1 tbd Sun Microsystems, Inc.
4 2.6.16.60-0.42.4 (SLES 10),
5 2.6.27.39-0.3.1 (SLES11),
6 2.6.18-164.6.1.el5 (RHEL 5)
7 2.6.18-164.6.1.0.1.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
15 Description: Make watchdog timer messages to be more clear and descriptive.
19 Description: quota type and version is not tolerant to resetting
22 Severity : enhancement
24 Description: Update kernel to SLES11 2.6.27.39-0.3.1.
28 Description: Clear imp_force_reconnect correctly in ptlrpc_connect_interpret()
32 Description: Allow non-root access for "lfs check".
33 Details : Added a check in obd_class_ioctl() for OBD_IOC_PING_TARGET.
35 Severity : enhancement
37 Description: Update kernel to RHEL5.4 2.6.18-164.6.1.el5 and
38 OEL5 2.6.18-164.6.1.0.1.el5(Both in-kernel OFED enabled).
42 Description: Sanity test_27p fails.
43 Details : OBD_FAIL_OST_ENOSPC has been moved to filter_statfs().
44 Sanity tests 27 were cleaned up and improved.
46 Severity : enhancement
48 Description: Adjust locks' extents on their first enqueue, so that at the time
49 they get granted, there is no need for another pass through the
50 queues since they are already shaped into the proper forms.
54 Description: Fix mds_shrink_intent_reply()/mds_intent_policy() to pass correct
55 arguments and prevent LBUG() in lustre_shrink_reply_v2().
59 Description: Change tunefs.lustre and mkfs.lustre --mountfsoptions so that
60 exactly the specified mount options are used. Leaving off
61 any "mandatory" mount options is an error. Leaving off any
62 default mount options causes a warning, but is allowed.
63 Change errors=remount-ro from mandatory to default. Sanitize
64 the mount string before storing it. Update man pages accordingly.
68 Description: mds_getattr() should return 0, even if mds_fid2entry() fails with
69 -ENOENT. Also fix in ptlrpc_expire_one_request() to print signed
72 Severity : enhancement
74 Description: Remove set_info(KEY_UNLINKED) from MDS/OSC
78 Description: Drop unnecessary __GFP_NOMEMALLOC flag from filter_get_page()
82 Description: Conf-sanity.sh test to check client reconnection to a busy server.
84 Severity : enhancement
86 Description: Build kernels (RHEL5, OEL5 and SLES10/11) using the vendor's own
91 Description: optimize the extent lock finding during read/write at client side
95 Description: can't stat file in some situation.
96 Details : improve initialize osc date when target is added to mds and
97 ability to resend too big getattr request is client isn't have info
100 Severity : enhancement
102 Description: Update kernel to SLES11 2.6.27.29-0.1.1.
106 Description: Improved cleanup procedure for conf-sanity test_46a.
110 Description: Prevent inconsistences between linux and lustre mount structures.
111 Details : Wait indefinitely in server_wait_finished() until mnt_count drops.
112 Make the sleep interruptible.
114 Severity : enhancement
116 Description: Enhancement for lfs(1) command to use numeric uid/gid.
118 Severity : enhancement
120 Description: Communicate OST degraded/readonly state via statfs to MDS
121 Details : Flags in the statfs returned from OSTs indicate whether the
122 OST is in a degraded RAID state, or if the filesystem has
123 turned read-only after a filesystem error is detected.
125 Severity : enhancement
127 Description: Vanilla kernel 2.6.30 patchless client support.
132 Description: don't panic if EPROTO was hit when reading symlink
133 Details : correctly handling request reference in error cases.
138 Description: open sometimes returns ENOENT instead of EACCES
139 Details : checking permission should be part of open part of mds_open, not
140 lookup part. so server should be set DISP_OPEN_OPEN disposition
141 before starting permission check.
142 Also not need revalidate dentry if client already have LOOKUP lock.
146 Description: enable client interface failover
147 Frequency : on servers with multiple network interfaces
148 Details : When a child reconnects from another NID, properly update export
149 nid hash position and ldlm reverse import.
151 Severity : enhancement
153 Description: implemented direct I/O with arbitrary (nonaligned) memory
154 addresses and file offsets.
156 Severity : enhancement
158 Description: added more recovery timeout options.
160 Severity : enhancement
162 Description: added llapi_file_open, llapi_file_create, llapi_file_get_stripe
167 Description: Avoid deadlock for local client writes
168 Frequency : only on systems with clients writing to an OST on the same node
169 Details : Use new OBD_BRW_MEMALLOC flag to notify OST about writes in the
170 memory freeing context. This allows OST threads to set the
171 PF_MEMALLOC flag on task structures in order to allocate memory
172 from reserved pools and complete IO.
173 Use GFP_HIGHUSER for OST allocations for non-local client writes,
174 so that the OST threads generate memory pressure and allow
175 inactive pages to be reclaimed.
180 Descriptoin: lock ordering violation between &cli->cl_sem and _lprocfs_lock
181 Details : move ldlm namespace creation in setup phase to avoid grab
182 _lprocfs_lock with cli_sem held.
186 Frequency : only during format of test systems
187 Description: Unable to run several mkfs.lustre on loop devices at the same time
188 Details : mkfs.lustre returns error 256 on the concurrent loop devices
189 formatting. The solution is to proper handle the error.
191 Severity : enhancement
193 Description: implement async create (obd_async_create) method for osc, to avoid
194 too long waiting new ost objects with holding ldlm lock.
196 Severity : enhancement
198 Descriptoin: OST grant too much space to client even there are not enough space.
199 Details : Client will shrink its grant cache to OST if there are no write
200 activity over 6 mins (GRANT_SHRINK_INTERVAL), and OST will retrieve
201 this grant cache if there are already not enough avaible space
202 (left_space < total_clients * 32M).
206 Frequency : occasionally during network problems
207 Description: client not allowed to reconnect to OST because of active request
208 Details : abort bulk requests received by the OST once the client has timed
209 out since the client will resend the request anyway. The client
210 also now retries to reconnect to the same server if a connect
211 request failed with EBUSY or -EAGAIN.
215 Frequency : occasionally during network problems
216 Description: client not allowed to reconnect to OST because of active request
217 Details : The client also now retries to reconnect to the same server if a
218 connect request failed with EBUSY or -EAGAIN. The patch has not
219 been landed in the previous commits.
222 Frequency : rare, if used wide striped file and one ost in down.
224 Descriptoin: don't return error if we created a subset of objects for file.
225 Details : lov_update_create_set() uses set->set_success as index for created
226 objects, so if some requests failed, they will have hole at end of
227 array and we can use qos_shrink_lsm for allocate correct lsm.
231 Description: quotacheck performance/scaling issues
232 Details : reduce quotacheck time on empty filesystem by skipping uninit
237 Description: Slow stale export processing during normal start up
238 Details : The global mgc lock prevents OST setup to be run in parallel.
239 Replace the global lock with a per-config_llog_data semaphore.
243 Description: Out or order replies might be lost on replay
244 Details : In ptlrpc_retain_replayable_request if we cannot find retained
245 request with tid smaller then one currently being added, add it
246 to the start, not end of the list.
250 Description: bad entry in directory xxx: inode out of bounds
251 Details : fix locking issue in the rename path which could race with any
252 other operations updating the same directory.
256 Description: BUG: soft lockup - CPU#1 stuck for 10s! [ll_mdt_07:4523]
257 Details : add cond_resched() calls to avoid hogging the cpu for too long
260 Severity : enhancement
262 Description: Performance improvements for debug messages with D_RPCTRACE,
263 D_LDLM, D_QUOTA options.
265 -------------------------------------------------------------------------------
266 2009-10-16 Sun Microsystems, Inc.
268 * Support for kernels:
269 2.6.16.60-0.42.4 (SLES 10),
270 2.6.27.29-0.1 (SLES11, i686 & x84_64 only),
271 2.6.18-128.7.1.el5 (RHEL 5),
272 2.6.18-128.7.1.el5 (OEL 5),
273 * Client support for unpatched kernels:
274 (see http://wiki.lustre.org/index.php?title=Patchless_Client)
275 2.6.16 - 2.6.27 vanilla (kernel.org)
276 * Recommended e2fsprogs version: 1.41.6.sun1
277 * ext4 support for RHEL5 is experimental and thus should not be
280 Severity : enhancement
282 Description: Add OEL5 support.
284 Severity : enhancement
286 Description: Update kernel to SLES11 2.6.27.29-0.1.
290 Description: File checksum failures with OST read cache on
291 Details : Disable page poisoning when the bulk transfer has to be aborted
292 because the client got evicted.
296 Description: Don't allow make backward step on assiging osc next id.
297 Details : race between allocation next id and ll_sync thread can be cause
298 of set wrong osc next id and can be kill valid ost objects.
300 Severity : enhancement
302 Description: Update kernel to RHEL5 2.6.18-128.7.1.el5.
304 Severity : enhancement
306 Description: Update kernel to SLES10 SP2 2.6.16.60-0.42.4.
310 Description: Changes in raid5-large-io-rhel5.patch to calculate sectors properly
314 Description: Increase the default BLK_DEF_MAX_SECTORS value for RHEL5 and SLES11
318 Description: Error handling in osc_statfs_interpret() has been improved.
319 Details : Check in osc_statfs_interpret() for EBADR.
323 Description: Do not send statfs() requests to OSTs disabled by administrator.
324 Details : Check in lov_prep_statfs_set() for non-NULL ltd_exp.
328 Description: Do not update ctime for the deleted inode.
329 Details : Check in mds_reint_unlink() before calling fsfilt_setattr().
333 Description: Increase of the size of the LDLM resource hash.
334 Details : Bump up RES_HASH_BITS=12.
338 Description: correctly send lsm on open replay
339 Details : MDS is trust to LSM size on replay open, but client can set wrong size
344 Description: Deadlock between filter_destroy() and filter_commitrw_write().
345 Details : filter_destroy() does not hold the DLM lock over the whole
346 operation. If the DLM lock is dropped, filter_commitrw() can go
347 through, causing the deadlock between page lock and i_mutex.
348 The i_alloc_sem should also be hold in filter_destroy() while
353 Description: truncate starts GFP_FS allocation under transaction causing deadlock
354 Details : ldiskfs_truncate calls grab_cache_page which may start page
355 allocation under an open transaction. This may lead to
356 calling prune_icache with consequent lustre reentrance.
359 Frequency : only when down/upgrading the MDS to 1.6/1.8 while 1.8 clients are
360 still up and when the OST pool feature is used
362 Description: interop testing got LBUG when run dd with OST pool
363 :LustreError: 30032:0:(llite_lib.c:1913:ll_replace_lsm()) LBUG
364 Details : down/upgrading the MDS to a version that doesn't/does support OST
365 pool can cause clients to crash because the lsm has changed
370 Description: missing tree_status on 1.8.1 RPM build
371 Details : make rpms failed due because the tree_status file is missing.
375 Description: continuing LustreError "mds adjust qunit failed!"
376 Details : don't print message on the console when ->adjust_qunit fails.
380 Description: don't increase ldlm timeout if previous client was evicted
381 Details : if a client doesn't respond to a blocking callback within the
382 adaptive ldlm enqueue timeout, don't adjust the adaptive estimate
383 when the lock is next granted.
387 Description: ost is being unmounted w/o all writes to last_rcvd landing on disk.
388 affects recovery negatively.
389 Details : make sure all exports have been properly destroyed by the zombie
390 thread processed before stopping the target.
394 Description: Performance degradation with O_DIRECT between 1.6 & 1.8.1 b190
395 Details : disable write barrier for ext4/SLES11.
399 Description: Kernel panic - not syncing: Out of memory and no killable
400 processes... on OSS when iozone
401 Details : fix memory leak in the journal checksum patch.
405 Description: group quota "too many blocks" OSS crashes
406 Details : we should keep the same uid/gid for lquota_chkquota() and
407 lquota_pending_commit()
411 Description: LustreError: 9153:0:(quota_context.c:622:dqacq_completion()) LBUG
412 Details : don't LBUG on release quota error. Just a workaround until the
413 problem is understood.
415 ------------------------------------------------------------------------------
416 2009-07-31 Sun Microsystems, Inc.
418 * Support for kernels:
419 2.6.16.60-0.39.3 (SLES 10),
420 2.6.27.23-0.1 (SLES11, i686 & x84_64 only),
421 2.6.18-128.1.14.el5 (RHEL 5),
422 * Client support for unpatched kernels:
423 (see http://wiki.lustre.org/index.php?title=Patchless_Client)
424 2.6.16 - 2.6.27 vanilla (kernel.org)
425 * Recommended e2fsprogs version: 1.41.6.sun1
426 * File join has been disabled in this release, refer to Bugzilla 16929.
427 * NFS export disabled when stack size < 8192. Since the NFSv4 export
428 of Lustre filesystem with 4K stack may cause a stack overflow. For
429 more information, please refer to bugzilla 17630.
430 * ext4 support for RHEL5 is experimental and thus should not be
433 Severity : enhancement
435 Description: Update kernel to SLES10 SP2 2.6.16.60-0.39.3.
438 Frequency : with 1.8 server and 1.6 clients
440 Description: correctly shrink reply for avoid send too big message to client.
441 Details : 1.8 mds is allocate to big buffer to LOV EA data and this produce
442 some problems with sending this reply to 1.6 client.
446 Description: Repeated atomic allocation failures.
447 Details : Use GFP_HIGHUSER | __GFP_NOMEMALLOC flags for memory allocations
448 to generate memory pressure and allow reclaiming of inactive pages.
449 At the same time, do not allow to exhaust emergency pools.
450 For local clients the use of GFP_NOFS will be introduced in 1.8.2
452 Severity : enhancement
453 Bugzilla : 19846, 18289
454 Description: Update kernel to RHEL5 2.6.18-128.1.14.el5.
456 Severity : enhancement
457 Bugzilla : 19625, 16893, 18668, 19848
458 Description: Add support for SLES11 2.6.27.23-0.1.
460 Severity : enhancement
462 Description: Update client support to vanila kernels up to 2.6.27.
464 Severity : enhancement
466 Description: Update kernel to SLES10 SP2 2.6.16.60-0.37.
468 Severity : enhancement
470 Description: Compile with -Werror by default for i686 and x86_64.
474 Description: resolve race between obd_disconnect and class_disconnect_exports
475 Details : if obd_disconnect will be called to already disconnected export he
476 forget release one reference and osc module can't unloaded.
478 Severity : enhancement
480 Description: move AT tunable parameters for more consistent usage
481 Details : add AT tunables under /proc/sys/lustre, add to conf_param parsing
485 Descriptoin: correctly skip time estimate if in recovery
486 Details : rq_send_state insn't bitmask so using bitwise ops is forbid.
490 Descriptoin: OSS DeadLock
491 Details : Use trylock to prevent deadlock when shrink icache.
493 Severity : enhancement
495 Description: Allow tuning service thread via /proc
496 Details : For each service a new
497 /proc/fs/lustre/{service}/*/thread_{min,max,started} entry is
498 created that can be used to set min/max thread counts, and get the
499 current number of running threads.
501 Severity : enhancement
503 Description: Add state history info file, enhance import info file
504 Details : Track import connection state changes in a new osc/mdc proc file;
505 add overview-type data to the osc/mdc import proc file.
509 Description: Reduce small size read RPC
510 Details : Set read-ahead limite for every file and only do read-ahead when
511 available read-ahead pages are bigger than 1M to avoid small size
516 Description: free_entry erroneously used groups_free instead of put_group_info
518 Severity : enhancement
520 Description: Make read-ahead stripe size aligned.
522 Severity : enhancement
524 Description: MDS create should not wait for statfs RPC while holding DLM lock.
527 Frequency : rare, connect and disconnect target at same time
529 Descriptoin: ASSERTION(atomic_read(&imp->imp_inflight) == 0
530 Details : don't call obd_disconnect under lov_lock. this long time
531 operation and can block ptlrpcd which answer to connect request.
534 Frequency : start MDS on uncleanly shutdowned MDS device
536 Descriptoin: ll_sync thread stay in waiting mds<>ost recovery finished
537 Details : stay in waiting mds<>ost recovery finished produce random bugs
538 due race between two ll_sync thread for one lov target. send
539 ACTIVATE event only if connect realy finished and import have
543 Frequency : start MDS on uncleanly shutdowned MDS device
545 Descriptoin: aborting recovery hang on MDS
546 Details : don't throttle destroy RPCs for the MDT.
550 Description: Slow reads beyond 8Tb offsets.
551 Details : Page index integer overflow in ll_read_ahead_page
555 Description: Soft lockup on OSS after MDS failover
556 Details : MSG_CONNECT_INITIAL is not set on the initial MDS->OST connect.
557 As a conseqence, the patch from bug 18224 is not operational
558 and the MDS export cannot be reused on the OSTs until it gets
562 Frequency : rare, only if using MMP with Linux RAID
564 Description: MMP doesn't work with Linux RAID
565 Details : While using HA for Lustre servers with Linux RAID, it is possible
566 that MMP will not detect multiple mounts. To make this work we
567 need to unplug the device queue in RAID when the MMP block is being
568 written. Also while reading the MMP block, we should read it from
569 disk and not the cached one.
572 Frequency : rare, during recovery
574 Description: Assertion failure in ldlm_lock_put
575 Details : Do not put cancelled locks into replay list, hold references on
580 Description: 1.6.5 mdsrate performance is slower than 1.4.11/12 (MDS is not cpu bound!)
581 Details : create_count always drops to the min value (=32) because grow_count
582 is being changed before the precreate RPC completes.
585 Frequency : Only in RHEL5 when mounting multiple ext3 filesystems
588 Description: "kmem_cache_create: duplicate cache jbd_4k" error message
589 Details : add proper locking for creation of jbd_4k slab cache
593 Description: MMP check in ext3_remount() fails without displaying any error
594 Details : When multiple mount protection fails during remount, proper error
599 Description: Rare Client crash on resend if the file was deleted.
600 Details : When file is opened, but open reply is lost and file is
601 subsequently deleted before resend, resend processing logic
602 breaks trying to open the file again, should not try to open.
606 Description: add check for >8TB ldiskfs filesystems
607 Details : ext3-based ldiskfs does not support greater than 8TB LUNs.
608 Don't allow >8TB ldiskfs filesystems to be mounted without
609 force_over_8tb mount option
613 Description: Client locked up when running multiple instances of an app. on
614 multiple mount points
615 Details : ll_shrink_cache() can sleep while holding the ll_sb_lock.
616 Convert ll_sb_lock to a read/write semaphore to fix the problem.
620 Description: Cannot acces an NFS-mounted Lustre filesystem
621 Details : An NFS client cannot access the Lustre filesystem NFS-mounted
622 from a Lustre-client exporting the Lustre filesystem via NFS.
626 Description: panic in ll_statahead_thread
627 Details : grab dentry reference in parent process.
630 -------------------------------------------------------------------------------
632 tbd Sun Microsystems, Inc.
634 * Support for kernels:
635 2.6.16.60-0.31 (SLES 10),
636 2.6.18-128.1.6.el5 (RHEL 5),
637 2.6.22.14 vanilla (kernel.org)
638 * Client support for unpatched kernels:
639 (see http://wiki.lustre.org/index.php?title=Patchless_Client)
640 2.6.16 - 2.6.22 vanilla (kernel.org)
641 * Recommended e2fsprogs version: 1.40.11-sun1
642 * File join has been disabled, refer to Bugzilla 16929.
643 * A new Lustre ADIO driver is available for MPICH2-1.0.7.
644 * NFS export disabled when stack size < 8192. Since the NFSv4 export of
645 Lustre filesystem with 4K stack may cause a stack overflow. For more
646 information, please refer to bugzilla 17630.
648 Severity : enhancement
650 Description: Update to RHEL5.3 kernel-2.6.18-128.1.6.el5.
652 Severity : enhancement
654 Description: Update OFED release to 1.4.1 RC4
656 Severity : major, only with big OST
658 Description: Very poor metadata performance on Infiniband lustre configuration
659 Details : OST object precreation becomes very slow on big OSTs. This is due
660 to the ialloc patch spending too much time scanning groups.
663 Frequency : during recovery
665 Description: don't mix llog inodes with normal.
666 Details : allocate inodes for log in last inode group
671 Description: fix lqs' reference which won't be put in some situations
672 Details : This patch fixes:
673 1. In quota_check_common(), this function will check quota
674 for user and group, but only send one return via "pending".
675 In most cases, the pendings should be same. But that is not
677 2. If quotaoff runs between lquota_chkquota() and
678 lquota_pending_commit(), the same thing will happen too.
679 That is why it comes:
680 - if (!ll_sb_any_quota_active(qctxt->lqc_sb))
683 -------------------------------------------------------------------------------
685 2008-12-31 Sun Microsystems, Inc.
687 * Support for kernels:
688 2.6.5-7.314 (SLES 9),
689 2.6.9-67.0.22.EL (RHEL 4),
690 2.6.16.60-0.31 (SLES 10),
691 2.6.18-92.1.17.el5 (RHEL 5),
692 2.6.22.14 vanilla (kernel.org)
693 * Client support for unpatched kernels:
694 (see http://wiki.lustre.org/index.php?title=Patchless_Client)
695 2.6.16 - 2.6.22 vanilla (kernel.org)
696 * Client support for unpatched kernels:
697 we do not recommend using patchless RHEL4 clients with kernels
698 prior to 2.6.9-55EL (RHEL4U5).
699 * Recommended e2fsprogs version: 1.40.11-sun1
700 * Note that reiserfs quotas are disabled on SLES 10 in this kernel.
701 * RHEL 4 and RHEL 5/SLES 10 clients behaves differently on 'cd' to a
702 removed cwd "./" (refer to Bugzilla 14399).
703 * A new quota file format has been introduced in 1.6.5.
704 The format conversion from prior releases is handled transparently,
705 but releases older than 1.4.12/1.6.5 don't understand this new
706 format. The automatic format conversion can be avoided by running
707 the following command on the MDS:
708 'tunefs.lustre --param="mdt.quota_type=ug1" $MDTDEV'.
709 For more information, please refer to bugzilla 13904.
710 * A new quota file format was introduced in 1.6.6/1.8.0 (kernels 2.6.16+).
711 The format conversion from prior releases is handled transparently,
712 but releases older than 1.6.6/1.8.0 don't understand this new
713 format. The automatic format conversion can be avoided by running
714 the following commands on the MDS and OSS servers (for
715 pre 1.4.12-1.6.5 quota files):
716 'tunefs.lustre --param="mdt.quota_type=ug1" $MDTDEV',
717 'tunefs.lustre --param="ost.quota_type=ug1" $MDTDEV'
718 or (for 1.4.12/1.6.5 quota files)
719 'tunefs.lustre --param="mdt.quota_type=ug2" $MDTDEV',
720 'tunefs.lustre --param="ost.quota_type=ug2" $MDTDEV'
721 For more information, please refer to bugzilla 13904.
722 * Output of lfs quota has been made less detailed by default,
723 old (verbose) output can be obtained by using -v option.
724 * File join has been disabled in this release, refer to Bugzilla 16929.
725 * A new Lustre ADIO driver is available for MPICH2-1.0.7.
726 * NFS export disabled when stack size < 8192. Since the NFSv4 export of
727 Lustre filesystem with 4K stack may cause a stack overflow. For more
728 information, please refer to bugzilla 17630.
730 Severity : enhancement
732 Description: Caching OSS
733 Details : introduce data caching on the OSS. The OSS now relies on the linux
734 kernel page cache to keep recently accessed data in memory.
735 It is worth noting that all write requests are still flushed
736 synchronously as in lustre 1.6.
738 Severity : enhancement
740 Description: Version based recovery
741 Details : introduce finer grained recovery able to detect transaction
742 dependencies and can deal with transaction gaps caused by clients
743 failing at the same time as the server.
745 Severity : enhancement
747 Description: Enable adaptive timeouts by default
748 Details : The Lustre timeout value in /proc/sys/lustre/timeout is now
749 managed dynamically based on server load and should not need
750 to be tuned manually based on cluster size. This allows Lustre
751 to work under a wider variety of system sizes and loads, without
752 unnecessarily causing lengthy recovery times.
754 Severity : enhancement
756 Description: Add OST Pools support
757 Details : File striping can now be set to use an arbitrary pool of OSTs
759 Severity : enhancement
761 Description: add lazystatfs mount option to allow statfs(2) to skip down OSTs
762 Details : allow skip disconnected ost for send statfs request and hide error
766 Frequency : rare, on llog test 6
768 Descriptoin: don't allow connect to already connected import
769 Details : allowing connect to already connected import is hide connecting problem.
772 Frequency : rare, on failed llog setup
774 Descriptoin: don't leak obd reference on failed llog setup
775 Details : for failed llog setup - mgc forget call class_destroy_import for
776 client import, move destroy import to more generic place.
781 Descriptoin: allow kill process which wait statahead result
782 Details : for some reasons 'ls' can stick in waiting result from statahead,
783 in this case need way for kill this process.
786 Frequency : rare, at shutdown
788 Descriptoin: panic at umount
789 Details : llap_shrinker can be raced with killing super block from list and
790 this produce panic with access to already freeded pointer
795 Descriptoin: don't lose wakeup for imp_recovery_waitq
796 Details : recover_import_no_retry or invalidate_import and import_close can
797 both sleep on imp_recovery_waitq, but we was send only one wakeup
803 Descriptoin: panic in mds_open
804 Details : don't confuse mds_finish_transno() with PTR_ERR(-ENOENT)
809 Descriptoin: stuck in cache_remove_extent() or panic with accessing to already
811 Details : release lock refernce only after add page to pages list.
814 Frequency : always with long access acl
816 Descriptoin: mds can't pack reply with long acl.
817 Details : mds don't control size of acl but they limited by reint/getattr
822 Frequency : on remount
824 Description: external journal device not working after the remount
825 Details : clear dev_rdonly flag for external journal devices in
831 Description: shutdown vs evict race
832 Details : client_disconnect_export vs connect request race.
833 if client will evicted at this time - we start invalidate
834 thread without referece to import and import can be freed
840 Description: shrink LOV EAs before replying
841 Details : correctly adjust LOV EA buffer for reply.
846 Description: don't skip ost target if they assigned to file
847 Details : Drop slow OSCs if we can, but not for requested start idx.
848 This means "if OSC is slow and it is not the requested
849 start OST, then it can be skipped, otherwise skip it only
850 if it is inactive/recovering/out-of-space.
852 Severity : enhancement
854 Description: Update to RHEL5 kernel-2.6.18-92.1.17.el5.
857 Frequency : rare, need acl's on inode.
859 Description: client can't handle ost additional correctly
860 Details : if ost was added after client connected to mds client can have
861 hit lnet_try_match_md ... to big messages to wide striped files.
862 in this case need teach client to handle config events about add
863 lov target and update client max ea size at that event.
865 Severity : enhancement
867 Description: Update to sles9 kernel-2.6.5-7.314.
869 Severity : enhancement
871 Description: Update to SLES10 SP2 kernel-2.6.16.60-0.31.
874 Frequency : Create a symlink file with a very long name
876 Description: ldlm_cancel_pack()) ASSERTION(max >= dlm->lock_count + count)
877 Details : If there is no extra space in the request for early cancels,
878 ldlm_req_handles_avail() returns 0 instead of a negative value.
883 Description: mds is deadlocked
884 Details : in rare cases, inode in catalog can have i_no less than have parent
885 i_no, this produce wrong order for locking during open, and
886 parallel unlink can be lock open. this need teach mds_open to grab
887 locks in resource id order, not at parent -> child order.
889 Severity : enhancement
891 Description: Add /proc entry for import status
892 Details : The mdc, osc, and mgc import directories now have
893 an import directory that contains useful import data for debugging
896 Severity : enhancement
898 Description: Re-disable certain /proc logging
899 Details : Enable and disable client's offset_stats, extents_stats and
900 extents_stats_per_process stats logging on the fly.
903 Frequency : Only on FC kernels 2.6.22+
905 Description: oops in statahead
906 Details : Do not drop reference count for the dentry from VFS when lookup,
907 VFS will do that by itself.
909 Severity : enhancement
911 Description: Generic /proc file permissions
912 Details : Set /Proc file permissions in a more generic way to enable non-
913 root users operate on some /proc files.
917 Description: Hitting mdc_commit_close() ASSERTION
918 Details : Properly handle request reference release in
919 ll_release_openhandle().
923 Frequency : only patchless client
924 Description: add workaround for race between add/remove dentry from hash
926 Severity : enhancement
928 Description: Allow OST glimpses to return PW locks
932 Description: LBUG when llog conf file is full
933 Details : When llog bitmap is full, ENOSPC should be returned for plain
938 Description: Prevent import from entering FULL state when server in recovery
942 Description: service mount cannot take device name with ":"
943 Details : Only when device name contains ":/" will mount treat it as
949 Description: replace ptlrpcd with the statahead thread to interpret the async
950 statahead RPC callback
954 Frequency : on recovery
955 Description: I/O failures after umount during fail back
956 Details : if client reconnected to restarted server we need join to recovery
957 instead of find server handler is changed and process self eviction
958 with cancel all locks.
960 Severity : enhancement
962 Description: Update to RHEL5 kernel-2.6.18-92.1.10.el5.
966 Description: Kernel BUG tries to release flock
967 Details : Lustre does not destroy flock lock before last reference goes
968 away. So always drop flock locks when client is evicted and
969 perform unlock regardless of successfulness of speaking to MDS.
971 Severity : enhancement
973 Description: Update to SLES10 SP2 kernel-2.6.16.60-0.27.
975 Severity : enhancement
977 Description: Upcall on Lustre log has been dumped
978 Details : Allow for a user mode script to be called once a Lustre log has
979 been dumped. It passes the filename of the dumped log to the
980 script, the location of the script can be specified via
981 /proc/sys/lnet/debug_log_upcall.
986 Description: avoid messages about idr_remove called for id that is not allocated
987 Details : Move assigment s_dev for clustered nfs to end of initialization,
988 for avoid problem with error handling.
993 Description: avoid Already found the key in hash [CONN_UNUSED_HASH] messages
994 Details : When connection is reused this not moved from CONN_UNUSED_HASH into
995 CONN_USED_HASH and this prodice warning when put connection again
1001 Description: avoid ASSERTION(client_stat->nid_exp_ref_count == 0) failed
1002 Details : release reference to stats when client disconnected, not
1003 when export destroyed for avoid races when client destroyed
1004 after main ost export.
1008 Description: more cleanup in mds_lov
1009 Details : add workaround for get valid ost count for avoid warnings about
1010 drop too big messages, not init llog cat under semphore which
1011 can be blocked on reconnect and break normal replay, fix access
1014 Severity : enhancement
1016 Description: Export bytes_read/bytes_write count on OSC/OST.
1020 Description: Early reply size mismatch, MGC loses connection
1021 Details : Apply the MGS_CONNECT_SUPPORTED mask at reconnect time so
1022 the connect flags are properly negotiated.
1026 Description: Properly propagate oinfo flags from lov to osc for statfs
1027 Details : restore missing copy oi_flags to lov requests.
1031 Description: exports in /proc are broken
1032 Details : recreate /proc entries for clients when they reconnect.
1034 Severity : enhancement
1036 Description: Add man pages for llobdstat(8), llstat(8), plot-llstat(8),
1037 : l_getgroups(8), lst(8), routerstat(8)
1038 Details : included man pages for llobdstat(8), llstat(8),
1039 : plot-llstat(8), l_getgroups(8), lst(8), routerstat(8)
1041 Severity : enhancement
1043 Description: Implement lustre ll_show_options method.
1045 Severity : enhancement
1047 Description: Update to SLES9 kernel-2.6.5-7.312.
1049 Severity : enhancement
1051 Description: Update to RHEL4 kernel-2.6.9-67.0.22.EL.
1055 Description: exports in /proc are broken
1056 Details : recreate /proc entries for clients when they reconnect.
1060 Description: don't fail open with -ERANGE
1061 Details : if client connected until mds will be know about real ost count
1062 get LOV EA can be fail because mds not allocate enougth buffer
1067 Description: Resolve device initialization race
1068 Details : Prevent proc handler from accessing devices added to the
1069 obd_devs array but yet be intialized.
1071 Severity : enhancement
1073 Description: configure's --enable-quota should check the
1074 : kernel .config for CONFIG_QUOTA
1075 Details : configure is terminated if --enable-quota is passed but
1076 : no quota support is in kernel
1078 Severity : enhancement
1080 Description: Update to SLES10 SP2 kernel-2.6.16.60-0.23.
1082 Severity : enhancement
1084 Description: Update to RHEL5 kernel-2.6.18-92.1.6.el5.
1088 Frequency : rare, on PPC clients
1089 Description: don't swab ost objects in response about directory, because
1091 Details : bug similar bug 14856, but in different function.
1093 Severity : enhancement
1095 Description: lfs quota tool enhancement
1096 Details : added units specifiers support for setquota, default to
1097 current uid/gid for quota report, short quota stats by
1098 default, nonpositional parameters for setquota, added
1099 llapi_quotactl manual page.
1101 Severity : enhancement
1103 Description: *optional* service tags registration
1104 Details : if the "service tags" package is installed on a Lustre node
1105 When the filesystem is mounted, a local-node service tag will
1106 be created. See http://inventory.sun.com/ for more information
1107 about the Service Tags asset management system.
1111 Description: Client runs out of low memory
1112 Details : Consider only lowmem when counting initial number of llap pages
1115 Frequency : occasional
1117 Description: add refcount for osc callbacks, so avoid panic on shutdown
1119 Severity : enhancement
1121 Description: Update to RHEL4 kernel-2.6.9-67.0.20.
1124 Frequency : testing only
1126 Description: sanity test 65a fails if stripecount of -1 is set
1127 Details : handle -1 striping on filesystem in ll_dirstripe_verify
1130 Frequency : only in unusual configurations
1132 Description: Kernel panic with find ost index.
1133 Details : lov_obd have panic if some OST's have sparse indexes.
1135 Severity : enhancement
1137 Description: Update to RHEL5 kernel-2.6.18-53.1.21.el5.
1140 Frequency : rarely, if filesystem is mounted with -o flock
1142 Description: do not process already freed flock
1143 Details : flock can possibly be freed by another thread before it reaches
1144 to ldlm_flock_completion_ast.
1147 Frequency : rarely, if filesystem is mounted with -o flock
1149 Description: LBUG during stress test
1150 Details : Need properly lock accesses the flock deadlock detection list.
1153 Frequency : rarely, if binaries are being run from Lustre
1155 Description: oops in page fault handler
1156 Details : kernel page fault handler can return two special 'pages' in
1157 error case, don't try dereference NOPAGE_SIGBUS and NOPAGE_OMM.
1160 Frequency : rarely, during shutdown
1162 Description: timeout with invalidate import.
1163 Details : ptlrpcd_check call obd_zombie_impexp_cull and wait request which
1164 should be handled by ptlrpcd. This produce long age waiting and
1165 -ETIMEOUT ptlrpc_invalidate_import and as result LASSERT.
1171 Description: ASSERTION(CheckWriteback(page,cmd)) failed
1172 Details : badly clear PG_Writeback bit in ll_ap_completion can produce false
1176 Frequency : only with broken builds/installations
1178 Description: no LBUG if lquota.ko and fsfilt_ldiskfs.ko are different versions
1179 Details : just return an error to a user, put a console error message
1181 Severity : enhancement
1183 Description: Update to RHEL5 kernel-2.6.18-53.1.19.el5.
1185 Severity : enhancement
1187 Description: Update to RHEL4 kernel-2.6.9-67.0.15.
1189 Severity : enhancement
1191 Description: enable MGS and MDT services start separately
1192 Details : add a 'nomgs' option in mount.lustre to enable start a MDT with
1193 a co-located MGS without starting the MGS, which is a complement
1194 to 'nosvc' mount option.
1197 Frequency : always, on big-endian systems
1199 Description: cleanup in ptlrpc code, related to PPC platform
1200 Details : store magic in native order avoid panic's in recovery on PPC
1201 node and forbid from this error in future. Also fix posibily
1202 of twice swab data. Fix get lov striping to userpace.
1205 Frequency : rarely, if replay get lost on server
1207 Description: server incorrectly drop resent replays lead to recovery failure.
1208 Details : do not drop replay according to msg flags, instead we check the
1209 per-export recovery request queue for duplication of transno.
1212 Frequency : after recovery
1214 Description: precreate to many object's after del orphan.
1215 Details : del orphan st in oscc last_id == next_id and this triger growing
1216 count of precreated objects. Set flag LOW to skip increase count
1217 of precreated objects.
1220 Frequency : rare, on clear nid stats
1222 Description: ASSERTION(client_stat->nid_exp_ref_count == 0)
1223 Details : when clean nid stats sometimes try destroy live entry,
1224 and this produce panic in free.
1227 Frequency : occasionally since 1.6.4
1229 Description: Stack overflow during MDS log replay
1230 Details : ease stack pressure by using a thread dealing llog_process.
1233 Frequency : very rare
1235 Description: MDT cannot be unmounted, reporting "Mount still busy"
1236 Details : Mountpoint references were being leaked during open reply
1237 reconstruction after an MDS restart. Drop mountpoint reference
1238 in reconstruct_open() and free dentry reference also.
1243 Description: wait until IO finished before start new when do lock cancel.
1244 Details : VM protocol want old IO finished before start new, in this case
1245 need wait until PG_writeback is cleared until check dirty flag
1246 and call writepages in lock cancel callback.
1251 Description: mds_mfd_close() ASSERTION(rc == 0)
1252 Details : In mds_mfd_close(), we need protect inode's writecount change
1253 within its orphan write semaphore to prevent possible races.
1256 Frequency : rare, on shutdown ost
1258 Description: don't hit live lock with umount ost.
1259 Details : shrink_dcache_parent can be in long loop with destroy dentries,
1260 use shrink_dcache_sb instead.
1263 Frequency : only when echo_client is used
1265 Description: don't panic with use echo_client
1266 Details : echo client pass NULL as client nid pointer and this produce NULL
1267 pointer dereference.
1270 Frequency : Always on 32-bit PowerPC systems
1272 Description: fix build on PPC32
1273 Details : compile code with -m64 flag produce wrong object file for PPC32.
1278 Description: MDS LBUG: ASSERTION(!IS_ERR(dchild))
1279 Details : In reconstruct_* functions, LASSERTs on both the data supplied
1280 by a client, and the data on disk are dangerous and incorrect.
1281 Change them with client eviction.
1283 Severity : enhancement
1285 Description: skiplist implementation simplification
1286 Details : skiplists are used to group compatible locks on granted list
1287 that was implemented as tracking first and last lock of each lock
1288 group the patch changes that to using doubly linked lists
1292 Description: delete compatibility for 32bit qdata
1293 Details : as planned, when lustre is beyond b1_8, lquota won't support 32bit
1294 qunit. That means servers of b1_4 and servers of b1_8 can't be
1295 used together if users want to use quota.
1298 Frequency : only with administrator action
1300 Description: mount failure if config log has invalid conf_param setting
1301 Details : If administrator specified an incorrect configuration parameter
1302 with "lctl conf_param" this would cause an error during future
1303 client mounts. Instead, ignore the bad configuration parameter.
1306 Frequency : blocks per group < blocksize*8 and uninit_groups is enabled
1308 Description: ldiskfs error: XXX blocks in bitmap, YYY in gd
1309 Details : If blocks per group is less than blocksize*8, set rest of the
1313 Frequency : Application do stride read on lustre
1315 Description: The read performance will drop a lot if the application does
1317 Details : Because the stride_start_offset are missing in stride read-ahead,
1318 it will cause clients read a lot of unused pages in read-ahead,
1319 then the read-performance drops.
1323 Description: more ldlm soft lockups
1324 Details : In ldlm_resource_add_lock(), call to ldlm_resource_dump()
1325 starve other threads from the resource lock for a long time in
1326 case of long waiting queue, so change the debug level from
1327 D_OTHER to the less frequently used D_INFO.
1329 Severity : enhancement
1331 Description: add -gid, -group, -uid, -user options to lfs find
1333 Severity : enhancement
1335 Description: ll_recover_lost_found_objs - recover objects in lost+found
1336 Details : OST corruption and subsequent e2fsck can leave objects in the
1337 lost+found directory. Using the "ll_recover_lost_found_objs"
1338 tool, these objects can be retrieved and data can be salvaged
1339 by using the object ID saved in the fid EA on each object.
1344 Description: this bug _only_ happens when inode quota limitation is very low
1345 (less than 12), so that inode quota unit is 1 at initialization.
1346 Details : if remaining quota equates 1, it is a sign to demonstate that quota
1347 is effective now. So least quota qunit should be 2.
1351 Description: Hung threads in invalidate_inode_pages2_range
1352 Details : The direct IO path doesn't call check_rpcs to submit a new RPC once
1353 one is completed. As a result, some RPCs are stuck in the queue
1358 Description: Procfs and llog threads access destoryed import sometimes.
1359 Details : Sync the import destoryed process with procfs and llog threads by
1360 the import refcount and semaphore.
1364 Description: mds fails to respond, threads stuck in ldlm_completion_ast
1365 Details : Sort source/child resource pair after updating child resource.
1370 Description: kernel BUG at ldiskfs2_ext_new_extent_cb
1371 Details : If insertion of an extent fails, then discard the inode
1372 preallocation and free data blocks else it can lead to duplicate
1377 Description: don't always update ctime in ext3_xattr_set_handle()
1378 Details : Current xattr code updates inode ctime in ext3_xattr_set_handle()
1379 In some cases the ctime should not be updated, for example for
1380 2.0->1.8 compatibility it is necessary to delete an xattr and it
1381 should not update the ctime.
1385 Description: add quota statistics
1386 Details : 1. sort out quota proc entries and proc code.
1387 2. add quota statistics
1392 Description: quotas are not honored with O_DIRECT
1393 Details : all writes with the flag O_DIRECT will use grants which leads to
1394 this problem. Now using OBD_BRW_SYNC to guard this.
1398 Bugzilla : 15713/16362
1399 Description: Assertion in iopen_connect_dentry in 1.6.3
1400 Details : looking up an inode via iopen with the wrong generation number can
1401 populate the dcache with a disconneced dentry while the inode
1402 number is in the process of being reallocated. This causes an
1403 assertion failure in iopen since the inode's dentry list contains
1404 both a connected and disconnected dentry.
1408 Description: assertion failure in ldlm_handle2lock()
1409 Details : fix a race between class_handle_unhash() and class_handle2object()
1410 introduced in lustre 1.6.5 by bug 13622.
1412 Severity : enhancement
1414 Description: superblock lock contention with many SMP cores on one client
1415 Details : several client filesystem locks were highly contended on SMP
1416 NUMA systems with 8 or more cores. Per-CPU datastructures
1417 and more efficient locking implemented to reduce contention.
1422 Description: Kernel BUG: sd_iostats_bump: unexpected disk index
1423 Details : remove the limit of 256 scsi disks in the sd_iostat patch
1428 Description: oops in sd_iostats_seq_show()
1429 Details : unloading/reloading the scsi low level driver triggers a kernel
1430 bug when trying to access the sd iostat file.
1435 Description: Kernel panics during QLogic driver reload
1436 Details : REQ_BLOCK_PC requests are not handled properly in the sd iostat
1437 patch, causing memory corruption.
1442 Description: journal_dev option does not work in b1_6
1443 Details : pass mount option during pre-mount.
1445 Severity : enhancement
1447 Description: Add a FIEMAP(FIle Extent MAP) ioctl for ldiskfs
1448 Details : FIEMAP ioctl will allow an application to efficiently fetch the
1449 extent information of a file. It can be used to map logical blocks
1450 in a file to physical blocks in the block device.
1453 Frequency : only with adaptive timeout enabled
1455 Description: DEBUG_REQ() bad paging request
1456 Details : ptlrpc_at_recv_early_reply() should not modify req->rq_repmsg
1457 because it can be accessed by reply_in_callback() without the
1461 Frequency : only on Cray X2
1463 Description: X2 build failures
1464 Details : fix build failures on Cray X2.
1468 Description: xid & resent requests
1469 Details : Initialize RPC XID from clock at startup (randomly if clock is
1474 Description: quota recovery deadlock during mds failover
1475 Details : This patch includes att18982, att18236, att18237 in bz14840.
1477 1. fix osts hang when mds does failover with quotaon
1478 2. prevent watchdog storm when osts threads wait for the
1483 Description: kernel panic on racer
1484 Details : Do not access dchild->d_inode when IS_ERR(dchild) is true.
1486 Severity : enhancement
1488 Description: Add lustre_start utility to start or stop multiple Lustre servers
1493 Description: Lustre GPF in {:ptlrpc:ptlrpc_server_free_request+373}
1494 Details : In case of memory pressure, list_del() can be called twice on
1495 req->rq_history_list, causing a kernel oops.
1499 Description: (ptllnd_peer.c:557:kptllnd_peer_check_sends()) ASSERTION(!in_interrupt()) failed
1500 Details : fix stack overflow in the distributed lock manager by defering
1501 export eviction after a failed ast to the elt thread instead of
1502 handling it in the dlm interpret routine.
1504 Severity : enhancement
1506 Description: More exported tunables for mballoc
1507 Details : Add support for tunable preallocation window and new tunables for
1508 large/small requests
1512 Description: Detect corruption of block bitmap and checking for preallocations
1513 Details : Checks validity of on-disk block bitmap. Also it does better
1514 checking of number of applied preallocations. When corruption is
1515 found, it turns filesystem readonly to prevent further corruptions.
1519 Frequency : only for big-endian servers
1520 Description: Check for big-endian system while mounting fs with extents feature
1521 Details : Mounting a filesystem with extents feature will fail on big-endian
1522 systems since ext3-based ldiskfs is not supported on big-endian
1523 systems. Can be over-ridden with "bigendian_extents" mount option.
1527 Description: Excessive recovery window
1528 Details : With AT enabled, the recovery window can be excessively long (6000+
1529 seconds). To address this problem, we no longer use
1530 OBD_RECOVERY_FACTOR when extending the recovery window (the connect
1531 timeout no longer depends on the service time, it is set to
1532 INITIAL_CONNECT_TIMEOUT now) and clients report the old service
1533 time via pb_service_time.
1536 Descriptoin: Don't sync journal after every i/o
1537 Details : Implement write RPC replay to allow server replies for write RPCs
1538 before data is on disk. However, this feature is disabled by
1539 default since some issues leading to data corruptions have been
1540 found during recovery (e.g. bug 19128). This feature can be enabled
1541 by running the following command on the OSSs:
1542 lctl set_param obdfilter.*.sync_journal=0
1546 Description: Watchdog triggered on MDS failover
1547 Details : enable OBD_CONNECT_MDT flag when connecting from the MDS so that
1548 the OSTs know that the MDS "UUID" can be reused for the same export
1549 from a different NID, so we do not need to wait for the export to
1554 Descriptoin: Lustre detected file system corruption with inode out of bounds
1555 Details : don't update i_size on MDS_CLOSE for directories. This causes
1556 directory corruptions on the MDT.
1558 -------------------------------------------------------------------------------
1560 2008-05-26 Sun Microsystems, Inc.
1562 * Support for kernels:
1563 2.6.5-7.311 (SLES 9),
1564 2.6.9-67.0.7.EL (RHEL 4),
1565 2.6.16.54-0.2.5 (SLES 10),
1566 2.6.18-53.1.14.el5 (RHEL 5),
1567 2.6.22.14 vanilla (kernel.org)
1568 * Client support for unpatched kernels:
1569 (see http://wiki.lustre.org/index.php?title=Patchless_Client)
1570 2.6.16 - 2.6.22 vanilla (kernel.org)
1571 * Due to problems with nested symlinks and FMODE_EXEC (bug 12652),
1572 we do not recommend using patchless RHEL4 clients with kernels
1573 prior to 2.6.9-55EL (RHEL4U5).
1574 * Recommended e2fsprogs version: 1.40.7-sun1
1575 * Note that reiserfs quotas are disabled on SLES 10 in this kernel.
1576 * RHEL 4 and RHEL 5/SLES 10 clients behaves differently on 'cd' to a
1577 removed cwd "./" (refer to Bugzilla 14399).
1578 * A new quota file format has been introduced in 1.6.5.
1579 The format conversion from prior releases is handled transparently,
1580 but releases older than 1.4.12/1.6.5 will not understand this new
1581 format. The automatic format conversion can be avoided by running
1582 the following command on the MDS before upgrading:
1583 'tunefs.lustre --param="mdt.quota_type=ug1" $MDTDEV'.
1584 For more information, please refer to bugzilla 13904.
1588 Description: quota performance fix
1589 Details : quota data is written in journalled mode instead of ordered to
1590 increase performance
1594 Description: lfs support for human-readable quota grace time strings
1595 Details : lfs setquota -t and lfs quota -t represent quota grace times
1596 in "XXwXXdXXhXXmXXs" format instead of large values in seconds
1599 Frequency : always with o2ib 1.3 and sles10
1601 Description: fix build with SLES10 and o2ib v3.
1602 Details : sles10 uses diffrent name for Module.symver file but configure
1603 assume this file has same name on RHEL/SLES/vanila kernels.
1606 Frequency : very rare, if additional xattrs are used on kernels >= 2.6.12
1608 Description: MDS may lose file striping (and hence file data) in some cases
1609 Details : If there are additional extended attributes stored on the MDS,
1610 in particular ACLs, SELinux, or user attributes (if user_xattr
1611 is specified for the client mount options) then there is a risk
1612 of attribute loss. Additionally, the Lustre file striping
1613 needs to be larger than default (e.g. striped over all OSTs),
1614 and an additional attribute must be stored initially in the
1615 inode and then increase in size enough to be moved to the
1616 external attribute block (e.g. ACL growing in size) for file
1619 Severity : enhancement
1621 Description: add message levels for liblustreapi
1626 Description: MDT cannot be unmounted, reporting "Mount still busy"
1627 Details : Mountpoint references were being leaked during open reply
1628 reconstruction after an MDS restart. Drop mountpoint reference
1629 in reconstruct_open() and free dentry reference also.
1634 Description: fix for occasional failure case of -ENOSPC in recovery-small tests
1635 Details : Move the 'good_osts' check before the 'total_bavail' check. This
1636 will result in an -EAGAIN and in the exit call path we call
1637 alloc_rr() which will with increasing aggressiveness attempt to
1638 aquire precreated objects on the minimum number of required OSCs.
1642 Description: Use old size assignment to avoid deadlock
1643 Details : This reverts the changes in bugs 2369 and bug 14138 that introduced
1644 the scheduling while holding a spinlock. We do not need locking
1645 for size in ll_update_inode() because size is only updated from
1646 the MDS for directories or files without objects, so there is no
1647 other place to do the update, and concurrent access to such inodes
1648 are protected by the inode lock.
1652 Description: Use __u64 instead of int for valid bits
1656 Description: resolve "_IOWR redefined" build error on SLES10
1660 Description: dump the memory debugging after all modules are unloaded to
1661 suppress false negative in conf_sanity test 39
1665 Description: the recovery timer never expires
1666 Details : for new client connect request, the recovery timer should not be
1667 reset, otherwise recovery timer will never expired, if the old
1668 client never come. Only old client connect and first connection
1669 req should trigger recovery timer reset.
1673 Description: the min numbers of lproc stats are wrong
1674 Details : adding a new constant LC_MIN_INIT and use it for initialization
1678 Frequency : always with interactive lfs
1680 Description: Reinitialize optind to 0 so that interactive lfs works in all cases
1683 Frequency : with multiple concurrent readdir processes in same directory
1684 Bugzilla : 15406, 15169, 15175
1685 Description: misc fixes for directory readahead.
1686 Details : prevent previous statahead async RPC callback from processing the
1687 current "statahead_info", race condition between async RPC callback
1688 add dentry into dentry hash table and "ls" thread revalidate such
1689 dentry, statahead his/miss control for hidden items, and so on.
1691 Severity : enhancement
1693 Description: build kernel-ib packages for OFED 1.3 in our release cycle
1697 Description: incore types cleaning in quota code (with respect to 64-bit limits)
1698 Details : several u32 variables declarations are replaced with u64 declarations
1703 Description: fix SLES kernel versioning
1704 Details : the kernel version for our SLES 10 kernel did not include a "-"
1705 before the "smp" at the end. while this was not a problem in
1706 general, it did mean that software trying to use the kernel
1707 version to try to detect a vendor specific kernel would fail.
1708 this was most evident by the OFED build scripts.
1713 Description: Don't update lov_desc members until making sure they are valid
1714 Details : When updating lov_desc members via proc fs, need fix their
1715 validities before doing the real update.
1718 Frequency : very rare
1720 Description: don't put request into delay list while invalidate in flight.
1721 Details : ptlrpc_delay_request sometimes put in delay list while invalidate
1722 import in flight. this produce timeout for invalidate and sometimes
1723 can cause stale data.
1725 Severity : enhancement
1727 Description: Update kernel to SLES9 2.6.5-7.311.
1729 Severity : enhancement
1731 Description: Update kernel to RHEL4 2.6.9-67.0.7.
1736 Frequency : on PPC only
1737 Description: not convert ost objects for directory because it's not exist.
1738 Details : ll_dir_getstripe assume dirrectory has ost objects but this wrong.
1740 Severity : enhancement
1742 Description: Fix warnings with compile liblustre at sles10/rhel5 which have
1743 __u64 as usingied long long type.
1746 Frequency : rare, on shutdown
1748 Description: race process ast vs remove callback
1749 Details : removing callback before disconnect import open race with
1750 processing callback.
1752 Severity : enhancement
1754 Description: Update kernel to SLES9 2.6.5-7.311.
1756 Severity : enhancement
1758 Description: Files open for execute are not marked busy on SLES10
1759 Details : Add FMODE_EXEC to SLES10 SP1 server kernel series.
1761 Severity : enhancement
1763 Description: Add server support for vanilla-2.6.22.14.
1766 Frequency : occasional
1768 Description: Avoid lov_create() getting stuck in obd_statfs_rqset()
1769 Details : If an OST is down the MDS will hang indefinitely in
1770 obd_statfs_rqset() waiting for the statfs data. While for
1771 MDS QOS usage of statfs, it should not stuck in waiting.
1773 Severity : enhancement
1775 Description: Disable adaptive timeouts by default
1778 Frequency : on network error
1780 Description: panic with double free request if network error
1781 Details : mdc_finish_enqueue is finish request if any network error occuring,
1782 but it's true only for synchronus enqueue, for async enqueue
1783 (via ptlrpcd) this incorrect and ptlrpcd want finish request
1787 Frequency : rare, on recovery
1789 Description: read procfs can produce deadlock in some situation
1790 Details : Holding lprocfs lock with send rpc can produce block for destroy
1791 obd objects and this also block reconnect with -EALREADY.
1792 This isn't fix all lprocfs bugs - but make it rare.
1794 Severity : enhancement
1796 Description: Update kernel to RHEL5 2.6.18-53.1.14.el5.
1799 Frequency : frequent on X2 node
1801 Description: mdc_set_open_replay_data LBUG
1802 Details : Set replay data for requests that are eligible for replay.
1807 Description: lustre_mgs: operation 101 on unconnected MGS
1808 Details : When MGC is disconnected from MGS long enough, MGS will evict the
1809 MGC, and late on MGC cannot successfully connect to MGS and a lot
1810 of the error messages complaining that MGS is not connected.
1813 Frequency : rare, depends on device drivers and load
1815 Description: MDS or OSS nodes crash due to stack overflow
1816 Details : Code changes in 1.6.4 increased the stack usage of some functions.
1817 In some cases, in conjunction with device drivers that use a lot
1818 of stack, the MDS (or possibly OSS) service threads could overflow
1819 the stack. One change which was identified to consume additional
1820 stack has been reworked to avoid the extra stack usage.
1822 Severity : enhancement
1824 Description: Update to RHEL5 latest kernel-2.6.18-53.1.13.el5.
1826 Severity : enhancement
1828 Description: Update to SLES10 SP1 latest kernel-2.6.16.54-0.2.5.
1830 Severity : enhancement
1832 Description: Update to RHEL5 latest kernel-2.6.18-53.1.6.el5.
1834 Serverity : enhancement
1836 Description: Update RHEL4 kernel to 2.6.9-67.0.4.
1839 Frequency : rare on shutdown OST
1841 Description: Don't allow skipping OSTs if index has been specified.
1842 Details : Don't allow skipping OSTs if index has been specified, make locking
1843 in internal create lots better.
1848 Description: ASSERTION(!PageDirty(page)) failed
1849 Details : Wrong check could lead to an assertion failure under specific
1855 Description: LBUG in ptlrpc_check_set() bad phase ebc0de00
1856 Details : access to bitfield in structure is always rounded to long
1857 and this produce problem with not atomic change any bit.
1862 Description: Lustre make rpms failed.
1863 Details : Remove ldiskfs spec file to avoids rpmbuild be confused when
1864 builds Lustre rpms from tarball.
1866 Severity : enhancement
1868 Description: Update to SLES9 SP4 kernel-2.6.5-7.308.
1871 Frequency : rare on shutdown OST
1873 Description: If llog cancel was not send before clean_exports phase, this can
1874 produce deadlock in llog code.
1875 Details : If llog thread has last reference to obd and call class_import_put
1876 this produce deadlock because llog_cleanup_commit_master wait when
1877 last llog_commit_thread exited, but this never success because was
1878 called from llog_commit_thread.
1881 Frequency : only if OST index is skipped
1883 Description: NULL lov_tgts causing MDS oops
1884 Details : more safe checks for NULL lov_tgts for avoid oops.
1886 Severity : enhancement
1888 Description: Update to RHEL4 latest kernel-2.6.9-67.0.1.EL.
1890 Severity : enhancement
1892 Description: Update to RHEL5 latest kernel-2.6.18-53.1.4.el5.
1897 Description: make mgs_setparam() handle fsname containing dash
1898 Details : fsname containing a dash does not work with lctl conf_param
1900 Severity : enhancement
1902 Description: Update to RHEL4 Update-6 kernel-2.6.9-67.EL.
1905 Frequency : rare, in recovery and (or) destroy lovobjid file.
1907 Description: rewrite lov objid code.
1908 Details : Cleanup for lov objid code, remove scability problems and wrong
1909 locking. Fix sending last_id into ost.
1911 Severity : enhancement
1913 Description: Update to SLES10 SP1 latest kernel-2.6.16.54-0.2.3.
1915 Severity : enhancement
1917 Description: Update to RHEL5 Update-1 kernel 2.6.18-53.el5.
1918 Details : Use d_move_locked instead of __d_move.
1921 Frequency : rare, at shutdown
1923 Description: access already free / zero obd_namespace.
1924 Details : if client_disconnect_export was called without force flag set,
1925 and exist connect request in flight, this can produce access to
1926 NULL pointer (or already free pointer) when connect_interpret
1927 store ocd flags in obd_namespace.
1930 Frequency : only at startup
1932 Description: not alloc memory with spinlock held.
1933 Details : allocation memory with GFP_KERNEL can produce sleep deadlock,
1934 if any spinlock held.
1939 Description: lfs find does not continue on file error
1940 Details : Continue other files processing when a file/dir is absent.
1944 Description: Inconsistent usage of lustre_pack_reply()
1945 Details : Standardize the usage of lustre_pack_reply() such that it
1946 always generate a CERROR on failure.
1949 Frequency : very rare
1951 Description: Fix replay if there is an un-replied request and open
1952 Details : In some cases, older replay request will revert the
1953 mcd->mcd_last_xid on MDS which is used to record the client's
1954 latest sent request.
1956 Severity : enhancement
1958 Description: Update to RHEL5 kernel 2.6.18-8.1.15.el5.
1960 Severity : enhancement
1962 Description: Update to SLES10 SP1 kernel 2.6.16.53-0.16
1964 Severity : enhancement
1966 Description: Update to SLES9 kernel-2.6.5-7.287.3.
1968 Severity : enhancement
1970 Description: Update to RHEL4 kernel-2.6.9-55.0.12.EL.
1972 Severity : enhancement
1974 Description: Build SLES10 patchless client fails
1975 Details : The configure was broken by run ./configure with
1976 --with-linux-obj=.... argument for patchless client. When the
1977 configure use --with-linux-obj, the LINUXINCLUDE= -Iinclude
1978 can't search header adequately. Use absolute path such as
1979 -I($LINUX)/include instead.
1981 Severity : enhancement
1983 Description: Lustre Page Accounting
1984 Details : New macros for page alloc and free which enable accounting
1985 of page allocation of Lustre. Use percpu counters to store memory
1986 and page statistics.
1989 Frequency : only if debugging is disabled
1991 Description: LASSERT_{REQ,REP}SWAB macros are buggy
1992 Details : If SWAB_PARANOIA is disabled, the LASSERT_REQSWAB and
1993 LASSERT_REPSWAB macros become no-ops, which is incorrect. Drop
1994 these macros and replace them with their definitions instead.
1999 Description: interrupt oig_wait produce painc on resend.
2000 Details : brw_redo_request can be used for resend requests from ptlrpcd and
2001 private set, and this produce situation when rq_ptlrpcd_data not
2002 copyed to new allocated request and triggered LBUG on assert
2003 req->rq_ptlrpcd_data != NULL. But this member used only for wakeup
2004 ptlrpcd set if request is changed and can be safety changed to use
2007 Severity : enhancement
2009 Description: organize the server-side client stats on per-nid basis
2010 Details : Change the structure of stats under obdfilter and mds to
2018 The "uuid"s file would list the uuids of _active_ exports.
2019 And the clear entry is to clear all stats and stale nids.
2024 Description: Processes looping in ll_readdir() on Lustre clients finally causing
2025 a full node pseudo-hang
2026 Details : Concurrent access to the same directory from multiple clients with
2027 intensive file creation/removal can cause a client node to spin in
2028 ll_readdir(). i_version must be increased every time the lock is
2029 cancelled to ensure a revalidate is done.
2034 Description: touch file failed when fs is not full
2035 Details : OST in recovery should not be discarded by MDS in alloc_qos(),
2036 otherwise we can get ENOSP while fs is not full.
2039 Frequency : only for Cray XT3
2040 Bugzilla : 12829/13455
2041 Description: Changing primary group doesn't change the group lustre assigns to
2043 Details : When CRAY_XT3 is defined, the fsgid supplied by the client is
2044 overridden with the primary group provided by the group upcall,
2045 whereas the supplied fsgid can be trusted if it is in the list of
2046 supplementary groups returned by the group upcall.
2048 Severity : enhancement
2050 Description: Root Squash Functionality
2051 Details : Implementation of NFS-like root squash capability. Specifically,
2052 don't allow someone with root access on a client node to be able
2053 to manipulate files owned by root on a server node.
2055 Severity : enhancement
2057 Description: Slow trucate/writes to huge files at high offsets.
2058 Details : Directly associate cached pages to lock that protect those pages,
2059 this allows us to quickly find what pages to write and remove
2060 once lock callback is received.
2065 Description: Too many locks accumulating on client during NFS usage
2066 Details : mds_open improperly used accmode to find out access mode to a
2067 file. Also mdc_intent_lock logic to find out if we already have
2068 lock similar to just received was flawed since introduction of
2069 skiplists - locks are now added to the front of the granted
2074 Description: Hit ASSERTION(obd->obd_stopping == 1) failed in some setup failed
2076 Details : In obd setup failure handler, obd_stopping will not necessarily to
2077 be 1, and obd_set_up should also be checked to make sure whether
2078 obd is completely setup.
2080 Severity : enhancement
2082 Description: Allow masking D_WARNING, D_ERROR messages from console
2083 Details : Console messages can now be disabled via lnet.printk.
2088 Description: User code with malformed file open parameter crashes client node
2089 Details : Before packing join_file req, all the related reference should be
2090 checked carefully in case some malformed flags cause fake join_file
2096 Description: shrink/enlarge qunit size when needed; fix the problem of coarse
2097 grain of quota doing harm to quota's accuracy
2098 Details : qunit size will be changed when quota limitation is too low/high;
2099 record the pending quota write in order to get more accureate
2100 quota; delete the patch for bug12588, which is unnecessary when
2101 this patch is landed. This bug also contains fixes for bug 14526,
2102 14299, 14601 and 13794.
2106 Description: LDLM_ENQUEUE races with LDLM_CP_CALLBACK
2107 Details : ldlm_completion_ast() assumes that a lock is granted when the req
2108 mode is equal to the granted mode. However, it should also check
2109 that LDLM_FL_CP_REQD is not set.
2113 Description: Heavy nfs access might result in deadlocks
2114 Details : After ELC code landed, it is now improper to enqueue any mds
2115 locks under och_sem, because enqueue might want to decide to
2116 cancel open locks for same inode we are holding och_sem for.
2120 Description: 35% write performance drop with ldiskfs2 when quotas are on
2121 Details : Enable ext3 journalled quota by default to improve performance
2122 when quotas are turned on.
2126 Description: Client eviction while running blogbench
2127 Details : A lot of unlink operations with concurrent I/O can lead to a
2128 deadlock causing evictions. To address the problem, the number of
2129 oustanding OST_DESTROY requests is now throttled to
2130 max_rpcs_in_flight per OSC and LDLM_FL_DISCARD_DATA blocking
2131 callbacks are processed in priority.
2134 Frequency : RHEL4 only
2136 Description: mkfs is very slow on IA64/RHEL4
2137 Details : A performance regression has been discovered in the MPT Fusion
2138 driver between versions 3.02.73rh and 3.02.99.00rh. As a
2139 consequence, we have downgraded the MPT Fusion driver in the RHEL4
2140 kernel from 3.02.99.00 to 3.02.73 until this problem is fixed.
2143 Frequency : PPC/PPC64 only
2145 Description: conflicts between asm-ppc64/types.h and lustre_types.h
2146 Details : fix duplicated definitions between asm-ppc64/types.h and
2147 lustre_types.h on PPC.
2150 Frequency : PPC/PPC64 only
2152 Description: asm-ppc/segment.h does not exist
2153 Details : fix compile issue on PPC.
2157 Description: data checksumming impacts single node performance
2158 Details : add support for several checksum algorithms. Currently, CRC32 and
2159 Adler-32 are supported. The checksum type can be changed on the fly
2160 through /proc/fs/lustre/osc/*/checksum_type.
2164 Description: use adler32 for page checksums
2165 Details : when available, use the Adler-32 algorithm instead of CRC32 for
2170 Description: better handle error messages in extents code
2172 Severity : enhancement
2174 Description: SNMP support enhancement
2175 Details : Adding total number of sampled request for an MDS node in snmp
2178 Severity : enhancement
2180 Description: Optimize ldlm waiting list processing for PR extent locks
2181 Details : When processing waiting list for read extent lock and meeting read
2182 lock that is same or wider to it that is not contended, skip
2183 processing rest of the list and immediatelly return current
2184 status of conflictness, since we are guaranteed there are no
2185 conflicting locks in the rest of the list.
2189 Description: Time out and refuse to reconnect
2190 Details : When the failover node is the primary node, it is possible
2191 to have two identical connections in imp_conn_list. We must
2192 compare not conn's pointers but NIDs, otherwise we
2193 can defeat connection throttling.
2197 Description: Client not clear own cache if answer to reconnect is lost.
2198 Details : Client gets evicted from server. Now client also thinks it is
2199 disconnected (or gets ENOTCONN on its operation) and decides to
2200 reconnect. Server receives reconnect message, but cannot find
2201 export. New export is created that is fully valid (new cookie!),
2202 but reply is lost and not reported to client. Client reconnects
2203 again and gets back a just-created connection, but it is not new
2204 so client thinks it was not evicted and does not do recovery.
2208 Description: Detect stride IO mode in read-ahead
2209 Details : When a client does stride read, read-ahead should detect that and
2210 read-ahead pages according to the detected stride pattern.
2214 Description: build for x2 fails
2215 Details : fix compile issue on Cray systems.
2217 Severity : enhancement
2219 Description: implement readv/writev APIs(aio_read/aio_writes in newer kernels)
2220 Details : This greatly improves speed of NFS writes on 2.6 kernels.
2223 Frequency : only on PPC/SLES10
2225 Description: "BITS_PER_LONG is not 32 or 64" in linux/idr.h
2226 Details : On SLES10/PPC, fs.h includes idr.h which requires BITS_PER_LONG to
2227 be defined. Add a hack in mkfs_lustre.c to work around this compile
2232 Description: LASSERT on MDS when client holding flock lock dies
2233 Details : ldlm pool logic depends on number of granted locks equal to
2234 number of released locks which is not true for flock locks, so
2235 just exclude such locks from consideration.
2239 Description: MDS deadlock with many ll_sync_lov threads and I/O stalled
2240 Details : Use fsfilt_sync() for both the whole filesystem sync and
2241 individual file sync to eliminate dangerous inode locking
2242 with I_LOCK that can lead to a deadlock.
2246 Description: Update an obsolete wirecheck.c generator
2247 Details : Update wirecheck.c/wirehdr.c and regenerate wiretest.c
2251 Description: Client can panic on open sometimes
2252 Details : It is possible that we try to free already freed request in
2253 ll_file_open in some error cases when we send request from
2258 Description: performance in 1.6.3
2259 Details : Force q->max_phys_segments to MAX_PHYS_SEGMENTS on SLES10 to be
2260 sure that 1MB requests are not fragmented by the block layer.
2264 Description: LDLM soft lockups - improvement
2265 Details : It is be possible to send the lock handle along with each read
2266 or write request because the client is already doing a lock match
2267 itself so there isn't any reason the OST should have to re-do that
2273 Description: lfs quota fails with deactivated OSTS
2274 Details : With this patch, three improvements are included:
2275 1. detete the softlimit in mds and osts when use "lfs quota".
2276 2. display the inaccurate data in the output of "lfs quota".
2277 3. try to get quota info when "lfs quota" is executed.
2282 Description: Extent locks not granted with no conflicts sometimes.
2283 Details : When race occurs in glimpse handler and nothing is returned,
2284 we do not reprocess the queue after lock cancel, and that leads
2285 to a stall until next activity on a resource
2288 Frequency : failover with quotaon
2290 Description: during mds failovers with quota on, OSTs got into deadlock state
2291 and causing dumpstack.
2292 Details : for every quota slave, at any time, there is only one quota req
2293 is sent to quota master for every uid/gid. Before that quota req
2294 returns, all the thread relative to the same uid/gid will wait.
2295 So if the quota req is lost because mds failovers or any other
2296 reasons, this bug will be hit. Now, dqacq_interpret() will handle
2297 quota reqs who time out.
2299 Severity : enhancement
2302 Description: when quota slave checks if quota is enough, there is an unnecessary
2304 Details : place this wait on necessary place instead of always waiting.
2306 --------------------------------------------------------------------------------
2308 2007-12-07 Cluster File Systems, Inc. <info@clusterfs.com>
2310 * Support for kernels:
2311 2.6.5-7.286 (SLES 9),
2312 2.6.9-55.0.9.EL (RHEL 4),
2313 2.6.16.53-0.8 (SLES 10),
2314 2.6.18-8.1.14.el5 (RHEL 5),
2315 2.6.18.8 vanilla (kernel.org)
2316 * Client support for unpatched kernels:
2317 (see http://wiki.lustre.org/index.php?title=Patchless_Client)
2318 2.6.16 - 2.6.22 vanilla (kernel.org)
2319 * Due to recently discovered recovery problems, we do not recommend
2320 using patchless RHEL 4 clients with this or any earlier release.
2321 * Recommended e2fsprogs version: 1.40.2-cfs1
2322 * Note that reiserfs quotas are disabled on SLES 10 in this kernel.
2325 Frequency : occasional
2327 Description: MDS hang or stay in waiting lock
2328 Details : If client receive lock with CBPENDING flag ldlm need send lock
2329 cancel as separate rpc, to avoid situation when cancel request
2330 can't processed due all i/o threads stay in wait lock.
2333 Frequency : occasional
2335 Description: Do not fail import if osc_interpret_create gets -EAGAIN
2336 Details : If osc_interpret_create got -EAGAIN it immediately exits and
2337 wakeup oscc_waitq. After wakeup oscc_wait_for_objects call
2338 oscc_has_objects and see OSC has no objests and call
2339 oscc_internal_create to resend create request.
2341 Severity : enhancement
2343 Description: Update kernel patches for SLES10 2.6.16.53-0.8.
2344 Details : Update which_patch & target file for SLES10 latest kernel.
2346 Severity : enhancement
2348 Description: add --type and --size parameters to lfs find
2349 Details : Enhance lfs find by adding filetype and filesize parameters. Also
2350 multiple OBDs can now be specified for the --obd option.
2352 Severity : enhancement
2354 Description: eliminate client locks in face of contention
2355 Details : file contention detection and lockless i/o implementation
2356 for contended files.
2358 Severity : enhancement
2360 Description: Remove client patches from SLES 10 kernel.
2361 Details : This causes SLES 10 clients to behave as patchless clients
2362 even on a Lustre-patched (server) kernel.
2364 Severity : enhancement
2366 Description: use i_size_read and i_size_write in 2.6 port
2367 Details : replace inode->i_size access with i_size_read/write()
2370 Frequency : when removing large files
2372 Description: scheduling issue during removal of large Lustre files
2373 Details : Don't take the BKL in fsfilt_ext3_setattr() for 2.6 kernels.
2374 It causes scheduling issues when removing large files (17TB in the
2380 Description: 1.4.11 Can't handle directories with stripe set and extended ACLs
2381 Details : Impossible (EPROTO is returned) to access a directory that has a
2382 non-default striping and ACLs.
2385 Frequency : only on ppc
2387 Description: /proc/fs/lustre/devices broken on ppc
2388 Details : The patch as applied to 1.6.2 doesn't look correct for all arches.
2389 We should make sure the type of 'index' is loff_t and then cast
2390 explicitly as needed below. Do not assign an explicitly cast
2394 Frequency : only for rhel5
2396 Description: Kernel patches update for RHEL5 2.6.18-8.1.10.el5.
2397 Details : Modify the target file & which_kernel.
2400 Frequency : if the uninit_groups feature is enabled on ldiskfs
2402 Description: e2fsck reports "invalid unused inodes count"
2403 Details : If a new ldiskfs filesystem is created with the "uninit_groups"
2404 feature and only a single inode is created in a group then the
2405 "bg_unused_inodes" count is incorrectly updated. Creating a
2406 second inode in that group would update it correctly.
2411 Description: buffer overruns could theoretically occur
2412 Details : llapi_semantic_traverse() modifies the "path" argument by
2413 appending values to the end of the origin string, and a buffer
2414 overrun may occur. Adding buffer overrun check in liblustreapi.
2416 Severity : enhancement
2418 Description: Add jbd statistics patch for RHEL5 and 2.6.18-vanilla.
2421 Frequency : only if filesystem is inconsistent
2423 Description: handle "serious error: objid * already exists" more gracefully
2424 Details : If LAST_ID value on disk is smaller than the objects existing in
2425 the O/0/d* directories, it indicates disk corruption and causes an
2426 LBUG(). If the object is 0-length, then we should use the existing
2427 object. This will help to avoid a full fsck in most cases.
2429 Severity : enhancement
2431 Description: Kernel patches update for RHEL4 2.6.9-55.0.6.
2432 Details : Modify vm-tunables-rhel4.patch.
2434 Severity : enhancement
2436 Description: Kernel config for 2.6.18-vanilla.
2437 Details : Modify targets/2.6-vanilla.target.in.
2438 Add config file kernel-2.6.18-2.6-vanilla-i686.config.
2439 Add config file kernel-2.6.18-2.6-vanilla-i686-smp.config.
2440 Add config file kernel-2.6.18-2.6-vanilla-x86_64.config.
2441 Add config file kernel-2.6.18-2.6-vanilla-x86_64-smp.config.
2444 Frequency : occasional
2446 Description: improve handling recoverable errors
2447 Details : If request processed with error which can be recoverable on server
2448 request should be resend, otherwise page released from cache and
2454 Description: Kernel patches update for RHEL5 2.6.18-8.1.14.el5.
2455 Details : Modify target file & which_patch.
2456 A flaw was found in the IA32 system call emulation provided
2457 on AMD64 and Intel 64 platforms. An improperly validated 64-bit
2458 value could be stored in the %RAX register, which could trigger an
2459 out-of-bounds system call table access. An untrusted local user
2460 could exploit this flaw to run code in the kernel
2461 (ie a root privilege escalation). (CVE-2007-4573).
2465 Description: change order of libsysio includes
2466 Details : '#include sysio.h' should always come before '#include xtio.h'
2468 Severity : enhancement
2470 Description: adapt the lustre_config script to support the upgrade case
2471 Details : Add "-u" option for lustre_config script to support upgrading 1.4
2472 server targets to 1.6 in parallel.
2477 Description: To avoid grant space > avaible space when the disk is almost
2478 full. Without this patch you might see the error "grant XXXX >
2479 available" or some LBUG about grant, when the disk is almost
2481 Details : In filter_check_grant, for non_grant cache write, we should
2482 check the left space by if (*left > ungranted + bytes), instead
2483 of (*left > ungranted), because only we are sure the left space
2484 is enough for another "bytes", then the ungrant space should be
2485 increase. In client, we should update cl_avail_grant only there is
2486 OBD_MD_FLGRANT in the reply.
2491 Description: Update RHEL 4 kernel to fix local root privilege escalation.
2492 Details : Update to the latest RHEL 4 kernel to fix the vulnerability
2493 described in CVE-2007-4573. This problem could allow untrusted
2494 local users to gain root access.
2497 Frequency : when using O_DIRECT and quotas
2499 Description: Incorrect file ownership on O_DIRECT output files
2500 Details : block usage reported by 'lfs quota' does not take into account
2501 files that have been written with O_DIRECT.
2505 Description: (rw.c:1323:ll_read_ahead_pages()) ASSERTION(page_idx > ria->ria_stoff) failed
2506 Details : Once the unmatched stride IO mode is detected, shrink the stride-ahead
2507 window to 0. If it does hit cache miss, and read-pattern is still
2508 stride-io mode, does not reset the stride window, but also does not
2509 increase the stride window length in this case.
2511 --------------------------------------------------------------------------------
2513 2007-09-27 Cluster File Systems, Inc. <info@clusterfs.com>
2515 * Support for kernels:
2516 2.6.5-7.286 (SLES 9),
2517 2.6.9-55.0.2.EL (RHEL 4),
2518 2.6.16.46-0.14 (SLES 10),
2519 2.6.18-8.1.8.el5 (RHEL 5),
2520 2.6.18.8 vanilla (kernel.org)
2521 * Client support for unpatched kernels:
2522 (see http://wiki.lustre.org/index.php?title=Patchless_Client)
2523 2.6.16 - 2.6.21 vanilla (kernel.org)
2524 * Due to recently discovered recovery problems, we do not recommend
2525 using patchless RHEL 4 clients with this or any earlier release.
2526 * Recommended e2fsprogs version: 1.40.2-cfs1
2527 * Note that reiserfs quotas are disabled on SLES 10 in this kernel.
2531 Description: Fix errors in lfs documentation
2532 Details : Fixes man pages
2534 Severity : enhancement
2536 Description: Adaptive timeouts
2537 Details : RPC timeouts adapt to changing server load and network
2538 conditions to reduce resend attempts and improve recovery time.
2540 Severity : enhancement
2542 Description: llapi_file_create() does not allow some changes
2543 Details : add llapi_file_open() that allows specifying the file creation
2544 mode and open flags, and also returns an open file handle.
2546 Severity : enhancement
2548 Description: df doesn't work properly if diskfs blocksize != 4K
2549 Details : Choose biggest blocksize of OST's as the LOV's blocksize.
2552 Frequency : always on directories with default striping set
2554 Description: lfs find on -1 stripe looping in lsm_lmm_verify_common()
2555 Details : Avoid lov_verify_lmm_common() on directory with -1 stripe count.
2557 Severity : enhancement
2559 Description: merge and cleanup kernel patches.
2560 Details : Remove mnt_lustre_list in vfs_intent-2.6-rhel4.patch.
2562 Severity : enhancement
2564 Description: RedHat Update kernel for RHEL5
2565 Details : Kernel config file for RHEL5.
2567 Severity : enhancement
2569 Description: OSS needs mutliple precreate threads
2570 Details : Add ability to start more than one create thread per OSS.
2573 Frequency : only with quota on the root user
2575 Description: mds_obd_create error creating tmp object
2576 Details : When the user sets quota on root, llog will be affected and can't
2577 create files and write files.
2580 Frequency : Always on ia64 patchless client, and possibly others.
2582 Description: Add EXPORT_SYMBOL check for node_to_cpumask symbol.
2583 Details : This allows the patchless client to be loaded on architectures
2584 without this export.
2589 Description: disorder of journal start and llog_add cause deadlock.
2590 Details : in llog_origin_connect, journal start should happen before
2591 llog_add keep the same order as other functions to avoid
2594 Severity : enhancement
2596 Description: RedHat Update kernel for RHEL5
2597 Details : Modify the kernel config file more closer RHEL5.
2599 Severity : enhancement
2601 Description: Build failure against Centos5 (RHEL5)
2602 Details : Define PAGE_SIZE when it isn't present.
2605 Frequency : occasionally when using NFS
2607 Description: "ll_intent_file_open()) lock enqueue: err: -13" with nfs
2608 Details : with NFS, the anon dentry's parent was set to itself in
2609 d_alloc_anon(), so in MDS, we use rec->ur_fid1 to find the
2610 corresponding dentry other than use rec->ur_name.
2612 Severity : enhancement
2614 Description: enable data checksumming by default
2615 Details : enable checksum by default, allow --disable-checksum
2616 configure option and "-o nochecksum" mount option. Checksums
2617 can also be disabled at runtime via $LPROC/osc/*/checksum_pages.
2620 Frequency : Occasionally with failover
2622 Description: Client eviction due to failover config
2623 Details : after a connection loss, the lustre client should attempt to
2624 reconnect to the last active server first before trying the
2625 other potential connections.
2627 Severity : enhancement
2629 Description: client-side metadata stat-ahead during readdir(directory readahead)
2630 Details : perform client-side metadata stat-ahead when the client detects
2631 readdir and sequential stat of dir entries therein
2636 Description: when mds and osts use different quota unit(32bit and 64bit),
2637 quota will be released repeatly.
2638 Details : void sending multiple quota reqs to mds, which will keep the
2639 status between the reqs.
2642 Frequency : only with liblustre clients on XT3
2644 Description: evictions taking too long
2645 Details : allow llrd to evict clients directly on OSTs
2649 Description: osts not allocated evenly to files
2650 Details : change the condition to increase offset_idx
2652 Severity : enhancement
2654 Description: self-adjustable client's lru lists
2655 Details : use adaptive algorithm for managing client cached locks lru
2656 lists according to current server load, other client's work
2657 pattern, memory activities, etc. Both, server and client
2658 side namespaces provide number of proc tunables for controlling
2663 Description: rewrite ext2-derived code in llite/dir.c and obdclass/uuid.c
2664 Details : rewrite inherited code (uuid parsing code from ext2 utils and
2665 readdir code from ext3) from scratch preserving functionality.
2669 Description: Only those disconnect error should be returned by rq_status.
2670 Details : In open/enqueue processs, Some errors, which will cause client
2671 disconnected, should be returned by rq_status, while other
2672 errors should still be returned by intent, then mdc or llite will
2675 Severity : enhancement
2677 Description: Tune the kernel for good SCSI performance.
2678 Details : Set the value of /sys/block/{dev}/queue/max_sectors_kb
2679 to the value of /sys/block/{dev}/queue/max_hw_sectors_kb
2683 Frequency : Always for filesystems larger than 2TB on 32-bit systems.
2684 Bugzilla : 13547, 13627
2685 Description: Data corruption for OSTs that are formatted larger than 2TB
2687 Details : When generating the bio request for lustre file writes the
2688 sector number would overflow a temporary variable before being
2689 used for the IO. The data reads correctly from Lustre (which
2690 will overflow in a similar manner) but other file data or
2691 filesystem metadata may be corrupted in some cases.
2695 Description: "lfs find -obd UUID" prints directories
2696 Details : "lfs find -obd UUID" will return all directory names instead
2697 of just file names. It is incorrect because the directories
2698 do not reside on the OSTs.
2701 Bugzilla : 13596, 13614
2702 Description: MDS hang after unclean shutdown of lots of clients
2703 Details : Never resend AST requests.
2707 Frequency : Always, for kernels after 2.6.16
2708 Description: Fix warning idr_remove called for id=.. which is not allocated.
2709 Details : Last kernels save old s_dev before kill super and not allow
2710 to restore from callback - restore it before call kill_anon_super.
2714 Description: Fix error on 'ls .' at the top of the Lustre mount.
2715 Details : Don't revalidate dentry if it is a root dentry.
2717 --------------------------------------------------------------------------------
2719 2007-08-27 Cluster File Systems, Inc. <info@clusterfs.com>
2721 * Support for kernels:
2722 2.6.5-7.286 (SLES 9),
2723 2.6.9-55.0.2.EL (RHEL 4),
2724 2.6.16.46-0.14 (SLES 10),
2725 2.6.18-8.1.8.el5 (RHEL 5),
2726 2.6.18.8 vanilla (kernel.org)
2727 * Client support for unpatched kernels:
2728 (see http://wiki.lustre.org/index.php?title=Patchless_Client)
2729 2.6.16 - 2.6.22 vanilla (kernel.org)
2730 * Due to recently discovered recovery problems, we do not recommend
2731 using patchless RHEL 4 clients with this or any earlier release.
2732 * Recommended e2fsprogs version: 1.39.cfs8
2733 * Note that reiserfs quotas are disabled on SLES 10 in this kernel.
2738 Description: block reactivating mgc import until all deactivates complete
2739 Details : Fix race when failing back MDT/MGS to itself (testing)
2741 Severity : enhancement
2743 Description: lfs setstripe enhancement
2744 Details : Make lfs setstripe understand 'k', 'm' and 'g' for stripe size.
2747 Frequency : mds/oss recovery
2749 Description: llog ctxt is refrenced after it has been freed.
2750 Details : llog ctxt refcount was added to avoide the race between ctxt free
2751 and llog recovery process. Each llog user must hold ctxt refcount
2752 before it access the llog. And the llog ctxt can only be freed
2753 when its refcount is zero.
2755 Severity : enhancement
2757 Description: randomly memory allocation failure util
2758 Details : Make lustre randomly failed allocating memory for testing purpose.
2761 Frequency : only for SLES10
2763 Description: Update kernel patch for SLES10 SP1
2764 Details : Add patch blkdev_tunables-2.6-sles10.patch to 2.6-sles10.series.
2766 Severity : enhancement
2768 Description: omit set fsid for export NFS
2769 Details : fix set/restore device id for avoid EMFILE error and mark lustre fs
2770 as FS_REQUIRES_DEV for avoid problems with generate fsid.
2773 Frequency : after network failures
2775 Description: Add sync option to mount_lustre.c
2776 Details : Client loses data written to lustre after a network interruption.
2778 Severity : enhancement
2780 Description: Error message improvement.
2781 Details : Merging of two LCONSOLE_ERROR_MSG into one.
2783 Severity : enhancement
2785 Description: don't use GFP_* in generic Lustre code.
2786 Details : Use cfs_alloc_* functions and CFS_* flags for code portability.
2788 Severity : enhancement
2790 Description: obdclass is limited by single OBD_ALLOC(idarray)
2791 Details : replace OBD_ALLOC/OBD_FREE with OBD_VMALLOC/OBD_VFREE
2793 Severity : enhancement
2795 Description: updated patchess for new RHEL4 kernel
2796 Details : Fixed ext3-unlink-race.patch per Kalpak's comment.
2798 Severity : enhancement
2800 Description: warnings with build patchless client with vanila 2.6.19 and up
2801 Details : change old ctl_table style and replace ctl_table/ctl_table_header
2802 with cfs_sysctl_table_t/cfs_sysctl_table_header_t
2804 Severity : enhancement
2806 Description: O_DIRECT bypasses client statistics.
2807 Details : When running with O_DIRECT I/O, neither the client rpc_stats nor
2808 read_ahead_stats were updated. Copied stats section from
2809 osc_send_oap_rpc() into async_internal().
2812 Frequency : only for Cray XT3
2814 Description: peer credits not enough on many OST per OSS systems.
2815 Details : Use new lnet way to add credits as we need those for pings and ASTs
2818 Frequency : only with liblustre
2820 Description: Liblustre is not releasing flock locks on file close.
2821 Details : Release flock locks on file close.
2824 Frequency : only for RHEL4
2826 Description: Update kernel patches for kernel-2.6.9-55.0.2.EL
2827 Details : Remove inode-nr_unused-2.6.9-rhel4.patch from 2.6-rhel4.series
2828 Update target file and kernel config.
2832 Description: lustre support for RHEL5
2833 Details : Add support for RHEL5.
2838 Description: ASSERTION(export != NULL) failed in target_handle_connect
2839 Details : Assetion hit is result of rare race between disconnect and connet
2840 to same nid. target_handle_connect found old connect cockie and
2841 tried to reconnect, but can't find export for this cockie.
2846 Description: umount blocks forever on error
2847 Details : In result of wrong using obd_no_recov and obd_force flags client
2848 can hand if cancel or some other requests is lost.
2851 Frequency : Only for SLES
2853 Description: sanity_quota fail test_1
2854 Details : There are multiple occurences of $TSTUSR in SLES's /etc/group
2855 file, which makes TSTID[2] inunique.
2857 Severity : enhancement
2859 Description: Kernel patches for SLES9 2.6.5-7.286 kernel
2860 Details : Update target/ChangeLog/which_patch .
2862 Severity : enhancement
2864 Description: jbd statistics
2865 Details : Port older jbd statistics patch for sles10
2870 Description: lvbo_init failed for resource with missing objects.
2871 Details : Fix returning error if we do stat for file with missing/corrupted
2872 objects and i_size set to all sum of size of all avaible objects.
2873 if we to truncate/write to missing object - it is recreated.
2878 Description: Oops in read and write path when failing to allocate lock.
2879 Details : Check if lock allocation failed and return error back.
2882 Frequency : When flocks are used.
2884 Description: assertion failure in ldlm_cli_enquque_fini for non NULL lock.
2885 Details : Flock locks might destroy just granted lock if it could be merged
2886 with another existing flock, this is done in completion handler,
2887 so teach ldlm_cli_enquque_fini that this is a valid case for
2893 Description: reply_lock_interpret crash due to race with it and lock cancel.
2894 Details : Do not replay locks that are being cancelled. Do not reference
2895 locks by their address during replay, just by their handle.
2898 Frequency : only with deactivated OSTs
2900 Description: lstripe command fails for valid OST index
2901 Details : The stripe offset is compared to 'lov->desc.ld_tgt_count'
2902 instead of lov->desc.ld_active_tgt_count.
2904 Severity : enhancement
2906 Description: Build failure against Centos5 (RHEL5)
2907 Details : Use getpagesize() instead of PAGE_SIZE.
2909 --------------------------------------------------------------------------------
2911 2007-07-30 Cluster File Systems, Inc. <info@clusterfs.com>
2913 * Support for kernels:
2914 2.6.5-7.283 (SLES 9),
2915 2.6.9-55.EL (RHEL 4),
2916 2.6.16.46-0.14 (SLES 10),
2917 2.6.18.8 vanilla (kernel.org)
2918 * Client support for unpatched kernels:
2919 (see http://wiki.lustre.org/index.php?title=Patchless_Client)
2920 2.6.16 - 2.6.22 vanilla (kernel.org)
2921 * Due to recently discovered recovery problems, we do not recommend
2922 using patchless RHEL 4 clients with this or any earlier release.
2923 * Recommended e2fsprogs version: 1.39.cfs8
2924 * Note that reiserfs quotas are disabled on SLES 10 in this kernel.
2925 * Starting with this release, the ldiskfs backing filesystem required
2926 by Lustre is now in its own package, lustre-ldiskfs. This package
2927 should be installed. It is versioned separately from Lustre and
2928 may be released separately in future.
2933 Description: server LBUG when shutting down
2934 Details : Block umount forever until the mount refcount is zero rather
2935 than giving up after an arbitrary timeout.
2937 Severity : enhancement
2939 Description: add optional extra BUILD_VERSION info
2940 Details : add a new environment variable (namely LUSTRE_VERS) which allows
2941 to override the lustre version.
2944 Frequency : 2.6.18 servers only
2946 Description: ll_kern_mount() doesn't release the module reference
2947 Details : The ldiskfs module reference count never drops down to 0
2948 because ll_kern_mount() doesn't release the module reference.
2953 Description: server LBUG when using old ost_num_threads parameter
2954 Details : Accept the old ost_num_threads parameter but warn that it
2955 is deprecated, and fix an off-by-one error that caused an LBUG.
2960 Description: Transient SCSI error results in persistent IO issue
2961 Details : iobuf->dr_error is not reinitialized to 0 between two
2965 Frequency : sometimes when underlying device returns I/O errors
2967 Description: OSTs not going read-only during write failures
2968 Details : OSTs are not remounted read-only when the journal commit threads
2969 get I/O errors because fsfilt_ext3 calls journal_start/stop()
2970 instead of the ext3 wrappers.
2973 Frequency : only with 10000 clients or more
2975 Description: poor connect scaling with increasing client count
2976 Details : Don't run filter_grant_sanity_check for more than 100 exports
2977 to improve scaling for large numbers of clients.
2980 Frequency : SLES10 only
2982 Description: sanity-quota.sh quotacheck failed: rc = -22
2983 Details : Quotas cannot be enabled on SLES10.
2986 Frequency : liblustre clients only
2988 Description: getdirentries does not give error when run on compute nodes
2989 Details : getdirentries does not fail when the size specified as an argument
2990 is too small to contain at least one entry
2992 Severity : enhancement
2994 Description: Add LNET router traceability for debug purposes
2995 Details : If a checksum failure occurs with a router as part of the
2996 IO path, the NID of the last router that forwarded the bulk data
2997 is printed so it can be identified.
3002 Description: OST "spontaneously" evicts client; client has imp_pingable == 0
3003 Details : Due to a race condition, liblustre clients were occasionally
3004 evicted incorrectly.
3006 Severity : enhancement
3008 Description: lfs setstripe use optional parameters instead of postional
3011 Severity : enhancement
3013 Description: Nanosecond timestamp support for ldiskfs
3014 Details : The on-disk ldiskfs filesystem has added support for nanosecond
3015 resolution timestamps. There is not yet support for this at
3016 the Lustre filesystem level.
3019 Frequency : during server recovery
3021 Description: MDS failing to send precreate requests due to OSCC_FLAG_RECOVERING
3022 Details : request with rq_no_resend flag not awake l_wait_event if they get a
3026 Frequency : nfs export on patchless client
3028 Description: connectathon hang when test nfs export over patchless client
3029 Details : Disconnected dentry cannot be found with lookup, so we do not need
3030 to unhash it or make it invalid
3033 Description: fix llapi_lov_get_uuids() to allow many OSTs to be returned
3034 Details: : Change llapi_lov_get_uuids() to read the UUIDs from /proc instead
3035 of using an ioctl. This allows lfsck for > 160 OSTs to succeed.
3040 Description: open req refcounting wrong on reconnect
3041 Details : If reconnect happened between getting open reply from server and
3042 call to mdc_set_replay_data in ll_file_open, we will schedule
3043 replay for unreferenced request that we are about to free.
3044 Subsequent close will crash in variety of ways.
3045 Check that request is still eligible for replay in
3046 mdc_set_replay_data().
3051 Description: disable writes to filesystem when reading health_check file
3052 Details : the default for reading the health_check proc file has changed
3053 to NOT do a journal transaction and write to disk, because this
3054 can cause reads of the /proc file to hang and block HA state
3055 checking on a healthy but otherwise heavily loaded system. It
3056 is possible to return to the previous behaviour during configure
3057 with --enable-health-write.
3059 Severity : enhancement
3061 Description: 64-bit inode version
3062 Details: : Add a on-disk 64-bit inode version for ext3 to track changes made
3063 to the inode. This will be required for version-based recovery.
3068 Description: MDS fails to start if a duplicate client export is detected
3069 Details : in some rare cases it was possible for a client to connect to
3070 an MDS multiple times. Upon recovery the MDS would detect this
3071 and fail during startup. Handle this more gracefully.
3073 Severity : enhancement
3075 Description: Add -o localflock option to simulate old noflock behaviour.
3076 Details : This will achieve local-only flock/fcntl locks coherentness.
3081 Description: log_commit_thread vs filter_destroy race leads to crash
3082 Details : Take import reference before releasing llog record semaphore
3087 Description: Wrong request locking in request set processing
3088 Details : ptlrpc_check_set wrongly uses req->rq_lock for proctect add to
3089 imp_delayed_list, in this place should be used imp_lock.
3092 Frequency : when reconnecting
3094 Description: Grant leak when OSC reconnect to OST
3095 Details : When osc reconnect ost, OST(filter) should check whether it
3096 should grant more space to client by comparing fed_grant and
3097 cl_avail_grant, and return the granted space to client instead
3098 of "new granted" space, because client will call osc_init_grant
3099 to update the client grant space info.
3102 Frequency : when client reconnects to OST
3104 Description: Grant leak when OSC does a resend and replays bulk write
3105 Details : When osc reconnect to OST, OST(filter) should clear grant info of
3106 bulk write request, because the grant info while be sync between
3107 OSC and OST when reconnect, and we should ignore the grant info
3108 these of resend/replay write req.
3113 Description: Grant space more than avaiable space sometimes.
3114 Details : When then OST is about to be full, if two bulk writing from
3115 different clients came to OST. Accord the avaliable space of the
3116 OST, the first req should be permitted, and the second one
3117 should be denied by ENOSPC. But if the seconde arrived before
3118 the first one is commited. The OST might wrongly permit second
3119 writing, which will cause grant space > avaiable space.
3122 Frequency : when client is evicted
3124 Description: Grant might be wrongly erased when osc is evicted by OST
3125 Details : when the import is evicted by server, it will fork another
3126 thread ptlrpc_invalidate_import_thread to invalidate the
3127 import, where the grant will be set to 0. While the original
3128 thread will update the grant it got when connecting. So if
3129 the former happened latter, the grant will be wrongly errased
3130 because of this race.
3135 Description: Checking Stale with correct fid
3136 Details : ll_revalidate_it should uses de_inode instead of op_data.fid2
3137 to check whether it is stale, because sometimes, we want the
3138 enqueue happened anyway, and op_data.fid2 will not be initialized.
3140 Severity : enhancement
3142 Description: update patchless client
3143 Details : Add support for patchless client with 2.6.20, 2.6.21 and RHEL 5
3146 Frequency : only with 2.4 kernel
3148 Description: random memory corruption
3149 Details : size of struct ll_inode_info is to big for union inode.u and this
3150 can be cause of random memory corruption.
3155 Description: Memory leak in recovery
3156 Details : Lov_mds_md was not free in an error handler in mds_create_object.
3157 It should also check obd_fail before fsfilt_start, otherwise if
3158 fsfilt_start return -EROFS,(failover mds during mds recovery).
3159 then the req will return with repmsg->transno = 0 and rc = EROFS.
3160 and we met hit the assert LASSERT(req->rq_reqmsg->transno ==
3161 req->rq_repmsg->transno) in ptlrpc_replay_interpret. Fcc should
3162 be freed no matter whether fsfilt_commit success or not.
3165 Frequency : only with large numbers of cores on a single node
3167 Description: Prevents from taking the superblock lock in llap_from_page for
3169 Details : using LL_ORIGIN_REMOVEPAGE origin flag instead of LL_ORIGIN_UNKNOW
3170 for llap_from_page call in ll_removepage() prevents from taking
3171 the superblock lock for a soon died page.
3176 Description: Not check open intent error before release open handle
3177 Details : in some rare cases, the open intent error is not checked before
3178 release open handle, which may cause
3179 ASSERTION(open_req->rq_transno != 0), because it tries to release
3180 the failed open handle.
3185 Description: Set cat log bitmap only after create log success.
3186 Details : in some rare cases, the cat log bitmap is set too early. and it
3187 should be set only after create log success.
3191 Description: Accessing a block bevice can re-enable I/O when Lustre is
3192 tearing down a device.
3193 Details : dev_clear_rdonly(bdev) must be called in kill_bdev() instead of
3198 Description: service threads may hog cpus when there are a lot of requests
3199 Details : Insert cond_resched to give other threads a chance to use some CPU
3204 Description: the cat log was not initialized in recovery
3205 Details : When mds(mgs) do recovery, the tgt_count might be zero, so the
3206 unlink log on mds will not be initialized until mds post
3207 recovery. And also in mds post recovery, the unlink log will
3208 initialization will be done asynchronausly, so there will be race
3209 between add unlink log and unlink log initialization.
3213 Description: brw_stats were being printed incorrectly
3214 Details : brw_stats were being printed as log2 but all of them were not
3215 recorded as log2. Also remove some code duplication arising from
3216 filter_tally_{read,write}.
3220 Frequency : rare, only in recovery.
3221 Description: ASSERTION(req->rq_type != LI_POISON) failed
3222 Details : imp_lock should be held while iterating over imp_sending_list for
3223 prevent destroy request after get timeout in ptlrpc_queue_wait.
3227 Description: replay-single.sh test 52 fails
3228 Details : A lock's skiplist need to be cleanup when it being unlinked
3229 from its resource list.
3234 Description: Short directio read returns full requested size rather than
3236 Details : Direct I/O operations should return actual amount of bytes
3237 transferred rather than requested size.
3239 Severity : enhancement
3241 Description: metadata RPC reduction (e.g. for rm performance)
3242 Details : decrease the amount of synchronous RPC between clients and servers
3243 by canceling conflicing lock before the operation on the client
3244 and packing thier handles into the main operation RPC to server.
3246 Severity : enhancement
3248 Description: add #ifdef HAVE_KERNEL_CONFIG_H
3249 Details : kernels from 2.6.19 not need include linux/config.h, but add
3250 include linux/autoconf.h in commpiler command line.
3252 Severity : enhancement
3254 Description: patchless client support for 2.6.22 kernel
3255 Details : 2.6.22 has only one visble change, SLAB_CTOR_* constants is
3256 removed. In this case we need drop using os depended interface to
3257 kmem_cache and use cfs_mem_cache API.
3262 Description: fix mal-formatted messages
3263 Details : fix some mal-formatted DEBUG_REQ and LCONSOLE_ERROR_MSG messages
3267 Frequency : always in liblustre
3268 Description: wrong IS_ERR implementation in liblustre.h
3269 Details : fix IS_ERR implementation in liblustre.h for right detect errors.
3274 Description: Correct condition for output debug message.
3275 Details : inode i_nlink equal zero is not enough for output message about
3276 disk corruption, i_ctime and i_mode should be also checked.
3280 Frequency : always in patchless client
3281 Description: add configure check for truncate_complete_page
3282 Details : improve checks for exported symbols. This allow run check without
3283 sources, but with Module.symvers shipped with kernel distribution.
3284 add check for truncate_complete_page used by patchless client.
3288 Description: sanity.sh test_77h fails with "test_77h file compare failed"
3289 Details : test_77h uses a file which was messed by other test case.
3293 Description: Not Check whether lov_tgts is NULL in some lov functions
3294 Details : Checking whether lov_tgts is NULL in some functions.
3298 Description: replace obdo_alloc() with OBDO_ALLOC macro
3299 Details : nothing special is done in obdo_alloc() function, and for
3300 debugging purpose, it needs to be replaced with macros.
3304 Description: bad return value and errno from fcntl call
3305 Details : In liblustre API, errno should be a negative value if error
3310 Description: ptlrpc_check_set() LBUG
3311 Details : In case of positive reply from server and failed client bulk
3312 callback after bulk transfer shouldn't LBUG, but process this
3313 request as erroneous.
3315 Severity : enhancement
3317 Description: VFS operations stats tool.
3318 Details : Tool which collects stats by tracking value written in pid,
3319 ppid, gid and uses llstat to generate output to plot graph using
3321 Updated lustre/utils/Makefile.am
3322 Added lustre/utils/ltrack_stats.c
3324 Severity : enhancement
3326 Description: 2.6.18 server support (lustre 1.6.1)
3327 Details : Support for 2.6.18 kernels on the server side.
3332 Description: ASSERTION(imp->imp_conn_current) failed
3333 Details : an assertion failure is hit if a client node boots and attempts to
3334 mount a lustre filesystem faster than RECONNECT_INTERVAL seconds.
3337 Frequency : only for i686
3339 Description: 1.4.11 RC1 build fails for RHEL 4, i686
3340 Details : Fixed config variable for build.
3345 Description: Updated patchess for new RHEL4 kernel
3346 Details : Updated patch inode-nr_unused-2.6.9-rhel4.patch
3347 Updated patch jbd-stats-2.6.9.patch
3348 Updated patch qsnet-rhel4-2.6.patch
3349 Updated patch quota-deadlock-on-pagelock-core.patch
3350 Updated patch vfs_intent-2.6-rhel4.patch
3351 Updated patch vfs_races-2.6-rhel4.patch
3352 Updated series file 2.6-rhel4-titech.series
3353 Updated series file 2.6-rhel4.series
3354 Updated kernel config files
3359 Description: lquota slave complains LBUG when reconnecting with mds
3361 Details : quota slave depends qctxt->lqc_import to send its quota request.
3362 This pointer will be invalid if mds did failover or broke its
3363 connect to osts, which leads to LBUG.
3366 Frequency : when qunit size is too small(less than 20M)
3368 Description: write is stopped by improper -EDQUOT
3369 Details : If the master is busy and qunit size is small enough(let's say 1M),
3370 the slave can not get quota from master on time, which will lead
3371 slave to trigger a -EQUOTA to client.
3376 Description: Deadlock during metadata tests
3377 Details : in prune_dir_dentries(), shrink_dcache_parent() should not be
3378 called with the per-dentry lock held.
3381 Frequency : SLES9 only
3383 Description: Lustre patched kernel for SLES9 SP3 has NR_CPUS set to 8
3384 Details : set CONFIG_NR_CPUS to 128 instead of 8.
3386 Severity : enhancement
3388 Description: remove fs_prep_san_write operation and related patches
3389 Details : remove the ext3-san-jdike patches which are no longer useful.
3394 Description: LDISKFS-fs error (device sdc): ldiskfs_free_blocks
3395 Details : a disk corruption can cause the mballoc code to assert on a
3396 double free or other extent corruptions. Handle these with
3397 ext3_error() instead of assertions.
3400 Frequency : only with mballoc3 code and deep extent trees
3402 Description: ldiskfs_ext_search_right: bad header in inode: unexpected eh_depth
3403 Details : a wrong check of extent headers in ldiskfs_ext_search_right()
3404 can cause the filesystem to be remounted read-only.
3409 Description: lfsck built against 1.4.x cannot run against 1.6.0 lustre
3410 Details : the definition for OBD_IOC_GETNAME changed in 1.6.0. One of the
3411 few external users of this ioctl number is lfsck's call to
3412 llapi_lov_get_uuids() and this caused lfsck to fail at startup.
3413 Add the old ioctl number to the handler so both old and new
3418 Description: parallel lock callbacks
3419 Details : Instead of sending blocking and completion callbacks as separated
3420 requests, adding them to a set and sending in parallel.
3424 Description: Disable most debugging by default
3425 Details : To improve performance, disable most logging (for debug purposes)
3426 by default. VFSTRACE, RPCTRACE, and DLMTRACE are now off by
3427 default, and HA includes fewer messages.
3431 Frequency : only run on patchless client.
3432 Description: use do_facet on sanity.sh for test handling recoverables errors
3433 Details : use do_facet instead of direct use sysctl for set fail_loc on OST
3437 Description: hash tables for lists of nids, connections and uuids
3438 Details : Hash tables noticeably help when a lot of clients connect to a
3439 server, to faster identify duplicate connections or reconnects,
3440 also to faster find export to evict in manual eviction case.
3444 Description: Sometimes, when the server evict a client, and the client will
3445 not be evicted as soon as possible.
3446 Details : In enqueue req, the error was returned by intent, instead of
3447 rq_status which make ptlrpc layer not detect this error, and
3448 does not evict the client. So enqueue error should be returned
3452 Frequency : only at startup
3454 Description: Delay client connections to MDT until fist MDT->OST connect
3455 Details : If a client tried to create a new file before the MDT had
3456 connected to any OSTs, the create would return EIO. Now
3457 the client will simply block until the MDT connects to the
3458 first OST and the create can succeed.
3461 Frequency : only for SLES9
3463 Description: Routinely utilize latest Quadrics drivers in CFS releases
3464 Details : Update patch qsnet-suse-2.6.patch.
3467 Frequency : only for sles10
3469 Description: Update patches for SLES 10 SP1 kernel.
3470 Details : Update the patch vfs_intent-2.6-sles10.patch.
3475 Description: Routinely utilize latest Quadrics drivers in CFS releases
3476 Details : Update patch qsnet-rhel4-2.6.patch.
3479 Frequency : at statup only
3481 Description: mds_lov_synchronize race leads to various problems
3482 Details : simultaneous MDT->OST connections at startup can cause the
3483 sync to abort, leaving the OSC in a bad state.
3488 Description: Using wrong pointer in osc_brw_prep_request
3489 Details : Access to array[-1] can produce panic if kernel compiled with
3490 CONFIG_PAGE_ALLOC enabled
3492 Severity : enhancement
3494 Description: Async OSC create to avoid the blocking unnecessarily.
3495 Details : If a OST has no remain object, system will block on the creating
3496 when need to create a new object on this OST. Now, ways use
3497 pre-created objects when available, instead of blocking on an
3498 empty osc while others are not empty. If we must block, we block
3499 for the shortest possible period of time.
3503 Frequency : only in recovery
3504 Description: Mark OST as early accessible if his start SYNC.
3505 Details : osc_precreate return flag early accessible if oscc marked as
3506 OSCC_FLAG_SYNC_IN_PROGRESS.
3511 Description: Sometimes precreate code can triger create object on wrong ost
3512 Details : Wrong protected or not not restored variables aftre precreate loop
3513 can produce creation object on wrong ost.
3516 Frequency : oss recovery
3518 Description: llog_commit_thread cleanup should sync with llog_commit_thread
3520 Details : llog_commit_thread_count should be synced between llog_commit
3521 start and cleanup, so new llog_commit thread should not be started
3522 when llog_commit threads being stopped to avoid accessing some
3525 Severity : enhancement
3527 Description: Add printing inode info into message about error in writepage.
3529 --------------------------------------------------------------------------------
3531 2007-05-03 Cluster File Systems, Inc. <info@clusterfs.com>
3536 Frequency : on some architectures
3538 Description: 1.6 client sometimes fails to mount from a 1.4 MDT
3539 Details : Uninitialized flags sometimes cause configuration commands to
3543 Frequency : patchless clients only
3545 Description: missing __iget() symbol export
3546 Details : The __iget() symbol export is missing. To avoid the need for
3547 this on patchless clients the deathrow inode reaper is turned
3548 off, and we depend on the VM to clean up old inodes. This
3549 dependency was during via the fix for bug 12181.
3554 Description: sanity.sh fail: test_52b
3555 Details : The ll_inode_to_ext_flags() has a glitch which makes MDS return
3556 incorrect inode's flags to client.
3558 --------------------------------------------------------------------------------
3560 2007-04-19 Cluster File Systems, Inc. <info@clusterfs.com>
3562 * CONFIGURATION CHANGE. This version of Lustre WILL NOT
3563 INTEROPERATE with older versions automatically. In many cases a
3564 special upgrade step is needed. Please read the
3565 user documentation before upgrading any part of a 1.4.x system.
3566 * WARNING: Lustre configuration and startup changes are required with
3567 this release. See https://mail.clusterfs.com/wikis/lustre/MountConf
3569 * Support for kernels:
3570 2.4.21-47.0.1.EL (RHEL 3),
3571 2.6.5-7.283 (SLES 9),
3572 2.6.9-42.0.10.EL (RHEL 4),
3573 2.6.12.6 vanilla (kernel.org),
3574 2.6.16.27-0.9 (SLES10)
3575 * Client support for unpatched kernels:
3576 (see https://mail.clusterfs.com/wikis/lustre/PatchlessClient)
3577 2.6.16 - 2.6.19 vanilla (kernel.org),
3578 2.6.9-42.0.8EL (RHEL 4)
3579 * Recommended e2fsprogs version: 1.39.cfs6
3580 * Note that reiserfs quotas are disabled on SLES 10 in this kernel
3583 Severity : enhancement
3585 Description: Async OSC create to avoid the blocking unnecessarily.
3586 Details : If a OST has no remain object, system will block on the creating
3587 when need to create a new object on this OST. Now, ways use
3588 pre-created objects when available, instead of blocking on an
3589 empty osc while others are not empty. If we must block, we block
3590 for the shortest possible period of time.
3592 Severity : enhancement
3594 Description: MountConf
3595 Details : Lustre configuration is now managed via mkfs and mount
3596 commands instead of lmc and lconf. New obd types (MGS, MGC)
3597 are added for dynamic configuration management. See
3598 https://mail.clusterfs.com/wikis/lustre/MountConf for
3601 Severity : enhancement
3603 Description: dynamic OST addition
3604 Details : OSTs can now be added to a live filesystem
3606 Severity : enhancement
3608 Description: startup order invariance
3609 Details : MDTs and OSTs can be started in any order. Clients only
3610 require the MDT to complete startup.
3612 Severity : enhancement
3614 Description: parallel, asynchronous orphan cleanup
3615 Details : orphan cleanup is now performed in separate threads for each
3616 OST, allowing parallel non-blocking operation.
3618 Severity : enhancement
3620 Description: optimized stripe assignment
3621 Details : stripe assignments are now made based on ost space available,
3622 ost previous usage, and OSS previous usage, in order to try
3623 to optimize storage space and networking resources.
3625 Severity : enhancement
3627 Description: Permanently set tunables
3628 Details : All writable /proc/fs/lustre tunables can now be permanently
3629 set on a per-server basis, at mkfs time or on a live
3632 Severity : enhancement
3634 Description: Lustre message v2
3635 Details : Add lustre message format v2.
3637 Severity : enhancement
3639 Description: client OST exclusion list
3640 Details : Clients can be started with a list of OSTs that should be
3641 declared "inactive" for known non-responsive OSTs.
3645 Description: ENOENT returned for valid filehandle during dbench.
3646 Details : Check if a directory has children when invalidating dentries
3647 associated with an inode during lock cancellation. This fixes
3648 an incorrect ENOENT sometimes seen for valid filehandles during
3649 testing with dbench.
3652 Frequency : SFS test only (otherwise harmless)
3654 Description: SPEC SFS validation failure on NFS v2 over lustre.
3655 Details : Changes the blocksize for regular files to be 2x RPC size,
3656 and not depend on stripe size.
3658 Severity : enhancement
3660 Description: fine-grained SMP locking inside DLM
3661 Details : Improve DLM performance on SMP systems by removing the single
3662 per-namespace lock and replace it with per-resource locks.
3664 Severity : enhancement
3666 Description: don't hold multiple extent locks at one time
3667 Details : To avoid client eviction during large writes, locks are not
3668 held on multiple stripes at one time or for very large writes.
3669 Otherwise, clients can block waiting for a lock on a failed OST
3670 while holding locks on other OSTs and be evicted.
3672 Severity : enhancement
3674 Description: Multiple MD RPCs in flight.
3675 Details : Further unserialise some read-only MDT RPCs - learn about intents.
3676 To avoid overly-overloading MDT, introduce a limit on number of
3677 MDT RPCs in flight for a single client and add /proc controls
3678 to adjust this limit.
3680 Severity : enhancement
3682 Description: client read/write statistics
3683 Details : Add client read/write call usage stats for performance
3684 analysis of user processes.
3685 /proc/fs/lustre/llite/*/offset_stats shows non-sequential
3686 file access. extents_stats shows chunk size distribution.
3687 extents_stats_per_process show chunk size distribution per
3690 Severity : enhancement
3692 Description: per-client statistics on server
3693 Details : Add ldlm and operations statistics for each client in
3694 /proc/fs/lustre/mds|obdfilter/*/exports/
3696 Severity : enhancement
3698 Description: improved MDT statistics
3699 Details : Add detailed MDT operations statistics in
3700 /proc/fs/lustre/mds/*/stats
3702 Severity : enhancement
3704 Description: VFS operations stats
3705 Details : Add client VFS call stats, trackable by pid, ppid, or gid
3706 /proc/fs/lustre/llite/*/stats_track_[pid|ppid|gid]
3711 Description: Fix client-side osc byte counters
3712 Details : The osc read/write byte counters in
3713 /proc/fs/lustre/osc/*/stats are now working
3716 Frequency : always as root on SLES
3718 Description: Failure of copying files with lustre special EAs.
3719 Details : Client side always return success for setxattr call for lustre
3720 special xattr (currently only "trusted.lov").
3725 Description: Refcount LNET uuids
3726 Details : The global LNET uuid list grew linearly with every startup;
3727 refcount repeated list entries instead of always adding to
3730 Severity : enhancement
3732 Description: Dynamic service threads
3733 Details : Within a small range, start extra service threads
3734 automatically when the request queue builds up.
3737 Frequency : mixed-endian client/server environments
3739 Description: mixed-endian crashes
3740 Details : The new msg_v2 system had some failures in mixed-endian
3743 Severity : enhancement
3745 Description: Easy OST removal
3746 Details : OSTs can be permanently deactivated with e.g. 'lctl
3747 conf_param lustre-OST0001.osc.active=0'
3749 Severity : enhancement
3751 Description: MGS proc entries
3752 Details : Added basic proc entries for the MGS showing what filesystems
3755 Severity : enhancement
3757 Description: provide MGS failover
3758 Details : Added config lock reacquisition after MGS server failover.
3760 Severity : enhancement
3762 Description: add Linux 2.4 support
3763 Details : Added support for RHEL 2.4.21 kernel for 1.6 servers and clients
3767 Description: a large application tries to do I/O to the same resource and dies
3768 in the middle of it.
3769 Details : Check the req->rq_arrival time after the call to
3770 ost_brw_lock_get(), but before we do anything about
3771 processing it & sending the BULK transfer request. This
3772 should help move old stale pending locks off the queue as
3773 quickly as obd_timeout.
3776 Frequency : when an incorrect nid is specified during startup
3778 Description: ptlrpc connect to non-existant node causes kernel crash
3779 Details : LNET can't be re-entered from an event callback, which
3780 happened when we expire a message after the export has been
3781 cleaned up. Instead, hand the zombie cleanup off to another
3784 Severity : enhancement
3786 Description: plain/inodebits lock performance improvement
3787 Details : Grouping plain/inodebits in granted list by their request modes
3788 and bits policy, thus improving the performance of search through
3792 Frequency : only if OST filesystem is corrupted
3794 Description: client incorrectly hits assertion in ptlrpc_replay_req()
3795 Details : for a short time RPCs with bulk IO are in the replay list,
3796 but replay of bulk IOs is unimplemented. If the OST filesystem
3797 is corrupted due to disk cache incoherency and then replay is
3798 started it is possible to trip an assertion. Avoid putting
3799 committed RPCs into the replay list at all to avoid this issue.
3802 Frequency : liblustre (e.g. catamount) on a large cluster with >= 8 OSTs/OSS
3804 Description: System hang on startup
3805 Details : This bug allowed the liblustre (e.g. catamount) client to
3806 return to the app before handling all startup RPCs. This
3807 could leave the node unresponsive to lustre network traffic
3808 and manifested as a server ptllnd timeout.
3810 Severity : enhancement
3812 Description: Add "/proc/sys/lustre/debug_peer_on_timeout"
3813 Details : liblustre envirable: LIBLUSTRE_DEBUG_PEER_ON_TIMEOUT
3814 boolean to control whether to print peer debug info when a
3815 client's RPC times out.
3818 Frequency : only for kernels with patches from Lustre below 1.4.3
3820 Description: Remove old rdonly API
3821 Details : Remove old rdonly API which unused from at least lustre 1.4.3
3824 Frequency : only for devices with external journals
3826 Description: Set external device read-only also
3827 Details : During a commanded failover stop, we set the disk device
3828 read-only while the server shuts down. We now also set any
3829 external journal device read-only at the same time.
3832 Frequency : when upgrading from 1.4 while trying to change parameters
3834 Description: The wrong (new) MDC name was used when setting parameters for
3835 upgraded MDT's. Also allows changing of OSC (and MDC)
3836 parameters if --writeconf is specified at tunefs upgrade time.
3839 Frequency : when setting specific ost indicies
3841 Description: QOS code breaks on skipped indicies
3842 Details : Add checks for missing OST indicies in the QOS code, so OSTs
3843 created with --index need not be sequential.
3845 Severity : enhancement
3847 Description: Add uninit_groups feature to ldiskfs2 to speed up e2fsck
3848 Details : The uninit_groups feature works in conjunction with the kernel
3849 filesystem code (ldiskfs2 only) and e2fsprogs-1.39-cfs6 to speed
3850 up the pass1 processing of e2fsck. This is a read-only feature
3851 in ldiskfs2 only, so older kernels and current ldiskfs cannot
3852 mount filesystems that have had this feature enabled.
3854 Severity : enhancement
3856 Description: Improve multi-block allocation algorithm to avoid fragmentation
3857 Details : The mballoc3 code (ldiskfs2 only) adds new mechanisms to improve
3858 allocation locality and avoid filesystem fragmentation.
3860 ------------------------------------------------------------------------------
3862 2007-04-01 Cluster File Systems, Inc. <info@clusterfs.com>
3864 * Support for kernels:
3865 2.4.21-47.0.1.EL (RHEL 3)
3866 2.6.5-7.283 (SLES 9)
3867 2.6.9-42.0.10.EL (RHEL 4)
3868 2.6.12.6 vanilla (kernel.org)
3869 2.6.16.27-0.9 (SLES 10)
3870 * Recommended e2fsprogs version: 1.39.cfs5
3872 * Note that reiserfs quotas are disabled on SLES 10 in this kernel
3876 Frequency : occasional, depends on client load and configuration
3877 Bugzilla : 12181, 12203
3878 Description: data loss for recently-modified files
3880 Details : In some cases it is possible that recently written or created
3881 files may not be written to disk in a timely manner (this should
3882 normally be within 30s unless client IO load is very high).
3883 The problem appears as zero-length files or files that are a
3884 multiple of 1MB in size after a client crash or client eviction
3885 that are missing data at the end of the file.
3887 This problem is more likely to be hit on clients where files are
3888 repeatedly created and unlinked in the same directory, clients
3889 have a large amount of RAM, have many CPUs, the filesystem has
3890 many OSTs, the clients are rebooted frequently, and/or the files
3891 are not accessed by other nodes after being written.
3893 The presence of the problem can be detected by looking at
3894 /proc/sys/fs/inode-state. If the first number (nr_inodes) is
3895 smaller than the second (nr_unused) then dirty files will not
3896 be flushed automatically to disk. "sync; sleep 10" should be
3897 run several times on the node before unmounting it to update
3898 Lustre (this is also safe to run on nodes without this problem).
3900 There is also a related kernel bug in the RHEL4 4 2.6.9 kernel
3901 that can cause this same problem, so customers using that kernel
3902 also need to update the kernel in addition to Lustre. In order
3903 to properly fix this bug, the RHEL3 2.4.21 kernel is also updated.
3905 It is normal that files written just before a client crash (less
3906 than 30s) may not yet have been flushed to disk, even for local
3910 Frequency : frequent on thin XT3 nodes
3912 Description: UUID collision on thin XT3 Linux nodes
3913 Details : UUIDs on Compute Node Linux XT3 nodes were not generated
3914 randomly, since we relied on an insufficiently-seeded PRNG.
3919 Description: OSS hangs after "All ost request buffers busy"
3920 Details : A deadlock between quota and journal operations caused OSS
3921 hangs after printing "All ost request buffers busy."
3924 Frequency : always on liblustre builds
3926 Description: Cleanup compiler warnings on liblustre
3929 Frequency : always on liblustre builds on XT3
3931 Description: LC_CONFIG_CDEBUG don't run while build liblustre on XT3.
3935 Description: Addition of EXT3_FEATURE_RO_COMPAT_DIR_NLINKS flag for
3936 > 32000 subdirectories
3937 Details : Add EXT3_FEATURE_RO_COMPAT_DIR_NLINK flag to
3938 EXT3_FEATURE_RO_COMPAT_SUPP. This flag will be set whenever
3939 subdirectory count crosses 32000. This will aid e2fsck to
3940 correctly handle more than 32000 subdirectories.
3943 Frequency : liblustre (e.g. catamount) on a large cluster with >= 8 OSTs/OSS
3945 Description: System hang on startup
3946 Details : This bug allowed the liblustre (e.g. catamount) client to
3947 return to the app before handling all startup RPCs. This
3948 could leave the node unresponsive to lustre network traffic
3949 and manifested as a server ptllnd timeout.
3951 Severity : enhancement
3953 Description: Add "/proc/sys/lustre/debug_peer_on_timeout"
3954 (liblustre envirable: LIBLUSTRE_DEBUG_PEER_ON_TIMEOUT)
3955 boolean to control whether to print peer debug info when a
3956 client's RPC times out.
3961 Description: make O_SYNC working on 2.6 kernels
3962 Details : 2.6 kernels use different method for mark pages for write,
3963 so need add a code to lustre for O_SYNC work.
3968 Description: Failure to close file and release space on NFS
3969 Details : Put inode details into lock acquired in ll_intent_file_open.
3970 Use mdc_intent_lock in ll_intent_open to properly
3971 detect all kind of errors unhandled by mdc_enqueue.
3976 Description: proc file read during shutdown sometimes raced obd removal,
3978 Details : Add lock to prevent obd access after proc file removal.
3981 Frequency : Only for files larger than 4GB on 32-bit clients.
3983 Description: improperly doing page alignment of locks
3984 Details : Modify lustre core code to use CFS_PAGE_* defines instead of
3985 PAGE_*. Make CFS_PAGE_MASK a 64-bit mask.
3990 Description: RPCs being resent when they shouldn't be
3991 Details : Some RPCs that should not be resent are being resent. This
3992 can cause inconsistencies in the RPC state machine. Do not
3993 resend such requests.
3996 Frequency : rare, only with NFS export
3998 Description: Crash on NFS re-export node
3999 Details : under very unusual load conditions an assertion is hit in
4000 ll_intent_file_open()
4003 Frequency : only if OST filesystem is corrupted
4005 Description: client incorrectly hits assertion in ptlrpc_replay_req()
4006 Details : for a short time RPCs with bulk IO are in the replay list,
4007 but replay of bulk IOs is unimplemented. If the OST filesystem
4008 is corrupted due to disk cache incoherency and then replay is
4009 started it is possible to trip an assertion. Avoid putting
4010 committed RPCs into the replay list at all to avoid this issue.
4015 Description: large O_DIRECT requests fail under memory pressure/fragmentation
4016 Details : Large single O_DIRECT read and write calls can fail to allocate
4017 a sufficiently large buffer to process the request. In case of
4018 allocation failure the allocation is retried with a smaller
4019 buffer and broken into smaller requests.
4021 Severity : enhancement
4023 Description: Add -o localflock option to simulate old noflock behaviour.
4024 Details : This will achieve local-only flock/fcntl locks coherentness.
4029 Description: versioning check is incomplete
4030 Details : Checking the version difference of client vs. server, report
4031 error if the gap is too big.
4036 Description: add support PG_writeback bit
4037 Details : add support for PG_writeback bit for Lustre, for more carefull
4038 work with page cache in 2.6 kernel. This also fix some deadlocks
4039 and remove hack for work O_SYNC with 2.6 kernel.
4041 Severity : enhancement
4043 Description: Add uninit_groups feature to ldiskfs2 to speed up e2fsck
4044 Details : The uninit_groups feature works in conjunction with the kernel
4045 filesystem code (ldiskfs2 only) and e2fsprogs-1.39-cfs6 to speed
4046 up the pass1 processing of e2fsck. This is a read-only feature
4047 in ldiskfs2 only, so older kernels and current ldiskfs cannot
4048 mount filesystems that have had this feature enabled.
4050 Severity : enhancement
4052 Description: Improve multi-block allocation algorithm to avoid fragmentation
4053 Details : The mballoc3 code (ldiskfs2 only) adds new mechanisms to improve
4054 allocation locality and avoid filesystem fragmentation.
4056 ------------------------------------------------------------------------------
4058 2007-02-09 Cluster File Systems, Inc. <info@clusterfs.com>
4060 * Support for kernels:
4061 2.6.9-42.0.3.EL (RHEL 4)
4062 2.6.5-7.276 (SLES 9)
4063 2.4.21-47.0.1.EL (RHEL 3)
4064 2.6.12.6 vanilla (kernel.org)
4065 2.6.16.21-0.8 (SLES10)
4066 * Recommended e2fsprogs version: 1.39.cfs2-0
4068 * The backwards-compatible /proc/sys/portals symlink has been removed
4069 in this release. Before upgrading, please ensure that you change
4070 any configuration scripts or /etc/sysctl.conf files that access
4071 /proc/sys/portals/* or sysctl portals.* to use the corresponding
4072 entry in /proc/sys/lnet or sysctl lnet.*. This change can be made
4073 in advance of the upgrade on any system running Lustre 1.4.6 or
4074 newer, since /proc/sys/lnet was added in that version.
4075 * Note that reiserfs quotas are disabled on SLES 10 in this kernel
4079 Frequency : only when quota is used
4081 Description: avoid scanning export list for quota master
4082 Details : Change the algorithms to avoid scanning export list in order
4083 to improve the efficiency.
4086 Frequency : MDS failover only, very rarely
4088 Description: "went back in time" messages on mds failover
4089 Details : The greatest transno may be lost when the current operation
4090 finishes with an error (transno==0) and the client's last_rcvd
4091 record is over-written. Save the greatest transno in the
4092 mds_last_transno for this case.
4095 Frequency : always for specific kernels and striping counts
4097 Description: client may get "Matching packet too big" without ACL support
4098 Details : Clients compiled without CONFIG_FS_POSIX_ACL get an error message
4099 when trying to access files in certain configurations. The
4100 clients should in fact be denied when mounting because they do
4101 not understand ACLs.
4104 Frequency : Cray XT3 with more than 4000 clients and multiple jobs
4106 Description: many clients connecting with IO in progress causes connect timeouts
4107 Details : Avoid synchronous journal commits to avoid delays caused by many
4108 clients connecting/disconnecting when bulk IO is in progress.
4109 Queue liblustre connect requests on OST_REQUEST_PORTAL instead of
4110 OST_IO_PORTAL to avoid delays behind potentially many pending
4114 Frequency : occasionally with multiple writers to a single file
4116 Description: shared writes to file may result in wrong size reported by stat()
4117 Details : Allow growing of kms when extent lock is cancelled
4120 Frequency : always with random mmap IO to multi-striped file
4122 Description: mmap write might be lost if we are writing to a 'hole' in stripe
4123 Details : Only if the hole is at the end of OST object so that kms is too
4124 small. Fix is to increase kms accordingly in ll_nopage.
4127 Frequency : rare, only if OST filesystem is inconsistent with MDS filesystem
4129 Description: writes to a missing object would leak memory on the OST
4130 Details : If there is an inconsistency between the MDS and OST filesystems,
4131 such that the MDS references an object that doesn't exist, writes
4132 to that object will leak memory due to incorrect cleanup in the
4133 error handling path, eventually running out of memory on the OST.
4138 Description: Creating too long symlink causes lustre errors
4139 Details : Check symlink and name lengths before sending requests to MDS.
4142 Frequency : only if flock is enabled (not on by default)
4144 Description: posix locks not released on fd closure on 2.6.9+
4145 Details : We failed to add posix locks to list of inode locks on 2.6.9+
4146 kernels, this caused such locks not to be released on fd close and
4147 then assertions on fs unmount about still used locks.
4150 Frequency : MDS failover only, very rarely
4152 Description: clients may get ASSERTION(granted_lock != NULL)
4153 Details : When request was taking a long time, and a client was resending
4154 a getattr by name lock request. The were multiple lock requests
4155 with the same client lock handle and
4156 mds_getattr_name->fixup_handle_for_resent_request found one of the
4157 lock handles but later failed with ASSERTION(granted_lock != NULL).
4162 Description: handle->h_buffer_credits > 0, assertion failure
4163 Details : h_buffer_credits is zero after truncate, causing assertion
4164 failure. This patch extends the transaction or creates a new
4168 Frequency : NFS re-export or patchless client
4169 Bugzilla : 11179, 10796
4170 Description: Crash on NFS re-export node (__d_move)
4171 Details : We do not want to hash the dentry if we don't have a lock.
4172 But if this dentry is later used in d_move, we'd hit uninitialised
4173 list head d_hash, so we just do this to init d_hash field but
4174 leave dentry unhashed.
4177 Frequency : NFS re-export or patchless client
4179 Description: NFS exports has problem with symbolic link
4180 Details : lustre client didn't properly install dentry when re-exported
4181 to NFS or running patchless client.
4184 Frequency : NFS re-export or patchless client
4186 Description: Various nfs/patchless fixes.
4187 Details : fixes reuse disconected alias for lookup process - this fixes
4188 warning "find_exported_dentry: npd != pd",
4189 fix permission error with open files at nfs.
4190 fix apply umask when do revalidate.
4193 Frequency : occasional
4195 Description: Crash on NFS re-export node
4196 Details : calling clear_page() on the wrong pointer triggered oops in
4197 generic_mapping_read().
4200 Frequency : rarely, using O_DIRECT IO
4202 Description: unaligned directio crashes client with LASSERT
4203 Details : check for unaligned buffers before trying any requests.
4206 Frequency : rarely, using CFS RAID5 patches in non-standard kernel series
4208 Description: stale data returned from RAID cache
4209 Details : If only a small amount of IO is done to the RAID device before
4210 reading it again it is possible to get stale data from the RAID
4211 cache instead of reading it from disk.
4214 Frequency : always for sles10 kernel
4216 Description: sles10 support
4217 Details : ll_follow_link: compile fixes and using of nd_set_link
4218 under newer kernels.
4221 Frequency : depends on arch, kernel and compiler version, always on sles10
4224 Description: recursive or deep enough symlinks cause stack overflow
4225 Details : getting rid of large stack-allocated variable in
4229 Frequency : depends on hardware
4231 Description: lustre write performance loss in the SLES10 kernel
4232 Details : the performance loss is caused by using of write barriers in the
4233 ext3 code. The SLES10 kernel turns barrier support on by
4234 default. The fix is to undo that change for ldiskfs.
4236 ------------------------------------------------------------------------------
4238 2006-12-09 Cluster File Systems, Inc. <info@clusterfs.com>
4240 * Support for kernels:
4241 2.6.9-42.0.3EL (RHEL 4)
4242 2.6.5-7.276 (SLES 9)
4243 2.4.21-47.0.1.EL (RHEL 3)
4244 2.6.12.6 vanilla (kernel.org)
4248 Frequency : quota enabled and large files being deleted
4250 Description: releasing more than 4GB of quota at once hangs OST
4251 Details : If a user deletes more than 4GB of files on a single OST it
4252 will cause the OST to spin in an infinite loop. Release
4253 quota in < 4GB chunks, or use a 64-bit value for 1.4.7.1+.
4258 Description: statfs data retrieved from /proc may be stale or zero
4259 Details : When reading per-device statfs data from /proc, in the
4260 {kbytes,files}_{total,free,avail} files, it may appear
4261 as zero or be out of date.
4264 Frequency : systems with MD RAID1 external journal devices
4266 Description: lconf's call to blkid is confused by RAID1 journal devices
4267 Details : Use the "blkid -l" flag to locate the MD RAID device instead
4268 of returning all block devices that match the journal UUID.
4271 Frequency : always, for aggregate stripe size over 4GB
4273 Description: "lfs setstripe" fails assertion when setting 4GB+ stripe width
4274 Details : Using "lfs setstripe" to set stripe size * stripe count over 4GB
4275 will fail the kernel with "ASSERTION(lsm->lsm_xfersize != 0)"
4278 Frequency : always if "lfs find" used on a local file/directory
4280 Description: "lfs find" segfaults if used on a local file/directory
4281 Details : The case where a directory component was not specified wasn't
4282 handled correctly. Handle this properly.
4285 Frequency : always on ppc64
4287 Description: the write to an ext3 filesystem mounted with mballoc got stuck
4288 Details : ext3_mb_generate_buddy() uses find_next_bit() which does not
4289 perform endianness conversion.
4292 Frequency : rarely (truncate to non-zero file size after write under load)
4293 Bugzilla : 10730, 10687
4294 Description: Files padded with zeros to next 4K multiple
4295 Details : With filesystems mounted using the "extents" option (2.6 kernels)
4296 it is possible that files that are truncated to a non-zero size
4297 immediately after being written are filled with zero bytes beyond
4298 the truncated size. No file data is lost.
4300 Severity : enhancement
4302 Description: Allow recovery/failover for liblustre clients.
4303 Details : liblustre clients were unaware of failover configurations until
4306 Severity : enhancement
4308 Description: user file locks should fail when not mounting with flock option
4309 Details : Set up an error-returning stub in ll_file_operations.lock field
4310 to prevent incorrect behaviour when client is mounted without
4311 flock option. Also, set up properly f_op->flock field for
4315 Frequency : always on ia64
4317 Description: "lfs df" loops on printing out MDS statfs information
4318 Details : The obd_ioctl_data was not initialized and in some systems
4319 this caused a failure during the ioctl that did not return
4320 an error. Initialize the struct and return an error on failure.
4323 Frequency : SLES 9 only
4325 Description: Error of copying files with lustre special EAs as root
4326 Details : Client side always return success for setxattr call for lustre
4327 special xattr (currently only "trusted.lov").
4330 Frequency : rarely on clusters with both ia64+i386 clients
4332 Description: ia64+i686 clients doing shared IO on the same file may LBUG
4333 Details : In rare cases when both ia64+i686 (or other mixed-PAGE_SIZE)
4334 clients are doing concurrent writes to the same file it is
4335 possible that the ia64 clients may LASSERT because the OST
4336 extent locks are not PAGE_SIZE aligned. Ensure that grown
4337 locks are always aligned on the request boundary.
4340 Frequency : specific use, occasional
4342 Description: Overwriting in use executable truncates on-disk binary image
4343 Details : If one node attempts to overwrite an executable in use by
4344 another node, we now correctly return ETXTBSY instead of
4345 truncating the file.
4347 Severity : enhancement
4349 Description: Async OSC create to avoid the blocking unnecessarily.
4350 Details : If a OST has no remain object, system will block on the creating
4351 when need to create a new object on this OST. Now, ways use
4352 pre-created objects when available, instead of blocking on an
4353 empty osc while others are not empty. If we must block, we block
4354 for the shortest possible period of time.
4359 Description: chmod on Lustre root is propagated to other clients
4360 Details : Re-validate root's dentry in ll_lookup_it to avoid having it
4361 invalid by the follow_mount time.
4364 Frequency : liblustre clients only
4366 Description: Race in 'instant cancel' lock handling could lead to such locks
4367 never to be granted in case of SMP MDS
4368 Details : Do not destroy not yet granted but cbpending locks in
4372 Frequency : replay/resend of open
4374 Description: non null lock assetion failure in mds_intent_policy
4375 Details : Trying to replay/resend lockless open requests resulted in
4376 mds_open() returning 0 with no lock. Now it sets a flag if
4377 a lock is going to be returned.
4379 Severity : enhancement
4381 Description: Checksum enhancements
4382 Details : New checksum enhancements allow for resending RPCs that failed
4385 Severity : enhancement
4387 Description: Tunables on number of dirty pages in cacche
4388 Details : Allow to set limit on number of dirty pages cached.
4393 Description: client crash on unmount - lock still has references
4394 Details : In some error handling cases it was possible to leak a lock
4395 reference on a client while accessing a file. This was not
4396 harmful to the client during operation, but would cause the
4397 client to crash when the filesystem is unmounted.
4400 Frequency : specific case, rare
4402 Description: ETXTBSY on mds though file not in use by client
4403 Details : ETXTBSY is no longer incorrectly returned when attempting to
4404 chmod or chown a directory that the user previously tried to
4405 execute or a currently-executing binary.
4408 Frequency : extremely rare except on liblustre-based clients
4410 Description: Lustre space not freed when files are deleted
4411 Details : Clean up open-unlinked files after client eviction. Previously
4412 the unlink was skipped and the files remained as orphans.
4417 Description: OST failure "would be an LBUG" in waiting_locks_callback()
4418 Details : In some cases it was possible to send a blocking callback to
4419 a client doing a glimpse, even though that client didn't get
4420 a lock granted. When the glimpse lock is cancelled on the OST
4421 the freed lock is left on the waiting list and corrupted the list.
4424 Frequency : all core dumps
4426 Description: Broke core dumps to lustre
4427 Details : Negative dentry may be unhashed if parent does not have UPDATE
4428 lock, but some callers, e.g. do_coredump, expect dentry to be
4429 hashed after successful create, hash it in ll_create_it.
4431 ------------------------------------------------------------------------------
4433 2006-09-13 Cluster File Systems, Inc. <info@clusterfs.com>
4435 * Support for kernels:
4436 2.6.9-42.0.2.EL (RHEL 4)
4437 2.6.5-7.276 (SLES 9)
4438 2.4.21-40.EL (RHEL 3)
4439 2.6.12.6 vanilla (kernel.org)
4443 Frequency : always on RHEL 3
4445 Description: Number of open files grows over time
4446 Details : The number of open files grows over time, whether or not
4447 Lustre is started. This was due to a filp leak introduced
4448 by one of our kernel patches.
4450 ------------------------------------------------------------------------------
4452 2006-08-20 Cluster File Systems, Inc. <info@clusterfs.com>
4454 * Support for kernels:
4455 2.6.9-42.EL (RHEL 4)
4456 2.6.5-7.267 (SLES 9)
4457 2.4.21-40.EL (RHEL 3)
4458 2.6.12.6 vanilla (kernel.org)
4463 Bugzilla : 5719, 9635, 9792, 9684
4464 Description: OST (or MDS) trips assertions in (re)connection under heavy load
4465 Details : If a server is under heavy load and cannot reply to new
4466 connection requests before the client resends the (re)connect,
4467 the connection handling code can behave badly if two service
4468 threads are concurrently handing separate (re)connections from
4469 the same client. Add better locking to the connection handling
4470 code, and ensure that only a single connection will be processed
4471 for a given client UUID, even if the lock is dropped.
4473 Severity : enhancement
4475 Description: add TCP zero-copy support to kernel
4476 Details : Add support to the kernel TCP stack to allow zero-copy bulk
4477 sends if the hardware supports scatter-gather and checksumming.
4478 This allows socklnd to do client-write and server-read more
4479 efficiently and reduce CPU utilization from skbuf copying.
4482 Frequency : only if NFS exporting from client
4484 Description: NULL pointer deref in ll_iocontrol() if chattr mknod file
4485 Details : If setting attributes on a file created under NFS that had
4486 never been opened it would be possible to oops the client
4487 if the file had no objects.
4491 Bugzilla : 9326, 10402, 10897
4492 Description: client crash in ptlrpcd_wake() thread when sending async RPC
4493 Details : It is possible that ptlrpcd_wake() dereferences a freed async
4494 RPC. In rare cases the ptlrpcd thread alread processed the RPC
4495 before ptlrpcd_wake() was called and the request was freed.
4498 Frequency : always for liblustre
4500 Description: liblustre client does MDS+OSTs setattr RPC for each write
4501 Details : When doing a write from a liblustre client, the client
4502 incorrectly issued an RPC to the MDS and each OST the file was
4503 striped over in order to update the timestamps. When writing
4504 with small chunks and many clients this could overwhelm the MDS
4505 with RPCs. In all cases it would slow down the write because
4506 these RPCs are unnecessary.
4508 Severity : enhancement
4510 Description: allow number of MDS service threads to be changed at module load
4511 Details : It is now possible to change the number of MDS service threads
4512 running. Adding "options mds mds_num_threads={N}" to the MDS's
4513 /etc/modprobe.conf will set the number of threads for the next
4514 time Lustre is restarted (assuming the "mds" module is also
4515 reloaded at that time). The default number of threads will
4516 stay the same, 32 for most systems.
4521 Description: OST crash if filesystem is unformatted or corrupt
4522 Details : If an OST is started on a device that has never been formatted
4523 or if the filesystem is corrupt and cannot even mount then the
4524 error handling cleanup routines would dereference a NULL pointer.
4529 Description: NULL pointer deref in llap_from_page.
4530 Details : get_cache_page_nowait can return a page with NULL (or otherwise
4531 incorrect) mapping if the page was truncated/reclaimed while it was
4532 searched for. Check for this condition and skip such pages when
4533 doing readahead. Introduce extra check to llap_from_page() to
4534 verify page->mapping->host is non-NULL (so page is not anonymous).
4537 Frequency : Sometimes when using sys_sendfile
4539 Description: "page not covered by a lock" warnings from ll_readpage
4540 Details : sendfile called ll_readpage without right page locks present.
4541 Now we introduced ll_file_sendfile that does necessary locking
4542 around call to generic_file_sendfile() much like we do in
4546 Frequency : with certain MDS communication failures at client mount time
4548 Description: NULL pointer deref after failed client mount
4549 Details : a client connection request may delayed by the network layer
4550 and not be sent until after the PTLRPC layer has timed out the
4551 request. If the client fails the mount immediately it will try
4552 to clean up before the network times out the request. Add a
4553 reference from the request import to the obd device and delay
4554 the cleanup until the network drops the request.
4557 Frequency : occasionally during client (re)connect
4559 Description: assertion failure during client (re)connect
4560 Details : processing a client connection request may be delayed by the
4561 client or server longer than the client connect timeout. This
4562 causes the client to resend the connection request. If the
4563 original connection request is replied in this interval, the
4564 client may trip an assertion failure in ptlrpc_connect_interpret()
4565 which thought it would be the only running connect process.
4568 Frequency : only with obd_echo servers and clients that are rebooted
4570 Description: kernel BUG accessing uninitialized data structure
4571 Details : When running an obd_echo server it did not start the ping_evictor
4572 thread, and when a client was evicted an uninitialized data
4573 structure was accessed. Start the ping_evictor in the RPC
4574 service startup instead of the OBD startup.
4576 Severity : enhancement
4577 Bugzilla : 10193 (patchless)
4578 Description: Remove dependency on various unexported kernel interfaces.
4579 Details : No longer need reparent_to_init, exit_mm, exit_files,
4580 sock_getsockopt, filemap_populate, FMODE_EXEC, put_filp.
4583 Frequency : rare (only users of deprecated and unsupported LDAP config)
4585 Description: write_conf for zeroconf mount queried LDAP incorrectly for client
4586 Details : LDAP apparently contains 'lustreName' attributes instead of
4587 'name'. A simple remapping of the name is sufficient.
4590 Frequency : rare (only with non-default dump_on_timeout debug enabled)
4592 Description: waiting_locks_callback trips kernel BUG if client is evicted
4593 Details : Running with the dump_on_timeout debug flag turned on makes
4594 it possible that the waiting_locks_callback() can try to dump
4595 the Lustre kernel debug logs from an interrupt handler. Defer
4596 this log dumping to the expired_lock_main() thread.
4598 Severity : enhancement
4600 Description: Support NFS exporting on 2.6 kernels.
4601 Details : Implement non-rawops metadata methods for NFS server to use without
4602 changing NFS server code.
4605 Frequency : very rare (synthetic metadata workload only)
4607 Description: two racing renames might cause an MDS thread to deadlock
4608 Details : Running the "racer" program may cause one MDS thread to rename
4609 a file from being the source of a rename to being the target of
4610 a rename at exactly the same time that another thread is doing
4611 so, and the second thread has already enqueued these locks after
4612 doing a lookup of the target and is trying to relock them in
4613 order. Ensure that we don't try to re-lock the same resource.
4616 Frequency : only very large systems with liblustre clients
4618 Description: slow eviction of liblustre clients with the "evict_by_nid" RPC
4619 Details : Use asynchronous set_info RPCs to send the "evict_by_nid" to
4620 all OSTs in parallel. This allows the eviction of stale liblustre
4621 clients to proceed much faster than if they were done in series,
4622 and also offers similar improvements for other set_info RPCs.
4627 Description: excessive CPU usage during initial read phase on client
4628 Details : During the initial read phase on a client, it would agressively
4629 retry readahead on the file, consuming too much CPU and impacting
4630 performance (since 1.4.5.8). Improve the readahead algorithm
4631 to avoid this, and also improve some other common cases (read
4632 of small files in particular, where "small" is files smaller than
4633 /proc/fs/lustre/llite/*/max_read_ahead_whole_mb, 2MB by default).
4638 Description: MDS crash when receiving packet with unknown intent.
4639 Details : Do not LBUG in unknown intent case, just return -EFAULT
4641 Severity : enhancement
4642 Bugzilla : 9293, 9385
4643 Description: MDS RPCs are serialised on client. This is unnecessary for some.
4644 Details : Do not serialize getattr (non-intent version) and statfs.
4647 Frequency : occasional, when OST network is overloaded/intermittent
4649 Description: client evicted by OST after bulk IO timeout
4650 Details : If a client sends a bulk IO request (read or write) the OST
4651 may evict the client if it is unresposive to its data GET/PUT
4652 request. This is incorrect if the network is overloaded (takes
4653 too long to transfer the RPC data) or dropped the OST GET/PUT
4654 request. There is no need to evict the client at all, since
4655 the pinger and/or lock callbacks will handle this, and the
4656 client can restart the bulk request.
4659 Frequency : Always when mmapping file with no objects
4661 Description: client crashes when mmapping file with no objects
4662 Details : Check that we actually have objects in a file before doing any
4663 operations on objects in ll_vm_open, ll_vm_close and
4669 Description: Request leak when working with deleted CWD
4670 Details : Introduce advanced request refcount tracking for requests
4671 referenced from lustre intent.
4673 Severity : Enhancement
4675 Description: Cache open file handles on client.
4676 Details : MDS now will return special lock along with openhandle, if
4677 requested and client is allowed to hold openhandle, even if unused,
4678 until such a lock is revoked. Helps NFS a lot, since NFS is opening
4679 closing files for every read/write openration.
4681 Severity : Enhancement
4683 Description: Cache open negative dentries on client when possible.
4684 Details : Guard negative dentries with UPDATE lock on parent dir, drop
4685 negative dentries on lock revocation.
4690 Description: Remounting a client read-only wasn't possible with a zconf mount
4691 Details : It wasn't possible to remount a client read-only with llmount.
4693 Severity : enhancement
4694 Description: Include MPICH 1.2.6 Lustre ADIO interface patch
4695 Details : In lustre/contrib/ or /usr/share/lustre in RPM a patch for
4696 MPICH is included to add Lustre-specific ADIO interfaces.
4697 This is based closely on the UFS ADIO layer and only differs
4698 in file creation, in order to allow the OST striping to be set.
4699 This is user-contributed code and not supported by CFS.
4704 Description: extended inode attributes (immutable, append-only) work improperly
4705 when 2.4 and 2.6 kernels are used on client/server or vice versa
4706 Details : Introduce kernel-independent values for these flags.
4708 Severity : enhancement
4711 Description: Allow fractional MB tunings for lustre in /proc/ filesystem.
4712 Details : Many of the /proc/ tunables can only be tuned at a megabyte
4713 granularity. Now, Fractional MB granularity is be supported,
4714 this is very useful for low memory system.
4716 Severity : enhancement
4718 Description: Getattr by fid
4719 Details : Getting a file attributes by its fid, obtaining UPDATE|LOOKUP
4720 locks, avoids extra getattr rpc requests to MDS, allows '/' to
4721 have locks and avoids getattr rpc requests for it on every stat.
4724 Frequency : Always, for filesystems larger than 2TB
4726 Description: ldiskfs crash at mount for filesystem larger than 2TB with mballoc
4727 Details : Kenrel kmalloc limits allocations to 128kB and this prevents
4728 filesystems larger than 2TB to be mounted with mballoc enabled.
4731 Frequency : Always, for 32-bit kernel without CONFIG_LBD and filesystem > 2TB
4733 Description: filesystem corruption for non-standard kernels and very large OSTs
4734 Details : If a 32-bit kernel is compiled without CONFIG_LBD enabled and a
4735 filesystems larger than 2TB is mounted then the kernel will
4736 silently corrupt the start of the filesystem. CONFIG_LBD is
4737 enabled for all CFS-supported kernels, but the possibility of
4738 this happening with a modified kernel config exists.
4740 Severity : enhancement
4742 Description: add client O_DIRECT support for 2.6 kernels
4743 Details : It is now possible to do O_DIRECT reads and writes to files
4744 in the Lustre client mountpoint on 2.6 kernel clients.
4746 Severity : enhancement
4748 Description: parallel glimpse, setattr, statfs, punch, destroy requests
4749 Details : Sends glimpse, setattr, statfs, punch, destroy requests to OSTs in
4750 parallel, not waiting for response from every OST before sending
4751 a rpc to the next OST.
4756 Description: setattr vs write race when updating file timestamps
4757 Details : Client processes that update a file timestamp into the past
4758 right after writing to the file (e.g. tar) it is possible that
4759 the updated file modification time can be reset to the current
4760 time due to a race between processing the setattr and write RPC.
4762 Severity : enhancement
4764 Description: Bring 'lfs find' closer in line with regular Linux find.
4765 Details : lfs find util supports -atime, -mtime, -ctime, -maxdepth, -print,
4766 -print0 options and obtains all the needed info through the lustre
4769 Severity : enhancement
4771 Description: support up to 1024 configured devices on one node
4772 Details : change obd_dev array from statically allocated to dynamically
4773 allocated structs as they are first used to reduce memory usage
4778 Description: Flush dirty partially truncated pages during truncate
4779 Details : Immediatelly flush partially truncated pages in filter_setattr,
4780 this way we completely avoid having any pages in page cache on OST
4781 and can retire ugly workarounds during writes to flush such pages.
4786 Description: i_sem vs transaction deadlock in mds_obd_destroy during unlink.
4787 Details : protect inode from truncation within vfs_unlink() context
4788 just take a reference before calling vfs_unlink() and release it
4789 when parent's i_sem is free.
4792 Frequency : always, if extents are used on OSTs
4794 Description: index ei_leaf_hi (48-bit extension) is not zeroed in extent index
4795 Details : OSTs using the extents format would not zero the high 16 bits of
4796 the index physical block number. This is not a problem for any
4797 OST filesystems smaller than 16TB, and no kernels support ext3
4798 filesystems larger than 16TB yet. This is fixed in 1.4.7 (all
4799 new/modified files) and can be fixed for existing filesystems
4800 with e2fsprogs-1.39-cfs1.
4805 Description: import connection selection may be incorrect if timer wraps
4806 Details : Using a 32-bit jiffies timer with HZ=1000 may cause backup
4807 import connections to be ignored if the 32-bit jiffies counter
4808 wraps. Use a 64-bit jiffies counter.
4811 Frequency : during server recovery
4813 Description: crash after server is denying duplicate export
4814 Details : If clients are resending connect requests to the server, the
4815 server refuses to allow a client to connect multiple times.
4816 Fixed a bug in the handling of this case.
4819 Frequency : very large clusters immediately after boot
4821 Description: LNET request buffers exhausted under heavy short-term load
4822 Details : If a large number of client requests are generated on a service
4823 that has previously never seen so many requests it is possible
4824 that the request buffer growth cannot keep up with the spike in
4825 demand. Instead of dropping incoming requests, they are held in
4826 the LND until the RPC service can accept more requests.
4829 Frequency : Sometimes during replay
4831 Description: Assertion failure in ll_local_open after replay.
4832 Details : If replay happened on an open request reply before we were able
4833 to set replay handler, reply will become not swabbed tripping the
4834 assertion in ll_local_open. Now we set the handler right after
4835 recognising of open request
4838 Frequency : very rare
4840 Description: kernel reports "badness in vsnprintf"
4841 Details : Reading from the "recovery_status" /proc file in small chunks
4842 may cause a negative length in lprocfs_obd_rd_recovery_status()
4843 call to vsnprintf() (which is otherwise harmless). Exit early
4844 if there is no more space in the output buffer.
4846 Severity : enhancement
4848 Description: clear OBD RPC statistics by writing to them
4849 Details : It is now possible to clear the OBD RPC statistics by writing
4850 to the "stats" file.
4855 Description: Client mtime is not the same on different clients after utimes
4856 Details : In some cases, the client was using the utimes() syscall on
4857 a file cached on another node. The clients now validate the
4858 ctime from the MDS + OSTs to determine which one is right.
4863 Description: Inability to activate failout mode
4864 Details : lconf script incorrectly assumed that in python string's numeric
4865 value is used in comparisons.
4868 Frequency : always with multiple stripes per file
4870 Description: Inefficient object allocation for mutli-stripe files
4871 Details : When selecting which OSTs to stripe files over, for files with
4872 a stripe count that divides evenly into the number of OSTs,
4873 the MDS is always picking the same starting OST for each file.
4874 Return the OST selection heuristic to the original design.
4879 Description: mount failures may take full timeout to return an error
4880 Details : Under some heavy load conditions it is possible that a
4881 failed mount can wait for the full obd_timeout interval,
4882 possibly several minutes, before reporting an error.
4883 Instead return an error as soon as the status is known.
4885 ------------------------------------------------------------------------------
4887 2006-02-14 Cluster File Systems, Inc. <info@clusterfs.com>
4889 * WIRE PROTOCOL CHANGE. This version of Lustre networking WILL NOT
4890 INTEROPERATE with older versions automatically. Please read the
4891 user documentation before upgrading any part of a live system.
4892 * WARNING: Lustre networking configuration changes are required with
4893 this release. See https://bugzilla.clusterfs.com/show_bug.cgi?id=10052
4896 * Support for kernels:
4897 2.6.9-22.0.2.EL (RHEL 4)
4898 2.6.5-7.244 (SLES 9)
4899 2.6.12.6 vanilla (kernel.org)
4902 Severity : enhancement
4903 Bugzilla : 7981/8208
4904 Description: Introduced Lustre Networking (LNET)
4905 Details : LNET is new networking infrastructure for Lustre, it includes
4906 a reorganized network configuration mode (see the user
4907 documentation for full details) as well as support for routing
4908 between different network fabrics. Lustre Networking Devices
4909 (LNDs) for the supported network fabrics have also been
4910 created for this new infrastructure.
4912 Severity : enhancement
4913 Description: Introduced Access control lists
4914 Details : clients can set ACLs on files and directories in order to have
4915 more fine-grained permissions than the standard Unix UGO+RWX.
4916 The MDS must be started with the "-o acl" mount option.
4918 Severity : enhancement
4919 Description: Introduced filesystem quotas
4920 Details : Administrators may now establish per-user quotas on the
4923 Severity : enhancement
4925 Description: Configuration change for the XT3
4926 The PTLLND is now used to run Lustre over Portals on the XT3
4927 The configure option(s) --with-cray-portals are no longer used.
4928 Rather --with-portals=<path-to-portals-includes> is used to
4929 enable building on the XT3. In addition to enable XT3 specific
4930 features the option --enable-cray-xt3 must be used.
4935 Description: Running on many-way SMP OSTs can trigger oops in llcd_send()
4936 Details : A race between allocating a new llcd and re-getting the llcd_lock
4937 allowed another thread to grab newly-allocated llcd.
4939 Severity : enhancement
4941 Description: 2.6 OST async journal commit and locking fix to improve performance
4942 Details : The filter_direct_io()+filter_commitrw_write() journal commits for
4943 2.6 kernels are now async as they already were in 2.4 kernels so
4944 that they can commit concurrently with the network bulk transfer.
4945 For block-allocated files the filter allocation semaphore is held
4946 to avoid filesystem fragmentation during allocation. BKL lock
4947 removed for 2.6 xattr operations where it is no longer needed.
4952 Description: lconf incorrectly determined whether two IP networks could talk
4953 Details : In some more complicated routing and multiple-network
4954 configurations, lconf will avoid trying to make a network
4955 connection to a disjoint part of the IP space. It was doing the
4956 math incorrectly for one set of cases.
4961 Description: Fix for potential infinite loop processing records in an llog.
4962 Details : If an llog record is corrupted/zeroed, it is possible to loop
4963 forever in llog_process(). Validate the llog record length
4964 and skip the remainder of the block on error.
4967 Frequency : occasional (liblustre only)
4969 Description: liblustre could not open files whose last component is a symlink
4970 Details : sysio_path_walk() would incorrectly pass the open intent to
4971 intermediate path components.
4974 Frequency : rare (liblustre only with non-standard tuning)
4975 Bugzilla : 7201 (7350)
4976 Description: Tuning the MDC DLM LRU size to zero triggers client LASSERT
4977 Details : llu_lookup_finish_locks() tries to set lock data on a lock
4978 after it has been released, only do this for referenced locks
4980 Severity : enhancement
4982 Description: specifying an (invalid) directory default stripe_size of -1
4983 would reset the directory default striping
4984 Details : stripe_size -1 was used internally to signal directory stripe
4985 removal, now use "all default" to signal dir stripe removal
4986 as a directory striping of "all default" is not useful
4989 Frequency : common for large clusters running liblustre clients
4991 Description: doing an ls when liblustre clients are running is slow
4992 Details : sending a glimpse AST to a liblustre client waits for every AST
4993 to time out, as liblustre clients will not respond. Since they
4994 cannot cache data we refresh the OST lock LVB from disk instead.
4996 Severity : enhancement
4998 Description: doing an ls at the same time as file IO can be slow
4999 Details : enqueue and other "small" requests can be blocked behind many
5000 large IO requests. Create a new OST IO portal for non-IO
5001 requests so they can be processed faster.
5004 Frequency : rare (only HPUX clients mounting unsupported re-exported NFS vol)
5006 Description: an HPUX NFS client would get -EACCESS when ftruncate()ing a newly
5007 created file with mode 000
5008 Details : the Linux NFS server relies on an MDS_OPEN_OWNEROVERRIDE hack to
5009 allow an ftruncate() as a non-root user to a file with mode 000.
5010 Lustre now respects this flag to disable mode checks when
5011 truncating a file owned by the user
5014 Frequency : liblustre-only, when liblustre client dies unexpectedly or becomes
5017 Description: Revoking locks from clients that went dead or catatonic might take
5019 Details : New lock flags FL_CANCEL_ON_BLOCK used by liblustre makes
5020 cancellation of such locks instant on servers without waiting for
5021 any reply from clients. Clients drops these locks when cancel
5022 notification from server is received without replying.
5025 Frequency : liblustre-only, when liblustre client dies or becomes busy
5027 Description: Doing ls on Linux clients can take a long time with active
5029 Details : Liblustre client cannot handle ASTs in timely manner, so avoid
5030 granting such locks to it in the first place if possible. Locks
5031 are taken by proxy on the OST during the read or write and
5032 dropped immediately afterward. Add connect flags handling, do
5033 not grant locks to liblustre clients for glimpse ASTs.
5035 Severity : enhancement
5037 Description: Improve read-ahead algorithm to avoid excessive IO for random reads
5038 Details : Existing read-ahead algorithm is tuned for the case of streamlined
5039 sequential reads and behaves badly with applications doing random
5040 reads. Improve it by reading ahead at least read region, and
5041 avoiding excessive large RPC for small reads.
5043 Severity : enhancement
5045 Description: Creating more than 1000 files for a single job may cause a load
5046 imbalance on the OSTs if there are also a large number of OSTs.
5047 Details : qos_prep_create() uses an OST index reseed value that is an
5048 even multiple of the number of available OSTs so that if the
5049 reseed happens in the middle of the object allocation it will
5050 still utilize the OSTs as uniformly as possible.
5055 Description: OST or MDS may oops in ping_evictor_main()
5056 Details : ping_evictor_main() drops obd_dev_lock if deleting a stale export
5057 but doesn't restart at beginning of obd_exports_timed list
5060 Severity : enhancement
5062 Description: improve by-nid export eviction on the MDS and OST
5063 Details : allow multiple exports with the same NID to be evicted at one
5064 time without re-searching the exports list.
5067 Frequency : rare, only with supplementary groups enabled on SMP 2.6 kernels
5069 Description: MDS may oops in groups_free()
5070 Details : in rare race conditions a newly allocated group_info struct is
5071 freed again, and this can be NULL. The 2.4 compatibility code
5072 for groups_free() checked for a NULL pointer, but 2.6 did not.
5075 Frequency : common for liblustre clients doing little filesystem IO
5076 Bugzilla : 9352, 7313
5077 Description: server may evict liblustre clients accessing contended locks
5078 Details : if a client is granted a lock or receives a completion AST
5079 with a blocking AST already set it would not reply to the AST
5080 for LDLM_FL_CANCEL_ON_BLOCK locks. It now replies to such ASTs.
5083 Frequency : lfs setstripe, only systems with more than 160 OSTs
5085 Description: unable to set striping with a starting offset beyond OST 160
5086 Details : llapi_create_file() incorrectly limited the starting stripe
5087 index to the maximum single-file stripe count.
5090 Frequency : LDAP users only
5092 Description: lconf did not handle in-kernel recovery with LDAP properly
5093 Details : lconf/LustreDB get_refs() is searching the wrong namespace
5095 Severity : enhancement
5097 Description: bind OST threads to NUMA nodes to improve performance
5098 Details : all OST threads are uniformly bound to CPUs on a single NUMA
5099 node and do their allocations there to localize memory access
5101 Severity : enhancement
5103 Description: llmount can determine client NID directly from Myrinet (GM)
5104 Details : the client NID code from gmnalnid was moved directly into
5105 llmount, removing the need to use this or specifying the
5106 client NID explicitly when mounting GM clients with zeroconf
5109 Frequency : if client is started with down MDS
5111 Description: if client is started with down MDS mount hangs in ptlrpc_queue_wait
5112 Details : Having an LWI_INTR() wait event (interruptible, but no timeout)
5113 will wait indefinitely in ptlrpc_queue_wait->l_wait_event() after
5114 ptlrpc_import_delayed_req() because we didn't check if the
5115 request was interrupted, and we also didn't break out of the
5116 event loop if there was no timeout
5121 Description: data loss during non-page-aligned writes to a single file from
5122 both multiple nodes and multiple threads on one node at same time
5123 Details : updates to KMS and lsm weren't protected by common lock. Resulting
5124 inconsistency led to false short-reads, that were cached and later
5125 used by ->prepare_write() to fill in partially written page,
5126 leading to data loss.
5129 Frequency : always, if lconf --abort_recovery used
5131 Description: lconf --abort_recovery fails with 'Operation not supported'
5132 Details : lconf was attempting to abort recovery on the MDT device and not
5135 Severity : enhancement
5137 Description: remove cleanup logs
5138 Details : replace lconf-generated cleanup logs with lustre internal
5139 cleanup routines. Eliminates the need for client-cleanup and
5142 Severity : enhancement
5144 Description: add support for EAs (user and system) on lustre filesystems
5145 Details : it is now possible to store extended attributes in the Lustre
5146 client filesystem, and with the user_xattr mount option it
5147 is possible to allow users to store EAs on their files also
5149 Severity : enhancement
5151 Description: Add possibility (config option) to show minimal available OST free
5153 Details : When compiled with --enable-mindf configure option, statfs(2)
5154 (and so, df) will return least minimal free space available from
5155 all OSTs as amount of free space on FS, instead of summary of
5156 free spaces of all OSTs.
5158 Severity : enhancement
5160 Description: do not expand extent locks acquired on OST-side
5161 Details : Modify ldlm_extent_policy() to not expand local locks, acquired
5162 by server: they are not cached anyway.
5165 Frequency : when mmap is used/binaries executed from Lustre
5167 Description: Unmmap pages before throwing them away from read cache.
5168 Details : llap_shrink cache now attempts to unmap pages before discarding
5169 them (if unmapping failed - do not discard). SLES9 kernel has
5170 extra checks that trigger if this unmapping is not done first.
5175 Description: lconf didn't resolve symlinks before checking to see whether a
5176 given mountpoint was already in use
5179 Frequency : when migrating failover services
5180 Bugzilla : 6395, 9514
5181 Description: When migrating a subset of services from a node (e.g. failback
5182 from a failover service node) the remaining services would
5183 time out and evict clients.
5184 Details : lconf --force (implied by --failover) sets the global obd_timeout
5185 to 5 seconds in order to quickly disconnect, but this caused
5186 other RPCs to time out too quickly. Do not change the global
5187 obd_timeout for force cleanup, only set it for DISCONNECT RPCs.
5189 Severity : enhancement
5190 Frequency : if MDS is started with down OST
5191 Bugzilla : 9439,5706
5192 Description: Allow startup/shutdown of an MDS without depending on the
5193 availability of the OSTs.
5194 Details : Asynchronously call mds_lov_synchronize during MDS startup.
5195 Add appropriate locking and lov-osc refcounts for safe
5196 cleaning. Add osc abort_inflight calls in case the
5197 synchronize never started.
5200 Frequency : occasional (Cray XT3 only)
5202 Description: root not authorized to access files in CRAY_PORTALS environment
5203 Details : The client process capabilities were not honoured on the MDS in
5204 a CRAY_PORTALS/CRAY_XT3 environment. If the file had previously
5205 been accessed by an authorized user then root was able to access
5206 the file on the local client also. The root user capabilities
5207 are now allowed on the MDS, as this environment has secure UID.
5210 Frequency : occasional
5212 Description: ldiskfs "too long searching" message happens too often
5213 Details : A debugging message (otherwise harmless) prints too often on
5214 the OST console. This has been reduced to only happen when
5215 there are fragmentation problems on the filesystem.
5220 Description: Division by zero in statfs when all OSCs are inactive
5221 Details : lov_get_stripecnt() returns zero due to incorrect order of checks,
5222 lov_statfs divides by value returned by lov_get_stripecnt().
5226 Bugzilla : 9489, 3273
5227 Description: First write from each client to each OST was only 4kB in size,
5228 to initialize client writeback cache, which caused sub-optimal
5229 RPCs and poor layout on disk for the first writen file.
5230 Details : Clients now request an initial cache grant at (re)connect time
5231 and so that they can start streaming writes to the cache right
5232 away and always do full-sized RPCs if there is enough data.
5233 If the OST is rebooted the client also re-establishes its grant
5234 so that client cached writes will be honoured under the grant.
5239 Description: Slow ls (and stat(2) syscall) on files residing on IO-loaded OSTs
5240 Details : Now I/O RPCs go to different portal number and (presumably) fast
5241 lock requests (and glimses) and other RPCs get their own service
5242 threads pool that should be able to service those RPCs
5245 Severity : enhancement
5247 Description: Ability to exchange lustre version between client and servers and
5248 issue warnings at client side if client is too old. Also for
5249 liblustre clients there is ability to refuse connection of too old
5251 Details : New 'version' field is added to connect data structure that is
5252 filled with version info. That info is later checked by server and
5256 Frequency : rare, liblustre only.
5257 Bugzilla : 9296, 9581
5258 Description: Two simultaneous writes from liblustre at offset within same page
5259 might proceed at the same time overwriting eachother with stale
5261 Details : I/O lock withing llu_file_prwv was released too early, before data
5262 actually was hitting the wire. Extended lock-holding time until
5263 server acknowledges receiving data.
5266 Frequency : extremely rare. Never observed in practice.
5268 Description: avoid generating lustre_handle cookie of 0.
5269 Details : class_handle_hash() generates handle cookies by incrementing
5270 global counter, and can hit 0 occasionaly (this is unlikely, but
5271 not impossible, because initial value of cookie counter is
5272 selected randonly). Value of 0 is used as a sentinel meaning
5273 "unassigned handle" --- avoid it. Also coalesce two critical
5274 sections in this function into one.
5276 Severity : enhancement
5278 Description: allow liblustre clients to delegate truncate locking to OST
5279 Details : To avoid overhead of locking, liblustre client instructs OST to
5280 take extent lock in ost_punch() on client's behalf. New connection
5281 flag is added to handle backward compatibility.
5283 Severity : enhancement
5284 Bugzilla : 4928, 7341, 9758
5285 Description: allow number of OST service threads to be specified
5286 Details : a module parameter allows the number of OST service threads
5287 to be specified via "options ost ost_num_threads={N}" in the
5288 OSS's /etc/modules.conf or /etc/modprobe.conf.
5292 Bugzilla : 6146, 9635, 9895
5293 Description: servers crash with bad pointer in target_handle_connect()
5294 Details : In rare cases when a client is reconnecting it was possible that
5295 the connection request was the last reference for that export.
5296 We would temporarily drop the export reference and get a new
5297 one, but this may have been the last reference and the export
5298 was just destroyed. Get new reference before dropping old one.
5300 Severity : enhancement
5301 Frequency : if client is started with failover MDS
5303 Description: Allow multiple MDS hostnames in the mount command
5304 Details : Try to read the configuration from all specified MDS
5305 hostnames during a client mount in case the "primary"
5308 Severity : enhancement
5310 Description: Stop sending data to evicted clients as soon as possible.
5311 Details : Check if the client we are about to send or are sending data to
5312 was evicted already. (Check is done every second of waiting,
5313 for which l_wait_event interface was extended to allow checking
5314 of exit condition at specified intervals).
5317 Frequency : rare, normally only when NFS exporting is done from client
5319 Description: 'bad disk LOV MAGIC: 0x00000000' error when chown'ing files
5321 Details : Make mds_get_md() recognise empty md case and set lmm size to 0.
5324 Frequency : always, if srand() is called before liblustre initialization
5326 Description: Liblustre uses system PRNG disturbing its usage by user application
5327 Details : Introduce internal to lustre fast and high-quality PRNG for
5328 lustre usage and make liblustre and some other places in generic
5329 lustre code to use it.
5331 Severity : enhancement
5332 Bugzilla : 9477, 9557, 9870
5333 Description: Verify that the MDS configuration logs are updated when xml is
5334 Details : Check if the .xml configuration logs are newer than the config
5335 logs stored on the MDS and report an error if this is the case.
5336 Request --write-conf, or allow starting with --old_conf.
5338 Severity : enhancement
5340 Description: Handle symlinks in the path when checking if Lustre is mounted.
5341 Details : Resolve intermediate symlinks when checking if a client has
5342 mounted a filesystem to avoid duplicate client mounts.
5347 Description: lconf can hit an error exception but still return success.
5348 Details : The lconf command catches the Command error exception at the top
5349 level script context and will exit with the associated exit
5350 status, but doesn't ensure that this exit status is non-zero.
5355 Description: failure of ptlrpc thread startup can cause oops
5356 Details : Starting a ptlrpc service thread can fail if there are a large
5357 number of threads or the server memory is very fragmented.
5358 Handle this without oopsing.
5361 Frequency : always, only if liblustre and non-default acceptor port was used
5363 Description: liblustre cannot connect to servers with non-default acceptor port
5364 Details : tcpnal_set_default_params() was not called and was therefore
5365 ignoring the environment varaible TCPNAL_PORT, as well as other
5366 TCPNAL_ environment variables
5371 Description: two objects could be created on the same OST for a single file
5372 Details : If an OST is down, in some cases it was possible to create two
5373 objects on a single OST for a single file. No problems other
5374 than potential performance impact and spurious error messages.
5378 Bugzilla : 5681, 9562
5379 Description: Client may oops in ll_unhash_aliases
5380 Details : Client dcache may become inconsistent in race condition.
5381 In some cases "getcwd" can fail if the current directory is
5387 Description: Inode refcounting problems in NFS export code
5388 Details : link_raw functions used to call d_instantiate without obtaining
5389 extra inode reference first.
5393 Bugzilla : 9942, 9903
5394 Description: Referencing freed requests leading to crash, memleaks with NFS.
5395 Details : We used to require that call to ll_revalidate_it was always
5396 followed by ll_lookup_it. Also with revalidate_special() it is
5397 possible to call ll_revalidate_it() twice for the same dentry
5398 even if first occurence returned success. This fix changes semantic
5399 between DISP_ENQ_COMPLETE disposition flag to mean there is extra
5400 reference on a request referred from the intent.
5401 ll_intent_release() then releases such a request.
5404 Frequency : rare, normally benchmark loads only
5406 Description: unlinked inodes were kept in memory on the client
5407 Details : If a client is repeatedly creating and unlinking files it
5408 can accumulate a lot of stale inodes in the inode slab cache.
5409 If there is no other client load running this can cause the
5410 client node to run out of memory. Instead flush old inodes
5411 from client cache that have the same inode number as a new inode.
5414 Frequency : SLES9 2.6.5 kernel and long filenames only
5415 Bugzilla : 9969, 10379
5416 Description: utime reports stale NFS file handle
5417 Details : SLES9 uses out-of-dentry names in some cases, which confused
5418 the lustre dentry revalidation. Change it to always use the
5422 Frequency : rare, unless heavy write-truncate concurrency is continuous
5423 Bugzilla : 4180, 6984, 7171, 9963, 9331
5424 Description: OST becomes very slow and/or deadlocked during object unlink
5425 Details : filter_destroy() was holding onto the parent directory lock
5426 while truncating+unlinking objects. For very large objects this
5427 may block other threads for a long time and slow overall OST
5428 responsiveness. It may also be possible to get a lock ordering
5429 deadlock in this case, or run out of journal credits because of
5430 the combined truncate+unlink. Solution is to do object truncate
5431 first in one transaction without parent lock, and then do the
5432 final unlink in a new transaction with the parent lock. This
5433 reduces the lock hold time dramatically.
5436 Frequency : rare, 2.4 kernels only
5438 Description: MDS or OST cleanup may trip kernel BUG when dropping kernel lock
5439 Details : mds_cleanup() and filter_cleanup() need to drop the kernel lock
5440 before unmounting their filesystem in order to avoid deadlock.
5441 The kernel_locked() function in 2.4 kernels only checks whether
5442 the kernel lock is held, not whether it is this process that is
5443 holding it as 2.6 kernels do.
5448 Description: MDS or OST may oops/LBUG if a client is connecting multiple times
5449 Details : The client ptlrpc code may be trying to reconnect to a down
5450 server before a previous connection attempt has timed out.
5451 Increase the reconnect interval to be longer than the connection
5452 timeout interval to avoid sending duplicate connections to
5456 Frequency : echo_client brw_test command
5458 Description: fix echo_client to work with OST preallocated code
5459 Details : OST preallocation code (5137) didn't take echo_client IO path
5460 into account: echo_client calls filter methods outside of any
5461 OST thread and, hence, there is no per-thread preallocated
5462 pages and buffers to use. Solution: hijack pga pages for IO. As
5463 a byproduct, this avoids unnecessary data copying.
5467 Bugzilla : 3555, 5962, 6025, 6155, 6296, 9574
5468 Description: Client can oops in mdc_commit_close() after open replay
5469 Details : It was possible for the MDS to return an open request with no
5470 transaction number in mds_finish_transno() if the client was
5471 evicted, but without actually returning an error. Clients
5472 would later try to replay that open and may trip an assertion
5473 Simplify the client close codepath, and always return an error
5474 from the MDS in case the open is not successful.
5477 Frequency : rare, 2.6 OSTs only
5479 Description: OST may deadlock under high load on fragmented files
5480 Details : If there was a heavy load and highly-fragmented OST filesystems
5481 it was possible to have all the OST threads deadlock waiting on
5482 allocation of biovecs, because the biovecs were not released
5483 until the entire RPC IO was completed. Instead, release biovecs
5484 as soon as they are complete to ensure forward IO progress.
5486 Severity : enhancement
5488 Description: Support for specifying external journal device at mount
5489 Details : If an OST or MDS device is formatted with an external journal
5490 device, this device major/minor is stored in the ext3 superblock
5491 and may not be valid for failover. Allow detecting and
5492 specifying the external journal at mount time.
5497 Description: Mounting an MDS with pending unlinked files may cause oops
5498 Details : target_finish_recovery() calls mds_postrecov() which returned
5499 the number of orphans unlinked. mds_lov_connect->mds_postsetup()
5500 considers this an error and immediately begins cleaning up the
5501 lov, just after starting the mds_lov process
5503 Severity : enhancement
5505 Description: Implement 'lfs df' to report actual free space on per-OST basis
5506 Details : Add sub-command 'df' on 'lfs' to report the disk space usage of
5507 MDS/OSDs. Usage: lfs df [-i][-h]. Command Options: '-i' to report
5508 usage of objects; '-h' to report in human readable format.
5510 ------------------------------------------------------------------------------
5512 2005-08-26 Cluster File Systems, Inc. <info@clusterfs.com>
5519 Description: Mounting an ldiskfs file system with mballoc may crash OST node.
5520 Details : ldiskfs mballoc code may reference an uninitialized buddy struct
5521 at startup during orphan unlinking. Instead, skip buddy update
5522 before setup, as it will be regenerated after recovery is complete.
5527 Description: If an OST is inactive, its locks might reference stale inodes.
5528 Details : lov_change_cbdata() must iterate over all namespaces, even if
5529 they are inactive to clear inode references from the lock.
5531 Severity : enhancement
5532 Frequency : occasional, if non-standard max_dirty_mb used
5534 Description: Client will block write RPCs if not enough grant
5535 Details : If a client has max_dirty_mb smaller than max_rpcs_in_flight,
5536 then the client will block writes while waiting for another RPC
5537 to complete instead of consuming its dirty limit. With change
5538 we get improved performance when max_dirty_mb is small.
5540 Severity : enhancement
5541 Bugzilla : 3389, 6253
5542 Description: Add support for supplementary groups on the MDS.
5543 Details : The MDS has an upcall /proc/fs/lustre/mds/{mds}/group_upcall
5544 (set to /usr/sbin/l_getgroups if enabled) which will do MDS-side
5545 lookups for user supplementary groups into a cache.
5549 Description: O_CREAT|O_EXCL open flags in liblustre always return -EEXIST
5550 Details : Make libsysio to not enforce O_EXCL by clearing the flag,
5551 for liblustre O_EXCL is enforced by MDS.
5555 Description: readdir never returns NULL in liblustre.
5556 Details : Corrected llu_iop_getdirentries logic, to return offset of next
5557 dentry in struct dirent.
5561 Frequency : liblustre only, depends on application IO pattern
5562 Description: liblustre clients evicted if not contacting servers
5563 Details : Don't put liblustre clients into the ping_evictor list, so
5564 they will not be evicted by the pinger ever.
5566 Severity : enhancement
5568 Description: Add ability to evict clients by NID from MDS.
5569 Details : By echoing "nid:$NID" string into
5570 /proc/fs/lustre/mds/.../evict_client client with nid that equals to
5571 $NID would be instantly evicted from this MDS and from all active
5572 OSTs connected to it.
5576 Description: Do not query file size twice, somewhat slowing stat(2) calls.
5577 Details : lookup_it_finish() used to query file size from OSTs that was not
5582 Description: service threads change working directory to that of init
5583 Details : Starting lustre service threads may pin the working directory
5584 of the parent thread, making that filesystem busy. Threads
5585 now change to the working directory of init to avoid this.
5589 Frequency : during shutdown only
5590 Description: shutdown with a failed MDS or OST can cause unmount to hang
5591 Details : Don't resend DISCONNECT messages in ptlrpc_disconnect_import()
5596 Frequency : 2.6 only
5597 Description: chmod/chown may include an extra supplementary group
5598 Details : ll{,u}_mdc_pack_op_data() does not properly initialize the
5599 supplementary group and if none is specified this is used.
5602 Bugzilla : 5479 (6816)
5604 Description: Racing open + rm can assert client in mdc_set_open_replay_data()
5605 Details : If lookup is in progress on a file that is unlinked we might try
5606 to revalidate the inode and fail in revalidate after lookup is
5607 complete and ll_file_open() enqueues the open again but
5608 it_open_error() was not checking DISP_OPEN_OPEN errors correctly.
5611 Frequency : always, if lconf --abort_recovery used
5613 Description: lconf --abort_recovery fails with 'Operation not supported'
5614 Details : lconf was attempting to abort recovery on the MDT device and not
5617 ------------------------------------------------------------------------------
5619 2005-08-08 Cluster File Systems, Inc. <info@clusterfs.com>
5624 Frequency : rare (only unsupported configurations with a node running as an
5626 Bugzilla : 6514, 5137
5627 Description: Mounting a Lustre file system on a node running as an OST could
5629 Details : OSTs now preallocates memory needed to write out data at
5630 startup, instead of when needed, to avoid having to
5631 allocate memory in possibly low memory situations.
5632 Specifically, if the file system is mounted on on OST,
5633 memory pressure could force it to try to write out data,
5634 which it needed to allocate memory to do. Due to the low
5635 memory, it would be unable to do so and the node would
5636 become unresponsive.
5638 Severity : enhancement
5640 Description: Addition of lconf --service command line option
5641 Details : lconf now accepts a '--service <arg>' option, which is
5642 shorthand for 'lconf --group <arg> --select <arg>=<hostname>'
5644 Severity : enhancement
5646 Description: Failover mode is now the default for OSTs.
5647 Details : By default, OSTs will now run in failover mode. To return to
5648 the old behaviour, add '--failout' to the lmc line for OSTs.
5650 Severity : enhancement
5652 Description: Health checks are now provided for MDS and OSTs
5653 Details : Additional detailed health check information on MSD and OSTs
5654 is now provided through the procfs health_check value.
5657 Frequency : occasional, depends on IO load
5659 Description: Disk fragmentation on the OSTs could eventually cause slowdowns
5660 after numerous create/delete cycles
5661 Details : The ext3 inode allocation policy would not allocate new inodes
5662 very well on the OSTs because there are no new directories
5663 being created. Instead we look for groups with free space if
5664 the parent directories are nearly full.
5669 Description: Network or server problems during mount may cause partially
5670 mounted clients instead of returning an error.
5671 Details : The config llog parsing code may overwrite the error return
5672 code during mount error handling, returning success instead
5678 Description: MDS can fail to allocate large reply buffers
5679 Details : After long uptimes the MDS can fail to allocate large reply
5680 buffers (e.g. zconf client mount config records) due to memory
5681 fragmentation or consumption by the buffer cache. Preallocate
5682 some large reply buffers so that these replies can be sent even
5683 under memory pressure.
5687 Frequency : rare (liblustre)
5688 Description: fsx running with liblustre complained that using truncate() to
5689 extend the file doesn't work. This patch corrects that issue.
5690 Details : This is the liblustre equivalent of the fix for bug 6196. Fixes
5691 ATTR_SIZE and lsm use in llu_setattr_raw.
5695 Frequency : rare, only 2.6 kernels
5696 Description: Unusual file access patterns on the MDS may result in inode
5697 data being lost in very rare circumstances.
5698 Details : Bad interaction between the ea-in-inode patch and the "no-read"
5699 code in the 2.6 kernel caused the inode and/or EA data not to
5700 be read from disk, causing single-file corruption.
5704 Frequency : rare, only 2.6 filesystems using extents
5705 Description: Heavy concurrent write and delete load may cause data corruption.
5706 Details : It was possible under high-load situations to have an extent
5707 metadata block in the block device cache from a just-unlinked
5708 file overwrite a newly-allocated data block. We now unmap any
5709 metadata buffers that alias just-allocated data blocks.
5713 Frequency : filesystems with default stripe_count larger than 77
5714 Description: lconf+mke2fs fail when formatting filesystem with > 77 stripes
5715 Details : lconf specifies an inode size of 4096 bytes when the default
5716 stripe_count is larger than 77. This conflicts with the default
5717 inode density of 1 per 4096 bytes. Allocate smaller inodes in
5718 this case to avoid pinning too much memory for large EAs.
5720 ------------------------------------------------------------------------------
5722 2005-07-07 Cluster File Systems, Inc. <info@clusterfs.com>
5727 Frequency : rare (extremely heavy IO load with hundreds of clients)
5729 Description: Client is evicted, gets IO error writing to file
5730 Details : lock ordering changes for bug 5492 reintroduced bug 3267 and
5731 caused clients to be evicted for AST timeouts. The fixes in
5732 bug 5192 mean we no longer need to have such short AST timeouts
5733 so ldlm_timeout has been increased.
5736 Frequency : occasional during --force or --failover shutdown under load
5737 Bugzilla : 5949, 4834
5738 Description: Server oops/LBUG if stopped with --force or --failover under load
5739 Details : a collection of import/export refcount and cleanup ordering
5740 issues fixed for safer force cleanup
5743 Frequency : only filesystems larger than 120 OSTs
5744 Bugzilla : 5990, 6223
5745 Description: lfs getstripe would oops on a very large filesystem
5746 Details : lov_getconfig used kfree on vmalloc'd memory
5749 Frequency : only filesystems exporting via NFS to Solaris 10 clients
5750 Bugzilla : 6242, 6243
5751 Description: reading from files that had been truncated to a non-zero size
5752 but never opened returned no data
5753 Details : ll_file_read() reads zeros from no-object files to EOF
5758 Description: A bug in MDS/OSS recovery could cause the OSS to fail an assertion
5759 Details : There's little harm in aborting MDS/OSS recovery and letting it
5760 try again, so I removed the LASSERT and return an error instead.
5762 Severity : enhancement
5764 Description: New debugging infrastructure for tracking down data corruption
5765 Details : The I/O checksum code was replaced to: (a) control it at runtime,
5766 (b) cover more of the client-side code path, and (c) try to narrow
5767 down where problems occurred
5771 Bugzilla : 3819, 4364, 4397, 6313
5772 Description: Racing close and eviction MDS could cause assertion in mds_close
5773 Details : It was possible to get multiple mfd references during close and
5774 client eviction, leading to one thread referencing a freed mfd.
5776 Severity: : enhancement
5777 Bugzilla : 3262, 6359
5778 Description: Attempts to reconnect to servers are now more aggressive.
5779 Details : This builds on the enhanced upcall-less recovery that was added
5780 in 1.4.2. When trying to reconnect to servers, clients will
5781 now try each server in the failover group every 10 seconds. By
5782 default, clients would previously try one server every 25 seconds.
5787 Description: After recovery, certain operations trigger a failed
5788 assertion on a client.
5789 Details : Failing over an mds, using lconf -d --failover, while a
5790 client was doing a readdir() call would cause the client to
5791 LBUG after recovery completed and the readdir() was resent.
5793 Severity : enhancement
5795 Description: Default groups are now added by lconf
5796 Details : You can now run lconf --group <servicename> without having to
5797 manually add groups with lmc.
5800 Frequency : occasional
5802 Description: Nodes with an elan id of 0 trigger a failed assertion
5805 Frequency : always when accessing e.g. tty/console device nodes
5807 Description: tty and some other devices nodes cannot be used on lustre
5808 Details : file's private_data field is used by device data and lustre
5809 values in there got lost. New field was added to struct file to
5810 store fs-specific private data.
5813 Frequency : when exporting Lustre via NFS
5815 Description: NFSD failed occasionally when looking up a path component
5816 Details : NFSD is looking up ".." which was broken in ext3 directories
5817 that had grown large enough to become hashed.
5820 Frequency : Clusters with multiple interfaces not on the same subnet
5822 Description: Nodes will repeatedly try to reconnect to an interface which it
5823 cannot reach and report an error to the log.
5824 Details : Extra peer list entries will be created by lconf with some peers
5825 unreachable. lconf now validates the peer before adding it.
5828 Frequency : Only if a default stripe is set on the filesystem root.
5830 Description: Setting a default stripe on the filesystem root prevented the
5831 filesystem from being remounted.
5832 Details : The client was sending extra request flags in the root getattr
5833 request and did not allocate a reply buffer for the dir EA.
5836 Frequency : occasional, higher if lots of files are accessed by one client
5837 Bugzilla : 6159, 6097
5838 Description: Client trips assertion regarding lsm mismatch/magic
5839 Details : While revalidating inodes the VFS looks up inodes with ifind()
5840 and in rare cases can find an inode that is being freed.
5841 The ll_test_inode() code will free the lsm during ifind()
5842 when it finds an existing inode and then the VFS later attaches
5843 this free lsm to a new inode.
5847 Bugzilla : 6422, 7030
5848 Description: MDS deadlock between mkdir and client eviction
5849 Details : Creating a new file via mkdir or mknod (starting a transaction
5850 and getting the ns lock) can deadlock with client eviction
5851 (gets ns lock and trying to finish a synchronous transaction).
5854 Frequency : occasional
5855 Description: While starting a server, the fsfilt_ext3 module could not be
5857 Details : CFS's improved ext3 filesystem is named ldiskfs for 2.6
5858 kernels. Previously, lconf would still use the ext3 name
5859 when trying to load modules. Now, it will correctly use
5860 ext3 on 2.4 and ldiskfs on 2.6.
5862 Severity : enhancement
5863 Description: The default stripe count has been changed to 1
5864 Details : The interpretation of the default stripe count (0, to lfs
5865 or lmc) has been changed to mean striping across a single
5866 OST, rather than all available. For general usage we have
5867 found a stripe count of 1 or 2 works best.
5869 Severity : enhancement
5870 Description: Add support for compiling against Cray portals.
5871 Details : Conditional compiling for some areas that are different
5875 Frequency : occasional
5876 Bugzilla : 6409, 6834
5877 Description: Creating files with an explicit stripe count may lead to
5878 a failed assertion on the MDS
5879 Details : If some OSTs are full or unavailable, creating files may
5880 trigger a failed assertion on the MDS. Now, Lustre will
5881 try to use other servers or return an error to the
5885 Frequency : occasional
5887 Description: Multiple concurrent overlapping read+write on multiple SMP nodes
5888 caused lock timeout during readahead (since 1.4.2).
5889 Details : Processes doing readahead might match a lock that hasn't been
5890 granted yet if there are overlapping and conflicting lock
5891 requests. The readahead process waits on ungranted lock
5892 (original lock is CBPENDING), while OST waits for that process
5893 to cancel CBPENDING read lock and eventually evicts client.
5895 Severity : enhancement
5897 Description: Initial enabling of flock support for clients
5898 Details : Implements fcntl advisory locking and file status functions.
5899 This feature is provided as an optional mount flag (default
5900 off), and is NOT CURRENTLY SUPPORTED. Not all types of record
5901 locking are implemented yet, and those that are are not guaranteed
5902 to be completely correct in production environments.
5903 mount -t lustre -o [flock|noflock] ...
5906 Frequency : occasional
5908 Description: OSTs running 2.4 kernels but with extents enabled might trip an
5909 assertion in the ext3 JBD (journaling) layer.
5910 Details : The b_committed_data struct is protected by the big kernel lock
5911 in 2.4 kernels, serializing journal_commit_transaction() and
5912 ext3_get_block_handle->ext3_new_block->find_next_usable_block()
5913 access to this struct. In 2.6 kernels there is finer grained
5914 locking to improve SMP performance of the JBD layer.
5918 Description: Changes the "SCSI I/O Stats" kernel patch to default to "enabled"
5920 -----------------------------------------------------------------------------
5922 2005-05-05 Cluster File Systems, Inc. <info@clusterfs.com>
5924 NOTE: Lustre 1.4.2 uses an incompatible network protocol than previous
5925 versions of Lustre. Please update all servers and clients to
5926 version 1.4.2 or later at the same time. You must also run
5927 "lconf --write-conf {config}.xml" on the MDS while it is stopped
5928 to update the configuration logs.
5930 - fix for HPUX NFS client breakage when NFS exporting Lustre (5781)
5931 - mdc_enqueue does not need max_mds_easize request buffer on send (5707)
5932 - swab llog records of type '0' so we get proper header size/idx (5861)
5933 - send llog cancel req to DLM cancel portal instead of cb portal (5515)
5934 - fix rename of one directory over another leaking an inode (5953)
5935 - avoid SetPageDirty on 2.6 (5981)
5936 - don't re-add just-being-destroyed locks to the waiting list (5653)
5937 - when creating new directories, inherit the parent's custom
5938 striping settings if present parent (3048)
5939 - flush buffers from cache before direct IO in 2.6 obdfilter (4982)
5940 - don't hold i_size_sem in ll_nopage() and ll_ap_refresh_count (6077)
5941 - don't hold client locks on temporary worklist from l_lru (5666)
5942 - handle IO errors in 2.6 obdfilter bio completion routine (6046)
5943 - automatically evict dead clients (5921)
5944 - Update file size properly in create+truncate+fstat case (6196)
5945 - Do not unhash mountpoint dentries, do not allow removal of
5947 - Avoid lock ordering deadlock issue with write/truncate (6203,5654)
5948 - reserve enough journal credits in fsfilt_start_log for setattr (4554)
5949 - ldlm_enqueue freed-export error path would always LBUG (6149,6184)
5950 - don't reference lr_lvb_data until after we hold lr_lvb_sem (6170)
5951 - don't overwrite last_rcvd if there is a *_client_add() error (6086)
5952 - Correctly handle reads of files with no objects (6243)
5953 - lctl recover will also mark a device active if deactivate used (5933)
5955 - by default create 1 inode per 4kB space on MDS, per 16kB on OSTs
5956 - allow --write-conf on an MDS with different nettype than client (5619)
5957 - don't write config llogs to MDS for mounts not from that MDS (5617)
5958 - lconf should create multiple TCP connections from a client (5201)
5959 - init scripts are now turned off by default; run chkconfig --on
5960 lustre and chkconfig --on lustrefs to use them
5961 - upcalls are no longer needed for clients to recover to failover
5963 - add --abort-recovery option to lconf to abort recovery on device
5965 - add support for an arbitrary number of OSTs (3026)
5966 - Quota support protocol changes.
5967 - forward compatibility changes to wire structs (6007)
5968 - rmmod NALs that might be loaded because of /etc/modules.conf (6133)
5969 - support for mountfsoptions and clientoptions to the Lustre LDAP (5873)
5970 - improved "lustre status" script
5971 - initialize blocksize for non-regular files (6062)
5972 - added --disable-server and --disable-client configure options (5782)
5973 - introduce a lookup cache for lconf to avoid repeated DB scans (6204)
5974 - Vanilla 2.4.29 support
5975 - increase maximum number of obd devices to 520 (6242)
5976 - remove the tcp-zero-copy patch from the suse-2.4 series (5902)
5977 - Quadrics Elan drivers are now included for the RHEL 3 2.4.21 and
5978 SLES 9 2.6.5 kernels
5979 - limit stripes per file to 160 (the maximum EA size) (6093)
5981 2005-03-22 Cluster File Systems, Inc. <info@clusterfs.com>
5984 - don't LASSERT in ll_release on NULL lld with NFS export (4655, 5760)
5985 - hold NS lock when calling handle_ast_error->del_waiting_lock (5746)
5986 - fix setattr mtime regression from lovcleanup merge (4829, 5669)
5987 - workaround for 2.6 crash in ll_unhash_aliases (5687, 5210)
5988 - small ext3 extents cleanups and fixes (5733)
5989 - improved mballoc code, several small races and bugs fixed (5733, 5638)
5990 - kernel version 43 - fix remove_suid bugs in both 2.4 and 2.6 (5695)
5991 - avoid needless client->OST connect, fix handle mismatch (5317)
5992 - fix DLM error path that led to out-of-sync client, long delays (5779)
5993 - support common vfs-enforced mount options (nodev,nosuid,noexec) (5637)
5994 - fix several locking issues related to i_size (5492,5624,5654,5672)
5995 - don't move pending lock onto export if it is already evicted (5683)
5996 - fix kernel oops when creating .foo in unlinked directory (5548)
5997 - fix deadlock in obdfilter statistics vs. object create (5811)
5998 - use time_{before,after} to avoid timer jiffies wrap (5882)
5999 - shutdown --force/--failover stability (3607,3651,4797,5203,4834)
6000 - Do not leak request if server was not able to process it (5154)
6001 - If mds_open unable to find parent dir, make that negative lookup(5154)
6002 - don't create new directories with extent-mapping (5909, 5936)
6004 - fix lustre/lustrefs init scripts for SuSE (patch from Scali, 5702)
6005 - don't hold the pinger_sem in ptlrpc_pinger_sending_on_import
6006 - change obd_increase_kms to obd_adjust_kms (up or down) (5654)
6007 - lconf, lmc search both /usr/lib and /usr/lib64 for Python libs (5800)
6008 - support for RHEL4 kernel on i686 (5773)
6009 - provide error messages when incompatible logs are encountered (5898)
6011 2005-02-18 Cluster File Systems, Inc. <info@clusterfs.com>
6012 * version 1.4.0.10 (1.4.1 release candidate 1)
6014 - don't keep a lock reference when lock is not granted (4238)
6015 - unsafe list practices (rarely) led to infinite eviction loop (4908)
6016 - add per-fs limit of Lustre pages in page cache, avoid OOM (4699)
6017 - drop import inflight refcount on signal_completed_replay error (5255)
6018 - unlock page after async write error during send (3677)
6019 - handle missing objects in filter_preprw_read properly (5265)
6020 - no transno return for symlink open, don't save no-trasno open (3440)
6021 - don't try to complete elan receive that already failed (4012)
6022 - free RPC server reply state on error (5406)
6023 - clean up thread from ptlrpc_start_thread() on error (5160)
6024 - readahead could read extra page into cache that wasn't ejected (5388)
6025 - prevent races in class_attach/setup/cleanup/detach (5260)
6026 - don't dereference de->d_inode after l_dput of de (5458)
6027 - use "int" for stripe value returned from lock_to_stripe (5544)
6028 - mballoc allocation and error-checking fixes in 2.6 (5504)
6029 - block device patches to fix I/O request sizes in 2.6 (5482)
6030 - look up hostnames for IB nals (5602)
6031 - 2.6 changed lock ordering of 2 semaphores, caused deadlock (5654)
6032 - don't start multiple acceptors for the same port (5277)
6033 - fix incorrect LASSERT in mds_getattr_name (5635)
6034 - export a proc file for general "ping" checking (5628)
6035 - fix "lfs check" to not block when the MDS is down (5628)
6037 - service request history (4965)
6038 - put {ll,lov,osc}_async_page structs in a single slab (4699)
6039 - create an "evict_client" /proc entry on OSTs, like the MDS has
6040 - fix mount usage message, return errors per mount(8) (5168)
6041 - change grep [] to grep "[]" in tests so they work in more UMLs
6042 - fix ppc64/x86_64 spec to use %{_libdir} instead of /usr/lib (5389)
6043 - remove ancient LOV_MAGIC_V0 EA support (5047)
6044 - add "disk I/Os in flight" and "I/O req time" stats in obdfilter
6045 - align r/w RPCs to PTLRPC_MAX_BRW_SIZE boundary for performance (3451)
6046 - allow readahead allocations to fail when low on memory (5383)
6047 - mmap locking landed again, after considerable improvement (2828)
6048 - add get_hostaddr() to lustreDB.py for LDAP support (5459)
6050 2004-11-23 Cluster File Systems, Inc. <info@clusterfs.com>
6053 - send OST transaction number in read/write reply to free req (4966)
6054 - don't ASSERT in ptl_send_rpc() if we run out of memory (5119)
6055 - lock /proc/sys/portals/routes internal state, avoiding oops (4827)
6056 - the watchdog thread now runs as interruptible (5246)
6057 - flock/lockf fixes (but it's still disabled, pending 5135)
6058 - don't use EXT3 constants in llite code (5094)
6059 - memory shortage at startup could cause assertion (5176)
6061 - reorganization of lov code
6062 - single portals codebase
6064 - add extents/mballoc support (5025)
6065 - direct I/O reads in the obdfilter (4048)
6066 - kernel patches from LNXI for 2.6 (bluesmoke, perfctr, mtd, kexec)
6068 tbd Cluster File Systems, Inc. <info@clusterfs.com>
6071 - send OST transaction number in read/write reply to free req (4966)
6072 - don't ASSERT in ptl_send_rpc() if we run out of memory (5119)
6073 - lock /proc/sys/portals/routes internal state, avoiding oops (4827)
6074 - the watchdog thread now runs as interruptible (5246)
6075 - handle missing objects in filter_preprw_read properly (5265)
6076 - unsafe list practices (rarely) led to infinite eviction loop (4908)
6077 - drop import inflight refcount on signal_completed_replay error (5255)
6078 - unlock page after async write error during send (3677)
6079 - return original error code on reconstructed replies (3761)
6080 - no transno return for symlink open, don't save no-trasno open (3440)
6082 - add pid to ldlm debugging output (4922)
6083 - bump the watchdog timeouts -- we can't handle 30sec yet
6084 - extra debugging for orphan dentry/inode bug (5259)
6086 2004-11-16 Cluster File Systems, Inc. <info@clusterfs.com>
6089 - fix TCP_NODELAY bug, which caused extreme perf regression (5134)
6090 - allocate qswnal tx descriptors singly to avoid fragmentation (4504)
6091 - don't LBUG on obdo_alloc() failure, use OBD_SLAB_ALLOC() (4800)
6092 - fix NULL dereference in /proc/sys/portals/routes (4827)
6093 - allow failed mdc_close() operations to be interrupted (4561)
6094 - stop precreate on OST before MDS would time out on it (4778)
6095 - don't send partial-page writes before EOF from client (4410)
6096 - discard client grant for sub-page writes on large-page clients (4520)
6097 - don't free dentries not owned by NFS code, check generation (4806)
6098 - fix lsm leak if mds_create_objects() fails (4801)
6099 - limit debug_daemon file size, always print CERROR messages (4789)
6100 - use transno after validating reply (3892)
6101 - process timed out requests if import state changes (3754)
6102 - update mtime on OST during writes, return in glimpse (4829)
6103 - add mkfsoptions to LDAP (4679)
6104 - use ->max_readahead method instead of zapping global ra (5039)
6105 - don't interrupt __l_wait_event() during strace
6107 - add software watchdogs to catch hung threads quickly (4941)
6108 - make lustrefs init script start after nfs is mounted
6109 - fix CWARN/ERROR duplication (4930)
6110 - return async write errors to application if possible (2248)
6111 - add /proc/sys/portal/memused (bytes allocated by PORTALS_ALLOC)
6112 - print NAL number in %x format (4645)
6113 - update barely-supported suse-2.4.21-171 series (4842)
6114 - support for sles 9 %post scripts
6115 - support for building 2.6 kernel-source packages
6116 - support for sles km_* packages
6118 2004-10-07 Cluster File Systems, Inc. <info@clusterfs.com>
6121 - ignore -ENOENT errors in osc_destroy (3639)
6122 - notify osc create thread that OSC is being cleaned up (4600)
6123 - add nettype argument for llmount in #5d in conf-sanity.sh (3936)
6124 - reconstruct ost_handle() like mds_handle() (4657)
6125 - create a new thread to do import eviction to avoid deadlock (3969)
6126 - let lconf resolve symlinked-to devices (4629)
6127 - don't unlink "objects" from directory with default EA (4554)
6128 - hold socknal file ref over connect in case target is down (4394)
6129 - allow more than 32000 subdirectories in a single directory (3244)
6130 - fix blocks count for O_DIRECT writes (3751)
6131 - OST returns ENOSPC from object create when no space left (4539)
6132 - don't send truncate RPC if file size isn't changing (4410)
6133 - limit OSC precreate to 1/2 of value OST considers bogus (4778)
6134 - bind to privileged port in socknal and tcpnal (3689)
6136 - rate limit CERROR/CWARN console message to avoid overload (4519)
6137 - GETFILEINFO dir ioctl returns LOV EA + MDS stat in 1 call (3327)
6138 - basic mmap support (3918)
6139 - kernel patch series update from b1_4 (4711)
6141 2004-09-16 Cluster File Systems, Inc. <info@clusterfs.com>
6144 - avoid crash during MDS cleanup with OST shut down (2775)
6145 - fix loi_list_lock/oig_lock inversion on interrupted IO (4136)
6146 - don't use bad inodes on the MDS (3744)
6147 - dynamic object preallocation to improve recovery speed (4236)
6148 - don't hold spinlock over lock dumping or change debug flags (4401)
6149 - don't zero obd_dev when it is force cleaned (3651)
6150 - print grants to console if they go negative (4431)
6151 - "lctl deactivate" will stop automatic recovery attempts (3406)
6152 - look for existing locks in ldlm_handle_enqueue() (3764)
6153 - don't resolve lock handle twice in recovery avoiding race (4401)
6154 - revalidate should check working dir is a directory (4134)
6156 - don't always mark "slow" obdfilter messages as errors (4418)
6158 2004-08-24 Cluster File Systems, Inc. <info@clusterfs.com>
6161 - don't close LustreDB during write_conf until it is done (3860)
6162 - fix typo in lconf for_each_profile (3821)
6163 - allow dumping logs from multiple threads at one time (3820)
6164 - don't allow multiple threads in OSC recovery (3812)
6165 - fix debug_size parameters (3864)
6166 - fix mds_postrecov to initialize import for llog ctxt (3121)
6167 - replace config semaphore with spinlock (3306)
6168 - be sure to send a reply for a CANCEL rpc with bad export (3863)
6169 - don't allow enqueue to complete on a destroyed export (3822)
6170 - down write_lock before checking llog header bitmap (3825)
6171 - recover from lock replay timeout (3764)
6172 - up llog sem before sending rpc (3652)
6173 - reduce ns lock hold times when setting kms (3267)
6174 - change a dlm LBUG to LASSERTF, to maybe learn something (4228)
6175 - fix NULL deref and obd_dev leak on setup error (3312)
6176 - replace some LBUG about llog ops with error handling (3841)
6177 - don't match INVALID dentries from d_lookup and spin (3784)
6178 - hold dcache_lock while marking dentries INVALID and hashing (4255)
6179 - fix invalid assertion in ptlrpc_set_wait (3880)
6181 - add libwrap support for the TCP acceptor (3996)
6182 - add /proc/sys/portals/routes for non-root route listing (3994)
6183 - allow setting MDS UUID in .xml (2580)
6184 - print the stack of a process that LBUGs (4228)
6186 2004-07-14 Cluster File Systems, Inc. <info@clusterfs.com>
6189 - don't cleanup request in ll_file_open() on failed MDS open (3430)
6190 - make sure to unset replay flag from failed open requests (3440)
6191 - if default stripe count is 0, use OST count for inode size (3636)
6192 - update parent mtime/ctime on client for create/unlink (2611)
6193 - drop dentry ref in ext3_add_link from open_connect_dentry (3266)
6194 - free recovery state on server during a forced cleanup (3571)
6195 - unregister_reply for resent reqs (3063)
6196 - loop back devices mounting and status check on 2.6 (3563)
6197 - fix resource-creation race that can provoke i_size == 0 (3513)
6198 - don't try to use bad inodes returned from MDS/OST fs lookup (3688)
6199 - more debugging for page-accounting assertion (3746)
6200 - return -ENOENT instead of asserting if ost getattr+unlink race (3558)
6201 - avoid deadlock after precreation failure (3758)
6202 - fix race and lock order deadlock in orphan handling (3450, 3750)
6203 - add validity checks when grabbing inodes from l_ast_data (3599)
6205 - add /proc/.../recovery_status to obdfilter (3428)
6206 - lightweight CDEBUG infrastructure, debug daemon (3668)
6207 - change default OSC RPC parameters to be better on small clusters
6208 - turn off OST read cache for files smaller than 32MB
6209 - install man pages and include them in rpms (3100)
6210 - add new init script for (un)mounting lustre filesystems (2593)
6211 - run chkconfig in %post for init scripts (3701)
6212 - drop scimac NAL (unmaintained)
6214 2004-06-17 Cluster File Systems, Inc. <info@clusterfs.com>
6217 - clean kiobufs before and after use (3485)
6218 - strip trailing '/'s before comparing paths with /proc/mounts (3486)
6219 - remove assertions to work around "in-flight rpcs" recovery bug (3063)
6220 - change init script to fail more clearly if not run as root (1528)
6221 - allow clients to reconnect during replay (1742)
6222 - fix ns_lock/i_sem lock ordering deadlock for kms update (3477)
6223 - don't do DNS lookups on NIDs too small for IP addresses (3442)
6224 - re-awaken ptlrpcd if new requests arrive during check_set (3554)
6225 - fix cond_resched (3554)
6226 - only evict unfinished clients after recovery (3515)
6227 - allow bulk resend, prevent data loss (3570)
6228 - dynamic ptlrpc request buffer allocation (2102)
6229 - don't allow unlinking open directory if it isn't empty (2904)
6230 - set MDS/OST threads to umask 0 to not clobber client modes (3359)
6231 - remove extraneous obd dereference causing LASSERT failure (3334)
6232 - don't use get_cycles() when creating temp. files on the mds (3156)
6233 - hold i_sem when setting i_size in ll_extent_lock() (3564)
6234 - handle EEXIST for set-stripe, set proper directory name (3336)
6236 - servers can dump a log evicting a client - lustre.dump_on_timeout=1
6237 - fix ksocknal_fmb_callback() error messages (2918)
6239 2004-05-27 Cluster File Systems, Inc. <info@clusterfs.com>
6242 - don't copy lvb into (possibly NULL) reply on error (2983)
6243 - don't deref dentry after dput, don't free lvb on error (2922)
6244 - use the kms to determine writeback rpc length (2947)
6245 - increment oti_logcookies when osc is inactive (2948)
6246 - update client's i_blocks count via lvb messages (2543)
6247 - handle intent open/close of special files properly (1557)
6248 - mount MDS with errors=remount-ro, like obdfilter (2009)
6249 - initialize lock handle to avoid ASSERT on error cleanup (3057)
6250 - don't use cancelling-locks' kms values (2947)
6251 - use highest lock extent for kms, not last one (2925)
6252 - don't dereference ERR_PTR() dentry in error handling path (3107)
6253 - fix thread race in portals_debug_dumplog() (3122)
6254 - create lprocfs device entries at setup instead of at attach (1519)
6255 - common AST error handler, don't evict client on completion race (3145)
6256 - zero nameidata in detach_mnt in 2.6 (3118)
6257 - verify d_inode after revalidate_special is valid in 2.6 (3116)
6258 - use lustre_put_super() to handle zconf unmounts in 2.6 (3064)
6259 - initialize RPC timeout timer earlier for 2.6 (3219)
6260 - don't dereference NULL reply buffer if mdc_close was never sent (2410)
6261 - print nal/nid for unknown nid (3258)
6262 - additional checks for oscc recovery before doing precreate (3284)
6263 - fix ll_extent_lock() error return code for 64-bit systems (3043)
6264 - don't crash in mdc_close for bad permissions on open (3285)
6265 - zero i_rdev for non-device files (3147)
6266 - clear page->private before handing to FS, better assertion (3119)
6267 - tune the read pipeline (3236)
6268 - fix incorrect decref of invalidated dentry (2350)
6269 - provide read-ahead stats and refine rpc in flight stats (3328)
6270 - don't hold journal transaction open across create RPC (3313)
6271 - update atime on MDS at close time (3265)
6272 - close LDAP connection when recovering to avoid server load (3315)
6273 - update iopen-2.6 patch with fixes from 2399,2517,2904 (3301)
6274 - don't leak open file on MDS after open resend (3325)
6275 - serialize filter_precreate and filter_destroy_precreated (3329)
6276 - loop device shouldn't call sync_dev() for nul device (3092)
6277 - clear page cache after eviction (2766)
6278 - resynchronize MDS->OST in background (2824)
6279 - refuse to mount the same filesystem twice on same mountpoint (3394)
6280 - allow llmount to create routes for mounting behind routers (3320)
6281 - push lock cancellation to blocking thread for glimpse ASTs (3409)
6282 - don't call osc_set_data_with_check() for TEST_LOCK matches (3159)
6283 - fix rare problem with rename on htree directories (3417)
6285 - allow default OST striping configuration per directory (1414)
6286 - fix compilation for qswnal for 2.6 kernels (3125)
6287 - increase maximum number of MDS request buffers for large systems
6288 - change liblustreapi to be useful for external progs like lfsck (3098)
6289 - increase local configuration timeout for slow disks (3353)
6290 - allow configuring ldlm AST timeout - lustre.ldlm_timeout=<seconds>
6292 2004-03-22 Cluster File Systems, Inc. <info@clusterfs.com>
6295 - fixes for glimpse AST timeouts / incorrectly 0-sized files (2818)
6296 - don't overwrite extent policy data in reply if lock was blocked (2901)
6297 - drop filter export grants atomically with removal from device (2663)
6298 - del obd_self_export from work_list in class_disconnect_exports (2908)
6299 - don't LBUG if MDS recovery times out during orphan cleanup (2530)
6300 - swab reply message in mdc_close, other PPC fixes (2464)
6301 - fix destroying of named logs (2325)
6302 - overwrite old logs when running lconf --write_conf (2264)
6303 - bump LLOG_CHUNKSIZE to 8k to allow for larger clusters (2306)
6304 - fix race in target_handle_connect (2898)
6305 - mds_reint_create() should take same inode create lock (2926)
6306 - correct journal credits calculated for CANCEL_UNLINK_LOG (2931)
6307 - don't close files for self_export to avoid uninitialized obd (2936)
6308 - allow MDS with the same name as client node (2939)
6309 - hold dentry reference for closed log files for unlink (2325)
6310 - reserve space for all logs during transactions (2059)
6311 - don't evict page beyond end of stripe extent (2925)
6312 - don't oops on a deleted current working directory (2399)
6313 - handle hard links to targets without a parent properly (2517)
6314 - don't dereference NULL lock when racing during eviction (2867)
6315 - don't grow lock extents when lots of conflicting locks (2919)
6317 2004-03-04 Cluster File Systems, Inc. <info@clusterfs.com>
6320 - account for cache space usage on clients to avoid data loss (974)
6321 - lfsck support in lustre kernel code (2349)
6322 - reduce journal credits needed for BRW writes (2370)
6323 - orphan handling to avoid losing space on client/server crashes
6324 - ptlrpcd can be blocked, stopping ALL progress (2477)
6325 - use lock value blocks to assist in proper KMS, faster stat (1021)
6326 - takes i_sem instead of DLM locks internally on obdfilter (2720)
6327 - recovery for initial connections (2355)
6328 - fixes for mds_cleanup_orphans (1934)
6329 - abort_recovery crashes MDS in b_eq (mds_unlink_orphan) (2584)
6330 - block all file creations until orphan recovery completes (1901)
6331 - client remove rq_connection from request struct (2423)
6332 - conf-sanity test_5, proper cleanup in umount log not availale (2640)
6333 - recovery timer race (2670)
6334 - mdc_close recovey bug (2532)
6335 - ptlrpc cleanup bug (2710)
6336 - mds timeout on local locks (2588)
6337 - namespace lock held during RPCs (2431)
6338 - handle interrupted sync write properly (2503)
6339 - don't try to handle a message that hasn't been replied to (2699)
6340 - client assert failure during cleanup after abort recovery (2701)
6341 - leak mdc device after failed mount (2712)
6342 - ptlrpc_check_set allows timedout requests to complete (2714)
6343 - wait for inflight reqs when ptlrpcd finishes (2710)
6344 - make sure unregistered services are removed from the srv_list
6345 - reset bulk XID's when resending them (caught by 1138 test)
6346 - unregister_bulk after timeout
6347 - fix lconf error (2694)
6348 - handle write after unfinished setstripe, stripe-only getstripe (2388)
6349 - readahead locks pages, leaves pending causing memory pressure (2673)
6350 - increase OST request buffers to 4096 on large machines (2729)
6351 - fix up permission of existing directories in simple_mkdir (2661)
6352 - init deleted item, add assertions ptlrpc_abort_inflight() (2725)
6353 - don't assign transno to errored transactions (2742)
6354 - don't delete objects on OST if given a bogus objid from MDS (2751)
6355 - handle large client PAGE_SIZE readdir on small PAGE_SIZE MDS (2777)
6356 - if rq_no_resend, then timeout request after recovery (2432)
6357 - fix MDS llog_logid record size, 64-bit array alignment (2733)
6358 - don't call usermode_helper from ptlrpcd, DEFAULT upcall (2773)
6359 - put magic in mount.lustre data, check for bad/NULL mount data (2529)
6360 - MDS recovery shouldn't delete objects that it has given out (2730)
6361 - if enqueue arrives after completion, don't clobber LVB (2819)
6362 - don't unlock pages twice when trigger_group_io returns error (2814)
6363 - don't deref NULL rq_repmsg if ldlm_handle_enqueue failed (2822)
6364 - don't write pages to disk if there was an error (1450)
6365 - don't ping imports that have recovery disabled (2676)
6366 - take buffered bytes into account when balancing socknal conn (2817)
6367 - hold a DLM lock over readdir always, use truncate_inode_pages (2706)
6368 - reconnect unlink llog connection after MDS reconnects to OST (2816)
6369 - remove little-endian swabbing of llog records (1987)
6370 - set/limit i_blksize to LL_MAX_BLKSIZE on client (2884)
6371 - retry reposting request buffers if they fail (1191)
6372 - grow extent at grant time to avoid granting a revoked lock (2809)
6373 - lock revoke doesn't evict page if covered by a second lock (2765)
6374 - disable VM readahead to avoid reading outside lock extents (2805)
6376 - return LL_SUPER_MAGIC from statfs for the filesystem type (1972)
6377 - updated kernel patches for hp-2.4.20 kernel (2681)
6379 2004-02-07 Cluster File Systems, Inc. <info@clusterfs.com>
6382 - fix truncated write corruption (2366)
6383 - fix for failed assertion in iopen_connect_dentry (1792,2517)
6385 - don't flag the ptlrpcd thread with PF_MEMALLOC (2636)
6386 - ensure len(uuid) < 37 in lmc (1171)
6387 - fix ia64 OOPS in llog_test (2255)
6388 - zero end of page at obdfilter for partial page writes (2648)
6389 - don't leave stale dentries around after renames (bug 2428)
6390 - fix timeouts when evicting a client with a single lock held (2642)
6391 - set deadline for the initial HELLO message to drain (2634)
6392 - print out dotted-quad IP addresses in the socknal (2302)
6394 - additional debugging for MDS client eviction problem (2443)
6395 - fix mkfsoptions support for osts (2603, 2604)
6397 2004-01-27 Cluster File Systems, Inc. <info@clusterfs.com>
6400 - add series for the vanilla 2.6.0 kernel
6401 - add series for the vanilla 2.4.24 kernel
6402 - add series for a cray x86/64 UL kernel drop
6403 - fix xattr patches for the vanilla 2.4.19 series
6405 - generate true UUIDs in lmc (1171)
6406 - have portals stack dumping break in UML (2466)
6407 - avoid bad dchild deref; avoid inum lock w/o creation (2362)
6408 - allocate with _NOFS in ldlm to avoid deadlock (1933)
6409 - wake callback waiting threads on client eviction (2460)
6410 - Add --ptldebug and --subsystem to lmc (1719)
6411 - update assertion to allow safe interrupt allocation
6412 - set rq_no_resend for cancel requests (2432)
6413 - recalculate ptlrpcd timeout after resend (2494)
6414 - call vfs_rmdir when removing pending directories (2368)
6415 - fix renaming a file to itself (2429)
6416 - lmc creates a default one-stripe lov (2454)
6417 - expand procfs space to handle large clusters (2326)
6418 - increase UML stack to avoid overflow
6419 - update lconf's list of debug and subsystem masks
6420 - fix lfs find --obd (2510)
6421 - /proc tunable for disabling filter read caching (2591)
6422 - stop rpm packages from altering slapd.conf (2301)
6423 - disable nagle in the socknal under 0conf (2578)
6424 - choose mds inode size based on stripe count (2572)
6425 - fix kernel-source rpm problems (2516)
6427 - add --disable-doc to avoid pdf generation (2421)
6428 - update documentation, tests, type-os, comments
6429 - avoid format warnings on ia64
6430 - remove the TOE NAL
6431 - tiny code cleanups by removing unused fields
6433 2004-01-07 Cluster File Systems, Inc. <info@clusterfs.com>
6436 - fix obvious semaphore misuse in as-yet-unused setattr path (2348)
6437 - remove the most blatant lies from BUILDING file (2371)
6438 - change default debug level to reasonable production setting
6439 - reduce client side cache size to reduce cache flush time
6440 - reduce max RPCs in flight to avoid unnecessary file fragmentation
6441 - make TCP zerocopy and pinger support enabled by default (2476)
6442 - sync writes completed after process exits caused crashes (2319)
6443 - maintain correct mount count on the MDS (2356)
6444 - backout 1557, because 2316 wasn't really fixed
6445 - better file I/O statistics gathering in /proc
6446 - don't take unnecessary, deadlock-inducing bug in readpage (2383)
6447 - another kernel patch to fix zero-copy TCP function export
6448 - don't take duplicate lock when processing re-sent getattr (2420)
6449 - lctl uses obd_self_export instead of creating new conn (2353)
6450 - MDS/OST recovery case which requires object creation asserted (2425)
6451 - move lfs from /usr/sbin to /usr/bin in packages
6452 - fix race between mds_client_add and mds_client_free (2417)
6453 - use kmalloc instead of slabs in portals (2430)
6454 - don't create duplicate records when a failover MDS is present (2442)
6455 - remove unnecessary mount age check (2332)
6456 - don't remove directory inodes from locks prematurely (2451)
6457 - don't break if MDS service name is the same as hostname (2103)
6458 - fix races in client write RPC generation when cache full (2482)
6460 2003-12-13 Cluster File Systems, Inc. <info@clusterfs.com>
6463 - remove now-unused request->rq_obd (278)
6464 - if an allocation fails, print out how much memory we've used (1933)
6465 - use PORTAL_SLAB_ALLOC for structures, to get GFP_MEMALLOC (1933)
6466 - add the "configurable stack size" patch to most series files (1256)
6467 - ability to write large log records, for 100+ OST configs (2306)
6468 - fix NULL deref when filter_prep fails (2314)
6469 - fix operator precedence error in filter_sync
6470 - dynamic allocation of socknal TX descriptors (2315)
6471 - fix a missed case in the GFP_MEMALLOC patch, can cause deadlock (2310)
6472 - fix gcc 2.96 compilation problem in xattr kernel patch (2294)
6473 - ensure that CWARN messages in Portals always get to the syslog
6474 - __init/__exit are not for prototype decls (ldlm_init/exit)
6475 - x86-64 compile warning fixes
6476 - fix gateway LMC keyword conflict (2318)
6477 - fix MDS lock inversions in getattr/reint paths (1844)
6478 - fix a rare lock re-ordering bug, which caused deadlock (2322)
6479 - fix i_sem/journal inversion in fsfilt_ext3_write_record (2306)
6480 - DLM race condition prevented some lock evictions (2328)
6481 - ENOMEM detection and retry on socknal sends (2230)
6482 - use GFP_NOFS throughout Lustre, to combat ENOMEM (2230)
6483 - move osc_rpcd into ptlrpc, for use in MDC and others (2329)
6484 - protect MDS inode fsdata with stronger locking; fixes assertion (2313)
6485 - better error messages when a client is rejected during recovery (1505)
6486 - avoid cancelling locks which were never granted, after failure (2330)
6487 - fix i_sem/journal inversion in mds_client_add (2333)
6488 - fix truncate/getattr lock cycle deadlock (2334)
6489 - use rpcd to send close; allows resend after timeout, avoid leak (1897)
6490 - fix two rare exit paths which could leak an l_lock() ref (2321)
6491 - fencepost error in MDS/OST orphan recovery (2226)
6492 - make log record alignment 8 bytes (1988)
6493 - lstripe now fails when requested offset > ost_count (2237)
6494 - ensure that all kernel series have a complete list.h (1607)
6495 - fix crashes in special-file operations (2316)
6496 - lctl create/brw OID mismatch, caused by obsolete filter loop (2339)
6498 - allow configurable automake binary, for testing new versions
6499 - small update to the lfs documentation
6501 2003-12-03 Cluster File Systems, Inc. <info@clusterfs.com>
6503 * fix negative export reference count in fsfilt_sync (2312)
6505 2003-12-01 Cluster File Systems, Inc. <info@clusterfs.com>
6506 * release candidate 0.9.1
6508 - orphans are moved into the PENDING directory for possible recovery
6509 - replayed opens now open by fid for orphan/rename safety (1042)
6510 - last close of an orphan inode generates a transno (683)
6511 - chdir() and mount() now pin the directory entry (1020)
6512 - avoid CERROR in normal ll_setattr_raw() error case (1500)
6513 - discard very old requests without processing them (1502)
6514 - remove some common, well-understood CERRORs (1505)
6515 - require O_DIRECT I/O to be page-sized to workaround IA64 crash (1609)
6516 - clear "grant" flags in OST replies until OST grant code lands (1644)
6517 - fix read performance by not clobbering i_blksize on client (1598)
6518 - fix __ldlm_handle2lock oops by not dereferencing lock after PUT (1625)
6519 - make LRU size a /proc tunable, clears locks when reduced (707)
6520 - fix some lprocfs rot that prevented ptlbd from loading (1732)
6521 - server locks take references on exports now (1558)
6522 - build fixes for 2.4.20-rh trees (1663)
6523 - return an error from lov_create if all OSCs are inactive (1751)
6524 - fix import levels when a reconnect happens without a timeout (1597)
6525 - exit early from mds_open if we get a lookup error (1749)
6526 - partial page read at EOF wouldn't wait for disk before sending (1642)
6527 - avoid NULL deref in obdfilter when reading page past EOF (1592)
6528 - avoid LASSERT in ll_intent_lock if server failed very early (1090)
6529 - fix LBUG in ll_it_open_error with rc = -2 (1861)
6530 - write/truncate lock inversion (1639)
6531 - Don't auto-load obdclass, portals modules during cleanup (1495)
6532 - fix timestamps from jumping to "now" (1763)
6533 - extra journal assertions (1648)
6534 - add an extra multiunlink test (1771)
6535 - fix read_record/write_record API (1776)
6536 - fix leak of offset_extent, possible incorrect i_size later (1772)
6537 - fix lasserts in mis-matched transnos during open-unlink testing (1541)
6538 - Debugging for the kqswnal_get_idle_tx problems (1820)
6539 - Allow recovery to be attempted multiple times (1536)
6540 - Write out MDS last_rcvd file after it is first created (1600)
6541 - Fix tx_descriptor leak in failed transmit situations (1827)
6542 - ext3 journaling fixes for assertion failure after IO error (1871)
6543 - class_export_put() on freed export after completion AST error (1896)
6544 - Fix revalidate looping in VFS (1322)
6545 - Don't access a freed export during MDS_REINT timeout (1521)
6546 - Add open-unlink recovery support on the MDS (1673,1764)
6547 - Return an error if no MDS data was read from last_rcvd (1946)
6548 - Fix for lookup "." or ".." crash on error (1932,1931,1935)
6549 - Don't setup a disk device that doesn't match exported UUID (317)
6550 - Reduce bulk RPC timeout to avoid cascading client/OST failures (1845)
6551 - avoid committing NULL handle in force close
6552 - local.sh is now a one-stripe LOV configuration
6553 - POSIX utime.4 -EPERM on FIFO not owned by user (56)
6554 - fix ext3 htree duplicate directory entry corruption (1516)
6555 - POSIX creat.13, fstat.1, open.18, stat.3 new file atime/mtime (2020)
6556 - update to new LOV EA format (2097)
6557 - interoperability for different PAGE_SIZE/wordsize (686,1821,1343,2042)
6559 2003-06-15 Phil Schwan <phil@clusterfs.com>
6562 - imports and exports cleanup too early, need refcounts (349, 879, 1045)
6563 - per-import/export recovery handling (958, 931, 959)
6564 - multiple last-rcvd slots, for serving multiple FSes (949)
6565 - connections are again shared between multiple imp/exports (963, 964)
6566 - "umount -f" would hang if any requests needed to be sent (393, 978)
6567 - avoid pinning large req buffer by copying for queued messages (989)
6568 - add "uuid" to "lctl device" command to help upcalls (991)
6569 - "open" RPCs with transnos would confuse recovery counters (1037)
6570 - do proper endian conversion of all wire messages (288, 340, 891)
6571 - remove OST bulk get LBUGs, fix ost_brw_write cleanup (1126)
6572 - call waiting locks callback from LDLM recovery thread (1127, 1151)
6573 - fix ptlrpc_connection leak in target_handle_connect (1174)
6574 - fix import refcounting bug in OST and MDS cleanup (1134)
6575 - if an invalid-at-open-time OSC returned before close(), LBUG (1150)
6576 - fix very unlikely obd_types race condition (501)
6577 - remove osc_open hack for echo_client (1187)
6578 - we leaked exports/dlmimps for forcibly disconnected clients (1143)
6579 - a failure in read_inode2 leads to deadlock (1139)
6580 - cancel ack-locks as soon as transaction is committed (1072)
6581 - fix major leaks and crashes in the bulk I/O path (937, 1057)
6582 - make sure to commitrw after any preprw to avoid deadlock (1162)
6583 - failing to execute a file in a lustre FS would lock inode (1203)
6584 - small DEBUG_REQ fix to avoid dereferencing a NULL (1227)
6585 - don't ASSERT while cleaning up an incompletely-setup obd (1248)
6586 - obd_uuid2tgt would walk off the end of the list (1255)
6587 - on IA64 the osc would give portals incorrect bulk size (1258)
6588 - fix debug daemon ioctl interface; allows daemon on ia64 (1274)
6589 - fix lock inversion caused by new llite matching code (1282)
6590 - limit the number of dirty pages on a client to 10MB (1286)
6591 - timed out locks were not being corrected cancelled (1289)
6592 - fix O_DIRECT above 4GB on IA-32 (1292)
6593 * major user-visible changes
6594 - fail out/fail over policy now controlled by the upcall (993)
6596 - add OBD_PING to check server availability and failure (954)
6597 - lustre messages are now sent in sending host order (288, 340, 891)
6598 - add eadatalen to MDS getattr reply (340)
6599 - OST read replies may contain second buffer, with per-page status (593)
6601 2003-03-11 Phil Schwan <phil@clusterfs.com>
6604 - LDLM_DEBUG macro fix, for gcc 3.2 (850)
6605 - failed open()s could cause deadlock; fixed (867, 869)
6606 - stop cancelling OST locks when files are closed (481)
6607 - overlapping XID spaces caused network corruption (851, 853)
6608 - fix unsafe fsfilt counter arithmetic; change to atomic_t
6609 - setattr_raw added, to do single-RPC, server-side setattrs
6610 - lmc/lconf syntax change for OST UUIDs
6611 - fix crashy race condition between ptlrpc_free_req and osc_close
6612 - don't use request in mdc_enqueue if we hit a timeout (889)
6613 - don't set the inode i_size for regular files from the MDS (896)
6614 - handle out of order completion AST (842)
6615 - don't LBUG if a lock request times out after receiving AST (913)
6616 - avoid d_rehash race in ll_find_alias by rehashing inside dcache_lock
6617 - if a bad lock AST arrives, send an error instead of dropping entirely
6618 - return 0 from revalidate2 if ll_intent_lock returns -EINTR (912)
6619 - fix leak in bulk IO when only partially completed (899, 900, 926)
6620 - fix O_DIRECT for ia64 (55)
6621 - (almost) eliminate Lustre-kernel-thread effects on load average (722)
6622 - C-z after timeout could hang a process forever; fixed (977)
6624 - client-side I/O cache (678, 924, 929, 941, 970)
6626 - READPAGE and SETATTRs which don't take server-side locks get
6629 2003-02-11 Phil Schwan <phil@clusterfs.com>
6632 - Fix ldlm_lock_match on the MDS to avoid matching remote locks (592)
6633 - Fix fsfilt_extN_readpage() to read a full page of directory
6634 entries, or fake the remainder if PAGE_SIZE != blocksize (500)
6635 - Avoid extra mdc_getattr() in ll_intent_lock when possible (534, 604)
6636 - Fix imbalanced LOV object allocation and out-of-bound access (469)
6637 - Most intent operations were removed, in favour of a new RPC mode
6638 that does a single RPC to the server and bypasses most of the VFS
6639 - All LDLM resource ID arrays were removed in favour of ldlm_res_id
6640 - Aggressively cancel local locks on DLM servers
6641 - mds_reint_unlink sends EA to the client if it's the last nlink.
6642 client uses that EA to unlink OST objects.
6643 - mds_reint_{rename,unlink,link} were rewritten to take ordered locks
6644 - recursive symlinks were fixed (439)
6645 - fixed NULL deref in DEBUG_REQ
6646 - filter_update_lastobjid no longer calls sync, which annoyed extN
6647 - fixed multi-client small-writes to a single file problem (445)
6648 - fixed mtime updates during file writes (607)
6649 - fixed vector writes on obdfilter causing problems when ENOSPC (670)
6650 - fixed bug in obd_brw_read/write() (under guise of testing 367)
6651 - fixed Linux OST size reporting problem (444, 656)
6652 - OST now updates object mtime with writes or setattr (607, 619)
6653 - client verifies file size before zeroing page past EOF (445)
6654 - OST now writes last allocated objid to disk with allocation (108)
6655 - LOV on echo now works (409)
6657 - mds_reint_unlink sends a new buffer, with the EA included. this
6658 buffer is only valid if body->valid & OBD_MD_FLEASIZE, which is only
6659 set if a regular file was being unlinked, and it was the last link
6660 - use PtlGet from the target for bulk writes (315)
6661 - OST now updates object mtime with writes or setattr (607, 619)
6662 - LDLM now has a grant-time callback to revalidate locked items, if
6664 - Many MDS operations were reorganized to combat race conditions
6666 - Merge b_intel branch (updated lprocfs code) - now at /proc/fs/lustre
6667 - configure check to avoid gcc version 2.96 20000731-2.96-98) (606)
6669 2003-01-06 Andreas Dilger <adilger@clusterfs.com>
6672 - Fully reactivate OST imports after reconnection (512, others)
6673 - Make sure client sees our -ENOTCONN from mds_handle (513 - partial)
6674 - More graceful error handling for truncating on dead OST (515)
6675 - Don't error out unless we're actually accessing dead stripes (474)
6676 - Fix garbage sizes when stripes are missing (410)
6677 - LRU counters were broken, causing constant lock purge (433, 432)
6678 - garbage on read from stripes with failed OSTs (441)
6679 - mark OSCs as active before reconnecting during recovery (438)
6680 - lov_enqueue and lov_cancel need to handle inactive OSTs (403)
6681 - lfind did not preserve OST order in output (443)
6682 - symlinks cause hung clients, incorrect data (439)
6683 - stop dereferencing request after dropping refcount (457)
6684 - don't LASSERT(spin_is_locked) on non-SMP (455)
6685 - fixes for many rename() bugs
6686 - fstat didn't correctly synchronize attributes (399)
6687 - server must handle lock cancellation during blocking AST prep (487)
6688 - bulk descriptors were free()d too soon (511)
6689 - fix paths in lconf, which would load incorrect modules (451, 507)
6690 - fix confusing lconf 'host not found' error message (386)
6691 - fix lock order deadlock on OST (O/R i_sem before journal ops, 478)
6692 - fix race condition in mdc_blocking_ast() for inode access (526)
6693 - fix lov_unpackmd() unpacking wrong number of stripes (537)
6694 - fix lov_set_osc_active() marking wrong OSC inactive (440)
6695 - fix bad lstripe lov_unpackmd() assertion (fix layering too) (527)
6696 - fix multiple writes of stripe MD to MDS (358, maybe 519)
6697 - fix lstripe in several ways (kernel side) (527)
6698 - fix request leak in ldlm_cli_enqueue (262)
6699 - incorrect OSC was marked inactive after OST failure
6700 - call mds_fs_cleanup before unmounting filesystem (524)
6701 - fix races between taking ns_lock and ldlm_lock_change_resource
6702 - fix races updating LOV export open file list
6703 - fix lov_enqueue error path, avoid decref-ing bad lock handle
6704 - fix recovery NULL deref in ldlm_cli_cancel_unused
6705 - fix some DLM races by using new hash table for lock handles (419)
6706 - permit the client to specify desired inodes, at replay
6707 - duplicate requests when we queue them for replay reintegration
6708 - fix last_rcvd offset calculation
6709 - sync after each recovered transaction, so we always make progress
6710 - never, not always, ERESTART requests without transnos
6711 - store the lov_desc in the MDS, so we don't depend on getlovinfo to
6713 - skip replay if the MDS says that the client is already connected
6714 - don't check for a recovery-enabled export to match lctl's UUID
6715 - don't INC_USE_COUNT for phantom exports
6716 - don't crash when cleaning up phantom exports (567)
6717 - don't double-finish or set replay data for errored mdc_open requests
6718 - abort requests when they time out, so we don't get old replies
6719 - send/receive replies for AST messages again
6720 - if the client says that it doesn't have the lock, cancel it on the
6722 - if we timeout during I/O, don't try to cancel an in-use lock; instead
6723 mark it as destroyed, it will all work out when decref is called
6724 - fix module use counts (22, 581)
6726 - ASTs now expect a reply (server cancels lock on error reply)
6728 2002-12-02 Andreas Dilger <adilger@clusterfs.com>
6731 - fix many simultaneous client startup (392)
6732 - fix dentry->d_it clobbering
6733 - credentials weren't being shipped for readdir/getattr operations
6734 - remove invalid assertions triggered during some concurrent MD
6736 - proper Lustre versions added (336, 389)
6737 - fix memory leak for create error case (398)
6738 - fix LOV locking bug that would get cli/srv out of sync
6739 - fix echo client over LOV (409)
6740 - fix dbench 2, extN refcount problem (170, 258, 356, 418)
6741 - fix double-O_EXCL intent crash (424)
6742 - avoid sending multiple lock CANCELs (352)
6744 - MDS can do multi-client recovery (modulo bugs in new code)
6746 - many updates, edits, cleanups
6748 2002-11-18 Phil Schwan <phil@clusterfs.com>
6751 - fix null d_it dereference (346)
6752 - fix full OST/dbench hang (333)
6753 - fix permission problem with file removal (286)
6754 - fix removal of OSCs from LOV when they fail
6755 - fix NULL deref during bulk timeout (214)
6756 - fix problems related to multiple filesystems on one MDS (241)
6757 - fixed serious subtle metadata locking bugs
6758 - free locks on clients when inodes are removed due to memory
6760 - fix inode pointer in lock data (285)
6761 - partial support for multiple MDS on a single host (241)
6762 - data locks weren't cancelled at clear_inode time (290, 311)
6763 - intent locks could lead to unbounded lock growth (205)
6764 - added a maximum lock count, an LRU list, and a flusher
6765 - fix multiple rename (365)
6766 - properly abstracted the echo client
6767 - OSC locked 1 byte too many; fixed
6768 - rewrote brw callback code:
6769 - fixed recovery bugs related to LOVs (306)
6770 - fixed too-many-pages-in-one-write crash (191)
6771 - fixed (again) crash in sync_io_timeout (214)
6772 - probably fixed callback-related race (385)
6774 - Add capability to MDS protocol
6775 - LDLM cancellations and callbacks on different portals
6777 2002-10-28 Andreas Dilger <adilger@clusterfs.com>
6780 - limit client IOV size to PTL_MD_MAX_IOV (611336, 191)
6781 - defer open object destruction to close time (601981, 138)
6782 - open/close OST file handle in obdo (OBD_MD_FLHANDLE) (601981, 138)
6783 - move LDLM_ENQUEUE/CONVERT back to MDS portal (625069)
6784 - abstract ll_lookup2, fix ll_revalidate2 to use abstraction (256)
6785 - don't call obd_setattr in ll_file_release for destroyed objects
6786 * protocol change to lustre_msg: move |version| and add |flags|
6787 * protocol change to osc_punch: "start" in "o_size", "end" in "o_blocks"
6788 * lock replay: for LDLM_FL_REPLAY trust client to do right thing
6789 * added replay of create, unlink, link and rename operations during
6790 MDS failover; recovery should be much more robust now
6791 * remove failed OSCs from LOVs (only lov_create uses this so far)
6792 * the lustre-HOWTO was brought (more) up to date (582544)
6794 2002-10-23 Phil Schwan <phil@clusterfs.com>
6797 - in-use dentries weren't being reused properly (617851)
6798 - prevent multiple LDLM setup (599178)
6799 - fix LOV size calculations for truncate (617853)
6800 - fix client handling of MDS intent errors (POSIX)
6801 - fix permission bug in lovstripe.c test (624321)
6802 - fix MDS thread deadlock - move LDLM handler to DLM portal (625069)
6803 - truncate past end of file could corrupt data
6804 - proper cleanup after timeouts, crashes, etc (592524, 550815)
6805 - a race in recovery could return ETIMEDOUT to apps (623947)
6806 - building outside the source directory was fixed
6807 * the lustre-HOWTO was brought (more) up to date (582544)
6808 * major progress was made on recovery functionality
6810 2002-10-10 Phil Schwan <phil@clusterfs.com>
6813 - recovery deadlock fix
6814 - rm -rf causes LBUG fix (617817)
6815 - file open by multiple tasks fix (618962)
6816 - directory permissions bugs (602707 and 620007)
6817 - journal_stop fixed with locking (611313)
6818 - O_APPEND failures resolved (618273, perhaps 614459)
6819 - lconf PATH fix (619770)
6820 - IA64 build fix (621450)
6821 - RPC buffer sizes scale with amount of memory
6823 2002-10-01 Phil Schwan <phil@clusterfs.com>
6826 - locks would be cancelled without throwing away data pages,
6827 resulting in inconsistent data (605627)
6828 - inode attributes were not always being refreshed (605627, 612449)
6829 - lconf now continues to cleanup after lctl reports an error
6830 - MDS now enforces user permissions (602707)
6831 - lprocfs cleanup fixed, but not yet enabled (614157)
6832 - fixed infinite server hang, should a client not respond to an AST
6833 - avoid going into recovery if user calls readlink() with a buffer
6834 that's too small (613941)
6835 - AST RPCs no longer require replies (614867) -- this may be changed
6836 - don't crash server if client sends an IOV that's too big (611336)
6837 - fixed lock conversion deadlock (611892)
6838 - fixed the following of symlinks (614622)
6839 * recovery: the server can remove locks from a client that dies, other
6840 clients can make progress
6841 * more extN patch fixes
6842 * compile-time configurable ptlrpc buffer allocations
6844 - collaborative read cache document
6845 - Lustre Lite Performance CDR document-in-progress
6847 2002-09-20 Andreas Dilger <adilger@clusterfs.com>
6850 - fix typo in patch-2.4.18
6852 2002-09-20 Andreas Dilger <adilger@clusterfs.com>
6855 - clear ptlrpc request each time in handle_incoming_request()
6856 - unlink of files now destroys the object on the OST
6858 2002-09-19 Peter Braam <braam@clusterfs.com>
6860 * add hard link support
6861 * change obdfile creation method
6862 * kernel patch changed
6864 2002-09-19 Peter Braam <braam@clusterfs.com>
6867 - stack overflow bug in extN fixed
6869 2002-09-18 Andreas Dilger <adilger@clusterfs.com>
6871 * documentation updates
6872 - add man pages for config tools
6873 - update tests/README to describe testing with new config tools
6874 - finish metadata API descriptions
6875 * bug fixes and cleanups
6876 - statfs workaround for 16TB limit
6877 - LOV stripe allocation improved, can stripe on subset of OSTs
6878 - LOV file size/IO offset was wrong for files > 4GB in size
6879 - object EA data was being dropped, caused files to be unreadable
6880 - memory overflow with non-LOV OST caused memory corruption
6881 - fixed regression tests to work with new config tools, obdfilter
6882 - fixed bug when directory size became larger than 1 block
6883 - fixed bug (for single client case) when PWD was deleted
6884 - invalidate local directory pages when doing intent-based ops
6885 - avoid LDLM oops when lock callback contained bad data
6887 2002-09-09 Andreas Dilger <adilger@clusterfs.com>
6889 * documentation updates
6890 * bug fixes and cleanups
6891 - configuration tools
6894 - 64-bit compile warnings
6895 - 64-bit internal statfs data
6897 * test_brw on persistent OST devices
6899 * lprocfs (disabled)
6901 2002-09-04 Andreas Dilger <adilger@clusterfs.com>
6903 * documentation updates
6904 * bug fixes and cleanups
6905 * configuration tools
6907 2002-08-30 Peter J. Braam <braam@clusterfs.com>
6910 * many small fixes to 0_5_4
6911 * io/network handling
6912 * thinkos in MDS operations
6914 2002-08-24 Peter J. Braam <braam@clusterfs.com>
6917 * crucial basic fixes to 0.5.3
6918 * IOR, Iozone work over Elan
6921 2002-08-07 Phil Schwan <phil@clusterfs.com>
6922 * version 0_5_3, our first alpha
6923 * we use the new Portals iovs
6924 * documentation updates
6925 * bug fixes and cleanups
6926 * small changes in the DLM wire protocol
6928 2002-07-25 Peter J. Braam <braam@clusterfs.com>
6929 * version 0_5_1 with some initial stability,
6930 * locking on MD and file I/O.
6931 * documentation updates
6932 * several bug fixes since 0.5.0
6933 * small changes in wire protocol
6935 2002-07-18 Phil Schwan <phil@clusterfs.com>
6937 * delivered as Lustre Light Alpha
6938 * fixed a crash after handling invalid MDS requests
6939 * fixed directory pages for architectures with non-4k pages sizes
6941 2002-07-11 Andreas Dilger <adilger@clusterfs.com>
6942 * release version v0_4_4
6943 * Moves TCP acceptor to be on port 2432 (unused Coda port) instead
6945 * Fixes a number of interruption problems with OST operations.
6946 * Update documentation for portals header changes
6947 * Move all wire protocol structs/defines to lustre_idl.h
6948 * Fixes symlink length bug.
6949 * Add tcpdump to repository.
6951 2002-07-05 Andreas Dilger <adilger@clusterfs.com>
6952 * release version v0_4_3
6953 * Fixes statfs for inodes on extN.
6954 * Fixes bug in runtests which would delete /etc/hosts.
6955 * Use 64-bit object IDs wherever possible (not into VFS though)
6956 Remove ost_get_info, which is unused by lustre, and out of date.
6958 2002-07-03 Peter Braam <braam@clusterfs.com>
6959 * release version v0_4_2 Fixes a lookup error (type not passed)
6960 * move forward to head of Portals
6961 * move forward to latest Lustre kernel
6963 2002-06-25 Peter Braam <braam@clusterfs.com>
6964 * release version v0_4_1. Hopefully stable on single node use.