1 2010-01-29 Sun Microsystems, Inc.
4 2.6.16.60-0.42.8 (SLES 10),
5 2.6.27.39-0.3.1 (SLES11),
6 2.6.18-164.11.1.el5 (RHEL 5)
7 2.6.18-164.6.1.0.1.el5 (OEL 5)
8 * Client support for unpatched kernels:
9 (see http://wiki.lustre.org/index.php?title=Patchless_Client)
10 2.6.16 - 2.6.30 vanilla (kernel.org)
11 * Recommended e2fsprogs version: 1.41.6.sun1
12 * The async journal commit feature (bug 19128) and the cancel
13 lock before replay feature (bug 16774) are disabled by default.
15 Severity : enhancement
17 Description: Update RHEL5.4 kernel to 2.6.18-164.11.1.el5.
19 Severity : enhancement
20 Bugzilla : 21511/19848
21 Description: Update SLES11 kernel to 2.6.27.39-0.3.1.
23 Severity : enhancement
25 Description: Update supported SLES10 kernel to 2.6.16.60-0.42.8.
27 Severity : enhancement
29 Description: Update kernel to RHEL5.4 2.6.18-164.6.1.el5 and
30 OEL5 2.6.18-164.6.1.0.1.el5(Both in-kernel OFED enabled).
32 Severity : enhancement
34 Description: Build kernels (RHEL5, OEL5 and SLES10/11) using the vendor's own
37 Severity : enhancement
39 Description: Vanilla kernel 2.6.30 patchless client support.
44 Description: bad entry in directory xxx: inode out of bounds
45 Details : fix locking issue in the rename path which could race with any
46 other operations updating the same directory.
48 Severity : enhancement
50 Description: Make watchdog timer messages to be more clear and descriptive.
54 Description: cp -p command does not preserve the dates and timestamp
55 Details : mtime could be spoiled by a write callback
59 Description: Clear imp_force_reconnect correctly in ptlrpc_connect_interpret()
63 Description: Allow non-root access for "lfs check".
64 Details : Added a check in obd_class_ioctl() for OBD_IOC_PING_TARGET.
66 Severity : enhancement
68 Description: quotacheck performance/scaling issues
69 Details : reduce quotacheck time on empty filesystem by skipping uninit
72 Severity : enhancement
74 Description: Enhancement for lfs(1) command to use numeric uid/gid.
76 Severity : enhancement
78 Description: Adjust locks' extents on their first enqueue, so that at the time
79 they get granted, there is no need for another pass through the
80 queues since they are already shaped into the proper forms.
84 Description: Fix mds_shrink_intent_reply()/mds_intent_policy() to pass correct
85 arguments and prevent LBUG() in lustre_shrink_reply_v2().
89 Description: Change tunefs.lustre and mkfs.lustre --mountfsoptions so that
90 exactly the specified mount options are used. Leaving off
91 any "mandatory" mount options is an error. Leaving off any
92 default mount options causes a warning, but is allowed.
93 Change errors=remount-ro from mandatory to default. Sanitize
94 the mount string before storing it. Update man pages accordingly.
98 Description: mds_getattr() should return 0, even if mds_fid2entry() fails with
99 -ENOENT. Also fix in ptlrpc_expire_one_request() to print signed
102 Severity : enhancement
104 Description: Remove set_info(KEY_UNLINKED) from MDS/OSC
106 Severity : enhancement
108 Description: Clients can replay thousands of unused locks during recovery
109 Details : Don't replay unused locks (only read locks for now) during
110 recovery. This feature is disabled by default and can be
111 enabled by running the following command on the clients:
112 lctl get_param ldlm.cancel_unused_locks_before_replay
116 Description: can't stat file in some situation.
117 Details : improve initialize osc date when target is added to mds and
118 ability to resend too big getattr request is client isn't have info
123 Description: Prevent inconsistences between linux and lustre mount structures.
124 Details : Wait indefinitely in server_wait_finished() until mnt_count drops.
125 Make the sleep interruptible.
127 Severity : enhancement
129 Description: Communicate OST degraded/readonly state via statfs to MDS
130 Details : Flags in the statfs returned from OSTs indicate whether the
131 OST is in a degraded RAID state, or if the filesystem has
132 turned read-only after a filesystem error is detected.
137 Description: don't panic if EPROTO was hit when reading symlink
138 Details : correctly handling request reference in error cases.
143 Description: open sometimes returns ENOENT instead of EACCES
144 Details : checking permission should be part of open part of mds_open, not
145 lookup part. so server should be set DISP_OPEN_OPEN disposition
146 before starting permission check.
147 Also not need revalidate dentry if client already have LOOKUP lock.
151 Description: enable client interface failover
152 Frequency : on servers with multiple network interfaces
153 Details : When a child reconnects from another NID, properly update export
154 nid hash position and ldlm reverse import.
156 Severity : enhancement
158 Description: implemented direct I/O with arbitrary (nonaligned) memory
159 addresses and file offsets.
161 Severity : enhancement
163 Description: added more recovery timeout options.
165 Severity : enhancement
167 Description: added llapi_file_open, llapi_file_create, llapi_file_get_stripe
172 Description: Avoid deadlock for local client writes
173 Frequency : only on systems with clients writing to an OST on the same node
174 Details : Use new OBD_BRW_MEMALLOC flag to notify OST about writes in the
175 memory freeing context. This allows OST threads to set the
176 PF_MEMALLOC flag on task structures in order to allocate memory
177 from reserved pools and complete IO.
178 Use GFP_HIGHUSER for OST allocations for non-local client writes,
179 so that the OST threads generate memory pressure and allow
180 inactive pages to be reclaimed.
185 Descriptoin: lock ordering violation between &cli->cl_sem and _lprocfs_lock
186 Details : move ldlm namespace creation in setup phase to avoid grab
187 _lprocfs_lock with cli_sem held.
191 Frequency : only during format of test systems
192 Description: Unable to run several mkfs.lustre on loop devices at the same time
193 Details : mkfs.lustre returns error 256 on the concurrent loop devices
194 formatting. The solution is to proper handle the error.
196 Severity : enhancement
198 Description: implement async create (obd_async_create) method for osc, to avoid
199 too long waiting new ost objects with holding ldlm lock.
203 Frequency : occasionally during network problems
204 Description: client not allowed to reconnect to OST because of active request
205 Details : abort bulk requests received by the OST once the client has timed
206 out since the client will resend the request anyway. The client
207 also now retries to reconnect to the same server if a connect
208 request failed with EBUSY or -EAGAIN.
211 Frequency : rare, if used wide striped file and one ost in down.
213 Descriptoin: don't return error if we created a subset of objects for file.
214 Details : lov_update_create_set() uses set->set_success as index for created
215 objects, so if some requests failed, they will have hole at end of
216 array and we can use qos_shrink_lsm for allocate correct lsm.
220 Description: Slow stale export processing during normal start up
221 Details : The global mgc lock prevents OST setup to be run in parallel.
222 Replace the global lock with a per-config_llog_data semaphore.
226 Description: Out or order replies might be lost on replay
227 Details : In ptlrpc_retain_replayable_request if we cannot find retained
228 request with tid smaller then one currently being added, add it
229 to the start, not end of the list.
233 Description: BUG: soft lockup - CPU#1 stuck for 10s! [ll_mdt_07:4523]
234 Details : add cond_resched() calls to avoid hogging the cpu for too long
235 in the hash code. Make also lustre_hash_for_each_empty() more
238 Severity : enhancement
240 Description: Performance improvements for debug messages with D_RPCTRACE,
241 D_LDLM, D_QUOTA options.
244 Frequency : only with NFS export
246 Description: (lov_merge.c:74:lov_merge_lvb())
247 ASSERTION(spin_is_locked(&lsm->lsm_lock)) failed (SR 71691004)
248 Details : Fix a race in the nfs export code by populating inode
249 info while the new inode is still locked
251 Severity : enhancement
253 Description: add a new file in procfs called force_lbug. Writting to this
254 file triggers a LBUG. Only for test purpose.
258 Description: OOM killer causes node hang
259 Details : really interrupt the sleep in osc_enter_cache on signals
263 Description: LustreError: 9153:0:(quota_context.c:622:dqacq_completion()) LBUG
264 Details : fix race during quota release on the slave.
266 Severity : enhancement
268 Description: smaller hash bucket sizes, cleanups
269 Details : increase hash table sizes and enabled rehashing for pools, uuid,
272 Severity : enhancement
274 Description: Add ldiskfs maxdirsize mount option
275 Details : add max_dir size mount option
279 Description: panic in ll_statahead_thread
280 Details : prevent parent thread to be killed before its child
283 Frequency : only with 16TB device
285 Description: unable to perform "mount -t lustre" of 16TB OST device
286 Details : Mounting 16TB LUNs failed due to three bugs in mkfs.lustre.
290 Description: ASSERTION(atomic_read(&imp->imp_inflight) == 0) failed
291 Details : unregistering should be zero if no RPC inflight.
295 Description: hyperion: Oops during metabench
296 Details : Correct the refcount of lov_request_set
298 Severity : enhancement
300 Description: Add mptlinux and nxge drivers to Lustre builds
302 Severity : enhancement
304 Description: Fix watchdog timer message to be more clear
305 Details : Make watchdog timer messages more clear and descriptive.
309 Description: LNET soft lockups in socknal_cd thread
310 Details : don't hog CPU for active-connecting if another connd is
311 accepting connecting-requst from the same peer
315 Description: recovery-small test_17 hang
316 Details : Land several AT improvements & fixes.
320 Description: MDS panic and hanging client processes
321 Details : Replace exp_ops_stats with exp_nid_stats->nid_stats
325 Description: OSS stuck in recovery.
326 Details : fix race during recovery. class_unlink_export,
327 class_set_export_delayed and target_queue_last_replay_reply
328 may race while increasing/decreasing obd_recoverable_clients
329 and obd_delayed_clients, causing recovery to wait forever.
331 Severity : enhancement
333 Description: add cascading_rw.c to lustre/tests
337 Description: filter_last_id() NULL deref
338 Details : lprocfs_filter_rd_last_id() should check for the fully
339 setup obd device, before proceeding further.
341 Severity : enhancement
343 Description: Loadgen improvements
344 Details : stacksize and locking fixes for loadgen
347 Description: Quiet CERROR("dirty %d > system dirty_max %d\n"
348 Details : The atomic_read() allowing the atomic_inc() are not covered
349 by a lock. Thus they may safely race and trip this CERROR()
350 unless we add in a small fudge factor (+1).
352 Severity : enhancement
354 Description: shrink_slab: nr=-9223362083340912175
355 Details : fix spurious message from shrink_slab reporing negative nr
358 Description: Quiet bogus previously committed transno error
359 Details : suppress the "server went back in time" error message which
360 is always printed even in the common case after a client eviction
363 Description: Parallel statfs() calls result in client eviction
364 Details : cache statfs data for 1s.
367 Description: parallel-scale test_compilebench: @@@@@@ FAIL: compilebench
369 Details : fix serveral issues in pinger code causing clients not to ping
370 servers for too long, resulting in evictions.
373 Description: e2fsck should warn when MMP update interval is extended
374 Details : print mmp_check_interval and make it possible to abort mount
375 operation in case it takes too long.
378 Description: mdsrate-create-large.sh, BUG: soft lockup - CPU#0 stuck for 10s!
379 Details : fix bug in the RHEL5's jbd2 callback patch.
382 Description: drop number of active requests when queued for recovery
383 Details : Now that we take a reference on the original request instead of
384 making a copy of it for recovery. We need to drop the number of
385 active requests or the queued requests will prevent all request
386 processing when they exceed (srv->srv_threads_running - 1).
389 Description: refuse to invalidate operational quota files when they are in use
390 Details : an attempt to invalidate operational quota files on the quota
391 master is not actually permitted by VFS (returning -EPERM), but we
392 should not depend on that and should return the error earlier.
394 -------------------------------------------------------------------------------
396 2009-10-16 Sun Microsystems, Inc.
398 * Support for kernels:
399 2.6.16.60-0.42.4 (SLES 10),
400 2.6.27.29-0.1 (SLES11, i686 & x84_64 only),
401 2.6.18-128.7.1.el5 (RHEL 5),
402 2.6.18-128.7.1.el5 (OEL 5),
403 * Client support for unpatched kernels:
404 (see http://wiki.lustre.org/index.php?title=Patchless_Client)
405 2.6.16 - 2.6.27 vanilla (kernel.org)
406 * Recommended e2fsprogs version: 1.41.6.sun1
407 * ext4 support for RHEL5 is experimental and thus should not be
410 Severity : enhancement
412 Description: Add OEL5 support.
414 Severity : enhancement
416 Description: Update kernel to SLES11 2.6.27.29-0.1.
420 Description: File checksum failures with OST read cache on
421 Details : Disable page poisoning when the bulk transfer has to be aborted
422 because the client got evicted.
426 Description: Don't allow make backward step on assiging osc next id.
427 Details : race between allocation next id and ll_sync thread can be cause
428 of set wrong osc next id and can be kill valid ost objects.
430 Severity : enhancement
432 Description: Update kernel to RHEL5 2.6.18-128.7.1.el5.
434 Severity : enhancement
436 Description: Update kernel to SLES10 SP2 2.6.16.60-0.42.4.
440 Description: Changes in raid5-large-io-rhel5.patch to calculate sectors properly
444 Description: Increase the default BLK_DEF_MAX_SECTORS value for RHEL5 and SLES11
448 Description: Error handling in osc_statfs_interpret() has been improved.
449 Details : Check in osc_statfs_interpret() for EBADR.
453 Description: Do not send statfs() requests to OSTs disabled by administrator.
454 Details : Check in lov_prep_statfs_set() for non-NULL ltd_exp.
458 Description: Do not update ctime for the deleted inode.
459 Details : Check in mds_reint_unlink() before calling fsfilt_setattr().
463 Description: Increase of the size of the LDLM resource hash.
464 Details : Bump up RES_HASH_BITS=12.
468 Description: correctly send lsm on open replay
469 Details : MDS is trust to LSM size on replay open, but client can set wrong size
474 Description: Deadlock between filter_destroy() and filter_commitrw_write().
475 Details : filter_destroy() does not hold the DLM lock over the whole
476 operation. If the DLM lock is dropped, filter_commitrw() can go
477 through, causing the deadlock between page lock and i_mutex.
478 The i_alloc_sem should also be hold in filter_destroy() while
483 Description: truncate starts GFP_FS allocation under transaction causing deadlock
484 Details : ldiskfs_truncate calls grab_cache_page which may start page
485 allocation under an open transaction. This may lead to
486 calling prune_icache with consequent lustre reentrance.
489 Frequency : only when down/upgrading the MDS to 1.6/1.8 while 1.8 clients are
490 still up and when the OST pool feature is used
492 Description: interop testing got LBUG when run dd with OST pool
493 :LustreError: 30032:0:(llite_lib.c:1913:ll_replace_lsm()) LBUG
494 Details : down/upgrading the MDS to a version that doesn't/does support OST
495 pool can cause clients to crash because the lsm has changed
500 Description: missing tree_status on 1.8.1 RPM build
501 Details : make rpms failed due because the tree_status file is missing.
505 Description: continuing LustreError "mds adjust qunit failed!"
506 Details : don't print message on the console when ->adjust_qunit fails.
510 Description: don't increase ldlm timeout if previous client was evicted
511 Details : if a client doesn't respond to a blocking callback within the
512 adaptive ldlm enqueue timeout, don't adjust the adaptive estimate
513 when the lock is next granted.
517 Description: ost is being unmounted w/o all writes to last_rcvd landing on disk.
518 affects recovery negatively.
519 Details : make sure all exports have been properly destroyed by the zombie
520 thread processed before stopping the target.
524 Description: Performance degradation with O_DIRECT between 1.6 & 1.8.1 b190
525 Details : disable write barrier for ext4/SLES11.
529 Description: Kernel panic - not syncing: Out of memory and no killable
530 processes... on OSS when iozone
531 Details : fix memory leak in the journal checksum patch.
535 Description: group quota "too many blocks" OSS crashes
536 Details : we should keep the same uid/gid for lquota_chkquota() and
537 lquota_pending_commit()
541 Description: LustreError: 9153:0:(quota_context.c:622:dqacq_completion()) LBUG
542 Details : don't LBUG on release quota error. Just a workaround until the
543 problem is understood.
545 ------------------------------------------------------------------------------
546 2009-07-31 Sun Microsystems, Inc.
548 * Support for kernels:
549 2.6.16.60-0.39.3 (SLES 10),
550 2.6.27.23-0.1 (SLES11, i686 & x84_64 only),
551 2.6.18-128.1.14.el5 (RHEL 5),
552 * Client support for unpatched kernels:
553 (see http://wiki.lustre.org/index.php?title=Patchless_Client)
554 2.6.16 - 2.6.27 vanilla (kernel.org)
555 * Recommended e2fsprogs version: 1.41.6.sun1
556 * File join has been disabled in this release, refer to Bugzilla 16929.
557 * NFS export disabled when stack size < 8192. Since the NFSv4 export
558 of Lustre filesystem with 4K stack may cause a stack overflow. For
559 more information, please refer to bugzilla 17630.
560 * ext4 support for RHEL5 is experimental and thus should not be
563 Severity : enhancement
565 Description: Update kernel to SLES10 SP2 2.6.16.60-0.39.3.
568 Frequency : with 1.8 server and 1.6 clients
570 Description: correctly shrink reply for avoid send too big message to client.
571 Details : 1.8 mds is allocate to big buffer to LOV EA data and this produce
572 some problems with sending this reply to 1.6 client.
576 Description: Repeated atomic allocation failures.
577 Details : Use GFP_HIGHUSER | __GFP_NOMEMALLOC flags for memory allocations
578 to generate memory pressure and allow reclaiming of inactive pages.
579 At the same time, do not allow to exhaust emergency pools.
580 For local clients the use of GFP_NOFS will be introduced in 1.8.2
582 Severity : enhancement
583 Bugzilla : 19846, 18289
584 Description: Update kernel to RHEL5 2.6.18-128.1.14.el5.
586 Severity : enhancement
587 Bugzilla : 19625, 16893, 18668, 19848
588 Description: Add support for SLES11 2.6.27.23-0.1.
590 Severity : enhancement
592 Description: Update client support to vanila kernels up to 2.6.27.
594 Severity : enhancement
596 Description: Update kernel to SLES10 SP2 2.6.16.60-0.37.
598 Severity : enhancement
600 Description: Compile with -Werror by default for i686 and x86_64.
604 Description: resolve race between obd_disconnect and class_disconnect_exports
605 Details : if obd_disconnect will be called to already disconnected export he
606 forget release one reference and osc module can't unloaded.
608 Severity : enhancement
610 Description: move AT tunable parameters for more consistent usage
611 Details : add AT tunables under /proc/sys/lustre, add to conf_param parsing
615 Descriptoin: correctly skip time estimate if in recovery
616 Details : rq_send_state insn't bitmask so using bitwise ops is forbid.
620 Descriptoin: OSS DeadLock
621 Details : Use trylock to prevent deadlock when shrink icache.
623 Severity : enhancement
625 Description: Allow tuning service thread via /proc
626 Details : For each service a new
627 /proc/fs/lustre/{service}/*/thread_{min,max,started} entry is
628 created that can be used to set min/max thread counts, and get the
629 current number of running threads.
631 Severity : enhancement
633 Description: Add state history info file, enhance import info file
634 Details : Track import connection state changes in a new osc/mdc proc file;
635 add overview-type data to the osc/mdc import proc file.
639 Description: Reduce small size read RPC
640 Details : Set read-ahead limite for every file and only do read-ahead when
641 available read-ahead pages are bigger than 1M to avoid small size
646 Description: free_entry erroneously used groups_free instead of put_group_info
648 Severity : enhancement
650 Description: Make read-ahead stripe size aligned.
652 Severity : enhancement
654 Description: MDS create should not wait for statfs RPC while holding DLM lock.
657 Frequency : rare, connect and disconnect target at same time
659 Descriptoin: ASSERTION(atomic_read(&imp->imp_inflight) == 0
660 Details : don't call obd_disconnect under lov_lock. this long time
661 operation and can block ptlrpcd which answer to connect request.
664 Frequency : start MDS on uncleanly shutdowned MDS device
666 Descriptoin: ll_sync thread stay in waiting mds<>ost recovery finished
667 Details : stay in waiting mds<>ost recovery finished produce random bugs
668 due race between two ll_sync thread for one lov target. send
669 ACTIVATE event only if connect realy finished and import have
673 Frequency : start MDS on uncleanly shutdowned MDS device
675 Descriptoin: aborting recovery hang on MDS
676 Details : don't throttle destroy RPCs for the MDT.
680 Description: Slow reads beyond 8Tb offsets.
681 Details : Page index integer overflow in ll_read_ahead_page
685 Description: Soft lockup on OSS after MDS failover
686 Details : MSG_CONNECT_INITIAL is not set on the initial MDS->OST connect.
687 As a conseqence, the patch from bug 18224 is not operational
688 and the MDS export cannot be reused on the OSTs until it gets
692 Frequency : rare, only if using MMP with Linux RAID
694 Description: MMP doesn't work with Linux RAID
695 Details : While using HA for Lustre servers with Linux RAID, it is possible
696 that MMP will not detect multiple mounts. To make this work we
697 need to unplug the device queue in RAID when the MMP block is being
698 written. Also while reading the MMP block, we should read it from
699 disk and not the cached one.
702 Frequency : rare, during recovery
704 Description: Assertion failure in ldlm_lock_put
705 Details : Do not put cancelled locks into replay list, hold references on
710 Description: 1.6.5 mdsrate performance is slower than 1.4.11/12 (MDS is not cpu bound!)
711 Details : create_count always drops to the min value (=32) because grow_count
712 is being changed before the precreate RPC completes.
715 Frequency : Only in RHEL5 when mounting multiple ext3 filesystems
718 Description: "kmem_cache_create: duplicate cache jbd_4k" error message
719 Details : add proper locking for creation of jbd_4k slab cache
723 Description: MMP check in ext3_remount() fails without displaying any error
724 Details : When multiple mount protection fails during remount, proper error
729 Description: Rare Client crash on resend if the file was deleted.
730 Details : When file is opened, but open reply is lost and file is
731 subsequently deleted before resend, resend processing logic
732 breaks trying to open the file again, should not try to open.
736 Description: add check for >8TB ldiskfs filesystems
737 Details : ext3-based ldiskfs does not support greater than 8TB LUNs.
738 Don't allow >8TB ldiskfs filesystems to be mounted without
739 force_over_8tb mount option
743 Description: Client locked up when running multiple instances of an app. on
744 multiple mount points
745 Details : ll_shrink_cache() can sleep while holding the ll_sb_lock.
746 Convert ll_sb_lock to a read/write semaphore to fix the problem.
750 Description: Cannot acces an NFS-mounted Lustre filesystem
751 Details : An NFS client cannot access the Lustre filesystem NFS-mounted
752 from a Lustre-client exporting the Lustre filesystem via NFS.
756 Description: panic in ll_statahead_thread
757 Details : grab dentry reference in parent process.
760 -------------------------------------------------------------------------------
762 tbd Sun Microsystems, Inc.
764 * Support for kernels:
765 2.6.16.60-0.31 (SLES 10),
766 2.6.18-128.1.6.el5 (RHEL 5),
767 2.6.22.14 vanilla (kernel.org)
768 * Client support for unpatched kernels:
769 (see http://wiki.lustre.org/index.php?title=Patchless_Client)
770 2.6.16 - 2.6.22 vanilla (kernel.org)
771 * Recommended e2fsprogs version: 1.40.11-sun1
772 * File join has been disabled, refer to Bugzilla 16929.
773 * A new Lustre ADIO driver is available for MPICH2-1.0.7.
774 * NFS export disabled when stack size < 8192. Since the NFSv4 export of
775 Lustre filesystem with 4K stack may cause a stack overflow. For more
776 information, please refer to bugzilla 17630.
778 Severity : enhancement
780 Description: Update to RHEL5.3 kernel-2.6.18-128.1.6.el5.
782 Severity : enhancement
784 Description: Update OFED release to 1.4.1 RC4
786 Severity : major, only with big OST
788 Description: Very poor metadata performance on Infiniband lustre configuration
789 Details : OST object precreation becomes very slow on big OSTs. This is due
790 to the ialloc patch spending too much time scanning groups.
793 Frequency : during recovery
795 Description: don't mix llog inodes with normal.
796 Details : allocate inodes for log in last inode group
801 Description: fix lqs' reference which won't be put in some situations
802 Details : This patch fixes:
803 1. In quota_check_common(), this function will check quota
804 for user and group, but only send one return via "pending".
805 In most cases, the pendings should be same. But that is not
807 2. If quotaoff runs between lquota_chkquota() and
808 lquota_pending_commit(), the same thing will happen too.
809 That is why it comes:
810 - if (!ll_sb_any_quota_active(qctxt->lqc_sb))
813 -------------------------------------------------------------------------------
815 2008-12-31 Sun Microsystems, Inc.
817 * Support for kernels:
818 2.6.5-7.314 (SLES 9),
819 2.6.9-67.0.22.EL (RHEL 4),
820 2.6.16.60-0.31 (SLES 10),
821 2.6.18-92.1.17.el5 (RHEL 5),
822 2.6.22.14 vanilla (kernel.org)
823 * Client support for unpatched kernels:
824 (see http://wiki.lustre.org/index.php?title=Patchless_Client)
825 2.6.16 - 2.6.22 vanilla (kernel.org)
826 * Client support for unpatched kernels:
827 we do not recommend using patchless RHEL4 clients with kernels
828 prior to 2.6.9-55EL (RHEL4U5).
829 * Recommended e2fsprogs version: 1.40.11-sun1
830 * Note that reiserfs quotas are disabled on SLES 10 in this kernel.
831 * RHEL 4 and RHEL 5/SLES 10 clients behaves differently on 'cd' to a
832 removed cwd "./" (refer to Bugzilla 14399).
833 * A new quota file format has been introduced in 1.6.5.
834 The format conversion from prior releases is handled transparently,
835 but releases older than 1.4.12/1.6.5 don't understand this new
836 format. The automatic format conversion can be avoided by running
837 the following command on the MDS:
838 'tunefs.lustre --param="mdt.quota_type=ug1" $MDTDEV'.
839 For more information, please refer to bugzilla 13904.
840 * A new quota file format was introduced in 1.6.6/1.8.0 (kernels 2.6.16+).
841 The format conversion from prior releases is handled transparently,
842 but releases older than 1.6.6/1.8.0 don't understand this new
843 format. The automatic format conversion can be avoided by running
844 the following commands on the MDS and OSS servers (for
845 pre 1.4.12-1.6.5 quota files):
846 'tunefs.lustre --param="mdt.quota_type=ug1" $MDTDEV',
847 'tunefs.lustre --param="ost.quota_type=ug1" $MDTDEV'
848 or (for 1.4.12/1.6.5 quota files)
849 'tunefs.lustre --param="mdt.quota_type=ug2" $MDTDEV',
850 'tunefs.lustre --param="ost.quota_type=ug2" $MDTDEV'
851 For more information, please refer to bugzilla 13904.
852 * Output of lfs quota has been made less detailed by default,
853 old (verbose) output can be obtained by using -v option.
854 * File join has been disabled in this release, refer to Bugzilla 16929.
855 * A new Lustre ADIO driver is available for MPICH2-1.0.7.
856 * NFS export disabled when stack size < 8192. Since the NFSv4 export of
857 Lustre filesystem with 4K stack may cause a stack overflow. For more
858 information, please refer to bugzilla 17630.
860 Severity : enhancement
862 Description: Caching OSS
863 Details : introduce data caching on the OSS. The OSS now relies on the linux
864 kernel page cache to keep recently accessed data in memory.
865 It is worth noting that all write requests are still flushed
866 synchronously as in lustre 1.6.
868 Severity : enhancement
870 Description: Version based recovery
871 Details : introduce finer grained recovery able to detect transaction
872 dependencies and can deal with transaction gaps caused by clients
873 failing at the same time as the server.
875 Severity : enhancement
877 Description: Enable adaptive timeouts by default
878 Details : The Lustre timeout value in /proc/sys/lustre/timeout is now
879 managed dynamically based on server load and should not need
880 to be tuned manually based on cluster size. This allows Lustre
881 to work under a wider variety of system sizes and loads, without
882 unnecessarily causing lengthy recovery times.
884 Severity : enhancement
886 Description: Add OST Pools support
887 Details : File striping can now be set to use an arbitrary pool of OSTs
889 Severity : enhancement
891 Description: add lazystatfs mount option to allow statfs(2) to skip down OSTs
892 Details : allow skip disconnected ost for send statfs request and hide error
896 Frequency : rare, on llog test 6
898 Descriptoin: don't allow connect to already connected import
899 Details : allowing connect to already connected import is hide connecting problem.
902 Frequency : rare, on failed llog setup
904 Descriptoin: don't leak obd reference on failed llog setup
905 Details : for failed llog setup - mgc forget call class_destroy_import for
906 client import, move destroy import to more generic place.
911 Descriptoin: allow kill process which wait statahead result
912 Details : for some reasons 'ls' can stick in waiting result from statahead,
913 in this case need way for kill this process.
916 Frequency : rare, at shutdown
918 Descriptoin: panic at umount
919 Details : llap_shrinker can be raced with killing super block from list and
920 this produce panic with access to already freeded pointer
925 Descriptoin: don't lose wakeup for imp_recovery_waitq
926 Details : recover_import_no_retry or invalidate_import and import_close can
927 both sleep on imp_recovery_waitq, but we was send only one wakeup
933 Descriptoin: panic in mds_open
934 Details : don't confuse mds_finish_transno() with PTR_ERR(-ENOENT)
939 Descriptoin: stuck in cache_remove_extent() or panic with accessing to already
941 Details : release lock refernce only after add page to pages list.
944 Frequency : always with long access acl
946 Descriptoin: mds can't pack reply with long acl.
947 Details : mds don't control size of acl but they limited by reint/getattr
952 Frequency : on remount
954 Description: external journal device not working after the remount
955 Details : clear dev_rdonly flag for external journal devices in
961 Description: shutdown vs evict race
962 Details : client_disconnect_export vs connect request race.
963 if client will evicted at this time - we start invalidate
964 thread without referece to import and import can be freed
970 Description: shrink LOV EAs before replying
971 Details : correctly adjust LOV EA buffer for reply.
976 Description: don't skip ost target if they assigned to file
977 Details : Drop slow OSCs if we can, but not for requested start idx.
978 This means "if OSC is slow and it is not the requested
979 start OST, then it can be skipped, otherwise skip it only
980 if it is inactive/recovering/out-of-space.
982 Severity : enhancement
984 Description: Update to RHEL5 kernel-2.6.18-92.1.17.el5.
987 Frequency : rare, need acl's on inode.
989 Description: client can't handle ost additional correctly
990 Details : if ost was added after client connected to mds client can have
991 hit lnet_try_match_md ... to big messages to wide striped files.
992 in this case need teach client to handle config events about add
993 lov target and update client max ea size at that event.
995 Severity : enhancement
997 Description: Update to sles9 kernel-2.6.5-7.314.
999 Severity : enhancement
1001 Description: Update to SLES10 SP2 kernel-2.6.16.60-0.31.
1004 Frequency : Create a symlink file with a very long name
1006 Description: ldlm_cancel_pack()) ASSERTION(max >= dlm->lock_count + count)
1007 Details : If there is no extra space in the request for early cancels,
1008 ldlm_req_handles_avail() returns 0 instead of a negative value.
1013 Description: mds is deadlocked
1014 Details : in rare cases, inode in catalog can have i_no less than have parent
1015 i_no, this produce wrong order for locking during open, and
1016 parallel unlink can be lock open. this need teach mds_open to grab
1017 locks in resource id order, not at parent -> child order.
1019 Severity : enhancement
1021 Description: Add /proc entry for import status
1022 Details : The mdc, osc, and mgc import directories now have
1023 an import directory that contains useful import data for debugging
1024 connection problems.
1026 Severity : enhancement
1028 Description: Re-disable certain /proc logging
1029 Details : Enable and disable client's offset_stats, extents_stats and
1030 extents_stats_per_process stats logging on the fly.
1033 Frequency : Only on FC kernels 2.6.22+
1035 Description: oops in statahead
1036 Details : Do not drop reference count for the dentry from VFS when lookup,
1037 VFS will do that by itself.
1039 Severity : enhancement
1041 Description: Generic /proc file permissions
1042 Details : Set /Proc file permissions in a more generic way to enable non-
1043 root users operate on some /proc files.
1047 Description: Hitting mdc_commit_close() ASSERTION
1048 Details : Properly handle request reference release in
1049 ll_release_openhandle().
1053 Frequency : only patchless client
1054 Description: add workaround for race between add/remove dentry from hash
1056 Severity : enhancement
1058 Description: Allow OST glimpses to return PW locks
1062 Description: LBUG when llog conf file is full
1063 Details : When llog bitmap is full, ENOSPC should be returned for plain
1068 Description: Prevent import from entering FULL state when server in recovery
1072 Description: service mount cannot take device name with ":"
1073 Details : Only when device name contains ":/" will mount treat it as
1079 Description: replace ptlrpcd with the statahead thread to interpret the async
1080 statahead RPC callback
1084 Frequency : on recovery
1085 Description: I/O failures after umount during fail back
1086 Details : if client reconnected to restarted server we need join to recovery
1087 instead of find server handler is changed and process self eviction
1088 with cancel all locks.
1090 Severity : enhancement
1092 Description: Update to RHEL5 kernel-2.6.18-92.1.10.el5.
1096 Description: Kernel BUG tries to release flock
1097 Details : Lustre does not destroy flock lock before last reference goes
1098 away. So always drop flock locks when client is evicted and
1099 perform unlock regardless of successfulness of speaking to MDS.
1101 Severity : enhancement
1103 Description: Update to SLES10 SP2 kernel-2.6.16.60-0.27.
1105 Severity : enhancement
1107 Description: Upcall on Lustre log has been dumped
1108 Details : Allow for a user mode script to be called once a Lustre log has
1109 been dumped. It passes the filename of the dumped log to the
1110 script, the location of the script can be specified via
1111 /proc/sys/lnet/debug_log_upcall.
1116 Description: avoid messages about idr_remove called for id that is not allocated
1117 Details : Move assigment s_dev for clustered nfs to end of initialization,
1118 for avoid problem with error handling.
1123 Description: avoid Already found the key in hash [CONN_UNUSED_HASH] messages
1124 Details : When connection is reused this not moved from CONN_UNUSED_HASH into
1125 CONN_USED_HASH and this prodice warning when put connection again
1131 Description: avoid ASSERTION(client_stat->nid_exp_ref_count == 0) failed
1132 Details : release reference to stats when client disconnected, not
1133 when export destroyed for avoid races when client destroyed
1134 after main ost export.
1138 Description: more cleanup in mds_lov
1139 Details : add workaround for get valid ost count for avoid warnings about
1140 drop too big messages, not init llog cat under semphore which
1141 can be blocked on reconnect and break normal replay, fix access
1144 Severity : enhancement
1146 Description: Export bytes_read/bytes_write count on OSC/OST.
1150 Description: Early reply size mismatch, MGC loses connection
1151 Details : Apply the MGS_CONNECT_SUPPORTED mask at reconnect time so
1152 the connect flags are properly negotiated.
1156 Description: Properly propagate oinfo flags from lov to osc for statfs
1157 Details : restore missing copy oi_flags to lov requests.
1161 Description: exports in /proc are broken
1162 Details : recreate /proc entries for clients when they reconnect.
1164 Severity : enhancement
1166 Description: Add man pages for llobdstat(8), llstat(8), plot-llstat(8),
1167 : l_getgroups(8), lst(8), routerstat(8)
1168 Details : included man pages for llobdstat(8), llstat(8),
1169 : plot-llstat(8), l_getgroups(8), lst(8), routerstat(8)
1171 Severity : enhancement
1173 Description: Implement lustre ll_show_options method.
1175 Severity : enhancement
1177 Description: Update to SLES9 kernel-2.6.5-7.312.
1179 Severity : enhancement
1181 Description: Update to RHEL4 kernel-2.6.9-67.0.22.EL.
1185 Description: exports in /proc are broken
1186 Details : recreate /proc entries for clients when they reconnect.
1190 Description: don't fail open with -ERANGE
1191 Details : if client connected until mds will be know about real ost count
1192 get LOV EA can be fail because mds not allocate enougth buffer
1197 Description: Resolve device initialization race
1198 Details : Prevent proc handler from accessing devices added to the
1199 obd_devs array but yet be intialized.
1201 Severity : enhancement
1203 Description: configure's --enable-quota should check the
1204 : kernel .config for CONFIG_QUOTA
1205 Details : configure is terminated if --enable-quota is passed but
1206 : no quota support is in kernel
1208 Severity : enhancement
1210 Description: Update to SLES10 SP2 kernel-2.6.16.60-0.23.
1212 Severity : enhancement
1214 Description: Update to RHEL5 kernel-2.6.18-92.1.6.el5.
1218 Frequency : rare, on PPC clients
1219 Description: don't swab ost objects in response about directory, because
1221 Details : bug similar bug 14856, but in different function.
1223 Severity : enhancement
1225 Description: lfs quota tool enhancement
1226 Details : added units specifiers support for setquota, default to
1227 current uid/gid for quota report, short quota stats by
1228 default, nonpositional parameters for setquota, added
1229 llapi_quotactl manual page.
1231 Severity : enhancement
1233 Description: *optional* service tags registration
1234 Details : if the "service tags" package is installed on a Lustre node
1235 When the filesystem is mounted, a local-node service tag will
1236 be created. See http://inventory.sun.com/ for more information
1237 about the Service Tags asset management system.
1241 Description: Client runs out of low memory
1242 Details : Consider only lowmem when counting initial number of llap pages
1245 Frequency : occasional
1247 Description: add refcount for osc callbacks, so avoid panic on shutdown
1249 Severity : enhancement
1251 Description: Update to RHEL4 kernel-2.6.9-67.0.20.
1254 Frequency : testing only
1256 Description: sanity test 65a fails if stripecount of -1 is set
1257 Details : handle -1 striping on filesystem in ll_dirstripe_verify
1260 Frequency : only in unusual configurations
1262 Description: Kernel panic with find ost index.
1263 Details : lov_obd have panic if some OST's have sparse indexes.
1265 Severity : enhancement
1267 Description: Update to RHEL5 kernel-2.6.18-53.1.21.el5.
1270 Frequency : rarely, if filesystem is mounted with -o flock
1272 Description: do not process already freed flock
1273 Details : flock can possibly be freed by another thread before it reaches
1274 to ldlm_flock_completion_ast.
1277 Frequency : rarely, if filesystem is mounted with -o flock
1279 Description: LBUG during stress test
1280 Details : Need properly lock accesses the flock deadlock detection list.
1283 Frequency : rarely, if binaries are being run from Lustre
1285 Description: oops in page fault handler
1286 Details : kernel page fault handler can return two special 'pages' in
1287 error case, don't try dereference NOPAGE_SIGBUS and NOPAGE_OMM.
1290 Frequency : rarely, during shutdown
1292 Description: timeout with invalidate import.
1293 Details : ptlrpcd_check call obd_zombie_impexp_cull and wait request which
1294 should be handled by ptlrpcd. This produce long age waiting and
1295 -ETIMEOUT ptlrpc_invalidate_import and as result LASSERT.
1301 Description: ASSERTION(CheckWriteback(page,cmd)) failed
1302 Details : badly clear PG_Writeback bit in ll_ap_completion can produce false
1306 Frequency : only with broken builds/installations
1308 Description: no LBUG if lquota.ko and fsfilt_ldiskfs.ko are different versions
1309 Details : just return an error to a user, put a console error message
1311 Severity : enhancement
1313 Description: Update to RHEL5 kernel-2.6.18-53.1.19.el5.
1315 Severity : enhancement
1317 Description: Update to RHEL4 kernel-2.6.9-67.0.15.
1319 Severity : enhancement
1321 Description: enable MGS and MDT services start separately
1322 Details : add a 'nomgs' option in mount.lustre to enable start a MDT with
1323 a co-located MGS without starting the MGS, which is a complement
1324 to 'nosvc' mount option.
1327 Frequency : always, on big-endian systems
1329 Description: cleanup in ptlrpc code, related to PPC platform
1330 Details : store magic in native order avoid panic's in recovery on PPC
1331 node and forbid from this error in future. Also fix posibily
1332 of twice swab data. Fix get lov striping to userpace.
1335 Frequency : rarely, if replay get lost on server
1337 Description: server incorrectly drop resent replays lead to recovery failure.
1338 Details : do not drop replay according to msg flags, instead we check the
1339 per-export recovery request queue for duplication of transno.
1342 Frequency : after recovery
1344 Description: precreate to many object's after del orphan.
1345 Details : del orphan st in oscc last_id == next_id and this triger growing
1346 count of precreated objects. Set flag LOW to skip increase count
1347 of precreated objects.
1350 Frequency : rare, on clear nid stats
1352 Description: ASSERTION(client_stat->nid_exp_ref_count == 0)
1353 Details : when clean nid stats sometimes try destroy live entry,
1354 and this produce panic in free.
1357 Frequency : occasionally since 1.6.4
1359 Description: Stack overflow during MDS log replay
1360 Details : ease stack pressure by using a thread dealing llog_process.
1363 Frequency : very rare
1365 Description: MDT cannot be unmounted, reporting "Mount still busy"
1366 Details : Mountpoint references were being leaked during open reply
1367 reconstruction after an MDS restart. Drop mountpoint reference
1368 in reconstruct_open() and free dentry reference also.
1373 Description: wait until IO finished before start new when do lock cancel.
1374 Details : VM protocol want old IO finished before start new, in this case
1375 need wait until PG_writeback is cleared until check dirty flag
1376 and call writepages in lock cancel callback.
1381 Description: mds_mfd_close() ASSERTION(rc == 0)
1382 Details : In mds_mfd_close(), we need protect inode's writecount change
1383 within its orphan write semaphore to prevent possible races.
1386 Frequency : rare, on shutdown ost
1388 Description: don't hit live lock with umount ost.
1389 Details : shrink_dcache_parent can be in long loop with destroy dentries,
1390 use shrink_dcache_sb instead.
1393 Frequency : only when echo_client is used
1395 Description: don't panic with use echo_client
1396 Details : echo client pass NULL as client nid pointer and this produce NULL
1397 pointer dereference.
1400 Frequency : Always on 32-bit PowerPC systems
1402 Description: fix build on PPC32
1403 Details : compile code with -m64 flag produce wrong object file for PPC32.
1408 Description: MDS LBUG: ASSERTION(!IS_ERR(dchild))
1409 Details : In reconstruct_* functions, LASSERTs on both the data supplied
1410 by a client, and the data on disk are dangerous and incorrect.
1411 Change them with client eviction.
1413 Severity : enhancement
1415 Description: skiplist implementation simplification
1416 Details : skiplists are used to group compatible locks on granted list
1417 that was implemented as tracking first and last lock of each lock
1418 group the patch changes that to using doubly linked lists
1422 Description: delete compatibility for 32bit qdata
1423 Details : as planned, when lustre is beyond b1_8, lquota won't support 32bit
1424 qunit. That means servers of b1_4 and servers of b1_8 can't be
1425 used together if users want to use quota.
1428 Frequency : only with administrator action
1430 Description: mount failure if config log has invalid conf_param setting
1431 Details : If administrator specified an incorrect configuration parameter
1432 with "lctl conf_param" this would cause an error during future
1433 client mounts. Instead, ignore the bad configuration parameter.
1436 Frequency : blocks per group < blocksize*8 and uninit_groups is enabled
1438 Description: ldiskfs error: XXX blocks in bitmap, YYY in gd
1439 Details : If blocks per group is less than blocksize*8, set rest of the
1443 Frequency : Application do stride read on lustre
1445 Description: The read performance will drop a lot if the application does
1447 Details : Because the stride_start_offset are missing in stride read-ahead,
1448 it will cause clients read a lot of unused pages in read-ahead,
1449 then the read-performance drops.
1453 Description: more ldlm soft lockups
1454 Details : In ldlm_resource_add_lock(), call to ldlm_resource_dump()
1455 starve other threads from the resource lock for a long time in
1456 case of long waiting queue, so change the debug level from
1457 D_OTHER to the less frequently used D_INFO.
1459 Severity : enhancement
1461 Description: add -gid, -group, -uid, -user options to lfs find
1463 Severity : enhancement
1465 Description: ll_recover_lost_found_objs - recover objects in lost+found
1466 Details : OST corruption and subsequent e2fsck can leave objects in the
1467 lost+found directory. Using the "ll_recover_lost_found_objs"
1468 tool, these objects can be retrieved and data can be salvaged
1469 by using the object ID saved in the fid EA on each object.
1474 Description: this bug _only_ happens when inode quota limitation is very low
1475 (less than 12), so that inode quota unit is 1 at initialization.
1476 Details : if remaining quota equates 1, it is a sign to demonstate that quota
1477 is effective now. So least quota qunit should be 2.
1481 Description: Hung threads in invalidate_inode_pages2_range
1482 Details : The direct IO path doesn't call check_rpcs to submit a new RPC once
1483 one is completed. As a result, some RPCs are stuck in the queue
1488 Description: Procfs and llog threads access destoryed import sometimes.
1489 Details : Sync the import destoryed process with procfs and llog threads by
1490 the import refcount and semaphore.
1494 Description: mds fails to respond, threads stuck in ldlm_completion_ast
1495 Details : Sort source/child resource pair after updating child resource.
1500 Description: kernel BUG at ldiskfs2_ext_new_extent_cb
1501 Details : If insertion of an extent fails, then discard the inode
1502 preallocation and free data blocks else it can lead to duplicate
1507 Description: don't always update ctime in ext3_xattr_set_handle()
1508 Details : Current xattr code updates inode ctime in ext3_xattr_set_handle()
1509 In some cases the ctime should not be updated, for example for
1510 2.0->1.8 compatibility it is necessary to delete an xattr and it
1511 should not update the ctime.
1515 Description: add quota statistics
1516 Details : 1. sort out quota proc entries and proc code.
1517 2. add quota statistics
1522 Description: quotas are not honored with O_DIRECT
1523 Details : all writes with the flag O_DIRECT will use grants which leads to
1524 this problem. Now using OBD_BRW_SYNC to guard this.
1528 Bugzilla : 15713/16362
1529 Description: Assertion in iopen_connect_dentry in 1.6.3
1530 Details : looking up an inode via iopen with the wrong generation number can
1531 populate the dcache with a disconneced dentry while the inode
1532 number is in the process of being reallocated. This causes an
1533 assertion failure in iopen since the inode's dentry list contains
1534 both a connected and disconnected dentry.
1538 Description: assertion failure in ldlm_handle2lock()
1539 Details : fix a race between class_handle_unhash() and class_handle2object()
1540 introduced in lustre 1.6.5 by bug 13622.
1542 Severity : enhancement
1544 Description: superblock lock contention with many SMP cores on one client
1545 Details : several client filesystem locks were highly contended on SMP
1546 NUMA systems with 8 or more cores. Per-CPU datastructures
1547 and more efficient locking implemented to reduce contention.
1552 Description: Kernel BUG: sd_iostats_bump: unexpected disk index
1553 Details : remove the limit of 256 scsi disks in the sd_iostat patch
1558 Description: oops in sd_iostats_seq_show()
1559 Details : unloading/reloading the scsi low level driver triggers a kernel
1560 bug when trying to access the sd iostat file.
1565 Description: Kernel panics during QLogic driver reload
1566 Details : REQ_BLOCK_PC requests are not handled properly in the sd iostat
1567 patch, causing memory corruption.
1572 Description: journal_dev option does not work in b1_6
1573 Details : pass mount option during pre-mount.
1575 Severity : enhancement
1577 Description: Add a FIEMAP(FIle Extent MAP) ioctl for ldiskfs
1578 Details : FIEMAP ioctl will allow an application to efficiently fetch the
1579 extent information of a file. It can be used to map logical blocks
1580 in a file to physical blocks in the block device.
1583 Frequency : only with adaptive timeout enabled
1585 Description: DEBUG_REQ() bad paging request
1586 Details : ptlrpc_at_recv_early_reply() should not modify req->rq_repmsg
1587 because it can be accessed by reply_in_callback() without the
1591 Frequency : only on Cray X2
1593 Description: X2 build failures
1594 Details : fix build failures on Cray X2.
1598 Description: xid & resent requests
1599 Details : Initialize RPC XID from clock at startup (randomly if clock is
1604 Description: quota recovery deadlock during mds failover
1605 Details : This patch includes att18982, att18236, att18237 in bz14840.
1607 1. fix osts hang when mds does failover with quotaon
1608 2. prevent watchdog storm when osts threads wait for the
1613 Description: kernel panic on racer
1614 Details : Do not access dchild->d_inode when IS_ERR(dchild) is true.
1616 Severity : enhancement
1618 Description: Add lustre_start utility to start or stop multiple Lustre servers
1623 Description: Lustre GPF in {:ptlrpc:ptlrpc_server_free_request+373}
1624 Details : In case of memory pressure, list_del() can be called twice on
1625 req->rq_history_list, causing a kernel oops.
1629 Description: (ptllnd_peer.c:557:kptllnd_peer_check_sends()) ASSERTION(!in_interrupt()) failed
1630 Details : fix stack overflow in the distributed lock manager by defering
1631 export eviction after a failed ast to the elt thread instead of
1632 handling it in the dlm interpret routine.
1634 Severity : enhancement
1636 Description: More exported tunables for mballoc
1637 Details : Add support for tunable preallocation window and new tunables for
1638 large/small requests
1642 Description: Detect corruption of block bitmap and checking for preallocations
1643 Details : Checks validity of on-disk block bitmap. Also it does better
1644 checking of number of applied preallocations. When corruption is
1645 found, it turns filesystem readonly to prevent further corruptions.
1649 Frequency : only for big-endian servers
1650 Description: Check for big-endian system while mounting fs with extents feature
1651 Details : Mounting a filesystem with extents feature will fail on big-endian
1652 systems since ext3-based ldiskfs is not supported on big-endian
1653 systems. Can be over-ridden with "bigendian_extents" mount option.
1657 Description: Excessive recovery window
1658 Details : With AT enabled, the recovery window can be excessively long (6000+
1659 seconds). To address this problem, we no longer use
1660 OBD_RECOVERY_FACTOR when extending the recovery window (the connect
1661 timeout no longer depends on the service time, it is set to
1662 INITIAL_CONNECT_TIMEOUT now) and clients report the old service
1663 time via pb_service_time.
1666 Descriptoin: Don't sync journal after every i/o
1667 Details : Implement write RPC replay to allow server replies for write RPCs
1668 before data is on disk. However, this feature is disabled by
1669 default since some issues leading to data corruptions have been
1670 found during recovery (e.g. bug 19128). This feature can be enabled
1671 by running the following command on the OSSs:
1672 lctl set_param obdfilter.*.sync_journal=0
1676 Description: Watchdog triggered on MDS failover
1677 Details : enable OBD_CONNECT_MDT flag when connecting from the MDS so that
1678 the OSTs know that the MDS "UUID" can be reused for the same export
1679 from a different NID, so we do not need to wait for the export to
1684 Descriptoin: Lustre detected file system corruption with inode out of bounds
1685 Details : don't update i_size on MDS_CLOSE for directories. This causes
1686 directory corruptions on the MDT.
1688 -------------------------------------------------------------------------------
1690 2008-05-26 Sun Microsystems, Inc.
1692 * Support for kernels:
1693 2.6.5-7.311 (SLES 9),
1694 2.6.9-67.0.7.EL (RHEL 4),
1695 2.6.16.54-0.2.5 (SLES 10),
1696 2.6.18-53.1.14.el5 (RHEL 5),
1697 2.6.22.14 vanilla (kernel.org)
1698 * Client support for unpatched kernels:
1699 (see http://wiki.lustre.org/index.php?title=Patchless_Client)
1700 2.6.16 - 2.6.22 vanilla (kernel.org)
1701 * Due to problems with nested symlinks and FMODE_EXEC (bug 12652),
1702 we do not recommend using patchless RHEL4 clients with kernels
1703 prior to 2.6.9-55EL (RHEL4U5).
1704 * Recommended e2fsprogs version: 1.40.7-sun1
1705 * Note that reiserfs quotas are disabled on SLES 10 in this kernel.
1706 * RHEL 4 and RHEL 5/SLES 10 clients behaves differently on 'cd' to a
1707 removed cwd "./" (refer to Bugzilla 14399).
1708 * A new quota file format has been introduced in 1.6.5.
1709 The format conversion from prior releases is handled transparently,
1710 but releases older than 1.4.12/1.6.5 will not understand this new
1711 format. The automatic format conversion can be avoided by running
1712 the following command on the MDS before upgrading:
1713 'tunefs.lustre --param="mdt.quota_type=ug1" $MDTDEV'.
1714 For more information, please refer to bugzilla 13904.
1718 Description: quota performance fix
1719 Details : quota data is written in journalled mode instead of ordered to
1720 increase performance
1724 Description: lfs support for human-readable quota grace time strings
1725 Details : lfs setquota -t and lfs quota -t represent quota grace times
1726 in "XXwXXdXXhXXmXXs" format instead of large values in seconds
1729 Frequency : always with o2ib 1.3 and sles10
1731 Description: fix build with SLES10 and o2ib v3.
1732 Details : sles10 uses diffrent name for Module.symver file but configure
1733 assume this file has same name on RHEL/SLES/vanila kernels.
1736 Frequency : very rare, if additional xattrs are used on kernels >= 2.6.12
1738 Description: MDS may lose file striping (and hence file data) in some cases
1739 Details : If there are additional extended attributes stored on the MDS,
1740 in particular ACLs, SELinux, or user attributes (if user_xattr
1741 is specified for the client mount options) then there is a risk
1742 of attribute loss. Additionally, the Lustre file striping
1743 needs to be larger than default (e.g. striped over all OSTs),
1744 and an additional attribute must be stored initially in the
1745 inode and then increase in size enough to be moved to the
1746 external attribute block (e.g. ACL growing in size) for file
1749 Severity : enhancement
1751 Description: add message levels for liblustreapi
1756 Description: MDT cannot be unmounted, reporting "Mount still busy"
1757 Details : Mountpoint references were being leaked during open reply
1758 reconstruction after an MDS restart. Drop mountpoint reference
1759 in reconstruct_open() and free dentry reference also.
1764 Description: fix for occasional failure case of -ENOSPC in recovery-small tests
1765 Details : Move the 'good_osts' check before the 'total_bavail' check. This
1766 will result in an -EAGAIN and in the exit call path we call
1767 alloc_rr() which will with increasing aggressiveness attempt to
1768 aquire precreated objects on the minimum number of required OSCs.
1772 Description: Use old size assignment to avoid deadlock
1773 Details : This reverts the changes in bugs 2369 and bug 14138 that introduced
1774 the scheduling while holding a spinlock. We do not need locking
1775 for size in ll_update_inode() because size is only updated from
1776 the MDS for directories or files without objects, so there is no
1777 other place to do the update, and concurrent access to such inodes
1778 are protected by the inode lock.
1782 Description: Use __u64 instead of int for valid bits
1786 Description: resolve "_IOWR redefined" build error on SLES10
1790 Description: dump the memory debugging after all modules are unloaded to
1791 suppress false negative in conf_sanity test 39
1795 Description: the recovery timer never expires
1796 Details : for new client connect request, the recovery timer should not be
1797 reset, otherwise recovery timer will never expired, if the old
1798 client never come. Only old client connect and first connection
1799 req should trigger recovery timer reset.
1803 Description: the min numbers of lproc stats are wrong
1804 Details : adding a new constant LC_MIN_INIT and use it for initialization
1808 Frequency : always with interactive lfs
1810 Description: Reinitialize optind to 0 so that interactive lfs works in all cases
1813 Frequency : with multiple concurrent readdir processes in same directory
1814 Bugzilla : 15406, 15169, 15175
1815 Description: misc fixes for directory readahead.
1816 Details : prevent previous statahead async RPC callback from processing the
1817 current "statahead_info", race condition between async RPC callback
1818 add dentry into dentry hash table and "ls" thread revalidate such
1819 dentry, statahead his/miss control for hidden items, and so on.
1821 Severity : enhancement
1823 Description: build kernel-ib packages for OFED 1.3 in our release cycle
1827 Description: incore types cleaning in quota code (with respect to 64-bit limits)
1828 Details : several u32 variables declarations are replaced with u64 declarations
1833 Description: fix SLES kernel versioning
1834 Details : the kernel version for our SLES 10 kernel did not include a "-"
1835 before the "smp" at the end. while this was not a problem in
1836 general, it did mean that software trying to use the kernel
1837 version to try to detect a vendor specific kernel would fail.
1838 this was most evident by the OFED build scripts.
1843 Description: Don't update lov_desc members until making sure they are valid
1844 Details : When updating lov_desc members via proc fs, need fix their
1845 validities before doing the real update.
1848 Frequency : very rare
1850 Description: don't put request into delay list while invalidate in flight.
1851 Details : ptlrpc_delay_request sometimes put in delay list while invalidate
1852 import in flight. this produce timeout for invalidate and sometimes
1853 can cause stale data.
1855 Severity : enhancement
1857 Description: Update kernel to SLES9 2.6.5-7.311.
1859 Severity : enhancement
1861 Description: Update kernel to RHEL4 2.6.9-67.0.7.
1866 Frequency : on PPC only
1867 Description: not convert ost objects for directory because it's not exist.
1868 Details : ll_dir_getstripe assume dirrectory has ost objects but this wrong.
1870 Severity : enhancement
1872 Description: Fix warnings with compile liblustre at sles10/rhel5 which have
1873 __u64 as usingied long long type.
1876 Frequency : rare, on shutdown
1878 Description: race process ast vs remove callback
1879 Details : removing callback before disconnect import open race with
1880 processing callback.
1882 Severity : enhancement
1884 Description: Update kernel to SLES9 2.6.5-7.311.
1886 Severity : enhancement
1888 Description: Files open for execute are not marked busy on SLES10
1889 Details : Add FMODE_EXEC to SLES10 SP1 server kernel series.
1891 Severity : enhancement
1893 Description: Add server support for vanilla-2.6.22.14.
1896 Frequency : occasional
1898 Description: Avoid lov_create() getting stuck in obd_statfs_rqset()
1899 Details : If an OST is down the MDS will hang indefinitely in
1900 obd_statfs_rqset() waiting for the statfs data. While for
1901 MDS QOS usage of statfs, it should not stuck in waiting.
1903 Severity : enhancement
1905 Description: Disable adaptive timeouts by default
1908 Frequency : on network error
1910 Description: panic with double free request if network error
1911 Details : mdc_finish_enqueue is finish request if any network error occuring,
1912 but it's true only for synchronus enqueue, for async enqueue
1913 (via ptlrpcd) this incorrect and ptlrpcd want finish request
1917 Frequency : rare, on recovery
1919 Description: read procfs can produce deadlock in some situation
1920 Details : Holding lprocfs lock with send rpc can produce block for destroy
1921 obd objects and this also block reconnect with -EALREADY.
1922 This isn't fix all lprocfs bugs - but make it rare.
1924 Severity : enhancement
1926 Description: Update kernel to RHEL5 2.6.18-53.1.14.el5.
1929 Frequency : frequent on X2 node
1931 Description: mdc_set_open_replay_data LBUG
1932 Details : Set replay data for requests that are eligible for replay.
1937 Description: lustre_mgs: operation 101 on unconnected MGS
1938 Details : When MGC is disconnected from MGS long enough, MGS will evict the
1939 MGC, and late on MGC cannot successfully connect to MGS and a lot
1940 of the error messages complaining that MGS is not connected.
1943 Frequency : rare, depends on device drivers and load
1945 Description: MDS or OSS nodes crash due to stack overflow
1946 Details : Code changes in 1.6.4 increased the stack usage of some functions.
1947 In some cases, in conjunction with device drivers that use a lot
1948 of stack, the MDS (or possibly OSS) service threads could overflow
1949 the stack. One change which was identified to consume additional
1950 stack has been reworked to avoid the extra stack usage.
1952 Severity : enhancement
1954 Description: Update to RHEL5 latest kernel-2.6.18-53.1.13.el5.
1956 Severity : enhancement
1958 Description: Update to SLES10 SP1 latest kernel-2.6.16.54-0.2.5.
1960 Severity : enhancement
1962 Description: Update to RHEL5 latest kernel-2.6.18-53.1.6.el5.
1964 Serverity : enhancement
1966 Description: Update RHEL4 kernel to 2.6.9-67.0.4.
1969 Frequency : rare on shutdown OST
1971 Description: Don't allow skipping OSTs if index has been specified.
1972 Details : Don't allow skipping OSTs if index has been specified, make locking
1973 in internal create lots better.
1978 Description: ASSERTION(!PageDirty(page)) failed
1979 Details : Wrong check could lead to an assertion failure under specific
1985 Description: LBUG in ptlrpc_check_set() bad phase ebc0de00
1986 Details : access to bitfield in structure is always rounded to long
1987 and this produce problem with not atomic change any bit.
1992 Description: Lustre make rpms failed.
1993 Details : Remove ldiskfs spec file to avoids rpmbuild be confused when
1994 builds Lustre rpms from tarball.
1996 Severity : enhancement
1998 Description: Update to SLES9 SP4 kernel-2.6.5-7.308.
2001 Frequency : rare on shutdown OST
2003 Description: If llog cancel was not send before clean_exports phase, this can
2004 produce deadlock in llog code.
2005 Details : If llog thread has last reference to obd and call class_import_put
2006 this produce deadlock because llog_cleanup_commit_master wait when
2007 last llog_commit_thread exited, but this never success because was
2008 called from llog_commit_thread.
2011 Frequency : only if OST index is skipped
2013 Description: NULL lov_tgts causing MDS oops
2014 Details : more safe checks for NULL lov_tgts for avoid oops.
2016 Severity : enhancement
2018 Description: Update to RHEL4 latest kernel-2.6.9-67.0.1.EL.
2020 Severity : enhancement
2022 Description: Update to RHEL5 latest kernel-2.6.18-53.1.4.el5.
2027 Description: make mgs_setparam() handle fsname containing dash
2028 Details : fsname containing a dash does not work with lctl conf_param
2030 Severity : enhancement
2032 Description: Update to RHEL4 Update-6 kernel-2.6.9-67.EL.
2035 Frequency : rare, in recovery and (or) destroy lovobjid file.
2037 Description: rewrite lov objid code.
2038 Details : Cleanup for lov objid code, remove scability problems and wrong
2039 locking. Fix sending last_id into ost.
2041 Severity : enhancement
2043 Description: Update to SLES10 SP1 latest kernel-2.6.16.54-0.2.3.
2045 Severity : enhancement
2047 Description: Update to RHEL5 Update-1 kernel 2.6.18-53.el5.
2048 Details : Use d_move_locked instead of __d_move.
2051 Frequency : rare, at shutdown
2053 Description: access already free / zero obd_namespace.
2054 Details : if client_disconnect_export was called without force flag set,
2055 and exist connect request in flight, this can produce access to
2056 NULL pointer (or already free pointer) when connect_interpret
2057 store ocd flags in obd_namespace.
2060 Frequency : only at startup
2062 Description: not alloc memory with spinlock held.
2063 Details : allocation memory with GFP_KERNEL can produce sleep deadlock,
2064 if any spinlock held.
2069 Description: lfs find does not continue on file error
2070 Details : Continue other files processing when a file/dir is absent.
2074 Description: Inconsistent usage of lustre_pack_reply()
2075 Details : Standardize the usage of lustre_pack_reply() such that it
2076 always generate a CERROR on failure.
2079 Frequency : very rare
2081 Description: Fix replay if there is an un-replied request and open
2082 Details : In some cases, older replay request will revert the
2083 mcd->mcd_last_xid on MDS which is used to record the client's
2084 latest sent request.
2086 Severity : enhancement
2088 Description: Update to RHEL5 kernel 2.6.18-8.1.15.el5.
2090 Severity : enhancement
2092 Description: Update to SLES10 SP1 kernel 2.6.16.53-0.16
2094 Severity : enhancement
2096 Description: Update to SLES9 kernel-2.6.5-7.287.3.
2098 Severity : enhancement
2100 Description: Update to RHEL4 kernel-2.6.9-55.0.12.EL.
2102 Severity : enhancement
2104 Description: Build SLES10 patchless client fails
2105 Details : The configure was broken by run ./configure with
2106 --with-linux-obj=.... argument for patchless client. When the
2107 configure use --with-linux-obj, the LINUXINCLUDE= -Iinclude
2108 can't search header adequately. Use absolute path such as
2109 -I($LINUX)/include instead.
2111 Severity : enhancement
2113 Description: Lustre Page Accounting
2114 Details : New macros for page alloc and free which enable accounting
2115 of page allocation of Lustre. Use percpu counters to store memory
2116 and page statistics.
2119 Frequency : only if debugging is disabled
2121 Description: LASSERT_{REQ,REP}SWAB macros are buggy
2122 Details : If SWAB_PARANOIA is disabled, the LASSERT_REQSWAB and
2123 LASSERT_REPSWAB macros become no-ops, which is incorrect. Drop
2124 these macros and replace them with their definitions instead.
2129 Description: interrupt oig_wait produce painc on resend.
2130 Details : brw_redo_request can be used for resend requests from ptlrpcd and
2131 private set, and this produce situation when rq_ptlrpcd_data not
2132 copyed to new allocated request and triggered LBUG on assert
2133 req->rq_ptlrpcd_data != NULL. But this member used only for wakeup
2134 ptlrpcd set if request is changed and can be safety changed to use
2137 Severity : enhancement
2139 Description: organize the server-side client stats on per-nid basis
2140 Details : Change the structure of stats under obdfilter and mds to
2148 The "uuid"s file would list the uuids of _active_ exports.
2149 And the clear entry is to clear all stats and stale nids.
2154 Description: Processes looping in ll_readdir() on Lustre clients finally causing
2155 a full node pseudo-hang
2156 Details : Concurrent access to the same directory from multiple clients with
2157 intensive file creation/removal can cause a client node to spin in
2158 ll_readdir(). i_version must be increased every time the lock is
2159 cancelled to ensure a revalidate is done.
2164 Description: touch file failed when fs is not full
2165 Details : OST in recovery should not be discarded by MDS in alloc_qos(),
2166 otherwise we can get ENOSP while fs is not full.
2169 Frequency : only for Cray XT3
2170 Bugzilla : 12829/13455
2171 Description: Changing primary group doesn't change the group lustre assigns to
2173 Details : When CRAY_XT3 is defined, the fsgid supplied by the client is
2174 overridden with the primary group provided by the group upcall,
2175 whereas the supplied fsgid can be trusted if it is in the list of
2176 supplementary groups returned by the group upcall.
2178 Severity : enhancement
2180 Description: Root Squash Functionality
2181 Details : Implementation of NFS-like root squash capability. Specifically,
2182 don't allow someone with root access on a client node to be able
2183 to manipulate files owned by root on a server node.
2185 Severity : enhancement
2187 Description: Slow trucate/writes to huge files at high offsets.
2188 Details : Directly associate cached pages to lock that protect those pages,
2189 this allows us to quickly find what pages to write and remove
2190 once lock callback is received.
2195 Description: Too many locks accumulating on client during NFS usage
2196 Details : mds_open improperly used accmode to find out access mode to a
2197 file. Also mdc_intent_lock logic to find out if we already have
2198 lock similar to just received was flawed since introduction of
2199 skiplists - locks are now added to the front of the granted
2204 Description: Hit ASSERTION(obd->obd_stopping == 1) failed in some setup failed
2206 Details : In obd setup failure handler, obd_stopping will not necessarily to
2207 be 1, and obd_set_up should also be checked to make sure whether
2208 obd is completely setup.
2210 Severity : enhancement
2212 Description: Allow masking D_WARNING, D_ERROR messages from console
2213 Details : Console messages can now be disabled via lnet.printk.
2218 Description: User code with malformed file open parameter crashes client node
2219 Details : Before packing join_file req, all the related reference should be
2220 checked carefully in case some malformed flags cause fake join_file
2226 Description: shrink/enlarge qunit size when needed; fix the problem of coarse
2227 grain of quota doing harm to quota's accuracy
2228 Details : qunit size will be changed when quota limitation is too low/high;
2229 record the pending quota write in order to get more accureate
2230 quota; delete the patch for bug12588, which is unnecessary when
2231 this patch is landed. This bug also contains fixes for bug 14526,
2232 14299, 14601 and 13794.
2236 Description: LDLM_ENQUEUE races with LDLM_CP_CALLBACK
2237 Details : ldlm_completion_ast() assumes that a lock is granted when the req
2238 mode is equal to the granted mode. However, it should also check
2239 that LDLM_FL_CP_REQD is not set.
2243 Description: Heavy nfs access might result in deadlocks
2244 Details : After ELC code landed, it is now improper to enqueue any mds
2245 locks under och_sem, because enqueue might want to decide to
2246 cancel open locks for same inode we are holding och_sem for.
2250 Description: 35% write performance drop with ldiskfs2 when quotas are on
2251 Details : Enable ext3 journalled quota by default to improve performance
2252 when quotas are turned on.
2256 Description: Client eviction while running blogbench
2257 Details : A lot of unlink operations with concurrent I/O can lead to a
2258 deadlock causing evictions. To address the problem, the number of
2259 oustanding OST_DESTROY requests is now throttled to
2260 max_rpcs_in_flight per OSC and LDLM_FL_DISCARD_DATA blocking
2261 callbacks are processed in priority.
2264 Frequency : RHEL4 only
2266 Description: mkfs is very slow on IA64/RHEL4
2267 Details : A performance regression has been discovered in the MPT Fusion
2268 driver between versions 3.02.73rh and 3.02.99.00rh. As a
2269 consequence, we have downgraded the MPT Fusion driver in the RHEL4
2270 kernel from 3.02.99.00 to 3.02.73 until this problem is fixed.
2273 Frequency : PPC/PPC64 only
2275 Description: conflicts between asm-ppc64/types.h and lustre_types.h
2276 Details : fix duplicated definitions between asm-ppc64/types.h and
2277 lustre_types.h on PPC.
2280 Frequency : PPC/PPC64 only
2282 Description: asm-ppc/segment.h does not exist
2283 Details : fix compile issue on PPC.
2287 Description: data checksumming impacts single node performance
2288 Details : add support for several checksum algorithms. Currently, CRC32 and
2289 Adler-32 are supported. The checksum type can be changed on the fly
2290 through /proc/fs/lustre/osc/*/checksum_type.
2294 Description: use adler32 for page checksums
2295 Details : when available, use the Adler-32 algorithm instead of CRC32 for
2300 Description: better handle error messages in extents code
2302 Severity : enhancement
2304 Description: SNMP support enhancement
2305 Details : Adding total number of sampled request for an MDS node in snmp
2308 Severity : enhancement
2310 Description: Optimize ldlm waiting list processing for PR extent locks
2311 Details : When processing waiting list for read extent lock and meeting read
2312 lock that is same or wider to it that is not contended, skip
2313 processing rest of the list and immediatelly return current
2314 status of conflictness, since we are guaranteed there are no
2315 conflicting locks in the rest of the list.
2319 Description: Time out and refuse to reconnect
2320 Details : When the failover node is the primary node, it is possible
2321 to have two identical connections in imp_conn_list. We must
2322 compare not conn's pointers but NIDs, otherwise we
2323 can defeat connection throttling.
2327 Description: Client not clear own cache if answer to reconnect is lost.
2328 Details : Client gets evicted from server. Now client also thinks it is
2329 disconnected (or gets ENOTCONN on its operation) and decides to
2330 reconnect. Server receives reconnect message, but cannot find
2331 export. New export is created that is fully valid (new cookie!),
2332 but reply is lost and not reported to client. Client reconnects
2333 again and gets back a just-created connection, but it is not new
2334 so client thinks it was not evicted and does not do recovery.
2338 Description: Detect stride IO mode in read-ahead
2339 Details : When a client does stride read, read-ahead should detect that and
2340 read-ahead pages according to the detected stride pattern.
2344 Description: build for x2 fails
2345 Details : fix compile issue on Cray systems.
2347 Severity : enhancement
2349 Description: implement readv/writev APIs(aio_read/aio_writes in newer kernels)
2350 Details : This greatly improves speed of NFS writes on 2.6 kernels.
2353 Frequency : only on PPC/SLES10
2355 Description: "BITS_PER_LONG is not 32 or 64" in linux/idr.h
2356 Details : On SLES10/PPC, fs.h includes idr.h which requires BITS_PER_LONG to
2357 be defined. Add a hack in mkfs_lustre.c to work around this compile
2362 Description: LASSERT on MDS when client holding flock lock dies
2363 Details : ldlm pool logic depends on number of granted locks equal to
2364 number of released locks which is not true for flock locks, so
2365 just exclude such locks from consideration.
2369 Description: MDS deadlock with many ll_sync_lov threads and I/O stalled
2370 Details : Use fsfilt_sync() for both the whole filesystem sync and
2371 individual file sync to eliminate dangerous inode locking
2372 with I_LOCK that can lead to a deadlock.
2376 Description: Update an obsolete wirecheck.c generator
2377 Details : Update wirecheck.c/wirehdr.c and regenerate wiretest.c
2381 Description: Client can panic on open sometimes
2382 Details : It is possible that we try to free already freed request in
2383 ll_file_open in some error cases when we send request from
2388 Description: performance in 1.6.3
2389 Details : Force q->max_phys_segments to MAX_PHYS_SEGMENTS on SLES10 to be
2390 sure that 1MB requests are not fragmented by the block layer.
2394 Description: LDLM soft lockups - improvement
2395 Details : It is be possible to send the lock handle along with each read
2396 or write request because the client is already doing a lock match
2397 itself so there isn't any reason the OST should have to re-do that
2403 Description: lfs quota fails with deactivated OSTS
2404 Details : With this patch, three improvements are included:
2405 1. detete the softlimit in mds and osts when use "lfs quota".
2406 2. display the inaccurate data in the output of "lfs quota".
2407 3. try to get quota info when "lfs quota" is executed.
2412 Description: Extent locks not granted with no conflicts sometimes.
2413 Details : When race occurs in glimpse handler and nothing is returned,
2414 we do not reprocess the queue after lock cancel, and that leads
2415 to a stall until next activity on a resource
2418 Frequency : failover with quotaon
2420 Description: during mds failovers with quota on, OSTs got into deadlock state
2421 and causing dumpstack.
2422 Details : for every quota slave, at any time, there is only one quota req
2423 is sent to quota master for every uid/gid. Before that quota req
2424 returns, all the thread relative to the same uid/gid will wait.
2425 So if the quota req is lost because mds failovers or any other
2426 reasons, this bug will be hit. Now, dqacq_interpret() will handle
2427 quota reqs who time out.
2429 Severity : enhancement
2432 Description: when quota slave checks if quota is enough, there is an unnecessary
2434 Details : place this wait on necessary place instead of always waiting.
2436 --------------------------------------------------------------------------------
2438 2007-12-07 Cluster File Systems, Inc. <info@clusterfs.com>
2440 * Support for kernels:
2441 2.6.5-7.286 (SLES 9),
2442 2.6.9-55.0.9.EL (RHEL 4),
2443 2.6.16.53-0.8 (SLES 10),
2444 2.6.18-8.1.14.el5 (RHEL 5),
2445 2.6.18.8 vanilla (kernel.org)
2446 * Client support for unpatched kernels:
2447 (see http://wiki.lustre.org/index.php?title=Patchless_Client)
2448 2.6.16 - 2.6.22 vanilla (kernel.org)
2449 * Due to recently discovered recovery problems, we do not recommend
2450 using patchless RHEL 4 clients with this or any earlier release.
2451 * Recommended e2fsprogs version: 1.40.2-cfs1
2452 * Note that reiserfs quotas are disabled on SLES 10 in this kernel.
2455 Frequency : occasional
2457 Description: MDS hang or stay in waiting lock
2458 Details : If client receive lock with CBPENDING flag ldlm need send lock
2459 cancel as separate rpc, to avoid situation when cancel request
2460 can't processed due all i/o threads stay in wait lock.
2463 Frequency : occasional
2465 Description: Do not fail import if osc_interpret_create gets -EAGAIN
2466 Details : If osc_interpret_create got -EAGAIN it immediately exits and
2467 wakeup oscc_waitq. After wakeup oscc_wait_for_objects call
2468 oscc_has_objects and see OSC has no objests and call
2469 oscc_internal_create to resend create request.
2471 Severity : enhancement
2473 Description: Update kernel patches for SLES10 2.6.16.53-0.8.
2474 Details : Update which_patch & target file for SLES10 latest kernel.
2476 Severity : enhancement
2478 Description: add --type and --size parameters to lfs find
2479 Details : Enhance lfs find by adding filetype and filesize parameters. Also
2480 multiple OBDs can now be specified for the --obd option.
2482 Severity : enhancement
2484 Description: eliminate client locks in face of contention
2485 Details : file contention detection and lockless i/o implementation
2486 for contended files.
2488 Severity : enhancement
2490 Description: Remove client patches from SLES 10 kernel.
2491 Details : This causes SLES 10 clients to behave as patchless clients
2492 even on a Lustre-patched (server) kernel.
2494 Severity : enhancement
2496 Description: use i_size_read and i_size_write in 2.6 port
2497 Details : replace inode->i_size access with i_size_read/write()
2500 Frequency : when removing large files
2502 Description: scheduling issue during removal of large Lustre files
2503 Details : Don't take the BKL in fsfilt_ext3_setattr() for 2.6 kernels.
2504 It causes scheduling issues when removing large files (17TB in the
2510 Description: 1.4.11 Can't handle directories with stripe set and extended ACLs
2511 Details : Impossible (EPROTO is returned) to access a directory that has a
2512 non-default striping and ACLs.
2515 Frequency : only on ppc
2517 Description: /proc/fs/lustre/devices broken on ppc
2518 Details : The patch as applied to 1.6.2 doesn't look correct for all arches.
2519 We should make sure the type of 'index' is loff_t and then cast
2520 explicitly as needed below. Do not assign an explicitly cast
2524 Frequency : only for rhel5
2526 Description: Kernel patches update for RHEL5 2.6.18-8.1.10.el5.
2527 Details : Modify the target file & which_kernel.
2530 Frequency : if the uninit_groups feature is enabled on ldiskfs
2532 Description: e2fsck reports "invalid unused inodes count"
2533 Details : If a new ldiskfs filesystem is created with the "uninit_groups"
2534 feature and only a single inode is created in a group then the
2535 "bg_unused_inodes" count is incorrectly updated. Creating a
2536 second inode in that group would update it correctly.
2541 Description: buffer overruns could theoretically occur
2542 Details : llapi_semantic_traverse() modifies the "path" argument by
2543 appending values to the end of the origin string, and a buffer
2544 overrun may occur. Adding buffer overrun check in liblustreapi.
2546 Severity : enhancement
2548 Description: Add jbd statistics patch for RHEL5 and 2.6.18-vanilla.
2551 Frequency : only if filesystem is inconsistent
2553 Description: handle "serious error: objid * already exists" more gracefully
2554 Details : If LAST_ID value on disk is smaller than the objects existing in
2555 the O/0/d* directories, it indicates disk corruption and causes an
2556 LBUG(). If the object is 0-length, then we should use the existing
2557 object. This will help to avoid a full fsck in most cases.
2559 Severity : enhancement
2561 Description: Kernel patches update for RHEL4 2.6.9-55.0.6.
2562 Details : Modify vm-tunables-rhel4.patch.
2564 Severity : enhancement
2566 Description: Kernel config for 2.6.18-vanilla.
2567 Details : Modify targets/2.6-vanilla.target.in.
2568 Add config file kernel-2.6.18-2.6-vanilla-i686.config.
2569 Add config file kernel-2.6.18-2.6-vanilla-i686-smp.config.
2570 Add config file kernel-2.6.18-2.6-vanilla-x86_64.config.
2571 Add config file kernel-2.6.18-2.6-vanilla-x86_64-smp.config.
2574 Frequency : occasional
2576 Description: improve handling recoverable errors
2577 Details : If request processed with error which can be recoverable on server
2578 request should be resend, otherwise page released from cache and
2584 Description: Kernel patches update for RHEL5 2.6.18-8.1.14.el5.
2585 Details : Modify target file & which_patch.
2586 A flaw was found in the IA32 system call emulation provided
2587 on AMD64 and Intel 64 platforms. An improperly validated 64-bit
2588 value could be stored in the %RAX register, which could trigger an
2589 out-of-bounds system call table access. An untrusted local user
2590 could exploit this flaw to run code in the kernel
2591 (ie a root privilege escalation). (CVE-2007-4573).
2595 Description: change order of libsysio includes
2596 Details : '#include sysio.h' should always come before '#include xtio.h'
2598 Severity : enhancement
2600 Description: adapt the lustre_config script to support the upgrade case
2601 Details : Add "-u" option for lustre_config script to support upgrading 1.4
2602 server targets to 1.6 in parallel.
2607 Description: To avoid grant space > avaible space when the disk is almost
2608 full. Without this patch you might see the error "grant XXXX >
2609 available" or some LBUG about grant, when the disk is almost
2611 Details : In filter_check_grant, for non_grant cache write, we should
2612 check the left space by if (*left > ungranted + bytes), instead
2613 of (*left > ungranted), because only we are sure the left space
2614 is enough for another "bytes", then the ungrant space should be
2615 increase. In client, we should update cl_avail_grant only there is
2616 OBD_MD_FLGRANT in the reply.
2621 Description: Update RHEL 4 kernel to fix local root privilege escalation.
2622 Details : Update to the latest RHEL 4 kernel to fix the vulnerability
2623 described in CVE-2007-4573. This problem could allow untrusted
2624 local users to gain root access.
2627 Frequency : when using O_DIRECT and quotas
2629 Description: Incorrect file ownership on O_DIRECT output files
2630 Details : block usage reported by 'lfs quota' does not take into account
2631 files that have been written with O_DIRECT.
2635 Description: (rw.c:1323:ll_read_ahead_pages()) ASSERTION(page_idx > ria->ria_stoff) failed
2636 Details : Once the unmatched stride IO mode is detected, shrink the stride-ahead
2637 window to 0. If it does hit cache miss, and read-pattern is still
2638 stride-io mode, does not reset the stride window, but also does not
2639 increase the stride window length in this case.
2641 --------------------------------------------------------------------------------
2643 2007-09-27 Cluster File Systems, Inc. <info@clusterfs.com>
2645 * Support for kernels:
2646 2.6.5-7.286 (SLES 9),
2647 2.6.9-55.0.2.EL (RHEL 4),
2648 2.6.16.46-0.14 (SLES 10),
2649 2.6.18-8.1.8.el5 (RHEL 5),
2650 2.6.18.8 vanilla (kernel.org)
2651 * Client support for unpatched kernels:
2652 (see http://wiki.lustre.org/index.php?title=Patchless_Client)
2653 2.6.16 - 2.6.21 vanilla (kernel.org)
2654 * Due to recently discovered recovery problems, we do not recommend
2655 using patchless RHEL 4 clients with this or any earlier release.
2656 * Recommended e2fsprogs version: 1.40.2-cfs1
2657 * Note that reiserfs quotas are disabled on SLES 10 in this kernel.
2661 Description: Fix errors in lfs documentation
2662 Details : Fixes man pages
2664 Severity : enhancement
2666 Description: Adaptive timeouts
2667 Details : RPC timeouts adapt to changing server load and network
2668 conditions to reduce resend attempts and improve recovery time.
2670 Severity : enhancement
2672 Description: llapi_file_create() does not allow some changes
2673 Details : add llapi_file_open() that allows specifying the file creation
2674 mode and open flags, and also returns an open file handle.
2676 Severity : enhancement
2678 Description: df doesn't work properly if diskfs blocksize != 4K
2679 Details : Choose biggest blocksize of OST's as the LOV's blocksize.
2682 Frequency : always on directories with default striping set
2684 Description: lfs find on -1 stripe looping in lsm_lmm_verify_common()
2685 Details : Avoid lov_verify_lmm_common() on directory with -1 stripe count.
2687 Severity : enhancement
2689 Description: merge and cleanup kernel patches.
2690 Details : Remove mnt_lustre_list in vfs_intent-2.6-rhel4.patch.
2692 Severity : enhancement
2694 Description: RedHat Update kernel for RHEL5
2695 Details : Kernel config file for RHEL5.
2697 Severity : enhancement
2699 Description: OSS needs mutliple precreate threads
2700 Details : Add ability to start more than one create thread per OSS.
2703 Frequency : only with quota on the root user
2705 Description: mds_obd_create error creating tmp object
2706 Details : When the user sets quota on root, llog will be affected and can't
2707 create files and write files.
2710 Frequency : Always on ia64 patchless client, and possibly others.
2712 Description: Add EXPORT_SYMBOL check for node_to_cpumask symbol.
2713 Details : This allows the patchless client to be loaded on architectures
2714 without this export.
2719 Description: disorder of journal start and llog_add cause deadlock.
2720 Details : in llog_origin_connect, journal start should happen before
2721 llog_add keep the same order as other functions to avoid
2724 Severity : enhancement
2726 Description: RedHat Update kernel for RHEL5
2727 Details : Modify the kernel config file more closer RHEL5.
2729 Severity : enhancement
2731 Description: Build failure against Centos5 (RHEL5)
2732 Details : Define PAGE_SIZE when it isn't present.
2735 Frequency : occasionally when using NFS
2737 Description: "ll_intent_file_open()) lock enqueue: err: -13" with nfs
2738 Details : with NFS, the anon dentry's parent was set to itself in
2739 d_alloc_anon(), so in MDS, we use rec->ur_fid1 to find the
2740 corresponding dentry other than use rec->ur_name.
2742 Severity : enhancement
2744 Description: enable data checksumming by default
2745 Details : enable checksum by default, allow --disable-checksum
2746 configure option and "-o nochecksum" mount option. Checksums
2747 can also be disabled at runtime via $LPROC/osc/*/checksum_pages.
2750 Frequency : Occasionally with failover
2752 Description: Client eviction due to failover config
2753 Details : after a connection loss, the lustre client should attempt to
2754 reconnect to the last active server first before trying the
2755 other potential connections.
2757 Severity : enhancement
2759 Description: client-side metadata stat-ahead during readdir(directory readahead)
2760 Details : perform client-side metadata stat-ahead when the client detects
2761 readdir and sequential stat of dir entries therein
2766 Description: when mds and osts use different quota unit(32bit and 64bit),
2767 quota will be released repeatly.
2768 Details : void sending multiple quota reqs to mds, which will keep the
2769 status between the reqs.
2772 Frequency : only with liblustre clients on XT3
2774 Description: evictions taking too long
2775 Details : allow llrd to evict clients directly on OSTs
2779 Description: osts not allocated evenly to files
2780 Details : change the condition to increase offset_idx
2782 Severity : enhancement
2784 Description: self-adjustable client's lru lists
2785 Details : use adaptive algorithm for managing client cached locks lru
2786 lists according to current server load, other client's work
2787 pattern, memory activities, etc. Both, server and client
2788 side namespaces provide number of proc tunables for controlling
2793 Description: rewrite ext2-derived code in llite/dir.c and obdclass/uuid.c
2794 Details : rewrite inherited code (uuid parsing code from ext2 utils and
2795 readdir code from ext3) from scratch preserving functionality.
2799 Description: Only those disconnect error should be returned by rq_status.
2800 Details : In open/enqueue processs, Some errors, which will cause client
2801 disconnected, should be returned by rq_status, while other
2802 errors should still be returned by intent, then mdc or llite will
2805 Severity : enhancement
2807 Description: Tune the kernel for good SCSI performance.
2808 Details : Set the value of /sys/block/{dev}/queue/max_sectors_kb
2809 to the value of /sys/block/{dev}/queue/max_hw_sectors_kb
2813 Frequency : Always for filesystems larger than 2TB on 32-bit systems.
2814 Bugzilla : 13547, 13627
2815 Description: Data corruption for OSTs that are formatted larger than 2TB
2817 Details : When generating the bio request for lustre file writes the
2818 sector number would overflow a temporary variable before being
2819 used for the IO. The data reads correctly from Lustre (which
2820 will overflow in a similar manner) but other file data or
2821 filesystem metadata may be corrupted in some cases.
2825 Description: "lfs find -obd UUID" prints directories
2826 Details : "lfs find -obd UUID" will return all directory names instead
2827 of just file names. It is incorrect because the directories
2828 do not reside on the OSTs.
2831 Bugzilla : 13596, 13614
2832 Description: MDS hang after unclean shutdown of lots of clients
2833 Details : Never resend AST requests.
2837 Frequency : Always, for kernels after 2.6.16
2838 Description: Fix warning idr_remove called for id=.. which is not allocated.
2839 Details : Last kernels save old s_dev before kill super and not allow
2840 to restore from callback - restore it before call kill_anon_super.
2844 Description: Fix error on 'ls .' at the top of the Lustre mount.
2845 Details : Don't revalidate dentry if it is a root dentry.
2847 --------------------------------------------------------------------------------
2849 2007-08-27 Cluster File Systems, Inc. <info@clusterfs.com>
2851 * Support for kernels:
2852 2.6.5-7.286 (SLES 9),
2853 2.6.9-55.0.2.EL (RHEL 4),
2854 2.6.16.46-0.14 (SLES 10),
2855 2.6.18-8.1.8.el5 (RHEL 5),
2856 2.6.18.8 vanilla (kernel.org)
2857 * Client support for unpatched kernels:
2858 (see http://wiki.lustre.org/index.php?title=Patchless_Client)
2859 2.6.16 - 2.6.22 vanilla (kernel.org)
2860 * Due to recently discovered recovery problems, we do not recommend
2861 using patchless RHEL 4 clients with this or any earlier release.
2862 * Recommended e2fsprogs version: 1.39.cfs8
2863 * Note that reiserfs quotas are disabled on SLES 10 in this kernel.
2868 Description: block reactivating mgc import until all deactivates complete
2869 Details : Fix race when failing back MDT/MGS to itself (testing)
2871 Severity : enhancement
2873 Description: lfs setstripe enhancement
2874 Details : Make lfs setstripe understand 'k', 'm' and 'g' for stripe size.
2877 Frequency : mds/oss recovery
2879 Description: llog ctxt is refrenced after it has been freed.
2880 Details : llog ctxt refcount was added to avoide the race between ctxt free
2881 and llog recovery process. Each llog user must hold ctxt refcount
2882 before it access the llog. And the llog ctxt can only be freed
2883 when its refcount is zero.
2885 Severity : enhancement
2887 Description: randomly memory allocation failure util
2888 Details : Make lustre randomly failed allocating memory for testing purpose.
2891 Frequency : only for SLES10
2893 Description: Update kernel patch for SLES10 SP1
2894 Details : Add patch blkdev_tunables-2.6-sles10.patch to 2.6-sles10.series.
2896 Severity : enhancement
2898 Description: omit set fsid for export NFS
2899 Details : fix set/restore device id for avoid EMFILE error and mark lustre fs
2900 as FS_REQUIRES_DEV for avoid problems with generate fsid.
2903 Frequency : after network failures
2905 Description: Add sync option to mount_lustre.c
2906 Details : Client loses data written to lustre after a network interruption.
2908 Severity : enhancement
2910 Description: Error message improvement.
2911 Details : Merging of two LCONSOLE_ERROR_MSG into one.
2913 Severity : enhancement
2915 Description: don't use GFP_* in generic Lustre code.
2916 Details : Use cfs_alloc_* functions and CFS_* flags for code portability.
2918 Severity : enhancement
2920 Description: obdclass is limited by single OBD_ALLOC(idarray)
2921 Details : replace OBD_ALLOC/OBD_FREE with OBD_VMALLOC/OBD_VFREE
2923 Severity : enhancement
2925 Description: updated patchess for new RHEL4 kernel
2926 Details : Fixed ext3-unlink-race.patch per Kalpak's comment.
2928 Severity : enhancement
2930 Description: warnings with build patchless client with vanila 2.6.19 and up
2931 Details : change old ctl_table style and replace ctl_table/ctl_table_header
2932 with cfs_sysctl_table_t/cfs_sysctl_table_header_t
2934 Severity : enhancement
2936 Description: O_DIRECT bypasses client statistics.
2937 Details : When running with O_DIRECT I/O, neither the client rpc_stats nor
2938 read_ahead_stats were updated. Copied stats section from
2939 osc_send_oap_rpc() into async_internal().
2942 Frequency : only for Cray XT3
2944 Description: peer credits not enough on many OST per OSS systems.
2945 Details : Use new lnet way to add credits as we need those for pings and ASTs
2948 Frequency : only with liblustre
2950 Description: Liblustre is not releasing flock locks on file close.
2951 Details : Release flock locks on file close.
2954 Frequency : only for RHEL4
2956 Description: Update kernel patches for kernel-2.6.9-55.0.2.EL
2957 Details : Remove inode-nr_unused-2.6.9-rhel4.patch from 2.6-rhel4.series
2958 Update target file and kernel config.
2962 Description: lustre support for RHEL5
2963 Details : Add support for RHEL5.
2968 Description: ASSERTION(export != NULL) failed in target_handle_connect
2969 Details : Assetion hit is result of rare race between disconnect and connet
2970 to same nid. target_handle_connect found old connect cockie and
2971 tried to reconnect, but can't find export for this cockie.
2976 Description: umount blocks forever on error
2977 Details : In result of wrong using obd_no_recov and obd_force flags client
2978 can hand if cancel or some other requests is lost.
2981 Frequency : Only for SLES
2983 Description: sanity_quota fail test_1
2984 Details : There are multiple occurences of $TSTUSR in SLES's /etc/group
2985 file, which makes TSTID[2] inunique.
2987 Severity : enhancement
2989 Description: Kernel patches for SLES9 2.6.5-7.286 kernel
2990 Details : Update target/ChangeLog/which_patch .
2992 Severity : enhancement
2994 Description: jbd statistics
2995 Details : Port older jbd statistics patch for sles10
3000 Description: lvbo_init failed for resource with missing objects.
3001 Details : Fix returning error if we do stat for file with missing/corrupted
3002 objects and i_size set to all sum of size of all avaible objects.
3003 if we to truncate/write to missing object - it is recreated.
3008 Description: Oops in read and write path when failing to allocate lock.
3009 Details : Check if lock allocation failed and return error back.
3012 Frequency : When flocks are used.
3014 Description: assertion failure in ldlm_cli_enquque_fini for non NULL lock.
3015 Details : Flock locks might destroy just granted lock if it could be merged
3016 with another existing flock, this is done in completion handler,
3017 so teach ldlm_cli_enquque_fini that this is a valid case for
3023 Description: reply_lock_interpret crash due to race with it and lock cancel.
3024 Details : Do not replay locks that are being cancelled. Do not reference
3025 locks by their address during replay, just by their handle.
3028 Frequency : only with deactivated OSTs
3030 Description: lstripe command fails for valid OST index
3031 Details : The stripe offset is compared to 'lov->desc.ld_tgt_count'
3032 instead of lov->desc.ld_active_tgt_count.
3034 Severity : enhancement
3036 Description: Build failure against Centos5 (RHEL5)
3037 Details : Use getpagesize() instead of PAGE_SIZE.
3039 --------------------------------------------------------------------------------
3041 2007-07-30 Cluster File Systems, Inc. <info@clusterfs.com>
3043 * Support for kernels:
3044 2.6.5-7.283 (SLES 9),
3045 2.6.9-55.EL (RHEL 4),
3046 2.6.16.46-0.14 (SLES 10),
3047 2.6.18.8 vanilla (kernel.org)
3048 * Client support for unpatched kernels:
3049 (see http://wiki.lustre.org/index.php?title=Patchless_Client)
3050 2.6.16 - 2.6.22 vanilla (kernel.org)
3051 * Due to recently discovered recovery problems, we do not recommend
3052 using patchless RHEL 4 clients with this or any earlier release.
3053 * Recommended e2fsprogs version: 1.39.cfs8
3054 * Note that reiserfs quotas are disabled on SLES 10 in this kernel.
3055 * Starting with this release, the ldiskfs backing filesystem required
3056 by Lustre is now in its own package, lustre-ldiskfs. This package
3057 should be installed. It is versioned separately from Lustre and
3058 may be released separately in future.
3063 Description: server LBUG when shutting down
3064 Details : Block umount forever until the mount refcount is zero rather
3065 than giving up after an arbitrary timeout.
3067 Severity : enhancement
3069 Description: add optional extra BUILD_VERSION info
3070 Details : add a new environment variable (namely LUSTRE_VERS) which allows
3071 to override the lustre version.
3074 Frequency : 2.6.18 servers only
3076 Description: ll_kern_mount() doesn't release the module reference
3077 Details : The ldiskfs module reference count never drops down to 0
3078 because ll_kern_mount() doesn't release the module reference.
3083 Description: server LBUG when using old ost_num_threads parameter
3084 Details : Accept the old ost_num_threads parameter but warn that it
3085 is deprecated, and fix an off-by-one error that caused an LBUG.
3090 Description: Transient SCSI error results in persistent IO issue
3091 Details : iobuf->dr_error is not reinitialized to 0 between two
3095 Frequency : sometimes when underlying device returns I/O errors
3097 Description: OSTs not going read-only during write failures
3098 Details : OSTs are not remounted read-only when the journal commit threads
3099 get I/O errors because fsfilt_ext3 calls journal_start/stop()
3100 instead of the ext3 wrappers.
3103 Frequency : only with 10000 clients or more
3105 Description: poor connect scaling with increasing client count
3106 Details : Don't run filter_grant_sanity_check for more than 100 exports
3107 to improve scaling for large numbers of clients.
3110 Frequency : SLES10 only
3112 Description: sanity-quota.sh quotacheck failed: rc = -22
3113 Details : Quotas cannot be enabled on SLES10.
3116 Frequency : liblustre clients only
3118 Description: getdirentries does not give error when run on compute nodes
3119 Details : getdirentries does not fail when the size specified as an argument
3120 is too small to contain at least one entry
3122 Severity : enhancement
3124 Description: Add LNET router traceability for debug purposes
3125 Details : If a checksum failure occurs with a router as part of the
3126 IO path, the NID of the last router that forwarded the bulk data
3127 is printed so it can be identified.
3132 Description: OST "spontaneously" evicts client; client has imp_pingable == 0
3133 Details : Due to a race condition, liblustre clients were occasionally
3134 evicted incorrectly.
3136 Severity : enhancement
3138 Description: lfs setstripe use optional parameters instead of postional
3141 Severity : enhancement
3143 Description: Nanosecond timestamp support for ldiskfs
3144 Details : The on-disk ldiskfs filesystem has added support for nanosecond
3145 resolution timestamps. There is not yet support for this at
3146 the Lustre filesystem level.
3149 Frequency : during server recovery
3151 Description: MDS failing to send precreate requests due to OSCC_FLAG_RECOVERING
3152 Details : request with rq_no_resend flag not awake l_wait_event if they get a
3156 Frequency : nfs export on patchless client
3158 Description: connectathon hang when test nfs export over patchless client
3159 Details : Disconnected dentry cannot be found with lookup, so we do not need
3160 to unhash it or make it invalid
3163 Description: fix llapi_lov_get_uuids() to allow many OSTs to be returned
3164 Details: : Change llapi_lov_get_uuids() to read the UUIDs from /proc instead
3165 of using an ioctl. This allows lfsck for > 160 OSTs to succeed.
3170 Description: open req refcounting wrong on reconnect
3171 Details : If reconnect happened between getting open reply from server and
3172 call to mdc_set_replay_data in ll_file_open, we will schedule
3173 replay for unreferenced request that we are about to free.
3174 Subsequent close will crash in variety of ways.
3175 Check that request is still eligible for replay in
3176 mdc_set_replay_data().
3181 Description: disable writes to filesystem when reading health_check file
3182 Details : the default for reading the health_check proc file has changed
3183 to NOT do a journal transaction and write to disk, because this
3184 can cause reads of the /proc file to hang and block HA state
3185 checking on a healthy but otherwise heavily loaded system. It
3186 is possible to return to the previous behaviour during configure
3187 with --enable-health-write.
3189 Severity : enhancement
3191 Description: 64-bit inode version
3192 Details: : Add a on-disk 64-bit inode version for ext3 to track changes made
3193 to the inode. This will be required for version-based recovery.
3198 Description: MDS fails to start if a duplicate client export is detected
3199 Details : in some rare cases it was possible for a client to connect to
3200 an MDS multiple times. Upon recovery the MDS would detect this
3201 and fail during startup. Handle this more gracefully.
3203 Severity : enhancement
3205 Description: Add -o localflock option to simulate old noflock behaviour.
3206 Details : This will achieve local-only flock/fcntl locks coherentness.
3211 Description: log_commit_thread vs filter_destroy race leads to crash
3212 Details : Take import reference before releasing llog record semaphore
3217 Description: Wrong request locking in request set processing
3218 Details : ptlrpc_check_set wrongly uses req->rq_lock for proctect add to
3219 imp_delayed_list, in this place should be used imp_lock.
3222 Frequency : when reconnecting
3224 Description: Grant leak when OSC reconnect to OST
3225 Details : When osc reconnect ost, OST(filter) should check whether it
3226 should grant more space to client by comparing fed_grant and
3227 cl_avail_grant, and return the granted space to client instead
3228 of "new granted" space, because client will call osc_init_grant
3229 to update the client grant space info.
3232 Frequency : when client reconnects to OST
3234 Description: Grant leak when OSC does a resend and replays bulk write
3235 Details : When osc reconnect to OST, OST(filter) should clear grant info of
3236 bulk write request, because the grant info while be sync between
3237 OSC and OST when reconnect, and we should ignore the grant info
3238 these of resend/replay write req.
3243 Description: Grant space more than avaiable space sometimes.
3244 Details : When then OST is about to be full, if two bulk writing from
3245 different clients came to OST. Accord the avaliable space of the
3246 OST, the first req should be permitted, and the second one
3247 should be denied by ENOSPC. But if the seconde arrived before
3248 the first one is commited. The OST might wrongly permit second
3249 writing, which will cause grant space > avaiable space.
3252 Frequency : when client is evicted
3254 Description: Grant might be wrongly erased when osc is evicted by OST
3255 Details : when the import is evicted by server, it will fork another
3256 thread ptlrpc_invalidate_import_thread to invalidate the
3257 import, where the grant will be set to 0. While the original
3258 thread will update the grant it got when connecting. So if
3259 the former happened latter, the grant will be wrongly errased
3260 because of this race.
3265 Description: Checking Stale with correct fid
3266 Details : ll_revalidate_it should uses de_inode instead of op_data.fid2
3267 to check whether it is stale, because sometimes, we want the
3268 enqueue happened anyway, and op_data.fid2 will not be initialized.
3270 Severity : enhancement
3272 Description: update patchless client
3273 Details : Add support for patchless client with 2.6.20, 2.6.21 and RHEL 5
3276 Frequency : only with 2.4 kernel
3278 Description: random memory corruption
3279 Details : size of struct ll_inode_info is to big for union inode.u and this
3280 can be cause of random memory corruption.
3285 Description: Memory leak in recovery
3286 Details : Lov_mds_md was not free in an error handler in mds_create_object.
3287 It should also check obd_fail before fsfilt_start, otherwise if
3288 fsfilt_start return -EROFS,(failover mds during mds recovery).
3289 then the req will return with repmsg->transno = 0 and rc = EROFS.
3290 and we met hit the assert LASSERT(req->rq_reqmsg->transno ==
3291 req->rq_repmsg->transno) in ptlrpc_replay_interpret. Fcc should
3292 be freed no matter whether fsfilt_commit success or not.
3295 Frequency : only with large numbers of cores on a single node
3297 Description: Prevents from taking the superblock lock in llap_from_page for
3299 Details : using LL_ORIGIN_REMOVEPAGE origin flag instead of LL_ORIGIN_UNKNOW
3300 for llap_from_page call in ll_removepage() prevents from taking
3301 the superblock lock for a soon died page.
3306 Description: Not check open intent error before release open handle
3307 Details : in some rare cases, the open intent error is not checked before
3308 release open handle, which may cause
3309 ASSERTION(open_req->rq_transno != 0), because it tries to release
3310 the failed open handle.
3315 Description: Set cat log bitmap only after create log success.
3316 Details : in some rare cases, the cat log bitmap is set too early. and it
3317 should be set only after create log success.
3321 Description: Accessing a block bevice can re-enable I/O when Lustre is
3322 tearing down a device.
3323 Details : dev_clear_rdonly(bdev) must be called in kill_bdev() instead of
3328 Description: service threads may hog cpus when there are a lot of requests
3329 Details : Insert cond_resched to give other threads a chance to use some CPU
3334 Description: the cat log was not initialized in recovery
3335 Details : When mds(mgs) do recovery, the tgt_count might be zero, so the
3336 unlink log on mds will not be initialized until mds post
3337 recovery. And also in mds post recovery, the unlink log will
3338 initialization will be done asynchronausly, so there will be race
3339 between add unlink log and unlink log initialization.
3343 Description: brw_stats were being printed incorrectly
3344 Details : brw_stats were being printed as log2 but all of them were not
3345 recorded as log2. Also remove some code duplication arising from
3346 filter_tally_{read,write}.
3350 Frequency : rare, only in recovery.
3351 Description: ASSERTION(req->rq_type != LI_POISON) failed
3352 Details : imp_lock should be held while iterating over imp_sending_list for
3353 prevent destroy request after get timeout in ptlrpc_queue_wait.
3357 Description: replay-single.sh test 52 fails
3358 Details : A lock's skiplist need to be cleanup when it being unlinked
3359 from its resource list.
3364 Description: Short directio read returns full requested size rather than
3366 Details : Direct I/O operations should return actual amount of bytes
3367 transferred rather than requested size.
3369 Severity : enhancement
3371 Description: metadata RPC reduction (e.g. for rm performance)
3372 Details : decrease the amount of synchronous RPC between clients and servers
3373 by canceling conflicing lock before the operation on the client
3374 and packing thier handles into the main operation RPC to server.
3376 Severity : enhancement
3378 Description: add #ifdef HAVE_KERNEL_CONFIG_H
3379 Details : kernels from 2.6.19 not need include linux/config.h, but add
3380 include linux/autoconf.h in commpiler command line.
3382 Severity : enhancement
3384 Description: patchless client support for 2.6.22 kernel
3385 Details : 2.6.22 has only one visble change, SLAB_CTOR_* constants is
3386 removed. In this case we need drop using os depended interface to
3387 kmem_cache and use cfs_mem_cache API.
3392 Description: fix mal-formatted messages
3393 Details : fix some mal-formatted DEBUG_REQ and LCONSOLE_ERROR_MSG messages
3397 Frequency : always in liblustre
3398 Description: wrong IS_ERR implementation in liblustre.h
3399 Details : fix IS_ERR implementation in liblustre.h for right detect errors.
3404 Description: Correct condition for output debug message.
3405 Details : inode i_nlink equal zero is not enough for output message about
3406 disk corruption, i_ctime and i_mode should be also checked.
3410 Frequency : always in patchless client
3411 Description: add configure check for truncate_complete_page
3412 Details : improve checks for exported symbols. This allow run check without
3413 sources, but with Module.symvers shipped with kernel distribution.
3414 add check for truncate_complete_page used by patchless client.
3418 Description: sanity.sh test_77h fails with "test_77h file compare failed"
3419 Details : test_77h uses a file which was messed by other test case.
3423 Description: Not Check whether lov_tgts is NULL in some lov functions
3424 Details : Checking whether lov_tgts is NULL in some functions.
3428 Description: replace obdo_alloc() with OBDO_ALLOC macro
3429 Details : nothing special is done in obdo_alloc() function, and for
3430 debugging purpose, it needs to be replaced with macros.
3434 Description: bad return value and errno from fcntl call
3435 Details : In liblustre API, errno should be a negative value if error
3440 Description: ptlrpc_check_set() LBUG
3441 Details : In case of positive reply from server and failed client bulk
3442 callback after bulk transfer shouldn't LBUG, but process this
3443 request as erroneous.
3445 Severity : enhancement
3447 Description: VFS operations stats tool.
3448 Details : Tool which collects stats by tracking value written in pid,
3449 ppid, gid and uses llstat to generate output to plot graph using
3451 Updated lustre/utils/Makefile.am
3452 Added lustre/utils/ltrack_stats.c
3454 Severity : enhancement
3456 Description: 2.6.18 server support (lustre 1.6.1)
3457 Details : Support for 2.6.18 kernels on the server side.
3462 Description: ASSERTION(imp->imp_conn_current) failed
3463 Details : an assertion failure is hit if a client node boots and attempts to
3464 mount a lustre filesystem faster than RECONNECT_INTERVAL seconds.
3467 Frequency : only for i686
3469 Description: 1.4.11 RC1 build fails for RHEL 4, i686
3470 Details : Fixed config variable for build.
3475 Description: Updated patchess for new RHEL4 kernel
3476 Details : Updated patch inode-nr_unused-2.6.9-rhel4.patch
3477 Updated patch jbd-stats-2.6.9.patch
3478 Updated patch qsnet-rhel4-2.6.patch
3479 Updated patch quota-deadlock-on-pagelock-core.patch
3480 Updated patch vfs_intent-2.6-rhel4.patch
3481 Updated patch vfs_races-2.6-rhel4.patch
3482 Updated series file 2.6-rhel4-titech.series
3483 Updated series file 2.6-rhel4.series
3484 Updated kernel config files
3489 Description: lquota slave complains LBUG when reconnecting with mds
3491 Details : quota slave depends qctxt->lqc_import to send its quota request.
3492 This pointer will be invalid if mds did failover or broke its
3493 connect to osts, which leads to LBUG.
3496 Frequency : when qunit size is too small(less than 20M)
3498 Description: write is stopped by improper -EDQUOT
3499 Details : If the master is busy and qunit size is small enough(let's say 1M),
3500 the slave can not get quota from master on time, which will lead
3501 slave to trigger a -EQUOTA to client.
3506 Description: Deadlock during metadata tests
3507 Details : in prune_dir_dentries(), shrink_dcache_parent() should not be
3508 called with the per-dentry lock held.
3511 Frequency : SLES9 only
3513 Description: Lustre patched kernel for SLES9 SP3 has NR_CPUS set to 8
3514 Details : set CONFIG_NR_CPUS to 128 instead of 8.
3516 Severity : enhancement
3518 Description: remove fs_prep_san_write operation and related patches
3519 Details : remove the ext3-san-jdike patches which are no longer useful.
3524 Description: LDISKFS-fs error (device sdc): ldiskfs_free_blocks
3525 Details : a disk corruption can cause the mballoc code to assert on a
3526 double free or other extent corruptions. Handle these with
3527 ext3_error() instead of assertions.
3530 Frequency : only with mballoc3 code and deep extent trees
3532 Description: ldiskfs_ext_search_right: bad header in inode: unexpected eh_depth
3533 Details : a wrong check of extent headers in ldiskfs_ext_search_right()
3534 can cause the filesystem to be remounted read-only.
3539 Description: lfsck built against 1.4.x cannot run against 1.6.0 lustre
3540 Details : the definition for OBD_IOC_GETNAME changed in 1.6.0. One of the
3541 few external users of this ioctl number is lfsck's call to
3542 llapi_lov_get_uuids() and this caused lfsck to fail at startup.
3543 Add the old ioctl number to the handler so both old and new
3548 Description: parallel lock callbacks
3549 Details : Instead of sending blocking and completion callbacks as separated
3550 requests, adding them to a set and sending in parallel.
3554 Description: Disable most debugging by default
3555 Details : To improve performance, disable most logging (for debug purposes)
3556 by default. VFSTRACE, RPCTRACE, and DLMTRACE are now off by
3557 default, and HA includes fewer messages.
3561 Frequency : only run on patchless client.
3562 Description: use do_facet on sanity.sh for test handling recoverables errors
3563 Details : use do_facet instead of direct use sysctl for set fail_loc on OST
3567 Description: hash tables for lists of nids, connections and uuids
3568 Details : Hash tables noticeably help when a lot of clients connect to a
3569 server, to faster identify duplicate connections or reconnects,
3570 also to faster find export to evict in manual eviction case.
3574 Description: Sometimes, when the server evict a client, and the client will
3575 not be evicted as soon as possible.
3576 Details : In enqueue req, the error was returned by intent, instead of
3577 rq_status which make ptlrpc layer not detect this error, and
3578 does not evict the client. So enqueue error should be returned
3582 Frequency : only at startup
3584 Description: Delay client connections to MDT until fist MDT->OST connect
3585 Details : If a client tried to create a new file before the MDT had
3586 connected to any OSTs, the create would return EIO. Now
3587 the client will simply block until the MDT connects to the
3588 first OST and the create can succeed.
3591 Frequency : only for SLES9
3593 Description: Routinely utilize latest Quadrics drivers in CFS releases
3594 Details : Update patch qsnet-suse-2.6.patch.
3597 Frequency : only for sles10
3599 Description: Update patches for SLES 10 SP1 kernel.
3600 Details : Update the patch vfs_intent-2.6-sles10.patch.
3605 Description: Routinely utilize latest Quadrics drivers in CFS releases
3606 Details : Update patch qsnet-rhel4-2.6.patch.
3609 Frequency : at statup only
3611 Description: mds_lov_synchronize race leads to various problems
3612 Details : simultaneous MDT->OST connections at startup can cause the
3613 sync to abort, leaving the OSC in a bad state.
3618 Description: Using wrong pointer in osc_brw_prep_request
3619 Details : Access to array[-1] can produce panic if kernel compiled with
3620 CONFIG_PAGE_ALLOC enabled
3622 Severity : enhancement
3624 Description: Async OSC create to avoid the blocking unnecessarily.
3625 Details : If a OST has no remain object, system will block on the creating
3626 when need to create a new object on this OST. Now, ways use
3627 pre-created objects when available, instead of blocking on an
3628 empty osc while others are not empty. If we must block, we block
3629 for the shortest possible period of time.
3633 Frequency : only in recovery
3634 Description: Mark OST as early accessible if his start SYNC.
3635 Details : osc_precreate return flag early accessible if oscc marked as
3636 OSCC_FLAG_SYNC_IN_PROGRESS.
3641 Description: Sometimes precreate code can triger create object on wrong ost
3642 Details : Wrong protected or not not restored variables aftre precreate loop
3643 can produce creation object on wrong ost.
3646 Frequency : oss recovery
3648 Description: llog_commit_thread cleanup should sync with llog_commit_thread
3650 Details : llog_commit_thread_count should be synced between llog_commit
3651 start and cleanup, so new llog_commit thread should not be started
3652 when llog_commit threads being stopped to avoid accessing some
3655 Severity : enhancement
3657 Description: Add printing inode info into message about error in writepage.
3659 --------------------------------------------------------------------------------
3661 2007-05-03 Cluster File Systems, Inc. <info@clusterfs.com>
3666 Frequency : on some architectures
3668 Description: 1.6 client sometimes fails to mount from a 1.4 MDT
3669 Details : Uninitialized flags sometimes cause configuration commands to
3673 Frequency : patchless clients only
3675 Description: missing __iget() symbol export
3676 Details : The __iget() symbol export is missing. To avoid the need for
3677 this on patchless clients the deathrow inode reaper is turned
3678 off, and we depend on the VM to clean up old inodes. This
3679 dependency was during via the fix for bug 12181.
3684 Description: sanity.sh fail: test_52b
3685 Details : The ll_inode_to_ext_flags() has a glitch which makes MDS return
3686 incorrect inode's flags to client.
3688 --------------------------------------------------------------------------------
3690 2007-04-19 Cluster File Systems, Inc. <info@clusterfs.com>
3692 * CONFIGURATION CHANGE. This version of Lustre WILL NOT
3693 INTEROPERATE with older versions automatically. In many cases a
3694 special upgrade step is needed. Please read the
3695 user documentation before upgrading any part of a 1.4.x system.
3696 * WARNING: Lustre configuration and startup changes are required with
3697 this release. See https://mail.clusterfs.com/wikis/lustre/MountConf
3699 * Support for kernels:
3700 2.4.21-47.0.1.EL (RHEL 3),
3701 2.6.5-7.283 (SLES 9),
3702 2.6.9-42.0.10.EL (RHEL 4),
3703 2.6.12.6 vanilla (kernel.org),
3704 2.6.16.27-0.9 (SLES10)
3705 * Client support for unpatched kernels:
3706 (see https://mail.clusterfs.com/wikis/lustre/PatchlessClient)
3707 2.6.16 - 2.6.19 vanilla (kernel.org),
3708 2.6.9-42.0.8EL (RHEL 4)
3709 * Recommended e2fsprogs version: 1.39.cfs6
3710 * Note that reiserfs quotas are disabled on SLES 10 in this kernel
3713 Severity : enhancement
3715 Description: Async OSC create to avoid the blocking unnecessarily.
3716 Details : If a OST has no remain object, system will block on the creating
3717 when need to create a new object on this OST. Now, ways use
3718 pre-created objects when available, instead of blocking on an
3719 empty osc while others are not empty. If we must block, we block
3720 for the shortest possible period of time.
3722 Severity : enhancement
3724 Description: MountConf
3725 Details : Lustre configuration is now managed via mkfs and mount
3726 commands instead of lmc and lconf. New obd types (MGS, MGC)
3727 are added for dynamic configuration management. See
3728 https://mail.clusterfs.com/wikis/lustre/MountConf for
3731 Severity : enhancement
3733 Description: dynamic OST addition
3734 Details : OSTs can now be added to a live filesystem
3736 Severity : enhancement
3738 Description: startup order invariance
3739 Details : MDTs and OSTs can be started in any order. Clients only
3740 require the MDT to complete startup.
3742 Severity : enhancement
3744 Description: parallel, asynchronous orphan cleanup
3745 Details : orphan cleanup is now performed in separate threads for each
3746 OST, allowing parallel non-blocking operation.
3748 Severity : enhancement
3750 Description: optimized stripe assignment
3751 Details : stripe assignments are now made based on ost space available,
3752 ost previous usage, and OSS previous usage, in order to try
3753 to optimize storage space and networking resources.
3755 Severity : enhancement
3757 Description: Permanently set tunables
3758 Details : All writable /proc/fs/lustre tunables can now be permanently
3759 set on a per-server basis, at mkfs time or on a live
3762 Severity : enhancement
3764 Description: Lustre message v2
3765 Details : Add lustre message format v2.
3767 Severity : enhancement
3769 Description: client OST exclusion list
3770 Details : Clients can be started with a list of OSTs that should be
3771 declared "inactive" for known non-responsive OSTs.
3775 Description: ENOENT returned for valid filehandle during dbench.
3776 Details : Check if a directory has children when invalidating dentries
3777 associated with an inode during lock cancellation. This fixes
3778 an incorrect ENOENT sometimes seen for valid filehandles during
3779 testing with dbench.
3782 Frequency : SFS test only (otherwise harmless)
3784 Description: SPEC SFS validation failure on NFS v2 over lustre.
3785 Details : Changes the blocksize for regular files to be 2x RPC size,
3786 and not depend on stripe size.
3788 Severity : enhancement
3790 Description: fine-grained SMP locking inside DLM
3791 Details : Improve DLM performance on SMP systems by removing the single
3792 per-namespace lock and replace it with per-resource locks.
3794 Severity : enhancement
3796 Description: don't hold multiple extent locks at one time
3797 Details : To avoid client eviction during large writes, locks are not
3798 held on multiple stripes at one time or for very large writes.
3799 Otherwise, clients can block waiting for a lock on a failed OST
3800 while holding locks on other OSTs and be evicted.
3802 Severity : enhancement
3804 Description: Multiple MD RPCs in flight.
3805 Details : Further unserialise some read-only MDT RPCs - learn about intents.
3806 To avoid overly-overloading MDT, introduce a limit on number of
3807 MDT RPCs in flight for a single client and add /proc controls
3808 to adjust this limit.
3810 Severity : enhancement
3812 Description: client read/write statistics
3813 Details : Add client read/write call usage stats for performance
3814 analysis of user processes.
3815 /proc/fs/lustre/llite/*/offset_stats shows non-sequential
3816 file access. extents_stats shows chunk size distribution.
3817 extents_stats_per_process show chunk size distribution per
3820 Severity : enhancement
3822 Description: per-client statistics on server
3823 Details : Add ldlm and operations statistics for each client in
3824 /proc/fs/lustre/mds|obdfilter/*/exports/
3826 Severity : enhancement
3828 Description: improved MDT statistics
3829 Details : Add detailed MDT operations statistics in
3830 /proc/fs/lustre/mds/*/stats
3832 Severity : enhancement
3834 Description: VFS operations stats
3835 Details : Add client VFS call stats, trackable by pid, ppid, or gid
3836 /proc/fs/lustre/llite/*/stats_track_[pid|ppid|gid]
3841 Description: Fix client-side osc byte counters
3842 Details : The osc read/write byte counters in
3843 /proc/fs/lustre/osc/*/stats are now working
3846 Frequency : always as root on SLES
3848 Description: Failure of copying files with lustre special EAs.
3849 Details : Client side always return success for setxattr call for lustre
3850 special xattr (currently only "trusted.lov").
3855 Description: Refcount LNET uuids
3856 Details : The global LNET uuid list grew linearly with every startup;
3857 refcount repeated list entries instead of always adding to
3860 Severity : enhancement
3862 Description: Dynamic service threads
3863 Details : Within a small range, start extra service threads
3864 automatically when the request queue builds up.
3867 Frequency : mixed-endian client/server environments
3869 Description: mixed-endian crashes
3870 Details : The new msg_v2 system had some failures in mixed-endian
3873 Severity : enhancement
3875 Description: Easy OST removal
3876 Details : OSTs can be permanently deactivated with e.g. 'lctl
3877 conf_param lustre-OST0001.osc.active=0'
3879 Severity : enhancement
3881 Description: MGS proc entries
3882 Details : Added basic proc entries for the MGS showing what filesystems
3885 Severity : enhancement
3887 Description: provide MGS failover
3888 Details : Added config lock reacquisition after MGS server failover.
3890 Severity : enhancement
3892 Description: add Linux 2.4 support
3893 Details : Added support for RHEL 2.4.21 kernel for 1.6 servers and clients
3897 Description: a large application tries to do I/O to the same resource and dies
3898 in the middle of it.
3899 Details : Check the req->rq_arrival time after the call to
3900 ost_brw_lock_get(), but before we do anything about
3901 processing it & sending the BULK transfer request. This
3902 should help move old stale pending locks off the queue as
3903 quickly as obd_timeout.
3906 Frequency : when an incorrect nid is specified during startup
3908 Description: ptlrpc connect to non-existant node causes kernel crash
3909 Details : LNET can't be re-entered from an event callback, which
3910 happened when we expire a message after the export has been
3911 cleaned up. Instead, hand the zombie cleanup off to another
3914 Severity : enhancement
3916 Description: plain/inodebits lock performance improvement
3917 Details : Grouping plain/inodebits in granted list by their request modes
3918 and bits policy, thus improving the performance of search through
3922 Frequency : only if OST filesystem is corrupted
3924 Description: client incorrectly hits assertion in ptlrpc_replay_req()
3925 Details : for a short time RPCs with bulk IO are in the replay list,
3926 but replay of bulk IOs is unimplemented. If the OST filesystem
3927 is corrupted due to disk cache incoherency and then replay is
3928 started it is possible to trip an assertion. Avoid putting
3929 committed RPCs into the replay list at all to avoid this issue.
3932 Frequency : liblustre (e.g. catamount) on a large cluster with >= 8 OSTs/OSS
3934 Description: System hang on startup
3935 Details : This bug allowed the liblustre (e.g. catamount) client to
3936 return to the app before handling all startup RPCs. This
3937 could leave the node unresponsive to lustre network traffic
3938 and manifested as a server ptllnd timeout.
3940 Severity : enhancement
3942 Description: Add "/proc/sys/lustre/debug_peer_on_timeout"
3943 Details : liblustre envirable: LIBLUSTRE_DEBUG_PEER_ON_TIMEOUT
3944 boolean to control whether to print peer debug info when a
3945 client's RPC times out.
3948 Frequency : only for kernels with patches from Lustre below 1.4.3
3950 Description: Remove old rdonly API
3951 Details : Remove old rdonly API which unused from at least lustre 1.4.3
3954 Frequency : only for devices with external journals
3956 Description: Set external device read-only also
3957 Details : During a commanded failover stop, we set the disk device
3958 read-only while the server shuts down. We now also set any
3959 external journal device read-only at the same time.
3962 Frequency : when upgrading from 1.4 while trying to change parameters
3964 Description: The wrong (new) MDC name was used when setting parameters for
3965 upgraded MDT's. Also allows changing of OSC (and MDC)
3966 parameters if --writeconf is specified at tunefs upgrade time.
3969 Frequency : when setting specific ost indicies
3971 Description: QOS code breaks on skipped indicies
3972 Details : Add checks for missing OST indicies in the QOS code, so OSTs
3973 created with --index need not be sequential.
3975 Severity : enhancement
3977 Description: Add uninit_groups feature to ldiskfs2 to speed up e2fsck
3978 Details : The uninit_groups feature works in conjunction with the kernel
3979 filesystem code (ldiskfs2 only) and e2fsprogs-1.39-cfs6 to speed
3980 up the pass1 processing of e2fsck. This is a read-only feature
3981 in ldiskfs2 only, so older kernels and current ldiskfs cannot
3982 mount filesystems that have had this feature enabled.
3984 Severity : enhancement
3986 Description: Improve multi-block allocation algorithm to avoid fragmentation
3987 Details : The mballoc3 code (ldiskfs2 only) adds new mechanisms to improve
3988 allocation locality and avoid filesystem fragmentation.
3990 ------------------------------------------------------------------------------
3992 2007-04-01 Cluster File Systems, Inc. <info@clusterfs.com>
3994 * Support for kernels:
3995 2.4.21-47.0.1.EL (RHEL 3)
3996 2.6.5-7.283 (SLES 9)
3997 2.6.9-42.0.10.EL (RHEL 4)
3998 2.6.12.6 vanilla (kernel.org)
3999 2.6.16.27-0.9 (SLES 10)
4000 * Recommended e2fsprogs version: 1.39.cfs5
4002 * Note that reiserfs quotas are disabled on SLES 10 in this kernel
4006 Frequency : occasional, depends on client load and configuration
4007 Bugzilla : 12181, 12203
4008 Description: data loss for recently-modified files
4010 Details : In some cases it is possible that recently written or created
4011 files may not be written to disk in a timely manner (this should
4012 normally be within 30s unless client IO load is very high).
4013 The problem appears as zero-length files or files that are a
4014 multiple of 1MB in size after a client crash or client eviction
4015 that are missing data at the end of the file.
4017 This problem is more likely to be hit on clients where files are
4018 repeatedly created and unlinked in the same directory, clients
4019 have a large amount of RAM, have many CPUs, the filesystem has
4020 many OSTs, the clients are rebooted frequently, and/or the files
4021 are not accessed by other nodes after being written.
4023 The presence of the problem can be detected by looking at
4024 /proc/sys/fs/inode-state. If the first number (nr_inodes) is
4025 smaller than the second (nr_unused) then dirty files will not
4026 be flushed automatically to disk. "sync; sleep 10" should be
4027 run several times on the node before unmounting it to update
4028 Lustre (this is also safe to run on nodes without this problem).
4030 There is also a related kernel bug in the RHEL4 4 2.6.9 kernel
4031 that can cause this same problem, so customers using that kernel
4032 also need to update the kernel in addition to Lustre. In order
4033 to properly fix this bug, the RHEL3 2.4.21 kernel is also updated.
4035 It is normal that files written just before a client crash (less
4036 than 30s) may not yet have been flushed to disk, even for local
4040 Frequency : frequent on thin XT3 nodes
4042 Description: UUID collision on thin XT3 Linux nodes
4043 Details : UUIDs on Compute Node Linux XT3 nodes were not generated
4044 randomly, since we relied on an insufficiently-seeded PRNG.
4049 Description: OSS hangs after "All ost request buffers busy"
4050 Details : A deadlock between quota and journal operations caused OSS
4051 hangs after printing "All ost request buffers busy."
4054 Frequency : always on liblustre builds
4056 Description: Cleanup compiler warnings on liblustre
4059 Frequency : always on liblustre builds on XT3
4061 Description: LC_CONFIG_CDEBUG don't run while build liblustre on XT3.
4065 Description: Addition of EXT3_FEATURE_RO_COMPAT_DIR_NLINKS flag for
4066 > 32000 subdirectories
4067 Details : Add EXT3_FEATURE_RO_COMPAT_DIR_NLINK flag to
4068 EXT3_FEATURE_RO_COMPAT_SUPP. This flag will be set whenever
4069 subdirectory count crosses 32000. This will aid e2fsck to
4070 correctly handle more than 32000 subdirectories.
4073 Frequency : liblustre (e.g. catamount) on a large cluster with >= 8 OSTs/OSS
4075 Description: System hang on startup
4076 Details : This bug allowed the liblustre (e.g. catamount) client to
4077 return to the app before handling all startup RPCs. This
4078 could leave the node unresponsive to lustre network traffic
4079 and manifested as a server ptllnd timeout.
4081 Severity : enhancement
4083 Description: Add "/proc/sys/lustre/debug_peer_on_timeout"
4084 (liblustre envirable: LIBLUSTRE_DEBUG_PEER_ON_TIMEOUT)
4085 boolean to control whether to print peer debug info when a
4086 client's RPC times out.
4091 Description: make O_SYNC working on 2.6 kernels
4092 Details : 2.6 kernels use different method for mark pages for write,
4093 so need add a code to lustre for O_SYNC work.
4098 Description: Failure to close file and release space on NFS
4099 Details : Put inode details into lock acquired in ll_intent_file_open.
4100 Use mdc_intent_lock in ll_intent_open to properly
4101 detect all kind of errors unhandled by mdc_enqueue.
4106 Description: proc file read during shutdown sometimes raced obd removal,
4108 Details : Add lock to prevent obd access after proc file removal.
4111 Frequency : Only for files larger than 4GB on 32-bit clients.
4113 Description: improperly doing page alignment of locks
4114 Details : Modify lustre core code to use CFS_PAGE_* defines instead of
4115 PAGE_*. Make CFS_PAGE_MASK a 64-bit mask.
4120 Description: RPCs being resent when they shouldn't be
4121 Details : Some RPCs that should not be resent are being resent. This
4122 can cause inconsistencies in the RPC state machine. Do not
4123 resend such requests.
4126 Frequency : rare, only with NFS export
4128 Description: Crash on NFS re-export node
4129 Details : under very unusual load conditions an assertion is hit in
4130 ll_intent_file_open()
4133 Frequency : only if OST filesystem is corrupted
4135 Description: client incorrectly hits assertion in ptlrpc_replay_req()
4136 Details : for a short time RPCs with bulk IO are in the replay list,
4137 but replay of bulk IOs is unimplemented. If the OST filesystem
4138 is corrupted due to disk cache incoherency and then replay is
4139 started it is possible to trip an assertion. Avoid putting
4140 committed RPCs into the replay list at all to avoid this issue.
4145 Description: large O_DIRECT requests fail under memory pressure/fragmentation
4146 Details : Large single O_DIRECT read and write calls can fail to allocate
4147 a sufficiently large buffer to process the request. In case of
4148 allocation failure the allocation is retried with a smaller
4149 buffer and broken into smaller requests.
4151 Severity : enhancement
4153 Description: Add -o localflock option to simulate old noflock behaviour.
4154 Details : This will achieve local-only flock/fcntl locks coherentness.
4159 Description: versioning check is incomplete
4160 Details : Checking the version difference of client vs. server, report
4161 error if the gap is too big.
4166 Description: add support PG_writeback bit
4167 Details : add support for PG_writeback bit for Lustre, for more carefull
4168 work with page cache in 2.6 kernel. This also fix some deadlocks
4169 and remove hack for work O_SYNC with 2.6 kernel.
4171 Severity : enhancement
4173 Description: Add uninit_groups feature to ldiskfs2 to speed up e2fsck
4174 Details : The uninit_groups feature works in conjunction with the kernel
4175 filesystem code (ldiskfs2 only) and e2fsprogs-1.39-cfs6 to speed
4176 up the pass1 processing of e2fsck. This is a read-only feature
4177 in ldiskfs2 only, so older kernels and current ldiskfs cannot
4178 mount filesystems that have had this feature enabled.
4180 Severity : enhancement
4182 Description: Improve multi-block allocation algorithm to avoid fragmentation
4183 Details : The mballoc3 code (ldiskfs2 only) adds new mechanisms to improve
4184 allocation locality and avoid filesystem fragmentation.
4186 ------------------------------------------------------------------------------
4188 2007-02-09 Cluster File Systems, Inc. <info@clusterfs.com>
4190 * Support for kernels:
4191 2.6.9-42.0.3.EL (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)
4195 2.6.16.21-0.8 (SLES10)
4196 * Recommended e2fsprogs version: 1.39.cfs2-0
4198 * The backwards-compatible /proc/sys/portals symlink has been removed
4199 in this release. Before upgrading, please ensure that you change
4200 any configuration scripts or /etc/sysctl.conf files that access
4201 /proc/sys/portals/* or sysctl portals.* to use the corresponding
4202 entry in /proc/sys/lnet or sysctl lnet.*. This change can be made
4203 in advance of the upgrade on any system running Lustre 1.4.6 or
4204 newer, since /proc/sys/lnet was added in that version.
4205 * Note that reiserfs quotas are disabled on SLES 10 in this kernel
4209 Frequency : only when quota is used
4211 Description: avoid scanning export list for quota master
4212 Details : Change the algorithms to avoid scanning export list in order
4213 to improve the efficiency.
4216 Frequency : MDS failover only, very rarely
4218 Description: "went back in time" messages on mds failover
4219 Details : The greatest transno may be lost when the current operation
4220 finishes with an error (transno==0) and the client's last_rcvd
4221 record is over-written. Save the greatest transno in the
4222 mds_last_transno for this case.
4225 Frequency : always for specific kernels and striping counts
4227 Description: client may get "Matching packet too big" without ACL support
4228 Details : Clients compiled without CONFIG_FS_POSIX_ACL get an error message
4229 when trying to access files in certain configurations. The
4230 clients should in fact be denied when mounting because they do
4231 not understand ACLs.
4234 Frequency : Cray XT3 with more than 4000 clients and multiple jobs
4236 Description: many clients connecting with IO in progress causes connect timeouts
4237 Details : Avoid synchronous journal commits to avoid delays caused by many
4238 clients connecting/disconnecting when bulk IO is in progress.
4239 Queue liblustre connect requests on OST_REQUEST_PORTAL instead of
4240 OST_IO_PORTAL to avoid delays behind potentially many pending
4244 Frequency : occasionally with multiple writers to a single file
4246 Description: shared writes to file may result in wrong size reported by stat()
4247 Details : Allow growing of kms when extent lock is cancelled
4250 Frequency : always with random mmap IO to multi-striped file
4252 Description: mmap write might be lost if we are writing to a 'hole' in stripe
4253 Details : Only if the hole is at the end of OST object so that kms is too
4254 small. Fix is to increase kms accordingly in ll_nopage.
4257 Frequency : rare, only if OST filesystem is inconsistent with MDS filesystem
4259 Description: writes to a missing object would leak memory on the OST
4260 Details : If there is an inconsistency between the MDS and OST filesystems,
4261 such that the MDS references an object that doesn't exist, writes
4262 to that object will leak memory due to incorrect cleanup in the
4263 error handling path, eventually running out of memory on the OST.
4268 Description: Creating too long symlink causes lustre errors
4269 Details : Check symlink and name lengths before sending requests to MDS.
4272 Frequency : only if flock is enabled (not on by default)
4274 Description: posix locks not released on fd closure on 2.6.9+
4275 Details : We failed to add posix locks to list of inode locks on 2.6.9+
4276 kernels, this caused such locks not to be released on fd close and
4277 then assertions on fs unmount about still used locks.
4280 Frequency : MDS failover only, very rarely
4282 Description: clients may get ASSERTION(granted_lock != NULL)
4283 Details : When request was taking a long time, and a client was resending
4284 a getattr by name lock request. The were multiple lock requests
4285 with the same client lock handle and
4286 mds_getattr_name->fixup_handle_for_resent_request found one of the
4287 lock handles but later failed with ASSERTION(granted_lock != NULL).
4292 Description: handle->h_buffer_credits > 0, assertion failure
4293 Details : h_buffer_credits is zero after truncate, causing assertion
4294 failure. This patch extends the transaction or creates a new
4298 Frequency : NFS re-export or patchless client
4299 Bugzilla : 11179, 10796
4300 Description: Crash on NFS re-export node (__d_move)
4301 Details : We do not want to hash the dentry if we don't have a lock.
4302 But if this dentry is later used in d_move, we'd hit uninitialised
4303 list head d_hash, so we just do this to init d_hash field but
4304 leave dentry unhashed.
4307 Frequency : NFS re-export or patchless client
4309 Description: NFS exports has problem with symbolic link
4310 Details : lustre client didn't properly install dentry when re-exported
4311 to NFS or running patchless client.
4314 Frequency : NFS re-export or patchless client
4316 Description: Various nfs/patchless fixes.
4317 Details : fixes reuse disconected alias for lookup process - this fixes
4318 warning "find_exported_dentry: npd != pd",
4319 fix permission error with open files at nfs.
4320 fix apply umask when do revalidate.
4323 Frequency : occasional
4325 Description: Crash on NFS re-export node
4326 Details : calling clear_page() on the wrong pointer triggered oops in
4327 generic_mapping_read().
4330 Frequency : rarely, using O_DIRECT IO
4332 Description: unaligned directio crashes client with LASSERT
4333 Details : check for unaligned buffers before trying any requests.
4336 Frequency : rarely, using CFS RAID5 patches in non-standard kernel series
4338 Description: stale data returned from RAID cache
4339 Details : If only a small amount of IO is done to the RAID device before
4340 reading it again it is possible to get stale data from the RAID
4341 cache instead of reading it from disk.
4344 Frequency : always for sles10 kernel
4346 Description: sles10 support
4347 Details : ll_follow_link: compile fixes and using of nd_set_link
4348 under newer kernels.
4351 Frequency : depends on arch, kernel and compiler version, always on sles10
4354 Description: recursive or deep enough symlinks cause stack overflow
4355 Details : getting rid of large stack-allocated variable in
4359 Frequency : depends on hardware
4361 Description: lustre write performance loss in the SLES10 kernel
4362 Details : the performance loss is caused by using of write barriers in the
4363 ext3 code. The SLES10 kernel turns barrier support on by
4364 default. The fix is to undo that change for ldiskfs.
4366 ------------------------------------------------------------------------------
4368 2006-12-09 Cluster File Systems, Inc. <info@clusterfs.com>
4370 * Support for kernels:
4371 2.6.9-42.0.3EL (RHEL 4)
4372 2.6.5-7.276 (SLES 9)
4373 2.4.21-47.0.1.EL (RHEL 3)
4374 2.6.12.6 vanilla (kernel.org)
4378 Frequency : quota enabled and large files being deleted
4380 Description: releasing more than 4GB of quota at once hangs OST
4381 Details : If a user deletes more than 4GB of files on a single OST it
4382 will cause the OST to spin in an infinite loop. Release
4383 quota in < 4GB chunks, or use a 64-bit value for 1.4.7.1+.
4388 Description: statfs data retrieved from /proc may be stale or zero
4389 Details : When reading per-device statfs data from /proc, in the
4390 {kbytes,files}_{total,free,avail} files, it may appear
4391 as zero or be out of date.
4394 Frequency : systems with MD RAID1 external journal devices
4396 Description: lconf's call to blkid is confused by RAID1 journal devices
4397 Details : Use the "blkid -l" flag to locate the MD RAID device instead
4398 of returning all block devices that match the journal UUID.
4401 Frequency : always, for aggregate stripe size over 4GB
4403 Description: "lfs setstripe" fails assertion when setting 4GB+ stripe width
4404 Details : Using "lfs setstripe" to set stripe size * stripe count over 4GB
4405 will fail the kernel with "ASSERTION(lsm->lsm_xfersize != 0)"
4408 Frequency : always if "lfs find" used on a local file/directory
4410 Description: "lfs find" segfaults if used on a local file/directory
4411 Details : The case where a directory component was not specified wasn't
4412 handled correctly. Handle this properly.
4415 Frequency : always on ppc64
4417 Description: the write to an ext3 filesystem mounted with mballoc got stuck
4418 Details : ext3_mb_generate_buddy() uses find_next_bit() which does not
4419 perform endianness conversion.
4422 Frequency : rarely (truncate to non-zero file size after write under load)
4423 Bugzilla : 10730, 10687
4424 Description: Files padded with zeros to next 4K multiple
4425 Details : With filesystems mounted using the "extents" option (2.6 kernels)
4426 it is possible that files that are truncated to a non-zero size
4427 immediately after being written are filled with zero bytes beyond
4428 the truncated size. No file data is lost.
4430 Severity : enhancement
4432 Description: Allow recovery/failover for liblustre clients.
4433 Details : liblustre clients were unaware of failover configurations until
4436 Severity : enhancement
4438 Description: user file locks should fail when not mounting with flock option
4439 Details : Set up an error-returning stub in ll_file_operations.lock field
4440 to prevent incorrect behaviour when client is mounted without
4441 flock option. Also, set up properly f_op->flock field for
4445 Frequency : always on ia64
4447 Description: "lfs df" loops on printing out MDS statfs information
4448 Details : The obd_ioctl_data was not initialized and in some systems
4449 this caused a failure during the ioctl that did not return
4450 an error. Initialize the struct and return an error on failure.
4453 Frequency : SLES 9 only
4455 Description: Error of copying files with lustre special EAs as root
4456 Details : Client side always return success for setxattr call for lustre
4457 special xattr (currently only "trusted.lov").
4460 Frequency : rarely on clusters with both ia64+i386 clients
4462 Description: ia64+i686 clients doing shared IO on the same file may LBUG
4463 Details : In rare cases when both ia64+i686 (or other mixed-PAGE_SIZE)
4464 clients are doing concurrent writes to the same file it is
4465 possible that the ia64 clients may LASSERT because the OST
4466 extent locks are not PAGE_SIZE aligned. Ensure that grown
4467 locks are always aligned on the request boundary.
4470 Frequency : specific use, occasional
4472 Description: Overwriting in use executable truncates on-disk binary image
4473 Details : If one node attempts to overwrite an executable in use by
4474 another node, we now correctly return ETXTBSY instead of
4475 truncating the file.
4477 Severity : enhancement
4479 Description: Async OSC create to avoid the blocking unnecessarily.
4480 Details : If a OST has no remain object, system will block on the creating
4481 when need to create a new object on this OST. Now, ways use
4482 pre-created objects when available, instead of blocking on an
4483 empty osc while others are not empty. If we must block, we block
4484 for the shortest possible period of time.
4489 Description: chmod on Lustre root is propagated to other clients
4490 Details : Re-validate root's dentry in ll_lookup_it to avoid having it
4491 invalid by the follow_mount time.
4494 Frequency : liblustre clients only
4496 Description: Race in 'instant cancel' lock handling could lead to such locks
4497 never to be granted in case of SMP MDS
4498 Details : Do not destroy not yet granted but cbpending locks in
4502 Frequency : replay/resend of open
4504 Description: non null lock assetion failure in mds_intent_policy
4505 Details : Trying to replay/resend lockless open requests resulted in
4506 mds_open() returning 0 with no lock. Now it sets a flag if
4507 a lock is going to be returned.
4509 Severity : enhancement
4511 Description: Checksum enhancements
4512 Details : New checksum enhancements allow for resending RPCs that failed
4515 Severity : enhancement
4517 Description: Tunables on number of dirty pages in cacche
4518 Details : Allow to set limit on number of dirty pages cached.
4523 Description: client crash on unmount - lock still has references
4524 Details : In some error handling cases it was possible to leak a lock
4525 reference on a client while accessing a file. This was not
4526 harmful to the client during operation, but would cause the
4527 client to crash when the filesystem is unmounted.
4530 Frequency : specific case, rare
4532 Description: ETXTBSY on mds though file not in use by client
4533 Details : ETXTBSY is no longer incorrectly returned when attempting to
4534 chmod or chown a directory that the user previously tried to
4535 execute or a currently-executing binary.
4538 Frequency : extremely rare except on liblustre-based clients
4540 Description: Lustre space not freed when files are deleted
4541 Details : Clean up open-unlinked files after client eviction. Previously
4542 the unlink was skipped and the files remained as orphans.
4547 Description: OST failure "would be an LBUG" in waiting_locks_callback()
4548 Details : In some cases it was possible to send a blocking callback to
4549 a client doing a glimpse, even though that client didn't get
4550 a lock granted. When the glimpse lock is cancelled on the OST
4551 the freed lock is left on the waiting list and corrupted the list.
4554 Frequency : all core dumps
4556 Description: Broke core dumps to lustre
4557 Details : Negative dentry may be unhashed if parent does not have UPDATE
4558 lock, but some callers, e.g. do_coredump, expect dentry to be
4559 hashed after successful create, hash it in ll_create_it.
4561 ------------------------------------------------------------------------------
4563 2006-09-13 Cluster File Systems, Inc. <info@clusterfs.com>
4565 * Support for kernels:
4566 2.6.9-42.0.2.EL (RHEL 4)
4567 2.6.5-7.276 (SLES 9)
4568 2.4.21-40.EL (RHEL 3)
4569 2.6.12.6 vanilla (kernel.org)
4573 Frequency : always on RHEL 3
4575 Description: Number of open files grows over time
4576 Details : The number of open files grows over time, whether or not
4577 Lustre is started. This was due to a filp leak introduced
4578 by one of our kernel patches.
4580 ------------------------------------------------------------------------------
4582 2006-08-20 Cluster File Systems, Inc. <info@clusterfs.com>
4584 * Support for kernels:
4585 2.6.9-42.EL (RHEL 4)
4586 2.6.5-7.267 (SLES 9)
4587 2.4.21-40.EL (RHEL 3)
4588 2.6.12.6 vanilla (kernel.org)
4593 Bugzilla : 5719, 9635, 9792, 9684
4594 Description: OST (or MDS) trips assertions in (re)connection under heavy load
4595 Details : If a server is under heavy load and cannot reply to new
4596 connection requests before the client resends the (re)connect,
4597 the connection handling code can behave badly if two service
4598 threads are concurrently handing separate (re)connections from
4599 the same client. Add better locking to the connection handling
4600 code, and ensure that only a single connection will be processed
4601 for a given client UUID, even if the lock is dropped.
4603 Severity : enhancement
4605 Description: add TCP zero-copy support to kernel
4606 Details : Add support to the kernel TCP stack to allow zero-copy bulk
4607 sends if the hardware supports scatter-gather and checksumming.
4608 This allows socklnd to do client-write and server-read more
4609 efficiently and reduce CPU utilization from skbuf copying.
4612 Frequency : only if NFS exporting from client
4614 Description: NULL pointer deref in ll_iocontrol() if chattr mknod file
4615 Details : If setting attributes on a file created under NFS that had
4616 never been opened it would be possible to oops the client
4617 if the file had no objects.
4621 Bugzilla : 9326, 10402, 10897
4622 Description: client crash in ptlrpcd_wake() thread when sending async RPC
4623 Details : It is possible that ptlrpcd_wake() dereferences a freed async
4624 RPC. In rare cases the ptlrpcd thread alread processed the RPC
4625 before ptlrpcd_wake() was called and the request was freed.
4628 Frequency : always for liblustre
4630 Description: liblustre client does MDS+OSTs setattr RPC for each write
4631 Details : When doing a write from a liblustre client, the client
4632 incorrectly issued an RPC to the MDS and each OST the file was
4633 striped over in order to update the timestamps. When writing
4634 with small chunks and many clients this could overwhelm the MDS
4635 with RPCs. In all cases it would slow down the write because
4636 these RPCs are unnecessary.
4638 Severity : enhancement
4640 Description: allow number of MDS service threads to be changed at module load
4641 Details : It is now possible to change the number of MDS service threads
4642 running. Adding "options mds mds_num_threads={N}" to the MDS's
4643 /etc/modprobe.conf will set the number of threads for the next
4644 time Lustre is restarted (assuming the "mds" module is also
4645 reloaded at that time). The default number of threads will
4646 stay the same, 32 for most systems.
4651 Description: OST crash if filesystem is unformatted or corrupt
4652 Details : If an OST is started on a device that has never been formatted
4653 or if the filesystem is corrupt and cannot even mount then the
4654 error handling cleanup routines would dereference a NULL pointer.
4659 Description: NULL pointer deref in llap_from_page.
4660 Details : get_cache_page_nowait can return a page with NULL (or otherwise
4661 incorrect) mapping if the page was truncated/reclaimed while it was
4662 searched for. Check for this condition and skip such pages when
4663 doing readahead. Introduce extra check to llap_from_page() to
4664 verify page->mapping->host is non-NULL (so page is not anonymous).
4667 Frequency : Sometimes when using sys_sendfile
4669 Description: "page not covered by a lock" warnings from ll_readpage
4670 Details : sendfile called ll_readpage without right page locks present.
4671 Now we introduced ll_file_sendfile that does necessary locking
4672 around call to generic_file_sendfile() much like we do in
4676 Frequency : with certain MDS communication failures at client mount time
4678 Description: NULL pointer deref after failed client mount
4679 Details : a client connection request may delayed by the network layer
4680 and not be sent until after the PTLRPC layer has timed out the
4681 request. If the client fails the mount immediately it will try
4682 to clean up before the network times out the request. Add a
4683 reference from the request import to the obd device and delay
4684 the cleanup until the network drops the request.
4687 Frequency : occasionally during client (re)connect
4689 Description: assertion failure during client (re)connect
4690 Details : processing a client connection request may be delayed by the
4691 client or server longer than the client connect timeout. This
4692 causes the client to resend the connection request. If the
4693 original connection request is replied in this interval, the
4694 client may trip an assertion failure in ptlrpc_connect_interpret()
4695 which thought it would be the only running connect process.
4698 Frequency : only with obd_echo servers and clients that are rebooted
4700 Description: kernel BUG accessing uninitialized data structure
4701 Details : When running an obd_echo server it did not start the ping_evictor
4702 thread, and when a client was evicted an uninitialized data
4703 structure was accessed. Start the ping_evictor in the RPC
4704 service startup instead of the OBD startup.
4706 Severity : enhancement
4707 Bugzilla : 10193 (patchless)
4708 Description: Remove dependency on various unexported kernel interfaces.
4709 Details : No longer need reparent_to_init, exit_mm, exit_files,
4710 sock_getsockopt, filemap_populate, FMODE_EXEC, put_filp.
4713 Frequency : rare (only users of deprecated and unsupported LDAP config)
4715 Description: write_conf for zeroconf mount queried LDAP incorrectly for client
4716 Details : LDAP apparently contains 'lustreName' attributes instead of
4717 'name'. A simple remapping of the name is sufficient.
4720 Frequency : rare (only with non-default dump_on_timeout debug enabled)
4722 Description: waiting_locks_callback trips kernel BUG if client is evicted
4723 Details : Running with the dump_on_timeout debug flag turned on makes
4724 it possible that the waiting_locks_callback() can try to dump
4725 the Lustre kernel debug logs from an interrupt handler. Defer
4726 this log dumping to the expired_lock_main() thread.
4728 Severity : enhancement
4730 Description: Support NFS exporting on 2.6 kernels.
4731 Details : Implement non-rawops metadata methods for NFS server to use without
4732 changing NFS server code.
4735 Frequency : very rare (synthetic metadata workload only)
4737 Description: two racing renames might cause an MDS thread to deadlock
4738 Details : Running the "racer" program may cause one MDS thread to rename
4739 a file from being the source of a rename to being the target of
4740 a rename at exactly the same time that another thread is doing
4741 so, and the second thread has already enqueued these locks after
4742 doing a lookup of the target and is trying to relock them in
4743 order. Ensure that we don't try to re-lock the same resource.
4746 Frequency : only very large systems with liblustre clients
4748 Description: slow eviction of liblustre clients with the "evict_by_nid" RPC
4749 Details : Use asynchronous set_info RPCs to send the "evict_by_nid" to
4750 all OSTs in parallel. This allows the eviction of stale liblustre
4751 clients to proceed much faster than if they were done in series,
4752 and also offers similar improvements for other set_info RPCs.
4757 Description: excessive CPU usage during initial read phase on client
4758 Details : During the initial read phase on a client, it would agressively
4759 retry readahead on the file, consuming too much CPU and impacting
4760 performance (since 1.4.5.8). Improve the readahead algorithm
4761 to avoid this, and also improve some other common cases (read
4762 of small files in particular, where "small" is files smaller than
4763 /proc/fs/lustre/llite/*/max_read_ahead_whole_mb, 2MB by default).
4768 Description: MDS crash when receiving packet with unknown intent.
4769 Details : Do not LBUG in unknown intent case, just return -EFAULT
4771 Severity : enhancement
4772 Bugzilla : 9293, 9385
4773 Description: MDS RPCs are serialised on client. This is unnecessary for some.
4774 Details : Do not serialize getattr (non-intent version) and statfs.
4777 Frequency : occasional, when OST network is overloaded/intermittent
4779 Description: client evicted by OST after bulk IO timeout
4780 Details : If a client sends a bulk IO request (read or write) the OST
4781 may evict the client if it is unresposive to its data GET/PUT
4782 request. This is incorrect if the network is overloaded (takes
4783 too long to transfer the RPC data) or dropped the OST GET/PUT
4784 request. There is no need to evict the client at all, since
4785 the pinger and/or lock callbacks will handle this, and the
4786 client can restart the bulk request.
4789 Frequency : Always when mmapping file with no objects
4791 Description: client crashes when mmapping file with no objects
4792 Details : Check that we actually have objects in a file before doing any
4793 operations on objects in ll_vm_open, ll_vm_close and
4799 Description: Request leak when working with deleted CWD
4800 Details : Introduce advanced request refcount tracking for requests
4801 referenced from lustre intent.
4803 Severity : Enhancement
4805 Description: Cache open file handles on client.
4806 Details : MDS now will return special lock along with openhandle, if
4807 requested and client is allowed to hold openhandle, even if unused,
4808 until such a lock is revoked. Helps NFS a lot, since NFS is opening
4809 closing files for every read/write openration.
4811 Severity : Enhancement
4813 Description: Cache open negative dentries on client when possible.
4814 Details : Guard negative dentries with UPDATE lock on parent dir, drop
4815 negative dentries on lock revocation.
4820 Description: Remounting a client read-only wasn't possible with a zconf mount
4821 Details : It wasn't possible to remount a client read-only with llmount.
4823 Severity : enhancement
4824 Description: Include MPICH 1.2.6 Lustre ADIO interface patch
4825 Details : In lustre/contrib/ or /usr/share/lustre in RPM a patch for
4826 MPICH is included to add Lustre-specific ADIO interfaces.
4827 This is based closely on the UFS ADIO layer and only differs
4828 in file creation, in order to allow the OST striping to be set.
4829 This is user-contributed code and not supported by CFS.
4834 Description: extended inode attributes (immutable, append-only) work improperly
4835 when 2.4 and 2.6 kernels are used on client/server or vice versa
4836 Details : Introduce kernel-independent values for these flags.
4838 Severity : enhancement
4841 Description: Allow fractional MB tunings for lustre in /proc/ filesystem.
4842 Details : Many of the /proc/ tunables can only be tuned at a megabyte
4843 granularity. Now, Fractional MB granularity is be supported,
4844 this is very useful for low memory system.
4846 Severity : enhancement
4848 Description: Getattr by fid
4849 Details : Getting a file attributes by its fid, obtaining UPDATE|LOOKUP
4850 locks, avoids extra getattr rpc requests to MDS, allows '/' to
4851 have locks and avoids getattr rpc requests for it on every stat.
4854 Frequency : Always, for filesystems larger than 2TB
4856 Description: ldiskfs crash at mount for filesystem larger than 2TB with mballoc
4857 Details : Kenrel kmalloc limits allocations to 128kB and this prevents
4858 filesystems larger than 2TB to be mounted with mballoc enabled.
4861 Frequency : Always, for 32-bit kernel without CONFIG_LBD and filesystem > 2TB
4863 Description: filesystem corruption for non-standard kernels and very large OSTs
4864 Details : If a 32-bit kernel is compiled without CONFIG_LBD enabled and a
4865 filesystems larger than 2TB is mounted then the kernel will
4866 silently corrupt the start of the filesystem. CONFIG_LBD is
4867 enabled for all CFS-supported kernels, but the possibility of
4868 this happening with a modified kernel config exists.
4870 Severity : enhancement
4872 Description: add client O_DIRECT support for 2.6 kernels
4873 Details : It is now possible to do O_DIRECT reads and writes to files
4874 in the Lustre client mountpoint on 2.6 kernel clients.
4876 Severity : enhancement
4878 Description: parallel glimpse, setattr, statfs, punch, destroy requests
4879 Details : Sends glimpse, setattr, statfs, punch, destroy requests to OSTs in
4880 parallel, not waiting for response from every OST before sending
4881 a rpc to the next OST.
4886 Description: setattr vs write race when updating file timestamps
4887 Details : Client processes that update a file timestamp into the past
4888 right after writing to the file (e.g. tar) it is possible that
4889 the updated file modification time can be reset to the current
4890 time due to a race between processing the setattr and write RPC.
4892 Severity : enhancement
4894 Description: Bring 'lfs find' closer in line with regular Linux find.
4895 Details : lfs find util supports -atime, -mtime, -ctime, -maxdepth, -print,
4896 -print0 options and obtains all the needed info through the lustre
4899 Severity : enhancement
4901 Description: support up to 1024 configured devices on one node
4902 Details : change obd_dev array from statically allocated to dynamically
4903 allocated structs as they are first used to reduce memory usage
4908 Description: Flush dirty partially truncated pages during truncate
4909 Details : Immediatelly flush partially truncated pages in filter_setattr,
4910 this way we completely avoid having any pages in page cache on OST
4911 and can retire ugly workarounds during writes to flush such pages.
4916 Description: i_sem vs transaction deadlock in mds_obd_destroy during unlink.
4917 Details : protect inode from truncation within vfs_unlink() context
4918 just take a reference before calling vfs_unlink() and release it
4919 when parent's i_sem is free.
4922 Frequency : always, if extents are used on OSTs
4924 Description: index ei_leaf_hi (48-bit extension) is not zeroed in extent index
4925 Details : OSTs using the extents format would not zero the high 16 bits of
4926 the index physical block number. This is not a problem for any
4927 OST filesystems smaller than 16TB, and no kernels support ext3
4928 filesystems larger than 16TB yet. This is fixed in 1.4.7 (all
4929 new/modified files) and can be fixed for existing filesystems
4930 with e2fsprogs-1.39-cfs1.
4935 Description: import connection selection may be incorrect if timer wraps
4936 Details : Using a 32-bit jiffies timer with HZ=1000 may cause backup
4937 import connections to be ignored if the 32-bit jiffies counter
4938 wraps. Use a 64-bit jiffies counter.
4941 Frequency : during server recovery
4943 Description: crash after server is denying duplicate export
4944 Details : If clients are resending connect requests to the server, the
4945 server refuses to allow a client to connect multiple times.
4946 Fixed a bug in the handling of this case.
4949 Frequency : very large clusters immediately after boot
4951 Description: LNET request buffers exhausted under heavy short-term load
4952 Details : If a large number of client requests are generated on a service
4953 that has previously never seen so many requests it is possible
4954 that the request buffer growth cannot keep up with the spike in
4955 demand. Instead of dropping incoming requests, they are held in
4956 the LND until the RPC service can accept more requests.
4959 Frequency : Sometimes during replay
4961 Description: Assertion failure in ll_local_open after replay.
4962 Details : If replay happened on an open request reply before we were able
4963 to set replay handler, reply will become not swabbed tripping the
4964 assertion in ll_local_open. Now we set the handler right after
4965 recognising of open request
4968 Frequency : very rare
4970 Description: kernel reports "badness in vsnprintf"
4971 Details : Reading from the "recovery_status" /proc file in small chunks
4972 may cause a negative length in lprocfs_obd_rd_recovery_status()
4973 call to vsnprintf() (which is otherwise harmless). Exit early
4974 if there is no more space in the output buffer.
4976 Severity : enhancement
4978 Description: clear OBD RPC statistics by writing to them
4979 Details : It is now possible to clear the OBD RPC statistics by writing
4980 to the "stats" file.
4985 Description: Client mtime is not the same on different clients after utimes
4986 Details : In some cases, the client was using the utimes() syscall on
4987 a file cached on another node. The clients now validate the
4988 ctime from the MDS + OSTs to determine which one is right.
4993 Description: Inability to activate failout mode
4994 Details : lconf script incorrectly assumed that in python string's numeric
4995 value is used in comparisons.
4998 Frequency : always with multiple stripes per file
5000 Description: Inefficient object allocation for mutli-stripe files
5001 Details : When selecting which OSTs to stripe files over, for files with
5002 a stripe count that divides evenly into the number of OSTs,
5003 the MDS is always picking the same starting OST for each file.
5004 Return the OST selection heuristic to the original design.
5009 Description: mount failures may take full timeout to return an error
5010 Details : Under some heavy load conditions it is possible that a
5011 failed mount can wait for the full obd_timeout interval,
5012 possibly several minutes, before reporting an error.
5013 Instead return an error as soon as the status is known.
5015 ------------------------------------------------------------------------------
5017 2006-02-14 Cluster File Systems, Inc. <info@clusterfs.com>
5019 * WIRE PROTOCOL CHANGE. This version of Lustre networking WILL NOT
5020 INTEROPERATE with older versions automatically. Please read the
5021 user documentation before upgrading any part of a live system.
5022 * WARNING: Lustre networking configuration changes are required with
5023 this release. See https://bugzilla.clusterfs.com/show_bug.cgi?id=10052
5026 * Support for kernels:
5027 2.6.9-22.0.2.EL (RHEL 4)
5028 2.6.5-7.244 (SLES 9)
5029 2.6.12.6 vanilla (kernel.org)
5032 Severity : enhancement
5033 Bugzilla : 7981/8208
5034 Description: Introduced Lustre Networking (LNET)
5035 Details : LNET is new networking infrastructure for Lustre, it includes
5036 a reorganized network configuration mode (see the user
5037 documentation for full details) as well as support for routing
5038 between different network fabrics. Lustre Networking Devices
5039 (LNDs) for the supported network fabrics have also been
5040 created for this new infrastructure.
5042 Severity : enhancement
5043 Description: Introduced Access control lists
5044 Details : clients can set ACLs on files and directories in order to have
5045 more fine-grained permissions than the standard Unix UGO+RWX.
5046 The MDS must be started with the "-o acl" mount option.
5048 Severity : enhancement
5049 Description: Introduced filesystem quotas
5050 Details : Administrators may now establish per-user quotas on the
5053 Severity : enhancement
5055 Description: Configuration change for the XT3
5056 The PTLLND is now used to run Lustre over Portals on the XT3
5057 The configure option(s) --with-cray-portals are no longer used.
5058 Rather --with-portals=<path-to-portals-includes> is used to
5059 enable building on the XT3. In addition to enable XT3 specific
5060 features the option --enable-cray-xt3 must be used.
5065 Description: Running on many-way SMP OSTs can trigger oops in llcd_send()
5066 Details : A race between allocating a new llcd and re-getting the llcd_lock
5067 allowed another thread to grab newly-allocated llcd.
5069 Severity : enhancement
5071 Description: 2.6 OST async journal commit and locking fix to improve performance
5072 Details : The filter_direct_io()+filter_commitrw_write() journal commits for
5073 2.6 kernels are now async as they already were in 2.4 kernels so
5074 that they can commit concurrently with the network bulk transfer.
5075 For block-allocated files the filter allocation semaphore is held
5076 to avoid filesystem fragmentation during allocation. BKL lock
5077 removed for 2.6 xattr operations where it is no longer needed.
5082 Description: lconf incorrectly determined whether two IP networks could talk
5083 Details : In some more complicated routing and multiple-network
5084 configurations, lconf will avoid trying to make a network
5085 connection to a disjoint part of the IP space. It was doing the
5086 math incorrectly for one set of cases.
5091 Description: Fix for potential infinite loop processing records in an llog.
5092 Details : If an llog record is corrupted/zeroed, it is possible to loop
5093 forever in llog_process(). Validate the llog record length
5094 and skip the remainder of the block on error.
5097 Frequency : occasional (liblustre only)
5099 Description: liblustre could not open files whose last component is a symlink
5100 Details : sysio_path_walk() would incorrectly pass the open intent to
5101 intermediate path components.
5104 Frequency : rare (liblustre only with non-standard tuning)
5105 Bugzilla : 7201 (7350)
5106 Description: Tuning the MDC DLM LRU size to zero triggers client LASSERT
5107 Details : llu_lookup_finish_locks() tries to set lock data on a lock
5108 after it has been released, only do this for referenced locks
5110 Severity : enhancement
5112 Description: specifying an (invalid) directory default stripe_size of -1
5113 would reset the directory default striping
5114 Details : stripe_size -1 was used internally to signal directory stripe
5115 removal, now use "all default" to signal dir stripe removal
5116 as a directory striping of "all default" is not useful
5119 Frequency : common for large clusters running liblustre clients
5121 Description: doing an ls when liblustre clients are running is slow
5122 Details : sending a glimpse AST to a liblustre client waits for every AST
5123 to time out, as liblustre clients will not respond. Since they
5124 cannot cache data we refresh the OST lock LVB from disk instead.
5126 Severity : enhancement
5128 Description: doing an ls at the same time as file IO can be slow
5129 Details : enqueue and other "small" requests can be blocked behind many
5130 large IO requests. Create a new OST IO portal for non-IO
5131 requests so they can be processed faster.
5134 Frequency : rare (only HPUX clients mounting unsupported re-exported NFS vol)
5136 Description: an HPUX NFS client would get -EACCESS when ftruncate()ing a newly
5137 created file with mode 000
5138 Details : the Linux NFS server relies on an MDS_OPEN_OWNEROVERRIDE hack to
5139 allow an ftruncate() as a non-root user to a file with mode 000.
5140 Lustre now respects this flag to disable mode checks when
5141 truncating a file owned by the user
5144 Frequency : liblustre-only, when liblustre client dies unexpectedly or becomes
5147 Description: Revoking locks from clients that went dead or catatonic might take
5149 Details : New lock flags FL_CANCEL_ON_BLOCK used by liblustre makes
5150 cancellation of such locks instant on servers without waiting for
5151 any reply from clients. Clients drops these locks when cancel
5152 notification from server is received without replying.
5155 Frequency : liblustre-only, when liblustre client dies or becomes busy
5157 Description: Doing ls on Linux clients can take a long time with active
5159 Details : Liblustre client cannot handle ASTs in timely manner, so avoid
5160 granting such locks to it in the first place if possible. Locks
5161 are taken by proxy on the OST during the read or write and
5162 dropped immediately afterward. Add connect flags handling, do
5163 not grant locks to liblustre clients for glimpse ASTs.
5165 Severity : enhancement
5167 Description: Improve read-ahead algorithm to avoid excessive IO for random reads
5168 Details : Existing read-ahead algorithm is tuned for the case of streamlined
5169 sequential reads and behaves badly with applications doing random
5170 reads. Improve it by reading ahead at least read region, and
5171 avoiding excessive large RPC for small reads.
5173 Severity : enhancement
5175 Description: Creating more than 1000 files for a single job may cause a load
5176 imbalance on the OSTs if there are also a large number of OSTs.
5177 Details : qos_prep_create() uses an OST index reseed value that is an
5178 even multiple of the number of available OSTs so that if the
5179 reseed happens in the middle of the object allocation it will
5180 still utilize the OSTs as uniformly as possible.
5185 Description: OST or MDS may oops in ping_evictor_main()
5186 Details : ping_evictor_main() drops obd_dev_lock if deleting a stale export
5187 but doesn't restart at beginning of obd_exports_timed list
5190 Severity : enhancement
5192 Description: improve by-nid export eviction on the MDS and OST
5193 Details : allow multiple exports with the same NID to be evicted at one
5194 time without re-searching the exports list.
5197 Frequency : rare, only with supplementary groups enabled on SMP 2.6 kernels
5199 Description: MDS may oops in groups_free()
5200 Details : in rare race conditions a newly allocated group_info struct is
5201 freed again, and this can be NULL. The 2.4 compatibility code
5202 for groups_free() checked for a NULL pointer, but 2.6 did not.
5205 Frequency : common for liblustre clients doing little filesystem IO
5206 Bugzilla : 9352, 7313
5207 Description: server may evict liblustre clients accessing contended locks
5208 Details : if a client is granted a lock or receives a completion AST
5209 with a blocking AST already set it would not reply to the AST
5210 for LDLM_FL_CANCEL_ON_BLOCK locks. It now replies to such ASTs.
5213 Frequency : lfs setstripe, only systems with more than 160 OSTs
5215 Description: unable to set striping with a starting offset beyond OST 160
5216 Details : llapi_create_file() incorrectly limited the starting stripe
5217 index to the maximum single-file stripe count.
5220 Frequency : LDAP users only
5222 Description: lconf did not handle in-kernel recovery with LDAP properly
5223 Details : lconf/LustreDB get_refs() is searching the wrong namespace
5225 Severity : enhancement
5227 Description: bind OST threads to NUMA nodes to improve performance
5228 Details : all OST threads are uniformly bound to CPUs on a single NUMA
5229 node and do their allocations there to localize memory access
5231 Severity : enhancement
5233 Description: llmount can determine client NID directly from Myrinet (GM)
5234 Details : the client NID code from gmnalnid was moved directly into
5235 llmount, removing the need to use this or specifying the
5236 client NID explicitly when mounting GM clients with zeroconf
5239 Frequency : if client is started with down MDS
5241 Description: if client is started with down MDS mount hangs in ptlrpc_queue_wait
5242 Details : Having an LWI_INTR() wait event (interruptible, but no timeout)
5243 will wait indefinitely in ptlrpc_queue_wait->l_wait_event() after
5244 ptlrpc_import_delayed_req() because we didn't check if the
5245 request was interrupted, and we also didn't break out of the
5246 event loop if there was no timeout
5251 Description: data loss during non-page-aligned writes to a single file from
5252 both multiple nodes and multiple threads on one node at same time
5253 Details : updates to KMS and lsm weren't protected by common lock. Resulting
5254 inconsistency led to false short-reads, that were cached and later
5255 used by ->prepare_write() to fill in partially written page,
5256 leading to data loss.
5259 Frequency : always, if lconf --abort_recovery used
5261 Description: lconf --abort_recovery fails with 'Operation not supported'
5262 Details : lconf was attempting to abort recovery on the MDT device and not
5265 Severity : enhancement
5267 Description: remove cleanup logs
5268 Details : replace lconf-generated cleanup logs with lustre internal
5269 cleanup routines. Eliminates the need for client-cleanup and
5272 Severity : enhancement
5274 Description: add support for EAs (user and system) on lustre filesystems
5275 Details : it is now possible to store extended attributes in the Lustre
5276 client filesystem, and with the user_xattr mount option it
5277 is possible to allow users to store EAs on their files also
5279 Severity : enhancement
5281 Description: Add possibility (config option) to show minimal available OST free
5283 Details : When compiled with --enable-mindf configure option, statfs(2)
5284 (and so, df) will return least minimal free space available from
5285 all OSTs as amount of free space on FS, instead of summary of
5286 free spaces of all OSTs.
5288 Severity : enhancement
5290 Description: do not expand extent locks acquired on OST-side
5291 Details : Modify ldlm_extent_policy() to not expand local locks, acquired
5292 by server: they are not cached anyway.
5295 Frequency : when mmap is used/binaries executed from Lustre
5297 Description: Unmmap pages before throwing them away from read cache.
5298 Details : llap_shrink cache now attempts to unmap pages before discarding
5299 them (if unmapping failed - do not discard). SLES9 kernel has
5300 extra checks that trigger if this unmapping is not done first.
5305 Description: lconf didn't resolve symlinks before checking to see whether a
5306 given mountpoint was already in use
5309 Frequency : when migrating failover services
5310 Bugzilla : 6395, 9514
5311 Description: When migrating a subset of services from a node (e.g. failback
5312 from a failover service node) the remaining services would
5313 time out and evict clients.
5314 Details : lconf --force (implied by --failover) sets the global obd_timeout
5315 to 5 seconds in order to quickly disconnect, but this caused
5316 other RPCs to time out too quickly. Do not change the global
5317 obd_timeout for force cleanup, only set it for DISCONNECT RPCs.
5319 Severity : enhancement
5320 Frequency : if MDS is started with down OST
5321 Bugzilla : 9439,5706
5322 Description: Allow startup/shutdown of an MDS without depending on the
5323 availability of the OSTs.
5324 Details : Asynchronously call mds_lov_synchronize during MDS startup.
5325 Add appropriate locking and lov-osc refcounts for safe
5326 cleaning. Add osc abort_inflight calls in case the
5327 synchronize never started.
5330 Frequency : occasional (Cray XT3 only)
5332 Description: root not authorized to access files in CRAY_PORTALS environment
5333 Details : The client process capabilities were not honoured on the MDS in
5334 a CRAY_PORTALS/CRAY_XT3 environment. If the file had previously
5335 been accessed by an authorized user then root was able to access
5336 the file on the local client also. The root user capabilities
5337 are now allowed on the MDS, as this environment has secure UID.
5340 Frequency : occasional
5342 Description: ldiskfs "too long searching" message happens too often
5343 Details : A debugging message (otherwise harmless) prints too often on
5344 the OST console. This has been reduced to only happen when
5345 there are fragmentation problems on the filesystem.
5350 Description: Division by zero in statfs when all OSCs are inactive
5351 Details : lov_get_stripecnt() returns zero due to incorrect order of checks,
5352 lov_statfs divides by value returned by lov_get_stripecnt().
5356 Bugzilla : 9489, 3273
5357 Description: First write from each client to each OST was only 4kB in size,
5358 to initialize client writeback cache, which caused sub-optimal
5359 RPCs and poor layout on disk for the first writen file.
5360 Details : Clients now request an initial cache grant at (re)connect time
5361 and so that they can start streaming writes to the cache right
5362 away and always do full-sized RPCs if there is enough data.
5363 If the OST is rebooted the client also re-establishes its grant
5364 so that client cached writes will be honoured under the grant.
5369 Description: Slow ls (and stat(2) syscall) on files residing on IO-loaded OSTs
5370 Details : Now I/O RPCs go to different portal number and (presumably) fast
5371 lock requests (and glimses) and other RPCs get their own service
5372 threads pool that should be able to service those RPCs
5375 Severity : enhancement
5377 Description: Ability to exchange lustre version between client and servers and
5378 issue warnings at client side if client is too old. Also for
5379 liblustre clients there is ability to refuse connection of too old
5381 Details : New 'version' field is added to connect data structure that is
5382 filled with version info. That info is later checked by server and
5386 Frequency : rare, liblustre only.
5387 Bugzilla : 9296, 9581
5388 Description: Two simultaneous writes from liblustre at offset within same page
5389 might proceed at the same time overwriting eachother with stale
5391 Details : I/O lock withing llu_file_prwv was released too early, before data
5392 actually was hitting the wire. Extended lock-holding time until
5393 server acknowledges receiving data.
5396 Frequency : extremely rare. Never observed in practice.
5398 Description: avoid generating lustre_handle cookie of 0.
5399 Details : class_handle_hash() generates handle cookies by incrementing
5400 global counter, and can hit 0 occasionaly (this is unlikely, but
5401 not impossible, because initial value of cookie counter is
5402 selected randonly). Value of 0 is used as a sentinel meaning
5403 "unassigned handle" --- avoid it. Also coalesce two critical
5404 sections in this function into one.
5406 Severity : enhancement
5408 Description: allow liblustre clients to delegate truncate locking to OST
5409 Details : To avoid overhead of locking, liblustre client instructs OST to
5410 take extent lock in ost_punch() on client's behalf. New connection
5411 flag is added to handle backward compatibility.
5413 Severity : enhancement
5414 Bugzilla : 4928, 7341, 9758
5415 Description: allow number of OST service threads to be specified
5416 Details : a module parameter allows the number of OST service threads
5417 to be specified via "options ost ost_num_threads={N}" in the
5418 OSS's /etc/modules.conf or /etc/modprobe.conf.
5422 Bugzilla : 6146, 9635, 9895
5423 Description: servers crash with bad pointer in target_handle_connect()
5424 Details : In rare cases when a client is reconnecting it was possible that
5425 the connection request was the last reference for that export.
5426 We would temporarily drop the export reference and get a new
5427 one, but this may have been the last reference and the export
5428 was just destroyed. Get new reference before dropping old one.
5430 Severity : enhancement
5431 Frequency : if client is started with failover MDS
5433 Description: Allow multiple MDS hostnames in the mount command
5434 Details : Try to read the configuration from all specified MDS
5435 hostnames during a client mount in case the "primary"
5438 Severity : enhancement
5440 Description: Stop sending data to evicted clients as soon as possible.
5441 Details : Check if the client we are about to send or are sending data to
5442 was evicted already. (Check is done every second of waiting,
5443 for which l_wait_event interface was extended to allow checking
5444 of exit condition at specified intervals).
5447 Frequency : rare, normally only when NFS exporting is done from client
5449 Description: 'bad disk LOV MAGIC: 0x00000000' error when chown'ing files
5451 Details : Make mds_get_md() recognise empty md case and set lmm size to 0.
5454 Frequency : always, if srand() is called before liblustre initialization
5456 Description: Liblustre uses system PRNG disturbing its usage by user application
5457 Details : Introduce internal to lustre fast and high-quality PRNG for
5458 lustre usage and make liblustre and some other places in generic
5459 lustre code to use it.
5461 Severity : enhancement
5462 Bugzilla : 9477, 9557, 9870
5463 Description: Verify that the MDS configuration logs are updated when xml is
5464 Details : Check if the .xml configuration logs are newer than the config
5465 logs stored on the MDS and report an error if this is the case.
5466 Request --write-conf, or allow starting with --old_conf.
5468 Severity : enhancement
5470 Description: Handle symlinks in the path when checking if Lustre is mounted.
5471 Details : Resolve intermediate symlinks when checking if a client has
5472 mounted a filesystem to avoid duplicate client mounts.
5477 Description: lconf can hit an error exception but still return success.
5478 Details : The lconf command catches the Command error exception at the top
5479 level script context and will exit with the associated exit
5480 status, but doesn't ensure that this exit status is non-zero.
5485 Description: failure of ptlrpc thread startup can cause oops
5486 Details : Starting a ptlrpc service thread can fail if there are a large
5487 number of threads or the server memory is very fragmented.
5488 Handle this without oopsing.
5491 Frequency : always, only if liblustre and non-default acceptor port was used
5493 Description: liblustre cannot connect to servers with non-default acceptor port
5494 Details : tcpnal_set_default_params() was not called and was therefore
5495 ignoring the environment varaible TCPNAL_PORT, as well as other
5496 TCPNAL_ environment variables
5501 Description: two objects could be created on the same OST for a single file
5502 Details : If an OST is down, in some cases it was possible to create two
5503 objects on a single OST for a single file. No problems other
5504 than potential performance impact and spurious error messages.
5508 Bugzilla : 5681, 9562
5509 Description: Client may oops in ll_unhash_aliases
5510 Details : Client dcache may become inconsistent in race condition.
5511 In some cases "getcwd" can fail if the current directory is
5517 Description: Inode refcounting problems in NFS export code
5518 Details : link_raw functions used to call d_instantiate without obtaining
5519 extra inode reference first.
5523 Bugzilla : 9942, 9903
5524 Description: Referencing freed requests leading to crash, memleaks with NFS.
5525 Details : We used to require that call to ll_revalidate_it was always
5526 followed by ll_lookup_it. Also with revalidate_special() it is
5527 possible to call ll_revalidate_it() twice for the same dentry
5528 even if first occurence returned success. This fix changes semantic
5529 between DISP_ENQ_COMPLETE disposition flag to mean there is extra
5530 reference on a request referred from the intent.
5531 ll_intent_release() then releases such a request.
5534 Frequency : rare, normally benchmark loads only
5536 Description: unlinked inodes were kept in memory on the client
5537 Details : If a client is repeatedly creating and unlinking files it
5538 can accumulate a lot of stale inodes in the inode slab cache.
5539 If there is no other client load running this can cause the
5540 client node to run out of memory. Instead flush old inodes
5541 from client cache that have the same inode number as a new inode.
5544 Frequency : SLES9 2.6.5 kernel and long filenames only
5545 Bugzilla : 9969, 10379
5546 Description: utime reports stale NFS file handle
5547 Details : SLES9 uses out-of-dentry names in some cases, which confused
5548 the lustre dentry revalidation. Change it to always use the
5552 Frequency : rare, unless heavy write-truncate concurrency is continuous
5553 Bugzilla : 4180, 6984, 7171, 9963, 9331
5554 Description: OST becomes very slow and/or deadlocked during object unlink
5555 Details : filter_destroy() was holding onto the parent directory lock
5556 while truncating+unlinking objects. For very large objects this
5557 may block other threads for a long time and slow overall OST
5558 responsiveness. It may also be possible to get a lock ordering
5559 deadlock in this case, or run out of journal credits because of
5560 the combined truncate+unlink. Solution is to do object truncate
5561 first in one transaction without parent lock, and then do the
5562 final unlink in a new transaction with the parent lock. This
5563 reduces the lock hold time dramatically.
5566 Frequency : rare, 2.4 kernels only
5568 Description: MDS or OST cleanup may trip kernel BUG when dropping kernel lock
5569 Details : mds_cleanup() and filter_cleanup() need to drop the kernel lock
5570 before unmounting their filesystem in order to avoid deadlock.
5571 The kernel_locked() function in 2.4 kernels only checks whether
5572 the kernel lock is held, not whether it is this process that is
5573 holding it as 2.6 kernels do.
5578 Description: MDS or OST may oops/LBUG if a client is connecting multiple times
5579 Details : The client ptlrpc code may be trying to reconnect to a down
5580 server before a previous connection attempt has timed out.
5581 Increase the reconnect interval to be longer than the connection
5582 timeout interval to avoid sending duplicate connections to
5586 Frequency : echo_client brw_test command
5588 Description: fix echo_client to work with OST preallocated code
5589 Details : OST preallocation code (5137) didn't take echo_client IO path
5590 into account: echo_client calls filter methods outside of any
5591 OST thread and, hence, there is no per-thread preallocated
5592 pages and buffers to use. Solution: hijack pga pages for IO. As
5593 a byproduct, this avoids unnecessary data copying.
5597 Bugzilla : 3555, 5962, 6025, 6155, 6296, 9574
5598 Description: Client can oops in mdc_commit_close() after open replay
5599 Details : It was possible for the MDS to return an open request with no
5600 transaction number in mds_finish_transno() if the client was
5601 evicted, but without actually returning an error. Clients
5602 would later try to replay that open and may trip an assertion
5603 Simplify the client close codepath, and always return an error
5604 from the MDS in case the open is not successful.
5607 Frequency : rare, 2.6 OSTs only
5609 Description: OST may deadlock under high load on fragmented files
5610 Details : If there was a heavy load and highly-fragmented OST filesystems
5611 it was possible to have all the OST threads deadlock waiting on
5612 allocation of biovecs, because the biovecs were not released
5613 until the entire RPC IO was completed. Instead, release biovecs
5614 as soon as they are complete to ensure forward IO progress.
5616 Severity : enhancement
5618 Description: Support for specifying external journal device at mount
5619 Details : If an OST or MDS device is formatted with an external journal
5620 device, this device major/minor is stored in the ext3 superblock
5621 and may not be valid for failover. Allow detecting and
5622 specifying the external journal at mount time.
5627 Description: Mounting an MDS with pending unlinked files may cause oops
5628 Details : target_finish_recovery() calls mds_postrecov() which returned
5629 the number of orphans unlinked. mds_lov_connect->mds_postsetup()
5630 considers this an error and immediately begins cleaning up the
5631 lov, just after starting the mds_lov process
5633 Severity : enhancement
5635 Description: Implement 'lfs df' to report actual free space on per-OST basis
5636 Details : Add sub-command 'df' on 'lfs' to report the disk space usage of
5637 MDS/OSDs. Usage: lfs df [-i][-h]. Command Options: '-i' to report
5638 usage of objects; '-h' to report in human readable format.
5640 ------------------------------------------------------------------------------
5642 2005-08-26 Cluster File Systems, Inc. <info@clusterfs.com>
5649 Description: Mounting an ldiskfs file system with mballoc may crash OST node.
5650 Details : ldiskfs mballoc code may reference an uninitialized buddy struct
5651 at startup during orphan unlinking. Instead, skip buddy update
5652 before setup, as it will be regenerated after recovery is complete.
5657 Description: If an OST is inactive, its locks might reference stale inodes.
5658 Details : lov_change_cbdata() must iterate over all namespaces, even if
5659 they are inactive to clear inode references from the lock.
5661 Severity : enhancement
5662 Frequency : occasional, if non-standard max_dirty_mb used
5664 Description: Client will block write RPCs if not enough grant
5665 Details : If a client has max_dirty_mb smaller than max_rpcs_in_flight,
5666 then the client will block writes while waiting for another RPC
5667 to complete instead of consuming its dirty limit. With change
5668 we get improved performance when max_dirty_mb is small.
5670 Severity : enhancement
5671 Bugzilla : 3389, 6253
5672 Description: Add support for supplementary groups on the MDS.
5673 Details : The MDS has an upcall /proc/fs/lustre/mds/{mds}/group_upcall
5674 (set to /usr/sbin/l_getgroups if enabled) which will do MDS-side
5675 lookups for user supplementary groups into a cache.
5679 Description: O_CREAT|O_EXCL open flags in liblustre always return -EEXIST
5680 Details : Make libsysio to not enforce O_EXCL by clearing the flag,
5681 for liblustre O_EXCL is enforced by MDS.
5685 Description: readdir never returns NULL in liblustre.
5686 Details : Corrected llu_iop_getdirentries logic, to return offset of next
5687 dentry in struct dirent.
5691 Frequency : liblustre only, depends on application IO pattern
5692 Description: liblustre clients evicted if not contacting servers
5693 Details : Don't put liblustre clients into the ping_evictor list, so
5694 they will not be evicted by the pinger ever.
5696 Severity : enhancement
5698 Description: Add ability to evict clients by NID from MDS.
5699 Details : By echoing "nid:$NID" string into
5700 /proc/fs/lustre/mds/.../evict_client client with nid that equals to
5701 $NID would be instantly evicted from this MDS and from all active
5702 OSTs connected to it.
5706 Description: Do not query file size twice, somewhat slowing stat(2) calls.
5707 Details : lookup_it_finish() used to query file size from OSTs that was not
5712 Description: service threads change working directory to that of init
5713 Details : Starting lustre service threads may pin the working directory
5714 of the parent thread, making that filesystem busy. Threads
5715 now change to the working directory of init to avoid this.
5719 Frequency : during shutdown only
5720 Description: shutdown with a failed MDS or OST can cause unmount to hang
5721 Details : Don't resend DISCONNECT messages in ptlrpc_disconnect_import()
5726 Frequency : 2.6 only
5727 Description: chmod/chown may include an extra supplementary group
5728 Details : ll{,u}_mdc_pack_op_data() does not properly initialize the
5729 supplementary group and if none is specified this is used.
5732 Bugzilla : 5479 (6816)
5734 Description: Racing open + rm can assert client in mdc_set_open_replay_data()
5735 Details : If lookup is in progress on a file that is unlinked we might try
5736 to revalidate the inode and fail in revalidate after lookup is
5737 complete and ll_file_open() enqueues the open again but
5738 it_open_error() was not checking DISP_OPEN_OPEN errors correctly.
5741 Frequency : always, if lconf --abort_recovery used
5743 Description: lconf --abort_recovery fails with 'Operation not supported'
5744 Details : lconf was attempting to abort recovery on the MDT device and not
5747 ------------------------------------------------------------------------------
5749 2005-08-08 Cluster File Systems, Inc. <info@clusterfs.com>
5754 Frequency : rare (only unsupported configurations with a node running as an
5756 Bugzilla : 6514, 5137
5757 Description: Mounting a Lustre file system on a node running as an OST could
5759 Details : OSTs now preallocates memory needed to write out data at
5760 startup, instead of when needed, to avoid having to
5761 allocate memory in possibly low memory situations.
5762 Specifically, if the file system is mounted on on OST,
5763 memory pressure could force it to try to write out data,
5764 which it needed to allocate memory to do. Due to the low
5765 memory, it would be unable to do so and the node would
5766 become unresponsive.
5768 Severity : enhancement
5770 Description: Addition of lconf --service command line option
5771 Details : lconf now accepts a '--service <arg>' option, which is
5772 shorthand for 'lconf --group <arg> --select <arg>=<hostname>'
5774 Severity : enhancement
5776 Description: Failover mode is now the default for OSTs.
5777 Details : By default, OSTs will now run in failover mode. To return to
5778 the old behaviour, add '--failout' to the lmc line for OSTs.
5780 Severity : enhancement
5782 Description: Health checks are now provided for MDS and OSTs
5783 Details : Additional detailed health check information on MSD and OSTs
5784 is now provided through the procfs health_check value.
5787 Frequency : occasional, depends on IO load
5789 Description: Disk fragmentation on the OSTs could eventually cause slowdowns
5790 after numerous create/delete cycles
5791 Details : The ext3 inode allocation policy would not allocate new inodes
5792 very well on the OSTs because there are no new directories
5793 being created. Instead we look for groups with free space if
5794 the parent directories are nearly full.
5799 Description: Network or server problems during mount may cause partially
5800 mounted clients instead of returning an error.
5801 Details : The config llog parsing code may overwrite the error return
5802 code during mount error handling, returning success instead
5808 Description: MDS can fail to allocate large reply buffers
5809 Details : After long uptimes the MDS can fail to allocate large reply
5810 buffers (e.g. zconf client mount config records) due to memory
5811 fragmentation or consumption by the buffer cache. Preallocate
5812 some large reply buffers so that these replies can be sent even
5813 under memory pressure.
5817 Frequency : rare (liblustre)
5818 Description: fsx running with liblustre complained that using truncate() to
5819 extend the file doesn't work. This patch corrects that issue.
5820 Details : This is the liblustre equivalent of the fix for bug 6196. Fixes
5821 ATTR_SIZE and lsm use in llu_setattr_raw.
5825 Frequency : rare, only 2.6 kernels
5826 Description: Unusual file access patterns on the MDS may result in inode
5827 data being lost in very rare circumstances.
5828 Details : Bad interaction between the ea-in-inode patch and the "no-read"
5829 code in the 2.6 kernel caused the inode and/or EA data not to
5830 be read from disk, causing single-file corruption.
5834 Frequency : rare, only 2.6 filesystems using extents
5835 Description: Heavy concurrent write and delete load may cause data corruption.
5836 Details : It was possible under high-load situations to have an extent
5837 metadata block in the block device cache from a just-unlinked
5838 file overwrite a newly-allocated data block. We now unmap any
5839 metadata buffers that alias just-allocated data blocks.
5843 Frequency : filesystems with default stripe_count larger than 77
5844 Description: lconf+mke2fs fail when formatting filesystem with > 77 stripes
5845 Details : lconf specifies an inode size of 4096 bytes when the default
5846 stripe_count is larger than 77. This conflicts with the default
5847 inode density of 1 per 4096 bytes. Allocate smaller inodes in
5848 this case to avoid pinning too much memory for large EAs.
5850 ------------------------------------------------------------------------------
5852 2005-07-07 Cluster File Systems, Inc. <info@clusterfs.com>
5857 Frequency : rare (extremely heavy IO load with hundreds of clients)
5859 Description: Client is evicted, gets IO error writing to file
5860 Details : lock ordering changes for bug 5492 reintroduced bug 3267 and
5861 caused clients to be evicted for AST timeouts. The fixes in
5862 bug 5192 mean we no longer need to have such short AST timeouts
5863 so ldlm_timeout has been increased.
5866 Frequency : occasional during --force or --failover shutdown under load
5867 Bugzilla : 5949, 4834
5868 Description: Server oops/LBUG if stopped with --force or --failover under load
5869 Details : a collection of import/export refcount and cleanup ordering
5870 issues fixed for safer force cleanup
5873 Frequency : only filesystems larger than 120 OSTs
5874 Bugzilla : 5990, 6223
5875 Description: lfs getstripe would oops on a very large filesystem
5876 Details : lov_getconfig used kfree on vmalloc'd memory
5879 Frequency : only filesystems exporting via NFS to Solaris 10 clients
5880 Bugzilla : 6242, 6243
5881 Description: reading from files that had been truncated to a non-zero size
5882 but never opened returned no data
5883 Details : ll_file_read() reads zeros from no-object files to EOF
5888 Description: A bug in MDS/OSS recovery could cause the OSS to fail an assertion
5889 Details : There's little harm in aborting MDS/OSS recovery and letting it
5890 try again, so I removed the LASSERT and return an error instead.
5892 Severity : enhancement
5894 Description: New debugging infrastructure for tracking down data corruption
5895 Details : The I/O checksum code was replaced to: (a) control it at runtime,
5896 (b) cover more of the client-side code path, and (c) try to narrow
5897 down where problems occurred
5901 Bugzilla : 3819, 4364, 4397, 6313
5902 Description: Racing close and eviction MDS could cause assertion in mds_close
5903 Details : It was possible to get multiple mfd references during close and
5904 client eviction, leading to one thread referencing a freed mfd.
5906 Severity: : enhancement
5907 Bugzilla : 3262, 6359
5908 Description: Attempts to reconnect to servers are now more aggressive.
5909 Details : This builds on the enhanced upcall-less recovery that was added
5910 in 1.4.2. When trying to reconnect to servers, clients will
5911 now try each server in the failover group every 10 seconds. By
5912 default, clients would previously try one server every 25 seconds.
5917 Description: After recovery, certain operations trigger a failed
5918 assertion on a client.
5919 Details : Failing over an mds, using lconf -d --failover, while a
5920 client was doing a readdir() call would cause the client to
5921 LBUG after recovery completed and the readdir() was resent.
5923 Severity : enhancement
5925 Description: Default groups are now added by lconf
5926 Details : You can now run lconf --group <servicename> without having to
5927 manually add groups with lmc.
5930 Frequency : occasional
5932 Description: Nodes with an elan id of 0 trigger a failed assertion
5935 Frequency : always when accessing e.g. tty/console device nodes
5937 Description: tty and some other devices nodes cannot be used on lustre
5938 Details : file's private_data field is used by device data and lustre
5939 values in there got lost. New field was added to struct file to
5940 store fs-specific private data.
5943 Frequency : when exporting Lustre via NFS
5945 Description: NFSD failed occasionally when looking up a path component
5946 Details : NFSD is looking up ".." which was broken in ext3 directories
5947 that had grown large enough to become hashed.
5950 Frequency : Clusters with multiple interfaces not on the same subnet
5952 Description: Nodes will repeatedly try to reconnect to an interface which it
5953 cannot reach and report an error to the log.
5954 Details : Extra peer list entries will be created by lconf with some peers
5955 unreachable. lconf now validates the peer before adding it.
5958 Frequency : Only if a default stripe is set on the filesystem root.
5960 Description: Setting a default stripe on the filesystem root prevented the
5961 filesystem from being remounted.
5962 Details : The client was sending extra request flags in the root getattr
5963 request and did not allocate a reply buffer for the dir EA.
5966 Frequency : occasional, higher if lots of files are accessed by one client
5967 Bugzilla : 6159, 6097
5968 Description: Client trips assertion regarding lsm mismatch/magic
5969 Details : While revalidating inodes the VFS looks up inodes with ifind()
5970 and in rare cases can find an inode that is being freed.
5971 The ll_test_inode() code will free the lsm during ifind()
5972 when it finds an existing inode and then the VFS later attaches
5973 this free lsm to a new inode.
5977 Bugzilla : 6422, 7030
5978 Description: MDS deadlock between mkdir and client eviction
5979 Details : Creating a new file via mkdir or mknod (starting a transaction
5980 and getting the ns lock) can deadlock with client eviction
5981 (gets ns lock and trying to finish a synchronous transaction).
5984 Frequency : occasional
5985 Description: While starting a server, the fsfilt_ext3 module could not be
5987 Details : CFS's improved ext3 filesystem is named ldiskfs for 2.6
5988 kernels. Previously, lconf would still use the ext3 name
5989 when trying to load modules. Now, it will correctly use
5990 ext3 on 2.4 and ldiskfs on 2.6.
5992 Severity : enhancement
5993 Description: The default stripe count has been changed to 1
5994 Details : The interpretation of the default stripe count (0, to lfs
5995 or lmc) has been changed to mean striping across a single
5996 OST, rather than all available. For general usage we have
5997 found a stripe count of 1 or 2 works best.
5999 Severity : enhancement
6000 Description: Add support for compiling against Cray portals.
6001 Details : Conditional compiling for some areas that are different
6005 Frequency : occasional
6006 Bugzilla : 6409, 6834
6007 Description: Creating files with an explicit stripe count may lead to
6008 a failed assertion on the MDS
6009 Details : If some OSTs are full or unavailable, creating files may
6010 trigger a failed assertion on the MDS. Now, Lustre will
6011 try to use other servers or return an error to the
6015 Frequency : occasional
6017 Description: Multiple concurrent overlapping read+write on multiple SMP nodes
6018 caused lock timeout during readahead (since 1.4.2).
6019 Details : Processes doing readahead might match a lock that hasn't been
6020 granted yet if there are overlapping and conflicting lock
6021 requests. The readahead process waits on ungranted lock
6022 (original lock is CBPENDING), while OST waits for that process
6023 to cancel CBPENDING read lock and eventually evicts client.
6025 Severity : enhancement
6027 Description: Initial enabling of flock support for clients
6028 Details : Implements fcntl advisory locking and file status functions.
6029 This feature is provided as an optional mount flag (default
6030 off), and is NOT CURRENTLY SUPPORTED. Not all types of record
6031 locking are implemented yet, and those that are are not guaranteed
6032 to be completely correct in production environments.
6033 mount -t lustre -o [flock|noflock] ...
6036 Frequency : occasional
6038 Description: OSTs running 2.4 kernels but with extents enabled might trip an
6039 assertion in the ext3 JBD (journaling) layer.
6040 Details : The b_committed_data struct is protected by the big kernel lock
6041 in 2.4 kernels, serializing journal_commit_transaction() and
6042 ext3_get_block_handle->ext3_new_block->find_next_usable_block()
6043 access to this struct. In 2.6 kernels there is finer grained
6044 locking to improve SMP performance of the JBD layer.
6048 Description: Changes the "SCSI I/O Stats" kernel patch to default to "enabled"
6050 -----------------------------------------------------------------------------
6052 2005-05-05 Cluster File Systems, Inc. <info@clusterfs.com>
6054 NOTE: Lustre 1.4.2 uses an incompatible network protocol than previous
6055 versions of Lustre. Please update all servers and clients to
6056 version 1.4.2 or later at the same time. You must also run
6057 "lconf --write-conf {config}.xml" on the MDS while it is stopped
6058 to update the configuration logs.
6060 - fix for HPUX NFS client breakage when NFS exporting Lustre (5781)
6061 - mdc_enqueue does not need max_mds_easize request buffer on send (5707)
6062 - swab llog records of type '0' so we get proper header size/idx (5861)
6063 - send llog cancel req to DLM cancel portal instead of cb portal (5515)
6064 - fix rename of one directory over another leaking an inode (5953)
6065 - avoid SetPageDirty on 2.6 (5981)
6066 - don't re-add just-being-destroyed locks to the waiting list (5653)
6067 - when creating new directories, inherit the parent's custom
6068 striping settings if present parent (3048)
6069 - flush buffers from cache before direct IO in 2.6 obdfilter (4982)
6070 - don't hold i_size_sem in ll_nopage() and ll_ap_refresh_count (6077)
6071 - don't hold client locks on temporary worklist from l_lru (5666)
6072 - handle IO errors in 2.6 obdfilter bio completion routine (6046)
6073 - automatically evict dead clients (5921)
6074 - Update file size properly in create+truncate+fstat case (6196)
6075 - Do not unhash mountpoint dentries, do not allow removal of
6077 - Avoid lock ordering deadlock issue with write/truncate (6203,5654)
6078 - reserve enough journal credits in fsfilt_start_log for setattr (4554)
6079 - ldlm_enqueue freed-export error path would always LBUG (6149,6184)
6080 - don't reference lr_lvb_data until after we hold lr_lvb_sem (6170)
6081 - don't overwrite last_rcvd if there is a *_client_add() error (6086)
6082 - Correctly handle reads of files with no objects (6243)
6083 - lctl recover will also mark a device active if deactivate used (5933)
6085 - by default create 1 inode per 4kB space on MDS, per 16kB on OSTs
6086 - allow --write-conf on an MDS with different nettype than client (5619)
6087 - don't write config llogs to MDS for mounts not from that MDS (5617)
6088 - lconf should create multiple TCP connections from a client (5201)
6089 - init scripts are now turned off by default; run chkconfig --on
6090 lustre and chkconfig --on lustrefs to use them
6091 - upcalls are no longer needed for clients to recover to failover
6093 - add --abort-recovery option to lconf to abort recovery on device
6095 - add support for an arbitrary number of OSTs (3026)
6096 - Quota support protocol changes.
6097 - forward compatibility changes to wire structs (6007)
6098 - rmmod NALs that might be loaded because of /etc/modules.conf (6133)
6099 - support for mountfsoptions and clientoptions to the Lustre LDAP (5873)
6100 - improved "lustre status" script
6101 - initialize blocksize for non-regular files (6062)
6102 - added --disable-server and --disable-client configure options (5782)
6103 - introduce a lookup cache for lconf to avoid repeated DB scans (6204)
6104 - Vanilla 2.4.29 support
6105 - increase maximum number of obd devices to 520 (6242)
6106 - remove the tcp-zero-copy patch from the suse-2.4 series (5902)
6107 - Quadrics Elan drivers are now included for the RHEL 3 2.4.21 and
6108 SLES 9 2.6.5 kernels
6109 - limit stripes per file to 160 (the maximum EA size) (6093)
6111 2005-03-22 Cluster File Systems, Inc. <info@clusterfs.com>
6114 - don't LASSERT in ll_release on NULL lld with NFS export (4655, 5760)
6115 - hold NS lock when calling handle_ast_error->del_waiting_lock (5746)
6116 - fix setattr mtime regression from lovcleanup merge (4829, 5669)
6117 - workaround for 2.6 crash in ll_unhash_aliases (5687, 5210)
6118 - small ext3 extents cleanups and fixes (5733)
6119 - improved mballoc code, several small races and bugs fixed (5733, 5638)
6120 - kernel version 43 - fix remove_suid bugs in both 2.4 and 2.6 (5695)
6121 - avoid needless client->OST connect, fix handle mismatch (5317)
6122 - fix DLM error path that led to out-of-sync client, long delays (5779)
6123 - support common vfs-enforced mount options (nodev,nosuid,noexec) (5637)
6124 - fix several locking issues related to i_size (5492,5624,5654,5672)
6125 - don't move pending lock onto export if it is already evicted (5683)
6126 - fix kernel oops when creating .foo in unlinked directory (5548)
6127 - fix deadlock in obdfilter statistics vs. object create (5811)
6128 - use time_{before,after} to avoid timer jiffies wrap (5882)
6129 - shutdown --force/--failover stability (3607,3651,4797,5203,4834)
6130 - Do not leak request if server was not able to process it (5154)
6131 - If mds_open unable to find parent dir, make that negative lookup(5154)
6132 - don't create new directories with extent-mapping (5909, 5936)
6134 - fix lustre/lustrefs init scripts for SuSE (patch from Scali, 5702)
6135 - don't hold the pinger_sem in ptlrpc_pinger_sending_on_import
6136 - change obd_increase_kms to obd_adjust_kms (up or down) (5654)
6137 - lconf, lmc search both /usr/lib and /usr/lib64 for Python libs (5800)
6138 - support for RHEL4 kernel on i686 (5773)
6139 - provide error messages when incompatible logs are encountered (5898)
6141 2005-02-18 Cluster File Systems, Inc. <info@clusterfs.com>
6142 * version 1.4.0.10 (1.4.1 release candidate 1)
6144 - don't keep a lock reference when lock is not granted (4238)
6145 - unsafe list practices (rarely) led to infinite eviction loop (4908)
6146 - add per-fs limit of Lustre pages in page cache, avoid OOM (4699)
6147 - drop import inflight refcount on signal_completed_replay error (5255)
6148 - unlock page after async write error during send (3677)
6149 - handle missing objects in filter_preprw_read properly (5265)
6150 - no transno return for symlink open, don't save no-trasno open (3440)
6151 - don't try to complete elan receive that already failed (4012)
6152 - free RPC server reply state on error (5406)
6153 - clean up thread from ptlrpc_start_thread() on error (5160)
6154 - readahead could read extra page into cache that wasn't ejected (5388)
6155 - prevent races in class_attach/setup/cleanup/detach (5260)
6156 - don't dereference de->d_inode after l_dput of de (5458)
6157 - use "int" for stripe value returned from lock_to_stripe (5544)
6158 - mballoc allocation and error-checking fixes in 2.6 (5504)
6159 - block device patches to fix I/O request sizes in 2.6 (5482)
6160 - look up hostnames for IB nals (5602)
6161 - 2.6 changed lock ordering of 2 semaphores, caused deadlock (5654)
6162 - don't start multiple acceptors for the same port (5277)
6163 - fix incorrect LASSERT in mds_getattr_name (5635)
6164 - export a proc file for general "ping" checking (5628)
6165 - fix "lfs check" to not block when the MDS is down (5628)
6167 - service request history (4965)
6168 - put {ll,lov,osc}_async_page structs in a single slab (4699)
6169 - create an "evict_client" /proc entry on OSTs, like the MDS has
6170 - fix mount usage message, return errors per mount(8) (5168)
6171 - change grep [] to grep "[]" in tests so they work in more UMLs
6172 - fix ppc64/x86_64 spec to use %{_libdir} instead of /usr/lib (5389)
6173 - remove ancient LOV_MAGIC_V0 EA support (5047)
6174 - add "disk I/Os in flight" and "I/O req time" stats in obdfilter
6175 - align r/w RPCs to PTLRPC_MAX_BRW_SIZE boundary for performance (3451)
6176 - allow readahead allocations to fail when low on memory (5383)
6177 - mmap locking landed again, after considerable improvement (2828)
6178 - add get_hostaddr() to lustreDB.py for LDAP support (5459)
6180 2004-11-23 Cluster File Systems, Inc. <info@clusterfs.com>
6183 - send OST transaction number in read/write reply to free req (4966)
6184 - don't ASSERT in ptl_send_rpc() if we run out of memory (5119)
6185 - lock /proc/sys/portals/routes internal state, avoiding oops (4827)
6186 - the watchdog thread now runs as interruptible (5246)
6187 - flock/lockf fixes (but it's still disabled, pending 5135)
6188 - don't use EXT3 constants in llite code (5094)
6189 - memory shortage at startup could cause assertion (5176)
6191 - reorganization of lov code
6192 - single portals codebase
6194 - add extents/mballoc support (5025)
6195 - direct I/O reads in the obdfilter (4048)
6196 - kernel patches from LNXI for 2.6 (bluesmoke, perfctr, mtd, kexec)
6198 tbd Cluster File Systems, Inc. <info@clusterfs.com>
6201 - send OST transaction number in read/write reply to free req (4966)
6202 - don't ASSERT in ptl_send_rpc() if we run out of memory (5119)
6203 - lock /proc/sys/portals/routes internal state, avoiding oops (4827)
6204 - the watchdog thread now runs as interruptible (5246)
6205 - handle missing objects in filter_preprw_read properly (5265)
6206 - unsafe list practices (rarely) led to infinite eviction loop (4908)
6207 - drop import inflight refcount on signal_completed_replay error (5255)
6208 - unlock page after async write error during send (3677)
6209 - return original error code on reconstructed replies (3761)
6210 - no transno return for symlink open, don't save no-trasno open (3440)
6212 - add pid to ldlm debugging output (4922)
6213 - bump the watchdog timeouts -- we can't handle 30sec yet
6214 - extra debugging for orphan dentry/inode bug (5259)
6216 2004-11-16 Cluster File Systems, Inc. <info@clusterfs.com>
6219 - fix TCP_NODELAY bug, which caused extreme perf regression (5134)
6220 - allocate qswnal tx descriptors singly to avoid fragmentation (4504)
6221 - don't LBUG on obdo_alloc() failure, use OBD_SLAB_ALLOC() (4800)
6222 - fix NULL dereference in /proc/sys/portals/routes (4827)
6223 - allow failed mdc_close() operations to be interrupted (4561)
6224 - stop precreate on OST before MDS would time out on it (4778)
6225 - don't send partial-page writes before EOF from client (4410)
6226 - discard client grant for sub-page writes on large-page clients (4520)
6227 - don't free dentries not owned by NFS code, check generation (4806)
6228 - fix lsm leak if mds_create_objects() fails (4801)
6229 - limit debug_daemon file size, always print CERROR messages (4789)
6230 - use transno after validating reply (3892)
6231 - process timed out requests if import state changes (3754)
6232 - update mtime on OST during writes, return in glimpse (4829)
6233 - add mkfsoptions to LDAP (4679)
6234 - use ->max_readahead method instead of zapping global ra (5039)
6235 - don't interrupt __l_wait_event() during strace
6237 - add software watchdogs to catch hung threads quickly (4941)
6238 - make lustrefs init script start after nfs is mounted
6239 - fix CWARN/ERROR duplication (4930)
6240 - return async write errors to application if possible (2248)
6241 - add /proc/sys/portal/memused (bytes allocated by PORTALS_ALLOC)
6242 - print NAL number in %x format (4645)
6243 - update barely-supported suse-2.4.21-171 series (4842)
6244 - support for sles 9 %post scripts
6245 - support for building 2.6 kernel-source packages
6246 - support for sles km_* packages
6248 2004-10-07 Cluster File Systems, Inc. <info@clusterfs.com>
6251 - ignore -ENOENT errors in osc_destroy (3639)
6252 - notify osc create thread that OSC is being cleaned up (4600)
6253 - add nettype argument for llmount in #5d in conf-sanity.sh (3936)
6254 - reconstruct ost_handle() like mds_handle() (4657)
6255 - create a new thread to do import eviction to avoid deadlock (3969)
6256 - let lconf resolve symlinked-to devices (4629)
6257 - don't unlink "objects" from directory with default EA (4554)
6258 - hold socknal file ref over connect in case target is down (4394)
6259 - allow more than 32000 subdirectories in a single directory (3244)
6260 - fix blocks count for O_DIRECT writes (3751)
6261 - OST returns ENOSPC from object create when no space left (4539)
6262 - don't send truncate RPC if file size isn't changing (4410)
6263 - limit OSC precreate to 1/2 of value OST considers bogus (4778)
6264 - bind to privileged port in socknal and tcpnal (3689)
6266 - rate limit CERROR/CWARN console message to avoid overload (4519)
6267 - GETFILEINFO dir ioctl returns LOV EA + MDS stat in 1 call (3327)
6268 - basic mmap support (3918)
6269 - kernel patch series update from b1_4 (4711)
6271 2004-09-16 Cluster File Systems, Inc. <info@clusterfs.com>
6274 - avoid crash during MDS cleanup with OST shut down (2775)
6275 - fix loi_list_lock/oig_lock inversion on interrupted IO (4136)
6276 - don't use bad inodes on the MDS (3744)
6277 - dynamic object preallocation to improve recovery speed (4236)
6278 - don't hold spinlock over lock dumping or change debug flags (4401)
6279 - don't zero obd_dev when it is force cleaned (3651)
6280 - print grants to console if they go negative (4431)
6281 - "lctl deactivate" will stop automatic recovery attempts (3406)
6282 - look for existing locks in ldlm_handle_enqueue() (3764)
6283 - don't resolve lock handle twice in recovery avoiding race (4401)
6284 - revalidate should check working dir is a directory (4134)
6286 - don't always mark "slow" obdfilter messages as errors (4418)
6288 2004-08-24 Cluster File Systems, Inc. <info@clusterfs.com>
6291 - don't close LustreDB during write_conf until it is done (3860)
6292 - fix typo in lconf for_each_profile (3821)
6293 - allow dumping logs from multiple threads at one time (3820)
6294 - don't allow multiple threads in OSC recovery (3812)
6295 - fix debug_size parameters (3864)
6296 - fix mds_postrecov to initialize import for llog ctxt (3121)
6297 - replace config semaphore with spinlock (3306)
6298 - be sure to send a reply for a CANCEL rpc with bad export (3863)
6299 - don't allow enqueue to complete on a destroyed export (3822)
6300 - down write_lock before checking llog header bitmap (3825)
6301 - recover from lock replay timeout (3764)
6302 - up llog sem before sending rpc (3652)
6303 - reduce ns lock hold times when setting kms (3267)
6304 - change a dlm LBUG to LASSERTF, to maybe learn something (4228)
6305 - fix NULL deref and obd_dev leak on setup error (3312)
6306 - replace some LBUG about llog ops with error handling (3841)
6307 - don't match INVALID dentries from d_lookup and spin (3784)
6308 - hold dcache_lock while marking dentries INVALID and hashing (4255)
6309 - fix invalid assertion in ptlrpc_set_wait (3880)
6311 - add libwrap support for the TCP acceptor (3996)
6312 - add /proc/sys/portals/routes for non-root route listing (3994)
6313 - allow setting MDS UUID in .xml (2580)
6314 - print the stack of a process that LBUGs (4228)
6316 2004-07-14 Cluster File Systems, Inc. <info@clusterfs.com>
6319 - don't cleanup request in ll_file_open() on failed MDS open (3430)
6320 - make sure to unset replay flag from failed open requests (3440)
6321 - if default stripe count is 0, use OST count for inode size (3636)
6322 - update parent mtime/ctime on client for create/unlink (2611)
6323 - drop dentry ref in ext3_add_link from open_connect_dentry (3266)
6324 - free recovery state on server during a forced cleanup (3571)
6325 - unregister_reply for resent reqs (3063)
6326 - loop back devices mounting and status check on 2.6 (3563)
6327 - fix resource-creation race that can provoke i_size == 0 (3513)
6328 - don't try to use bad inodes returned from MDS/OST fs lookup (3688)
6329 - more debugging for page-accounting assertion (3746)
6330 - return -ENOENT instead of asserting if ost getattr+unlink race (3558)
6331 - avoid deadlock after precreation failure (3758)
6332 - fix race and lock order deadlock in orphan handling (3450, 3750)
6333 - add validity checks when grabbing inodes from l_ast_data (3599)
6335 - add /proc/.../recovery_status to obdfilter (3428)
6336 - lightweight CDEBUG infrastructure, debug daemon (3668)
6337 - change default OSC RPC parameters to be better on small clusters
6338 - turn off OST read cache for files smaller than 32MB
6339 - install man pages and include them in rpms (3100)
6340 - add new init script for (un)mounting lustre filesystems (2593)
6341 - run chkconfig in %post for init scripts (3701)
6342 - drop scimac NAL (unmaintained)
6344 2004-06-17 Cluster File Systems, Inc. <info@clusterfs.com>
6347 - clean kiobufs before and after use (3485)
6348 - strip trailing '/'s before comparing paths with /proc/mounts (3486)
6349 - remove assertions to work around "in-flight rpcs" recovery bug (3063)
6350 - change init script to fail more clearly if not run as root (1528)
6351 - allow clients to reconnect during replay (1742)
6352 - fix ns_lock/i_sem lock ordering deadlock for kms update (3477)
6353 - don't do DNS lookups on NIDs too small for IP addresses (3442)
6354 - re-awaken ptlrpcd if new requests arrive during check_set (3554)
6355 - fix cond_resched (3554)
6356 - only evict unfinished clients after recovery (3515)
6357 - allow bulk resend, prevent data loss (3570)
6358 - dynamic ptlrpc request buffer allocation (2102)
6359 - don't allow unlinking open directory if it isn't empty (2904)
6360 - set MDS/OST threads to umask 0 to not clobber client modes (3359)
6361 - remove extraneous obd dereference causing LASSERT failure (3334)
6362 - don't use get_cycles() when creating temp. files on the mds (3156)
6363 - hold i_sem when setting i_size in ll_extent_lock() (3564)
6364 - handle EEXIST for set-stripe, set proper directory name (3336)
6366 - servers can dump a log evicting a client - lustre.dump_on_timeout=1
6367 - fix ksocknal_fmb_callback() error messages (2918)
6369 2004-05-27 Cluster File Systems, Inc. <info@clusterfs.com>
6372 - don't copy lvb into (possibly NULL) reply on error (2983)
6373 - don't deref dentry after dput, don't free lvb on error (2922)
6374 - use the kms to determine writeback rpc length (2947)
6375 - increment oti_logcookies when osc is inactive (2948)
6376 - update client's i_blocks count via lvb messages (2543)
6377 - handle intent open/close of special files properly (1557)
6378 - mount MDS with errors=remount-ro, like obdfilter (2009)
6379 - initialize lock handle to avoid ASSERT on error cleanup (3057)
6380 - don't use cancelling-locks' kms values (2947)
6381 - use highest lock extent for kms, not last one (2925)
6382 - don't dereference ERR_PTR() dentry in error handling path (3107)
6383 - fix thread race in portals_debug_dumplog() (3122)
6384 - create lprocfs device entries at setup instead of at attach (1519)
6385 - common AST error handler, don't evict client on completion race (3145)
6386 - zero nameidata in detach_mnt in 2.6 (3118)
6387 - verify d_inode after revalidate_special is valid in 2.6 (3116)
6388 - use lustre_put_super() to handle zconf unmounts in 2.6 (3064)
6389 - initialize RPC timeout timer earlier for 2.6 (3219)
6390 - don't dereference NULL reply buffer if mdc_close was never sent (2410)
6391 - print nal/nid for unknown nid (3258)
6392 - additional checks for oscc recovery before doing precreate (3284)
6393 - fix ll_extent_lock() error return code for 64-bit systems (3043)
6394 - don't crash in mdc_close for bad permissions on open (3285)
6395 - zero i_rdev for non-device files (3147)
6396 - clear page->private before handing to FS, better assertion (3119)
6397 - tune the read pipeline (3236)
6398 - fix incorrect decref of invalidated dentry (2350)
6399 - provide read-ahead stats and refine rpc in flight stats (3328)
6400 - don't hold journal transaction open across create RPC (3313)
6401 - update atime on MDS at close time (3265)
6402 - close LDAP connection when recovering to avoid server load (3315)
6403 - update iopen-2.6 patch with fixes from 2399,2517,2904 (3301)
6404 - don't leak open file on MDS after open resend (3325)
6405 - serialize filter_precreate and filter_destroy_precreated (3329)
6406 - loop device shouldn't call sync_dev() for nul device (3092)
6407 - clear page cache after eviction (2766)
6408 - resynchronize MDS->OST in background (2824)
6409 - refuse to mount the same filesystem twice on same mountpoint (3394)
6410 - allow llmount to create routes for mounting behind routers (3320)
6411 - push lock cancellation to blocking thread for glimpse ASTs (3409)
6412 - don't call osc_set_data_with_check() for TEST_LOCK matches (3159)
6413 - fix rare problem with rename on htree directories (3417)
6415 - allow default OST striping configuration per directory (1414)
6416 - fix compilation for qswnal for 2.6 kernels (3125)
6417 - increase maximum number of MDS request buffers for large systems
6418 - change liblustreapi to be useful for external progs like lfsck (3098)
6419 - increase local configuration timeout for slow disks (3353)
6420 - allow configuring ldlm AST timeout - lustre.ldlm_timeout=<seconds>
6422 2004-03-22 Cluster File Systems, Inc. <info@clusterfs.com>
6425 - fixes for glimpse AST timeouts / incorrectly 0-sized files (2818)
6426 - don't overwrite extent policy data in reply if lock was blocked (2901)
6427 - drop filter export grants atomically with removal from device (2663)
6428 - del obd_self_export from work_list in class_disconnect_exports (2908)
6429 - don't LBUG if MDS recovery times out during orphan cleanup (2530)
6430 - swab reply message in mdc_close, other PPC fixes (2464)
6431 - fix destroying of named logs (2325)
6432 - overwrite old logs when running lconf --write_conf (2264)
6433 - bump LLOG_CHUNKSIZE to 8k to allow for larger clusters (2306)
6434 - fix race in target_handle_connect (2898)
6435 - mds_reint_create() should take same inode create lock (2926)
6436 - correct journal credits calculated for CANCEL_UNLINK_LOG (2931)
6437 - don't close files for self_export to avoid uninitialized obd (2936)
6438 - allow MDS with the same name as client node (2939)
6439 - hold dentry reference for closed log files for unlink (2325)
6440 - reserve space for all logs during transactions (2059)
6441 - don't evict page beyond end of stripe extent (2925)
6442 - don't oops on a deleted current working directory (2399)
6443 - handle hard links to targets without a parent properly (2517)
6444 - don't dereference NULL lock when racing during eviction (2867)
6445 - don't grow lock extents when lots of conflicting locks (2919)
6447 2004-03-04 Cluster File Systems, Inc. <info@clusterfs.com>
6450 - account for cache space usage on clients to avoid data loss (974)
6451 - lfsck support in lustre kernel code (2349)
6452 - reduce journal credits needed for BRW writes (2370)
6453 - orphan handling to avoid losing space on client/server crashes
6454 - ptlrpcd can be blocked, stopping ALL progress (2477)
6455 - use lock value blocks to assist in proper KMS, faster stat (1021)
6456 - takes i_sem instead of DLM locks internally on obdfilter (2720)
6457 - recovery for initial connections (2355)
6458 - fixes for mds_cleanup_orphans (1934)
6459 - abort_recovery crashes MDS in b_eq (mds_unlink_orphan) (2584)
6460 - block all file creations until orphan recovery completes (1901)
6461 - client remove rq_connection from request struct (2423)
6462 - conf-sanity test_5, proper cleanup in umount log not availale (2640)
6463 - recovery timer race (2670)
6464 - mdc_close recovey bug (2532)
6465 - ptlrpc cleanup bug (2710)
6466 - mds timeout on local locks (2588)
6467 - namespace lock held during RPCs (2431)
6468 - handle interrupted sync write properly (2503)
6469 - don't try to handle a message that hasn't been replied to (2699)
6470 - client assert failure during cleanup after abort recovery (2701)
6471 - leak mdc device after failed mount (2712)
6472 - ptlrpc_check_set allows timedout requests to complete (2714)
6473 - wait for inflight reqs when ptlrpcd finishes (2710)
6474 - make sure unregistered services are removed from the srv_list
6475 - reset bulk XID's when resending them (caught by 1138 test)
6476 - unregister_bulk after timeout
6477 - fix lconf error (2694)
6478 - handle write after unfinished setstripe, stripe-only getstripe (2388)
6479 - readahead locks pages, leaves pending causing memory pressure (2673)
6480 - increase OST request buffers to 4096 on large machines (2729)
6481 - fix up permission of existing directories in simple_mkdir (2661)
6482 - init deleted item, add assertions ptlrpc_abort_inflight() (2725)
6483 - don't assign transno to errored transactions (2742)
6484 - don't delete objects on OST if given a bogus objid from MDS (2751)
6485 - handle large client PAGE_SIZE readdir on small PAGE_SIZE MDS (2777)
6486 - if rq_no_resend, then timeout request after recovery (2432)
6487 - fix MDS llog_logid record size, 64-bit array alignment (2733)
6488 - don't call usermode_helper from ptlrpcd, DEFAULT upcall (2773)
6489 - put magic in mount.lustre data, check for bad/NULL mount data (2529)
6490 - MDS recovery shouldn't delete objects that it has given out (2730)
6491 - if enqueue arrives after completion, don't clobber LVB (2819)
6492 - don't unlock pages twice when trigger_group_io returns error (2814)
6493 - don't deref NULL rq_repmsg if ldlm_handle_enqueue failed (2822)
6494 - don't write pages to disk if there was an error (1450)
6495 - don't ping imports that have recovery disabled (2676)
6496 - take buffered bytes into account when balancing socknal conn (2817)
6497 - hold a DLM lock over readdir always, use truncate_inode_pages (2706)
6498 - reconnect unlink llog connection after MDS reconnects to OST (2816)
6499 - remove little-endian swabbing of llog records (1987)
6500 - set/limit i_blksize to LL_MAX_BLKSIZE on client (2884)
6501 - retry reposting request buffers if they fail (1191)
6502 - grow extent at grant time to avoid granting a revoked lock (2809)
6503 - lock revoke doesn't evict page if covered by a second lock (2765)
6504 - disable VM readahead to avoid reading outside lock extents (2805)
6506 - return LL_SUPER_MAGIC from statfs for the filesystem type (1972)
6507 - updated kernel patches for hp-2.4.20 kernel (2681)
6509 2004-02-07 Cluster File Systems, Inc. <info@clusterfs.com>
6512 - fix truncated write corruption (2366)
6513 - fix for failed assertion in iopen_connect_dentry (1792,2517)
6515 - don't flag the ptlrpcd thread with PF_MEMALLOC (2636)
6516 - ensure len(uuid) < 37 in lmc (1171)
6517 - fix ia64 OOPS in llog_test (2255)
6518 - zero end of page at obdfilter for partial page writes (2648)
6519 - don't leave stale dentries around after renames (bug 2428)
6520 - fix timeouts when evicting a client with a single lock held (2642)
6521 - set deadline for the initial HELLO message to drain (2634)
6522 - print out dotted-quad IP addresses in the socknal (2302)
6524 - additional debugging for MDS client eviction problem (2443)
6525 - fix mkfsoptions support for osts (2603, 2604)
6527 2004-01-27 Cluster File Systems, Inc. <info@clusterfs.com>
6530 - add series for the vanilla 2.6.0 kernel
6531 - add series for the vanilla 2.4.24 kernel
6532 - add series for a cray x86/64 UL kernel drop
6533 - fix xattr patches for the vanilla 2.4.19 series
6535 - generate true UUIDs in lmc (1171)
6536 - have portals stack dumping break in UML (2466)
6537 - avoid bad dchild deref; avoid inum lock w/o creation (2362)
6538 - allocate with _NOFS in ldlm to avoid deadlock (1933)
6539 - wake callback waiting threads on client eviction (2460)
6540 - Add --ptldebug and --subsystem to lmc (1719)
6541 - update assertion to allow safe interrupt allocation
6542 - set rq_no_resend for cancel requests (2432)
6543 - recalculate ptlrpcd timeout after resend (2494)
6544 - call vfs_rmdir when removing pending directories (2368)
6545 - fix renaming a file to itself (2429)
6546 - lmc creates a default one-stripe lov (2454)
6547 - expand procfs space to handle large clusters (2326)
6548 - increase UML stack to avoid overflow
6549 - update lconf's list of debug and subsystem masks
6550 - fix lfs find --obd (2510)
6551 - /proc tunable for disabling filter read caching (2591)
6552 - stop rpm packages from altering slapd.conf (2301)
6553 - disable nagle in the socknal under 0conf (2578)
6554 - choose mds inode size based on stripe count (2572)
6555 - fix kernel-source rpm problems (2516)
6557 - add --disable-doc to avoid pdf generation (2421)
6558 - update documentation, tests, type-os, comments
6559 - avoid format warnings on ia64
6560 - remove the TOE NAL
6561 - tiny code cleanups by removing unused fields
6563 2004-01-07 Cluster File Systems, Inc. <info@clusterfs.com>
6566 - fix obvious semaphore misuse in as-yet-unused setattr path (2348)
6567 - remove the most blatant lies from BUILDING file (2371)
6568 - change default debug level to reasonable production setting
6569 - reduce client side cache size to reduce cache flush time
6570 - reduce max RPCs in flight to avoid unnecessary file fragmentation
6571 - make TCP zerocopy and pinger support enabled by default (2476)
6572 - sync writes completed after process exits caused crashes (2319)
6573 - maintain correct mount count on the MDS (2356)
6574 - backout 1557, because 2316 wasn't really fixed
6575 - better file I/O statistics gathering in /proc
6576 - don't take unnecessary, deadlock-inducing bug in readpage (2383)
6577 - another kernel patch to fix zero-copy TCP function export
6578 - don't take duplicate lock when processing re-sent getattr (2420)
6579 - lctl uses obd_self_export instead of creating new conn (2353)
6580 - MDS/OST recovery case which requires object creation asserted (2425)
6581 - move lfs from /usr/sbin to /usr/bin in packages
6582 - fix race between mds_client_add and mds_client_free (2417)
6583 - use kmalloc instead of slabs in portals (2430)
6584 - don't create duplicate records when a failover MDS is present (2442)
6585 - remove unnecessary mount age check (2332)
6586 - don't remove directory inodes from locks prematurely (2451)
6587 - don't break if MDS service name is the same as hostname (2103)
6588 - fix races in client write RPC generation when cache full (2482)
6590 2003-12-13 Cluster File Systems, Inc. <info@clusterfs.com>
6593 - remove now-unused request->rq_obd (278)
6594 - if an allocation fails, print out how much memory we've used (1933)
6595 - use PORTAL_SLAB_ALLOC for structures, to get GFP_MEMALLOC (1933)
6596 - add the "configurable stack size" patch to most series files (1256)
6597 - ability to write large log records, for 100+ OST configs (2306)
6598 - fix NULL deref when filter_prep fails (2314)
6599 - fix operator precedence error in filter_sync
6600 - dynamic allocation of socknal TX descriptors (2315)
6601 - fix a missed case in the GFP_MEMALLOC patch, can cause deadlock (2310)
6602 - fix gcc 2.96 compilation problem in xattr kernel patch (2294)
6603 - ensure that CWARN messages in Portals always get to the syslog
6604 - __init/__exit are not for prototype decls (ldlm_init/exit)
6605 - x86-64 compile warning fixes
6606 - fix gateway LMC keyword conflict (2318)
6607 - fix MDS lock inversions in getattr/reint paths (1844)
6608 - fix a rare lock re-ordering bug, which caused deadlock (2322)
6609 - fix i_sem/journal inversion in fsfilt_ext3_write_record (2306)
6610 - DLM race condition prevented some lock evictions (2328)
6611 - ENOMEM detection and retry on socknal sends (2230)
6612 - use GFP_NOFS throughout Lustre, to combat ENOMEM (2230)
6613 - move osc_rpcd into ptlrpc, for use in MDC and others (2329)
6614 - protect MDS inode fsdata with stronger locking; fixes assertion (2313)
6615 - better error messages when a client is rejected during recovery (1505)
6616 - avoid cancelling locks which were never granted, after failure (2330)
6617 - fix i_sem/journal inversion in mds_client_add (2333)
6618 - fix truncate/getattr lock cycle deadlock (2334)
6619 - use rpcd to send close; allows resend after timeout, avoid leak (1897)
6620 - fix two rare exit paths which could leak an l_lock() ref (2321)
6621 - fencepost error in MDS/OST orphan recovery (2226)
6622 - make log record alignment 8 bytes (1988)
6623 - lstripe now fails when requested offset > ost_count (2237)
6624 - ensure that all kernel series have a complete list.h (1607)
6625 - fix crashes in special-file operations (2316)
6626 - lctl create/brw OID mismatch, caused by obsolete filter loop (2339)
6628 - allow configurable automake binary, for testing new versions
6629 - small update to the lfs documentation
6631 2003-12-03 Cluster File Systems, Inc. <info@clusterfs.com>
6633 * fix negative export reference count in fsfilt_sync (2312)
6635 2003-12-01 Cluster File Systems, Inc. <info@clusterfs.com>
6636 * release candidate 0.9.1
6638 - orphans are moved into the PENDING directory for possible recovery
6639 - replayed opens now open by fid for orphan/rename safety (1042)
6640 - last close of an orphan inode generates a transno (683)
6641 - chdir() and mount() now pin the directory entry (1020)
6642 - avoid CERROR in normal ll_setattr_raw() error case (1500)
6643 - discard very old requests without processing them (1502)
6644 - remove some common, well-understood CERRORs (1505)
6645 - require O_DIRECT I/O to be page-sized to workaround IA64 crash (1609)
6646 - clear "grant" flags in OST replies until OST grant code lands (1644)
6647 - fix read performance by not clobbering i_blksize on client (1598)
6648 - fix __ldlm_handle2lock oops by not dereferencing lock after PUT (1625)
6649 - make LRU size a /proc tunable, clears locks when reduced (707)
6650 - fix some lprocfs rot that prevented ptlbd from loading (1732)
6651 - server locks take references on exports now (1558)
6652 - build fixes for 2.4.20-rh trees (1663)
6653 - return an error from lov_create if all OSCs are inactive (1751)
6654 - fix import levels when a reconnect happens without a timeout (1597)
6655 - exit early from mds_open if we get a lookup error (1749)
6656 - partial page read at EOF wouldn't wait for disk before sending (1642)
6657 - avoid NULL deref in obdfilter when reading page past EOF (1592)
6658 - avoid LASSERT in ll_intent_lock if server failed very early (1090)
6659 - fix LBUG in ll_it_open_error with rc = -2 (1861)
6660 - write/truncate lock inversion (1639)
6661 - Don't auto-load obdclass, portals modules during cleanup (1495)
6662 - fix timestamps from jumping to "now" (1763)
6663 - extra journal assertions (1648)
6664 - add an extra multiunlink test (1771)
6665 - fix read_record/write_record API (1776)
6666 - fix leak of offset_extent, possible incorrect i_size later (1772)
6667 - fix lasserts in mis-matched transnos during open-unlink testing (1541)
6668 - Debugging for the kqswnal_get_idle_tx problems (1820)
6669 - Allow recovery to be attempted multiple times (1536)
6670 - Write out MDS last_rcvd file after it is first created (1600)
6671 - Fix tx_descriptor leak in failed transmit situations (1827)
6672 - ext3 journaling fixes for assertion failure after IO error (1871)
6673 - class_export_put() on freed export after completion AST error (1896)
6674 - Fix revalidate looping in VFS (1322)
6675 - Don't access a freed export during MDS_REINT timeout (1521)
6676 - Add open-unlink recovery support on the MDS (1673,1764)
6677 - Return an error if no MDS data was read from last_rcvd (1946)
6678 - Fix for lookup "." or ".." crash on error (1932,1931,1935)
6679 - Don't setup a disk device that doesn't match exported UUID (317)
6680 - Reduce bulk RPC timeout to avoid cascading client/OST failures (1845)
6681 - avoid committing NULL handle in force close
6682 - local.sh is now a one-stripe LOV configuration
6683 - POSIX utime.4 -EPERM on FIFO not owned by user (56)
6684 - fix ext3 htree duplicate directory entry corruption (1516)
6685 - POSIX creat.13, fstat.1, open.18, stat.3 new file atime/mtime (2020)
6686 - update to new LOV EA format (2097)
6687 - interoperability for different PAGE_SIZE/wordsize (686,1821,1343,2042)
6689 2003-06-15 Phil Schwan <phil@clusterfs.com>
6692 - imports and exports cleanup too early, need refcounts (349, 879, 1045)
6693 - per-import/export recovery handling (958, 931, 959)
6694 - multiple last-rcvd slots, for serving multiple FSes (949)
6695 - connections are again shared between multiple imp/exports (963, 964)
6696 - "umount -f" would hang if any requests needed to be sent (393, 978)
6697 - avoid pinning large req buffer by copying for queued messages (989)
6698 - add "uuid" to "lctl device" command to help upcalls (991)
6699 - "open" RPCs with transnos would confuse recovery counters (1037)
6700 - do proper endian conversion of all wire messages (288, 340, 891)
6701 - remove OST bulk get LBUGs, fix ost_brw_write cleanup (1126)
6702 - call waiting locks callback from LDLM recovery thread (1127, 1151)
6703 - fix ptlrpc_connection leak in target_handle_connect (1174)
6704 - fix import refcounting bug in OST and MDS cleanup (1134)
6705 - if an invalid-at-open-time OSC returned before close(), LBUG (1150)
6706 - fix very unlikely obd_types race condition (501)
6707 - remove osc_open hack for echo_client (1187)
6708 - we leaked exports/dlmimps for forcibly disconnected clients (1143)
6709 - a failure in read_inode2 leads to deadlock (1139)
6710 - cancel ack-locks as soon as transaction is committed (1072)
6711 - fix major leaks and crashes in the bulk I/O path (937, 1057)
6712 - make sure to commitrw after any preprw to avoid deadlock (1162)
6713 - failing to execute a file in a lustre FS would lock inode (1203)
6714 - small DEBUG_REQ fix to avoid dereferencing a NULL (1227)
6715 - don't ASSERT while cleaning up an incompletely-setup obd (1248)
6716 - obd_uuid2tgt would walk off the end of the list (1255)
6717 - on IA64 the osc would give portals incorrect bulk size (1258)
6718 - fix debug daemon ioctl interface; allows daemon on ia64 (1274)
6719 - fix lock inversion caused by new llite matching code (1282)
6720 - limit the number of dirty pages on a client to 10MB (1286)
6721 - timed out locks were not being corrected cancelled (1289)
6722 - fix O_DIRECT above 4GB on IA-32 (1292)
6723 * major user-visible changes
6724 - fail out/fail over policy now controlled by the upcall (993)
6726 - add OBD_PING to check server availability and failure (954)
6727 - lustre messages are now sent in sending host order (288, 340, 891)
6728 - add eadatalen to MDS getattr reply (340)
6729 - OST read replies may contain second buffer, with per-page status (593)
6731 2003-03-11 Phil Schwan <phil@clusterfs.com>
6734 - LDLM_DEBUG macro fix, for gcc 3.2 (850)
6735 - failed open()s could cause deadlock; fixed (867, 869)
6736 - stop cancelling OST locks when files are closed (481)
6737 - overlapping XID spaces caused network corruption (851, 853)
6738 - fix unsafe fsfilt counter arithmetic; change to atomic_t
6739 - setattr_raw added, to do single-RPC, server-side setattrs
6740 - lmc/lconf syntax change for OST UUIDs
6741 - fix crashy race condition between ptlrpc_free_req and osc_close
6742 - don't use request in mdc_enqueue if we hit a timeout (889)
6743 - don't set the inode i_size for regular files from the MDS (896)
6744 - handle out of order completion AST (842)
6745 - don't LBUG if a lock request times out after receiving AST (913)
6746 - avoid d_rehash race in ll_find_alias by rehashing inside dcache_lock
6747 - if a bad lock AST arrives, send an error instead of dropping entirely
6748 - return 0 from revalidate2 if ll_intent_lock returns -EINTR (912)
6749 - fix leak in bulk IO when only partially completed (899, 900, 926)
6750 - fix O_DIRECT for ia64 (55)
6751 - (almost) eliminate Lustre-kernel-thread effects on load average (722)
6752 - C-z after timeout could hang a process forever; fixed (977)
6754 - client-side I/O cache (678, 924, 929, 941, 970)
6756 - READPAGE and SETATTRs which don't take server-side locks get
6759 2003-02-11 Phil Schwan <phil@clusterfs.com>
6762 - Fix ldlm_lock_match on the MDS to avoid matching remote locks (592)
6763 - Fix fsfilt_extN_readpage() to read a full page of directory
6764 entries, or fake the remainder if PAGE_SIZE != blocksize (500)
6765 - Avoid extra mdc_getattr() in ll_intent_lock when possible (534, 604)
6766 - Fix imbalanced LOV object allocation and out-of-bound access (469)
6767 - Most intent operations were removed, in favour of a new RPC mode
6768 that does a single RPC to the server and bypasses most of the VFS
6769 - All LDLM resource ID arrays were removed in favour of ldlm_res_id
6770 - Aggressively cancel local locks on DLM servers
6771 - mds_reint_unlink sends EA to the client if it's the last nlink.
6772 client uses that EA to unlink OST objects.
6773 - mds_reint_{rename,unlink,link} were rewritten to take ordered locks
6774 - recursive symlinks were fixed (439)
6775 - fixed NULL deref in DEBUG_REQ
6776 - filter_update_lastobjid no longer calls sync, which annoyed extN
6777 - fixed multi-client small-writes to a single file problem (445)
6778 - fixed mtime updates during file writes (607)
6779 - fixed vector writes on obdfilter causing problems when ENOSPC (670)
6780 - fixed bug in obd_brw_read/write() (under guise of testing 367)
6781 - fixed Linux OST size reporting problem (444, 656)
6782 - OST now updates object mtime with writes or setattr (607, 619)
6783 - client verifies file size before zeroing page past EOF (445)
6784 - OST now writes last allocated objid to disk with allocation (108)
6785 - LOV on echo now works (409)
6787 - mds_reint_unlink sends a new buffer, with the EA included. this
6788 buffer is only valid if body->valid & OBD_MD_FLEASIZE, which is only
6789 set if a regular file was being unlinked, and it was the last link
6790 - use PtlGet from the target for bulk writes (315)
6791 - OST now updates object mtime with writes or setattr (607, 619)
6792 - LDLM now has a grant-time callback to revalidate locked items, if
6794 - Many MDS operations were reorganized to combat race conditions
6796 - Merge b_intel branch (updated lprocfs code) - now at /proc/fs/lustre
6797 - configure check to avoid gcc version 2.96 20000731-2.96-98) (606)
6799 2003-01-06 Andreas Dilger <adilger@clusterfs.com>
6802 - Fully reactivate OST imports after reconnection (512, others)
6803 - Make sure client sees our -ENOTCONN from mds_handle (513 - partial)
6804 - More graceful error handling for truncating on dead OST (515)
6805 - Don't error out unless we're actually accessing dead stripes (474)
6806 - Fix garbage sizes when stripes are missing (410)
6807 - LRU counters were broken, causing constant lock purge (433, 432)
6808 - garbage on read from stripes with failed OSTs (441)
6809 - mark OSCs as active before reconnecting during recovery (438)
6810 - lov_enqueue and lov_cancel need to handle inactive OSTs (403)
6811 - lfind did not preserve OST order in output (443)
6812 - symlinks cause hung clients, incorrect data (439)
6813 - stop dereferencing request after dropping refcount (457)
6814 - don't LASSERT(spin_is_locked) on non-SMP (455)
6815 - fixes for many rename() bugs
6816 - fstat didn't correctly synchronize attributes (399)
6817 - server must handle lock cancellation during blocking AST prep (487)
6818 - bulk descriptors were free()d too soon (511)
6819 - fix paths in lconf, which would load incorrect modules (451, 507)
6820 - fix confusing lconf 'host not found' error message (386)
6821 - fix lock order deadlock on OST (O/R i_sem before journal ops, 478)
6822 - fix race condition in mdc_blocking_ast() for inode access (526)
6823 - fix lov_unpackmd() unpacking wrong number of stripes (537)
6824 - fix lov_set_osc_active() marking wrong OSC inactive (440)
6825 - fix bad lstripe lov_unpackmd() assertion (fix layering too) (527)
6826 - fix multiple writes of stripe MD to MDS (358, maybe 519)
6827 - fix lstripe in several ways (kernel side) (527)
6828 - fix request leak in ldlm_cli_enqueue (262)
6829 - incorrect OSC was marked inactive after OST failure
6830 - call mds_fs_cleanup before unmounting filesystem (524)
6831 - fix races between taking ns_lock and ldlm_lock_change_resource
6832 - fix races updating LOV export open file list
6833 - fix lov_enqueue error path, avoid decref-ing bad lock handle
6834 - fix recovery NULL deref in ldlm_cli_cancel_unused
6835 - fix some DLM races by using new hash table for lock handles (419)
6836 - permit the client to specify desired inodes, at replay
6837 - duplicate requests when we queue them for replay reintegration
6838 - fix last_rcvd offset calculation
6839 - sync after each recovered transaction, so we always make progress
6840 - never, not always, ERESTART requests without transnos
6841 - store the lov_desc in the MDS, so we don't depend on getlovinfo to
6843 - skip replay if the MDS says that the client is already connected
6844 - don't check for a recovery-enabled export to match lctl's UUID
6845 - don't INC_USE_COUNT for phantom exports
6846 - don't crash when cleaning up phantom exports (567)
6847 - don't double-finish or set replay data for errored mdc_open requests
6848 - abort requests when they time out, so we don't get old replies
6849 - send/receive replies for AST messages again
6850 - if the client says that it doesn't have the lock, cancel it on the
6852 - if we timeout during I/O, don't try to cancel an in-use lock; instead
6853 mark it as destroyed, it will all work out when decref is called
6854 - fix module use counts (22, 581)
6856 - ASTs now expect a reply (server cancels lock on error reply)
6858 2002-12-02 Andreas Dilger <adilger@clusterfs.com>
6861 - fix many simultaneous client startup (392)
6862 - fix dentry->d_it clobbering
6863 - credentials weren't being shipped for readdir/getattr operations
6864 - remove invalid assertions triggered during some concurrent MD
6866 - proper Lustre versions added (336, 389)
6867 - fix memory leak for create error case (398)
6868 - fix LOV locking bug that would get cli/srv out of sync
6869 - fix echo client over LOV (409)
6870 - fix dbench 2, extN refcount problem (170, 258, 356, 418)
6871 - fix double-O_EXCL intent crash (424)
6872 - avoid sending multiple lock CANCELs (352)
6874 - MDS can do multi-client recovery (modulo bugs in new code)
6876 - many updates, edits, cleanups
6878 2002-11-18 Phil Schwan <phil@clusterfs.com>
6881 - fix null d_it dereference (346)
6882 - fix full OST/dbench hang (333)
6883 - fix permission problem with file removal (286)
6884 - fix removal of OSCs from LOV when they fail
6885 - fix NULL deref during bulk timeout (214)
6886 - fix problems related to multiple filesystems on one MDS (241)
6887 - fixed serious subtle metadata locking bugs
6888 - free locks on clients when inodes are removed due to memory
6890 - fix inode pointer in lock data (285)
6891 - partial support for multiple MDS on a single host (241)
6892 - data locks weren't cancelled at clear_inode time (290, 311)
6893 - intent locks could lead to unbounded lock growth (205)
6894 - added a maximum lock count, an LRU list, and a flusher
6895 - fix multiple rename (365)
6896 - properly abstracted the echo client
6897 - OSC locked 1 byte too many; fixed
6898 - rewrote brw callback code:
6899 - fixed recovery bugs related to LOVs (306)
6900 - fixed too-many-pages-in-one-write crash (191)
6901 - fixed (again) crash in sync_io_timeout (214)
6902 - probably fixed callback-related race (385)
6904 - Add capability to MDS protocol
6905 - LDLM cancellations and callbacks on different portals
6907 2002-10-28 Andreas Dilger <adilger@clusterfs.com>
6910 - limit client IOV size to PTL_MD_MAX_IOV (611336, 191)
6911 - defer open object destruction to close time (601981, 138)
6912 - open/close OST file handle in obdo (OBD_MD_FLHANDLE) (601981, 138)
6913 - move LDLM_ENQUEUE/CONVERT back to MDS portal (625069)
6914 - abstract ll_lookup2, fix ll_revalidate2 to use abstraction (256)
6915 - don't call obd_setattr in ll_file_release for destroyed objects
6916 * protocol change to lustre_msg: move |version| and add |flags|
6917 * protocol change to osc_punch: "start" in "o_size", "end" in "o_blocks"
6918 * lock replay: for LDLM_FL_REPLAY trust client to do right thing
6919 * added replay of create, unlink, link and rename operations during
6920 MDS failover; recovery should be much more robust now
6921 * remove failed OSCs from LOVs (only lov_create uses this so far)
6922 * the lustre-HOWTO was brought (more) up to date (582544)
6924 2002-10-23 Phil Schwan <phil@clusterfs.com>
6927 - in-use dentries weren't being reused properly (617851)
6928 - prevent multiple LDLM setup (599178)
6929 - fix LOV size calculations for truncate (617853)
6930 - fix client handling of MDS intent errors (POSIX)
6931 - fix permission bug in lovstripe.c test (624321)
6932 - fix MDS thread deadlock - move LDLM handler to DLM portal (625069)
6933 - truncate past end of file could corrupt data
6934 - proper cleanup after timeouts, crashes, etc (592524, 550815)
6935 - a race in recovery could return ETIMEDOUT to apps (623947)
6936 - building outside the source directory was fixed
6937 * the lustre-HOWTO was brought (more) up to date (582544)
6938 * major progress was made on recovery functionality
6940 2002-10-10 Phil Schwan <phil@clusterfs.com>
6943 - recovery deadlock fix
6944 - rm -rf causes LBUG fix (617817)
6945 - file open by multiple tasks fix (618962)
6946 - directory permissions bugs (602707 and 620007)
6947 - journal_stop fixed with locking (611313)
6948 - O_APPEND failures resolved (618273, perhaps 614459)
6949 - lconf PATH fix (619770)
6950 - IA64 build fix (621450)
6951 - RPC buffer sizes scale with amount of memory
6953 2002-10-01 Phil Schwan <phil@clusterfs.com>
6956 - locks would be cancelled without throwing away data pages,
6957 resulting in inconsistent data (605627)
6958 - inode attributes were not always being refreshed (605627, 612449)
6959 - lconf now continues to cleanup after lctl reports an error
6960 - MDS now enforces user permissions (602707)
6961 - lprocfs cleanup fixed, but not yet enabled (614157)
6962 - fixed infinite server hang, should a client not respond to an AST
6963 - avoid going into recovery if user calls readlink() with a buffer
6964 that's too small (613941)
6965 - AST RPCs no longer require replies (614867) -- this may be changed
6966 - don't crash server if client sends an IOV that's too big (611336)
6967 - fixed lock conversion deadlock (611892)
6968 - fixed the following of symlinks (614622)
6969 * recovery: the server can remove locks from a client that dies, other
6970 clients can make progress
6971 * more extN patch fixes
6972 * compile-time configurable ptlrpc buffer allocations
6974 - collaborative read cache document
6975 - Lustre Lite Performance CDR document-in-progress
6977 2002-09-20 Andreas Dilger <adilger@clusterfs.com>
6980 - fix typo in patch-2.4.18
6982 2002-09-20 Andreas Dilger <adilger@clusterfs.com>
6985 - clear ptlrpc request each time in handle_incoming_request()
6986 - unlink of files now destroys the object on the OST
6988 2002-09-19 Peter Braam <braam@clusterfs.com>
6990 * add hard link support
6991 * change obdfile creation method
6992 * kernel patch changed
6994 2002-09-19 Peter Braam <braam@clusterfs.com>
6997 - stack overflow bug in extN fixed
6999 2002-09-18 Andreas Dilger <adilger@clusterfs.com>
7001 * documentation updates
7002 - add man pages for config tools
7003 - update tests/README to describe testing with new config tools
7004 - finish metadata API descriptions
7005 * bug fixes and cleanups
7006 - statfs workaround for 16TB limit
7007 - LOV stripe allocation improved, can stripe on subset of OSTs
7008 - LOV file size/IO offset was wrong for files > 4GB in size
7009 - object EA data was being dropped, caused files to be unreadable
7010 - memory overflow with non-LOV OST caused memory corruption
7011 - fixed regression tests to work with new config tools, obdfilter
7012 - fixed bug when directory size became larger than 1 block
7013 - fixed bug (for single client case) when PWD was deleted
7014 - invalidate local directory pages when doing intent-based ops
7015 - avoid LDLM oops when lock callback contained bad data
7017 2002-09-09 Andreas Dilger <adilger@clusterfs.com>
7019 * documentation updates
7020 * bug fixes and cleanups
7021 - configuration tools
7024 - 64-bit compile warnings
7025 - 64-bit internal statfs data
7027 * test_brw on persistent OST devices
7029 * lprocfs (disabled)
7031 2002-09-04 Andreas Dilger <adilger@clusterfs.com>
7033 * documentation updates
7034 * bug fixes and cleanups
7035 * configuration tools
7037 2002-08-30 Peter J. Braam <braam@clusterfs.com>
7040 * many small fixes to 0_5_4
7041 * io/network handling
7042 * thinkos in MDS operations
7044 2002-08-24 Peter J. Braam <braam@clusterfs.com>
7047 * crucial basic fixes to 0.5.3
7048 * IOR, Iozone work over Elan
7051 2002-08-07 Phil Schwan <phil@clusterfs.com>
7052 * version 0_5_3, our first alpha
7053 * we use the new Portals iovs
7054 * documentation updates
7055 * bug fixes and cleanups
7056 * small changes in the DLM wire protocol
7058 2002-07-25 Peter J. Braam <braam@clusterfs.com>
7059 * version 0_5_1 with some initial stability,
7060 * locking on MD and file I/O.
7061 * documentation updates
7062 * several bug fixes since 0.5.0
7063 * small changes in wire protocol
7065 2002-07-18 Phil Schwan <phil@clusterfs.com>
7067 * delivered as Lustre Light Alpha
7068 * fixed a crash after handling invalid MDS requests
7069 * fixed directory pages for architectures with non-4k pages sizes
7071 2002-07-11 Andreas Dilger <adilger@clusterfs.com>
7072 * release version v0_4_4
7073 * Moves TCP acceptor to be on port 2432 (unused Coda port) instead
7075 * Fixes a number of interruption problems with OST operations.
7076 * Update documentation for portals header changes
7077 * Move all wire protocol structs/defines to lustre_idl.h
7078 * Fixes symlink length bug.
7079 * Add tcpdump to repository.
7081 2002-07-05 Andreas Dilger <adilger@clusterfs.com>
7082 * release version v0_4_3
7083 * Fixes statfs for inodes on extN.
7084 * Fixes bug in runtests which would delete /etc/hosts.
7085 * Use 64-bit object IDs wherever possible (not into VFS though)
7086 Remove ost_get_info, which is unused by lustre, and out of date.
7088 2002-07-03 Peter Braam <braam@clusterfs.com>
7089 * release version v0_4_2 Fixes a lookup error (type not passed)
7090 * move forward to head of Portals
7091 * move forward to latest Lustre kernel
7093 2002-06-25 Peter Braam <braam@clusterfs.com>
7094 * release version v0_4_1. Hopefully stable on single node use.