1 tbd Sun Microsystems, Inc.
4 2.6.16.60-0.42.4 (SLES 10),
5 2.6.27.29-0.1 (SLES11),
6 2.6.18-128.7.1.el5 (RHEL 5)
7 2.6.18-128.7.1.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
13 Severity : enhancement
15 Description: Adjust locks' extents on their first enqueue, so that at the time
16 they get granted, there is no need for another pass through the
17 queues since they are already shaped into the proper forms.
21 Description: Fix mds_shrink_intent_reply()/mds_intent_policy() to pass correct
22 arguments and prevent LBUG() in lustre_shrink_reply_v2().
26 Description: Change tunefs.lustre and mkfs.lustre --mountfsoptions so that
27 exactly the specified mount options are used. Leaving off
28 any "mandatory" mount options is an error. Leaving off any
29 default mount options causes a warning, but is allowed.
30 Change errors=remount-ro from mandatory to default. Sanitize
31 the mount string before storing it. Update man pages accordingly.
35 Description: mds_getattr() should return 0, even if mds_fid2entry() fails with
36 -ENOENT. Also fix in ptlrpc_expire_one_request() to print signed
39 Severity : enhancement
41 Description: Remove set_info(KEY_UNLINKED) from MDS/OSC
45 Description: Drop unnecessary __GFP_NOMEMALLOC flag from filter_get_page()
49 Description: Conf-sanity.sh test to check client reconnection to a busy server.
51 Severity : enhancement
53 Description: Build kernels (RHEL5, OEL5 and SLES10/11) using the vendor's own
58 Description: optimize the extent lock finding during read/write at client side
62 Description: can't stat file in some situation.
63 Details : improve initialize osc date when target is added to mds and
64 ability to resend too big getattr request is client isn't have info
67 Severity : enhancement
69 Description: Update kernel to SLES11 2.6.27.29-0.1.1.
73 Description: Improved cleanup procedure for conf-sanity test_46a.
77 Description: Prevent inconsistences between linux and lustre mount structures.
78 Details : Wait indefinitely in server_wait_finished() until mnt_count drops.
79 Make the sleep interruptible.
81 Severity : enhancement
83 Description: Enhancement for lfs(1) command to use numeric uid/gid.
85 Severity : enhancement
87 Description: Communicate OST degraded/readonly state via statfs to MDS
88 Details : Flags in the statfs returned from OSTs indicate whether the
89 OST is in a degraded RAID state, or if the filesystem has
90 turned read-only after a filesystem error is detected.
92 Severity : enhancement
94 Description: Vanilla kernel 2.6.30 patchless client support.
99 Description: don't panic if EPROTO was hit when reading symlink
100 Details : correctly handling request reference in error cases.
105 Description: open sometimes returns ENOENT instead of EACCES
106 Details : checking permission should be part of open part of mds_open, not
107 lookup part. so server should be set DISP_OPEN_OPEN disposition
108 before starting permission check.
109 Also not need revalidate dentry if client already have LOOKUP lock.
113 Description: enable client interface failover
114 Frequency : on servers with multiple network interfaces
115 Details : When a child reconnects from another NID, properly update export
116 nid hash position and ldlm reverse import.
118 Severity : enhancement
120 Description: implemented direct I/O with arbitrary (nonaligned) memory
121 addresses and file offsets.
123 Severity : enhancement
125 Description: added more recovery timeout options.
127 Severity : enhancement
129 Description: added llapi_file_open, llapi_file_create, llapi_file_get_stripe
134 Description: Avoid deadlock for local client writes
135 Frequency : only on systems with clients writing to an OST on the same node
136 Details : Use new OBD_BRW_MEMALLOC flag to notify OST about writes in the
137 memory freeing context. This allows OST threads to set the
138 PF_MEMALLOC flag on task structures in order to allocate memory
139 from reserved pools and complete IO.
140 Use GFP_HIGHUSER for OST allocations for non-local client writes,
141 so that the OST threads generate memory pressure and allow
142 inactive pages to be reclaimed.
147 Descriptoin: lock ordering violation between &cli->cl_sem and _lprocfs_lock
148 Details : move ldlm namespace creation in setup phase to avoid grab
149 _lprocfs_lock with cli_sem held.
153 Frequency : only during format of test systems
154 Description: Unable to run several mkfs.lustre on loop devices at the same time
155 Details : mkfs.lustre returns error 256 on the concurrent loop devices
156 formatting. The solution is to proper handle the error.
158 Severity : enhancement
160 Description: implement async create (obd_async_create) method for osc, to avoid
161 too long waiting new ost objects with holding ldlm lock.
163 Severity : enhancement
165 Descriptoin: OST grant too much space to client even there are not enough space.
166 Details : Client will shrink its grant cache to OST if there are no write
167 activity over 6 mins (GRANT_SHRINK_INTERVAL), and OST will retrieve
168 this grant cache if there are already not enough avaible space
169 (left_space < total_clients * 32M).
173 Frequency : occasionally during network problems
174 Description: client not allowed to reconnect to OST because of active request
175 Details : abort bulk requests received by the OST once the client has timed
176 out since the client will resend the request anyway. The client
177 also now retries to reconnect to the same server if a connect
178 request failed with EBUSY or -EAGAIN.
182 Frequency : occasionally during network problems
183 Description: client not allowed to reconnect to OST because of active request
184 Details : The client also now retries to reconnect to the same server if a
185 connect request failed with EBUSY or -EAGAIN. The patch has not
186 been landed in the previous commits.
189 Frequency : rare, if used wide striped file and one ost in down.
191 Descriptoin: don't return error if we created a subset of objects for file.
192 Details : lov_update_create_set() uses set->set_success as index for created
193 objects, so if some requests failed, they will have hole at end of
194 array and we can use qos_shrink_lsm for allocate correct lsm.
198 Description: quotacheck performance/scaling issues
199 Details : reduce quotacheck time on empty filesystem by skipping uninit
204 Description: Slow stale export processing during normal start up
205 Details : The global mgc lock prevents OST setup to be run in parallel.
206 Replace the global lock with a per-config_llog_data semaphore.
210 Description: Out or order replies might be lost on replay
211 Details : In ptlrpc_retain_replayable_request if we cannot find retained
212 request with tid smaller then one currently being added, add it
213 to the start, not end of the list.
215 -------------------------------------------------------------------------------
216 2009-10-16 Sun Microsystems, Inc.
218 * Support for kernels:
219 2.6.16.60-0.42.4 (SLES 10),
220 2.6.27.29-0.1 (SLES11, i686 & x84_64 only),
221 2.6.18-128.7.1.el5 (RHEL 5),
222 2.6.18-128.7.1.el5 (OEL 5),
223 * Client support for unpatched kernels:
224 (see http://wiki.lustre.org/index.php?title=Patchless_Client)
225 2.6.16 - 2.6.27 vanilla (kernel.org)
226 * Recommended e2fsprogs version: 1.41.6.sun1
227 * ext4 support for RHEL5 is experimental and thus should not be
230 Severity : enhancement
232 Description: Add OEL5 support.
234 Severity : enhancement
236 Description: Update kernel to SLES11 2.6.27.29-0.1.
240 Description: File checksum failures with OST read cache on
241 Details : Disable page poisoning when the bulk transfer has to be aborted
242 because the client got evicted.
246 Description: Don't allow make backward step on assiging osc next id.
247 Details : race between allocation next id and ll_sync thread can be cause
248 of set wrong osc next id and can be kill valid ost objects.
250 Severity : enhancement
252 Description: Update kernel to RHEL5 2.6.18-128.7.1.el5.
254 Severity : enhancement
256 Description: Update kernel to SLES10 SP2 2.6.16.60-0.42.4.
260 Description: Changes in raid5-large-io-rhel5.patch to calculate sectors properly
264 Description: Increase the default BLK_DEF_MAX_SECTORS value for RHEL5 and SLES11
268 Description: Error handling in osc_statfs_interpret() has been improved.
269 Details : Check in osc_statfs_interpret() for EBADR.
273 Description: Do not send statfs() requests to OSTs disabled by administrator.
274 Details : Check in lov_prep_statfs_set() for non-NULL ltd_exp.
278 Description: Do not update ctime for the deleted inode.
279 Details : Check in mds_reint_unlink() before calling fsfilt_setattr().
283 Description: Increase of the size of the LDLM resource hash.
284 Details : Bump up RES_HASH_BITS=12.
288 Description: correctly send lsm on open replay
289 Details : MDS is trust to LSM size on replay open, but client can set wrong size
294 Description: Deadlock between filter_destroy() and filter_commitrw_write().
295 Details : filter_destroy() does not hold the DLM lock over the whole
296 operation. If the DLM lock is dropped, filter_commitrw() can go
297 through, causing the deadlock between page lock and i_mutex.
298 The i_alloc_sem should also be hold in filter_destroy() while
303 Description: truncate starts GFP_FS allocation under transaction causing deadlock
304 Details : ldiskfs_truncate calls grab_cache_page which may start page
305 allocation under an open transaction. This may lead to
306 calling prune_icache with consequent lustre reentrance.
309 Frequency : only when down/upgrading the MDS to 1.6/1.8 while 1.8 clients are
310 still up and when the OST pool feature is used
312 Description: interop testing got LBUG when run dd with OST pool
313 :LustreError: 30032:0:(llite_lib.c:1913:ll_replace_lsm()) LBUG
314 Details : down/upgrading the MDS to a version that doesn't/does support OST
315 pool can cause clients to crash because the lsm has changed
320 Description: missing tree_status on 1.8.1 RPM build
321 Details : make rpms failed due because the tree_status file is missing.
325 Description: continuing LustreError "mds adjust qunit failed!"
326 Details : don't print message on the console when ->adjust_qunit fails.
330 Description: don't increase ldlm timeout if previous client was evicted
331 Details : if a client doesn't respond to a blocking callback within the
332 adaptive ldlm enqueue timeout, don't adjust the adaptive estimate
333 when the lock is next granted.
337 Description: ost is being unmounted w/o all writes to last_rcvd landing on disk.
338 affects recovery negatively.
339 Details : make sure all exports have been properly destroyed by the zombie
340 thread processed before stopping the target.
344 Description: Performance degradation with O_DIRECT between 1.6 & 1.8.1 b190
345 Details : disable write barrier for ext4/SLES11.
349 Description: Kernel panic - not syncing: Out of memory and no killable
350 processes... on OSS when iozone
351 Details : fix memory leak in the journal checksum patch.
355 Description: group quota "too many blocks" OSS crashes
356 Details : we should keep the same uid/gid for lquota_chkquota() and
357 lquota_pending_commit()
361 Description: LustreError: 9153:0:(quota_context.c:622:dqacq_completion()) LBUG
362 Details : don't LBUG on release quota error. Just a workaround until the
363 problem is understood.
365 ------------------------------------------------------------------------------
366 2009-07-31 Sun Microsystems, Inc.
368 * Support for kernels:
369 2.6.16.60-0.39.3 (SLES 10),
370 2.6.27.23-0.1 (SLES11, i686 & x84_64 only),
371 2.6.18-128.1.14.el5 (RHEL 5),
372 * Client support for unpatched kernels:
373 (see http://wiki.lustre.org/index.php?title=Patchless_Client)
374 2.6.16 - 2.6.27 vanilla (kernel.org)
375 * Recommended e2fsprogs version: 1.41.6.sun1
376 * File join has been disabled in this release, refer to Bugzilla 16929.
377 * NFS export disabled when stack size < 8192. Since the NFSv4 export
378 of Lustre filesystem with 4K stack may cause a stack overflow. For
379 more information, please refer to bugzilla 17630.
380 * ext4 support for RHEL5 is experimental and thus should not be
383 Severity : enhancement
385 Description: Update kernel to SLES10 SP2 2.6.16.60-0.39.3.
388 Frequency : with 1.8 server and 1.6 clients
390 Description: correctly shrink reply for avoid send too big message to client.
391 Details : 1.8 mds is allocate to big buffer to LOV EA data and this produce
392 some problems with sending this reply to 1.6 client.
396 Description: Repeated atomic allocation failures.
397 Details : Use GFP_HIGHUSER | __GFP_NOMEMALLOC flags for memory allocations
398 to generate memory pressure and allow reclaiming of inactive pages.
399 At the same time, do not allow to exhaust emergency pools.
400 For local clients the use of GFP_NOFS will be introduced in 1.8.2
402 Severity : enhancement
403 Bugzilla : 19846, 18289
404 Description: Update kernel to RHEL5 2.6.18-128.1.14.el5.
406 Severity : enhancement
407 Bugzilla : 19625, 16893, 18668, 19848
408 Description: Add support for SLES11 2.6.27.23-0.1.
410 Severity : enhancement
412 Description: Update client support to vanila kernels up to 2.6.27.
414 Severity : enhancement
416 Description: Update kernel to SLES10 SP2 2.6.16.60-0.37.
418 Severity : enhancement
420 Description: Compile with -Werror by default for i686 and x86_64.
424 Description: resolve race between obd_disconnect and class_disconnect_exports
425 Details : if obd_disconnect will be called to already disconnected export he
426 forget release one reference and osc module can't unloaded.
428 Severity : enhancement
430 Description: move AT tunable parameters for more consistent usage
431 Details : add AT tunables under /proc/sys/lustre, add to conf_param parsing
435 Descriptoin: correctly skip time estimate if in recovery
436 Details : rq_send_state insn't bitmask so using bitwise ops is forbid.
440 Descriptoin: OSS DeadLock
441 Details : Use trylock to prevent deadlock when shrink icache.
443 Severity : enhancement
445 Description: Allow tuning service thread via /proc
446 Details : For each service a new
447 /proc/fs/lustre/{service}/*/thread_{min,max,started} entry is
448 created that can be used to set min/max thread counts, and get the
449 current number of running threads.
451 Severity : enhancement
453 Description: Add state history info file, enhance import info file
454 Details : Track import connection state changes in a new osc/mdc proc file;
455 add overview-type data to the osc/mdc import proc file.
459 Description: Reduce small size read RPC
460 Details : Set read-ahead limite for every file and only do read-ahead when
461 available read-ahead pages are bigger than 1M to avoid small size
466 Description: free_entry erroneously used groups_free instead of put_group_info
468 Severity : enhancement
470 Description: Make read-ahead stripe size aligned.
472 Severity : enhancement
474 Description: MDS create should not wait for statfs RPC while holding DLM lock.
477 Frequency : rare, connect and disconnect target at same time
479 Descriptoin: ASSERTION(atomic_read(&imp->imp_inflight) == 0
480 Details : don't call obd_disconnect under lov_lock. this long time
481 operation and can block ptlrpcd which answer to connect request.
484 Frequency : start MDS on uncleanly shutdowned MDS device
486 Descriptoin: ll_sync thread stay in waiting mds<>ost recovery finished
487 Details : stay in waiting mds<>ost recovery finished produce random bugs
488 due race between two ll_sync thread for one lov target. send
489 ACTIVATE event only if connect realy finished and import have
493 Frequency : start MDS on uncleanly shutdowned MDS device
495 Descriptoin: aborting recovery hang on MDS
496 Details : don't throttle destroy RPCs for the MDT.
500 Description: Slow reads beyond 8Tb offsets.
501 Details : Page index integer overflow in ll_read_ahead_page
505 Description: Soft lockup on OSS after MDS failover
506 Details : MSG_CONNECT_INITIAL is not set on the initial MDS->OST connect.
507 As a conseqence, the patch from bug 18224 is not operational
508 and the MDS export cannot be reused on the OSTs until it gets
512 Frequency : rare, only if using MMP with Linux RAID
514 Description: MMP doesn't work with Linux RAID
515 Details : While using HA for Lustre servers with Linux RAID, it is possible
516 that MMP will not detect multiple mounts. To make this work we
517 need to unplug the device queue in RAID when the MMP block is being
518 written. Also while reading the MMP block, we should read it from
519 disk and not the cached one.
522 Frequency : rare, during recovery
524 Description: Assertion failure in ldlm_lock_put
525 Details : Do not put cancelled locks into replay list, hold references on
530 Description: 1.6.5 mdsrate performance is slower than 1.4.11/12 (MDS is not cpu bound!)
531 Details : create_count always drops to the min value (=32) because grow_count
532 is being changed before the precreate RPC completes.
535 Frequency : Only in RHEL5 when mounting multiple ext3 filesystems
538 Description: "kmem_cache_create: duplicate cache jbd_4k" error message
539 Details : add proper locking for creation of jbd_4k slab cache
543 Description: MMP check in ext3_remount() fails without displaying any error
544 Details : When multiple mount protection fails during remount, proper error
549 Description: Rare Client crash on resend if the file was deleted.
550 Details : When file is opened, but open reply is lost and file is
551 subsequently deleted before resend, resend processing logic
552 breaks trying to open the file again, should not try to open.
556 Description: add check for >8TB ldiskfs filesystems
557 Details : ext3-based ldiskfs does not support greater than 8TB LUNs.
558 Don't allow >8TB ldiskfs filesystems to be mounted without
559 force_over_8tb mount option
563 Description: Client locked up when running multiple instances of an app. on
564 multiple mount points
565 Details : ll_shrink_cache() can sleep while holding the ll_sb_lock.
566 Convert ll_sb_lock to a read/write semaphore to fix the problem.
570 Description: Cannot acces an NFS-mounted Lustre filesystem
571 Details : An NFS client cannot access the Lustre filesystem NFS-mounted
572 from a Lustre-client exporting the Lustre filesystem via NFS.
576 Description: panic in ll_statahead_thread
577 Details : grab dentry reference in parent process.
580 -------------------------------------------------------------------------------
582 tbd Sun Microsystems, Inc.
584 * Support for kernels:
585 2.6.16.60-0.31 (SLES 10),
586 2.6.18-128.1.6.el5 (RHEL 5),
587 2.6.22.14 vanilla (kernel.org)
588 * Client support for unpatched kernels:
589 (see http://wiki.lustre.org/index.php?title=Patchless_Client)
590 2.6.16 - 2.6.22 vanilla (kernel.org)
591 * Recommended e2fsprogs version: 1.40.11-sun1
592 * File join has been disabled, refer to Bugzilla 16929.
593 * A new Lustre ADIO driver is available for MPICH2-1.0.7.
594 * NFS export disabled when stack size < 8192. Since the NFSv4 export of
595 Lustre filesystem with 4K stack may cause a stack overflow. For more
596 information, please refer to bugzilla 17630.
598 Severity : enhancement
600 Description: Update to RHEL5.3 kernel-2.6.18-128.1.6.el5.
602 Severity : enhancement
604 Description: Update OFED release to 1.4.1 RC4
606 Severity : major, only with big OST
608 Description: Very poor metadata performance on Infiniband lustre configuration
609 Details : OST object precreation becomes very slow on big OSTs. This is due
610 to the ialloc patch spending too much time scanning groups.
613 Frequency : during recovery
615 Description: don't mix llog inodes with normal.
616 Details : allocate inodes for log in last inode group
621 Description: fix lqs' reference which won't be put in some situations
622 Details : This patch fixes:
623 1. In quota_check_common(), this function will check quota
624 for user and group, but only send one return via "pending".
625 In most cases, the pendings should be same. But that is not
627 2. If quotaoff runs between lquota_chkquota() and
628 lquota_pending_commit(), the same thing will happen too.
629 That is why it comes:
630 - if (!ll_sb_any_quota_active(qctxt->lqc_sb))
633 -------------------------------------------------------------------------------
635 2008-12-31 Sun Microsystems, Inc.
637 * Support for kernels:
638 2.6.5-7.314 (SLES 9),
639 2.6.9-67.0.22.EL (RHEL 4),
640 2.6.16.60-0.31 (SLES 10),
641 2.6.18-92.1.17.el5 (RHEL 5),
642 2.6.22.14 vanilla (kernel.org)
643 * Client support for unpatched kernels:
644 (see http://wiki.lustre.org/index.php?title=Patchless_Client)
645 2.6.16 - 2.6.22 vanilla (kernel.org)
646 * Client support for unpatched kernels:
647 we do not recommend using patchless RHEL4 clients with kernels
648 prior to 2.6.9-55EL (RHEL4U5).
649 * Recommended e2fsprogs version: 1.40.11-sun1
650 * Note that reiserfs quotas are disabled on SLES 10 in this kernel.
651 * RHEL 4 and RHEL 5/SLES 10 clients behaves differently on 'cd' to a
652 removed cwd "./" (refer to Bugzilla 14399).
653 * A new quota file format has been introduced in 1.6.5.
654 The format conversion from prior releases is handled transparently,
655 but releases older than 1.4.12/1.6.5 don't understand this new
656 format. The automatic format conversion can be avoided by running
657 the following command on the MDS:
658 'tunefs.lustre --param="mdt.quota_type=ug1" $MDTDEV'.
659 For more information, please refer to bugzilla 13904.
660 * A new quota file format was introduced in 1.6.6/1.8.0 (kernels 2.6.16+).
661 The format conversion from prior releases is handled transparently,
662 but releases older than 1.6.6/1.8.0 don't understand this new
663 format. The automatic format conversion can be avoided by running
664 the following commands on the MDS and OSS servers (for
665 pre 1.4.12-1.6.5 quota files):
666 'tunefs.lustre --param="mdt.quota_type=ug1" $MDTDEV',
667 'tunefs.lustre --param="ost.quota_type=ug1" $MDTDEV'
668 or (for 1.4.12/1.6.5 quota files)
669 'tunefs.lustre --param="mdt.quota_type=ug2" $MDTDEV',
670 'tunefs.lustre --param="ost.quota_type=ug2" $MDTDEV'
671 For more information, please refer to bugzilla 13904.
672 * Output of lfs quota has been made less detailed by default,
673 old (verbose) output can be obtained by using -v option.
674 * File join has been disabled in this release, refer to Bugzilla 16929.
675 * A new Lustre ADIO driver is available for MPICH2-1.0.7.
676 * NFS export disabled when stack size < 8192. Since the NFSv4 export of
677 Lustre filesystem with 4K stack may cause a stack overflow. For more
678 information, please refer to bugzilla 17630.
680 Severity : enhancement
682 Description: Caching OSS
683 Details : introduce data caching on the OSS. The OSS now relies on the linux
684 kernel page cache to keep recently accessed data in memory.
685 It is worth noting that all write requests are still flushed
686 synchronously as in lustre 1.6.
688 Severity : enhancement
690 Description: Version based recovery
691 Details : introduce finer grained recovery able to detect transaction
692 dependencies and can deal with transaction gaps caused by clients
693 failing at the same time as the server.
695 Severity : enhancement
697 Description: Enable adaptive timeouts by default
698 Details : The Lustre timeout value in /proc/sys/lustre/timeout is now
699 managed dynamically based on server load and should not need
700 to be tuned manually based on cluster size. This allows Lustre
701 to work under a wider variety of system sizes and loads, without
702 unnecessarily causing lengthy recovery times.
704 Severity : enhancement
706 Description: Add OST Pools support
707 Details : File striping can now be set to use an arbitrary pool of OSTs
709 Severity : enhancement
711 Description: add lazystatfs mount option to allow statfs(2) to skip down OSTs
712 Details : allow skip disconnected ost for send statfs request and hide error
716 Frequency : rare, on llog test 6
718 Descriptoin: don't allow connect to already connected import
719 Details : allowing connect to already connected import is hide connecting problem.
722 Frequency : rare, on failed llog setup
724 Descriptoin: don't leak obd reference on failed llog setup
725 Details : for failed llog setup - mgc forget call class_destroy_import for
726 client import, move destroy import to more generic place.
731 Descriptoin: allow kill process which wait statahead result
732 Details : for some reasons 'ls' can stick in waiting result from statahead,
733 in this case need way for kill this process.
736 Frequency : rare, at shutdown
738 Descriptoin: panic at umount
739 Details : llap_shrinker can be raced with killing super block from list and
740 this produce panic with access to already freeded pointer
745 Descriptoin: don't lose wakeup for imp_recovery_waitq
746 Details : recover_import_no_retry or invalidate_import and import_close can
747 both sleep on imp_recovery_waitq, but we was send only one wakeup
753 Descriptoin: panic in mds_open
754 Details : don't confuse mds_finish_transno() with PTR_ERR(-ENOENT)
759 Descriptoin: stuck in cache_remove_extent() or panic with accessing to already
761 Details : release lock refernce only after add page to pages list.
764 Frequency : always with long access acl
766 Descriptoin: mds can't pack reply with long acl.
767 Details : mds don't control size of acl but they limited by reint/getattr
772 Frequency : on remount
774 Description: external journal device not working after the remount
775 Details : clear dev_rdonly flag for external journal devices in
781 Description: shutdown vs evict race
782 Details : client_disconnect_export vs connect request race.
783 if client will evicted at this time - we start invalidate
784 thread without referece to import and import can be freed
790 Description: shrink LOV EAs before replying
791 Details : correctly adjust LOV EA buffer for reply.
796 Description: don't skip ost target if they assigned to file
797 Details : Drop slow OSCs if we can, but not for requested start idx.
798 This means "if OSC is slow and it is not the requested
799 start OST, then it can be skipped, otherwise skip it only
800 if it is inactive/recovering/out-of-space.
802 Severity : enhancement
804 Description: Update to RHEL5 kernel-2.6.18-92.1.17.el5.
807 Frequency : rare, need acl's on inode.
809 Description: client can't handle ost additional correctly
810 Details : if ost was added after client connected to mds client can have
811 hit lnet_try_match_md ... to big messages to wide striped files.
812 in this case need teach client to handle config events about add
813 lov target and update client max ea size at that event.
815 Severity : enhancement
817 Description: Update to sles9 kernel-2.6.5-7.314.
819 Severity : enhancement
821 Description: Update to SLES10 SP2 kernel-2.6.16.60-0.31.
824 Frequency : Create a symlink file with a very long name
826 Description: ldlm_cancel_pack()) ASSERTION(max >= dlm->lock_count + count)
827 Details : If there is no extra space in the request for early cancels,
828 ldlm_req_handles_avail() returns 0 instead of a negative value.
833 Description: mds is deadlocked
834 Details : in rare cases, inode in catalog can have i_no less than have parent
835 i_no, this produce wrong order for locking during open, and
836 parallel unlink can be lock open. this need teach mds_open to grab
837 locks in resource id order, not at parent -> child order.
839 Severity : enhancement
841 Description: Add /proc entry for import status
842 Details : The mdc, osc, and mgc import directories now have
843 an import directory that contains useful import data for debugging
846 Severity : enhancement
848 Description: Re-disable certain /proc logging
849 Details : Enable and disable client's offset_stats, extents_stats and
850 extents_stats_per_process stats logging on the fly.
853 Frequency : Only on FC kernels 2.6.22+
855 Description: oops in statahead
856 Details : Do not drop reference count for the dentry from VFS when lookup,
857 VFS will do that by itself.
859 Severity : enhancement
861 Description: Generic /proc file permissions
862 Details : Set /Proc file permissions in a more generic way to enable non-
863 root users operate on some /proc files.
867 Description: Hitting mdc_commit_close() ASSERTION
868 Details : Properly handle request reference release in
869 ll_release_openhandle().
873 Frequency : only patchless client
874 Description: add workaround for race between add/remove dentry from hash
876 Severity : enhancement
878 Description: Allow OST glimpses to return PW locks
882 Description: LBUG when llog conf file is full
883 Details : When llog bitmap is full, ENOSPC should be returned for plain
888 Description: Prevent import from entering FULL state when server in recovery
892 Description: service mount cannot take device name with ":"
893 Details : Only when device name contains ":/" will mount treat it as
899 Description: replace ptlrpcd with the statahead thread to interpret the async
900 statahead RPC callback
904 Frequency : on recovery
905 Description: I/O failures after umount during fail back
906 Details : if client reconnected to restarted server we need join to recovery
907 instead of find server handler is changed and process self eviction
908 with cancel all locks.
910 Severity : enhancement
912 Description: Update to RHEL5 kernel-2.6.18-92.1.10.el5.
916 Description: Kernel BUG tries to release flock
917 Details : Lustre does not destroy flock lock before last reference goes
918 away. So always drop flock locks when client is evicted and
919 perform unlock regardless of successfulness of speaking to MDS.
921 Severity : enhancement
923 Description: Update to SLES10 SP2 kernel-2.6.16.60-0.27.
925 Severity : enhancement
927 Description: Upcall on Lustre log has been dumped
928 Details : Allow for a user mode script to be called once a Lustre log has
929 been dumped. It passes the filename of the dumped log to the
930 script, the location of the script can be specified via
931 /proc/sys/lnet/debug_log_upcall.
936 Description: avoid messages about idr_remove called for id that is not allocated
937 Details : Move assigment s_dev for clustered nfs to end of initialization,
938 for avoid problem with error handling.
943 Description: avoid Already found the key in hash [CONN_UNUSED_HASH] messages
944 Details : When connection is reused this not moved from CONN_UNUSED_HASH into
945 CONN_USED_HASH and this prodice warning when put connection again
951 Description: avoid ASSERTION(client_stat->nid_exp_ref_count == 0) failed
952 Details : release reference to stats when client disconnected, not
953 when export destroyed for avoid races when client destroyed
954 after main ost export.
958 Description: more cleanup in mds_lov
959 Details : add workaround for get valid ost count for avoid warnings about
960 drop too big messages, not init llog cat under semphore which
961 can be blocked on reconnect and break normal replay, fix access
964 Severity : enhancement
966 Description: Export bytes_read/bytes_write count on OSC/OST.
970 Description: Early reply size mismatch, MGC loses connection
971 Details : Apply the MGS_CONNECT_SUPPORTED mask at reconnect time so
972 the connect flags are properly negotiated.
976 Description: Properly propagate oinfo flags from lov to osc for statfs
977 Details : restore missing copy oi_flags to lov requests.
981 Description: exports in /proc are broken
982 Details : recreate /proc entries for clients when they reconnect.
984 Severity : enhancement
986 Description: Add man pages for llobdstat(8), llstat(8), plot-llstat(8),
987 : l_getgroups(8), lst(8), routerstat(8)
988 Details : included man pages for llobdstat(8), llstat(8),
989 : plot-llstat(8), l_getgroups(8), lst(8), routerstat(8)
991 Severity : enhancement
993 Description: Implement lustre ll_show_options method.
995 Severity : enhancement
997 Description: Update to SLES9 kernel-2.6.5-7.312.
999 Severity : enhancement
1001 Description: Update to RHEL4 kernel-2.6.9-67.0.22.EL.
1005 Description: exports in /proc are broken
1006 Details : recreate /proc entries for clients when they reconnect.
1010 Description: don't fail open with -ERANGE
1011 Details : if client connected until mds will be know about real ost count
1012 get LOV EA can be fail because mds not allocate enougth buffer
1017 Description: Resolve device initialization race
1018 Details : Prevent proc handler from accessing devices added to the
1019 obd_devs array but yet be intialized.
1021 Severity : enhancement
1023 Description: configure's --enable-quota should check the
1024 : kernel .config for CONFIG_QUOTA
1025 Details : configure is terminated if --enable-quota is passed but
1026 : no quota support is in kernel
1028 Severity : enhancement
1030 Description: Update to SLES10 SP2 kernel-2.6.16.60-0.23.
1032 Severity : enhancement
1034 Description: Update to RHEL5 kernel-2.6.18-92.1.6.el5.
1038 Frequency : rare, on PPC clients
1039 Description: don't swab ost objects in response about directory, because
1041 Details : bug similar bug 14856, but in different function.
1043 Severity : enhancement
1045 Description: lfs quota tool enhancement
1046 Details : added units specifiers support for setquota, default to
1047 current uid/gid for quota report, short quota stats by
1048 default, nonpositional parameters for setquota, added
1049 llapi_quotactl manual page.
1051 Severity : enhancement
1053 Description: *optional* service tags registration
1054 Details : if the "service tags" package is installed on a Lustre node
1055 When the filesystem is mounted, a local-node service tag will
1056 be created. See http://inventory.sun.com/ for more information
1057 about the Service Tags asset management system.
1061 Description: Client runs out of low memory
1062 Details : Consider only lowmem when counting initial number of llap pages
1065 Frequency : occasional
1067 Description: add refcount for osc callbacks, so avoid panic on shutdown
1069 Severity : enhancement
1071 Description: Update to RHEL4 kernel-2.6.9-67.0.20.
1074 Frequency : testing only
1076 Description: sanity test 65a fails if stripecount of -1 is set
1077 Details : handle -1 striping on filesystem in ll_dirstripe_verify
1080 Frequency : only in unusual configurations
1082 Description: Kernel panic with find ost index.
1083 Details : lov_obd have panic if some OST's have sparse indexes.
1085 Severity : enhancement
1087 Description: Update to RHEL5 kernel-2.6.18-53.1.21.el5.
1090 Frequency : rarely, if filesystem is mounted with -o flock
1092 Description: do not process already freed flock
1093 Details : flock can possibly be freed by another thread before it reaches
1094 to ldlm_flock_completion_ast.
1097 Frequency : rarely, if filesystem is mounted with -o flock
1099 Description: LBUG during stress test
1100 Details : Need properly lock accesses the flock deadlock detection list.
1103 Frequency : rarely, if binaries are being run from Lustre
1105 Description: oops in page fault handler
1106 Details : kernel page fault handler can return two special 'pages' in
1107 error case, don't try dereference NOPAGE_SIGBUS and NOPAGE_OMM.
1110 Frequency : rarely, during shutdown
1112 Description: timeout with invalidate import.
1113 Details : ptlrpcd_check call obd_zombie_impexp_cull and wait request which
1114 should be handled by ptlrpcd. This produce long age waiting and
1115 -ETIMEOUT ptlrpc_invalidate_import and as result LASSERT.
1121 Description: ASSERTION(CheckWriteback(page,cmd)) failed
1122 Details : badly clear PG_Writeback bit in ll_ap_completion can produce false
1126 Frequency : only with broken builds/installations
1128 Description: no LBUG if lquota.ko and fsfilt_ldiskfs.ko are different versions
1129 Details : just return an error to a user, put a console error message
1131 Severity : enhancement
1133 Description: Update to RHEL5 kernel-2.6.18-53.1.19.el5.
1135 Severity : enhancement
1137 Description: Update to RHEL4 kernel-2.6.9-67.0.15.
1139 Severity : enhancement
1141 Description: enable MGS and MDT services start separately
1142 Details : add a 'nomgs' option in mount.lustre to enable start a MDT with
1143 a co-located MGS without starting the MGS, which is a complement
1144 to 'nosvc' mount option.
1147 Frequency : always, on big-endian systems
1149 Description: cleanup in ptlrpc code, related to PPC platform
1150 Details : store magic in native order avoid panic's in recovery on PPC
1151 node and forbid from this error in future. Also fix posibily
1152 of twice swab data. Fix get lov striping to userpace.
1155 Frequency : rarely, if replay get lost on server
1157 Description: server incorrectly drop resent replays lead to recovery failure.
1158 Details : do not drop replay according to msg flags, instead we check the
1159 per-export recovery request queue for duplication of transno.
1162 Frequency : after recovery
1164 Description: precreate to many object's after del orphan.
1165 Details : del orphan st in oscc last_id == next_id and this triger growing
1166 count of precreated objects. Set flag LOW to skip increase count
1167 of precreated objects.
1170 Frequency : rare, on clear nid stats
1172 Description: ASSERTION(client_stat->nid_exp_ref_count == 0)
1173 Details : when clean nid stats sometimes try destroy live entry,
1174 and this produce panic in free.
1177 Frequency : occasionally since 1.6.4
1179 Description: Stack overflow during MDS log replay
1180 Details : ease stack pressure by using a thread dealing llog_process.
1183 Frequency : very rare
1185 Description: MDT cannot be unmounted, reporting "Mount still busy"
1186 Details : Mountpoint references were being leaked during open reply
1187 reconstruction after an MDS restart. Drop mountpoint reference
1188 in reconstruct_open() and free dentry reference also.
1193 Description: wait until IO finished before start new when do lock cancel.
1194 Details : VM protocol want old IO finished before start new, in this case
1195 need wait until PG_writeback is cleared until check dirty flag
1196 and call writepages in lock cancel callback.
1201 Description: mds_mfd_close() ASSERTION(rc == 0)
1202 Details : In mds_mfd_close(), we need protect inode's writecount change
1203 within its orphan write semaphore to prevent possible races.
1206 Frequency : rare, on shutdown ost
1208 Description: don't hit live lock with umount ost.
1209 Details : shrink_dcache_parent can be in long loop with destroy dentries,
1210 use shrink_dcache_sb instead.
1213 Frequency : only when echo_client is used
1215 Description: don't panic with use echo_client
1216 Details : echo client pass NULL as client nid pointer and this produce NULL
1217 pointer dereference.
1220 Frequency : Always on 32-bit PowerPC systems
1222 Description: fix build on PPC32
1223 Details : compile code with -m64 flag produce wrong object file for PPC32.
1228 Description: MDS LBUG: ASSERTION(!IS_ERR(dchild))
1229 Details : In reconstruct_* functions, LASSERTs on both the data supplied
1230 by a client, and the data on disk are dangerous and incorrect.
1231 Change them with client eviction.
1233 Severity : enhancement
1235 Description: skiplist implementation simplification
1236 Details : skiplists are used to group compatible locks on granted list
1237 that was implemented as tracking first and last lock of each lock
1238 group the patch changes that to using doubly linked lists
1242 Description: delete compatibility for 32bit qdata
1243 Details : as planned, when lustre is beyond b1_8, lquota won't support 32bit
1244 qunit. That means servers of b1_4 and servers of b1_8 can't be
1245 used together if users want to use quota.
1248 Frequency : only with administrator action
1250 Description: mount failure if config log has invalid conf_param setting
1251 Details : If administrator specified an incorrect configuration parameter
1252 with "lctl conf_param" this would cause an error during future
1253 client mounts. Instead, ignore the bad configuration parameter.
1256 Frequency : blocks per group < blocksize*8 and uninit_groups is enabled
1258 Description: ldiskfs error: XXX blocks in bitmap, YYY in gd
1259 Details : If blocks per group is less than blocksize*8, set rest of the
1263 Frequency : Application do stride read on lustre
1265 Description: The read performance will drop a lot if the application does
1267 Details : Because the stride_start_offset are missing in stride read-ahead,
1268 it will cause clients read a lot of unused pages in read-ahead,
1269 then the read-performance drops.
1273 Description: more ldlm soft lockups
1274 Details : In ldlm_resource_add_lock(), call to ldlm_resource_dump()
1275 starve other threads from the resource lock for a long time in
1276 case of long waiting queue, so change the debug level from
1277 D_OTHER to the less frequently used D_INFO.
1279 Severity : enhancement
1281 Description: add -gid, -group, -uid, -user options to lfs find
1283 Severity : enhancement
1285 Description: ll_recover_lost_found_objs - recover objects in lost+found
1286 Details : OST corruption and subsequent e2fsck can leave objects in the
1287 lost+found directory. Using the "ll_recover_lost_found_objs"
1288 tool, these objects can be retrieved and data can be salvaged
1289 by using the object ID saved in the fid EA on each object.
1294 Description: this bug _only_ happens when inode quota limitation is very low
1295 (less than 12), so that inode quota unit is 1 at initialization.
1296 Details : if remaining quota equates 1, it is a sign to demonstate that quota
1297 is effective now. So least quota qunit should be 2.
1301 Description: Hung threads in invalidate_inode_pages2_range
1302 Details : The direct IO path doesn't call check_rpcs to submit a new RPC once
1303 one is completed. As a result, some RPCs are stuck in the queue
1308 Description: Procfs and llog threads access destoryed import sometimes.
1309 Details : Sync the import destoryed process with procfs and llog threads by
1310 the import refcount and semaphore.
1314 Description: mds fails to respond, threads stuck in ldlm_completion_ast
1315 Details : Sort source/child resource pair after updating child resource.
1320 Description: kernel BUG at ldiskfs2_ext_new_extent_cb
1321 Details : If insertion of an extent fails, then discard the inode
1322 preallocation and free data blocks else it can lead to duplicate
1327 Description: don't always update ctime in ext3_xattr_set_handle()
1328 Details : Current xattr code updates inode ctime in ext3_xattr_set_handle()
1329 In some cases the ctime should not be updated, for example for
1330 2.0->1.8 compatibility it is necessary to delete an xattr and it
1331 should not update the ctime.
1335 Description: add quota statistics
1336 Details : 1. sort out quota proc entries and proc code.
1337 2. add quota statistics
1342 Description: quotas are not honored with O_DIRECT
1343 Details : all writes with the flag O_DIRECT will use grants which leads to
1344 this problem. Now using OBD_BRW_SYNC to guard this.
1348 Bugzilla : 15713/16362
1349 Description: Assertion in iopen_connect_dentry in 1.6.3
1350 Details : looking up an inode via iopen with the wrong generation number can
1351 populate the dcache with a disconneced dentry while the inode
1352 number is in the process of being reallocated. This causes an
1353 assertion failure in iopen since the inode's dentry list contains
1354 both a connected and disconnected dentry.
1358 Description: assertion failure in ldlm_handle2lock()
1359 Details : fix a race between class_handle_unhash() and class_handle2object()
1360 introduced in lustre 1.6.5 by bug 13622.
1362 Severity : enhancement
1364 Description: superblock lock contention with many SMP cores on one client
1365 Details : several client filesystem locks were highly contended on SMP
1366 NUMA systems with 8 or more cores. Per-CPU datastructures
1367 and more efficient locking implemented to reduce contention.
1372 Description: Kernel BUG: sd_iostats_bump: unexpected disk index
1373 Details : remove the limit of 256 scsi disks in the sd_iostat patch
1378 Description: oops in sd_iostats_seq_show()
1379 Details : unloading/reloading the scsi low level driver triggers a kernel
1380 bug when trying to access the sd iostat file.
1385 Description: Kernel panics during QLogic driver reload
1386 Details : REQ_BLOCK_PC requests are not handled properly in the sd iostat
1387 patch, causing memory corruption.
1392 Description: journal_dev option does not work in b1_6
1393 Details : pass mount option during pre-mount.
1395 Severity : enhancement
1397 Description: Add a FIEMAP(FIle Extent MAP) ioctl for ldiskfs
1398 Details : FIEMAP ioctl will allow an application to efficiently fetch the
1399 extent information of a file. It can be used to map logical blocks
1400 in a file to physical blocks in the block device.
1403 Frequency : only with adaptive timeout enabled
1405 Description: DEBUG_REQ() bad paging request
1406 Details : ptlrpc_at_recv_early_reply() should not modify req->rq_repmsg
1407 because it can be accessed by reply_in_callback() without the
1411 Frequency : only on Cray X2
1413 Description: X2 build failures
1414 Details : fix build failures on Cray X2.
1418 Description: xid & resent requests
1419 Details : Initialize RPC XID from clock at startup (randomly if clock is
1424 Description: quota recovery deadlock during mds failover
1425 Details : This patch includes att18982, att18236, att18237 in bz14840.
1427 1. fix osts hang when mds does failover with quotaon
1428 2. prevent watchdog storm when osts threads wait for the
1433 Description: kernel panic on racer
1434 Details : Do not access dchild->d_inode when IS_ERR(dchild) is true.
1436 Severity : enhancement
1438 Description: Add lustre_start utility to start or stop multiple Lustre servers
1443 Description: Lustre GPF in {:ptlrpc:ptlrpc_server_free_request+373}
1444 Details : In case of memory pressure, list_del() can be called twice on
1445 req->rq_history_list, causing a kernel oops.
1449 Description: (ptllnd_peer.c:557:kptllnd_peer_check_sends()) ASSERTION(!in_interrupt()) failed
1450 Details : fix stack overflow in the distributed lock manager by defering
1451 export eviction after a failed ast to the elt thread instead of
1452 handling it in the dlm interpret routine.
1454 Severity : enhancement
1456 Description: More exported tunables for mballoc
1457 Details : Add support for tunable preallocation window and new tunables for
1458 large/small requests
1462 Description: Detect corruption of block bitmap and checking for preallocations
1463 Details : Checks validity of on-disk block bitmap. Also it does better
1464 checking of number of applied preallocations. When corruption is
1465 found, it turns filesystem readonly to prevent further corruptions.
1469 Frequency : only for big-endian servers
1470 Description: Check for big-endian system while mounting fs with extents feature
1471 Details : Mounting a filesystem with extents feature will fail on big-endian
1472 systems since ext3-based ldiskfs is not supported on big-endian
1473 systems. Can be over-ridden with "bigendian_extents" mount option.
1477 Description: Excessive recovery window
1478 Details : With AT enabled, the recovery window can be excessively long (6000+
1479 seconds). To address this problem, we no longer use
1480 OBD_RECOVERY_FACTOR when extending the recovery window (the connect
1481 timeout no longer depends on the service time, it is set to
1482 INITIAL_CONNECT_TIMEOUT now) and clients report the old service
1483 time via pb_service_time.
1486 Descriptoin: Don't sync journal after every i/o
1487 Details : Implement write RPC replay to allow server replies for write RPCs
1488 before data is on disk. However, this feature is disabled by
1489 default since some issues leading to data corruptions have been
1490 found during recovery (e.g. bug 19128). This feature can be enabled
1491 by running the following command on the OSSs:
1492 lctl set_param obdfilter.*.sync_journal=0
1496 Description: Watchdog triggered on MDS failover
1497 Details : enable OBD_CONNECT_MDT flag when connecting from the MDS so that
1498 the OSTs know that the MDS "UUID" can be reused for the same export
1499 from a different NID, so we do not need to wait for the export to
1504 Descriptoin: Lustre detected file system corruption with inode out of bounds
1505 Details : don't update i_size on MDS_CLOSE for directories. This causes
1506 directory corruptions on the MDT.
1508 -------------------------------------------------------------------------------
1510 2008-05-26 Sun Microsystems, Inc.
1512 * Support for kernels:
1513 2.6.5-7.311 (SLES 9),
1514 2.6.9-67.0.7.EL (RHEL 4),
1515 2.6.16.54-0.2.5 (SLES 10),
1516 2.6.18-53.1.14.el5 (RHEL 5),
1517 2.6.22.14 vanilla (kernel.org)
1518 * Client support for unpatched kernels:
1519 (see http://wiki.lustre.org/index.php?title=Patchless_Client)
1520 2.6.16 - 2.6.22 vanilla (kernel.org)
1521 * Due to problems with nested symlinks and FMODE_EXEC (bug 12652),
1522 we do not recommend using patchless RHEL4 clients with kernels
1523 prior to 2.6.9-55EL (RHEL4U5).
1524 * Recommended e2fsprogs version: 1.40.7-sun1
1525 * Note that reiserfs quotas are disabled on SLES 10 in this kernel.
1526 * RHEL 4 and RHEL 5/SLES 10 clients behaves differently on 'cd' to a
1527 removed cwd "./" (refer to Bugzilla 14399).
1528 * A new quota file format has been introduced in 1.6.5.
1529 The format conversion from prior releases is handled transparently,
1530 but releases older than 1.4.12/1.6.5 will not understand this new
1531 format. The automatic format conversion can be avoided by running
1532 the following command on the MDS before upgrading:
1533 'tunefs.lustre --param="mdt.quota_type=ug1" $MDTDEV'.
1534 For more information, please refer to bugzilla 13904.
1538 Description: quota performance fix
1539 Details : quota data is written in journalled mode instead of ordered to
1540 increase performance
1544 Description: lfs support for human-readable quota grace time strings
1545 Details : lfs setquota -t and lfs quota -t represent quota grace times
1546 in "XXwXXdXXhXXmXXs" format instead of large values in seconds
1549 Frequency : always with o2ib 1.3 and sles10
1551 Description: fix build with SLES10 and o2ib v3.
1552 Details : sles10 uses diffrent name for Module.symver file but configure
1553 assume this file has same name on RHEL/SLES/vanila kernels.
1556 Frequency : very rare, if additional xattrs are used on kernels >= 2.6.12
1558 Description: MDS may lose file striping (and hence file data) in some cases
1559 Details : If there are additional extended attributes stored on the MDS,
1560 in particular ACLs, SELinux, or user attributes (if user_xattr
1561 is specified for the client mount options) then there is a risk
1562 of attribute loss. Additionally, the Lustre file striping
1563 needs to be larger than default (e.g. striped over all OSTs),
1564 and an additional attribute must be stored initially in the
1565 inode and then increase in size enough to be moved to the
1566 external attribute block (e.g. ACL growing in size) for file
1569 Severity : enhancement
1571 Description: add message levels for liblustreapi
1576 Description: MDT cannot be unmounted, reporting "Mount still busy"
1577 Details : Mountpoint references were being leaked during open reply
1578 reconstruction after an MDS restart. Drop mountpoint reference
1579 in reconstruct_open() and free dentry reference also.
1584 Description: fix for occasional failure case of -ENOSPC in recovery-small tests
1585 Details : Move the 'good_osts' check before the 'total_bavail' check. This
1586 will result in an -EAGAIN and in the exit call path we call
1587 alloc_rr() which will with increasing aggressiveness attempt to
1588 aquire precreated objects on the minimum number of required OSCs.
1592 Description: Use old size assignment to avoid deadlock
1593 Details : This reverts the changes in bugs 2369 and bug 14138 that introduced
1594 the scheduling while holding a spinlock. We do not need locking
1595 for size in ll_update_inode() because size is only updated from
1596 the MDS for directories or files without objects, so there is no
1597 other place to do the update, and concurrent access to such inodes
1598 are protected by the inode lock.
1602 Description: Use __u64 instead of int for valid bits
1606 Description: resolve "_IOWR redefined" build error on SLES10
1610 Description: dump the memory debugging after all modules are unloaded to
1611 suppress false negative in conf_sanity test 39
1615 Description: the recovery timer never expires
1616 Details : for new client connect request, the recovery timer should not be
1617 reset, otherwise recovery timer will never expired, if the old
1618 client never come. Only old client connect and first connection
1619 req should trigger recovery timer reset.
1623 Description: the min numbers of lproc stats are wrong
1624 Details : adding a new constant LC_MIN_INIT and use it for initialization
1628 Frequency : always with interactive lfs
1630 Description: Reinitialize optind to 0 so that interactive lfs works in all cases
1633 Frequency : with multiple concurrent readdir processes in same directory
1634 Bugzilla : 15406, 15169, 15175
1635 Description: misc fixes for directory readahead.
1636 Details : prevent previous statahead async RPC callback from processing the
1637 current "statahead_info", race condition between async RPC callback
1638 add dentry into dentry hash table and "ls" thread revalidate such
1639 dentry, statahead his/miss control for hidden items, and so on.
1641 Severity : enhancement
1643 Description: build kernel-ib packages for OFED 1.3 in our release cycle
1647 Description: incore types cleaning in quota code (with respect to 64-bit limits)
1648 Details : several u32 variables declarations are replaced with u64 declarations
1653 Description: fix SLES kernel versioning
1654 Details : the kernel version for our SLES 10 kernel did not include a "-"
1655 before the "smp" at the end. while this was not a problem in
1656 general, it did mean that software trying to use the kernel
1657 version to try to detect a vendor specific kernel would fail.
1658 this was most evident by the OFED build scripts.
1663 Description: Don't update lov_desc members until making sure they are valid
1664 Details : When updating lov_desc members via proc fs, need fix their
1665 validities before doing the real update.
1668 Frequency : very rare
1670 Description: don't put request into delay list while invalidate in flight.
1671 Details : ptlrpc_delay_request sometimes put in delay list while invalidate
1672 import in flight. this produce timeout for invalidate and sometimes
1673 can cause stale data.
1675 Severity : enhancement
1677 Description: Update kernel to SLES9 2.6.5-7.311.
1679 Severity : enhancement
1681 Description: Update kernel to RHEL4 2.6.9-67.0.7.
1686 Frequency : on PPC only
1687 Description: not convert ost objects for directory because it's not exist.
1688 Details : ll_dir_getstripe assume dirrectory has ost objects but this wrong.
1690 Severity : enhancement
1692 Description: Fix warnings with compile liblustre at sles10/rhel5 which have
1693 __u64 as usingied long long type.
1696 Frequency : rare, on shutdown
1698 Description: race process ast vs remove callback
1699 Details : removing callback before disconnect import open race with
1700 processing callback.
1702 Severity : enhancement
1704 Description: Update kernel to SLES9 2.6.5-7.311.
1706 Severity : enhancement
1708 Description: Files open for execute are not marked busy on SLES10
1709 Details : Add FMODE_EXEC to SLES10 SP1 server kernel series.
1711 Severity : enhancement
1713 Description: Add server support for vanilla-2.6.22.14.
1716 Frequency : occasional
1718 Description: Avoid lov_create() getting stuck in obd_statfs_rqset()
1719 Details : If an OST is down the MDS will hang indefinitely in
1720 obd_statfs_rqset() waiting for the statfs data. While for
1721 MDS QOS usage of statfs, it should not stuck in waiting.
1723 Severity : enhancement
1725 Description: Disable adaptive timeouts by default
1728 Frequency : on network error
1730 Description: panic with double free request if network error
1731 Details : mdc_finish_enqueue is finish request if any network error occuring,
1732 but it's true only for synchronus enqueue, for async enqueue
1733 (via ptlrpcd) this incorrect and ptlrpcd want finish request
1737 Frequency : rare, on recovery
1739 Description: read procfs can produce deadlock in some situation
1740 Details : Holding lprocfs lock with send rpc can produce block for destroy
1741 obd objects and this also block reconnect with -EALREADY.
1742 This isn't fix all lprocfs bugs - but make it rare.
1744 Severity : enhancement
1746 Description: Update kernel to RHEL5 2.6.18-53.1.14.el5.
1749 Frequency : frequent on X2 node
1751 Description: mdc_set_open_replay_data LBUG
1752 Details : Set replay data for requests that are eligible for replay.
1757 Description: lustre_mgs: operation 101 on unconnected MGS
1758 Details : When MGC is disconnected from MGS long enough, MGS will evict the
1759 MGC, and late on MGC cannot successfully connect to MGS and a lot
1760 of the error messages complaining that MGS is not connected.
1763 Frequency : rare, depends on device drivers and load
1765 Description: MDS or OSS nodes crash due to stack overflow
1766 Details : Code changes in 1.6.4 increased the stack usage of some functions.
1767 In some cases, in conjunction with device drivers that use a lot
1768 of stack, the MDS (or possibly OSS) service threads could overflow
1769 the stack. One change which was identified to consume additional
1770 stack has been reworked to avoid the extra stack usage.
1772 Severity : enhancement
1774 Description: Update to RHEL5 latest kernel-2.6.18-53.1.13.el5.
1776 Severity : enhancement
1778 Description: Update to SLES10 SP1 latest kernel-2.6.16.54-0.2.5.
1780 Severity : enhancement
1782 Description: Update to RHEL5 latest kernel-2.6.18-53.1.6.el5.
1784 Serverity : enhancement
1786 Description: Update RHEL4 kernel to 2.6.9-67.0.4.
1789 Frequency : rare on shutdown OST
1791 Description: Don't allow skipping OSTs if index has been specified.
1792 Details : Don't allow skipping OSTs if index has been specified, make locking
1793 in internal create lots better.
1798 Description: ASSERTION(!PageDirty(page)) failed
1799 Details : Wrong check could lead to an assertion failure under specific
1805 Description: LBUG in ptlrpc_check_set() bad phase ebc0de00
1806 Details : access to bitfield in structure is always rounded to long
1807 and this produce problem with not atomic change any bit.
1812 Description: Lustre make rpms failed.
1813 Details : Remove ldiskfs spec file to avoids rpmbuild be confused when
1814 builds Lustre rpms from tarball.
1816 Severity : enhancement
1818 Description: Update to SLES9 SP4 kernel-2.6.5-7.308.
1821 Frequency : rare on shutdown OST
1823 Description: If llog cancel was not send before clean_exports phase, this can
1824 produce deadlock in llog code.
1825 Details : If llog thread has last reference to obd and call class_import_put
1826 this produce deadlock because llog_cleanup_commit_master wait when
1827 last llog_commit_thread exited, but this never success because was
1828 called from llog_commit_thread.
1831 Frequency : only if OST index is skipped
1833 Description: NULL lov_tgts causing MDS oops
1834 Details : more safe checks for NULL lov_tgts for avoid oops.
1836 Severity : enhancement
1838 Description: Update to RHEL4 latest kernel-2.6.9-67.0.1.EL.
1840 Severity : enhancement
1842 Description: Update to RHEL5 latest kernel-2.6.18-53.1.4.el5.
1847 Description: make mgs_setparam() handle fsname containing dash
1848 Details : fsname containing a dash does not work with lctl conf_param
1850 Severity : enhancement
1852 Description: Update to RHEL4 Update-6 kernel-2.6.9-67.EL.
1855 Frequency : rare, in recovery and (or) destroy lovobjid file.
1857 Description: rewrite lov objid code.
1858 Details : Cleanup for lov objid code, remove scability problems and wrong
1859 locking. Fix sending last_id into ost.
1861 Severity : enhancement
1863 Description: Update to SLES10 SP1 latest kernel-2.6.16.54-0.2.3.
1865 Severity : enhancement
1867 Description: Update to RHEL5 Update-1 kernel 2.6.18-53.el5.
1868 Details : Use d_move_locked instead of __d_move.
1871 Frequency : rare, at shutdown
1873 Description: access already free / zero obd_namespace.
1874 Details : if client_disconnect_export was called without force flag set,
1875 and exist connect request in flight, this can produce access to
1876 NULL pointer (or already free pointer) when connect_interpret
1877 store ocd flags in obd_namespace.
1880 Frequency : only at startup
1882 Description: not alloc memory with spinlock held.
1883 Details : allocation memory with GFP_KERNEL can produce sleep deadlock,
1884 if any spinlock held.
1889 Description: lfs find does not continue on file error
1890 Details : Continue other files processing when a file/dir is absent.
1894 Description: Inconsistent usage of lustre_pack_reply()
1895 Details : Standardize the usage of lustre_pack_reply() such that it
1896 always generate a CERROR on failure.
1899 Frequency : very rare
1901 Description: Fix replay if there is an un-replied request and open
1902 Details : In some cases, older replay request will revert the
1903 mcd->mcd_last_xid on MDS which is used to record the client's
1904 latest sent request.
1906 Severity : enhancement
1908 Description: Update to RHEL5 kernel 2.6.18-8.1.15.el5.
1910 Severity : enhancement
1912 Description: Update to SLES10 SP1 kernel 2.6.16.53-0.16
1914 Severity : enhancement
1916 Description: Update to SLES9 kernel-2.6.5-7.287.3.
1918 Severity : enhancement
1920 Description: Update to RHEL4 kernel-2.6.9-55.0.12.EL.
1922 Severity : enhancement
1924 Description: Build SLES10 patchless client fails
1925 Details : The configure was broken by run ./configure with
1926 --with-linux-obj=.... argument for patchless client. When the
1927 configure use --with-linux-obj, the LINUXINCLUDE= -Iinclude
1928 can't search header adequately. Use absolute path such as
1929 -I($LINUX)/include instead.
1931 Severity : enhancement
1933 Description: Lustre Page Accounting
1934 Details : New macros for page alloc and free which enable accounting
1935 of page allocation of Lustre. Use percpu counters to store memory
1936 and page statistics.
1939 Frequency : only if debugging is disabled
1941 Description: LASSERT_{REQ,REP}SWAB macros are buggy
1942 Details : If SWAB_PARANOIA is disabled, the LASSERT_REQSWAB and
1943 LASSERT_REPSWAB macros become no-ops, which is incorrect. Drop
1944 these macros and replace them with their definitions instead.
1949 Description: interrupt oig_wait produce painc on resend.
1950 Details : brw_redo_request can be used for resend requests from ptlrpcd and
1951 private set, and this produce situation when rq_ptlrpcd_data not
1952 copyed to new allocated request and triggered LBUG on assert
1953 req->rq_ptlrpcd_data != NULL. But this member used only for wakeup
1954 ptlrpcd set if request is changed and can be safety changed to use
1957 Severity : enhancement
1959 Description: organize the server-side client stats on per-nid basis
1960 Details : Change the structure of stats under obdfilter and mds to
1968 The "uuid"s file would list the uuids of _active_ exports.
1969 And the clear entry is to clear all stats and stale nids.
1974 Description: Processes looping in ll_readdir() on Lustre clients finally causing
1975 a full node pseudo-hang
1976 Details : Concurrent access to the same directory from multiple clients with
1977 intensive file creation/removal can cause a client node to spin in
1978 ll_readdir(). i_version must be increased every time the lock is
1979 cancelled to ensure a revalidate is done.
1984 Description: touch file failed when fs is not full
1985 Details : OST in recovery should not be discarded by MDS in alloc_qos(),
1986 otherwise we can get ENOSP while fs is not full.
1989 Frequency : only for Cray XT3
1990 Bugzilla : 12829/13455
1991 Description: Changing primary group doesn't change the group lustre assigns to
1993 Details : When CRAY_XT3 is defined, the fsgid supplied by the client is
1994 overridden with the primary group provided by the group upcall,
1995 whereas the supplied fsgid can be trusted if it is in the list of
1996 supplementary groups returned by the group upcall.
1998 Severity : enhancement
2000 Description: Root Squash Functionality
2001 Details : Implementation of NFS-like root squash capability. Specifically,
2002 don't allow someone with root access on a client node to be able
2003 to manipulate files owned by root on a server node.
2005 Severity : enhancement
2007 Description: Slow trucate/writes to huge files at high offsets.
2008 Details : Directly associate cached pages to lock that protect those pages,
2009 this allows us to quickly find what pages to write and remove
2010 once lock callback is received.
2015 Description: Too many locks accumulating on client during NFS usage
2016 Details : mds_open improperly used accmode to find out access mode to a
2017 file. Also mdc_intent_lock logic to find out if we already have
2018 lock similar to just received was flawed since introduction of
2019 skiplists - locks are now added to the front of the granted
2024 Description: Hit ASSERTION(obd->obd_stopping == 1) failed in some setup failed
2026 Details : In obd setup failure handler, obd_stopping will not necessarily to
2027 be 1, and obd_set_up should also be checked to make sure whether
2028 obd is completely setup.
2030 Severity : enhancement
2032 Description: Allow masking D_WARNING, D_ERROR messages from console
2033 Details : Console messages can now be disabled via lnet.printk.
2038 Description: User code with malformed file open parameter crashes client node
2039 Details : Before packing join_file req, all the related reference should be
2040 checked carefully in case some malformed flags cause fake join_file
2046 Description: shrink/enlarge qunit size when needed; fix the problem of coarse
2047 grain of quota doing harm to quota's accuracy
2048 Details : qunit size will be changed when quota limitation is too low/high;
2049 record the pending quota write in order to get more accureate
2050 quota; delete the patch for bug12588, which is unnecessary when
2051 this patch is landed. This bug also contains fixes for bug 14526,
2052 14299, 14601 and 13794.
2056 Description: LDLM_ENQUEUE races with LDLM_CP_CALLBACK
2057 Details : ldlm_completion_ast() assumes that a lock is granted when the req
2058 mode is equal to the granted mode. However, it should also check
2059 that LDLM_FL_CP_REQD is not set.
2063 Description: Heavy nfs access might result in deadlocks
2064 Details : After ELC code landed, it is now improper to enqueue any mds
2065 locks under och_sem, because enqueue might want to decide to
2066 cancel open locks for same inode we are holding och_sem for.
2070 Description: 35% write performance drop with ldiskfs2 when quotas are on
2071 Details : Enable ext3 journalled quota by default to improve performance
2072 when quotas are turned on.
2076 Description: Client eviction while running blogbench
2077 Details : A lot of unlink operations with concurrent I/O can lead to a
2078 deadlock causing evictions. To address the problem, the number of
2079 oustanding OST_DESTROY requests is now throttled to
2080 max_rpcs_in_flight per OSC and LDLM_FL_DISCARD_DATA blocking
2081 callbacks are processed in priority.
2084 Frequency : RHEL4 only
2086 Description: mkfs is very slow on IA64/RHEL4
2087 Details : A performance regression has been discovered in the MPT Fusion
2088 driver between versions 3.02.73rh and 3.02.99.00rh. As a
2089 consequence, we have downgraded the MPT Fusion driver in the RHEL4
2090 kernel from 3.02.99.00 to 3.02.73 until this problem is fixed.
2093 Frequency : PPC/PPC64 only
2095 Description: conflicts between asm-ppc64/types.h and lustre_types.h
2096 Details : fix duplicated definitions between asm-ppc64/types.h and
2097 lustre_types.h on PPC.
2100 Frequency : PPC/PPC64 only
2102 Description: asm-ppc/segment.h does not exist
2103 Details : fix compile issue on PPC.
2107 Description: data checksumming impacts single node performance
2108 Details : add support for several checksum algorithms. Currently, CRC32 and
2109 Adler-32 are supported. The checksum type can be changed on the fly
2110 through /proc/fs/lustre/osc/*/checksum_type.
2114 Description: use adler32 for page checksums
2115 Details : when available, use the Adler-32 algorithm instead of CRC32 for
2120 Description: better handle error messages in extents code
2122 Severity : enhancement
2124 Description: SNMP support enhancement
2125 Details : Adding total number of sampled request for an MDS node in snmp
2128 Severity : enhancement
2130 Description: Optimize ldlm waiting list processing for PR extent locks
2131 Details : When processing waiting list for read extent lock and meeting read
2132 lock that is same or wider to it that is not contended, skip
2133 processing rest of the list and immediatelly return current
2134 status of conflictness, since we are guaranteed there are no
2135 conflicting locks in the rest of the list.
2139 Description: Time out and refuse to reconnect
2140 Details : When the failover node is the primary node, it is possible
2141 to have two identical connections in imp_conn_list. We must
2142 compare not conn's pointers but NIDs, otherwise we
2143 can defeat connection throttling.
2147 Description: Client not clear own cache if answer to reconnect is lost.
2148 Details : Client gets evicted from server. Now client also thinks it is
2149 disconnected (or gets ENOTCONN on its operation) and decides to
2150 reconnect. Server receives reconnect message, but cannot find
2151 export. New export is created that is fully valid (new cookie!),
2152 but reply is lost and not reported to client. Client reconnects
2153 again and gets back a just-created connection, but it is not new
2154 so client thinks it was not evicted and does not do recovery.
2158 Description: Detect stride IO mode in read-ahead
2159 Details : When a client does stride read, read-ahead should detect that and
2160 read-ahead pages according to the detected stride pattern.
2164 Description: build for x2 fails
2165 Details : fix compile issue on Cray systems.
2167 Severity : enhancement
2169 Description: implement readv/writev APIs(aio_read/aio_writes in newer kernels)
2170 Details : This greatly improves speed of NFS writes on 2.6 kernels.
2173 Frequency : only on PPC/SLES10
2175 Description: "BITS_PER_LONG is not 32 or 64" in linux/idr.h
2176 Details : On SLES10/PPC, fs.h includes idr.h which requires BITS_PER_LONG to
2177 be defined. Add a hack in mkfs_lustre.c to work around this compile
2182 Description: LASSERT on MDS when client holding flock lock dies
2183 Details : ldlm pool logic depends on number of granted locks equal to
2184 number of released locks which is not true for flock locks, so
2185 just exclude such locks from consideration.
2189 Description: MDS deadlock with many ll_sync_lov threads and I/O stalled
2190 Details : Use fsfilt_sync() for both the whole filesystem sync and
2191 individual file sync to eliminate dangerous inode locking
2192 with I_LOCK that can lead to a deadlock.
2196 Description: Update an obsolete wirecheck.c generator
2197 Details : Update wirecheck.c/wirehdr.c and regenerate wiretest.c
2201 Description: Client can panic on open sometimes
2202 Details : It is possible that we try to free already freed request in
2203 ll_file_open in some error cases when we send request from
2208 Description: performance in 1.6.3
2209 Details : Force q->max_phys_segments to MAX_PHYS_SEGMENTS on SLES10 to be
2210 sure that 1MB requests are not fragmented by the block layer.
2214 Description: LDLM soft lockups - improvement
2215 Details : It is be possible to send the lock handle along with each read
2216 or write request because the client is already doing a lock match
2217 itself so there isn't any reason the OST should have to re-do that
2223 Description: lfs quota fails with deactivated OSTS
2224 Details : With this patch, three improvements are included:
2225 1. detete the softlimit in mds and osts when use "lfs quota".
2226 2. display the inaccurate data in the output of "lfs quota".
2227 3. try to get quota info when "lfs quota" is executed.
2232 Description: Extent locks not granted with no conflicts sometimes.
2233 Details : When race occurs in glimpse handler and nothing is returned,
2234 we do not reprocess the queue after lock cancel, and that leads
2235 to a stall until next activity on a resource
2238 Frequency : failover with quotaon
2240 Description: during mds failovers with quota on, OSTs got into deadlock state
2241 and causing dumpstack.
2242 Details : for every quota slave, at any time, there is only one quota req
2243 is sent to quota master for every uid/gid. Before that quota req
2244 returns, all the thread relative to the same uid/gid will wait.
2245 So if the quota req is lost because mds failovers or any other
2246 reasons, this bug will be hit. Now, dqacq_interpret() will handle
2247 quota reqs who time out.
2249 Severity : enhancement
2252 Description: when quota slave checks if quota is enough, there is an unnecessary
2254 Details : place this wait on necessary place instead of always waiting.
2256 --------------------------------------------------------------------------------
2258 2007-12-07 Cluster File Systems, Inc. <info@clusterfs.com>
2260 * Support for kernels:
2261 2.6.5-7.286 (SLES 9),
2262 2.6.9-55.0.9.EL (RHEL 4),
2263 2.6.16.53-0.8 (SLES 10),
2264 2.6.18-8.1.14.el5 (RHEL 5),
2265 2.6.18.8 vanilla (kernel.org)
2266 * Client support for unpatched kernels:
2267 (see http://wiki.lustre.org/index.php?title=Patchless_Client)
2268 2.6.16 - 2.6.22 vanilla (kernel.org)
2269 * Due to recently discovered recovery problems, we do not recommend
2270 using patchless RHEL 4 clients with this or any earlier release.
2271 * Recommended e2fsprogs version: 1.40.2-cfs1
2272 * Note that reiserfs quotas are disabled on SLES 10 in this kernel.
2275 Frequency : occasional
2277 Description: MDS hang or stay in waiting lock
2278 Details : If client receive lock with CBPENDING flag ldlm need send lock
2279 cancel as separate rpc, to avoid situation when cancel request
2280 can't processed due all i/o threads stay in wait lock.
2283 Frequency : occasional
2285 Description: Do not fail import if osc_interpret_create gets -EAGAIN
2286 Details : If osc_interpret_create got -EAGAIN it immediately exits and
2287 wakeup oscc_waitq. After wakeup oscc_wait_for_objects call
2288 oscc_has_objects and see OSC has no objests and call
2289 oscc_internal_create to resend create request.
2291 Severity : enhancement
2293 Description: Update kernel patches for SLES10 2.6.16.53-0.8.
2294 Details : Update which_patch & target file for SLES10 latest kernel.
2296 Severity : enhancement
2298 Description: add --type and --size parameters to lfs find
2299 Details : Enhance lfs find by adding filetype and filesize parameters. Also
2300 multiple OBDs can now be specified for the --obd option.
2302 Severity : enhancement
2304 Description: eliminate client locks in face of contention
2305 Details : file contention detection and lockless i/o implementation
2306 for contended files.
2308 Severity : enhancement
2310 Description: Remove client patches from SLES 10 kernel.
2311 Details : This causes SLES 10 clients to behave as patchless clients
2312 even on a Lustre-patched (server) kernel.
2314 Severity : enhancement
2316 Description: use i_size_read and i_size_write in 2.6 port
2317 Details : replace inode->i_size access with i_size_read/write()
2320 Frequency : when removing large files
2322 Description: scheduling issue during removal of large Lustre files
2323 Details : Don't take the BKL in fsfilt_ext3_setattr() for 2.6 kernels.
2324 It causes scheduling issues when removing large files (17TB in the
2330 Description: 1.4.11 Can't handle directories with stripe set and extended ACLs
2331 Details : Impossible (EPROTO is returned) to access a directory that has a
2332 non-default striping and ACLs.
2335 Frequency : only on ppc
2337 Description: /proc/fs/lustre/devices broken on ppc
2338 Details : The patch as applied to 1.6.2 doesn't look correct for all arches.
2339 We should make sure the type of 'index' is loff_t and then cast
2340 explicitly as needed below. Do not assign an explicitly cast
2344 Frequency : only for rhel5
2346 Description: Kernel patches update for RHEL5 2.6.18-8.1.10.el5.
2347 Details : Modify the target file & which_kernel.
2350 Frequency : if the uninit_groups feature is enabled on ldiskfs
2352 Description: e2fsck reports "invalid unused inodes count"
2353 Details : If a new ldiskfs filesystem is created with the "uninit_groups"
2354 feature and only a single inode is created in a group then the
2355 "bg_unused_inodes" count is incorrectly updated. Creating a
2356 second inode in that group would update it correctly.
2361 Description: buffer overruns could theoretically occur
2362 Details : llapi_semantic_traverse() modifies the "path" argument by
2363 appending values to the end of the origin string, and a buffer
2364 overrun may occur. Adding buffer overrun check in liblustreapi.
2366 Severity : enhancement
2368 Description: Add jbd statistics patch for RHEL5 and 2.6.18-vanilla.
2371 Frequency : only if filesystem is inconsistent
2373 Description: handle "serious error: objid * already exists" more gracefully
2374 Details : If LAST_ID value on disk is smaller than the objects existing in
2375 the O/0/d* directories, it indicates disk corruption and causes an
2376 LBUG(). If the object is 0-length, then we should use the existing
2377 object. This will help to avoid a full fsck in most cases.
2379 Severity : enhancement
2381 Description: Kernel patches update for RHEL4 2.6.9-55.0.6.
2382 Details : Modify vm-tunables-rhel4.patch.
2384 Severity : enhancement
2386 Description: Kernel config for 2.6.18-vanilla.
2387 Details : Modify targets/2.6-vanilla.target.in.
2388 Add config file kernel-2.6.18-2.6-vanilla-i686.config.
2389 Add config file kernel-2.6.18-2.6-vanilla-i686-smp.config.
2390 Add config file kernel-2.6.18-2.6-vanilla-x86_64.config.
2391 Add config file kernel-2.6.18-2.6-vanilla-x86_64-smp.config.
2394 Frequency : occasional
2396 Description: improve handling recoverable errors
2397 Details : If request processed with error which can be recoverable on server
2398 request should be resend, otherwise page released from cache and
2404 Description: Kernel patches update for RHEL5 2.6.18-8.1.14.el5.
2405 Details : Modify target file & which_patch.
2406 A flaw was found in the IA32 system call emulation provided
2407 on AMD64 and Intel 64 platforms. An improperly validated 64-bit
2408 value could be stored in the %RAX register, which could trigger an
2409 out-of-bounds system call table access. An untrusted local user
2410 could exploit this flaw to run code in the kernel
2411 (ie a root privilege escalation). (CVE-2007-4573).
2415 Description: change order of libsysio includes
2416 Details : '#include sysio.h' should always come before '#include xtio.h'
2418 Severity : enhancement
2420 Description: adapt the lustre_config script to support the upgrade case
2421 Details : Add "-u" option for lustre_config script to support upgrading 1.4
2422 server targets to 1.6 in parallel.
2427 Description: To avoid grant space > avaible space when the disk is almost
2428 full. Without this patch you might see the error "grant XXXX >
2429 available" or some LBUG about grant, when the disk is almost
2431 Details : In filter_check_grant, for non_grant cache write, we should
2432 check the left space by if (*left > ungranted + bytes), instead
2433 of (*left > ungranted), because only we are sure the left space
2434 is enough for another "bytes", then the ungrant space should be
2435 increase. In client, we should update cl_avail_grant only there is
2436 OBD_MD_FLGRANT in the reply.
2441 Description: Update RHEL 4 kernel to fix local root privilege escalation.
2442 Details : Update to the latest RHEL 4 kernel to fix the vulnerability
2443 described in CVE-2007-4573. This problem could allow untrusted
2444 local users to gain root access.
2447 Frequency : when using O_DIRECT and quotas
2449 Description: Incorrect file ownership on O_DIRECT output files
2450 Details : block usage reported by 'lfs quota' does not take into account
2451 files that have been written with O_DIRECT.
2455 Description: (rw.c:1323:ll_read_ahead_pages()) ASSERTION(page_idx > ria->ria_stoff) failed
2456 Details : Once the unmatched stride IO mode is detected, shrink the stride-ahead
2457 window to 0. If it does hit cache miss, and read-pattern is still
2458 stride-io mode, does not reset the stride window, but also does not
2459 increase the stride window length in this case.
2461 --------------------------------------------------------------------------------
2463 2007-09-27 Cluster File Systems, Inc. <info@clusterfs.com>
2465 * Support for kernels:
2466 2.6.5-7.286 (SLES 9),
2467 2.6.9-55.0.2.EL (RHEL 4),
2468 2.6.16.46-0.14 (SLES 10),
2469 2.6.18-8.1.8.el5 (RHEL 5),
2470 2.6.18.8 vanilla (kernel.org)
2471 * Client support for unpatched kernels:
2472 (see http://wiki.lustre.org/index.php?title=Patchless_Client)
2473 2.6.16 - 2.6.21 vanilla (kernel.org)
2474 * Due to recently discovered recovery problems, we do not recommend
2475 using patchless RHEL 4 clients with this or any earlier release.
2476 * Recommended e2fsprogs version: 1.40.2-cfs1
2477 * Note that reiserfs quotas are disabled on SLES 10 in this kernel.
2481 Description: Fix errors in lfs documentation
2482 Details : Fixes man pages
2484 Severity : enhancement
2486 Description: Adaptive timeouts
2487 Details : RPC timeouts adapt to changing server load and network
2488 conditions to reduce resend attempts and improve recovery time.
2490 Severity : enhancement
2492 Description: llapi_file_create() does not allow some changes
2493 Details : add llapi_file_open() that allows specifying the file creation
2494 mode and open flags, and also returns an open file handle.
2496 Severity : enhancement
2498 Description: df doesn't work properly if diskfs blocksize != 4K
2499 Details : Choose biggest blocksize of OST's as the LOV's blocksize.
2502 Frequency : always on directories with default striping set
2504 Description: lfs find on -1 stripe looping in lsm_lmm_verify_common()
2505 Details : Avoid lov_verify_lmm_common() on directory with -1 stripe count.
2507 Severity : enhancement
2509 Description: merge and cleanup kernel patches.
2510 Details : Remove mnt_lustre_list in vfs_intent-2.6-rhel4.patch.
2512 Severity : enhancement
2514 Description: RedHat Update kernel for RHEL5
2515 Details : Kernel config file for RHEL5.
2517 Severity : enhancement
2519 Description: OSS needs mutliple precreate threads
2520 Details : Add ability to start more than one create thread per OSS.
2523 Frequency : only with quota on the root user
2525 Description: mds_obd_create error creating tmp object
2526 Details : When the user sets quota on root, llog will be affected and can't
2527 create files and write files.
2530 Frequency : Always on ia64 patchless client, and possibly others.
2532 Description: Add EXPORT_SYMBOL check for node_to_cpumask symbol.
2533 Details : This allows the patchless client to be loaded on architectures
2534 without this export.
2539 Description: disorder of journal start and llog_add cause deadlock.
2540 Details : in llog_origin_connect, journal start should happen before
2541 llog_add keep the same order as other functions to avoid
2544 Severity : enhancement
2546 Description: RedHat Update kernel for RHEL5
2547 Details : Modify the kernel config file more closer RHEL5.
2549 Severity : enhancement
2551 Description: Build failure against Centos5 (RHEL5)
2552 Details : Define PAGE_SIZE when it isn't present.
2555 Frequency : occasionally when using NFS
2557 Description: "ll_intent_file_open()) lock enqueue: err: -13" with nfs
2558 Details : with NFS, the anon dentry's parent was set to itself in
2559 d_alloc_anon(), so in MDS, we use rec->ur_fid1 to find the
2560 corresponding dentry other than use rec->ur_name.
2562 Severity : enhancement
2564 Description: enable data checksumming by default
2565 Details : enable checksum by default, allow --disable-checksum
2566 configure option and "-o nochecksum" mount option. Checksums
2567 can also be disabled at runtime via $LPROC/osc/*/checksum_pages.
2570 Frequency : Occasionally with failover
2572 Description: Client eviction due to failover config
2573 Details : after a connection loss, the lustre client should attempt to
2574 reconnect to the last active server first before trying the
2575 other potential connections.
2577 Severity : enhancement
2579 Description: client-side metadata stat-ahead during readdir(directory readahead)
2580 Details : perform client-side metadata stat-ahead when the client detects
2581 readdir and sequential stat of dir entries therein
2586 Description: when mds and osts use different quota unit(32bit and 64bit),
2587 quota will be released repeatly.
2588 Details : void sending multiple quota reqs to mds, which will keep the
2589 status between the reqs.
2592 Frequency : only with liblustre clients on XT3
2594 Description: evictions taking too long
2595 Details : allow llrd to evict clients directly on OSTs
2599 Description: osts not allocated evenly to files
2600 Details : change the condition to increase offset_idx
2602 Severity : enhancement
2604 Description: self-adjustable client's lru lists
2605 Details : use adaptive algorithm for managing client cached locks lru
2606 lists according to current server load, other client's work
2607 pattern, memory activities, etc. Both, server and client
2608 side namespaces provide number of proc tunables for controlling
2613 Description: rewrite ext2-derived code in llite/dir.c and obdclass/uuid.c
2614 Details : rewrite inherited code (uuid parsing code from ext2 utils and
2615 readdir code from ext3) from scratch preserving functionality.
2619 Description: Only those disconnect error should be returned by rq_status.
2620 Details : In open/enqueue processs, Some errors, which will cause client
2621 disconnected, should be returned by rq_status, while other
2622 errors should still be returned by intent, then mdc or llite will
2625 Severity : enhancement
2627 Description: Tune the kernel for good SCSI performance.
2628 Details : Set the value of /sys/block/{dev}/queue/max_sectors_kb
2629 to the value of /sys/block/{dev}/queue/max_hw_sectors_kb
2633 Frequency : Always for filesystems larger than 2TB on 32-bit systems.
2634 Bugzilla : 13547, 13627
2635 Description: Data corruption for OSTs that are formatted larger than 2TB
2637 Details : When generating the bio request for lustre file writes the
2638 sector number would overflow a temporary variable before being
2639 used for the IO. The data reads correctly from Lustre (which
2640 will overflow in a similar manner) but other file data or
2641 filesystem metadata may be corrupted in some cases.
2645 Description: "lfs find -obd UUID" prints directories
2646 Details : "lfs find -obd UUID" will return all directory names instead
2647 of just file names. It is incorrect because the directories
2648 do not reside on the OSTs.
2651 Bugzilla : 13596, 13614
2652 Description: MDS hang after unclean shutdown of lots of clients
2653 Details : Never resend AST requests.
2657 Frequency : Always, for kernels after 2.6.16
2658 Description: Fix warning idr_remove called for id=.. which is not allocated.
2659 Details : Last kernels save old s_dev before kill super and not allow
2660 to restore from callback - restore it before call kill_anon_super.
2664 Description: Fix error on 'ls .' at the top of the Lustre mount.
2665 Details : Don't revalidate dentry if it is a root dentry.
2667 --------------------------------------------------------------------------------
2669 2007-08-27 Cluster File Systems, Inc. <info@clusterfs.com>
2671 * Support for kernels:
2672 2.6.5-7.286 (SLES 9),
2673 2.6.9-55.0.2.EL (RHEL 4),
2674 2.6.16.46-0.14 (SLES 10),
2675 2.6.18-8.1.8.el5 (RHEL 5),
2676 2.6.18.8 vanilla (kernel.org)
2677 * Client support for unpatched kernels:
2678 (see http://wiki.lustre.org/index.php?title=Patchless_Client)
2679 2.6.16 - 2.6.22 vanilla (kernel.org)
2680 * Due to recently discovered recovery problems, we do not recommend
2681 using patchless RHEL 4 clients with this or any earlier release.
2682 * Recommended e2fsprogs version: 1.39.cfs8
2683 * Note that reiserfs quotas are disabled on SLES 10 in this kernel.
2688 Description: block reactivating mgc import until all deactivates complete
2689 Details : Fix race when failing back MDT/MGS to itself (testing)
2691 Severity : enhancement
2693 Description: lfs setstripe enhancement
2694 Details : Make lfs setstripe understand 'k', 'm' and 'g' for stripe size.
2697 Frequency : mds/oss recovery
2699 Description: llog ctxt is refrenced after it has been freed.
2700 Details : llog ctxt refcount was added to avoide the race between ctxt free
2701 and llog recovery process. Each llog user must hold ctxt refcount
2702 before it access the llog. And the llog ctxt can only be freed
2703 when its refcount is zero.
2705 Severity : enhancement
2707 Description: randomly memory allocation failure util
2708 Details : Make lustre randomly failed allocating memory for testing purpose.
2711 Frequency : only for SLES10
2713 Description: Update kernel patch for SLES10 SP1
2714 Details : Add patch blkdev_tunables-2.6-sles10.patch to 2.6-sles10.series.
2716 Severity : enhancement
2718 Description: omit set fsid for export NFS
2719 Details : fix set/restore device id for avoid EMFILE error and mark lustre fs
2720 as FS_REQUIRES_DEV for avoid problems with generate fsid.
2723 Frequency : after network failures
2725 Description: Add sync option to mount_lustre.c
2726 Details : Client loses data written to lustre after a network interruption.
2728 Severity : enhancement
2730 Description: Error message improvement.
2731 Details : Merging of two LCONSOLE_ERROR_MSG into one.
2733 Severity : enhancement
2735 Description: don't use GFP_* in generic Lustre code.
2736 Details : Use cfs_alloc_* functions and CFS_* flags for code portability.
2738 Severity : enhancement
2740 Description: obdclass is limited by single OBD_ALLOC(idarray)
2741 Details : replace OBD_ALLOC/OBD_FREE with OBD_VMALLOC/OBD_VFREE
2743 Severity : enhancement
2745 Description: updated patchess for new RHEL4 kernel
2746 Details : Fixed ext3-unlink-race.patch per Kalpak's comment.
2748 Severity : enhancement
2750 Description: warnings with build patchless client with vanila 2.6.19 and up
2751 Details : change old ctl_table style and replace ctl_table/ctl_table_header
2752 with cfs_sysctl_table_t/cfs_sysctl_table_header_t
2754 Severity : enhancement
2756 Description: O_DIRECT bypasses client statistics.
2757 Details : When running with O_DIRECT I/O, neither the client rpc_stats nor
2758 read_ahead_stats were updated. Copied stats section from
2759 osc_send_oap_rpc() into async_internal().
2762 Frequency : only for Cray XT3
2764 Description: peer credits not enough on many OST per OSS systems.
2765 Details : Use new lnet way to add credits as we need those for pings and ASTs
2768 Frequency : only with liblustre
2770 Description: Liblustre is not releasing flock locks on file close.
2771 Details : Release flock locks on file close.
2774 Frequency : only for RHEL4
2776 Description: Update kernel patches for kernel-2.6.9-55.0.2.EL
2777 Details : Remove inode-nr_unused-2.6.9-rhel4.patch from 2.6-rhel4.series
2778 Update target file and kernel config.
2782 Description: lustre support for RHEL5
2783 Details : Add support for RHEL5.
2788 Description: ASSERTION(export != NULL) failed in target_handle_connect
2789 Details : Assetion hit is result of rare race between disconnect and connet
2790 to same nid. target_handle_connect found old connect cockie and
2791 tried to reconnect, but can't find export for this cockie.
2796 Description: umount blocks forever on error
2797 Details : In result of wrong using obd_no_recov and obd_force flags client
2798 can hand if cancel or some other requests is lost.
2801 Frequency : Only for SLES
2803 Description: sanity_quota fail test_1
2804 Details : There are multiple occurences of $TSTUSR in SLES's /etc/group
2805 file, which makes TSTID[2] inunique.
2807 Severity : enhancement
2809 Description: Kernel patches for SLES9 2.6.5-7.286 kernel
2810 Details : Update target/ChangeLog/which_patch .
2812 Severity : enhancement
2814 Description: jbd statistics
2815 Details : Port older jbd statistics patch for sles10
2820 Description: lvbo_init failed for resource with missing objects.
2821 Details : Fix returning error if we do stat for file with missing/corrupted
2822 objects and i_size set to all sum of size of all avaible objects.
2823 if we to truncate/write to missing object - it is recreated.
2828 Description: Oops in read and write path when failing to allocate lock.
2829 Details : Check if lock allocation failed and return error back.
2832 Frequency : When flocks are used.
2834 Description: assertion failure in ldlm_cli_enquque_fini for non NULL lock.
2835 Details : Flock locks might destroy just granted lock if it could be merged
2836 with another existing flock, this is done in completion handler,
2837 so teach ldlm_cli_enquque_fini that this is a valid case for
2843 Description: reply_lock_interpret crash due to race with it and lock cancel.
2844 Details : Do not replay locks that are being cancelled. Do not reference
2845 locks by their address during replay, just by their handle.
2848 Frequency : only with deactivated OSTs
2850 Description: lstripe command fails for valid OST index
2851 Details : The stripe offset is compared to 'lov->desc.ld_tgt_count'
2852 instead of lov->desc.ld_active_tgt_count.
2854 Severity : enhancement
2856 Description: Build failure against Centos5 (RHEL5)
2857 Details : Use getpagesize() instead of PAGE_SIZE.
2859 --------------------------------------------------------------------------------
2861 2007-07-30 Cluster File Systems, Inc. <info@clusterfs.com>
2863 * Support for kernels:
2864 2.6.5-7.283 (SLES 9),
2865 2.6.9-55.EL (RHEL 4),
2866 2.6.16.46-0.14 (SLES 10),
2867 2.6.18.8 vanilla (kernel.org)
2868 * Client support for unpatched kernels:
2869 (see http://wiki.lustre.org/index.php?title=Patchless_Client)
2870 2.6.16 - 2.6.22 vanilla (kernel.org)
2871 * Due to recently discovered recovery problems, we do not recommend
2872 using patchless RHEL 4 clients with this or any earlier release.
2873 * Recommended e2fsprogs version: 1.39.cfs8
2874 * Note that reiserfs quotas are disabled on SLES 10 in this kernel.
2875 * Starting with this release, the ldiskfs backing filesystem required
2876 by Lustre is now in its own package, lustre-ldiskfs. This package
2877 should be installed. It is versioned separately from Lustre and
2878 may be released separately in future.
2883 Description: server LBUG when shutting down
2884 Details : Block umount forever until the mount refcount is zero rather
2885 than giving up after an arbitrary timeout.
2887 Severity : enhancement
2889 Description: add optional extra BUILD_VERSION info
2890 Details : add a new environment variable (namely LUSTRE_VERS) which allows
2891 to override the lustre version.
2894 Frequency : 2.6.18 servers only
2896 Description: ll_kern_mount() doesn't release the module reference
2897 Details : The ldiskfs module reference count never drops down to 0
2898 because ll_kern_mount() doesn't release the module reference.
2903 Description: server LBUG when using old ost_num_threads parameter
2904 Details : Accept the old ost_num_threads parameter but warn that it
2905 is deprecated, and fix an off-by-one error that caused an LBUG.
2910 Description: Transient SCSI error results in persistent IO issue
2911 Details : iobuf->dr_error is not reinitialized to 0 between two
2915 Frequency : sometimes when underlying device returns I/O errors
2917 Description: OSTs not going read-only during write failures
2918 Details : OSTs are not remounted read-only when the journal commit threads
2919 get I/O errors because fsfilt_ext3 calls journal_start/stop()
2920 instead of the ext3 wrappers.
2923 Frequency : only with 10000 clients or more
2925 Description: poor connect scaling with increasing client count
2926 Details : Don't run filter_grant_sanity_check for more than 100 exports
2927 to improve scaling for large numbers of clients.
2930 Frequency : SLES10 only
2932 Description: sanity-quota.sh quotacheck failed: rc = -22
2933 Details : Quotas cannot be enabled on SLES10.
2936 Frequency : liblustre clients only
2938 Description: getdirentries does not give error when run on compute nodes
2939 Details : getdirentries does not fail when the size specified as an argument
2940 is too small to contain at least one entry
2942 Severity : enhancement
2944 Description: Add LNET router traceability for debug purposes
2945 Details : If a checksum failure occurs with a router as part of the
2946 IO path, the NID of the last router that forwarded the bulk data
2947 is printed so it can be identified.
2952 Description: OST "spontaneously" evicts client; client has imp_pingable == 0
2953 Details : Due to a race condition, liblustre clients were occasionally
2954 evicted incorrectly.
2956 Severity : enhancement
2958 Description: lfs setstripe use optional parameters instead of postional
2961 Severity : enhancement
2963 Description: Nanosecond timestamp support for ldiskfs
2964 Details : The on-disk ldiskfs filesystem has added support for nanosecond
2965 resolution timestamps. There is not yet support for this at
2966 the Lustre filesystem level.
2969 Frequency : during server recovery
2971 Description: MDS failing to send precreate requests due to OSCC_FLAG_RECOVERING
2972 Details : request with rq_no_resend flag not awake l_wait_event if they get a
2976 Frequency : nfs export on patchless client
2978 Description: connectathon hang when test nfs export over patchless client
2979 Details : Disconnected dentry cannot be found with lookup, so we do not need
2980 to unhash it or make it invalid
2983 Description: fix llapi_lov_get_uuids() to allow many OSTs to be returned
2984 Details: : Change llapi_lov_get_uuids() to read the UUIDs from /proc instead
2985 of using an ioctl. This allows lfsck for > 160 OSTs to succeed.
2990 Description: open req refcounting wrong on reconnect
2991 Details : If reconnect happened between getting open reply from server and
2992 call to mdc_set_replay_data in ll_file_open, we will schedule
2993 replay for unreferenced request that we are about to free.
2994 Subsequent close will crash in variety of ways.
2995 Check that request is still eligible for replay in
2996 mdc_set_replay_data().
3001 Description: disable writes to filesystem when reading health_check file
3002 Details : the default for reading the health_check proc file has changed
3003 to NOT do a journal transaction and write to disk, because this
3004 can cause reads of the /proc file to hang and block HA state
3005 checking on a healthy but otherwise heavily loaded system. It
3006 is possible to return to the previous behaviour during configure
3007 with --enable-health-write.
3009 Severity : enhancement
3011 Description: 64-bit inode version
3012 Details: : Add a on-disk 64-bit inode version for ext3 to track changes made
3013 to the inode. This will be required for version-based recovery.
3018 Description: MDS fails to start if a duplicate client export is detected
3019 Details : in some rare cases it was possible for a client to connect to
3020 an MDS multiple times. Upon recovery the MDS would detect this
3021 and fail during startup. Handle this more gracefully.
3023 Severity : enhancement
3025 Description: Add -o localflock option to simulate old noflock behaviour.
3026 Details : This will achieve local-only flock/fcntl locks coherentness.
3031 Description: log_commit_thread vs filter_destroy race leads to crash
3032 Details : Take import reference before releasing llog record semaphore
3037 Description: Wrong request locking in request set processing
3038 Details : ptlrpc_check_set wrongly uses req->rq_lock for proctect add to
3039 imp_delayed_list, in this place should be used imp_lock.
3042 Frequency : when reconnecting
3044 Description: Grant leak when OSC reconnect to OST
3045 Details : When osc reconnect ost, OST(filter) should check whether it
3046 should grant more space to client by comparing fed_grant and
3047 cl_avail_grant, and return the granted space to client instead
3048 of "new granted" space, because client will call osc_init_grant
3049 to update the client grant space info.
3052 Frequency : when client reconnects to OST
3054 Description: Grant leak when OSC does a resend and replays bulk write
3055 Details : When osc reconnect to OST, OST(filter) should clear grant info of
3056 bulk write request, because the grant info while be sync between
3057 OSC and OST when reconnect, and we should ignore the grant info
3058 these of resend/replay write req.
3063 Description: Grant space more than avaiable space sometimes.
3064 Details : When then OST is about to be full, if two bulk writing from
3065 different clients came to OST. Accord the avaliable space of the
3066 OST, the first req should be permitted, and the second one
3067 should be denied by ENOSPC. But if the seconde arrived before
3068 the first one is commited. The OST might wrongly permit second
3069 writing, which will cause grant space > avaiable space.
3072 Frequency : when client is evicted
3074 Description: Grant might be wrongly erased when osc is evicted by OST
3075 Details : when the import is evicted by server, it will fork another
3076 thread ptlrpc_invalidate_import_thread to invalidate the
3077 import, where the grant will be set to 0. While the original
3078 thread will update the grant it got when connecting. So if
3079 the former happened latter, the grant will be wrongly errased
3080 because of this race.
3085 Description: Checking Stale with correct fid
3086 Details : ll_revalidate_it should uses de_inode instead of op_data.fid2
3087 to check whether it is stale, because sometimes, we want the
3088 enqueue happened anyway, and op_data.fid2 will not be initialized.
3090 Severity : enhancement
3092 Description: update patchless client
3093 Details : Add support for patchless client with 2.6.20, 2.6.21 and RHEL 5
3096 Frequency : only with 2.4 kernel
3098 Description: random memory corruption
3099 Details : size of struct ll_inode_info is to big for union inode.u and this
3100 can be cause of random memory corruption.
3105 Description: Memory leak in recovery
3106 Details : Lov_mds_md was not free in an error handler in mds_create_object.
3107 It should also check obd_fail before fsfilt_start, otherwise if
3108 fsfilt_start return -EROFS,(failover mds during mds recovery).
3109 then the req will return with repmsg->transno = 0 and rc = EROFS.
3110 and we met hit the assert LASSERT(req->rq_reqmsg->transno ==
3111 req->rq_repmsg->transno) in ptlrpc_replay_interpret. Fcc should
3112 be freed no matter whether fsfilt_commit success or not.
3115 Frequency : only with large numbers of cores on a single node
3117 Description: Prevents from taking the superblock lock in llap_from_page for
3119 Details : using LL_ORIGIN_REMOVEPAGE origin flag instead of LL_ORIGIN_UNKNOW
3120 for llap_from_page call in ll_removepage() prevents from taking
3121 the superblock lock for a soon died page.
3126 Description: Not check open intent error before release open handle
3127 Details : in some rare cases, the open intent error is not checked before
3128 release open handle, which may cause
3129 ASSERTION(open_req->rq_transno != 0), because it tries to release
3130 the failed open handle.
3135 Description: Set cat log bitmap only after create log success.
3136 Details : in some rare cases, the cat log bitmap is set too early. and it
3137 should be set only after create log success.
3141 Description: Accessing a block bevice can re-enable I/O when Lustre is
3142 tearing down a device.
3143 Details : dev_clear_rdonly(bdev) must be called in kill_bdev() instead of
3148 Description: service threads may hog cpus when there are a lot of requests
3149 Details : Insert cond_resched to give other threads a chance to use some CPU
3154 Description: the cat log was not initialized in recovery
3155 Details : When mds(mgs) do recovery, the tgt_count might be zero, so the
3156 unlink log on mds will not be initialized until mds post
3157 recovery. And also in mds post recovery, the unlink log will
3158 initialization will be done asynchronausly, so there will be race
3159 between add unlink log and unlink log initialization.
3163 Description: brw_stats were being printed incorrectly
3164 Details : brw_stats were being printed as log2 but all of them were not
3165 recorded as log2. Also remove some code duplication arising from
3166 filter_tally_{read,write}.
3170 Frequency : rare, only in recovery.
3171 Description: ASSERTION(req->rq_type != LI_POISON) failed
3172 Details : imp_lock should be held while iterating over imp_sending_list for
3173 prevent destroy request after get timeout in ptlrpc_queue_wait.
3177 Description: replay-single.sh test 52 fails
3178 Details : A lock's skiplist need to be cleanup when it being unlinked
3179 from its resource list.
3184 Description: Short directio read returns full requested size rather than
3186 Details : Direct I/O operations should return actual amount of bytes
3187 transferred rather than requested size.
3189 Severity : enhancement
3191 Description: metadata RPC reduction (e.g. for rm performance)
3192 Details : decrease the amount of synchronous RPC between clients and servers
3193 by canceling conflicing lock before the operation on the client
3194 and packing thier handles into the main operation RPC to server.
3196 Severity : enhancement
3198 Description: add #ifdef HAVE_KERNEL_CONFIG_H
3199 Details : kernels from 2.6.19 not need include linux/config.h, but add
3200 include linux/autoconf.h in commpiler command line.
3202 Severity : enhancement
3204 Description: patchless client support for 2.6.22 kernel
3205 Details : 2.6.22 has only one visble change, SLAB_CTOR_* constants is
3206 removed. In this case we need drop using os depended interface to
3207 kmem_cache and use cfs_mem_cache API.
3212 Description: fix mal-formatted messages
3213 Details : fix some mal-formatted DEBUG_REQ and LCONSOLE_ERROR_MSG messages
3217 Frequency : always in liblustre
3218 Description: wrong IS_ERR implementation in liblustre.h
3219 Details : fix IS_ERR implementation in liblustre.h for right detect errors.
3224 Description: Correct condition for output debug message.
3225 Details : inode i_nlink equal zero is not enough for output message about
3226 disk corruption, i_ctime and i_mode should be also checked.
3230 Frequency : always in patchless client
3231 Description: add configure check for truncate_complete_page
3232 Details : improve checks for exported symbols. This allow run check without
3233 sources, but with Module.symvers shipped with kernel distribution.
3234 add check for truncate_complete_page used by patchless client.
3238 Description: sanity.sh test_77h fails with "test_77h file compare failed"
3239 Details : test_77h uses a file which was messed by other test case.
3243 Description: Not Check whether lov_tgts is NULL in some lov functions
3244 Details : Checking whether lov_tgts is NULL in some functions.
3248 Description: replace obdo_alloc() with OBDO_ALLOC macro
3249 Details : nothing special is done in obdo_alloc() function, and for
3250 debugging purpose, it needs to be replaced with macros.
3254 Description: bad return value and errno from fcntl call
3255 Details : In liblustre API, errno should be a negative value if error
3260 Description: ptlrpc_check_set() LBUG
3261 Details : In case of positive reply from server and failed client bulk
3262 callback after bulk transfer shouldn't LBUG, but process this
3263 request as erroneous.
3265 Severity : enhancement
3267 Description: VFS operations stats tool.
3268 Details : Tool which collects stats by tracking value written in pid,
3269 ppid, gid and uses llstat to generate output to plot graph using
3271 Updated lustre/utils/Makefile.am
3272 Added lustre/utils/ltrack_stats.c
3274 Severity : enhancement
3276 Description: 2.6.18 server support (lustre 1.6.1)
3277 Details : Support for 2.6.18 kernels on the server side.
3282 Description: ASSERTION(imp->imp_conn_current) failed
3283 Details : an assertion failure is hit if a client node boots and attempts to
3284 mount a lustre filesystem faster than RECONNECT_INTERVAL seconds.
3287 Frequency : only for i686
3289 Description: 1.4.11 RC1 build fails for RHEL 4, i686
3290 Details : Fixed config variable for build.
3295 Description: Updated patchess for new RHEL4 kernel
3296 Details : Updated patch inode-nr_unused-2.6.9-rhel4.patch
3297 Updated patch jbd-stats-2.6.9.patch
3298 Updated patch qsnet-rhel4-2.6.patch
3299 Updated patch quota-deadlock-on-pagelock-core.patch
3300 Updated patch vfs_intent-2.6-rhel4.patch
3301 Updated patch vfs_races-2.6-rhel4.patch
3302 Updated series file 2.6-rhel4-titech.series
3303 Updated series file 2.6-rhel4.series
3304 Updated kernel config files
3309 Description: lquota slave complains LBUG when reconnecting with mds
3311 Details : quota slave depends qctxt->lqc_import to send its quota request.
3312 This pointer will be invalid if mds did failover or broke its
3313 connect to osts, which leads to LBUG.
3316 Frequency : when qunit size is too small(less than 20M)
3318 Description: write is stopped by improper -EDQUOT
3319 Details : If the master is busy and qunit size is small enough(let's say 1M),
3320 the slave can not get quota from master on time, which will lead
3321 slave to trigger a -EQUOTA to client.
3326 Description: Deadlock during metadata tests
3327 Details : in prune_dir_dentries(), shrink_dcache_parent() should not be
3328 called with the per-dentry lock held.
3331 Frequency : SLES9 only
3333 Description: Lustre patched kernel for SLES9 SP3 has NR_CPUS set to 8
3334 Details : set CONFIG_NR_CPUS to 128 instead of 8.
3336 Severity : enhancement
3338 Description: remove fs_prep_san_write operation and related patches
3339 Details : remove the ext3-san-jdike patches which are no longer useful.
3344 Description: LDISKFS-fs error (device sdc): ldiskfs_free_blocks
3345 Details : a disk corruption can cause the mballoc code to assert on a
3346 double free or other extent corruptions. Handle these with
3347 ext3_error() instead of assertions.
3350 Frequency : only with mballoc3 code and deep extent trees
3352 Description: ldiskfs_ext_search_right: bad header in inode: unexpected eh_depth
3353 Details : a wrong check of extent headers in ldiskfs_ext_search_right()
3354 can cause the filesystem to be remounted read-only.
3359 Description: lfsck built against 1.4.x cannot run against 1.6.0 lustre
3360 Details : the definition for OBD_IOC_GETNAME changed in 1.6.0. One of the
3361 few external users of this ioctl number is lfsck's call to
3362 llapi_lov_get_uuids() and this caused lfsck to fail at startup.
3363 Add the old ioctl number to the handler so both old and new
3368 Description: parallel lock callbacks
3369 Details : Instead of sending blocking and completion callbacks as separated
3370 requests, adding them to a set and sending in parallel.
3374 Description: Disable most debugging by default
3375 Details : To improve performance, disable most logging (for debug purposes)
3376 by default. VFSTRACE, RPCTRACE, and DLMTRACE are now off by
3377 default, and HA includes fewer messages.
3381 Frequency : only run on patchless client.
3382 Description: use do_facet on sanity.sh for test handling recoverables errors
3383 Details : use do_facet instead of direct use sysctl for set fail_loc on OST
3387 Description: hash tables for lists of nids, connections and uuids
3388 Details : Hash tables noticeably help when a lot of clients connect to a
3389 server, to faster identify duplicate connections or reconnects,
3390 also to faster find export to evict in manual eviction case.
3394 Description: Sometimes, when the server evict a client, and the client will
3395 not be evicted as soon as possible.
3396 Details : In enqueue req, the error was returned by intent, instead of
3397 rq_status which make ptlrpc layer not detect this error, and
3398 does not evict the client. So enqueue error should be returned
3402 Frequency : only at startup
3404 Description: Delay client connections to MDT until fist MDT->OST connect
3405 Details : If a client tried to create a new file before the MDT had
3406 connected to any OSTs, the create would return EIO. Now
3407 the client will simply block until the MDT connects to the
3408 first OST and the create can succeed.
3411 Frequency : only for SLES9
3413 Description: Routinely utilize latest Quadrics drivers in CFS releases
3414 Details : Update patch qsnet-suse-2.6.patch.
3417 Frequency : only for sles10
3419 Description: Update patches for SLES 10 SP1 kernel.
3420 Details : Update the patch vfs_intent-2.6-sles10.patch.
3425 Description: Routinely utilize latest Quadrics drivers in CFS releases
3426 Details : Update patch qsnet-rhel4-2.6.patch.
3429 Frequency : at statup only
3431 Description: mds_lov_synchronize race leads to various problems
3432 Details : simultaneous MDT->OST connections at startup can cause the
3433 sync to abort, leaving the OSC in a bad state.
3438 Description: Using wrong pointer in osc_brw_prep_request
3439 Details : Access to array[-1] can produce panic if kernel compiled with
3440 CONFIG_PAGE_ALLOC enabled
3442 Severity : enhancement
3444 Description: Async OSC create to avoid the blocking unnecessarily.
3445 Details : If a OST has no remain object, system will block on the creating
3446 when need to create a new object on this OST. Now, ways use
3447 pre-created objects when available, instead of blocking on an
3448 empty osc while others are not empty. If we must block, we block
3449 for the shortest possible period of time.
3453 Frequency : only in recovery
3454 Description: Mark OST as early accessible if his start SYNC.
3455 Details : osc_precreate return flag early accessible if oscc marked as
3456 OSCC_FLAG_SYNC_IN_PROGRESS.
3461 Description: Sometimes precreate code can triger create object on wrong ost
3462 Details : Wrong protected or not not restored variables aftre precreate loop
3463 can produce creation object on wrong ost.
3466 Frequency : oss recovery
3468 Description: llog_commit_thread cleanup should sync with llog_commit_thread
3470 Details : llog_commit_thread_count should be synced between llog_commit
3471 start and cleanup, so new llog_commit thread should not be started
3472 when llog_commit threads being stopped to avoid accessing some
3475 Severity : enhancement
3477 Description: Add printing inode info into message about error in writepage.
3479 --------------------------------------------------------------------------------
3481 2007-05-03 Cluster File Systems, Inc. <info@clusterfs.com>
3486 Frequency : on some architectures
3488 Description: 1.6 client sometimes fails to mount from a 1.4 MDT
3489 Details : Uninitialized flags sometimes cause configuration commands to
3493 Frequency : patchless clients only
3495 Description: missing __iget() symbol export
3496 Details : The __iget() symbol export is missing. To avoid the need for
3497 this on patchless clients the deathrow inode reaper is turned
3498 off, and we depend on the VM to clean up old inodes. This
3499 dependency was during via the fix for bug 12181.
3504 Description: sanity.sh fail: test_52b
3505 Details : The ll_inode_to_ext_flags() has a glitch which makes MDS return
3506 incorrect inode's flags to client.
3508 --------------------------------------------------------------------------------
3510 2007-04-19 Cluster File Systems, Inc. <info@clusterfs.com>
3512 * CONFIGURATION CHANGE. This version of Lustre WILL NOT
3513 INTEROPERATE with older versions automatically. In many cases a
3514 special upgrade step is needed. Please read the
3515 user documentation before upgrading any part of a 1.4.x system.
3516 * WARNING: Lustre configuration and startup changes are required with
3517 this release. See https://mail.clusterfs.com/wikis/lustre/MountConf
3519 * Support for kernels:
3520 2.4.21-47.0.1.EL (RHEL 3),
3521 2.6.5-7.283 (SLES 9),
3522 2.6.9-42.0.10.EL (RHEL 4),
3523 2.6.12.6 vanilla (kernel.org),
3524 2.6.16.27-0.9 (SLES10)
3525 * Client support for unpatched kernels:
3526 (see https://mail.clusterfs.com/wikis/lustre/PatchlessClient)
3527 2.6.16 - 2.6.19 vanilla (kernel.org),
3528 2.6.9-42.0.8EL (RHEL 4)
3529 * Recommended e2fsprogs version: 1.39.cfs6
3530 * Note that reiserfs quotas are disabled on SLES 10 in this kernel
3533 Severity : enhancement
3535 Description: Async OSC create to avoid the blocking unnecessarily.
3536 Details : If a OST has no remain object, system will block on the creating
3537 when need to create a new object on this OST. Now, ways use
3538 pre-created objects when available, instead of blocking on an
3539 empty osc while others are not empty. If we must block, we block
3540 for the shortest possible period of time.
3542 Severity : enhancement
3544 Description: MountConf
3545 Details : Lustre configuration is now managed via mkfs and mount
3546 commands instead of lmc and lconf. New obd types (MGS, MGC)
3547 are added for dynamic configuration management. See
3548 https://mail.clusterfs.com/wikis/lustre/MountConf for
3551 Severity : enhancement
3553 Description: dynamic OST addition
3554 Details : OSTs can now be added to a live filesystem
3556 Severity : enhancement
3558 Description: startup order invariance
3559 Details : MDTs and OSTs can be started in any order. Clients only
3560 require the MDT to complete startup.
3562 Severity : enhancement
3564 Description: parallel, asynchronous orphan cleanup
3565 Details : orphan cleanup is now performed in separate threads for each
3566 OST, allowing parallel non-blocking operation.
3568 Severity : enhancement
3570 Description: optimized stripe assignment
3571 Details : stripe assignments are now made based on ost space available,
3572 ost previous usage, and OSS previous usage, in order to try
3573 to optimize storage space and networking resources.
3575 Severity : enhancement
3577 Description: Permanently set tunables
3578 Details : All writable /proc/fs/lustre tunables can now be permanently
3579 set on a per-server basis, at mkfs time or on a live
3582 Severity : enhancement
3584 Description: Lustre message v2
3585 Details : Add lustre message format v2.
3587 Severity : enhancement
3589 Description: client OST exclusion list
3590 Details : Clients can be started with a list of OSTs that should be
3591 declared "inactive" for known non-responsive OSTs.
3595 Description: ENOENT returned for valid filehandle during dbench.
3596 Details : Check if a directory has children when invalidating dentries
3597 associated with an inode during lock cancellation. This fixes
3598 an incorrect ENOENT sometimes seen for valid filehandles during
3599 testing with dbench.
3602 Frequency : SFS test only (otherwise harmless)
3604 Description: SPEC SFS validation failure on NFS v2 over lustre.
3605 Details : Changes the blocksize for regular files to be 2x RPC size,
3606 and not depend on stripe size.
3608 Severity : enhancement
3610 Description: fine-grained SMP locking inside DLM
3611 Details : Improve DLM performance on SMP systems by removing the single
3612 per-namespace lock and replace it with per-resource locks.
3614 Severity : enhancement
3616 Description: don't hold multiple extent locks at one time
3617 Details : To avoid client eviction during large writes, locks are not
3618 held on multiple stripes at one time or for very large writes.
3619 Otherwise, clients can block waiting for a lock on a failed OST
3620 while holding locks on other OSTs and be evicted.
3622 Severity : enhancement
3624 Description: Multiple MD RPCs in flight.
3625 Details : Further unserialise some read-only MDT RPCs - learn about intents.
3626 To avoid overly-overloading MDT, introduce a limit on number of
3627 MDT RPCs in flight for a single client and add /proc controls
3628 to adjust this limit.
3630 Severity : enhancement
3632 Description: client read/write statistics
3633 Details : Add client read/write call usage stats for performance
3634 analysis of user processes.
3635 /proc/fs/lustre/llite/*/offset_stats shows non-sequential
3636 file access. extents_stats shows chunk size distribution.
3637 extents_stats_per_process show chunk size distribution per
3640 Severity : enhancement
3642 Description: per-client statistics on server
3643 Details : Add ldlm and operations statistics for each client in
3644 /proc/fs/lustre/mds|obdfilter/*/exports/
3646 Severity : enhancement
3648 Description: improved MDT statistics
3649 Details : Add detailed MDT operations statistics in
3650 /proc/fs/lustre/mds/*/stats
3652 Severity : enhancement
3654 Description: VFS operations stats
3655 Details : Add client VFS call stats, trackable by pid, ppid, or gid
3656 /proc/fs/lustre/llite/*/stats_track_[pid|ppid|gid]
3661 Description: Fix client-side osc byte counters
3662 Details : The osc read/write byte counters in
3663 /proc/fs/lustre/osc/*/stats are now working
3666 Frequency : always as root on SLES
3668 Description: Failure of copying files with lustre special EAs.
3669 Details : Client side always return success for setxattr call for lustre
3670 special xattr (currently only "trusted.lov").
3675 Description: Refcount LNET uuids
3676 Details : The global LNET uuid list grew linearly with every startup;
3677 refcount repeated list entries instead of always adding to
3680 Severity : enhancement
3682 Description: Dynamic service threads
3683 Details : Within a small range, start extra service threads
3684 automatically when the request queue builds up.
3687 Frequency : mixed-endian client/server environments
3689 Description: mixed-endian crashes
3690 Details : The new msg_v2 system had some failures in mixed-endian
3693 Severity : enhancement
3695 Description: Easy OST removal
3696 Details : OSTs can be permanently deactivated with e.g. 'lctl
3697 conf_param lustre-OST0001.osc.active=0'
3699 Severity : enhancement
3701 Description: MGS proc entries
3702 Details : Added basic proc entries for the MGS showing what filesystems
3705 Severity : enhancement
3707 Description: provide MGS failover
3708 Details : Added config lock reacquisition after MGS server failover.
3710 Severity : enhancement
3712 Description: add Linux 2.4 support
3713 Details : Added support for RHEL 2.4.21 kernel for 1.6 servers and clients
3717 Description: a large application tries to do I/O to the same resource and dies
3718 in the middle of it.
3719 Details : Check the req->rq_arrival time after the call to
3720 ost_brw_lock_get(), but before we do anything about
3721 processing it & sending the BULK transfer request. This
3722 should help move old stale pending locks off the queue as
3723 quickly as obd_timeout.
3726 Frequency : when an incorrect nid is specified during startup
3728 Description: ptlrpc connect to non-existant node causes kernel crash
3729 Details : LNET can't be re-entered from an event callback, which
3730 happened when we expire a message after the export has been
3731 cleaned up. Instead, hand the zombie cleanup off to another
3734 Severity : enhancement
3736 Description: plain/inodebits lock performance improvement
3737 Details : Grouping plain/inodebits in granted list by their request modes
3738 and bits policy, thus improving the performance of search through
3742 Frequency : only if OST filesystem is corrupted
3744 Description: client incorrectly hits assertion in ptlrpc_replay_req()
3745 Details : for a short time RPCs with bulk IO are in the replay list,
3746 but replay of bulk IOs is unimplemented. If the OST filesystem
3747 is corrupted due to disk cache incoherency and then replay is
3748 started it is possible to trip an assertion. Avoid putting
3749 committed RPCs into the replay list at all to avoid this issue.
3752 Frequency : liblustre (e.g. catamount) on a large cluster with >= 8 OSTs/OSS
3754 Description: System hang on startup
3755 Details : This bug allowed the liblustre (e.g. catamount) client to
3756 return to the app before handling all startup RPCs. This
3757 could leave the node unresponsive to lustre network traffic
3758 and manifested as a server ptllnd timeout.
3760 Severity : enhancement
3762 Description: Add "/proc/sys/lustre/debug_peer_on_timeout"
3763 Details : liblustre envirable: LIBLUSTRE_DEBUG_PEER_ON_TIMEOUT
3764 boolean to control whether to print peer debug info when a
3765 client's RPC times out.
3768 Frequency : only for kernels with patches from Lustre below 1.4.3
3770 Description: Remove old rdonly API
3771 Details : Remove old rdonly API which unused from at least lustre 1.4.3
3774 Frequency : only for devices with external journals
3776 Description: Set external device read-only also
3777 Details : During a commanded failover stop, we set the disk device
3778 read-only while the server shuts down. We now also set any
3779 external journal device read-only at the same time.
3782 Frequency : when upgrading from 1.4 while trying to change parameters
3784 Description: The wrong (new) MDC name was used when setting parameters for
3785 upgraded MDT's. Also allows changing of OSC (and MDC)
3786 parameters if --writeconf is specified at tunefs upgrade time.
3789 Frequency : when setting specific ost indicies
3791 Description: QOS code breaks on skipped indicies
3792 Details : Add checks for missing OST indicies in the QOS code, so OSTs
3793 created with --index need not be sequential.
3795 Severity : enhancement
3797 Description: Add uninit_groups feature to ldiskfs2 to speed up e2fsck
3798 Details : The uninit_groups feature works in conjunction with the kernel
3799 filesystem code (ldiskfs2 only) and e2fsprogs-1.39-cfs6 to speed
3800 up the pass1 processing of e2fsck. This is a read-only feature
3801 in ldiskfs2 only, so older kernels and current ldiskfs cannot
3802 mount filesystems that have had this feature enabled.
3804 Severity : enhancement
3806 Description: Improve multi-block allocation algorithm to avoid fragmentation
3807 Details : The mballoc3 code (ldiskfs2 only) adds new mechanisms to improve
3808 allocation locality and avoid filesystem fragmentation.
3810 ------------------------------------------------------------------------------
3812 2007-04-01 Cluster File Systems, Inc. <info@clusterfs.com>
3814 * Support for kernels:
3815 2.4.21-47.0.1.EL (RHEL 3)
3816 2.6.5-7.283 (SLES 9)
3817 2.6.9-42.0.10.EL (RHEL 4)
3818 2.6.12.6 vanilla (kernel.org)
3819 2.6.16.27-0.9 (SLES 10)
3820 * Recommended e2fsprogs version: 1.39.cfs5
3822 * Note that reiserfs quotas are disabled on SLES 10 in this kernel
3826 Frequency : occasional, depends on client load and configuration
3827 Bugzilla : 12181, 12203
3828 Description: data loss for recently-modified files
3830 Details : In some cases it is possible that recently written or created
3831 files may not be written to disk in a timely manner (this should
3832 normally be within 30s unless client IO load is very high).
3833 The problem appears as zero-length files or files that are a
3834 multiple of 1MB in size after a client crash or client eviction
3835 that are missing data at the end of the file.
3837 This problem is more likely to be hit on clients where files are
3838 repeatedly created and unlinked in the same directory, clients
3839 have a large amount of RAM, have many CPUs, the filesystem has
3840 many OSTs, the clients are rebooted frequently, and/or the files
3841 are not accessed by other nodes after being written.
3843 The presence of the problem can be detected by looking at
3844 /proc/sys/fs/inode-state. If the first number (nr_inodes) is
3845 smaller than the second (nr_unused) then dirty files will not
3846 be flushed automatically to disk. "sync; sleep 10" should be
3847 run several times on the node before unmounting it to update
3848 Lustre (this is also safe to run on nodes without this problem).
3850 There is also a related kernel bug in the RHEL4 4 2.6.9 kernel
3851 that can cause this same problem, so customers using that kernel
3852 also need to update the kernel in addition to Lustre. In order
3853 to properly fix this bug, the RHEL3 2.4.21 kernel is also updated.
3855 It is normal that files written just before a client crash (less
3856 than 30s) may not yet have been flushed to disk, even for local
3860 Frequency : frequent on thin XT3 nodes
3862 Description: UUID collision on thin XT3 Linux nodes
3863 Details : UUIDs on Compute Node Linux XT3 nodes were not generated
3864 randomly, since we relied on an insufficiently-seeded PRNG.
3869 Description: OSS hangs after "All ost request buffers busy"
3870 Details : A deadlock between quota and journal operations caused OSS
3871 hangs after printing "All ost request buffers busy."
3874 Frequency : always on liblustre builds
3876 Description: Cleanup compiler warnings on liblustre
3879 Frequency : always on liblustre builds on XT3
3881 Description: LC_CONFIG_CDEBUG don't run while build liblustre on XT3.
3885 Description: Addition of EXT3_FEATURE_RO_COMPAT_DIR_NLINKS flag for
3886 > 32000 subdirectories
3887 Details : Add EXT3_FEATURE_RO_COMPAT_DIR_NLINK flag to
3888 EXT3_FEATURE_RO_COMPAT_SUPP. This flag will be set whenever
3889 subdirectory count crosses 32000. This will aid e2fsck to
3890 correctly handle more than 32000 subdirectories.
3893 Frequency : liblustre (e.g. catamount) on a large cluster with >= 8 OSTs/OSS
3895 Description: System hang on startup
3896 Details : This bug allowed the liblustre (e.g. catamount) client to
3897 return to the app before handling all startup RPCs. This
3898 could leave the node unresponsive to lustre network traffic
3899 and manifested as a server ptllnd timeout.
3901 Severity : enhancement
3903 Description: Add "/proc/sys/lustre/debug_peer_on_timeout"
3904 (liblustre envirable: LIBLUSTRE_DEBUG_PEER_ON_TIMEOUT)
3905 boolean to control whether to print peer debug info when a
3906 client's RPC times out.
3911 Description: make O_SYNC working on 2.6 kernels
3912 Details : 2.6 kernels use different method for mark pages for write,
3913 so need add a code to lustre for O_SYNC work.
3918 Description: Failure to close file and release space on NFS
3919 Details : Put inode details into lock acquired in ll_intent_file_open.
3920 Use mdc_intent_lock in ll_intent_open to properly
3921 detect all kind of errors unhandled by mdc_enqueue.
3926 Description: proc file read during shutdown sometimes raced obd removal,
3928 Details : Add lock to prevent obd access after proc file removal.
3931 Frequency : Only for files larger than 4GB on 32-bit clients.
3933 Description: improperly doing page alignment of locks
3934 Details : Modify lustre core code to use CFS_PAGE_* defines instead of
3935 PAGE_*. Make CFS_PAGE_MASK a 64-bit mask.
3940 Description: RPCs being resent when they shouldn't be
3941 Details : Some RPCs that should not be resent are being resent. This
3942 can cause inconsistencies in the RPC state machine. Do not
3943 resend such requests.
3946 Frequency : rare, only with NFS export
3948 Description: Crash on NFS re-export node
3949 Details : under very unusual load conditions an assertion is hit in
3950 ll_intent_file_open()
3953 Frequency : only if OST filesystem is corrupted
3955 Description: client incorrectly hits assertion in ptlrpc_replay_req()
3956 Details : for a short time RPCs with bulk IO are in the replay list,
3957 but replay of bulk IOs is unimplemented. If the OST filesystem
3958 is corrupted due to disk cache incoherency and then replay is
3959 started it is possible to trip an assertion. Avoid putting
3960 committed RPCs into the replay list at all to avoid this issue.
3965 Description: large O_DIRECT requests fail under memory pressure/fragmentation
3966 Details : Large single O_DIRECT read and write calls can fail to allocate
3967 a sufficiently large buffer to process the request. In case of
3968 allocation failure the allocation is retried with a smaller
3969 buffer and broken into smaller requests.
3971 Severity : enhancement
3973 Description: Add -o localflock option to simulate old noflock behaviour.
3974 Details : This will achieve local-only flock/fcntl locks coherentness.
3979 Description: versioning check is incomplete
3980 Details : Checking the version difference of client vs. server, report
3981 error if the gap is too big.
3986 Description: add support PG_writeback bit
3987 Details : add support for PG_writeback bit for Lustre, for more carefull
3988 work with page cache in 2.6 kernel. This also fix some deadlocks
3989 and remove hack for work O_SYNC with 2.6 kernel.
3991 Severity : enhancement
3993 Description: Add uninit_groups feature to ldiskfs2 to speed up e2fsck
3994 Details : The uninit_groups feature works in conjunction with the kernel
3995 filesystem code (ldiskfs2 only) and e2fsprogs-1.39-cfs6 to speed
3996 up the pass1 processing of e2fsck. This is a read-only feature
3997 in ldiskfs2 only, so older kernels and current ldiskfs cannot
3998 mount filesystems that have had this feature enabled.
4000 Severity : enhancement
4002 Description: Improve multi-block allocation algorithm to avoid fragmentation
4003 Details : The mballoc3 code (ldiskfs2 only) adds new mechanisms to improve
4004 allocation locality and avoid filesystem fragmentation.
4006 ------------------------------------------------------------------------------
4008 2007-02-09 Cluster File Systems, Inc. <info@clusterfs.com>
4010 * Support for kernels:
4011 2.6.9-42.0.3.EL (RHEL 4)
4012 2.6.5-7.276 (SLES 9)
4013 2.4.21-47.0.1.EL (RHEL 3)
4014 2.6.12.6 vanilla (kernel.org)
4015 2.6.16.21-0.8 (SLES10)
4016 * Recommended e2fsprogs version: 1.39.cfs2-0
4018 * The backwards-compatible /proc/sys/portals symlink has been removed
4019 in this release. Before upgrading, please ensure that you change
4020 any configuration scripts or /etc/sysctl.conf files that access
4021 /proc/sys/portals/* or sysctl portals.* to use the corresponding
4022 entry in /proc/sys/lnet or sysctl lnet.*. This change can be made
4023 in advance of the upgrade on any system running Lustre 1.4.6 or
4024 newer, since /proc/sys/lnet was added in that version.
4025 * Note that reiserfs quotas are disabled on SLES 10 in this kernel
4029 Frequency : only when quota is used
4031 Description: avoid scanning export list for quota master
4032 Details : Change the algorithms to avoid scanning export list in order
4033 to improve the efficiency.
4036 Frequency : MDS failover only, very rarely
4038 Description: "went back in time" messages on mds failover
4039 Details : The greatest transno may be lost when the current operation
4040 finishes with an error (transno==0) and the client's last_rcvd
4041 record is over-written. Save the greatest transno in the
4042 mds_last_transno for this case.
4045 Frequency : always for specific kernels and striping counts
4047 Description: client may get "Matching packet too big" without ACL support
4048 Details : Clients compiled without CONFIG_FS_POSIX_ACL get an error message
4049 when trying to access files in certain configurations. The
4050 clients should in fact be denied when mounting because they do
4051 not understand ACLs.
4054 Frequency : Cray XT3 with more than 4000 clients and multiple jobs
4056 Description: many clients connecting with IO in progress causes connect timeouts
4057 Details : Avoid synchronous journal commits to avoid delays caused by many
4058 clients connecting/disconnecting when bulk IO is in progress.
4059 Queue liblustre connect requests on OST_REQUEST_PORTAL instead of
4060 OST_IO_PORTAL to avoid delays behind potentially many pending
4064 Frequency : occasionally with multiple writers to a single file
4066 Description: shared writes to file may result in wrong size reported by stat()
4067 Details : Allow growing of kms when extent lock is cancelled
4070 Frequency : always with random mmap IO to multi-striped file
4072 Description: mmap write might be lost if we are writing to a 'hole' in stripe
4073 Details : Only if the hole is at the end of OST object so that kms is too
4074 small. Fix is to increase kms accordingly in ll_nopage.
4077 Frequency : rare, only if OST filesystem is inconsistent with MDS filesystem
4079 Description: writes to a missing object would leak memory on the OST
4080 Details : If there is an inconsistency between the MDS and OST filesystems,
4081 such that the MDS references an object that doesn't exist, writes
4082 to that object will leak memory due to incorrect cleanup in the
4083 error handling path, eventually running out of memory on the OST.
4088 Description: Creating too long symlink causes lustre errors
4089 Details : Check symlink and name lengths before sending requests to MDS.
4092 Frequency : only if flock is enabled (not on by default)
4094 Description: posix locks not released on fd closure on 2.6.9+
4095 Details : We failed to add posix locks to list of inode locks on 2.6.9+
4096 kernels, this caused such locks not to be released on fd close and
4097 then assertions on fs unmount about still used locks.
4100 Frequency : MDS failover only, very rarely
4102 Description: clients may get ASSERTION(granted_lock != NULL)
4103 Details : When request was taking a long time, and a client was resending
4104 a getattr by name lock request. The were multiple lock requests
4105 with the same client lock handle and
4106 mds_getattr_name->fixup_handle_for_resent_request found one of the
4107 lock handles but later failed with ASSERTION(granted_lock != NULL).
4112 Description: handle->h_buffer_credits > 0, assertion failure
4113 Details : h_buffer_credits is zero after truncate, causing assertion
4114 failure. This patch extends the transaction or creates a new
4118 Frequency : NFS re-export or patchless client
4119 Bugzilla : 11179, 10796
4120 Description: Crash on NFS re-export node (__d_move)
4121 Details : We do not want to hash the dentry if we don't have a lock.
4122 But if this dentry is later used in d_move, we'd hit uninitialised
4123 list head d_hash, so we just do this to init d_hash field but
4124 leave dentry unhashed.
4127 Frequency : NFS re-export or patchless client
4129 Description: NFS exports has problem with symbolic link
4130 Details : lustre client didn't properly install dentry when re-exported
4131 to NFS or running patchless client.
4134 Frequency : NFS re-export or patchless client
4136 Description: Various nfs/patchless fixes.
4137 Details : fixes reuse disconected alias for lookup process - this fixes
4138 warning "find_exported_dentry: npd != pd",
4139 fix permission error with open files at nfs.
4140 fix apply umask when do revalidate.
4143 Frequency : occasional
4145 Description: Crash on NFS re-export node
4146 Details : calling clear_page() on the wrong pointer triggered oops in
4147 generic_mapping_read().
4150 Frequency : rarely, using O_DIRECT IO
4152 Description: unaligned directio crashes client with LASSERT
4153 Details : check for unaligned buffers before trying any requests.
4156 Frequency : rarely, using CFS RAID5 patches in non-standard kernel series
4158 Description: stale data returned from RAID cache
4159 Details : If only a small amount of IO is done to the RAID device before
4160 reading it again it is possible to get stale data from the RAID
4161 cache instead of reading it from disk.
4164 Frequency : always for sles10 kernel
4166 Description: sles10 support
4167 Details : ll_follow_link: compile fixes and using of nd_set_link
4168 under newer kernels.
4171 Frequency : depends on arch, kernel and compiler version, always on sles10
4174 Description: recursive or deep enough symlinks cause stack overflow
4175 Details : getting rid of large stack-allocated variable in
4179 Frequency : depends on hardware
4181 Description: lustre write performance loss in the SLES10 kernel
4182 Details : the performance loss is caused by using of write barriers in the
4183 ext3 code. The SLES10 kernel turns barrier support on by
4184 default. The fix is to undo that change for ldiskfs.
4186 ------------------------------------------------------------------------------
4188 2006-12-09 Cluster File Systems, Inc. <info@clusterfs.com>
4190 * Support for kernels:
4191 2.6.9-42.0.3EL (RHEL 4)
4192 2.6.5-7.276 (SLES 9)
4193 2.4.21-47.0.1.EL (RHEL 3)
4194 2.6.12.6 vanilla (kernel.org)
4198 Frequency : quota enabled and large files being deleted
4200 Description: releasing more than 4GB of quota at once hangs OST
4201 Details : If a user deletes more than 4GB of files on a single OST it
4202 will cause the OST to spin in an infinite loop. Release
4203 quota in < 4GB chunks, or use a 64-bit value for 1.4.7.1+.
4208 Description: statfs data retrieved from /proc may be stale or zero
4209 Details : When reading per-device statfs data from /proc, in the
4210 {kbytes,files}_{total,free,avail} files, it may appear
4211 as zero or be out of date.
4214 Frequency : systems with MD RAID1 external journal devices
4216 Description: lconf's call to blkid is confused by RAID1 journal devices
4217 Details : Use the "blkid -l" flag to locate the MD RAID device instead
4218 of returning all block devices that match the journal UUID.
4221 Frequency : always, for aggregate stripe size over 4GB
4223 Description: "lfs setstripe" fails assertion when setting 4GB+ stripe width
4224 Details : Using "lfs setstripe" to set stripe size * stripe count over 4GB
4225 will fail the kernel with "ASSERTION(lsm->lsm_xfersize != 0)"
4228 Frequency : always if "lfs find" used on a local file/directory
4230 Description: "lfs find" segfaults if used on a local file/directory
4231 Details : The case where a directory component was not specified wasn't
4232 handled correctly. Handle this properly.
4235 Frequency : always on ppc64
4237 Description: the write to an ext3 filesystem mounted with mballoc got stuck
4238 Details : ext3_mb_generate_buddy() uses find_next_bit() which does not
4239 perform endianness conversion.
4242 Frequency : rarely (truncate to non-zero file size after write under load)
4243 Bugzilla : 10730, 10687
4244 Description: Files padded with zeros to next 4K multiple
4245 Details : With filesystems mounted using the "extents" option (2.6 kernels)
4246 it is possible that files that are truncated to a non-zero size
4247 immediately after being written are filled with zero bytes beyond
4248 the truncated size. No file data is lost.
4250 Severity : enhancement
4252 Description: Allow recovery/failover for liblustre clients.
4253 Details : liblustre clients were unaware of failover configurations until
4256 Severity : enhancement
4258 Description: user file locks should fail when not mounting with flock option
4259 Details : Set up an error-returning stub in ll_file_operations.lock field
4260 to prevent incorrect behaviour when client is mounted without
4261 flock option. Also, set up properly f_op->flock field for
4265 Frequency : always on ia64
4267 Description: "lfs df" loops on printing out MDS statfs information
4268 Details : The obd_ioctl_data was not initialized and in some systems
4269 this caused a failure during the ioctl that did not return
4270 an error. Initialize the struct and return an error on failure.
4273 Frequency : SLES 9 only
4275 Description: Error of copying files with lustre special EAs as root
4276 Details : Client side always return success for setxattr call for lustre
4277 special xattr (currently only "trusted.lov").
4280 Frequency : rarely on clusters with both ia64+i386 clients
4282 Description: ia64+i686 clients doing shared IO on the same file may LBUG
4283 Details : In rare cases when both ia64+i686 (or other mixed-PAGE_SIZE)
4284 clients are doing concurrent writes to the same file it is
4285 possible that the ia64 clients may LASSERT because the OST
4286 extent locks are not PAGE_SIZE aligned. Ensure that grown
4287 locks are always aligned on the request boundary.
4290 Frequency : specific use, occasional
4292 Description: Overwriting in use executable truncates on-disk binary image
4293 Details : If one node attempts to overwrite an executable in use by
4294 another node, we now correctly return ETXTBSY instead of
4295 truncating the file.
4297 Severity : enhancement
4299 Description: Async OSC create to avoid the blocking unnecessarily.
4300 Details : If a OST has no remain object, system will block on the creating
4301 when need to create a new object on this OST. Now, ways use
4302 pre-created objects when available, instead of blocking on an
4303 empty osc while others are not empty. If we must block, we block
4304 for the shortest possible period of time.
4309 Description: chmod on Lustre root is propagated to other clients
4310 Details : Re-validate root's dentry in ll_lookup_it to avoid having it
4311 invalid by the follow_mount time.
4314 Frequency : liblustre clients only
4316 Description: Race in 'instant cancel' lock handling could lead to such locks
4317 never to be granted in case of SMP MDS
4318 Details : Do not destroy not yet granted but cbpending locks in
4322 Frequency : replay/resend of open
4324 Description: non null lock assetion failure in mds_intent_policy
4325 Details : Trying to replay/resend lockless open requests resulted in
4326 mds_open() returning 0 with no lock. Now it sets a flag if
4327 a lock is going to be returned.
4329 Severity : enhancement
4331 Description: Checksum enhancements
4332 Details : New checksum enhancements allow for resending RPCs that failed
4335 Severity : enhancement
4337 Description: Tunables on number of dirty pages in cacche
4338 Details : Allow to set limit on number of dirty pages cached.
4343 Description: client crash on unmount - lock still has references
4344 Details : In some error handling cases it was possible to leak a lock
4345 reference on a client while accessing a file. This was not
4346 harmful to the client during operation, but would cause the
4347 client to crash when the filesystem is unmounted.
4350 Frequency : specific case, rare
4352 Description: ETXTBSY on mds though file not in use by client
4353 Details : ETXTBSY is no longer incorrectly returned when attempting to
4354 chmod or chown a directory that the user previously tried to
4355 execute or a currently-executing binary.
4358 Frequency : extremely rare except on liblustre-based clients
4360 Description: Lustre space not freed when files are deleted
4361 Details : Clean up open-unlinked files after client eviction. Previously
4362 the unlink was skipped and the files remained as orphans.
4367 Description: OST failure "would be an LBUG" in waiting_locks_callback()
4368 Details : In some cases it was possible to send a blocking callback to
4369 a client doing a glimpse, even though that client didn't get
4370 a lock granted. When the glimpse lock is cancelled on the OST
4371 the freed lock is left on the waiting list and corrupted the list.
4374 Frequency : all core dumps
4376 Description: Broke core dumps to lustre
4377 Details : Negative dentry may be unhashed if parent does not have UPDATE
4378 lock, but some callers, e.g. do_coredump, expect dentry to be
4379 hashed after successful create, hash it in ll_create_it.
4381 ------------------------------------------------------------------------------
4383 2006-09-13 Cluster File Systems, Inc. <info@clusterfs.com>
4385 * Support for kernels:
4386 2.6.9-42.0.2.EL (RHEL 4)
4387 2.6.5-7.276 (SLES 9)
4388 2.4.21-40.EL (RHEL 3)
4389 2.6.12.6 vanilla (kernel.org)
4393 Frequency : always on RHEL 3
4395 Description: Number of open files grows over time
4396 Details : The number of open files grows over time, whether or not
4397 Lustre is started. This was due to a filp leak introduced
4398 by one of our kernel patches.
4400 ------------------------------------------------------------------------------
4402 2006-08-20 Cluster File Systems, Inc. <info@clusterfs.com>
4404 * Support for kernels:
4405 2.6.9-42.EL (RHEL 4)
4406 2.6.5-7.267 (SLES 9)
4407 2.4.21-40.EL (RHEL 3)
4408 2.6.12.6 vanilla (kernel.org)
4413 Bugzilla : 5719, 9635, 9792, 9684
4414 Description: OST (or MDS) trips assertions in (re)connection under heavy load
4415 Details : If a server is under heavy load and cannot reply to new
4416 connection requests before the client resends the (re)connect,
4417 the connection handling code can behave badly if two service
4418 threads are concurrently handing separate (re)connections from
4419 the same client. Add better locking to the connection handling
4420 code, and ensure that only a single connection will be processed
4421 for a given client UUID, even if the lock is dropped.
4423 Severity : enhancement
4425 Description: add TCP zero-copy support to kernel
4426 Details : Add support to the kernel TCP stack to allow zero-copy bulk
4427 sends if the hardware supports scatter-gather and checksumming.
4428 This allows socklnd to do client-write and server-read more
4429 efficiently and reduce CPU utilization from skbuf copying.
4432 Frequency : only if NFS exporting from client
4434 Description: NULL pointer deref in ll_iocontrol() if chattr mknod file
4435 Details : If setting attributes on a file created under NFS that had
4436 never been opened it would be possible to oops the client
4437 if the file had no objects.
4441 Bugzilla : 9326, 10402, 10897
4442 Description: client crash in ptlrpcd_wake() thread when sending async RPC
4443 Details : It is possible that ptlrpcd_wake() dereferences a freed async
4444 RPC. In rare cases the ptlrpcd thread alread processed the RPC
4445 before ptlrpcd_wake() was called and the request was freed.
4448 Frequency : always for liblustre
4450 Description: liblustre client does MDS+OSTs setattr RPC for each write
4451 Details : When doing a write from a liblustre client, the client
4452 incorrectly issued an RPC to the MDS and each OST the file was
4453 striped over in order to update the timestamps. When writing
4454 with small chunks and many clients this could overwhelm the MDS
4455 with RPCs. In all cases it would slow down the write because
4456 these RPCs are unnecessary.
4458 Severity : enhancement
4460 Description: allow number of MDS service threads to be changed at module load
4461 Details : It is now possible to change the number of MDS service threads
4462 running. Adding "options mds mds_num_threads={N}" to the MDS's
4463 /etc/modprobe.conf will set the number of threads for the next
4464 time Lustre is restarted (assuming the "mds" module is also
4465 reloaded at that time). The default number of threads will
4466 stay the same, 32 for most systems.
4471 Description: OST crash if filesystem is unformatted or corrupt
4472 Details : If an OST is started on a device that has never been formatted
4473 or if the filesystem is corrupt and cannot even mount then the
4474 error handling cleanup routines would dereference a NULL pointer.
4479 Description: NULL pointer deref in llap_from_page.
4480 Details : get_cache_page_nowait can return a page with NULL (or otherwise
4481 incorrect) mapping if the page was truncated/reclaimed while it was
4482 searched for. Check for this condition and skip such pages when
4483 doing readahead. Introduce extra check to llap_from_page() to
4484 verify page->mapping->host is non-NULL (so page is not anonymous).
4487 Frequency : Sometimes when using sys_sendfile
4489 Description: "page not covered by a lock" warnings from ll_readpage
4490 Details : sendfile called ll_readpage without right page locks present.
4491 Now we introduced ll_file_sendfile that does necessary locking
4492 around call to generic_file_sendfile() much like we do in
4496 Frequency : with certain MDS communication failures at client mount time
4498 Description: NULL pointer deref after failed client mount
4499 Details : a client connection request may delayed by the network layer
4500 and not be sent until after the PTLRPC layer has timed out the
4501 request. If the client fails the mount immediately it will try
4502 to clean up before the network times out the request. Add a
4503 reference from the request import to the obd device and delay
4504 the cleanup until the network drops the request.
4507 Frequency : occasionally during client (re)connect
4509 Description: assertion failure during client (re)connect
4510 Details : processing a client connection request may be delayed by the
4511 client or server longer than the client connect timeout. This
4512 causes the client to resend the connection request. If the
4513 original connection request is replied in this interval, the
4514 client may trip an assertion failure in ptlrpc_connect_interpret()
4515 which thought it would be the only running connect process.
4518 Frequency : only with obd_echo servers and clients that are rebooted
4520 Description: kernel BUG accessing uninitialized data structure
4521 Details : When running an obd_echo server it did not start the ping_evictor
4522 thread, and when a client was evicted an uninitialized data
4523 structure was accessed. Start the ping_evictor in the RPC
4524 service startup instead of the OBD startup.
4526 Severity : enhancement
4527 Bugzilla : 10193 (patchless)
4528 Description: Remove dependency on various unexported kernel interfaces.
4529 Details : No longer need reparent_to_init, exit_mm, exit_files,
4530 sock_getsockopt, filemap_populate, FMODE_EXEC, put_filp.
4533 Frequency : rare (only users of deprecated and unsupported LDAP config)
4535 Description: write_conf for zeroconf mount queried LDAP incorrectly for client
4536 Details : LDAP apparently contains 'lustreName' attributes instead of
4537 'name'. A simple remapping of the name is sufficient.
4540 Frequency : rare (only with non-default dump_on_timeout debug enabled)
4542 Description: waiting_locks_callback trips kernel BUG if client is evicted
4543 Details : Running with the dump_on_timeout debug flag turned on makes
4544 it possible that the waiting_locks_callback() can try to dump
4545 the Lustre kernel debug logs from an interrupt handler. Defer
4546 this log dumping to the expired_lock_main() thread.
4548 Severity : enhancement
4550 Description: Support NFS exporting on 2.6 kernels.
4551 Details : Implement non-rawops metadata methods for NFS server to use without
4552 changing NFS server code.
4555 Frequency : very rare (synthetic metadata workload only)
4557 Description: two racing renames might cause an MDS thread to deadlock
4558 Details : Running the "racer" program may cause one MDS thread to rename
4559 a file from being the source of a rename to being the target of
4560 a rename at exactly the same time that another thread is doing
4561 so, and the second thread has already enqueued these locks after
4562 doing a lookup of the target and is trying to relock them in
4563 order. Ensure that we don't try to re-lock the same resource.
4566 Frequency : only very large systems with liblustre clients
4568 Description: slow eviction of liblustre clients with the "evict_by_nid" RPC
4569 Details : Use asynchronous set_info RPCs to send the "evict_by_nid" to
4570 all OSTs in parallel. This allows the eviction of stale liblustre
4571 clients to proceed much faster than if they were done in series,
4572 and also offers similar improvements for other set_info RPCs.
4577 Description: excessive CPU usage during initial read phase on client
4578 Details : During the initial read phase on a client, it would agressively
4579 retry readahead on the file, consuming too much CPU and impacting
4580 performance (since 1.4.5.8). Improve the readahead algorithm
4581 to avoid this, and also improve some other common cases (read
4582 of small files in particular, where "small" is files smaller than
4583 /proc/fs/lustre/llite/*/max_read_ahead_whole_mb, 2MB by default).
4588 Description: MDS crash when receiving packet with unknown intent.
4589 Details : Do not LBUG in unknown intent case, just return -EFAULT
4591 Severity : enhancement
4592 Bugzilla : 9293, 9385
4593 Description: MDS RPCs are serialised on client. This is unnecessary for some.
4594 Details : Do not serialize getattr (non-intent version) and statfs.
4597 Frequency : occasional, when OST network is overloaded/intermittent
4599 Description: client evicted by OST after bulk IO timeout
4600 Details : If a client sends a bulk IO request (read or write) the OST
4601 may evict the client if it is unresposive to its data GET/PUT
4602 request. This is incorrect if the network is overloaded (takes
4603 too long to transfer the RPC data) or dropped the OST GET/PUT
4604 request. There is no need to evict the client at all, since
4605 the pinger and/or lock callbacks will handle this, and the
4606 client can restart the bulk request.
4609 Frequency : Always when mmapping file with no objects
4611 Description: client crashes when mmapping file with no objects
4612 Details : Check that we actually have objects in a file before doing any
4613 operations on objects in ll_vm_open, ll_vm_close and
4619 Description: Request leak when working with deleted CWD
4620 Details : Introduce advanced request refcount tracking for requests
4621 referenced from lustre intent.
4623 Severity : Enhancement
4625 Description: Cache open file handles on client.
4626 Details : MDS now will return special lock along with openhandle, if
4627 requested and client is allowed to hold openhandle, even if unused,
4628 until such a lock is revoked. Helps NFS a lot, since NFS is opening
4629 closing files for every read/write openration.
4631 Severity : Enhancement
4633 Description: Cache open negative dentries on client when possible.
4634 Details : Guard negative dentries with UPDATE lock on parent dir, drop
4635 negative dentries on lock revocation.
4640 Description: Remounting a client read-only wasn't possible with a zconf mount
4641 Details : It wasn't possible to remount a client read-only with llmount.
4643 Severity : enhancement
4644 Description: Include MPICH 1.2.6 Lustre ADIO interface patch
4645 Details : In lustre/contrib/ or /usr/share/lustre in RPM a patch for
4646 MPICH is included to add Lustre-specific ADIO interfaces.
4647 This is based closely on the UFS ADIO layer and only differs
4648 in file creation, in order to allow the OST striping to be set.
4649 This is user-contributed code and not supported by CFS.
4654 Description: extended inode attributes (immutable, append-only) work improperly
4655 when 2.4 and 2.6 kernels are used on client/server or vice versa
4656 Details : Introduce kernel-independent values for these flags.
4658 Severity : enhancement
4661 Description: Allow fractional MB tunings for lustre in /proc/ filesystem.
4662 Details : Many of the /proc/ tunables can only be tuned at a megabyte
4663 granularity. Now, Fractional MB granularity is be supported,
4664 this is very useful for low memory system.
4666 Severity : enhancement
4668 Description: Getattr by fid
4669 Details : Getting a file attributes by its fid, obtaining UPDATE|LOOKUP
4670 locks, avoids extra getattr rpc requests to MDS, allows '/' to
4671 have locks and avoids getattr rpc requests for it on every stat.
4674 Frequency : Always, for filesystems larger than 2TB
4676 Description: ldiskfs crash at mount for filesystem larger than 2TB with mballoc
4677 Details : Kenrel kmalloc limits allocations to 128kB and this prevents
4678 filesystems larger than 2TB to be mounted with mballoc enabled.
4681 Frequency : Always, for 32-bit kernel without CONFIG_LBD and filesystem > 2TB
4683 Description: filesystem corruption for non-standard kernels and very large OSTs
4684 Details : If a 32-bit kernel is compiled without CONFIG_LBD enabled and a
4685 filesystems larger than 2TB is mounted then the kernel will
4686 silently corrupt the start of the filesystem. CONFIG_LBD is
4687 enabled for all CFS-supported kernels, but the possibility of
4688 this happening with a modified kernel config exists.
4690 Severity : enhancement
4692 Description: add client O_DIRECT support for 2.6 kernels
4693 Details : It is now possible to do O_DIRECT reads and writes to files
4694 in the Lustre client mountpoint on 2.6 kernel clients.
4696 Severity : enhancement
4698 Description: parallel glimpse, setattr, statfs, punch, destroy requests
4699 Details : Sends glimpse, setattr, statfs, punch, destroy requests to OSTs in
4700 parallel, not waiting for response from every OST before sending
4701 a rpc to the next OST.
4706 Description: setattr vs write race when updating file timestamps
4707 Details : Client processes that update a file timestamp into the past
4708 right after writing to the file (e.g. tar) it is possible that
4709 the updated file modification time can be reset to the current
4710 time due to a race between processing the setattr and write RPC.
4712 Severity : enhancement
4714 Description: Bring 'lfs find' closer in line with regular Linux find.
4715 Details : lfs find util supports -atime, -mtime, -ctime, -maxdepth, -print,
4716 -print0 options and obtains all the needed info through the lustre
4719 Severity : enhancement
4721 Description: support up to 1024 configured devices on one node
4722 Details : change obd_dev array from statically allocated to dynamically
4723 allocated structs as they are first used to reduce memory usage
4728 Description: Flush dirty partially truncated pages during truncate
4729 Details : Immediatelly flush partially truncated pages in filter_setattr,
4730 this way we completely avoid having any pages in page cache on OST
4731 and can retire ugly workarounds during writes to flush such pages.
4736 Description: i_sem vs transaction deadlock in mds_obd_destroy during unlink.
4737 Details : protect inode from truncation within vfs_unlink() context
4738 just take a reference before calling vfs_unlink() and release it
4739 when parent's i_sem is free.
4742 Frequency : always, if extents are used on OSTs
4744 Description: index ei_leaf_hi (48-bit extension) is not zeroed in extent index
4745 Details : OSTs using the extents format would not zero the high 16 bits of
4746 the index physical block number. This is not a problem for any
4747 OST filesystems smaller than 16TB, and no kernels support ext3
4748 filesystems larger than 16TB yet. This is fixed in 1.4.7 (all
4749 new/modified files) and can be fixed for existing filesystems
4750 with e2fsprogs-1.39-cfs1.
4755 Description: import connection selection may be incorrect if timer wraps
4756 Details : Using a 32-bit jiffies timer with HZ=1000 may cause backup
4757 import connections to be ignored if the 32-bit jiffies counter
4758 wraps. Use a 64-bit jiffies counter.
4761 Frequency : during server recovery
4763 Description: crash after server is denying duplicate export
4764 Details : If clients are resending connect requests to the server, the
4765 server refuses to allow a client to connect multiple times.
4766 Fixed a bug in the handling of this case.
4769 Frequency : very large clusters immediately after boot
4771 Description: LNET request buffers exhausted under heavy short-term load
4772 Details : If a large number of client requests are generated on a service
4773 that has previously never seen so many requests it is possible
4774 that the request buffer growth cannot keep up with the spike in
4775 demand. Instead of dropping incoming requests, they are held in
4776 the LND until the RPC service can accept more requests.
4779 Frequency : Sometimes during replay
4781 Description: Assertion failure in ll_local_open after replay.
4782 Details : If replay happened on an open request reply before we were able
4783 to set replay handler, reply will become not swabbed tripping the
4784 assertion in ll_local_open. Now we set the handler right after
4785 recognising of open request
4788 Frequency : very rare
4790 Description: kernel reports "badness in vsnprintf"
4791 Details : Reading from the "recovery_status" /proc file in small chunks
4792 may cause a negative length in lprocfs_obd_rd_recovery_status()
4793 call to vsnprintf() (which is otherwise harmless). Exit early
4794 if there is no more space in the output buffer.
4796 Severity : enhancement
4798 Description: clear OBD RPC statistics by writing to them
4799 Details : It is now possible to clear the OBD RPC statistics by writing
4800 to the "stats" file.
4805 Description: Client mtime is not the same on different clients after utimes
4806 Details : In some cases, the client was using the utimes() syscall on
4807 a file cached on another node. The clients now validate the
4808 ctime from the MDS + OSTs to determine which one is right.
4813 Description: Inability to activate failout mode
4814 Details : lconf script incorrectly assumed that in python string's numeric
4815 value is used in comparisons.
4818 Frequency : always with multiple stripes per file
4820 Description: Inefficient object allocation for mutli-stripe files
4821 Details : When selecting which OSTs to stripe files over, for files with
4822 a stripe count that divides evenly into the number of OSTs,
4823 the MDS is always picking the same starting OST for each file.
4824 Return the OST selection heuristic to the original design.
4829 Description: mount failures may take full timeout to return an error
4830 Details : Under some heavy load conditions it is possible that a
4831 failed mount can wait for the full obd_timeout interval,
4832 possibly several minutes, before reporting an error.
4833 Instead return an error as soon as the status is known.
4835 ------------------------------------------------------------------------------
4837 2006-02-14 Cluster File Systems, Inc. <info@clusterfs.com>
4839 * WIRE PROTOCOL CHANGE. This version of Lustre networking WILL NOT
4840 INTEROPERATE with older versions automatically. Please read the
4841 user documentation before upgrading any part of a live system.
4842 * WARNING: Lustre networking configuration changes are required with
4843 this release. See https://bugzilla.clusterfs.com/show_bug.cgi?id=10052
4846 * Support for kernels:
4847 2.6.9-22.0.2.EL (RHEL 4)
4848 2.6.5-7.244 (SLES 9)
4849 2.6.12.6 vanilla (kernel.org)
4852 Severity : enhancement
4853 Bugzilla : 7981/8208
4854 Description: Introduced Lustre Networking (LNET)
4855 Details : LNET is new networking infrastructure for Lustre, it includes
4856 a reorganized network configuration mode (see the user
4857 documentation for full details) as well as support for routing
4858 between different network fabrics. Lustre Networking Devices
4859 (LNDs) for the supported network fabrics have also been
4860 created for this new infrastructure.
4862 Severity : enhancement
4863 Description: Introduced Access control lists
4864 Details : clients can set ACLs on files and directories in order to have
4865 more fine-grained permissions than the standard Unix UGO+RWX.
4866 The MDS must be started with the "-o acl" mount option.
4868 Severity : enhancement
4869 Description: Introduced filesystem quotas
4870 Details : Administrators may now establish per-user quotas on the
4873 Severity : enhancement
4875 Description: Configuration change for the XT3
4876 The PTLLND is now used to run Lustre over Portals on the XT3
4877 The configure option(s) --with-cray-portals are no longer used.
4878 Rather --with-portals=<path-to-portals-includes> is used to
4879 enable building on the XT3. In addition to enable XT3 specific
4880 features the option --enable-cray-xt3 must be used.
4885 Description: Running on many-way SMP OSTs can trigger oops in llcd_send()
4886 Details : A race between allocating a new llcd and re-getting the llcd_lock
4887 allowed another thread to grab newly-allocated llcd.
4889 Severity : enhancement
4891 Description: 2.6 OST async journal commit and locking fix to improve performance
4892 Details : The filter_direct_io()+filter_commitrw_write() journal commits for
4893 2.6 kernels are now async as they already were in 2.4 kernels so
4894 that they can commit concurrently with the network bulk transfer.
4895 For block-allocated files the filter allocation semaphore is held
4896 to avoid filesystem fragmentation during allocation. BKL lock
4897 removed for 2.6 xattr operations where it is no longer needed.
4902 Description: lconf incorrectly determined whether two IP networks could talk
4903 Details : In some more complicated routing and multiple-network
4904 configurations, lconf will avoid trying to make a network
4905 connection to a disjoint part of the IP space. It was doing the
4906 math incorrectly for one set of cases.
4911 Description: Fix for potential infinite loop processing records in an llog.
4912 Details : If an llog record is corrupted/zeroed, it is possible to loop
4913 forever in llog_process(). Validate the llog record length
4914 and skip the remainder of the block on error.
4917 Frequency : occasional (liblustre only)
4919 Description: liblustre could not open files whose last component is a symlink
4920 Details : sysio_path_walk() would incorrectly pass the open intent to
4921 intermediate path components.
4924 Frequency : rare (liblustre only with non-standard tuning)
4925 Bugzilla : 7201 (7350)
4926 Description: Tuning the MDC DLM LRU size to zero triggers client LASSERT
4927 Details : llu_lookup_finish_locks() tries to set lock data on a lock
4928 after it has been released, only do this for referenced locks
4930 Severity : enhancement
4932 Description: specifying an (invalid) directory default stripe_size of -1
4933 would reset the directory default striping
4934 Details : stripe_size -1 was used internally to signal directory stripe
4935 removal, now use "all default" to signal dir stripe removal
4936 as a directory striping of "all default" is not useful
4939 Frequency : common for large clusters running liblustre clients
4941 Description: doing an ls when liblustre clients are running is slow
4942 Details : sending a glimpse AST to a liblustre client waits for every AST
4943 to time out, as liblustre clients will not respond. Since they
4944 cannot cache data we refresh the OST lock LVB from disk instead.
4946 Severity : enhancement
4948 Description: doing an ls at the same time as file IO can be slow
4949 Details : enqueue and other "small" requests can be blocked behind many
4950 large IO requests. Create a new OST IO portal for non-IO
4951 requests so they can be processed faster.
4954 Frequency : rare (only HPUX clients mounting unsupported re-exported NFS vol)
4956 Description: an HPUX NFS client would get -EACCESS when ftruncate()ing a newly
4957 created file with mode 000
4958 Details : the Linux NFS server relies on an MDS_OPEN_OWNEROVERRIDE hack to
4959 allow an ftruncate() as a non-root user to a file with mode 000.
4960 Lustre now respects this flag to disable mode checks when
4961 truncating a file owned by the user
4964 Frequency : liblustre-only, when liblustre client dies unexpectedly or becomes
4967 Description: Revoking locks from clients that went dead or catatonic might take
4969 Details : New lock flags FL_CANCEL_ON_BLOCK used by liblustre makes
4970 cancellation of such locks instant on servers without waiting for
4971 any reply from clients. Clients drops these locks when cancel
4972 notification from server is received without replying.
4975 Frequency : liblustre-only, when liblustre client dies or becomes busy
4977 Description: Doing ls on Linux clients can take a long time with active
4979 Details : Liblustre client cannot handle ASTs in timely manner, so avoid
4980 granting such locks to it in the first place if possible. Locks
4981 are taken by proxy on the OST during the read or write and
4982 dropped immediately afterward. Add connect flags handling, do
4983 not grant locks to liblustre clients for glimpse ASTs.
4985 Severity : enhancement
4987 Description: Improve read-ahead algorithm to avoid excessive IO for random reads
4988 Details : Existing read-ahead algorithm is tuned for the case of streamlined
4989 sequential reads and behaves badly with applications doing random
4990 reads. Improve it by reading ahead at least read region, and
4991 avoiding excessive large RPC for small reads.
4993 Severity : enhancement
4995 Description: Creating more than 1000 files for a single job may cause a load
4996 imbalance on the OSTs if there are also a large number of OSTs.
4997 Details : qos_prep_create() uses an OST index reseed value that is an
4998 even multiple of the number of available OSTs so that if the
4999 reseed happens in the middle of the object allocation it will
5000 still utilize the OSTs as uniformly as possible.
5005 Description: OST or MDS may oops in ping_evictor_main()
5006 Details : ping_evictor_main() drops obd_dev_lock if deleting a stale export
5007 but doesn't restart at beginning of obd_exports_timed list
5010 Severity : enhancement
5012 Description: improve by-nid export eviction on the MDS and OST
5013 Details : allow multiple exports with the same NID to be evicted at one
5014 time without re-searching the exports list.
5017 Frequency : rare, only with supplementary groups enabled on SMP 2.6 kernels
5019 Description: MDS may oops in groups_free()
5020 Details : in rare race conditions a newly allocated group_info struct is
5021 freed again, and this can be NULL. The 2.4 compatibility code
5022 for groups_free() checked for a NULL pointer, but 2.6 did not.
5025 Frequency : common for liblustre clients doing little filesystem IO
5026 Bugzilla : 9352, 7313
5027 Description: server may evict liblustre clients accessing contended locks
5028 Details : if a client is granted a lock or receives a completion AST
5029 with a blocking AST already set it would not reply to the AST
5030 for LDLM_FL_CANCEL_ON_BLOCK locks. It now replies to such ASTs.
5033 Frequency : lfs setstripe, only systems with more than 160 OSTs
5035 Description: unable to set striping with a starting offset beyond OST 160
5036 Details : llapi_create_file() incorrectly limited the starting stripe
5037 index to the maximum single-file stripe count.
5040 Frequency : LDAP users only
5042 Description: lconf did not handle in-kernel recovery with LDAP properly
5043 Details : lconf/LustreDB get_refs() is searching the wrong namespace
5045 Severity : enhancement
5047 Description: bind OST threads to NUMA nodes to improve performance
5048 Details : all OST threads are uniformly bound to CPUs on a single NUMA
5049 node and do their allocations there to localize memory access
5051 Severity : enhancement
5053 Description: llmount can determine client NID directly from Myrinet (GM)
5054 Details : the client NID code from gmnalnid was moved directly into
5055 llmount, removing the need to use this or specifying the
5056 client NID explicitly when mounting GM clients with zeroconf
5059 Frequency : if client is started with down MDS
5061 Description: if client is started with down MDS mount hangs in ptlrpc_queue_wait
5062 Details : Having an LWI_INTR() wait event (interruptible, but no timeout)
5063 will wait indefinitely in ptlrpc_queue_wait->l_wait_event() after
5064 ptlrpc_import_delayed_req() because we didn't check if the
5065 request was interrupted, and we also didn't break out of the
5066 event loop if there was no timeout
5071 Description: data loss during non-page-aligned writes to a single file from
5072 both multiple nodes and multiple threads on one node at same time
5073 Details : updates to KMS and lsm weren't protected by common lock. Resulting
5074 inconsistency led to false short-reads, that were cached and later
5075 used by ->prepare_write() to fill in partially written page,
5076 leading to data loss.
5079 Frequency : always, if lconf --abort_recovery used
5081 Description: lconf --abort_recovery fails with 'Operation not supported'
5082 Details : lconf was attempting to abort recovery on the MDT device and not
5085 Severity : enhancement
5087 Description: remove cleanup logs
5088 Details : replace lconf-generated cleanup logs with lustre internal
5089 cleanup routines. Eliminates the need for client-cleanup and
5092 Severity : enhancement
5094 Description: add support for EAs (user and system) on lustre filesystems
5095 Details : it is now possible to store extended attributes in the Lustre
5096 client filesystem, and with the user_xattr mount option it
5097 is possible to allow users to store EAs on their files also
5099 Severity : enhancement
5101 Description: Add possibility (config option) to show minimal available OST free
5103 Details : When compiled with --enable-mindf configure option, statfs(2)
5104 (and so, df) will return least minimal free space available from
5105 all OSTs as amount of free space on FS, instead of summary of
5106 free spaces of all OSTs.
5108 Severity : enhancement
5110 Description: do not expand extent locks acquired on OST-side
5111 Details : Modify ldlm_extent_policy() to not expand local locks, acquired
5112 by server: they are not cached anyway.
5115 Frequency : when mmap is used/binaries executed from Lustre
5117 Description: Unmmap pages before throwing them away from read cache.
5118 Details : llap_shrink cache now attempts to unmap pages before discarding
5119 them (if unmapping failed - do not discard). SLES9 kernel has
5120 extra checks that trigger if this unmapping is not done first.
5125 Description: lconf didn't resolve symlinks before checking to see whether a
5126 given mountpoint was already in use
5129 Frequency : when migrating failover services
5130 Bugzilla : 6395, 9514
5131 Description: When migrating a subset of services from a node (e.g. failback
5132 from a failover service node) the remaining services would
5133 time out and evict clients.
5134 Details : lconf --force (implied by --failover) sets the global obd_timeout
5135 to 5 seconds in order to quickly disconnect, but this caused
5136 other RPCs to time out too quickly. Do not change the global
5137 obd_timeout for force cleanup, only set it for DISCONNECT RPCs.
5139 Severity : enhancement
5140 Frequency : if MDS is started with down OST
5141 Bugzilla : 9439,5706
5142 Description: Allow startup/shutdown of an MDS without depending on the
5143 availability of the OSTs.
5144 Details : Asynchronously call mds_lov_synchronize during MDS startup.
5145 Add appropriate locking and lov-osc refcounts for safe
5146 cleaning. Add osc abort_inflight calls in case the
5147 synchronize never started.
5150 Frequency : occasional (Cray XT3 only)
5152 Description: root not authorized to access files in CRAY_PORTALS environment
5153 Details : The client process capabilities were not honoured on the MDS in
5154 a CRAY_PORTALS/CRAY_XT3 environment. If the file had previously
5155 been accessed by an authorized user then root was able to access
5156 the file on the local client also. The root user capabilities
5157 are now allowed on the MDS, as this environment has secure UID.
5160 Frequency : occasional
5162 Description: ldiskfs "too long searching" message happens too often
5163 Details : A debugging message (otherwise harmless) prints too often on
5164 the OST console. This has been reduced to only happen when
5165 there are fragmentation problems on the filesystem.
5170 Description: Division by zero in statfs when all OSCs are inactive
5171 Details : lov_get_stripecnt() returns zero due to incorrect order of checks,
5172 lov_statfs divides by value returned by lov_get_stripecnt().
5176 Bugzilla : 9489, 3273
5177 Description: First write from each client to each OST was only 4kB in size,
5178 to initialize client writeback cache, which caused sub-optimal
5179 RPCs and poor layout on disk for the first writen file.
5180 Details : Clients now request an initial cache grant at (re)connect time
5181 and so that they can start streaming writes to the cache right
5182 away and always do full-sized RPCs if there is enough data.
5183 If the OST is rebooted the client also re-establishes its grant
5184 so that client cached writes will be honoured under the grant.
5189 Description: Slow ls (and stat(2) syscall) on files residing on IO-loaded OSTs
5190 Details : Now I/O RPCs go to different portal number and (presumably) fast
5191 lock requests (and glimses) and other RPCs get their own service
5192 threads pool that should be able to service those RPCs
5195 Severity : enhancement
5197 Description: Ability to exchange lustre version between client and servers and
5198 issue warnings at client side if client is too old. Also for
5199 liblustre clients there is ability to refuse connection of too old
5201 Details : New 'version' field is added to connect data structure that is
5202 filled with version info. That info is later checked by server and
5206 Frequency : rare, liblustre only.
5207 Bugzilla : 9296, 9581
5208 Description: Two simultaneous writes from liblustre at offset within same page
5209 might proceed at the same time overwriting eachother with stale
5211 Details : I/O lock withing llu_file_prwv was released too early, before data
5212 actually was hitting the wire. Extended lock-holding time until
5213 server acknowledges receiving data.
5216 Frequency : extremely rare. Never observed in practice.
5218 Description: avoid generating lustre_handle cookie of 0.
5219 Details : class_handle_hash() generates handle cookies by incrementing
5220 global counter, and can hit 0 occasionaly (this is unlikely, but
5221 not impossible, because initial value of cookie counter is
5222 selected randonly). Value of 0 is used as a sentinel meaning
5223 "unassigned handle" --- avoid it. Also coalesce two critical
5224 sections in this function into one.
5226 Severity : enhancement
5228 Description: allow liblustre clients to delegate truncate locking to OST
5229 Details : To avoid overhead of locking, liblustre client instructs OST to
5230 take extent lock in ost_punch() on client's behalf. New connection
5231 flag is added to handle backward compatibility.
5233 Severity : enhancement
5234 Bugzilla : 4928, 7341, 9758
5235 Description: allow number of OST service threads to be specified
5236 Details : a module parameter allows the number of OST service threads
5237 to be specified via "options ost ost_num_threads={N}" in the
5238 OSS's /etc/modules.conf or /etc/modprobe.conf.
5242 Bugzilla : 6146, 9635, 9895
5243 Description: servers crash with bad pointer in target_handle_connect()
5244 Details : In rare cases when a client is reconnecting it was possible that
5245 the connection request was the last reference for that export.
5246 We would temporarily drop the export reference and get a new
5247 one, but this may have been the last reference and the export
5248 was just destroyed. Get new reference before dropping old one.
5250 Severity : enhancement
5251 Frequency : if client is started with failover MDS
5253 Description: Allow multiple MDS hostnames in the mount command
5254 Details : Try to read the configuration from all specified MDS
5255 hostnames during a client mount in case the "primary"
5258 Severity : enhancement
5260 Description: Stop sending data to evicted clients as soon as possible.
5261 Details : Check if the client we are about to send or are sending data to
5262 was evicted already. (Check is done every second of waiting,
5263 for which l_wait_event interface was extended to allow checking
5264 of exit condition at specified intervals).
5267 Frequency : rare, normally only when NFS exporting is done from client
5269 Description: 'bad disk LOV MAGIC: 0x00000000' error when chown'ing files
5271 Details : Make mds_get_md() recognise empty md case and set lmm size to 0.
5274 Frequency : always, if srand() is called before liblustre initialization
5276 Description: Liblustre uses system PRNG disturbing its usage by user application
5277 Details : Introduce internal to lustre fast and high-quality PRNG for
5278 lustre usage and make liblustre and some other places in generic
5279 lustre code to use it.
5281 Severity : enhancement
5282 Bugzilla : 9477, 9557, 9870
5283 Description: Verify that the MDS configuration logs are updated when xml is
5284 Details : Check if the .xml configuration logs are newer than the config
5285 logs stored on the MDS and report an error if this is the case.
5286 Request --write-conf, or allow starting with --old_conf.
5288 Severity : enhancement
5290 Description: Handle symlinks in the path when checking if Lustre is mounted.
5291 Details : Resolve intermediate symlinks when checking if a client has
5292 mounted a filesystem to avoid duplicate client mounts.
5297 Description: lconf can hit an error exception but still return success.
5298 Details : The lconf command catches the Command error exception at the top
5299 level script context and will exit with the associated exit
5300 status, but doesn't ensure that this exit status is non-zero.
5305 Description: failure of ptlrpc thread startup can cause oops
5306 Details : Starting a ptlrpc service thread can fail if there are a large
5307 number of threads or the server memory is very fragmented.
5308 Handle this without oopsing.
5311 Frequency : always, only if liblustre and non-default acceptor port was used
5313 Description: liblustre cannot connect to servers with non-default acceptor port
5314 Details : tcpnal_set_default_params() was not called and was therefore
5315 ignoring the environment varaible TCPNAL_PORT, as well as other
5316 TCPNAL_ environment variables
5321 Description: two objects could be created on the same OST for a single file
5322 Details : If an OST is down, in some cases it was possible to create two
5323 objects on a single OST for a single file. No problems other
5324 than potential performance impact and spurious error messages.
5328 Bugzilla : 5681, 9562
5329 Description: Client may oops in ll_unhash_aliases
5330 Details : Client dcache may become inconsistent in race condition.
5331 In some cases "getcwd" can fail if the current directory is
5337 Description: Inode refcounting problems in NFS export code
5338 Details : link_raw functions used to call d_instantiate without obtaining
5339 extra inode reference first.
5343 Bugzilla : 9942, 9903
5344 Description: Referencing freed requests leading to crash, memleaks with NFS.
5345 Details : We used to require that call to ll_revalidate_it was always
5346 followed by ll_lookup_it. Also with revalidate_special() it is
5347 possible to call ll_revalidate_it() twice for the same dentry
5348 even if first occurence returned success. This fix changes semantic
5349 between DISP_ENQ_COMPLETE disposition flag to mean there is extra
5350 reference on a request referred from the intent.
5351 ll_intent_release() then releases such a request.
5354 Frequency : rare, normally benchmark loads only
5356 Description: unlinked inodes were kept in memory on the client
5357 Details : If a client is repeatedly creating and unlinking files it
5358 can accumulate a lot of stale inodes in the inode slab cache.
5359 If there is no other client load running this can cause the
5360 client node to run out of memory. Instead flush old inodes
5361 from client cache that have the same inode number as a new inode.
5364 Frequency : SLES9 2.6.5 kernel and long filenames only
5365 Bugzilla : 9969, 10379
5366 Description: utime reports stale NFS file handle
5367 Details : SLES9 uses out-of-dentry names in some cases, which confused
5368 the lustre dentry revalidation. Change it to always use the
5372 Frequency : rare, unless heavy write-truncate concurrency is continuous
5373 Bugzilla : 4180, 6984, 7171, 9963, 9331
5374 Description: OST becomes very slow and/or deadlocked during object unlink
5375 Details : filter_destroy() was holding onto the parent directory lock
5376 while truncating+unlinking objects. For very large objects this
5377 may block other threads for a long time and slow overall OST
5378 responsiveness. It may also be possible to get a lock ordering
5379 deadlock in this case, or run out of journal credits because of
5380 the combined truncate+unlink. Solution is to do object truncate
5381 first in one transaction without parent lock, and then do the
5382 final unlink in a new transaction with the parent lock. This
5383 reduces the lock hold time dramatically.
5386 Frequency : rare, 2.4 kernels only
5388 Description: MDS or OST cleanup may trip kernel BUG when dropping kernel lock
5389 Details : mds_cleanup() and filter_cleanup() need to drop the kernel lock
5390 before unmounting their filesystem in order to avoid deadlock.
5391 The kernel_locked() function in 2.4 kernels only checks whether
5392 the kernel lock is held, not whether it is this process that is
5393 holding it as 2.6 kernels do.
5398 Description: MDS or OST may oops/LBUG if a client is connecting multiple times
5399 Details : The client ptlrpc code may be trying to reconnect to a down
5400 server before a previous connection attempt has timed out.
5401 Increase the reconnect interval to be longer than the connection
5402 timeout interval to avoid sending duplicate connections to
5406 Frequency : echo_client brw_test command
5408 Description: fix echo_client to work with OST preallocated code
5409 Details : OST preallocation code (5137) didn't take echo_client IO path
5410 into account: echo_client calls filter methods outside of any
5411 OST thread and, hence, there is no per-thread preallocated
5412 pages and buffers to use. Solution: hijack pga pages for IO. As
5413 a byproduct, this avoids unnecessary data copying.
5417 Bugzilla : 3555, 5962, 6025, 6155, 6296, 9574
5418 Description: Client can oops in mdc_commit_close() after open replay
5419 Details : It was possible for the MDS to return an open request with no
5420 transaction number in mds_finish_transno() if the client was
5421 evicted, but without actually returning an error. Clients
5422 would later try to replay that open and may trip an assertion
5423 Simplify the client close codepath, and always return an error
5424 from the MDS in case the open is not successful.
5427 Frequency : rare, 2.6 OSTs only
5429 Description: OST may deadlock under high load on fragmented files
5430 Details : If there was a heavy load and highly-fragmented OST filesystems
5431 it was possible to have all the OST threads deadlock waiting on
5432 allocation of biovecs, because the biovecs were not released
5433 until the entire RPC IO was completed. Instead, release biovecs
5434 as soon as they are complete to ensure forward IO progress.
5436 Severity : enhancement
5438 Description: Support for specifying external journal device at mount
5439 Details : If an OST or MDS device is formatted with an external journal
5440 device, this device major/minor is stored in the ext3 superblock
5441 and may not be valid for failover. Allow detecting and
5442 specifying the external journal at mount time.
5447 Description: Mounting an MDS with pending unlinked files may cause oops
5448 Details : target_finish_recovery() calls mds_postrecov() which returned
5449 the number of orphans unlinked. mds_lov_connect->mds_postsetup()
5450 considers this an error and immediately begins cleaning up the
5451 lov, just after starting the mds_lov process
5453 Severity : enhancement
5455 Description: Implement 'lfs df' to report actual free space on per-OST basis
5456 Details : Add sub-command 'df' on 'lfs' to report the disk space usage of
5457 MDS/OSDs. Usage: lfs df [-i][-h]. Command Options: '-i' to report
5458 usage of objects; '-h' to report in human readable format.
5460 ------------------------------------------------------------------------------
5462 2005-08-26 Cluster File Systems, Inc. <info@clusterfs.com>
5469 Description: Mounting an ldiskfs file system with mballoc may crash OST node.
5470 Details : ldiskfs mballoc code may reference an uninitialized buddy struct
5471 at startup during orphan unlinking. Instead, skip buddy update
5472 before setup, as it will be regenerated after recovery is complete.
5477 Description: If an OST is inactive, its locks might reference stale inodes.
5478 Details : lov_change_cbdata() must iterate over all namespaces, even if
5479 they are inactive to clear inode references from the lock.
5481 Severity : enhancement
5482 Frequency : occasional, if non-standard max_dirty_mb used
5484 Description: Client will block write RPCs if not enough grant
5485 Details : If a client has max_dirty_mb smaller than max_rpcs_in_flight,
5486 then the client will block writes while waiting for another RPC
5487 to complete instead of consuming its dirty limit. With change
5488 we get improved performance when max_dirty_mb is small.
5490 Severity : enhancement
5491 Bugzilla : 3389, 6253
5492 Description: Add support for supplementary groups on the MDS.
5493 Details : The MDS has an upcall /proc/fs/lustre/mds/{mds}/group_upcall
5494 (set to /usr/sbin/l_getgroups if enabled) which will do MDS-side
5495 lookups for user supplementary groups into a cache.
5499 Description: O_CREAT|O_EXCL open flags in liblustre always return -EEXIST
5500 Details : Make libsysio to not enforce O_EXCL by clearing the flag,
5501 for liblustre O_EXCL is enforced by MDS.
5505 Description: readdir never returns NULL in liblustre.
5506 Details : Corrected llu_iop_getdirentries logic, to return offset of next
5507 dentry in struct dirent.
5511 Frequency : liblustre only, depends on application IO pattern
5512 Description: liblustre clients evicted if not contacting servers
5513 Details : Don't put liblustre clients into the ping_evictor list, so
5514 they will not be evicted by the pinger ever.
5516 Severity : enhancement
5518 Description: Add ability to evict clients by NID from MDS.
5519 Details : By echoing "nid:$NID" string into
5520 /proc/fs/lustre/mds/.../evict_client client with nid that equals to
5521 $NID would be instantly evicted from this MDS and from all active
5522 OSTs connected to it.
5526 Description: Do not query file size twice, somewhat slowing stat(2) calls.
5527 Details : lookup_it_finish() used to query file size from OSTs that was not
5532 Description: service threads change working directory to that of init
5533 Details : Starting lustre service threads may pin the working directory
5534 of the parent thread, making that filesystem busy. Threads
5535 now change to the working directory of init to avoid this.
5539 Frequency : during shutdown only
5540 Description: shutdown with a failed MDS or OST can cause unmount to hang
5541 Details : Don't resend DISCONNECT messages in ptlrpc_disconnect_import()
5546 Frequency : 2.6 only
5547 Description: chmod/chown may include an extra supplementary group
5548 Details : ll{,u}_mdc_pack_op_data() does not properly initialize the
5549 supplementary group and if none is specified this is used.
5552 Bugzilla : 5479 (6816)
5554 Description: Racing open + rm can assert client in mdc_set_open_replay_data()
5555 Details : If lookup is in progress on a file that is unlinked we might try
5556 to revalidate the inode and fail in revalidate after lookup is
5557 complete and ll_file_open() enqueues the open again but
5558 it_open_error() was not checking DISP_OPEN_OPEN errors correctly.
5561 Frequency : always, if lconf --abort_recovery used
5563 Description: lconf --abort_recovery fails with 'Operation not supported'
5564 Details : lconf was attempting to abort recovery on the MDT device and not
5567 ------------------------------------------------------------------------------
5569 2005-08-08 Cluster File Systems, Inc. <info@clusterfs.com>
5574 Frequency : rare (only unsupported configurations with a node running as an
5576 Bugzilla : 6514, 5137
5577 Description: Mounting a Lustre file system on a node running as an OST could
5579 Details : OSTs now preallocates memory needed to write out data at
5580 startup, instead of when needed, to avoid having to
5581 allocate memory in possibly low memory situations.
5582 Specifically, if the file system is mounted on on OST,
5583 memory pressure could force it to try to write out data,
5584 which it needed to allocate memory to do. Due to the low
5585 memory, it would be unable to do so and the node would
5586 become unresponsive.
5588 Severity : enhancement
5590 Description: Addition of lconf --service command line option
5591 Details : lconf now accepts a '--service <arg>' option, which is
5592 shorthand for 'lconf --group <arg> --select <arg>=<hostname>'
5594 Severity : enhancement
5596 Description: Failover mode is now the default for OSTs.
5597 Details : By default, OSTs will now run in failover mode. To return to
5598 the old behaviour, add '--failout' to the lmc line for OSTs.
5600 Severity : enhancement
5602 Description: Health checks are now provided for MDS and OSTs
5603 Details : Additional detailed health check information on MSD and OSTs
5604 is now provided through the procfs health_check value.
5607 Frequency : occasional, depends on IO load
5609 Description: Disk fragmentation on the OSTs could eventually cause slowdowns
5610 after numerous create/delete cycles
5611 Details : The ext3 inode allocation policy would not allocate new inodes
5612 very well on the OSTs because there are no new directories
5613 being created. Instead we look for groups with free space if
5614 the parent directories are nearly full.
5619 Description: Network or server problems during mount may cause partially
5620 mounted clients instead of returning an error.
5621 Details : The config llog parsing code may overwrite the error return
5622 code during mount error handling, returning success instead
5628 Description: MDS can fail to allocate large reply buffers
5629 Details : After long uptimes the MDS can fail to allocate large reply
5630 buffers (e.g. zconf client mount config records) due to memory
5631 fragmentation or consumption by the buffer cache. Preallocate
5632 some large reply buffers so that these replies can be sent even
5633 under memory pressure.
5637 Frequency : rare (liblustre)
5638 Description: fsx running with liblustre complained that using truncate() to
5639 extend the file doesn't work. This patch corrects that issue.
5640 Details : This is the liblustre equivalent of the fix for bug 6196. Fixes
5641 ATTR_SIZE and lsm use in llu_setattr_raw.
5645 Frequency : rare, only 2.6 kernels
5646 Description: Unusual file access patterns on the MDS may result in inode
5647 data being lost in very rare circumstances.
5648 Details : Bad interaction between the ea-in-inode patch and the "no-read"
5649 code in the 2.6 kernel caused the inode and/or EA data not to
5650 be read from disk, causing single-file corruption.
5654 Frequency : rare, only 2.6 filesystems using extents
5655 Description: Heavy concurrent write and delete load may cause data corruption.
5656 Details : It was possible under high-load situations to have an extent
5657 metadata block in the block device cache from a just-unlinked
5658 file overwrite a newly-allocated data block. We now unmap any
5659 metadata buffers that alias just-allocated data blocks.
5663 Frequency : filesystems with default stripe_count larger than 77
5664 Description: lconf+mke2fs fail when formatting filesystem with > 77 stripes
5665 Details : lconf specifies an inode size of 4096 bytes when the default
5666 stripe_count is larger than 77. This conflicts with the default
5667 inode density of 1 per 4096 bytes. Allocate smaller inodes in
5668 this case to avoid pinning too much memory for large EAs.
5670 ------------------------------------------------------------------------------
5672 2005-07-07 Cluster File Systems, Inc. <info@clusterfs.com>
5677 Frequency : rare (extremely heavy IO load with hundreds of clients)
5679 Description: Client is evicted, gets IO error writing to file
5680 Details : lock ordering changes for bug 5492 reintroduced bug 3267 and
5681 caused clients to be evicted for AST timeouts. The fixes in
5682 bug 5192 mean we no longer need to have such short AST timeouts
5683 so ldlm_timeout has been increased.
5686 Frequency : occasional during --force or --failover shutdown under load
5687 Bugzilla : 5949, 4834
5688 Description: Server oops/LBUG if stopped with --force or --failover under load
5689 Details : a collection of import/export refcount and cleanup ordering
5690 issues fixed for safer force cleanup
5693 Frequency : only filesystems larger than 120 OSTs
5694 Bugzilla : 5990, 6223
5695 Description: lfs getstripe would oops on a very large filesystem
5696 Details : lov_getconfig used kfree on vmalloc'd memory
5699 Frequency : only filesystems exporting via NFS to Solaris 10 clients
5700 Bugzilla : 6242, 6243
5701 Description: reading from files that had been truncated to a non-zero size
5702 but never opened returned no data
5703 Details : ll_file_read() reads zeros from no-object files to EOF
5708 Description: A bug in MDS/OSS recovery could cause the OSS to fail an assertion
5709 Details : There's little harm in aborting MDS/OSS recovery and letting it
5710 try again, so I removed the LASSERT and return an error instead.
5712 Severity : enhancement
5714 Description: New debugging infrastructure for tracking down data corruption
5715 Details : The I/O checksum code was replaced to: (a) control it at runtime,
5716 (b) cover more of the client-side code path, and (c) try to narrow
5717 down where problems occurred
5721 Bugzilla : 3819, 4364, 4397, 6313
5722 Description: Racing close and eviction MDS could cause assertion in mds_close
5723 Details : It was possible to get multiple mfd references during close and
5724 client eviction, leading to one thread referencing a freed mfd.
5726 Severity: : enhancement
5727 Bugzilla : 3262, 6359
5728 Description: Attempts to reconnect to servers are now more aggressive.
5729 Details : This builds on the enhanced upcall-less recovery that was added
5730 in 1.4.2. When trying to reconnect to servers, clients will
5731 now try each server in the failover group every 10 seconds. By
5732 default, clients would previously try one server every 25 seconds.
5737 Description: After recovery, certain operations trigger a failed
5738 assertion on a client.
5739 Details : Failing over an mds, using lconf -d --failover, while a
5740 client was doing a readdir() call would cause the client to
5741 LBUG after recovery completed and the readdir() was resent.
5743 Severity : enhancement
5745 Description: Default groups are now added by lconf
5746 Details : You can now run lconf --group <servicename> without having to
5747 manually add groups with lmc.
5750 Frequency : occasional
5752 Description: Nodes with an elan id of 0 trigger a failed assertion
5755 Frequency : always when accessing e.g. tty/console device nodes
5757 Description: tty and some other devices nodes cannot be used on lustre
5758 Details : file's private_data field is used by device data and lustre
5759 values in there got lost. New field was added to struct file to
5760 store fs-specific private data.
5763 Frequency : when exporting Lustre via NFS
5765 Description: NFSD failed occasionally when looking up a path component
5766 Details : NFSD is looking up ".." which was broken in ext3 directories
5767 that had grown large enough to become hashed.
5770 Frequency : Clusters with multiple interfaces not on the same subnet
5772 Description: Nodes will repeatedly try to reconnect to an interface which it
5773 cannot reach and report an error to the log.
5774 Details : Extra peer list entries will be created by lconf with some peers
5775 unreachable. lconf now validates the peer before adding it.
5778 Frequency : Only if a default stripe is set on the filesystem root.
5780 Description: Setting a default stripe on the filesystem root prevented the
5781 filesystem from being remounted.
5782 Details : The client was sending extra request flags in the root getattr
5783 request and did not allocate a reply buffer for the dir EA.
5786 Frequency : occasional, higher if lots of files are accessed by one client
5787 Bugzilla : 6159, 6097
5788 Description: Client trips assertion regarding lsm mismatch/magic
5789 Details : While revalidating inodes the VFS looks up inodes with ifind()
5790 and in rare cases can find an inode that is being freed.
5791 The ll_test_inode() code will free the lsm during ifind()
5792 when it finds an existing inode and then the VFS later attaches
5793 this free lsm to a new inode.
5797 Bugzilla : 6422, 7030
5798 Description: MDS deadlock between mkdir and client eviction
5799 Details : Creating a new file via mkdir or mknod (starting a transaction
5800 and getting the ns lock) can deadlock with client eviction
5801 (gets ns lock and trying to finish a synchronous transaction).
5804 Frequency : occasional
5805 Description: While starting a server, the fsfilt_ext3 module could not be
5807 Details : CFS's improved ext3 filesystem is named ldiskfs for 2.6
5808 kernels. Previously, lconf would still use the ext3 name
5809 when trying to load modules. Now, it will correctly use
5810 ext3 on 2.4 and ldiskfs on 2.6.
5812 Severity : enhancement
5813 Description: The default stripe count has been changed to 1
5814 Details : The interpretation of the default stripe count (0, to lfs
5815 or lmc) has been changed to mean striping across a single
5816 OST, rather than all available. For general usage we have
5817 found a stripe count of 1 or 2 works best.
5819 Severity : enhancement
5820 Description: Add support for compiling against Cray portals.
5821 Details : Conditional compiling for some areas that are different
5825 Frequency : occasional
5826 Bugzilla : 6409, 6834
5827 Description: Creating files with an explicit stripe count may lead to
5828 a failed assertion on the MDS
5829 Details : If some OSTs are full or unavailable, creating files may
5830 trigger a failed assertion on the MDS. Now, Lustre will
5831 try to use other servers or return an error to the
5835 Frequency : occasional
5837 Description: Multiple concurrent overlapping read+write on multiple SMP nodes
5838 caused lock timeout during readahead (since 1.4.2).
5839 Details : Processes doing readahead might match a lock that hasn't been
5840 granted yet if there are overlapping and conflicting lock
5841 requests. The readahead process waits on ungranted lock
5842 (original lock is CBPENDING), while OST waits for that process
5843 to cancel CBPENDING read lock and eventually evicts client.
5845 Severity : enhancement
5847 Description: Initial enabling of flock support for clients
5848 Details : Implements fcntl advisory locking and file status functions.
5849 This feature is provided as an optional mount flag (default
5850 off), and is NOT CURRENTLY SUPPORTED. Not all types of record
5851 locking are implemented yet, and those that are are not guaranteed
5852 to be completely correct in production environments.
5853 mount -t lustre -o [flock|noflock] ...
5856 Frequency : occasional
5858 Description: OSTs running 2.4 kernels but with extents enabled might trip an
5859 assertion in the ext3 JBD (journaling) layer.
5860 Details : The b_committed_data struct is protected by the big kernel lock
5861 in 2.4 kernels, serializing journal_commit_transaction() and
5862 ext3_get_block_handle->ext3_new_block->find_next_usable_block()
5863 access to this struct. In 2.6 kernels there is finer grained
5864 locking to improve SMP performance of the JBD layer.
5868 Description: Changes the "SCSI I/O Stats" kernel patch to default to "enabled"
5870 -----------------------------------------------------------------------------
5872 2005-05-05 Cluster File Systems, Inc. <info@clusterfs.com>
5874 NOTE: Lustre 1.4.2 uses an incompatible network protocol than previous
5875 versions of Lustre. Please update all servers and clients to
5876 version 1.4.2 or later at the same time. You must also run
5877 "lconf --write-conf {config}.xml" on the MDS while it is stopped
5878 to update the configuration logs.
5880 - fix for HPUX NFS client breakage when NFS exporting Lustre (5781)
5881 - mdc_enqueue does not need max_mds_easize request buffer on send (5707)
5882 - swab llog records of type '0' so we get proper header size/idx (5861)
5883 - send llog cancel req to DLM cancel portal instead of cb portal (5515)
5884 - fix rename of one directory over another leaking an inode (5953)
5885 - avoid SetPageDirty on 2.6 (5981)
5886 - don't re-add just-being-destroyed locks to the waiting list (5653)
5887 - when creating new directories, inherit the parent's custom
5888 striping settings if present parent (3048)
5889 - flush buffers from cache before direct IO in 2.6 obdfilter (4982)
5890 - don't hold i_size_sem in ll_nopage() and ll_ap_refresh_count (6077)
5891 - don't hold client locks on temporary worklist from l_lru (5666)
5892 - handle IO errors in 2.6 obdfilter bio completion routine (6046)
5893 - automatically evict dead clients (5921)
5894 - Update file size properly in create+truncate+fstat case (6196)
5895 - Do not unhash mountpoint dentries, do not allow removal of
5897 - Avoid lock ordering deadlock issue with write/truncate (6203,5654)
5898 - reserve enough journal credits in fsfilt_start_log for setattr (4554)
5899 - ldlm_enqueue freed-export error path would always LBUG (6149,6184)
5900 - don't reference lr_lvb_data until after we hold lr_lvb_sem (6170)
5901 - don't overwrite last_rcvd if there is a *_client_add() error (6086)
5902 - Correctly handle reads of files with no objects (6243)
5903 - lctl recover will also mark a device active if deactivate used (5933)
5905 - by default create 1 inode per 4kB space on MDS, per 16kB on OSTs
5906 - allow --write-conf on an MDS with different nettype than client (5619)
5907 - don't write config llogs to MDS for mounts not from that MDS (5617)
5908 - lconf should create multiple TCP connections from a client (5201)
5909 - init scripts are now turned off by default; run chkconfig --on
5910 lustre and chkconfig --on lustrefs to use them
5911 - upcalls are no longer needed for clients to recover to failover
5913 - add --abort-recovery option to lconf to abort recovery on device
5915 - add support for an arbitrary number of OSTs (3026)
5916 - Quota support protocol changes.
5917 - forward compatibility changes to wire structs (6007)
5918 - rmmod NALs that might be loaded because of /etc/modules.conf (6133)
5919 - support for mountfsoptions and clientoptions to the Lustre LDAP (5873)
5920 - improved "lustre status" script
5921 - initialize blocksize for non-regular files (6062)
5922 - added --disable-server and --disable-client configure options (5782)
5923 - introduce a lookup cache for lconf to avoid repeated DB scans (6204)
5924 - Vanilla 2.4.29 support
5925 - increase maximum number of obd devices to 520 (6242)
5926 - remove the tcp-zero-copy patch from the suse-2.4 series (5902)
5927 - Quadrics Elan drivers are now included for the RHEL 3 2.4.21 and
5928 SLES 9 2.6.5 kernels
5929 - limit stripes per file to 160 (the maximum EA size) (6093)
5931 2005-03-22 Cluster File Systems, Inc. <info@clusterfs.com>
5934 - don't LASSERT in ll_release on NULL lld with NFS export (4655, 5760)
5935 - hold NS lock when calling handle_ast_error->del_waiting_lock (5746)
5936 - fix setattr mtime regression from lovcleanup merge (4829, 5669)
5937 - workaround for 2.6 crash in ll_unhash_aliases (5687, 5210)
5938 - small ext3 extents cleanups and fixes (5733)
5939 - improved mballoc code, several small races and bugs fixed (5733, 5638)
5940 - kernel version 43 - fix remove_suid bugs in both 2.4 and 2.6 (5695)
5941 - avoid needless client->OST connect, fix handle mismatch (5317)
5942 - fix DLM error path that led to out-of-sync client, long delays (5779)
5943 - support common vfs-enforced mount options (nodev,nosuid,noexec) (5637)
5944 - fix several locking issues related to i_size (5492,5624,5654,5672)
5945 - don't move pending lock onto export if it is already evicted (5683)
5946 - fix kernel oops when creating .foo in unlinked directory (5548)
5947 - fix deadlock in obdfilter statistics vs. object create (5811)
5948 - use time_{before,after} to avoid timer jiffies wrap (5882)
5949 - shutdown --force/--failover stability (3607,3651,4797,5203,4834)
5950 - Do not leak request if server was not able to process it (5154)
5951 - If mds_open unable to find parent dir, make that negative lookup(5154)
5952 - don't create new directories with extent-mapping (5909, 5936)
5954 - fix lustre/lustrefs init scripts for SuSE (patch from Scali, 5702)
5955 - don't hold the pinger_sem in ptlrpc_pinger_sending_on_import
5956 - change obd_increase_kms to obd_adjust_kms (up or down) (5654)
5957 - lconf, lmc search both /usr/lib and /usr/lib64 for Python libs (5800)
5958 - support for RHEL4 kernel on i686 (5773)
5959 - provide error messages when incompatible logs are encountered (5898)
5961 2005-02-18 Cluster File Systems, Inc. <info@clusterfs.com>
5962 * version 1.4.0.10 (1.4.1 release candidate 1)
5964 - don't keep a lock reference when lock is not granted (4238)
5965 - unsafe list practices (rarely) led to infinite eviction loop (4908)
5966 - add per-fs limit of Lustre pages in page cache, avoid OOM (4699)
5967 - drop import inflight refcount on signal_completed_replay error (5255)
5968 - unlock page after async write error during send (3677)
5969 - handle missing objects in filter_preprw_read properly (5265)
5970 - no transno return for symlink open, don't save no-trasno open (3440)
5971 - don't try to complete elan receive that already failed (4012)
5972 - free RPC server reply state on error (5406)
5973 - clean up thread from ptlrpc_start_thread() on error (5160)
5974 - readahead could read extra page into cache that wasn't ejected (5388)
5975 - prevent races in class_attach/setup/cleanup/detach (5260)
5976 - don't dereference de->d_inode after l_dput of de (5458)
5977 - use "int" for stripe value returned from lock_to_stripe (5544)
5978 - mballoc allocation and error-checking fixes in 2.6 (5504)
5979 - block device patches to fix I/O request sizes in 2.6 (5482)
5980 - look up hostnames for IB nals (5602)
5981 - 2.6 changed lock ordering of 2 semaphores, caused deadlock (5654)
5982 - don't start multiple acceptors for the same port (5277)
5983 - fix incorrect LASSERT in mds_getattr_name (5635)
5984 - export a proc file for general "ping" checking (5628)
5985 - fix "lfs check" to not block when the MDS is down (5628)
5987 - service request history (4965)
5988 - put {ll,lov,osc}_async_page structs in a single slab (4699)
5989 - create an "evict_client" /proc entry on OSTs, like the MDS has
5990 - fix mount usage message, return errors per mount(8) (5168)
5991 - change grep [] to grep "[]" in tests so they work in more UMLs
5992 - fix ppc64/x86_64 spec to use %{_libdir} instead of /usr/lib (5389)
5993 - remove ancient LOV_MAGIC_V0 EA support (5047)
5994 - add "disk I/Os in flight" and "I/O req time" stats in obdfilter
5995 - align r/w RPCs to PTLRPC_MAX_BRW_SIZE boundary for performance (3451)
5996 - allow readahead allocations to fail when low on memory (5383)
5997 - mmap locking landed again, after considerable improvement (2828)
5998 - add get_hostaddr() to lustreDB.py for LDAP support (5459)
6000 2004-11-23 Cluster File Systems, Inc. <info@clusterfs.com>
6003 - send OST transaction number in read/write reply to free req (4966)
6004 - don't ASSERT in ptl_send_rpc() if we run out of memory (5119)
6005 - lock /proc/sys/portals/routes internal state, avoiding oops (4827)
6006 - the watchdog thread now runs as interruptible (5246)
6007 - flock/lockf fixes (but it's still disabled, pending 5135)
6008 - don't use EXT3 constants in llite code (5094)
6009 - memory shortage at startup could cause assertion (5176)
6011 - reorganization of lov code
6012 - single portals codebase
6014 - add extents/mballoc support (5025)
6015 - direct I/O reads in the obdfilter (4048)
6016 - kernel patches from LNXI for 2.6 (bluesmoke, perfctr, mtd, kexec)
6018 tbd Cluster File Systems, Inc. <info@clusterfs.com>
6021 - send OST transaction number in read/write reply to free req (4966)
6022 - don't ASSERT in ptl_send_rpc() if we run out of memory (5119)
6023 - lock /proc/sys/portals/routes internal state, avoiding oops (4827)
6024 - the watchdog thread now runs as interruptible (5246)
6025 - handle missing objects in filter_preprw_read properly (5265)
6026 - unsafe list practices (rarely) led to infinite eviction loop (4908)
6027 - drop import inflight refcount on signal_completed_replay error (5255)
6028 - unlock page after async write error during send (3677)
6029 - return original error code on reconstructed replies (3761)
6030 - no transno return for symlink open, don't save no-trasno open (3440)
6032 - add pid to ldlm debugging output (4922)
6033 - bump the watchdog timeouts -- we can't handle 30sec yet
6034 - extra debugging for orphan dentry/inode bug (5259)
6036 2004-11-16 Cluster File Systems, Inc. <info@clusterfs.com>
6039 - fix TCP_NODELAY bug, which caused extreme perf regression (5134)
6040 - allocate qswnal tx descriptors singly to avoid fragmentation (4504)
6041 - don't LBUG on obdo_alloc() failure, use OBD_SLAB_ALLOC() (4800)
6042 - fix NULL dereference in /proc/sys/portals/routes (4827)
6043 - allow failed mdc_close() operations to be interrupted (4561)
6044 - stop precreate on OST before MDS would time out on it (4778)
6045 - don't send partial-page writes before EOF from client (4410)
6046 - discard client grant for sub-page writes on large-page clients (4520)
6047 - don't free dentries not owned by NFS code, check generation (4806)
6048 - fix lsm leak if mds_create_objects() fails (4801)
6049 - limit debug_daemon file size, always print CERROR messages (4789)
6050 - use transno after validating reply (3892)
6051 - process timed out requests if import state changes (3754)
6052 - update mtime on OST during writes, return in glimpse (4829)
6053 - add mkfsoptions to LDAP (4679)
6054 - use ->max_readahead method instead of zapping global ra (5039)
6055 - don't interrupt __l_wait_event() during strace
6057 - add software watchdogs to catch hung threads quickly (4941)
6058 - make lustrefs init script start after nfs is mounted
6059 - fix CWARN/ERROR duplication (4930)
6060 - return async write errors to application if possible (2248)
6061 - add /proc/sys/portal/memused (bytes allocated by PORTALS_ALLOC)
6062 - print NAL number in %x format (4645)
6063 - update barely-supported suse-2.4.21-171 series (4842)
6064 - support for sles 9 %post scripts
6065 - support for building 2.6 kernel-source packages
6066 - support for sles km_* packages
6068 2004-10-07 Cluster File Systems, Inc. <info@clusterfs.com>
6071 - ignore -ENOENT errors in osc_destroy (3639)
6072 - notify osc create thread that OSC is being cleaned up (4600)
6073 - add nettype argument for llmount in #5d in conf-sanity.sh (3936)
6074 - reconstruct ost_handle() like mds_handle() (4657)
6075 - create a new thread to do import eviction to avoid deadlock (3969)
6076 - let lconf resolve symlinked-to devices (4629)
6077 - don't unlink "objects" from directory with default EA (4554)
6078 - hold socknal file ref over connect in case target is down (4394)
6079 - allow more than 32000 subdirectories in a single directory (3244)
6080 - fix blocks count for O_DIRECT writes (3751)
6081 - OST returns ENOSPC from object create when no space left (4539)
6082 - don't send truncate RPC if file size isn't changing (4410)
6083 - limit OSC precreate to 1/2 of value OST considers bogus (4778)
6084 - bind to privileged port in socknal and tcpnal (3689)
6086 - rate limit CERROR/CWARN console message to avoid overload (4519)
6087 - GETFILEINFO dir ioctl returns LOV EA + MDS stat in 1 call (3327)
6088 - basic mmap support (3918)
6089 - kernel patch series update from b1_4 (4711)
6091 2004-09-16 Cluster File Systems, Inc. <info@clusterfs.com>
6094 - avoid crash during MDS cleanup with OST shut down (2775)
6095 - fix loi_list_lock/oig_lock inversion on interrupted IO (4136)
6096 - don't use bad inodes on the MDS (3744)
6097 - dynamic object preallocation to improve recovery speed (4236)
6098 - don't hold spinlock over lock dumping or change debug flags (4401)
6099 - don't zero obd_dev when it is force cleaned (3651)
6100 - print grants to console if they go negative (4431)
6101 - "lctl deactivate" will stop automatic recovery attempts (3406)
6102 - look for existing locks in ldlm_handle_enqueue() (3764)
6103 - don't resolve lock handle twice in recovery avoiding race (4401)
6104 - revalidate should check working dir is a directory (4134)
6106 - don't always mark "slow" obdfilter messages as errors (4418)
6108 2004-08-24 Cluster File Systems, Inc. <info@clusterfs.com>
6111 - don't close LustreDB during write_conf until it is done (3860)
6112 - fix typo in lconf for_each_profile (3821)
6113 - allow dumping logs from multiple threads at one time (3820)
6114 - don't allow multiple threads in OSC recovery (3812)
6115 - fix debug_size parameters (3864)
6116 - fix mds_postrecov to initialize import for llog ctxt (3121)
6117 - replace config semaphore with spinlock (3306)
6118 - be sure to send a reply for a CANCEL rpc with bad export (3863)
6119 - don't allow enqueue to complete on a destroyed export (3822)
6120 - down write_lock before checking llog header bitmap (3825)
6121 - recover from lock replay timeout (3764)
6122 - up llog sem before sending rpc (3652)
6123 - reduce ns lock hold times when setting kms (3267)
6124 - change a dlm LBUG to LASSERTF, to maybe learn something (4228)
6125 - fix NULL deref and obd_dev leak on setup error (3312)
6126 - replace some LBUG about llog ops with error handling (3841)
6127 - don't match INVALID dentries from d_lookup and spin (3784)
6128 - hold dcache_lock while marking dentries INVALID and hashing (4255)
6129 - fix invalid assertion in ptlrpc_set_wait (3880)
6131 - add libwrap support for the TCP acceptor (3996)
6132 - add /proc/sys/portals/routes for non-root route listing (3994)
6133 - allow setting MDS UUID in .xml (2580)
6134 - print the stack of a process that LBUGs (4228)
6136 2004-07-14 Cluster File Systems, Inc. <info@clusterfs.com>
6139 - don't cleanup request in ll_file_open() on failed MDS open (3430)
6140 - make sure to unset replay flag from failed open requests (3440)
6141 - if default stripe count is 0, use OST count for inode size (3636)
6142 - update parent mtime/ctime on client for create/unlink (2611)
6143 - drop dentry ref in ext3_add_link from open_connect_dentry (3266)
6144 - free recovery state on server during a forced cleanup (3571)
6145 - unregister_reply for resent reqs (3063)
6146 - loop back devices mounting and status check on 2.6 (3563)
6147 - fix resource-creation race that can provoke i_size == 0 (3513)
6148 - don't try to use bad inodes returned from MDS/OST fs lookup (3688)
6149 - more debugging for page-accounting assertion (3746)
6150 - return -ENOENT instead of asserting if ost getattr+unlink race (3558)
6151 - avoid deadlock after precreation failure (3758)
6152 - fix race and lock order deadlock in orphan handling (3450, 3750)
6153 - add validity checks when grabbing inodes from l_ast_data (3599)
6155 - add /proc/.../recovery_status to obdfilter (3428)
6156 - lightweight CDEBUG infrastructure, debug daemon (3668)
6157 - change default OSC RPC parameters to be better on small clusters
6158 - turn off OST read cache for files smaller than 32MB
6159 - install man pages and include them in rpms (3100)
6160 - add new init script for (un)mounting lustre filesystems (2593)
6161 - run chkconfig in %post for init scripts (3701)
6162 - drop scimac NAL (unmaintained)
6164 2004-06-17 Cluster File Systems, Inc. <info@clusterfs.com>
6167 - clean kiobufs before and after use (3485)
6168 - strip trailing '/'s before comparing paths with /proc/mounts (3486)
6169 - remove assertions to work around "in-flight rpcs" recovery bug (3063)
6170 - change init script to fail more clearly if not run as root (1528)
6171 - allow clients to reconnect during replay (1742)
6172 - fix ns_lock/i_sem lock ordering deadlock for kms update (3477)
6173 - don't do DNS lookups on NIDs too small for IP addresses (3442)
6174 - re-awaken ptlrpcd if new requests arrive during check_set (3554)
6175 - fix cond_resched (3554)
6176 - only evict unfinished clients after recovery (3515)
6177 - allow bulk resend, prevent data loss (3570)
6178 - dynamic ptlrpc request buffer allocation (2102)
6179 - don't allow unlinking open directory if it isn't empty (2904)
6180 - set MDS/OST threads to umask 0 to not clobber client modes (3359)
6181 - remove extraneous obd dereference causing LASSERT failure (3334)
6182 - don't use get_cycles() when creating temp. files on the mds (3156)
6183 - hold i_sem when setting i_size in ll_extent_lock() (3564)
6184 - handle EEXIST for set-stripe, set proper directory name (3336)
6186 - servers can dump a log evicting a client - lustre.dump_on_timeout=1
6187 - fix ksocknal_fmb_callback() error messages (2918)
6189 2004-05-27 Cluster File Systems, Inc. <info@clusterfs.com>
6192 - don't copy lvb into (possibly NULL) reply on error (2983)
6193 - don't deref dentry after dput, don't free lvb on error (2922)
6194 - use the kms to determine writeback rpc length (2947)
6195 - increment oti_logcookies when osc is inactive (2948)
6196 - update client's i_blocks count via lvb messages (2543)
6197 - handle intent open/close of special files properly (1557)
6198 - mount MDS with errors=remount-ro, like obdfilter (2009)
6199 - initialize lock handle to avoid ASSERT on error cleanup (3057)
6200 - don't use cancelling-locks' kms values (2947)
6201 - use highest lock extent for kms, not last one (2925)
6202 - don't dereference ERR_PTR() dentry in error handling path (3107)
6203 - fix thread race in portals_debug_dumplog() (3122)
6204 - create lprocfs device entries at setup instead of at attach (1519)
6205 - common AST error handler, don't evict client on completion race (3145)
6206 - zero nameidata in detach_mnt in 2.6 (3118)
6207 - verify d_inode after revalidate_special is valid in 2.6 (3116)
6208 - use lustre_put_super() to handle zconf unmounts in 2.6 (3064)
6209 - initialize RPC timeout timer earlier for 2.6 (3219)
6210 - don't dereference NULL reply buffer if mdc_close was never sent (2410)
6211 - print nal/nid for unknown nid (3258)
6212 - additional checks for oscc recovery before doing precreate (3284)
6213 - fix ll_extent_lock() error return code for 64-bit systems (3043)
6214 - don't crash in mdc_close for bad permissions on open (3285)
6215 - zero i_rdev for non-device files (3147)
6216 - clear page->private before handing to FS, better assertion (3119)
6217 - tune the read pipeline (3236)
6218 - fix incorrect decref of invalidated dentry (2350)
6219 - provide read-ahead stats and refine rpc in flight stats (3328)
6220 - don't hold journal transaction open across create RPC (3313)
6221 - update atime on MDS at close time (3265)
6222 - close LDAP connection when recovering to avoid server load (3315)
6223 - update iopen-2.6 patch with fixes from 2399,2517,2904 (3301)
6224 - don't leak open file on MDS after open resend (3325)
6225 - serialize filter_precreate and filter_destroy_precreated (3329)
6226 - loop device shouldn't call sync_dev() for nul device (3092)
6227 - clear page cache after eviction (2766)
6228 - resynchronize MDS->OST in background (2824)
6229 - refuse to mount the same filesystem twice on same mountpoint (3394)
6230 - allow llmount to create routes for mounting behind routers (3320)
6231 - push lock cancellation to blocking thread for glimpse ASTs (3409)
6232 - don't call osc_set_data_with_check() for TEST_LOCK matches (3159)
6233 - fix rare problem with rename on htree directories (3417)
6235 - allow default OST striping configuration per directory (1414)
6236 - fix compilation for qswnal for 2.6 kernels (3125)
6237 - increase maximum number of MDS request buffers for large systems
6238 - change liblustreapi to be useful for external progs like lfsck (3098)
6239 - increase local configuration timeout for slow disks (3353)
6240 - allow configuring ldlm AST timeout - lustre.ldlm_timeout=<seconds>
6242 2004-03-22 Cluster File Systems, Inc. <info@clusterfs.com>
6245 - fixes for glimpse AST timeouts / incorrectly 0-sized files (2818)
6246 - don't overwrite extent policy data in reply if lock was blocked (2901)
6247 - drop filter export grants atomically with removal from device (2663)
6248 - del obd_self_export from work_list in class_disconnect_exports (2908)
6249 - don't LBUG if MDS recovery times out during orphan cleanup (2530)
6250 - swab reply message in mdc_close, other PPC fixes (2464)
6251 - fix destroying of named logs (2325)
6252 - overwrite old logs when running lconf --write_conf (2264)
6253 - bump LLOG_CHUNKSIZE to 8k to allow for larger clusters (2306)
6254 - fix race in target_handle_connect (2898)
6255 - mds_reint_create() should take same inode create lock (2926)
6256 - correct journal credits calculated for CANCEL_UNLINK_LOG (2931)
6257 - don't close files for self_export to avoid uninitialized obd (2936)
6258 - allow MDS with the same name as client node (2939)
6259 - hold dentry reference for closed log files for unlink (2325)
6260 - reserve space for all logs during transactions (2059)
6261 - don't evict page beyond end of stripe extent (2925)
6262 - don't oops on a deleted current working directory (2399)
6263 - handle hard links to targets without a parent properly (2517)
6264 - don't dereference NULL lock when racing during eviction (2867)
6265 - don't grow lock extents when lots of conflicting locks (2919)
6267 2004-03-04 Cluster File Systems, Inc. <info@clusterfs.com>
6270 - account for cache space usage on clients to avoid data loss (974)
6271 - lfsck support in lustre kernel code (2349)
6272 - reduce journal credits needed for BRW writes (2370)
6273 - orphan handling to avoid losing space on client/server crashes
6274 - ptlrpcd can be blocked, stopping ALL progress (2477)
6275 - use lock value blocks to assist in proper KMS, faster stat (1021)
6276 - takes i_sem instead of DLM locks internally on obdfilter (2720)
6277 - recovery for initial connections (2355)
6278 - fixes for mds_cleanup_orphans (1934)
6279 - abort_recovery crashes MDS in b_eq (mds_unlink_orphan) (2584)
6280 - block all file creations until orphan recovery completes (1901)
6281 - client remove rq_connection from request struct (2423)
6282 - conf-sanity test_5, proper cleanup in umount log not availale (2640)
6283 - recovery timer race (2670)
6284 - mdc_close recovey bug (2532)
6285 - ptlrpc cleanup bug (2710)
6286 - mds timeout on local locks (2588)
6287 - namespace lock held during RPCs (2431)
6288 - handle interrupted sync write properly (2503)
6289 - don't try to handle a message that hasn't been replied to (2699)
6290 - client assert failure during cleanup after abort recovery (2701)
6291 - leak mdc device after failed mount (2712)
6292 - ptlrpc_check_set allows timedout requests to complete (2714)
6293 - wait for inflight reqs when ptlrpcd finishes (2710)
6294 - make sure unregistered services are removed from the srv_list
6295 - reset bulk XID's when resending them (caught by 1138 test)
6296 - unregister_bulk after timeout
6297 - fix lconf error (2694)
6298 - handle write after unfinished setstripe, stripe-only getstripe (2388)
6299 - readahead locks pages, leaves pending causing memory pressure (2673)
6300 - increase OST request buffers to 4096 on large machines (2729)
6301 - fix up permission of existing directories in simple_mkdir (2661)
6302 - init deleted item, add assertions ptlrpc_abort_inflight() (2725)
6303 - don't assign transno to errored transactions (2742)
6304 - don't delete objects on OST if given a bogus objid from MDS (2751)
6305 - handle large client PAGE_SIZE readdir on small PAGE_SIZE MDS (2777)
6306 - if rq_no_resend, then timeout request after recovery (2432)
6307 - fix MDS llog_logid record size, 64-bit array alignment (2733)
6308 - don't call usermode_helper from ptlrpcd, DEFAULT upcall (2773)
6309 - put magic in mount.lustre data, check for bad/NULL mount data (2529)
6310 - MDS recovery shouldn't delete objects that it has given out (2730)
6311 - if enqueue arrives after completion, don't clobber LVB (2819)
6312 - don't unlock pages twice when trigger_group_io returns error (2814)
6313 - don't deref NULL rq_repmsg if ldlm_handle_enqueue failed (2822)
6314 - don't write pages to disk if there was an error (1450)
6315 - don't ping imports that have recovery disabled (2676)
6316 - take buffered bytes into account when balancing socknal conn (2817)
6317 - hold a DLM lock over readdir always, use truncate_inode_pages (2706)
6318 - reconnect unlink llog connection after MDS reconnects to OST (2816)
6319 - remove little-endian swabbing of llog records (1987)
6320 - set/limit i_blksize to LL_MAX_BLKSIZE on client (2884)
6321 - retry reposting request buffers if they fail (1191)
6322 - grow extent at grant time to avoid granting a revoked lock (2809)
6323 - lock revoke doesn't evict page if covered by a second lock (2765)
6324 - disable VM readahead to avoid reading outside lock extents (2805)
6326 - return LL_SUPER_MAGIC from statfs for the filesystem type (1972)
6327 - updated kernel patches for hp-2.4.20 kernel (2681)
6329 2004-02-07 Cluster File Systems, Inc. <info@clusterfs.com>
6332 - fix truncated write corruption (2366)
6333 - fix for failed assertion in iopen_connect_dentry (1792,2517)
6335 - don't flag the ptlrpcd thread with PF_MEMALLOC (2636)
6336 - ensure len(uuid) < 37 in lmc (1171)
6337 - fix ia64 OOPS in llog_test (2255)
6338 - zero end of page at obdfilter for partial page writes (2648)
6339 - don't leave stale dentries around after renames (bug 2428)
6340 - fix timeouts when evicting a client with a single lock held (2642)
6341 - set deadline for the initial HELLO message to drain (2634)
6342 - print out dotted-quad IP addresses in the socknal (2302)
6344 - additional debugging for MDS client eviction problem (2443)
6345 - fix mkfsoptions support for osts (2603, 2604)
6347 2004-01-27 Cluster File Systems, Inc. <info@clusterfs.com>
6350 - add series for the vanilla 2.6.0 kernel
6351 - add series for the vanilla 2.4.24 kernel
6352 - add series for a cray x86/64 UL kernel drop
6353 - fix xattr patches for the vanilla 2.4.19 series
6355 - generate true UUIDs in lmc (1171)
6356 - have portals stack dumping break in UML (2466)
6357 - avoid bad dchild deref; avoid inum lock w/o creation (2362)
6358 - allocate with _NOFS in ldlm to avoid deadlock (1933)
6359 - wake callback waiting threads on client eviction (2460)
6360 - Add --ptldebug and --subsystem to lmc (1719)
6361 - update assertion to allow safe interrupt allocation
6362 - set rq_no_resend for cancel requests (2432)
6363 - recalculate ptlrpcd timeout after resend (2494)
6364 - call vfs_rmdir when removing pending directories (2368)
6365 - fix renaming a file to itself (2429)
6366 - lmc creates a default one-stripe lov (2454)
6367 - expand procfs space to handle large clusters (2326)
6368 - increase UML stack to avoid overflow
6369 - update lconf's list of debug and subsystem masks
6370 - fix lfs find --obd (2510)
6371 - /proc tunable for disabling filter read caching (2591)
6372 - stop rpm packages from altering slapd.conf (2301)
6373 - disable nagle in the socknal under 0conf (2578)
6374 - choose mds inode size based on stripe count (2572)
6375 - fix kernel-source rpm problems (2516)
6377 - add --disable-doc to avoid pdf generation (2421)
6378 - update documentation, tests, type-os, comments
6379 - avoid format warnings on ia64
6380 - remove the TOE NAL
6381 - tiny code cleanups by removing unused fields
6383 2004-01-07 Cluster File Systems, Inc. <info@clusterfs.com>
6386 - fix obvious semaphore misuse in as-yet-unused setattr path (2348)
6387 - remove the most blatant lies from BUILDING file (2371)
6388 - change default debug level to reasonable production setting
6389 - reduce client side cache size to reduce cache flush time
6390 - reduce max RPCs in flight to avoid unnecessary file fragmentation
6391 - make TCP zerocopy and pinger support enabled by default (2476)
6392 - sync writes completed after process exits caused crashes (2319)
6393 - maintain correct mount count on the MDS (2356)
6394 - backout 1557, because 2316 wasn't really fixed
6395 - better file I/O statistics gathering in /proc
6396 - don't take unnecessary, deadlock-inducing bug in readpage (2383)
6397 - another kernel patch to fix zero-copy TCP function export
6398 - don't take duplicate lock when processing re-sent getattr (2420)
6399 - lctl uses obd_self_export instead of creating new conn (2353)
6400 - MDS/OST recovery case which requires object creation asserted (2425)
6401 - move lfs from /usr/sbin to /usr/bin in packages
6402 - fix race between mds_client_add and mds_client_free (2417)
6403 - use kmalloc instead of slabs in portals (2430)
6404 - don't create duplicate records when a failover MDS is present (2442)
6405 - remove unnecessary mount age check (2332)
6406 - don't remove directory inodes from locks prematurely (2451)
6407 - don't break if MDS service name is the same as hostname (2103)
6408 - fix races in client write RPC generation when cache full (2482)
6410 2003-12-13 Cluster File Systems, Inc. <info@clusterfs.com>
6413 - remove now-unused request->rq_obd (278)
6414 - if an allocation fails, print out how much memory we've used (1933)
6415 - use PORTAL_SLAB_ALLOC for structures, to get GFP_MEMALLOC (1933)
6416 - add the "configurable stack size" patch to most series files (1256)
6417 - ability to write large log records, for 100+ OST configs (2306)
6418 - fix NULL deref when filter_prep fails (2314)
6419 - fix operator precedence error in filter_sync
6420 - dynamic allocation of socknal TX descriptors (2315)
6421 - fix a missed case in the GFP_MEMALLOC patch, can cause deadlock (2310)
6422 - fix gcc 2.96 compilation problem in xattr kernel patch (2294)
6423 - ensure that CWARN messages in Portals always get to the syslog
6424 - __init/__exit are not for prototype decls (ldlm_init/exit)
6425 - x86-64 compile warning fixes
6426 - fix gateway LMC keyword conflict (2318)
6427 - fix MDS lock inversions in getattr/reint paths (1844)
6428 - fix a rare lock re-ordering bug, which caused deadlock (2322)
6429 - fix i_sem/journal inversion in fsfilt_ext3_write_record (2306)
6430 - DLM race condition prevented some lock evictions (2328)
6431 - ENOMEM detection and retry on socknal sends (2230)
6432 - use GFP_NOFS throughout Lustre, to combat ENOMEM (2230)
6433 - move osc_rpcd into ptlrpc, for use in MDC and others (2329)
6434 - protect MDS inode fsdata with stronger locking; fixes assertion (2313)
6435 - better error messages when a client is rejected during recovery (1505)
6436 - avoid cancelling locks which were never granted, after failure (2330)
6437 - fix i_sem/journal inversion in mds_client_add (2333)
6438 - fix truncate/getattr lock cycle deadlock (2334)
6439 - use rpcd to send close; allows resend after timeout, avoid leak (1897)
6440 - fix two rare exit paths which could leak an l_lock() ref (2321)
6441 - fencepost error in MDS/OST orphan recovery (2226)
6442 - make log record alignment 8 bytes (1988)
6443 - lstripe now fails when requested offset > ost_count (2237)
6444 - ensure that all kernel series have a complete list.h (1607)
6445 - fix crashes in special-file operations (2316)
6446 - lctl create/brw OID mismatch, caused by obsolete filter loop (2339)
6448 - allow configurable automake binary, for testing new versions
6449 - small update to the lfs documentation
6451 2003-12-03 Cluster File Systems, Inc. <info@clusterfs.com>
6453 * fix negative export reference count in fsfilt_sync (2312)
6455 2003-12-01 Cluster File Systems, Inc. <info@clusterfs.com>
6456 * release candidate 0.9.1
6458 - orphans are moved into the PENDING directory for possible recovery
6459 - replayed opens now open by fid for orphan/rename safety (1042)
6460 - last close of an orphan inode generates a transno (683)
6461 - chdir() and mount() now pin the directory entry (1020)
6462 - avoid CERROR in normal ll_setattr_raw() error case (1500)
6463 - discard very old requests without processing them (1502)
6464 - remove some common, well-understood CERRORs (1505)
6465 - require O_DIRECT I/O to be page-sized to workaround IA64 crash (1609)
6466 - clear "grant" flags in OST replies until OST grant code lands (1644)
6467 - fix read performance by not clobbering i_blksize on client (1598)
6468 - fix __ldlm_handle2lock oops by not dereferencing lock after PUT (1625)
6469 - make LRU size a /proc tunable, clears locks when reduced (707)
6470 - fix some lprocfs rot that prevented ptlbd from loading (1732)
6471 - server locks take references on exports now (1558)
6472 - build fixes for 2.4.20-rh trees (1663)
6473 - return an error from lov_create if all OSCs are inactive (1751)
6474 - fix import levels when a reconnect happens without a timeout (1597)
6475 - exit early from mds_open if we get a lookup error (1749)
6476 - partial page read at EOF wouldn't wait for disk before sending (1642)
6477 - avoid NULL deref in obdfilter when reading page past EOF (1592)
6478 - avoid LASSERT in ll_intent_lock if server failed very early (1090)
6479 - fix LBUG in ll_it_open_error with rc = -2 (1861)
6480 - write/truncate lock inversion (1639)
6481 - Don't auto-load obdclass, portals modules during cleanup (1495)
6482 - fix timestamps from jumping to "now" (1763)
6483 - extra journal assertions (1648)
6484 - add an extra multiunlink test (1771)
6485 - fix read_record/write_record API (1776)
6486 - fix leak of offset_extent, possible incorrect i_size later (1772)
6487 - fix lasserts in mis-matched transnos during open-unlink testing (1541)
6488 - Debugging for the kqswnal_get_idle_tx problems (1820)
6489 - Allow recovery to be attempted multiple times (1536)
6490 - Write out MDS last_rcvd file after it is first created (1600)
6491 - Fix tx_descriptor leak in failed transmit situations (1827)
6492 - ext3 journaling fixes for assertion failure after IO error (1871)
6493 - class_export_put() on freed export after completion AST error (1896)
6494 - Fix revalidate looping in VFS (1322)
6495 - Don't access a freed export during MDS_REINT timeout (1521)
6496 - Add open-unlink recovery support on the MDS (1673,1764)
6497 - Return an error if no MDS data was read from last_rcvd (1946)
6498 - Fix for lookup "." or ".." crash on error (1932,1931,1935)
6499 - Don't setup a disk device that doesn't match exported UUID (317)
6500 - Reduce bulk RPC timeout to avoid cascading client/OST failures (1845)
6501 - avoid committing NULL handle in force close
6502 - local.sh is now a one-stripe LOV configuration
6503 - POSIX utime.4 -EPERM on FIFO not owned by user (56)
6504 - fix ext3 htree duplicate directory entry corruption (1516)
6505 - POSIX creat.13, fstat.1, open.18, stat.3 new file atime/mtime (2020)
6506 - update to new LOV EA format (2097)
6507 - interoperability for different PAGE_SIZE/wordsize (686,1821,1343,2042)
6509 2003-06-15 Phil Schwan <phil@clusterfs.com>
6512 - imports and exports cleanup too early, need refcounts (349, 879, 1045)
6513 - per-import/export recovery handling (958, 931, 959)
6514 - multiple last-rcvd slots, for serving multiple FSes (949)
6515 - connections are again shared between multiple imp/exports (963, 964)
6516 - "umount -f" would hang if any requests needed to be sent (393, 978)
6517 - avoid pinning large req buffer by copying for queued messages (989)
6518 - add "uuid" to "lctl device" command to help upcalls (991)
6519 - "open" RPCs with transnos would confuse recovery counters (1037)
6520 - do proper endian conversion of all wire messages (288, 340, 891)
6521 - remove OST bulk get LBUGs, fix ost_brw_write cleanup (1126)
6522 - call waiting locks callback from LDLM recovery thread (1127, 1151)
6523 - fix ptlrpc_connection leak in target_handle_connect (1174)
6524 - fix import refcounting bug in OST and MDS cleanup (1134)
6525 - if an invalid-at-open-time OSC returned before close(), LBUG (1150)
6526 - fix very unlikely obd_types race condition (501)
6527 - remove osc_open hack for echo_client (1187)
6528 - we leaked exports/dlmimps for forcibly disconnected clients (1143)
6529 - a failure in read_inode2 leads to deadlock (1139)
6530 - cancel ack-locks as soon as transaction is committed (1072)
6531 - fix major leaks and crashes in the bulk I/O path (937, 1057)
6532 - make sure to commitrw after any preprw to avoid deadlock (1162)
6533 - failing to execute a file in a lustre FS would lock inode (1203)
6534 - small DEBUG_REQ fix to avoid dereferencing a NULL (1227)
6535 - don't ASSERT while cleaning up an incompletely-setup obd (1248)
6536 - obd_uuid2tgt would walk off the end of the list (1255)
6537 - on IA64 the osc would give portals incorrect bulk size (1258)
6538 - fix debug daemon ioctl interface; allows daemon on ia64 (1274)
6539 - fix lock inversion caused by new llite matching code (1282)
6540 - limit the number of dirty pages on a client to 10MB (1286)
6541 - timed out locks were not being corrected cancelled (1289)
6542 - fix O_DIRECT above 4GB on IA-32 (1292)
6543 * major user-visible changes
6544 - fail out/fail over policy now controlled by the upcall (993)
6546 - add OBD_PING to check server availability and failure (954)
6547 - lustre messages are now sent in sending host order (288, 340, 891)
6548 - add eadatalen to MDS getattr reply (340)
6549 - OST read replies may contain second buffer, with per-page status (593)
6551 2003-03-11 Phil Schwan <phil@clusterfs.com>
6554 - LDLM_DEBUG macro fix, for gcc 3.2 (850)
6555 - failed open()s could cause deadlock; fixed (867, 869)
6556 - stop cancelling OST locks when files are closed (481)
6557 - overlapping XID spaces caused network corruption (851, 853)
6558 - fix unsafe fsfilt counter arithmetic; change to atomic_t
6559 - setattr_raw added, to do single-RPC, server-side setattrs
6560 - lmc/lconf syntax change for OST UUIDs
6561 - fix crashy race condition between ptlrpc_free_req and osc_close
6562 - don't use request in mdc_enqueue if we hit a timeout (889)
6563 - don't set the inode i_size for regular files from the MDS (896)
6564 - handle out of order completion AST (842)
6565 - don't LBUG if a lock request times out after receiving AST (913)
6566 - avoid d_rehash race in ll_find_alias by rehashing inside dcache_lock
6567 - if a bad lock AST arrives, send an error instead of dropping entirely
6568 - return 0 from revalidate2 if ll_intent_lock returns -EINTR (912)
6569 - fix leak in bulk IO when only partially completed (899, 900, 926)
6570 - fix O_DIRECT for ia64 (55)
6571 - (almost) eliminate Lustre-kernel-thread effects on load average (722)
6572 - C-z after timeout could hang a process forever; fixed (977)
6574 - client-side I/O cache (678, 924, 929, 941, 970)
6576 - READPAGE and SETATTRs which don't take server-side locks get
6579 2003-02-11 Phil Schwan <phil@clusterfs.com>
6582 - Fix ldlm_lock_match on the MDS to avoid matching remote locks (592)
6583 - Fix fsfilt_extN_readpage() to read a full page of directory
6584 entries, or fake the remainder if PAGE_SIZE != blocksize (500)
6585 - Avoid extra mdc_getattr() in ll_intent_lock when possible (534, 604)
6586 - Fix imbalanced LOV object allocation and out-of-bound access (469)
6587 - Most intent operations were removed, in favour of a new RPC mode
6588 that does a single RPC to the server and bypasses most of the VFS
6589 - All LDLM resource ID arrays were removed in favour of ldlm_res_id
6590 - Aggressively cancel local locks on DLM servers
6591 - mds_reint_unlink sends EA to the client if it's the last nlink.
6592 client uses that EA to unlink OST objects.
6593 - mds_reint_{rename,unlink,link} were rewritten to take ordered locks
6594 - recursive symlinks were fixed (439)
6595 - fixed NULL deref in DEBUG_REQ
6596 - filter_update_lastobjid no longer calls sync, which annoyed extN
6597 - fixed multi-client small-writes to a single file problem (445)
6598 - fixed mtime updates during file writes (607)
6599 - fixed vector writes on obdfilter causing problems when ENOSPC (670)
6600 - fixed bug in obd_brw_read/write() (under guise of testing 367)
6601 - fixed Linux OST size reporting problem (444, 656)
6602 - OST now updates object mtime with writes or setattr (607, 619)
6603 - client verifies file size before zeroing page past EOF (445)
6604 - OST now writes last allocated objid to disk with allocation (108)
6605 - LOV on echo now works (409)
6607 - mds_reint_unlink sends a new buffer, with the EA included. this
6608 buffer is only valid if body->valid & OBD_MD_FLEASIZE, which is only
6609 set if a regular file was being unlinked, and it was the last link
6610 - use PtlGet from the target for bulk writes (315)
6611 - OST now updates object mtime with writes or setattr (607, 619)
6612 - LDLM now has a grant-time callback to revalidate locked items, if
6614 - Many MDS operations were reorganized to combat race conditions
6616 - Merge b_intel branch (updated lprocfs code) - now at /proc/fs/lustre
6617 - configure check to avoid gcc version 2.96 20000731-2.96-98) (606)
6619 2003-01-06 Andreas Dilger <adilger@clusterfs.com>
6622 - Fully reactivate OST imports after reconnection (512, others)
6623 - Make sure client sees our -ENOTCONN from mds_handle (513 - partial)
6624 - More graceful error handling for truncating on dead OST (515)
6625 - Don't error out unless we're actually accessing dead stripes (474)
6626 - Fix garbage sizes when stripes are missing (410)
6627 - LRU counters were broken, causing constant lock purge (433, 432)
6628 - garbage on read from stripes with failed OSTs (441)
6629 - mark OSCs as active before reconnecting during recovery (438)
6630 - lov_enqueue and lov_cancel need to handle inactive OSTs (403)
6631 - lfind did not preserve OST order in output (443)
6632 - symlinks cause hung clients, incorrect data (439)
6633 - stop dereferencing request after dropping refcount (457)
6634 - don't LASSERT(spin_is_locked) on non-SMP (455)
6635 - fixes for many rename() bugs
6636 - fstat didn't correctly synchronize attributes (399)
6637 - server must handle lock cancellation during blocking AST prep (487)
6638 - bulk descriptors were free()d too soon (511)
6639 - fix paths in lconf, which would load incorrect modules (451, 507)
6640 - fix confusing lconf 'host not found' error message (386)
6641 - fix lock order deadlock on OST (O/R i_sem before journal ops, 478)
6642 - fix race condition in mdc_blocking_ast() for inode access (526)
6643 - fix lov_unpackmd() unpacking wrong number of stripes (537)
6644 - fix lov_set_osc_active() marking wrong OSC inactive (440)
6645 - fix bad lstripe lov_unpackmd() assertion (fix layering too) (527)
6646 - fix multiple writes of stripe MD to MDS (358, maybe 519)
6647 - fix lstripe in several ways (kernel side) (527)
6648 - fix request leak in ldlm_cli_enqueue (262)
6649 - incorrect OSC was marked inactive after OST failure
6650 - call mds_fs_cleanup before unmounting filesystem (524)
6651 - fix races between taking ns_lock and ldlm_lock_change_resource
6652 - fix races updating LOV export open file list
6653 - fix lov_enqueue error path, avoid decref-ing bad lock handle
6654 - fix recovery NULL deref in ldlm_cli_cancel_unused
6655 - fix some DLM races by using new hash table for lock handles (419)
6656 - permit the client to specify desired inodes, at replay
6657 - duplicate requests when we queue them for replay reintegration
6658 - fix last_rcvd offset calculation
6659 - sync after each recovered transaction, so we always make progress
6660 - never, not always, ERESTART requests without transnos
6661 - store the lov_desc in the MDS, so we don't depend on getlovinfo to
6663 - skip replay if the MDS says that the client is already connected
6664 - don't check for a recovery-enabled export to match lctl's UUID
6665 - don't INC_USE_COUNT for phantom exports
6666 - don't crash when cleaning up phantom exports (567)
6667 - don't double-finish or set replay data for errored mdc_open requests
6668 - abort requests when they time out, so we don't get old replies
6669 - send/receive replies for AST messages again
6670 - if the client says that it doesn't have the lock, cancel it on the
6672 - if we timeout during I/O, don't try to cancel an in-use lock; instead
6673 mark it as destroyed, it will all work out when decref is called
6674 - fix module use counts (22, 581)
6676 - ASTs now expect a reply (server cancels lock on error reply)
6678 2002-12-02 Andreas Dilger <adilger@clusterfs.com>
6681 - fix many simultaneous client startup (392)
6682 - fix dentry->d_it clobbering
6683 - credentials weren't being shipped for readdir/getattr operations
6684 - remove invalid assertions triggered during some concurrent MD
6686 - proper Lustre versions added (336, 389)
6687 - fix memory leak for create error case (398)
6688 - fix LOV locking bug that would get cli/srv out of sync
6689 - fix echo client over LOV (409)
6690 - fix dbench 2, extN refcount problem (170, 258, 356, 418)
6691 - fix double-O_EXCL intent crash (424)
6692 - avoid sending multiple lock CANCELs (352)
6694 - MDS can do multi-client recovery (modulo bugs in new code)
6696 - many updates, edits, cleanups
6698 2002-11-18 Phil Schwan <phil@clusterfs.com>
6701 - fix null d_it dereference (346)
6702 - fix full OST/dbench hang (333)
6703 - fix permission problem with file removal (286)
6704 - fix removal of OSCs from LOV when they fail
6705 - fix NULL deref during bulk timeout (214)
6706 - fix problems related to multiple filesystems on one MDS (241)
6707 - fixed serious subtle metadata locking bugs
6708 - free locks on clients when inodes are removed due to memory
6710 - fix inode pointer in lock data (285)
6711 - partial support for multiple MDS on a single host (241)
6712 - data locks weren't cancelled at clear_inode time (290, 311)
6713 - intent locks could lead to unbounded lock growth (205)
6714 - added a maximum lock count, an LRU list, and a flusher
6715 - fix multiple rename (365)
6716 - properly abstracted the echo client
6717 - OSC locked 1 byte too many; fixed
6718 - rewrote brw callback code:
6719 - fixed recovery bugs related to LOVs (306)
6720 - fixed too-many-pages-in-one-write crash (191)
6721 - fixed (again) crash in sync_io_timeout (214)
6722 - probably fixed callback-related race (385)
6724 - Add capability to MDS protocol
6725 - LDLM cancellations and callbacks on different portals
6727 2002-10-28 Andreas Dilger <adilger@clusterfs.com>
6730 - limit client IOV size to PTL_MD_MAX_IOV (611336, 191)
6731 - defer open object destruction to close time (601981, 138)
6732 - open/close OST file handle in obdo (OBD_MD_FLHANDLE) (601981, 138)
6733 - move LDLM_ENQUEUE/CONVERT back to MDS portal (625069)
6734 - abstract ll_lookup2, fix ll_revalidate2 to use abstraction (256)
6735 - don't call obd_setattr in ll_file_release for destroyed objects
6736 * protocol change to lustre_msg: move |version| and add |flags|
6737 * protocol change to osc_punch: "start" in "o_size", "end" in "o_blocks"
6738 * lock replay: for LDLM_FL_REPLAY trust client to do right thing
6739 * added replay of create, unlink, link and rename operations during
6740 MDS failover; recovery should be much more robust now
6741 * remove failed OSCs from LOVs (only lov_create uses this so far)
6742 * the lustre-HOWTO was brought (more) up to date (582544)
6744 2002-10-23 Phil Schwan <phil@clusterfs.com>
6747 - in-use dentries weren't being reused properly (617851)
6748 - prevent multiple LDLM setup (599178)
6749 - fix LOV size calculations for truncate (617853)
6750 - fix client handling of MDS intent errors (POSIX)
6751 - fix permission bug in lovstripe.c test (624321)
6752 - fix MDS thread deadlock - move LDLM handler to DLM portal (625069)
6753 - truncate past end of file could corrupt data
6754 - proper cleanup after timeouts, crashes, etc (592524, 550815)
6755 - a race in recovery could return ETIMEDOUT to apps (623947)
6756 - building outside the source directory was fixed
6757 * the lustre-HOWTO was brought (more) up to date (582544)
6758 * major progress was made on recovery functionality
6760 2002-10-10 Phil Schwan <phil@clusterfs.com>
6763 - recovery deadlock fix
6764 - rm -rf causes LBUG fix (617817)
6765 - file open by multiple tasks fix (618962)
6766 - directory permissions bugs (602707 and 620007)
6767 - journal_stop fixed with locking (611313)
6768 - O_APPEND failures resolved (618273, perhaps 614459)
6769 - lconf PATH fix (619770)
6770 - IA64 build fix (621450)
6771 - RPC buffer sizes scale with amount of memory
6773 2002-10-01 Phil Schwan <phil@clusterfs.com>
6776 - locks would be cancelled without throwing away data pages,
6777 resulting in inconsistent data (605627)
6778 - inode attributes were not always being refreshed (605627, 612449)
6779 - lconf now continues to cleanup after lctl reports an error
6780 - MDS now enforces user permissions (602707)
6781 - lprocfs cleanup fixed, but not yet enabled (614157)
6782 - fixed infinite server hang, should a client not respond to an AST
6783 - avoid going into recovery if user calls readlink() with a buffer
6784 that's too small (613941)
6785 - AST RPCs no longer require replies (614867) -- this may be changed
6786 - don't crash server if client sends an IOV that's too big (611336)
6787 - fixed lock conversion deadlock (611892)
6788 - fixed the following of symlinks (614622)
6789 * recovery: the server can remove locks from a client that dies, other
6790 clients can make progress
6791 * more extN patch fixes
6792 * compile-time configurable ptlrpc buffer allocations
6794 - collaborative read cache document
6795 - Lustre Lite Performance CDR document-in-progress
6797 2002-09-20 Andreas Dilger <adilger@clusterfs.com>
6800 - fix typo in patch-2.4.18
6802 2002-09-20 Andreas Dilger <adilger@clusterfs.com>
6805 - clear ptlrpc request each time in handle_incoming_request()
6806 - unlink of files now destroys the object on the OST
6808 2002-09-19 Peter Braam <braam@clusterfs.com>
6810 * add hard link support
6811 * change obdfile creation method
6812 * kernel patch changed
6814 2002-09-19 Peter Braam <braam@clusterfs.com>
6817 - stack overflow bug in extN fixed
6819 2002-09-18 Andreas Dilger <adilger@clusterfs.com>
6821 * documentation updates
6822 - add man pages for config tools
6823 - update tests/README to describe testing with new config tools
6824 - finish metadata API descriptions
6825 * bug fixes and cleanups
6826 - statfs workaround for 16TB limit
6827 - LOV stripe allocation improved, can stripe on subset of OSTs
6828 - LOV file size/IO offset was wrong for files > 4GB in size
6829 - object EA data was being dropped, caused files to be unreadable
6830 - memory overflow with non-LOV OST caused memory corruption
6831 - fixed regression tests to work with new config tools, obdfilter
6832 - fixed bug when directory size became larger than 1 block
6833 - fixed bug (for single client case) when PWD was deleted
6834 - invalidate local directory pages when doing intent-based ops
6835 - avoid LDLM oops when lock callback contained bad data
6837 2002-09-09 Andreas Dilger <adilger@clusterfs.com>
6839 * documentation updates
6840 * bug fixes and cleanups
6841 - configuration tools
6844 - 64-bit compile warnings
6845 - 64-bit internal statfs data
6847 * test_brw on persistent OST devices
6849 * lprocfs (disabled)
6851 2002-09-04 Andreas Dilger <adilger@clusterfs.com>
6853 * documentation updates
6854 * bug fixes and cleanups
6855 * configuration tools
6857 2002-08-30 Peter J. Braam <braam@clusterfs.com>
6860 * many small fixes to 0_5_4
6861 * io/network handling
6862 * thinkos in MDS operations
6864 2002-08-24 Peter J. Braam <braam@clusterfs.com>
6867 * crucial basic fixes to 0.5.3
6868 * IOR, Iozone work over Elan
6871 2002-08-07 Phil Schwan <phil@clusterfs.com>
6872 * version 0_5_3, our first alpha
6873 * we use the new Portals iovs
6874 * documentation updates
6875 * bug fixes and cleanups
6876 * small changes in the DLM wire protocol
6878 2002-07-25 Peter J. Braam <braam@clusterfs.com>
6879 * version 0_5_1 with some initial stability,
6880 * locking on MD and file I/O.
6881 * documentation updates
6882 * several bug fixes since 0.5.0
6883 * small changes in wire protocol
6885 2002-07-18 Phil Schwan <phil@clusterfs.com>
6887 * delivered as Lustre Light Alpha
6888 * fixed a crash after handling invalid MDS requests
6889 * fixed directory pages for architectures with non-4k pages sizes
6891 2002-07-11 Andreas Dilger <adilger@clusterfs.com>
6892 * release version v0_4_4
6893 * Moves TCP acceptor to be on port 2432 (unused Coda port) instead
6895 * Fixes a number of interruption problems with OST operations.
6896 * Update documentation for portals header changes
6897 * Move all wire protocol structs/defines to lustre_idl.h
6898 * Fixes symlink length bug.
6899 * Add tcpdump to repository.
6901 2002-07-05 Andreas Dilger <adilger@clusterfs.com>
6902 * release version v0_4_3
6903 * Fixes statfs for inodes on extN.
6904 * Fixes bug in runtests which would delete /etc/hosts.
6905 * Use 64-bit object IDs wherever possible (not into VFS though)
6906 Remove ost_get_info, which is unused by lustre, and out of date.
6908 2002-07-03 Peter Braam <braam@clusterfs.com>
6909 * release version v0_4_2 Fixes a lookup error (type not passed)
6910 * move forward to head of Portals
6911 * move forward to latest Lustre kernel
6913 2002-06-25 Peter Braam <braam@clusterfs.com>
6914 * release version v0_4_1. Hopefully stable on single node use.