device data and lustre
6006 values in there got lost. New field was added to struct file to
6007 store fs-specific private data.
6010 Frequency : when exporting Lustre via NFS
6012 Description: NFSD failed occasionally when looking up a path component
6013 Details : NFSD is looking up ".." which was broken in ext3 directories
6014 that had grown large enough to become hashed.
6017 Frequency : Clusters with multiple interfaces not on the same subnet
6019 Description: Nodes will repeatedly try to reconnect to an interface which it
6020 cannot reach and report an error to the log.
6021 Details : Extra peer list entries will be created by lconf with some peers
6022 unreachable. lconf now validates the peer before adding it.
6025 Frequency : Only if a default stripe is set on the filesystem root.
6027 Description: Setting a default stripe on the filesystem root prevented the
6028 filesystem from being remounted.
6029 Details : The client was sending extra request flags in the root getattr
6030 request and did not allocate a reply buffer for the dir EA.
6033 Frequency : occasional, higher if lots of files are accessed by one client
6034 Bugzilla : 6159, 6097
6035 Description: Client trips assertion regarding lsm mismatch/magic
6036 Details : While revalidating inodes the VFS looks up inodes with ifind()
6037 and in rare cases can find an inode that is being freed.
6038 The ll_test_inode() code will free the lsm during ifind()
6039 when it finds an existing inode and then the VFS later attaches
6040 this free lsm to a new inode.
6044 Bugzilla : 6422, 7030
6045 Description: MDS deadlock between mkdir and client eviction
6046 Details : Creating a new file via mkdir or mknod (starting a transaction
6047 and getting the ns lock) can deadlock with client eviction
6048 (gets ns lock and trying to finish a synchronous transaction).
6051 Frequency : occasional
6052 Description: While starting a server, the fsfilt_ext3 module could not be
6054 Details : CFS's improved ext3 filesystem is named ldiskfs for 2.6
6055 kernels. Previously, lconf would still use the ext3 name
6056 when trying to load modules. Now, it will correctly use
6057 ext3 on 2.4 and ldiskfs on 2.6.
6059 Severity : enhancement
6060 Description: The default stripe count has been changed to 1
6061 Details : The interpretation of the default stripe count (0, to lfs
6062 or lmc) has been changed to mean striping across a single
6063 OST, rather than all available. For general usage we have
6064 found a stripe count of 1 or 2 works best.
6066 Severity : enhancement
6067 Description: Add support for compiling against Cray portals.
6068 Details : Conditional compiling for some areas that are different
6072 Frequency : occasional
6073 Bugzilla : 6409, 6834
6074 Description: Creating files with an explicit stripe count may lead to
6075 a failed assertion on the MDS
6076 Details : If some OSTs are full or unavailable, creating files may
6077 trigger a failed assertion on the MDS. Now, Lustre will
6078 try to use other servers or return an error to the
6082 Frequency : occasional
6084 Description: Multiple concurrent overlapping read+write on multiple SMP nodes
6085 caused lock timeout during readahead (since 1.4.2).
6086 Details : Processes doing readahead might match a lock that hasn't been
6087 granted yet if there are overlapping and conflicting lock
6088 requests. The readahead process waits on ungranted lock
6089 (original lock is CBPENDING), while OST waits for that process
6090 to cancel CBPENDING read lock and eventually evicts client.
6092 Severity : enhancement
6094 Description: Initial enabling of flock support for clients
6095 Details : Implements fcntl advisory locking and file status functions.
6096 This feature is provided as an optional mount flag (default
6097 off), and is NOT CURRENTLY SUPPORTED. Not all types of record
6098 locking are implemented yet, and those that are are not guaranteed
6099 to be completely correct in production environments.
6100 mount -t lustre -o [flock|noflock] ...
6103 Frequency : occasional
6105 Description: OSTs running 2.4 kernels but with extents enabled might trip an
6106 assertion in the ext3 JBD (journaling) layer.
6107 Details : The b_committed_data struct is protected by the big kernel lock
6108 in 2.4 kernels, serializing journal_commit_transaction() and
6109 ext3_get_block_handle->ext3_new_block->find_next_usable_block()
6110 access to this struct. In 2.6 kernels there is finer grained
6111 locking to improve SMP performance of the JBD layer.
6115 Description: Changes the "SCSI I/O Stats" kernel patch to default to "enabled"
6117 -----------------------------------------------------------------------------
6119 2005-05-05 Cluster File Systems, Inc. <info@clusterfs.com>
6121 NOTE: Lustre 1.4.2 uses an incompatible network protocol than previous
6122 versions of Lustre. Please update all servers and clients to
6123 version 1.4.2 or later at the same time. You must also run
6124 "lconf --write-conf {config}.xml" on the MDS while it is stopped
6125 to update the configuration logs.
6127 - fix for HPUX NFS client breakage when NFS exporting Lustre (5781)
6128 - mdc_enqueue does not need max_mds_easize request buffer on send (5707)
6129 - swab llog records of type '0' so we get proper header size/idx (5861)
6130 - send llog cancel req to DLM cancel portal instead of cb portal (5515)
6131 - fix rename of one directory over another leaking an inode (5953)
6132 - avoid SetPageDirty on 2.6 (5981)
6133 - don't re-add just-being-destroyed locks to the waiting list (5653)
6134 - when creating new directories, inherit the parent's custom
6135 striping settings if present parent (3048)
6136 - flush buffers from cache before direct IO in 2.6 obdfilter (4982)
6137 - don't hold i_size_sem in ll_nopage() and ll_ap_refresh_count (6077)
6138 - don't hold client locks on temporary worklist from l_lru (5666)
6139 - handle IO errors in 2.6 obdfilter bio completion routine (6046)
6140 - automatically evict dead clients (5921)
6141 - Update file size properly in create+truncate+fstat case (6196)
6142 - Do not unhash mountpoint dentries, do not allow removal of
6144 - Avoid lock ordering deadlock issue with write/truncate (6203,5654)
6145 - reserve enough journal credits in fsfilt_start_log for setattr (4554)
6146 - ldlm_enqueue freed-export error path would always LBUG (6149,6184)
6147 - don't reference lr_lvb_data until after we hold lr_lvb_sem (6170)
6148 - don't overwrite last_rcvd if there is a *_client_add() error (6086)
6149 - Correctly handle reads of files with no objects (6243)
6150 - lctl recover will also mark a device active if deactivate used (5933)
6152 - by default create 1 inode per 4kB space on MDS, per 16kB on OSTs
6153 - allow --write-conf on an MDS with different nettype than client (5619)
6154 - don't write config llogs to MDS for mounts not from that MDS (5617)
6155 - lconf should create multiple TCP connections from a client (5201)
6156 - init scripts are now turned off by default; run chkconfig --on
6157 lustre and chkconfig --on lustrefs to use them
6158 - upcalls are no longer needed for clients to recover to failover
6160 - add --abort-recovery option to lconf to abort recovery on device
6162 - add support for an arbitrary number of OSTs (3026)
6163 - Quota support protocol changes.
6164 - forward compatibility changes to wire structs (6007)
6165 - rmmod NALs that might be loaded because of /etc/modules.conf (6133)
6166 - support for mountfsoptions and clientoptions to the Lustre LDAP (5873)
6167 - improved "lustre status" script
6168 - initialize blocksize for non-regular files (6062)
6169 - added --disable-server and --disable-client configure options (5782)
6170 - introduce a lookup cache for lconf to avoid repeated DB scans (6204)
6171 - Vanilla 2.4.29 support
6172 - increase maximum number of obd devices to 520 (6242)
6173 - remove the tcp-zero-copy patch from the suse-2.4 series (5902)
6174 - Quadrics Elan drivers are now included for the RHEL 3 2.4.21 and
6175 SLES 9 2.6.5 kernels
6176 - limit stripes per file to 160 (the maximum EA size) (6093)
6178 2005-03-22 Cluster File Systems, Inc. <info@clusterfs.com>
6181 - don't LASSERT in ll_release on NULL lld with NFS export (4655, 5760)
6182 - hold NS lock when calling handle_ast_error->del_waiting_lock (5746)
6183 - fix setattr mtime regression from lovcleanup merge (4829, 5669)
6184 - workaround for 2.6 crash in ll_unhash_aliases (5687, 5210)
6185 - small ext3 extents cleanups and fixes (5733)
6186 - improved mballoc code, several small races and bugs fixed (5733, 5638)
6187 - kernel version 43 - fix remove_suid bugs in both 2.4 and 2.6 (5695)
6188 - avoid needless client->OST connect, fix handle mismatch (5317)
6189 - fix DLM error path that led to out-of-sync client, long delays (5779)
6190 - support common vfs-enforced mount options (nodev,nosuid,noexec) (5637)
6191 - fix several locking issues related to i_size (5492,5624,5654,5672)
6192 - don't move pending lock onto export if it is already evicted (5683)
6193 - fix kernel oops when creating .foo in unlinked directory (5548)
6194 - fix deadlock in obdfilter statistics vs. object create (5811)
6195 - use time_{before,after} to avoid timer jiffies wrap (5882)
6196 - shutdown --force/--failover stability (3607,3651,4797,5203,4834)
6197 - Do not leak request if server was not able to process it (5154)
6198 - If mds_open unable to find parent dir, make that negative lookup(5154)
6199 - don't create new directories with extent-mapping (5909, 5936)
6201 - fix lustre/lustrefs init scripts for SuSE (patch from Scali, 5702)
6202 - don't hold the pinger_sem in ptlrpc_pinger_sending_on_import
6203 - change obd_increase_kms to obd_adjust_kms (up or down) (5654)
6204 - lconf, lmc search both /usr/lib and /usr/lib64 for Python libs (5800)
6205 - support for RHEL4 kernel on i686 (5773)
6206 - provide error messages when incompatible logs are encountered (5898)
6208 2005-02-18 Cluster File Systems, Inc. <info@clusterfs.com>
6209 * version 1.4.0.10 (1.4.1 release candidate 1)
6211 - don't keep a lock reference when lock is not granted (4238)
6212 - unsafe list practices (rarely) led to infinite eviction loop (4908)
6213 - add per-fs limit of Lustre pages in page cache, avoid OOM (4699)
6214 - drop import inflight refcount on signal_completed_replay error (5255)
6215 - unlock page after async write error during send (3677)
6216 - handle missing objects in filter_preprw_read properly (5265)
6217 - no transno return for symlink open, don't save no-trasno open (3440)
6218 - don't try to complete elan receive that already failed (4012)
6219 - free RPC server reply state on error (5406)
6220 - clean up thread from ptlrpc_start_thread() on error (5160)
6221 - readahead could read extra page into cache that wasn't ejected (5388)
6222 - prevent races in class_attach/setup/cleanup/detach (5260)
6223 - don't dereference de->d_inode after l_dput of de (5458)
6224 - use "int" for stripe value returned from lock_to_stripe (5544)
6225 - mballoc allocation and error-checking fixes in 2.6 (5504)
6226 - block device patches to fix I/O request sizes in 2.6 (5482)
6227 - look up hostnames for IB nals (5602)
6228 - 2.6 changed lock ordering of 2 semaphores, caused deadlock (5654)
6229 - don't start multiple acceptors for the same port (5277)
6230 - fix incorrect LASSERT in mds_getattr_name (5635)
6231 - export a proc file for general "ping" checking (5628)
6232 - fix "lfs check" to not block when the MDS is down (5628)
6234 - service request history (4965)
6235 - put {ll,lov,osc}_async_page structs in a single slab (4699)
6236 - create an "evict_client" /proc entry on OSTs, like the MDS has
6237 - fix mount usage message, return errors per mount(8) (5168)
6238 - change grep [] to grep "[]" in tests so they work in more UMLs
6239 - fix ppc64/x86_64 spec to use %{_libdir} instead of /usr/lib (5389)
6240 - remove ancient LOV_MAGIC_V0 EA support (5047)
6241 - add "disk I/Os in flight" and "I/O req time" stats in obdfilter
6242 - align r/w RPCs to PTLRPC_MAX_BRW_SIZE boundary for performance (3451)
6243 - allow readahead allocations to fail when low on memory (5383)
6244 - mmap locking landed again, after considerable improvement (2828)
6245 - add get_hostaddr() to lustreDB.py for LDAP support (5459)
6247 2004-11-23 Cluster File Systems, Inc. <info@clusterfs.com>
6250 - send OST transaction number in read/write reply to free req (4966)
6251 - don't ASSERT in ptl_send_rpc() if we run out of memory (5119)
6252 - lock /proc/sys/portals/routes internal state, avoiding oops (4827)
6253 - the watchdog thread now runs as interruptible (5246)
6254 - flock/lockf fixes (but it's still disabled, pending 5135)
6255 - don't use EXT3 constants in llite code (5094)
6256 - memory shortage at startup could cause assertion (5176)
6258 - reorganization of lov code
6259 - single portals codebase
6261 - add extents/mballoc support (5025)
6262 - direct I/O reads in the obdfilter (4048)
6263 - kernel patches from LNXI for 2.6 (bluesmoke, perfctr, mtd, kexec)
6265 tbd Cluster File Systems, Inc. <info@clusterfs.com>
6268 - send OST transaction number in read/write reply to free req (4966)
6269 - don't ASSERT in ptl_send_rpc() if we run out of memory (5119)
6270 - lock /proc/sys/portals/routes internal state, avoiding oops (4827)
6271 - the watchdog thread now runs as interruptible (5246)
6272 - handle missing objects in filter_preprw_read properly (5265)
6273 - unsafe list practices (rarely) led to infinite eviction loop (4908)
6274 - drop import inflight refcount on signal_completed_replay error (5255)
6275 - unlock page after async write error during send (3677)
6276 - return original error code on reconstructed replies (3761)
6277 - no transno return for symlink open, don't save no-trasno open (3440)
6279 - add pid to ldlm debugging output (4922)
6280 - bump the watchdog timeouts -- we can't handle 30sec yet
6281 - extra debugging for orphan dentry/inode bug (5259)
6283 2004-11-16 Cluster File Systems, Inc. <info@clusterfs.com>
6286 - fix TCP_NODELAY bug, which caused extreme perf regression (5134)
6287 - allocate qswnal tx descriptors singly to avoid fragmentation (4504)
6288 - don't LBUG on obdo_alloc() failure, use OBD_SLAB_ALLOC() (4800)
6289 - fix NULL dereference in /proc/sys/portals/routes (4827)
6290 - allow failed mdc_close() operations to be interrupted (4561)
6291 - stop precreate on OST before MDS would time out on it (4778)
6292 - don't send partial-page writes before EOF from client (4410)
6293 - discard client grant for sub-page writes on large-page clients (4520)
6294 - don't free dentries not owned by NFS code, check generation (4806)
6295 - fix lsm leak if mds_create_objects() fails (4801)
6296 - limit debug_daemon file size, always print CERROR messages (4789)
6297 - use transno after validating reply (3892)
6298 - process timed out requests if import state changes (3754)
6299 - update mtime on OST during writes, return in glimpse (4829)
6300 - add mkfsoptions to LDAP (4679)
6301 - use ->max_readahead method instead of zapping global ra (5039)
6302 - don't interrupt __l_wait_event() during strace
6304 - add software watchdogs to catch hung threads quickly (4941)
6305 - make lustrefs init script start after nfs is mounted
6306 - fix CWARN/ERROR duplication (4930)
6307 - return async write errors to application if possible (2248)
6308 - add /proc/sys/portal/memused (bytes allocated by PORTALS_ALLOC)
6309 - print NAL number in %x format (4645)
6310 - update barely-supported suse-2.4.21-171 series (4842)
6311 - support for sles 9 %post scripts
6312 - support for building 2.6 kernel-source packages
6313 - support for sles km_* packages
6315 2004-10-07 Cluster File Systems, Inc. <info@clusterfs.com>
6318 - ignore -ENOENT errors in osc_destroy (3639)
6319 - notify osc create thread that OSC is being cleaned up (4600)
6320 - add nettype argument for llmount in #5d in conf-sanity.sh (3936)
6321 - reconstruct ost_handle() like mds_handle() (4657)
6322 - create a new thread to do import eviction to avoid deadlock (3969)
6323 - let lconf resolve symlinked-to devices (4629)
6324 - don't unlink "objects" from directory with default EA (4554)
6325 - hold socknal file ref over connect in case target is down (4394)
6326 - allow more than 32000 subdirectories in a single directory (3244)
6327 - fix blocks count for O_DIRECT writes (3751)
6328 - OST returns ENOSPC from object create when no space left (4539)
6329 - don't send truncate RPC if file size isn't changing (4410)
6330 - limit OSC precreate to 1/2 of value OST considers bogus (4778)
6331 - bind to privileged port in socknal and tcpnal (3689)
6333 - rate limit CERROR/CWARN console message to avoid overload (4519)
6334 - GETFILEINFO dir ioctl returns LOV EA + MDS stat in 1 call (3327)
6335 - basic mmap support (3918)
6336 - kernel patch series update from b1_4 (4711)
6338 2004-09-16 Cluster File Systems, Inc. <info@clusterfs.com>
6341 - avoid crash during MDS cleanup with OST shut down (2775)
6342 - fix loi_list_lock/oig_lock inversion on interrupted IO (4136)
6343 - don't use bad inodes on the MDS (3744)
6344 - dynamic object preallocation to improve recovery speed (4236)
6345 - don't hold spinlock over lock dumping or change debug flags (4401)
6346 - don't zero obd_dev when it is force cleaned (3651)
6347 - print grants to console if they go negative (4431)
6348 - "lctl deactivate" will stop automatic recovery attempts (3406)
6349 - look for existing locks in ldlm_handle_enqueue() (3764)
6350 - don't resolve lock handle twice in recovery avoiding race (4401)
6351 - revalidate should check working dir is a directory (4134)
6353 - don't always mark "slow" obdfilter messages as errors (4418)
6355 2004-08-24 Cluster File Systems, Inc. <info@clusterfs.com>
6358 - don't close LustreDB during write_conf until it is done (3860)
6359 - fix typo in lconf for_each_profile (3821)
6360 - allow dumping logs from multiple threads at one time (3820)
6361 - don't allow multiple threads in OSC recovery (3812)
6362 - fix debug_size parameters (3864)
6363 - fix mds_postrecov to initialize import for llog ctxt (3121)
6364 - replace config semaphore with spinlock (3306)
6365 - be sure to send a reply for a CANCEL rpc with bad export (3863)
6366 - don't allow enqueue to complete on a destroyed export (3822)
6367 - down write_lock before checking llog header bitmap (3825)
6368 - recover from lock replay timeout (3764)
6369 - up llog sem before sending rpc (3652)
6370 - reduce ns lock hold times when setting kms (3267)
6371 - change a dlm LBUG to LASSERTF, to maybe learn something (4228)
6372 - fix NULL deref and obd_dev leak on setup error (3312)
6373 - replace some LBUG about llog ops with error handling (3841)
6374 - don't match INVALID dentries from d_lookup and spin (3784)
6375 - hold dcache_lock while marking dentries INVALID and hashing (4255)
6376 - fix invalid assertion in ptlrpc_set_wait (3880)
6378 - add libwrap support for the TCP acceptor (3996)
6379 - add /proc/sys/portals/routes for non-root route listing (3994)
6380 - allow setting MDS UUID in .xml (2580)
6381 - print the stack of a process that LBUGs (4228)
6383 2004-07-14 Cluster File Systems, Inc. <info@clusterfs.com>
6386 - don't cleanup request in ll_file_open() on failed MDS open (3430)
6387 - make sure to unset replay flag from failed open requests (3440)
6388 - if default stripe count is 0, use OST count for inode size (3636)
6389 - update parent mtime/ctime on client for create/unlink (2611)
6390 - drop dentry ref in ext3_add_link from open_connect_dentry (3266)
6391 - free recovery state on server during a forced cleanup (3571)
6392 - unregister_reply for resent reqs (3063)
6393 - loop back devices mounting and status check on 2.6 (3563)
6394 - fix resource-creation race that can provoke i_size == 0 (3513)
6395 - don't try to use bad inodes returned from MDS/OST fs lookup (3688)
6396 - more debugging for page-accounting assertion (3746)
6397 - return -ENOENT instead of asserting if ost getattr+unlink race (3558)
6398 - avoid deadlock after precreation failure (3758)
6399 - fix race and lock order deadlock in orphan handling (3450, 3750)
6400 - add validity checks when grabbing inodes from l_ast_data (3599)
6402 - add /proc/.../recovery_status to obdfilter (3428)
6403 - lightweight CDEBUG infrastructure, debug daemon (3668)
6404 - change default OSC RPC parameters to be better on small clusters
6405 - turn off OST read cache for files smaller than 32MB
6406 - install man pages and include them in rpms (3100)
6407 - add new init script for (un)mounting lustre filesystems (2593)
6408 - run chkconfig in %post for init scripts (3701)
6409 - drop scimac NAL (unmaintained)
6411 2004-06-17 Cluster File Systems, Inc. <info@clusterfs.com>
6414 - clean kiobufs before and after use (3485)
6415 - strip trailing '/'s before comparing paths with /proc/mounts (3486)
6416 - remove assertions to work around "in-flight rpcs" recovery bug (3063)
6417 - change init script to fail more clearly if not run as root (1528)
6418 - allow clients to reconnect during replay (1742)
6419 - fix ns_lock/i_sem lock ordering deadlock for kms update (3477)
6420 - don't do DNS lookups on NIDs too small for IP addresses (3442)
6421 - re-awaken ptlrpcd if new requests arrive during check_set (3554)
6422 - fix cond_resched (3554)
6423 - only evict unfinished clients after recovery (3515)
6424 - allow bulk resend, prevent data loss (3570)
6425 - dynamic ptlrpc request buffer allocation (2102)
6426 - don't allow unlinking open directory if it isn't empty (2904)
6427 - set MDS/OST threads to umask 0 to not clobber client modes (3359)
6428 - remove extraneous obd dereference causing LASSERT failure (3334)
6429 - don't use get_cycles() when creating temp. files on the mds (3156)
6430 - hold i_sem when setting i_size in ll_extent_lock() (3564)
6431 - handle EEXIST for set-stripe, set proper directory name (3336)
6433 - servers can dump a log evicting a client - lustre.dump_on_timeout=1
6434 - fix ksocknal_fmb_callback() error messages (2918)
6436 2004-05-27 Cluster File Systems, Inc. <info@clusterfs.com>
6439 - don't copy lvb into (possibly NULL) reply on error (2983)
6440 - don't deref dentry after dput, don't free lvb on error (2922)
6441 - use the kms to determine writeback rpc length (2947)
6442 - increment oti_logcookies when osc is inactive (2948)
6443 - update client's i_blocks count via lvb messages (2543)
6444 - handle intent open/close of special files properly (1557)
6445 - mount MDS with errors=remount-ro, like obdfilter (2009)
6446 - initialize lock handle to avoid ASSERT on error cleanup (3057)
6447 - don't use cancelling-locks' kms values (2947)
6448 - use highest lock extent for kms, not last one (2925)
6449 - don't dereference ERR_PTR() dentry in error handling path (3107)
6450 - fix thread race in portals_debug_dumplog() (3122)
6451 - create lprocfs device entries at setup instead of at attach (1519)
6452 - common AST error handler, don't evict client on completion race (3145)
6453 - zero nameidata in detach_mnt in 2.6 (3118)
6454 - verify d_inode after revalidate_special is valid in 2.6 (3116)
6455 - use lustre_put_super() to handle zconf unmounts in 2.6 (3064)
6456 - initialize RPC timeout timer earlier for 2.6 (3219)
6457 - don't dereference NULL reply buffer if mdc_close was never sent (2410)
6458 - print nal/nid for unknown nid (3258)
6459 - additional checks for oscc recovery before doing precreate (3284)
6460 - fix ll_extent_lock() error return code for 64-bit systems (3043)
6461 - don't crash in mdc_close for bad permissions on open (3285)
6462 - zero i_rdev for non-device files (3147)
6463 - clear page->private before handing to FS, better assertion (3119)
6464 - tune the read pipeline (3236)
6465 - fix incorrect decref of invalidated dentry (2350)
6466 - provide read-ahead stats and refine rpc in flight stats (3328)
6467 - don't hold journal transaction open across create RPC (3313)
6468 - update atime on MDS at close time (3265)
6469 - close LDAP connection when recovering to avoid server load (3315)
6470 - update iopen-2.6 patch with fixes from 2399,2517,2904 (3301)
6471 - don't leak open file on MDS after open resend (3325)
6472 - serialize filter_precreate and filter_destroy_precreated (3329)
6473 - loop device shouldn't call sync_dev() for nul device (3092)
6474 - clear page cache after eviction (2766)
6475 - resynchronize MDS->OST in background (2824)
6476 - refuse to mount the same filesystem twice on same mountpoint (3394)
6477 - allow llmount to create routes for mounting behind routers (3320)
6478 - push lock cancellation to blocking thread for glimpse ASTs (3409)
6479 - don't call osc_set_data_with_check() for TEST_LOCK matches (3159)
6480 - fix rare problem with rename on htree directories (3417)
6482 - allow default OST striping configuration per directory (1414)
6483 - fix compilation for qswnal for 2.6 kernels (3125)
6484 - increase maximum number of MDS request buffers for large systems
6485 - change liblustreapi to be useful for external progs like lfsck (3098)
6486 - increase local configuration timeout for slow disks (3353)
6487 - allow configuring ldlm AST timeout - lustre.ldlm_timeout=<seconds>
6489 2004-03-22 Cluster File Systems, Inc. <info@clusterfs.com>
6492 - fixes for glimpse AST timeouts / incorrectly 0-sized files (2818)
6493 - don't overwrite extent policy data in reply if lock was blocked (2901)
6494 - drop filter export grants atomically with removal from device (2663)
6495 - del obd_self_export from work_list in class_disconnect_exports (2908)
6496 - don't LBUG if MDS recovery times out during orphan cleanup (2530)
6497 - swab reply message in mdc_close, other PPC fixes (2464)
6498 - fix destroying of named logs (2325)
6499 - overwrite old logs when running lconf --write_conf (2264)
6500 - bump LLOG_CHUNKSIZE to 8k to allow for larger clusters (2306)
6501 - fix race in target_handle_connect (2898)
6502 - mds_reint_create() should take same inode create lock (2926)
6503 - correct journal credits calculated for CANCEL_UNLINK_LOG (2931)
6504 - don't close files for self_export to avoid uninitialized obd (2936)
6505 - allow MDS with the same name as client node (2939)
6506 - hold dentry reference for closed log files for unlink (2325)
6507 - reserve space for all logs during transactions (2059)
6508 - don't evict page beyond end of stripe extent (2925)
6509 - don't oops on a deleted current working directory (2399)
6510 - handle hard links to targets without a parent properly (2517)
6511 - don't dereference NULL lock when racing during eviction (2867)
6512 - don't grow lock extents when lots of conflicting locks (2919)
6514 2004-03-04 Cluster File Systems, Inc. <info@clusterfs.com>
6517 - account for cache space usage on clients to avoid data loss (974)
6518 - lfsck support in lustre kernel code (2349)
6519 - reduce journal credits needed for BRW writes (2370)
6520 - orphan handling to avoid losing space on client/server crashes
6521 - ptlrpcd can be blocked, stopping ALL progress (2477)
6522 - use lock value blocks to assist in proper KMS, faster stat (1021)
6523 - takes i_sem instead of DLM locks internally on obdfilter (2720)
6524 - recovery for initial connections (2355)
6525 - fixes for mds_cleanup_orphans (1934)
6526 - abort_recovery crashes MDS in b_eq (mds_unlink_orphan) (2584)
6527 - block all file creations until orphan recovery completes (1901)
6528 - client remove rq_connection from request struct (2423)
6529 - conf-sanity test_5, proper cleanup in umount log not availale (2640)
6530 - recovery timer race (2670)
6531 - mdc_close recovey bug (2532)
6532 - ptlrpc cleanup bug (2710)
6533 - mds timeout on local locks (2588)
6534 - namespace lock held during RPCs (2431)
6535 - handle interrupted sync write properly (2503)
6536 - don't try to handle a message that hasn't been replied to (2699)
6537 - client assert failure during cleanup after abort recovery (2701)
6538 - leak mdc device after failed mount (2712)
6539 - ptlrpc_check_set allows timedout requests to complete (2714)
6540 - wait for inflight reqs when ptlrpcd finishes (2710)
6541 - make sure unregistered services are removed from the srv_list
6542 - reset bulk XID's when resending them (caught by 1138 test)
6543 - unregister_bulk after timeout
6544 - fix lconf error (2694)
6545 - handle write after unfinished setstripe, stripe-only getstripe (2388)
6546 - readahead locks pages, leaves pending causing memory pressure (2673)
6547 - increase OST request buffers to 4096 on large machines (2729)
6548 - fix up permission of existing directories in simple_mkdir (2661)
6549 - init deleted item, add assertions ptlrpc_abort_inflight() (2725)
6550 - don't assign transno to errored transactions (2742)
6551 - don't delete objects on OST if given a bogus objid from MDS (2751)
6552 - handle large client PAGE_SIZE readdir on small PAGE_SIZE MDS (2777)
6553 - if rq_no_resend, then timeout request after recovery (2432)
6554 - fix MDS llog_logid record size, 64-bit array alignment (2733)
6555 - don't call usermode_helper from ptlrpcd, DEFAULT upcall (2773)
6556 - put magic in mount.lustre data, check for bad/NULL mount data (2529)
6557 - MDS recovery shouldn't delete objects that it has given out (2730)
6558 - if enqueue arrives after completion, don't clobber LVB (2819)
6559 - don't unlock pages twice when trigger_group_io returns error (2814)
6560 - don't deref NULL rq_repmsg if ldlm_handle_enqueue failed (2822)
6561 - don't write pages to disk if there was an error (1450)
6562 - don't ping imports that have recovery disabled (2676)
6563 - take buffered bytes into account when balancing socknal conn (2817)
6564 - hold a DLM lock over readdir always, use truncate_inode_pages (2706)
6565 - reconnect unlink llog connection after MDS reconnects to OST (2816)
6566 - remove little-endian swabbing of llog records (1987)
6567 - set/limit i_blksize to LL_MAX_BLKSIZE on client (2884)
6568 - retry reposting request buffers if they fail (1191)
6569 - grow extent at grant time to avoid granting a revoked lock (2809)
6570 - lock revoke doesn't evict page if covered by a second lock (2765)
6571 - disable VM readahead to avoid reading outside lock extents (2805)
6573 - return LL_SUPER_MAGIC from statfs for the filesystem type (1972)
6574 - updated kernel patches for hp-2.4.20 kernel (2681)
6576 2004-02-07 Cluster File Systems, Inc. <info@clusterfs.com>
6579 - fix truncated write corruption (2366)
6580 - fix for failed assertion in iopen_connect_dentry (1792,2517)
6582 - don't flag the ptlrpcd thread with PF_MEMALLOC (2636)
6583 - ensure len(uuid) < 37 in lmc (1171)
6584 - fix ia64 OOPS in llog_test (2255)
6585 - zero end of page at obdfilter for partial page writes (2648)
6586 - don't leave stale dentries around after renames (bug 2428)
6587 - fix timeouts when evicting a client with a single lock held (2642)
6588 - set deadline for the initial HELLO message to drain (2634)
6589 - print out dotted-quad IP addresses in the socknal (2302)
6591 - additional debugging for MDS client eviction problem (2443)
6592 - fix mkfsoptions support for osts (2603, 2604)
6594 2004-01-27 Cluster File Systems, Inc. <info@clusterfs.com>
6597 - add series for the vanilla 2.6.0 kernel
6598 - add series for the vanilla 2.4.24 kernel
6599 - add series for a cray x86/64 UL kernel drop
6600 - fix xattr patches for the vanilla 2.4.19 series
6602 - generate true UUIDs in lmc (1171)
6603 - have portals stack dumping break in UML (2466)
6604 - avoid bad dchild deref; avoid inum lock w/o creation (2362)
6605 - allocate with _NOFS in ldlm to avoid deadlock (1933)
6606 - wake callback waiting threads on client eviction (2460)
6607 - Add --ptldebug and --subsystem to lmc (1719)
6608 - update assertion to allow safe interrupt allocation
6609 - set rq_no_resend for cancel requests (2432)
6610 - recalculate ptlrpcd timeout after resend (2494)
6611 - call vfs_rmdir when removing pending directories (2368)
6612 - fix renaming a file to itself (2429)
6613 - lmc creates a default one-stripe lov (2454)
6614 - expand procfs space to handle large clusters (2326)
6615 - increase UML stack to avoid overflow
6616 - update lconf's list of debug and subsystem masks
6617 - fix lfs find --obd (2510)
6618 - /proc tunable for disabling filter read caching (2591)
6619 - stop rpm packages from altering slapd.conf (2301)
6620 - disable nagle in the socknal under 0conf (2578)
6621 - choose mds inode size based on stripe count (2572)
6622 - fix kernel-source rpm problems (2516)
6624 - add --disable-doc to avoid pdf generation (2421)
6625 - update documentation, tests, type-os, comments
6626 - avoid format warnings on ia64
6627 - remove the TOE NAL
6628 - tiny code cleanups by removing unused fields
6630 2004-01-07 Cluster File Systems, Inc. <info@clusterfs.com>
6633 - fix obvious semaphore misuse in as-yet-unused setattr path (2348)
6634 - remove the most blatant lies from BUILDING file (2371)
6635 - change default debug level to reasonable production setting
6636 - reduce client side cache size to reduce cache flush time
6637 - reduce max RPCs in flight to avoid unnecessary file fragmentation
6638 - make TCP zerocopy and pinger support enabled by default (2476)
6639 - sync writes completed after process exits caused crashes (2319)
6640 - maintain correct mount count on the MDS (2356)
6641 - backout 1557, because 2316 wasn't really fixed
6642 - better file I/O statistics gathering in /proc
6643 - don't take unnecessary, deadlock-inducing bug in readpage (2383)
6644 - another kernel patch to fix zero-copy TCP function export
6645 - don't take duplicate lock when processing re-sent getattr (2420)
6646 - lctl uses obd_self_export instead of creating new conn (2353)
6647 - MDS/OST recovery case which requires object creation asserted (2425)
6648 - move lfs from /usr/sbin to /usr/bin in packages
6649 - fix race between mds_client_add and mds_client_free (2417)
6650 - use kmalloc instead of slabs in portals (2430)
6651 - don't create duplicate records when a failover MDS is present (2442)
6652 - remove unnecessary mount age check (2332)
6653 - don't remove directory inodes from locks prematurely (2451)
6654 - don't break if MDS service name is the same as hostname (2103)
6655 - fix races in client write RPC generation when cache full (2482)
6657 2003-12-13 Cluster File Systems, Inc. <info@clusterfs.com>
6660 - remove now-unused request->rq_obd (278)
6661 - if an allocation fails, print out how much memory we've used (1933)
6662 - use PORTAL_SLAB_ALLOC for structures, to get GFP_MEMALLOC (1933)
6663 - add the "configurable stack size" patch to most series files (1256)
6664 - ability to write large log records, for 100+ OST configs (2306)
6665 - fix NULL deref when filter_prep fails (2314)
6666 - fix operator precedence error in filter_sync
6667 - dynamic allocation of socknal TX descriptors (2315)
6668 - fix a missed case in the GFP_MEMALLOC patch, can cause deadlock (2310)
6669 - fix gcc 2.96 compilation problem in xattr kernel patch (2294)
6670 - ensure that CWARN messages in Portals always get to the syslog
6671 - __init/__exit are not for prototype decls (ldlm_init/exit)
6672 - x86-64 compile warning fixes
6673 - fix gateway LMC keyword conflict (2318)
6674 - fix MDS lock inversions in getattr/reint paths (1844)
6675 - fix a rare lock re-ordering bug, which caused deadlock (2322)
6676 - fix i_sem/journal inversion in fsfilt_ext3_write_record (2306)
6677 - DLM race condition prevented some lock evictions (2328)
6678 - ENOMEM detection and retry on socknal sends (2230)
6679 - use GFP_NOFS throughout Lustre, to combat ENOMEM (2230)
6680 - move osc_rpcd into ptlrpc, for use in MDC and others (2329)
6681 - protect MDS inode fsdata with stronger locking; fixes assertion (2313)
6682 - better error messages when a client is rejected during recovery (1505)
6683 - avoid cancelling locks which were never granted, after failure (2330)
6684 - fix i_sem/journal inversion in mds_client_add (2333)
6685 - fix truncate/getattr lock cycle deadlock (2334)
6686 - use rpcd to send close; allows resend after timeout, avoid leak (1897)
6687 - fix two rare exit paths which could leak an l_lock() ref (2321)
6688 - fencepost error in MDS/OST orphan recovery (2226)
6689 - make log record alignment 8 bytes (1988)
6690 - lstripe now fails when requested offset > ost_count (2237)
6691 - ensure that all kernel series have a complete list.h (1607)
6692 - fix crashes in special-file operations (2316)
6693 - lctl create/brw OID mismatch, caused by obsolete filter loop (2339)
6695 - allow configurable automake binary, for testing new versions
6696 - small update to the lfs documentation
6698 2003-12-03 Cluster File Systems, Inc. <info@clusterfs.com>
6700 * fix negative export reference count in fsfilt_sync (2312)
6702 2003-12-01 Cluster File Systems, Inc. <info@clusterfs.com>
6703 * release candidate 0.9.1
6705 - orphans are moved into the PENDING directory for possible recovery
6706 - replayed opens now open by fid for orphan/rename safety (1042)
6707 - last close of an orphan inode generates a transno (683)
6708 - chdir() and mount() now pin the directory entry (1020)
6709 - avoid CERROR in normal ll_setattr_raw() error case (1500)
6710 - discard very old requests without processing them (1502)
6711 - remove some common, well-understood CERRORs (1505)
6712 - require O_DIRECT I/O to be page-sized to workaround IA64 crash (1609)
6713 - clear "grant" flags in OST replies until OST grant code lands (1644)
6714 - fix read performance by not clobbering i_blksize on client (1598)
6715 - fix __ldlm_handle2lock oops by not dereferencing lock after PUT (1625)
6716 - make LRU size a /proc tunable, clears locks when reduced (707)
6717 - fix some lprocfs rot that prevented ptlbd from loading (1732)
6718 - server locks take references on exports now (1558)
6719 - build fixes for 2.4.20-rh trees (1663)
6720 - return an error from lov_create if all OSCs are inactive (1751)
6721 - fix import levels when a reconnect happens without a timeout (1597)
6722 - exit early from mds_open if we get a lookup error (1749)
6723 - partial page read at EOF wouldn't wait for disk before sending (1642)
6724 - avoid NULL deref in obdfilter when reading page past EOF (1592)
6725 - avoid LASSERT in ll_intent_lock if server failed very early (1090)
6726 - fix LBUG in ll_it_open_error with rc = -2 (1861)
6727 - write/truncate lock inversion (1639)
6728 - Don't auto-load obdclass, portals modules during cleanup (1495)
6729 - fix timestamps from jumping to "now" (1763)
6730 - extra journal assertions (1648)
6731 - add an extra multiunlink test (1771)
6732 - fix read_record/write_record API (1776)
6733 - fix leak of offset_extent, possible incorrect i_size later (1772)
6734 - fix lasserts in mis-matched transnos during open-unlink testing (1541)
6735 - Debugging for the kqswnal_get_idle_tx problems (1820)
6736 - Allow recovery to be attempted multiple times (1536)
6737 - Write out MDS last_rcvd file after it is first created (1600)
6738 - Fix tx_descriptor leak in failed transmit situations (1827)
6739 - ext3 journaling fixes for assertion failure after IO error (1871)
6740 - class_export_put() on freed export after completion AST error (1896)
6741 - Fix revalidate looping in VFS (1322)
6742 - Don't access a freed export during MDS_REINT timeout (1521)
6743 - Add open-unlink recovery support on the MDS (1673,1764)
6744 - Return an error if no MDS data was read from last_rcvd (1946)
6745 - Fix for lookup "." or ".." crash on error (1932,1931,1935)
6746 - Don't setup a disk device that doesn't match exported UUID (317)
6747 - Reduce bulk RPC timeout to avoid cascading client/OST failures (1845)
6748 - avoid committing NULL handle in force close
6749 - local.sh is now a one-stripe LOV configuration
6750 - POSIX utime.4 -EPERM on FIFO not owned by user (56)
6751 - fix ext3 htree duplicate directory entry corruption (1516)
6752 - POSIX creat.13, fstat.1, open.18, stat.3 new file atime/mtime (2020)
6753 - update to new LOV EA format (2097)
6754 - interoperability for different PAGE_SIZE/wordsize (686,1821,1343,2042)
6756 2003-06-15 Phil Schwan <phil@clusterfs.com>
6759 - imports and exports cleanup too early, need refcounts (349, 879, 1045)
6760 - per-import/export recovery handling (958, 931, 959)
6761 - multiple last-rcvd slots, for serving multiple FSes (949)
6762 - connections are again shared between multiple imp/exports (963, 964)
6763 - "umount -f" would hang if any requests needed to be sent (393, 978)
6764 - avoid pinning large req buffer by copying for queued messages (989)
6765 - add "uuid" to "lctl device" command to help upcalls (991)
6766 - "open" RPCs with transnos would confuse recovery counters (1037)
6767 - do proper endian conversion of all wire messages (288, 340, 891)
6768 - remove OST bulk get LBUGs, fix ost_brw_write cleanup (1126)
6769 - call waiting locks callback from LDLM recovery thread (1127, 1151)
6770 - fix ptlrpc_connection leak in target_handle_connect (1174)
6771 - fix import refcounting bug in OST and MDS cleanup (1134)
6772 - if an invalid-at-open-time OSC returned before close(), LBUG (1150)
6773 - fix very unlikely obd_types race condition (501)
6774 - remove osc_open hack for echo_client (1187)
6775 - we leaked exports/dlmimps for forcibly disconnected clients (1143)
6776 - a failure in read_inode2 leads to deadlock (1139)
6777 - cancel ack-locks as soon as transaction is committed (1072)
6778 - fix major leaks and crashes in the bulk I/O path (937, 1057)
6779 - make sure to commitrw after any preprw to avoid deadlock (1162)
6780 - failing to execute a file in a lustre FS would lock inode (1203)
6781 - small DEBUG_REQ fix to avoid dereferencing a NULL (1227)
6782 - don't ASSERT while cleaning up an incompletely-setup obd (1248)
6783 - obd_uuid2tgt would walk off the end of the list (1255)
6784 - on IA64 the osc would give portals incorrect bulk size (1258)
6785 - fix debug daemon ioctl interface; allows daemon on ia64 (1274)
6786 - fix lock inversion caused by new llite matching code (1282)
6787 - limit the number of dirty pages on a client to 10MB (1286)
6788 - timed out locks were not being corrected cancelled (1289)
6789 - fix O_DIRECT above 4GB on IA-32 (1292)
6790 * major user-visible changes
6791 - fail out/fail over policy now controlled by the upcall (993)
6793 - add OBD_PING to check server availability and failure (954)
6794 - lustre messages are now sent in sending host order (288, 340, 891)
6795 - add eadatalen to MDS getattr reply (340)
6796 - OST read replies may contain second buffer, with per-page status (593)
6798 2003-03-11 Phil Schwan <phil@clusterfs.com>
6801 - LDLM_DEBUG macro fix, for gcc 3.2 (850)
6802 - failed open()s could cause deadlock; fixed (867, 869)
6803 - stop cancelling OST locks when files are closed (481)
6804 - overlapping XID spaces caused network corruption (851, 853)
6805 - fix unsafe fsfilt counter arithmetic; change to atomic_t
6806 - setattr_raw added, to do single-RPC, server-side setattrs
6807 - lmc/lconf syntax change for OST UUIDs
6808 - fix crashy race condition between ptlrpc_free_req and osc_close
6809 - don't use request in mdc_enqueue if we hit a timeout (889)
6810 - don't set the inode i_size for regular files from the MDS (896)
6811 - handle out of order completion AST (842)
6812 - don't LBUG if a lock request times out after receiving AST (913)
6813 - avoid d_rehash race in ll_find_alias by rehashing inside dcache_lock
6814 - if a bad lock AST arrives, send an error instead of dropping entirely
6815 - return 0 from revalidate2 if ll_intent_lock returns -EINTR (912)
6816 - fix leak in bulk IO when only partially completed (899, 900, 926)
6817 - fix O_DIRECT for ia64 (55)
6818 - (almost) eliminate Lustre-kernel-thread effects on load average (722)
6819 - C-z after timeout could hang a process forever; fixed (977)
6821 - client-side I/O cache (678, 924, 929, 941, 970)
6823 - READPAGE and SETATTRs which don't take server-side locks get
6826 2003-02-11 Phil Schwan <phil@clusterfs.com>
6829 - Fix ldlm_lock_match on the MDS to avoid matching remote locks (592)
6830 - Fix fsfilt_extN_readpage() to read a full page of directory
6831 entries, or fake the remainder if PAGE_SIZE != blocksize (500)
6832 - Avoid extra mdc_getattr() in ll_intent_lock when possible (534, 604)
6833 - Fix imbalanced LOV object allocation and out-of-bound access (469)
6834 - Most intent operations were removed, in favour of a new RPC mode
6835 that does a single RPC to the server and bypasses most of the VFS
6836 - All LDLM resource ID arrays were removed in favour of ldlm_res_id
6837 - Aggressively cancel local locks on DLM servers
6838 - mds_reint_unlink sends EA to the client if it's the last nlink.
6839 client uses that EA to unlink OST objects.
6840 - mds_reint_{rename,unlink,link} were rewritten to take ordered locks
6841 - recursive symlinks were fixed (439)
6842 - fixed NULL deref in DEBUG_REQ
6843 - filter_update_lastobjid no longer calls sync, which annoyed extN
6844 - fixed multi-client small-writes to a single file problem (445)
6845 - fixed mtime updates during file writes (607)
6846 - fixed vector writes on obdfilter causing problems when ENOSPC (670)
6847 - fixed bug in obd_brw_read/write() (under guise of testing 367)
6848 - fixed Linux OST size reporting problem (444, 656)
6849 - OST now updates object mtime with writes or setattr (607, 619)
6850 - client verifies file size before zeroing page past EOF (445)
6851 - OST now writes last allocated objid to disk with allocation (108)
6852 - LOV on echo now works (409)
6854 - mds_reint_unlink sends a new buffer, with the EA included. this
6855 buffer is only valid if body->valid & OBD_MD_FLEASIZE, which is only
6856 set if a regular file was being unlinked, and it was the last link
6857 - use PtlGet from the target for bulk writes (315)
6858 - OST now updates object mtime with writes or setattr (607, 619)
6859 - LDLM now has a grant-time callback to revalidate locked items, if
6861 - Many MDS operations were reorganized to combat race conditions
6863 - Merge b_intel branch (updated lprocfs code) - now at /proc/fs/lustre
6864 - configure check to avoid gcc version 2.96 20000731-2.96-98) (606)
6866 2003-01-06 Andreas Dilger <adilger@clusterfs.com>
6869 - Fully reactivate OST imports after reconnection (512, others)
6870 - Make sure client sees our -ENOTCONN from mds_handle (513 - partial)
6871 - More graceful error handling for truncating on dead OST (515)
6872 - Don't error out unless we're actually accessing dead stripes (474)
6873 - Fix garbage sizes when stripes are missing (410)
6874 - LRU counters were broken, causing constant lock purge (433, 432)
6875 - garbage on read from stripes with failed OSTs (441)
6876 - mark OSCs as active before reconnecting during recovery (438)
6877 - lov_enqueue and lov_cancel need to handle inactive OSTs (403)
6878 - lfind did not preserve OST order in output (443)
6879 - symlinks cause hung clients, incorrect data (439)
6880 - stop dereferencing request after dropping refcount (457)
6881 - don't LASSERT(spin_is_locked) on non-SMP (455)
6882 - fixes for many rename() bugs
6883 - fstat didn't correctly synchronize attributes (399)
6884 - server must handle lock cancellation during blocking AST prep (487)
6885 - bulk descriptors were free()d too soon (511)
6886 - fix paths in lconf, which would load incorrect modules (451, 507)
6887 - fix confusing lconf 'host not found' error message (386)
6888 - fix lock order deadlock on OST (O/R i_sem before journal ops, 478)
6889 - fix race condition in mdc_blocking_ast() for inode access (526)
6890 - fix lov_unpackmd() unpacking wrong number of stripes (537)
6891 - fix lov_set_osc_active() marking wrong OSC inactive (440)
6892 - fix bad lstripe lov_unpackmd() assertion (fix layering too) (527)
6893 - fix multiple writes of stripe MD to MDS (358, maybe 519)
6894 - fix lstripe in several ways (kernel side) (527)
6895 - fix request leak in ldlm_cli_enqueue (262)
6896 - incorrect OSC was marked inactive after OST failure
6897 - call mds_fs_cleanup before unmounting filesystem (524)
6898 - fix races between taking ns_lock and ldlm_lock_change_resource
6899 - fix races updating LOV export open file list
6900 - fix lov_enqueue error path, avoid decref-ing bad lock handle
6901 - fix recovery NULL deref in ldlm_cli_cancel_unused
6902 - fix some DLM races by using new hash table for lock handles (419)
6903 - permit the client to specify desired inodes, at replay
6904 - duplicate requests when we queue them for replay reintegration
6905 - fix last_rcvd offset calculation
6906 - sync after each recovered transaction, so we always make progress
6907 - never, not always, ERESTART requests without transnos
6908 - store the lov_desc in the MDS, so we don't depend on getlovinfo to
6910 - skip replay if the MDS says that the client is already connected
6911 - don't check for a recovery-enabled export to match lctl's UUID
6912 - don't INC_USE_COUNT for phantom exports
6913 - don't crash when cleaning up phantom exports (567)
6914 - don't double-finish or set replay data for errored mdc_open requests
6915 - abort requests when they time out, so we don't get old replies
6916 - send/receive replies for AST messages again
6917 - if the client says that it doesn't have the lock, cancel it on the
6919 - if we timeout during I/O, don't try to cancel an in-use lock; instead
6920 mark it as destroyed, it will all work out when decref is called
6921 - fix module use counts (22, 581)
6923 - ASTs now expect a reply (server cancels lock on error reply)
6925 2002-12-02 Andreas Dilger <adilger@clusterfs.com>
6928 - fix many simultaneous client startup (392)
6929 - fix dentry->d_it clobbering
6930 - credentials weren't being shipped for readdir/getattr operations
6931 - remove invalid assertions triggered during some concurrent MD
6933 - proper Lustre versions added (336, 389)
6934 - fix memory leak for create error case (398)
6935 - fix LOV locking bug that would get cli/srv out of sync
6936 - fix echo client over LOV (409)
6937 - fix dbench 2, extN refcount problem (170, 258, 356, 418)
6938 - fix double-O_EXCL intent crash (424)
6939 - avoid sending multiple lock CANCELs (352)
6941 - MDS can do multi-client recovery (modulo bugs in new code)
6943 - many updates, edits, cleanups
6945 2002-11-18 Phil Schwan <phil@clusterfs.com>
6948 - fix null d_it dereference (346)
6949 - fix full OST/dbench hang (333)
6950 - fix permission problem with file removal (286)
6951 - fix removal of OSCs from LOV when they fail
6952 - fix NULL deref during bulk timeout (214)
6953 - fix problems related to multiple filesystems on one MDS (241)
6954 - fixed serious subtle metadata locking bugs
6955 - free locks on clients when inodes are removed due to memory
6957 - fix inode pointer in lock data (285)
6958 - partial support for multiple MDS on a single host (241)
6959 - data locks weren't cancelled at clear_inode time (290, 311)
6960 - intent locks could lead to unbounded lock growth (205)
6961 - added a maximum lock count, an LRU list, and a flusher
6962 - fix multiple rename (365)
6963 - properly abstracted the echo client
6964 - OSC locked 1 byte too many; fixed
6965 - rewrote brw callback code:
6966 - fixed recovery bugs related to LOVs (306)
6967 - fixed too-many-pages-in-one-write crash (191)
6968 - fixed (again) crash in sync_io_timeout (214)
6969 - probably fixed callback-related race (385)
6971 - Add capability to MDS protocol
6972 - LDLM cancellations and callbacks on different portals
6974 2002-10-28 Andreas Dilger <adilger@clusterfs.com>
6977 - limit client IOV size to PTL_MD_MAX_IOV (611336, 191)
6978 - defer open object destruction to close time (601981, 138)
6979 - open/close OST file handle in obdo (OBD_MD_FLHANDLE) (601981, 138)
6980 - move LDLM_ENQUEUE/CONVERT back to MDS portal (625069)
6981 - abstract ll_lookup2, fix ll_revalidate2 to use abstraction (256)
6982 - don't call obd_setattr in ll_file_release for destroyed objects
6983 * protocol change to lustre_msg: move |version| and add |flags|
6984 * protocol change to osc_punch: "start" in "o_size", "end" in "o_blocks"
6985 * lock replay: for LDLM_FL_REPLAY trust client to do right thing
6986 * added replay of create, unlink, link and rename operations during
6987 MDS failover; recovery should be much more robust now
6988 * remove failed OSCs from LOVs (only lov_create uses this so far)
6989 * the lustre-HOWTO was brought (more) up to date (582544)
6991 2002-10-23 Phil Schwan <phil@clusterfs.com>
6994 - in-use dentries weren't being reused properly (617851)
6995 - prevent multiple LDLM setup (599178)
6996 - fix LOV size calculations for truncate (617853)
6997 - fix client handling of MDS intent errors (POSIX)
6998 - fix permission bug in lovstripe.c test (624321)
6999 - fix MDS thread deadlock - move LDLM handler to DLM portal (625069)
7000 - truncate past end of file could corrupt data
7001 - proper cleanup after timeouts, crashes, etc (592524, 550815)
7002 - a race in recovery could return ETIMEDOUT to apps (623947)
7003 - building outside the source directory was fixed
7004 * the lustre-HOWTO was brought (more) up to date (582544)
7005 * major progress was made on recovery functionality
7007 2002-10-10 Phil Schwan <phil@clusterfs.com>
7010 - recovery deadlock fix
7011 - rm -rf causes LBUG fix (617817)
7012 - file open by multiple tasks fix (618962)
7013 - directory permissions bugs (602707 and 620007)
7014 - journal_stop fixed with locking (611313)
7015 - O_APPEND failures resolved (618273, perhaps 614459)
7016 - lconf PATH fix (619770)
7017 - IA64 build fix (621450)
7018 - RPC buffer sizes scale with amount of memory
7020 2002-10-01 Phil Schwan <phil@clusterfs.com>
7023 - locks would be cancelled without throwing away data pages,
7024 resulting in inconsistent data (605627)
7025 - inode attributes were not always being refreshed (605627, 612449)
7026 - lconf now continues to cleanup after lctl reports an error
7027 - MDS now enforces user permissions (602707)
7028 - lprocfs cleanup fixed, but not yet enabled (614157)
7029 - fixed infinite server hang, should a client not respond to an AST
7030 - avoid going into recovery if user calls readlink() with a buffer
7031 that's too small (613941)
7032 - AST RPCs no longer require replies (614867) -- this may be changed
7033 - don't crash server if client sends an IOV that's too big (611336)
7034 - fixed lock conversion deadlock (611892)
7035 - fixed the following of symlinks (614622)
7036 * recovery: the server can remove locks from a client that dies, other
7037 clients can make progress
7038 * more extN patch fixes
7039 * compile-time configurable ptlrpc buffer allocations
7041 - collaborative read cache document
7042 - Lustre Lite Performance CDR document-in-progress
7044 2002-09-20 Andreas Dilger <adilger@clusterfs.com>
7047 - fix typo in patch-2.4.18
7049 2002-09-20 Andreas Dilger <adilger@clusterfs.com>
7052 - clear ptlrpc request each time in handle_incoming_request()
7053 - unlink of files now destroys the object on the OST
7055 2002-09-19 Peter Braam <braam@clusterfs.com>
7057 * add hard link support
7058 * change obdfile creation method
7059 * kernel patch changed
7061 2002-09-19 Peter Braam <braam@clusterfs.com>
7064 - stack overflow bug in extN fixed
7066 2002-09-18 Andreas Dilger <adilger@clusterfs.com>
7068 * documentation updates
7069 - add man pages for config tools
7070 - update tests/README to describe testing with new config tools
7071 - finish metadata API descriptions
7072 * bug fixes and cleanups
7073 - statfs workaround for 16TB limit
7074 - LOV stripe allocation improved, can stripe on subset of OSTs
7075 - LOV file size/IO offset was wrong for files > 4GB in size
7076 - object EA data was being dropped, caused files to be unreadable
7077 - memory overflow with non-LOV OST caused memory corruption
7078 - fixed regression tests to work with new config tools, obdfilter
7079 - fixed bug when directory size became larger than 1 block
7080 - fixed bug (for single client case) when PWD was deleted
7081 - invalidate local directory pages when doing intent-based ops
7082 - avoid LDLM oops when lock callback contained bad data
7084 2002-09-09 Andreas Dilger <adilger@clusterfs.com>
7086 * documentation updates
7087 * bug fixes and cleanups
7088 - configuration tools
7091 - 64-bit compile warnings
7092 - 64-bit internal statfs data
7094 * test_brw on persistent OST devices
7096 * lprocfs (disabled)
7098 2002-09-04 Andreas Dilger <adilger@clusterfs.com>
7100 * documentation updates
7101 * bug fixes and cleanups
7102 * configuration tools
7104 2002-08-30 Peter J. Braam <braam@clusterfs.com>
7107 * many small fixes to 0_5_4
7108 * io/network handling
7109 * thinkos in MDS operations
7111 2002-08-24 Peter J. Braam <braam@clusterfs.com>
7114 * crucial basic fixes to 0.5.3
7115 * IOR, Iozone work over Elan
7118 2002-08-07 Phil Schwan <phil@clusterfs.com>
7119 * version 0_5_3, our first alpha
7120 * we use the new Portals iovs
7121 * documentation updates
7122 * bug fixes and cleanups
7123 * small changes in the DLM wire protocol
7125 2002-07-25 Peter J. Braam <braam@clusterfs.com>
7126 * version 0_5_1 with some initial stability,
7127 * locking on MD and file I/O.
7128 * documentation updates
7129 * several bug fixes since 0.5.0
7130 * small changes in wire protocol
7132 2002-07-18 Phil Schwan <phil@clusterfs.com>
7134 * delivered as Lustre Light Alpha
7135 * fixed a crash after handling invalid MDS requests
7136 * fixed directory pages for architectures with non-4k pages sizes
7138 2002-07-11 Andreas Dilger <adilger@clusterfs.com>
7139 * release version v0_4_4
7140 * Moves TCP acceptor to be on port 2432 (unused Coda port) instead
7142 * Fixes a number of interruption problems with OST operations.
7143 * Update documentation for portals header changes
7144 * Move all wire protocol structs/defines to lustre_idl.h
7145 * Fixes symlink length bug.
7146 * Add tcpdump to repository.
7148 2002-07-05 Andreas Dilger <adilger@clusterfs.com>
7149 * release version v0_4_3
7150 * Fixes statfs for inodes on extN.
7151 * Fixes bug in runtests which would delete /etc/hosts.
7152 * Use 64-bit object IDs wherever possible (not into VFS though)
7153 Remove ost_get_info, which is unused by lustre, and out of date.
7155 2002-07-03 Peter Braam <braam@clusterfs.com>
7156 * release version v0_4_2 Fixes a lookup error (type not passed)
7157 * move forward to head of Portals
7158 * move forward to latest Lustre kernel
7160 2002-06-25 Peter Braam <braam@clusterfs.com>
7161 * release version v0_4_1. Hopefully stable on single node use.