1 tbd Sun Microsystems, Inc.
4 2.6.16.60-0.37 (SLES 10),
5 2.6.18-128.1.6.el5 (RHEL 5),
6 2.6.22.14 vanilla (kernel.org)
7 * Client support for unpatched kernels:
8 (see http://wiki.lustre.org/index.php?title=Patchless_Client)
9 2.6.16 - 2.6.27 vanilla (kernel.org)
10 * Recommended e2fsprogs version: 1.41.5.sun2
11 * File join has been disabled in this release, refer to Bugzilla 16929.
12 * A new Lustre ADIO driver is available for MPICH2-1.0.7.
13 * NFS export disabled when stack size < 8192. Since the NFSv4 export
14 of Lustre filesystem with 4K stack may cause a stack overflow. For
15 more information, please refer to bugzilla 17630.
19 Description: abort bulk too early if client is reconnected
22 Frequency : rare, if used wide striped file and one ost in down.
24 Descriptoin: don't return error if we created a subset of objects for file.
25 Details : lov_update_create_set() uses set->set_success as index for created
26 objects, so if some requests failed, they will have hole at end of
27 array and we can use qos_shrink_lsm for allocate correct lsm.
29 Severity : enhancement
31 Description: implement async create (obd_async_create) method for osc, to avoid
32 too long waiting new ost objects with holding ldlm lock.
37 Description: fix lqs' reference which won't be put in some situations
38 Details : This patch fixes:
39 1. In quota_check_common(), this function will check quota
40 for user and group, but only send one return via "pending".
41 In most cases, the pendings should be same. But that is not
43 2. If quotaoff runs between lquota_chkquota() and
44 lquota_pending_commit(), the same thing will happen too.
46 - if (!ll_sb_any_quota_active(qctxt->lqc_sb))
49 ------------------------------------------------------------------------------
50 tbd Sun Microsystems, Inc.
52 * Support for kernels:
53 2.6.16.60-0.37 (SLES 10),
54 2.6.18-128.1.6.el5 (RHEL 5),
55 2.6.22.14 vanilla (kernel.org)
56 * Client support for unpatched kernels:
57 (see http://wiki.lustre.org/index.php?title=Patchless_Client)
58 2.6.16 - 2.6.27 vanilla (kernel.org)
59 * Recommended e2fsprogs version: 1.41.5-sun2
60 * File join has been disabled in this release, refer to Bugzilla 16929.
61 * A new Lustre ADIO driver is available for MPICH2-1.0.7.
62 * NFS export disabled when stack size < 8192. Since the NFSv4 export
63 of Lustre filesystem with 4K stack may cause a stack overflow. For
64 more information, please refer to bugzilla 17630.
66 Severity : enhancement
68 Description: move AT tunable parameters for more consistent usage
69 Details : add AT tunables under /proc/sys/lustre, add to conf_param parsing
71 Severity : enhancement
73 Description: Update kernel to RHEL5.3 2.6.18-128.1.6.el5.
75 Severity : enhancement
77 Description: Update kernel to SLES10 SP2 2.6.16.60-0.37.
79 Severity : enhancement
81 Description: Caching OSS
82 Details : introduce data caching on the OSS. The OSS now relies on the linux
83 kernel page cache to keep recently accessed data in memory.
84 It is worth noting that all write requests are still flushed
85 synchronously as in lustre 1.6.
87 Severity : enhancement
89 Description: version based recovery
90 Details : introduce finer grained recovery able to detect transaction
91 dependencies and can deal with transaction gaps.
93 Severity : enhancement
95 Description: Enable adaptive timeouts by default
99 Descriptoin: correctly skip time estimate if in recovery
100 Details : rq_send_state insn't bitmask so using bitwise ops is forbid.
104 Descriptoin: fix goal inodes
105 Details : Allocate inodes for llog in last inode group for avoid broken
110 Descriptoin: OSS DeadLock
111 Details : Use trylock to prevent deadlock when shrink icache.
113 Severity : enhancement
115 Description: Allow tuning service thread via /proc
116 Details : For each service a new
117 /proc/fs/lustre/{service}/*/thread_{min,max,started} entry is
118 created that can be used to set min/max thread counts, and get the
119 current number of running threads.
121 Severity : enhancement
123 Description: Add state history info file, enhance import info file
124 Details : Track import connection state changes in a new osc/mdc proc file;
125 add overview-type data to the osc/mdc import proc file.
129 Description: Reduce small size read RPC
130 Details : Set read-ahead limite for every file and only do read-ahead when
131 available read-ahead pages are bigger than 1M to avoid small size
136 Description: free_entry erroneously used groups_free instead of put_group_info
138 Severity : enhancement
140 Description: Make read-ahead stripe size aligned.
142 Severity : enhancement
144 Description: MDS create should not wait for statfs RPC while holding DLM lock.
149 Descriptoin: OST grant too much space to client even there are not enough space.
150 Details : Client will shrink its grant cache to OST if there are no write
151 activity over 6 mins (GRANT_SHRINK_INTERVAL), and OST will retrieve
152 this grant cache if there are already not enough avaible space
153 (left_space < total_clients * 32M).
155 Severity : enhancement
157 Description: Update client support to vanilla kernels up to 2.6.27.
160 Frequency : rare, connect and disconnect target at same time
162 Descriptoin: ASSERTION(atomic_read(&imp->imp_inflight) == 0
163 Details : Don't call obd_disconnect under lov_lock. this long time
164 operation and can block ptlrpcd which answer to connect request.
166 Severity : enhancement
168 Description: Update to sles9 kernel-2.6.5-7.315.
171 Frequency : start MDS on uncleanly shutdowned MDS device
173 Descriptoin: ll_sync thread stay in waiting mds<>ost recovery finished
174 Details : Stay in waiting mds<>ost recovery finished produce random bugs
175 due race between two ll_sync thread for one lov target. send
176 ACTIVATE event only if connect realy finished and import have
180 Frequency : start MDS on uncleanly shutdowned MDS device
182 Descriptoin: aborting recovery hang on MDS
183 Details : Don't throttle destroy RPCs for the MDT.
185 Severity : enhancement
187 Descriptoin: Don't sync journal after every i/o
188 Details : Implement write RPC replay to allow server replies for write RPCs
189 before data is on disk. However, this feature is disabled by
190 default since some issues leading to data corruptions have been
191 found during recovery (e.g. bug 19128). This feature can be
192 enabled by running the following command on the OSSs:
193 lctl set_param obdfilter.*.sync_journal=0
197 Description: Slow reads beyond 8Tb offsets.
198 Details : Page index integer overflow in ll_read_ahead_page
202 Description: Soft lockup on OSS after MDS failover
203 Details : MSG_CONNECT_INITIAL is not set on the initial MDS->OST connect.
204 As a conseqence, the patch from bug 18224 is not operational
205 and the MDS export cannot be reused on the OSTs until it gets
209 Frequency : rare, only if using MMP with Linux RAID
211 Description: MMP doesn't work with Linux RAID
212 Details : While using HA for Lustre servers with Linux RAID, it is possible
213 that MMP will not detect multiple mounts. To make this work we
214 need to unplug the device queue in RAID when the MMP block is being
215 written. Also while reading the MMP block, we should read it from
216 disk and not the cached one.
219 Frequency : rare, during recovery
221 Description: Assertion failure in ldlm_lock_put
222 Details : Do not put cancelled locks into replay list, hold references on
227 Description: 1.6.5 mdsrate performance is slower than 1.4.11 (MDS not cpu bound)
228 Details : create_count always drops to the min value (=32) because grow_count
229 is being changed before the precreate RPC completes.
232 Frequency : Only in RHEL5 when mounting multiple ext3 filesystems
235 Description: "kmem_cache_create: duplicate cache jbd_4k" error message
236 Details : add proper locking for creation of jbd_4k slab cache
240 Description: MMP check in ext3_remount() fails without displaying any error
241 Details : When multiple mount protection fails during remount, proper error
244 -------------------------------------------------------------------------------
245 2008-12-31 Sun Microsystems, Inc.
247 * Support for kernels:
248 2.6.5-7.314 (SLES 9),
249 2.6.9-67.0.22.EL (RHEL 4),
250 2.6.16.60-0.31 (SLES 10),
251 2.6.18-92.1.17.el5 (RHEL 5),
252 2.6.22.14 vanilla (kernel.org)
253 * Client support for unpatched kernels:
254 (see http://wiki.lustre.org/index.php?title=Patchless_Client)
255 2.6.16 - 2.6.22 vanilla (kernel.org)
256 * Client support for unpatched kernels:
257 we do not recommend using patchless RHEL4 clients with kernels
258 prior to 2.6.9-55EL (RHEL4U5).
259 * Recommended e2fsprogs version: 1.40.11-sun1
260 * Note that reiserfs quotas are disabled on SLES 10 in this kernel.
261 * RHEL 4 and RHEL 5/SLES 10 clients behaves differently on 'cd' to a
262 removed cwd "./" (refer to Bugzilla 14399).
263 * A new quota file format has been introduced in 1.6.5.
264 The format conversion from prior releases is handled transparently,
265 but releases older than 1.4.12/1.6.5 don't understand this new
266 format. The automatic format conversion can be avoided by running
267 the following command on the MDS:
268 'tunefs.lustre --param="mdt.quota_type=ug1" $MDTDEV'.
269 For more information, please refer to bugzilla 13904.
270 * A new quota file format was introduced in 1.6.6/1.8.0 (kernel 2.6.16+)
271 The format conversion from prior releases is handled transparently,
272 but releases older than 1.6.6/1.8.0 don't understand this new
273 format. The automatic format conversion can be avoided by running
274 the following commands on the MDS and OSS servers (for
275 pre 1.4.12-1.6.5 quota files):
276 'tunefs.lustre --param="mdt.quota_type=ug1" $MDTDEV',
277 'tunefs.lustre --param="ost.quota_type=ug1" $MDTDEV'
278 or (for 1.4.12/1.6.5 quota files)
279 'tunefs.lustre --param="mdt.quota_type=ug2" $MDTDEV',
280 'tunefs.lustre --param="ost.quota_type=ug2" $MDTDEV'
281 For more information, please refer to bugzilla 13904.
282 * Output of lfs quota has been made less detailed by default,
283 old (verbose) output can be obtained by using -v option.
284 * File join has been disabled in this release, refer to Bugzilla 16929.
285 * A new Lustre ADIO driver is available for MPICH2-1.0.7.
286 * NFS export disabled when stack size < 8192. Since the NFSv4 export of
287 Lustre filesystem with 4K stack may cause a stack overflow. For more
288 information, please refer to bugzilla 17630.
290 Severity : enhancement
292 Description: add lazystatfs mount option to allow statfs(2) to skip down OSTs
293 Details : Allow skip disconnected ost for send statfs request and hide error
297 Frequency : rare, on llog test 6
299 Descriptoin: don't allow connect to already connected import
300 Details : Connect to already connected import hides connection problem.
303 Frequency : rare, on failed llog setup
305 Descriptoin: don't leak obd reference on failed llog setup
306 Details : For failed llog setup - mgc forget call class_destroy_import for
307 client import, move destroy import to more generic place.
312 Descriptoin: allow kill process which wait statahead result
313 Details : For some reasons 'ls' can stick in waiting result from statahead,
314 in this case need way for kill this process.
317 Frequency : rare, at shutdown
319 Descriptoin: panic at umount
320 Details : llap_shrinker can be raced with killing super block from list and
321 this produce panic with access to already freeded pointer
326 Descriptoin: don't lose wakeup for imp_recovery_waitq
327 Details : Recover_import_no_retry or invalidate_import and import_close can
328 both sleep on imp_recovery_waitq, but we was send only one wakeup
334 Descriptoin: panic in mds_open
335 Details : don't confuse mds_finish_transno() with PTR_ERR(-ENOENT)
340 Descriptoin: stuck in cache_remove_extent() or panic with accessing to already
342 Details : Release lock refernce only after add page to pages list.
345 Frequency : always with long access acl
347 Descriptoin: mds can't pack reply with long acl.
348 Details : mds don't control size of acl but they limited by reint/getattr
353 Frequency : on remount
355 Description: external journal device not working after the remount
356 Details : clear dev_rdonly flag for external journal devices in
362 Description: shutdown vs evict race
363 Details : client_disconnect_export vs connect request race.
364 if client will evicted at this time - we start invalidate
365 thread without referece to import and import can be freed
371 Description: shrink LOV EAs before replying
372 Details : correctly adjust LOV EA buffer for reply.
377 Description: don't skip ost target if they assigned to file
378 Details : Drop slow OSCs if we can, but not for requested start idx.
379 This means "if OSC is slow and it is not the requested
380 start OST, then it can be skipped, otherwise skip it only
381 if it is inactive/recovering/out-of-space.
383 Severity : enhancement
385 Description: Update to RHEL5 kernel-2.6.18-92.1.17.el5.
388 Frequency : rare, need acl's on inode.
390 Description: client can't handle ost additional correctly
391 Details : if ost was added after client connected to mds client can have
392 hit lnet_try_match_md ... to big messages to wide striped files.
393 in this case need teach client to handle config events about add
394 lov target and update client max ea size at that event.
396 Severity : enhancement
398 Description: Update to sles9 kernel-2.6.5-7.314.
400 Severity : enhancement
402 Description: Update to SLES10 SP2 kernel-2.6.16.60-0.31.
405 Frequency : Create a symlink file with a very long name
407 Description: ldlm_cancel_pack()) ASSERTION(max >= dlm->lock_count + count)
408 Details : If there is no extra space in the request for early cancels,
409 ldlm_req_handles_avail() returns 0 instead of a negative value.
414 Description: mds is deadlocked
415 Details : In rare cases, inode in catalog can have i_no less than have parent
416 i_no, this produce wrong order for locking during open, and
417 parallel unlink can be lock open. this need teach mds_open to grab
418 locks in resource id order, not at parent -> child order.
420 Severity : enhancement
422 Description: Add /proc entry for import status
423 Details : The mdc, osc, and mgc import directories now have
424 an import directory that contains useful import data for debugging
427 Severity : enhancement
429 Description: Re-disable certain /proc logging
430 Details : Enable and disable client's offset_stats, extents_stats and
431 extents_stats_per_process stats logging on the fly.
434 Frequency : Only on FC kernels 2.6.22+
436 Description: oops in statahead
437 Details : Do not drop reference count for the dentry from VFS when lookup,
438 VFS will do that by itself.
440 Severity : enhancement
442 Description: Generic /proc file permissions
443 Details : Set /Proc file permissions in a more generic way to enable non-
444 root users operate on some /proc files.
448 Description: Hitting mdc_commit_close() ASSERTION
449 Details : Properly handle request reference release in
450 ll_release_openhandle().
454 Frequency : only patchless client
455 Description: add workaround for race between add/remove dentry from hash
457 Severity : enhancement
459 Description: Allow OST glimpses to return PW locks
463 Description: LBUG when llog conf file is full
464 Details : When llog bitmap is full, ENOSPC should be returned for plain
469 Description: Prevent import from entering FULL state when server in recovery
473 Description: service mount cannot take device name with ":"
474 Details : Only when device name contains ":/" will mount treat it as
480 Description: Replace ptlrpcd with the statahead thread to interpret the async
481 statahead RPC callback
485 Frequency : on recovery
486 Description: I/O failures after umount during fail back
487 Details : If client reconnected to restarted server we need join to recovery
488 instead of find server handler is changed and process self eviction
489 with cancel all locks.
491 Severity : enhancement
493 Description: Update to RHEL5 kernel-2.6.18-92.1.10.el5.
497 Description: Kernel BUG tries to release flock
498 Details : Lustre does not destroy flock lock before last reference goes
499 away. So always drop flock locks when client is evicted and
500 perform unlock regardless of successfulness of speaking to MDS.
502 Severity : enhancement
504 Description: Update to SLES10 SP2 kernel-2.6.16.60-0.27.
506 Severity : enhancement
508 Description: Upcall on Lustre log has been dumped
509 Details : Allow for a user mode script to be called once a Lustre log has
510 been dumped. It passes the filename of the dumped log to the
511 script, the location of the script can be specified via
512 /proc/sys/lnet/debug_log_upcall.
517 Description: avoid messages about idr_remove called for id that is not allocated
518 Details : Move assigment s_dev for clustered nfs to end of initialization,
519 for avoid problem with error handling.
524 Description: avoid Already found the key in hash [CONN_UNUSED_HASH] messages
525 Details : When connection is reused this not moved from CONN_UNUSED_HASH into
526 CONN_USED_HASH and this prodice warning when put connection again
532 Description: avoid ASSERTION(client_stat->nid_exp_ref_count == 0) failed
533 Details : Release reference to stats when client disconnected, not
534 when export destroyed for avoid races when client destroyed
535 after main ost export.
539 Description: more cleanup in mds_lov
540 Details : Add workaround for get valid ost count for avoid warnings about
541 drop too big messages, not init llog cat under semphore which
542 can be blocked on reconnect and break normal replay, fix access
545 Severity : enhancement
547 Description: File striping can now be set to use an arbitrary pool of OSTs.
549 Severity : enhancement
551 Description: Export bytes_read/bytes_write count on OSC/OST.
555 Description: Early reply size mismatch, MGC loses connection
556 Details : Apply the MGS_CONNECT_SUPPORTED mask at reconnect time so
557 the connect flags are properly negotiated.
561 Description: Properly propagate oinfo flags from lov to osc for statfs
562 Details : Restore missing copy oi_flags to lov requests.
566 Description: exports in /proc are broken
567 Details : recreate /proc entries for clients when they reconnect.
569 Severity : enhancement
571 Description: Add man pages for llobdstat(8), llstat(8), plot-llstat(8),
572 : l_getgroups(8), lst(8), routerstat(8)
573 Details : Included man pages for llobdstat(8), llstat(8),
574 : plot-llstat(8), l_getgroups(8), lst(8), routerstat(8)
576 Severity : enhancement
578 Description: Implement lustre ll_show_options method.
580 Severity : enhancement
582 Description: Update to SLES9 kernel-2.6.5-7.312.
584 Severity : enhancement
586 Description: Update to RHEL4 kernel-2.6.9-67.0.22.EL.
590 Description: exports in /proc are broken
591 Details : Recreate /proc entries for clients when they reconnect.
595 Description: don't fail open with -ERANGE
596 Details : If client connected until mds will be know about real ost count
597 get LOV EA can be fail because mds not allocate enougth buffer
602 Description: Resolve device initialization race
603 Details : Prevent proc handler from accessing devices added to the
604 obd_devs array but yet be intialized.
606 Severity : enhancement
608 Description: configure's --enable-quota should check the
609 kernel .config for CONFIG_QUOTA
610 Details : Configure is terminated if --enable-quota is passed but
611 no quota support is in kernel
613 Severity : enhancement
615 Description: Update to SLES10 SP2 kernel-2.6.16.60-0.23.
617 Severity : enhancement
619 Description: Update to RHEL5 kernel-2.6.18-92.1.6.el5.
623 Frequency : rare, on PPC clients
624 Description: don't swab ost objects in response about directory, because
626 Details : Bug similar bug 14856, but in different function.
628 Severity : enhancement
630 Description: lfs quota tool enhancement
631 Details : Added units specifiers support for setquota, default to
632 current uid/gid for quota report, short quota stats by
633 default, nonpositional parameters for setquota, added
634 llapi_quotactl manual page.
636 Severity : enhancement
638 Description: *optional* service tags registration
639 Details : If the "service tags" package is installed on a Lustre node
640 When the filesystem is mounted, a local-node service tag will
641 be created. See http://inventory.sun.com/ for more information
642 about the Service Tags asset management system.
646 Description: Client runs out of low memory
647 Details : Consider only lowmem when counting initial number of llap pages
650 Description: Kernel BUG tries to release flock
651 Details : Lustre does not destroy flock lock before last reference goes
652 away. So always drop flock locks when client is evicted and
653 perform unlock regardless of successfulness of speaking to MDS.
656 Frequency : occasional
658 Description: add refcount for osc callbacks, so avoid panic on shutdown
660 Severity : enhancement
662 Description: Update to RHEL4 kernel-2.6.9-67.0.20.
665 Frequency : testing only
667 Description: sanity test 65a fails if stripecount of -1 is set
668 Details : Handle -1 striping on filesystem in ll_dirstripe_verify
671 Frequency : only in unusual configurations
673 Description: Kernel panic with find ost index.
674 Details : lov_obd have panic if some OST's have sparse indexes.
676 Severity : enhancement
678 Description: Update to RHEL5 kernel-2.6.18-53.1.21.el5.
681 Frequency : rarely, if filesystem is mounted with -o flock
683 Description: do not process already freed flock
684 Details : Flock can possibly be freed by another thread before it reaches
685 to ldlm_flock_completion_ast.
688 Frequency : rarely, if filesystem is mounted with -o flock
690 Description: LBUG during stress test
691 Details : Need properly lock accesses the flock deadlock detection list.
694 Frequency : rarely, if binaries are being run from Lustre
696 Description: oops in page fault handler
697 Details : Kernel page fault handler can return two special 'pages' in
698 error case, don't try dereference NOPAGE_SIGBUS and NOPAGE_OMM.
701 Frequency : rarely, during shutdown
703 Description: timeout with invalidate import.
704 Details : ptlrpcd_check call obd_zombie_impexp_cull and wait request which
705 should be handled by ptlrpcd. This produce long age waiting and
706 -ETIMEOUT ptlrpc_invalidate_import and as result LASSERT.
712 Description: ASSERTION(CheckWriteback(page,cmd)) failed
713 Details : Badly clear PG_Writeback bit in ll_ap_completion can produce false
717 Frequency : only with broken builds/installations
719 Description: no LBUG if lquota.ko and fsfilt_ldiskfs.ko are different versions
720 Details : Just return an error to a user, put a console error message.
722 Severity : enhancement
724 Description: Update to RHEL5 kernel-2.6.18-53.1.19.el5.
726 Severity : enhancement
728 Description: Update to RHEL4 kernel-2.6.9-67.0.15.
730 Severity : enhancement
732 Description: enable MGS and MDT services start separately
733 Details : add a 'nomgs' option in mount.lustre to enable start a MDT with
734 a co-located MGS without starting the MGS, which is a complement
735 to 'nosvc' mount option.
738 Frequency : always, on big-endian systems
740 Description: cleanup in ptlrpc code, related to PPC platform
741 Details : store magic in native order avoid panic's in recovery on PPC
742 node and forbid from this error in future. Also fix posibily
743 of twice swab data. Fix get lov striping to userpace.
746 Frequency : rarely, if replay get lost on server
748 Description: server incorrectly drop resent replays lead to recovery failure.
749 Details : do not drop replay according to msg flags, instead we check the
750 per-export recovery request queue for duplication of transno.
753 Frequency : after recovery
755 Description: precreate to many object's after del orphan.
756 Details : del orphan st in oscc last_id == next_id and this triger growing
757 count of precreated objects. Set flag LOW to skip increase count
758 of precreated objects.
761 Frequency : rare, on clear nid stats
763 Description: ASSERTION(client_stat->nid_exp_ref_count == 0)
764 Details : when clean nid stats sometimes try destroy live entry,
765 and this produce panic in free.
768 Frequency : occasionally since 1.6.4
770 Description: Stack overflow during MDS log replay
771 Details : ease stack pressure by using a thread dealing llog_process.
774 Frequency : very rare
776 Description: MDT cannot be unmounted, reporting "Mount still busy"
777 Details : Mountpoint references were being leaked during open reply
778 reconstruction after an MDS restart. Drop mountpoint reference
779 in reconstruct_open() and free dentry reference also.
784 Description: wait until IO finished before start new when do lock cancel.
785 Details : VM protocol want old IO finished before start new, in this case
786 need wait until PG_writeback is cleared until check dirty flag
787 and call writepages in lock cancel callback.
792 Description: mds_mfd_close() ASSERTION(rc == 0)
793 Details : In mds_mfd_close(), we need protect inode's writecount change
794 within its orphan write semaphore to prevent possible races.
797 Frequency : rare, on shutdown ost
799 Description: don't hit live lock with umount ost.
800 Details : shrink_dcache_parent can be in long loop with destroy dentries,
801 use shrink_dcache_sb instead.
804 Frequency : only when echo_client is used
806 Description: don't panic with use echo_client
807 Details : echo client pass NULL as client nid pointer and this produce NULL
811 Frequency : Always on 32-bit PowerPC systems
813 Description: fix build on PPC32
814 Details : compile code with -m64 flag produce wrong object file for PPC32.
819 Description: MDS LBUG: ASSERTION(!IS_ERR(dchild))
820 Details : In reconstruct_* functions, LASSERTs on both the data supplied
821 by a client, and the data on disk are dangerous and incorrect.
822 Change them with client eviction.
824 Severity : enhancement
826 Description: skiplist implementation simplification
827 Details : skiplists are used to group compatible locks on granted list
828 that was implemented as tracking first and last lock of each lock
829 group the patch changes that to using doubly linked lists
833 Description: delete compatibility for 32bit qdata
834 Details : as planned, when lustre is beyond b1_8, lquota won't support 32bit
835 qunit. That means servers of b1_4 and servers of b1_8 can't be
836 used together if users want to use quota.
839 Frequency : only with administrator action
841 Description: mount failure if config log has invalid conf_param setting
842 Details : If administrator specified an incorrect configuration parameter
843 with "lctl conf_param" this would cause an error during future
844 client mounts. Instead, ignore the bad configuration parameter.
847 Frequency : blocks per group < blocksize*8 and uninit_groups is enabled
849 Description: ldiskfs error: XXX blocks in bitmap, YYY in gd
850 Details : If blocks per group is less than blocksize*8, set rest of the
854 Frequency : Application do stride read on lustre
856 Description: The read performance will drop a lot if the application does
858 Details : Because the stride_start_offset are missing in stride read-ahead,
859 it will cause clients read a lot of unused pages in read-ahead,
860 then the read-performance drops.
864 Description: more ldlm soft lockups
865 Details : In ldlm_resource_add_lock(), call to ldlm_resource_dump()
866 starve other threads from the resource lock for a long time in
867 case of long waiting queue, so change the debug level from
868 D_OTHER to the less frequently used D_INFO.
870 Severity : enhancement
872 Description: add -gid, -group, -uid, -user options to lfs find
874 Severity : enhancement
876 Description: ll_recover_lost_found_objs - recover objects in lost+found
877 Details : OST corruption and subsequent e2fsck can leave objects in the
878 lost+found directory. Using the "ll_recover_lost_found_objs"
879 tool, these objects can be retrieved and data can be salvaged
880 by using the object ID saved in the fid EA on each object.
885 Description: this bug _only_ happens when inode quota limitation is very low
886 (less than 12), so that inode quota unit is 1 at initialization.
887 Details : if remaining quota equates 1, it is a sign to demonstate that quota
888 is effective now. So least quota qunit should be 2.
892 Description: Hung threads in invalidate_inode_pages2_range
893 Details : The direct IO path doesn't call check_rpcs to submit a new RPC once
894 one is completed. As a result, some RPCs are stuck in the queue
899 Description: Procfs and llog threads access destoryed import sometimes.
900 Details : Sync the import destoryed process with procfs and llog threads by
901 the import refcount and semaphore.
905 Description: mds fails to respond, threads stuck in ldlm_completion_ast
906 Details : Sort source/child resource pair after updating child resource.
911 Description: kernel BUG at ldiskfs2_ext_new_extent_cb
912 Details : If insertion of an extent fails, then discard the inode
913 preallocation and free data blocks else it can lead to duplicate
918 Description: don't always update ctime in ext3_xattr_set_handle()
919 Details : Current xattr code updates inode ctime in ext3_xattr_set_handle()
920 In some cases the ctime should not be updated, for example for
921 2.0->1.8 compatibility it is necessary to delete an xattr and it
922 should not update the ctime.
926 Description: add quota statistics
927 Details : 1. sort out quota proc entries and proc code.
928 2. add quota statistics
933 Description: quotas are not honored with O_DIRECT
934 Details : all writes with the flag O_DIRECT will use grants which leads to
935 this problem. Now using OBD_BRW_SYNC to guard this.
939 Bugzilla : 15713/16362
940 Description: Assertion in iopen_connect_dentry in 1.6.3
941 Details : looking up an inode via iopen with the wrong generation number can
942 populate the dcache with a disconneced dentry while the inode
943 number is in the process of being reallocated. This causes an
944 assertion failure in iopen since the inode's dentry list contains
945 both a connected and disconnected dentry.
949 Description: assertion failure in ldlm_handle2lock()
950 Details : fix a race between class_handle_unhash() and class_handle2object()
951 introduced in lustre 1.6.5 by bug 13622.
953 Severity : enhancement
955 Description: superblock lock contention with many SMP cores on one client
956 Details : several client filesystem locks were highly contended on SMP
957 NUMA systems with 8 or more cores. Per-CPU datastructures
958 and more efficient locking implemented to reduce contention.
963 Description: Kernel BUG: sd_iostats_bump: unexpected disk index
964 Details : remove the limit of 256 scsi disks in the sd_iostat patch
969 Description: oops in sd_iostats_seq_show()
970 Details : unloading/reloading the scsi low level driver triggers a kernel
971 bug when trying to access the sd iostat file.
976 Description: Kernel panics during QLogic driver reload
977 Details : REQ_BLOCK_PC requests are not handled properly in the sd iostat
978 patch, causing memory corruption.
983 Description: journal_dev option does not work in b1_6
984 Details : pass mount option during pre-mount.
986 Severity : enhancement
988 Description: Add a FIEMAP(FIle Extent MAP) ioctl for ldiskfs
989 Details : FIEMAP ioctl will allow an application to efficiently fetch the
990 extent information of a file. It can be used to map logical blocks
991 in a file to physical blocks in the block device.
994 Frequency : only with adaptive timeout enabled
996 Description: DEBUG_REQ() bad paging request
997 Details : ptlrpc_at_recv_early_reply() should not modify req->rq_repmsg
998 because it can be accessed by reply_in_callback() without the
1002 Frequency : only on Cray X2
1004 Description: X2 build failures
1005 Details : fix build failures on Cray X2.
1009 Description: xid & resent requests
1010 Details : Initialize RPC XID from clock at startup (randomly if clock is
1015 Description: quota recovery deadlock during mds failover
1016 Details : This patch includes att18982, att18236, att18237 in bz14840.
1018 1. fix osts hang when mds does failover with quotaon
1019 2. prevent watchdog storm when osts threads wait for the
1024 Description: kernel panic on racer
1025 Details : Do not access dchild->d_inode when IS_ERR(dchild) is true.
1027 Severity : enhancement
1029 Description: Add lustre_start utility to start or stop multiple Lustre servers
1034 Description: Lustre GPF in {:ptlrpc:ptlrpc_server_free_request+373}
1035 Details : In case of memory pressure, list_del() can be called twice on
1036 req->rq_history_list, causing a kernel oops.
1040 Description: (ptllnd_peer.c:557:kptllnd_peer_check_sends()) ASSERTION(!in_interrupt()) failed
1041 Details : fix stack overflow in the distributed lock manager by defering
1042 export eviction after a failed ast to the elt thread instead of
1043 handling it in the dlm interpret routine.
1045 Severity : enhancement
1047 Description: More exported tunables for mballoc
1048 Details : Add support for tunable preallocation window and new tunables for
1049 large/small requests
1053 Description: Detect corruption of block bitmap and checking for preallocations
1054 Details : Checks validity of on-disk block bitmap. Also it does better
1055 checking of number of applied preallocations. When corruption is
1056 found, it turns filesystem readonly to prevent further corruptions.
1060 Frequency : only for big-endian servers
1061 Description: Check for big-endian system while mounting fs with extents feature
1062 Details : Mounting a filesystem with extents feature will fail on big-endian
1063 systems since ext3-based ldiskfs is not supported on big-endian
1064 systems. Can be over-ridden with "bigendian_extents" mount option.
1068 Description: Excessive recovery window
1069 Details : With AT enabled, the recovery window can be excessively long (6000+
1070 seconds). To address this problem, we no longer use
1071 OBD_RECOVERY_FACTOR when extending the recovery window (the connect
1072 timeout no longer depends on the service time, it is set to
1073 INITIAL_CONNECT_TIMEOUT now) and clients report the old service
1074 time via pb_service_time.
1078 Description: Watchdog triggered on MDS failover
1079 Details : enable OBD_CONNECT_MDT flag when connecting from the MDS so that
1080 the OSTs know that the MDS "UUID" can be reused for the same export
1081 from a different NID, so we do not need to wait for the export to
1086 Descriptoin: Lustre detected file system corruption with inode out of bounds
1087 Details : don't update i_size on MDS_CLOSE for directories. This causes
1088 directory corruptions on the MDT.
1090 -------------------------------------------------------------------------------
1092 2008-05-26 Sun Microsystems, Inc.
1094 * Support for kernels:
1095 2.6.5-7.311 (SLES 9),
1096 2.6.9-67.0.7.EL (RHEL 4),
1097 2.6.16.54-0.2.5 (SLES 10),
1098 2.6.18-53.1.14.el5 (RHEL 5),
1099 2.6.22.14 vanilla (kernel.org)
1100 * Client support for unpatched kernels:
1101 (see http://wiki.lustre.org/index.php?title=Patchless_Client)
1102 2.6.16 - 2.6.22 vanilla (kernel.org)
1103 * Due to problems with nested symlinks and FMODE_EXEC (bug 12652),
1104 we do not recommend using patchless RHEL4 clients with kernels
1105 prior to 2.6.9-55EL (RHEL4U5).
1106 * Recommended e2fsprogs version: 1.40.7-sun1
1107 * Note that reiserfs quotas are disabled on SLES 10 in this kernel.
1108 * RHEL 4 and RHEL 5/SLES 10 clients behaves differently on 'cd' to a
1109 removed cwd "./" (refer to Bugzilla 14399).
1110 * A new quota file format has been introduced in 1.6.5.
1111 The format conversion from prior releases is handled transparently,
1112 but releases older than 1.4.12/1.6.5 will not understand this new
1113 format. The automatic format conversion can be avoided by running
1114 the following command on the MDS before upgrading:
1115 'tunefs.lustre --param="mdt.quota_type=ug1" $MDTDEV'.
1116 For more information, please refer to bugzilla 13904.
1120 Description: quota performance fix
1121 Details : quota data is written in journalled mode instead of ordered to
1122 increase performance
1126 Description: lfs support for human-readable quota grace time strings
1127 Details : lfs setquota -t and lfs quota -t represent quota grace times
1128 in "XXwXXdXXhXXmXXs" format instead of large values in seconds
1131 Frequency : always with o2ib 1.3 and sles10
1133 Description: fix build with SLES10 and o2ib v3.
1134 Details : sles10 uses diffrent name for Module.symver file but configure
1135 assume this file has same name on RHEL/SLES/vanila kernels.
1138 Frequency : very rare, if additional xattrs are used on kernels >= 2.6.12
1140 Description: MDS may lose file striping (and hence file data) in some cases
1141 Details : If there are additional extended attributes stored on the MDS,
1142 in particular ACLs, SELinux, or user attributes (if user_xattr
1143 is specified for the client mount options) then there is a risk
1144 of attribute loss. Additionally, the Lustre file striping
1145 needs to be larger than default (e.g. striped over all OSTs),
1146 and an additional attribute must be stored initially in the
1147 inode and then increase in size enough to be moved to the
1148 external attribute block (e.g. ACL growing in size) for file
1151 Severity : enhancement
1153 Description: add message levels for liblustreapi
1158 Description: MDT cannot be unmounted, reporting "Mount still busy"
1159 Details : Mountpoint references were being leaked during open reply
1160 reconstruction after an MDS restart. Drop mountpoint reference
1161 in reconstruct_open() and free dentry reference also.
1166 Description: fix for occasional failure case of -ENOSPC in recovery-small tests
1167 Details : Move the 'good_osts' check before the 'total_bavail' check. This
1168 will result in an -EAGAIN and in the exit call path we call
1169 alloc_rr() which will with increasing aggressiveness attempt to
1170 aquire precreated objects on the minimum number of required OSCs.
1174 Description: Use old size assignment to avoid deadlock
1175 Details : This reverts the changes in bugs 2369 and bug 14138 that introduced
1176 the scheduling while holding a spinlock. We do not need locking
1177 for size in ll_update_inode() because size is only updated from
1178 the MDS for directories or files without objects, so there is no
1179 other place to do the update, and concurrent access to such inodes
1180 are protected by the inode lock.
1184 Description: Use __u64 instead of int for valid bits
1188 Description: resolve "_IOWR redefined" build error on SLES10
1192 Description: dump the memory debugging after all modules are unloaded to
1193 suppress false negative in conf_sanity test 39
1197 Description: the recovery timer never expires
1198 Details : for new client connect request, the recovery timer should not be
1199 reset, otherwise recovery timer will never expired, if the old
1200 client never come. Only old client connect and first connection
1201 req should trigger recovery timer reset.
1205 Description: the min numbers of lproc stats are wrong
1206 Details : adding a new constant LC_MIN_INIT and use it for initialization
1210 Frequency : always with interactive lfs
1212 Description: Reinitialize optind to 0 so that interactive lfs works in all cases
1215 Frequency : with multiple concurrent readdir processes in same directory
1216 Bugzilla : 15406, 15169, 15175
1217 Description: misc fixes for directory readahead.
1218 Details : prevent previous statahead async RPC callback from processing the
1219 current "statahead_info", race condition between async RPC callback
1220 add dentry into dentry hash table and "ls" thread revalidate such
1221 dentry, statahead his/miss control for hidden items, and so on.
1223 Severity : enhancement
1225 Description: build kernel-ib packages for OFED 1.3 in our release cycle
1229 Description: incore types cleaning in quota code (with respect to 64-bit limits)
1230 Details : several u32 variables declarations are replaced with u64 declarations
1235 Description: fix SLES kernel versioning
1236 Details : the kernel version for our SLES 10 kernel did not include a "-"
1237 before the "smp" at the end. while this was not a problem in
1238 general, it did mean that software trying to use the kernel
1239 version to try to detect a vendor specific kernel would fail.
1240 this was most evident by the OFED build scripts.
1245 Description: Don't update lov_desc members until making sure they are valid
1246 Details : When updating lov_desc members via proc fs, need fix their
1247 validities before doing the real update.
1250 Frequency : very rare
1252 Description: don't put request into delay list while invalidate in flight.
1253 Details : ptlrpc_delay_request sometimes put in delay list while invalidate
1254 import in flight. this produce timeout for invalidate and sometimes
1255 can cause stale data.
1257 Severity : enhancement
1259 Description: Update kernel to SLES9 2.6.5-7.311.
1261 Severity : enhancement
1263 Description: Update kernel to RHEL4 2.6.9-67.0.7.
1268 Frequency : on PPC only
1269 Description: not convert ost objects for directory because it's not exist.
1270 Details : ll_dir_getstripe assume dirrectory has ost objects but this wrong.
1272 Severity : enhancement
1274 Description: Fix warnings with compile liblustre at sles10/rhel5 which have
1275 __u64 as usingied long long type.
1278 Frequency : rare, on shutdown
1280 Description: race process ast vs remove callback
1281 Details : removing callback before disconnect import open race with
1282 processing callback.
1284 Severity : enhancement
1286 Description: Update kernel to SLES9 2.6.5-7.311.
1288 Severity : enhancement
1290 Description: Files open for execute are not marked busy on SLES10
1291 Details : Add FMODE_EXEC to SLES10 SP1 server kernel series.
1293 Severity : enhancement
1295 Description: Add server support for vanilla-2.6.22.14.
1298 Frequency : occasional
1300 Description: Avoid lov_create() getting stuck in obd_statfs_rqset()
1301 Details : If an OST is down the MDS will hang indefinitely in
1302 obd_statfs_rqset() waiting for the statfs data. While for
1303 MDS QOS usage of statfs, it should not stuck in waiting.
1305 Severity : enhancement
1307 Description: Disable adaptive timeouts by default
1310 Frequency : on network error
1312 Description: panic with double free request if network error
1313 Details : mdc_finish_enqueue is finish request if any network error occuring,
1314 but it's true only for synchronus enqueue, for async enqueue
1315 (via ptlrpcd) this incorrect and ptlrpcd want finish request
1319 Frequency : rare, on recovery
1321 Description: read procfs can produce deadlock in some situation
1322 Details : Holding lprocfs lock with send rpc can produce block for destroy
1323 obd objects and this also block reconnect with -EALREADY.
1324 This isn't fix all lprocfs bugs - but make it rare.
1326 Severity : enhancement
1328 Description: Update kernel to RHEL5 2.6.18-53.1.14.el5.
1331 Frequency : frequent on X2 node
1333 Description: mdc_set_open_replay_data LBUG
1334 Details : Set replay data for requests that are eligible for replay.
1339 Description: lustre_mgs: operation 101 on unconnected MGS
1340 Details : When MGC is disconnected from MGS long enough, MGS will evict the
1341 MGC, and late on MGC cannot successfully connect to MGS and a lot
1342 of the error messages complaining that MGS is not connected.
1345 Frequency : rare, depends on device drivers and load
1347 Description: MDS or OSS nodes crash due to stack overflow
1348 Details : Code changes in 1.6.4 increased the stack usage of some functions.
1349 In some cases, in conjunction with device drivers that use a lot
1350 of stack, the MDS (or possibly OSS) service threads could overflow
1351 the stack. One change which was identified to consume additional
1352 stack has been reworked to avoid the extra stack usage.
1354 Severity : enhancement
1356 Description: Update to RHEL5 latest kernel-2.6.18-53.1.13.el5.
1358 Severity : enhancement
1360 Description: Update to SLES10 SP1 latest kernel-2.6.16.54-0.2.5.
1362 Severity : enhancement
1364 Description: Update to RHEL5 latest kernel-2.6.18-53.1.6.el5.
1366 Serverity : enhancement
1368 Description: Update RHEL4 kernel to 2.6.9-67.0.4.
1371 Frequency : rare on shutdown OST
1373 Description: Don't allow skipping OSTs if index has been specified.
1374 Details : Don't allow skipping OSTs if index has been specified, make locking
1375 in internal create lots better.
1380 Description: ASSERTION(!PageDirty(page)) failed
1381 Details : Wrong check could lead to an assertion failure under specific
1387 Description: LBUG in ptlrpc_check_set() bad phase ebc0de00
1388 Details : access to bitfield in structure is always rounded to long
1389 and this produce problem with not atomic change any bit.
1394 Description: Lustre make rpms failed.
1395 Details : Remove ldiskfs spec file to avoids rpmbuild be confused when
1396 builds Lustre rpms from tarball.
1398 Severity : enhancement
1400 Description: Update to SLES9 SP4 kernel-2.6.5-7.308.
1403 Frequency : rare on shutdown OST
1405 Description: If llog cancel was not send before clean_exports phase, this can
1406 produce deadlock in llog code.
1407 Details : If llog thread has last reference to obd and call class_import_put
1408 this produce deadlock because llog_cleanup_commit_master wait when
1409 last llog_commit_thread exited, but this never success because was
1410 called from llog_commit_thread.
1413 Frequency : only if OST index is skipped
1415 Description: NULL lov_tgts causing MDS oops
1416 Details : more safe checks for NULL lov_tgts for avoid oops.
1418 Severity : enhancement
1420 Description: Update to RHEL4 latest kernel-2.6.9-67.0.1.EL.
1422 Severity : enhancement
1424 Description: Update to RHEL5 latest kernel-2.6.18-53.1.4.el5.
1429 Description: make mgs_setparam() handle fsname containing dash
1430 Details : fsname containing a dash does not work with lctl conf_param
1432 Severity : enhancement
1434 Description: Update to RHEL4 Update-6 kernel-2.6.9-67.EL.
1437 Frequency : rare, in recovery and (or) destroy lovobjid file.
1439 Description: rewrite lov objid code.
1440 Details : Cleanup for lov objid code, remove scability problems and wrong
1441 locking. Fix sending last_id into ost.
1443 Severity : enhancement
1445 Description: Update to SLES10 SP1 latest kernel-2.6.16.54-0.2.3.
1447 Severity : enhancement
1449 Description: Update to RHEL5 Update-1 kernel 2.6.18-53.el5.
1450 Details : Use d_move_locked instead of __d_move.
1453 Frequency : rare, at shutdown
1455 Description: access already free / zero obd_namespace.
1456 Details : if client_disconnect_export was called without force flag set,
1457 and exist connect request in flight, this can produce access to
1458 NULL pointer (or already free pointer) when connect_interpret
1459 store ocd flags in obd_namespace.
1462 Frequency : only at startup
1464 Description: not alloc memory with spinlock held.
1465 Details : allocation memory with GFP_KERNEL can produce sleep deadlock,
1466 if any spinlock held.
1471 Description: lfs find does not continue on file error
1472 Details : Continue other files processing when a file/dir is absent.
1476 Description: Inconsistent usage of lustre_pack_reply()
1477 Details : Standardize the usage of lustre_pack_reply() such that it
1478 always generate a CERROR on failure.
1481 Frequency : very rare
1483 Description: Fix replay if there is an un-replied request and open
1484 Details : In some cases, older replay request will revert the
1485 mcd->mcd_last_xid on MDS which is used to record the client's
1486 latest sent request.
1488 Severity : enhancement
1490 Description: Update to RHEL5 kernel 2.6.18-8.1.15.el5.
1492 Severity : enhancement
1494 Description: Update to SLES10 SP1 kernel 2.6.16.53-0.16
1496 Severity : enhancement
1498 Description: Update to SLES9 kernel-2.6.5-7.287.3.
1500 Severity : enhancement
1502 Description: Update to RHEL4 kernel-2.6.9-55.0.12.EL.
1504 Severity : enhancement
1506 Description: Build SLES10 patchless client fails
1507 Details : The configure was broken by run ./configure with
1508 --with-linux-obj=.... argument for patchless client. When the
1509 configure use --with-linux-obj, the LINUXINCLUDE= -Iinclude
1510 can't search header adequately. Use absolute path such as
1511 -I($LINUX)/include instead.
1513 Severity : enhancement
1515 Description: Lustre Page Accounting
1516 Details : New macros for page alloc and free which enable accounting
1517 of page allocation of Lustre. Use percpu counters to store memory
1518 and page statistics.
1521 Frequency : only if debugging is disabled
1523 Description: LASSERT_{REQ,REP}SWAB macros are buggy
1524 Details : If SWAB_PARANOIA is disabled, the LASSERT_REQSWAB and
1525 LASSERT_REPSWAB macros become no-ops, which is incorrect. Drop
1526 these macros and replace them with their definitions instead.
1531 Description: interrupt oig_wait produce painc on resend.
1532 Details : brw_redo_request can be used for resend requests from ptlrpcd and
1533 private set, and this produce situation when rq_ptlrpcd_data not
1534 copyed to new allocated request and triggered LBUG on assert
1535 req->rq_ptlrpcd_data != NULL. But this member used only for wakeup
1536 ptlrpcd set if request is changed and can be safety changed to use
1539 Severity : enhancement
1541 Description: organize the server-side client stats on per-nid basis
1542 Details : Change the structure of stats under obdfilter and mds to
1550 The "uuid"s file would list the uuids of _active_ exports.
1551 And the clear entry is to clear all stats and stale nids.
1556 Description: Processes looping in ll_readdir() on Lustre clients finally causing
1557 a full node pseudo-hang
1558 Details : Concurrent access to the same directory from multiple clients with
1559 intensive file creation/removal can cause a client node to spin in
1560 ll_readdir(). i_version must be increased every time the lock is
1561 cancelled to ensure a revalidate is done.
1566 Description: touch file failed when fs is not full
1567 Details : OST in recovery should not be discarded by MDS in alloc_qos(),
1568 otherwise we can get ENOSP while fs is not full.
1571 Frequency : only for Cray XT3
1572 Bugzilla : 12829/13455
1573 Description: Changing primary group doesn't change the group lustre assigns to
1575 Details : When CRAY_XT3 is defined, the fsgid supplied by the client is
1576 overridden with the primary group provided by the group upcall,
1577 whereas the supplied fsgid can be trusted if it is in the list of
1578 supplementary groups returned by the group upcall.
1580 Severity : enhancement
1582 Description: Root Squash Functionality
1583 Details : Implementation of NFS-like root squash capability. Specifically,
1584 don't allow someone with root access on a client node to be able
1585 to manipulate files owned by root on a server node.
1587 Severity : enhancement
1589 Description: Slow trucate/writes to huge files at high offsets.
1590 Details : Directly associate cached pages to lock that protect those pages,
1591 this allows us to quickly find what pages to write and remove
1592 once lock callback is received.
1597 Description: Too many locks accumulating on client during NFS usage
1598 Details : mds_open improperly used accmode to find out access mode to a
1599 file. Also mdc_intent_lock logic to find out if we already have
1600 lock similar to just received was flawed since introduction of
1601 skiplists - locks are now added to the front of the granted
1606 Description: Hit ASSERTION(obd->obd_stopping == 1) failed in some setup failed
1608 Details : In obd setup failure handler, obd_stopping will not necessarily to
1609 be 1, and obd_set_up should also be checked to make sure whether
1610 obd is completely setup.
1612 Severity : enhancement
1614 Description: Allow masking D_WARNING, D_ERROR messages from console
1615 Details : Console messages can now be disabled via lnet.printk.
1620 Description: User code with malformed file open parameter crashes client node
1621 Details : Before packing join_file req, all the related reference should be
1622 checked carefully in case some malformed flags cause fake join_file
1628 Description: shrink/enlarge qunit size when needed; fix the problem of coarse
1629 grain of quota doing harm to quota's accuracy
1630 Details : qunit size will be changed when quota limitation is too low/high;
1631 record the pending quota write in order to get more accureate
1632 quota; delete the patch for bug12588, which is unnecessary when
1633 this patch is landed. This bug also contains fixes for bug 14526,
1634 14299, 14601 and 13794.
1638 Description: LDLM_ENQUEUE races with LDLM_CP_CALLBACK
1639 Details : ldlm_completion_ast() assumes that a lock is granted when the req
1640 mode is equal to the granted mode. However, it should also check
1641 that LDLM_FL_CP_REQD is not set.
1645 Description: Heavy nfs access might result in deadlocks
1646 Details : After ELC code landed, it is now improper to enqueue any mds
1647 locks under och_sem, because enqueue might want to decide to
1648 cancel open locks for same inode we are holding och_sem for.
1652 Description: 35% write performance drop with ldiskfs2 when quotas are on
1653 Details : Enable ext3 journalled quota by default to improve performance
1654 when quotas are turned on.
1658 Description: Client eviction while running blogbench
1659 Details : A lot of unlink operations with concurrent I/O can lead to a
1660 deadlock causing evictions. To address the problem, the number of
1661 oustanding OST_DESTROY requests is now throttled to
1662 max_rpcs_in_flight per OSC and LDLM_FL_DISCARD_DATA blocking
1663 callbacks are processed in priority.
1666 Frequency : RHEL4 only
1668 Description: mkfs is very slow on IA64/RHEL4
1669 Details : A performance regression has been discovered in the MPT Fusion
1670 driver between versions 3.02.73rh and 3.02.99.00rh. As a
1671 consequence, we have downgraded the MPT Fusion driver in the RHEL4
1672 kernel from 3.02.99.00 to 3.02.73 until this problem is fixed.
1675 Frequency : PPC/PPC64 only
1677 Description: conflicts between asm-ppc64/types.h and lustre_types.h
1678 Details : fix duplicated definitions between asm-ppc64/types.h and
1679 lustre_types.h on PPC.
1682 Frequency : PPC/PPC64 only
1684 Description: asm-ppc/segment.h does not exist
1685 Details : fix compile issue on PPC.
1689 Description: data checksumming impacts single node performance
1690 Details : add support for several checksum algorithms. Currently, CRC32 and
1691 Adler-32 are supported. The checksum type can be changed on the fly
1692 through /proc/fs/lustre/osc/*/checksum_type.
1696 Description: use adler32 for page checksums
1697 Details : when available, use the Adler-32 algorithm instead of CRC32 for
1702 Description: better handle error messages in extents code
1704 Severity : enhancement
1706 Description: SNMP support enhancement
1707 Details : Adding total number of sampled request for an MDS node in snmp
1710 Severity : enhancement
1712 Description: Optimize ldlm waiting list processing for PR extent locks
1713 Details : When processing waiting list for read extent lock and meeting read
1714 lock that is same or wider to it that is not contended, skip
1715 processing rest of the list and immediatelly return current
1716 status of conflictness, since we are guaranteed there are no
1717 conflicting locks in the rest of the list.
1721 Description: Time out and refuse to reconnect
1722 Details : When the failover node is the primary node, it is possible
1723 to have two identical connections in imp_conn_list. We must
1724 compare not conn's pointers but NIDs, otherwise we
1725 can defeat connection throttling.
1729 Description: Client not clear own cache if answer to reconnect is lost.
1730 Details : Client gets evicted from server. Now client also thinks it is
1731 disconnected (or gets ENOTCONN on its operation) and decides to
1732 reconnect. Server receives reconnect message, but cannot find
1733 export. New export is created that is fully valid (new cookie!),
1734 but reply is lost and not reported to client. Client reconnects
1735 again and gets back a just-created connection, but it is not new
1736 so client thinks it was not evicted and does not do recovery.
1740 Description: Detect stride IO mode in read-ahead
1741 Details : When a client does stride read, read-ahead should detect that and
1742 read-ahead pages according to the detected stride pattern.
1746 Description: build for x2 fails
1747 Details : fix compile issue on Cray systems.
1749 Severity : enhancement
1751 Description: implement readv/writev APIs(aio_read/aio_writes in newer kernels)
1752 Details : This greatly improves speed of NFS writes on 2.6 kernels.
1755 Frequency : only on PPC/SLES10
1757 Description: "BITS_PER_LONG is not 32 or 64" in linux/idr.h
1758 Details : On SLES10/PPC, fs.h includes idr.h which requires BITS_PER_LONG to
1759 be defined. Add a hack in mkfs_lustre.c to work around this compile
1764 Description: LASSERT on MDS when client holding flock lock dies
1765 Details : ldlm pool logic depends on number of granted locks equal to
1766 number of released locks which is not true for flock locks, so
1767 just exclude such locks from consideration.
1771 Description: MDS deadlock with many ll_sync_lov threads and I/O stalled
1772 Details : Use fsfilt_sync() for both the whole filesystem sync and
1773 individual file sync to eliminate dangerous inode locking
1774 with I_LOCK that can lead to a deadlock.
1778 Description: Update an obsolete wirecheck.c generator
1779 Details : Update wirecheck.c/wirehdr.c and regenerate wiretest.c
1783 Description: Client can panic on open sometimes
1784 Details : It is possible that we try to free already freed request in
1785 ll_file_open in some error cases when we send request from
1790 Description: performance in 1.6.3
1791 Details : Force q->max_phys_segments to MAX_PHYS_SEGMENTS on SLES10 to be
1792 sure that 1MB requests are not fragmented by the block layer.
1796 Description: LDLM soft lockups - improvement
1797 Details : It is be possible to send the lock handle along with each read
1798 or write request because the client is already doing a lock match
1799 itself so there isn't any reason the OST should have to re-do that
1805 Description: lfs quota fails with deactivated OSTS
1806 Details : With this patch, three improvements are included:
1807 1. detete the softlimit in mds and osts when use "lfs quota".
1808 2. display the inaccurate data in the output of "lfs quota".
1809 3. try to get quota info when "lfs quota" is executed.
1814 Description: Extent locks not granted with no conflicts sometimes.
1815 Details : When race occurs in glimpse handler and nothing is returned,
1816 we do not reprocess the queue after lock cancel, and that leads
1817 to a stall until next activity on a resource
1820 Frequency : failover with quotaon
1822 Description: during mds failovers with quota on, OSTs got into deadlock state
1823 and causing dumpstack.
1824 Details : for every quota slave, at any time, there is only one quota req
1825 is sent to quota master for every uid/gid. Before that quota req
1826 returns, all the thread relative to the same uid/gid will wait.
1827 So if the quota req is lost because mds failovers or any other
1828 reasons, this bug will be hit. Now, dqacq_interpret() will handle
1829 quota reqs who time out.
1831 Severity : enhancement
1834 Description: when quota slave checks if quota is enough, there is an unnecessary
1836 Details : place this wait on necessary place instead of always waiting.
1838 --------------------------------------------------------------------------------
1840 2007-12-07 Cluster File Systems, Inc. <info@clusterfs.com>
1842 * Support for kernels:
1843 2.6.5-7.286 (SLES 9),
1844 2.6.9-55.0.9.EL (RHEL 4),
1845 2.6.16.53-0.8 (SLES 10),
1846 2.6.18-8.1.14.el5 (RHEL 5),
1847 2.6.18.8 vanilla (kernel.org)
1848 * Client support for unpatched kernels:
1849 (see http://wiki.lustre.org/index.php?title=Patchless_Client)
1850 2.6.16 - 2.6.22 vanilla (kernel.org)
1851 * Due to recently discovered recovery problems, we do not recommend
1852 using patchless RHEL 4 clients with this or any earlier release.
1853 * Recommended e2fsprogs version: 1.40.2-cfs1
1854 * Note that reiserfs quotas are disabled on SLES 10 in this kernel.
1857 Frequency : occasional
1859 Description: MDS hang or stay in waiting lock
1860 Details : If client receive lock with CBPENDING flag ldlm need send lock
1861 cancel as separate rpc, to avoid situation when cancel request
1862 can't processed due all i/o threads stay in wait lock.
1865 Frequency : occasional
1867 Description: Do not fail import if osc_interpret_create gets -EAGAIN
1868 Details : If osc_interpret_create got -EAGAIN it immediately exits and
1869 wakeup oscc_waitq. After wakeup oscc_wait_for_objects call
1870 oscc_has_objects and see OSC has no objests and call
1871 oscc_internal_create to resend create request.
1873 Severity : enhancement
1875 Description: Update kernel patches for SLES10 2.6.16.53-0.8.
1876 Details : Update which_patch & target file for SLES10 latest kernel.
1878 Severity : enhancement
1880 Description: add --type and --size parameters to lfs find
1881 Details : Enhance lfs find by adding filetype and filesize parameters. Also
1882 multiple OBDs can now be specified for the --obd option.
1884 Severity : enhancement
1886 Description: eliminate client locks in face of contention
1887 Details : file contention detection and lockless i/o implementation
1888 for contended files.
1890 Severity : enhancement
1892 Description: Remove client patches from SLES 10 kernel.
1893 Details : This causes SLES 10 clients to behave as patchless clients
1894 even on a Lustre-patched (server) kernel.
1896 Severity : enhancement
1898 Description: use i_size_read and i_size_write in 2.6 port
1899 Details : replace inode->i_size access with i_size_read/write()
1902 Frequency : when removing large files
1904 Description: scheduling issue during removal of large Lustre files
1905 Details : Don't take the BKL in fsfilt_ext3_setattr() for 2.6 kernels.
1906 It causes scheduling issues when removing large files (17TB in the
1912 Description: 1.4.11 Can't handle directories with stripe set and extended ACLs
1913 Details : Impossible (EPROTO is returned) to access a directory that has a
1914 non-default striping and ACLs.
1917 Frequency : only on ppc
1919 Description: /proc/fs/lustre/devices broken on ppc
1920 Details : The patch as applied to 1.6.2 doesn't look correct for all arches.
1921 We should make sure the type of 'index' is loff_t and then cast
1922 explicitly as needed below. Do not assign an explicitly cast
1926 Frequency : only for rhel5
1928 Description: Kernel patches update for RHEL5 2.6.18-8.1.10.el5.
1929 Details : Modify the target file & which_kernel.
1932 Frequency : if the uninit_groups feature is enabled on ldiskfs
1934 Description: e2fsck reports "invalid unused inodes count"
1935 Details : If a new ldiskfs filesystem is created with the "uninit_groups"
1936 feature and only a single inode is created in a group then the
1937 "bg_unused_inodes" count is incorrectly updated. Creating a
1938 second inode in that group would update it correctly.
1943 Description: buffer overruns could theoretically occur
1944 Details : llapi_semantic_traverse() modifies the "path" argument by
1945 appending values to the end of the origin string, and a buffer
1946 overrun may occur. Adding buffer overrun check in liblustreapi.
1948 Severity : enhancement
1950 Description: Add jbd statistics patch for RHEL5 and 2.6.18-vanilla.
1953 Frequency : only if filesystem is inconsistent
1955 Description: handle "serious error: objid * already exists" more gracefully
1956 Details : If LAST_ID value on disk is smaller than the objects existing in
1957 the O/0/d* directories, it indicates disk corruption and causes an
1958 LBUG(). If the object is 0-length, then we should use the existing
1959 object. This will help to avoid a full fsck in most cases.
1961 Severity : enhancement
1963 Description: Kernel patches update for RHEL4 2.6.9-55.0.6.
1964 Details : Modify vm-tunables-rhel4.patch.
1966 Severity : enhancement
1968 Description: Kernel config for 2.6.18-vanilla.
1969 Details : Modify targets/2.6-vanilla.target.in.
1970 Add config file kernel-2.6.18-2.6-vanilla-i686.config.
1971 Add config file kernel-2.6.18-2.6-vanilla-i686-smp.config.
1972 Add config file kernel-2.6.18-2.6-vanilla-x86_64.config.
1973 Add config file kernel-2.6.18-2.6-vanilla-x86_64-smp.config.
1976 Frequency : occasional
1978 Description: improve handling recoverable errors
1979 Details : If request processed with error which can be recoverable on server
1980 request should be resend, otherwise page released from cache and
1986 Description: Kernel patches update for RHEL5 2.6.18-8.1.14.el5.
1987 Details : Modify target file & which_patch.
1988 A flaw was found in the IA32 system call emulation provided
1989 on AMD64 and Intel 64 platforms. An improperly validated 64-bit
1990 value could be stored in the %RAX register, which could trigger an
1991 out-of-bounds system call table access. An untrusted local user
1992 could exploit this flaw to run code in the kernel
1993 (ie a root privilege escalation). (CVE-2007-4573).
1997 Description: change order of libsysio includes
1998 Details : '#include sysio.h' should always come before '#include xtio.h'
2000 Severity : enhancement
2002 Description: adapt the lustre_config script to support the upgrade case
2003 Details : Add "-u" option for lustre_config script to support upgrading 1.4
2004 server targets to 1.6 in parallel.
2009 Description: To avoid grant space > avaible space when the disk is almost
2010 full. Without this patch you might see the error "grant XXXX >
2011 available" or some LBUG about grant, when the disk is almost
2013 Details : In filter_check_grant, for non_grant cache write, we should
2014 check the left space by if (*left > ungranted + bytes), instead
2015 of (*left > ungranted), because only we are sure the left space
2016 is enough for another "bytes", then the ungrant space should be
2017 increase. In client, we should update cl_avail_grant only there is
2018 OBD_MD_FLGRANT in the reply.
2023 Description: Update RHEL 4 kernel to fix local root privilege escalation.
2024 Details : Update to the latest RHEL 4 kernel to fix the vulnerability
2025 described in CVE-2007-4573. This problem could allow untrusted
2026 local users to gain root access.
2029 Frequency : when using O_DIRECT and quotas
2031 Description: Incorrect file ownership on O_DIRECT output files
2032 Details : block usage reported by 'lfs quota' does not take into account
2033 files that have been written with O_DIRECT.
2037 Description: (rw.c:1323:ll_read_ahead_pages()) ASSERTION(page_idx > ria->ria_stoff) failed
2038 Details : Once the unmatched stride IO mode is detected, shrink the stride-ahead
2039 window to 0. If it does hit cache miss, and read-pattern is still
2040 stride-io mode, does not reset the stride window, but also does not
2041 increase the stride window length in this case.
2043 --------------------------------------------------------------------------------
2045 2007-09-27 Cluster File Systems, Inc. <info@clusterfs.com>
2047 * Support for kernels:
2048 2.6.5-7.286 (SLES 9),
2049 2.6.9-55.0.2.EL (RHEL 4),
2050 2.6.16.46-0.14 (SLES 10),
2051 2.6.18-8.1.8.el5 (RHEL 5),
2052 2.6.18.8 vanilla (kernel.org)
2053 * Client support for unpatched kernels:
2054 (see http://wiki.lustre.org/index.php?title=Patchless_Client)
2055 2.6.16 - 2.6.21 vanilla (kernel.org)
2056 * Due to recently discovered recovery problems, we do not recommend
2057 using patchless RHEL 4 clients with this or any earlier release.
2058 * Recommended e2fsprogs version: 1.40.2-cfs1
2059 * Note that reiserfs quotas are disabled on SLES 10 in this kernel.
2063 Description: Fix errors in lfs documentation
2064 Details : Fixes man pages
2066 Severity : enhancement
2068 Description: Adaptive timeouts
2069 Details : RPC timeouts adapt to changing server load and network
2070 conditions to reduce resend attempts and improve recovery time.
2072 Severity : enhancement
2074 Description: llapi_file_create() does not allow some changes
2075 Details : add llapi_file_open() that allows specifying the file creation
2076 mode and open flags, and also returns an open file handle.
2078 Severity : enhancement
2080 Description: df doesn't work properly if diskfs blocksize != 4K
2081 Details : Choose biggest blocksize of OST's as the LOV's blocksize.
2084 Frequency : always on directories with default striping set
2086 Description: lfs find on -1 stripe looping in lsm_lmm_verify_common()
2087 Details : Avoid lov_verify_lmm_common() on directory with -1 stripe count.
2089 Severity : enhancement
2091 Description: merge and cleanup kernel patches.
2092 Details : Remove mnt_lustre_list in vfs_intent-2.6-rhel4.patch.
2094 Severity : enhancement
2096 Description: RedHat Update kernel for RHEL5
2097 Details : Kernel config file for RHEL5.
2099 Severity : enhancement
2101 Description: OSS needs mutliple precreate threads
2102 Details : Add ability to start more than one create thread per OSS.
2105 Frequency : only with quota on the root user
2107 Description: mds_obd_create error creating tmp object
2108 Details : When the user sets quota on root, llog will be affected and can't
2109 create files and write files.
2112 Frequency : Always on ia64 patchless client, and possibly others.
2114 Description: Add EXPORT_SYMBOL check for node_to_cpumask symbol.
2115 Details : This allows the patchless client to be loaded on architectures
2116 without this export.
2121 Description: disorder of journal start and llog_add cause deadlock.
2122 Details : in llog_origin_connect, journal start should happen before
2123 llog_add keep the same order as other functions to avoid
2126 Severity : enhancement
2128 Description: RedHat Update kernel for RHEL5
2129 Details : Modify the kernel config file more closer RHEL5.
2131 Severity : enhancement
2133 Description: Build failure against Centos5 (RHEL5)
2134 Details : Define PAGE_SIZE when it isn't present.
2137 Frequency : occasionally when using NFS
2139 Description: "ll_intent_file_open()) lock enqueue: err: -13" with nfs
2140 Details : with NFS, the anon dentry's parent was set to itself in
2141 d_alloc_anon(), so in MDS, we use rec->ur_fid1 to find the
2142 corresponding dentry other than use rec->ur_name.
2144 Severity : enhancement
2146 Description: enable data checksumming by default
2147 Details : enable checksum by default, allow --disable-checksum
2148 configure option and "-o nochecksum" mount option. Checksums
2149 can also be disabled at runtime via $LPROC/osc/*/checksum_pages.
2152 Frequency : Occasionally with failover
2154 Description: Client eviction due to failover config
2155 Details : after a connection loss, the lustre client should attempt to
2156 reconnect to the last active server first before trying the
2157 other potential connections.
2159 Severity : enhancement
2161 Description: client-side metadata stat-ahead during readdir(directory readahead)
2162 Details : perform client-side metadata stat-ahead when the client detects
2163 readdir and sequential stat of dir entries therein
2168 Description: when mds and osts use different quota unit(32bit and 64bit),
2169 quota will be released repeatly.
2170 Details : void sending multiple quota reqs to mds, which will keep the
2171 status between the reqs.
2174 Frequency : only with liblustre clients on XT3
2176 Description: evictions taking too long
2177 Details : allow llrd to evict clients directly on OSTs
2181 Description: osts not allocated evenly to files
2182 Details : change the condition to increase offset_idx
2184 Severity : enhancement
2186 Description: self-adjustable client's lru lists
2187 Details : use adaptive algorithm for managing client cached locks lru
2188 lists according to current server load, other client's work
2189 pattern, memory activities, etc. Both, server and client
2190 side namespaces provide number of proc tunables for controlling
2195 Description: rewrite ext2-derived code in llite/dir.c and obdclass/uuid.c
2196 Details : rewrite inherited code (uuid parsing code from ext2 utils and
2197 readdir code from ext3) from scratch preserving functionality.
2201 Description: Only those disconnect error should be returned by rq_status.
2202 Details : In open/enqueue processs, Some errors, which will cause client
2203 disconnected, should be returned by rq_status, while other
2204 errors should still be returned by intent, then mdc or llite will
2207 Severity : enhancement
2209 Description: Tune the kernel for good SCSI performance.
2210 Details : Set the value of /sys/block/{dev}/queue/max_sectors_kb
2211 to the value of /sys/block/{dev}/queue/max_hw_sectors_kb
2215 Frequency : Always for filesystems larger than 2TB on 32-bit systems.
2216 Bugzilla : 13547, 13627
2217 Description: Data corruption for OSTs that are formatted larger than 2TB
2219 Details : When generating the bio request for lustre file writes the
2220 sector number would overflow a temporary variable before being
2221 used for the IO. The data reads correctly from Lustre (which
2222 will overflow in a similar manner) but other file data or
2223 filesystem metadata may be corrupted in some cases.
2227 Description: "lfs find -obd UUID" prints directories
2228 Details : "lfs find -obd UUID" will return all directory names instead
2229 of just file names. It is incorrect because the directories
2230 do not reside on the OSTs.
2233 Bugzilla : 13596, 13614
2234 Description: MDS hang after unclean shutdown of lots of clients
2235 Details : Never resend AST requests.
2239 Frequency : Always, for kernels after 2.6.16
2240 Description: Fix warning idr_remove called for id=.. which is not allocated.
2241 Details : Last kernels save old s_dev before kill super and not allow
2242 to restore from callback - restore it before call kill_anon_super.
2246 Description: Fix error on 'ls .' at the top of the Lustre mount.
2247 Details : Don't revalidate dentry if it is a root dentry.
2249 --------------------------------------------------------------------------------
2251 2007-08-27 Cluster File Systems, Inc. <info@clusterfs.com>
2253 * Support for kernels:
2254 2.6.5-7.286 (SLES 9),
2255 2.6.9-55.0.2.EL (RHEL 4),
2256 2.6.16.46-0.14 (SLES 10),
2257 2.6.18-8.1.8.el5 (RHEL 5),
2258 2.6.18.8 vanilla (kernel.org)
2259 * Client support for unpatched kernels:
2260 (see http://wiki.lustre.org/index.php?title=Patchless_Client)
2261 2.6.16 - 2.6.22 vanilla (kernel.org)
2262 * Due to recently discovered recovery problems, we do not recommend
2263 using patchless RHEL 4 clients with this or any earlier release.
2264 * Recommended e2fsprogs version: 1.39.cfs8
2265 * Note that reiserfs quotas are disabled on SLES 10 in this kernel.
2270 Description: block reactivating mgc import until all deactivates complete
2271 Details : Fix race when failing back MDT/MGS to itself (testing)
2273 Severity : enhancement
2275 Description: lfs setstripe enhancement
2276 Details : Make lfs setstripe understand 'k', 'm' and 'g' for stripe size.
2279 Frequency : mds/oss recovery
2281 Description: llog ctxt is refrenced after it has been freed.
2282 Details : llog ctxt refcount was added to avoide the race between ctxt free
2283 and llog recovery process. Each llog user must hold ctxt refcount
2284 before it access the llog. And the llog ctxt can only be freed
2285 when its refcount is zero.
2287 Severity : enhancement
2289 Description: randomly memory allocation failure util
2290 Details : Make lustre randomly failed allocating memory for testing purpose.
2293 Frequency : only for SLES10
2295 Description: Update kernel patch for SLES10 SP1
2296 Details : Add patch blkdev_tunables-2.6-sles10.patch to 2.6-sles10.series.
2298 Severity : enhancement
2300 Description: omit set fsid for export NFS
2301 Details : fix set/restore device id for avoid EMFILE error and mark lustre fs
2302 as FS_REQUIRES_DEV for avoid problems with generate fsid.
2305 Frequency : after network failures
2307 Description: Add sync option to mount_lustre.c
2308 Details : Client loses data written to lustre after a network interruption.
2310 Severity : enhancement
2312 Description: Error message improvement.
2313 Details : Merging of two LCONSOLE_ERROR_MSG into one.
2315 Severity : enhancement
2317 Description: don't use GFP_* in generic Lustre code.
2318 Details : Use cfs_alloc_* functions and CFS_* flags for code portability.
2320 Severity : enhancement
2322 Description: obdclass is limited by single OBD_ALLOC(idarray)
2323 Details : replace OBD_ALLOC/OBD_FREE with OBD_VMALLOC/OBD_VFREE
2325 Severity : enhancement
2327 Description: updated patchess for new RHEL4 kernel
2328 Details : Fixed ext3-unlink-race.patch per Kalpak's comment.
2330 Severity : enhancement
2332 Description: warnings with build patchless client with vanila 2.6.19 and up
2333 Details : change old ctl_table style and replace ctl_table/ctl_table_header
2334 with cfs_sysctl_table_t/cfs_sysctl_table_header_t
2336 Severity : enhancement
2338 Description: O_DIRECT bypasses client statistics.
2339 Details : When running with O_DIRECT I/O, neither the client rpc_stats nor
2340 read_ahead_stats were updated. Copied stats section from
2341 osc_send_oap_rpc() into async_internal().
2344 Frequency : only for Cray XT3
2346 Description: peer credits not enough on many OST per OSS systems.
2347 Details : Use new lnet way to add credits as we need those for pings and ASTs
2350 Frequency : only with liblustre
2352 Description: Liblustre is not releasing flock locks on file close.
2353 Details : Release flock locks on file close.
2356 Frequency : only for RHEL4
2358 Description: Update kernel patches for kernel-2.6.9-55.0.2.EL
2359 Details : Remove inode-nr_unused-2.6.9-rhel4.patch from 2.6-rhel4.series
2360 Update target file and kernel config.
2364 Description: lustre support for RHEL5
2365 Details : Add support for RHEL5.
2370 Description: ASSERTION(export != NULL) failed in target_handle_connect
2371 Details : Assetion hit is result of rare race between disconnect and connet
2372 to same nid. target_handle_connect found old connect cockie and
2373 tried to reconnect, but can't find export for this cockie.
2378 Description: umount blocks forever on error
2379 Details : In result of wrong using obd_no_recov and obd_force flags client
2380 can hand if cancel or some other requests is lost.
2383 Frequency : Only for SLES
2385 Description: sanity_quota fail test_1
2386 Details : There are multiple occurences of $TSTUSR in SLES's /etc/group
2387 file, which makes TSTID[2] inunique.
2389 Severity : enhancement
2391 Description: Kernel patches for SLES9 2.6.5-7.286 kernel
2392 Details : Update target/ChangeLog/which_patch .
2394 Severity : enhancement
2396 Description: jbd statistics
2397 Details : Port older jbd statistics patch for sles10
2402 Description: lvbo_init failed for resource with missing objects.
2403 Details : Fix returning error if we do stat for file with missing/corrupted
2404 objects and i_size set to all sum of size of all avaible objects.
2405 if we to truncate/write to missing object - it is recreated.
2410 Description: Oops in read and write path when failing to allocate lock.
2411 Details : Check if lock allocation failed and return error back.
2414 Frequency : When flocks are used.
2416 Description: assertion failure in ldlm_cli_enquque_fini for non NULL lock.
2417 Details : Flock locks might destroy just granted lock if it could be merged
2418 with another existing flock, this is done in completion handler,
2419 so teach ldlm_cli_enquque_fini that this is a valid case for
2425 Description: reply_lock_interpret crash due to race with it and lock cancel.
2426 Details : Do not replay locks that are being cancelled. Do not reference
2427 locks by their address during replay, just by their handle.
2430 Frequency : only with deactivated OSTs
2432 Description: lstripe command fails for valid OST index
2433 Details : The stripe offset is compared to 'lov->desc.ld_tgt_count'
2434 instead of lov->desc.ld_active_tgt_count.
2436 Severity : enhancement
2438 Description: Build failure against Centos5 (RHEL5)
2439 Details : Use getpagesize() instead of PAGE_SIZE.
2441 --------------------------------------------------------------------------------
2443 2007-07-30 Cluster File Systems, Inc. <info@clusterfs.com>
2445 * Support for kernels:
2446 2.6.5-7.283 (SLES 9),
2447 2.6.9-55.EL (RHEL 4),
2448 2.6.16.46-0.14 (SLES 10),
2449 2.6.18.8 vanilla (kernel.org)
2450 * Client support for unpatched kernels:
2451 (see http://wiki.lustre.org/index.php?title=Patchless_Client)
2452 2.6.16 - 2.6.22 vanilla (kernel.org)
2453 * Due to recently discovered recovery problems, we do not recommend
2454 using patchless RHEL 4 clients with this or any earlier release.
2455 * Recommended e2fsprogs version: 1.39.cfs8
2456 * Note that reiserfs quotas are disabled on SLES 10 in this kernel.
2457 * Starting with this release, the ldiskfs backing filesystem required
2458 by Lustre is now in its own package, lustre-ldiskfs. This package
2459 should be installed. It is versioned separately from Lustre and
2460 may be released separately in future.
2465 Description: server LBUG when shutting down
2466 Details : Block umount forever until the mount refcount is zero rather
2467 than giving up after an arbitrary timeout.
2469 Severity : enhancement
2471 Description: add optional extra BUILD_VERSION info
2472 Details : add a new environment variable (namely LUSTRE_VERS) which allows
2473 to override the lustre version.
2476 Frequency : 2.6.18 servers only
2478 Description: ll_kern_mount() doesn't release the module reference
2479 Details : The ldiskfs module reference count never drops down to 0
2480 because ll_kern_mount() doesn't release the module reference.
2485 Description: server LBUG when using old ost_num_threads parameter
2486 Details : Accept the old ost_num_threads parameter but warn that it
2487 is deprecated, and fix an off-by-one error that caused an LBUG.
2492 Description: Transient SCSI error results in persistent IO issue
2493 Details : iobuf->dr_error is not reinitialized to 0 between two
2497 Frequency : sometimes when underlying device returns I/O errors
2499 Description: OSTs not going read-only during write failures
2500 Details : OSTs are not remounted read-only when the journal commit threads
2501 get I/O errors because fsfilt_ext3 calls journal_start/stop()
2502 instead of the ext3 wrappers.
2505 Frequency : only with 10000 clients or more
2507 Description: poor connect scaling with increasing client count
2508 Details : Don't run filter_grant_sanity_check for more than 100 exports
2509 to improve scaling for large numbers of clients.
2512 Frequency : SLES10 only
2514 Description: sanity-quota.sh quotacheck failed: rc = -22
2515 Details : Quotas cannot be enabled on SLES10.
2518 Frequency : liblustre clients only
2520 Description: getdirentries does not give error when run on compute nodes
2521 Details : getdirentries does not fail when the size specified as an argument
2522 is too small to contain at least one entry
2524 Severity : enhancement
2526 Description: Add LNET router traceability for debug purposes
2527 Details : If a checksum failure occurs with a router as part of the
2528 IO path, the NID of the last router that forwarded the bulk data
2529 is printed so it can be identified.
2534 Description: OST "spontaneously" evicts client; client has imp_pingable == 0
2535 Details : Due to a race condition, liblustre clients were occasionally
2536 evicted incorrectly.
2538 Severity : enhancement
2540 Description: lfs setstripe use optional parameters instead of postional
2543 Severity : enhancement
2545 Description: Nanosecond timestamp support for ldiskfs
2546 Details : The on-disk ldiskfs filesystem has added support for nanosecond
2547 resolution timestamps. There is not yet support for this at
2548 the Lustre filesystem level.
2551 Frequency : during server recovery
2553 Description: MDS failing to send precreate requests due to OSCC_FLAG_RECOVERING
2554 Details : request with rq_no_resend flag not awake l_wait_event if they get a
2558 Frequency : nfs export on patchless client
2560 Description: connectathon hang when test nfs export over patchless client
2561 Details : Disconnected dentry cannot be found with lookup, so we do not need
2562 to unhash it or make it invalid
2565 Description: fix llapi_lov_get_uuids() to allow many OSTs to be returned
2566 Details: : Change llapi_lov_get_uuids() to read the UUIDs from /proc instead
2567 of using an ioctl. This allows lfsck for > 160 OSTs to succeed.
2572 Description: open req refcounting wrong on reconnect
2573 Details : If reconnect happened between getting open reply from server and
2574 call to mdc_set_replay_data in ll_file_open, we will schedule
2575 replay for unreferenced request that we are about to free.
2576 Subsequent close will crash in variety of ways.
2577 Check that request is still eligible for replay in
2578 mdc_set_replay_data().
2583 Description: disable writes to filesystem when reading health_check file
2584 Details : the default for reading the health_check proc file has changed
2585 to NOT do a journal transaction and write to disk, because this
2586 can cause reads of the /proc file to hang and block HA state
2587 checking on a healthy but otherwise heavily loaded system. It
2588 is possible to return to the previous behaviour during configure
2589 with --enable-health-write.
2591 Severity : enhancement
2593 Description: 64-bit inode version
2594 Details: : Add a on-disk 64-bit inode version for ext3 to track changes made
2595 to the inode. This will be required for version-based recovery.
2600 Description: MDS fails to start if a duplicate client export is detected
2601 Details : in some rare cases it was possible for a client to connect to
2602 an MDS multiple times. Upon recovery the MDS would detect this
2603 and fail during startup. Handle this more gracefully.
2605 Severity : enhancement
2607 Description: Add -o localflock option to simulate old noflock behaviour.
2608 Details : This will achieve local-only flock/fcntl locks coherentness.
2613 Description: log_commit_thread vs filter_destroy race leads to crash
2614 Details : Take import reference before releasing llog record semaphore
2619 Description: Wrong request locking in request set processing
2620 Details : ptlrpc_check_set wrongly uses req->rq_lock for proctect add to
2621 imp_delayed_list, in this place should be used imp_lock.
2624 Frequency : when reconnecting
2626 Description: Grant leak when OSC reconnect to OST
2627 Details : When osc reconnect ost, OST(filter) should check whether it
2628 should grant more space to client by comparing fed_grant and
2629 cl_avail_grant, and return the granted space to client instead
2630 of "new granted" space, because client will call osc_init_grant
2631 to update the client grant space info.
2634 Frequency : when client reconnects to OST
2636 Description: Grant leak when OSC does a resend and replays bulk write
2637 Details : When osc reconnect to OST, OST(filter) should clear grant info of
2638 bulk write request, because the grant info while be sync between
2639 OSC and OST when reconnect, and we should ignore the grant info
2640 these of resend/replay write req.
2645 Description: Grant space more than avaiable space sometimes.
2646 Details : When then OST is about to be full, if two bulk writing from
2647 different clients came to OST. Accord the avaliable space of the
2648 OST, the first req should be permitted, and the second one
2649 should be denied by ENOSPC. But if the seconde arrived before
2650 the first one is commited. The OST might wrongly permit second
2651 writing, which will cause grant space > avaiable space.
2654 Frequency : when client is evicted
2656 Description: Grant might be wrongly erased when osc is evicted by OST
2657 Details : when the import is evicted by server, it will fork another
2658 thread ptlrpc_invalidate_import_thread to invalidate the
2659 import, where the grant will be set to 0. While the original
2660 thread will update the grant it got when connecting. So if
2661 the former happened latter, the grant will be wrongly errased
2662 because of this race.
2667 Description: Checking Stale with correct fid
2668 Details : ll_revalidate_it should uses de_inode instead of op_data.fid2
2669 to check whether it is stale, because sometimes, we want the
2670 enqueue happened anyway, and op_data.fid2 will not be initialized.
2672 Severity : enhancement
2674 Description: update patchless client
2675 Details : Add support for patchless client with 2.6.20, 2.6.21 and RHEL 5
2678 Frequency : only with 2.4 kernel
2680 Description: random memory corruption
2681 Details : size of struct ll_inode_info is to big for union inode.u and this
2682 can be cause of random memory corruption.
2687 Description: Memory leak in recovery
2688 Details : Lov_mds_md was not free in an error handler in mds_create_object.
2689 It should also check obd_fail before fsfilt_start, otherwise if
2690 fsfilt_start return -EROFS,(failover mds during mds recovery).
2691 then the req will return with repmsg->transno = 0 and rc = EROFS.
2692 and we met hit the assert LASSERT(req->rq_reqmsg->transno ==
2693 req->rq_repmsg->transno) in ptlrpc_replay_interpret. Fcc should
2694 be freed no matter whether fsfilt_commit success or not.
2697 Frequency : only with large numbers of cores on a single node
2699 Description: Prevents from taking the superblock lock in llap_from_page for
2701 Details : using LL_ORIGIN_REMOVEPAGE origin flag instead of LL_ORIGIN_UNKNOW
2702 for llap_from_page call in ll_removepage() prevents from taking
2703 the superblock lock for a soon died page.
2708 Description: Not check open intent error before release open handle
2709 Details : in some rare cases, the open intent error is not checked before
2710 release open handle, which may cause
2711 ASSERTION(open_req->rq_transno != 0), because it tries to release
2712 the failed open handle.
2717 Description: Set cat log bitmap only after create log success.
2718 Details : in some rare cases, the cat log bitmap is set too early. and it
2719 should be set only after create log success.
2723 Description: Accessing a block bevice can re-enable I/O when Lustre is
2724 tearing down a device.
2725 Details : dev_clear_rdonly(bdev) must be called in kill_bdev() instead of
2730 Description: service threads may hog cpus when there are a lot of requests
2731 Details : Insert cond_resched to give other threads a chance to use some CPU
2736 Description: the cat log was not initialized in recovery
2737 Details : When mds(mgs) do recovery, the tgt_count might be zero, so the
2738 unlink log on mds will not be initialized until mds post
2739 recovery. And also in mds post recovery, the unlink log will
2740 initialization will be done asynchronausly, so there will be race
2741 between add unlink log and unlink log initialization.
2745 Description: brw_stats were being printed incorrectly
2746 Details : brw_stats were being printed as log2 but all of them were not
2747 recorded as log2. Also remove some code duplication arising from
2748 filter_tally_{read,write}.
2752 Frequency : rare, only in recovery.
2753 Description: ASSERTION(req->rq_type != LI_POISON) failed
2754 Details : imp_lock should be held while iterating over imp_sending_list for
2755 prevent destroy request after get timeout in ptlrpc_queue_wait.
2759 Description: replay-single.sh test 52 fails
2760 Details : A lock's skiplist need to be cleanup when it being unlinked
2761 from its resource list.
2766 Description: Short directio read returns full requested size rather than
2768 Details : Direct I/O operations should return actual amount of bytes
2769 transferred rather than requested size.
2771 Severity : enhancement
2773 Description: metadata RPC reduction (e.g. for rm performance)
2774 Details : decrease the amount of synchronous RPC between clients and servers
2775 by canceling conflicing lock before the operation on the client
2776 and packing thier handles into the main operation RPC to server.
2778 Severity : enhancement
2780 Description: add #ifdef HAVE_KERNEL_CONFIG_H
2781 Details : kernels from 2.6.19 not need include linux/config.h, but add
2782 include linux/autoconf.h in commpiler command line.
2784 Severity : enhancement
2786 Description: patchless client support for 2.6.22 kernel
2787 Details : 2.6.22 has only one visble change, SLAB_CTOR_* constants is
2788 removed. In this case we need drop using os depended interface to
2789 kmem_cache and use cfs_mem_cache API.
2794 Description: fix mal-formatted messages
2795 Details : fix some mal-formatted DEBUG_REQ and LCONSOLE_ERROR_MSG messages
2799 Frequency : always in liblustre
2800 Description: wrong IS_ERR implementation in liblustre.h
2801 Details : fix IS_ERR implementation in liblustre.h for right detect errors.
2806 Description: Correct condition for output debug message.
2807 Details : inode i_nlink equal zero is not enough for output message about
2808 disk corruption, i_ctime and i_mode should be also checked.
2812 Frequency : always in patchless client
2813 Description: add configure check for truncate_complete_page
2814 Details : improve checks for exported symbols. This allow run check without
2815 sources, but with Module.symvers shipped with kernel distribution.
2816 add check for truncate_complete_page used by patchless client.
2820 Description: sanity.sh test_77h fails with "test_77h file compare failed"
2821 Details : test_77h uses a file which was messed by other test case.
2825 Description: Not Check whether lov_tgts is NULL in some lov functions
2826 Details : Checking whether lov_tgts is NULL in some functions.
2830 Description: replace obdo_alloc() with OBDO_ALLOC macro
2831 Details : nothing special is done in obdo_alloc() function, and for
2832 debugging purpose, it needs to be replaced with macros.
2836 Description: bad return value and errno from fcntl call
2837 Details : In liblustre API, errno should be a negative value if error
2842 Description: ptlrpc_check_set() LBUG
2843 Details : In case of positive reply from server and failed client bulk
2844 callback after bulk transfer shouldn't LBUG, but process this
2845 request as erroneous.
2847 Severity : enhancement
2849 Description: VFS operations stats tool.
2850 Details : Tool which collects stats by tracking value written in pid,
2851 ppid, gid and uses llstat to generate output to plot graph using
2853 Updated lustre/utils/Makefile.am
2854 Added lustre/utils/ltrack_stats.c
2856 Severity : enhancement
2858 Description: 2.6.18 server support (lustre 1.6.1)
2859 Details : Support for 2.6.18 kernels on the server side.
2864 Description: ASSERTION(imp->imp_conn_current) failed
2865 Details : an assertion failure is hit if a client node boots and attempts to
2866 mount a lustre filesystem faster than RECONNECT_INTERVAL seconds.
2869 Frequency : only for i686
2871 Description: 1.4.11 RC1 build fails for RHEL 4, i686
2872 Details : Fixed config variable for build.
2877 Description: Updated patchess for new RHEL4 kernel
2878 Details : Updated patch inode-nr_unused-2.6.9-rhel4.patch
2879 Updated patch jbd-stats-2.6.9.patch
2880 Updated patch qsnet-rhel4-2.6.patch
2881 Updated patch quota-deadlock-on-pagelock-core.patch
2882 Updated patch vfs_intent-2.6-rhel4.patch
2883 Updated patch vfs_races-2.6-rhel4.patch
2884 Updated series file 2.6-rhel4-titech.series
2885 Updated series file 2.6-rhel4.series
2886 Updated kernel config files
2891 Description: lquota slave complains LBUG when reconnecting with mds
2893 Details : quota slave depends qctxt->lqc_import to send its quota request.
2894 This pointer will be invalid if mds did failover or broke its
2895 connect to osts, which leads to LBUG.
2898 Frequency : when qunit size is too small(less than 20M)
2900 Description: write is stopped by improper -EDQUOT
2901 Details : If the master is busy and qunit size is small enough(let's say 1M),
2902 the slave can not get quota from master on time, which will lead
2903 slave to trigger a -EQUOTA to client.
2908 Description: Deadlock during metadata tests
2909 Details : in prune_dir_dentries(), shrink_dcache_parent() should not be
2910 called with the per-dentry lock held.
2913 Frequency : SLES9 only
2915 Description: Lustre patched kernel for SLES9 SP3 has NR_CPUS set to 8
2916 Details : set CONFIG_NR_CPUS to 128 instead of 8.
2918 Severity : enhancement
2920 Description: remove fs_prep_san_write operation and related patches
2921 Details : remove the ext3-san-jdike patches which are no longer useful.
2926 Description: LDISKFS-fs error (device sdc): ldiskfs_free_blocks
2927 Details : a disk corruption can cause the mballoc code to assert on a
2928 double free or other extent corruptions. Handle these with
2929 ext3_error() instead of assertions.
2932 Frequency : only with mballoc3 code and deep extent trees
2934 Description: ldiskfs_ext_search_right: bad header in inode: unexpected eh_depth
2935 Details : a wrong check of extent headers in ldiskfs_ext_search_right()
2936 can cause the filesystem to be remounted read-only.
2941 Description: lfsck built against 1.4.x cannot run against 1.6.0 lustre
2942 Details : the definition for OBD_IOC_GETNAME changed in 1.6.0. One of the
2943 few external users of this ioctl number is lfsck's call to
2944 llapi_lov_get_uuids() and this caused lfsck to fail at startup.
2945 Add the old ioctl number to the handler so both old and new
2950 Description: parallel lock callbacks
2951 Details : Instead of sending blocking and completion callbacks as separated
2952 requests, adding them to a set and sending in parallel.
2956 Description: Disable most debugging by default
2957 Details : To improve performance, disable most logging (for debug purposes)
2958 by default. VFSTRACE, RPCTRACE, and DLMTRACE are now off by
2959 default, and HA includes fewer messages.
2963 Frequency : only run on patchless client.
2964 Description: use do_facet on sanity.sh for test handling recoverables errors
2965 Details : use do_facet instead of direct use sysctl for set fail_loc on OST
2969 Description: hash tables for lists of nids, connections and uuids
2970 Details : Hash tables noticeably help when a lot of clients connect to a
2971 server, to faster identify duplicate connections or reconnects,
2972 also to faster find export to evict in manual eviction case.
2976 Description: Sometimes, when the server evict a client, and the client will
2977 not be evicted as soon as possible.
2978 Details : In enqueue req, the error was returned by intent, instead of
2979 rq_status which make ptlrpc layer not detect this error, and
2980 does not evict the client. So enqueue error should be returned
2984 Frequency : only at startup
2986 Description: Delay client connections to MDT until fist MDT->OST connect
2987 Details : If a client tried to create a new file before the MDT had
2988 connected to any OSTs, the create would return EIO. Now
2989 the client will simply block until the MDT connects to the
2990 first OST and the create can succeed.
2993 Frequency : only for SLES9
2995 Description: Routinely utilize latest Quadrics drivers in CFS releases
2996 Details : Update patch qsnet-suse-2.6.patch.
2999 Frequency : only for sles10
3001 Description: Update patches for SLES 10 SP1 kernel.
3002 Details : Update the patch vfs_intent-2.6-sles10.patch.
3007 Description: Routinely utilize latest Quadrics drivers in CFS releases
3008 Details : Update patch qsnet-rhel4-2.6.patch.
3011 Frequency : at statup only
3013 Description: mds_lov_synchronize race leads to various problems
3014 Details : simultaneous MDT->OST connections at startup can cause the
3015 sync to abort, leaving the OSC in a bad state.
3020 Description: Using wrong pointer in osc_brw_prep_request
3021 Details : Access to array[-1] can produce panic if kernel compiled with
3022 CONFIG_PAGE_ALLOC enabled
3024 Severity : enhancement
3026 Description: Async OSC create to avoid the blocking unnecessarily.
3027 Details : If a OST has no remain object, system will block on the creating
3028 when need to create a new object on this OST. Now, ways use
3029 pre-created objects when available, instead of blocking on an
3030 empty osc while others are not empty. If we must block, we block
3031 for the shortest possible period of time.
3035 Frequency : only in recovery
3036 Description: Mark OST as early accessible if his start SYNC.
3037 Details : osc_precreate return flag early accessible if oscc marked as
3038 OSCC_FLAG_SYNC_IN_PROGRESS.
3043 Description: Sometimes precreate code can triger create object on wrong ost
3044 Details : Wrong protected or not not restored variables aftre precreate loop
3045 can produce creation object on wrong ost.
3048 Frequency : oss recovery
3050 Description: llog_commit_thread cleanup should sync with llog_commit_thread
3052 Details : llog_commit_thread_count should be synced between llog_commit
3053 start and cleanup, so new llog_commit thread should not be started
3054 when llog_commit threads being stopped to avoid accessing some
3057 Severity : enhancement
3059 Description: Add printing inode info into message about error in writepage.
3061 --------------------------------------------------------------------------------
3063 2007-05-03 Cluster File Systems, Inc. <info@clusterfs.com>
3068 Frequency : on some architectures
3070 Description: 1.6 client sometimes fails to mount from a 1.4 MDT
3071 Details : Uninitialized flags sometimes cause configuration commands to
3075 Frequency : patchless clients only
3077 Description: missing __iget() symbol export
3078 Details : The __iget() symbol export is missing. To avoid the need for
3079 this on patchless clients the deathrow inode reaper is turned
3080 off, and we depend on the VM to clean up old inodes. This
3081 dependency was during via the fix for bug 12181.
3086 Description: sanity.sh fail: test_52b
3087 Details : The ll_inode_to_ext_flags() has a glitch which makes MDS return
3088 incorrect inode's flags to client.
3090 --------------------------------------------------------------------------------
3092 2007-04-19 Cluster File Systems, Inc. <info@clusterfs.com>
3094 * CONFIGURATION CHANGE. This version of Lustre WILL NOT
3095 INTEROPERATE with older versions automatically. In many cases a
3096 special upgrade step is needed. Please read the
3097 user documentation before upgrading any part of a 1.4.x system.
3098 * WARNING: Lustre configuration and startup changes are required with
3099 this release. See https://mail.clusterfs.com/wikis/lustre/MountConf
3101 * Support for kernels:
3102 2.4.21-47.0.1.EL (RHEL 3),
3103 2.6.5-7.283 (SLES 9),
3104 2.6.9-42.0.10.EL (RHEL 4),
3105 2.6.12.6 vanilla (kernel.org),
3106 2.6.16.27-0.9 (SLES10)
3107 * Client support for unpatched kernels:
3108 (see https://mail.clusterfs.com/wikis/lustre/PatchlessClient)
3109 2.6.16 - 2.6.19 vanilla (kernel.org),
3110 2.6.9-42.0.8EL (RHEL 4)
3111 * Recommended e2fsprogs version: 1.39.cfs6
3112 * Note that reiserfs quotas are disabled on SLES 10 in this kernel
3115 Severity : enhancement
3117 Description: Async OSC create to avoid the blocking unnecessarily.
3118 Details : If a OST has no remain object, system will block on the creating
3119 when need to create a new object on this OST. Now, ways use
3120 pre-created objects when available, instead of blocking on an
3121 empty osc while others are not empty. If we must block, we block
3122 for the shortest possible period of time.
3124 Severity : enhancement
3126 Description: MountConf
3127 Details : Lustre configuration is now managed via mkfs and mount
3128 commands instead of lmc and lconf. New obd types (MGS, MGC)
3129 are added for dynamic configuration management. See
3130 https://mail.clusterfs.com/wikis/lustre/MountConf for
3133 Severity : enhancement
3135 Description: dynamic OST addition
3136 Details : OSTs can now be added to a live filesystem
3138 Severity : enhancement
3140 Description: startup order invariance
3141 Details : MDTs and OSTs can be started in any order. Clients only
3142 require the MDT to complete startup.
3144 Severity : enhancement
3146 Description: parallel, asynchronous orphan cleanup
3147 Details : orphan cleanup is now performed in separate threads for each
3148 OST, allowing parallel non-blocking operation.
3150 Severity : enhancement
3152 Description: optimized stripe assignment
3153 Details : stripe assignments are now made based on ost space available,
3154 ost previous usage, and OSS previous usage, in order to try
3155 to optimize storage space and networking resources.
3157 Severity : enhancement
3159 Description: Permanently set tunables
3160 Details : All writable /proc/fs/lustre tunables can now be permanently
3161 set on a per-server basis, at mkfs time or on a live
3164 Severity : enhancement
3166 Description: Lustre message v2
3167 Details : Add lustre message format v2.
3169 Severity : enhancement
3171 Description: client OST exclusion list
3172 Details : Clients can be started with a list of OSTs that should be
3173 declared "inactive" for known non-responsive OSTs.
3177 Description: ENOENT returned for valid filehandle during dbench.
3178 Details : Check if a directory has children when invalidating dentries
3179 associated with an inode during lock cancellation. This fixes
3180 an incorrect ENOENT sometimes seen for valid filehandles during
3181 testing with dbench.
3184 Frequency : SFS test only (otherwise harmless)
3186 Description: SPEC SFS validation failure on NFS v2 over lustre.
3187 Details : Changes the blocksize for regular files to be 2x RPC size,
3188 and not depend on stripe size.
3190 Severity : enhancement
3192 Description: fine-grained SMP locking inside DLM
3193 Details : Improve DLM performance on SMP systems by removing the single
3194 per-namespace lock and replace it with per-resource locks.
3196 Severity : enhancement
3198 Description: don't hold multiple extent locks at one time
3199 Details : To avoid client eviction during large writes, locks are not
3200 held on multiple stripes at one time or for very large writes.
3201 Otherwise, clients can block waiting for a lock on a failed OST
3202 while holding locks on other OSTs and be evicted.
3204 Severity : enhancement
3206 Description: Multiple MD RPCs in flight.
3207 Details : Further unserialise some read-only MDT RPCs - learn about intents.
3208 To avoid overly-overloading MDT, introduce a limit on number of
3209 MDT RPCs in flight for a single client and add /proc controls
3210 to adjust this limit.
3212 Severity : enhancement
3214 Description: client read/write statistics
3215 Details : Add client read/write call usage stats for performance
3216 analysis of user processes.
3217 /proc/fs/lustre/llite/*/offset_stats shows non-sequential
3218 file access. extents_stats shows chunk size distribution.
3219 extents_stats_per_process show chunk size distribution per
3222 Severity : enhancement
3224 Description: per-client statistics on server
3225 Details : Add ldlm and operations statistics for each client in
3226 /proc/fs/lustre/mds|obdfilter/*/exports/
3228 Severity : enhancement
3230 Description: improved MDT statistics
3231 Details : Add detailed MDT operations statistics in
3232 /proc/fs/lustre/mds/*/stats
3234 Severity : enhancement
3236 Description: VFS operations stats
3237 Details : Add client VFS call stats, trackable by pid, ppid, or gid
3238 /proc/fs/lustre/llite/*/stats_track_[pid|ppid|gid]
3243 Description: Fix client-side osc byte counters
3244 Details : The osc read/write byte counters in
3245 /proc/fs/lustre/osc/*/stats are now working
3248 Frequency : always as root on SLES
3250 Description: Failure of copying files with lustre special EAs.
3251 Details : Client side always return success for setxattr call for lustre
3252 special xattr (currently only "trusted.lov").
3257 Description: Refcount LNET uuids
3258 Details : The global LNET uuid list grew linearly with every startup;
3259 refcount repeated list entries instead of always adding to
3262 Severity : enhancement
3264 Description: Dynamic service threads
3265 Details : Within a small range, start extra service threads
3266 automatically when the request queue builds up.
3269 Frequency : mixed-endian client/server environments
3271 Description: mixed-endian crashes
3272 Details : The new msg_v2 system had some failures in mixed-endian
3275 Severity : enhancement
3277 Description: Easy OST removal
3278 Details : OSTs can be permanently deactivated with e.g. 'lctl
3279 conf_param lustre-OST0001.osc.active=0'
3281 Severity : enhancement
3283 Description: MGS proc entries
3284 Details : Added basic proc entries for the MGS showing what filesystems
3287 Severity : enhancement
3289 Description: provide MGS failover
3290 Details : Added config lock reacquisition after MGS server failover.
3292 Severity : enhancement
3294 Description: add Linux 2.4 support
3295 Details : Added support for RHEL 2.4.21 kernel for 1.6 servers and clients
3299 Description: a large application tries to do I/O to the same resource and dies
3300 in the middle of it.
3301 Details : Check the req->rq_arrival time after the call to
3302 ost_brw_lock_get(), but before we do anything about
3303 processing it & sending the BULK transfer request. This
3304 should help move old stale pending locks off the queue as
3305 quickly as obd_timeout.
3308 Frequency : when an incorrect nid is specified during startup
3310 Description: ptlrpc connect to non-existant node causes kernel crash
3311 Details : LNET can't be re-entered from an event callback, which
3312 happened when we expire a message after the export has been
3313 cleaned up. Instead, hand the zombie cleanup off to another
3316 Severity : enhancement
3318 Description: plain/inodebits lock performance improvement
3319 Details : Grouping plain/inodebits in granted list by their request modes
3320 and bits policy, thus improving the performance of search through
3324 Frequency : only if OST filesystem is corrupted
3326 Description: client incorrectly hits assertion in ptlrpc_replay_req()
3327 Details : for a short time RPCs with bulk IO are in the replay list,
3328 but replay of bulk IOs is unimplemented. If the OST filesystem
3329 is corrupted due to disk cache incoherency and then replay is
3330 started it is possible to trip an assertion. Avoid putting
3331 committed RPCs into the replay list at all to avoid this issue.
3334 Frequency : liblustre (e.g. catamount) on a large cluster with >= 8 OSTs/OSS
3336 Description: System hang on startup
3337 Details : This bug allowed the liblustre (e.g. catamount) client to
3338 return to the app before handling all startup RPCs. This
3339 could leave the node unresponsive to lustre network traffic
3340 and manifested as a server ptllnd timeout.
3342 Severity : enhancement
3344 Description: Add "/proc/sys/lustre/debug_peer_on_timeout"
3345 Details : liblustre envirable: LIBLUSTRE_DEBUG_PEER_ON_TIMEOUT
3346 boolean to control whether to print peer debug info when a
3347 client's RPC times out.
3350 Frequency : only for kernels with patches from Lustre below 1.4.3
3352 Description: Remove old rdonly API
3353 Details : Remove old rdonly API which unused from at least lustre 1.4.3
3356 Frequency : only for devices with external journals
3358 Description: Set external device read-only also
3359 Details : During a commanded failover stop, we set the disk device
3360 read-only while the server shuts down. We now also set any
3361 external journal device read-only at the same time.
3364 Frequency : when upgrading from 1.4 while trying to change parameters
3366 Description: The wrong (new) MDC name was used when setting parameters for
3367 upgraded MDT's. Also allows changing of OSC (and MDC)
3368 parameters if --writeconf is specified at tunefs upgrade time.
3371 Frequency : when setting specific ost indicies
3373 Description: QOS code breaks on skipped indicies
3374 Details : Add checks for missing OST indicies in the QOS code, so OSTs
3375 created with --index need not be sequential.
3377 Severity : enhancement
3379 Description: Add uninit_groups feature to ldiskfs2 to speed up e2fsck
3380 Details : The uninit_groups feature works in conjunction with the kernel
3381 filesystem code (ldiskfs2 only) and e2fsprogs-1.39-cfs6 to speed
3382 up the pass1 processing of e2fsck. This is a read-only feature
3383 in ldiskfs2 only, so older kernels and current ldiskfs cannot
3384 mount filesystems that have had this feature enabled.
3386 Severity : enhancement
3388 Description: Improve multi-block allocation algorithm to avoid fragmentation
3389 Details : The mballoc3 code (ldiskfs2 only) adds new mechanisms to improve
3390 allocation locality and avoid filesystem fragmentation.
3392 ------------------------------------------------------------------------------
3394 2007-04-01 Cluster File Systems, Inc. <info@clusterfs.com>
3396 * Support for kernels:
3397 2.4.21-47.0.1.EL (RHEL 3)
3398 2.6.5-7.283 (SLES 9)
3399 2.6.9-42.0.10.EL (RHEL 4)
3400 2.6.12.6 vanilla (kernel.org)
3401 2.6.16.27-0.9 (SLES 10)
3402 * Recommended e2fsprogs version: 1.39.cfs5
3404 * Note that reiserfs quotas are disabled on SLES 10 in this kernel
3408 Frequency : occasional, depends on client load and configuration
3409 Bugzilla : 12181, 12203
3410 Description: data loss for recently-modified files
3412 Details : In some cases it is possible that recently written or created
3413 files may not be written to disk in a timely manner (this should
3414 normally be within 30s unless client IO load is very high).
3415 The problem appears as zero-length files or files that are a
3416 multiple of 1MB in size after a client crash or client eviction
3417 that are missing data at the end of the file.
3419 This problem is more likely to be hit on clients where files are
3420 repeatedly created and unlinked in the same directory, clients
3421 have a large amount of RAM, have many CPUs, the filesystem has
3422 many OSTs, the clients are rebooted frequently, and/or the files
3423 are not accessed by other nodes after being written.
3425 The presence of the problem can be detected by looking at
3426 /proc/sys/fs/inode-state. If the first number (nr_inodes) is
3427 smaller than the second (nr_unused) then dirty files will not
3428 be flushed automatically to disk. "sync; sleep 10" should be
3429 run several times on the node before unmounting it to update
3430 Lustre (this is also safe to run on nodes without this problem).
3432 There is also a related kernel bug in the RHEL4 4 2.6.9 kernel
3433 that can cause this same problem, so customers using that kernel
3434 also need to update the kernel in addition to Lustre. In order
3435 to properly fix this bug, the RHEL3 2.4.21 kernel is also updated.
3437 It is normal that files written just before a client crash (less
3438 than 30s) may not yet have been flushed to disk, even for local
3442 Frequency : frequent on thin XT3 nodes
3444 Description: UUID collision on thin XT3 Linux nodes
3445 Details : UUIDs on Compute Node Linux XT3 nodes were not generated
3446 randomly, since we relied on an insufficiently-seeded PRNG.
3451 Description: OSS hangs after "All ost request buffers busy"
3452 Details : A deadlock between quota and journal operations caused OSS
3453 hangs after printing "All ost request buffers busy."
3456 Frequency : always on liblustre builds
3458 Description: Cleanup compiler warnings on liblustre
3461 Frequency : always on liblustre builds on XT3
3463 Description: LC_CONFIG_CDEBUG don't run while build liblustre on XT3.
3467 Description: Addition of EXT3_FEATURE_RO_COMPAT_DIR_NLINKS flag for
3468 > 32000 subdirectories
3469 Details : Add EXT3_FEATURE_RO_COMPAT_DIR_NLINK flag to
3470 EXT3_FEATURE_RO_COMPAT_SUPP. This flag will be set whenever
3471 subdirectory count crosses 32000. This will aid e2fsck to
3472 correctly handle more than 32000 subdirectories.
3475 Frequency : liblustre (e.g. catamount) on a large cluster with >= 8 OSTs/OSS
3477 Description: System hang on startup
3478 Details : This bug allowed the liblustre (e.g. catamount) client to
3479 return to the app before handling all startup RPCs. This
3480 could leave the node unresponsive to lustre network traffic
3481 and manifested as a server ptllnd timeout.
3483 Severity : enhancement
3485 Description: Add "/proc/sys/lustre/debug_peer_on_timeout"
3486 (liblustre envirable: LIBLUSTRE_DEBUG_PEER_ON_TIMEOUT)
3487 boolean to control whether to print peer debug info when a
3488 client's RPC times out.
3493 Description: make O_SYNC working on 2.6 kernels
3494 Details : 2.6 kernels use different method for mark pages for write,
3495 so need add a code to lustre for O_SYNC work.
3500 Description: Failure to close file and release space on NFS
3501 Details : Put inode details into lock acquired in ll_intent_file_open.
3502 Use mdc_intent_lock in ll_intent_open to properly
3503 detect all kind of errors unhandled by mdc_enqueue.
3508 Description: proc file read during shutdown sometimes raced obd removal,
3510 Details : Add lock to prevent obd access after proc file removal.
3513 Frequency : Only for files larger than 4GB on 32-bit clients.
3515 Description: improperly doing page alignment of locks
3516 Details : Modify lustre core code to use CFS_PAGE_* defines instead of
3517 PAGE_*. Make CFS_PAGE_MASK a 64-bit mask.
3522 Description: RPCs being resent when they shouldn't be
3523 Details : Some RPCs that should not be resent are being resent. This
3524 can cause inconsistencies in the RPC state machine. Do not
3525 resend such requests.
3528 Frequency : rare, only with NFS export
3530 Description: Crash on NFS re-export node
3531 Details : under very unusual load conditions an assertion is hit in
3532 ll_intent_file_open()
3535 Frequency : only if OST filesystem is corrupted
3537 Description: client incorrectly hits assertion in ptlrpc_replay_req()
3538 Details : for a short time RPCs with bulk IO are in the replay list,
3539 but replay of bulk IOs is unimplemented. If the OST filesystem
3540 is corrupted due to disk cache incoherency and then replay is
3541 started it is possible to trip an assertion. Avoid putting
3542 committed RPCs into the replay list at all to avoid this issue.
3547 Description: large O_DIRECT requests fail under memory pressure/fragmentation
3548 Details : Large single O_DIRECT read and write calls can fail to allocate
3549 a sufficiently large buffer to process the request. In case of
3550 allocation failure the allocation is retried with a smaller
3551 buffer and broken into smaller requests.
3553 Severity : enhancement
3555 Description: Add -o localflock option to simulate old noflock behaviour.
3556 Details : This will achieve local-only flock/fcntl locks coherentness.
3561 Description: versioning check is incomplete
3562 Details : Checking the version difference of client vs. server, report
3563 error if the gap is too big.
3568 Description: add support PG_writeback bit
3569 Details : add support for PG_writeback bit for Lustre, for more carefull
3570 work with page cache in 2.6 kernel. This also fix some deadlocks
3571 and remove hack for work O_SYNC with 2.6 kernel.
3573 Severity : enhancement
3575 Description: Add uninit_groups feature to ldiskfs2 to speed up e2fsck
3576 Details : The uninit_groups feature works in conjunction with the kernel
3577 filesystem code (ldiskfs2 only) and e2fsprogs-1.39-cfs6 to speed
3578 up the pass1 processing of e2fsck. This is a read-only feature
3579 in ldiskfs2 only, so older kernels and current ldiskfs cannot
3580 mount filesystems that have had this feature enabled.
3582 Severity : enhancement
3584 Description: Improve multi-block allocation algorithm to avoid fragmentation
3585 Details : The mballoc3 code (ldiskfs2 only) adds new mechanisms to improve
3586 allocation locality and avoid filesystem fragmentation.
3588 ------------------------------------------------------------------------------
3590 2007-02-09 Cluster File Systems, Inc. <info@clusterfs.com>
3592 * Support for kernels:
3593 2.6.9-42.0.3.EL (RHEL 4)
3594 2.6.5-7.276 (SLES 9)
3595 2.4.21-47.0.1.EL (RHEL 3)
3596 2.6.12.6 vanilla (kernel.org)
3597 2.6.16.21-0.8 (SLES10)
3598 * Recommended e2fsprogs version: 1.39.cfs2-0
3600 * The backwards-compatible /proc/sys/portals symlink has been removed
3601 in this release. Before upgrading, please ensure that you change
3602 any configuration scripts or /etc/sysctl.conf files that access
3603 /proc/sys/portals/* or sysctl portals.* to use the corresponding
3604 entry in /proc/sys/lnet or sysctl lnet.*. This change can be made
3605 in advance of the upgrade on any system running Lustre 1.4.6 or
3606 newer, since /proc/sys/lnet was added in that version.
3607 * Note that reiserfs quotas are disabled on SLES 10 in this kernel
3611 Frequency : only when quota is used
3613 Description: avoid scanning export list for quota master
3614 Details : Change the algorithms to avoid scanning export list in order
3615 to improve the efficiency.
3618 Frequency : MDS failover only, very rarely
3620 Description: "went back in time" messages on mds failover
3621 Details : The greatest transno may be lost when the current operation
3622 finishes with an error (transno==0) and the client's last_rcvd
3623 record is over-written. Save the greatest transno in the
3624 mds_last_transno for this case.
3627 Frequency : always for specific kernels and striping counts
3629 Description: client may get "Matching packet too big" without ACL support
3630 Details : Clients compiled without CONFIG_FS_POSIX_ACL get an error message
3631 when trying to access files in certain configurations. The
3632 clients should in fact be denied when mounting because they do
3633 not understand ACLs.
3636 Frequency : Cray XT3 with more than 4000 clients and multiple jobs
3638 Description: many clients connecting with IO in progress causes connect timeouts
3639 Details : Avoid synchronous journal commits to avoid delays caused by many
3640 clients connecting/disconnecting when bulk IO is in progress.
3641 Queue liblustre connect requests on OST_REQUEST_PORTAL instead of
3642 OST_IO_PORTAL to avoid delays behind potentially many pending
3646 Frequency : occasionally with multiple writers to a single file
3648 Description: shared writes to file may result in wrong size reported by stat()
3649 Details : Allow growing of kms when extent lock is cancelled
3652 Frequency : always with random mmap IO to multi-striped file
3654 Description: mmap write might be lost if we are writing to a 'hole' in stripe
3655 Details : Only if the hole is at the end of OST object so that kms is too
3656 small. Fix is to increase kms accordingly in ll_nopage.
3659 Frequency : rare, only if OST filesystem is inconsistent with MDS filesystem
3661 Description: writes to a missing object would leak memory on the OST
3662 Details : If there is an inconsistency between the MDS and OST filesystems,
3663 such that the MDS references an object that doesn't exist, writes
3664 to that object will leak memory due to incorrect cleanup in the
3665 error handling path, eventually running out of memory on the OST.
3670 Description: Creating too long symlink causes lustre errors
3671 Details : Check symlink and name lengths before sending requests to MDS.
3674 Frequency : only if flock is enabled (not on by default)
3676 Description: posix locks not released on fd closure on 2.6.9+
3677 Details : We failed to add posix locks to list of inode locks on 2.6.9+
3678 kernels, this caused such locks not to be released on fd close and
3679 then assertions on fs unmount about still used locks.
3682 Frequency : MDS failover only, very rarely
3684 Description: clients may get ASSERTION(granted_lock != NULL)
3685 Details : When request was taking a long time, and a client was resending
3686 a getattr by name lock request. The were multiple lock requests
3687 with the same client lock handle and
3688 mds_getattr_name->fixup_handle_for_resent_request found one of the
3689 lock handles but later failed with ASSERTION(granted_lock != NULL).
3694 Description: handle->h_buffer_credits > 0, assertion failure
3695 Details : h_buffer_credits is zero after truncate, causing assertion
3696 failure. This patch extends the transaction or creates a new
3700 Frequency : NFS re-export or patchless client
3701 Bugzilla : 11179, 10796
3702 Description: Crash on NFS re-export node (__d_move)
3703 Details : We do not want to hash the dentry if we don't have a lock.
3704 But if this dentry is later used in d_move, we'd hit uninitialised
3705 list head d_hash, so we just do this to init d_hash field but
3706 leave dentry unhashed.
3709 Frequency : NFS re-export or patchless client
3711 Description: NFS exports has problem with symbolic link
3712 Details : lustre client didn't properly install dentry when re-exported
3713 to NFS or running patchless client.
3716 Frequency : NFS re-export or patchless client
3718 Description: Various nfs/patchless fixes.
3719 Details : fixes reuse disconected alias for lookup process - this fixes
3720 warning "find_exported_dentry: npd != pd",
3721 fix permission error with open files at nfs.
3722 fix apply umask when do revalidate.
3725 Frequency : occasional
3727 Description: Crash on NFS re-export node
3728 Details : calling clear_page() on the wrong pointer triggered oops in
3729 generic_mapping_read().
3732 Frequency : rarely, using O_DIRECT IO
3734 Description: unaligned directio crashes client with LASSERT
3735 Details : check for unaligned buffers before trying any requests.
3738 Frequency : rarely, using CFS RAID5 patches in non-standard kernel series
3740 Description: stale data returned from RAID cache
3741 Details : If only a small amount of IO is done to the RAID device before
3742 reading it again it is possible to get stale data from the RAID
3743 cache instead of reading it from disk.
3746 Frequency : always for sles10 kernel
3748 Description: sles10 support
3749 Details : ll_follow_link: compile fixes and using of nd_set_link
3750 under newer kernels.
3753 Frequency : depends on arch, kernel and compiler version, always on sles10
3756 Description: recursive or deep enough symlinks cause stack overflow
3757 Details : getting rid of large stack-allocated variable in
3761 Frequency : depends on hardware
3763 Description: lustre write performance loss in the SLES10 kernel
3764 Details : the performance loss is caused by using of write barriers in the
3765 ext3 code. The SLES10 kernel turns barrier support on by
3766 default. The fix is to undo that change for ldiskfs.
3768 ------------------------------------------------------------------------------
3770 2006-12-09 Cluster File Systems, Inc. <info@clusterfs.com>
3772 * Support for kernels:
3773 2.6.9-42.0.3EL (RHEL 4)
3774 2.6.5-7.276 (SLES 9)
3775 2.4.21-47.0.1.EL (RHEL 3)
3776 2.6.12.6 vanilla (kernel.org)
3780 Frequency : quota enabled and large files being deleted
3782 Description: releasing more than 4GB of quota at once hangs OST
3783 Details : If a user deletes more than 4GB of files on a single OST it
3784 will cause the OST to spin in an infinite loop. Release
3785 quota in < 4GB chunks, or use a 64-bit value for 1.4.7.1+.
3790 Description: statfs data retrieved from /proc may be stale or zero
3791 Details : When reading per-device statfs data from /proc, in the
3792 {kbytes,files}_{total,free,avail} files, it may appear
3793 as zero or be out of date.
3796 Frequency : systems with MD RAID1 external journal devices
3798 Description: lconf's call to blkid is confused by RAID1 journal devices
3799 Details : Use the "blkid -l" flag to locate the MD RAID device instead
3800 of returning all block devices that match the journal UUID.
3803 Frequency : always, for aggregate stripe size over 4GB
3805 Description: "lfs setstripe" fails assertion when setting 4GB+ stripe width
3806 Details : Using "lfs setstripe" to set stripe size * stripe count over 4GB
3807 will fail the kernel with "ASSERTION(lsm->lsm_xfersize != 0)"
3810 Frequency : always if "lfs find" used on a local file/directory
3812 Description: "lfs find" segfaults if used on a local file/directory
3813 Details : The case where a directory component was not specified wasn't
3814 handled correctly. Handle this properly.
3817 Frequency : always on ppc64
3819 Description: the write to an ext3 filesystem mounted with mballoc got stuck
3820 Details : ext3_mb_generate_buddy() uses find_next_bit() which does not
3821 perform endianness conversion.
3824 Frequency : rarely (truncate to non-zero file size after write under load)
3825 Bugzilla : 10730, 10687
3826 Description: Files padded with zeros to next 4K multiple
3827 Details : With filesystems mounted using the "extents" option (2.6 kernels)
3828 it is possible that files that are truncated to a non-zero size
3829 immediately after being written are filled with zero bytes beyond
3830 the truncated size. No file data is lost.
3832 Severity : enhancement
3834 Description: Allow recovery/failover for liblustre clients.
3835 Details : liblustre clients were unaware of failover configurations until
3838 Severity : enhancement
3840 Description: user file locks should fail when not mounting with flock option
3841 Details : Set up an error-returning stub in ll_file_operations.lock field
3842 to prevent incorrect behaviour when client is mounted without
3843 flock option. Also, set up properly f_op->flock field for
3847 Frequency : always on ia64
3849 Description: "lfs df" loops on printing out MDS statfs information
3850 Details : The obd_ioctl_data was not initialized and in some systems
3851 this caused a failure during the ioctl that did not return
3852 an error. Initialize the struct and return an error on failure.
3855 Frequency : SLES 9 only
3857 Description: Error of copying files with lustre special EAs as root
3858 Details : Client side always return success for setxattr call for lustre
3859 special xattr (currently only "trusted.lov").
3862 Frequency : rarely on clusters with both ia64+i386 clients
3864 Description: ia64+i686 clients doing shared IO on the same file may LBUG
3865 Details : In rare cases when both ia64+i686 (or other mixed-PAGE_SIZE)
3866 clients are doing concurrent writes to the same file it is
3867 possible that the ia64 clients may LASSERT because the OST
3868 extent locks are not PAGE_SIZE aligned. Ensure that grown
3869 locks are always aligned on the request boundary.
3872 Frequency : specific use, occasional
3874 Description: Overwriting in use executable truncates on-disk binary image
3875 Details : If one node attempts to overwrite an executable in use by
3876 another node, we now correctly return ETXTBSY instead of
3877 truncating the file.
3879 Severity : enhancement
3881 Description: Async OSC create to avoid the blocking unnecessarily.
3882 Details : If a OST has no remain object, system will block on the creating
3883 when need to create a new object on this OST. Now, ways use
3884 pre-created objects when available, instead of blocking on an
3885 empty osc while others are not empty. If we must block, we block
3886 for the shortest possible period of time.
3891 Description: chmod on Lustre root is propagated to other clients
3892 Details : Re-validate root's dentry in ll_lookup_it to avoid having it
3893 invalid by the follow_mount time.
3896 Frequency : liblustre clients only
3898 Description: Race in 'instant cancel' lock handling could lead to such locks
3899 never to be granted in case of SMP MDS
3900 Details : Do not destroy not yet granted but cbpending locks in
3904 Frequency : replay/resend of open
3906 Description: non null lock assetion failure in mds_intent_policy
3907 Details : Trying to replay/resend lockless open requests resulted in
3908 mds_open() returning 0 with no lock. Now it sets a flag if
3909 a lock is going to be returned.
3911 Severity : enhancement
3913 Description: Checksum enhancements
3914 Details : New checksum enhancements allow for resending RPCs that failed
3917 Severity : enhancement
3919 Description: Tunables on number of dirty pages in cacche
3920 Details : Allow to set limit on number of dirty pages cached.
3925 Description: client crash on unmount - lock still has references
3926 Details : In some error handling cases it was possible to leak a lock
3927 reference on a client while accessing a file. This was not
3928 harmful to the client during operation, but would cause the
3929 client to crash when the filesystem is unmounted.
3932 Frequency : specific case, rare
3934 Description: ETXTBSY on mds though file not in use by client
3935 Details : ETXTBSY is no longer incorrectly returned when attempting to
3936 chmod or chown a directory that the user previously tried to
3937 execute or a currently-executing binary.
3940 Frequency : extremely rare except on liblustre-based clients
3942 Description: Lustre space not freed when files are deleted
3943 Details : Clean up open-unlinked files after client eviction. Previously
3944 the unlink was skipped and the files remained as orphans.
3949 Description: OST failure "would be an LBUG" in waiting_locks_callback()
3950 Details : In some cases it was possible to send a blocking callback to
3951 a client doing a glimpse, even though that client didn't get
3952 a lock granted. When the glimpse lock is cancelled on the OST
3953 the freed lock is left on the waiting list and corrupted the list.
3956 Frequency : all core dumps
3958 Description: Broke core dumps to lustre
3959 Details : Negative dentry may be unhashed if parent does not have UPDATE
3960 lock, but some callers, e.g. do_coredump, expect dentry to be
3961 hashed after successful create, hash it in ll_create_it.
3963 ------------------------------------------------------------------------------
3965 2006-09-13 Cluster File Systems, Inc. <info@clusterfs.com>
3967 * Support for kernels:
3968 2.6.9-42.0.2.EL (RHEL 4)
3969 2.6.5-7.276 (SLES 9)
3970 2.4.21-40.EL (RHEL 3)
3971 2.6.12.6 vanilla (kernel.org)
3975 Frequency : always on RHEL 3
3977 Description: Number of open files grows over time
3978 Details : The number of open files grows over time, whether or not
3979 Lustre is started. This was due to a filp leak introduced
3980 by one of our kernel patches.
3982 ------------------------------------------------------------------------------
3984 2006-08-20 Cluster File Systems, Inc. <info@clusterfs.com>
3986 * Support for kernels:
3987 2.6.9-42.EL (RHEL 4)
3988 2.6.5-7.267 (SLES 9)
3989 2.4.21-40.EL (RHEL 3)
3990 2.6.12.6 vanilla (kernel.org)
3995 Bugzilla : 5719, 9635, 9792, 9684
3996 Description: OST (or MDS) trips assertions in (re)connection under heavy load
3997 Details : If a server is under heavy load and cannot reply to new
3998 connection requests before the client resends the (re)connect,
3999 the connection handling code can behave badly if two service
4000 threads are concurrently handing separate (re)connections from
4001 the same client. Add better locking to the connection handling
4002 code, and ensure that only a single connection will be processed
4003 for a given client UUID, even if the lock is dropped.
4005 Severity : enhancement
4007 Description: add TCP zero-copy support to kernel
4008 Details : Add support to the kernel TCP stack to allow zero-copy bulk
4009 sends if the hardware supports scatter-gather and checksumming.
4010 This allows socklnd to do client-write and server-read more
4011 efficiently and reduce CPU utilization from skbuf copying.
4014 Frequency : only if NFS exporting from client
4016 Description: NULL pointer deref in ll_iocontrol() if chattr mknod file
4017 Details : If setting attributes on a file created under NFS that had
4018 never been opened it would be possible to oops the client
4019 if the file had no objects.
4023 Bugzilla : 9326, 10402, 10897
4024 Description: client crash in ptlrpcd_wake() thread when sending async RPC
4025 Details : It is possible that ptlrpcd_wake() dereferences a freed async
4026 RPC. In rare cases the ptlrpcd thread alread processed the RPC
4027 before ptlrpcd_wake() was called and the request was freed.
4030 Frequency : always for liblustre
4032 Description: liblustre client does MDS+OSTs setattr RPC for each write
4033 Details : When doing a write from a liblustre client, the client
4034 incorrectly issued an RPC to the MDS and each OST the file was
4035 striped over in order to update the timestamps. When writing
4036 with small chunks and many clients this could overwhelm the MDS
4037 with RPCs. In all cases it would slow down the write because
4038 these RPCs are unnecessary.
4040 Severity : enhancement
4042 Description: allow number of MDS service threads to be changed at module load
4043 Details : It is now possible to change the number of MDS service threads
4044 running. Adding "options mds mds_num_threads={N}" to the MDS's
4045 /etc/modprobe.conf will set the number of threads for the next
4046 time Lustre is restarted (assuming the "mds" module is also
4047 reloaded at that time). The default number of threads will
4048 stay the same, 32 for most systems.
4053 Description: OST crash if filesystem is unformatted or corrupt
4054 Details : If an OST is started on a device that has never been formatted
4055 or if the filesystem is corrupt and cannot even mount then the
4056 error handling cleanup routines would dereference a NULL pointer.
4061 Description: NULL pointer deref in llap_from_page.
4062 Details : get_cache_page_nowait can return a page with NULL (or otherwise
4063 incorrect) mapping if the page was truncated/reclaimed while it was
4064 searched for. Check for this condition and skip such pages when
4065 doing readahead. Introduce extra check to llap_from_page() to
4066 verify page->mapping->host is non-NULL (so page is not anonymous).
4069 Frequency : Sometimes when using sys_sendfile
4071 Description: "page not covered by a lock" warnings from ll_readpage
4072 Details : sendfile called ll_readpage without right page locks present.
4073 Now we introduced ll_file_sendfile that does necessary locking
4074 around call to generic_file_sendfile() much like we do in
4078 Frequency : with certain MDS communication failures at client mount time
4080 Description: NULL pointer deref after failed client mount
4081 Details : a client connection request may delayed by the network layer
4082 and not be sent until after the PTLRPC layer has timed out the
4083 request. If the client fails the mount immediately it will try
4084 to clean up before the network times out the request. Add a
4085 reference from the request import to the obd device and delay
4086 the cleanup until the network drops the request.
4089 Frequency : occasionally during client (re)connect
4091 Description: assertion failure during client (re)connect
4092 Details : processing a client connection request may be delayed by the
4093 client or server longer than the client connect timeout. This
4094 causes the client to resend the connection request. If the
4095 original connection request is replied in this interval, the
4096 client may trip an assertion failure in ptlrpc_connect_interpret()
4097 which thought it would be the only running connect process.
4100 Frequency : only with obd_echo servers and clients that are rebooted
4102 Description: kernel BUG accessing uninitialized data structure
4103 Details : When running an obd_echo server it did not start the ping_evictor
4104 thread, and when a client was evicted an uninitialized data
4105 structure was accessed. Start the ping_evictor in the RPC
4106 service startup instead of the OBD startup.
4108 Severity : enhancement
4109 Bugzilla : 10193 (patchless)
4110 Description: Remove dependency on various unexported kernel interfaces.
4111 Details : No longer need reparent_to_init, exit_mm, exit_files,
4112 sock_getsockopt, filemap_populate, FMODE_EXEC, put_filp.
4115 Frequency : rare (only users of deprecated and unsupported LDAP config)
4117 Description: write_conf for zeroconf mount queried LDAP incorrectly for client
4118 Details : LDAP apparently contains 'lustreName' attributes instead of
4119 'name'. A simple remapping of the name is sufficient.
4122 Frequency : rare (only with non-default dump_on_timeout debug enabled)
4124 Description: waiting_locks_callback trips kernel BUG if client is evicted
4125 Details : Running with the dump_on_timeout debug flag turned on makes
4126 it possible that the waiting_locks_callback() can try to dump
4127 the Lustre kernel debug logs from an interrupt handler. Defer
4128 this log dumping to the expired_lock_main() thread.
4130 Severity : enhancement
4132 Description: Support NFS exporting on 2.6 kernels.
4133 Details : Implement non-rawops metadata methods for NFS server to use without
4134 changing NFS server code.
4137 Frequency : very rare (synthetic metadata workload only)
4139 Description: two racing renames might cause an MDS thread to deadlock
4140 Details : Running the "racer" program may cause one MDS thread to rename
4141 a file from being the source of a rename to being the target of
4142 a rename at exactly the same time that another thread is doing
4143 so, and the second thread has already enqueued these locks after
4144 doing a lookup of the target and is trying to relock them in
4145 order. Ensure that we don't try to re-lock the same resource.
4148 Frequency : only very large systems with liblustre clients
4150 Description: slow eviction of liblustre clients with the "evict_by_nid" RPC
4151 Details : Use asynchronous set_info RPCs to send the "evict_by_nid" to
4152 all OSTs in parallel. This allows the eviction of stale liblustre
4153 clients to proceed much faster than if they were done in series,
4154 and also offers similar improvements for other set_info RPCs.
4159 Description: excessive CPU usage during initial read phase on client
4160 Details : During the initial read phase on a client, it would agressively
4161 retry readahead on the file, consuming too much CPU and impacting
4162 performance (since 1.4.5.8). Improve the readahead algorithm
4163 to avoid this, and also improve some other common cases (read
4164 of small files in particular, where "small" is files smaller than
4165 /proc/fs/lustre/llite/*/max_read_ahead_whole_mb, 2MB by default).
4170 Description: MDS crash when receiving packet with unknown intent.
4171 Details : Do not LBUG in unknown intent case, just return -EFAULT
4173 Severity : enhancement
4174 Bugzilla : 9293, 9385
4175 Description: MDS RPCs are serialised on client. This is unnecessary for some.
4176 Details : Do not serialize getattr (non-intent version) and statfs.
4179 Frequency : occasional, when OST network is overloaded/intermittent
4181 Description: client evicted by OST after bulk IO timeout
4182 Details : If a client sends a bulk IO request (read or write) the OST
4183 may evict the client if it is unresposive to its data GET/PUT
4184 request. This is incorrect if the network is overloaded (takes
4185 too long to transfer the RPC data) or dropped the OST GET/PUT
4186 request. There is no need to evict the client at all, since
4187 the pinger and/or lock callbacks will handle this, and the
4188 client can restart the bulk request.
4191 Frequency : Always when mmapping file with no objects
4193 Description: client crashes when mmapping file with no objects
4194 Details : Check that we actually have objects in a file before doing any
4195 operations on objects in ll_vm_open, ll_vm_close and
4201 Description: Request leak when working with deleted CWD
4202 Details : Introduce advanced request refcount tracking for requests
4203 referenced from lustre intent.
4205 Severity : Enhancement
4207 Description: Cache open file handles on client.
4208 Details : MDS now will return special lock along with openhandle, if
4209 requested and client is allowed to hold openhandle, even if unused,
4210 until such a lock is revoked. Helps NFS a lot, since NFS is opening
4211 closing files for every read/write openration.
4213 Severity : Enhancement
4215 Description: Cache open negative dentries on client when possible.
4216 Details : Guard negative dentries with UPDATE lock on parent dir, drop
4217 negative dentries on lock revocation.
4222 Description: Remounting a client read-only wasn't possible with a zconf mount
4223 Details : It wasn't possible to remount a client read-only with llmount.
4225 Severity : enhancement
4226 Description: Include MPICH 1.2.6 Lustre ADIO interface patch
4227 Details : In lustre/contrib/ or /usr/share/lustre in RPM a patch for
4228 MPICH is included to add Lustre-specific ADIO interfaces.
4229 This is based closely on the UFS ADIO layer and only differs
4230 in file creation, in order to allow the OST striping to be set.
4231 This is user-contributed code and not supported by CFS.
4236 Description: extended inode attributes (immutable, append-only) work improperly
4237 when 2.4 and 2.6 kernels are used on client/server or vice versa
4238 Details : Introduce kernel-independent values for these flags.
4240 Severity : enhancement
4243 Description: Allow fractional MB tunings for lustre in /proc/ filesystem.
4244 Details : Many of the /proc/ tunables can only be tuned at a megabyte
4245 granularity. Now, Fractional MB granularity is be supported,
4246 this is very useful for low memory system.
4248 Severity : enhancement
4250 Description: Getattr by fid
4251 Details : Getting a file attributes by its fid, obtaining UPDATE|LOOKUP
4252 locks, avoids extra getattr rpc requests to MDS, allows '/' to
4253 have locks and avoids getattr rpc requests for it on every stat.
4256 Frequency : Always, for filesystems larger than 2TB
4258 Description: ldiskfs crash at mount for filesystem larger than 2TB with mballoc
4259 Details : Kenrel kmalloc limits allocations to 128kB and this prevents
4260 filesystems larger than 2TB to be mounted with mballoc enabled.
4263 Frequency : Always, for 32-bit kernel without CONFIG_LBD and filesystem > 2TB
4265 Description: filesystem corruption for non-standard kernels and very large OSTs
4266 Details : If a 32-bit kernel is compiled without CONFIG_LBD enabled and a
4267 filesystems larger than 2TB is mounted then the kernel will
4268 silently corrupt the start of the filesystem. CONFIG_LBD is
4269 enabled for all CFS-supported kernels, but the possibility of
4270 this happening with a modified kernel config exists.
4272 Severity : enhancement
4274 Description: add client O_DIRECT support for 2.6 kernels
4275 Details : It is now possible to do O_DIRECT reads and writes to files
4276 in the Lustre client mountpoint on 2.6 kernel clients.
4278 Severity : enhancement
4280 Description: parallel glimpse, setattr, statfs, punch, destroy requests
4281 Details : Sends glimpse, setattr, statfs, punch, destroy requests to OSTs in
4282 parallel, not waiting for response from every OST before sending
4283 a rpc to the next OST.
4288 Description: setattr vs write race when updating file timestamps
4289 Details : Client processes that update a file timestamp into the past
4290 right after writing to the file (e.g. tar) it is possible that
4291 the updated file modification time can be reset to the current
4292 time due to a race between processing the setattr and write RPC.
4294 Severity : enhancement
4296 Description: Bring 'lfs find' closer in line with regular Linux find.
4297 Details : lfs find util supports -atime, -mtime, -ctime, -maxdepth, -print,
4298 -print0 options and obtains all the needed info through the lustre
4301 Severity : enhancement
4303 Description: support up to 1024 configured devices on one node
4304 Details : change obd_dev array from statically allocated to dynamically
4305 allocated structs as they are first used to reduce memory usage
4310 Description: Flush dirty partially truncated pages during truncate
4311 Details : Immediatelly flush partially truncated pages in filter_setattr,
4312 this way we completely avoid having any pages in page cache on OST
4313 and can retire ugly workarounds during writes to flush such pages.
4318 Description: i_sem vs transaction deadlock in mds_obd_destroy during unlink.
4319 Details : protect inode from truncation within vfs_unlink() context
4320 just take a reference before calling vfs_unlink() and release it
4321 when parent's i_sem is free.
4324 Frequency : always, if extents are used on OSTs
4326 Description: index ei_leaf_hi (48-bit extension) is not zeroed in extent index
4327 Details : OSTs using the extents format would not zero the high 16 bits of
4328 the index physical block number. This is not a problem for any
4329 OST filesystems smaller than 16TB, and no kernels support ext3
4330 filesystems larger than 16TB yet. This is fixed in 1.4.7 (all
4331 new/modified files) and can be fixed for existing filesystems
4332 with e2fsprogs-1.39-cfs1.
4337 Description: import connection selection may be incorrect if timer wraps
4338 Details : Using a 32-bit jiffies timer with HZ=1000 may cause backup
4339 import connections to be ignored if the 32-bit jiffies counter
4340 wraps. Use a 64-bit jiffies counter.
4343 Frequency : during server recovery
4345 Description: crash after server is denying duplicate export
4346 Details : If clients are resending connect requests to the server, the
4347 server refuses to allow a client to connect multiple times.
4348 Fixed a bug in the handling of this case.
4351 Frequency : very large clusters immediately after boot
4353 Description: LNET request buffers exhausted under heavy short-term load
4354 Details : If a large number of client requests are generated on a service
4355 that has previously never seen so many requests it is possible
4356 that the request buffer growth cannot keep up with the spike in
4357 demand. Instead of dropping incoming requests, they are held in
4358 the LND until the RPC service can accept more requests.
4361 Frequency : Sometimes during replay
4363 Description: Assertion failure in ll_local_open after replay.
4364 Details : If replay happened on an open request reply before we were able
4365 to set replay handler, reply will become not swabbed tripping the
4366 assertion in ll_local_open. Now we set the handler right after
4367 recognising of open request
4370 Frequency : very rare
4372 Description: kernel reports "badness in vsnprintf"
4373 Details : Reading from the "recovery_status" /proc file in small chunks
4374 may cause a negative length in lprocfs_obd_rd_recovery_status()
4375 call to vsnprintf() (which is otherwise harmless). Exit early
4376 if there is no more space in the output buffer.
4378 Severity : enhancement
4380 Description: clear OBD RPC statistics by writing to them
4381 Details : It is now possible to clear the OBD RPC statistics by writing
4382 to the "stats" file.
4387 Description: Client mtime is not the same on different clients after utimes
4388 Details : In some cases, the client was using the utimes() syscall on
4389 a file cached on another node. The clients now validate the
4390 ctime from the MDS + OSTs to determine which one is right.
4395 Description: Inability to activate failout mode
4396 Details : lconf script incorrectly assumed that in python string's numeric
4397 value is used in comparisons.
4400 Frequency : always with multiple stripes per file
4402 Description: Inefficient object allocation for mutli-stripe files
4403 Details : When selecting which OSTs to stripe files over, for files with
4404 a stripe count that divides evenly into the number of OSTs,
4405 the MDS is always picking the same starting OST for each file.
4406 Return the OST selection heuristic to the original design.
4411 Description: mount failures may take full timeout to return an error
4412 Details : Under some heavy load conditions it is possible that a
4413 failed mount can wait for the full obd_timeout interval,
4414 possibly several minutes, before reporting an error.
4415 Instead return an error as soon as the status is known.
4417 ------------------------------------------------------------------------------
4419 2006-02-14 Cluster File Systems, Inc. <info@clusterfs.com>
4421 * WIRE PROTOCOL CHANGE. This version of Lustre networking WILL NOT
4422 INTEROPERATE with older versions automatically. Please read the
4423 user documentation before upgrading any part of a live system.
4424 * WARNING: Lustre networking configuration changes are required with
4425 this release. See https://bugzilla.clusterfs.com/show_bug.cgi?id=10052
4428 * Support for kernels:
4429 2.6.9-22.0.2.EL (RHEL 4)
4430 2.6.5-7.244 (SLES 9)
4431 2.6.12.6 vanilla (kernel.org)
4434 Severity : enhancement
4435 Bugzilla : 7981/8208
4436 Description: Introduced Lustre Networking (LNET)
4437 Details : LNET is new networking infrastructure for Lustre, it includes
4438 a reorganized network configuration mode (see the user
4439 documentation for full details) as well as support for routing
4440 between different network fabrics. Lustre Networking Devices
4441 (LNDs) for the supported network fabrics have also been
4442 created for this new infrastructure.
4444 Severity : enhancement
4445 Description: Introduced Access control lists
4446 Details : clients can set ACLs on files and directories in order to have
4447 more fine-grained permissions than the standard Unix UGO+RWX.
4448 The MDS must be started with the "-o acl" mount option.
4450 Severity : enhancement
4451 Description: Introduced filesystem quotas
4452 Details : Administrators may now establish per-user quotas on the
4455 Severity : enhancement
4457 Description: Configuration change for the XT3
4458 The PTLLND is now used to run Lustre over Portals on the XT3
4459 The configure option(s) --with-cray-portals are no longer used.
4460 Rather --with-portals=<path-to-portals-includes> is used to
4461 enable building on the XT3. In addition to enable XT3 specific
4462 features the option --enable-cray-xt3 must be used.
4467 Description: Running on many-way SMP OSTs can trigger oops in llcd_send()
4468 Details : A race between allocating a new llcd and re-getting the llcd_lock
4469 allowed another thread to grab newly-allocated llcd.
4471 Severity : enhancement
4473 Description: 2.6 OST async journal commit and locking fix to improve performance
4474 Details : The filter_direct_io()+filter_commitrw_write() journal commits for
4475 2.6 kernels are now async as they already were in 2.4 kernels so
4476 that they can commit concurrently with the network bulk transfer.
4477 For block-allocated files the filter allocation semaphore is held
4478 to avoid filesystem fragmentation during allocation. BKL lock
4479 removed for 2.6 xattr operations where it is no longer needed.
4484 Description: lconf incorrectly determined whether two IP networks could talk
4485 Details : In some more complicated routing and multiple-network
4486 configurations, lconf will avoid trying to make a network
4487 connection to a disjoint part of the IP space. It was doing the
4488 math incorrectly for one set of cases.
4493 Description: Fix for potential infinite loop processing records in an llog.
4494 Details : If an llog record is corrupted/zeroed, it is possible to loop
4495 forever in llog_process(). Validate the llog record length
4496 and skip the remainder of the block on error.
4499 Frequency : occasional (liblustre only)
4501 Description: liblustre could not open files whose last component is a symlink
4502 Details : sysio_path_walk() would incorrectly pass the open intent to
4503 intermediate path components.
4506 Frequency : rare (liblustre only with non-standard tuning)
4507 Bugzilla : 7201 (7350)
4508 Description: Tuning the MDC DLM LRU size to zero triggers client LASSERT
4509 Details : llu_lookup_finish_locks() tries to set lock data on a lock
4510 after it has been released, only do this for referenced locks
4512 Severity : enhancement
4514 Description: specifying an (invalid) directory default stripe_size of -1
4515 would reset the directory default striping
4516 Details : stripe_size -1 was used internally to signal directory stripe
4517 removal, now use "all default" to signal dir stripe removal
4518 as a directory striping of "all default" is not useful
4521 Frequency : common for large clusters running liblustre clients
4523 Description: doing an ls when liblustre clients are running is slow
4524 Details : sending a glimpse AST to a liblustre client waits for every AST
4525 to time out, as liblustre clients will not respond. Since they
4526 cannot cache data we refresh the OST lock LVB from disk instead.
4528 Severity : enhancement
4530 Description: doing an ls at the same time as file IO can be slow
4531 Details : enqueue and other "small" requests can be blocked behind many
4532 large IO requests. Create a new OST IO portal for non-IO
4533 requests so they can be processed faster.
4536 Frequency : rare (only HPUX clients mounting unsupported re-exported NFS vol)
4538 Description: an HPUX NFS client would get -EACCESS when ftruncate()ing a newly
4539 created file with mode 000
4540 Details : the Linux NFS server relies on an MDS_OPEN_OWNEROVERRIDE hack to
4541 allow an ftruncate() as a non-root user to a file with mode 000.
4542 Lustre now respects this flag to disable mode checks when
4543 truncating a file owned by the user
4546 Frequency : liblustre-only, when liblustre client dies unexpectedly or becomes
4549 Description: Revoking locks from clients that went dead or catatonic might take
4551 Details : New lock flags FL_CANCEL_ON_BLOCK used by liblustre makes
4552 cancellation of such locks instant on servers without waiting for
4553 any reply from clients. Clients drops these locks when cancel
4554 notification from server is received without replying.
4557 Frequency : liblustre-only, when liblustre client dies or becomes busy
4559 Description: Doing ls on Linux clients can take a long time with active
4561 Details : Liblustre client cannot handle ASTs in timely manner, so avoid
4562 granting such locks to it in the first place if possible. Locks
4563 are taken by proxy on the OST during the read or write and
4564 dropped immediately afterward. Add connect flags handling, do
4565 not grant locks to liblustre clients for glimpse ASTs.
4567 Severity : enhancement
4569 Description: Improve read-ahead algorithm to avoid excessive IO for random reads
4570 Details : Existing read-ahead algorithm is tuned for the case of streamlined
4571 sequential reads and behaves badly with applications doing random
4572 reads. Improve it by reading ahead at least read region, and
4573 avoiding excessive large RPC for small reads.
4575 Severity : enhancement
4577 Description: Creating more than 1000 files for a single job may cause a load
4578 imbalance on the OSTs if there are also a large number of OSTs.
4579 Details : qos_prep_create() uses an OST index reseed value that is an
4580 even multiple of the number of available OSTs so that if the
4581 reseed happens in the middle of the object allocation it will
4582 still utilize the OSTs as uniformly as possible.
4587 Description: OST or MDS may oops in ping_evictor_main()
4588 Details : ping_evictor_main() drops obd_dev_lock if deleting a stale export
4589 but doesn't restart at beginning of obd_exports_timed list
4592 Severity : enhancement
4594 Description: improve by-nid export eviction on the MDS and OST
4595 Details : allow multiple exports with the same NID to be evicted at one
4596 time without re-searching the exports list.
4599 Frequency : rare, only with supplementary groups enabled on SMP 2.6 kernels
4601 Description: MDS may oops in groups_free()
4602 Details : in rare race conditions a newly allocated group_info struct is
4603 freed again, and this can be NULL. The 2.4 compatibility code
4604 for groups_free() checked for a NULL pointer, but 2.6 did not.
4607 Frequency : common for liblustre clients doing little filesystem IO
4608 Bugzilla : 9352, 7313
4609 Description: server may evict liblustre clients accessing contended locks
4610 Details : if a client is granted a lock or receives a completion AST
4611 with a blocking AST already set it would not reply to the AST
4612 for LDLM_FL_CANCEL_ON_BLOCK locks. It now replies to such ASTs.
4615 Frequency : lfs setstripe, only systems with more than 160 OSTs
4617 Description: unable to set striping with a starting offset beyond OST 160
4618 Details : llapi_create_file() incorrectly limited the starting stripe
4619 index to the maximum single-file stripe count.
4622 Frequency : LDAP users only
4624 Description: lconf did not handle in-kernel recovery with LDAP properly
4625 Details : lconf/LustreDB get_refs() is searching the wrong namespace
4627 Severity : enhancement
4629 Description: bind OST threads to NUMA nodes to improve performance
4630 Details : all OST threads are uniformly bound to CPUs on a single NUMA
4631 node and do their allocations there to localize memory access
4633 Severity : enhancement
4635 Description: llmount can determine client NID directly from Myrinet (GM)
4636 Details : the client NID code from gmnalnid was moved directly into
4637 llmount, removing the need to use this or specifying the
4638 client NID explicitly when mounting GM clients with zeroconf
4641 Frequency : if client is started with down MDS
4643 Description: if client is started with down MDS mount hangs in ptlrpc_queue_wait
4644 Details : Having an LWI_INTR() wait event (interruptible, but no timeout)
4645 will wait indefinitely in ptlrpc_queue_wait->l_wait_event() after
4646 ptlrpc_import_delayed_req() because we didn't check if the
4647 request was interrupted, and we also didn't break out of the
4648 event loop if there was no timeout
4653 Description: data loss during non-page-aligned writes to a single file from
4654 both multiple nodes and multiple threads on one node at same time
4655 Details : updates to KMS and lsm weren't protected by common lock. Resulting
4656 inconsistency led to false short-reads, that were cached and later
4657 used by ->prepare_write() to fill in partially written page,
4658 leading to data loss.
4661 Frequency : always, if lconf --abort_recovery used
4663 Description: lconf --abort_recovery fails with 'Operation not supported'
4664 Details : lconf was attempting to abort recovery on the MDT device and not
4667 Severity : enhancement
4669 Description: remove cleanup logs
4670 Details : replace lconf-generated cleanup logs with lustre internal
4671 cleanup routines. Eliminates the need for client-cleanup and
4674 Severity : enhancement
4676 Description: add support for EAs (user and system) on lustre filesystems
4677 Details : it is now possible to store extended attributes in the Lustre
4678 client filesystem, and with the user_xattr mount option it
4679 is possible to allow users to store EAs on their files also
4681 Severity : enhancement
4683 Description: Add possibility (config option) to show minimal available OST free
4685 Details : When compiled with --enable-mindf configure option, statfs(2)
4686 (and so, df) will return least minimal free space available from
4687 all OSTs as amount of free space on FS, instead of summary of
4688 free spaces of all OSTs.
4690 Severity : enhancement
4692 Description: do not expand extent locks acquired on OST-side
4693 Details : Modify ldlm_extent_policy() to not expand local locks, acquired
4694 by server: they are not cached anyway.
4697 Frequency : when mmap is used/binaries executed from Lustre
4699 Description: Unmmap pages before throwing them away from read cache.
4700 Details : llap_shrink cache now attempts to unmap pages before discarding
4701 them (if unmapping failed - do not discard). SLES9 kernel has
4702 extra checks that trigger if this unmapping is not done first.
4707 Description: lconf didn't resolve symlinks before checking to see whether a
4708 given mountpoint was already in use
4711 Frequency : when migrating failover services
4712 Bugzilla : 6395, 9514
4713 Description: When migrating a subset of services from a node (e.g. failback
4714 from a failover service node) the remaining services would
4715 time out and evict clients.
4716 Details : lconf --force (implied by --failover) sets the global obd_timeout
4717 to 5 seconds in order to quickly disconnect, but this caused
4718 other RPCs to time out too quickly. Do not change the global
4719 obd_timeout for force cleanup, only set it for DISCONNECT RPCs.
4721 Severity : enhancement
4722 Frequency : if MDS is started with down OST
4723 Bugzilla : 9439,5706
4724 Description: Allow startup/shutdown of an MDS without depending on the
4725 availability of the OSTs.
4726 Details : Asynchronously call mds_lov_synchronize during MDS startup.
4727 Add appropriate locking and lov-osc refcounts for safe
4728 cleaning. Add osc abort_inflight calls in case the
4729 synchronize never started.
4732 Frequency : occasional (Cray XT3 only)
4734 Description: root not authorized to access files in CRAY_PORTALS environment
4735 Details : The client process capabilities were not honoured on the MDS in
4736 a CRAY_PORTALS/CRAY_XT3 environment. If the file had previously
4737 been accessed by an authorized user then root was able to access
4738 the file on the local client also. The root user capabilities
4739 are now allowed on the MDS, as this environment has secure UID.
4742 Frequency : occasional
4744 Description: ldiskfs "too long searching" message happens too often
4745 Details : A debugging message (otherwise harmless) prints too often on
4746 the OST console. This has been reduced to only happen when
4747 there are fragmentation problems on the filesystem.
4752 Description: Division by zero in statfs when all OSCs are inactive
4753 Details : lov_get_stripecnt() returns zero due to incorrect order of checks,
4754 lov_statfs divides by value returned by lov_get_stripecnt().
4758 Bugzilla : 9489, 3273
4759 Description: First write from each client to each OST was only 4kB in size,
4760 to initialize client writeback cache, which caused sub-optimal
4761 RPCs and poor layout on disk for the first writen file.
4762 Details : Clients now request an initial cache grant at (re)connect time
4763 and so that they can start streaming writes to the cache right
4764 away and always do full-sized RPCs if there is enough data.
4765 If the OST is rebooted the client also re-establishes its grant
4766 so that client cached writes will be honoured under the grant.
4771 Description: Slow ls (and stat(2) syscall) on files residing on IO-loaded OSTs
4772 Details : Now I/O RPCs go to different portal number and (presumably) fast
4773 lock requests (and glimses) and other RPCs get their own service
4774 threads pool that should be able to service those RPCs
4777 Severity : enhancement
4779 Description: Ability to exchange lustre version between client and servers and
4780 issue warnings at client side if client is too old. Also for
4781 liblustre clients there is ability to refuse connection of too old
4783 Details : New 'version' field is added to connect data structure that is
4784 filled with version info. That info is later checked by server and
4788 Frequency : rare, liblustre only.
4789 Bugzilla : 9296, 9581
4790 Description: Two simultaneous writes from liblustre at offset within same page
4791 might proceed at the same time overwriting eachother with stale
4793 Details : I/O lock withing llu_file_prwv was released too early, before data
4794 actually was hitting the wire. Extended lock-holding time until
4795 server acknowledges receiving data.
4798 Frequency : extremely rare. Never observed in practice.
4800 Description: avoid generating lustre_handle cookie of 0.
4801 Details : class_handle_hash() generates handle cookies by incrementing
4802 global counter, and can hit 0 occasionaly (this is unlikely, but
4803 not impossible, because initial value of cookie counter is
4804 selected randonly). Value of 0 is used as a sentinel meaning
4805 "unassigned handle" --- avoid it. Also coalesce two critical
4806 sections in this function into one.
4808 Severity : enhancement
4810 Description: allow liblustre clients to delegate truncate locking to OST
4811 Details : To avoid overhead of locking, liblustre client instructs OST to
4812 take extent lock in ost_punch() on client's behalf. New connection
4813 flag is added to handle backward compatibility.
4815 Severity : enhancement
4816 Bugzilla : 4928, 7341, 9758
4817 Description: allow number of OST service threads to be specified
4818 Details : a module parameter allows the number of OST service threads
4819 to be specified via "options ost ost_num_threads={N}" in the
4820 OSS's /etc/modules.conf or /etc/modprobe.conf.
4824 Bugzilla : 6146, 9635, 9895
4825 Description: servers crash with bad pointer in target_handle_connect()
4826 Details : In rare cases when a client is reconnecting it was possible that
4827 the connection request was the last reference for that export.
4828 We would temporarily drop the export reference and get a new
4829 one, but this may have been the last reference and the export
4830 was just destroyed. Get new reference before dropping old one.
4832 Severity : enhancement
4833 Frequency : if client is started with failover MDS
4835 Description: Allow multiple MDS hostnames in the mount command
4836 Details : Try to read the configuration from all specified MDS
4837 hostnames during a client mount in case the "primary"
4840 Severity : enhancement
4842 Description: Stop sending data to evicted clients as soon as possible.
4843 Details : Check if the client we are about to send or are sending data to
4844 was evicted already. (Check is done every second of waiting,
4845 for which l_wait_event interface was extended to allow checking
4846 of exit condition at specified intervals).
4849 Frequency : rare, normally only when NFS exporting is done from client
4851 Description: 'bad disk LOV MAGIC: 0x00000000' error when chown'ing files
4853 Details : Make mds_get_md() recognise empty md case and set lmm size to 0.
4856 Frequency : always, if srand() is called before liblustre initialization
4858 Description: Liblustre uses system PRNG disturbing its usage by user application
4859 Details : Introduce internal to lustre fast and high-quality PRNG for
4860 lustre usage and make liblustre and some other places in generic
4861 lustre code to use it.
4863 Severity : enhancement
4864 Bugzilla : 9477, 9557, 9870
4865 Description: Verify that the MDS configuration logs are updated when xml is
4866 Details : Check if the .xml configuration logs are newer than the config
4867 logs stored on the MDS and report an error if this is the case.
4868 Request --write-conf, or allow starting with --old_conf.
4870 Severity : enhancement
4872 Description: Handle symlinks in the path when checking if Lustre is mounted.
4873 Details : Resolve intermediate symlinks when checking if a client has
4874 mounted a filesystem to avoid duplicate client mounts.
4879 Description: lconf can hit an error exception but still return success.
4880 Details : The lconf command catches the Command error exception at the top
4881 level script context and will exit with the associated exit
4882 status, but doesn't ensure that this exit status is non-zero.
4887 Description: failure of ptlrpc thread startup can cause oops
4888 Details : Starting a ptlrpc service thread can fail if there are a large
4889 number of threads or the server memory is very fragmented.
4890 Handle this without oopsing.
4893 Frequency : always, only if liblustre and non-default acceptor port was used
4895 Description: liblustre cannot connect to servers with non-default acceptor port
4896 Details : tcpnal_set_default_params() was not called and was therefore
4897 ignoring the environment varaible TCPNAL_PORT, as well as other
4898 TCPNAL_ environment variables
4903 Description: two objects could be created on the same OST for a single file
4904 Details : If an OST is down, in some cases it was possible to create two
4905 objects on a single OST for a single file. No problems other
4906 than potential performance impact and spurious error messages.
4910 Bugzilla : 5681, 9562
4911 Description: Client may oops in ll_unhash_aliases
4912 Details : Client dcache may become inconsistent in race condition.
4913 In some cases "getcwd" can fail if the current directory is
4919 Description: Inode refcounting problems in NFS export code
4920 Details : link_raw functions used to call d_instantiate without obtaining
4921 extra inode reference first.
4925 Bugzilla : 9942, 9903
4926 Description: Referencing freed requests leading to crash, memleaks with NFS.
4927 Details : We used to require that call to ll_revalidate_it was always
4928 followed by ll_lookup_it. Also with revalidate_special() it is
4929 possible to call ll_revalidate_it() twice for the same dentry
4930 even if first occurence returned success. This fix changes semantic
4931 between DISP_ENQ_COMPLETE disposition flag to mean there is extra
4932 reference on a request referred from the intent.
4933 ll_intent_release() then releases such a request.
4936 Frequency : rare, normally benchmark loads only
4938 Description: unlinked inodes were kept in memory on the client
4939 Details : If a client is repeatedly creating and unlinking files it
4940 can accumulate a lot of stale inodes in the inode slab cache.
4941 If there is no other client load running this can cause the
4942 client node to run out of memory. Instead flush old inodes
4943 from client cache that have the same inode number as a new inode.
4946 Frequency : SLES9 2.6.5 kernel and long filenames only
4947 Bugzilla : 9969, 10379
4948 Description: utime reports stale NFS file handle
4949 Details : SLES9 uses out-of-dentry names in some cases, which confused
4950 the lustre dentry revalidation. Change it to always use the
4954 Frequency : rare, unless heavy write-truncate concurrency is continuous
4955 Bugzilla : 4180, 6984, 7171, 9963, 9331
4956 Description: OST becomes very slow and/or deadlocked during object unlink
4957 Details : filter_destroy() was holding onto the parent directory lock
4958 while truncating+unlinking objects. For very large objects this
4959 may block other threads for a long time and slow overall OST
4960 responsiveness. It may also be possible to get a lock ordering
4961 deadlock in this case, or run out of journal credits because of
4962 the combined truncate+unlink. Solution is to do object truncate
4963 first in one transaction without parent lock, and then do the
4964 final unlink in a new transaction with the parent lock. This
4965 reduces the lock hold time dramatically.
4968 Frequency : rare, 2.4 kernels only
4970 Description: MDS or OST cleanup may trip kernel BUG when dropping kernel lock
4971 Details : mds_cleanup() and filter_cleanup() need to drop the kernel lock
4972 before unmounting their filesystem in order to avoid deadlock.
4973 The kernel_locked() function in 2.4 kernels only checks whether
4974 the kernel lock is held, not whether it is this process that is
4975 holding it as 2.6 kernels do.
4980 Description: MDS or OST may oops/LBUG if a client is connecting multiple times
4981 Details : The client ptlrpc code may be trying to reconnect to a down
4982 server before a previous connection attempt has timed out.
4983 Increase the reconnect interval to be longer than the connection
4984 timeout interval to avoid sending duplicate connections to
4988 Frequency : echo_client brw_test command
4990 Description: fix echo_client to work with OST preallocated code
4991 Details : OST preallocation code (5137) didn't take echo_client IO path
4992 into account: echo_client calls filter methods outside of any
4993 OST thread and, hence, there is no per-thread preallocated
4994 pages and buffers to use. Solution: hijack pga pages for IO. As
4995 a byproduct, this avoids unnecessary data copying.
4999 Bugzilla : 3555, 5962, 6025, 6155, 6296, 9574
5000 Description: Client can oops in mdc_commit_close() after open replay
5001 Details : It was possible for the MDS to return an open request with no
5002 transaction number in mds_finish_transno() if the client was
5003 evicted, but without actually returning an error. Clients
5004 would later try to replay that open and may trip an assertion
5005 Simplify the client close codepath, and always return an error
5006 from the MDS in case the open is not successful.
5009 Frequency : rare, 2.6 OSTs only
5011 Description: OST may deadlock under high load on fragmented files
5012 Details : If there was a heavy load and highly-fragmented OST filesystems
5013 it was possible to have all the OST threads deadlock waiting on
5014 allocation of biovecs, because the biovecs were not released
5015 until the entire RPC IO was completed. Instead, release biovecs
5016 as soon as they are complete to ensure forward IO progress.
5018 Severity : enhancement
5020 Description: Support for specifying external journal device at mount
5021 Details : If an OST or MDS device is formatted with an external journal
5022 device, this device major/minor is stored in the ext3 superblock
5023 and may not be valid for failover. Allow detecting and
5024 specifying the external journal at mount time.
5029 Description: Mounting an MDS with pending unlinked files may cause oops
5030 Details : target_finish_recovery() calls mds_postrecov() which returned
5031 the number of orphans unlinked. mds_lov_connect->mds_postsetup()
5032 considers this an error and immediately begins cleaning up the
5033 lov, just after starting the mds_lov process
5035 Severity : enhancement
5037 Description: Implement 'lfs df' to report actual free space on per-OST basis
5038 Details : Add sub-command 'df' on 'lfs' to report the disk space usage of
5039 MDS/OSDs. Usage: lfs df [-i][-h]. Command Options: '-i' to report
5040 usage of objects; '-h' to report in human readable format.
5042 ------------------------------------------------------------------------------
5044 2005-08-26 Cluster File Systems, Inc. <info@clusterfs.com>
5051 Description: Mounting an ldiskfs file system with mballoc may crash OST node.
5052 Details : ldiskfs mballoc code may reference an uninitialized buddy struct
5053 at startup during orphan unlinking. Instead, skip buddy update
5054 before setup, as it will be regenerated after recovery is complete.
5059 Description: If an OST is inactive, its locks might reference stale inodes.
5060 Details : lov_change_cbdata() must iterate over all namespaces, even if
5061 they are inactive to clear inode references from the lock.
5063 Severity : enhancement
5064 Frequency : occasional, if non-standard max_dirty_mb used
5066 Description: Client will block write RPCs if not enough grant
5067 Details : If a client has max_dirty_mb smaller than max_rpcs_in_flight,
5068 then the client will block writes while waiting for another RPC
5069 to complete instead of consuming its dirty limit. With change
5070 we get improved performance when max_dirty_mb is small.
5072 Severity : enhancement
5073 Bugzilla : 3389, 6253
5074 Description: Add support for supplementary groups on the MDS.
5075 Details : The MDS has an upcall /proc/fs/lustre/mds/{mds}/group_upcall
5076 (set to /usr/sbin/l_getgroups if enabled) which will do MDS-side
5077 lookups for user supplementary groups into a cache.
5081 Description: O_CREAT|O_EXCL open flags in liblustre always return -EEXIST
5082 Details : Make libsysio to not enforce O_EXCL by clearing the flag,
5083 for liblustre O_EXCL is enforced by MDS.
5087 Description: readdir never returns NULL in liblustre.
5088 Details : Corrected llu_iop_getdirentries logic, to return offset of next
5089 dentry in struct dirent.
5093 Frequency : liblustre only, depends on application IO pattern
5094 Description: liblustre clients evicted if not contacting servers
5095 Details : Don't put liblustre clients into the ping_evictor list, so
5096 they will not be evicted by the pinger ever.
5098 Severity : enhancement
5100 Description: Add ability to evict clients by NID from MDS.
5101 Details : By echoing "nid:$NID" string into
5102 /proc/fs/lustre/mds/.../evict_client client with nid that equals to
5103 $NID would be instantly evicted from this MDS and from all active
5104 OSTs connected to it.
5108 Description: Do not query file size twice, somewhat slowing stat(2) calls.
5109 Details : lookup_it_finish() used to query file size from OSTs that was not
5114 Description: service threads change working directory to that of init
5115 Details : Starting lustre service threads may pin the working directory
5116 of the parent thread, making that filesystem busy. Threads
5117 now change to the working directory of init to avoid this.
5121 Frequency : during shutdown only
5122 Description: shutdown with a failed MDS or OST can cause unmount to hang
5123 Details : Don't resend DISCONNECT messages in ptlrpc_disconnect_import()
5128 Frequency : 2.6 only
5129 Description: chmod/chown may include an extra supplementary group
5130 Details : ll{,u}_mdc_pack_op_data() does not properly initialize the
5131 supplementary group and if none is specified this is used.
5134 Bugzilla : 5479 (6816)
5136 Description: Racing open + rm can assert client in mdc_set_open_replay_data()
5137 Details : If lookup is in progress on a file that is unlinked we might try
5138 to revalidate the inode and fail in revalidate after lookup is
5139 complete and ll_file_open() enqueues the open again but
5140 it_open_error() was not checking DISP_OPEN_OPEN errors correctly.
5143 Frequency : always, if lconf --abort_recovery used
5145 Description: lconf --abort_recovery fails with 'Operation not supported'
5146 Details : lconf was attempting to abort recovery on the MDT device and not
5149 ------------------------------------------------------------------------------
5151 2005-08-08 Cluster File Systems, Inc. <info@clusterfs.com>
5156 Frequency : rare (only unsupported configurations with a node running as an
5158 Bugzilla : 6514, 5137
5159 Description: Mounting a Lustre file system on a node running as an OST could
5161 Details : OSTs now preallocates memory needed to write out data at
5162 startup, instead of when needed, to avoid having to
5163 allocate memory in possibly low memory situations.
5164 Specifically, if the file system is mounted on on OST,
5165 memory pressure could force it to try to write out data,
5166 which it needed to allocate memory to do. Due to the low
5167 memory, it would be unable to do so and the node would
5168 become unresponsive.
5170 Severity : enhancement
5172 Description: Addition of lconf --service command line option
5173 Details : lconf now accepts a '--service <arg>' option, which is
5174 shorthand for 'lconf --group <arg> --select <arg>=<hostname>'
5176 Severity : enhancement
5178 Description: Failover mode is now the default for OSTs.
5179 Details : By default, OSTs will now run in failover mode. To return to
5180 the old behaviour, add '--failout' to the lmc line for OSTs.
5182 Severity : enhancement
5184 Description: Health checks are now provided for MDS and OSTs
5185 Details : Additional detailed health check information on MSD and OSTs
5186 is now provided through the procfs health_check value.
5189 Frequency : occasional, depends on IO load
5191 Description: Disk fragmentation on the OSTs could eventually cause slowdowns
5192 after numerous create/delete cycles
5193 Details : The ext3 inode allocation policy would not allocate new inodes
5194 very well on the OSTs because there are no new directories
5195 being created. Instead we look for groups with free space if
5196 the parent directories are nearly full.
5201 Description: Network or server problems during mount may cause partially
5202 mounted clients instead of returning an error.
5203 Details : The config llog parsing code may overwrite the error return
5204 code during mount error handling, returning success instead
5210 Description: MDS can fail to allocate large reply buffers
5211 Details : After long uptimes the MDS can fail to allocate large reply
5212 buffers (e.g. zconf client mount config records) due to memory
5213 fragmentation or consumption by the buffer cache. Preallocate
5214 some large reply buffers so that these replies can be sent even
5215 under memory pressure.
5219 Frequency : rare (liblustre)
5220 Description: fsx running with liblustre complained that using truncate() to
5221 extend the file doesn't work. This patch corrects that issue.
5222 Details : This is the liblustre equivalent of the fix for bug 6196. Fixes
5223 ATTR_SIZE and lsm use in llu_setattr_raw.
5227 Frequency : rare, only 2.6 kernels
5228 Description: Unusual file access patterns on the MDS may result in inode
5229 data being lost in very rare circumstances.
5230 Details : Bad interaction between the ea-in-inode patch and the "no-read"
5231 code in the 2.6 kernel caused the inode and/or EA data not to
5232 be read from disk, causing single-file corruption.
5236 Frequency : rare, only 2.6 filesystems using extents
5237 Description: Heavy concurrent write and delete load may cause data corruption.
5238 Details : It was possible under high-load situations to have an extent
5239 metadata block in the block device cache from a just-unlinked
5240 file overwrite a newly-allocated data block. We now unmap any
5241 metadata buffers that alias just-allocated data blocks.
5245 Frequency : filesystems with default stripe_count larger than 77
5246 Description: lconf+mke2fs fail when formatting filesystem with > 77 stripes
5247 Details : lconf specifies an inode size of 4096 bytes when the default
5248 stripe_count is larger than 77. This conflicts with the default
5249 inode density of 1 per 4096 bytes. Allocate smaller inodes in
5250 this case to avoid pinning too much memory for large EAs.
5252 ------------------------------------------------------------------------------
5254 2005-07-07 Cluster File Systems, Inc. <info@clusterfs.com>
5259 Frequency : rare (extremely heavy IO load with hundreds of clients)
5261 Description: Client is evicted, gets IO error writing to file
5262 Details : lock ordering changes for bug 5492 reintroduced bug 3267 and
5263 caused clients to be evicted for AST timeouts. The fixes in
5264 bug 5192 mean we no longer need to have such short AST timeouts
5265 so ldlm_timeout has been increased.
5268 Frequency : occasional during --force or --failover shutdown under load
5269 Bugzilla : 5949, 4834
5270 Description: Server oops/LBUG if stopped with --force or --failover under load
5271 Details : a collection of import/export refcount and cleanup ordering
5272 issues fixed for safer force cleanup
5275 Frequency : only filesystems larger than 120 OSTs
5276 Bugzilla : 5990, 6223
5277 Description: lfs getstripe would oops on a very large filesystem
5278 Details : lov_getconfig used kfree on vmalloc'd memory
5281 Frequency : only filesystems exporting via NFS to Solaris 10 clients
5282 Bugzilla : 6242, 6243
5283 Description: reading from files that had been truncated to a non-zero size
5284 but never opened returned no data
5285 Details : ll_file_read() reads zeros from no-object files to EOF
5290 Description: A bug in MDS/OSS recovery could cause the OSS to fail an assertion
5291 Details : There's little harm in aborting MDS/OSS recovery and letting it
5292 try again, so I removed the LASSERT and return an error instead.
5294 Severity : enhancement
5296 Description: New debugging infrastructure for tracking down data corruption
5297 Details : The I/O checksum code was replaced to: (a) control it at runtime,
5298 (b) cover more of the client-side code path, and (c) try to narrow
5299 down where problems occurred
5303 Bugzilla : 3819, 4364, 4397, 6313
5304 Description: Racing close and eviction MDS could cause assertion in mds_close
5305 Details : It was possible to get multiple mfd references during close and
5306 client eviction, leading to one thread referencing a freed mfd.
5308 Severity: : enhancement
5309 Bugzilla : 3262, 6359
5310 Description: Attempts to reconnect to servers are now more aggressive.
5311 Details : This builds on the enhanced upcall-less recovery that was added
5312 in 1.4.2. When trying to reconnect to servers, clients will
5313 now try each server in the failover group every 10 seconds. By
5314 default, clients would previously try one server every 25 seconds.
5319 Description: After recovery, certain operations trigger a failed
5320 assertion on a client.
5321 Details : Failing over an mds, using lconf -d --failover, while a
5322 client was doing a readdir() call would cause the client to
5323 LBUG after recovery completed and the readdir() was resent.
5325 Severity : enhancement
5327 Description: Default groups are now added by lconf
5328 Details : You can now run lconf --group <servicename> without having to
5329 manually add groups with lmc.
5332 Frequency : occasional
5334 Description: Nodes with an elan id of 0 trigger a failed assertion
5337 Frequency : always when accessing e.g. tty/console device nodes
5339 Description: tty and some other devices nodes cannot be used on lustre
5340 Details : file's private_data field is used by device data and lustre
5341 values in there got lost. New field was added to struct file to
5342 store fs-specific private data.
5345 Frequency : when exporting Lustre via NFS
5347 Description: NFSD failed occasionally when looking up a path component
5348 Details : NFSD is looking up ".." which was broken in ext3 directories
5349 that had grown large enough to become hashed.
5352 Frequency : Clusters with multiple interfaces not on the same subnet
5354 Description: Nodes will repeatedly try to reconnect to an interface which it
5355 cannot reach and report an error to the log.
5356 Details : Extra peer list entries will be created by lconf with some peers
5357 unreachable. lconf now validates the peer before adding it.
5360 Frequency : Only if a default stripe is set on the filesystem root.
5362 Description: Setting a default stripe on the filesystem root prevented the
5363 filesystem from being remounted.
5364 Details : The client was sending extra request flags in the root getattr
5365 request and did not allocate a reply buffer for the dir EA.
5368 Frequency : occasional, higher if lots of files are accessed by one client
5369 Bugzilla : 6159, 6097
5370 Description: Client trips assertion regarding lsm mismatch/magic
5371 Details : While revalidating inodes the VFS looks up inodes with ifind()
5372 and in rare cases can find an inode that is being freed.
5373 The ll_test_inode() code will free the lsm during ifind()
5374 when it finds an existing inode and then the VFS later attaches
5375 this free lsm to a new inode.
5379 Bugzilla : 6422, 7030
5380 Description: MDS deadlock between mkdir and client eviction
5381 Details : Creating a new file via mkdir or mknod (starting a transaction
5382 and getting the ns lock) can deadlock with client eviction
5383 (gets ns lock and trying to finish a synchronous transaction).
5386 Frequency : occasional
5387 Description: While starting a server, the fsfilt_ext3 module could not be
5389 Details : CFS's improved ext3 filesystem is named ldiskfs for 2.6
5390 kernels. Previously, lconf would still use the ext3 name
5391 when trying to load modules. Now, it will correctly use
5392 ext3 on 2.4 and ldiskfs on 2.6.
5394 Severity : enhancement
5395 Description: The default stripe count has been changed to 1
5396 Details : The interpretation of the default stripe count (0, to lfs
5397 or lmc) has been changed to mean striping across a single
5398 OST, rather than all available. For general usage we have
5399 found a stripe count of 1 or 2 works best.
5401 Severity : enhancement
5402 Description: Add support for compiling against Cray portals.
5403 Details : Conditional compiling for some areas that are different
5407 Frequency : occasional
5408 Bugzilla : 6409, 6834
5409 Description: Creating files with an explicit stripe count may lead to
5410 a failed assertion on the MDS
5411 Details : If some OSTs are full or unavailable, creating files may
5412 trigger a failed assertion on the MDS. Now, Lustre will
5413 try to use other servers or return an error to the
5417 Frequency : occasional
5419 Description: Multiple concurrent overlapping read+write on multiple SMP nodes
5420 caused lock timeout during readahead (since 1.4.2).
5421 Details : Processes doing readahead might match a lock that hasn't been
5422 granted yet if there are overlapping and conflicting lock
5423 requests. The readahead process waits on ungranted lock
5424 (original lock is CBPENDING), while OST waits for that process
5425 to cancel CBPENDING read lock and eventually evicts client.
5427 Severity : enhancement
5429 Description: Initial enabling of flock support for clients
5430 Details : Implements fcntl advisory locking and file status functions.
5431 This feature is provided as an optional mount flag (default
5432 off), and is NOT CURRENTLY SUPPORTED. Not all types of record
5433 locking are implemented yet, and those that are are not guaranteed
5434 to be completely correct in production environments.
5435 mount -t lustre -o [flock|noflock] ...
5438 Frequency : occasional
5440 Description: OSTs running 2.4 kernels but with extents enabled might trip an
5441 assertion in the ext3 JBD (journaling) layer.
5442 Details : The b_committed_data struct is protected by the big kernel lock
5443 in 2.4 kernels, serializing journal_commit_transaction() and
5444 ext3_get_block_handle->ext3_new_block->find_next_usable_block()
5445 access to this struct. In 2.6 kernels there is finer grained
5446 locking to improve SMP performance of the JBD layer.
5450 Description: Changes the "SCSI I/O Stats" kernel patch to default to "enabled"
5452 -----------------------------------------------------------------------------
5454 2005-05-05 Cluster File Systems, Inc. <info@clusterfs.com>
5456 NOTE: Lustre 1.4.2 uses an incompatible network protocol than previous
5457 versions of Lustre. Please update all servers and clients to
5458 version 1.4.2 or later at the same time. You must also run
5459 "lconf --write-conf {config}.xml" on the MDS while it is stopped
5460 to update the configuration logs.
5462 - fix for HPUX NFS client breakage when NFS exporting Lustre (5781)
5463 - mdc_enqueue does not need max_mds_easize request buffer on send (5707)
5464 - swab llog records of type '0' so we get proper header size/idx (5861)
5465 - send llog cancel req to DLM cancel portal instead of cb portal (5515)
5466 - fix rename of one directory over another leaking an inode (5953)
5467 - avoid SetPageDirty on 2.6 (5981)
5468 - don't re-add just-being-destroyed locks to the waiting list (5653)
5469 - when creating new directories, inherit the parent's custom
5470 striping settings if present parent (3048)
5471 - flush buffers from cache before direct IO in 2.6 obdfilter (4982)
5472 - don't hold i_size_sem in ll_nopage() and ll_ap_refresh_count (6077)
5473 - don't hold client locks on temporary worklist from l_lru (5666)
5474 - handle IO errors in 2.6 obdfilter bio completion routine (6046)
5475 - automatically evict dead clients (5921)
5476 - Update file size properly in create+truncate+fstat case (6196)
5477 - Do not unhash mountpoint dentries, do not allow removal of
5479 - Avoid lock ordering deadlock issue with write/truncate (6203,5654)
5480 - reserve enough journal credits in fsfilt_start_log for setattr (4554)
5481 - ldlm_enqueue freed-export error path would always LBUG (6149,6184)
5482 - don't reference lr_lvb_data until after we hold lr_lvb_sem (6170)
5483 - don't overwrite last_rcvd if there is a *_client_add() error (6086)
5484 - Correctly handle reads of files with no objects (6243)
5485 - lctl recover will also mark a device active if deactivate used (5933)
5487 - by default create 1 inode per 4kB space on MDS, per 16kB on OSTs
5488 - allow --write-conf on an MDS with different nettype than client (5619)
5489 - don't write config llogs to MDS for mounts not from that MDS (5617)
5490 - lconf should create multiple TCP connections from a client (5201)
5491 - init scripts are now turned off by default; run chkconfig --on
5492 lustre and chkconfig --on lustrefs to use them
5493 - upcalls are no longer needed for clients to recover to failover
5495 - add --abort-recovery option to lconf to abort recovery on device
5497 - add support for an arbitrary number of OSTs (3026)
5498 - Quota support protocol changes.
5499 - forward compatibility changes to wire structs (6007)
5500 - rmmod NALs that might be loaded because of /etc/modules.conf (6133)
5501 - support for mountfsoptions and clientoptions to the Lustre LDAP (5873)
5502 - improved "lustre status" script
5503 - initialize blocksize for non-regular files (6062)
5504 - added --disable-server and --disable-client configure options (5782)
5505 - introduce a lookup cache for lconf to avoid repeated DB scans (6204)
5506 - Vanilla 2.4.29 support
5507 - increase maximum number of obd devices to 520 (6242)
5508 - remove the tcp-zero-copy patch from the suse-2.4 series (5902)
5509 - Quadrics Elan drivers are now included for the RHEL 3 2.4.21 and
5510 SLES 9 2.6.5 kernels
5511 - limit stripes per file to 160 (the maximum EA size) (6093)
5513 2005-03-22 Cluster File Systems, Inc. <info@clusterfs.com>
5516 - don't LASSERT in ll_release on NULL lld with NFS export (4655, 5760)
5517 - hold NS lock when calling handle_ast_error->del_waiting_lock (5746)
5518 - fix setattr mtime regression from lovcleanup merge (4829, 5669)
5519 - workaround for 2.6 crash in ll_unhash_aliases (5687, 5210)
5520 - small ext3 extents cleanups and fixes (5733)
5521 - improved mballoc code, several small races and bugs fixed (5733, 5638)
5522 - kernel version 43 - fix remove_suid bugs in both 2.4 and 2.6 (5695)
5523 - avoid needless client->OST connect, fix handle mismatch (5317)
5524 - fix DLM error path that led to out-of-sync client, long delays (5779)
5525 - support common vfs-enforced mount options (nodev,nosuid,noexec) (5637)
5526 - fix several locking issues related to i_size (5492,5624,5654,5672)
5527 - don't move pending lock onto export if it is already evicted (5683)
5528 - fix kernel oops when creating .foo in unlinked directory (5548)
5529 - fix deadlock in obdfilter statistics vs. object create (5811)
5530 - use time_{before,after} to avoid timer jiffies wrap (5882)
5531 - shutdown --force/--failover stability (3607,3651,4797,5203,4834)
5532 - Do not leak request if server was not able to process it (5154)
5533 - If mds_open unable to find parent dir, make that negative lookup(5154)
5534 - don't create new directories with extent-mapping (5909, 5936)
5536 - fix lustre/lustrefs init scripts for SuSE (patch from Scali, 5702)
5537 - don't hold the pinger_sem in ptlrpc_pinger_sending_on_import
5538 - change obd_increase_kms to obd_adjust_kms (up or down) (5654)
5539 - lconf, lmc search both /usr/lib and /usr/lib64 for Python libs (5800)
5540 - support for RHEL4 kernel on i686 (5773)
5541 - provide error messages when incompatible logs are encountered (5898)
5543 2005-02-18 Cluster File Systems, Inc. <info@clusterfs.com>
5544 * version 1.4.0.10 (1.4.1 release candidate 1)
5546 - don't keep a lock reference when lock is not granted (4238)
5547 - unsafe list practices (rarely) led to infinite eviction loop (4908)
5548 - add per-fs limit of Lustre pages in page cache, avoid OOM (4699)
5549 - drop import inflight refcount on signal_completed_replay error (5255)
5550 - unlock page after async write error during send (3677)
5551 - handle missing objects in filter_preprw_read properly (5265)
5552 - no transno return for symlink open, don't save no-trasno open (3440)
5553 - don't try to complete elan receive that already failed (4012)
5554 - free RPC server reply state on error (5406)
5555 - clean up thread from ptlrpc_start_thread() on error (5160)
5556 - readahead could read extra page into cache that wasn't ejected (5388)
5557 - prevent races in class_attach/setup/cleanup/detach (5260)
5558 - don't dereference de->d_inode after l_dput of de (5458)
5559 - use "int" for stripe value returned from lock_to_stripe (5544)
5560 - mballoc allocation and error-checking fixes in 2.6 (5504)
5561 - block device patches to fix I/O request sizes in 2.6 (5482)
5562 - look up hostnames for IB nals (5602)
5563 - 2.6 changed lock ordering of 2 semaphores, caused deadlock (5654)
5564 - don't start multiple acceptors for the same port (5277)
5565 - fix incorrect LASSERT in mds_getattr_name (5635)
5566 - export a proc file for general "ping" checking (5628)
5567 - fix "lfs check" to not block when the MDS is down (5628)
5569 - service request history (4965)
5570 - put {ll,lov,osc}_async_page structs in a single slab (4699)
5571 - create an "evict_client" /proc entry on OSTs, like the MDS has
5572 - fix mount usage message, return errors per mount(8) (5168)
5573 - change grep [] to grep "[]" in tests so they work in more UMLs
5574 - fix ppc64/x86_64 spec to use %{_libdir} instead of /usr/lib (5389)
5575 - remove ancient LOV_MAGIC_V0 EA support (5047)
5576 - add "disk I/Os in flight" and "I/O req time" stats in obdfilter
5577 - align r/w RPCs to PTLRPC_MAX_BRW_SIZE boundary for performance (3451)
5578 - allow readahead allocations to fail when low on memory (5383)
5579 - mmap locking landed again, after considerable improvement (2828)
5580 - add get_hostaddr() to lustreDB.py for LDAP support (5459)
5582 2004-11-23 Cluster File Systems, Inc. <info@clusterfs.com>
5585 - send OST transaction number in read/write reply to free req (4966)
5586 - don't ASSERT in ptl_send_rpc() if we run out of memory (5119)
5587 - lock /proc/sys/portals/routes internal state, avoiding oops (4827)
5588 - the watchdog thread now runs as interruptible (5246)
5589 - flock/lockf fixes (but it's still disabled, pending 5135)
5590 - don't use EXT3 constants in llite code (5094)
5591 - memory shortage at startup could cause assertion (5176)
5593 - reorganization of lov code
5594 - single portals codebase
5596 - add extents/mballoc support (5025)
5597 - direct I/O reads in the obdfilter (4048)
5598 - kernel patches from LNXI for 2.6 (bluesmoke, perfctr, mtd, kexec)
5600 tbd Cluster File Systems, Inc. <info@clusterfs.com>
5603 - send OST transaction number in read/write reply to free req (4966)
5604 - don't ASSERT in ptl_send_rpc() if we run out of memory (5119)
5605 - lock /proc/sys/portals/routes internal state, avoiding oops (4827)
5606 - the watchdog thread now runs as interruptible (5246)
5607 - handle missing objects in filter_preprw_read properly (5265)
5608 - unsafe list practices (rarely) led to infinite eviction loop (4908)
5609 - drop import inflight refcount on signal_completed_replay error (5255)
5610 - unlock page after async write error during send (3677)
5611 - return original error code on reconstructed replies (3761)
5612 - no transno return for symlink open, don't save no-trasno open (3440)
5614 - add pid to ldlm debugging output (4922)
5615 - bump the watchdog timeouts -- we can't handle 30sec yet
5616 - extra debugging for orphan dentry/inode bug (5259)
5618 2004-11-16 Cluster File Systems, Inc. <info@clusterfs.com>
5621 - fix TCP_NODELAY bug, which caused extreme perf regression (5134)
5622 - allocate qswnal tx descriptors singly to avoid fragmentation (4504)
5623 - don't LBUG on obdo_alloc() failure, use OBD_SLAB_ALLOC() (4800)
5624 - fix NULL dereference in /proc/sys/portals/routes (4827)
5625 - allow failed mdc_close() operations to be interrupted (4561)
5626 - stop precreate on OST before MDS would time out on it (4778)
5627 - don't send partial-page writes before EOF from client (4410)
5628 - discard client grant for sub-page writes on large-page clients (4520)
5629 - don't free dentries not owned by NFS code, check generation (4806)
5630 - fix lsm leak if mds_create_objects() fails (4801)
5631 - limit debug_daemon file size, always print CERROR messages (4789)
5632 - use transno after validating reply (3892)
5633 - process timed out requests if import state changes (3754)
5634 - update mtime on OST during writes, return in glimpse (4829)
5635 - add mkfsoptions to LDAP (4679)
5636 - use ->max_readahead method instead of zapping global ra (5039)
5637 - don't interrupt __l_wait_event() during strace
5639 - add software watchdogs to catch hung threads quickly (4941)
5640 - make lustrefs init script start after nfs is mounted
5641 - fix CWARN/ERROR duplication (4930)
5642 - return async write errors to application if possible (2248)
5643 - add /proc/sys/portal/memused (bytes allocated by PORTALS_ALLOC)
5644 - print NAL number in %x format (4645)
5645 - update barely-supported suse-2.4.21-171 series (4842)
5646 - support for sles 9 %post scripts
5647 - support for building 2.6 kernel-source packages
5648 - support for sles km_* packages
5650 2004-10-07 Cluster File Systems, Inc. <info@clusterfs.com>
5653 - ignore -ENOENT errors in osc_destroy (3639)
5654 - notify osc create thread that OSC is being cleaned up (4600)
5655 - add nettype argument for llmount in #5d in conf-sanity.sh (3936)
5656 - reconstruct ost_handle() like mds_handle() (4657)
5657 - create a new thread to do import eviction to avoid deadlock (3969)
5658 - let lconf resolve symlinked-to devices (4629)
5659 - don't unlink "objects" from directory with default EA (4554)
5660 - hold socknal file ref over connect in case target is down (4394)
5661 - allow more than 32000 subdirectories in a single directory (3244)
5662 - fix blocks count for O_DIRECT writes (3751)
5663 - OST returns ENOSPC from object create when no space left (4539)
5664 - don't send truncate RPC if file size isn't changing (4410)
5665 - limit OSC precreate to 1/2 of value OST considers bogus (4778)
5666 - bind to privileged port in socknal and tcpnal (3689)
5668 - rate limit CERROR/CWARN console message to avoid overload (4519)
5669 - GETFILEINFO dir ioctl returns LOV EA + MDS stat in 1 call (3327)
5670 - basic mmap support (3918)
5671 - kernel patch series update from b1_4 (4711)
5673 2004-09-16 Cluster File Systems, Inc. <info@clusterfs.com>
5676 - avoid crash during MDS cleanup with OST shut down (2775)
5677 - fix loi_list_lock/oig_lock inversion on interrupted IO (4136)
5678 - don't use bad inodes on the MDS (3744)
5679 - dynamic object preallocation to improve recovery speed (4236)
5680 - don't hold spinlock over lock dumping or change debug flags (4401)
5681 - don't zero obd_dev when it is force cleaned (3651)
5682 - print grants to console if they go negative (4431)
5683 - "lctl deactivate" will stop automatic recovery attempts (3406)
5684 - look for existing locks in ldlm_handle_enqueue() (3764)
5685 - don't resolve lock handle twice in recovery avoiding race (4401)
5686 - revalidate should check working dir is a directory (4134)
5688 - don't always mark "slow" obdfilter messages as errors (4418)
5690 2004-08-24 Cluster File Systems, Inc. <info@clusterfs.com>
5693 - don't close LustreDB during write_conf until it is done (3860)
5694 - fix typo in lconf for_each_profile (3821)
5695 - allow dumping logs from multiple threads at one time (3820)
5696 - don't allow multiple threads in OSC recovery (3812)
5697 - fix debug_size parameters (3864)
5698 - fix mds_postrecov to initialize import for llog ctxt (3121)
5699 - replace config semaphore with spinlock (3306)
5700 - be sure to send a reply for a CANCEL rpc with bad export (3863)
5701 - don't allow enqueue to complete on a destroyed export (3822)
5702 - down write_lock before checking llog header bitmap (3825)
5703 - recover from lock replay timeout (3764)
5704 - up llog sem before sending rpc (3652)
5705 - reduce ns lock hold times when setting kms (3267)
5706 - change a dlm LBUG to LASSERTF, to maybe learn something (4228)
5707 - fix NULL deref and obd_dev leak on setup error (3312)
5708 - replace some LBUG about llog ops with error handling (3841)
5709 - don't match INVALID dentries from d_lookup and spin (3784)
5710 - hold dcache_lock while marking dentries INVALID and hashing (4255)
5711 - fix invalid assertion in ptlrpc_set_wait (3880)
5713 - add libwrap support for the TCP acceptor (3996)
5714 - add /proc/sys/portals/routes for non-root route listing (3994)
5715 - allow setting MDS UUID in .xml (2580)
5716 - print the stack of a process that LBUGs (4228)
5718 2004-07-14 Cluster File Systems, Inc. <info@clusterfs.com>
5721 - don't cleanup request in ll_file_open() on failed MDS open (3430)
5722 - make sure to unset replay flag from failed open requests (3440)
5723 - if default stripe count is 0, use OST count for inode size (3636)
5724 - update parent mtime/ctime on client for create/unlink (2611)
5725 - drop dentry ref in ext3_add_link from open_connect_dentry (3266)
5726 - free recovery state on server during a forced cleanup (3571)
5727 - unregister_reply for resent reqs (3063)
5728 - loop back devices mounting and status check on 2.6 (3563)
5729 - fix resource-creation race that can provoke i_size == 0 (3513)
5730 - don't try to use bad inodes returned from MDS/OST fs lookup (3688)
5731 - more debugging for page-accounting assertion (3746)
5732 - return -ENOENT instead of asserting if ost getattr+unlink race (3558)
5733 - avoid deadlock after precreation failure (3758)
5734 - fix race and lock order deadlock in orphan handling (3450, 3750)
5735 - add validity checks when grabbing inodes from l_ast_data (3599)
5737 - add /proc/.../recovery_status to obdfilter (3428)
5738 - lightweight CDEBUG infrastructure, debug daemon (3668)
5739 - change default OSC RPC parameters to be better on small clusters
5740 - turn off OST read cache for files smaller than 32MB
5741 - install man pages and include them in rpms (3100)
5742 - add new init script for (un)mounting lustre filesystems (2593)
5743 - run chkconfig in %post for init scripts (3701)
5744 - drop scimac NAL (unmaintained)
5746 2004-06-17 Cluster File Systems, Inc. <info@clusterfs.com>
5749 - clean kiobufs before and after use (3485)
5750 - strip trailing '/'s before comparing paths with /proc/mounts (3486)
5751 - remove assertions to work around "in-flight rpcs" recovery bug (3063)
5752 - change init script to fail more clearly if not run as root (1528)
5753 - allow clients to reconnect during replay (1742)
5754 - fix ns_lock/i_sem lock ordering deadlock for kms update (3477)
5755 - don't do DNS lookups on NIDs too small for IP addresses (3442)
5756 - re-awaken ptlrpcd if new requests arrive during check_set (3554)
5757 - fix cond_resched (3554)
5758 - only evict unfinished clients after recovery (3515)
5759 - allow bulk resend, prevent data loss (3570)
5760 - dynamic ptlrpc request buffer allocation (2102)
5761 - don't allow unlinking open directory if it isn't empty (2904)
5762 - set MDS/OST threads to umask 0 to not clobber client modes (3359)
5763 - remove extraneous obd dereference causing LASSERT failure (3334)
5764 - don't use get_cycles() when creating temp. files on the mds (3156)
5765 - hold i_sem when setting i_size in ll_extent_lock() (3564)
5766 - handle EEXIST for set-stripe, set proper directory name (3336)
5768 - servers can dump a log evicting a client - lustre.dump_on_timeout=1
5769 - fix ksocknal_fmb_callback() error messages (2918)
5771 2004-05-27 Cluster File Systems, Inc. <info@clusterfs.com>
5774 - don't copy lvb into (possibly NULL) reply on error (2983)
5775 - don't deref dentry after dput, don't free lvb on error (2922)
5776 - use the kms to determine writeback rpc length (2947)
5777 - increment oti_logcookies when osc is inactive (2948)
5778 - update client's i_blocks count via lvb messages (2543)
5779 - handle intent open/close of special files properly (1557)
5780 - mount MDS with errors=remount-ro, like obdfilter (2009)
5781 - initialize lock handle to avoid ASSERT on error cleanup (3057)
5782 - don't use cancelling-locks' kms values (2947)
5783 - use highest lock extent for kms, not last one (2925)
5784 - don't dereference ERR_PTR() dentry in error handling path (3107)
5785 - fix thread race in portals_debug_dumplog() (3122)
5786 - create lprocfs device entries at setup instead of at attach (1519)
5787 - common AST error handler, don't evict client on completion race (3145)
5788 - zero nameidata in detach_mnt in 2.6 (3118)
5789 - verify d_inode after revalidate_special is valid in 2.6 (3116)
5790 - use lustre_put_super() to handle zconf unmounts in 2.6 (3064)
5791 - initialize RPC timeout timer earlier for 2.6 (3219)
5792 - don't dereference NULL reply buffer if mdc_close was never sent (2410)
5793 - print nal/nid for unknown nid (3258)
5794 - additional checks for oscc recovery before doing precreate (3284)
5795 - fix ll_extent_lock() error return code for 64-bit systems (3043)
5796 - don't crash in mdc_close for bad permissions on open (3285)
5797 - zero i_rdev for non-device files (3147)
5798 - clear page->private before handing to FS, better assertion (3119)
5799 - tune the read pipeline (3236)
5800 - fix incorrect decref of invalidated dentry (2350)
5801 - provide read-ahead stats and refine rpc in flight stats (3328)
5802 - don't hold journal transaction open across create RPC (3313)
5803 - update atime on MDS at close time (3265)
5804 - close LDAP connection when recovering to avoid server load (3315)
5805 - update iopen-2.6 patch with fixes from 2399,2517,2904 (3301)
5806 - don't leak open file on MDS after open resend (3325)
5807 - serialize filter_precreate and filter_destroy_precreated (3329)
5808 - loop device shouldn't call sync_dev() for nul device (3092)
5809 - clear page cache after eviction (2766)
5810 - resynchronize MDS->OST in background (2824)
5811 - refuse to mount the same filesystem twice on same mountpoint (3394)
5812 - allow llmount to create routes for mounting behind routers (3320)
5813 - push lock cancellation to blocking thread for glimpse ASTs (3409)
5814 - don't call osc_set_data_with_check() for TEST_LOCK matches (3159)
5815 - fix rare problem with rename on htree directories (3417)
5817 - allow default OST striping configuration per directory (1414)
5818 - fix compilation for qswnal for 2.6 kernels (3125)
5819 - increase maximum number of MDS request buffers for large systems
5820 - change liblustreapi to be useful for external progs like lfsck (3098)
5821 - increase local configuration timeout for slow disks (3353)
5822 - allow configuring ldlm AST timeout - lustre.ldlm_timeout=<seconds>
5824 2004-03-22 Cluster File Systems, Inc. <info@clusterfs.com>
5827 - fixes for glimpse AST timeouts / incorrectly 0-sized files (2818)
5828 - don't overwrite extent policy data in reply if lock was blocked (2901)
5829 - drop filter export grants atomically with removal from device (2663)
5830 - del obd_self_export from work_list in class_disconnect_exports (2908)
5831 - don't LBUG if MDS recovery times out during orphan cleanup (2530)
5832 - swab reply message in mdc_close, other PPC fixes (2464)
5833 - fix destroying of named logs (2325)
5834 - overwrite old logs when running lconf --write_conf (2264)
5835 - bump LLOG_CHUNKSIZE to 8k to allow for larger clusters (2306)
5836 - fix race in target_handle_connect (2898)
5837 - mds_reint_create() should take same inode create lock (2926)
5838 - correct journal credits calculated for CANCEL_UNLINK_LOG (2931)
5839 - don't close files for self_export to avoid uninitialized obd (2936)
5840 - allow MDS with the same name as client node (2939)
5841 - hold dentry reference for closed log files for unlink (2325)
5842 - reserve space for all logs during transactions (2059)
5843 - don't evict page beyond end of stripe extent (2925)
5844 - don't oops on a deleted current working directory (2399)
5845 - handle hard links to targets without a parent properly (2517)
5846 - don't dereference NULL lock when racing during eviction (2867)
5847 - don't grow lock extents when lots of conflicting locks (2919)
5849 2004-03-04 Cluster File Systems, Inc. <info@clusterfs.com>
5852 - account for cache space usage on clients to avoid data loss (974)
5853 - lfsck support in lustre kernel code (2349)
5854 - reduce journal credits needed for BRW writes (2370)
5855 - orphan handling to avoid losing space on client/server crashes
5856 - ptlrpcd can be blocked, stopping ALL progress (2477)
5857 - use lock value blocks to assist in proper KMS, faster stat (1021)
5858 - takes i_sem instead of DLM locks internally on obdfilter (2720)
5859 - recovery for initial connections (2355)
5860 - fixes for mds_cleanup_orphans (1934)
5861 - abort_recovery crashes MDS in b_eq (mds_unlink_orphan) (2584)
5862 - block all file creations until orphan recovery completes (1901)
5863 - client remove rq_connection from request struct (2423)
5864 - conf-sanity test_5, proper cleanup in umount log not availale (2640)
5865 - recovery timer race (2670)
5866 - mdc_close recovey bug (2532)
5867 - ptlrpc cleanup bug (2710)
5868 - mds timeout on local locks (2588)
5869 - namespace lock held during RPCs (2431)
5870 - handle interrupted sync write properly (2503)
5871 - don't try to handle a message that hasn't been replied to (2699)
5872 - client assert failure during cleanup after abort recovery (2701)
5873 - leak mdc device after failed mount (2712)
5874 - ptlrpc_check_set allows timedout requests to complete (2714)
5875 - wait for inflight reqs when ptlrpcd finishes (2710)
5876 - make sure unregistered services are removed from the srv_list
5877 - reset bulk XID's when resending them (caught by 1138 test)
5878 - unregister_bulk after timeout
5879 - fix lconf error (2694)
5880 - handle write after unfinished setstripe, stripe-only getstripe (2388)
5881 - readahead locks pages, leaves pending causing memory pressure (2673)
5882 - increase OST request buffers to 4096 on large machines (2729)
5883 - fix up permission of existing directories in simple_mkdir (2661)
5884 - init deleted item, add assertions ptlrpc_abort_inflight() (2725)
5885 - don't assign transno to errored transactions (2742)
5886 - don't delete objects on OST if given a bogus objid from MDS (2751)
5887 - handle large client PAGE_SIZE readdir on small PAGE_SIZE MDS (2777)
5888 - if rq_no_resend, then timeout request after recovery (2432)
5889 - fix MDS llog_logid record size, 64-bit array alignment (2733)
5890 - don't call usermode_helper from ptlrpcd, DEFAULT upcall (2773)
5891 - put magic in mount.lustre data, check for bad/NULL mount data (2529)
5892 - MDS recovery shouldn't delete objects that it has given out (2730)
5893 - if enqueue arrives after completion, don't clobber LVB (2819)
5894 - don't unlock pages twice when trigger_group_io returns error (2814)
5895 - don't deref NULL rq_repmsg if ldlm_handle_enqueue failed (2822)
5896 - don't write pages to disk if there was an error (1450)
5897 - don't ping imports that have recovery disabled (2676)
5898 - take buffered bytes into account when balancing socknal conn (2817)
5899 - hold a DLM lock over readdir always, use truncate_inode_pages (2706)
5900 - reconnect unlink llog connection after MDS reconnects to OST (2816)
5901 - remove little-endian swabbing of llog records (1987)
5902 - set/limit i_blksize to LL_MAX_BLKSIZE on client (2884)
5903 - retry reposting request buffers if they fail (1191)
5904 - grow extent at grant time to avoid granting a revoked lock (2809)
5905 - lock revoke doesn't evict page if covered by a second lock (2765)
5906 - disable VM readahead to avoid reading outside lock extents (2805)
5908 - return LL_SUPER_MAGIC from statfs for the filesystem type (1972)
5909 - updated kernel patches for hp-2.4.20 kernel (2681)
5911 2004-02-07 Cluster File Systems, Inc. <info@clusterfs.com>
5914 - fix truncated write corruption (2366)
5915 - fix for failed assertion in iopen_connect_dentry (1792,2517)
5917 - don't flag the ptlrpcd thread with PF_MEMALLOC (2636)
5918 - ensure len(uuid) < 37 in lmc (1171)
5919 - fix ia64 OOPS in llog_test (2255)
5920 - zero end of page at obdfilter for partial page writes (2648)
5921 - don't leave stale dentries around after renames (bug 2428)
5922 - fix timeouts when evicting a client with a single lock held (2642)
5923 - set deadline for the initial HELLO message to drain (2634)
5924 - print out dotted-quad IP addresses in the socknal (2302)
5926 - additional debugging for MDS client eviction problem (2443)
5927 - fix mkfsoptions support for osts (2603, 2604)
5929 2004-01-27 Cluster File Systems, Inc. <info@clusterfs.com>
5932 - add series for the vanilla 2.6.0 kernel
5933 - add series for the vanilla 2.4.24 kernel
5934 - add series for a cray x86/64 UL kernel drop
5935 - fix xattr patches for the vanilla 2.4.19 series
5937 - generate true UUIDs in lmc (1171)
5938 - have portals stack dumping break in UML (2466)
5939 - avoid bad dchild deref; avoid inum lock w/o creation (2362)
5940 - allocate with _NOFS in ldlm to avoid deadlock (1933)
5941 - wake callback waiting threads on client eviction (2460)
5942 - Add --ptldebug and --subsystem to lmc (1719)
5943 - update assertion to allow safe interrupt allocation
5944 - set rq_no_resend for cancel requests (2432)
5945 - recalculate ptlrpcd timeout after resend (2494)
5946 - call vfs_rmdir when removing pending directories (2368)
5947 - fix renaming a file to itself (2429)
5948 - lmc creates a default one-stripe lov (2454)
5949 - expand procfs space to handle large clusters (2326)
5950 - increase UML stack to avoid overflow
5951 - update lconf's list of debug and subsystem masks
5952 - fix lfs find --obd (2510)
5953 - /proc tunable for disabling filter read caching (2591)
5954 - stop rpm packages from altering slapd.conf (2301)
5955 - disable nagle in the socknal under 0conf (2578)
5956 - choose mds inode size based on stripe count (2572)
5957 - fix kernel-source rpm problems (2516)
5959 - add --disable-doc to avoid pdf generation (2421)
5960 - update documentation, tests, type-os, comments
5961 - avoid format warnings on ia64
5962 - remove the TOE NAL
5963 - tiny code cleanups by removing unused fields
5965 2004-01-07 Cluster File Systems, Inc. <info@clusterfs.com>
5968 - fix obvious semaphore misuse in as-yet-unused setattr path (2348)
5969 - remove the most blatant lies from BUILDING file (2371)
5970 - change default debug level to reasonable production setting
5971 - reduce client side cache size to reduce cache flush time
5972 - reduce max RPCs in flight to avoid unnecessary file fragmentation
5973 - make TCP zerocopy and pinger support enabled by default (2476)
5974 - sync writes completed after process exits caused crashes (2319)
5975 - maintain correct mount count on the MDS (2356)
5976 - backout 1557, because 2316 wasn't really fixed
5977 - better file I/O statistics gathering in /proc
5978 - don't take unnecessary, deadlock-inducing bug in readpage (2383)
5979 - another kernel patch to fix zero-copy TCP function export
5980 - don't take duplicate lock when processing re-sent getattr (2420)
5981 - lctl uses obd_self_export instead of creating new conn (2353)
5982 - MDS/OST recovery case which requires object creation asserted (2425)
5983 - move lfs from /usr/sbin to /usr/bin in packages
5984 - fix race between mds_client_add and mds_client_free (2417)
5985 - use kmalloc instead of slabs in portals (2430)
5986 - don't create duplicate records when a failover MDS is present (2442)
5987 - remove unnecessary mount age check (2332)
5988 - don't remove directory inodes from locks prematurely (2451)
5989 - don't break if MDS service name is the same as hostname (2103)
5990 - fix races in client write RPC generation when cache full (2482)
5992 2003-12-13 Cluster File Systems, Inc. <info@clusterfs.com>
5995 - remove now-unused request->rq_obd (278)
5996 - if an allocation fails, print out how much memory we've used (1933)
5997 - use PORTAL_SLAB_ALLOC for structures, to get GFP_MEMALLOC (1933)
5998 - add the "configurable stack size" patch to most series files (1256)
5999 - ability to write large log records, for 100+ OST configs (2306)
6000 - fix NULL deref when filter_prep fails (2314)
6001 - fix operator precedence error in filter_sync
6002 - dynamic allocation of socknal TX descriptors (2315)
6003 - fix a missed case in the GFP_MEMALLOC patch, can cause deadlock (2310)
6004 - fix gcc 2.96 compilation problem in xattr kernel patch (2294)
6005 - ensure that CWARN messages in Portals always get to the syslog
6006 - __init/__exit are not for prototype decls (ldlm_init/exit)
6007 - x86-64 compile warning fixes
6008 - fix gateway LMC keyword conflict (2318)
6009 - fix MDS lock inversions in getattr/reint paths (1844)
6010 - fix a rare lock re-ordering bug, which caused deadlock (2322)
6011 - fix i_sem/journal inversion in fsfilt_ext3_write_record (2306)
6012 - DLM race condition prevented some lock evictions (2328)
6013 - ENOMEM detection and retry on socknal sends (2230)
6014 - use GFP_NOFS throughout Lustre, to combat ENOMEM (2230)
6015 - move osc_rpcd into ptlrpc, for use in MDC and others (2329)
6016 - protect MDS inode fsdata with stronger locking; fixes assertion (2313)
6017 - better error messages when a client is rejected during recovery (1505)
6018 - avoid cancelling locks which were never granted, after failure (2330)
6019 - fix i_sem/journal inversion in mds_client_add (2333)
6020 - fix truncate/getattr lock cycle deadlock (2334)
6021 - use rpcd to send close; allows resend after timeout, avoid leak (1897)
6022 - fix two rare exit paths which could leak an l_lock() ref (2321)
6023 - fencepost error in MDS/OST orphan recovery (2226)
6024 - make log record alignment 8 bytes (1988)
6025 - lstripe now fails when requested offset > ost_count (2237)
6026 - ensure that all kernel series have a complete list.h (1607)
6027 - fix crashes in special-file operations (2316)
6028 - lctl create/brw OID mismatch, caused by obsolete filter loop (2339)
6030 - allow configurable automake binary, for testing new versions
6031 - small update to the lfs documentation
6033 2003-12-03 Cluster File Systems, Inc. <info@clusterfs.com>
6035 * fix negative export reference count in fsfilt_sync (2312)
6037 2003-12-01 Cluster File Systems, Inc. <info@clusterfs.com>
6038 * release candidate 0.9.1
6040 - orphans are moved into the PENDING directory for possible recovery
6041 - replayed opens now open by fid for orphan/rename safety (1042)
6042 - last close of an orphan inode generates a transno (683)
6043 - chdir() and mount() now pin the directory entry (1020)
6044 - avoid CERROR in normal ll_setattr_raw() error case (1500)
6045 - discard very old requests without processing them (1502)
6046 - remove some common, well-understood CERRORs (1505)
6047 - require O_DIRECT I/O to be page-sized to workaround IA64 crash (1609)
6048 - clear "grant" flags in OST replies until OST grant code lands (1644)
6049 - fix read performance by not clobbering i_blksize on client (1598)
6050 - fix __ldlm_handle2lock oops by not dereferencing lock after PUT (1625)
6051 - make LRU size a /proc tunable, clears locks when reduced (707)
6052 - fix some lprocfs rot that prevented ptlbd from loading (1732)
6053 - server locks take references on exports now (1558)
6054 - build fixes for 2.4.20-rh trees (1663)
6055 - return an error from lov_create if all OSCs are inactive (1751)
6056 - fix import levels when a reconnect happens without a timeout (1597)
6057 - exit early from mds_open if we get a lookup error (1749)
6058 - partial page read at EOF wouldn't wait for disk before sending (1642)
6059 - avoid NULL deref in obdfilter when reading page past EOF (1592)
6060 - avoid LASSERT in ll_intent_lock if server failed very early (1090)
6061 - fix LBUG in ll_it_open_error with rc = -2 (1861)
6062 - write/truncate lock inversion (1639)
6063 - Don't auto-load obdclass, portals modules during cleanup (1495)
6064 - fix timestamps from jumping to "now" (1763)
6065 - extra journal assertions (1648)
6066 - add an extra multiunlink test (1771)
6067 - fix read_record/write_record API (1776)
6068 - fix leak of offset_extent, possible incorrect i_size later (1772)
6069 - fix lasserts in mis-matched transnos during open-unlink testing (1541)
6070 - Debugging for the kqswnal_get_idle_tx problems (1820)
6071 - Allow recovery to be attempted multiple times (1536)
6072 - Write out MDS last_rcvd file after it is first created (1600)
6073 - Fix tx_descriptor leak in failed transmit situations (1827)
6074 - ext3 journaling fixes for assertion failure after IO error (1871)
6075 - class_export_put() on freed export after completion AST error (1896)
6076 - Fix revalidate looping in VFS (1322)
6077 - Don't access a freed export during MDS_REINT timeout (1521)
6078 - Add open-unlink recovery support on the MDS (1673,1764)
6079 - Return an error if no MDS data was read from last_rcvd (1946)
6080 - Fix for lookup "." or ".." crash on error (1932,1931,1935)
6081 - Don't setup a disk device that doesn't match exported UUID (317)
6082 - Reduce bulk RPC timeout to avoid cascading client/OST failures (1845)
6083 - avoid committing NULL handle in force close
6084 - local.sh is now a one-stripe LOV configuration
6085 - POSIX utime.4 -EPERM on FIFO not owned by user (56)
6086 - fix ext3 htree duplicate directory entry corruption (1516)
6087 - POSIX creat.13, fstat.1, open.18, stat.3 new file atime/mtime (2020)
6088 - update to new LOV EA format (2097)
6089 - interoperability for different PAGE_SIZE/wordsize (686,1821,1343,2042)
6091 2003-06-15 Phil Schwan <phil@clusterfs.com>
6094 - imports and exports cleanup too early, need refcounts (349, 879, 1045)
6095 - per-import/export recovery handling (958, 931, 959)
6096 - multiple last-rcvd slots, for serving multiple FSes (949)
6097 - connections are again shared between multiple imp/exports (963, 964)
6098 - "umount -f" would hang if any requests needed to be sent (393, 978)
6099 - avoid pinning large req buffer by copying for queued messages (989)
6100 - add "uuid" to "lctl device" command to help upcalls (991)
6101 - "open" RPCs with transnos would confuse recovery counters (1037)
6102 - do proper endian conversion of all wire messages (288, 340, 891)
6103 - remove OST bulk get LBUGs, fix ost_brw_write cleanup (1126)
6104 - call waiting locks callback from LDLM recovery thread (1127, 1151)
6105 - fix ptlrpc_connection leak in target_handle_connect (1174)
6106 - fix import refcounting bug in OST and MDS cleanup (1134)
6107 - if an invalid-at-open-time OSC returned before close(), LBUG (1150)
6108 - fix very unlikely obd_types race condition (501)
6109 - remove osc_open hack for echo_client (1187)
6110 - we leaked exports/dlmimps for forcibly disconnected clients (1143)
6111 - a failure in read_inode2 leads to deadlock (1139)
6112 - cancel ack-locks as soon as transaction is committed (1072)
6113 - fix major leaks and crashes in the bulk I/O path (937, 1057)
6114 - make sure to commitrw after any preprw to avoid deadlock (1162)
6115 - failing to execute a file in a lustre FS would lock inode (1203)
6116 - small DEBUG_REQ fix to avoid dereferencing a NULL (1227)
6117 - don't ASSERT while cleaning up an incompletely-setup obd (1248)
6118 - obd_uuid2tgt would walk off the end of the list (1255)
6119 - on IA64 the osc would give portals incorrect bulk size (1258)
6120 - fix debug daemon ioctl interface; allows daemon on ia64 (1274)
6121 - fix lock inversion caused by new llite matching code (1282)
6122 - limit the number of dirty pages on a client to 10MB (1286)
6123 - timed out locks were not being corrected cancelled (1289)
6124 - fix O_DIRECT above 4GB on IA-32 (1292)
6125 * major user-visible changes
6126 - fail out/fail over policy now controlled by the upcall (993)
6128 - add OBD_PING to check server availability and failure (954)
6129 - lustre messages are now sent in sending host order (288, 340, 891)
6130 - add eadatalen to MDS getattr reply (340)
6131 - OST read replies may contain second buffer, with per-page status (593)
6133 2003-03-11 Phil Schwan <phil@clusterfs.com>
6136 - LDLM_DEBUG macro fix, for gcc 3.2 (850)
6137 - failed open()s could cause deadlock; fixed (867, 869)
6138 - stop cancelling OST locks when files are closed (481)
6139 - overlapping XID spaces caused network corruption (851, 853)
6140 - fix unsafe fsfilt counter arithmetic; change to atomic_t
6141 - setattr_raw added, to do single-RPC, server-side setattrs
6142 - lmc/lconf syntax change for OST UUIDs
6143 - fix crashy race condition between ptlrpc_free_req and osc_close
6144 - don't use request in mdc_enqueue if we hit a timeout (889)
6145 - don't set the inode i_size for regular files from the MDS (896)
6146 - handle out of order completion AST (842)
6147 - don't LBUG if a lock request times out after receiving AST (913)
6148 - avoid d_rehash race in ll_find_alias by rehashing inside dcache_lock
6149 - if a bad lock AST arrives, send an error instead of dropping entirely
6150 - return 0 from revalidate2 if ll_intent_lock returns -EINTR (912)
6151 - fix leak in bulk IO when only partially completed (899, 900, 926)
6152 - fix O_DIRECT for ia64 (55)
6153 - (almost) eliminate Lustre-kernel-thread effects on load average (722)
6154 - C-z after timeout could hang a process forever; fixed (977)
6156 - client-side I/O cache (678, 924, 929, 941, 970)
6158 - READPAGE and SETATTRs which don't take server-side locks get
6161 2003-02-11 Phil Schwan <phil@clusterfs.com>
6164 - Fix ldlm_lock_match on the MDS to avoid matching remote locks (592)
6165 - Fix fsfilt_extN_readpage() to read a full page of directory
6166 entries, or fake the remainder if PAGE_SIZE != blocksize (500)
6167 - Avoid extra mdc_getattr() in ll_intent_lock when possible (534, 604)
6168 - Fix imbalanced LOV object allocation and out-of-bound access (469)
6169 - Most intent operations were removed, in favour of a new RPC mode
6170 that does a single RPC to the server and bypasses most of the VFS
6171 - All LDLM resource ID arrays were removed in favour of ldlm_res_id
6172 - Aggressively cancel local locks on DLM servers
6173 - mds_reint_unlink sends EA to the client if it's the last nlink.
6174 client uses that EA to unlink OST objects.
6175 - mds_reint_{rename,unlink,link} were rewritten to take ordered locks
6176 - recursive symlinks were fixed (439)
6177 - fixed NULL deref in DEBUG_REQ
6178 - filter_update_lastobjid no longer calls sync, which annoyed extN
6179 - fixed multi-client small-writes to a single file problem (445)
6180 - fixed mtime updates during file writes (607)
6181 - fixed vector writes on obdfilter causing problems when ENOSPC (670)
6182 - fixed bug in obd_brw_read/write() (under guise of testing 367)
6183 - fixed Linux OST size reporting problem (444, 656)
6184 - OST now updates object mtime with writes or setattr (607, 619)
6185 - client verifies file size before zeroing page past EOF (445)
6186 - OST now writes last allocated objid to disk with allocation (108)
6187 - LOV on echo now works (409)
6189 - mds_reint_unlink sends a new buffer, with the EA included. this
6190 buffer is only valid if body->valid & OBD_MD_FLEASIZE, which is only
6191 set if a regular file was being unlinked, and it was the last link
6192 - use PtlGet from the target for bulk writes (315)
6193 - OST now updates object mtime with writes or setattr (607, 619)
6194 - LDLM now has a grant-time callback to revalidate locked items, if
6196 - Many MDS operations were reorganized to combat race conditions
6198 - Merge b_intel branch (updated lprocfs code) - now at /proc/fs/lustre
6199 - configure check to avoid gcc version 2.96 20000731-2.96-98) (606)
6201 2003-01-06 Andreas Dilger <adilger@clusterfs.com>
6204 - Fully reactivate OST imports after reconnection (512, others)
6205 - Make sure client sees our -ENOTCONN from mds_handle (513 - partial)
6206 - More graceful error handling for truncating on dead OST (515)
6207 - Don't error out unless we're actually accessing dead stripes (474)
6208 - Fix garbage sizes when stripes are missing (410)
6209 - LRU counters were broken, causing constant lock purge (433, 432)
6210 - garbage on read from stripes with failed OSTs (441)
6211 - mark OSCs as active before reconnecting during recovery (438)
6212 - lov_enqueue and lov_cancel need to handle inactive OSTs (403)
6213 - lfind did not preserve OST order in output (443)
6214 - symlinks cause hung clients, incorrect data (439)
6215 - stop dereferencing request after dropping refcount (457)
6216 - don't LASSERT(spin_is_locked) on non-SMP (455)
6217 - fixes for many rename() bugs
6218 - fstat didn't correctly synchronize attributes (399)
6219 - server must handle lock cancellation during blocking AST prep (487)
6220 - bulk descriptors were free()d too soon (511)
6221 - fix paths in lconf, which would load incorrect modules (451, 507)
6222 - fix confusing lconf 'host not found' error message (386)
6223 - fix lock order deadlock on OST (O/R i_sem before journal ops, 478)
6224 - fix race condition in mdc_blocking_ast() for inode access (526)
6225 - fix lov_unpackmd() unpacking wrong number of stripes (537)
6226 - fix lov_set_osc_active() marking wrong OSC inactive (440)
6227 - fix bad lstripe lov_unpackmd() assertion (fix layering too) (527)
6228 - fix multiple writes of stripe MD to MDS (358, maybe 519)
6229 - fix lstripe in several ways (kernel side) (527)
6230 - fix request leak in ldlm_cli_enqueue (262)
6231 - incorrect OSC was marked inactive after OST failure
6232 - call mds_fs_cleanup before unmounting filesystem (524)
6233 - fix races between taking ns_lock and ldlm_lock_change_resource
6234 - fix races updating LOV export open file list
6235 - fix lov_enqueue error path, avoid decref-ing bad lock handle
6236 - fix recovery NULL deref in ldlm_cli_cancel_unused
6237 - fix some DLM races by using new hash table for lock handles (419)
6238 - permit the client to specify desired inodes, at replay
6239 - duplicate requests when we queue them for replay reintegration
6240 - fix last_rcvd offset calculation
6241 - sync after each recovered transaction, so we always make progress
6242 - never, not always, ERESTART requests without transnos
6243 - store the lov_desc in the MDS, so we don't depend on getlovinfo to
6245 - skip replay if the MDS says that the client is already connected
6246 - don't check for a recovery-enabled export to match lctl's UUID
6247 - don't INC_USE_COUNT for phantom exports
6248 - don't crash when cleaning up phantom exports (567)
6249 - don't double-finish or set replay data for errored mdc_open requests
6250 - abort requests when they time out, so we don't get old replies
6251 - send/receive replies for AST messages again
6252 - if the client says that it doesn't have the lock, cancel it on the
6254 - if we timeout during I/O, don't try to cancel an in-use lock; instead
6255 mark it as destroyed, it will all work out when decref is called
6256 - fix module use counts (22, 581)
6258 - ASTs now expect a reply (server cancels lock on error reply)
6260 2002-12-02 Andreas Dilger <adilger@clusterfs.com>
6263 - fix many simultaneous client startup (392)
6264 - fix dentry->d_it clobbering
6265 - credentials weren't being shipped for readdir/getattr operations
6266 - remove invalid assertions triggered during some concurrent MD
6268 - proper Lustre versions added (336, 389)
6269 - fix memory leak for create error case (398)
6270 - fix LOV locking bug that would get cli/srv out of sync
6271 - fix echo client over LOV (409)
6272 - fix dbench 2, extN refcount problem (170, 258, 356, 418)
6273 - fix double-O_EXCL intent crash (424)
6274 - avoid sending multiple lock CANCELs (352)
6276 - MDS can do multi-client recovery (modulo bugs in new code)
6278 - many updates, edits, cleanups
6280 2002-11-18 Phil Schwan <phil@clusterfs.com>
6283 - fix null d_it dereference (346)
6284 - fix full OST/dbench hang (333)
6285 - fix permission problem with file removal (286)
6286 - fix removal of OSCs from LOV when they fail
6287 - fix NULL deref during bulk timeout (214)
6288 - fix problems related to multiple filesystems on one MDS (241)
6289 - fixed serious subtle metadata locking bugs
6290 - free locks on clients when inodes are removed due to memory
6292 - fix inode pointer in lock data (285)
6293 - partial support for multiple MDS on a single host (241)
6294 - data locks weren't cancelled at clear_inode time (290, 311)
6295 - intent locks could lead to unbounded lock growth (205)
6296 - added a maximum lock count, an LRU list, and a flusher
6297 - fix multiple rename (365)
6298 - properly abstracted the echo client
6299 - OSC locked 1 byte too many; fixed
6300 - rewrote brw callback code:
6301 - fixed recovery bugs related to LOVs (306)
6302 - fixed too-many-pages-in-one-write crash (191)
6303 - fixed (again) crash in sync_io_timeout (214)
6304 - probably fixed callback-related race (385)
6306 - Add capability to MDS protocol
6307 - LDLM cancellations and callbacks on different portals
6309 2002-10-28 Andreas Dilger <adilger@clusterfs.com>
6312 - limit client IOV size to PTL_MD_MAX_IOV (611336, 191)
6313 - defer open object destruction to close time (601981, 138)
6314 - open/close OST file handle in obdo (OBD_MD_FLHANDLE) (601981, 138)
6315 - move LDLM_ENQUEUE/CONVERT back to MDS portal (625069)
6316 - abstract ll_lookup2, fix ll_revalidate2 to use abstraction (256)
6317 - don't call obd_setattr in ll_file_release for destroyed objects
6318 * protocol change to lustre_msg: move |version| and add |flags|
6319 * protocol change to osc_punch: "start" in "o_size", "end" in "o_blocks"
6320 * lock replay: for LDLM_FL_REPLAY trust client to do right thing
6321 * added replay of create, unlink, link and rename operations during
6322 MDS failover; recovery should be much more robust now
6323 * remove failed OSCs from LOVs (only lov_create uses this so far)
6324 * the lustre-HOWTO was brought (more) up to date (582544)
6326 2002-10-23 Phil Schwan <phil@clusterfs.com>
6329 - in-use dentries weren't being reused properly (617851)
6330 - prevent multiple LDLM setup (599178)
6331 - fix LOV size calculations for truncate (617853)
6332 - fix client handling of MDS intent errors (POSIX)
6333 - fix permission bug in lovstripe.c test (624321)
6334 - fix MDS thread deadlock - move LDLM handler to DLM portal (625069)
6335 - truncate past end of file could corrupt data
6336 - proper cleanup after timeouts, crashes, etc (592524, 550815)
6337 - a race in recovery could return ETIMEDOUT to apps (623947)
6338 - building outside the source directory was fixed
6339 * the lustre-HOWTO was brought (more) up to date (582544)
6340 * major progress was made on recovery functionality
6342 2002-10-10 Phil Schwan <phil@clusterfs.com>
6345 - recovery deadlock fix
6346 - rm -rf causes LBUG fix (617817)
6347 - file open by multiple tasks fix (618962)
6348 - directory permissions bugs (602707 and 620007)
6349 - journal_stop fixed with locking (611313)
6350 - O_APPEND failures resolved (618273, perhaps 614459)
6351 - lconf PATH fix (619770)
6352 - IA64 build fix (621450)
6353 - RPC buffer sizes scale with amount of memory
6355 2002-10-01 Phil Schwan <phil@clusterfs.com>
6358 - locks would be cancelled without throwing away data pages,
6359 resulting in inconsistent data (605627)
6360 - inode attributes were not always being refreshed (605627, 612449)
6361 - lconf now continues to cleanup after lctl reports an error
6362 - MDS now enforces user permissions (602707)
6363 - lprocfs cleanup fixed, but not yet enabled (614157)
6364 - fixed infinite server hang, should a client not respond to an AST
6365 - avoid going into recovery if user calls readlink() with a buffer
6366 that's too small (613941)
6367 - AST RPCs no longer require replies (614867) -- this may be changed
6368 - don't crash server if client sends an IOV that's too big (611336)
6369 - fixed lock conversion deadlock (611892)
6370 - fixed the following of symlinks (614622)
6371 * recovery: the server can remove locks from a client that dies, other
6372 clients can make progress
6373 * more extN patch fixes
6374 * compile-time configurable ptlrpc buffer allocations
6376 - collaborative read cache document
6377 - Lustre Lite Performance CDR document-in-progress
6379 2002-09-20 Andreas Dilger <adilger@clusterfs.com>
6382 - fix typo in patch-2.4.18
6384 2002-09-20 Andreas Dilger <adilger@clusterfs.com>
6387 - clear ptlrpc request each time in handle_incoming_request()
6388 - unlink of files now destroys the object on the OST
6390 2002-09-19 Peter Braam <braam@clusterfs.com>
6392 * add hard link support
6393 * change obdfile creation method
6394 * kernel patch changed
6396 2002-09-19 Peter Braam <braam@clusterfs.com>
6399 - stack overflow bug in extN fixed
6401 2002-09-18 Andreas Dilger <adilger@clusterfs.com>
6403 * documentation updates
6404 - add man pages for config tools
6405 - update tests/README to describe testing with new config tools
6406 - finish metadata API descriptions
6407 * bug fixes and cleanups
6408 - statfs workaround for 16TB limit
6409 - LOV stripe allocation improved, can stripe on subset of OSTs
6410 - LOV file size/IO offset was wrong for files > 4GB in size
6411 - object EA data was being dropped, caused files to be unreadable
6412 - memory overflow with non-LOV OST caused memory corruption
6413 - fixed regression tests to work with new config tools, obdfilter
6414 - fixed bug when directory size became larger than 1 block
6415 - fixed bug (for single client case) when PWD was deleted
6416 - invalidate local directory pages when doing intent-based ops
6417 - avoid LDLM oops when lock callback contained bad data
6419 2002-09-09 Andreas Dilger <adilger@clusterfs.com>
6421 * documentation updates
6422 * bug fixes and cleanups
6423 - configuration tools
6426 - 64-bit compile warnings
6427 - 64-bit internal statfs data
6429 * test_brw on persistent OST devices
6431 * lprocfs (disabled)
6433 2002-09-04 Andreas Dilger <adilger@clusterfs.com>
6435 * documentation updates
6436 * bug fixes and cleanups
6437 * configuration tools
6439 2002-08-30 Peter J. Braam <braam@clusterfs.com>
6442 * many small fixes to 0_5_4
6443 * io/network handling
6444 * thinkos in MDS operations
6446 2002-08-24 Peter J. Braam <braam@clusterfs.com>
6449 * crucial basic fixes to 0.5.3
6450 * IOR, Iozone work over Elan
6453 2002-08-07 Phil Schwan <phil@clusterfs.com>
6454 * version 0_5_3, our first alpha
6455 * we use the new Portals iovs
6456 * documentation updates
6457 * bug fixes and cleanups
6458 * small changes in the DLM wire protocol
6460 2002-07-25 Peter J. Braam <braam@clusterfs.com>
6461 * version 0_5_1 with some initial stability,
6462 * locking on MD and file I/O.
6463 * documentation updates
6464 * several bug fixes since 0.5.0
6465 * small changes in wire protocol
6467 2002-07-18 Phil Schwan <phil@clusterfs.com>
6469 * delivered as Lustre Light Alpha
6470 * fixed a crash after handling invalid MDS requests
6471 * fixed directory pages for architectures with non-4k pages sizes
6473 2002-07-11 Andreas Dilger <adilger@clusterfs.com>
6474 * release version v0_4_4
6475 * Moves TCP acceptor to be on port 2432 (unused Coda port) instead
6477 * Fixes a number of interruption problems with OST operations.
6478 * Update documentation for portals header changes
6479 * Move all wire protocol structs/defines to lustre_idl.h
6480 * Fixes symlink length bug.
6481 * Add tcpdump to repository.
6483 2002-07-05 Andreas Dilger <adilger@clusterfs.com>
6484 * release version v0_4_3
6485 * Fixes statfs for inodes on extN.
6486 * Fixes bug in runtests which would delete /etc/hosts.
6487 * Use 64-bit object IDs wherever possible (not into VFS though)
6488 Remove ost_get_info, which is unused by lustre, and out of date.
6490 2002-07-03 Peter Braam <braam@clusterfs.com>
6491 * release version v0_4_2 Fixes a lookup error (type not passed)
6492 * move forward to head of Portals
6493 * move forward to latest Lustre kernel
6495 2002-06-25 Peter Braam <braam@clusterfs.com>
6496 * release version v0_4_1. Hopefully stable on single node use.