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: llobdstat fix and enhancement
18 Details : add a counter to set a limit to how many samples will be returned
19 fix a wildcard in the path to limit to obdfilter stats only
23 Description: recovery-small 51 hang - MDT can not stop, Mount still busy with
25 Details : abort lock enqueue processing sooner when we umount is in progress.
27 Severity : enhancement
29 Description: Update RHEL5.4 kernel to 2.6.18-164.11.1.el5 and
30 OEL5.4 kernel to 2.6.18-164.11.1.0.1.el5.
32 Severity : enhancement
34 Description: Simplify MDT/OST service start message.
37 Bugzilla : 21961/17914
38 Description: ignore trailing -mdc when determining index number
42 Description: decrease the usage of memory on clients.
43 Details : 1. On clients, recycle dentries and inodes unused.
44 2. Delete the code related to ll_deathrow(att 6215 in bug 1443). It
46 -------------------------------------------------------------------------------
48 2010-01-29 Sun Microsystems, Inc.
50 * Support for kernels:
51 2.6.16.60-0.42.8 (SLES 10),
52 2.6.27.39-0.3.1 (SLES11),
53 2.6.18-164.11.1.el5 (RHEL 5)
54 2.6.18-164.6.1.0.1.el5 (OEL 5)
55 * Client support for unpatched kernels:
56 (see http://wiki.lustre.org/index.php?title=Patchless_Client)
57 2.6.16 - 2.6.30 vanilla (kernel.org)
58 * Recommended e2fsprogs version: 1.41.6.sun1
59 * The async journal commit feature (bug 19128) and the cancel
60 lock before replay feature (bug 16774) are disabled by default.
62 Severity : enhancement
64 Description: Update RHEL5.4 kernel to 2.6.18-164.11.1.el5 and
65 OEL5.4 kernel to 2.6.18-164.11.1.0.1.el5.
67 Severity : enhancement
68 Bugzilla : 21511/19848
69 Description: Update SLES11 kernel to 2.6.27.39-0.3.1.
71 Severity : enhancement
73 Description: Update supported SLES10 kernel to 2.6.16.60-0.42.8.
75 Severity : enhancement
77 Description: Update kernel to RHEL5.4 2.6.18-164.6.1.el5 and
78 OEL5 2.6.18-164.6.1.0.1.el5(Both in-kernel OFED enabled).
80 Severity : enhancement
82 Description: Build kernels (RHEL5, OEL5 and SLES10/11) using the vendor's own
85 Severity : enhancement
87 Description: Vanilla kernel 2.6.30 patchless client support.
92 Description: bad entry in directory xxx: inode out of bounds
93 Details : fix locking issue in the rename path which could race with any
94 other operations updating the same directory.
96 Severity : enhancement
98 Description: Make watchdog timer messages to be more clear and descriptive.
102 Description: cp -p command does not preserve the dates and timestamp
103 Details : mtime could be spoiled by a write callback
107 Description: Clear imp_force_reconnect correctly in ptlrpc_connect_interpret()
111 Description: Allow non-root access for "lfs check".
112 Details : Added a check in obd_class_ioctl() for OBD_IOC_PING_TARGET.
114 Severity : enhancement
116 Description: quotacheck performance/scaling issues
117 Details : reduce quotacheck time on empty filesystem by skipping uninit
120 Severity : enhancement
122 Description: Enhancement for lfs(1) command to use numeric uid/gid.
124 Severity : enhancement
126 Description: Adjust locks' extents on their first enqueue, so that at the time
127 they get granted, there is no need for another pass through the
128 queues since they are already shaped into the proper forms.
132 Description: Fix mds_shrink_intent_reply()/mds_intent_policy() to pass correct
133 arguments and prevent LBUG() in lustre_shrink_reply_v2().
137 Description: Change tunefs.lustre and mkfs.lustre --mountfsoptions so that
138 exactly the specified mount options are used. Leaving off
139 any "mandatory" mount options is an error. Leaving off any
140 default mount options causes a warning, but is allowed.
141 Change errors=remount-ro from mandatory to default. Sanitize
142 the mount string before storing it. Update man pages accordingly.
146 Description: mds_getattr() should return 0, even if mds_fid2entry() fails with
147 -ENOENT. Also fix in ptlrpc_expire_one_request() to print signed
150 Severity : enhancement
152 Description: Remove set_info(KEY_UNLINKED) from MDS/OSC
154 Severity : enhancement
156 Description: Clients can replay thousands of unused locks during recovery
157 Details : Don't replay unused locks (only read locks for now) during
158 recovery. This feature is disabled by default and can be
159 enabled by running the following command on the clients:
160 lctl get_param ldlm.cancel_unused_locks_before_replay
164 Description: can't stat file in some situation.
165 Details : improve initialize osc date when target is added to mds and
166 ability to resend too big getattr request is client isn't have info
171 Description: Prevent inconsistences between linux and lustre mount structures.
172 Details : Wait indefinitely in server_wait_finished() until mnt_count drops.
173 Make the sleep interruptible.
175 Severity : enhancement
177 Description: Communicate OST degraded/readonly state via statfs to MDS
178 Details : Flags in the statfs returned from OSTs indicate whether the
179 OST is in a degraded RAID state, or if the filesystem has
180 turned read-only after a filesystem error is detected.
185 Description: don't panic if EPROTO was hit when reading symlink
186 Details : correctly handling request reference in error cases.
191 Description: open sometimes returns ENOENT instead of EACCES
192 Details : checking permission should be part of open part of mds_open, not
193 lookup part. so server should be set DISP_OPEN_OPEN disposition
194 before starting permission check.
195 Also not need revalidate dentry if client already have LOOKUP lock.
199 Description: enable client interface failover
200 Frequency : on servers with multiple network interfaces
201 Details : When a child reconnects from another NID, properly update export
202 nid hash position and ldlm reverse import.
204 Severity : enhancement
206 Description: implemented direct I/O with arbitrary (nonaligned) memory
207 addresses and file offsets.
209 Severity : enhancement
211 Description: added more recovery timeout options.
213 Severity : enhancement
215 Description: added llapi_file_open, llapi_file_create, llapi_file_get_stripe
220 Description: Avoid deadlock for local client writes
221 Frequency : only on systems with clients writing to an OST on the same node
222 Details : Use new OBD_BRW_MEMALLOC flag to notify OST about writes in the
223 memory freeing context. This allows OST threads to set the
224 PF_MEMALLOC flag on task structures in order to allocate memory
225 from reserved pools and complete IO.
226 Use GFP_HIGHUSER for OST allocations for non-local client writes,
227 so that the OST threads generate memory pressure and allow
228 inactive pages to be reclaimed.
233 Descriptoin: lock ordering violation between &cli->cl_sem and _lprocfs_lock
234 Details : move ldlm namespace creation in setup phase to avoid grab
235 _lprocfs_lock with cli_sem held.
239 Frequency : only during format of test systems
240 Description: Unable to run several mkfs.lustre on loop devices at the same time
241 Details : mkfs.lustre returns error 256 on the concurrent loop devices
242 formatting. The solution is to proper handle the error.
244 Severity : enhancement
246 Description: implement async create (obd_async_create) method for osc, to avoid
247 too long waiting new ost objects with holding ldlm lock.
251 Frequency : occasionally during network problems
252 Description: client not allowed to reconnect to OST because of active request
253 Details : abort bulk requests received by the OST once the client has timed
254 out since the client will resend the request anyway. The client
255 also now retries to reconnect to the same server if a connect
256 request failed with EBUSY or -EAGAIN.
259 Frequency : rare, if used wide striped file and one ost in down.
261 Descriptoin: don't return error if we created a subset of objects for file.
262 Details : lov_update_create_set() uses set->set_success as index for created
263 objects, so if some requests failed, they will have hole at end of
264 array and we can use qos_shrink_lsm for allocate correct lsm.
268 Description: Slow stale export processing during normal start up
269 Details : The global mgc lock prevents OST setup to be run in parallel.
270 Replace the global lock with a per-config_llog_data semaphore.
274 Description: Out or order replies might be lost on replay
275 Details : In ptlrpc_retain_replayable_request if we cannot find retained
276 request with tid smaller then one currently being added, add it
277 to the start, not end of the list.
281 Description: BUG: soft lockup - CPU#1 stuck for 10s! [ll_mdt_07:4523]
282 Details : add cond_resched() calls to avoid hogging the cpu for too long
283 in the hash code. Make also lustre_hash_for_each_empty() more
286 Severity : enhancement
288 Description: Performance improvements for debug messages with D_RPCTRACE,
289 D_LDLM, D_QUOTA options.
292 Frequency : only with NFS export
294 Description: (lov_merge.c:74:lov_merge_lvb())
295 ASSERTION(spin_is_locked(&lsm->lsm_lock)) failed (SR 71691004)
296 Details : Fix a race in the nfs export code by populating inode
297 info while the new inode is still locked
299 Severity : enhancement
301 Description: add a new file in procfs called force_lbug. Writting to this
302 file triggers a LBUG. Only for test purpose.
306 Description: OOM killer causes node hang
307 Details : really interrupt the sleep in osc_enter_cache on signals
311 Description: LustreError: 9153:0:(quota_context.c:622:dqacq_completion()) LBUG
312 Details : fix race during quota release on the slave.
314 Severity : enhancement
316 Description: smaller hash bucket sizes, cleanups
317 Details : increase hash table sizes and enabled rehashing for pools, uuid,
320 Severity : enhancement
322 Description: Add ldiskfs maxdirsize mount option
323 Details : add max_dir size mount option
327 Description: panic in ll_statahead_thread
328 Details : prevent parent thread to be killed before its child
331 Frequency : only with 16TB device
333 Description: unable to perform "mount -t lustre" of 16TB OST device
334 Details : Mounting 16TB LUNs failed due to three bugs in mkfs.lustre.
338 Description: ASSERTION(atomic_read(&imp->imp_inflight) == 0) failed
339 Details : unregistering should be zero if no RPC inflight.
343 Description: hyperion: Oops during metabench
344 Details : Correct the refcount of lov_request_set
346 Severity : enhancement
348 Description: Add mptlinux and nxge drivers to Lustre builds
350 Severity : enhancement
352 Description: Fix watchdog timer message to be more clear
353 Details : Make watchdog timer messages more clear and descriptive.
357 Description: LNET soft lockups in socknal_cd thread
358 Details : don't hog CPU for active-connecting if another connd is
359 accepting connecting-requst from the same peer
363 Description: recovery-small test_17 hang
364 Details : Land several AT improvements & fixes.
368 Description: MDS panic and hanging client processes
369 Details : Replace exp_ops_stats with exp_nid_stats->nid_stats
373 Description: OSS stuck in recovery.
374 Details : fix race during recovery. class_unlink_export,
375 class_set_export_delayed and target_queue_last_replay_reply
376 may race while increasing/decreasing obd_recoverable_clients
377 and obd_delayed_clients, causing recovery to wait forever.
379 Severity : enhancement
381 Description: add cascading_rw.c to lustre/tests
385 Description: filter_last_id() NULL deref
386 Details : lprocfs_filter_rd_last_id() should check for the fully
387 setup obd device, before proceeding further.
389 Severity : enhancement
391 Description: Loadgen improvements
392 Details : stacksize and locking fixes for loadgen
396 Description: Quiet CERROR("dirty %d > system dirty_max %d\n"
397 Details : The atomic_read() allowing the atomic_inc() are not covered
398 by a lock. Thus they may safely race and trip this CERROR()
399 unless we add in a small fudge factor (+1).
401 Severity : enhancement
403 Description: shrink_slab: nr=-9223362083340912175
404 Details : fix spurious message from shrink_slab reporing negative nr
408 Description: Quiet bogus previously committed transno error
409 Details : suppress the "server went back in time" error message which
410 is always printed even in the common case after a client eviction
412 Severity : enhancement
414 Description: Parallel statfs() calls result in client eviction
415 Details : cache statfs data for 1s.
419 Description: parallel-scale test_compilebench: @@@@@@ FAIL: compilebench
421 Details : fix serveral issues in pinger code causing clients not to ping
422 servers for too long, resulting in evictions.
426 Description: e2fsck should warn when MMP update interval is extended
427 Details : print mmp_check_interval and make it possible to abort mount
428 operation in case it takes too long.
432 Description: mdsrate-create-large.sh, BUG: soft lockup - CPU#0 stuck for 10s!
433 Details : fix bug in the RHEL5's jbd2 callback patch.
437 Description: drop number of active requests when queued for recovery
438 Details : Now that we take a reference on the original request instead of
439 making a copy of it for recovery. We need to drop the number of
440 active requests or the queued requests will prevent all request
441 processing when they exceed (srv->srv_threads_running - 1).
443 Severity : enhancement
445 Description: refuse to invalidate operational quota files when they are in use
446 Details : an attempt to invalidate operational quota files on the quota
447 master is not actually permitted by VFS (returning -EPERM), but we
448 should not depend on that and should return the error earlier.
452 Description: Applications stuck in jbd2_log_wait_commit during exit
453 Details : fix deadlock between kjournald2 trying to acquire the page lock
454 owned by an ost_io thread waiting for journal commit.
456 -------------------------------------------------------------------------------
458 2009-10-16 Sun Microsystems, Inc.
460 * Support for kernels:
461 2.6.16.60-0.42.4 (SLES 10),
462 2.6.27.29-0.1 (SLES11, i686 & x84_64 only),
463 2.6.18-128.7.1.el5 (RHEL 5),
464 2.6.18-128.7.1.el5 (OEL 5),
465 * Client support for unpatched kernels:
466 (see http://wiki.lustre.org/index.php?title=Patchless_Client)
467 2.6.16 - 2.6.27 vanilla (kernel.org)
468 * Recommended e2fsprogs version: 1.41.6.sun1
469 * ext4 support for RHEL5 is experimental and thus should not be
472 Severity : enhancement
474 Description: Add OEL5 support.
476 Severity : enhancement
478 Description: Update kernel to SLES11 2.6.27.29-0.1.
482 Description: File checksum failures with OST read cache on
483 Details : Disable page poisoning when the bulk transfer has to be aborted
484 because the client got evicted.
488 Description: Don't allow make backward step on assiging osc next id.
489 Details : race between allocation next id and ll_sync thread can be cause
490 of set wrong osc next id and can be kill valid ost objects.
492 Severity : enhancement
494 Description: Update kernel to RHEL5 2.6.18-128.7.1.el5.
496 Severity : enhancement
498 Description: Update kernel to SLES10 SP2 2.6.16.60-0.42.4.
502 Description: Changes in raid5-large-io-rhel5.patch to calculate sectors properly
506 Description: Increase the default BLK_DEF_MAX_SECTORS value for RHEL5 and SLES11
510 Description: Error handling in osc_statfs_interpret() has been improved.
511 Details : Check in osc_statfs_interpret() for EBADR.
515 Description: Do not send statfs() requests to OSTs disabled by administrator.
516 Details : Check in lov_prep_statfs_set() for non-NULL ltd_exp.
520 Description: Do not update ctime for the deleted inode.
521 Details : Check in mds_reint_unlink() before calling fsfilt_setattr().
525 Description: Increase of the size of the LDLM resource hash.
526 Details : Bump up RES_HASH_BITS=12.
530 Description: correctly send lsm on open replay
531 Details : MDS is trust to LSM size on replay open, but client can set wrong size
536 Description: Deadlock between filter_destroy() and filter_commitrw_write().
537 Details : filter_destroy() does not hold the DLM lock over the whole
538 operation. If the DLM lock is dropped, filter_commitrw() can go
539 through, causing the deadlock between page lock and i_mutex.
540 The i_alloc_sem should also be hold in filter_destroy() while
545 Description: truncate starts GFP_FS allocation under transaction causing deadlock
546 Details : ldiskfs_truncate calls grab_cache_page which may start page
547 allocation under an open transaction. This may lead to
548 calling prune_icache with consequent lustre reentrance.
551 Frequency : only when down/upgrading the MDS to 1.6/1.8 while 1.8 clients are
552 still up and when the OST pool feature is used
554 Description: interop testing got LBUG when run dd with OST pool
555 :LustreError: 30032:0:(llite_lib.c:1913:ll_replace_lsm()) LBUG
556 Details : down/upgrading the MDS to a version that doesn't/does support OST
557 pool can cause clients to crash because the lsm has changed
562 Description: missing tree_status on 1.8.1 RPM build
563 Details : make rpms failed due because the tree_status file is missing.
567 Description: continuing LustreError "mds adjust qunit failed!"
568 Details : don't print message on the console when ->adjust_qunit fails.
572 Description: don't increase ldlm timeout if previous client was evicted
573 Details : if a client doesn't respond to a blocking callback within the
574 adaptive ldlm enqueue timeout, don't adjust the adaptive estimate
575 when the lock is next granted.
579 Description: ost is being unmounted w/o all writes to last_rcvd landing on disk.
580 affects recovery negatively.
581 Details : make sure all exports have been properly destroyed by the zombie
582 thread processed before stopping the target.
586 Description: Performance degradation with O_DIRECT between 1.6 & 1.8.1 b190
587 Details : disable write barrier for ext4/SLES11.
591 Description: Kernel panic - not syncing: Out of memory and no killable
592 processes... on OSS when iozone
593 Details : fix memory leak in the journal checksum patch.
597 Description: group quota "too many blocks" OSS crashes
598 Details : we should keep the same uid/gid for lquota_chkquota() and
599 lquota_pending_commit()
603 Description: LustreError: 9153:0:(quota_context.c:622:dqacq_completion()) LBUG
604 Details : don't LBUG on release quota error. Just a workaround until the
605 problem is understood.
607 ------------------------------------------------------------------------------
608 2009-07-31 Sun Microsystems, Inc.
610 * Support for kernels:
611 2.6.16.60-0.39.3 (SLES 10),
612 2.6.27.23-0.1 (SLES11, i686 & x84_64 only),
613 2.6.18-128.1.14.el5 (RHEL 5),
614 * Client support for unpatched kernels:
615 (see http://wiki.lustre.org/index.php?title=Patchless_Client)
616 2.6.16 - 2.6.27 vanilla (kernel.org)
617 * Recommended e2fsprogs version: 1.41.6.sun1
618 * File join has been disabled in this release, refer to Bugzilla 16929.
619 * NFS export disabled when stack size < 8192. Since the NFSv4 export
620 of Lustre filesystem with 4K stack may cause a stack overflow. For
621 more information, please refer to bugzilla 17630.
622 * ext4 support for RHEL5 is experimental and thus should not be
625 Severity : enhancement
627 Description: Update kernel to SLES10 SP2 2.6.16.60-0.39.3.
630 Frequency : with 1.8 server and 1.6 clients
632 Description: correctly shrink reply for avoid send too big message to client.
633 Details : 1.8 mds is allocate to big buffer to LOV EA data and this produce
634 some problems with sending this reply to 1.6 client.
638 Description: Repeated atomic allocation failures.
639 Details : Use GFP_HIGHUSER | __GFP_NOMEMALLOC flags for memory allocations
640 to generate memory pressure and allow reclaiming of inactive pages.
641 At the same time, do not allow to exhaust emergency pools.
642 For local clients the use of GFP_NOFS will be introduced in 1.8.2
644 Severity : enhancement
645 Bugzilla : 19846, 18289
646 Description: Update kernel to RHEL5 2.6.18-128.1.14.el5.
648 Severity : enhancement
649 Bugzilla : 19625, 16893, 18668, 19848
650 Description: Add support for SLES11 2.6.27.23-0.1.
652 Severity : enhancement
654 Description: Update client support to vanila kernels up to 2.6.27.
656 Severity : enhancement
658 Description: Update kernel to SLES10 SP2 2.6.16.60-0.37.
660 Severity : enhancement
662 Description: Compile with -Werror by default for i686 and x86_64.
666 Description: resolve race between obd_disconnect and class_disconnect_exports
667 Details : if obd_disconnect will be called to already disconnected export he
668 forget release one reference and osc module can't unloaded.
670 Severity : enhancement
672 Description: move AT tunable parameters for more consistent usage
673 Details : add AT tunables under /proc/sys/lustre, add to conf_param parsing
677 Descriptoin: correctly skip time estimate if in recovery
678 Details : rq_send_state insn't bitmask so using bitwise ops is forbid.
682 Descriptoin: OSS DeadLock
683 Details : Use trylock to prevent deadlock when shrink icache.
685 Severity : enhancement
687 Description: Allow tuning service thread via /proc
688 Details : For each service a new
689 /proc/fs/lustre/{service}/*/thread_{min,max,started} entry is
690 created that can be used to set min/max thread counts, and get the
691 current number of running threads.
693 Severity : enhancement
695 Description: Add state history info file, enhance import info file
696 Details : Track import connection state changes in a new osc/mdc proc file;
697 add overview-type data to the osc/mdc import proc file.
701 Description: Reduce small size read RPC
702 Details : Set read-ahead limite for every file and only do read-ahead when
703 available read-ahead pages are bigger than 1M to avoid small size
708 Description: free_entry erroneously used groups_free instead of put_group_info
710 Severity : enhancement
712 Description: Make read-ahead stripe size aligned.
714 Severity : enhancement
716 Description: MDS create should not wait for statfs RPC while holding DLM lock.
719 Frequency : rare, connect and disconnect target at same time
721 Descriptoin: ASSERTION(atomic_read(&imp->imp_inflight) == 0
722 Details : don't call obd_disconnect under lov_lock. this long time
723 operation and can block ptlrpcd which answer to connect request.
726 Frequency : start MDS on uncleanly shutdowned MDS device
728 Descriptoin: ll_sync thread stay in waiting mds<>ost recovery finished
729 Details : stay in waiting mds<>ost recovery finished produce random bugs
730 due race between two ll_sync thread for one lov target. send
731 ACTIVATE event only if connect realy finished and import have
735 Frequency : start MDS on uncleanly shutdowned MDS device
737 Descriptoin: aborting recovery hang on MDS
738 Details : don't throttle destroy RPCs for the MDT.
742 Description: Slow reads beyond 8Tb offsets.
743 Details : Page index integer overflow in ll_read_ahead_page
747 Description: Soft lockup on OSS after MDS failover
748 Details : MSG_CONNECT_INITIAL is not set on the initial MDS->OST connect.
749 As a conseqence, the patch from bug 18224 is not operational
750 and the MDS export cannot be reused on the OSTs until it gets
754 Frequency : rare, only if using MMP with Linux RAID
756 Description: MMP doesn't work with Linux RAID
757 Details : While using HA for Lustre servers with Linux RAID, it is possible
758 that MMP will not detect multiple mounts. To make this work we
759 need to unplug the device queue in RAID when the MMP block is being
760 written. Also while reading the MMP block, we should read it from
761 disk and not the cached one.
764 Frequency : rare, during recovery
766 Description: Assertion failure in ldlm_lock_put
767 Details : Do not put cancelled locks into replay list, hold references on
772 Description: 1.6.5 mdsrate performance is slower than 1.4.11/12 (MDS is not cpu bound!)
773 Details : create_count always drops to the min value (=32) because grow_count
774 is being changed before the precreate RPC completes.
777 Frequency : Only in RHEL5 when mounting multiple ext3 filesystems
780 Description: "kmem_cache_create: duplicate cache jbd_4k" error message
781 Details : add proper locking for creation of jbd_4k slab cache
785 Description: MMP check in ext3_remount() fails without displaying any error
786 Details : When multiple mount protection fails during remount, proper error
791 Description: Rare Client crash on resend if the file was deleted.
792 Details : When file is opened, but open reply is lost and file is
793 subsequently deleted before resend, resend processing logic
794 breaks trying to open the file again, should not try to open.
798 Description: add check for >8TB ldiskfs filesystems
799 Details : ext3-based ldiskfs does not support greater than 8TB LUNs.
800 Don't allow >8TB ldiskfs filesystems to be mounted without
801 force_over_8tb mount option
805 Description: Client locked up when running multiple instances of an app. on
806 multiple mount points
807 Details : ll_shrink_cache() can sleep while holding the ll_sb_lock.
808 Convert ll_sb_lock to a read/write semaphore to fix the problem.
812 Description: Cannot acces an NFS-mounted Lustre filesystem
813 Details : An NFS client cannot access the Lustre filesystem NFS-mounted
814 from a Lustre-client exporting the Lustre filesystem via NFS.
818 Description: panic in ll_statahead_thread
819 Details : grab dentry reference in parent process.
822 -------------------------------------------------------------------------------
824 tbd Sun Microsystems, Inc.
826 * Support for kernels:
827 2.6.16.60-0.31 (SLES 10),
828 2.6.18-128.1.6.el5 (RHEL 5),
829 2.6.22.14 vanilla (kernel.org)
830 * Client support for unpatched kernels:
831 (see http://wiki.lustre.org/index.php?title=Patchless_Client)
832 2.6.16 - 2.6.22 vanilla (kernel.org)
833 * Recommended e2fsprogs version: 1.40.11-sun1
834 * File join has been disabled, refer to Bugzilla 16929.
835 * A new Lustre ADIO driver is available for MPICH2-1.0.7.
836 * NFS export disabled when stack size < 8192. Since the NFSv4 export of
837 Lustre filesystem with 4K stack may cause a stack overflow. For more
838 information, please refer to bugzilla 17630.
840 Severity : enhancement
842 Description: Update to RHEL5.3 kernel-2.6.18-128.1.6.el5.
844 Severity : enhancement
846 Description: Update OFED release to 1.4.1 RC4
848 Severity : major, only with big OST
850 Description: Very poor metadata performance on Infiniband lustre configuration
851 Details : OST object precreation becomes very slow on big OSTs. This is due
852 to the ialloc patch spending too much time scanning groups.
855 Frequency : during recovery
857 Description: don't mix llog inodes with normal.
858 Details : allocate inodes for log in last inode group
863 Description: fix lqs' reference which won't be put in some situations
864 Details : This patch fixes:
865 1. In quota_check_common(), this function will check quota
866 for user and group, but only send one return via "pending".
867 In most cases, the pendings should be same. But that is not
869 2. If quotaoff runs between lquota_chkquota() and
870 lquota_pending_commit(), the same thing will happen too.
871 That is why it comes:
872 - if (!ll_sb_any_quota_active(qctxt->lqc_sb))
875 -------------------------------------------------------------------------------
877 2008-12-31 Sun Microsystems, Inc.
879 * Support for kernels:
880 2.6.5-7.314 (SLES 9),
881 2.6.9-67.0.22.EL (RHEL 4),
882 2.6.16.60-0.31 (SLES 10),
883 2.6.18-92.1.17.el5 (RHEL 5),
884 2.6.22.14 vanilla (kernel.org)
885 * Client support for unpatched kernels:
886 (see http://wiki.lustre.org/index.php?title=Patchless_Client)
887 2.6.16 - 2.6.22 vanilla (kernel.org)
888 * Client support for unpatched kernels:
889 we do not recommend using patchless RHEL4 clients with kernels
890 prior to 2.6.9-55EL (RHEL4U5).
891 * Recommended e2fsprogs version: 1.40.11-sun1
892 * Note that reiserfs quotas are disabled on SLES 10 in this kernel.
893 * RHEL 4 and RHEL 5/SLES 10 clients behaves differently on 'cd' to a
894 removed cwd "./" (refer to Bugzilla 14399).
895 * A new quota file format has been introduced in 1.6.5.
896 The format conversion from prior releases is handled transparently,
897 but releases older than 1.4.12/1.6.5 don't understand this new
898 format. The automatic format conversion can be avoided by running
899 the following command on the MDS:
900 'tunefs.lustre --param="mdt.quota_type=ug1" $MDTDEV'.
901 For more information, please refer to bugzilla 13904.
902 * A new quota file format was introduced in 1.6.6/1.8.0 (kernels 2.6.16+).
903 The format conversion from prior releases is handled transparently,
904 but releases older than 1.6.6/1.8.0 don't understand this new
905 format. The automatic format conversion can be avoided by running
906 the following commands on the MDS and OSS servers (for
907 pre 1.4.12-1.6.5 quota files):
908 'tunefs.lustre --param="mdt.quota_type=ug1" $MDTDEV',
909 'tunefs.lustre --param="ost.quota_type=ug1" $MDTDEV'
910 or (for 1.4.12/1.6.5 quota files)
911 'tunefs.lustre --param="mdt.quota_type=ug2" $MDTDEV',
912 'tunefs.lustre --param="ost.quota_type=ug2" $MDTDEV'
913 For more information, please refer to bugzilla 13904.
914 * Output of lfs quota has been made less detailed by default,
915 old (verbose) output can be obtained by using -v option.
916 * File join has been disabled in this release, refer to Bugzilla 16929.
917 * A new Lustre ADIO driver is available for MPICH2-1.0.7.
918 * NFS export disabled when stack size < 8192. Since the NFSv4 export of
919 Lustre filesystem with 4K stack may cause a stack overflow. For more
920 information, please refer to bugzilla 17630.
922 Severity : enhancement
924 Description: Caching OSS
925 Details : introduce data caching on the OSS. The OSS now relies on the linux
926 kernel page cache to keep recently accessed data in memory.
927 It is worth noting that all write requests are still flushed
928 synchronously as in lustre 1.6.
930 Severity : enhancement
932 Description: Version based recovery
933 Details : introduce finer grained recovery able to detect transaction
934 dependencies and can deal with transaction gaps caused by clients
935 failing at the same time as the server.
937 Severity : enhancement
939 Description: Enable adaptive timeouts by default
940 Details : The Lustre timeout value in /proc/sys/lustre/timeout is now
941 managed dynamically based on server load and should not need
942 to be tuned manually based on cluster size. This allows Lustre
943 to work under a wider variety of system sizes and loads, without
944 unnecessarily causing lengthy recovery times.
946 Severity : enhancement
948 Description: Add OST Pools support
949 Details : File striping can now be set to use an arbitrary pool of OSTs
951 Severity : enhancement
953 Description: add lazystatfs mount option to allow statfs(2) to skip down OSTs
954 Details : allow skip disconnected ost for send statfs request and hide error
958 Frequency : rare, on llog test 6
960 Descriptoin: don't allow connect to already connected import
961 Details : allowing connect to already connected import is hide connecting problem.
964 Frequency : rare, on failed llog setup
966 Descriptoin: don't leak obd reference on failed llog setup
967 Details : for failed llog setup - mgc forget call class_destroy_import for
968 client import, move destroy import to more generic place.
973 Descriptoin: allow kill process which wait statahead result
974 Details : for some reasons 'ls' can stick in waiting result from statahead,
975 in this case need way for kill this process.
978 Frequency : rare, at shutdown
980 Descriptoin: panic at umount
981 Details : llap_shrinker can be raced with killing super block from list and
982 this produce panic with access to already freeded pointer
987 Descriptoin: don't lose wakeup for imp_recovery_waitq
988 Details : recover_import_no_retry or invalidate_import and import_close can
989 both sleep on imp_recovery_waitq, but we was send only one wakeup
995 Descriptoin: panic in mds_open
996 Details : don't confuse mds_finish_transno() with PTR_ERR(-ENOENT)
1001 Descriptoin: stuck in cache_remove_extent() or panic with accessing to already
1003 Details : release lock refernce only after add page to pages list.
1006 Frequency : always with long access acl
1008 Descriptoin: mds can't pack reply with long acl.
1009 Details : mds don't control size of acl but they limited by reint/getattr
1014 Frequency : on remount
1016 Description: external journal device not working after the remount
1017 Details : clear dev_rdonly flag for external journal devices in
1023 Description: shutdown vs evict race
1024 Details : client_disconnect_export vs connect request race.
1025 if client will evicted at this time - we start invalidate
1026 thread without referece to import and import can be freed
1032 Description: shrink LOV EAs before replying
1033 Details : correctly adjust LOV EA buffer for reply.
1038 Description: don't skip ost target if they assigned to file
1039 Details : Drop slow OSCs if we can, but not for requested start idx.
1040 This means "if OSC is slow and it is not the requested
1041 start OST, then it can be skipped, otherwise skip it only
1042 if it is inactive/recovering/out-of-space.
1044 Severity : enhancement
1046 Description: Update to RHEL5 kernel-2.6.18-92.1.17.el5.
1049 Frequency : rare, need acl's on inode.
1051 Description: client can't handle ost additional correctly
1052 Details : if ost was added after client connected to mds client can have
1053 hit lnet_try_match_md ... to big messages to wide striped files.
1054 in this case need teach client to handle config events about add
1055 lov target and update client max ea size at that event.
1057 Severity : enhancement
1059 Description: Update to sles9 kernel-2.6.5-7.314.
1061 Severity : enhancement
1063 Description: Update to SLES10 SP2 kernel-2.6.16.60-0.31.
1066 Frequency : Create a symlink file with a very long name
1068 Description: ldlm_cancel_pack()) ASSERTION(max >= dlm->lock_count + count)
1069 Details : If there is no extra space in the request for early cancels,
1070 ldlm_req_handles_avail() returns 0 instead of a negative value.
1075 Description: mds is deadlocked
1076 Details : in rare cases, inode in catalog can have i_no less than have parent
1077 i_no, this produce wrong order for locking during open, and
1078 parallel unlink can be lock open. this need teach mds_open to grab
1079 locks in resource id order, not at parent -> child order.
1081 Severity : enhancement
1083 Description: Add /proc entry for import status
1084 Details : The mdc, osc, and mgc import directories now have
1085 an import directory that contains useful import data for debugging
1086 connection problems.
1088 Severity : enhancement
1090 Description: Re-disable certain /proc logging
1091 Details : Enable and disable client's offset_stats, extents_stats and
1092 extents_stats_per_process stats logging on the fly.
1095 Frequency : Only on FC kernels 2.6.22+
1097 Description: oops in statahead
1098 Details : Do not drop reference count for the dentry from VFS when lookup,
1099 VFS will do that by itself.
1101 Severity : enhancement
1103 Description: Generic /proc file permissions
1104 Details : Set /Proc file permissions in a more generic way to enable non-
1105 root users operate on some /proc files.
1109 Description: Hitting mdc_commit_close() ASSERTION
1110 Details : Properly handle request reference release in
1111 ll_release_openhandle().
1115 Frequency : only patchless client
1116 Description: add workaround for race between add/remove dentry from hash
1118 Severity : enhancement
1120 Description: Allow OST glimpses to return PW locks
1124 Description: LBUG when llog conf file is full
1125 Details : When llog bitmap is full, ENOSPC should be returned for plain
1130 Description: Prevent import from entering FULL state when server in recovery
1134 Description: service mount cannot take device name with ":"
1135 Details : Only when device name contains ":/" will mount treat it as
1141 Description: replace ptlrpcd with the statahead thread to interpret the async
1142 statahead RPC callback
1146 Frequency : on recovery
1147 Description: I/O failures after umount during fail back
1148 Details : if client reconnected to restarted server we need join to recovery
1149 instead of find server handler is changed and process self eviction
1150 with cancel all locks.
1152 Severity : enhancement
1154 Description: Update to RHEL5 kernel-2.6.18-92.1.10.el5.
1158 Description: Kernel BUG tries to release flock
1159 Details : Lustre does not destroy flock lock before last reference goes
1160 away. So always drop flock locks when client is evicted and
1161 perform unlock regardless of successfulness of speaking to MDS.
1163 Severity : enhancement
1165 Description: Update to SLES10 SP2 kernel-2.6.16.60-0.27.
1167 Severity : enhancement
1169 Description: Upcall on Lustre log has been dumped
1170 Details : Allow for a user mode script to be called once a Lustre log has
1171 been dumped. It passes the filename of the dumped log to the
1172 script, the location of the script can be specified via
1173 /proc/sys/lnet/debug_log_upcall.
1178 Description: avoid messages about idr_remove called for id that is not allocated
1179 Details : Move assigment s_dev for clustered nfs to end of initialization,
1180 for avoid problem with error handling.
1185 Description: avoid Already found the key in hash [CONN_UNUSED_HASH] messages
1186 Details : When connection is reused this not moved from CONN_UNUSED_HASH into
1187 CONN_USED_HASH and this prodice warning when put connection again
1193 Description: avoid ASSERTION(client_stat->nid_exp_ref_count == 0) failed
1194 Details : release reference to stats when client disconnected, not
1195 when export destroyed for avoid races when client destroyed
1196 after main ost export.
1200 Description: more cleanup in mds_lov
1201 Details : add workaround for get valid ost count for avoid warnings about
1202 drop too big messages, not init llog cat under semphore which
1203 can be blocked on reconnect and break normal replay, fix access
1206 Severity : enhancement
1208 Description: Export bytes_read/bytes_write count on OSC/OST.
1212 Description: Early reply size mismatch, MGC loses connection
1213 Details : Apply the MGS_CONNECT_SUPPORTED mask at reconnect time so
1214 the connect flags are properly negotiated.
1218 Description: Properly propagate oinfo flags from lov to osc for statfs
1219 Details : restore missing copy oi_flags to lov requests.
1223 Description: exports in /proc are broken
1224 Details : recreate /proc entries for clients when they reconnect.
1226 Severity : enhancement
1228 Description: Add man pages for llobdstat(8), llstat(8), plot-llstat(8),
1229 : l_getgroups(8), lst(8), routerstat(8)
1230 Details : included man pages for llobdstat(8), llstat(8),
1231 : plot-llstat(8), l_getgroups(8), lst(8), routerstat(8)
1233 Severity : enhancement
1235 Description: Implement lustre ll_show_options method.
1237 Severity : enhancement
1239 Description: Update to SLES9 kernel-2.6.5-7.312.
1241 Severity : enhancement
1243 Description: Update to RHEL4 kernel-2.6.9-67.0.22.EL.
1247 Description: exports in /proc are broken
1248 Details : recreate /proc entries for clients when they reconnect.
1252 Description: don't fail open with -ERANGE
1253 Details : if client connected until mds will be know about real ost count
1254 get LOV EA can be fail because mds not allocate enougth buffer
1259 Description: Resolve device initialization race
1260 Details : Prevent proc handler from accessing devices added to the
1261 obd_devs array but yet be intialized.
1263 Severity : enhancement
1265 Description: configure's --enable-quota should check the
1266 : kernel .config for CONFIG_QUOTA
1267 Details : configure is terminated if --enable-quota is passed but
1268 : no quota support is in kernel
1270 Severity : enhancement
1272 Description: Update to SLES10 SP2 kernel-2.6.16.60-0.23.
1274 Severity : enhancement
1276 Description: Update to RHEL5 kernel-2.6.18-92.1.6.el5.
1280 Frequency : rare, on PPC clients
1281 Description: don't swab ost objects in response about directory, because
1283 Details : bug similar bug 14856, but in different function.
1285 Severity : enhancement
1287 Description: lfs quota tool enhancement
1288 Details : added units specifiers support for setquota, default to
1289 current uid/gid for quota report, short quota stats by
1290 default, nonpositional parameters for setquota, added
1291 llapi_quotactl manual page.
1293 Severity : enhancement
1295 Description: *optional* service tags registration
1296 Details : if the "service tags" package is installed on a Lustre node
1297 When the filesystem is mounted, a local-node service tag will
1298 be created. See http://inventory.sun.com/ for more information
1299 about the Service Tags asset management system.
1303 Description: Client runs out of low memory
1304 Details : Consider only lowmem when counting initial number of llap pages
1307 Frequency : occasional
1309 Description: add refcount for osc callbacks, so avoid panic on shutdown
1311 Severity : enhancement
1313 Description: Update to RHEL4 kernel-2.6.9-67.0.20.
1316 Frequency : testing only
1318 Description: sanity test 65a fails if stripecount of -1 is set
1319 Details : handle -1 striping on filesystem in ll_dirstripe_verify
1322 Frequency : only in unusual configurations
1324 Description: Kernel panic with find ost index.
1325 Details : lov_obd have panic if some OST's have sparse indexes.
1327 Severity : enhancement
1329 Description: Update to RHEL5 kernel-2.6.18-53.1.21.el5.
1332 Frequency : rarely, if filesystem is mounted with -o flock
1334 Description: do not process already freed flock
1335 Details : flock can possibly be freed by another thread before it reaches
1336 to ldlm_flock_completion_ast.
1339 Frequency : rarely, if filesystem is mounted with -o flock
1341 Description: LBUG during stress test
1342 Details : Need properly lock accesses the flock deadlock detection list.
1345 Frequency : rarely, if binaries are being run from Lustre
1347 Description: oops in page fault handler
1348 Details : kernel page fault handler can return two special 'pages' in
1349 error case, don't try dereference NOPAGE_SIGBUS and NOPAGE_OMM.
1352 Frequency : rarely, during shutdown
1354 Description: timeout with invalidate import.
1355 Details : ptlrpcd_check call obd_zombie_impexp_cull and wait request which
1356 should be handled by ptlrpcd. This produce long age waiting and
1357 -ETIMEOUT ptlrpc_invalidate_import and as result LASSERT.
1363 Description: ASSERTION(CheckWriteback(page,cmd)) failed
1364 Details : badly clear PG_Writeback bit in ll_ap_completion can produce false
1368 Frequency : only with broken builds/installations
1370 Description: no LBUG if lquota.ko and fsfilt_ldiskfs.ko are different versions
1371 Details : just return an error to a user, put a console error message
1373 Severity : enhancement
1375 Description: Update to RHEL5 kernel-2.6.18-53.1.19.el5.
1377 Severity : enhancement
1379 Description: Update to RHEL4 kernel-2.6.9-67.0.15.
1381 Severity : enhancement
1383 Description: enable MGS and MDT services start separately
1384 Details : add a 'nomgs' option in mount.lustre to enable start a MDT with
1385 a co-located MGS without starting the MGS, which is a complement
1386 to 'nosvc' mount option.
1389 Frequency : always, on big-endian systems
1391 Description: cleanup in ptlrpc code, related to PPC platform
1392 Details : store magic in native order avoid panic's in recovery on PPC
1393 node and forbid from this error in future. Also fix posibily
1394 of twice swab data. Fix get lov striping to userpace.
1397 Frequency : rarely, if replay get lost on server
1399 Description: server incorrectly drop resent replays lead to recovery failure.
1400 Details : do not drop replay according to msg flags, instead we check the
1401 per-export recovery request queue for duplication of transno.
1404 Frequency : after recovery
1406 Description: precreate to many object's after del orphan.
1407 Details : del orphan st in oscc last_id == next_id and this triger growing
1408 count of precreated objects. Set flag LOW to skip increase count
1409 of precreated objects.
1412 Frequency : rare, on clear nid stats
1414 Description: ASSERTION(client_stat->nid_exp_ref_count == 0)
1415 Details : when clean nid stats sometimes try destroy live entry,
1416 and this produce panic in free.
1419 Frequency : occasionally since 1.6.4
1421 Description: Stack overflow during MDS log replay
1422 Details : ease stack pressure by using a thread dealing llog_process.
1425 Frequency : very rare
1427 Description: MDT cannot be unmounted, reporting "Mount still busy"
1428 Details : Mountpoint references were being leaked during open reply
1429 reconstruction after an MDS restart. Drop mountpoint reference
1430 in reconstruct_open() and free dentry reference also.
1435 Description: wait until IO finished before start new when do lock cancel.
1436 Details : VM protocol want old IO finished before start new, in this case
1437 need wait until PG_writeback is cleared until check dirty flag
1438 and call writepages in lock cancel callback.
1443 Description: mds_mfd_close() ASSERTION(rc == 0)
1444 Details : In mds_mfd_close(), we need protect inode's writecount change
1445 within its orphan write semaphore to prevent possible races.
1448 Frequency : rare, on shutdown ost
1450 Description: don't hit live lock with umount ost.
1451 Details : shrink_dcache_parent can be in long loop with destroy dentries,
1452 use shrink_dcache_sb instead.
1455 Frequency : only when echo_client is used
1457 Description: don't panic with use echo_client
1458 Details : echo client pass NULL as client nid pointer and this produce NULL
1459 pointer dereference.
1462 Frequency : Always on 32-bit PowerPC systems
1464 Description: fix build on PPC32
1465 Details : compile code with -m64 flag produce wrong object file for PPC32.
1470 Description: MDS LBUG: ASSERTION(!IS_ERR(dchild))
1471 Details : In reconstruct_* functions, LASSERTs on both the data supplied
1472 by a client, and the data on disk are dangerous and incorrect.
1473 Change them with client eviction.
1475 Severity : enhancement
1477 Description: skiplist implementation simplification
1478 Details : skiplists are used to group compatible locks on granted list
1479 that was implemented as tracking first and last lock of each lock
1480 group the patch changes that to using doubly linked lists
1484 Description: delete compatibility for 32bit qdata
1485 Details : as planned, when lustre is beyond b1_8, lquota won't support 32bit
1486 qunit. That means servers of b1_4 and servers of b1_8 can't be
1487 used together if users want to use quota.
1490 Frequency : only with administrator action
1492 Description: mount failure if config log has invalid conf_param setting
1493 Details : If administrator specified an incorrect configuration parameter
1494 with "lctl conf_param" this would cause an error during future
1495 client mounts. Instead, ignore the bad configuration parameter.
1498 Frequency : blocks per group < blocksize*8 and uninit_groups is enabled
1500 Description: ldiskfs error: XXX blocks in bitmap, YYY in gd
1501 Details : If blocks per group is less than blocksize*8, set rest of the
1505 Frequency : Application do stride read on lustre
1507 Description: The read performance will drop a lot if the application does
1509 Details : Because the stride_start_offset are missing in stride read-ahead,
1510 it will cause clients read a lot of unused pages in read-ahead,
1511 then the read-performance drops.
1515 Description: more ldlm soft lockups
1516 Details : In ldlm_resource_add_lock(), call to ldlm_resource_dump()
1517 starve other threads from the resource lock for a long time in
1518 case of long waiting queue, so change the debug level from
1519 D_OTHER to the less frequently used D_INFO.
1521 Severity : enhancement
1523 Description: add -gid, -group, -uid, -user options to lfs find
1525 Severity : enhancement
1527 Description: ll_recover_lost_found_objs - recover objects in lost+found
1528 Details : OST corruption and subsequent e2fsck can leave objects in the
1529 lost+found directory. Using the "ll_recover_lost_found_objs"
1530 tool, these objects can be retrieved and data can be salvaged
1531 by using the object ID saved in the fid EA on each object.
1536 Description: this bug _only_ happens when inode quota limitation is very low
1537 (less than 12), so that inode quota unit is 1 at initialization.
1538 Details : if remaining quota equates 1, it is a sign to demonstate that quota
1539 is effective now. So least quota qunit should be 2.
1543 Description: Hung threads in invalidate_inode_pages2_range
1544 Details : The direct IO path doesn't call check_rpcs to submit a new RPC once
1545 one is completed. As a result, some RPCs are stuck in the queue
1550 Description: Procfs and llog threads access destoryed import sometimes.
1551 Details : Sync the import destoryed process with procfs and llog threads by
1552 the import refcount and semaphore.
1556 Description: mds fails to respond, threads stuck in ldlm_completion_ast
1557 Details : Sort source/child resource pair after updating child resource.
1562 Description: kernel BUG at ldiskfs2_ext_new_extent_cb
1563 Details : If insertion of an extent fails, then discard the inode
1564 preallocation and free data blocks else it can lead to duplicate
1569 Description: don't always update ctime in ext3_xattr_set_handle()
1570 Details : Current xattr code updates inode ctime in ext3_xattr_set_handle()
1571 In some cases the ctime should not be updated, for example for
1572 2.0->1.8 compatibility it is necessary to delete an xattr and it
1573 should not update the ctime.
1577 Description: add quota statistics
1578 Details : 1. sort out quota proc entries and proc code.
1579 2. add quota statistics
1584 Description: quotas are not honored with O_DIRECT
1585 Details : all writes with the flag O_DIRECT will use grants which leads to
1586 this problem. Now using OBD_BRW_SYNC to guard this.
1590 Bugzilla : 15713/16362
1591 Description: Assertion in iopen_connect_dentry in 1.6.3
1592 Details : looking up an inode via iopen with the wrong generation number can
1593 populate the dcache with a disconneced dentry while the inode
1594 number is in the process of being reallocated. This causes an
1595 assertion failure in iopen since the inode's dentry list contains
1596 both a connected and disconnected dentry.
1600 Description: assertion failure in ldlm_handle2lock()
1601 Details : fix a race between class_handle_unhash() and class_handle2object()
1602 introduced in lustre 1.6.5 by bug 13622.
1604 Severity : enhancement
1606 Description: superblock lock contention with many SMP cores on one client
1607 Details : several client filesystem locks were highly contended on SMP
1608 NUMA systems with 8 or more cores. Per-CPU datastructures
1609 and more efficient locking implemented to reduce contention.
1614 Description: Kernel BUG: sd_iostats_bump: unexpected disk index
1615 Details : remove the limit of 256 scsi disks in the sd_iostat patch
1620 Description: oops in sd_iostats_seq_show()
1621 Details : unloading/reloading the scsi low level driver triggers a kernel
1622 bug when trying to access the sd iostat file.
1627 Description: Kernel panics during QLogic driver reload
1628 Details : REQ_BLOCK_PC requests are not handled properly in the sd iostat
1629 patch, causing memory corruption.
1634 Description: journal_dev option does not work in b1_6
1635 Details : pass mount option during pre-mount.
1637 Severity : enhancement
1639 Description: Add a FIEMAP(FIle Extent MAP) ioctl for ldiskfs
1640 Details : FIEMAP ioctl will allow an application to efficiently fetch the
1641 extent information of a file. It can be used to map logical blocks
1642 in a file to physical blocks in the block device.
1645 Frequency : only with adaptive timeout enabled
1647 Description: DEBUG_REQ() bad paging request
1648 Details : ptlrpc_at_recv_early_reply() should not modify req->rq_repmsg
1649 because it can be accessed by reply_in_callback() without the
1653 Frequency : only on Cray X2
1655 Description: X2 build failures
1656 Details : fix build failures on Cray X2.
1660 Description: xid & resent requests
1661 Details : Initialize RPC XID from clock at startup (randomly if clock is
1666 Description: quota recovery deadlock during mds failover
1667 Details : This patch includes att18982, att18236, att18237 in bz14840.
1669 1. fix osts hang when mds does failover with quotaon
1670 2. prevent watchdog storm when osts threads wait for the
1675 Description: kernel panic on racer
1676 Details : Do not access dchild->d_inode when IS_ERR(dchild) is true.
1678 Severity : enhancement
1680 Description: Add lustre_start utility to start or stop multiple Lustre servers
1685 Description: Lustre GPF in {:ptlrpc:ptlrpc_server_free_request+373}
1686 Details : In case of memory pressure, list_del() can be called twice on
1687 req->rq_history_list, causing a kernel oops.
1691 Description: (ptllnd_peer.c:557:kptllnd_peer_check_sends()) ASSERTION(!in_interrupt()) failed
1692 Details : fix stack overflow in the distributed lock manager by defering
1693 export eviction after a failed ast to the elt thread instead of
1694 handling it in the dlm interpret routine.
1696 Severity : enhancement
1698 Description: More exported tunables for mballoc
1699 Details : Add support for tunable preallocation window and new tunables for
1700 large/small requests
1704 Description: Detect corruption of block bitmap and checking for preallocations
1705 Details : Checks validity of on-disk block bitmap. Also it does better
1706 checking of number of applied preallocations. When corruption is
1707 found, it turns filesystem readonly to prevent further corruptions.
1711 Frequency : only for big-endian servers
1712 Description: Check for big-endian system while mounting fs with extents feature
1713 Details : Mounting a filesystem with extents feature will fail on big-endian
1714 systems since ext3-based ldiskfs is not supported on big-endian
1715 systems. Can be over-ridden with "bigendian_extents" mount option.
1719 Description: Excessive recovery window
1720 Details : With AT enabled, the recovery window can be excessively long (6000+
1721 seconds). To address this problem, we no longer use
1722 OBD_RECOVERY_FACTOR when extending the recovery window (the connect
1723 timeout no longer depends on the service time, it is set to
1724 INITIAL_CONNECT_TIMEOUT now) and clients report the old service
1725 time via pb_service_time.
1728 Descriptoin: Don't sync journal after every i/o
1729 Details : Implement write RPC replay to allow server replies for write RPCs
1730 before data is on disk. However, this feature is disabled by
1731 default since some issues leading to data corruptions have been
1732 found during recovery (e.g. bug 19128). This feature can be enabled
1733 by running the following command on the OSSs:
1734 lctl set_param obdfilter.*.sync_journal=0
1738 Description: Watchdog triggered on MDS failover
1739 Details : enable OBD_CONNECT_MDT flag when connecting from the MDS so that
1740 the OSTs know that the MDS "UUID" can be reused for the same export
1741 from a different NID, so we do not need to wait for the export to
1746 Descriptoin: Lustre detected file system corruption with inode out of bounds
1747 Details : don't update i_size on MDS_CLOSE for directories. This causes
1748 directory corruptions on the MDT.
1750 -------------------------------------------------------------------------------
1752 2008-05-26 Sun Microsystems, Inc.
1754 * Support for kernels:
1755 2.6.5-7.311 (SLES 9),
1756 2.6.9-67.0.7.EL (RHEL 4),
1757 2.6.16.54-0.2.5 (SLES 10),
1758 2.6.18-53.1.14.el5 (RHEL 5),
1759 2.6.22.14 vanilla (kernel.org)
1760 * Client support for unpatched kernels:
1761 (see http://wiki.lustre.org/index.php?title=Patchless_Client)
1762 2.6.16 - 2.6.22 vanilla (kernel.org)
1763 * Due to problems with nested symlinks and FMODE_EXEC (bug 12652),
1764 we do not recommend using patchless RHEL4 clients with kernels
1765 prior to 2.6.9-55EL (RHEL4U5).
1766 * Recommended e2fsprogs version: 1.40.7-sun1
1767 * Note that reiserfs quotas are disabled on SLES 10 in this kernel.
1768 * RHEL 4 and RHEL 5/SLES 10 clients behaves differently on 'cd' to a
1769 removed cwd "./" (refer to Bugzilla 14399).
1770 * A new quota file format has been introduced in 1.6.5.
1771 The format conversion from prior releases is handled transparently,
1772 but releases older than 1.4.12/1.6.5 will not understand this new
1773 format. The automatic format conversion can be avoided by running
1774 the following command on the MDS before upgrading:
1775 'tunefs.lustre --param="mdt.quota_type=ug1" $MDTDEV'.
1776 For more information, please refer to bugzilla 13904.
1780 Description: quota performance fix
1781 Details : quota data is written in journalled mode instead of ordered to
1782 increase performance
1786 Description: lfs support for human-readable quota grace time strings
1787 Details : lfs setquota -t and lfs quota -t represent quota grace times
1788 in "XXwXXdXXhXXmXXs" format instead of large values in seconds
1791 Frequency : always with o2ib 1.3 and sles10
1793 Description: fix build with SLES10 and o2ib v3.
1794 Details : sles10 uses diffrent name for Module.symver file but configure
1795 assume this file has same name on RHEL/SLES/vanila kernels.
1798 Frequency : very rare, if additional xattrs are used on kernels >= 2.6.12
1800 Description: MDS may lose file striping (and hence file data) in some cases
1801 Details : If there are additional extended attributes stored on the MDS,
1802 in particular ACLs, SELinux, or user attributes (if user_xattr
1803 is specified for the client mount options) then there is a risk
1804 of attribute loss. Additionally, the Lustre file striping
1805 needs to be larger than default (e.g. striped over all OSTs),
1806 and an additional attribute must be stored initially in the
1807 inode and then increase in size enough to be moved to the
1808 external attribute block (e.g. ACL growing in size) for file
1811 Severity : enhancement
1813 Description: add message levels for liblustreapi
1818 Description: MDT cannot be unmounted, reporting "Mount still busy"
1819 Details : Mountpoint references were being leaked during open reply
1820 reconstruction after an MDS restart. Drop mountpoint reference
1821 in reconstruct_open() and free dentry reference also.
1826 Description: fix for occasional failure case of -ENOSPC in recovery-small tests
1827 Details : Move the 'good_osts' check before the 'total_bavail' check. This
1828 will result in an -EAGAIN and in the exit call path we call
1829 alloc_rr() which will with increasing aggressiveness attempt to
1830 aquire precreated objects on the minimum number of required OSCs.
1834 Description: Use old size assignment to avoid deadlock
1835 Details : This reverts the changes in bugs 2369 and bug 14138 that introduced
1836 the scheduling while holding a spinlock. We do not need locking
1837 for size in ll_update_inode() because size is only updated from
1838 the MDS for directories or files without objects, so there is no
1839 other place to do the update, and concurrent access to such inodes
1840 are protected by the inode lock.
1844 Description: Use __u64 instead of int for valid bits
1848 Description: resolve "_IOWR redefined" build error on SLES10
1852 Description: dump the memory debugging after all modules are unloaded to
1853 suppress false negative in conf_sanity test 39
1857 Description: the recovery timer never expires
1858 Details : for new client connect request, the recovery timer should not be
1859 reset, otherwise recovery timer will never expired, if the old
1860 client never come. Only old client connect and first connection
1861 req should trigger recovery timer reset.
1865 Description: the min numbers of lproc stats are wrong
1866 Details : adding a new constant LC_MIN_INIT and use it for initialization
1870 Frequency : always with interactive lfs
1872 Description: Reinitialize optind to 0 so that interactive lfs works in all cases
1875 Frequency : with multiple concurrent readdir processes in same directory
1876 Bugzilla : 15406, 15169, 15175
1877 Description: misc fixes for directory readahead.
1878 Details : prevent previous statahead async RPC callback from processing the
1879 current "statahead_info", race condition between async RPC callback
1880 add dentry into dentry hash table and "ls" thread revalidate such
1881 dentry, statahead his/miss control for hidden items, and so on.
1883 Severity : enhancement
1885 Description: build kernel-ib packages for OFED 1.3 in our release cycle
1889 Description: incore types cleaning in quota code (with respect to 64-bit limits)
1890 Details : several u32 variables declarations are replaced with u64 declarations
1895 Description: fix SLES kernel versioning
1896 Details : the kernel version for our SLES 10 kernel did not include a "-"
1897 before the "smp" at the end. while this was not a problem in
1898 general, it did mean that software trying to use the kernel
1899 version to try to detect a vendor specific kernel would fail.
1900 this was most evident by the OFED build scripts.
1905 Description: Don't update lov_desc members until making sure they are valid
1906 Details : When updating lov_desc members via proc fs, need fix their
1907 validities before doing the real update.
1910 Frequency : very rare
1912 Description: don't put request into delay list while invalidate in flight.
1913 Details : ptlrpc_delay_request sometimes put in delay list while invalidate
1914 import in flight. this produce timeout for invalidate and sometimes
1915 can cause stale data.
1917 Severity : enhancement
1919 Description: Update kernel to SLES9 2.6.5-7.311.
1921 Severity : enhancement
1923 Description: Update kernel to RHEL4 2.6.9-67.0.7.
1928 Frequency : on PPC only
1929 Description: not convert ost objects for directory because it's not exist.
1930 Details : ll_dir_getstripe assume dirrectory has ost objects but this wrong.
1932 Severity : enhancement
1934 Description: Fix warnings with compile liblustre at sles10/rhel5 which have
1935 __u64 as usingied long long type.
1938 Frequency : rare, on shutdown
1940 Description: race process ast vs remove callback
1941 Details : removing callback before disconnect import open race with
1942 processing callback.
1944 Severity : enhancement
1946 Description: Update kernel to SLES9 2.6.5-7.311.
1948 Severity : enhancement
1950 Description: Files open for execute are not marked busy on SLES10
1951 Details : Add FMODE_EXEC to SLES10 SP1 server kernel series.
1953 Severity : enhancement
1955 Description: Add server support for vanilla-2.6.22.14.
1958 Frequency : occasional
1960 Description: Avoid lov_create() getting stuck in obd_statfs_rqset()
1961 Details : If an OST is down the MDS will hang indefinitely in
1962 obd_statfs_rqset() waiting for the statfs data. While for
1963 MDS QOS usage of statfs, it should not stuck in waiting.
1965 Severity : enhancement
1967 Description: Disable adaptive timeouts by default
1970 Frequency : on network error
1972 Description: panic with double free request if network error
1973 Details : mdc_finish_enqueue is finish request if any network error occuring,
1974 but it's true only for synchronus enqueue, for async enqueue
1975 (via ptlrpcd) this incorrect and ptlrpcd want finish request
1979 Frequency : rare, on recovery
1981 Description: read procfs can produce deadlock in some situation
1982 Details : Holding lprocfs lock with send rpc can produce block for destroy
1983 obd objects and this also block reconnect with -EALREADY.
1984 This isn't fix all lprocfs bugs - but make it rare.
1986 Severity : enhancement
1988 Description: Update kernel to RHEL5 2.6.18-53.1.14.el5.
1991 Frequency : frequent on X2 node
1993 Description: mdc_set_open_replay_data LBUG
1994 Details : Set replay data for requests that are eligible for replay.
1999 Description: lustre_mgs: operation 101 on unconnected MGS
2000 Details : When MGC is disconnected from MGS long enough, MGS will evict the
2001 MGC, and late on MGC cannot successfully connect to MGS and a lot
2002 of the error messages complaining that MGS is not connected.
2005 Frequency : rare, depends on device drivers and load
2007 Description: MDS or OSS nodes crash due to stack overflow
2008 Details : Code changes in 1.6.4 increased the stack usage of some functions.
2009 In some cases, in conjunction with device drivers that use a lot
2010 of stack, the MDS (or possibly OSS) service threads could overflow
2011 the stack. One change which was identified to consume additional
2012 stack has been reworked to avoid the extra stack usage.
2014 Severity : enhancement
2016 Description: Update to RHEL5 latest kernel-2.6.18-53.1.13.el5.
2018 Severity : enhancement
2020 Description: Update to SLES10 SP1 latest kernel-2.6.16.54-0.2.5.
2022 Severity : enhancement
2024 Description: Update to RHEL5 latest kernel-2.6.18-53.1.6.el5.
2026 Serverity : enhancement
2028 Description: Update RHEL4 kernel to 2.6.9-67.0.4.
2031 Frequency : rare on shutdown OST
2033 Description: Don't allow skipping OSTs if index has been specified.
2034 Details : Don't allow skipping OSTs if index has been specified, make locking
2035 in internal create lots better.
2040 Description: ASSERTION(!PageDirty(page)) failed
2041 Details : Wrong check could lead to an assertion failure under specific
2047 Description: LBUG in ptlrpc_check_set() bad phase ebc0de00
2048 Details : access to bitfield in structure is always rounded to long
2049 and this produce problem with not atomic change any bit.
2054 Description: Lustre make rpms failed.
2055 Details : Remove ldiskfs spec file to avoids rpmbuild be confused when
2056 builds Lustre rpms from tarball.
2058 Severity : enhancement
2060 Description: Update to SLES9 SP4 kernel-2.6.5-7.308.
2063 Frequency : rare on shutdown OST
2065 Description: If llog cancel was not send before clean_exports phase, this can
2066 produce deadlock in llog code.
2067 Details : If llog thread has last reference to obd and call class_import_put
2068 this produce deadlock because llog_cleanup_commit_master wait when
2069 last llog_commit_thread exited, but this never success because was
2070 called from llog_commit_thread.
2073 Frequency : only if OST index is skipped
2075 Description: NULL lov_tgts causing MDS oops
2076 Details : more safe checks for NULL lov_tgts for avoid oops.
2078 Severity : enhancement
2080 Description: Update to RHEL4 latest kernel-2.6.9-67.0.1.EL.
2082 Severity : enhancement
2084 Description: Update to RHEL5 latest kernel-2.6.18-53.1.4.el5.
2089 Description: make mgs_setparam() handle fsname containing dash
2090 Details : fsname containing a dash does not work with lctl conf_param
2092 Severity : enhancement
2094 Description: Update to RHEL4 Update-6 kernel-2.6.9-67.EL.
2097 Frequency : rare, in recovery and (or) destroy lovobjid file.
2099 Description: rewrite lov objid code.
2100 Details : Cleanup for lov objid code, remove scability problems and wrong
2101 locking. Fix sending last_id into ost.
2103 Severity : enhancement
2105 Description: Update to SLES10 SP1 latest kernel-2.6.16.54-0.2.3.
2107 Severity : enhancement
2109 Description: Update to RHEL5 Update-1 kernel 2.6.18-53.el5.
2110 Details : Use d_move_locked instead of __d_move.
2113 Frequency : rare, at shutdown
2115 Description: access already free / zero obd_namespace.
2116 Details : if client_disconnect_export was called without force flag set,
2117 and exist connect request in flight, this can produce access to
2118 NULL pointer (or already free pointer) when connect_interpret
2119 store ocd flags in obd_namespace.
2122 Frequency : only at startup
2124 Description: not alloc memory with spinlock held.
2125 Details : allocation memory with GFP_KERNEL can produce sleep deadlock,
2126 if any spinlock held.
2131 Description: lfs find does not continue on file error
2132 Details : Continue other files processing when a file/dir is absent.
2136 Description: Inconsistent usage of lustre_pack_reply()
2137 Details : Standardize the usage of lustre_pack_reply() such that it
2138 always generate a CERROR on failure.
2141 Frequency : very rare
2143 Description: Fix replay if there is an un-replied request and open
2144 Details : In some cases, older replay request will revert the
2145 mcd->mcd_last_xid on MDS which is used to record the client's
2146 latest sent request.
2148 Severity : enhancement
2150 Description: Update to RHEL5 kernel 2.6.18-8.1.15.el5.
2152 Severity : enhancement
2154 Description: Update to SLES10 SP1 kernel 2.6.16.53-0.16
2156 Severity : enhancement
2158 Description: Update to SLES9 kernel-2.6.5-7.287.3.
2160 Severity : enhancement
2162 Description: Update to RHEL4 kernel-2.6.9-55.0.12.EL.
2164 Severity : enhancement
2166 Description: Build SLES10 patchless client fails
2167 Details : The configure was broken by run ./configure with
2168 --with-linux-obj=.... argument for patchless client. When the
2169 configure use --with-linux-obj, the LINUXINCLUDE= -Iinclude
2170 can't search header adequately. Use absolute path such as
2171 -I($LINUX)/include instead.
2173 Severity : enhancement
2175 Description: Lustre Page Accounting
2176 Details : New macros for page alloc and free which enable accounting
2177 of page allocation of Lustre. Use percpu counters to store memory
2178 and page statistics.
2181 Frequency : only if debugging is disabled
2183 Description: LASSERT_{REQ,REP}SWAB macros are buggy
2184 Details : If SWAB_PARANOIA is disabled, the LASSERT_REQSWAB and
2185 LASSERT_REPSWAB macros become no-ops, which is incorrect. Drop
2186 these macros and replace them with their definitions instead.
2191 Description: interrupt oig_wait produce painc on resend.
2192 Details : brw_redo_request can be used for resend requests from ptlrpcd and
2193 private set, and this produce situation when rq_ptlrpcd_data not
2194 copyed to new allocated request and triggered LBUG on assert
2195 req->rq_ptlrpcd_data != NULL. But this member used only for wakeup
2196 ptlrpcd set if request is changed and can be safety changed to use
2199 Severity : enhancement
2201 Description: organize the server-side client stats on per-nid basis
2202 Details : Change the structure of stats under obdfilter and mds to
2210 The "uuid"s file would list the uuids of _active_ exports.
2211 And the clear entry is to clear all stats and stale nids.
2216 Description: Processes looping in ll_readdir() on Lustre clients finally causing
2217 a full node pseudo-hang
2218 Details : Concurrent access to the same directory from multiple clients with
2219 intensive file creation/removal can cause a client node to spin in
2220 ll_readdir(). i_version must be increased every time the lock is
2221 cancelled to ensure a revalidate is done.
2226 Description: touch file failed when fs is not full
2227 Details : OST in recovery should not be discarded by MDS in alloc_qos(),
2228 otherwise we can get ENOSP while fs is not full.
2231 Frequency : only for Cray XT3
2232 Bugzilla : 12829/13455
2233 Description: Changing primary group doesn't change the group lustre assigns to
2235 Details : When CRAY_XT3 is defined, the fsgid supplied by the client is
2236 overridden with the primary group provided by the group upcall,
2237 whereas the supplied fsgid can be trusted if it is in the list of
2238 supplementary groups returned by the group upcall.
2240 Severity : enhancement
2242 Description: Root Squash Functionality
2243 Details : Implementation of NFS-like root squash capability. Specifically,
2244 don't allow someone with root access on a client node to be able
2245 to manipulate files owned by root on a server node.
2247 Severity : enhancement
2249 Description: Slow trucate/writes to huge files at high offsets.
2250 Details : Directly associate cached pages to lock that protect those pages,
2251 this allows us to quickly find what pages to write and remove
2252 once lock callback is received.
2257 Description: Too many locks accumulating on client during NFS usage
2258 Details : mds_open improperly used accmode to find out access mode to a
2259 file. Also mdc_intent_lock logic to find out if we already have
2260 lock similar to just received was flawed since introduction of
2261 skiplists - locks are now added to the front of the granted
2266 Description: Hit ASSERTION(obd->obd_stopping == 1) failed in some setup failed
2268 Details : In obd setup failure handler, obd_stopping will not necessarily to
2269 be 1, and obd_set_up should also be checked to make sure whether
2270 obd is completely setup.
2272 Severity : enhancement
2274 Description: Allow masking D_WARNING, D_ERROR messages from console
2275 Details : Console messages can now be disabled via lnet.printk.
2280 Description: User code with malformed file open parameter crashes client node
2281 Details : Before packing join_file req, all the related reference should be
2282 checked carefully in case some malformed flags cause fake join_file
2288 Description: shrink/enlarge qunit size when needed; fix the problem of coarse
2289 grain of quota doing harm to quota's accuracy
2290 Details : qunit size will be changed when quota limitation is too low/high;
2291 record the pending quota write in order to get more accureate
2292 quota; delete the patch for bug12588, which is unnecessary when
2293 this patch is landed. This bug also contains fixes for bug 14526,
2294 14299, 14601 and 13794.
2298 Description: LDLM_ENQUEUE races with LDLM_CP_CALLBACK
2299 Details : ldlm_completion_ast() assumes that a lock is granted when the req
2300 mode is equal to the granted mode. However, it should also check
2301 that LDLM_FL_CP_REQD is not set.
2305 Description: Heavy nfs access might result in deadlocks
2306 Details : After ELC code landed, it is now improper to enqueue any mds
2307 locks under och_sem, because enqueue might want to decide to
2308 cancel open locks for same inode we are holding och_sem for.
2312 Description: 35% write performance drop with ldiskfs2 when quotas are on
2313 Details : Enable ext3 journalled quota by default to improve performance
2314 when quotas are turned on.
2318 Description: Client eviction while running blogbench
2319 Details : A lot of unlink operations with concurrent I/O can lead to a
2320 deadlock causing evictions. To address the problem, the number of
2321 oustanding OST_DESTROY requests is now throttled to
2322 max_rpcs_in_flight per OSC and LDLM_FL_DISCARD_DATA blocking
2323 callbacks are processed in priority.
2326 Frequency : RHEL4 only
2328 Description: mkfs is very slow on IA64/RHEL4
2329 Details : A performance regression has been discovered in the MPT Fusion
2330 driver between versions 3.02.73rh and 3.02.99.00rh. As a
2331 consequence, we have downgraded the MPT Fusion driver in the RHEL4
2332 kernel from 3.02.99.00 to 3.02.73 until this problem is fixed.
2335 Frequency : PPC/PPC64 only
2337 Description: conflicts between asm-ppc64/types.h and lustre_types.h
2338 Details : fix duplicated definitions between asm-ppc64/types.h and
2339 lustre_types.h on PPC.
2342 Frequency : PPC/PPC64 only
2344 Description: asm-ppc/segment.h does not exist
2345 Details : fix compile issue on PPC.
2349 Description: data checksumming impacts single node performance
2350 Details : add support for several checksum algorithms. Currently, CRC32 and
2351 Adler-32 are supported. The checksum type can be changed on the fly
2352 through /proc/fs/lustre/osc/*/checksum_type.
2356 Description: use adler32 for page checksums
2357 Details : when available, use the Adler-32 algorithm instead of CRC32 for
2362 Description: better handle error messages in extents code
2364 Severity : enhancement
2366 Description: SNMP support enhancement
2367 Details : Adding total number of sampled request for an MDS node in snmp
2370 Severity : enhancement
2372 Description: Optimize ldlm waiting list processing for PR extent locks
2373 Details : When processing waiting list for read extent lock and meeting read
2374 lock that is same or wider to it that is not contended, skip
2375 processing rest of the list and immediatelly return current
2376 status of conflictness, since we are guaranteed there are no
2377 conflicting locks in the rest of the list.
2381 Description: Time out and refuse to reconnect
2382 Details : When the failover node is the primary node, it is possible
2383 to have two identical connections in imp_conn_list. We must
2384 compare not conn's pointers but NIDs, otherwise we
2385 can defeat connection throttling.
2389 Description: Client not clear own cache if answer to reconnect is lost.
2390 Details : Client gets evicted from server. Now client also thinks it is
2391 disconnected (or gets ENOTCONN on its operation) and decides to
2392 reconnect. Server receives reconnect message, but cannot find
2393 export. New export is created that is fully valid (new cookie!),
2394 but reply is lost and not reported to client. Client reconnects
2395 again and gets back a just-created connection, but it is not new
2396 so client thinks it was not evicted and does not do recovery.
2400 Description: Detect stride IO mode in read-ahead
2401 Details : When a client does stride read, read-ahead should detect that and
2402 read-ahead pages according to the detected stride pattern.
2406 Description: build for x2 fails
2407 Details : fix compile issue on Cray systems.
2409 Severity : enhancement
2411 Description: implement readv/writev APIs(aio_read/aio_writes in newer kernels)
2412 Details : This greatly improves speed of NFS writes on 2.6 kernels.
2415 Frequency : only on PPC/SLES10
2417 Description: "BITS_PER_LONG is not 32 or 64" in linux/idr.h
2418 Details : On SLES10/PPC, fs.h includes idr.h which requires BITS_PER_LONG to
2419 be defined. Add a hack in mkfs_lustre.c to work around this compile
2424 Description: LASSERT on MDS when client holding flock lock dies
2425 Details : ldlm pool logic depends on number of granted locks equal to
2426 number of released locks which is not true for flock locks, so
2427 just exclude such locks from consideration.
2431 Description: MDS deadlock with many ll_sync_lov threads and I/O stalled
2432 Details : Use fsfilt_sync() for both the whole filesystem sync and
2433 individual file sync to eliminate dangerous inode locking
2434 with I_LOCK that can lead to a deadlock.
2438 Description: Update an obsolete wirecheck.c generator
2439 Details : Update wirecheck.c/wirehdr.c and regenerate wiretest.c
2443 Description: Client can panic on open sometimes
2444 Details : It is possible that we try to free already freed request in
2445 ll_file_open in some error cases when we send request from
2450 Description: performance in 1.6.3
2451 Details : Force q->max_phys_segments to MAX_PHYS_SEGMENTS on SLES10 to be
2452 sure that 1MB requests are not fragmented by the block layer.
2456 Description: LDLM soft lockups - improvement
2457 Details : It is be possible to send the lock handle along with each read
2458 or write request because the client is already doing a lock match
2459 itself so there isn't any reason the OST should have to re-do that
2465 Description: lfs quota fails with deactivated OSTS
2466 Details : With this patch, three improvements are included:
2467 1. detete the softlimit in mds and osts when use "lfs quota".
2468 2. display the inaccurate data in the output of "lfs quota".
2469 3. try to get quota info when "lfs quota" is executed.
2474 Description: Extent locks not granted with no conflicts sometimes.
2475 Details : When race occurs in glimpse handler and nothing is returned,
2476 we do not reprocess the queue after lock cancel, and that leads
2477 to a stall until next activity on a resource
2480 Frequency : failover with quotaon
2482 Description: during mds failovers with quota on, OSTs got into deadlock state
2483 and causing dumpstack.
2484 Details : for every quota slave, at any time, there is only one quota req
2485 is sent to quota master for every uid/gid. Before that quota req
2486 returns, all the thread relative to the same uid/gid will wait.
2487 So if the quota req is lost because mds failovers or any other
2488 reasons, this bug will be hit. Now, dqacq_interpret() will handle
2489 quota reqs who time out.
2491 Severity : enhancement
2494 Description: when quota slave checks if quota is enough, there is an unnecessary
2496 Details : place this wait on necessary place instead of always waiting.
2498 --------------------------------------------------------------------------------
2500 2007-12-07 Cluster File Systems, Inc. <info@clusterfs.com>
2502 * Support for kernels:
2503 2.6.5-7.286 (SLES 9),
2504 2.6.9-55.0.9.EL (RHEL 4),
2505 2.6.16.53-0.8 (SLES 10),
2506 2.6.18-8.1.14.el5 (RHEL 5),
2507 2.6.18.8 vanilla (kernel.org)
2508 * Client support for unpatched kernels:
2509 (see http://wiki.lustre.org/index.php?title=Patchless_Client)
2510 2.6.16 - 2.6.22 vanilla (kernel.org)
2511 * Due to recently discovered recovery problems, we do not recommend
2512 using patchless RHEL 4 clients with this or any earlier release.
2513 * Recommended e2fsprogs version: 1.40.2-cfs1
2514 * Note that reiserfs quotas are disabled on SLES 10 in this kernel.
2517 Frequency : occasional
2519 Description: MDS hang or stay in waiting lock
2520 Details : If client receive lock with CBPENDING flag ldlm need send lock
2521 cancel as separate rpc, to avoid situation when cancel request
2522 can't processed due all i/o threads stay in wait lock.
2525 Frequency : occasional
2527 Description: Do not fail import if osc_interpret_create gets -EAGAIN
2528 Details : If osc_interpret_create got -EAGAIN it immediately exits and
2529 wakeup oscc_waitq. After wakeup oscc_wait_for_objects call
2530 oscc_has_objects and see OSC has no objests and call
2531 oscc_internal_create to resend create request.
2533 Severity : enhancement
2535 Description: Update kernel patches for SLES10 2.6.16.53-0.8.
2536 Details : Update which_patch & target file for SLES10 latest kernel.
2538 Severity : enhancement
2540 Description: add --type and --size parameters to lfs find
2541 Details : Enhance lfs find by adding filetype and filesize parameters. Also
2542 multiple OBDs can now be specified for the --obd option.
2544 Severity : enhancement
2546 Description: eliminate client locks in face of contention
2547 Details : file contention detection and lockless i/o implementation
2548 for contended files.
2550 Severity : enhancement
2552 Description: Remove client patches from SLES 10 kernel.
2553 Details : This causes SLES 10 clients to behave as patchless clients
2554 even on a Lustre-patched (server) kernel.
2556 Severity : enhancement
2558 Description: use i_size_read and i_size_write in 2.6 port
2559 Details : replace inode->i_size access with i_size_read/write()
2562 Frequency : when removing large files
2564 Description: scheduling issue during removal of large Lustre files
2565 Details : Don't take the BKL in fsfilt_ext3_setattr() for 2.6 kernels.
2566 It causes scheduling issues when removing large files (17TB in the
2572 Description: 1.4.11 Can't handle directories with stripe set and extended ACLs
2573 Details : Impossible (EPROTO is returned) to access a directory that has a
2574 non-default striping and ACLs.
2577 Frequency : only on ppc
2579 Description: /proc/fs/lustre/devices broken on ppc
2580 Details : The patch as applied to 1.6.2 doesn't look correct for all arches.
2581 We should make sure the type of 'index' is loff_t and then cast
2582 explicitly as needed below. Do not assign an explicitly cast
2586 Frequency : only for rhel5
2588 Description: Kernel patches update for RHEL5 2.6.18-8.1.10.el5.
2589 Details : Modify the target file & which_kernel.
2592 Frequency : if the uninit_groups feature is enabled on ldiskfs
2594 Description: e2fsck reports "invalid unused inodes count"
2595 Details : If a new ldiskfs filesystem is created with the "uninit_groups"
2596 feature and only a single inode is created in a group then the
2597 "bg_unused_inodes" count is incorrectly updated. Creating a
2598 second inode in that group would update it correctly.
2603 Description: buffer overruns could theoretically occur
2604 Details : llapi_semantic_traverse() modifies the "path" argument by
2605 appending values to the end of the origin string, and a buffer
2606 overrun may occur. Adding buffer overrun check in liblustreapi.
2608 Severity : enhancement
2610 Description: Add jbd statistics patch for RHEL5 and 2.6.18-vanilla.
2613 Frequency : only if filesystem is inconsistent
2615 Description: handle "serious error: objid * already exists" more gracefully
2616 Details : If LAST_ID value on disk is smaller than the objects existing in
2617 the O/0/d* directories, it indicates disk corruption and causes an
2618 LBUG(). If the object is 0-length, then we should use the existing
2619 object. This will help to avoid a full fsck in most cases.
2621 Severity : enhancement
2623 Description: Kernel patches update for RHEL4 2.6.9-55.0.6.
2624 Details : Modify vm-tunables-rhel4.patch.
2626 Severity : enhancement
2628 Description: Kernel config for 2.6.18-vanilla.
2629 Details : Modify targets/2.6-vanilla.target.in.
2630 Add config file kernel-2.6.18-2.6-vanilla-i686.config.
2631 Add config file kernel-2.6.18-2.6-vanilla-i686-smp.config.
2632 Add config file kernel-2.6.18-2.6-vanilla-x86_64.config.
2633 Add config file kernel-2.6.18-2.6-vanilla-x86_64-smp.config.
2636 Frequency : occasional
2638 Description: improve handling recoverable errors
2639 Details : If request processed with error which can be recoverable on server
2640 request should be resend, otherwise page released from cache and
2646 Description: Kernel patches update for RHEL5 2.6.18-8.1.14.el5.
2647 Details : Modify target file & which_patch.
2648 A flaw was found in the IA32 system call emulation provided
2649 on AMD64 and Intel 64 platforms. An improperly validated 64-bit
2650 value could be stored in the %RAX register, which could trigger an
2651 out-of-bounds system call table access. An untrusted local user
2652 could exploit this flaw to run code in the kernel
2653 (ie a root privilege escalation). (CVE-2007-4573).
2657 Description: change order of libsysio includes
2658 Details : '#include sysio.h' should always come before '#include xtio.h'
2660 Severity : enhancement
2662 Description: adapt the lustre_config script to support the upgrade case
2663 Details : Add "-u" option for lustre_config script to support upgrading 1.4
2664 server targets to 1.6 in parallel.
2669 Description: To avoid grant space > avaible space when the disk is almost
2670 full. Without this patch you might see the error "grant XXXX >
2671 available" or some LBUG about grant, when the disk is almost
2673 Details : In filter_check_grant, for non_grant cache write, we should
2674 check the left space by if (*left > ungranted + bytes), instead
2675 of (*left > ungranted), because only we are sure the left space
2676 is enough for another "bytes", then the ungrant space should be
2677 increase. In client, we should update cl_avail_grant only there is
2678 OBD_MD_FLGRANT in the reply.
2683 Description: Update RHEL 4 kernel to fix local root privilege escalation.
2684 Details : Update to the latest RHEL 4 kernel to fix the vulnerability
2685 described in CVE-2007-4573. This problem could allow untrusted
2686 local users to gain root access.
2689 Frequency : when using O_DIRECT and quotas
2691 Description: Incorrect file ownership on O_DIRECT output files
2692 Details : block usage reported by 'lfs quota' does not take into account
2693 files that have been written with O_DIRECT.
2697 Description: (rw.c:1323:ll_read_ahead_pages()) ASSERTION(page_idx > ria->ria_stoff) failed
2698 Details : Once the unmatched stride IO mode is detected, shrink the stride-ahead
2699 window to 0. If it does hit cache miss, and read-pattern is still
2700 stride-io mode, does not reset the stride window, but also does not
2701 increase the stride window length in this case.
2703 --------------------------------------------------------------------------------
2705 2007-09-27 Cluster File Systems, Inc. <info@clusterfs.com>
2707 * Support for kernels:
2708 2.6.5-7.286 (SLES 9),
2709 2.6.9-55.0.2.EL (RHEL 4),
2710 2.6.16.46-0.14 (SLES 10),
2711 2.6.18-8.1.8.el5 (RHEL 5),
2712 2.6.18.8 vanilla (kernel.org)
2713 * Client support for unpatched kernels:
2714 (see http://wiki.lustre.org/index.php?title=Patchless_Client)
2715 2.6.16 - 2.6.21 vanilla (kernel.org)
2716 * Due to recently discovered recovery problems, we do not recommend
2717 using patchless RHEL 4 clients with this or any earlier release.
2718 * Recommended e2fsprogs version: 1.40.2-cfs1
2719 * Note that reiserfs quotas are disabled on SLES 10 in this kernel.
2723 Description: Fix errors in lfs documentation
2724 Details : Fixes man pages
2726 Severity : enhancement
2728 Description: Adaptive timeouts
2729 Details : RPC timeouts adapt to changing server load and network
2730 conditions to reduce resend attempts and improve recovery time.
2732 Severity : enhancement
2734 Description: llapi_file_create() does not allow some changes
2735 Details : add llapi_file_open() that allows specifying the file creation
2736 mode and open flags, and also returns an open file handle.
2738 Severity : enhancement
2740 Description: df doesn't work properly if diskfs blocksize != 4K
2741 Details : Choose biggest blocksize of OST's as the LOV's blocksize.
2744 Frequency : always on directories with default striping set
2746 Description: lfs find on -1 stripe looping in lsm_lmm_verify_common()
2747 Details : Avoid lov_verify_lmm_common() on directory with -1 stripe count.
2749 Severity : enhancement
2751 Description: merge and cleanup kernel patches.
2752 Details : Remove mnt_lustre_list in vfs_intent-2.6-rhel4.patch.
2754 Severity : enhancement
2756 Description: RedHat Update kernel for RHEL5
2757 Details : Kernel config file for RHEL5.
2759 Severity : enhancement
2761 Description: OSS needs mutliple precreate threads
2762 Details : Add ability to start more than one create thread per OSS.
2765 Frequency : only with quota on the root user
2767 Description: mds_obd_create error creating tmp object
2768 Details : When the user sets quota on root, llog will be affected and can't
2769 create files and write files.
2772 Frequency : Always on ia64 patchless client, and possibly others.
2774 Description: Add EXPORT_SYMBOL check for node_to_cpumask symbol.
2775 Details : This allows the patchless client to be loaded on architectures
2776 without this export.
2781 Description: disorder of journal start and llog_add cause deadlock.
2782 Details : in llog_origin_connect, journal start should happen before
2783 llog_add keep the same order as other functions to avoid
2786 Severity : enhancement
2788 Description: RedHat Update kernel for RHEL5
2789 Details : Modify the kernel config file more closer RHEL5.
2791 Severity : enhancement
2793 Description: Build failure against Centos5 (RHEL5)
2794 Details : Define PAGE_SIZE when it isn't present.
2797 Frequency : occasionally when using NFS
2799 Description: "ll_intent_file_open()) lock enqueue: err: -13" with nfs
2800 Details : with NFS, the anon dentry's parent was set to itself in
2801 d_alloc_anon(), so in MDS, we use rec->ur_fid1 to find the
2802 corresponding dentry other than use rec->ur_name.
2804 Severity : enhancement
2806 Description: enable data checksumming by default
2807 Details : enable checksum by default, allow --disable-checksum
2808 configure option and "-o nochecksum" mount option. Checksums
2809 can also be disabled at runtime via $LPROC/osc/*/checksum_pages.
2812 Frequency : Occasionally with failover
2814 Description: Client eviction due to failover config
2815 Details : after a connection loss, the lustre client should attempt to
2816 reconnect to the last active server first before trying the
2817 other potential connections.
2819 Severity : enhancement
2821 Description: client-side metadata stat-ahead during readdir(directory readahead)
2822 Details : perform client-side metadata stat-ahead when the client detects
2823 readdir and sequential stat of dir entries therein
2828 Description: when mds and osts use different quota unit(32bit and 64bit),
2829 quota will be released repeatly.
2830 Details : void sending multiple quota reqs to mds, which will keep the
2831 status between the reqs.
2834 Frequency : only with liblustre clients on XT3
2836 Description: evictions taking too long
2837 Details : allow llrd to evict clients directly on OSTs
2841 Description: osts not allocated evenly to files
2842 Details : change the condition to increase offset_idx
2844 Severity : enhancement
2846 Description: self-adjustable client's lru lists
2847 Details : use adaptive algorithm for managing client cached locks lru
2848 lists according to current server load, other client's work
2849 pattern, memory activities, etc. Both, server and client
2850 side namespaces provide number of proc tunables for controlling
2855 Description: rewrite ext2-derived code in llite/dir.c and obdclass/uuid.c
2856 Details : rewrite inherited code (uuid parsing code from ext2 utils and
2857 readdir code from ext3) from scratch preserving functionality.
2861 Description: Only those disconnect error should be returned by rq_status.
2862 Details : In open/enqueue processs, Some errors, which will cause client
2863 disconnected, should be returned by rq_status, while other
2864 errors should still be returned by intent, then mdc or llite will
2867 Severity : enhancement
2869 Description: Tune the kernel for good SCSI performance.
2870 Details : Set the value of /sys/block/{dev}/queue/max_sectors_kb
2871 to the value of /sys/block/{dev}/queue/max_hw_sectors_kb
2875 Frequency : Always for filesystems larger than 2TB on 32-bit systems.
2876 Bugzilla : 13547, 13627
2877 Description: Data corruption for OSTs that are formatted larger than 2TB
2879 Details : When generating the bio request for lustre file writes the
2880 sector number would overflow a temporary variable before being
2881 used for the IO. The data reads correctly from Lustre (which
2882 will overflow in a similar manner) but other file data or
2883 filesystem metadata may be corrupted in some cases.
2887 Description: "lfs find -obd UUID" prints directories
2888 Details : "lfs find -obd UUID" will return all directory names instead
2889 of just file names. It is incorrect because the directories
2890 do not reside on the OSTs.
2893 Bugzilla : 13596, 13614
2894 Description: MDS hang after unclean shutdown of lots of clients
2895 Details : Never resend AST requests.
2899 Frequency : Always, for kernels after 2.6.16
2900 Description: Fix warning idr_remove called for id=.. which is not allocated.
2901 Details : Last kernels save old s_dev before kill super and not allow
2902 to restore from callback - restore it before call kill_anon_super.
2906 Description: Fix error on 'ls .' at the top of the Lustre mount.
2907 Details : Don't revalidate dentry if it is a root dentry.
2909 --------------------------------------------------------------------------------
2911 2007-08-27 Cluster File Systems, Inc. <info@clusterfs.com>
2913 * Support for kernels:
2914 2.6.5-7.286 (SLES 9),
2915 2.6.9-55.0.2.EL (RHEL 4),
2916 2.6.16.46-0.14 (SLES 10),
2917 2.6.18-8.1.8.el5 (RHEL 5),
2918 2.6.18.8 vanilla (kernel.org)
2919 * Client support for unpatched kernels:
2920 (see http://wiki.lustre.org/index.php?title=Patchless_Client)
2921 2.6.16 - 2.6.22 vanilla (kernel.org)
2922 * Due to recently discovered recovery problems, we do not recommend
2923 using patchless RHEL 4 clients with this or any earlier release.
2924 * Recommended e2fsprogs version: 1.39.cfs8
2925 * Note that reiserfs quotas are disabled on SLES 10 in this kernel.
2930 Description: block reactivating mgc import until all deactivates complete
2931 Details : Fix race when failing back MDT/MGS to itself (testing)
2933 Severity : enhancement
2935 Description: lfs setstripe enhancement
2936 Details : Make lfs setstripe understand 'k', 'm' and 'g' for stripe size.
2939 Frequency : mds/oss recovery
2941 Description: llog ctxt is refrenced after it has been freed.
2942 Details : llog ctxt refcount was added to avoide the race between ctxt free
2943 and llog recovery process. Each llog user must hold ctxt refcount
2944 before it access the llog. And the llog ctxt can only be freed
2945 when its refcount is zero.
2947 Severity : enhancement
2949 Description: randomly memory allocation failure util
2950 Details : Make lustre randomly failed allocating memory for testing purpose.
2953 Frequency : only for SLES10
2955 Description: Update kernel patch for SLES10 SP1
2956 Details : Add patch blkdev_tunables-2.6-sles10.patch to 2.6-sles10.series.
2958 Severity : enhancement
2960 Description: omit set fsid for export NFS
2961 Details : fix set/restore device id for avoid EMFILE error and mark lustre fs
2962 as FS_REQUIRES_DEV for avoid problems with generate fsid.
2965 Frequency : after network failures
2967 Description: Add sync option to mount_lustre.c
2968 Details : Client loses data written to lustre after a network interruption.
2970 Severity : enhancement
2972 Description: Error message improvement.
2973 Details : Merging of two LCONSOLE_ERROR_MSG into one.
2975 Severity : enhancement
2977 Description: don't use GFP_* in generic Lustre code.
2978 Details : Use cfs_alloc_* functions and CFS_* flags for code portability.
2980 Severity : enhancement
2982 Description: obdclass is limited by single OBD_ALLOC(idarray)
2983 Details : replace OBD_ALLOC/OBD_FREE with OBD_VMALLOC/OBD_VFREE
2985 Severity : enhancement
2987 Description: updated patchess for new RHEL4 kernel
2988 Details : Fixed ext3-unlink-race.patch per Kalpak's comment.
2990 Severity : enhancement
2992 Description: warnings with build patchless client with vanila 2.6.19 and up
2993 Details : change old ctl_table style and replace ctl_table/ctl_table_header
2994 with cfs_sysctl_table_t/cfs_sysctl_table_header_t
2996 Severity : enhancement
2998 Description: O_DIRECT bypasses client statistics.
2999 Details : When running with O_DIRECT I/O, neither the client rpc_stats nor
3000 read_ahead_stats were updated. Copied stats section from
3001 osc_send_oap_rpc() into async_internal().
3004 Frequency : only for Cray XT3
3006 Description: peer credits not enough on many OST per OSS systems.
3007 Details : Use new lnet way to add credits as we need those for pings and ASTs
3010 Frequency : only with liblustre
3012 Description: Liblustre is not releasing flock locks on file close.
3013 Details : Release flock locks on file close.
3016 Frequency : only for RHEL4
3018 Description: Update kernel patches for kernel-2.6.9-55.0.2.EL
3019 Details : Remove inode-nr_unused-2.6.9-rhel4.patch from 2.6-rhel4.series
3020 Update target file and kernel config.
3024 Description: lustre support for RHEL5
3025 Details : Add support for RHEL5.
3030 Description: ASSERTION(export != NULL) failed in target_handle_connect
3031 Details : Assetion hit is result of rare race between disconnect and connet
3032 to same nid. target_handle_connect found old connect cockie and
3033 tried to reconnect, but can't find export for this cockie.
3038 Description: umount blocks forever on error
3039 Details : In result of wrong using obd_no_recov and obd_force flags client
3040 can hand if cancel or some other requests is lost.
3043 Frequency : Only for SLES
3045 Description: sanity_quota fail test_1
3046 Details : There are multiple occurences of $TSTUSR in SLES's /etc/group
3047 file, which makes TSTID[2] inunique.
3049 Severity : enhancement
3051 Description: Kernel patches for SLES9 2.6.5-7.286 kernel
3052 Details : Update target/ChangeLog/which_patch .
3054 Severity : enhancement
3056 Description: jbd statistics
3057 Details : Port older jbd statistics patch for sles10
3062 Description: lvbo_init failed for resource with missing objects.
3063 Details : Fix returning error if we do stat for file with missing/corrupted
3064 objects and i_size set to all sum of size of all avaible objects.
3065 if we to truncate/write to missing object - it is recreated.
3070 Description: Oops in read and write path when failing to allocate lock.
3071 Details : Check if lock allocation failed and return error back.
3074 Frequency : When flocks are used.
3076 Description: assertion failure in ldlm_cli_enquque_fini for non NULL lock.
3077 Details : Flock locks might destroy just granted lock if it could be merged
3078 with another existing flock, this is done in completion handler,
3079 so teach ldlm_cli_enquque_fini that this is a valid case for
3085 Description: reply_lock_interpret crash due to race with it and lock cancel.
3086 Details : Do not replay locks that are being cancelled. Do not reference
3087 locks by their address during replay, just by their handle.
3090 Frequency : only with deactivated OSTs
3092 Description: lstripe command fails for valid OST index
3093 Details : The stripe offset is compared to 'lov->desc.ld_tgt_count'
3094 instead of lov->desc.ld_active_tgt_count.
3096 Severity : enhancement
3098 Description: Build failure against Centos5 (RHEL5)
3099 Details : Use getpagesize() instead of PAGE_SIZE.
3101 --------------------------------------------------------------------------------
3103 2007-07-30 Cluster File Systems, Inc. <info@clusterfs.com>
3105 * Support for kernels:
3106 2.6.5-7.283 (SLES 9),
3107 2.6.9-55.EL (RHEL 4),
3108 2.6.16.46-0.14 (SLES 10),
3109 2.6.18.8 vanilla (kernel.org)
3110 * Client support for unpatched kernels:
3111 (see http://wiki.lustre.org/index.php?title=Patchless_Client)
3112 2.6.16 - 2.6.22 vanilla (kernel.org)
3113 * Due to recently discovered recovery problems, we do not recommend
3114 using patchless RHEL 4 clients with this or any earlier release.
3115 * Recommended e2fsprogs version: 1.39.cfs8
3116 * Note that reiserfs quotas are disabled on SLES 10 in this kernel.
3117 * Starting with this release, the ldiskfs backing filesystem required
3118 by Lustre is now in its own package, lustre-ldiskfs. This package
3119 should be installed. It is versioned separately from Lustre and
3120 may be released separately in future.
3125 Description: server LBUG when shutting down
3126 Details : Block umount forever until the mount refcount is zero rather
3127 than giving up after an arbitrary timeout.
3129 Severity : enhancement
3131 Description: add optional extra BUILD_VERSION info
3132 Details : add a new environment variable (namely LUSTRE_VERS) which allows
3133 to override the lustre version.
3136 Frequency : 2.6.18 servers only
3138 Description: ll_kern_mount() doesn't release the module reference
3139 Details : The ldiskfs module reference count never drops down to 0
3140 because ll_kern_mount() doesn't release the module reference.
3145 Description: server LBUG when using old ost_num_threads parameter
3146 Details : Accept the old ost_num_threads parameter but warn that it
3147 is deprecated, and fix an off-by-one error that caused an LBUG.
3152 Description: Transient SCSI error results in persistent IO issue
3153 Details : iobuf->dr_error is not reinitialized to 0 between two
3157 Frequency : sometimes when underlying device returns I/O errors
3159 Description: OSTs not going read-only during write failures
3160 Details : OSTs are not remounted read-only when the journal commit threads
3161 get I/O errors because fsfilt_ext3 calls journal_start/stop()
3162 instead of the ext3 wrappers.
3165 Frequency : only with 10000 clients or more
3167 Description: poor connect scaling with increasing client count
3168 Details : Don't run filter_grant_sanity_check for more than 100 exports
3169 to improve scaling for large numbers of clients.
3172 Frequency : SLES10 only
3174 Description: sanity-quota.sh quotacheck failed: rc = -22
3175 Details : Quotas cannot be enabled on SLES10.
3178 Frequency : liblustre clients only
3180 Description: getdirentries does not give error when run on compute nodes
3181 Details : getdirentries does not fail when the size specified as an argument
3182 is too small to contain at least one entry
3184 Severity : enhancement
3186 Description: Add LNET router traceability for debug purposes
3187 Details : If a checksum failure occurs with a router as part of the
3188 IO path, the NID of the last router that forwarded the bulk data
3189 is printed so it can be identified.
3194 Description: OST "spontaneously" evicts client; client has imp_pingable == 0
3195 Details : Due to a race condition, liblustre clients were occasionally
3196 evicted incorrectly.
3198 Severity : enhancement
3200 Description: lfs setstripe use optional parameters instead of postional
3203 Severity : enhancement
3205 Description: Nanosecond timestamp support for ldiskfs
3206 Details : The on-disk ldiskfs filesystem has added support for nanosecond
3207 resolution timestamps. There is not yet support for this at
3208 the Lustre filesystem level.
3211 Frequency : during server recovery
3213 Description: MDS failing to send precreate requests due to OSCC_FLAG_RECOVERING
3214 Details : request with rq_no_resend flag not awake l_wait_event if they get a
3218 Frequency : nfs export on patchless client
3220 Description: connectathon hang when test nfs export over patchless client
3221 Details : Disconnected dentry cannot be found with lookup, so we do not need
3222 to unhash it or make it invalid
3225 Description: fix llapi_lov_get_uuids() to allow many OSTs to be returned
3226 Details: : Change llapi_lov_get_uuids() to read the UUIDs from /proc instead
3227 of using an ioctl. This allows lfsck for > 160 OSTs to succeed.
3232 Description: open req refcounting wrong on reconnect
3233 Details : If reconnect happened between getting open reply from server and
3234 call to mdc_set_replay_data in ll_file_open, we will schedule
3235 replay for unreferenced request that we are about to free.
3236 Subsequent close will crash in variety of ways.
3237 Check that request is still eligible for replay in
3238 mdc_set_replay_data().
3243 Description: disable writes to filesystem when reading health_check file
3244 Details : the default for reading the health_check proc file has changed
3245 to NOT do a journal transaction and write to disk, because this
3246 can cause reads of the /proc file to hang and block HA state
3247 checking on a healthy but otherwise heavily loaded system. It
3248 is possible to return to the previous behaviour during configure
3249 with --enable-health-write.
3251 Severity : enhancement
3253 Description: 64-bit inode version
3254 Details: : Add a on-disk 64-bit inode version for ext3 to track changes made
3255 to the inode. This will be required for version-based recovery.
3260 Description: MDS fails to start if a duplicate client export is detected
3261 Details : in some rare cases it was possible for a client to connect to
3262 an MDS multiple times. Upon recovery the MDS would detect this
3263 and fail during startup. Handle this more gracefully.
3265 Severity : enhancement
3267 Description: Add -o localflock option to simulate old noflock behaviour.
3268 Details : This will achieve local-only flock/fcntl locks coherentness.
3273 Description: log_commit_thread vs filter_destroy race leads to crash
3274 Details : Take import reference before releasing llog record semaphore
3279 Description: Wrong request locking in request set processing
3280 Details : ptlrpc_check_set wrongly uses req->rq_lock for proctect add to
3281 imp_delayed_list, in this place should be used imp_lock.
3284 Frequency : when reconnecting
3286 Description: Grant leak when OSC reconnect to OST
3287 Details : When osc reconnect ost, OST(filter) should check whether it
3288 should grant more space to client by comparing fed_grant and
3289 cl_avail_grant, and return the granted space to client instead
3290 of "new granted" space, because client will call osc_init_grant
3291 to update the client grant space info.
3294 Frequency : when client reconnects to OST
3296 Description: Grant leak when OSC does a resend and replays bulk write
3297 Details : When osc reconnect to OST, OST(filter) should clear grant info of
3298 bulk write request, because the grant info while be sync between
3299 OSC and OST when reconnect, and we should ignore the grant info
3300 these of resend/replay write req.
3305 Description: Grant space more than avaiable space sometimes.
3306 Details : When then OST is about to be full, if two bulk writing from
3307 different clients came to OST. Accord the avaliable space of the
3308 OST, the first req should be permitted, and the second one
3309 should be denied by ENOSPC. But if the seconde arrived before
3310 the first one is commited. The OST might wrongly permit second
3311 writing, which will cause grant space > avaiable space.
3314 Frequency : when client is evicted
3316 Description: Grant might be wrongly erased when osc is evicted by OST
3317 Details : when the import is evicted by server, it will fork another
3318 thread ptlrpc_invalidate_import_thread to invalidate the
3319 import, where the grant will be set to 0. While the original
3320 thread will update the grant it got when connecting. So if
3321 the former happened latter, the grant will be wrongly errased
3322 because of this race.
3327 Description: Checking Stale with correct fid
3328 Details : ll_revalidate_it should uses de_inode instead of op_data.fid2
3329 to check whether it is stale, because sometimes, we want the
3330 enqueue happened anyway, and op_data.fid2 will not be initialized.
3332 Severity : enhancement
3334 Description: update patchless client
3335 Details : Add support for patchless client with 2.6.20, 2.6.21 and RHEL 5
3338 Frequency : only with 2.4 kernel
3340 Description: random memory corruption
3341 Details : size of struct ll_inode_info is to big for union inode.u and this
3342 can be cause of random memory corruption.
3347 Description: Memory leak in recovery
3348 Details : Lov_mds_md was not free in an error handler in mds_create_object.
3349 It should also check obd_fail before fsfilt_start, otherwise if
3350 fsfilt_start return -EROFS,(failover mds during mds recovery).
3351 then the req will return with repmsg->transno = 0 and rc = EROFS.
3352 and we met hit the assert LASSERT(req->rq_reqmsg->transno ==
3353 req->rq_repmsg->transno) in ptlrpc_replay_interpret. Fcc should
3354 be freed no matter whether fsfilt_commit success or not.
3357 Frequency : only with large numbers of cores on a single node
3359 Description: Prevents from taking the superblock lock in llap_from_page for
3361 Details : using LL_ORIGIN_REMOVEPAGE origin flag instead of LL_ORIGIN_UNKNOW
3362 for llap_from_page call in ll_removepage() prevents from taking
3363 the superblock lock for a soon died page.
3368 Description: Not check open intent error before release open handle
3369 Details : in some rare cases, the open intent error is not checked before
3370 release open handle, which may cause
3371 ASSERTION(open_req->rq_transno != 0), because it tries to release
3372 the failed open handle.
3377 Description: Set cat log bitmap only after create log success.
3378 Details : in some rare cases, the cat log bitmap is set too early. and it
3379 should be set only after create log success.
3383 Description: Accessing a block bevice can re-enable I/O when Lustre is
3384 tearing down a device.
3385 Details : dev_clear_rdonly(bdev) must be called in kill_bdev() instead of
3390 Description: service threads may hog cpus when there are a lot of requests
3391 Details : Insert cond_resched to give other threads a chance to use some CPU
3396 Description: the cat log was not initialized in recovery
3397 Details : When mds(mgs) do recovery, the tgt_count might be zero, so the
3398 unlink log on mds will not be initialized until mds post
3399 recovery. And also in mds post recovery, the unlink log will
3400 initialization will be done asynchronausly, so there will be race
3401 between add unlink log and unlink log initialization.
3405 Description: brw_stats were being printed incorrectly
3406 Details : brw_stats were being printed as log2 but all of them were not
3407 recorded as log2. Also remove some code duplication arising from
3408 filter_tally_{read,write}.
3412 Frequency : rare, only in recovery.
3413 Description: ASSERTION(req->rq_type != LI_POISON) failed
3414 Details : imp_lock should be held while iterating over imp_sending_list for
3415 prevent destroy request after get timeout in ptlrpc_queue_wait.
3419 Description: replay-single.sh test 52 fails
3420 Details : A lock's skiplist need to be cleanup when it being unlinked
3421 from its resource list.
3426 Description: Short directio read returns full requested size rather than
3428 Details : Direct I/O operations should return actual amount of bytes
3429 transferred rather than requested size.
3431 Severity : enhancement
3433 Description: metadata RPC reduction (e.g. for rm performance)
3434 Details : decrease the amount of synchronous RPC between clients and servers
3435 by canceling conflicing lock before the operation on the client
3436 and packing thier handles into the main operation RPC to server.
3438 Severity : enhancement
3440 Description: add #ifdef HAVE_KERNEL_CONFIG_H
3441 Details : kernels from 2.6.19 not need include linux/config.h, but add
3442 include linux/autoconf.h in commpiler command line.
3444 Severity : enhancement
3446 Description: patchless client support for 2.6.22 kernel
3447 Details : 2.6.22 has only one visble change, SLAB_CTOR_* constants is
3448 removed. In this case we need drop using os depended interface to
3449 kmem_cache and use cfs_mem_cache API.
3454 Description: fix mal-formatted messages
3455 Details : fix some mal-formatted DEBUG_REQ and LCONSOLE_ERROR_MSG messages
3459 Frequency : always in liblustre
3460 Description: wrong IS_ERR implementation in liblustre.h
3461 Details : fix IS_ERR implementation in liblustre.h for right detect errors.
3466 Description: Correct condition for output debug message.
3467 Details : inode i_nlink equal zero is not enough for output message about
3468 disk corruption, i_ctime and i_mode should be also checked.
3472 Frequency : always in patchless client
3473 Description: add configure check for truncate_complete_page
3474 Details : improve checks for exported symbols. This allow run check without
3475 sources, but with Module.symvers shipped with kernel distribution.
3476 add check for truncate_complete_page used by patchless client.
3480 Description: sanity.sh test_77h fails with "test_77h file compare failed"
3481 Details : test_77h uses a file which was messed by other test case.
3485 Description: Not Check whether lov_tgts is NULL in some lov functions
3486 Details : Checking whether lov_tgts is NULL in some functions.
3490 Description: replace obdo_alloc() with OBDO_ALLOC macro
3491 Details : nothing special is done in obdo_alloc() function, and for
3492 debugging purpose, it needs to be replaced with macros.
3496 Description: bad return value and errno from fcntl call
3497 Details : In liblustre API, errno should be a negative value if error
3502 Description: ptlrpc_check_set() LBUG
3503 Details : In case of positive reply from server and failed client bulk
3504 callback after bulk transfer shouldn't LBUG, but process this
3505 request as erroneous.
3507 Severity : enhancement
3509 Description: VFS operations stats tool.
3510 Details : Tool which collects stats by tracking value written in pid,
3511 ppid, gid and uses llstat to generate output to plot graph using
3513 Updated lustre/utils/Makefile.am
3514 Added lustre/utils/ltrack_stats.c
3516 Severity : enhancement
3518 Description: 2.6.18 server support (lustre 1.6.1)
3519 Details : Support for 2.6.18 kernels on the server side.
3524 Description: ASSERTION(imp->imp_conn_current) failed
3525 Details : an assertion failure is hit if a client node boots and attempts to
3526 mount a lustre filesystem faster than RECONNECT_INTERVAL seconds.
3529 Frequency : only for i686
3531 Description: 1.4.11 RC1 build fails for RHEL 4, i686
3532 Details : Fixed config variable for build.
3537 Description: Updated patchess for new RHEL4 kernel
3538 Details : Updated patch inode-nr_unused-2.6.9-rhel4.patch
3539 Updated patch jbd-stats-2.6.9.patch
3540 Updated patch qsnet-rhel4-2.6.patch
3541 Updated patch quota-deadlock-on-pagelock-core.patch
3542 Updated patch vfs_intent-2.6-rhel4.patch
3543 Updated patch vfs_races-2.6-rhel4.patch
3544 Updated series file 2.6-rhel4-titech.series
3545 Updated series file 2.6-rhel4.series
3546 Updated kernel config files
3551 Description: lquota slave complains LBUG when reconnecting with mds
3553 Details : quota slave depends qctxt->lqc_import to send its quota request.
3554 This pointer will be invalid if mds did failover or broke its
3555 connect to osts, which leads to LBUG.
3558 Frequency : when qunit size is too small(less than 20M)
3560 Description: write is stopped by improper -EDQUOT
3561 Details : If the master is busy and qunit size is small enough(let's say 1M),
3562 the slave can not get quota from master on time, which will lead
3563 slave to trigger a -EQUOTA to client.
3568 Description: Deadlock during metadata tests
3569 Details : in prune_dir_dentries(), shrink_dcache_parent() should not be
3570 called with the per-dentry lock held.
3573 Frequency : SLES9 only
3575 Description: Lustre patched kernel for SLES9 SP3 has NR_CPUS set to 8
3576 Details : set CONFIG_NR_CPUS to 128 instead of 8.
3578 Severity : enhancement
3580 Description: remove fs_prep_san_write operation and related patches
3581 Details : remove the ext3-san-jdike patches which are no longer useful.
3586 Description: LDISKFS-fs error (device sdc): ldiskfs_free_blocks
3587 Details : a disk corruption can cause the mballoc code to assert on a
3588 double free or other extent corruptions. Handle these with
3589 ext3_error() instead of assertions.
3592 Frequency : only with mballoc3 code and deep extent trees
3594 Description: ldiskfs_ext_search_right: bad header in inode: unexpected eh_depth
3595 Details : a wrong check of extent headers in ldiskfs_ext_search_right()
3596 can cause the filesystem to be remounted read-only.
3601 Description: lfsck built against 1.4.x cannot run against 1.6.0 lustre
3602 Details : the definition for OBD_IOC_GETNAME changed in 1.6.0. One of the
3603 few external users of this ioctl number is lfsck's call to
3604 llapi_lov_get_uuids() and this caused lfsck to fail at startup.
3605 Add the old ioctl number to the handler so both old and new
3610 Description: parallel lock callbacks
3611 Details : Instead of sending blocking and completion callbacks as separated
3612 requests, adding them to a set and sending in parallel.
3616 Description: Disable most debugging by default
3617 Details : To improve performance, disable most logging (for debug purposes)
3618 by default. VFSTRACE, RPCTRACE, and DLMTRACE are now off by
3619 default, and HA includes fewer messages.
3623 Frequency : only run on patchless client.
3624 Description: use do_facet on sanity.sh for test handling recoverables errors
3625 Details : use do_facet instead of direct use sysctl for set fail_loc on OST
3629 Description: hash tables for lists of nids, connections and uuids
3630 Details : Hash tables noticeably help when a lot of clients connect to a
3631 server, to faster identify duplicate connections or reconnects,
3632 also to faster find export to evict in manual eviction case.
3636 Description: Sometimes, when the server evict a client, and the client will
3637 not be evicted as soon as possible.
3638 Details : In enqueue req, the error was returned by intent, instead of
3639 rq_status which make ptlrpc layer not detect this error, and
3640 does not evict the client. So enqueue error should be returned
3644 Frequency : only at startup
3646 Description: Delay client connections to MDT until fist MDT->OST connect
3647 Details : If a client tried to create a new file before the MDT had
3648 connected to any OSTs, the create would return EIO. Now
3649 the client will simply block until the MDT connects to the
3650 first OST and the create can succeed.
3653 Frequency : only for SLES9
3655 Description: Routinely utilize latest Quadrics drivers in CFS releases
3656 Details : Update patch qsnet-suse-2.6.patch.
3659 Frequency : only for sles10
3661 Description: Update patches for SLES 10 SP1 kernel.
3662 Details : Update the patch vfs_intent-2.6-sles10.patch.
3667 Description: Routinely utilize latest Quadrics drivers in CFS releases
3668 Details : Update patch qsnet-rhel4-2.6.patch.
3671 Frequency : at statup only
3673 Description: mds_lov_synchronize race leads to various problems
3674 Details : simultaneous MDT->OST connections at startup can cause the
3675 sync to abort, leaving the OSC in a bad state.
3680 Description: Using wrong pointer in osc_brw_prep_request
3681 Details : Access to array[-1] can produce panic if kernel compiled with
3682 CONFIG_PAGE_ALLOC enabled
3684 Severity : enhancement
3686 Description: Async OSC create to avoid the blocking unnecessarily.
3687 Details : If a OST has no remain object, system will block on the creating
3688 when need to create a new object on this OST. Now, ways use
3689 pre-created objects when available, instead of blocking on an
3690 empty osc while others are not empty. If we must block, we block
3691 for the shortest possible period of time.
3695 Frequency : only in recovery
3696 Description: Mark OST as early accessible if his start SYNC.
3697 Details : osc_precreate return flag early accessible if oscc marked as
3698 OSCC_FLAG_SYNC_IN_PROGRESS.
3703 Description: Sometimes precreate code can triger create object on wrong ost
3704 Details : Wrong protected or not not restored variables aftre precreate loop
3705 can produce creation object on wrong ost.
3708 Frequency : oss recovery
3710 Description: llog_commit_thread cleanup should sync with llog_commit_thread
3712 Details : llog_commit_thread_count should be synced between llog_commit
3713 start and cleanup, so new llog_commit thread should not be started
3714 when llog_commit threads being stopped to avoid accessing some
3717 Severity : enhancement
3719 Description: Add printing inode info into message about error in writepage.
3721 --------------------------------------------------------------------------------
3723 2007-05-03 Cluster File Systems, Inc. <info@clusterfs.com>
3728 Frequency : on some architectures
3730 Description: 1.6 client sometimes fails to mount from a 1.4 MDT
3731 Details : Uninitialized flags sometimes cause configuration commands to
3735 Frequency : patchless clients only
3737 Description: missing __iget() symbol export
3738 Details : The __iget() symbol export is missing. To avoid the need for
3739 this on patchless clients the deathrow inode reaper is turned
3740 off, and we depend on the VM to clean up old inodes. This
3741 dependency was during via the fix for bug 12181.
3746 Description: sanity.sh fail: test_52b
3747 Details : The ll_inode_to_ext_flags() has a glitch which makes MDS return
3748 incorrect inode's flags to client.
3750 --------------------------------------------------------------------------------
3752 2007-04-19 Cluster File Systems, Inc. <info@clusterfs.com>
3754 * CONFIGURATION CHANGE. This version of Lustre WILL NOT
3755 INTEROPERATE with older versions automatically. In many cases a
3756 special upgrade step is needed. Please read the
3757 user documentation before upgrading any part of a 1.4.x system.
3758 * WARNING: Lustre configuration and startup changes are required with
3759 this release. See https://mail.clusterfs.com/wikis/lustre/MountConf
3761 * Support for kernels:
3762 2.4.21-47.0.1.EL (RHEL 3),
3763 2.6.5-7.283 (SLES 9),
3764 2.6.9-42.0.10.EL (RHEL 4),
3765 2.6.12.6 vanilla (kernel.org),
3766 2.6.16.27-0.9 (SLES10)
3767 * Client support for unpatched kernels:
3768 (see https://mail.clusterfs.com/wikis/lustre/PatchlessClient)
3769 2.6.16 - 2.6.19 vanilla (kernel.org),
3770 2.6.9-42.0.8EL (RHEL 4)
3771 * Recommended e2fsprogs version: 1.39.cfs6
3772 * Note that reiserfs quotas are disabled on SLES 10 in this kernel
3775 Severity : enhancement
3777 Description: Async OSC create to avoid the blocking unnecessarily.
3778 Details : If a OST has no remain object, system will block on the creating
3779 when need to create a new object on this OST. Now, ways use
3780 pre-created objects when available, instead of blocking on an
3781 empty osc while others are not empty. If we must block, we block
3782 for the shortest possible period of time.
3784 Severity : enhancement
3786 Description: MountConf
3787 Details : Lustre configuration is now managed via mkfs and mount
3788 commands instead of lmc and lconf. New obd types (MGS, MGC)
3789 are added for dynamic configuration management. See
3790 https://mail.clusterfs.com/wikis/lustre/MountConf for
3793 Severity : enhancement
3795 Description: dynamic OST addition
3796 Details : OSTs can now be added to a live filesystem
3798 Severity : enhancement
3800 Description: startup order invariance
3801 Details : MDTs and OSTs can be started in any order. Clients only
3802 require the MDT to complete startup.
3804 Severity : enhancement
3806 Description: parallel, asynchronous orphan cleanup
3807 Details : orphan cleanup is now performed in separate threads for each
3808 OST, allowing parallel non-blocking operation.
3810 Severity : enhancement
3812 Description: optimized stripe assignment
3813 Details : stripe assignments are now made based on ost space available,
3814 ost previous usage, and OSS previous usage, in order to try
3815 to optimize storage space and networking resources.
3817 Severity : enhancement
3819 Description: Permanently set tunables
3820 Details : All writable /proc/fs/lustre tunables can now be permanently
3821 set on a per-server basis, at mkfs time or on a live
3824 Severity : enhancement
3826 Description: Lustre message v2
3827 Details : Add lustre message format v2.
3829 Severity : enhancement
3831 Description: client OST exclusion list
3832 Details : Clients can be started with a list of OSTs that should be
3833 declared "inactive" for known non-responsive OSTs.
3837 Description: ENOENT returned for valid filehandle during dbench.
3838 Details : Check if a directory has children when invalidating dentries
3839 associated with an inode during lock cancellation. This fixes
3840 an incorrect ENOENT sometimes seen for valid filehandles during
3841 testing with dbench.
3844 Frequency : SFS test only (otherwise harmless)
3846 Description: SPEC SFS validation failure on NFS v2 over lustre.
3847 Details : Changes the blocksize for regular files to be 2x RPC size,
3848 and not depend on stripe size.
3850 Severity : enhancement
3852 Description: fine-grained SMP locking inside DLM
3853 Details : Improve DLM performance on SMP systems by removing the single
3854 per-namespace lock and replace it with per-resource locks.
3856 Severity : enhancement
3858 Description: don't hold multiple extent locks at one time
3859 Details : To avoid client eviction during large writes, locks are not
3860 held on multiple stripes at one time or for very large writes.
3861 Otherwise, clients can block waiting for a lock on a failed OST
3862 while holding locks on other OSTs and be evicted.
3864 Severity : enhancement
3866 Description: Multiple MD RPCs in flight.
3867 Details : Further unserialise some read-only MDT RPCs - learn about intents.
3868 To avoid overly-overloading MDT, introduce a limit on number of
3869 MDT RPCs in flight for a single client and add /proc controls
3870 to adjust this limit.
3872 Severity : enhancement
3874 Description: client read/write statistics
3875 Details : Add client read/write call usage stats for performance
3876 analysis of user processes.
3877 /proc/fs/lustre/llite/*/offset_stats shows non-sequential
3878 file access. extents_stats shows chunk size distribution.
3879 extents_stats_per_process show chunk size distribution per
3882 Severity : enhancement
3884 Description: per-client statistics on server
3885 Details : Add ldlm and operations statistics for each client in
3886 /proc/fs/lustre/mds|obdfilter/*/exports/
3888 Severity : enhancement
3890 Description: improved MDT statistics
3891 Details : Add detailed MDT operations statistics in
3892 /proc/fs/lustre/mds/*/stats
3894 Severity : enhancement
3896 Description: VFS operations stats
3897 Details : Add client VFS call stats, trackable by pid, ppid, or gid
3898 /proc/fs/lustre/llite/*/stats_track_[pid|ppid|gid]
3903 Description: Fix client-side osc byte counters
3904 Details : The osc read/write byte counters in
3905 /proc/fs/lustre/osc/*/stats are now working
3908 Frequency : always as root on SLES
3910 Description: Failure of copying files with lustre special EAs.
3911 Details : Client side always return success for setxattr call for lustre
3912 special xattr (currently only "trusted.lov").
3917 Description: Refcount LNET uuids
3918 Details : The global LNET uuid list grew linearly with every startup;
3919 refcount repeated list entries instead of always adding to
3922 Severity : enhancement
3924 Description: Dynamic service threads
3925 Details : Within a small range, start extra service threads
3926 automatically when the request queue builds up.
3929 Frequency : mixed-endian client/server environments
3931 Description: mixed-endian crashes
3932 Details : The new msg_v2 system had some failures in mixed-endian
3935 Severity : enhancement
3937 Description: Easy OST removal
3938 Details : OSTs can be permanently deactivated with e.g. 'lctl
3939 conf_param lustre-OST0001.osc.active=0'
3941 Severity : enhancement
3943 Description: MGS proc entries
3944 Details : Added basic proc entries for the MGS showing what filesystems
3947 Severity : enhancement
3949 Description: provide MGS failover
3950 Details : Added config lock reacquisition after MGS server failover.
3952 Severity : enhancement
3954 Description: add Linux 2.4 support
3955 Details : Added support for RHEL 2.4.21 kernel for 1.6 servers and clients
3959 Description: a large application tries to do I/O to the same resource and dies
3960 in the middle of it.
3961 Details : Check the req->rq_arrival time after the call to
3962 ost_brw_lock_get(), but before we do anything about
3963 processing it & sending the BULK transfer request. This
3964 should help move old stale pending locks off the queue as
3965 quickly as obd_timeout.
3968 Frequency : when an incorrect nid is specified during startup
3970 Description: ptlrpc connect to non-existant node causes kernel crash
3971 Details : LNET can't be re-entered from an event callback, which
3972 happened when we expire a message after the export has been
3973 cleaned up. Instead, hand the zombie cleanup off to another
3976 Severity : enhancement
3978 Description: plain/inodebits lock performance improvement
3979 Details : Grouping plain/inodebits in granted list by their request modes
3980 and bits policy, thus improving the performance of search through
3984 Frequency : only if OST filesystem is corrupted
3986 Description: client incorrectly hits assertion in ptlrpc_replay_req()
3987 Details : for a short time RPCs with bulk IO are in the replay list,
3988 but replay of bulk IOs is unimplemented. If the OST filesystem
3989 is corrupted due to disk cache incoherency and then replay is
3990 started it is possible to trip an assertion. Avoid putting
3991 committed RPCs into the replay list at all to avoid this issue.
3994 Frequency : liblustre (e.g. catamount) on a large cluster with >= 8 OSTs/OSS
3996 Description: System hang on startup
3997 Details : This bug allowed the liblustre (e.g. catamount) client to
3998 return to the app before handling all startup RPCs. This
3999 could leave the node unresponsive to lustre network traffic
4000 and manifested as a server ptllnd timeout.
4002 Severity : enhancement
4004 Description: Add "/proc/sys/lustre/debug_peer_on_timeout"
4005 Details : liblustre envirable: LIBLUSTRE_DEBUG_PEER_ON_TIMEOUT
4006 boolean to control whether to print peer debug info when a
4007 client's RPC times out.
4010 Frequency : only for kernels with patches from Lustre below 1.4.3
4012 Description: Remove old rdonly API
4013 Details : Remove old rdonly API which unused from at least lustre 1.4.3
4016 Frequency : only for devices with external journals
4018 Description: Set external device read-only also
4019 Details : During a commanded failover stop, we set the disk device
4020 read-only while the server shuts down. We now also set any
4021 external journal device read-only at the same time.
4024 Frequency : when upgrading from 1.4 while trying to change parameters
4026 Description: The wrong (new) MDC name was used when setting parameters for
4027 upgraded MDT's. Also allows changing of OSC (and MDC)
4028 parameters if --writeconf is specified at tunefs upgrade time.
4031 Frequency : when setting specific ost indicies
4033 Description: QOS code breaks on skipped indicies
4034 Details : Add checks for missing OST indicies in the QOS code, so OSTs
4035 created with --index need not be sequential.
4037 Severity : enhancement
4039 Description: Add uninit_groups feature to ldiskfs2 to speed up e2fsck
4040 Details : The uninit_groups feature works in conjunction with the kernel
4041 filesystem code (ldiskfs2 only) and e2fsprogs-1.39-cfs6 to speed
4042 up the pass1 processing of e2fsck. This is a read-only feature
4043 in ldiskfs2 only, so older kernels and current ldiskfs cannot
4044 mount filesystems that have had this feature enabled.
4046 Severity : enhancement
4048 Description: Improve multi-block allocation algorithm to avoid fragmentation
4049 Details : The mballoc3 code (ldiskfs2 only) adds new mechanisms to improve
4050 allocation locality and avoid filesystem fragmentation.
4052 ------------------------------------------------------------------------------
4054 2007-04-01 Cluster File Systems, Inc. <info@clusterfs.com>
4056 * Support for kernels:
4057 2.4.21-47.0.1.EL (RHEL 3)
4058 2.6.5-7.283 (SLES 9)
4059 2.6.9-42.0.10.EL (RHEL 4)
4060 2.6.12.6 vanilla (kernel.org)
4061 2.6.16.27-0.9 (SLES 10)
4062 * Recommended e2fsprogs version: 1.39.cfs5
4064 * Note that reiserfs quotas are disabled on SLES 10 in this kernel
4068 Frequency : occasional, depends on client load and configuration
4069 Bugzilla : 12181, 12203
4070 Description: data loss for recently-modified files
4072 Details : In some cases it is possible that recently written or created
4073 files may not be written to disk in a timely manner (this should
4074 normally be within 30s unless client IO load is very high).
4075 The problem appears as zero-length files or files that are a
4076 multiple of 1MB in size after a client crash or client eviction
4077 that are missing data at the end of the file.
4079 This problem is more likely to be hit on clients where files are
4080 repeatedly created and unlinked in the same directory, clients
4081 have a large amount of RAM, have many CPUs, the filesystem has
4082 many OSTs, the clients are rebooted frequently, and/or the files
4083 are not accessed by other nodes after being written.
4085 The presence of the problem can be detected by looking at
4086 /proc/sys/fs/inode-state. If the first number (nr_inodes) is
4087 smaller than the second (nr_unused) then dirty files will not
4088 be flushed automatically to disk. "sync; sleep 10" should be
4089 run several times on the node before unmounting it to update
4090 Lustre (this is also safe to run on nodes without this problem).
4092 There is also a related kernel bug in the RHEL4 4 2.6.9 kernel
4093 that can cause this same problem, so customers using that kernel
4094 also need to update the kernel in addition to Lustre. In order
4095 to properly fix this bug, the RHEL3 2.4.21 kernel is also updated.
4097 It is normal that files written just before a client crash (less
4098 than 30s) may not yet have been flushed to disk, even for local
4102 Frequency : frequent on thin XT3 nodes
4104 Description: UUID collision on thin XT3 Linux nodes
4105 Details : UUIDs on Compute Node Linux XT3 nodes were not generated
4106 randomly, since we relied on an insufficiently-seeded PRNG.
4111 Description: OSS hangs after "All ost request buffers busy"
4112 Details : A deadlock between quota and journal operations caused OSS
4113 hangs after printing "All ost request buffers busy."
4116 Frequency : always on liblustre builds
4118 Description: Cleanup compiler warnings on liblustre
4121 Frequency : always on liblustre builds on XT3
4123 Description: LC_CONFIG_CDEBUG don't run while build liblustre on XT3.
4127 Description: Addition of EXT3_FEATURE_RO_COMPAT_DIR_NLINKS flag for
4128 > 32000 subdirectories
4129 Details : Add EXT3_FEATURE_RO_COMPAT_DIR_NLINK flag to
4130 EXT3_FEATURE_RO_COMPAT_SUPP. This flag will be set whenever
4131 subdirectory count crosses 32000. This will aid e2fsck to
4132 correctly handle more than 32000 subdirectories.
4135 Frequency : liblustre (e.g. catamount) on a large cluster with >= 8 OSTs/OSS
4137 Description: System hang on startup
4138 Details : This bug allowed the liblustre (e.g. catamount) client to
4139 return to the app before handling all startup RPCs. This
4140 could leave the node unresponsive to lustre network traffic
4141 and manifested as a server ptllnd timeout.
4143 Severity : enhancement
4145 Description: Add "/proc/sys/lustre/debug_peer_on_timeout"
4146 (liblustre envirable: LIBLUSTRE_DEBUG_PEER_ON_TIMEOUT)
4147 boolean to control whether to print peer debug info when a
4148 client's RPC times out.
4153 Description: make O_SYNC working on 2.6 kernels
4154 Details : 2.6 kernels use different method for mark pages for write,
4155 so need add a code to lustre for O_SYNC work.
4160 Description: Failure to close file and release space on NFS
4161 Details : Put inode details into lock acquired in ll_intent_file_open.
4162 Use mdc_intent_lock in ll_intent_open to properly
4163 detect all kind of errors unhandled by mdc_enqueue.
4168 Description: proc file read during shutdown sometimes raced obd removal,
4170 Details : Add lock to prevent obd access after proc file removal.
4173 Frequency : Only for files larger than 4GB on 32-bit clients.
4175 Description: improperly doing page alignment of locks
4176 Details : Modify lustre core code to use CFS_PAGE_* defines instead of
4177 PAGE_*. Make CFS_PAGE_MASK a 64-bit mask.
4182 Description: RPCs being resent when they shouldn't be
4183 Details : Some RPCs that should not be resent are being resent. This
4184 can cause inconsistencies in the RPC state machine. Do not
4185 resend such requests.
4188 Frequency : rare, only with NFS export
4190 Description: Crash on NFS re-export node
4191 Details : under very unusual load conditions an assertion is hit in
4192 ll_intent_file_open()
4195 Frequency : only if OST filesystem is corrupted
4197 Description: client incorrectly hits assertion in ptlrpc_replay_req()
4198 Details : for a short time RPCs with bulk IO are in the replay list,
4199 but replay of bulk IOs is unimplemented. If the OST filesystem
4200 is corrupted due to disk cache incoherency and then replay is
4201 started it is possible to trip an assertion. Avoid putting
4202 committed RPCs into the replay list at all to avoid this issue.
4207 Description: large O_DIRECT requests fail under memory pressure/fragmentation
4208 Details : Large single O_DIRECT read and write calls can fail to allocate
4209 a sufficiently large buffer to process the request. In case of
4210 allocation failure the allocation is retried with a smaller
4211 buffer and broken into smaller requests.
4213 Severity : enhancement
4215 Description: Add -o localflock option to simulate old noflock behaviour.
4216 Details : This will achieve local-only flock/fcntl locks coherentness.
4221 Description: versioning check is incomplete
4222 Details : Checking the version difference of client vs. server, report
4223 error if the gap is too big.
4228 Description: add support PG_writeback bit
4229 Details : add support for PG_writeback bit for Lustre, for more carefull
4230 work with page cache in 2.6 kernel. This also fix some deadlocks
4231 and remove hack for work O_SYNC with 2.6 kernel.
4233 Severity : enhancement
4235 Description: Add uninit_groups feature to ldiskfs2 to speed up e2fsck
4236 Details : The uninit_groups feature works in conjunction with the kernel
4237 filesystem code (ldiskfs2 only) and e2fsprogs-1.39-cfs6 to speed
4238 up the pass1 processing of e2fsck. This is a read-only feature
4239 in ldiskfs2 only, so older kernels and current ldiskfs cannot
4240 mount filesystems that have had this feature enabled.
4242 Severity : enhancement
4244 Description: Improve multi-block allocation algorithm to avoid fragmentation
4245 Details : The mballoc3 code (ldiskfs2 only) adds new mechanisms to improve
4246 allocation locality and avoid filesystem fragmentation.
4248 ------------------------------------------------------------------------------
4250 2007-02-09 Cluster File Systems, Inc. <info@clusterfs.com>
4252 * Support for kernels:
4253 2.6.9-42.0.3.EL (RHEL 4)
4254 2.6.5-7.276 (SLES 9)
4255 2.4.21-47.0.1.EL (RHEL 3)
4256 2.6.12.6 vanilla (kernel.org)
4257 2.6.16.21-0.8 (SLES10)
4258 * Recommended e2fsprogs version: 1.39.cfs2-0
4260 * The backwards-compatible /proc/sys/portals symlink has been removed
4261 in this release. Before upgrading, please ensure that you change
4262 any configuration scripts or /etc/sysctl.conf files that access
4263 /proc/sys/portals/* or sysctl portals.* to use the corresponding
4264 entry in /proc/sys/lnet or sysctl lnet.*. This change can be made
4265 in advance of the upgrade on any system running Lustre 1.4.6 or
4266 newer, since /proc/sys/lnet was added in that version.
4267 * Note that reiserfs quotas are disabled on SLES 10 in this kernel
4271 Frequency : only when quota is used
4273 Description: avoid scanning export list for quota master
4274 Details : Change the algorithms to avoid scanning export list in order
4275 to improve the efficiency.
4278 Frequency : MDS failover only, very rarely
4280 Description: "went back in time" messages on mds failover
4281 Details : The greatest transno may be lost when the current operation
4282 finishes with an error (transno==0) and the client's last_rcvd
4283 record is over-written. Save the greatest transno in the
4284 mds_last_transno for this case.
4287 Frequency : always for specific kernels and striping counts
4289 Description: client may get "Matching packet too big" without ACL support
4290 Details : Clients compiled without CONFIG_FS_POSIX_ACL get an error message
4291 when trying to access files in certain configurations. The
4292 clients should in fact be denied when mounting because they do
4293 not understand ACLs.
4296 Frequency : Cray XT3 with more than 4000 clients and multiple jobs
4298 Description: many clients connecting with IO in progress causes connect timeouts
4299 Details : Avoid synchronous journal commits to avoid delays caused by many
4300 clients connecting/disconnecting when bulk IO is in progress.
4301 Queue liblustre connect requests on OST_REQUEST_PORTAL instead of
4302 OST_IO_PORTAL to avoid delays behind potentially many pending
4306 Frequency : occasionally with multiple writers to a single file
4308 Description: shared writes to file may result in wrong size reported by stat()
4309 Details : Allow growing of kms when extent lock is cancelled
4312 Frequency : always with random mmap IO to multi-striped file
4314 Description: mmap write might be lost if we are writing to a 'hole' in stripe
4315 Details : Only if the hole is at the end of OST object so that kms is too
4316 small. Fix is to increase kms accordingly in ll_nopage.
4319 Frequency : rare, only if OST filesystem is inconsistent with MDS filesystem
4321 Description: writes to a missing object would leak memory on the OST
4322 Details : If there is an inconsistency between the MDS and OST filesystems,
4323 such that the MDS references an object that doesn't exist, writes
4324 to that object will leak memory due to incorrect cleanup in the
4325 error handling path, eventually running out of memory on the OST.
4330 Description: Creating too long symlink causes lustre errors
4331 Details : Check symlink and name lengths before sending requests to MDS.
4334 Frequency : only if flock is enabled (not on by default)
4336 Description: posix locks not released on fd closure on 2.6.9+
4337 Details : We failed to add posix locks to list of inode locks on 2.6.9+
4338 kernels, this caused such locks not to be released on fd close and
4339 then assertions on fs unmount about still used locks.
4342 Frequency : MDS failover only, very rarely
4344 Description: clients may get ASSERTION(granted_lock != NULL)
4345 Details : When request was taking a long time, and a client was resending
4346 a getattr by name lock request. The were multiple lock requests
4347 with the same client lock handle and
4348 mds_getattr_name->fixup_handle_for_resent_request found one of the
4349 lock handles but later failed with ASSERTION(granted_lock != NULL).
4354 Description: handle->h_buffer_credits > 0, assertion failure
4355 Details : h_buffer_credits is zero after truncate, causing assertion
4356 failure. This patch extends the transaction or creates a new
4360 Frequency : NFS re-export or patchless client
4361 Bugzilla : 11179, 10796
4362 Description: Crash on NFS re-export node (__d_move)
4363 Details : We do not want to hash the dentry if we don't have a lock.
4364 But if this dentry is later used in d_move, we'd hit uninitialised
4365 list head d_hash, so we just do this to init d_hash field but
4366 leave dentry unhashed.
4369 Frequency : NFS re-export or patchless client
4371 Description: NFS exports has problem with symbolic link
4372 Details : lustre client didn't properly install dentry when re-exported
4373 to NFS or running patchless client.
4376 Frequency : NFS re-export or patchless client
4378 Description: Various nfs/patchless fixes.
4379 Details : fixes reuse disconected alias for lookup process - this fixes
4380 warning "find_exported_dentry: npd != pd",
4381 fix permission error with open files at nfs.
4382 fix apply umask when do revalidate.
4385 Frequency : occasional
4387 Description: Crash on NFS re-export node
4388 Details : calling clear_page() on the wrong pointer triggered oops in
4389 generic_mapping_read().
4392 Frequency : rarely, using O_DIRECT IO
4394 Description: unaligned directio crashes client with LASSERT
4395 Details : check for unaligned buffers before trying any requests.
4398 Frequency : rarely, using CFS RAID5 patches in non-standard kernel series
4400 Description: stale data returned from RAID cache
4401 Details : If only a small amount of IO is done to the RAID device before
4402 reading it again it is possible to get stale data from the RAID
4403 cache instead of reading it from disk.
4406 Frequency : always for sles10 kernel
4408 Description: sles10 support
4409 Details : ll_follow_link: compile fixes and using of nd_set_link
4410 under newer kernels.
4413 Frequency : depends on arch, kernel and compiler version, always on sles10
4416 Description: recursive or deep enough symlinks cause stack overflow
4417 Details : getting rid of large stack-allocated variable in
4421 Frequency : depends on hardware
4423 Description: lustre write performance loss in the SLES10 kernel
4424 Details : the performance loss is caused by using of write barriers in the
4425 ext3 code. The SLES10 kernel turns barrier support on by
4426 default. The fix is to undo that change for ldiskfs.
4428 ------------------------------------------------------------------------------
4430 2006-12-09 Cluster File Systems, Inc. <info@clusterfs.com>
4432 * Support for kernels:
4433 2.6.9-42.0.3EL (RHEL 4)
4434 2.6.5-7.276 (SLES 9)
4435 2.4.21-47.0.1.EL (RHEL 3)
4436 2.6.12.6 vanilla (kernel.org)
4440 Frequency : quota enabled and large files being deleted
4442 Description: releasing more than 4GB of quota at once hangs OST
4443 Details : If a user deletes more than 4GB of files on a single OST it
4444 will cause the OST to spin in an infinite loop. Release
4445 quota in < 4GB chunks, or use a 64-bit value for 1.4.7.1+.
4450 Description: statfs data retrieved from /proc may be stale or zero
4451 Details : When reading per-device statfs data from /proc, in the
4452 {kbytes,files}_{total,free,avail} files, it may appear
4453 as zero or be out of date.
4456 Frequency : systems with MD RAID1 external journal devices
4458 Description: lconf's call to blkid is confused by RAID1 journal devices
4459 Details : Use the "blkid -l" flag to locate the MD RAID device instead
4460 of returning all block devices that match the journal UUID.
4463 Frequency : always, for aggregate stripe size over 4GB
4465 Description: "lfs setstripe" fails assertion when setting 4GB+ stripe width
4466 Details : Using "lfs setstripe" to set stripe size * stripe count over 4GB
4467 will fail the kernel with "ASSERTION(lsm->lsm_xfersize != 0)"
4470 Frequency : always if "lfs find" used on a local file/directory
4472 Description: "lfs find" segfaults if used on a local file/directory
4473 Details : The case where a directory component was not specified wasn't
4474 handled correctly. Handle this properly.
4477 Frequency : always on ppc64
4479 Description: the write to an ext3 filesystem mounted with mballoc got stuck
4480 Details : ext3_mb_generate_buddy() uses find_next_bit() which does not
4481 perform endianness conversion.
4484 Frequency : rarely (truncate to non-zero file size after write under load)
4485 Bugzilla : 10730, 10687
4486 Description: Files padded with zeros to next 4K multiple
4487 Details : With filesystems mounted using the "extents" option (2.6 kernels)
4488 it is possible that files that are truncated to a non-zero size
4489 immediately after being written are filled with zero bytes beyond
4490 the truncated size. No file data is lost.
4492 Severity : enhancement
4494 Description: Allow recovery/failover for liblustre clients.
4495 Details : liblustre clients were unaware of failover configurations until
4498 Severity : enhancement
4500 Description: user file locks should fail when not mounting with flock option
4501 Details : Set up an error-returning stub in ll_file_operations.lock field
4502 to prevent incorrect behaviour when client is mounted without
4503 flock option. Also, set up properly f_op->flock field for
4507 Frequency : always on ia64
4509 Description: "lfs df" loops on printing out MDS statfs information
4510 Details : The obd_ioctl_data was not initialized and in some systems
4511 this caused a failure during the ioctl that did not return
4512 an error. Initialize the struct and return an error on failure.
4515 Frequency : SLES 9 only
4517 Description: Error of copying files with lustre special EAs as root
4518 Details : Client side always return success for setxattr call for lustre
4519 special xattr (currently only "trusted.lov").
4522 Frequency : rarely on clusters with both ia64+i386 clients
4524 Description: ia64+i686 clients doing shared IO on the same file may LBUG
4525 Details : In rare cases when both ia64+i686 (or other mixed-PAGE_SIZE)
4526 clients are doing concurrent writes to the same file it is
4527 possible that the ia64 clients may LASSERT because the OST
4528 extent locks are not PAGE_SIZE aligned. Ensure that grown
4529 locks are always aligned on the request boundary.
4532 Frequency : specific use, occasional
4534 Description: Overwriting in use executable truncates on-disk binary image
4535 Details : If one node attempts to overwrite an executable in use by
4536 another node, we now correctly return ETXTBSY instead of
4537 truncating the file.
4539 Severity : enhancement
4541 Description: Async OSC create to avoid the blocking unnecessarily.
4542 Details : If a OST has no remain object, system will block on the creating
4543 when need to create a new object on this OST. Now, ways use
4544 pre-created objects when available, instead of blocking on an
4545 empty osc while others are not empty. If we must block, we block
4546 for the shortest possible period of time.
4551 Description: chmod on Lustre root is propagated to other clients
4552 Details : Re-validate root's dentry in ll_lookup_it to avoid having it
4553 invalid by the follow_mount time.
4556 Frequency : liblustre clients only
4558 Description: Race in 'instant cancel' lock handling could lead to such locks
4559 never to be granted in case of SMP MDS
4560 Details : Do not destroy not yet granted but cbpending locks in
4564 Frequency : replay/resend of open
4566 Description: non null lock assetion failure in mds_intent_policy
4567 Details : Trying to replay/resend lockless open requests resulted in
4568 mds_open() returning 0 with no lock. Now it sets a flag if
4569 a lock is going to be returned.
4571 Severity : enhancement
4573 Description: Checksum enhancements
4574 Details : New checksum enhancements allow for resending RPCs that failed
4577 Severity : enhancement
4579 Description: Tunables on number of dirty pages in cacche
4580 Details : Allow to set limit on number of dirty pages cached.
4585 Description: client crash on unmount - lock still has references
4586 Details : In some error handling cases it was possible to leak a lock
4587 reference on a client while accessing a file. This was not
4588 harmful to the client during operation, but would cause the
4589 client to crash when the filesystem is unmounted.
4592 Frequency : specific case, rare
4594 Description: ETXTBSY on mds though file not in use by client
4595 Details : ETXTBSY is no longer incorrectly returned when attempting to
4596 chmod or chown a directory that the user previously tried to
4597 execute or a currently-executing binary.
4600 Frequency : extremely rare except on liblustre-based clients
4602 Description: Lustre space not freed when files are deleted
4603 Details : Clean up open-unlinked files after client eviction. Previously
4604 the unlink was skipped and the files remained as orphans.
4609 Description: OST failure "would be an LBUG" in waiting_locks_callback()
4610 Details : In some cases it was possible to send a blocking callback to
4611 a client doing a glimpse, even though that client didn't get
4612 a lock granted. When the glimpse lock is cancelled on the OST
4613 the freed lock is left on the waiting list and corrupted the list.
4616 Frequency : all core dumps
4618 Description: Broke core dumps to lustre
4619 Details : Negative dentry may be unhashed if parent does not have UPDATE
4620 lock, but some callers, e.g. do_coredump, expect dentry to be
4621 hashed after successful create, hash it in ll_create_it.
4623 ------------------------------------------------------------------------------
4625 2006-09-13 Cluster File Systems, Inc. <info@clusterfs.com>
4627 * Support for kernels:
4628 2.6.9-42.0.2.EL (RHEL 4)
4629 2.6.5-7.276 (SLES 9)
4630 2.4.21-40.EL (RHEL 3)
4631 2.6.12.6 vanilla (kernel.org)
4635 Frequency : always on RHEL 3
4637 Description: Number of open files grows over time
4638 Details : The number of open files grows over time, whether or not
4639 Lustre is started. This was due to a filp leak introduced
4640 by one of our kernel patches.
4642 ------------------------------------------------------------------------------
4644 2006-08-20 Cluster File Systems, Inc. <info@clusterfs.com>
4646 * Support for kernels:
4647 2.6.9-42.EL (RHEL 4)
4648 2.6.5-7.267 (SLES 9)
4649 2.4.21-40.EL (RHEL 3)
4650 2.6.12.6 vanilla (kernel.org)
4655 Bugzilla : 5719, 9635, 9792, 9684
4656 Description: OST (or MDS) trips assertions in (re)connection under heavy load
4657 Details : If a server is under heavy load and cannot reply to new
4658 connection requests before the client resends the (re)connect,
4659 the connection handling code can behave badly if two service
4660 threads are concurrently handing separate (re)connections from
4661 the same client. Add better locking to the connection handling
4662 code, and ensure that only a single connection will be processed
4663 for a given client UUID, even if the lock is dropped.
4665 Severity : enhancement
4667 Description: add TCP zero-copy support to kernel
4668 Details : Add support to the kernel TCP stack to allow zero-copy bulk
4669 sends if the hardware supports scatter-gather and checksumming.
4670 This allows socklnd to do client-write and server-read more
4671 efficiently and reduce CPU utilization from skbuf copying.
4674 Frequency : only if NFS exporting from client
4676 Description: NULL pointer deref in ll_iocontrol() if chattr mknod file
4677 Details : If setting attributes on a file created under NFS that had
4678 never been opened it would be possible to oops the client
4679 if the file had no objects.
4683 Bugzilla : 9326, 10402, 10897
4684 Description: client crash in ptlrpcd_wake() thread when sending async RPC
4685 Details : It is possible that ptlrpcd_wake() dereferences a freed async
4686 RPC. In rare cases the ptlrpcd thread alread processed the RPC
4687 before ptlrpcd_wake() was called and the request was freed.
4690 Frequency : always for liblustre
4692 Description: liblustre client does MDS+OSTs setattr RPC for each write
4693 Details : When doing a write from a liblustre client, the client
4694 incorrectly issued an RPC to the MDS and each OST the file was
4695 striped over in order to update the timestamps. When writing
4696 with small chunks and many clients this could overwhelm the MDS
4697 with RPCs. In all cases it would slow down the write because
4698 these RPCs are unnecessary.
4700 Severity : enhancement
4702 Description: allow number of MDS service threads to be changed at module load
4703 Details : It is now possible to change the number of MDS service threads
4704 running. Adding "options mds mds_num_threads={N}" to the MDS's
4705 /etc/modprobe.conf will set the number of threads for the next
4706 time Lustre is restarted (assuming the "mds" module is also
4707 reloaded at that time). The default number of threads will
4708 stay the same, 32 for most systems.
4713 Description: OST crash if filesystem is unformatted or corrupt
4714 Details : If an OST is started on a device that has never been formatted
4715 or if the filesystem is corrupt and cannot even mount then the
4716 error handling cleanup routines would dereference a NULL pointer.
4721 Description: NULL pointer deref in llap_from_page.
4722 Details : get_cache_page_nowait can return a page with NULL (or otherwise
4723 incorrect) mapping if the page was truncated/reclaimed while it was
4724 searched for. Check for this condition and skip such pages when
4725 doing readahead. Introduce extra check to llap_from_page() to
4726 verify page->mapping->host is non-NULL (so page is not anonymous).
4729 Frequency : Sometimes when using sys_sendfile
4731 Description: "page not covered by a lock" warnings from ll_readpage
4732 Details : sendfile called ll_readpage without right page locks present.
4733 Now we introduced ll_file_sendfile that does necessary locking
4734 around call to generic_file_sendfile() much like we do in
4738 Frequency : with certain MDS communication failures at client mount time
4740 Description: NULL pointer deref after failed client mount
4741 Details : a client connection request may delayed by the network layer
4742 and not be sent until after the PTLRPC layer has timed out the
4743 request. If the client fails the mount immediately it will try
4744 to clean up before the network times out the request. Add a
4745 reference from the request import to the obd device and delay
4746 the cleanup until the network drops the request.
4749 Frequency : occasionally during client (re)connect
4751 Description: assertion failure during client (re)connect
4752 Details : processing a client connection request may be delayed by the
4753 client or server longer than the client connect timeout. This
4754 causes the client to resend the connection request. If the
4755 original connection request is replied in this interval, the
4756 client may trip an assertion failure in ptlrpc_connect_interpret()
4757 which thought it would be the only running connect process.
4760 Frequency : only with obd_echo servers and clients that are rebooted
4762 Description: kernel BUG accessing uninitialized data structure
4763 Details : When running an obd_echo server it did not start the ping_evictor
4764 thread, and when a client was evicted an uninitialized data
4765 structure was accessed. Start the ping_evictor in the RPC
4766 service startup instead of the OBD startup.
4768 Severity : enhancement
4769 Bugzilla : 10193 (patchless)
4770 Description: Remove dependency on various unexported kernel interfaces.
4771 Details : No longer need reparent_to_init, exit_mm, exit_files,
4772 sock_getsockopt, filemap_populate, FMODE_EXEC, put_filp.
4775 Frequency : rare (only users of deprecated and unsupported LDAP config)
4777 Description: write_conf for zeroconf mount queried LDAP incorrectly for client
4778 Details : LDAP apparently contains 'lustreName' attributes instead of
4779 'name'. A simple remapping of the name is sufficient.
4782 Frequency : rare (only with non-default dump_on_timeout debug enabled)
4784 Description: waiting_locks_callback trips kernel BUG if client is evicted
4785 Details : Running with the dump_on_timeout debug flag turned on makes
4786 it possible that the waiting_locks_callback() can try to dump
4787 the Lustre kernel debug logs from an interrupt handler. Defer
4788 this log dumping to the expired_lock_main() thread.
4790 Severity : enhancement
4792 Description: Support NFS exporting on 2.6 kernels.
4793 Details : Implement non-rawops metadata methods for NFS server to use without
4794 changing NFS server code.
4797 Frequency : very rare (synthetic metadata workload only)
4799 Description: two racing renames might cause an MDS thread to deadlock
4800 Details : Running the "racer" program may cause one MDS thread to rename
4801 a file from being the source of a rename to being the target of
4802 a rename at exactly the same time that another thread is doing
4803 so, and the second thread has already enqueued these locks after
4804 doing a lookup of the target and is trying to relock them in
4805 order. Ensure that we don't try to re-lock the same resource.
4808 Frequency : only very large systems with liblustre clients
4810 Description: slow eviction of liblustre clients with the "evict_by_nid" RPC
4811 Details : Use asynchronous set_info RPCs to send the "evict_by_nid" to
4812 all OSTs in parallel. This allows the eviction of stale liblustre
4813 clients to proceed much faster than if they were done in series,
4814 and also offers similar improvements for other set_info RPCs.
4819 Description: excessive CPU usage during initial read phase on client
4820 Details : During the initial read phase on a client, it would agressively
4821 retry readahead on the file, consuming too much CPU and impacting
4822 performance (since 1.4.5.8). Improve the readahead algorithm
4823 to avoid this, and also improve some other common cases (read
4824 of small files in particular, where "small" is files smaller than
4825 /proc/fs/lustre/llite/*/max_read_ahead_whole_mb, 2MB by default).
4830 Description: MDS crash when receiving packet with unknown intent.
4831 Details : Do not LBUG in unknown intent case, just return -EFAULT
4833 Severity : enhancement
4834 Bugzilla : 9293, 9385
4835 Description: MDS RPCs are serialised on client. This is unnecessary for some.
4836 Details : Do not serialize getattr (non-intent version) and statfs.
4839 Frequency : occasional, when OST network is overloaded/intermittent
4841 Description: client evicted by OST after bulk IO timeout
4842 Details : If a client sends a bulk IO request (read or write) the OST
4843 may evict the client if it is unresposive to its data GET/PUT
4844 request. This is incorrect if the network is overloaded (takes
4845 too long to transfer the RPC data) or dropped the OST GET/PUT
4846 request. There is no need to evict the client at all, since
4847 the pinger and/or lock callbacks will handle this, and the
4848 client can restart the bulk request.
4851 Frequency : Always when mmapping file with no objects
4853 Description: client crashes when mmapping file with no objects
4854 Details : Check that we actually have objects in a file before doing any
4855 operations on objects in ll_vm_open, ll_vm_close and
4861 Description: Request leak when working with deleted CWD
4862 Details : Introduce advanced request refcount tracking for requests
4863 referenced from lustre intent.
4865 Severity : Enhancement
4867 Description: Cache open file handles on client.
4868 Details : MDS now will return special lock along with openhandle, if
4869 requested and client is allowed to hold openhandle, even if unused,
4870 until such a lock is revoked. Helps NFS a lot, since NFS is opening
4871 closing files for every read/write openration.
4873 Severity : Enhancement
4875 Description: Cache open negative dentries on client when possible.
4876 Details : Guard negative dentries with UPDATE lock on parent dir, drop
4877 negative dentries on lock revocation.
4882 Description: Remounting a client read-only wasn't possible with a zconf mount
4883 Details : It wasn't possible to remount a client read-only with llmount.
4885 Severity : enhancement
4886 Description: Include MPICH 1.2.6 Lustre ADIO interface patch
4887 Details : In lustre/contrib/ or /usr/share/lustre in RPM a patch for
4888 MPICH is included to add Lustre-specific ADIO interfaces.
4889 This is based closely on the UFS ADIO layer and only differs
4890 in file creation, in order to allow the OST striping to be set.
4891 This is user-contributed code and not supported by CFS.
4896 Description: extended inode attributes (immutable, append-only) work improperly
4897 when 2.4 and 2.6 kernels are used on client/server or vice versa
4898 Details : Introduce kernel-independent values for these flags.
4900 Severity : enhancement
4903 Description: Allow fractional MB tunings for lustre in /proc/ filesystem.
4904 Details : Many of the /proc/ tunables can only be tuned at a megabyte
4905 granularity. Now, Fractional MB granularity is be supported,
4906 this is very useful for low memory system.
4908 Severity : enhancement
4910 Description: Getattr by fid
4911 Details : Getting a file attributes by its fid, obtaining UPDATE|LOOKUP
4912 locks, avoids extra getattr rpc requests to MDS, allows '/' to
4913 have locks and avoids getattr rpc requests for it on every stat.
4916 Frequency : Always, for filesystems larger than 2TB
4918 Description: ldiskfs crash at mount for filesystem larger than 2TB with mballoc
4919 Details : Kenrel kmalloc limits allocations to 128kB and this prevents
4920 filesystems larger than 2TB to be mounted with mballoc enabled.
4923 Frequency : Always, for 32-bit kernel without CONFIG_LBD and filesystem > 2TB
4925 Description: filesystem corruption for non-standard kernels and very large OSTs
4926 Details : If a 32-bit kernel is compiled without CONFIG_LBD enabled and a
4927 filesystems larger than 2TB is mounted then the kernel will
4928 silently corrupt the start of the filesystem. CONFIG_LBD is
4929 enabled for all CFS-supported kernels, but the possibility of
4930 this happening with a modified kernel config exists.
4932 Severity : enhancement
4934 Description: add client O_DIRECT support for 2.6 kernels
4935 Details : It is now possible to do O_DIRECT reads and writes to files
4936 in the Lustre client mountpoint on 2.6 kernel clients.
4938 Severity : enhancement
4940 Description: parallel glimpse, setattr, statfs, punch, destroy requests
4941 Details : Sends glimpse, setattr, statfs, punch, destroy requests to OSTs in
4942 parallel, not waiting for response from every OST before sending
4943 a rpc to the next OST.
4948 Description: setattr vs write race when updating file timestamps
4949 Details : Client processes that update a file timestamp into the past
4950 right after writing to the file (e.g. tar) it is possible that
4951 the updated file modification time can be reset to the current
4952 time due to a race between processing the setattr and write RPC.
4954 Severity : enhancement
4956 Description: Bring 'lfs find' closer in line with regular Linux find.
4957 Details : lfs find util supports -atime, -mtime, -ctime, -maxdepth, -print,
4958 -print0 options and obtains all the needed info through the lustre
4961 Severity : enhancement
4963 Description: support up to 1024 configured devices on one node
4964 Details : change obd_dev array from statically allocated to dynamically
4965 allocated structs as they are first used to reduce memory usage
4970 Description: Flush dirty partially truncated pages during truncate
4971 Details : Immediatelly flush partially truncated pages in filter_setattr,
4972 this way we completely avoid having any pages in page cache on OST
4973 and can retire ugly workarounds during writes to flush such pages.
4978 Description: i_sem vs transaction deadlock in mds_obd_destroy during unlink.
4979 Details : protect inode from truncation within vfs_unlink() context
4980 just take a reference before calling vfs_unlink() and release it
4981 when parent's i_sem is free.
4984 Frequency : always, if extents are used on OSTs
4986 Description: index ei_leaf_hi (48-bit extension) is not zeroed in extent index
4987 Details : OSTs using the extents format would not zero the high 16 bits of
4988 the index physical block number. This is not a problem for any
4989 OST filesystems smaller than 16TB, and no kernels support ext3
4990 filesystems larger than 16TB yet. This is fixed in 1.4.7 (all
4991 new/modified files) and can be fixed for existing filesystems
4992 with e2fsprogs-1.39-cfs1.
4997 Description: import connection selection may be incorrect if timer wraps
4998 Details : Using a 32-bit jiffies timer with HZ=1000 may cause backup
4999 import connections to be ignored if the 32-bit jiffies counter
5000 wraps. Use a 64-bit jiffies counter.
5003 Frequency : during server recovery
5005 Description: crash after server is denying duplicate export
5006 Details : If clients are resending connect requests to the server, the
5007 server refuses to allow a client to connect multiple times.
5008 Fixed a bug in the handling of this case.
5011 Frequency : very large clusters immediately after boot
5013 Description: LNET request buffers exhausted under heavy short-term load
5014 Details : If a large number of client requests are generated on a service
5015 that has previously never seen so many requests it is possible
5016 that the request buffer growth cannot keep up with the spike in
5017 demand. Instead of dropping incoming requests, they are held in
5018 the LND until the RPC service can accept more requests.
5021 Frequency : Sometimes during replay
5023 Description: Assertion failure in ll_local_open after replay.
5024 Details : If replay happened on an open request reply before we were able
5025 to set replay handler, reply will become not swabbed tripping the
5026 assertion in ll_local_open. Now we set the handler right after
5027 recognising of open request
5030 Frequency : very rare
5032 Description: kernel reports "badness in vsnprintf"
5033 Details : Reading from the "recovery_status" /proc file in small chunks
5034 may cause a negative length in lprocfs_obd_rd_recovery_status()
5035 call to vsnprintf() (which is otherwise harmless). Exit early
5036 if there is no more space in the output buffer.
5038 Severity : enhancement
5040 Description: clear OBD RPC statistics by writing to them
5041 Details : It is now possible to clear the OBD RPC statistics by writing
5042 to the "stats" file.
5047 Description: Client mtime is not the same on different clients after utimes
5048 Details : In some cases, the client was using the utimes() syscall on
5049 a file cached on another node. The clients now validate the
5050 ctime from the MDS + OSTs to determine which one is right.
5055 Description: Inability to activate failout mode
5056 Details : lconf script incorrectly assumed that in python string's numeric
5057 value is used in comparisons.
5060 Frequency : always with multiple stripes per file
5062 Description: Inefficient object allocation for mutli-stripe files
5063 Details : When selecting which OSTs to stripe files over, for files with
5064 a stripe count that divides evenly into the number of OSTs,
5065 the MDS is always picking the same starting OST for each file.
5066 Return the OST selection heuristic to the original design.
5071 Description: mount failures may take full timeout to return an error
5072 Details : Under some heavy load conditions it is possible that a
5073 failed mount can wait for the full obd_timeout interval,
5074 possibly several minutes, before reporting an error.
5075 Instead return an error as soon as the status is known.
5077 ------------------------------------------------------------------------------
5079 2006-02-14 Cluster File Systems, Inc. <info@clusterfs.com>
5081 * WIRE PROTOCOL CHANGE. This version of Lustre networking WILL NOT
5082 INTEROPERATE with older versions automatically. Please read the
5083 user documentation before upgrading any part of a live system.
5084 * WARNING: Lustre networking configuration changes are required with
5085 this release. See https://bugzilla.clusterfs.com/show_bug.cgi?id=10052
5088 * Support for kernels:
5089 2.6.9-22.0.2.EL (RHEL 4)
5090 2.6.5-7.244 (SLES 9)
5091 2.6.12.6 vanilla (kernel.org)
5094 Severity : enhancement
5095 Bugzilla : 7981/8208
5096 Description: Introduced Lustre Networking (LNET)
5097 Details : LNET is new networking infrastructure for Lustre, it includes
5098 a reorganized network configuration mode (see the user
5099 documentation for full details) as well as support for routing
5100 between different network fabrics. Lustre Networking Devices
5101 (LNDs) for the supported network fabrics have also been
5102 created for this new infrastructure.
5104 Severity : enhancement
5105 Description: Introduced Access control lists
5106 Details : clients can set ACLs on files and directories in order to have
5107 more fine-grained permissions than the standard Unix UGO+RWX.
5108 The MDS must be started with the "-o acl" mount option.
5110 Severity : enhancement
5111 Description: Introduced filesystem quotas
5112 Details : Administrators may now establish per-user quotas on the
5115 Severity : enhancement
5117 Description: Configuration change for the XT3
5118 The PTLLND is now used to run Lustre over Portals on the XT3
5119 The configure option(s) --with-cray-portals are no longer used.
5120 Rather --with-portals=<path-to-portals-includes> is used to
5121 enable building on the XT3. In addition to enable XT3 specific
5122 features the option --enable-cray-xt3 must be used.
5127 Description: Running on many-way SMP OSTs can trigger oops in llcd_send()
5128 Details : A race between allocating a new llcd and re-getting the llcd_lock
5129 allowed another thread to grab newly-allocated llcd.
5131 Severity : enhancement
5133 Description: 2.6 OST async journal commit and locking fix to improve performance
5134 Details : The filter_direct_io()+filter_commitrw_write() journal commits for
5135 2.6 kernels are now async as they already were in 2.4 kernels so
5136 that they can commit concurrently with the network bulk transfer.
5137 For block-allocated files the filter allocation semaphore is held
5138 to avoid filesystem fragmentation during allocation. BKL lock
5139 removed for 2.6 xattr operations where it is no longer needed.
5144 Description: lconf incorrectly determined whether two IP networks could talk
5145 Details : In some more complicated routing and multiple-network
5146 configurations, lconf will avoid trying to make a network
5147 connection to a disjoint part of the IP space. It was doing the
5148 math incorrectly for one set of cases.
5153 Description: Fix for potential infinite loop processing records in an llog.
5154 Details : If an llog record is corrupted/zeroed, it is possible to loop
5155 forever in llog_process(). Validate the llog record length
5156 and skip the remainder of the block on error.
5159 Frequency : occasional (liblustre only)
5161 Description: liblustre could not open files whose last component is a symlink
5162 Details : sysio_path_walk() would incorrectly pass the open intent to
5163 intermediate path components.
5166 Frequency : rare (liblustre only with non-standard tuning)
5167 Bugzilla : 7201 (7350)
5168 Description: Tuning the MDC DLM LRU size to zero triggers client LASSERT
5169 Details : llu_lookup_finish_locks() tries to set lock data on a lock
5170 after it has been released, only do this for referenced locks
5172 Severity : enhancement
5174 Description: specifying an (invalid) directory default stripe_size of -1
5175 would reset the directory default striping
5176 Details : stripe_size -1 was used internally to signal directory stripe
5177 removal, now use "all default" to signal dir stripe removal
5178 as a directory striping of "all default" is not useful
5181 Frequency : common for large clusters running liblustre clients
5183 Description: doing an ls when liblustre clients are running is slow
5184 Details : sending a glimpse AST to a liblustre client waits for every AST
5185 to time out, as liblustre clients will not respond. Since they
5186 cannot cache data we refresh the OST lock LVB from disk instead.
5188 Severity : enhancement
5190 Description: doing an ls at the same time as file IO can be slow
5191 Details : enqueue and other "small" requests can be blocked behind many
5192 large IO requests. Create a new OST IO portal for non-IO
5193 requests so they can be processed faster.
5196 Frequency : rare (only HPUX clients mounting unsupported re-exported NFS vol)
5198 Description: an HPUX NFS client would get -EACCESS when ftruncate()ing a newly
5199 created file with mode 000
5200 Details : the Linux NFS server relies on an MDS_OPEN_OWNEROVERRIDE hack to
5201 allow an ftruncate() as a non-root user to a file with mode 000.
5202 Lustre now respects this flag to disable mode checks when
5203 truncating a file owned by the user
5206 Frequency : liblustre-only, when liblustre client dies unexpectedly or becomes
5209 Description: Revoking locks from clients that went dead or catatonic might take
5211 Details : New lock flags FL_CANCEL_ON_BLOCK used by liblustre makes
5212 cancellation of such locks instant on servers without waiting for
5213 any reply from clients. Clients drops these locks when cancel
5214 notification from server is received without replying.
5217 Frequency : liblustre-only, when liblustre client dies or becomes busy
5219 Description: Doing ls on Linux clients can take a long time with active
5221 Details : Liblustre client cannot handle ASTs in timely manner, so avoid
5222 granting such locks to it in the first place if possible. Locks
5223 are taken by proxy on the OST during the read or write and
5224 dropped immediately afterward. Add connect flags handling, do
5225 not grant locks to liblustre clients for glimpse ASTs.
5227 Severity : enhancement
5229 Description: Improve read-ahead algorithm to avoid excessive IO for random reads
5230 Details : Existing read-ahead algorithm is tuned for the case of streamlined
5231 sequential reads and behaves badly with applications doing random
5232 reads. Improve it by reading ahead at least read region, and
5233 avoiding excessive large RPC for small reads.
5235 Severity : enhancement
5237 Description: Creating more than 1000 files for a single job may cause a load
5238 imbalance on the OSTs if there are also a large number of OSTs.
5239 Details : qos_prep_create() uses an OST index reseed value that is an
5240 even multiple of the number of available OSTs so that if the
5241 reseed happens in the middle of the object allocation it will
5242 still utilize the OSTs as uniformly as possible.
5247 Description: OST or MDS may oops in ping_evictor_main()
5248 Details : ping_evictor_main() drops obd_dev_lock if deleting a stale export
5249 but doesn't restart at beginning of obd_exports_timed list
5252 Severity : enhancement
5254 Description: improve by-nid export eviction on the MDS and OST
5255 Details : allow multiple exports with the same NID to be evicted at one
5256 time without re-searching the exports list.
5259 Frequency : rare, only with supplementary groups enabled on SMP 2.6 kernels
5261 Description: MDS may oops in groups_free()
5262 Details : in rare race conditions a newly allocated group_info struct is
5263 freed again, and this can be NULL. The 2.4 compatibility code
5264 for groups_free() checked for a NULL pointer, but 2.6 did not.
5267 Frequency : common for liblustre clients doing little filesystem IO
5268 Bugzilla : 9352, 7313
5269 Description: server may evict liblustre clients accessing contended locks
5270 Details : if a client is granted a lock or receives a completion AST
5271 with a blocking AST already set it would not reply to the AST
5272 for LDLM_FL_CANCEL_ON_BLOCK locks. It now replies to such ASTs.
5275 Frequency : lfs setstripe, only systems with more than 160 OSTs
5277 Description: unable to set striping with a starting offset beyond OST 160
5278 Details : llapi_create_file() incorrectly limited the starting stripe
5279 index to the maximum single-file stripe count.
5282 Frequency : LDAP users only
5284 Description: lconf did not handle in-kernel recovery with LDAP properly
5285 Details : lconf/LustreDB get_refs() is searching the wrong namespace
5287 Severity : enhancement
5289 Description: bind OST threads to NUMA nodes to improve performance
5290 Details : all OST threads are uniformly bound to CPUs on a single NUMA
5291 node and do their allocations there to localize memory access
5293 Severity : enhancement
5295 Description: llmount can determine client NID directly from Myrinet (GM)
5296 Details : the client NID code from gmnalnid was moved directly into
5297 llmount, removing the need to use this or specifying the
5298 client NID explicitly when mounting GM clients with zeroconf
5301 Frequency : if client is started with down MDS
5303 Description: if client is started with down MDS mount hangs in ptlrpc_queue_wait
5304 Details : Having an LWI_INTR() wait event (interruptible, but no timeout)
5305 will wait indefinitely in ptlrpc_queue_wait->l_wait_event() after
5306 ptlrpc_import_delayed_req() because we didn't check if the
5307 request was interrupted, and we also didn't break out of the
5308 event loop if there was no timeout
5313 Description: data loss during non-page-aligned writes to a single file from
5314 both multiple nodes and multiple threads on one node at same time
5315 Details : updates to KMS and lsm weren't protected by common lock. Resulting
5316 inconsistency led to false short-reads, that were cached and later
5317 used by ->prepare_write() to fill in partially written page,
5318 leading to data loss.
5321 Frequency : always, if lconf --abort_recovery used
5323 Description: lconf --abort_recovery fails with 'Operation not supported'
5324 Details : lconf was attempting to abort recovery on the MDT device and not
5327 Severity : enhancement
5329 Description: remove cleanup logs
5330 Details : replace lconf-generated cleanup logs with lustre internal
5331 cleanup routines. Eliminates the need for client-cleanup and
5334 Severity : enhancement
5336 Description: add support for EAs (user and system) on lustre filesystems
5337 Details : it is now possible to store extended attributes in the Lustre
5338 client filesystem, and with the user_xattr mount option it
5339 is possible to allow users to store EAs on their files also
5341 Severity : enhancement
5343 Description: Add possibility (config option) to show minimal available OST free
5345 Details : When compiled with --enable-mindf configure option, statfs(2)
5346 (and so, df) will return least minimal free space available from
5347 all OSTs as amount of free space on FS, instead of summary of
5348 free spaces of all OSTs.
5350 Severity : enhancement
5352 Description: do not expand extent locks acquired on OST-side
5353 Details : Modify ldlm_extent_policy() to not expand local locks, acquired
5354 by server: they are not cached anyway.
5357 Frequency : when mmap is used/binaries executed from Lustre
5359 Description: Unmmap pages before throwing them away from read cache.
5360 Details : llap_shrink cache now attempts to unmap pages before discarding
5361 them (if unmapping failed - do not discard). SLES9 kernel has
5362 extra checks that trigger if this unmapping is not done first.
5367 Description: lconf didn't resolve symlinks before checking to see whether a
5368 given mountpoint was already in use
5371 Frequency : when migrating failover services
5372 Bugzilla : 6395, 9514
5373 Description: When migrating a subset of services from a node (e.g. failback
5374 from a failover service node) the remaining services would
5375 time out and evict clients.
5376 Details : lconf --force (implied by --failover) sets the global obd_timeout
5377 to 5 seconds in order to quickly disconnect, but this caused
5378 other RPCs to time out too quickly. Do not change the global
5379 obd_timeout for force cleanup, only set it for DISCONNECT RPCs.
5381 Severity : enhancement
5382 Frequency : if MDS is started with down OST
5383 Bugzilla : 9439,5706
5384 Description: Allow startup/shutdown of an MDS without depending on the
5385 availability of the OSTs.
5386 Details : Asynchronously call mds_lov_synchronize during MDS startup.
5387 Add appropriate locking and lov-osc refcounts for safe
5388 cleaning. Add osc abort_inflight calls in case the
5389 synchronize never started.
5392 Frequency : occasional (Cray XT3 only)
5394 Description: root not authorized to access files in CRAY_PORTALS environment
5395 Details : The client process capabilities were not honoured on the MDS in
5396 a CRAY_PORTALS/CRAY_XT3 environment. If the file had previously
5397 been accessed by an authorized user then root was able to access
5398 the file on the local client also. The root user capabilities
5399 are now allowed on the MDS, as this environment has secure UID.
5402 Frequency : occasional
5404 Description: ldiskfs "too long searching" message happens too often
5405 Details : A debugging message (otherwise harmless) prints too often on
5406 the OST console. This has been reduced to only happen when
5407 there are fragmentation problems on the filesystem.
5412 Description: Division by zero in statfs when all OSCs are inactive
5413 Details : lov_get_stripecnt() returns zero due to incorrect order of checks,
5414 lov_statfs divides by value returned by lov_get_stripecnt().
5418 Bugzilla : 9489, 3273
5419 Description: First write from each client to each OST was only 4kB in size,
5420 to initialize client writeback cache, which caused sub-optimal
5421 RPCs and poor layout on disk for the first writen file.
5422 Details : Clients now request an initial cache grant at (re)connect time
5423 and so that they can start streaming writes to the cache right
5424 away and always do full-sized RPCs if there is enough data.
5425 If the OST is rebooted the client also re-establishes its grant
5426 so that client cached writes will be honoured under the grant.
5431 Description: Slow ls (and stat(2) syscall) on files residing on IO-loaded OSTs
5432 Details : Now I/O RPCs go to different portal number and (presumably) fast
5433 lock requests (and glimses) and other RPCs get their own service
5434 threads pool that should be able to service those RPCs
5437 Severity : enhancement
5439 Description: Ability to exchange lustre version between client and servers and
5440 issue warnings at client side if client is too old. Also for
5441 liblustre clients there is ability to refuse connection of too old
5443 Details : New 'version' field is added to connect data structure that is
5444 filled with version info. That info is later checked by server and
5448 Frequency : rare, liblustre only.
5449 Bugzilla : 9296, 9581
5450 Description: Two simultaneous writes from liblustre at offset within same page
5451 might proceed at the same time overwriting eachother with stale
5453 Details : I/O lock withing llu_file_prwv was released too early, before data
5454 actually was hitting the wire. Extended lock-holding time until
5455 server acknowledges receiving data.
5458 Frequency : extremely rare. Never observed in practice.
5460 Description: avoid generating lustre_handle cookie of 0.
5461 Details : class_handle_hash() generates handle cookies by incrementing
5462 global counter, and can hit 0 occasionaly (this is unlikely, but
5463 not impossible, because initial value of cookie counter is
5464 selected randonly). Value of 0 is used as a sentinel meaning
5465 "unassigned handle" --- avoid it. Also coalesce two critical
5466 sections in this function into one.
5468 Severity : enhancement
5470 Description: allow liblustre clients to delegate truncate locking to OST
5471 Details : To avoid overhead of locking, liblustre client instructs OST to
5472 take extent lock in ost_punch() on client's behalf. New connection
5473 flag is added to handle backward compatibility.
5475 Severity : enhancement
5476 Bugzilla : 4928, 7341, 9758
5477 Description: allow number of OST service threads to be specified
5478 Details : a module parameter allows the number of OST service threads
5479 to be specified via "options ost ost_num_threads={N}" in the
5480 OSS's /etc/modules.conf or /etc/modprobe.conf.
5484 Bugzilla : 6146, 9635, 9895
5485 Description: servers crash with bad pointer in target_handle_connect()
5486 Details : In rare cases when a client is reconnecting it was possible that
5487 the connection request was the last reference for that export.
5488 We would temporarily drop the export reference and get a new
5489 one, but this may have been the last reference and the export
5490 was just destroyed. Get new reference before dropping old one.
5492 Severity : enhancement
5493 Frequency : if client is started with failover MDS
5495 Description: Allow multiple MDS hostnames in the mount command
5496 Details : Try to read the configuration from all specified MDS
5497 hostnames during a client mount in case the "primary"
5500 Severity : enhancement
5502 Description: Stop sending data to evicted clients as soon as possible.
5503 Details : Check if the client we are about to send or are sending data to
5504 was evicted already. (Check is done every second of waiting,
5505 for which l_wait_event interface was extended to allow checking
5506 of exit condition at specified intervals).
5509 Frequency : rare, normally only when NFS exporting is done from client
5511 Description: 'bad disk LOV MAGIC: 0x00000000' error when chown'ing files
5513 Details : Make mds_get_md() recognise empty md case and set lmm size to 0.
5516 Frequency : always, if srand() is called before liblustre initialization
5518 Description: Liblustre uses system PRNG disturbing its usage by user application
5519 Details : Introduce internal to lustre fast and high-quality PRNG for
5520 lustre usage and make liblustre and some other places in generic
5521 lustre code to use it.
5523 Severity : enhancement
5524 Bugzilla : 9477, 9557, 9870
5525 Description: Verify that the MDS configuration logs are updated when xml is
5526 Details : Check if the .xml configuration logs are newer than the config
5527 logs stored on the MDS and report an error if this is the case.
5528 Request --write-conf, or allow starting with --old_conf.
5530 Severity : enhancement
5532 Description: Handle symlinks in the path when checking if Lustre is mounted.
5533 Details : Resolve intermediate symlinks when checking if a client has
5534 mounted a filesystem to avoid duplicate client mounts.
5539 Description: lconf can hit an error exception but still return success.
5540 Details : The lconf command catches the Command error exception at the top
5541 level script context and will exit with the associated exit
5542 status, but doesn't ensure that this exit status is non-zero.
5547 Description: failure of ptlrpc thread startup can cause oops
5548 Details : Starting a ptlrpc service thread can fail if there are a large
5549 number of threads or the server memory is very fragmented.
5550 Handle this without oopsing.
5553 Frequency : always, only if liblustre and non-default acceptor port was used
5555 Description: liblustre cannot connect to servers with non-default acceptor port
5556 Details : tcpnal_set_default_params() was not called and was therefore
5557 ignoring the environment varaible TCPNAL_PORT, as well as other
5558 TCPNAL_ environment variables
5563 Description: two objects could be created on the same OST for a single file
5564 Details : If an OST is down, in some cases it was possible to create two
5565 objects on a single OST for a single file. No problems other
5566 than potential performance impact and spurious error messages.
5570 Bugzilla : 5681, 9562
5571 Description: Client may oops in ll_unhash_aliases
5572 Details : Client dcache may become inconsistent in race condition.
5573 In some cases "getcwd" can fail if the current directory is
5579 Description: Inode refcounting problems in NFS export code
5580 Details : link_raw functions used to call d_instantiate without obtaining
5581 extra inode reference first.
5585 Bugzilla : 9942, 9903
5586 Description: Referencing freed requests leading to crash, memleaks with NFS.
5587 Details : We used to require that call to ll_revalidate_it was always
5588 followed by ll_lookup_it. Also with revalidate_special() it is
5589 possible to call ll_revalidate_it() twice for the same dentry
5590 even if first occurence returned success. This fix changes semantic
5591 between DISP_ENQ_COMPLETE disposition flag to mean there is extra
5592 reference on a request referred from the intent.
5593 ll_intent_release() then releases such a request.
5596 Frequency : rare, normally benchmark loads only
5598 Description: unlinked inodes were kept in memory on the client
5599 Details : If a client is repeatedly creating and unlinking files it
5600 can accumulate a lot of stale inodes in the inode slab cache.
5601 If there is no other client load running this can cause the
5602 client node to run out of memory. Instead flush old inodes
5603 from client cache that have the same inode number as a new inode.
5606 Frequency : SLES9 2.6.5 kernel and long filenames only
5607 Bugzilla : 9969, 10379
5608 Description: utime reports stale NFS file handle
5609 Details : SLES9 uses out-of-dentry names in some cases, which confused
5610 the lustre dentry revalidation. Change it to always use the
5614 Frequency : rare, unless heavy write-truncate concurrency is continuous
5615 Bugzilla : 4180, 6984, 7171, 9963, 9331
5616 Description: OST becomes very slow and/or deadlocked during object unlink
5617 Details : filter_destroy() was holding onto the parent directory lock
5618 while truncating+unlinking objects. For very large objects this
5619 may block other threads for a long time and slow overall OST
5620 responsiveness. It may also be possible to get a lock ordering
5621 deadlock in this case, or run out of journal credits because of
5622 the combined truncate+unlink. Solution is to do object truncate
5623 first in one transaction without parent lock, and then do the
5624 final unlink in a new transaction with the parent lock. This
5625 reduces the lock hold time dramatically.
5628 Frequency : rare, 2.4 kernels only
5630 Description: MDS or OST cleanup may trip kernel BUG when dropping kernel lock
5631 Details : mds_cleanup() and filter_cleanup() need to drop the kernel lock
5632 before unmounting their filesystem in order to avoid deadlock.
5633 The kernel_locked() function in 2.4 kernels only checks whether
5634 the kernel lock is held, not whether it is this process that is
5635 holding it as 2.6 kernels do.
5640 Description: MDS or OST may oops/LBUG if a client is connecting multiple times
5641 Details : The client ptlrpc code may be trying to reconnect to a down
5642 server before a previous connection attempt has timed out.
5643 Increase the reconnect interval to be longer than the connection
5644 timeout interval to avoid sending duplicate connections to
5648 Frequency : echo_client brw_test command
5650 Description: fix echo_client to work with OST preallocated code
5651 Details : OST preallocation code (5137) didn't take echo_client IO path
5652 into account: echo_client calls filter methods outside of any
5653 OST thread and, hence, there is no per-thread preallocated
5654 pages and buffers to use. Solution: hijack pga pages for IO. As
5655 a byproduct, this avoids unnecessary data copying.
5659 Bugzilla : 3555, 5962, 6025, 6155, 6296, 9574
5660 Description: Client can oops in mdc_commit_close() after open replay
5661 Details : It was possible for the MDS to return an open request with no
5662 transaction number in mds_finish_transno() if the client was
5663 evicted, but without actually returning an error. Clients
5664 would later try to replay that open and may trip an assertion
5665 Simplify the client close codepath, and always return an error
5666 from the MDS in case the open is not successful.
5669 Frequency : rare, 2.6 OSTs only
5671 Description: OST may deadlock under high load on fragmented files
5672 Details : If there was a heavy load and highly-fragmented OST filesystems
5673 it was possible to have all the OST threads deadlock waiting on
5674 allocation of biovecs, because the biovecs were not released
5675 until the entire RPC IO was completed. Instead, release biovecs
5676 as soon as they are complete to ensure forward IO progress.
5678 Severity : enhancement
5680 Description: Support for specifying external journal device at mount
5681 Details : If an OST or MDS device is formatted with an external journal
5682 device, this device major/minor is stored in the ext3 superblock
5683 and may not be valid for failover. Allow detecting and
5684 specifying the external journal at mount time.
5689 Description: Mounting an MDS with pending unlinked files may cause oops
5690 Details : target_finish_recovery() calls mds_postrecov() which returned
5691 the number of orphans unlinked. mds_lov_connect->mds_postsetup()
5692 considers this an error and immediately begins cleaning up the
5693 lov, just after starting the mds_lov process
5695 Severity : enhancement
5697 Description: Implement 'lfs df' to report actual free space on per-OST basis
5698 Details : Add sub-command 'df' on 'lfs' to report the disk space usage of
5699 MDS/OSDs. Usage: lfs df [-i][-h]. Command Options: '-i' to report
5700 usage of objects; '-h' to report in human readable format.
5702 ------------------------------------------------------------------------------
5704 2005-08-26 Cluster File Systems, Inc. <info@clusterfs.com>
5711 Description: Mounting an ldiskfs file system with mballoc may crash OST node.
5712 Details : ldiskfs mballoc code may reference an uninitialized buddy struct
5713 at startup during orphan unlinking. Instead, skip buddy update
5714 before setup, as it will be regenerated after recovery is complete.
5719 Description: If an OST is inactive, its locks might reference stale inodes.
5720 Details : lov_change_cbdata() must iterate over all namespaces, even if
5721 they are inactive to clear inode references from the lock.
5723 Severity : enhancement
5724 Frequency : occasional, if non-standard max_dirty_mb used
5726 Description: Client will block write RPCs if not enough grant
5727 Details : If a client has max_dirty_mb smaller than max_rpcs_in_flight,
5728 then the client will block writes while waiting for another RPC
5729 to complete instead of consuming its dirty limit. With change
5730 we get improved performance when max_dirty_mb is small.
5732 Severity : enhancement
5733 Bugzilla : 3389, 6253
5734 Description: Add support for supplementary groups on the MDS.
5735 Details : The MDS has an upcall /proc/fs/lustre/mds/{mds}/group_upcall
5736 (set to /usr/sbin/l_getgroups if enabled) which will do MDS-side
5737 lookups for user supplementary groups into a cache.
5741 Description: O_CREAT|O_EXCL open flags in liblustre always return -EEXIST
5742 Details : Make libsysio to not enforce O_EXCL by clearing the flag,
5743 for liblustre O_EXCL is enforced by MDS.
5747 Description: readdir never returns NULL in liblustre.
5748 Details : Corrected llu_iop_getdirentries logic, to return offset of next
5749 dentry in struct dirent.
5753 Frequency : liblustre only, depends on application IO pattern
5754 Description: liblustre clients evicted if not contacting servers
5755 Details : Don't put liblustre clients into the ping_evictor list, so
5756 they will not be evicted by the pinger ever.
5758 Severity : enhancement
5760 Description: Add ability to evict clients by NID from MDS.
5761 Details : By echoing "nid:$NID" string into
5762 /proc/fs/lustre/mds/.../evict_client client with nid that equals to
5763 $NID would be instantly evicted from this MDS and from all active
5764 OSTs connected to it.
5768 Description: Do not query file size twice, somewhat slowing stat(2) calls.
5769 Details : lookup_it_finish() used to query file size from OSTs that was not
5774 Description: service threads change working directory to that of init
5775 Details : Starting lustre service threads may pin the working directory
5776 of the parent thread, making that filesystem busy. Threads
5777 now change to the working directory of init to avoid this.
5781 Frequency : during shutdown only
5782 Description: shutdown with a failed MDS or OST can cause unmount to hang
5783 Details : Don't resend DISCONNECT messages in ptlrpc_disconnect_import()
5788 Frequency : 2.6 only
5789 Description: chmod/chown may include an extra supplementary group
5790 Details : ll{,u}_mdc_pack_op_data() does not properly initialize the
5791 supplementary group and if none is specified this is used.
5794 Bugzilla : 5479 (6816)
5796 Description: Racing open + rm can assert client in mdc_set_open_replay_data()
5797 Details : If lookup is in progress on a file that is unlinked we might try
5798 to revalidate the inode and fail in revalidate after lookup is
5799 complete and ll_file_open() enqueues the open again but
5800 it_open_error() was not checking DISP_OPEN_OPEN errors correctly.
5803 Frequency : always, if lconf --abort_recovery used
5805 Description: lconf --abort_recovery fails with 'Operation not supported'
5806 Details : lconf was attempting to abort recovery on the MDT device and not
5809 ------------------------------------------------------------------------------
5811 2005-08-08 Cluster File Systems, Inc. <info@clusterfs.com>
5816 Frequency : rare (only unsupported configurations with a node running as an
5818 Bugzilla : 6514, 5137
5819 Description: Mounting a Lustre file system on a node running as an OST could
5821 Details : OSTs now preallocates memory needed to write out data at
5822 startup, instead of when needed, to avoid having to
5823 allocate memory in possibly low memory situations.
5824 Specifically, if the file system is mounted on on OST,
5825 memory pressure could force it to try to write out data,
5826 which it needed to allocate memory to do. Due to the low
5827 memory, it would be unable to do so and the node would
5828 become unresponsive.
5830 Severity : enhancement
5832 Description: Addition of lconf --service command line option
5833 Details : lconf now accepts a '--service <arg>' option, which is
5834 shorthand for 'lconf --group <arg> --select <arg>=<hostname>'
5836 Severity : enhancement
5838 Description: Failover mode is now the default for OSTs.
5839 Details : By default, OSTs will now run in failover mode. To return to
5840 the old behaviour, add '--failout' to the lmc line for OSTs.
5842 Severity : enhancement
5844 Description: Health checks are now provided for MDS and OSTs
5845 Details : Additional detailed health check information on MSD and OSTs
5846 is now provided through the procfs health_check value.
5849 Frequency : occasional, depends on IO load
5851 Description: Disk fragmentation on the OSTs could eventually cause slowdowns
5852 after numerous create/delete cycles
5853 Details : The ext3 inode allocation policy would not allocate new inodes
5854 very well on the OSTs because there are no new directories
5855 being created. Instead we look for groups with free space if
5856 the parent directories are nearly full.
5861 Description: Network or server problems during mount may cause partially
5862 mounted clients instead of returning an error.
5863 Details : The config llog parsing code may overwrite the error return
5864 code during mount error handling, returning success instead
5870 Description: MDS can fail to allocate large reply buffers
5871 Details : After long uptimes the MDS can fail to allocate large reply
5872 buffers (e.g. zconf client mount config records) due to memory
5873 fragmentation or consumption by the buffer cache. Preallocate
5874 some large reply buffers so that these replies can be sent even
5875 under memory pressure.
5879 Frequency : rare (liblustre)
5880 Description: fsx running with liblustre complained that using truncate() to
5881 extend the file doesn't work. This patch corrects that issue.
5882 Details : This is the liblustre equivalent of the fix for bug 6196. Fixes
5883 ATTR_SIZE and lsm use in llu_setattr_raw.
5887 Frequency : rare, only 2.6 kernels
5888 Description: Unusual file access patterns on the MDS may result in inode
5889 data being lost in very rare circumstances.
5890 Details : Bad interaction between the ea-in-inode patch and the "no-read"
5891 code in the 2.6 kernel caused the inode and/or EA data not to
5892 be read from disk, causing single-file corruption.
5896 Frequency : rare, only 2.6 filesystems using extents
5897 Description: Heavy concurrent write and delete load may cause data corruption.
5898 Details : It was possible under high-load situations to have an extent
5899 metadata block in the block device cache from a just-unlinked
5900 file overwrite a newly-allocated data block. We now unmap any
5901 metadata buffers that alias just-allocated data blocks.
5905 Frequency : filesystems with default stripe_count larger than 77
5906 Description: lconf+mke2fs fail when formatting filesystem with > 77 stripes
5907 Details : lconf specifies an inode size of 4096 bytes when the default
5908 stripe_count is larger than 77. This conflicts with the default
5909 inode density of 1 per 4096 bytes. Allocate smaller inodes in
5910 this case to avoid pinning too much memory for large EAs.
5912 ------------------------------------------------------------------------------
5914 2005-07-07 Cluster File Systems, Inc. <info@clusterfs.com>
5919 Frequency : rare (extremely heavy IO load with hundreds of clients)
5921 Description: Client is evicted, gets IO error writing to file
5922 Details : lock ordering changes for bug 5492 reintroduced bug 3267 and
5923 caused clients to be evicted for AST timeouts. The fixes in
5924 bug 5192 mean we no longer need to have such short AST timeouts
5925 so ldlm_timeout has been increased.
5928 Frequency : occasional during --force or --failover shutdown under load
5929 Bugzilla : 5949, 4834
5930 Description: Server oops/LBUG if stopped with --force or --failover under load
5931 Details : a collection of import/export refcount and cleanup ordering
5932 issues fixed for safer force cleanup
5935 Frequency : only filesystems larger than 120 OSTs
5936 Bugzilla : 5990, 6223
5937 Description: lfs getstripe would oops on a very large filesystem
5938 Details : lov_getconfig used kfree on vmalloc'd memory
5941 Frequency : only filesystems exporting via NFS to Solaris 10 clients
5942 Bugzilla : 6242, 6243
5943 Description: reading from files that had been truncated to a non-zero size
5944 but never opened returned no data
5945 Details : ll_file_read() reads zeros from no-object files to EOF
5950 Description: A bug in MDS/OSS recovery could cause the OSS to fail an assertion
5951 Details : There's little harm in aborting MDS/OSS recovery and letting it
5952 try again, so I removed the LASSERT and return an error instead.
5954 Severity : enhancement
5956 Description: New debugging infrastructure for tracking down data corruption
5957 Details : The I/O checksum code was replaced to: (a) control it at runtime,
5958 (b) cover more of the client-side code path, and (c) try to narrow
5959 down where problems occurred
5963 Bugzilla : 3819, 4364, 4397, 6313
5964 Description: Racing close and eviction MDS could cause assertion in mds_close
5965 Details : It was possible to get multiple mfd references during close and
5966 client eviction, leading to one thread referencing a freed mfd.
5968 Severity: : enhancement
5969 Bugzilla : 3262, 6359
5970 Description: Attempts to reconnect to servers are now more aggressive.
5971 Details : This builds on the enhanced upcall-less recovery that was added
5972 in 1.4.2. When trying to reconnect to servers, clients will
5973 now try each server in the failover group every 10 seconds. By
5974 default, clients would previously try one server every 25 seconds.
5979 Description: After recovery, certain operations trigger a failed
5980 assertion on a client.
5981 Details : Failing over an mds, using lconf -d --failover, while a
5982 client was doing a readdir() call would cause the client to
5983 LBUG after recovery completed and the readdir() was resent.
5985 Severity : enhancement
5987 Description: Default groups are now added by lconf
5988 Details : You can now run lconf --group <servicename> without having to
5989 manually add groups with lmc.
5992 Frequency : occasional
5994 Description: Nodes with an elan id of 0 trigger a failed assertion
5997 Frequency : always when accessing e.g. tty/console device nodes
5999 Description: tty and some other devices nodes cannot be used on lustre
6000 Details : file's private_data field is used by device data and lustre
6001 values in there got lost. New field was added to struct file to
6002 store fs-specific private data.
6005 Frequency : when exporting Lustre via NFS
6007 Description: NFSD failed occasionally when looking up a path component
6008 Details : NFSD is looking up ".." which was broken in ext3 directories
6009 that had grown large enough to become hashed.
6012 Frequency : Clusters with multiple interfaces not on the same subnet
6014 Description: Nodes will repeatedly try to reconnect to an interface which it
6015 cannot reach and report an error to the log.
6016 Details : Extra peer list entries will be created by lconf with some peers
6017 unreachable. lconf now validates the peer before adding it.
6020 Frequency : Only if a default stripe is set on the filesystem root.
6022 Description: Setting a default stripe on the filesystem root prevented the
6023 filesystem from being remounted.
6024 Details : The client was sending extra request flags in the root getattr
6025 request and did not allocate a reply buffer for the dir EA.
6028 Frequency : occasional, higher if lots of files are accessed by one client
6029 Bugzilla : 6159, 6097
6030 Description: Client trips assertion regarding lsm mismatch/magic
6031 Details : While revalidating inodes the VFS looks up inodes with ifind()
6032 and in rare cases can find an inode that is being freed.
6033 The ll_test_inode() code will free the lsm during ifind()
6034 when it finds an existing inode and then the VFS later attaches
6035 this free lsm to a new inode.
6039 Bugzilla : 6422, 7030
6040 Description: MDS deadlock between mkdir and client eviction
6041 Details : Creating a new file via mkdir or mknod (starting a transaction
6042 and getting the ns lock) can deadlock with client eviction
6043 (gets ns lock and trying to finish a synchronous transaction).
6046 Frequency : occasional
6047 Description: While starting a server, the fsfilt_ext3 module could not be
6049 Details : CFS's improved ext3 filesystem is named ldiskfs for 2.6
6050 kernels. Previously, lconf would still use the ext3 name
6051 when trying to load modules. Now, it will correctly use
6052 ext3 on 2.4 and ldiskfs on 2.6.
6054 Severity : enhancement
6055 Description: The default stripe count has been changed to 1
6056 Details : The interpretation of the default stripe count (0, to lfs
6057 or lmc) has been changed to mean striping across a single
6058 OST, rather than all available. For general usage we have
6059 found a stripe count of 1 or 2 works best.
6061 Severity : enhancement
6062 Description: Add support for compiling against Cray portals.
6063 Details : Conditional compiling for some areas that are different
6067 Frequency : occasional
6068 Bugzilla : 6409, 6834
6069 Description: Creating files with an explicit stripe count may lead to
6070 a failed assertion on the MDS
6071 Details : If some OSTs are full or unavailable, creating files may
6072 trigger a failed assertion on the MDS. Now, Lustre will
6073 try to use other servers or return an error to the
6077 Frequency : occasional
6079 Description: Multiple concurrent overlapping read+write on multiple SMP nodes
6080 caused lock timeout during readahead (since 1.4.2).
6081 Details : Processes doing readahead might match a lock that hasn't been
6082 granted yet if there are overlapping and conflicting lock
6083 requests. The readahead process waits on ungranted lock
6084 (original lock is CBPENDING), while OST waits for that process
6085 to cancel CBPENDING read lock and eventually evicts client.
6087 Severity : enhancement
6089 Description: Initial enabling of flock support for clients
6090 Details : Implements fcntl advisory locking and file status functions.
6091 This feature is provided as an optional mount flag (default
6092 off), and is NOT CURRENTLY SUPPORTED. Not all types of record
6093 locking are implemented yet, and those that are are not guaranteed
6094 to be completely correct in production environments.
6095 mount -t lustre -o [flock|noflock] ...
6098 Frequency : occasional
6100 Description: OSTs running 2.4 kernels but with extents enabled might trip an
6101 assertion in the ext3 JBD (journaling) layer.
6102 Details : The b_committed_data struct is protected by the big kernel lock
6103 in 2.4 kernels, serializing journal_commit_transaction() and
6104 ext3_get_block_handle->ext3_new_block->find_next_usable_block()
6105 access to this struct. In 2.6 kernels there is finer grained
6106 locking to improve SMP performance of the JBD layer.
6110 Description: Changes the "SCSI I/O Stats" kernel patch to default to "enabled"
6112 -----------------------------------------------------------------------------
6114 2005-05-05 Cluster File Systems, Inc. <info@clusterfs.com>
6116 NOTE: Lustre 1.4.2 uses an incompatible network protocol than previous
6117 versions of Lustre. Please update all servers and clients to
6118 version 1.4.2 or later at the same time. You must also run
6119 "lconf --write-conf {config}.xml" on the MDS while it is stopped
6120 to update the configuration logs.
6122 - fix for HPUX NFS client breakage when NFS exporting Lustre (5781)
6123 - mdc_enqueue does not need max_mds_easize request buffer on send (5707)
6124 - swab llog records of type '0' so we get proper header size/idx (5861)
6125 - send llog cancel req to DLM cancel portal instead of cb portal (5515)
6126 - fix rename of one directory over another leaking an inode (5953)
6127 - avoid SetPageDirty on 2.6 (5981)
6128 - don't re-add just-being-destroyed locks to the waiting list (5653)
6129 - when creating new directories, inherit the parent's custom
6130 striping settings if present parent (3048)
6131 - flush buffers from cache before direct IO in 2.6 obdfilter (4982)
6132 - don't hold i_size_sem in ll_nopage() and ll_ap_refresh_count (6077)
6133 - don't hold client locks on temporary worklist from l_lru (5666)
6134 - handle IO errors in 2.6 obdfilter bio completion routine (6046)
6135 - automatically evict dead clients (5921)
6136 - Update file size properly in create+truncate+fstat case (6196)
6137 - Do not unhash mountpoint dentries, do not allow removal of
6139 - Avoid lock ordering deadlock issue with write/truncate (6203,5654)
6140 - reserve enough journal credits in fsfilt_start_log for setattr (4554)
6141 - ldlm_enqueue freed-export error path would always LBUG (6149,6184)
6142 - don't reference lr_lvb_data until after we hold lr_lvb_sem (6170)
6143 - don't overwrite last_rcvd if there is a *_client_add() error (6086)
6144 - Correctly handle reads of files with no objects (6243)
6145 - lctl recover will also mark a device active if deactivate used (5933)
6147 - by default create 1 inode per 4kB space on MDS, per 16kB on OSTs
6148 - allow --write-conf on an MDS with different nettype than client (5619)
6149 - don't write config llogs to MDS for mounts not from that MDS (5617)
6150 - lconf should create multiple TCP connections from a client (5201)
6151 - init scripts are now turned off by default; run chkconfig --on
6152 lustre and chkconfig --on lustrefs to use them
6153 - upcalls are no longer needed for clients to recover to failover
6155 - add --abort-recovery option to lconf to abort recovery on device
6157 - add support for an arbitrary number of OSTs (3026)
6158 - Quota support protocol changes.
6159 - forward compatibility changes to wire structs (6007)
6160 - rmmod NALs that might be loaded because of /etc/modules.conf (6133)
6161 - support for mountfsoptions and clientoptions to the Lustre LDAP (5873)
6162 - improved "lustre status" script
6163 - initialize blocksize for non-regular files (6062)
6164 - added --disable-server and --disable-client configure options (5782)
6165 - introduce a lookup cache for lconf to avoid repeated DB scans (6204)
6166 - Vanilla 2.4.29 support
6167 - increase maximum number of obd devices to 520 (6242)
6168 - remove the tcp-zero-copy patch from the suse-2.4 series (5902)
6169 - Quadrics Elan drivers are now included for the RHEL 3 2.4.21 and
6170 SLES 9 2.6.5 kernels
6171 - limit stripes per file to 160 (the maximum EA size) (6093)
6173 2005-03-22 Cluster File Systems, Inc. <info@clusterfs.com>
6176 - don't LASSERT in ll_release on NULL lld with NFS export (4655, 5760)
6177 - hold NS lock when calling handle_ast_error->del_waiting_lock (5746)
6178 - fix setattr mtime regression from lovcleanup merge (4829, 5669)
6179 - workaround for 2.6 crash in ll_unhash_aliases (5687, 5210)
6180 - small ext3 extents cleanups and fixes (5733)
6181 - improved mballoc code, several small races and bugs fixed (5733, 5638)
6182 - kernel version 43 - fix remove_suid bugs in both 2.4 and 2.6 (5695)
6183 - avoid needless client->OST connect, fix handle mismatch (5317)
6184 - fix DLM error path that led to out-of-sync client, long delays (5779)
6185 - support common vfs-enforced mount options (nodev,nosuid,noexec) (5637)
6186 - fix several locking issues related to i_size (5492,5624,5654,5672)
6187 - don't move pending lock onto export if it is already evicted (5683)
6188 - fix kernel oops when creating .foo in unlinked directory (5548)
6189 - fix deadlock in obdfilter statistics vs. object create (5811)
6190 - use time_{before,after} to avoid timer jiffies wrap (5882)
6191 - shutdown --force/--failover stability (3607,3651,4797,5203,4834)
6192 - Do not leak request if server was not able to process it (5154)
6193 - If mds_open unable to find parent dir, make that negative lookup(5154)
6194 - don't create new directories with extent-mapping (5909, 5936)
6196 - fix lustre/lustrefs init scripts for SuSE (patch from Scali, 5702)
6197 - don't hold the pinger_sem in ptlrpc_pinger_sending_on_import
6198 - change obd_increase_kms to obd_adjust_kms (up or down) (5654)
6199 - lconf, lmc search both /usr/lib and /usr/lib64 for Python libs (5800)
6200 - support for RHEL4 kernel on i686 (5773)
6201 - provide error messages when incompatible logs are encountered (5898)
6203 2005-02-18 Cluster File Systems, Inc. <info@clusterfs.com>
6204 * version 1.4.0.10 (1.4.1 release candidate 1)
6206 - don't keep a lock reference when lock is not granted (4238)
6207 - unsafe list practices (rarely) led to infinite eviction loop (4908)
6208 - add per-fs limit of Lustre pages in page cache, avoid OOM (4699)
6209 - drop import inflight refcount on signal_completed_replay error (5255)
6210 - unlock page after async write error during send (3677)
6211 - handle missing objects in filter_preprw_read properly (5265)
6212 - no transno return for symlink open, don't save no-trasno open (3440)
6213 - don't try to complete elan receive that already failed (4012)
6214 - free RPC server reply state on error (5406)
6215 - clean up thread from ptlrpc_start_thread() on error (5160)
6216 - readahead could read extra page into cache that wasn't ejected (5388)
6217 - prevent races in class_attach/setup/cleanup/detach (5260)
6218 - don't dereference de->d_inode after l_dput of de (5458)
6219 - use "int" for stripe value returned from lock_to_stripe (5544)
6220 - mballoc allocation and error-checking fixes in 2.6 (5504)
6221 - block device patches to fix I/O request sizes in 2.6 (5482)
6222 - look up hostnames for IB nals (5602)
6223 - 2.6 changed lock ordering of 2 semaphores, caused deadlock (5654)
6224 - don't start multiple acceptors for the same port (5277)
6225 - fix incorrect LASSERT in mds_getattr_name (5635)
6226 - export a proc file for general "ping" checking (5628)
6227 - fix "lfs check" to not block when the MDS is down (5628)
6229 - service request history (4965)
6230 - put {ll,lov,osc}_async_page structs in a single slab (4699)
6231 - create an "evict_client" /proc entry on OSTs, like the MDS has
6232 - fix mount usage message, return errors per mount(8) (5168)
6233 - change grep [] to grep "[]" in tests so they work in more UMLs
6234 - fix ppc64/x86_64 spec to use %{_libdir} instead of /usr/lib (5389)
6235 - remove ancient LOV_MAGIC_V0 EA support (5047)
6236 - add "disk I/Os in flight" and "I/O req time" stats in obdfilter
6237 - align r/w RPCs to PTLRPC_MAX_BRW_SIZE boundary for performance (3451)
6238 - allow readahead allocations to fail when low on memory (5383)
6239 - mmap locking landed again, after considerable improvement (2828)
6240 - add get_hostaddr() to lustreDB.py for LDAP support (5459)
6242 2004-11-23 Cluster File Systems, Inc. <info@clusterfs.com>
6245 - send OST transaction number in read/write reply to free req (4966)
6246 - don't ASSERT in ptl_send_rpc() if we run out of memory (5119)
6247 - lock /proc/sys/portals/routes internal state, avoiding oops (4827)
6248 - the watchdog thread now runs as interruptible (5246)
6249 - flock/lockf fixes (but it's still disabled, pending 5135)
6250 - don't use EXT3 constants in llite code (5094)
6251 - memory shortage at startup could cause assertion (5176)
6253 - reorganization of lov code
6254 - single portals codebase
6256 - add extents/mballoc support (5025)
6257 - direct I/O reads in the obdfilter (4048)
6258 - kernel patches from LNXI for 2.6 (bluesmoke, perfctr, mtd, kexec)
6260 tbd Cluster File Systems, Inc. <info@clusterfs.com>
6263 - send OST transaction number in read/write reply to free req (4966)
6264 - don't ASSERT in ptl_send_rpc() if we run out of memory (5119)
6265 - lock /proc/sys/portals/routes internal state, avoiding oops (4827)
6266 - the watchdog thread now runs as interruptible (5246)
6267 - handle missing objects in filter_preprw_read properly (5265)
6268 - unsafe list practices (rarely) led to infinite eviction loop (4908)
6269 - drop import inflight refcount on signal_completed_replay error (5255)
6270 - unlock page after async write error during send (3677)
6271 - return original error code on reconstructed replies (3761)
6272 - no transno return for symlink open, don't save no-trasno open (3440)
6274 - add pid to ldlm debugging output (4922)
6275 - bump the watchdog timeouts -- we can't handle 30sec yet
6276 - extra debugging for orphan dentry/inode bug (5259)
6278 2004-11-16 Cluster File Systems, Inc. <info@clusterfs.com>
6281 - fix TCP_NODELAY bug, which caused extreme perf regression (5134)
6282 - allocate qswnal tx descriptors singly to avoid fragmentation (4504)
6283 - don't LBUG on obdo_alloc() failure, use OBD_SLAB_ALLOC() (4800)
6284 - fix NULL dereference in /proc/sys/portals/routes (4827)
6285 - allow failed mdc_close() operations to be interrupted (4561)
6286 - stop precreate on OST before MDS would time out on it (4778)
6287 - don't send partial-page writes before EOF from client (4410)
6288 - discard client grant for sub-page writes on large-page clients (4520)
6289 - don't free dentries not owned by NFS code, check generation (4806)
6290 - fix lsm leak if mds_create_objects() fails (4801)
6291 - limit debug_daemon file size, always print CERROR messages (4789)
6292 - use transno after validating reply (3892)
6293 - process timed out requests if import state changes (3754)
6294 - update mtime on OST during writes, return in glimpse (4829)
6295 - add mkfsoptions to LDAP (4679)
6296 - use ->max_readahead method instead of zapping global ra (5039)
6297 - don't interrupt __l_wait_event() during strace
6299 - add software watchdogs to catch hung threads quickly (4941)
6300 - make lustrefs init script start after nfs is mounted
6301 - fix CWARN/ERROR duplication (4930)
6302 - return async write errors to application if possible (2248)
6303 - add /proc/sys/portal/memused (bytes allocated by PORTALS_ALLOC)
6304 - print NAL number in %x format (4645)
6305 - update barely-supported suse-2.4.21-171 series (4842)
6306 - support for sles 9 %post scripts
6307 - support for building 2.6 kernel-source packages
6308 - support for sles km_* packages
6310 2004-10-07 Cluster File Systems, Inc. <info@clusterfs.com>
6313 - ignore -ENOENT errors in osc_destroy (3639)
6314 - notify osc create thread that OSC is being cleaned up (4600)
6315 - add nettype argument for llmount in #5d in conf-sanity.sh (3936)
6316 - reconstruct ost_handle() like mds_handle() (4657)
6317 - create a new thread to do import eviction to avoid deadlock (3969)
6318 - let lconf resolve symlinked-to devices (4629)
6319 - don't unlink "objects" from directory with default EA (4554)
6320 - hold socknal file ref over connect in case target is down (4394)
6321 - allow more than 32000 subdirectories in a single directory (3244)
6322 - fix blocks count for O_DIRECT writes (3751)
6323 - OST returns ENOSPC from object create when no space left (4539)
6324 - don't send truncate RPC if file size isn't changing (4410)
6325 - limit OSC precreate to 1/2 of value OST considers bogus (4778)
6326 - bind to privileged port in socknal and tcpnal (3689)
6328 - rate limit CERROR/CWARN console message to avoid overload (4519)
6329 - GETFILEINFO dir ioctl returns LOV EA + MDS stat in 1 call (3327)
6330 - basic mmap support (3918)
6331 - kernel patch series update from b1_4 (4711)
6333 2004-09-16 Cluster File Systems, Inc. <info@clusterfs.com>
6336 - avoid crash during MDS cleanup with OST shut down (2775)
6337 - fix loi_list_lock/oig_lock inversion on interrupted IO (4136)
6338 - don't use bad inodes on the MDS (3744)
6339 - dynamic object preallocation to improve recovery speed (4236)
6340 - don't hold spinlock over lock dumping or change debug flags (4401)
6341 - don't zero obd_dev when it is force cleaned (3651)
6342 - print grants to console if they go negative (4431)
6343 - "lctl deactivate" will stop automatic recovery attempts (3406)
6344 - look for existing locks in ldlm_handle_enqueue() (3764)
6345 - don't resolve lock handle twice in recovery avoiding race (4401)
6346 - revalidate should check working dir is a directory (4134)
6348 - don't always mark "slow" obdfilter messages as errors (4418)
6350 2004-08-24 Cluster File Systems, Inc. <info@clusterfs.com>
6353 - don't close LustreDB during write_conf until it is done (3860)
6354 - fix typo in lconf for_each_profile (3821)
6355 - allow dumping logs from multiple threads at one time (3820)
6356 - don't allow multiple threads in OSC recovery (3812)
6357 - fix debug_size parameters (3864)
6358 - fix mds_postrecov to initialize import for llog ctxt (3121)
6359 - replace config semaphore with spinlock (3306)
6360 - be sure to send a reply for a CANCEL rpc with bad export (3863)
6361 - don't allow enqueue to complete on a destroyed export (3822)
6362 - down write_lock before checking llog header bitmap (3825)
6363 - recover from lock replay timeout (3764)
6364 - up llog sem before sending rpc (3652)
6365 - reduce ns lock hold times when setting kms (3267)
6366 - change a dlm LBUG to LASSERTF, to maybe learn something (4228)
6367 - fix NULL deref and obd_dev leak on setup error (3312)
6368 - replace some LBUG about llog ops with error handling (3841)
6369 - don't match INVALID dentries from d_lookup and spin (3784)
6370 - hold dcache_lock while marking dentries INVALID and hashing (4255)
6371 - fix invalid assertion in ptlrpc_set_wait (3880)
6373 - add libwrap support for the TCP acceptor (3996)
6374 - add /proc/sys/portals/routes for non-root route listing (3994)
6375 - allow setting MDS UUID in .xml (2580)
6376 - print the stack of a process that LBUGs (4228)
6378 2004-07-14 Cluster File Systems, Inc. <info@clusterfs.com>
6381 - don't cleanup request in ll_file_open() on failed MDS open (3430)
6382 - make sure to unset replay flag from failed open requests (3440)
6383 - if default stripe count is 0, use OST count for inode size (3636)
6384 - update parent mtime/ctime on client for create/unlink (2611)
6385 - drop dentry ref in ext3_add_link from open_connect_dentry (3266)
6386 - free recovery state on server during a forced cleanup (3571)
6387 - unregister_reply for resent reqs (3063)
6388 - loop back devices mounting and status check on 2.6 (3563)
6389 - fix resource-creation race that can provoke i_size == 0 (3513)
6390 - don't try to use bad inodes returned from MDS/OST fs lookup (3688)
6391 - more debugging for page-accounting assertion (3746)
6392 - return -ENOENT instead of asserting if ost getattr+unlink race (3558)
6393 - avoid deadlock after precreation failure (3758)
6394 - fix race and lock order deadlock in orphan handling (3450, 3750)
6395 - add validity checks when grabbing inodes from l_ast_data (3599)
6397 - add /proc/.../recovery_status to obdfilter (3428)
6398 - lightweight CDEBUG infrastructure, debug daemon (3668)
6399 - change default OSC RPC parameters to be better on small clusters
6400 - turn off OST read cache for files smaller than 32MB
6401 - install man pages and include them in rpms (3100)
6402 - add new init script for (un)mounting lustre filesystems (2593)
6403 - run chkconfig in %post for init scripts (3701)
6404 - drop scimac NAL (unmaintained)
6406 2004-06-17 Cluster File Systems, Inc. <info@clusterfs.com>
6409 - clean kiobufs before and after use (3485)
6410 - strip trailing '/'s before comparing paths with /proc/mounts (3486)
6411 - remove assertions to work around "in-flight rpcs" recovery bug (3063)
6412 - change init script to fail more clearly if not run as root (1528)
6413 - allow clients to reconnect during replay (1742)
6414 - fix ns_lock/i_sem lock ordering deadlock for kms update (3477)
6415 - don't do DNS lookups on NIDs too small for IP addresses (3442)
6416 - re-awaken ptlrpcd if new requests arrive during check_set (3554)
6417 - fix cond_resched (3554)
6418 - only evict unfinished clients after recovery (3515)
6419 - allow bulk resend, prevent data loss (3570)
6420 - dynamic ptlrpc request buffer allocation (2102)
6421 - don't allow unlinking open directory if it isn't empty (2904)
6422 - set MDS/OST threads to umask 0 to not clobber client modes (3359)
6423 - remove extraneous obd dereference causing LASSERT failure (3334)
6424 - don't use get_cycles() when creating temp. files on the mds (3156)
6425 - hold i_sem when setting i_size in ll_extent_lock() (3564)
6426 - handle EEXIST for set-stripe, set proper directory name (3336)
6428 - servers can dump a log evicting a client - lustre.dump_on_timeout=1
6429 - fix ksocknal_fmb_callback() error messages (2918)
6431 2004-05-27 Cluster File Systems, Inc. <info@clusterfs.com>
6434 - don't copy lvb into (possibly NULL) reply on error (2983)
6435 - don't deref dentry after dput, don't free lvb on error (2922)
6436 - use the kms to determine writeback rpc length (2947)
6437 - increment oti_logcookies when osc is inactive (2948)
6438 - update client's i_blocks count via lvb messages (2543)
6439 - handle intent open/close of special files properly (1557)
6440 - mount MDS with errors=remount-ro, like obdfilter (2009)
6441 - initialize lock handle to avoid ASSERT on error cleanup (3057)
6442 - don't use cancelling-locks' kms values (2947)
6443 - use highest lock extent for kms, not last one (2925)
6444 - don't dereference ERR_PTR() dentry in error handling path (3107)
6445 - fix thread race in portals_debug_dumplog() (3122)
6446 - create lprocfs device entries at setup instead of at attach (1519)
6447 - common AST error handler, don't evict client on completion race (3145)
6448 - zero nameidata in detach_mnt in 2.6 (3118)
6449 - verify d_inode after revalidate_special is valid in 2.6 (3116)
6450 - use lustre_put_super() to handle zconf unmounts in 2.6 (3064)
6451 - initialize RPC timeout timer earlier for 2.6 (3219)
6452 - don't dereference NULL reply buffer if mdc_close was never sent (2410)
6453 - print nal/nid for unknown nid (3258)
6454 - additional checks for oscc recovery before doing precreate (3284)
6455 - fix ll_extent_lock() error return code for 64-bit systems (3043)
6456 - don't crash in mdc_close for bad permissions on open (3285)
6457 - zero i_rdev for non-device files (3147)
6458 - clear page->private before handing to FS, better assertion (3119)
6459 - tune the read pipeline (3236)
6460 - fix incorrect decref of invalidated dentry (2350)
6461 - provide read-ahead stats and refine rpc in flight stats (3328)
6462 - don't hold journal transaction open across create RPC (3313)
6463 - update atime on MDS at close time (3265)
6464 - close LDAP connection when recovering to avoid server load (3315)
6465 - update iopen-2.6 patch with fixes from 2399,2517,2904 (3301)
6466 - don't leak open file on MDS after open resend (3325)
6467 - serialize filter_precreate and filter_destroy_precreated (3329)
6468 - loop device shouldn't call sync_dev() for nul device (3092)
6469 - clear page cache after eviction (2766)
6470 - resynchronize MDS->OST in background (2824)
6471 - refuse to mount the same filesystem twice on same mountpoint (3394)
6472 - allow llmount to create routes for mounting behind routers (3320)
6473 - push lock cancellation to blocking thread for glimpse ASTs (3409)
6474 - don't call osc_set_data_with_check() for TEST_LOCK matches (3159)
6475 - fix rare problem with rename on htree directories (3417)
6477 - allow default OST striping configuration per directory (1414)
6478 - fix compilation for qswnal for 2.6 kernels (3125)
6479 - increase maximum number of MDS request buffers for large systems
6480 - change liblustreapi to be useful for external progs like lfsck (3098)
6481 - increase local configuration timeout for slow disks (3353)
6482 - allow configuring ldlm AST timeout - lustre.ldlm_timeout=<seconds>
6484 2004-03-22 Cluster File Systems, Inc. <info@clusterfs.com>
6487 - fixes for glimpse AST timeouts / incorrectly 0-sized files (2818)
6488 - don't overwrite extent policy data in reply if lock was blocked (2901)
6489 - drop filter export grants atomically with removal from device (2663)
6490 - del obd_self_export from work_list in class_disconnect_exports (2908)
6491 - don't LBUG if MDS recovery times out during orphan cleanup (2530)
6492 - swab reply message in mdc_close, other PPC fixes (2464)
6493 - fix destroying of named logs (2325)
6494 - overwrite old logs when running lconf --write_conf (2264)
6495 - bump LLOG_CHUNKSIZE to 8k to allow for larger clusters (2306)
6496 - fix race in target_handle_connect (2898)
6497 - mds_reint_create() should take same inode create lock (2926)
6498 - correct journal credits calculated for CANCEL_UNLINK_LOG (2931)
6499 - don't close files for self_export to avoid uninitialized obd (2936)
6500 - allow MDS with the same name as client node (2939)
6501 - hold dentry reference for closed log files for unlink (2325)
6502 - reserve space for all logs during transactions (2059)
6503 - don't evict page beyond end of stripe extent (2925)
6504 - don't oops on a deleted current working directory (2399)
6505 - handle hard links to targets without a parent properly (2517)
6506 - don't dereference NULL lock when racing during eviction (2867)
6507 - don't grow lock extents when lots of conflicting locks (2919)
6509 2004-03-04 Cluster File Systems, Inc. <info@clusterfs.com>
6512 - account for cache space usage on clients to avoid data loss (974)
6513 - lfsck support in lustre kernel code (2349)
6514 - reduce journal credits needed for BRW writes (2370)
6515 - orphan handling to avoid losing space on client/server crashes
6516 - ptlrpcd can be blocked, stopping ALL progress (2477)
6517 - use lock value blocks to assist in proper KMS, faster stat (1021)
6518 - takes i_sem instead of DLM locks internally on obdfilter (2720)
6519 - recovery for initial connections (2355)
6520 - fixes for mds_cleanup_orphans (1934)
6521 - abort_recovery crashes MDS in b_eq (mds_unlink_orphan) (2584)
6522 - block all file creations until orphan recovery completes (1901)
6523 - client remove rq_connection from request struct (2423)
6524 - conf-sanity test_5, proper cleanup in umount log not availale (2640)
6525 - recovery timer race (2670)
6526 - mdc_close recovey bug (2532)
6527 - ptlrpc cleanup bug (2710)
6528 - mds timeout on local locks (2588)
6529 - namespace lock held during RPCs (2431)
6530 - handle interrupted sync write properly (2503)
6531 - don't try to handle a message that hasn't been replied to (2699)
6532 - client assert failure during cleanup after abort recovery (2701)
6533 - leak mdc device after failed mount (2712)
6534 - ptlrpc_check_set allows timedout requests to complete (2714)
6535 - wait for inflight reqs when ptlrpcd finishes (2710)
6536 - make sure unregistered services are removed from the srv_list
6537 - reset bulk XID's when resending them (caught by 1138 test)
6538 - unregister_bulk after timeout
6539 - fix lconf error (2694)
6540 - handle write after unfinished setstripe, stripe-only getstripe (2388)
6541 - readahead locks pages, leaves pending causing memory pressure (2673)
6542 - increase OST request buffers to 4096 on large machines (2729)
6543 - fix up permission of existing directories in simple_mkdir (2661)
6544 - init deleted item, add assertions ptlrpc_abort_inflight() (2725)
6545 - don't assign transno to errored transactions (2742)
6546 - don't delete objects on OST if given a bogus objid from MDS (2751)
6547 - handle large client PAGE_SIZE readdir on small PAGE_SIZE MDS (2777)
6548 - if rq_no_resend, then timeout request after recovery (2432)
6549 - fix MDS llog_logid record size, 64-bit array alignment (2733)
6550 - don't call usermode_helper from ptlrpcd, DEFAULT upcall (2773)
6551 - put magic in mount.lustre data, check for bad/NULL mount data (2529)
6552 - MDS recovery shouldn't delete objects that it has given out (2730)
6553 - if enqueue arrives after completion, don't clobber LVB (2819)
6554 - don't unlock pages twice when trigger_group_io returns error (2814)
6555 - don't deref NULL rq_repmsg if ldlm_handle_enqueue failed (2822)
6556 - don't write pages to disk if there was an error (1450)
6557 - don't ping imports that have recovery disabled (2676)
6558 - take buffered bytes into account when balancing socknal conn (2817)
6559 - hold a DLM lock over readdir always, use truncate_inode_pages (2706)
6560 - reconnect unlink llog connection after MDS reconnects to OST (2816)
6561 - remove little-endian swabbing of llog records (1987)
6562 - set/limit i_blksize to LL_MAX_BLKSIZE on client (2884)
6563 - retry reposting request buffers if they fail (1191)
6564 - grow extent at grant time to avoid granting a revoked lock (2809)
6565 - lock revoke doesn't evict page if covered by a second lock (2765)
6566 - disable VM readahead to avoid reading outside lock extents (2805)
6568 - return LL_SUPER_MAGIC from statfs for the filesystem type (1972)
6569 - updated kernel patches for hp-2.4.20 kernel (2681)
6571 2004-02-07 Cluster File Systems, Inc. <info@clusterfs.com>
6574 - fix truncated write corruption (2366)
6575 - fix for failed assertion in iopen_connect_dentry (1792,2517)
6577 - don't flag the ptlrpcd thread with PF_MEMALLOC (2636)
6578 - ensure len(uuid) < 37 in lmc (1171)
6579 - fix ia64 OOPS in llog_test (2255)
6580 - zero end of page at obdfilter for partial page writes (2648)
6581 - don't leave stale dentries around after renames (bug 2428)
6582 - fix timeouts when evicting a client with a single lock held (2642)
6583 - set deadline for the initial HELLO message to drain (2634)
6584 - print out dotted-quad IP addresses in the socknal (2302)
6586 - additional debugging for MDS client eviction problem (2443)
6587 - fix mkfsoptions support for osts (2603, 2604)
6589 2004-01-27 Cluster File Systems, Inc. <info@clusterfs.com>
6592 - add series for the vanilla 2.6.0 kernel
6593 - add series for the vanilla 2.4.24 kernel
6594 - add series for a cray x86/64 UL kernel drop
6595 - fix xattr patches for the vanilla 2.4.19 series
6597 - generate true UUIDs in lmc (1171)
6598 - have portals stack dumping break in UML (2466)
6599 - avoid bad dchild deref; avoid inum lock w/o creation (2362)
6600 - allocate with _NOFS in ldlm to avoid deadlock (1933)
6601 - wake callback waiting threads on client eviction (2460)
6602 - Add --ptldebug and --subsystem to lmc (1719)
6603 - update assertion to allow safe interrupt allocation
6604 - set rq_no_resend for cancel requests (2432)
6605 - recalculate ptlrpcd timeout after resend (2494)
6606 - call vfs_rmdir when removing pending directories (2368)
6607 - fix renaming a file to itself (2429)
6608 - lmc creates a default one-stripe lov (2454)
6609 - expand procfs space to handle large clusters (2326)
6610 - increase UML stack to avoid overflow
6611 - update lconf's list of debug and subsystem masks
6612 - fix lfs find --obd (2510)
6613 - /proc tunable for disabling filter read caching (2591)
6614 - stop rpm packages from altering slapd.conf (2301)
6615 - disable nagle in the socknal under 0conf (2578)
6616 - choose mds inode size based on stripe count (2572)
6617 - fix kernel-source rpm problems (2516)
6619 - add --disable-doc to avoid pdf generation (2421)
6620 - update documentation, tests, type-os, comments
6621 - avoid format warnings on ia64
6622 - remove the TOE NAL
6623 - tiny code cleanups by removing unused fields
6625 2004-01-07 Cluster File Systems, Inc. <info@clusterfs.com>
6628 - fix obvious semaphore misuse in as-yet-unused setattr path (2348)
6629 - remove the most blatant lies from BUILDING file (2371)
6630 - change default debug level to reasonable production setting
6631 - reduce client side cache size to reduce cache flush time
6632 - reduce max RPCs in flight to avoid unnecessary file fragmentation
6633 - make TCP zerocopy and pinger support enabled by default (2476)
6634 - sync writes completed after process exits caused crashes (2319)
6635 - maintain correct mount count on the MDS (2356)
6636 - backout 1557, because 2316 wasn't really fixed
6637 - better file I/O statistics gathering in /proc
6638 - don't take unnecessary, deadlock-inducing bug in readpage (2383)
6639 - another kernel patch to fix zero-copy TCP function export
6640 - don't take duplicate lock when processing re-sent getattr (2420)
6641 - lctl uses obd_self_export instead of creating new conn (2353)
6642 - MDS/OST recovery case which requires object creation asserted (2425)
6643 - move lfs from /usr/sbin to /usr/bin in packages
6644 - fix race between mds_client_add and mds_client_free (2417)
6645 - use kmalloc instead of slabs in portals (2430)
6646 - don't create duplicate records when a failover MDS is present (2442)
6647 - remove unnecessary mount age check (2332)
6648 - don't remove directory inodes from locks prematurely (2451)
6649 - don't break if MDS service name is the same as hostname (2103)
6650 - fix races in client write RPC generation when cache full (2482)
6652 2003-12-13 Cluster File Systems, Inc. <info@clusterfs.com>
6655 - remove now-unused request->rq_obd (278)
6656 - if an allocation fails, print out how much memory we've used (1933)
6657 - use PORTAL_SLAB_ALLOC for structures, to get GFP_MEMALLOC (1933)
6658 - add the "configurable stack size" patch to most series files (1256)
6659 - ability to write large log records, for 100+ OST configs (2306)
6660 - fix NULL deref when filter_prep fails (2314)
6661 - fix operator precedence error in filter_sync
6662 - dynamic allocation of socknal TX descriptors (2315)
6663 - fix a missed case in the GFP_MEMALLOC patch, can cause deadlock (2310)
6664 - fix gcc 2.96 compilation problem in xattr kernel patch (2294)
6665 - ensure that CWARN messages in Portals always get to the syslog
6666 - __init/__exit are not for prototype decls (ldlm_init/exit)
6667 - x86-64 compile warning fixes
6668 - fix gateway LMC keyword conflict (2318)
6669 - fix MDS lock inversions in getattr/reint paths (1844)
6670 - fix a rare lock re-ordering bug, which caused deadlock (2322)
6671 - fix i_sem/journal inversion in fsfilt_ext3_write_record (2306)
6672 - DLM race condition prevented some lock evictions (2328)
6673 - ENOMEM detection and retry on socknal sends (2230)
6674 - use GFP_NOFS throughout Lustre, to combat ENOMEM (2230)
6675 - move osc_rpcd into ptlrpc, for use in MDC and others (2329)
6676 - protect MDS inode fsdata with stronger locking; fixes assertion (2313)
6677 - better error messages when a client is rejected during recovery (1505)
6678 - avoid cancelling locks which were never granted, after failure (2330)
6679 - fix i_sem/journal inversion in mds_client_add (2333)
6680 - fix truncate/getattr lock cycle deadlock (2334)
6681 - use rpcd to send close; allows resend after timeout, avoid leak (1897)
6682 - fix two rare exit paths which could leak an l_lock() ref (2321)
6683 - fencepost error in MDS/OST orphan recovery (2226)
6684 - make log record alignment 8 bytes (1988)
6685 - lstripe now fails when requested offset > ost_count (2237)
6686 - ensure that all kernel series have a complete list.h (1607)
6687 - fix crashes in special-file operations (2316)
6688 - lctl create/brw OID mismatch, caused by obsolete filter loop (2339)
6690 - allow configurable automake binary, for testing new versions
6691 - small update to the lfs documentation
6693 2003-12-03 Cluster File Systems, Inc. <info@clusterfs.com>
6695 * fix negative export reference count in fsfilt_sync (2312)
6697 2003-12-01 Cluster File Systems, Inc. <info@clusterfs.com>
6698 * release candidate 0.9.1
6700 - orphans are moved into the PENDING directory for possible recovery
6701 - replayed opens now open by fid for orphan/rename safety (1042)
6702 - last close of an orphan inode generates a transno (683)
6703 - chdir() and mount() now pin the directory entry (1020)
6704 - avoid CERROR in normal ll_setattr_raw() error case (1500)
6705 - discard very old requests without processing them (1502)
6706 - remove some common, well-understood CERRORs (1505)
6707 - require O_DIRECT I/O to be page-sized to workaround IA64 crash (1609)
6708 - clear "grant" flags in OST replies until OST grant code lands (1644)
6709 - fix read performance by not clobbering i_blksize on client (1598)
6710 - fix __ldlm_handle2lock oops by not dereferencing lock after PUT (1625)
6711 - make LRU size a /proc tunable, clears locks when reduced (707)
6712 - fix some lprocfs rot that prevented ptlbd from loading (1732)
6713 - server locks take references on exports now (1558)
6714 - build fixes for 2.4.20-rh trees (1663)
6715 - return an error from lov_create if all OSCs are inactive (1751)
6716 - fix import levels when a reconnect happens without a timeout (1597)
6717 - exit early from mds_open if we get a lookup error (1749)
6718 - partial page read at EOF wouldn't wait for disk before sending (1642)
6719 - avoid NULL deref in obdfilter when reading page past EOF (1592)
6720 - avoid LASSERT in ll_intent_lock if server failed very early (1090)
6721 - fix LBUG in ll_it_open_error with rc = -2 (1861)
6722 - write/truncate lock inversion (1639)
6723 - Don't auto-load obdclass, portals modules during cleanup (1495)
6724 - fix timestamps from jumping to "now" (1763)
6725 - extra journal assertions (1648)
6726 - add an extra multiunlink test (1771)
6727 - fix read_record/write_record API (1776)
6728 - fix leak of offset_extent, possible incorrect i_size later (1772)
6729 - fix lasserts in mis-matched transnos during open-unlink testing (1541)
6730 - Debugging for the kqswnal_get_idle_tx problems (1820)
6731 - Allow recovery to be attempted multiple times (1536)
6732 - Write out MDS last_rcvd file after it is first created (1600)
6733 - Fix tx_descriptor leak in failed transmit situations (1827)
6734 - ext3 journaling fixes for assertion failure after IO error (1871)
6735 - class_export_put() on freed export after completion AST error (1896)
6736 - Fix revalidate looping in VFS (1322)
6737 - Don't access a freed export during MDS_REINT timeout (1521)
6738 - Add open-unlink recovery support on the MDS (1673,1764)
6739 - Return an error if no MDS data was read from last_rcvd (1946)
6740 - Fix for lookup "." or ".." crash on error (1932,1931,1935)
6741 - Don't setup a disk device that doesn't match exported UUID (317)
6742 - Reduce bulk RPC timeout to avoid cascading client/OST failures (1845)
6743 - avoid committing NULL handle in force close
6744 - local.sh is now a one-stripe LOV configuration
6745 - POSIX utime.4 -EPERM on FIFO not owned by user (56)
6746 - fix ext3 htree duplicate directory entry corruption (1516)
6747 - POSIX creat.13, fstat.1, open.18, stat.3 new file atime/mtime (2020)
6748 - update to new LOV EA format (2097)
6749 - interoperability for different PAGE_SIZE/wordsize (686,1821,1343,2042)
6751 2003-06-15 Phil Schwan <phil@clusterfs.com>
6754 - imports and exports cleanup too early, need refcounts (349, 879, 1045)
6755 - per-import/export recovery handling (958, 931, 959)
6756 - multiple last-rcvd slots, for serving multiple FSes (949)
6757 - connections are again shared between multiple imp/exports (963, 964)
6758 - "umount -f" would hang if any requests needed to be sent (393, 978)
6759 - avoid pinning large req buffer by copying for queued messages (989)
6760 - add "uuid" to "lctl device" command to help upcalls (991)
6761 - "open" RPCs with transnos would confuse recovery counters (1037)
6762 - do proper endian conversion of all wire messages (288, 340, 891)
6763 - remove OST bulk get LBUGs, fix ost_brw_write cleanup (1126)
6764 - call waiting locks callback from LDLM recovery thread (1127, 1151)
6765 - fix ptlrpc_connection leak in target_handle_connect (1174)
6766 - fix import refcounting bug in OST and MDS cleanup (1134)
6767 - if an invalid-at-open-time OSC returned before close(), LBUG (1150)
6768 - fix very unlikely obd_types race condition (501)
6769 - remove osc_open hack for echo_client (1187)
6770 - we leaked exports/dlmimps for forcibly disconnected clients (1143)
6771 - a failure in read_inode2 leads to deadlock (1139)
6772 - cancel ack-locks as soon as transaction is committed (1072)
6773 - fix major leaks and crashes in the bulk I/O path (937, 1057)
6774 - make sure to commitrw after any preprw to avoid deadlock (1162)
6775 - failing to execute a file in a lustre FS would lock inode (1203)
6776 - small DEBUG_REQ fix to avoid dereferencing a NULL (1227)
6777 - don't ASSERT while cleaning up an incompletely-setup obd (1248)
6778 - obd_uuid2tgt would walk off the end of the list (1255)
6779 - on IA64 the osc would give portals incorrect bulk size (1258)
6780 - fix debug daemon ioctl interface; allows daemon on ia64 (1274)
6781 - fix lock inversion caused by new llite matching code (1282)
6782 - limit the number of dirty pages on a client to 10MB (1286)
6783 - timed out locks were not being corrected cancelled (1289)
6784 - fix O_DIRECT above 4GB on IA-32 (1292)
6785 * major user-visible changes
6786 - fail out/fail over policy now controlled by the upcall (993)
6788 - add OBD_PING to check server availability and failure (954)
6789 - lustre messages are now sent in sending host order (288, 340, 891)
6790 - add eadatalen to MDS getattr reply (340)
6791 - OST read replies may contain second buffer, with per-page status (593)
6793 2003-03-11 Phil Schwan <phil@clusterfs.com>
6796 - LDLM_DEBUG macro fix, for gcc 3.2 (850)
6797 - failed open()s could cause deadlock; fixed (867, 869)
6798 - stop cancelling OST locks when files are closed (481)
6799 - overlapping XID spaces caused network corruption (851, 853)
6800 - fix unsafe fsfilt counter arithmetic; change to atomic_t
6801 - setattr_raw added, to do single-RPC, server-side setattrs
6802 - lmc/lconf syntax change for OST UUIDs
6803 - fix crashy race condition between ptlrpc_free_req and osc_close
6804 - don't use request in mdc_enqueue if we hit a timeout (889)
6805 - don't set the inode i_size for regular files from the MDS (896)
6806 - handle out of order completion AST (842)
6807 - don't LBUG if a lock request times out after receiving AST (913)
6808 - avoid d_rehash race in ll_find_alias by rehashing inside dcache_lock
6809 - if a bad lock AST arrives, send an error instead of dropping entirely
6810 - return 0 from revalidate2 if ll_intent_lock returns -EINTR (912)
6811 - fix leak in bulk IO when only partially completed (899, 900, 926)
6812 - fix O_DIRECT for ia64 (55)
6813 - (almost) eliminate Lustre-kernel-thread effects on load average (722)
6814 - C-z after timeout could hang a process forever; fixed (977)
6816 - client-side I/O cache (678, 924, 929, 941, 970)
6818 - READPAGE and SETATTRs which don't take server-side locks get
6821 2003-02-11 Phil Schwan <phil@clusterfs.com>
6824 - Fix ldlm_lock_match on the MDS to avoid matching remote locks (592)
6825 - Fix fsfilt_extN_readpage() to read a full page of directory
6826 entries, or fake the remainder if PAGE_SIZE != blocksize (500)
6827 - Avoid extra mdc_getattr() in ll_intent_lock when possible (534, 604)
6828 - Fix imbalanced LOV object allocation and out-of-bound access (469)
6829 - Most intent operations were removed, in favour of a new RPC mode
6830 that does a single RPC to the server and bypasses most of the VFS
6831 - All LDLM resource ID arrays were removed in favour of ldlm_res_id
6832 - Aggressively cancel local locks on DLM servers
6833 - mds_reint_unlink sends EA to the client if it's the last nlink.
6834 client uses that EA to unlink OST objects.
6835 - mds_reint_{rename,unlink,link} were rewritten to take ordered locks
6836 - recursive symlinks were fixed (439)
6837 - fixed NULL deref in DEBUG_REQ
6838 - filter_update_lastobjid no longer calls sync, which annoyed extN
6839 - fixed multi-client small-writes to a single file problem (445)
6840 - fixed mtime updates during file writes (607)
6841 - fixed vector writes on obdfilter causing problems when ENOSPC (670)
6842 - fixed bug in obd_brw_read/write() (under guise of testing 367)
6843 - fixed Linux OST size reporting problem (444, 656)
6844 - OST now updates object mtime with writes or setattr (607, 619)
6845 - client verifies file size before zeroing page past EOF (445)
6846 - OST now writes last allocated objid to disk with allocation (108)
6847 - LOV on echo now works (409)
6849 - mds_reint_unlink sends a new buffer, with the EA included. this
6850 buffer is only valid if body->valid & OBD_MD_FLEASIZE, which is only
6851 set if a regular file was being unlinked, and it was the last link
6852 - use PtlGet from the target for bulk writes (315)
6853 - OST now updates object mtime with writes or setattr (607, 619)
6854 - LDLM now has a grant-time callback to revalidate locked items, if
6856 - Many MDS operations were reorganized to combat race conditions
6858 - Merge b_intel branch (updated lprocfs code) - now at /proc/fs/lustre
6859 - configure check to avoid gcc version 2.96 20000731-2.96-98) (606)
6861 2003-01-06 Andreas Dilger <adilger@clusterfs.com>
6864 - Fully reactivate OST imports after reconnection (512, others)
6865 - Make sure client sees our -ENOTCONN from mds_handle (513 - partial)
6866 - More graceful error handling for truncating on dead OST (515)
6867 - Don't error out unless we're actually accessing dead stripes (474)
6868 - Fix garbage sizes when stripes are missing (410)
6869 - LRU counters were broken, causing constant lock purge (433, 432)
6870 - garbage on read from stripes with failed OSTs (441)
6871 - mark OSCs as active before reconnecting during recovery (438)
6872 - lov_enqueue and lov_cancel need to handle inactive OSTs (403)
6873 - lfind did not preserve OST order in output (443)
6874 - symlinks cause hung clients, incorrect data (439)
6875 - stop dereferencing request after dropping refcount (457)
6876 - don't LASSERT(spin_is_locked) on non-SMP (455)
6877 - fixes for many rename() bugs
6878 - fstat didn't correctly synchronize attributes (399)
6879 - server must handle lock cancellation during blocking AST prep (487)
6880 - bulk descriptors were free()d too soon (511)
6881 - fix paths in lconf, which would load incorrect modules (451, 507)
6882 - fix confusing lconf 'host not found' error message (386)
6883 - fix lock order deadlock on OST (O/R i_sem before journal ops, 478)
6884 - fix race condition in mdc_blocking_ast() for inode access (526)
6885 - fix lov_unpackmd() unpacking wrong number of stripes (537)
6886 - fix lov_set_osc_active() marking wrong OSC inactive (440)
6887 - fix bad lstripe lov_unpackmd() assertion (fix layering too) (527)
6888 - fix multiple writes of stripe MD to MDS (358, maybe 519)
6889 - fix lstripe in several ways (kernel side) (527)
6890 - fix request leak in ldlm_cli_enqueue (262)
6891 - incorrect OSC was marked inactive after OST failure
6892 - call mds_fs_cleanup before unmounting filesystem (524)
6893 - fix races between taking ns_lock and ldlm_lock_change_resource
6894 - fix races updating LOV export open file list
6895 - fix lov_enqueue error path, avoid decref-ing bad lock handle
6896 - fix recovery NULL deref in ldlm_cli_cancel_unused
6897 - fix some DLM races by using new hash table for lock handles (419)
6898 - permit the client to specify desired inodes, at replay
6899 - duplicate requests when we queue them for replay reintegration
6900 - fix last_rcvd offset calculation
6901 - sync after each recovered transaction, so we always make progress
6902 - never, not always, ERESTART requests without transnos
6903 - store the lov_desc in the MDS, so we don't depend on getlovinfo to
6905 - skip replay if the MDS says that the client is already connected
6906 - don't check for a recovery-enabled export to match lctl's UUID
6907 - don't INC_USE_COUNT for phantom exports
6908 - don't crash when cleaning up phantom exports (567)
6909 - don't double-finish or set replay data for errored mdc_open requests
6910 - abort requests when they time out, so we don't get old replies
6911 - send/receive replies for AST messages again
6912 - if the client says that it doesn't have the lock, cancel it on the
6914 - if we timeout during I/O, don't try to cancel an in-use lock; instead
6915 mark it as destroyed, it will all work out when decref is called
6916 - fix module use counts (22, 581)
6918 - ASTs now expect a reply (server cancels lock on error reply)
6920 2002-12-02 Andreas Dilger <adilger@clusterfs.com>
6923 - fix many simultaneous client startup (392)
6924 - fix dentry->d_it clobbering
6925 - credentials weren't being shipped for readdir/getattr operations
6926 - remove invalid assertions triggered during some concurrent MD
6928 - proper Lustre versions added (336, 389)
6929 - fix memory leak for create error case (398)
6930 - fix LOV locking bug that would get cli/srv out of sync
6931 - fix echo client over LOV (409)
6932 - fix dbench 2, extN refcount problem (170, 258, 356, 418)
6933 - fix double-O_EXCL intent crash (424)
6934 - avoid sending multiple lock CANCELs (352)
6936 - MDS can do multi-client recovery (modulo bugs in new code)
6938 - many updates, edits, cleanups
6940 2002-11-18 Phil Schwan <phil@clusterfs.com>
6943 - fix null d_it dereference (346)
6944 - fix full OST/dbench hang (333)
6945 - fix permission problem with file removal (286)
6946 - fix removal of OSCs from LOV when they fail
6947 - fix NULL deref during bulk timeout (214)
6948 - fix problems related to multiple filesystems on one MDS (241)
6949 - fixed serious subtle metadata locking bugs
6950 - free locks on clients when inodes are removed due to memory
6952 - fix inode pointer in lock data (285)
6953 - partial support for multiple MDS on a single host (241)
6954 - data locks weren't cancelled at clear_inode time (290, 311)
6955 - intent locks could lead to unbounded lock growth (205)
6956 - added a maximum lock count, an LRU list, and a flusher
6957 - fix multiple rename (365)
6958 - properly abstracted the echo client
6959 - OSC locked 1 byte too many; fixed
6960 - rewrote brw callback code:
6961 - fixed recovery bugs related to LOVs (306)
6962 - fixed too-many-pages-in-one-write crash (191)
6963 - fixed (again) crash in sync_io_timeout (214)
6964 - probably fixed callback-related race (385)
6966 - Add capability to MDS protocol
6967 - LDLM cancellations and callbacks on different portals
6969 2002-10-28 Andreas Dilger <adilger@clusterfs.com>
6972 - limit client IOV size to PTL_MD_MAX_IOV (611336, 191)
6973 - defer open object destruction to close time (601981, 138)
6974 - open/close OST file handle in obdo (OBD_MD_FLHANDLE) (601981, 138)
6975 - move LDLM_ENQUEUE/CONVERT back to MDS portal (625069)
6976 - abstract ll_lookup2, fix ll_revalidate2 to use abstraction (256)
6977 - don't call obd_setattr in ll_file_release for destroyed objects
6978 * protocol change to lustre_msg: move |version| and add |flags|
6979 * protocol change to osc_punch: "start" in "o_size", "end" in "o_blocks"
6980 * lock replay: for LDLM_FL_REPLAY trust client to do right thing
6981 * added replay of create, unlink, link and rename operations during
6982 MDS failover; recovery should be much more robust now
6983 * remove failed OSCs from LOVs (only lov_create uses this so far)
6984 * the lustre-HOWTO was brought (more) up to date (582544)
6986 2002-10-23 Phil Schwan <phil@clusterfs.com>
6989 - in-use dentries weren't being reused properly (617851)
6990 - prevent multiple LDLM setup (599178)
6991 - fix LOV size calculations for truncate (617853)
6992 - fix client handling of MDS intent errors (POSIX)
6993 - fix permission bug in lovstripe.c test (624321)
6994 - fix MDS thread deadlock - move LDLM handler to DLM portal (625069)
6995 - truncate past end of file could corrupt data
6996 - proper cleanup after timeouts, crashes, etc (592524, 550815)
6997 - a race in recovery could return ETIMEDOUT to apps (623947)
6998 - building outside the source directory was fixed
6999 * the lustre-HOWTO was brought (more) up to date (582544)
7000 * major progress was made on recovery functionality
7002 2002-10-10 Phil Schwan <phil@clusterfs.com>
7005 - recovery deadlock fix
7006 - rm -rf causes LBUG fix (617817)
7007 - file open by multiple tasks fix (618962)
7008 - directory permissions bugs (602707 and 620007)
7009 - journal_stop fixed with locking (611313)
7010 - O_APPEND failures resolved (618273, perhaps 614459)
7011 - lconf PATH fix (619770)
7012 - IA64 build fix (621450)
7013 - RPC buffer sizes scale with amount of memory
7015 2002-10-01 Phil Schwan <phil@clusterfs.com>
7018 - locks would be cancelled without throwing away data pages,
7019 resulting in inconsistent data (605627)
7020 - inode attributes were not always being refreshed (605627, 612449)
7021 - lconf now continues to cleanup after lctl reports an error
7022 - MDS now enforces user permissions (602707)
7023 - lprocfs cleanup fixed, but not yet enabled (614157)
7024 - fixed infinite server hang, should a client not respond to an AST
7025 - avoid going into recovery if user calls readlink() with a buffer
7026 that's too small (613941)
7027 - AST RPCs no longer require replies (614867) -- this may be changed
7028 - don't crash server if client sends an IOV that's too big (611336)
7029 - fixed lock conversion deadlock (611892)
7030 - fixed the following of symlinks (614622)
7031 * recovery: the server can remove locks from a client that dies, other
7032 clients can make progress
7033 * more extN patch fixes
7034 * compile-time configurable ptlrpc buffer allocations
7036 - collaborative read cache document
7037 - Lustre Lite Performance CDR document-in-progress
7039 2002-09-20 Andreas Dilger <adilger@clusterfs.com>
7042 - fix typo in patch-2.4.18
7044 2002-09-20 Andreas Dilger <adilger@clusterfs.com>
7047 - clear ptlrpc request each time in handle_incoming_request()
7048 - unlink of files now destroys the object on the OST
7050 2002-09-19 Peter Braam <braam@clusterfs.com>
7052 * add hard link support
7053 * change obdfile creation method
7054 * kernel patch changed
7056 2002-09-19 Peter Braam <braam@clusterfs.com>
7059 - stack overflow bug in extN fixed
7061 2002-09-18 Andreas Dilger <adilger@clusterfs.com>
7063 * documentation updates
7064 - add man pages for config tools
7065 - update tests/README to describe testing with new config tools
7066 - finish metadata API descriptions
7067 * bug fixes and cleanups
7068 - statfs workaround for 16TB limit
7069 - LOV stripe allocation improved, can stripe on subset of OSTs
7070 - LOV file size/IO offset was wrong for files > 4GB in size
7071 - object EA data was being dropped, caused files to be unreadable
7072 - memory overflow with non-LOV OST caused memory corruption
7073 - fixed regression tests to work with new config tools, obdfilter
7074 - fixed bug when directory size became larger than 1 block
7075 - fixed bug (for single client case) when PWD was deleted
7076 - invalidate local directory pages when doing intent-based ops
7077 - avoid LDLM oops when lock callback contained bad data
7079 2002-09-09 Andreas Dilger <adilger@clusterfs.com>
7081 * documentation updates
7082 * bug fixes and cleanups
7083 - configuration tools
7086 - 64-bit compile warnings
7087 - 64-bit internal statfs data
7089 * test_brw on persistent OST devices
7091 * lprocfs (disabled)
7093 2002-09-04 Andreas Dilger <adilger@clusterfs.com>
7095 * documentation updates
7096 * bug fixes and cleanups
7097 * configuration tools
7099 2002-08-30 Peter J. Braam <braam@clusterfs.com>
7102 * many small fixes to 0_5_4
7103 * io/network handling
7104 * thinkos in MDS operations
7106 2002-08-24 Peter J. Braam <braam@clusterfs.com>
7109 * crucial basic fixes to 0.5.3
7110 * IOR, Iozone work over Elan
7113 2002-08-07 Phil Schwan <phil@clusterfs.com>
7114 * version 0_5_3, our first alpha
7115 * we use the new Portals iovs
7116 * documentation updates
7117 * bug fixes and cleanups
7118 * small changes in the DLM wire protocol
7120 2002-07-25 Peter J. Braam <braam@clusterfs.com>
7121 * version 0_5_1 with some initial stability,
7122 * locking on MD and file I/O.
7123 * documentation updates
7124 * several bug fixes since 0.5.0
7125 * small changes in wire protocol
7127 2002-07-18 Phil Schwan <phil@clusterfs.com>
7129 * delivered as Lustre Light Alpha
7130 * fixed a crash after handling invalid MDS requests
7131 * fixed directory pages for architectures with non-4k pages sizes
7133 2002-07-11 Andreas Dilger <adilger@clusterfs.com>
7134 * release version v0_4_4
7135 * Moves TCP acceptor to be on port 2432 (unused Coda port) instead
7137 * Fixes a number of interruption problems with OST operations.
7138 * Update documentation for portals header changes
7139 * Move all wire protocol structs/defines to lustre_idl.h
7140 * Fixes symlink length bug.
7141 * Add tcpdump to repository.
7143 2002-07-05 Andreas Dilger <adilger@clusterfs.com>
7144 * release version v0_4_3
7145 * Fixes statfs for inodes on extN.
7146 * Fixes bug in runtests which would delete /etc/hosts.
7147 * Use 64-bit object IDs wherever possible (not into VFS though)
7148 Remove ost_get_info, which is unused by lustre, and out of date.
7150 2002-07-03 Peter Braam <braam@clusterfs.com>
7151 * release version v0_4_2 Fixes a lookup error (type not passed)
7152 * move forward to head of Portals
7153 * move forward to latest Lustre kernel
7155 2002-06-25 Peter Braam <braam@clusterfs.com>
7156 * release version v0_4_1. Hopefully stable on single node use.