1 tbd Sun Microsystems, Inc.
4 2.6.16.60-0.37 (SLES 10),
5 2.6.27.21-0.1 (SLES11),
6 2.6.18-128.1.6.el5 (RHEL 5)
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: 1.41.6.sun1
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.
17 Severity : enhancement
19 Description: implement async create (obd_async_create) method for osc, to avoid
20 too long waiting new ost objects with holding ldlm lock.
22 Severity : enhancement
24 Descriptoin: OST grant too much space to client even there are not enough space.
25 Details : Client will shrink its grant cache to OST if there are no write
26 activity over 6 mins (GRANT_SHRINK_INTERVAL), and OST will retrieve
27 this grant cache if there are already not enough avaible space
28 (left_space < total_clients * 32M).
31 Frequency : rare, if clients are evicted due to failure to return a lock
33 Description: don't increase ldlm timeout if previous client was evicted
34 Details : if a client doesn't respond to a blocking callback within the
35 adaptive ldlm enqueue timeout, don't adjust the adaptive estimate
36 when the lock is next granted.
40 Description: client not allowed to reconnect to OST because of active request
41 Details : abort bulk requests received by the OST once the client has timed
42 out since the client will resend the request anyway. The client
43 also now retries to reconnect to the same server if a connect
44 request failed with EBUSY or -EAGAIN.
47 Frequency : rare, if used wide striped file and one ost in down.
49 Descriptoin: don't return error if we created a subset of objects for file.
50 Details : lov_update_create_set() uses set->set_success as index for created
51 objects, so if some requests failed, they will have hole at end of
52 array and we can use qos_shrink_lsm for allocate correct lsm.
57 Description: fix lqs' reference which won't be put in some situations
58 Details : This patch fixes:
59 1. In quota_check_common(), this function will check quota
60 for user and group, but only send one return via "pending".
61 In most cases, the pendings should be same. But that is not
63 2. If quotaoff runs between lquota_chkquota() and
64 lquota_pending_commit(), the same thing will happen too.
66 - if (!ll_sb_any_quota_active(qctxt->lqc_sb))
71 Description: Rare Client crash on resend if the file was deleted.
72 Details : When file is opened, but open reply is lost and file is
73 subsequently deleted before resend, resend processing logic
74 breaks trying to open the file again, should not try to open.
78 Description: Rare dentry leakage on MDS on forceful shutdown and
80 Details : Sometimes on transaction errors mds_mfd_close did not release
81 mfd and dentry pointed to it. reconstruct_open did not release
82 parent dentry if no matching mfd was found and child lookup was
85 ------------------------------------------------------------------------------
86 tbd Sun Microsystems, Inc.
88 * Support for kernels:
89 2.6.16.60-0.37 (SLES 10),
90 2.6.27.21-0.1 (SLES11),
91 2.6.18-128.1.6.el5 (RHEL 5)
92 * Client support for unpatched kernels:
93 (see http://wiki.lustre.org/index.php?title=Patchless_Client)
94 2.6.16 - 2.6.27 vanilla (kernel.org)
95 * Recommended e2fsprogs version: 1.41.5-sun2
96 * File join has been disabled in this release, refer to Bugzilla 16929.
97 * A new Lustre ADIO driver is available for MPICH2-1.0.7.
98 * NFS export disabled when stack size < 8192. Since the NFSv4 export
99 of Lustre filesystem with 4K stack may cause a stack overflow. For
100 more information, please refer to bugzilla 17630.
104 Description: resolve race between obd_disconnect and class_disconnect_exports
105 Details : if obd_disconnect will be called to already disconnected export he
106 forget release one reference and osc module can't unloaded.
108 Severity : enhancement
110 Description: move AT tunable parameters for more consistent usage
111 Details : add AT tunables under /proc/sys/lustre, add to conf_param parsing
113 Severity : enhancement
115 Description: Update kernel to RHEL5.3 2.6.18-128.1.6.el5.
117 Severity : enhancement
119 Description: Update kernel to SLES10 SP2 2.6.16.60-0.37.
121 Severity : enhancement
123 Description: Caching OSS
124 Details : introduce data caching on the OSS. The OSS now relies on the linux
125 kernel page cache to keep recently accessed data in memory.
126 It is worth noting that all write requests are still flushed
127 synchronously as in lustre 1.6.
129 Severity : enhancement
131 Description: version based recovery
132 Details : introduce finer grained recovery able to detect transaction
133 dependencies and can deal with transaction gaps.
135 Severity : enhancement
137 Description: Enable adaptive timeouts by default
141 Descriptoin: correctly skip time estimate if in recovery
142 Details : rq_send_state insn't bitmask so using bitwise ops is forbid.
146 Descriptoin: fix goal inodes
147 Details : Allocate inodes for llog in last inode group for avoid broken
152 Descriptoin: OSS DeadLock
153 Details : Use trylock to prevent deadlock when shrink icache.
155 Severity : enhancement
157 Description: Allow tuning service thread via /proc
158 Details : For each service a new
159 /proc/fs/lustre/{service}/*/thread_{min,max,started} entry is
160 created that can be used to set min/max thread counts, and get the
161 current number of running threads.
163 Severity : enhancement
165 Description: Add state history info file, enhance import info file
166 Details : Track import connection state changes in a new osc/mdc proc file;
167 add overview-type data to the osc/mdc import proc file.
171 Description: Reduce small size read RPC
172 Details : Set read-ahead limite for every file and only do read-ahead when
173 available read-ahead pages are bigger than 1M to avoid small size
178 Description: free_entry erroneously used groups_free instead of put_group_info
180 Severity : enhancement
182 Description: Make read-ahead stripe size aligned.
184 Severity : enhancement
186 Description: MDS create should not wait for statfs RPC while holding DLM lock.
188 Severity : enhancement
190 Description: Update client support to vanilla kernels up to 2.6.27.
193 Frequency : rare, connect and disconnect target at same time
195 Descriptoin: ASSERTION(atomic_read(&imp->imp_inflight) == 0
196 Details : Don't call obd_disconnect under lov_lock. this long time
197 operation and can block ptlrpcd which answer to connect request.
199 Severity : enhancement
201 Description: Update to sles9 kernel-2.6.5-7.315.
204 Frequency : start MDS on uncleanly shutdowned MDS device
206 Descriptoin: ll_sync thread stay in waiting mds<>ost recovery finished
207 Details : Stay in waiting mds<>ost recovery finished produce random bugs
208 due race between two ll_sync thread for one lov target. send
209 ACTIVATE event only if connect realy finished and import have
213 Frequency : start MDS on uncleanly shutdowned MDS device
215 Descriptoin: aborting recovery hang on MDS
216 Details : Don't throttle destroy RPCs for the MDT.
218 Severity : enhancement
220 Descriptoin: Don't sync journal after every i/o
221 Details : Implement write RPC replay to allow server replies for write RPCs
222 before data is on disk. However, this feature is disabled by
223 default since some issues leading to data corruptions have been
224 found during recovery (e.g. bug 19128). This feature can be
225 enabled by running the following command on the OSSs:
226 lctl set_param obdfilter.*.sync_journal=0
230 Description: Slow reads beyond 8Tb offsets.
231 Details : Page index integer overflow in ll_read_ahead_page
235 Description: Soft lockup on OSS after MDS failover
236 Details : MSG_CONNECT_INITIAL is not set on the initial MDS->OST connect.
237 As a conseqence, the patch from bug 18224 is not operational
238 and the MDS export cannot be reused on the OSTs until it gets
242 Frequency : rare, only if using MMP with Linux RAID
244 Description: MMP doesn't work with Linux RAID
245 Details : While using HA for Lustre servers with Linux RAID, it is possible
246 that MMP will not detect multiple mounts. To make this work we
247 need to unplug the device queue in RAID when the MMP block is being
248 written. Also while reading the MMP block, we should read it from
249 disk and not the cached one.
252 Frequency : rare, during recovery
254 Description: Assertion failure in ldlm_lock_put
255 Details : Do not put cancelled locks into replay list, hold references on
260 Description: 1.6.5 mdsrate performance is slower than 1.4.11 (MDS not cpu bound)
261 Details : create_count always drops to the min value (=32) because grow_count
262 is being changed before the precreate RPC completes.
265 Frequency : Only in RHEL5 when mounting multiple ext3 filesystems
268 Description: "kmem_cache_create: duplicate cache jbd_4k" error message
269 Details : add proper locking for creation of jbd_4k slab cache
273 Description: MMP check in ext3_remount() fails without displaying any error
274 Details : When multiple mount protection fails during remount, proper error
277 -------------------------------------------------------------------------------
278 2008-12-31 Sun Microsystems, Inc.
280 * Support for kernels:
281 2.6.5-7.314 (SLES 9),
282 2.6.9-67.0.22.EL (RHEL 4),
283 2.6.16.60-0.31 (SLES 10),
284 2.6.18-92.1.17.el5 (RHEL 5),
285 2.6.22.14 vanilla (kernel.org)
286 * Client support for unpatched kernels:
287 (see http://wiki.lustre.org/index.php?title=Patchless_Client)
288 2.6.16 - 2.6.22 vanilla (kernel.org)
289 * Client support for unpatched kernels:
290 we do not recommend using patchless RHEL4 clients with kernels
291 prior to 2.6.9-55EL (RHEL4U5).
292 * Recommended e2fsprogs version: 1.40.11-sun1
293 * Note that reiserfs quotas are disabled on SLES 10 in this kernel.
294 * RHEL 4 and RHEL 5/SLES 10 clients behaves differently on 'cd' to a
295 removed cwd "./" (refer to Bugzilla 14399).
296 * A new quota file format has been introduced in 1.6.5.
297 The format conversion from prior releases is handled transparently,
298 but releases older than 1.4.12/1.6.5 don't understand this new
299 format. The automatic format conversion can be avoided by running
300 the following command on the MDS:
301 'tunefs.lustre --param="mdt.quota_type=ug1" $MDTDEV'.
302 For more information, please refer to bugzilla 13904.
303 * A new quota file format was introduced in 1.6.6/1.8.0 (kernel 2.6.16+)
304 The format conversion from prior releases is handled transparently,
305 but releases older than 1.6.6/1.8.0 don't understand this new
306 format. The automatic format conversion can be avoided by running
307 the following commands on the MDS and OSS servers (for
308 pre 1.4.12-1.6.5 quota files):
309 'tunefs.lustre --param="mdt.quota_type=ug1" $MDTDEV',
310 'tunefs.lustre --param="ost.quota_type=ug1" $MDTDEV'
311 or (for 1.4.12/1.6.5 quota files)
312 'tunefs.lustre --param="mdt.quota_type=ug2" $MDTDEV',
313 'tunefs.lustre --param="ost.quota_type=ug2" $MDTDEV'
314 For more information, please refer to bugzilla 13904.
315 * Output of lfs quota has been made less detailed by default,
316 old (verbose) output can be obtained by using -v option.
317 * File join has been disabled in this release, refer to Bugzilla 16929.
318 * A new Lustre ADIO driver is available for MPICH2-1.0.7.
319 * NFS export disabled when stack size < 8192. Since the NFSv4 export of
320 Lustre filesystem with 4K stack may cause a stack overflow. For more
321 information, please refer to bugzilla 17630.
323 Severity : enhancement
325 Description: add lazystatfs mount option to allow statfs(2) to skip down OSTs
326 Details : Allow skip disconnected ost for send statfs request and hide error
330 Frequency : rare, on llog test 6
332 Descriptoin: don't allow connect to already connected import
333 Details : Connect to already connected import hides connection problem.
336 Frequency : rare, on failed llog setup
338 Descriptoin: don't leak obd reference on failed llog setup
339 Details : For failed llog setup - mgc forget call class_destroy_import for
340 client import, move destroy import to more generic place.
345 Descriptoin: allow kill process which wait statahead result
346 Details : For some reasons 'ls' can stick in waiting result from statahead,
347 in this case need way for kill this process.
350 Frequency : rare, at shutdown
352 Descriptoin: panic at umount
353 Details : llap_shrinker can be raced with killing super block from list and
354 this produce panic with access to already freeded pointer
359 Descriptoin: don't lose wakeup for imp_recovery_waitq
360 Details : Recover_import_no_retry or invalidate_import and import_close can
361 both sleep on imp_recovery_waitq, but we was send only one wakeup
367 Descriptoin: panic in mds_open
368 Details : don't confuse mds_finish_transno() with PTR_ERR(-ENOENT)
373 Descriptoin: stuck in cache_remove_extent() or panic with accessing to already
375 Details : Release lock refernce only after add page to pages list.
378 Frequency : always with long access acl
380 Descriptoin: mds can't pack reply with long acl.
381 Details : mds don't control size of acl but they limited by reint/getattr
386 Frequency : on remount
388 Description: external journal device not working after the remount
389 Details : clear dev_rdonly flag for external journal devices in
395 Description: shutdown vs evict race
396 Details : client_disconnect_export vs connect request race.
397 if client will evicted at this time - we start invalidate
398 thread without referece to import and import can be freed
404 Description: shrink LOV EAs before replying
405 Details : correctly adjust LOV EA buffer for reply.
410 Description: don't skip ost target if they assigned to file
411 Details : Drop slow OSCs if we can, but not for requested start idx.
412 This means "if OSC is slow and it is not the requested
413 start OST, then it can be skipped, otherwise skip it only
414 if it is inactive/recovering/out-of-space.
416 Severity : enhancement
418 Description: Update to RHEL5 kernel-2.6.18-92.1.17.el5.
421 Frequency : rare, need acl's on inode.
423 Description: client can't handle ost additional correctly
424 Details : if ost was added after client connected to mds client can have
425 hit lnet_try_match_md ... to big messages to wide striped files.
426 in this case need teach client to handle config events about add
427 lov target and update client max ea size at that event.
429 Severity : enhancement
431 Description: Update to sles9 kernel-2.6.5-7.314.
433 Severity : enhancement
435 Description: Update to SLES10 SP2 kernel-2.6.16.60-0.31.
438 Frequency : Create a symlink file with a very long name
440 Description: ldlm_cancel_pack()) ASSERTION(max >= dlm->lock_count + count)
441 Details : If there is no extra space in the request for early cancels,
442 ldlm_req_handles_avail() returns 0 instead of a negative value.
447 Description: mds is deadlocked
448 Details : In rare cases, inode in catalog can have i_no less than have parent
449 i_no, this produce wrong order for locking during open, and
450 parallel unlink can be lock open. this need teach mds_open to grab
451 locks in resource id order, not at parent -> child order.
453 Severity : enhancement
455 Description: Add /proc entry for import status
456 Details : The mdc, osc, and mgc import directories now have
457 an import directory that contains useful import data for debugging
460 Severity : enhancement
462 Description: Re-disable certain /proc logging
463 Details : Enable and disable client's offset_stats, extents_stats and
464 extents_stats_per_process stats logging on the fly.
467 Frequency : Only on FC kernels 2.6.22+
469 Description: oops in statahead
470 Details : Do not drop reference count for the dentry from VFS when lookup,
471 VFS will do that by itself.
473 Severity : enhancement
475 Description: Generic /proc file permissions
476 Details : Set /Proc file permissions in a more generic way to enable non-
477 root users operate on some /proc files.
481 Description: Hitting mdc_commit_close() ASSERTION
482 Details : Properly handle request reference release in
483 ll_release_openhandle().
487 Frequency : only patchless client
488 Description: add workaround for race between add/remove dentry from hash
490 Severity : enhancement
492 Description: Allow OST glimpses to return PW locks
496 Description: LBUG when llog conf file is full
497 Details : When llog bitmap is full, ENOSPC should be returned for plain
502 Description: Prevent import from entering FULL state when server in recovery
506 Description: service mount cannot take device name with ":"
507 Details : Only when device name contains ":/" will mount treat it as
513 Description: Replace ptlrpcd with the statahead thread to interpret the async
514 statahead RPC callback
518 Frequency : on recovery
519 Description: I/O failures after umount during fail back
520 Details : If client reconnected to restarted server we need join to recovery
521 instead of find server handler is changed and process self eviction
522 with cancel all locks.
524 Severity : enhancement
526 Description: Update to RHEL5 kernel-2.6.18-92.1.10.el5.
530 Description: Kernel BUG tries to release flock
531 Details : Lustre does not destroy flock lock before last reference goes
532 away. So always drop flock locks when client is evicted and
533 perform unlock regardless of successfulness of speaking to MDS.
535 Severity : enhancement
537 Description: Update to SLES10 SP2 kernel-2.6.16.60-0.27.
539 Severity : enhancement
541 Description: Upcall on Lustre log has been dumped
542 Details : Allow for a user mode script to be called once a Lustre log has
543 been dumped. It passes the filename of the dumped log to the
544 script, the location of the script can be specified via
545 /proc/sys/lnet/debug_log_upcall.
550 Description: avoid messages about idr_remove called for id that is not allocated
551 Details : Move assigment s_dev for clustered nfs to end of initialization,
552 for avoid problem with error handling.
557 Description: avoid Already found the key in hash [CONN_UNUSED_HASH] messages
558 Details : When connection is reused this not moved from CONN_UNUSED_HASH into
559 CONN_USED_HASH and this prodice warning when put connection again
565 Description: avoid ASSERTION(client_stat->nid_exp_ref_count == 0) failed
566 Details : Release reference to stats when client disconnected, not
567 when export destroyed for avoid races when client destroyed
568 after main ost export.
572 Description: more cleanup in mds_lov
573 Details : Add workaround for get valid ost count for avoid warnings about
574 drop too big messages, not init llog cat under semphore which
575 can be blocked on reconnect and break normal replay, fix access
578 Severity : enhancement
580 Description: File striping can now be set to use an arbitrary pool of OSTs.
582 Severity : enhancement
584 Description: Export bytes_read/bytes_write count on OSC/OST.
588 Description: Early reply size mismatch, MGC loses connection
589 Details : Apply the MGS_CONNECT_SUPPORTED mask at reconnect time so
590 the connect flags are properly negotiated.
594 Description: Properly propagate oinfo flags from lov to osc for statfs
595 Details : Restore missing copy oi_flags to lov requests.
599 Description: exports in /proc are broken
600 Details : recreate /proc entries for clients when they reconnect.
602 Severity : enhancement
604 Description: Add man pages for llobdstat(8), llstat(8), plot-llstat(8),
605 : l_getgroups(8), lst(8), routerstat(8)
606 Details : Included man pages for llobdstat(8), llstat(8),
607 : plot-llstat(8), l_getgroups(8), lst(8), routerstat(8)
609 Severity : enhancement
611 Description: Implement lustre ll_show_options method.
613 Severity : enhancement
615 Description: Update to SLES9 kernel-2.6.5-7.312.
617 Severity : enhancement
619 Description: Update to RHEL4 kernel-2.6.9-67.0.22.EL.
623 Description: exports in /proc are broken
624 Details : Recreate /proc entries for clients when they reconnect.
628 Description: don't fail open with -ERANGE
629 Details : If client connected until mds will be know about real ost count
630 get LOV EA can be fail because mds not allocate enougth buffer
635 Description: Resolve device initialization race
636 Details : Prevent proc handler from accessing devices added to the
637 obd_devs array but yet be intialized.
639 Severity : enhancement
641 Description: configure's --enable-quota should check the
642 kernel .config for CONFIG_QUOTA
643 Details : Configure is terminated if --enable-quota is passed but
644 no quota support is in kernel
646 Severity : enhancement
648 Description: Update to SLES10 SP2 kernel-2.6.16.60-0.23.
650 Severity : enhancement
652 Description: Update to RHEL5 kernel-2.6.18-92.1.6.el5.
656 Frequency : rare, on PPC clients
657 Description: don't swab ost objects in response about directory, because
659 Details : Bug similar bug 14856, but in different function.
661 Severity : enhancement
663 Description: lfs quota tool enhancement
664 Details : Added units specifiers support for setquota, default to
665 current uid/gid for quota report, short quota stats by
666 default, nonpositional parameters for setquota, added
667 llapi_quotactl manual page.
669 Severity : enhancement
671 Description: *optional* service tags registration
672 Details : If the "service tags" package is installed on a Lustre node
673 When the filesystem is mounted, a local-node service tag will
674 be created. See http://inventory.sun.com/ for more information
675 about the Service Tags asset management system.
679 Description: Client runs out of low memory
680 Details : Consider only lowmem when counting initial number of llap pages
683 Description: Kernel BUG tries to release flock
684 Details : Lustre does not destroy flock lock before last reference goes
685 away. So always drop flock locks when client is evicted and
686 perform unlock regardless of successfulness of speaking to MDS.
689 Frequency : occasional
691 Description: add refcount for osc callbacks, so avoid panic on shutdown
693 Severity : enhancement
695 Description: Update to RHEL4 kernel-2.6.9-67.0.20.
698 Frequency : testing only
700 Description: sanity test 65a fails if stripecount of -1 is set
701 Details : Handle -1 striping on filesystem in ll_dirstripe_verify
704 Frequency : only in unusual configurations
706 Description: Kernel panic with find ost index.
707 Details : lov_obd have panic if some OST's have sparse indexes.
709 Severity : enhancement
711 Description: Update to RHEL5 kernel-2.6.18-53.1.21.el5.
714 Frequency : rarely, if filesystem is mounted with -o flock
716 Description: do not process already freed flock
717 Details : Flock can possibly be freed by another thread before it reaches
718 to ldlm_flock_completion_ast.
721 Frequency : rarely, if filesystem is mounted with -o flock
723 Description: LBUG during stress test
724 Details : Need properly lock accesses the flock deadlock detection list.
727 Frequency : rarely, if binaries are being run from Lustre
729 Description: oops in page fault handler
730 Details : Kernel page fault handler can return two special 'pages' in
731 error case, don't try dereference NOPAGE_SIGBUS and NOPAGE_OMM.
734 Frequency : rarely, during shutdown
736 Description: timeout with invalidate import.
737 Details : ptlrpcd_check call obd_zombie_impexp_cull and wait request which
738 should be handled by ptlrpcd. This produce long age waiting and
739 -ETIMEOUT ptlrpc_invalidate_import and as result LASSERT.
745 Description: ASSERTION(CheckWriteback(page,cmd)) failed
746 Details : Badly clear PG_Writeback bit in ll_ap_completion can produce false
750 Frequency : only with broken builds/installations
752 Description: no LBUG if lquota.ko and fsfilt_ldiskfs.ko are different versions
753 Details : Just return an error to a user, put a console error message.
755 Severity : enhancement
757 Description: Update to RHEL5 kernel-2.6.18-53.1.19.el5.
759 Severity : enhancement
761 Description: Update to RHEL4 kernel-2.6.9-67.0.15.
763 Severity : enhancement
765 Description: enable MGS and MDT services start separately
766 Details : add a 'nomgs' option in mount.lustre to enable start a MDT with
767 a co-located MGS without starting the MGS, which is a complement
768 to 'nosvc' mount option.
771 Frequency : always, on big-endian systems
773 Description: cleanup in ptlrpc code, related to PPC platform
774 Details : store magic in native order avoid panic's in recovery on PPC
775 node and forbid from this error in future. Also fix posibily
776 of twice swab data. Fix get lov striping to userpace.
779 Frequency : rarely, if replay get lost on server
781 Description: server incorrectly drop resent replays lead to recovery failure.
782 Details : do not drop replay according to msg flags, instead we check the
783 per-export recovery request queue for duplication of transno.
786 Frequency : after recovery
788 Description: precreate to many object's after del orphan.
789 Details : del orphan st in oscc last_id == next_id and this triger growing
790 count of precreated objects. Set flag LOW to skip increase count
791 of precreated objects.
794 Frequency : rare, on clear nid stats
796 Description: ASSERTION(client_stat->nid_exp_ref_count == 0)
797 Details : when clean nid stats sometimes try destroy live entry,
798 and this produce panic in free.
801 Frequency : occasionally since 1.6.4
803 Description: Stack overflow during MDS log replay
804 Details : ease stack pressure by using a thread dealing llog_process.
807 Frequency : very rare
809 Description: MDT cannot be unmounted, reporting "Mount still busy"
810 Details : Mountpoint references were being leaked during open reply
811 reconstruction after an MDS restart. Drop mountpoint reference
812 in reconstruct_open() and free dentry reference also.
817 Description: wait until IO finished before start new when do lock cancel.
818 Details : VM protocol want old IO finished before start new, in this case
819 need wait until PG_writeback is cleared until check dirty flag
820 and call writepages in lock cancel callback.
825 Description: mds_mfd_close() ASSERTION(rc == 0)
826 Details : In mds_mfd_close(), we need protect inode's writecount change
827 within its orphan write semaphore to prevent possible races.
830 Frequency : rare, on shutdown ost
832 Description: don't hit live lock with umount ost.
833 Details : shrink_dcache_parent can be in long loop with destroy dentries,
834 use shrink_dcache_sb instead.
837 Frequency : only when echo_client is used
839 Description: don't panic with use echo_client
840 Details : echo client pass NULL as client nid pointer and this produce NULL
844 Frequency : Always on 32-bit PowerPC systems
846 Description: fix build on PPC32
847 Details : compile code with -m64 flag produce wrong object file for PPC32.
852 Description: MDS LBUG: ASSERTION(!IS_ERR(dchild))
853 Details : In reconstruct_* functions, LASSERTs on both the data supplied
854 by a client, and the data on disk are dangerous and incorrect.
855 Change them with client eviction.
857 Severity : enhancement
859 Description: skiplist implementation simplification
860 Details : skiplists are used to group compatible locks on granted list
861 that was implemented as tracking first and last lock of each lock
862 group the patch changes that to using doubly linked lists
866 Description: delete compatibility for 32bit qdata
867 Details : as planned, when lustre is beyond b1_8, lquota won't support 32bit
868 qunit. That means servers of b1_4 and servers of b1_8 can't be
869 used together if users want to use quota.
872 Frequency : only with administrator action
874 Description: mount failure if config log has invalid conf_param setting
875 Details : If administrator specified an incorrect configuration parameter
876 with "lctl conf_param" this would cause an error during future
877 client mounts. Instead, ignore the bad configuration parameter.
880 Frequency : blocks per group < blocksize*8 and uninit_groups is enabled
882 Description: ldiskfs error: XXX blocks in bitmap, YYY in gd
883 Details : If blocks per group is less than blocksize*8, set rest of the
887 Frequency : Application do stride read on lustre
889 Description: The read performance will drop a lot if the application does
891 Details : Because the stride_start_offset are missing in stride read-ahead,
892 it will cause clients read a lot of unused pages in read-ahead,
893 then the read-performance drops.
897 Description: more ldlm soft lockups
898 Details : In ldlm_resource_add_lock(), call to ldlm_resource_dump()
899 starve other threads from the resource lock for a long time in
900 case of long waiting queue, so change the debug level from
901 D_OTHER to the less frequently used D_INFO.
903 Severity : enhancement
905 Description: add -gid, -group, -uid, -user options to lfs find
907 Severity : enhancement
909 Description: ll_recover_lost_found_objs - recover objects in lost+found
910 Details : OST corruption and subsequent e2fsck can leave objects in the
911 lost+found directory. Using the "ll_recover_lost_found_objs"
912 tool, these objects can be retrieved and data can be salvaged
913 by using the object ID saved in the fid EA on each object.
918 Description: this bug _only_ happens when inode quota limitation is very low
919 (less than 12), so that inode quota unit is 1 at initialization.
920 Details : if remaining quota equates 1, it is a sign to demonstate that quota
921 is effective now. So least quota qunit should be 2.
925 Description: Hung threads in invalidate_inode_pages2_range
926 Details : The direct IO path doesn't call check_rpcs to submit a new RPC once
927 one is completed. As a result, some RPCs are stuck in the queue
932 Description: Procfs and llog threads access destoryed import sometimes.
933 Details : Sync the import destoryed process with procfs and llog threads by
934 the import refcount and semaphore.
938 Description: mds fails to respond, threads stuck in ldlm_completion_ast
939 Details : Sort source/child resource pair after updating child resource.
944 Description: kernel BUG at ldiskfs2_ext_new_extent_cb
945 Details : If insertion of an extent fails, then discard the inode
946 preallocation and free data blocks else it can lead to duplicate
951 Description: don't always update ctime in ext3_xattr_set_handle()
952 Details : Current xattr code updates inode ctime in ext3_xattr_set_handle()
953 In some cases the ctime should not be updated, for example for
954 2.0->1.8 compatibility it is necessary to delete an xattr and it
955 should not update the ctime.
959 Description: add quota statistics
960 Details : 1. sort out quota proc entries and proc code.
961 2. add quota statistics
966 Description: quotas are not honored with O_DIRECT
967 Details : all writes with the flag O_DIRECT will use grants which leads to
968 this problem. Now using OBD_BRW_SYNC to guard this.
972 Bugzilla : 15713/16362
973 Description: Assertion in iopen_connect_dentry in 1.6.3
974 Details : looking up an inode via iopen with the wrong generation number can
975 populate the dcache with a disconneced dentry while the inode
976 number is in the process of being reallocated. This causes an
977 assertion failure in iopen since the inode's dentry list contains
978 both a connected and disconnected dentry.
982 Description: assertion failure in ldlm_handle2lock()
983 Details : fix a race between class_handle_unhash() and class_handle2object()
984 introduced in lustre 1.6.5 by bug 13622.
986 Severity : enhancement
988 Description: superblock lock contention with many SMP cores on one client
989 Details : several client filesystem locks were highly contended on SMP
990 NUMA systems with 8 or more cores. Per-CPU datastructures
991 and more efficient locking implemented to reduce contention.
996 Description: Kernel BUG: sd_iostats_bump: unexpected disk index
997 Details : remove the limit of 256 scsi disks in the sd_iostat patch
1002 Description: oops in sd_iostats_seq_show()
1003 Details : unloading/reloading the scsi low level driver triggers a kernel
1004 bug when trying to access the sd iostat file.
1009 Description: Kernel panics during QLogic driver reload
1010 Details : REQ_BLOCK_PC requests are not handled properly in the sd iostat
1011 patch, causing memory corruption.
1016 Description: journal_dev option does not work in b1_6
1017 Details : pass mount option during pre-mount.
1019 Severity : enhancement
1021 Description: Add a FIEMAP(FIle Extent MAP) ioctl for ldiskfs
1022 Details : FIEMAP ioctl will allow an application to efficiently fetch the
1023 extent information of a file. It can be used to map logical blocks
1024 in a file to physical blocks in the block device.
1027 Frequency : only with adaptive timeout enabled
1029 Description: DEBUG_REQ() bad paging request
1030 Details : ptlrpc_at_recv_early_reply() should not modify req->rq_repmsg
1031 because it can be accessed by reply_in_callback() without the
1035 Frequency : only on Cray X2
1037 Description: X2 build failures
1038 Details : fix build failures on Cray X2.
1042 Description: xid & resent requests
1043 Details : Initialize RPC XID from clock at startup (randomly if clock is
1048 Description: quota recovery deadlock during mds failover
1049 Details : This patch includes att18982, att18236, att18237 in bz14840.
1051 1. fix osts hang when mds does failover with quotaon
1052 2. prevent watchdog storm when osts threads wait for the
1057 Description: kernel panic on racer
1058 Details : Do not access dchild->d_inode when IS_ERR(dchild) is true.
1060 Severity : enhancement
1062 Description: Add lustre_start utility to start or stop multiple Lustre servers
1067 Description: Lustre GPF in {:ptlrpc:ptlrpc_server_free_request+373}
1068 Details : In case of memory pressure, list_del() can be called twice on
1069 req->rq_history_list, causing a kernel oops.
1073 Description: (ptllnd_peer.c:557:kptllnd_peer_check_sends()) ASSERTION(!in_interrupt()) failed
1074 Details : fix stack overflow in the distributed lock manager by defering
1075 export eviction after a failed ast to the elt thread instead of
1076 handling it in the dlm interpret routine.
1078 Severity : enhancement
1080 Description: More exported tunables for mballoc
1081 Details : Add support for tunable preallocation window and new tunables for
1082 large/small requests
1086 Description: Detect corruption of block bitmap and checking for preallocations
1087 Details : Checks validity of on-disk block bitmap. Also it does better
1088 checking of number of applied preallocations. When corruption is
1089 found, it turns filesystem readonly to prevent further corruptions.
1093 Frequency : only for big-endian servers
1094 Description: Check for big-endian system while mounting fs with extents feature
1095 Details : Mounting a filesystem with extents feature will fail on big-endian
1096 systems since ext3-based ldiskfs is not supported on big-endian
1097 systems. Can be over-ridden with "bigendian_extents" mount option.
1101 Description: Excessive recovery window
1102 Details : With AT enabled, the recovery window can be excessively long (6000+
1103 seconds). To address this problem, we no longer use
1104 OBD_RECOVERY_FACTOR when extending the recovery window (the connect
1105 timeout no longer depends on the service time, it is set to
1106 INITIAL_CONNECT_TIMEOUT now) and clients report the old service
1107 time via pb_service_time.
1111 Description: Watchdog triggered on MDS failover
1112 Details : enable OBD_CONNECT_MDT flag when connecting from the MDS so that
1113 the OSTs know that the MDS "UUID" can be reused for the same export
1114 from a different NID, so we do not need to wait for the export to
1119 Descriptoin: Lustre detected file system corruption with inode out of bounds
1120 Details : don't update i_size on MDS_CLOSE for directories. This causes
1121 directory corruptions on the MDT.
1123 -------------------------------------------------------------------------------
1125 2008-05-26 Sun Microsystems, Inc.
1127 * Support for kernels:
1128 2.6.5-7.311 (SLES 9),
1129 2.6.9-67.0.7.EL (RHEL 4),
1130 2.6.16.54-0.2.5 (SLES 10),
1131 2.6.18-53.1.14.el5 (RHEL 5),
1132 2.6.22.14 vanilla (kernel.org)
1133 * Client support for unpatched kernels:
1134 (see http://wiki.lustre.org/index.php?title=Patchless_Client)
1135 2.6.16 - 2.6.22 vanilla (kernel.org)
1136 * Due to problems with nested symlinks and FMODE_EXEC (bug 12652),
1137 we do not recommend using patchless RHEL4 clients with kernels
1138 prior to 2.6.9-55EL (RHEL4U5).
1139 * Recommended e2fsprogs version: 1.40.7-sun1
1140 * Note that reiserfs quotas are disabled on SLES 10 in this kernel.
1141 * RHEL 4 and RHEL 5/SLES 10 clients behaves differently on 'cd' to a
1142 removed cwd "./" (refer to Bugzilla 14399).
1143 * A new quota file format has been introduced in 1.6.5.
1144 The format conversion from prior releases is handled transparently,
1145 but releases older than 1.4.12/1.6.5 will not understand this new
1146 format. The automatic format conversion can be avoided by running
1147 the following command on the MDS before upgrading:
1148 'tunefs.lustre --param="mdt.quota_type=ug1" $MDTDEV'.
1149 For more information, please refer to bugzilla 13904.
1153 Description: quota performance fix
1154 Details : quota data is written in journalled mode instead of ordered to
1155 increase performance
1159 Description: lfs support for human-readable quota grace time strings
1160 Details : lfs setquota -t and lfs quota -t represent quota grace times
1161 in "XXwXXdXXhXXmXXs" format instead of large values in seconds
1164 Frequency : always with o2ib 1.3 and sles10
1166 Description: fix build with SLES10 and o2ib v3.
1167 Details : sles10 uses diffrent name for Module.symver file but configure
1168 assume this file has same name on RHEL/SLES/vanila kernels.
1171 Frequency : very rare, if additional xattrs are used on kernels >= 2.6.12
1173 Description: MDS may lose file striping (and hence file data) in some cases
1174 Details : If there are additional extended attributes stored on the MDS,
1175 in particular ACLs, SELinux, or user attributes (if user_xattr
1176 is specified for the client mount options) then there is a risk
1177 of attribute loss. Additionally, the Lustre file striping
1178 needs to be larger than default (e.g. striped over all OSTs),
1179 and an additional attribute must be stored initially in the
1180 inode and then increase in size enough to be moved to the
1181 external attribute block (e.g. ACL growing in size) for file
1184 Severity : enhancement
1186 Description: add message levels for liblustreapi
1191 Description: MDT cannot be unmounted, reporting "Mount still busy"
1192 Details : Mountpoint references were being leaked during open reply
1193 reconstruction after an MDS restart. Drop mountpoint reference
1194 in reconstruct_open() and free dentry reference also.
1199 Description: fix for occasional failure case of -ENOSPC in recovery-small tests
1200 Details : Move the 'good_osts' check before the 'total_bavail' check. This
1201 will result in an -EAGAIN and in the exit call path we call
1202 alloc_rr() which will with increasing aggressiveness attempt to
1203 aquire precreated objects on the minimum number of required OSCs.
1207 Description: Use old size assignment to avoid deadlock
1208 Details : This reverts the changes in bugs 2369 and bug 14138 that introduced
1209 the scheduling while holding a spinlock. We do not need locking
1210 for size in ll_update_inode() because size is only updated from
1211 the MDS for directories or files without objects, so there is no
1212 other place to do the update, and concurrent access to such inodes
1213 are protected by the inode lock.
1217 Description: Use __u64 instead of int for valid bits
1221 Description: resolve "_IOWR redefined" build error on SLES10
1225 Description: dump the memory debugging after all modules are unloaded to
1226 suppress false negative in conf_sanity test 39
1230 Description: the recovery timer never expires
1231 Details : for new client connect request, the recovery timer should not be
1232 reset, otherwise recovery timer will never expired, if the old
1233 client never come. Only old client connect and first connection
1234 req should trigger recovery timer reset.
1238 Description: the min numbers of lproc stats are wrong
1239 Details : adding a new constant LC_MIN_INIT and use it for initialization
1243 Frequency : always with interactive lfs
1245 Description: Reinitialize optind to 0 so that interactive lfs works in all cases
1248 Frequency : with multiple concurrent readdir processes in same directory
1249 Bugzilla : 15406, 15169, 15175
1250 Description: misc fixes for directory readahead.
1251 Details : prevent previous statahead async RPC callback from processing the
1252 current "statahead_info", race condition between async RPC callback
1253 add dentry into dentry hash table and "ls" thread revalidate such
1254 dentry, statahead his/miss control for hidden items, and so on.
1256 Severity : enhancement
1258 Description: build kernel-ib packages for OFED 1.3 in our release cycle
1262 Description: incore types cleaning in quota code (with respect to 64-bit limits)
1263 Details : several u32 variables declarations are replaced with u64 declarations
1268 Description: fix SLES kernel versioning
1269 Details : the kernel version for our SLES 10 kernel did not include a "-"
1270 before the "smp" at the end. while this was not a problem in
1271 general, it did mean that software trying to use the kernel
1272 version to try to detect a vendor specific kernel would fail.
1273 this was most evident by the OFED build scripts.
1278 Description: Don't update lov_desc members until making sure they are valid
1279 Details : When updating lov_desc members via proc fs, need fix their
1280 validities before doing the real update.
1283 Frequency : very rare
1285 Description: don't put request into delay list while invalidate in flight.
1286 Details : ptlrpc_delay_request sometimes put in delay list while invalidate
1287 import in flight. this produce timeout for invalidate and sometimes
1288 can cause stale data.
1290 Severity : enhancement
1292 Description: Update kernel to SLES9 2.6.5-7.311.
1294 Severity : enhancement
1296 Description: Update kernel to RHEL4 2.6.9-67.0.7.
1301 Frequency : on PPC only
1302 Description: not convert ost objects for directory because it's not exist.
1303 Details : ll_dir_getstripe assume dirrectory has ost objects but this wrong.
1305 Severity : enhancement
1307 Description: Fix warnings with compile liblustre at sles10/rhel5 which have
1308 __u64 as usingied long long type.
1311 Frequency : rare, on shutdown
1313 Description: race process ast vs remove callback
1314 Details : removing callback before disconnect import open race with
1315 processing callback.
1317 Severity : enhancement
1319 Description: Update kernel to SLES9 2.6.5-7.311.
1321 Severity : enhancement
1323 Description: Files open for execute are not marked busy on SLES10
1324 Details : Add FMODE_EXEC to SLES10 SP1 server kernel series.
1326 Severity : enhancement
1328 Description: Add server support for vanilla-2.6.22.14.
1331 Frequency : occasional
1333 Description: Avoid lov_create() getting stuck in obd_statfs_rqset()
1334 Details : If an OST is down the MDS will hang indefinitely in
1335 obd_statfs_rqset() waiting for the statfs data. While for
1336 MDS QOS usage of statfs, it should not stuck in waiting.
1338 Severity : enhancement
1340 Description: Disable adaptive timeouts by default
1343 Frequency : on network error
1345 Description: panic with double free request if network error
1346 Details : mdc_finish_enqueue is finish request if any network error occuring,
1347 but it's true only for synchronus enqueue, for async enqueue
1348 (via ptlrpcd) this incorrect and ptlrpcd want finish request
1352 Frequency : rare, on recovery
1354 Description: read procfs can produce deadlock in some situation
1355 Details : Holding lprocfs lock with send rpc can produce block for destroy
1356 obd objects and this also block reconnect with -EALREADY.
1357 This isn't fix all lprocfs bugs - but make it rare.
1359 Severity : enhancement
1361 Description: Update kernel to RHEL5 2.6.18-53.1.14.el5.
1364 Frequency : frequent on X2 node
1366 Description: mdc_set_open_replay_data LBUG
1367 Details : Set replay data for requests that are eligible for replay.
1372 Description: lustre_mgs: operation 101 on unconnected MGS
1373 Details : When MGC is disconnected from MGS long enough, MGS will evict the
1374 MGC, and late on MGC cannot successfully connect to MGS and a lot
1375 of the error messages complaining that MGS is not connected.
1378 Frequency : rare, depends on device drivers and load
1380 Description: MDS or OSS nodes crash due to stack overflow
1381 Details : Code changes in 1.6.4 increased the stack usage of some functions.
1382 In some cases, in conjunction with device drivers that use a lot
1383 of stack, the MDS (or possibly OSS) service threads could overflow
1384 the stack. One change which was identified to consume additional
1385 stack has been reworked to avoid the extra stack usage.
1387 Severity : enhancement
1389 Description: Update to RHEL5 latest kernel-2.6.18-53.1.13.el5.
1391 Severity : enhancement
1393 Description: Update to SLES10 SP1 latest kernel-2.6.16.54-0.2.5.
1395 Severity : enhancement
1397 Description: Update to RHEL5 latest kernel-2.6.18-53.1.6.el5.
1399 Serverity : enhancement
1401 Description: Update RHEL4 kernel to 2.6.9-67.0.4.
1404 Frequency : rare on shutdown OST
1406 Description: Don't allow skipping OSTs if index has been specified.
1407 Details : Don't allow skipping OSTs if index has been specified, make locking
1408 in internal create lots better.
1413 Description: ASSERTION(!PageDirty(page)) failed
1414 Details : Wrong check could lead to an assertion failure under specific
1420 Description: LBUG in ptlrpc_check_set() bad phase ebc0de00
1421 Details : access to bitfield in structure is always rounded to long
1422 and this produce problem with not atomic change any bit.
1427 Description: Lustre make rpms failed.
1428 Details : Remove ldiskfs spec file to avoids rpmbuild be confused when
1429 builds Lustre rpms from tarball.
1431 Severity : enhancement
1433 Description: Update to SLES9 SP4 kernel-2.6.5-7.308.
1436 Frequency : rare on shutdown OST
1438 Description: If llog cancel was not send before clean_exports phase, this can
1439 produce deadlock in llog code.
1440 Details : If llog thread has last reference to obd and call class_import_put
1441 this produce deadlock because llog_cleanup_commit_master wait when
1442 last llog_commit_thread exited, but this never success because was
1443 called from llog_commit_thread.
1446 Frequency : only if OST index is skipped
1448 Description: NULL lov_tgts causing MDS oops
1449 Details : more safe checks for NULL lov_tgts for avoid oops.
1451 Severity : enhancement
1453 Description: Update to RHEL4 latest kernel-2.6.9-67.0.1.EL.
1455 Severity : enhancement
1457 Description: Update to RHEL5 latest kernel-2.6.18-53.1.4.el5.
1462 Description: make mgs_setparam() handle fsname containing dash
1463 Details : fsname containing a dash does not work with lctl conf_param
1465 Severity : enhancement
1467 Description: Update to RHEL4 Update-6 kernel-2.6.9-67.EL.
1470 Frequency : rare, in recovery and (or) destroy lovobjid file.
1472 Description: rewrite lov objid code.
1473 Details : Cleanup for lov objid code, remove scability problems and wrong
1474 locking. Fix sending last_id into ost.
1476 Severity : enhancement
1478 Description: Update to SLES10 SP1 latest kernel-2.6.16.54-0.2.3.
1480 Severity : enhancement
1482 Description: Update to RHEL5 Update-1 kernel 2.6.18-53.el5.
1483 Details : Use d_move_locked instead of __d_move.
1486 Frequency : rare, at shutdown
1488 Description: access already free / zero obd_namespace.
1489 Details : if client_disconnect_export was called without force flag set,
1490 and exist connect request in flight, this can produce access to
1491 NULL pointer (or already free pointer) when connect_interpret
1492 store ocd flags in obd_namespace.
1495 Frequency : only at startup
1497 Description: not alloc memory with spinlock held.
1498 Details : allocation memory with GFP_KERNEL can produce sleep deadlock,
1499 if any spinlock held.
1504 Description: lfs find does not continue on file error
1505 Details : Continue other files processing when a file/dir is absent.
1509 Description: Inconsistent usage of lustre_pack_reply()
1510 Details : Standardize the usage of lustre_pack_reply() such that it
1511 always generate a CERROR on failure.
1514 Frequency : very rare
1516 Description: Fix replay if there is an un-replied request and open
1517 Details : In some cases, older replay request will revert the
1518 mcd->mcd_last_xid on MDS which is used to record the client's
1519 latest sent request.
1521 Severity : enhancement
1523 Description: Update to RHEL5 kernel 2.6.18-8.1.15.el5.
1525 Severity : enhancement
1527 Description: Update to SLES10 SP1 kernel 2.6.16.53-0.16
1529 Severity : enhancement
1531 Description: Update to SLES9 kernel-2.6.5-7.287.3.
1533 Severity : enhancement
1535 Description: Update to RHEL4 kernel-2.6.9-55.0.12.EL.
1537 Severity : enhancement
1539 Description: Build SLES10 patchless client fails
1540 Details : The configure was broken by run ./configure with
1541 --with-linux-obj=.... argument for patchless client. When the
1542 configure use --with-linux-obj, the LINUXINCLUDE= -Iinclude
1543 can't search header adequately. Use absolute path such as
1544 -I($LINUX)/include instead.
1546 Severity : enhancement
1548 Description: Lustre Page Accounting
1549 Details : New macros for page alloc and free which enable accounting
1550 of page allocation of Lustre. Use percpu counters to store memory
1551 and page statistics.
1554 Frequency : only if debugging is disabled
1556 Description: LASSERT_{REQ,REP}SWAB macros are buggy
1557 Details : If SWAB_PARANOIA is disabled, the LASSERT_REQSWAB and
1558 LASSERT_REPSWAB macros become no-ops, which is incorrect. Drop
1559 these macros and replace them with their definitions instead.
1564 Description: interrupt oig_wait produce painc on resend.
1565 Details : brw_redo_request can be used for resend requests from ptlrpcd and
1566 private set, and this produce situation when rq_ptlrpcd_data not
1567 copyed to new allocated request and triggered LBUG on assert
1568 req->rq_ptlrpcd_data != NULL. But this member used only for wakeup
1569 ptlrpcd set if request is changed and can be safety changed to use
1572 Severity : enhancement
1574 Description: organize the server-side client stats on per-nid basis
1575 Details : Change the structure of stats under obdfilter and mds to
1583 The "uuid"s file would list the uuids of _active_ exports.
1584 And the clear entry is to clear all stats and stale nids.
1589 Description: Processes looping in ll_readdir() on Lustre clients finally causing
1590 a full node pseudo-hang
1591 Details : Concurrent access to the same directory from multiple clients with
1592 intensive file creation/removal can cause a client node to spin in
1593 ll_readdir(). i_version must be increased every time the lock is
1594 cancelled to ensure a revalidate is done.
1599 Description: touch file failed when fs is not full
1600 Details : OST in recovery should not be discarded by MDS in alloc_qos(),
1601 otherwise we can get ENOSP while fs is not full.
1604 Frequency : only for Cray XT3
1605 Bugzilla : 12829/13455
1606 Description: Changing primary group doesn't change the group lustre assigns to
1608 Details : When CRAY_XT3 is defined, the fsgid supplied by the client is
1609 overridden with the primary group provided by the group upcall,
1610 whereas the supplied fsgid can be trusted if it is in the list of
1611 supplementary groups returned by the group upcall.
1613 Severity : enhancement
1615 Description: Root Squash Functionality
1616 Details : Implementation of NFS-like root squash capability. Specifically,
1617 don't allow someone with root access on a client node to be able
1618 to manipulate files owned by root on a server node.
1620 Severity : enhancement
1622 Description: Slow trucate/writes to huge files at high offsets.
1623 Details : Directly associate cached pages to lock that protect those pages,
1624 this allows us to quickly find what pages to write and remove
1625 once lock callback is received.
1630 Description: Too many locks accumulating on client during NFS usage
1631 Details : mds_open improperly used accmode to find out access mode to a
1632 file. Also mdc_intent_lock logic to find out if we already have
1633 lock similar to just received was flawed since introduction of
1634 skiplists - locks are now added to the front of the granted
1639 Description: Hit ASSERTION(obd->obd_stopping == 1) failed in some setup failed
1641 Details : In obd setup failure handler, obd_stopping will not necessarily to
1642 be 1, and obd_set_up should also be checked to make sure whether
1643 obd is completely setup.
1645 Severity : enhancement
1647 Description: Allow masking D_WARNING, D_ERROR messages from console
1648 Details : Console messages can now be disabled via lnet.printk.
1653 Description: User code with malformed file open parameter crashes client node
1654 Details : Before packing join_file req, all the related reference should be
1655 checked carefully in case some malformed flags cause fake join_file
1661 Description: shrink/enlarge qunit size when needed; fix the problem of coarse
1662 grain of quota doing harm to quota's accuracy
1663 Details : qunit size will be changed when quota limitation is too low/high;
1664 record the pending quota write in order to get more accureate
1665 quota; delete the patch for bug12588, which is unnecessary when
1666 this patch is landed. This bug also contains fixes for bug 14526,
1667 14299, 14601 and 13794.
1671 Description: LDLM_ENQUEUE races with LDLM_CP_CALLBACK
1672 Details : ldlm_completion_ast() assumes that a lock is granted when the req
1673 mode is equal to the granted mode. However, it should also check
1674 that LDLM_FL_CP_REQD is not set.
1678 Description: Heavy nfs access might result in deadlocks
1679 Details : After ELC code landed, it is now improper to enqueue any mds
1680 locks under och_sem, because enqueue might want to decide to
1681 cancel open locks for same inode we are holding och_sem for.
1685 Description: 35% write performance drop with ldiskfs2 when quotas are on
1686 Details : Enable ext3 journalled quota by default to improve performance
1687 when quotas are turned on.
1691 Description: Client eviction while running blogbench
1692 Details : A lot of unlink operations with concurrent I/O can lead to a
1693 deadlock causing evictions. To address the problem, the number of
1694 oustanding OST_DESTROY requests is now throttled to
1695 max_rpcs_in_flight per OSC and LDLM_FL_DISCARD_DATA blocking
1696 callbacks are processed in priority.
1699 Frequency : RHEL4 only
1701 Description: mkfs is very slow on IA64/RHEL4
1702 Details : A performance regression has been discovered in the MPT Fusion
1703 driver between versions 3.02.73rh and 3.02.99.00rh. As a
1704 consequence, we have downgraded the MPT Fusion driver in the RHEL4
1705 kernel from 3.02.99.00 to 3.02.73 until this problem is fixed.
1708 Frequency : PPC/PPC64 only
1710 Description: conflicts between asm-ppc64/types.h and lustre_types.h
1711 Details : fix duplicated definitions between asm-ppc64/types.h and
1712 lustre_types.h on PPC.
1715 Frequency : PPC/PPC64 only
1717 Description: asm-ppc/segment.h does not exist
1718 Details : fix compile issue on PPC.
1722 Description: data checksumming impacts single node performance
1723 Details : add support for several checksum algorithms. Currently, CRC32 and
1724 Adler-32 are supported. The checksum type can be changed on the fly
1725 through /proc/fs/lustre/osc/*/checksum_type.
1729 Description: use adler32 for page checksums
1730 Details : when available, use the Adler-32 algorithm instead of CRC32 for
1735 Description: better handle error messages in extents code
1737 Severity : enhancement
1739 Description: SNMP support enhancement
1740 Details : Adding total number of sampled request for an MDS node in snmp
1743 Severity : enhancement
1745 Description: Optimize ldlm waiting list processing for PR extent locks
1746 Details : When processing waiting list for read extent lock and meeting read
1747 lock that is same or wider to it that is not contended, skip
1748 processing rest of the list and immediatelly return current
1749 status of conflictness, since we are guaranteed there are no
1750 conflicting locks in the rest of the list.
1754 Description: Time out and refuse to reconnect
1755 Details : When the failover node is the primary node, it is possible
1756 to have two identical connections in imp_conn_list. We must
1757 compare not conn's pointers but NIDs, otherwise we
1758 can defeat connection throttling.
1762 Description: Client not clear own cache if answer to reconnect is lost.
1763 Details : Client gets evicted from server. Now client also thinks it is
1764 disconnected (or gets ENOTCONN on its operation) and decides to
1765 reconnect. Server receives reconnect message, but cannot find
1766 export. New export is created that is fully valid (new cookie!),
1767 but reply is lost and not reported to client. Client reconnects
1768 again and gets back a just-created connection, but it is not new
1769 so client thinks it was not evicted and does not do recovery.
1773 Description: Detect stride IO mode in read-ahead
1774 Details : When a client does stride read, read-ahead should detect that and
1775 read-ahead pages according to the detected stride pattern.
1779 Description: build for x2 fails
1780 Details : fix compile issue on Cray systems.
1782 Severity : enhancement
1784 Description: implement readv/writev APIs(aio_read/aio_writes in newer kernels)
1785 Details : This greatly improves speed of NFS writes on 2.6 kernels.
1788 Frequency : only on PPC/SLES10
1790 Description: "BITS_PER_LONG is not 32 or 64" in linux/idr.h
1791 Details : On SLES10/PPC, fs.h includes idr.h which requires BITS_PER_LONG to
1792 be defined. Add a hack in mkfs_lustre.c to work around this compile
1797 Description: LASSERT on MDS when client holding flock lock dies
1798 Details : ldlm pool logic depends on number of granted locks equal to
1799 number of released locks which is not true for flock locks, so
1800 just exclude such locks from consideration.
1804 Description: MDS deadlock with many ll_sync_lov threads and I/O stalled
1805 Details : Use fsfilt_sync() for both the whole filesystem sync and
1806 individual file sync to eliminate dangerous inode locking
1807 with I_LOCK that can lead to a deadlock.
1811 Description: Update an obsolete wirecheck.c generator
1812 Details : Update wirecheck.c/wirehdr.c and regenerate wiretest.c
1816 Description: Client can panic on open sometimes
1817 Details : It is possible that we try to free already freed request in
1818 ll_file_open in some error cases when we send request from
1823 Description: performance in 1.6.3
1824 Details : Force q->max_phys_segments to MAX_PHYS_SEGMENTS on SLES10 to be
1825 sure that 1MB requests are not fragmented by the block layer.
1829 Description: LDLM soft lockups - improvement
1830 Details : It is be possible to send the lock handle along with each read
1831 or write request because the client is already doing a lock match
1832 itself so there isn't any reason the OST should have to re-do that
1838 Description: lfs quota fails with deactivated OSTS
1839 Details : With this patch, three improvements are included:
1840 1. detete the softlimit in mds and osts when use "lfs quota".
1841 2. display the inaccurate data in the output of "lfs quota".
1842 3. try to get quota info when "lfs quota" is executed.
1847 Description: Extent locks not granted with no conflicts sometimes.
1848 Details : When race occurs in glimpse handler and nothing is returned,
1849 we do not reprocess the queue after lock cancel, and that leads
1850 to a stall until next activity on a resource
1853 Frequency : failover with quotaon
1855 Description: during mds failovers with quota on, OSTs got into deadlock state
1856 and causing dumpstack.
1857 Details : for every quota slave, at any time, there is only one quota req
1858 is sent to quota master for every uid/gid. Before that quota req
1859 returns, all the thread relative to the same uid/gid will wait.
1860 So if the quota req is lost because mds failovers or any other
1861 reasons, this bug will be hit. Now, dqacq_interpret() will handle
1862 quota reqs who time out.
1864 Severity : enhancement
1867 Description: when quota slave checks if quota is enough, there is an unnecessary
1869 Details : place this wait on necessary place instead of always waiting.
1871 --------------------------------------------------------------------------------
1873 2007-12-07 Cluster File Systems, Inc. <info@clusterfs.com>
1875 * Support for kernels:
1876 2.6.5-7.286 (SLES 9),
1877 2.6.9-55.0.9.EL (RHEL 4),
1878 2.6.16.53-0.8 (SLES 10),
1879 2.6.18-8.1.14.el5 (RHEL 5),
1880 2.6.18.8 vanilla (kernel.org)
1881 * Client support for unpatched kernels:
1882 (see http://wiki.lustre.org/index.php?title=Patchless_Client)
1883 2.6.16 - 2.6.22 vanilla (kernel.org)
1884 * Due to recently discovered recovery problems, we do not recommend
1885 using patchless RHEL 4 clients with this or any earlier release.
1886 * Recommended e2fsprogs version: 1.40.2-cfs1
1887 * Note that reiserfs quotas are disabled on SLES 10 in this kernel.
1890 Frequency : occasional
1892 Description: MDS hang or stay in waiting lock
1893 Details : If client receive lock with CBPENDING flag ldlm need send lock
1894 cancel as separate rpc, to avoid situation when cancel request
1895 can't processed due all i/o threads stay in wait lock.
1898 Frequency : occasional
1900 Description: Do not fail import if osc_interpret_create gets -EAGAIN
1901 Details : If osc_interpret_create got -EAGAIN it immediately exits and
1902 wakeup oscc_waitq. After wakeup oscc_wait_for_objects call
1903 oscc_has_objects and see OSC has no objests and call
1904 oscc_internal_create to resend create request.
1906 Severity : enhancement
1908 Description: Update kernel patches for SLES10 2.6.16.53-0.8.
1909 Details : Update which_patch & target file for SLES10 latest kernel.
1911 Severity : enhancement
1913 Description: add --type and --size parameters to lfs find
1914 Details : Enhance lfs find by adding filetype and filesize parameters. Also
1915 multiple OBDs can now be specified for the --obd option.
1917 Severity : enhancement
1919 Description: eliminate client locks in face of contention
1920 Details : file contention detection and lockless i/o implementation
1921 for contended files.
1923 Severity : enhancement
1925 Description: Remove client patches from SLES 10 kernel.
1926 Details : This causes SLES 10 clients to behave as patchless clients
1927 even on a Lustre-patched (server) kernel.
1929 Severity : enhancement
1931 Description: use i_size_read and i_size_write in 2.6 port
1932 Details : replace inode->i_size access with i_size_read/write()
1935 Frequency : when removing large files
1937 Description: scheduling issue during removal of large Lustre files
1938 Details : Don't take the BKL in fsfilt_ext3_setattr() for 2.6 kernels.
1939 It causes scheduling issues when removing large files (17TB in the
1945 Description: 1.4.11 Can't handle directories with stripe set and extended ACLs
1946 Details : Impossible (EPROTO is returned) to access a directory that has a
1947 non-default striping and ACLs.
1950 Frequency : only on ppc
1952 Description: /proc/fs/lustre/devices broken on ppc
1953 Details : The patch as applied to 1.6.2 doesn't look correct for all arches.
1954 We should make sure the type of 'index' is loff_t and then cast
1955 explicitly as needed below. Do not assign an explicitly cast
1959 Frequency : only for rhel5
1961 Description: Kernel patches update for RHEL5 2.6.18-8.1.10.el5.
1962 Details : Modify the target file & which_kernel.
1965 Frequency : if the uninit_groups feature is enabled on ldiskfs
1967 Description: e2fsck reports "invalid unused inodes count"
1968 Details : If a new ldiskfs filesystem is created with the "uninit_groups"
1969 feature and only a single inode is created in a group then the
1970 "bg_unused_inodes" count is incorrectly updated. Creating a
1971 second inode in that group would update it correctly.
1976 Description: buffer overruns could theoretically occur
1977 Details : llapi_semantic_traverse() modifies the "path" argument by
1978 appending values to the end of the origin string, and a buffer
1979 overrun may occur. Adding buffer overrun check in liblustreapi.
1981 Severity : enhancement
1983 Description: Add jbd statistics patch for RHEL5 and 2.6.18-vanilla.
1986 Frequency : only if filesystem is inconsistent
1988 Description: handle "serious error: objid * already exists" more gracefully
1989 Details : If LAST_ID value on disk is smaller than the objects existing in
1990 the O/0/d* directories, it indicates disk corruption and causes an
1991 LBUG(). If the object is 0-length, then we should use the existing
1992 object. This will help to avoid a full fsck in most cases.
1994 Severity : enhancement
1996 Description: Kernel patches update for RHEL4 2.6.9-55.0.6.
1997 Details : Modify vm-tunables-rhel4.patch.
1999 Severity : enhancement
2001 Description: Kernel config for 2.6.18-vanilla.
2002 Details : Modify targets/2.6-vanilla.target.in.
2003 Add config file kernel-2.6.18-2.6-vanilla-i686.config.
2004 Add config file kernel-2.6.18-2.6-vanilla-i686-smp.config.
2005 Add config file kernel-2.6.18-2.6-vanilla-x86_64.config.
2006 Add config file kernel-2.6.18-2.6-vanilla-x86_64-smp.config.
2009 Frequency : occasional
2011 Description: improve handling recoverable errors
2012 Details : If request processed with error which can be recoverable on server
2013 request should be resend, otherwise page released from cache and
2019 Description: Kernel patches update for RHEL5 2.6.18-8.1.14.el5.
2020 Details : Modify target file & which_patch.
2021 A flaw was found in the IA32 system call emulation provided
2022 on AMD64 and Intel 64 platforms. An improperly validated 64-bit
2023 value could be stored in the %RAX register, which could trigger an
2024 out-of-bounds system call table access. An untrusted local user
2025 could exploit this flaw to run code in the kernel
2026 (ie a root privilege escalation). (CVE-2007-4573).
2030 Description: change order of libsysio includes
2031 Details : '#include sysio.h' should always come before '#include xtio.h'
2033 Severity : enhancement
2035 Description: adapt the lustre_config script to support the upgrade case
2036 Details : Add "-u" option for lustre_config script to support upgrading 1.4
2037 server targets to 1.6 in parallel.
2042 Description: To avoid grant space > avaible space when the disk is almost
2043 full. Without this patch you might see the error "grant XXXX >
2044 available" or some LBUG about grant, when the disk is almost
2046 Details : In filter_check_grant, for non_grant cache write, we should
2047 check the left space by if (*left > ungranted + bytes), instead
2048 of (*left > ungranted), because only we are sure the left space
2049 is enough for another "bytes", then the ungrant space should be
2050 increase. In client, we should update cl_avail_grant only there is
2051 OBD_MD_FLGRANT in the reply.
2056 Description: Update RHEL 4 kernel to fix local root privilege escalation.
2057 Details : Update to the latest RHEL 4 kernel to fix the vulnerability
2058 described in CVE-2007-4573. This problem could allow untrusted
2059 local users to gain root access.
2062 Frequency : when using O_DIRECT and quotas
2064 Description: Incorrect file ownership on O_DIRECT output files
2065 Details : block usage reported by 'lfs quota' does not take into account
2066 files that have been written with O_DIRECT.
2070 Description: (rw.c:1323:ll_read_ahead_pages()) ASSERTION(page_idx > ria->ria_stoff) failed
2071 Details : Once the unmatched stride IO mode is detected, shrink the stride-ahead
2072 window to 0. If it does hit cache miss, and read-pattern is still
2073 stride-io mode, does not reset the stride window, but also does not
2074 increase the stride window length in this case.
2076 --------------------------------------------------------------------------------
2078 2007-09-27 Cluster File Systems, Inc. <info@clusterfs.com>
2080 * Support for kernels:
2081 2.6.5-7.286 (SLES 9),
2082 2.6.9-55.0.2.EL (RHEL 4),
2083 2.6.16.46-0.14 (SLES 10),
2084 2.6.18-8.1.8.el5 (RHEL 5),
2085 2.6.18.8 vanilla (kernel.org)
2086 * Client support for unpatched kernels:
2087 (see http://wiki.lustre.org/index.php?title=Patchless_Client)
2088 2.6.16 - 2.6.21 vanilla (kernel.org)
2089 * Due to recently discovered recovery problems, we do not recommend
2090 using patchless RHEL 4 clients with this or any earlier release.
2091 * Recommended e2fsprogs version: 1.40.2-cfs1
2092 * Note that reiserfs quotas are disabled on SLES 10 in this kernel.
2096 Description: Fix errors in lfs documentation
2097 Details : Fixes man pages
2099 Severity : enhancement
2101 Description: Adaptive timeouts
2102 Details : RPC timeouts adapt to changing server load and network
2103 conditions to reduce resend attempts and improve recovery time.
2105 Severity : enhancement
2107 Description: llapi_file_create() does not allow some changes
2108 Details : add llapi_file_open() that allows specifying the file creation
2109 mode and open flags, and also returns an open file handle.
2111 Severity : enhancement
2113 Description: df doesn't work properly if diskfs blocksize != 4K
2114 Details : Choose biggest blocksize of OST's as the LOV's blocksize.
2117 Frequency : always on directories with default striping set
2119 Description: lfs find on -1 stripe looping in lsm_lmm_verify_common()
2120 Details : Avoid lov_verify_lmm_common() on directory with -1 stripe count.
2122 Severity : enhancement
2124 Description: merge and cleanup kernel patches.
2125 Details : Remove mnt_lustre_list in vfs_intent-2.6-rhel4.patch.
2127 Severity : enhancement
2129 Description: RedHat Update kernel for RHEL5
2130 Details : Kernel config file for RHEL5.
2132 Severity : enhancement
2134 Description: OSS needs mutliple precreate threads
2135 Details : Add ability to start more than one create thread per OSS.
2138 Frequency : only with quota on the root user
2140 Description: mds_obd_create error creating tmp object
2141 Details : When the user sets quota on root, llog will be affected and can't
2142 create files and write files.
2145 Frequency : Always on ia64 patchless client, and possibly others.
2147 Description: Add EXPORT_SYMBOL check for node_to_cpumask symbol.
2148 Details : This allows the patchless client to be loaded on architectures
2149 without this export.
2154 Description: disorder of journal start and llog_add cause deadlock.
2155 Details : in llog_origin_connect, journal start should happen before
2156 llog_add keep the same order as other functions to avoid
2159 Severity : enhancement
2161 Description: RedHat Update kernel for RHEL5
2162 Details : Modify the kernel config file more closer RHEL5.
2164 Severity : enhancement
2166 Description: Build failure against Centos5 (RHEL5)
2167 Details : Define PAGE_SIZE when it isn't present.
2170 Frequency : occasionally when using NFS
2172 Description: "ll_intent_file_open()) lock enqueue: err: -13" with nfs
2173 Details : with NFS, the anon dentry's parent was set to itself in
2174 d_alloc_anon(), so in MDS, we use rec->ur_fid1 to find the
2175 corresponding dentry other than use rec->ur_name.
2177 Severity : enhancement
2179 Description: enable data checksumming by default
2180 Details : enable checksum by default, allow --disable-checksum
2181 configure option and "-o nochecksum" mount option. Checksums
2182 can also be disabled at runtime via $LPROC/osc/*/checksum_pages.
2185 Frequency : Occasionally with failover
2187 Description: Client eviction due to failover config
2188 Details : after a connection loss, the lustre client should attempt to
2189 reconnect to the last active server first before trying the
2190 other potential connections.
2192 Severity : enhancement
2194 Description: client-side metadata stat-ahead during readdir(directory readahead)
2195 Details : perform client-side metadata stat-ahead when the client detects
2196 readdir and sequential stat of dir entries therein
2201 Description: when mds and osts use different quota unit(32bit and 64bit),
2202 quota will be released repeatly.
2203 Details : void sending multiple quota reqs to mds, which will keep the
2204 status between the reqs.
2207 Frequency : only with liblustre clients on XT3
2209 Description: evictions taking too long
2210 Details : allow llrd to evict clients directly on OSTs
2214 Description: osts not allocated evenly to files
2215 Details : change the condition to increase offset_idx
2217 Severity : enhancement
2219 Description: self-adjustable client's lru lists
2220 Details : use adaptive algorithm for managing client cached locks lru
2221 lists according to current server load, other client's work
2222 pattern, memory activities, etc. Both, server and client
2223 side namespaces provide number of proc tunables for controlling
2228 Description: rewrite ext2-derived code in llite/dir.c and obdclass/uuid.c
2229 Details : rewrite inherited code (uuid parsing code from ext2 utils and
2230 readdir code from ext3) from scratch preserving functionality.
2234 Description: Only those disconnect error should be returned by rq_status.
2235 Details : In open/enqueue processs, Some errors, which will cause client
2236 disconnected, should be returned by rq_status, while other
2237 errors should still be returned by intent, then mdc or llite will
2240 Severity : enhancement
2242 Description: Tune the kernel for good SCSI performance.
2243 Details : Set the value of /sys/block/{dev}/queue/max_sectors_kb
2244 to the value of /sys/block/{dev}/queue/max_hw_sectors_kb
2248 Frequency : Always for filesystems larger than 2TB on 32-bit systems.
2249 Bugzilla : 13547, 13627
2250 Description: Data corruption for OSTs that are formatted larger than 2TB
2252 Details : When generating the bio request for lustre file writes the
2253 sector number would overflow a temporary variable before being
2254 used for the IO. The data reads correctly from Lustre (which
2255 will overflow in a similar manner) but other file data or
2256 filesystem metadata may be corrupted in some cases.
2260 Description: "lfs find -obd UUID" prints directories
2261 Details : "lfs find -obd UUID" will return all directory names instead
2262 of just file names. It is incorrect because the directories
2263 do not reside on the OSTs.
2266 Bugzilla : 13596, 13614
2267 Description: MDS hang after unclean shutdown of lots of clients
2268 Details : Never resend AST requests.
2272 Frequency : Always, for kernels after 2.6.16
2273 Description: Fix warning idr_remove called for id=.. which is not allocated.
2274 Details : Last kernels save old s_dev before kill super and not allow
2275 to restore from callback - restore it before call kill_anon_super.
2279 Description: Fix error on 'ls .' at the top of the Lustre mount.
2280 Details : Don't revalidate dentry if it is a root dentry.
2282 --------------------------------------------------------------------------------
2284 2007-08-27 Cluster File Systems, Inc. <info@clusterfs.com>
2286 * Support for kernels:
2287 2.6.5-7.286 (SLES 9),
2288 2.6.9-55.0.2.EL (RHEL 4),
2289 2.6.16.46-0.14 (SLES 10),
2290 2.6.18-8.1.8.el5 (RHEL 5),
2291 2.6.18.8 vanilla (kernel.org)
2292 * Client support for unpatched kernels:
2293 (see http://wiki.lustre.org/index.php?title=Patchless_Client)
2294 2.6.16 - 2.6.22 vanilla (kernel.org)
2295 * Due to recently discovered recovery problems, we do not recommend
2296 using patchless RHEL 4 clients with this or any earlier release.
2297 * Recommended e2fsprogs version: 1.39.cfs8
2298 * Note that reiserfs quotas are disabled on SLES 10 in this kernel.
2303 Description: block reactivating mgc import until all deactivates complete
2304 Details : Fix race when failing back MDT/MGS to itself (testing)
2306 Severity : enhancement
2308 Description: lfs setstripe enhancement
2309 Details : Make lfs setstripe understand 'k', 'm' and 'g' for stripe size.
2312 Frequency : mds/oss recovery
2314 Description: llog ctxt is refrenced after it has been freed.
2315 Details : llog ctxt refcount was added to avoide the race between ctxt free
2316 and llog recovery process. Each llog user must hold ctxt refcount
2317 before it access the llog. And the llog ctxt can only be freed
2318 when its refcount is zero.
2320 Severity : enhancement
2322 Description: randomly memory allocation failure util
2323 Details : Make lustre randomly failed allocating memory for testing purpose.
2326 Frequency : only for SLES10
2328 Description: Update kernel patch for SLES10 SP1
2329 Details : Add patch blkdev_tunables-2.6-sles10.patch to 2.6-sles10.series.
2331 Severity : enhancement
2333 Description: omit set fsid for export NFS
2334 Details : fix set/restore device id for avoid EMFILE error and mark lustre fs
2335 as FS_REQUIRES_DEV for avoid problems with generate fsid.
2338 Frequency : after network failures
2340 Description: Add sync option to mount_lustre.c
2341 Details : Client loses data written to lustre after a network interruption.
2343 Severity : enhancement
2345 Description: Error message improvement.
2346 Details : Merging of two LCONSOLE_ERROR_MSG into one.
2348 Severity : enhancement
2350 Description: don't use GFP_* in generic Lustre code.
2351 Details : Use cfs_alloc_* functions and CFS_* flags for code portability.
2353 Severity : enhancement
2355 Description: obdclass is limited by single OBD_ALLOC(idarray)
2356 Details : replace OBD_ALLOC/OBD_FREE with OBD_VMALLOC/OBD_VFREE
2358 Severity : enhancement
2360 Description: updated patchess for new RHEL4 kernel
2361 Details : Fixed ext3-unlink-race.patch per Kalpak's comment.
2363 Severity : enhancement
2365 Description: warnings with build patchless client with vanila 2.6.19 and up
2366 Details : change old ctl_table style and replace ctl_table/ctl_table_header
2367 with cfs_sysctl_table_t/cfs_sysctl_table_header_t
2369 Severity : enhancement
2371 Description: O_DIRECT bypasses client statistics.
2372 Details : When running with O_DIRECT I/O, neither the client rpc_stats nor
2373 read_ahead_stats were updated. Copied stats section from
2374 osc_send_oap_rpc() into async_internal().
2377 Frequency : only for Cray XT3
2379 Description: peer credits not enough on many OST per OSS systems.
2380 Details : Use new lnet way to add credits as we need those for pings and ASTs
2383 Frequency : only with liblustre
2385 Description: Liblustre is not releasing flock locks on file close.
2386 Details : Release flock locks on file close.
2389 Frequency : only for RHEL4
2391 Description: Update kernel patches for kernel-2.6.9-55.0.2.EL
2392 Details : Remove inode-nr_unused-2.6.9-rhel4.patch from 2.6-rhel4.series
2393 Update target file and kernel config.
2397 Description: lustre support for RHEL5
2398 Details : Add support for RHEL5.
2403 Description: ASSERTION(export != NULL) failed in target_handle_connect
2404 Details : Assetion hit is result of rare race between disconnect and connet
2405 to same nid. target_handle_connect found old connect cockie and
2406 tried to reconnect, but can't find export for this cockie.
2411 Description: umount blocks forever on error
2412 Details : In result of wrong using obd_no_recov and obd_force flags client
2413 can hand if cancel or some other requests is lost.
2416 Frequency : Only for SLES
2418 Description: sanity_quota fail test_1
2419 Details : There are multiple occurences of $TSTUSR in SLES's /etc/group
2420 file, which makes TSTID[2] inunique.
2422 Severity : enhancement
2424 Description: Kernel patches for SLES9 2.6.5-7.286 kernel
2425 Details : Update target/ChangeLog/which_patch .
2427 Severity : enhancement
2429 Description: jbd statistics
2430 Details : Port older jbd statistics patch for sles10
2435 Description: lvbo_init failed for resource with missing objects.
2436 Details : Fix returning error if we do stat for file with missing/corrupted
2437 objects and i_size set to all sum of size of all avaible objects.
2438 if we to truncate/write to missing object - it is recreated.
2443 Description: Oops in read and write path when failing to allocate lock.
2444 Details : Check if lock allocation failed and return error back.
2447 Frequency : When flocks are used.
2449 Description: assertion failure in ldlm_cli_enquque_fini for non NULL lock.
2450 Details : Flock locks might destroy just granted lock if it could be merged
2451 with another existing flock, this is done in completion handler,
2452 so teach ldlm_cli_enquque_fini that this is a valid case for
2458 Description: reply_lock_interpret crash due to race with it and lock cancel.
2459 Details : Do not replay locks that are being cancelled. Do not reference
2460 locks by their address during replay, just by their handle.
2463 Frequency : only with deactivated OSTs
2465 Description: lstripe command fails for valid OST index
2466 Details : The stripe offset is compared to 'lov->desc.ld_tgt_count'
2467 instead of lov->desc.ld_active_tgt_count.
2469 Severity : enhancement
2471 Description: Build failure against Centos5 (RHEL5)
2472 Details : Use getpagesize() instead of PAGE_SIZE.
2474 --------------------------------------------------------------------------------
2476 2007-07-30 Cluster File Systems, Inc. <info@clusterfs.com>
2478 * Support for kernels:
2479 2.6.5-7.283 (SLES 9),
2480 2.6.9-55.EL (RHEL 4),
2481 2.6.16.46-0.14 (SLES 10),
2482 2.6.18.8 vanilla (kernel.org)
2483 * Client support for unpatched kernels:
2484 (see http://wiki.lustre.org/index.php?title=Patchless_Client)
2485 2.6.16 - 2.6.22 vanilla (kernel.org)
2486 * Due to recently discovered recovery problems, we do not recommend
2487 using patchless RHEL 4 clients with this or any earlier release.
2488 * Recommended e2fsprogs version: 1.39.cfs8
2489 * Note that reiserfs quotas are disabled on SLES 10 in this kernel.
2490 * Starting with this release, the ldiskfs backing filesystem required
2491 by Lustre is now in its own package, lustre-ldiskfs. This package
2492 should be installed. It is versioned separately from Lustre and
2493 may be released separately in future.
2498 Description: server LBUG when shutting down
2499 Details : Block umount forever until the mount refcount is zero rather
2500 than giving up after an arbitrary timeout.
2502 Severity : enhancement
2504 Description: add optional extra BUILD_VERSION info
2505 Details : add a new environment variable (namely LUSTRE_VERS) which allows
2506 to override the lustre version.
2509 Frequency : 2.6.18 servers only
2511 Description: ll_kern_mount() doesn't release the module reference
2512 Details : The ldiskfs module reference count never drops down to 0
2513 because ll_kern_mount() doesn't release the module reference.
2518 Description: server LBUG when using old ost_num_threads parameter
2519 Details : Accept the old ost_num_threads parameter but warn that it
2520 is deprecated, and fix an off-by-one error that caused an LBUG.
2525 Description: Transient SCSI error results in persistent IO issue
2526 Details : iobuf->dr_error is not reinitialized to 0 between two
2530 Frequency : sometimes when underlying device returns I/O errors
2532 Description: OSTs not going read-only during write failures
2533 Details : OSTs are not remounted read-only when the journal commit threads
2534 get I/O errors because fsfilt_ext3 calls journal_start/stop()
2535 instead of the ext3 wrappers.
2538 Frequency : only with 10000 clients or more
2540 Description: poor connect scaling with increasing client count
2541 Details : Don't run filter_grant_sanity_check for more than 100 exports
2542 to improve scaling for large numbers of clients.
2545 Frequency : SLES10 only
2547 Description: sanity-quota.sh quotacheck failed: rc = -22
2548 Details : Quotas cannot be enabled on SLES10.
2551 Frequency : liblustre clients only
2553 Description: getdirentries does not give error when run on compute nodes
2554 Details : getdirentries does not fail when the size specified as an argument
2555 is too small to contain at least one entry
2557 Severity : enhancement
2559 Description: Add LNET router traceability for debug purposes
2560 Details : If a checksum failure occurs with a router as part of the
2561 IO path, the NID of the last router that forwarded the bulk data
2562 is printed so it can be identified.
2567 Description: OST "spontaneously" evicts client; client has imp_pingable == 0
2568 Details : Due to a race condition, liblustre clients were occasionally
2569 evicted incorrectly.
2571 Severity : enhancement
2573 Description: lfs setstripe use optional parameters instead of postional
2576 Severity : enhancement
2578 Description: Nanosecond timestamp support for ldiskfs
2579 Details : The on-disk ldiskfs filesystem has added support for nanosecond
2580 resolution timestamps. There is not yet support for this at
2581 the Lustre filesystem level.
2584 Frequency : during server recovery
2586 Description: MDS failing to send precreate requests due to OSCC_FLAG_RECOVERING
2587 Details : request with rq_no_resend flag not awake l_wait_event if they get a
2591 Frequency : nfs export on patchless client
2593 Description: connectathon hang when test nfs export over patchless client
2594 Details : Disconnected dentry cannot be found with lookup, so we do not need
2595 to unhash it or make it invalid
2598 Description: fix llapi_lov_get_uuids() to allow many OSTs to be returned
2599 Details: : Change llapi_lov_get_uuids() to read the UUIDs from /proc instead
2600 of using an ioctl. This allows lfsck for > 160 OSTs to succeed.
2605 Description: open req refcounting wrong on reconnect
2606 Details : If reconnect happened between getting open reply from server and
2607 call to mdc_set_replay_data in ll_file_open, we will schedule
2608 replay for unreferenced request that we are about to free.
2609 Subsequent close will crash in variety of ways.
2610 Check that request is still eligible for replay in
2611 mdc_set_replay_data().
2616 Description: disable writes to filesystem when reading health_check file
2617 Details : the default for reading the health_check proc file has changed
2618 to NOT do a journal transaction and write to disk, because this
2619 can cause reads of the /proc file to hang and block HA state
2620 checking on a healthy but otherwise heavily loaded system. It
2621 is possible to return to the previous behaviour during configure
2622 with --enable-health-write.
2624 Severity : enhancement
2626 Description: 64-bit inode version
2627 Details: : Add a on-disk 64-bit inode version for ext3 to track changes made
2628 to the inode. This will be required for version-based recovery.
2633 Description: MDS fails to start if a duplicate client export is detected
2634 Details : in some rare cases it was possible for a client to connect to
2635 an MDS multiple times. Upon recovery the MDS would detect this
2636 and fail during startup. Handle this more gracefully.
2638 Severity : enhancement
2640 Description: Add -o localflock option to simulate old noflock behaviour.
2641 Details : This will achieve local-only flock/fcntl locks coherentness.
2646 Description: log_commit_thread vs filter_destroy race leads to crash
2647 Details : Take import reference before releasing llog record semaphore
2652 Description: Wrong request locking in request set processing
2653 Details : ptlrpc_check_set wrongly uses req->rq_lock for proctect add to
2654 imp_delayed_list, in this place should be used imp_lock.
2657 Frequency : when reconnecting
2659 Description: Grant leak when OSC reconnect to OST
2660 Details : When osc reconnect ost, OST(filter) should check whether it
2661 should grant more space to client by comparing fed_grant and
2662 cl_avail_grant, and return the granted space to client instead
2663 of "new granted" space, because client will call osc_init_grant
2664 to update the client grant space info.
2667 Frequency : when client reconnects to OST
2669 Description: Grant leak when OSC does a resend and replays bulk write
2670 Details : When osc reconnect to OST, OST(filter) should clear grant info of
2671 bulk write request, because the grant info while be sync between
2672 OSC and OST when reconnect, and we should ignore the grant info
2673 these of resend/replay write req.
2678 Description: Grant space more than avaiable space sometimes.
2679 Details : When then OST is about to be full, if two bulk writing from
2680 different clients came to OST. Accord the avaliable space of the
2681 OST, the first req should be permitted, and the second one
2682 should be denied by ENOSPC. But if the seconde arrived before
2683 the first one is commited. The OST might wrongly permit second
2684 writing, which will cause grant space > avaiable space.
2687 Frequency : when client is evicted
2689 Description: Grant might be wrongly erased when osc is evicted by OST
2690 Details : when the import is evicted by server, it will fork another
2691 thread ptlrpc_invalidate_import_thread to invalidate the
2692 import, where the grant will be set to 0. While the original
2693 thread will update the grant it got when connecting. So if
2694 the former happened latter, the grant will be wrongly errased
2695 because of this race.
2700 Description: Checking Stale with correct fid
2701 Details : ll_revalidate_it should uses de_inode instead of op_data.fid2
2702 to check whether it is stale, because sometimes, we want the
2703 enqueue happened anyway, and op_data.fid2 will not be initialized.
2705 Severity : enhancement
2707 Description: update patchless client
2708 Details : Add support for patchless client with 2.6.20, 2.6.21 and RHEL 5
2711 Frequency : only with 2.4 kernel
2713 Description: random memory corruption
2714 Details : size of struct ll_inode_info is to big for union inode.u and this
2715 can be cause of random memory corruption.
2720 Description: Memory leak in recovery
2721 Details : Lov_mds_md was not free in an error handler in mds_create_object.
2722 It should also check obd_fail before fsfilt_start, otherwise if
2723 fsfilt_start return -EROFS,(failover mds during mds recovery).
2724 then the req will return with repmsg->transno = 0 and rc = EROFS.
2725 and we met hit the assert LASSERT(req->rq_reqmsg->transno ==
2726 req->rq_repmsg->transno) in ptlrpc_replay_interpret. Fcc should
2727 be freed no matter whether fsfilt_commit success or not.
2730 Frequency : only with large numbers of cores on a single node
2732 Description: Prevents from taking the superblock lock in llap_from_page for
2734 Details : using LL_ORIGIN_REMOVEPAGE origin flag instead of LL_ORIGIN_UNKNOW
2735 for llap_from_page call in ll_removepage() prevents from taking
2736 the superblock lock for a soon died page.
2741 Description: Not check open intent error before release open handle
2742 Details : in some rare cases, the open intent error is not checked before
2743 release open handle, which may cause
2744 ASSERTION(open_req->rq_transno != 0), because it tries to release
2745 the failed open handle.
2750 Description: Set cat log bitmap only after create log success.
2751 Details : in some rare cases, the cat log bitmap is set too early. and it
2752 should be set only after create log success.
2756 Description: Accessing a block bevice can re-enable I/O when Lustre is
2757 tearing down a device.
2758 Details : dev_clear_rdonly(bdev) must be called in kill_bdev() instead of
2763 Description: service threads may hog cpus when there are a lot of requests
2764 Details : Insert cond_resched to give other threads a chance to use some CPU
2769 Description: the cat log was not initialized in recovery
2770 Details : When mds(mgs) do recovery, the tgt_count might be zero, so the
2771 unlink log on mds will not be initialized until mds post
2772 recovery. And also in mds post recovery, the unlink log will
2773 initialization will be done asynchronausly, so there will be race
2774 between add unlink log and unlink log initialization.
2778 Description: brw_stats were being printed incorrectly
2779 Details : brw_stats were being printed as log2 but all of them were not
2780 recorded as log2. Also remove some code duplication arising from
2781 filter_tally_{read,write}.
2785 Frequency : rare, only in recovery.
2786 Description: ASSERTION(req->rq_type != LI_POISON) failed
2787 Details : imp_lock should be held while iterating over imp_sending_list for
2788 prevent destroy request after get timeout in ptlrpc_queue_wait.
2792 Description: replay-single.sh test 52 fails
2793 Details : A lock's skiplist need to be cleanup when it being unlinked
2794 from its resource list.
2799 Description: Short directio read returns full requested size rather than
2801 Details : Direct I/O operations should return actual amount of bytes
2802 transferred rather than requested size.
2804 Severity : enhancement
2806 Description: metadata RPC reduction (e.g. for rm performance)
2807 Details : decrease the amount of synchronous RPC between clients and servers
2808 by canceling conflicing lock before the operation on the client
2809 and packing thier handles into the main operation RPC to server.
2811 Severity : enhancement
2813 Description: add #ifdef HAVE_KERNEL_CONFIG_H
2814 Details : kernels from 2.6.19 not need include linux/config.h, but add
2815 include linux/autoconf.h in commpiler command line.
2817 Severity : enhancement
2819 Description: patchless client support for 2.6.22 kernel
2820 Details : 2.6.22 has only one visble change, SLAB_CTOR_* constants is
2821 removed. In this case we need drop using os depended interface to
2822 kmem_cache and use cfs_mem_cache API.
2827 Description: fix mal-formatted messages
2828 Details : fix some mal-formatted DEBUG_REQ and LCONSOLE_ERROR_MSG messages
2832 Frequency : always in liblustre
2833 Description: wrong IS_ERR implementation in liblustre.h
2834 Details : fix IS_ERR implementation in liblustre.h for right detect errors.
2839 Description: Correct condition for output debug message.
2840 Details : inode i_nlink equal zero is not enough for output message about
2841 disk corruption, i_ctime and i_mode should be also checked.
2845 Frequency : always in patchless client
2846 Description: add configure check for truncate_complete_page
2847 Details : improve checks for exported symbols. This allow run check without
2848 sources, but with Module.symvers shipped with kernel distribution.
2849 add check for truncate_complete_page used by patchless client.
2853 Description: sanity.sh test_77h fails with "test_77h file compare failed"
2854 Details : test_77h uses a file which was messed by other test case.
2858 Description: Not Check whether lov_tgts is NULL in some lov functions
2859 Details : Checking whether lov_tgts is NULL in some functions.
2863 Description: replace obdo_alloc() with OBDO_ALLOC macro
2864 Details : nothing special is done in obdo_alloc() function, and for
2865 debugging purpose, it needs to be replaced with macros.
2869 Description: bad return value and errno from fcntl call
2870 Details : In liblustre API, errno should be a negative value if error
2875 Description: ptlrpc_check_set() LBUG
2876 Details : In case of positive reply from server and failed client bulk
2877 callback after bulk transfer shouldn't LBUG, but process this
2878 request as erroneous.
2880 Severity : enhancement
2882 Description: VFS operations stats tool.
2883 Details : Tool which collects stats by tracking value written in pid,
2884 ppid, gid and uses llstat to generate output to plot graph using
2886 Updated lustre/utils/Makefile.am
2887 Added lustre/utils/ltrack_stats.c
2889 Severity : enhancement
2891 Description: 2.6.18 server support (lustre 1.6.1)
2892 Details : Support for 2.6.18 kernels on the server side.
2897 Description: ASSERTION(imp->imp_conn_current) failed
2898 Details : an assertion failure is hit if a client node boots and attempts to
2899 mount a lustre filesystem faster than RECONNECT_INTERVAL seconds.
2902 Frequency : only for i686
2904 Description: 1.4.11 RC1 build fails for RHEL 4, i686
2905 Details : Fixed config variable for build.
2910 Description: Updated patchess for new RHEL4 kernel
2911 Details : Updated patch inode-nr_unused-2.6.9-rhel4.patch
2912 Updated patch jbd-stats-2.6.9.patch
2913 Updated patch qsnet-rhel4-2.6.patch
2914 Updated patch quota-deadlock-on-pagelock-core.patch
2915 Updated patch vfs_intent-2.6-rhel4.patch
2916 Updated patch vfs_races-2.6-rhel4.patch
2917 Updated series file 2.6-rhel4-titech.series
2918 Updated series file 2.6-rhel4.series
2919 Updated kernel config files
2924 Description: lquota slave complains LBUG when reconnecting with mds
2926 Details : quota slave depends qctxt->lqc_import to send its quota request.
2927 This pointer will be invalid if mds did failover or broke its
2928 connect to osts, which leads to LBUG.
2931 Frequency : when qunit size is too small(less than 20M)
2933 Description: write is stopped by improper -EDQUOT
2934 Details : If the master is busy and qunit size is small enough(let's say 1M),
2935 the slave can not get quota from master on time, which will lead
2936 slave to trigger a -EQUOTA to client.
2941 Description: Deadlock during metadata tests
2942 Details : in prune_dir_dentries(), shrink_dcache_parent() should not be
2943 called with the per-dentry lock held.
2946 Frequency : SLES9 only
2948 Description: Lustre patched kernel for SLES9 SP3 has NR_CPUS set to 8
2949 Details : set CONFIG_NR_CPUS to 128 instead of 8.
2951 Severity : enhancement
2953 Description: remove fs_prep_san_write operation and related patches
2954 Details : remove the ext3-san-jdike patches which are no longer useful.
2959 Description: LDISKFS-fs error (device sdc): ldiskfs_free_blocks
2960 Details : a disk corruption can cause the mballoc code to assert on a
2961 double free or other extent corruptions. Handle these with
2962 ext3_error() instead of assertions.
2965 Frequency : only with mballoc3 code and deep extent trees
2967 Description: ldiskfs_ext_search_right: bad header in inode: unexpected eh_depth
2968 Details : a wrong check of extent headers in ldiskfs_ext_search_right()
2969 can cause the filesystem to be remounted read-only.
2974 Description: lfsck built against 1.4.x cannot run against 1.6.0 lustre
2975 Details : the definition for OBD_IOC_GETNAME changed in 1.6.0. One of the
2976 few external users of this ioctl number is lfsck's call to
2977 llapi_lov_get_uuids() and this caused lfsck to fail at startup.
2978 Add the old ioctl number to the handler so both old and new
2983 Description: parallel lock callbacks
2984 Details : Instead of sending blocking and completion callbacks as separated
2985 requests, adding them to a set and sending in parallel.
2989 Description: Disable most debugging by default
2990 Details : To improve performance, disable most logging (for debug purposes)
2991 by default. VFSTRACE, RPCTRACE, and DLMTRACE are now off by
2992 default, and HA includes fewer messages.
2996 Frequency : only run on patchless client.
2997 Description: use do_facet on sanity.sh for test handling recoverables errors
2998 Details : use do_facet instead of direct use sysctl for set fail_loc on OST
3002 Description: hash tables for lists of nids, connections and uuids
3003 Details : Hash tables noticeably help when a lot of clients connect to a
3004 server, to faster identify duplicate connections or reconnects,
3005 also to faster find export to evict in manual eviction case.
3009 Description: Sometimes, when the server evict a client, and the client will
3010 not be evicted as soon as possible.
3011 Details : In enqueue req, the error was returned by intent, instead of
3012 rq_status which make ptlrpc layer not detect this error, and
3013 does not evict the client. So enqueue error should be returned
3017 Frequency : only at startup
3019 Description: Delay client connections to MDT until fist MDT->OST connect
3020 Details : If a client tried to create a new file before the MDT had
3021 connected to any OSTs, the create would return EIO. Now
3022 the client will simply block until the MDT connects to the
3023 first OST and the create can succeed.
3026 Frequency : only for SLES9
3028 Description: Routinely utilize latest Quadrics drivers in CFS releases
3029 Details : Update patch qsnet-suse-2.6.patch.
3032 Frequency : only for sles10
3034 Description: Update patches for SLES 10 SP1 kernel.
3035 Details : Update the patch vfs_intent-2.6-sles10.patch.
3040 Description: Routinely utilize latest Quadrics drivers in CFS releases
3041 Details : Update patch qsnet-rhel4-2.6.patch.
3044 Frequency : at statup only
3046 Description: mds_lov_synchronize race leads to various problems
3047 Details : simultaneous MDT->OST connections at startup can cause the
3048 sync to abort, leaving the OSC in a bad state.
3053 Description: Using wrong pointer in osc_brw_prep_request
3054 Details : Access to array[-1] can produce panic if kernel compiled with
3055 CONFIG_PAGE_ALLOC enabled
3057 Severity : enhancement
3059 Description: Async OSC create to avoid the blocking unnecessarily.
3060 Details : If a OST has no remain object, system will block on the creating
3061 when need to create a new object on this OST. Now, ways use
3062 pre-created objects when available, instead of blocking on an
3063 empty osc while others are not empty. If we must block, we block
3064 for the shortest possible period of time.
3068 Frequency : only in recovery
3069 Description: Mark OST as early accessible if his start SYNC.
3070 Details : osc_precreate return flag early accessible if oscc marked as
3071 OSCC_FLAG_SYNC_IN_PROGRESS.
3076 Description: Sometimes precreate code can triger create object on wrong ost
3077 Details : Wrong protected or not not restored variables aftre precreate loop
3078 can produce creation object on wrong ost.
3081 Frequency : oss recovery
3083 Description: llog_commit_thread cleanup should sync with llog_commit_thread
3085 Details : llog_commit_thread_count should be synced between llog_commit
3086 start and cleanup, so new llog_commit thread should not be started
3087 when llog_commit threads being stopped to avoid accessing some
3090 Severity : enhancement
3092 Description: Add printing inode info into message about error in writepage.
3094 --------------------------------------------------------------------------------
3096 2007-05-03 Cluster File Systems, Inc. <info@clusterfs.com>
3101 Frequency : on some architectures
3103 Description: 1.6 client sometimes fails to mount from a 1.4 MDT
3104 Details : Uninitialized flags sometimes cause configuration commands to
3108 Frequency : patchless clients only
3110 Description: missing __iget() symbol export
3111 Details : The __iget() symbol export is missing. To avoid the need for
3112 this on patchless clients the deathrow inode reaper is turned
3113 off, and we depend on the VM to clean up old inodes. This
3114 dependency was during via the fix for bug 12181.
3119 Description: sanity.sh fail: test_52b
3120 Details : The ll_inode_to_ext_flags() has a glitch which makes MDS return
3121 incorrect inode's flags to client.
3123 --------------------------------------------------------------------------------
3125 2007-04-19 Cluster File Systems, Inc. <info@clusterfs.com>
3127 * CONFIGURATION CHANGE. This version of Lustre WILL NOT
3128 INTEROPERATE with older versions automatically. In many cases a
3129 special upgrade step is needed. Please read the
3130 user documentation before upgrading any part of a 1.4.x system.
3131 * WARNING: Lustre configuration and startup changes are required with
3132 this release. See https://mail.clusterfs.com/wikis/lustre/MountConf
3134 * Support for kernels:
3135 2.4.21-47.0.1.EL (RHEL 3),
3136 2.6.5-7.283 (SLES 9),
3137 2.6.9-42.0.10.EL (RHEL 4),
3138 2.6.12.6 vanilla (kernel.org),
3139 2.6.16.27-0.9 (SLES10)
3140 * Client support for unpatched kernels:
3141 (see https://mail.clusterfs.com/wikis/lustre/PatchlessClient)
3142 2.6.16 - 2.6.19 vanilla (kernel.org),
3143 2.6.9-42.0.8EL (RHEL 4)
3144 * Recommended e2fsprogs version: 1.39.cfs6
3145 * Note that reiserfs quotas are disabled on SLES 10 in this kernel
3148 Severity : enhancement
3150 Description: Async OSC create to avoid the blocking unnecessarily.
3151 Details : If a OST has no remain object, system will block on the creating
3152 when need to create a new object on this OST. Now, ways use
3153 pre-created objects when available, instead of blocking on an
3154 empty osc while others are not empty. If we must block, we block
3155 for the shortest possible period of time.
3157 Severity : enhancement
3159 Description: MountConf
3160 Details : Lustre configuration is now managed via mkfs and mount
3161 commands instead of lmc and lconf. New obd types (MGS, MGC)
3162 are added for dynamic configuration management. See
3163 https://mail.clusterfs.com/wikis/lustre/MountConf for
3166 Severity : enhancement
3168 Description: dynamic OST addition
3169 Details : OSTs can now be added to a live filesystem
3171 Severity : enhancement
3173 Description: startup order invariance
3174 Details : MDTs and OSTs can be started in any order. Clients only
3175 require the MDT to complete startup.
3177 Severity : enhancement
3179 Description: parallel, asynchronous orphan cleanup
3180 Details : orphan cleanup is now performed in separate threads for each
3181 OST, allowing parallel non-blocking operation.
3183 Severity : enhancement
3185 Description: optimized stripe assignment
3186 Details : stripe assignments are now made based on ost space available,
3187 ost previous usage, and OSS previous usage, in order to try
3188 to optimize storage space and networking resources.
3190 Severity : enhancement
3192 Description: Permanently set tunables
3193 Details : All writable /proc/fs/lustre tunables can now be permanently
3194 set on a per-server basis, at mkfs time or on a live
3197 Severity : enhancement
3199 Description: Lustre message v2
3200 Details : Add lustre message format v2.
3202 Severity : enhancement
3204 Description: client OST exclusion list
3205 Details : Clients can be started with a list of OSTs that should be
3206 declared "inactive" for known non-responsive OSTs.
3210 Description: ENOENT returned for valid filehandle during dbench.
3211 Details : Check if a directory has children when invalidating dentries
3212 associated with an inode during lock cancellation. This fixes
3213 an incorrect ENOENT sometimes seen for valid filehandles during
3214 testing with dbench.
3217 Frequency : SFS test only (otherwise harmless)
3219 Description: SPEC SFS validation failure on NFS v2 over lustre.
3220 Details : Changes the blocksize for regular files to be 2x RPC size,
3221 and not depend on stripe size.
3223 Severity : enhancement
3225 Description: fine-grained SMP locking inside DLM
3226 Details : Improve DLM performance on SMP systems by removing the single
3227 per-namespace lock and replace it with per-resource locks.
3229 Severity : enhancement
3231 Description: don't hold multiple extent locks at one time
3232 Details : To avoid client eviction during large writes, locks are not
3233 held on multiple stripes at one time or for very large writes.
3234 Otherwise, clients can block waiting for a lock on a failed OST
3235 while holding locks on other OSTs and be evicted.
3237 Severity : enhancement
3239 Description: Multiple MD RPCs in flight.
3240 Details : Further unserialise some read-only MDT RPCs - learn about intents.
3241 To avoid overly-overloading MDT, introduce a limit on number of
3242 MDT RPCs in flight for a single client and add /proc controls
3243 to adjust this limit.
3245 Severity : enhancement
3247 Description: client read/write statistics
3248 Details : Add client read/write call usage stats for performance
3249 analysis of user processes.
3250 /proc/fs/lustre/llite/*/offset_stats shows non-sequential
3251 file access. extents_stats shows chunk size distribution.
3252 extents_stats_per_process show chunk size distribution per
3255 Severity : enhancement
3257 Description: per-client statistics on server
3258 Details : Add ldlm and operations statistics for each client in
3259 /proc/fs/lustre/mds|obdfilter/*/exports/
3261 Severity : enhancement
3263 Description: improved MDT statistics
3264 Details : Add detailed MDT operations statistics in
3265 /proc/fs/lustre/mds/*/stats
3267 Severity : enhancement
3269 Description: VFS operations stats
3270 Details : Add client VFS call stats, trackable by pid, ppid, or gid
3271 /proc/fs/lustre/llite/*/stats_track_[pid|ppid|gid]
3276 Description: Fix client-side osc byte counters
3277 Details : The osc read/write byte counters in
3278 /proc/fs/lustre/osc/*/stats are now working
3281 Frequency : always as root on SLES
3283 Description: Failure of copying files with lustre special EAs.
3284 Details : Client side always return success for setxattr call for lustre
3285 special xattr (currently only "trusted.lov").
3290 Description: Refcount LNET uuids
3291 Details : The global LNET uuid list grew linearly with every startup;
3292 refcount repeated list entries instead of always adding to
3295 Severity : enhancement
3297 Description: Dynamic service threads
3298 Details : Within a small range, start extra service threads
3299 automatically when the request queue builds up.
3302 Frequency : mixed-endian client/server environments
3304 Description: mixed-endian crashes
3305 Details : The new msg_v2 system had some failures in mixed-endian
3308 Severity : enhancement
3310 Description: Easy OST removal
3311 Details : OSTs can be permanently deactivated with e.g. 'lctl
3312 conf_param lustre-OST0001.osc.active=0'
3314 Severity : enhancement
3316 Description: MGS proc entries
3317 Details : Added basic proc entries for the MGS showing what filesystems
3320 Severity : enhancement
3322 Description: provide MGS failover
3323 Details : Added config lock reacquisition after MGS server failover.
3325 Severity : enhancement
3327 Description: add Linux 2.4 support
3328 Details : Added support for RHEL 2.4.21 kernel for 1.6 servers and clients
3332 Description: a large application tries to do I/O to the same resource and dies
3333 in the middle of it.
3334 Details : Check the req->rq_arrival time after the call to
3335 ost_brw_lock_get(), but before we do anything about
3336 processing it & sending the BULK transfer request. This
3337 should help move old stale pending locks off the queue as
3338 quickly as obd_timeout.
3341 Frequency : when an incorrect nid is specified during startup
3343 Description: ptlrpc connect to non-existant node causes kernel crash
3344 Details : LNET can't be re-entered from an event callback, which
3345 happened when we expire a message after the export has been
3346 cleaned up. Instead, hand the zombie cleanup off to another
3349 Severity : enhancement
3351 Description: plain/inodebits lock performance improvement
3352 Details : Grouping plain/inodebits in granted list by their request modes
3353 and bits policy, thus improving the performance of search through
3357 Frequency : only if OST filesystem is corrupted
3359 Description: client incorrectly hits assertion in ptlrpc_replay_req()
3360 Details : for a short time RPCs with bulk IO are in the replay list,
3361 but replay of bulk IOs is unimplemented. If the OST filesystem
3362 is corrupted due to disk cache incoherency and then replay is
3363 started it is possible to trip an assertion. Avoid putting
3364 committed RPCs into the replay list at all to avoid this issue.
3367 Frequency : liblustre (e.g. catamount) on a large cluster with >= 8 OSTs/OSS
3369 Description: System hang on startup
3370 Details : This bug allowed the liblustre (e.g. catamount) client to
3371 return to the app before handling all startup RPCs. This
3372 could leave the node unresponsive to lustre network traffic
3373 and manifested as a server ptllnd timeout.
3375 Severity : enhancement
3377 Description: Add "/proc/sys/lustre/debug_peer_on_timeout"
3378 Details : liblustre envirable: LIBLUSTRE_DEBUG_PEER_ON_TIMEOUT
3379 boolean to control whether to print peer debug info when a
3380 client's RPC times out.
3383 Frequency : only for kernels with patches from Lustre below 1.4.3
3385 Description: Remove old rdonly API
3386 Details : Remove old rdonly API which unused from at least lustre 1.4.3
3389 Frequency : only for devices with external journals
3391 Description: Set external device read-only also
3392 Details : During a commanded failover stop, we set the disk device
3393 read-only while the server shuts down. We now also set any
3394 external journal device read-only at the same time.
3397 Frequency : when upgrading from 1.4 while trying to change parameters
3399 Description: The wrong (new) MDC name was used when setting parameters for
3400 upgraded MDT's. Also allows changing of OSC (and MDC)
3401 parameters if --writeconf is specified at tunefs upgrade time.
3404 Frequency : when setting specific ost indicies
3406 Description: QOS code breaks on skipped indicies
3407 Details : Add checks for missing OST indicies in the QOS code, so OSTs
3408 created with --index need not be sequential.
3410 Severity : enhancement
3412 Description: Add uninit_groups feature to ldiskfs2 to speed up e2fsck
3413 Details : The uninit_groups feature works in conjunction with the kernel
3414 filesystem code (ldiskfs2 only) and e2fsprogs-1.39-cfs6 to speed
3415 up the pass1 processing of e2fsck. This is a read-only feature
3416 in ldiskfs2 only, so older kernels and current ldiskfs cannot
3417 mount filesystems that have had this feature enabled.
3419 Severity : enhancement
3421 Description: Improve multi-block allocation algorithm to avoid fragmentation
3422 Details : The mballoc3 code (ldiskfs2 only) adds new mechanisms to improve
3423 allocation locality and avoid filesystem fragmentation.
3425 ------------------------------------------------------------------------------
3427 2007-04-01 Cluster File Systems, Inc. <info@clusterfs.com>
3429 * Support for kernels:
3430 2.4.21-47.0.1.EL (RHEL 3)
3431 2.6.5-7.283 (SLES 9)
3432 2.6.9-42.0.10.EL (RHEL 4)
3433 2.6.12.6 vanilla (kernel.org)
3434 2.6.16.27-0.9 (SLES 10)
3435 * Recommended e2fsprogs version: 1.39.cfs5
3437 * Note that reiserfs quotas are disabled on SLES 10 in this kernel
3441 Frequency : occasional, depends on client load and configuration
3442 Bugzilla : 12181, 12203
3443 Description: data loss for recently-modified files
3445 Details : In some cases it is possible that recently written or created
3446 files may not be written to disk in a timely manner (this should
3447 normally be within 30s unless client IO load is very high).
3448 The problem appears as zero-length files or files that are a
3449 multiple of 1MB in size after a client crash or client eviction
3450 that are missing data at the end of the file.
3452 This problem is more likely to be hit on clients where files are
3453 repeatedly created and unlinked in the same directory, clients
3454 have a large amount of RAM, have many CPUs, the filesystem has
3455 many OSTs, the clients are rebooted frequently, and/or the files
3456 are not accessed by other nodes after being written.
3458 The presence of the problem can be detected by looking at
3459 /proc/sys/fs/inode-state. If the first number (nr_inodes) is
3460 smaller than the second (nr_unused) then dirty files will not
3461 be flushed automatically to disk. "sync; sleep 10" should be
3462 run several times on the node before unmounting it to update
3463 Lustre (this is also safe to run on nodes without this problem).
3465 There is also a related kernel bug in the RHEL4 4 2.6.9 kernel
3466 that can cause this same problem, so customers using that kernel
3467 also need to update the kernel in addition to Lustre. In order
3468 to properly fix this bug, the RHEL3 2.4.21 kernel is also updated.
3470 It is normal that files written just before a client crash (less
3471 than 30s) may not yet have been flushed to disk, even for local
3475 Frequency : frequent on thin XT3 nodes
3477 Description: UUID collision on thin XT3 Linux nodes
3478 Details : UUIDs on Compute Node Linux XT3 nodes were not generated
3479 randomly, since we relied on an insufficiently-seeded PRNG.
3484 Description: OSS hangs after "All ost request buffers busy"
3485 Details : A deadlock between quota and journal operations caused OSS
3486 hangs after printing "All ost request buffers busy."
3489 Frequency : always on liblustre builds
3491 Description: Cleanup compiler warnings on liblustre
3494 Frequency : always on liblustre builds on XT3
3496 Description: LC_CONFIG_CDEBUG don't run while build liblustre on XT3.
3500 Description: Addition of EXT3_FEATURE_RO_COMPAT_DIR_NLINKS flag for
3501 > 32000 subdirectories
3502 Details : Add EXT3_FEATURE_RO_COMPAT_DIR_NLINK flag to
3503 EXT3_FEATURE_RO_COMPAT_SUPP. This flag will be set whenever
3504 subdirectory count crosses 32000. This will aid e2fsck to
3505 correctly handle more than 32000 subdirectories.
3508 Frequency : liblustre (e.g. catamount) on a large cluster with >= 8 OSTs/OSS
3510 Description: System hang on startup
3511 Details : This bug allowed the liblustre (e.g. catamount) client to
3512 return to the app before handling all startup RPCs. This
3513 could leave the node unresponsive to lustre network traffic
3514 and manifested as a server ptllnd timeout.
3516 Severity : enhancement
3518 Description: Add "/proc/sys/lustre/debug_peer_on_timeout"
3519 (liblustre envirable: LIBLUSTRE_DEBUG_PEER_ON_TIMEOUT)
3520 boolean to control whether to print peer debug info when a
3521 client's RPC times out.
3526 Description: make O_SYNC working on 2.6 kernels
3527 Details : 2.6 kernels use different method for mark pages for write,
3528 so need add a code to lustre for O_SYNC work.
3533 Description: Failure to close file and release space on NFS
3534 Details : Put inode details into lock acquired in ll_intent_file_open.
3535 Use mdc_intent_lock in ll_intent_open to properly
3536 detect all kind of errors unhandled by mdc_enqueue.
3541 Description: proc file read during shutdown sometimes raced obd removal,
3543 Details : Add lock to prevent obd access after proc file removal.
3546 Frequency : Only for files larger than 4GB on 32-bit clients.
3548 Description: improperly doing page alignment of locks
3549 Details : Modify lustre core code to use CFS_PAGE_* defines instead of
3550 PAGE_*. Make CFS_PAGE_MASK a 64-bit mask.
3555 Description: RPCs being resent when they shouldn't be
3556 Details : Some RPCs that should not be resent are being resent. This
3557 can cause inconsistencies in the RPC state machine. Do not
3558 resend such requests.
3561 Frequency : rare, only with NFS export
3563 Description: Crash on NFS re-export node
3564 Details : under very unusual load conditions an assertion is hit in
3565 ll_intent_file_open()
3568 Frequency : only if OST filesystem is corrupted
3570 Description: client incorrectly hits assertion in ptlrpc_replay_req()
3571 Details : for a short time RPCs with bulk IO are in the replay list,
3572 but replay of bulk IOs is unimplemented. If the OST filesystem
3573 is corrupted due to disk cache incoherency and then replay is
3574 started it is possible to trip an assertion. Avoid putting
3575 committed RPCs into the replay list at all to avoid this issue.
3580 Description: large O_DIRECT requests fail under memory pressure/fragmentation
3581 Details : Large single O_DIRECT read and write calls can fail to allocate
3582 a sufficiently large buffer to process the request. In case of
3583 allocation failure the allocation is retried with a smaller
3584 buffer and broken into smaller requests.
3586 Severity : enhancement
3588 Description: Add -o localflock option to simulate old noflock behaviour.
3589 Details : This will achieve local-only flock/fcntl locks coherentness.
3594 Description: versioning check is incomplete
3595 Details : Checking the version difference of client vs. server, report
3596 error if the gap is too big.
3601 Description: add support PG_writeback bit
3602 Details : add support for PG_writeback bit for Lustre, for more carefull
3603 work with page cache in 2.6 kernel. This also fix some deadlocks
3604 and remove hack for work O_SYNC with 2.6 kernel.
3606 Severity : enhancement
3608 Description: Add uninit_groups feature to ldiskfs2 to speed up e2fsck
3609 Details : The uninit_groups feature works in conjunction with the kernel
3610 filesystem code (ldiskfs2 only) and e2fsprogs-1.39-cfs6 to speed
3611 up the pass1 processing of e2fsck. This is a read-only feature
3612 in ldiskfs2 only, so older kernels and current ldiskfs cannot
3613 mount filesystems that have had this feature enabled.
3615 Severity : enhancement
3617 Description: Improve multi-block allocation algorithm to avoid fragmentation
3618 Details : The mballoc3 code (ldiskfs2 only) adds new mechanisms to improve
3619 allocation locality and avoid filesystem fragmentation.
3621 ------------------------------------------------------------------------------
3623 2007-02-09 Cluster File Systems, Inc. <info@clusterfs.com>
3625 * Support for kernels:
3626 2.6.9-42.0.3.EL (RHEL 4)
3627 2.6.5-7.276 (SLES 9)
3628 2.4.21-47.0.1.EL (RHEL 3)
3629 2.6.12.6 vanilla (kernel.org)
3630 2.6.16.21-0.8 (SLES10)
3631 * Recommended e2fsprogs version: 1.39.cfs2-0
3633 * The backwards-compatible /proc/sys/portals symlink has been removed
3634 in this release. Before upgrading, please ensure that you change
3635 any configuration scripts or /etc/sysctl.conf files that access
3636 /proc/sys/portals/* or sysctl portals.* to use the corresponding
3637 entry in /proc/sys/lnet or sysctl lnet.*. This change can be made
3638 in advance of the upgrade on any system running Lustre 1.4.6 or
3639 newer, since /proc/sys/lnet was added in that version.
3640 * Note that reiserfs quotas are disabled on SLES 10 in this kernel
3644 Frequency : only when quota is used
3646 Description: avoid scanning export list for quota master
3647 Details : Change the algorithms to avoid scanning export list in order
3648 to improve the efficiency.
3651 Frequency : MDS failover only, very rarely
3653 Description: "went back in time" messages on mds failover
3654 Details : The greatest transno may be lost when the current operation
3655 finishes with an error (transno==0) and the client's last_rcvd
3656 record is over-written. Save the greatest transno in the
3657 mds_last_transno for this case.
3660 Frequency : always for specific kernels and striping counts
3662 Description: client may get "Matching packet too big" without ACL support
3663 Details : Clients compiled without CONFIG_FS_POSIX_ACL get an error message
3664 when trying to access files in certain configurations. The
3665 clients should in fact be denied when mounting because they do
3666 not understand ACLs.
3669 Frequency : Cray XT3 with more than 4000 clients and multiple jobs
3671 Description: many clients connecting with IO in progress causes connect timeouts
3672 Details : Avoid synchronous journal commits to avoid delays caused by many
3673 clients connecting/disconnecting when bulk IO is in progress.
3674 Queue liblustre connect requests on OST_REQUEST_PORTAL instead of
3675 OST_IO_PORTAL to avoid delays behind potentially many pending
3679 Frequency : occasionally with multiple writers to a single file
3681 Description: shared writes to file may result in wrong size reported by stat()
3682 Details : Allow growing of kms when extent lock is cancelled
3685 Frequency : always with random mmap IO to multi-striped file
3687 Description: mmap write might be lost if we are writing to a 'hole' in stripe
3688 Details : Only if the hole is at the end of OST object so that kms is too
3689 small. Fix is to increase kms accordingly in ll_nopage.
3692 Frequency : rare, only if OST filesystem is inconsistent with MDS filesystem
3694 Description: writes to a missing object would leak memory on the OST
3695 Details : If there is an inconsistency between the MDS and OST filesystems,
3696 such that the MDS references an object that doesn't exist, writes
3697 to that object will leak memory due to incorrect cleanup in the
3698 error handling path, eventually running out of memory on the OST.
3703 Description: Creating too long symlink causes lustre errors
3704 Details : Check symlink and name lengths before sending requests to MDS.
3707 Frequency : only if flock is enabled (not on by default)
3709 Description: posix locks not released on fd closure on 2.6.9+
3710 Details : We failed to add posix locks to list of inode locks on 2.6.9+
3711 kernels, this caused such locks not to be released on fd close and
3712 then assertions on fs unmount about still used locks.
3715 Frequency : MDS failover only, very rarely
3717 Description: clients may get ASSERTION(granted_lock != NULL)
3718 Details : When request was taking a long time, and a client was resending
3719 a getattr by name lock request. The were multiple lock requests
3720 with the same client lock handle and
3721 mds_getattr_name->fixup_handle_for_resent_request found one of the
3722 lock handles but later failed with ASSERTION(granted_lock != NULL).
3727 Description: handle->h_buffer_credits > 0, assertion failure
3728 Details : h_buffer_credits is zero after truncate, causing assertion
3729 failure. This patch extends the transaction or creates a new
3733 Frequency : NFS re-export or patchless client
3734 Bugzilla : 11179, 10796
3735 Description: Crash on NFS re-export node (__d_move)
3736 Details : We do not want to hash the dentry if we don't have a lock.
3737 But if this dentry is later used in d_move, we'd hit uninitialised
3738 list head d_hash, so we just do this to init d_hash field but
3739 leave dentry unhashed.
3742 Frequency : NFS re-export or patchless client
3744 Description: NFS exports has problem with symbolic link
3745 Details : lustre client didn't properly install dentry when re-exported
3746 to NFS or running patchless client.
3749 Frequency : NFS re-export or patchless client
3751 Description: Various nfs/patchless fixes.
3752 Details : fixes reuse disconected alias for lookup process - this fixes
3753 warning "find_exported_dentry: npd != pd",
3754 fix permission error with open files at nfs.
3755 fix apply umask when do revalidate.
3758 Frequency : occasional
3760 Description: Crash on NFS re-export node
3761 Details : calling clear_page() on the wrong pointer triggered oops in
3762 generic_mapping_read().
3765 Frequency : rarely, using O_DIRECT IO
3767 Description: unaligned directio crashes client with LASSERT
3768 Details : check for unaligned buffers before trying any requests.
3771 Frequency : rarely, using CFS RAID5 patches in non-standard kernel series
3773 Description: stale data returned from RAID cache
3774 Details : If only a small amount of IO is done to the RAID device before
3775 reading it again it is possible to get stale data from the RAID
3776 cache instead of reading it from disk.
3779 Frequency : always for sles10 kernel
3781 Description: sles10 support
3782 Details : ll_follow_link: compile fixes and using of nd_set_link
3783 under newer kernels.
3786 Frequency : depends on arch, kernel and compiler version, always on sles10
3789 Description: recursive or deep enough symlinks cause stack overflow
3790 Details : getting rid of large stack-allocated variable in
3794 Frequency : depends on hardware
3796 Description: lustre write performance loss in the SLES10 kernel
3797 Details : the performance loss is caused by using of write barriers in the
3798 ext3 code. The SLES10 kernel turns barrier support on by
3799 default. The fix is to undo that change for ldiskfs.
3801 ------------------------------------------------------------------------------
3803 2006-12-09 Cluster File Systems, Inc. <info@clusterfs.com>
3805 * Support for kernels:
3806 2.6.9-42.0.3EL (RHEL 4)
3807 2.6.5-7.276 (SLES 9)
3808 2.4.21-47.0.1.EL (RHEL 3)
3809 2.6.12.6 vanilla (kernel.org)
3813 Frequency : quota enabled and large files being deleted
3815 Description: releasing more than 4GB of quota at once hangs OST
3816 Details : If a user deletes more than 4GB of files on a single OST it
3817 will cause the OST to spin in an infinite loop. Release
3818 quota in < 4GB chunks, or use a 64-bit value for 1.4.7.1+.
3823 Description: statfs data retrieved from /proc may be stale or zero
3824 Details : When reading per-device statfs data from /proc, in the
3825 {kbytes,files}_{total,free,avail} files, it may appear
3826 as zero or be out of date.
3829 Frequency : systems with MD RAID1 external journal devices
3831 Description: lconf's call to blkid is confused by RAID1 journal devices
3832 Details : Use the "blkid -l" flag to locate the MD RAID device instead
3833 of returning all block devices that match the journal UUID.
3836 Frequency : always, for aggregate stripe size over 4GB
3838 Description: "lfs setstripe" fails assertion when setting 4GB+ stripe width
3839 Details : Using "lfs setstripe" to set stripe size * stripe count over 4GB
3840 will fail the kernel with "ASSERTION(lsm->lsm_xfersize != 0)"
3843 Frequency : always if "lfs find" used on a local file/directory
3845 Description: "lfs find" segfaults if used on a local file/directory
3846 Details : The case where a directory component was not specified wasn't
3847 handled correctly. Handle this properly.
3850 Frequency : always on ppc64
3852 Description: the write to an ext3 filesystem mounted with mballoc got stuck
3853 Details : ext3_mb_generate_buddy() uses find_next_bit() which does not
3854 perform endianness conversion.
3857 Frequency : rarely (truncate to non-zero file size after write under load)
3858 Bugzilla : 10730, 10687
3859 Description: Files padded with zeros to next 4K multiple
3860 Details : With filesystems mounted using the "extents" option (2.6 kernels)
3861 it is possible that files that are truncated to a non-zero size
3862 immediately after being written are filled with zero bytes beyond
3863 the truncated size. No file data is lost.
3865 Severity : enhancement
3867 Description: Allow recovery/failover for liblustre clients.
3868 Details : liblustre clients were unaware of failover configurations until
3871 Severity : enhancement
3873 Description: user file locks should fail when not mounting with flock option
3874 Details : Set up an error-returning stub in ll_file_operations.lock field
3875 to prevent incorrect behaviour when client is mounted without
3876 flock option. Also, set up properly f_op->flock field for
3880 Frequency : always on ia64
3882 Description: "lfs df" loops on printing out MDS statfs information
3883 Details : The obd_ioctl_data was not initialized and in some systems
3884 this caused a failure during the ioctl that did not return
3885 an error. Initialize the struct and return an error on failure.
3888 Frequency : SLES 9 only
3890 Description: Error of copying files with lustre special EAs as root
3891 Details : Client side always return success for setxattr call for lustre
3892 special xattr (currently only "trusted.lov").
3895 Frequency : rarely on clusters with both ia64+i386 clients
3897 Description: ia64+i686 clients doing shared IO on the same file may LBUG
3898 Details : In rare cases when both ia64+i686 (or other mixed-PAGE_SIZE)
3899 clients are doing concurrent writes to the same file it is
3900 possible that the ia64 clients may LASSERT because the OST
3901 extent locks are not PAGE_SIZE aligned. Ensure that grown
3902 locks are always aligned on the request boundary.
3905 Frequency : specific use, occasional
3907 Description: Overwriting in use executable truncates on-disk binary image
3908 Details : If one node attempts to overwrite an executable in use by
3909 another node, we now correctly return ETXTBSY instead of
3910 truncating the file.
3912 Severity : enhancement
3914 Description: Async OSC create to avoid the blocking unnecessarily.
3915 Details : If a OST has no remain object, system will block on the creating
3916 when need to create a new object on this OST. Now, ways use
3917 pre-created objects when available, instead of blocking on an
3918 empty osc while others are not empty. If we must block, we block
3919 for the shortest possible period of time.
3924 Description: chmod on Lustre root is propagated to other clients
3925 Details : Re-validate root's dentry in ll_lookup_it to avoid having it
3926 invalid by the follow_mount time.
3929 Frequency : liblustre clients only
3931 Description: Race in 'instant cancel' lock handling could lead to such locks
3932 never to be granted in case of SMP MDS
3933 Details : Do not destroy not yet granted but cbpending locks in
3937 Frequency : replay/resend of open
3939 Description: non null lock assetion failure in mds_intent_policy
3940 Details : Trying to replay/resend lockless open requests resulted in
3941 mds_open() returning 0 with no lock. Now it sets a flag if
3942 a lock is going to be returned.
3944 Severity : enhancement
3946 Description: Checksum enhancements
3947 Details : New checksum enhancements allow for resending RPCs that failed
3950 Severity : enhancement
3952 Description: Tunables on number of dirty pages in cacche
3953 Details : Allow to set limit on number of dirty pages cached.
3958 Description: client crash on unmount - lock still has references
3959 Details : In some error handling cases it was possible to leak a lock
3960 reference on a client while accessing a file. This was not
3961 harmful to the client during operation, but would cause the
3962 client to crash when the filesystem is unmounted.
3965 Frequency : specific case, rare
3967 Description: ETXTBSY on mds though file not in use by client
3968 Details : ETXTBSY is no longer incorrectly returned when attempting to
3969 chmod or chown a directory that the user previously tried to
3970 execute or a currently-executing binary.
3973 Frequency : extremely rare except on liblustre-based clients
3975 Description: Lustre space not freed when files are deleted
3976 Details : Clean up open-unlinked files after client eviction. Previously
3977 the unlink was skipped and the files remained as orphans.
3982 Description: OST failure "would be an LBUG" in waiting_locks_callback()
3983 Details : In some cases it was possible to send a blocking callback to
3984 a client doing a glimpse, even though that client didn't get
3985 a lock granted. When the glimpse lock is cancelled on the OST
3986 the freed lock is left on the waiting list and corrupted the list.
3989 Frequency : all core dumps
3991 Description: Broke core dumps to lustre
3992 Details : Negative dentry may be unhashed if parent does not have UPDATE
3993 lock, but some callers, e.g. do_coredump, expect dentry to be
3994 hashed after successful create, hash it in ll_create_it.
3996 ------------------------------------------------------------------------------
3998 2006-09-13 Cluster File Systems, Inc. <info@clusterfs.com>
4000 * Support for kernels:
4001 2.6.9-42.0.2.EL (RHEL 4)
4002 2.6.5-7.276 (SLES 9)
4003 2.4.21-40.EL (RHEL 3)
4004 2.6.12.6 vanilla (kernel.org)
4008 Frequency : always on RHEL 3
4010 Description: Number of open files grows over time
4011 Details : The number of open files grows over time, whether or not
4012 Lustre is started. This was due to a filp leak introduced
4013 by one of our kernel patches.
4015 ------------------------------------------------------------------------------
4017 2006-08-20 Cluster File Systems, Inc. <info@clusterfs.com>
4019 * Support for kernels:
4020 2.6.9-42.EL (RHEL 4)
4021 2.6.5-7.267 (SLES 9)
4022 2.4.21-40.EL (RHEL 3)
4023 2.6.12.6 vanilla (kernel.org)
4028 Bugzilla : 5719, 9635, 9792, 9684
4029 Description: OST (or MDS) trips assertions in (re)connection under heavy load
4030 Details : If a server is under heavy load and cannot reply to new
4031 connection requests before the client resends the (re)connect,
4032 the connection handling code can behave badly if two service
4033 threads are concurrently handing separate (re)connections from
4034 the same client. Add better locking to the connection handling
4035 code, and ensure that only a single connection will be processed
4036 for a given client UUID, even if the lock is dropped.
4038 Severity : enhancement
4040 Description: add TCP zero-copy support to kernel
4041 Details : Add support to the kernel TCP stack to allow zero-copy bulk
4042 sends if the hardware supports scatter-gather and checksumming.
4043 This allows socklnd to do client-write and server-read more
4044 efficiently and reduce CPU utilization from skbuf copying.
4047 Frequency : only if NFS exporting from client
4049 Description: NULL pointer deref in ll_iocontrol() if chattr mknod file
4050 Details : If setting attributes on a file created under NFS that had
4051 never been opened it would be possible to oops the client
4052 if the file had no objects.
4056 Bugzilla : 9326, 10402, 10897
4057 Description: client crash in ptlrpcd_wake() thread when sending async RPC
4058 Details : It is possible that ptlrpcd_wake() dereferences a freed async
4059 RPC. In rare cases the ptlrpcd thread alread processed the RPC
4060 before ptlrpcd_wake() was called and the request was freed.
4063 Frequency : always for liblustre
4065 Description: liblustre client does MDS+OSTs setattr RPC for each write
4066 Details : When doing a write from a liblustre client, the client
4067 incorrectly issued an RPC to the MDS and each OST the file was
4068 striped over in order to update the timestamps. When writing
4069 with small chunks and many clients this could overwhelm the MDS
4070 with RPCs. In all cases it would slow down the write because
4071 these RPCs are unnecessary.
4073 Severity : enhancement
4075 Description: allow number of MDS service threads to be changed at module load
4076 Details : It is now possible to change the number of MDS service threads
4077 running. Adding "options mds mds_num_threads={N}" to the MDS's
4078 /etc/modprobe.conf will set the number of threads for the next
4079 time Lustre is restarted (assuming the "mds" module is also
4080 reloaded at that time). The default number of threads will
4081 stay the same, 32 for most systems.
4086 Description: OST crash if filesystem is unformatted or corrupt
4087 Details : If an OST is started on a device that has never been formatted
4088 or if the filesystem is corrupt and cannot even mount then the
4089 error handling cleanup routines would dereference a NULL pointer.
4094 Description: NULL pointer deref in llap_from_page.
4095 Details : get_cache_page_nowait can return a page with NULL (or otherwise
4096 incorrect) mapping if the page was truncated/reclaimed while it was
4097 searched for. Check for this condition and skip such pages when
4098 doing readahead. Introduce extra check to llap_from_page() to
4099 verify page->mapping->host is non-NULL (so page is not anonymous).
4102 Frequency : Sometimes when using sys_sendfile
4104 Description: "page not covered by a lock" warnings from ll_readpage
4105 Details : sendfile called ll_readpage without right page locks present.
4106 Now we introduced ll_file_sendfile that does necessary locking
4107 around call to generic_file_sendfile() much like we do in
4111 Frequency : with certain MDS communication failures at client mount time
4113 Description: NULL pointer deref after failed client mount
4114 Details : a client connection request may delayed by the network layer
4115 and not be sent until after the PTLRPC layer has timed out the
4116 request. If the client fails the mount immediately it will try
4117 to clean up before the network times out the request. Add a
4118 reference from the request import to the obd device and delay
4119 the cleanup until the network drops the request.
4122 Frequency : occasionally during client (re)connect
4124 Description: assertion failure during client (re)connect
4125 Details : processing a client connection request may be delayed by the
4126 client or server longer than the client connect timeout. This
4127 causes the client to resend the connection request. If the
4128 original connection request is replied in this interval, the
4129 client may trip an assertion failure in ptlrpc_connect_interpret()
4130 which thought it would be the only running connect process.
4133 Frequency : only with obd_echo servers and clients that are rebooted
4135 Description: kernel BUG accessing uninitialized data structure
4136 Details : When running an obd_echo server it did not start the ping_evictor
4137 thread, and when a client was evicted an uninitialized data
4138 structure was accessed. Start the ping_evictor in the RPC
4139 service startup instead of the OBD startup.
4141 Severity : enhancement
4142 Bugzilla : 10193 (patchless)
4143 Description: Remove dependency on various unexported kernel interfaces.
4144 Details : No longer need reparent_to_init, exit_mm, exit_files,
4145 sock_getsockopt, filemap_populate, FMODE_EXEC, put_filp.
4148 Frequency : rare (only users of deprecated and unsupported LDAP config)
4150 Description: write_conf for zeroconf mount queried LDAP incorrectly for client
4151 Details : LDAP apparently contains 'lustreName' attributes instead of
4152 'name'. A simple remapping of the name is sufficient.
4155 Frequency : rare (only with non-default dump_on_timeout debug enabled)
4157 Description: waiting_locks_callback trips kernel BUG if client is evicted
4158 Details : Running with the dump_on_timeout debug flag turned on makes
4159 it possible that the waiting_locks_callback() can try to dump
4160 the Lustre kernel debug logs from an interrupt handler. Defer
4161 this log dumping to the expired_lock_main() thread.
4163 Severity : enhancement
4165 Description: Support NFS exporting on 2.6 kernels.
4166 Details : Implement non-rawops metadata methods for NFS server to use without
4167 changing NFS server code.
4170 Frequency : very rare (synthetic metadata workload only)
4172 Description: two racing renames might cause an MDS thread to deadlock
4173 Details : Running the "racer" program may cause one MDS thread to rename
4174 a file from being the source of a rename to being the target of
4175 a rename at exactly the same time that another thread is doing
4176 so, and the second thread has already enqueued these locks after
4177 doing a lookup of the target and is trying to relock them in
4178 order. Ensure that we don't try to re-lock the same resource.
4181 Frequency : only very large systems with liblustre clients
4183 Description: slow eviction of liblustre clients with the "evict_by_nid" RPC
4184 Details : Use asynchronous set_info RPCs to send the "evict_by_nid" to
4185 all OSTs in parallel. This allows the eviction of stale liblustre
4186 clients to proceed much faster than if they were done in series,
4187 and also offers similar improvements for other set_info RPCs.
4192 Description: excessive CPU usage during initial read phase on client
4193 Details : During the initial read phase on a client, it would agressively
4194 retry readahead on the file, consuming too much CPU and impacting
4195 performance (since 1.4.5.8). Improve the readahead algorithm
4196 to avoid this, and also improve some other common cases (read
4197 of small files in particular, where "small" is files smaller than
4198 /proc/fs/lustre/llite/*/max_read_ahead_whole_mb, 2MB by default).
4203 Description: MDS crash when receiving packet with unknown intent.
4204 Details : Do not LBUG in unknown intent case, just return -EFAULT
4206 Severity : enhancement
4207 Bugzilla : 9293, 9385
4208 Description: MDS RPCs are serialised on client. This is unnecessary for some.
4209 Details : Do not serialize getattr (non-intent version) and statfs.
4212 Frequency : occasional, when OST network is overloaded/intermittent
4214 Description: client evicted by OST after bulk IO timeout
4215 Details : If a client sends a bulk IO request (read or write) the OST
4216 may evict the client if it is unresposive to its data GET/PUT
4217 request. This is incorrect if the network is overloaded (takes
4218 too long to transfer the RPC data) or dropped the OST GET/PUT
4219 request. There is no need to evict the client at all, since
4220 the pinger and/or lock callbacks will handle this, and the
4221 client can restart the bulk request.
4224 Frequency : Always when mmapping file with no objects
4226 Description: client crashes when mmapping file with no objects
4227 Details : Check that we actually have objects in a file before doing any
4228 operations on objects in ll_vm_open, ll_vm_close and
4234 Description: Request leak when working with deleted CWD
4235 Details : Introduce advanced request refcount tracking for requests
4236 referenced from lustre intent.
4238 Severity : Enhancement
4240 Description: Cache open file handles on client.
4241 Details : MDS now will return special lock along with openhandle, if
4242 requested and client is allowed to hold openhandle, even if unused,
4243 until such a lock is revoked. Helps NFS a lot, since NFS is opening
4244 closing files for every read/write openration.
4246 Severity : Enhancement
4248 Description: Cache open negative dentries on client when possible.
4249 Details : Guard negative dentries with UPDATE lock on parent dir, drop
4250 negative dentries on lock revocation.
4255 Description: Remounting a client read-only wasn't possible with a zconf mount
4256 Details : It wasn't possible to remount a client read-only with llmount.
4258 Severity : enhancement
4259 Description: Include MPICH 1.2.6 Lustre ADIO interface patch
4260 Details : In lustre/contrib/ or /usr/share/lustre in RPM a patch for
4261 MPICH is included to add Lustre-specific ADIO interfaces.
4262 This is based closely on the UFS ADIO layer and only differs
4263 in file creation, in order to allow the OST striping to be set.
4264 This is user-contributed code and not supported by CFS.
4269 Description: extended inode attributes (immutable, append-only) work improperly
4270 when 2.4 and 2.6 kernels are used on client/server or vice versa
4271 Details : Introduce kernel-independent values for these flags.
4273 Severity : enhancement
4276 Description: Allow fractional MB tunings for lustre in /proc/ filesystem.
4277 Details : Many of the /proc/ tunables can only be tuned at a megabyte
4278 granularity. Now, Fractional MB granularity is be supported,
4279 this is very useful for low memory system.
4281 Severity : enhancement
4283 Description: Getattr by fid
4284 Details : Getting a file attributes by its fid, obtaining UPDATE|LOOKUP
4285 locks, avoids extra getattr rpc requests to MDS, allows '/' to
4286 have locks and avoids getattr rpc requests for it on every stat.
4289 Frequency : Always, for filesystems larger than 2TB
4291 Description: ldiskfs crash at mount for filesystem larger than 2TB with mballoc
4292 Details : Kenrel kmalloc limits allocations to 128kB and this prevents
4293 filesystems larger than 2TB to be mounted with mballoc enabled.
4296 Frequency : Always, for 32-bit kernel without CONFIG_LBD and filesystem > 2TB
4298 Description: filesystem corruption for non-standard kernels and very large OSTs
4299 Details : If a 32-bit kernel is compiled without CONFIG_LBD enabled and a
4300 filesystems larger than 2TB is mounted then the kernel will
4301 silently corrupt the start of the filesystem. CONFIG_LBD is
4302 enabled for all CFS-supported kernels, but the possibility of
4303 this happening with a modified kernel config exists.
4305 Severity : enhancement
4307 Description: add client O_DIRECT support for 2.6 kernels
4308 Details : It is now possible to do O_DIRECT reads and writes to files
4309 in the Lustre client mountpoint on 2.6 kernel clients.
4311 Severity : enhancement
4313 Description: parallel glimpse, setattr, statfs, punch, destroy requests
4314 Details : Sends glimpse, setattr, statfs, punch, destroy requests to OSTs in
4315 parallel, not waiting for response from every OST before sending
4316 a rpc to the next OST.
4321 Description: setattr vs write race when updating file timestamps
4322 Details : Client processes that update a file timestamp into the past
4323 right after writing to the file (e.g. tar) it is possible that
4324 the updated file modification time can be reset to the current
4325 time due to a race between processing the setattr and write RPC.
4327 Severity : enhancement
4329 Description: Bring 'lfs find' closer in line with regular Linux find.
4330 Details : lfs find util supports -atime, -mtime, -ctime, -maxdepth, -print,
4331 -print0 options and obtains all the needed info through the lustre
4334 Severity : enhancement
4336 Description: support up to 1024 configured devices on one node
4337 Details : change obd_dev array from statically allocated to dynamically
4338 allocated structs as they are first used to reduce memory usage
4343 Description: Flush dirty partially truncated pages during truncate
4344 Details : Immediatelly flush partially truncated pages in filter_setattr,
4345 this way we completely avoid having any pages in page cache on OST
4346 and can retire ugly workarounds during writes to flush such pages.
4351 Description: i_sem vs transaction deadlock in mds_obd_destroy during unlink.
4352 Details : protect inode from truncation within vfs_unlink() context
4353 just take a reference before calling vfs_unlink() and release it
4354 when parent's i_sem is free.
4357 Frequency : always, if extents are used on OSTs
4359 Description: index ei_leaf_hi (48-bit extension) is not zeroed in extent index
4360 Details : OSTs using the extents format would not zero the high 16 bits of
4361 the index physical block number. This is not a problem for any
4362 OST filesystems smaller than 16TB, and no kernels support ext3
4363 filesystems larger than 16TB yet. This is fixed in 1.4.7 (all
4364 new/modified files) and can be fixed for existing filesystems
4365 with e2fsprogs-1.39-cfs1.
4370 Description: import connection selection may be incorrect if timer wraps
4371 Details : Using a 32-bit jiffies timer with HZ=1000 may cause backup
4372 import connections to be ignored if the 32-bit jiffies counter
4373 wraps. Use a 64-bit jiffies counter.
4376 Frequency : during server recovery
4378 Description: crash after server is denying duplicate export
4379 Details : If clients are resending connect requests to the server, the
4380 server refuses to allow a client to connect multiple times.
4381 Fixed a bug in the handling of this case.
4384 Frequency : very large clusters immediately after boot
4386 Description: LNET request buffers exhausted under heavy short-term load
4387 Details : If a large number of client requests are generated on a service
4388 that has previously never seen so many requests it is possible
4389 that the request buffer growth cannot keep up with the spike in
4390 demand. Instead of dropping incoming requests, they are held in
4391 the LND until the RPC service can accept more requests.
4394 Frequency : Sometimes during replay
4396 Description: Assertion failure in ll_local_open after replay.
4397 Details : If replay happened on an open request reply before we were able
4398 to set replay handler, reply will become not swabbed tripping the
4399 assertion in ll_local_open. Now we set the handler right after
4400 recognising of open request
4403 Frequency : very rare
4405 Description: kernel reports "badness in vsnprintf"
4406 Details : Reading from the "recovery_status" /proc file in small chunks
4407 may cause a negative length in lprocfs_obd_rd_recovery_status()
4408 call to vsnprintf() (which is otherwise harmless). Exit early
4409 if there is no more space in the output buffer.
4411 Severity : enhancement
4413 Description: clear OBD RPC statistics by writing to them
4414 Details : It is now possible to clear the OBD RPC statistics by writing
4415 to the "stats" file.
4420 Description: Client mtime is not the same on different clients after utimes
4421 Details : In some cases, the client was using the utimes() syscall on
4422 a file cached on another node. The clients now validate the
4423 ctime from the MDS + OSTs to determine which one is right.
4428 Description: Inability to activate failout mode
4429 Details : lconf script incorrectly assumed that in python string's numeric
4430 value is used in comparisons.
4433 Frequency : always with multiple stripes per file
4435 Description: Inefficient object allocation for mutli-stripe files
4436 Details : When selecting which OSTs to stripe files over, for files with
4437 a stripe count that divides evenly into the number of OSTs,
4438 the MDS is always picking the same starting OST for each file.
4439 Return the OST selection heuristic to the original design.
4444 Description: mount failures may take full timeout to return an error
4445 Details : Under some heavy load conditions it is possible that a
4446 failed mount can wait for the full obd_timeout interval,
4447 possibly several minutes, before reporting an error.
4448 Instead return an error as soon as the status is known.
4450 ------------------------------------------------------------------------------
4452 2006-02-14 Cluster File Systems, Inc. <info@clusterfs.com>
4454 * WIRE PROTOCOL CHANGE. This version of Lustre networking WILL NOT
4455 INTEROPERATE with older versions automatically. Please read the
4456 user documentation before upgrading any part of a live system.
4457 * WARNING: Lustre networking configuration changes are required with
4458 this release. See https://bugzilla.clusterfs.com/show_bug.cgi?id=10052
4461 * Support for kernels:
4462 2.6.9-22.0.2.EL (RHEL 4)
4463 2.6.5-7.244 (SLES 9)
4464 2.6.12.6 vanilla (kernel.org)
4467 Severity : enhancement
4468 Bugzilla : 7981/8208
4469 Description: Introduced Lustre Networking (LNET)
4470 Details : LNET is new networking infrastructure for Lustre, it includes
4471 a reorganized network configuration mode (see the user
4472 documentation for full details) as well as support for routing
4473 between different network fabrics. Lustre Networking Devices
4474 (LNDs) for the supported network fabrics have also been
4475 created for this new infrastructure.
4477 Severity : enhancement
4478 Description: Introduced Access control lists
4479 Details : clients can set ACLs on files and directories in order to have
4480 more fine-grained permissions than the standard Unix UGO+RWX.
4481 The MDS must be started with the "-o acl" mount option.
4483 Severity : enhancement
4484 Description: Introduced filesystem quotas
4485 Details : Administrators may now establish per-user quotas on the
4488 Severity : enhancement
4490 Description: Configuration change for the XT3
4491 The PTLLND is now used to run Lustre over Portals on the XT3
4492 The configure option(s) --with-cray-portals are no longer used.
4493 Rather --with-portals=<path-to-portals-includes> is used to
4494 enable building on the XT3. In addition to enable XT3 specific
4495 features the option --enable-cray-xt3 must be used.
4500 Description: Running on many-way SMP OSTs can trigger oops in llcd_send()
4501 Details : A race between allocating a new llcd and re-getting the llcd_lock
4502 allowed another thread to grab newly-allocated llcd.
4504 Severity : enhancement
4506 Description: 2.6 OST async journal commit and locking fix to improve performance
4507 Details : The filter_direct_io()+filter_commitrw_write() journal commits for
4508 2.6 kernels are now async as they already were in 2.4 kernels so
4509 that they can commit concurrently with the network bulk transfer.
4510 For block-allocated files the filter allocation semaphore is held
4511 to avoid filesystem fragmentation during allocation. BKL lock
4512 removed for 2.6 xattr operations where it is no longer needed.
4517 Description: lconf incorrectly determined whether two IP networks could talk
4518 Details : In some more complicated routing and multiple-network
4519 configurations, lconf will avoid trying to make a network
4520 connection to a disjoint part of the IP space. It was doing the
4521 math incorrectly for one set of cases.
4526 Description: Fix for potential infinite loop processing records in an llog.
4527 Details : If an llog record is corrupted/zeroed, it is possible to loop
4528 forever in llog_process(). Validate the llog record length
4529 and skip the remainder of the block on error.
4532 Frequency : occasional (liblustre only)
4534 Description: liblustre could not open files whose last component is a symlink
4535 Details : sysio_path_walk() would incorrectly pass the open intent to
4536 intermediate path components.
4539 Frequency : rare (liblustre only with non-standard tuning)
4540 Bugzilla : 7201 (7350)
4541 Description: Tuning the MDC DLM LRU size to zero triggers client LASSERT
4542 Details : llu_lookup_finish_locks() tries to set lock data on a lock
4543 after it has been released, only do this for referenced locks
4545 Severity : enhancement
4547 Description: specifying an (invalid) directory default stripe_size of -1
4548 would reset the directory default striping
4549 Details : stripe_size -1 was used internally to signal directory stripe
4550 removal, now use "all default" to signal dir stripe removal
4551 as a directory striping of "all default" is not useful
4554 Frequency : common for large clusters running liblustre clients
4556 Description: doing an ls when liblustre clients are running is slow
4557 Details : sending a glimpse AST to a liblustre client waits for every AST
4558 to time out, as liblustre clients will not respond. Since they
4559 cannot cache data we refresh the OST lock LVB from disk instead.
4561 Severity : enhancement
4563 Description: doing an ls at the same time as file IO can be slow
4564 Details : enqueue and other "small" requests can be blocked behind many
4565 large IO requests. Create a new OST IO portal for non-IO
4566 requests so they can be processed faster.
4569 Frequency : rare (only HPUX clients mounting unsupported re-exported NFS vol)
4571 Description: an HPUX NFS client would get -EACCESS when ftruncate()ing a newly
4572 created file with mode 000
4573 Details : the Linux NFS server relies on an MDS_OPEN_OWNEROVERRIDE hack to
4574 allow an ftruncate() as a non-root user to a file with mode 000.
4575 Lustre now respects this flag to disable mode checks when
4576 truncating a file owned by the user
4579 Frequency : liblustre-only, when liblustre client dies unexpectedly or becomes
4582 Description: Revoking locks from clients that went dead or catatonic might take
4584 Details : New lock flags FL_CANCEL_ON_BLOCK used by liblustre makes
4585 cancellation of such locks instant on servers without waiting for
4586 any reply from clients. Clients drops these locks when cancel
4587 notification from server is received without replying.
4590 Frequency : liblustre-only, when liblustre client dies or becomes busy
4592 Description: Doing ls on Linux clients can take a long time with active
4594 Details : Liblustre client cannot handle ASTs in timely manner, so avoid
4595 granting such locks to it in the first place if possible. Locks
4596 are taken by proxy on the OST during the read or write and
4597 dropped immediately afterward. Add connect flags handling, do
4598 not grant locks to liblustre clients for glimpse ASTs.
4600 Severity : enhancement
4602 Description: Improve read-ahead algorithm to avoid excessive IO for random reads
4603 Details : Existing read-ahead algorithm is tuned for the case of streamlined
4604 sequential reads and behaves badly with applications doing random
4605 reads. Improve it by reading ahead at least read region, and
4606 avoiding excessive large RPC for small reads.
4608 Severity : enhancement
4610 Description: Creating more than 1000 files for a single job may cause a load
4611 imbalance on the OSTs if there are also a large number of OSTs.
4612 Details : qos_prep_create() uses an OST index reseed value that is an
4613 even multiple of the number of available OSTs so that if the
4614 reseed happens in the middle of the object allocation it will
4615 still utilize the OSTs as uniformly as possible.
4620 Description: OST or MDS may oops in ping_evictor_main()
4621 Details : ping_evictor_main() drops obd_dev_lock if deleting a stale export
4622 but doesn't restart at beginning of obd_exports_timed list
4625 Severity : enhancement
4627 Description: improve by-nid export eviction on the MDS and OST
4628 Details : allow multiple exports with the same NID to be evicted at one
4629 time without re-searching the exports list.
4632 Frequency : rare, only with supplementary groups enabled on SMP 2.6 kernels
4634 Description: MDS may oops in groups_free()
4635 Details : in rare race conditions a newly allocated group_info struct is
4636 freed again, and this can be NULL. The 2.4 compatibility code
4637 for groups_free() checked for a NULL pointer, but 2.6 did not.
4640 Frequency : common for liblustre clients doing little filesystem IO
4641 Bugzilla : 9352, 7313
4642 Description: server may evict liblustre clients accessing contended locks
4643 Details : if a client is granted a lock or receives a completion AST
4644 with a blocking AST already set it would not reply to the AST
4645 for LDLM_FL_CANCEL_ON_BLOCK locks. It now replies to such ASTs.
4648 Frequency : lfs setstripe, only systems with more than 160 OSTs
4650 Description: unable to set striping with a starting offset beyond OST 160
4651 Details : llapi_create_file() incorrectly limited the starting stripe
4652 index to the maximum single-file stripe count.
4655 Frequency : LDAP users only
4657 Description: lconf did not handle in-kernel recovery with LDAP properly
4658 Details : lconf/LustreDB get_refs() is searching the wrong namespace
4660 Severity : enhancement
4662 Description: bind OST threads to NUMA nodes to improve performance
4663 Details : all OST threads are uniformly bound to CPUs on a single NUMA
4664 node and do their allocations there to localize memory access
4666 Severity : enhancement
4668 Description: llmount can determine client NID directly from Myrinet (GM)
4669 Details : the client NID code from gmnalnid was moved directly into
4670 llmount, removing the need to use this or specifying the
4671 client NID explicitly when mounting GM clients with zeroconf
4674 Frequency : if client is started with down MDS
4676 Description: if client is started with down MDS mount hangs in ptlrpc_queue_wait
4677 Details : Having an LWI_INTR() wait event (interruptible, but no timeout)
4678 will wait indefinitely in ptlrpc_queue_wait->l_wait_event() after
4679 ptlrpc_import_delayed_req() because we didn't check if the
4680 request was interrupted, and we also didn't break out of the
4681 event loop if there was no timeout
4686 Description: data loss during non-page-aligned writes to a single file from
4687 both multiple nodes and multiple threads on one node at same time
4688 Details : updates to KMS and lsm weren't protected by common lock. Resulting
4689 inconsistency led to false short-reads, that were cached and later
4690 used by ->prepare_write() to fill in partially written page,
4691 leading to data loss.
4694 Frequency : always, if lconf --abort_recovery used
4696 Description: lconf --abort_recovery fails with 'Operation not supported'
4697 Details : lconf was attempting to abort recovery on the MDT device and not
4700 Severity : enhancement
4702 Description: remove cleanup logs
4703 Details : replace lconf-generated cleanup logs with lustre internal
4704 cleanup routines. Eliminates the need for client-cleanup and
4707 Severity : enhancement
4709 Description: add support for EAs (user and system) on lustre filesystems
4710 Details : it is now possible to store extended attributes in the Lustre
4711 client filesystem, and with the user_xattr mount option it
4712 is possible to allow users to store EAs on their files also
4714 Severity : enhancement
4716 Description: Add possibility (config option) to show minimal available OST free
4718 Details : When compiled with --enable-mindf configure option, statfs(2)
4719 (and so, df) will return least minimal free space available from
4720 all OSTs as amount of free space on FS, instead of summary of
4721 free spaces of all OSTs.
4723 Severity : enhancement
4725 Description: do not expand extent locks acquired on OST-side
4726 Details : Modify ldlm_extent_policy() to not expand local locks, acquired
4727 by server: they are not cached anyway.
4730 Frequency : when mmap is used/binaries executed from Lustre
4732 Description: Unmmap pages before throwing them away from read cache.
4733 Details : llap_shrink cache now attempts to unmap pages before discarding
4734 them (if unmapping failed - do not discard). SLES9 kernel has
4735 extra checks that trigger if this unmapping is not done first.
4740 Description: lconf didn't resolve symlinks before checking to see whether a
4741 given mountpoint was already in use
4744 Frequency : when migrating failover services
4745 Bugzilla : 6395, 9514
4746 Description: When migrating a subset of services from a node (e.g. failback
4747 from a failover service node) the remaining services would
4748 time out and evict clients.
4749 Details : lconf --force (implied by --failover) sets the global obd_timeout
4750 to 5 seconds in order to quickly disconnect, but this caused
4751 other RPCs to time out too quickly. Do not change the global
4752 obd_timeout for force cleanup, only set it for DISCONNECT RPCs.
4754 Severity : enhancement
4755 Frequency : if MDS is started with down OST
4756 Bugzilla : 9439,5706
4757 Description: Allow startup/shutdown of an MDS without depending on the
4758 availability of the OSTs.
4759 Details : Asynchronously call mds_lov_synchronize during MDS startup.
4760 Add appropriate locking and lov-osc refcounts for safe
4761 cleaning. Add osc abort_inflight calls in case the
4762 synchronize never started.
4765 Frequency : occasional (Cray XT3 only)
4767 Description: root not authorized to access files in CRAY_PORTALS environment
4768 Details : The client process capabilities were not honoured on the MDS in
4769 a CRAY_PORTALS/CRAY_XT3 environment. If the file had previously
4770 been accessed by an authorized user then root was able to access
4771 the file on the local client also. The root user capabilities
4772 are now allowed on the MDS, as this environment has secure UID.
4775 Frequency : occasional
4777 Description: ldiskfs "too long searching" message happens too often
4778 Details : A debugging message (otherwise harmless) prints too often on
4779 the OST console. This has been reduced to only happen when
4780 there are fragmentation problems on the filesystem.
4785 Description: Division by zero in statfs when all OSCs are inactive
4786 Details : lov_get_stripecnt() returns zero due to incorrect order of checks,
4787 lov_statfs divides by value returned by lov_get_stripecnt().
4791 Bugzilla : 9489, 3273
4792 Description: First write from each client to each OST was only 4kB in size,
4793 to initialize client writeback cache, which caused sub-optimal
4794 RPCs and poor layout on disk for the first writen file.
4795 Details : Clients now request an initial cache grant at (re)connect time
4796 and so that they can start streaming writes to the cache right
4797 away and always do full-sized RPCs if there is enough data.
4798 If the OST is rebooted the client also re-establishes its grant
4799 so that client cached writes will be honoured under the grant.
4804 Description: Slow ls (and stat(2) syscall) on files residing on IO-loaded OSTs
4805 Details : Now I/O RPCs go to different portal number and (presumably) fast
4806 lock requests (and glimses) and other RPCs get their own service
4807 threads pool that should be able to service those RPCs
4810 Severity : enhancement
4812 Description: Ability to exchange lustre version between client and servers and
4813 issue warnings at client side if client is too old. Also for
4814 liblustre clients there is ability to refuse connection of too old
4816 Details : New 'version' field is added to connect data structure that is
4817 filled with version info. That info is later checked by server and
4821 Frequency : rare, liblustre only.
4822 Bugzilla : 9296, 9581
4823 Description: Two simultaneous writes from liblustre at offset within same page
4824 might proceed at the same time overwriting eachother with stale
4826 Details : I/O lock withing llu_file_prwv was released too early, before data
4827 actually was hitting the wire. Extended lock-holding time until
4828 server acknowledges receiving data.
4831 Frequency : extremely rare. Never observed in practice.
4833 Description: avoid generating lustre_handle cookie of 0.
4834 Details : class_handle_hash() generates handle cookies by incrementing
4835 global counter, and can hit 0 occasionaly (this is unlikely, but
4836 not impossible, because initial value of cookie counter is
4837 selected randonly). Value of 0 is used as a sentinel meaning
4838 "unassigned handle" --- avoid it. Also coalesce two critical
4839 sections in this function into one.
4841 Severity : enhancement
4843 Description: allow liblustre clients to delegate truncate locking to OST
4844 Details : To avoid overhead of locking, liblustre client instructs OST to
4845 take extent lock in ost_punch() on client's behalf. New connection
4846 flag is added to handle backward compatibility.
4848 Severity : enhancement
4849 Bugzilla : 4928, 7341, 9758
4850 Description: allow number of OST service threads to be specified
4851 Details : a module parameter allows the number of OST service threads
4852 to be specified via "options ost ost_num_threads={N}" in the
4853 OSS's /etc/modules.conf or /etc/modprobe.conf.
4857 Bugzilla : 6146, 9635, 9895
4858 Description: servers crash with bad pointer in target_handle_connect()
4859 Details : In rare cases when a client is reconnecting it was possible that
4860 the connection request was the last reference for that export.
4861 We would temporarily drop the export reference and get a new
4862 one, but this may have been the last reference and the export
4863 was just destroyed. Get new reference before dropping old one.
4865 Severity : enhancement
4866 Frequency : if client is started with failover MDS
4868 Description: Allow multiple MDS hostnames in the mount command
4869 Details : Try to read the configuration from all specified MDS
4870 hostnames during a client mount in case the "primary"
4873 Severity : enhancement
4875 Description: Stop sending data to evicted clients as soon as possible.
4876 Details : Check if the client we are about to send or are sending data to
4877 was evicted already. (Check is done every second of waiting,
4878 for which l_wait_event interface was extended to allow checking
4879 of exit condition at specified intervals).
4882 Frequency : rare, normally only when NFS exporting is done from client
4884 Description: 'bad disk LOV MAGIC: 0x00000000' error when chown'ing files
4886 Details : Make mds_get_md() recognise empty md case and set lmm size to 0.
4889 Frequency : always, if srand() is called before liblustre initialization
4891 Description: Liblustre uses system PRNG disturbing its usage by user application
4892 Details : Introduce internal to lustre fast and high-quality PRNG for
4893 lustre usage and make liblustre and some other places in generic
4894 lustre code to use it.
4896 Severity : enhancement
4897 Bugzilla : 9477, 9557, 9870
4898 Description: Verify that the MDS configuration logs are updated when xml is
4899 Details : Check if the .xml configuration logs are newer than the config
4900 logs stored on the MDS and report an error if this is the case.
4901 Request --write-conf, or allow starting with --old_conf.
4903 Severity : enhancement
4905 Description: Handle symlinks in the path when checking if Lustre is mounted.
4906 Details : Resolve intermediate symlinks when checking if a client has
4907 mounted a filesystem to avoid duplicate client mounts.
4912 Description: lconf can hit an error exception but still return success.
4913 Details : The lconf command catches the Command error exception at the top
4914 level script context and will exit with the associated exit
4915 status, but doesn't ensure that this exit status is non-zero.
4920 Description: failure of ptlrpc thread startup can cause oops
4921 Details : Starting a ptlrpc service thread can fail if there are a large
4922 number of threads or the server memory is very fragmented.
4923 Handle this without oopsing.
4926 Frequency : always, only if liblustre and non-default acceptor port was used
4928 Description: liblustre cannot connect to servers with non-default acceptor port
4929 Details : tcpnal_set_default_params() was not called and was therefore
4930 ignoring the environment varaible TCPNAL_PORT, as well as other
4931 TCPNAL_ environment variables
4936 Description: two objects could be created on the same OST for a single file
4937 Details : If an OST is down, in some cases it was possible to create two
4938 objects on a single OST for a single file. No problems other
4939 than potential performance impact and spurious error messages.
4943 Bugzilla : 5681, 9562
4944 Description: Client may oops in ll_unhash_aliases
4945 Details : Client dcache may become inconsistent in race condition.
4946 In some cases "getcwd" can fail if the current directory is
4952 Description: Inode refcounting problems in NFS export code
4953 Details : link_raw functions used to call d_instantiate without obtaining
4954 extra inode reference first.
4958 Bugzilla : 9942, 9903
4959 Description: Referencing freed requests leading to crash, memleaks with NFS.
4960 Details : We used to require that call to ll_revalidate_it was always
4961 followed by ll_lookup_it. Also with revalidate_special() it is
4962 possible to call ll_revalidate_it() twice for the same dentry
4963 even if first occurence returned success. This fix changes semantic
4964 between DISP_ENQ_COMPLETE disposition flag to mean there is extra
4965 reference on a request referred from the intent.
4966 ll_intent_release() then releases such a request.
4969 Frequency : rare, normally benchmark loads only
4971 Description: unlinked inodes were kept in memory on the client
4972 Details : If a client is repeatedly creating and unlinking files it
4973 can accumulate a lot of stale inodes in the inode slab cache.
4974 If there is no other client load running this can cause the
4975 client node to run out of memory. Instead flush old inodes
4976 from client cache that have the same inode number as a new inode.
4979 Frequency : SLES9 2.6.5 kernel and long filenames only
4980 Bugzilla : 9969, 10379
4981 Description: utime reports stale NFS file handle
4982 Details : SLES9 uses out-of-dentry names in some cases, which confused
4983 the lustre dentry revalidation. Change it to always use the
4987 Frequency : rare, unless heavy write-truncate concurrency is continuous
4988 Bugzilla : 4180, 6984, 7171, 9963, 9331
4989 Description: OST becomes very slow and/or deadlocked during object unlink
4990 Details : filter_destroy() was holding onto the parent directory lock
4991 while truncating+unlinking objects. For very large objects this
4992 may block other threads for a long time and slow overall OST
4993 responsiveness. It may also be possible to get a lock ordering
4994 deadlock in this case, or run out of journal credits because of
4995 the combined truncate+unlink. Solution is to do object truncate
4996 first in one transaction without parent lock, and then do the
4997 final unlink in a new transaction with the parent lock. This
4998 reduces the lock hold time dramatically.
5001 Frequency : rare, 2.4 kernels only
5003 Description: MDS or OST cleanup may trip kernel BUG when dropping kernel lock
5004 Details : mds_cleanup() and filter_cleanup() need to drop the kernel lock
5005 before unmounting their filesystem in order to avoid deadlock.
5006 The kernel_locked() function in 2.4 kernels only checks whether
5007 the kernel lock is held, not whether it is this process that is
5008 holding it as 2.6 kernels do.
5013 Description: MDS or OST may oops/LBUG if a client is connecting multiple times
5014 Details : The client ptlrpc code may be trying to reconnect to a down
5015 server before a previous connection attempt has timed out.
5016 Increase the reconnect interval to be longer than the connection
5017 timeout interval to avoid sending duplicate connections to
5021 Frequency : echo_client brw_test command
5023 Description: fix echo_client to work with OST preallocated code
5024 Details : OST preallocation code (5137) didn't take echo_client IO path
5025 into account: echo_client calls filter methods outside of any
5026 OST thread and, hence, there is no per-thread preallocated
5027 pages and buffers to use. Solution: hijack pga pages for IO. As
5028 a byproduct, this avoids unnecessary data copying.
5032 Bugzilla : 3555, 5962, 6025, 6155, 6296, 9574
5033 Description: Client can oops in mdc_commit_close() after open replay
5034 Details : It was possible for the MDS to return an open request with no
5035 transaction number in mds_finish_transno() if the client was
5036 evicted, but without actually returning an error. Clients
5037 would later try to replay that open and may trip an assertion
5038 Simplify the client close codepath, and always return an error
5039 from the MDS in case the open is not successful.
5042 Frequency : rare, 2.6 OSTs only
5044 Description: OST may deadlock under high load on fragmented files
5045 Details : If there was a heavy load and highly-fragmented OST filesystems
5046 it was possible to have all the OST threads deadlock waiting on
5047 allocation of biovecs, because the biovecs were not released
5048 until the entire RPC IO was completed. Instead, release biovecs
5049 as soon as they are complete to ensure forward IO progress.
5051 Severity : enhancement
5053 Description: Support for specifying external journal device at mount
5054 Details : If an OST or MDS device is formatted with an external journal
5055 device, this device major/minor is stored in the ext3 superblock
5056 and may not be valid for failover. Allow detecting and
5057 specifying the external journal at mount time.
5062 Description: Mounting an MDS with pending unlinked files may cause oops
5063 Details : target_finish_recovery() calls mds_postrecov() which returned
5064 the number of orphans unlinked. mds_lov_connect->mds_postsetup()
5065 considers this an error and immediately begins cleaning up the
5066 lov, just after starting the mds_lov process
5068 Severity : enhancement
5070 Description: Implement 'lfs df' to report actual free space on per-OST basis
5071 Details : Add sub-command 'df' on 'lfs' to report the disk space usage of
5072 MDS/OSDs. Usage: lfs df [-i][-h]. Command Options: '-i' to report
5073 usage of objects; '-h' to report in human readable format.
5075 ------------------------------------------------------------------------------
5077 2005-08-26 Cluster File Systems, Inc. <info@clusterfs.com>
5084 Description: Mounting an ldiskfs file system with mballoc may crash OST node.
5085 Details : ldiskfs mballoc code may reference an uninitialized buddy struct
5086 at startup during orphan unlinking. Instead, skip buddy update
5087 before setup, as it will be regenerated after recovery is complete.
5092 Description: If an OST is inactive, its locks might reference stale inodes.
5093 Details : lov_change_cbdata() must iterate over all namespaces, even if
5094 they are inactive to clear inode references from the lock.
5096 Severity : enhancement
5097 Frequency : occasional, if non-standard max_dirty_mb used
5099 Description: Client will block write RPCs if not enough grant
5100 Details : If a client has max_dirty_mb smaller than max_rpcs_in_flight,
5101 then the client will block writes while waiting for another RPC
5102 to complete instead of consuming its dirty limit. With change
5103 we get improved performance when max_dirty_mb is small.
5105 Severity : enhancement
5106 Bugzilla : 3389, 6253
5107 Description: Add support for supplementary groups on the MDS.
5108 Details : The MDS has an upcall /proc/fs/lustre/mds/{mds}/group_upcall
5109 (set to /usr/sbin/l_getgroups if enabled) which will do MDS-side
5110 lookups for user supplementary groups into a cache.
5114 Description: O_CREAT|O_EXCL open flags in liblustre always return -EEXIST
5115 Details : Make libsysio to not enforce O_EXCL by clearing the flag,
5116 for liblustre O_EXCL is enforced by MDS.
5120 Description: readdir never returns NULL in liblustre.
5121 Details : Corrected llu_iop_getdirentries logic, to return offset of next
5122 dentry in struct dirent.
5126 Frequency : liblustre only, depends on application IO pattern
5127 Description: liblustre clients evicted if not contacting servers
5128 Details : Don't put liblustre clients into the ping_evictor list, so
5129 they will not be evicted by the pinger ever.
5131 Severity : enhancement
5133 Description: Add ability to evict clients by NID from MDS.
5134 Details : By echoing "nid:$NID" string into
5135 /proc/fs/lustre/mds/.../evict_client client with nid that equals to
5136 $NID would be instantly evicted from this MDS and from all active
5137 OSTs connected to it.
5141 Description: Do not query file size twice, somewhat slowing stat(2) calls.
5142 Details : lookup_it_finish() used to query file size from OSTs that was not
5147 Description: service threads change working directory to that of init
5148 Details : Starting lustre service threads may pin the working directory
5149 of the parent thread, making that filesystem busy. Threads
5150 now change to the working directory of init to avoid this.
5154 Frequency : during shutdown only
5155 Description: shutdown with a failed MDS or OST can cause unmount to hang
5156 Details : Don't resend DISCONNECT messages in ptlrpc_disconnect_import()
5161 Frequency : 2.6 only
5162 Description: chmod/chown may include an extra supplementary group
5163 Details : ll{,u}_mdc_pack_op_data() does not properly initialize the
5164 supplementary group and if none is specified this is used.
5167 Bugzilla : 5479 (6816)
5169 Description: Racing open + rm can assert client in mdc_set_open_replay_data()
5170 Details : If lookup is in progress on a file that is unlinked we might try
5171 to revalidate the inode and fail in revalidate after lookup is
5172 complete and ll_file_open() enqueues the open again but
5173 it_open_error() was not checking DISP_OPEN_OPEN errors correctly.
5176 Frequency : always, if lconf --abort_recovery used
5178 Description: lconf --abort_recovery fails with 'Operation not supported'
5179 Details : lconf was attempting to abort recovery on the MDT device and not
5182 ------------------------------------------------------------------------------
5184 2005-08-08 Cluster File Systems, Inc. <info@clusterfs.com>
5189 Frequency : rare (only unsupported configurations with a node running as an
5191 Bugzilla : 6514, 5137
5192 Description: Mounting a Lustre file system on a node running as an OST could
5194 Details : OSTs now preallocates memory needed to write out data at
5195 startup, instead of when needed, to avoid having to
5196 allocate memory in possibly low memory situations.
5197 Specifically, if the file system is mounted on on OST,
5198 memory pressure could force it to try to write out data,
5199 which it needed to allocate memory to do. Due to the low
5200 memory, it would be unable to do so and the node would
5201 become unresponsive.
5203 Severity : enhancement
5205 Description: Addition of lconf --service command line option
5206 Details : lconf now accepts a '--service <arg>' option, which is
5207 shorthand for 'lconf --group <arg> --select <arg>=<hostname>'
5209 Severity : enhancement
5211 Description: Failover mode is now the default for OSTs.
5212 Details : By default, OSTs will now run in failover mode. To return to
5213 the old behaviour, add '--failout' to the lmc line for OSTs.
5215 Severity : enhancement
5217 Description: Health checks are now provided for MDS and OSTs
5218 Details : Additional detailed health check information on MSD and OSTs
5219 is now provided through the procfs health_check value.
5222 Frequency : occasional, depends on IO load
5224 Description: Disk fragmentation on the OSTs could eventually cause slowdowns
5225 after numerous create/delete cycles
5226 Details : The ext3 inode allocation policy would not allocate new inodes
5227 very well on the OSTs because there are no new directories
5228 being created. Instead we look for groups with free space if
5229 the parent directories are nearly full.
5234 Description: Network or server problems during mount may cause partially
5235 mounted clients instead of returning an error.
5236 Details : The config llog parsing code may overwrite the error return
5237 code during mount error handling, returning success instead
5243 Description: MDS can fail to allocate large reply buffers
5244 Details : After long uptimes the MDS can fail to allocate large reply
5245 buffers (e.g. zconf client mount config records) due to memory
5246 fragmentation or consumption by the buffer cache. Preallocate
5247 some large reply buffers so that these replies can be sent even
5248 under memory pressure.
5252 Frequency : rare (liblustre)
5253 Description: fsx running with liblustre complained that using truncate() to
5254 extend the file doesn't work. This patch corrects that issue.
5255 Details : This is the liblustre equivalent of the fix for bug 6196. Fixes
5256 ATTR_SIZE and lsm use in llu_setattr_raw.
5260 Frequency : rare, only 2.6 kernels
5261 Description: Unusual file access patterns on the MDS may result in inode
5262 data being lost in very rare circumstances.
5263 Details : Bad interaction between the ea-in-inode patch and the "no-read"
5264 code in the 2.6 kernel caused the inode and/or EA data not to
5265 be read from disk, causing single-file corruption.
5269 Frequency : rare, only 2.6 filesystems using extents
5270 Description: Heavy concurrent write and delete load may cause data corruption.
5271 Details : It was possible under high-load situations to have an extent
5272 metadata block in the block device cache from a just-unlinked
5273 file overwrite a newly-allocated data block. We now unmap any
5274 metadata buffers that alias just-allocated data blocks.
5278 Frequency : filesystems with default stripe_count larger than 77
5279 Description: lconf+mke2fs fail when formatting filesystem with > 77 stripes
5280 Details : lconf specifies an inode size of 4096 bytes when the default
5281 stripe_count is larger than 77. This conflicts with the default
5282 inode density of 1 per 4096 bytes. Allocate smaller inodes in
5283 this case to avoid pinning too much memory for large EAs.
5285 ------------------------------------------------------------------------------
5287 2005-07-07 Cluster File Systems, Inc. <info@clusterfs.com>
5292 Frequency : rare (extremely heavy IO load with hundreds of clients)
5294 Description: Client is evicted, gets IO error writing to file
5295 Details : lock ordering changes for bug 5492 reintroduced bug 3267 and
5296 caused clients to be evicted for AST timeouts. The fixes in
5297 bug 5192 mean we no longer need to have such short AST timeouts
5298 so ldlm_timeout has been increased.
5301 Frequency : occasional during --force or --failover shutdown under load
5302 Bugzilla : 5949, 4834
5303 Description: Server oops/LBUG if stopped with --force or --failover under load
5304 Details : a collection of import/export refcount and cleanup ordering
5305 issues fixed for safer force cleanup
5308 Frequency : only filesystems larger than 120 OSTs
5309 Bugzilla : 5990, 6223
5310 Description: lfs getstripe would oops on a very large filesystem
5311 Details : lov_getconfig used kfree on vmalloc'd memory
5314 Frequency : only filesystems exporting via NFS to Solaris 10 clients
5315 Bugzilla : 6242, 6243
5316 Description: reading from files that had been truncated to a non-zero size
5317 but never opened returned no data
5318 Details : ll_file_read() reads zeros from no-object files to EOF
5323 Description: A bug in MDS/OSS recovery could cause the OSS to fail an assertion
5324 Details : There's little harm in aborting MDS/OSS recovery and letting it
5325 try again, so I removed the LASSERT and return an error instead.
5327 Severity : enhancement
5329 Description: New debugging infrastructure for tracking down data corruption
5330 Details : The I/O checksum code was replaced to: (a) control it at runtime,
5331 (b) cover more of the client-side code path, and (c) try to narrow
5332 down where problems occurred
5336 Bugzilla : 3819, 4364, 4397, 6313
5337 Description: Racing close and eviction MDS could cause assertion in mds_close
5338 Details : It was possible to get multiple mfd references during close and
5339 client eviction, leading to one thread referencing a freed mfd.
5341 Severity: : enhancement
5342 Bugzilla : 3262, 6359
5343 Description: Attempts to reconnect to servers are now more aggressive.
5344 Details : This builds on the enhanced upcall-less recovery that was added
5345 in 1.4.2. When trying to reconnect to servers, clients will
5346 now try each server in the failover group every 10 seconds. By
5347 default, clients would previously try one server every 25 seconds.
5352 Description: After recovery, certain operations trigger a failed
5353 assertion on a client.
5354 Details : Failing over an mds, using lconf -d --failover, while a
5355 client was doing a readdir() call would cause the client to
5356 LBUG after recovery completed and the readdir() was resent.
5358 Severity : enhancement
5360 Description: Default groups are now added by lconf
5361 Details : You can now run lconf --group <servicename> without having to
5362 manually add groups with lmc.
5365 Frequency : occasional
5367 Description: Nodes with an elan id of 0 trigger a failed assertion
5370 Frequency : always when accessing e.g. tty/console device nodes
5372 Description: tty and some other devices nodes cannot be used on lustre
5373 Details : file's private_data field is used by device data and lustre
5374 values in there got lost. New field was added to struct file to
5375 store fs-specific private data.
5378 Frequency : when exporting Lustre via NFS
5380 Description: NFSD failed occasionally when looking up a path component
5381 Details : NFSD is looking up ".." which was broken in ext3 directories
5382 that had grown large enough to become hashed.
5385 Frequency : Clusters with multiple interfaces not on the same subnet
5387 Description: Nodes will repeatedly try to reconnect to an interface which it
5388 cannot reach and report an error to the log.
5389 Details : Extra peer list entries will be created by lconf with some peers
5390 unreachable. lconf now validates the peer before adding it.
5393 Frequency : Only if a default stripe is set on the filesystem root.
5395 Description: Setting a default stripe on the filesystem root prevented the
5396 filesystem from being remounted.
5397 Details : The client was sending extra request flags in the root getattr
5398 request and did not allocate a reply buffer for the dir EA.
5401 Frequency : occasional, higher if lots of files are accessed by one client
5402 Bugzilla : 6159, 6097
5403 Description: Client trips assertion regarding lsm mismatch/magic
5404 Details : While revalidating inodes the VFS looks up inodes with ifind()
5405 and in rare cases can find an inode that is being freed.
5406 The ll_test_inode() code will free the lsm during ifind()
5407 when it finds an existing inode and then the VFS later attaches
5408 this free lsm to a new inode.
5412 Bugzilla : 6422, 7030
5413 Description: MDS deadlock between mkdir and client eviction
5414 Details : Creating a new file via mkdir or mknod (starting a transaction
5415 and getting the ns lock) can deadlock with client eviction
5416 (gets ns lock and trying to finish a synchronous transaction).
5419 Frequency : occasional
5420 Description: While starting a server, the fsfilt_ext3 module could not be
5422 Details : CFS's improved ext3 filesystem is named ldiskfs for 2.6
5423 kernels. Previously, lconf would still use the ext3 name
5424 when trying to load modules. Now, it will correctly use
5425 ext3 on 2.4 and ldiskfs on 2.6.
5427 Severity : enhancement
5428 Description: The default stripe count has been changed to 1
5429 Details : The interpretation of the default stripe count (0, to lfs
5430 or lmc) has been changed to mean striping across a single
5431 OST, rather than all available. For general usage we have
5432 found a stripe count of 1 or 2 works best.
5434 Severity : enhancement
5435 Description: Add support for compiling against Cray portals.
5436 Details : Conditional compiling for some areas that are different
5440 Frequency : occasional
5441 Bugzilla : 6409, 6834
5442 Description: Creating files with an explicit stripe count may lead to
5443 a failed assertion on the MDS
5444 Details : If some OSTs are full or unavailable, creating files may
5445 trigger a failed assertion on the MDS. Now, Lustre will
5446 try to use other servers or return an error to the
5450 Frequency : occasional
5452 Description: Multiple concurrent overlapping read+write on multiple SMP nodes
5453 caused lock timeout during readahead (since 1.4.2).
5454 Details : Processes doing readahead might match a lock that hasn't been
5455 granted yet if there are overlapping and conflicting lock
5456 requests. The readahead process waits on ungranted lock
5457 (original lock is CBPENDING), while OST waits for that process
5458 to cancel CBPENDING read lock and eventually evicts client.
5460 Severity : enhancement
5462 Description: Initial enabling of flock support for clients
5463 Details : Implements fcntl advisory locking and file status functions.
5464 This feature is provided as an optional mount flag (default
5465 off), and is NOT CURRENTLY SUPPORTED. Not all types of record
5466 locking are implemented yet, and those that are are not guaranteed
5467 to be completely correct in production environments.
5468 mount -t lustre -o [flock|noflock] ...
5471 Frequency : occasional
5473 Description: OSTs running 2.4 kernels but with extents enabled might trip an
5474 assertion in the ext3 JBD (journaling) layer.
5475 Details : The b_committed_data struct is protected by the big kernel lock
5476 in 2.4 kernels, serializing journal_commit_transaction() and
5477 ext3_get_block_handle->ext3_new_block->find_next_usable_block()
5478 access to this struct. In 2.6 kernels there is finer grained
5479 locking to improve SMP performance of the JBD layer.
5483 Description: Changes the "SCSI I/O Stats" kernel patch to default to "enabled"
5485 -----------------------------------------------------------------------------
5487 2005-05-05 Cluster File Systems, Inc. <info@clusterfs.com>
5489 NOTE: Lustre 1.4.2 uses an incompatible network protocol than previous
5490 versions of Lustre. Please update all servers and clients to
5491 version 1.4.2 or later at the same time. You must also run
5492 "lconf --write-conf {config}.xml" on the MDS while it is stopped
5493 to update the configuration logs.
5495 - fix for HPUX NFS client breakage when NFS exporting Lustre (5781)
5496 - mdc_enqueue does not need max_mds_easize request buffer on send (5707)
5497 - swab llog records of type '0' so we get proper header size/idx (5861)
5498 - send llog cancel req to DLM cancel portal instead of cb portal (5515)
5499 - fix rename of one directory over another leaking an inode (5953)
5500 - avoid SetPageDirty on 2.6 (5981)
5501 - don't re-add just-being-destroyed locks to the waiting list (5653)
5502 - when creating new directories, inherit the parent's custom
5503 striping settings if present parent (3048)
5504 - flush buffers from cache before direct IO in 2.6 obdfilter (4982)
5505 - don't hold i_size_sem in ll_nopage() and ll_ap_refresh_count (6077)
5506 - don't hold client locks on temporary worklist from l_lru (5666)
5507 - handle IO errors in 2.6 obdfilter bio completion routine (6046)
5508 - automatically evict dead clients (5921)
5509 - Update file size properly in create+truncate+fstat case (6196)
5510 - Do not unhash mountpoint dentries, do not allow removal of
5512 - Avoid lock ordering deadlock issue with write/truncate (6203,5654)
5513 - reserve enough journal credits in fsfilt_start_log for setattr (4554)
5514 - ldlm_enqueue freed-export error path would always LBUG (6149,6184)
5515 - don't reference lr_lvb_data until after we hold lr_lvb_sem (6170)
5516 - don't overwrite last_rcvd if there is a *_client_add() error (6086)
5517 - Correctly handle reads of files with no objects (6243)
5518 - lctl recover will also mark a device active if deactivate used (5933)
5520 - by default create 1 inode per 4kB space on MDS, per 16kB on OSTs
5521 - allow --write-conf on an MDS with different nettype than client (5619)
5522 - don't write config llogs to MDS for mounts not from that MDS (5617)
5523 - lconf should create multiple TCP connections from a client (5201)
5524 - init scripts are now turned off by default; run chkconfig --on
5525 lustre and chkconfig --on lustrefs to use them
5526 - upcalls are no longer needed for clients to recover to failover
5528 - add --abort-recovery option to lconf to abort recovery on device
5530 - add support for an arbitrary number of OSTs (3026)
5531 - Quota support protocol changes.
5532 - forward compatibility changes to wire structs (6007)
5533 - rmmod NALs that might be loaded because of /etc/modules.conf (6133)
5534 - support for mountfsoptions and clientoptions to the Lustre LDAP (5873)
5535 - improved "lustre status" script
5536 - initialize blocksize for non-regular files (6062)
5537 - added --disable-server and --disable-client configure options (5782)
5538 - introduce a lookup cache for lconf to avoid repeated DB scans (6204)
5539 - Vanilla 2.4.29 support
5540 - increase maximum number of obd devices to 520 (6242)
5541 - remove the tcp-zero-copy patch from the suse-2.4 series (5902)
5542 - Quadrics Elan drivers are now included for the RHEL 3 2.4.21 and
5543 SLES 9 2.6.5 kernels
5544 - limit stripes per file to 160 (the maximum EA size) (6093)
5546 2005-03-22 Cluster File Systems, Inc. <info@clusterfs.com>
5549 - don't LASSERT in ll_release on NULL lld with NFS export (4655, 5760)
5550 - hold NS lock when calling handle_ast_error->del_waiting_lock (5746)
5551 - fix setattr mtime regression from lovcleanup merge (4829, 5669)
5552 - workaround for 2.6 crash in ll_unhash_aliases (5687, 5210)
5553 - small ext3 extents cleanups and fixes (5733)
5554 - improved mballoc code, several small races and bugs fixed (5733, 5638)
5555 - kernel version 43 - fix remove_suid bugs in both 2.4 and 2.6 (5695)
5556 - avoid needless client->OST connect, fix handle mismatch (5317)
5557 - fix DLM error path that led to out-of-sync client, long delays (5779)
5558 - support common vfs-enforced mount options (nodev,nosuid,noexec) (5637)
5559 - fix several locking issues related to i_size (5492,5624,5654,5672)
5560 - don't move pending lock onto export if it is already evicted (5683)
5561 - fix kernel oops when creating .foo in unlinked directory (5548)
5562 - fix deadlock in obdfilter statistics vs. object create (5811)
5563 - use time_{before,after} to avoid timer jiffies wrap (5882)
5564 - shutdown --force/--failover stability (3607,3651,4797,5203,4834)
5565 - Do not leak request if server was not able to process it (5154)
5566 - If mds_open unable to find parent dir, make that negative lookup(5154)
5567 - don't create new directories with extent-mapping (5909, 5936)
5569 - fix lustre/lustrefs init scripts for SuSE (patch from Scali, 5702)
5570 - don't hold the pinger_sem in ptlrpc_pinger_sending_on_import
5571 - change obd_increase_kms to obd_adjust_kms (up or down) (5654)
5572 - lconf, lmc search both /usr/lib and /usr/lib64 for Python libs (5800)
5573 - support for RHEL4 kernel on i686 (5773)
5574 - provide error messages when incompatible logs are encountered (5898)
5576 2005-02-18 Cluster File Systems, Inc. <info@clusterfs.com>
5577 * version 1.4.0.10 (1.4.1 release candidate 1)
5579 - don't keep a lock reference when lock is not granted (4238)
5580 - unsafe list practices (rarely) led to infinite eviction loop (4908)
5581 - add per-fs limit of Lustre pages in page cache, avoid OOM (4699)
5582 - drop import inflight refcount on signal_completed_replay error (5255)
5583 - unlock page after async write error during send (3677)
5584 - handle missing objects in filter_preprw_read properly (5265)
5585 - no transno return for symlink open, don't save no-trasno open (3440)
5586 - don't try to complete elan receive that already failed (4012)
5587 - free RPC server reply state on error (5406)
5588 - clean up thread from ptlrpc_start_thread() on error (5160)
5589 - readahead could read extra page into cache that wasn't ejected (5388)
5590 - prevent races in class_attach/setup/cleanup/detach (5260)
5591 - don't dereference de->d_inode after l_dput of de (5458)
5592 - use "int" for stripe value returned from lock_to_stripe (5544)
5593 - mballoc allocation and error-checking fixes in 2.6 (5504)
5594 - block device patches to fix I/O request sizes in 2.6 (5482)
5595 - look up hostnames for IB nals (5602)
5596 - 2.6 changed lock ordering of 2 semaphores, caused deadlock (5654)
5597 - don't start multiple acceptors for the same port (5277)
5598 - fix incorrect LASSERT in mds_getattr_name (5635)
5599 - export a proc file for general "ping" checking (5628)
5600 - fix "lfs check" to not block when the MDS is down (5628)
5602 - service request history (4965)
5603 - put {ll,lov,osc}_async_page structs in a single slab (4699)
5604 - create an "evict_client" /proc entry on OSTs, like the MDS has
5605 - fix mount usage message, return errors per mount(8) (5168)
5606 - change grep [] to grep "[]" in tests so they work in more UMLs
5607 - fix ppc64/x86_64 spec to use %{_libdir} instead of /usr/lib (5389)
5608 - remove ancient LOV_MAGIC_V0 EA support (5047)
5609 - add "disk I/Os in flight" and "I/O req time" stats in obdfilter
5610 - align r/w RPCs to PTLRPC_MAX_BRW_SIZE boundary for performance (3451)
5611 - allow readahead allocations to fail when low on memory (5383)
5612 - mmap locking landed again, after considerable improvement (2828)
5613 - add get_hostaddr() to lustreDB.py for LDAP support (5459)
5615 2004-11-23 Cluster File Systems, Inc. <info@clusterfs.com>
5618 - send OST transaction number in read/write reply to free req (4966)
5619 - don't ASSERT in ptl_send_rpc() if we run out of memory (5119)
5620 - lock /proc/sys/portals/routes internal state, avoiding oops (4827)
5621 - the watchdog thread now runs as interruptible (5246)
5622 - flock/lockf fixes (but it's still disabled, pending 5135)
5623 - don't use EXT3 constants in llite code (5094)
5624 - memory shortage at startup could cause assertion (5176)
5626 - reorganization of lov code
5627 - single portals codebase
5629 - add extents/mballoc support (5025)
5630 - direct I/O reads in the obdfilter (4048)
5631 - kernel patches from LNXI for 2.6 (bluesmoke, perfctr, mtd, kexec)
5633 tbd Cluster File Systems, Inc. <info@clusterfs.com>
5636 - send OST transaction number in read/write reply to free req (4966)
5637 - don't ASSERT in ptl_send_rpc() if we run out of memory (5119)
5638 - lock /proc/sys/portals/routes internal state, avoiding oops (4827)
5639 - the watchdog thread now runs as interruptible (5246)
5640 - handle missing objects in filter_preprw_read properly (5265)
5641 - unsafe list practices (rarely) led to infinite eviction loop (4908)
5642 - drop import inflight refcount on signal_completed_replay error (5255)
5643 - unlock page after async write error during send (3677)
5644 - return original error code on reconstructed replies (3761)
5645 - no transno return for symlink open, don't save no-trasno open (3440)
5647 - add pid to ldlm debugging output (4922)
5648 - bump the watchdog timeouts -- we can't handle 30sec yet
5649 - extra debugging for orphan dentry/inode bug (5259)
5651 2004-11-16 Cluster File Systems, Inc. <info@clusterfs.com>
5654 - fix TCP_NODELAY bug, which caused extreme perf regression (5134)
5655 - allocate qswnal tx descriptors singly to avoid fragmentation (4504)
5656 - don't LBUG on obdo_alloc() failure, use OBD_SLAB_ALLOC() (4800)
5657 - fix NULL dereference in /proc/sys/portals/routes (4827)
5658 - allow failed mdc_close() operations to be interrupted (4561)
5659 - stop precreate on OST before MDS would time out on it (4778)
5660 - don't send partial-page writes before EOF from client (4410)
5661 - discard client grant for sub-page writes on large-page clients (4520)
5662 - don't free dentries not owned by NFS code, check generation (4806)
5663 - fix lsm leak if mds_create_objects() fails (4801)
5664 - limit debug_daemon file size, always print CERROR messages (4789)
5665 - use transno after validating reply (3892)
5666 - process timed out requests if import state changes (3754)
5667 - update mtime on OST during writes, return in glimpse (4829)
5668 - add mkfsoptions to LDAP (4679)
5669 - use ->max_readahead method instead of zapping global ra (5039)
5670 - don't interrupt __l_wait_event() during strace
5672 - add software watchdogs to catch hung threads quickly (4941)
5673 - make lustrefs init script start after nfs is mounted
5674 - fix CWARN/ERROR duplication (4930)
5675 - return async write errors to application if possible (2248)
5676 - add /proc/sys/portal/memused (bytes allocated by PORTALS_ALLOC)
5677 - print NAL number in %x format (4645)
5678 - update barely-supported suse-2.4.21-171 series (4842)
5679 - support for sles 9 %post scripts
5680 - support for building 2.6 kernel-source packages
5681 - support for sles km_* packages
5683 2004-10-07 Cluster File Systems, Inc. <info@clusterfs.com>
5686 - ignore -ENOENT errors in osc_destroy (3639)
5687 - notify osc create thread that OSC is being cleaned up (4600)
5688 - add nettype argument for llmount in #5d in conf-sanity.sh (3936)
5689 - reconstruct ost_handle() like mds_handle() (4657)
5690 - create a new thread to do import eviction to avoid deadlock (3969)
5691 - let lconf resolve symlinked-to devices (4629)
5692 - don't unlink "objects" from directory with default EA (4554)
5693 - hold socknal file ref over connect in case target is down (4394)
5694 - allow more than 32000 subdirectories in a single directory (3244)
5695 - fix blocks count for O_DIRECT writes (3751)
5696 - OST returns ENOSPC from object create when no space left (4539)
5697 - don't send truncate RPC if file size isn't changing (4410)
5698 - limit OSC precreate to 1/2 of value OST considers bogus (4778)
5699 - bind to privileged port in socknal and tcpnal (3689)
5701 - rate limit CERROR/CWARN console message to avoid overload (4519)
5702 - GETFILEINFO dir ioctl returns LOV EA + MDS stat in 1 call (3327)
5703 - basic mmap support (3918)
5704 - kernel patch series update from b1_4 (4711)
5706 2004-09-16 Cluster File Systems, Inc. <info@clusterfs.com>
5709 - avoid crash during MDS cleanup with OST shut down (2775)
5710 - fix loi_list_lock/oig_lock inversion on interrupted IO (4136)
5711 - don't use bad inodes on the MDS (3744)
5712 - dynamic object preallocation to improve recovery speed (4236)
5713 - don't hold spinlock over lock dumping or change debug flags (4401)
5714 - don't zero obd_dev when it is force cleaned (3651)
5715 - print grants to console if they go negative (4431)
5716 - "lctl deactivate" will stop automatic recovery attempts (3406)
5717 - look for existing locks in ldlm_handle_enqueue() (3764)
5718 - don't resolve lock handle twice in recovery avoiding race (4401)
5719 - revalidate should check working dir is a directory (4134)
5721 - don't always mark "slow" obdfilter messages as errors (4418)
5723 2004-08-24 Cluster File Systems, Inc. <info@clusterfs.com>
5726 - don't close LustreDB during write_conf until it is done (3860)
5727 - fix typo in lconf for_each_profile (3821)
5728 - allow dumping logs from multiple threads at one time (3820)
5729 - don't allow multiple threads in OSC recovery (3812)
5730 - fix debug_size parameters (3864)
5731 - fix mds_postrecov to initialize import for llog ctxt (3121)
5732 - replace config semaphore with spinlock (3306)
5733 - be sure to send a reply for a CANCEL rpc with bad export (3863)
5734 - don't allow enqueue to complete on a destroyed export (3822)
5735 - down write_lock before checking llog header bitmap (3825)
5736 - recover from lock replay timeout (3764)
5737 - up llog sem before sending rpc (3652)
5738 - reduce ns lock hold times when setting kms (3267)
5739 - change a dlm LBUG to LASSERTF, to maybe learn something (4228)
5740 - fix NULL deref and obd_dev leak on setup error (3312)
5741 - replace some LBUG about llog ops with error handling (3841)
5742 - don't match INVALID dentries from d_lookup and spin (3784)
5743 - hold dcache_lock while marking dentries INVALID and hashing (4255)
5744 - fix invalid assertion in ptlrpc_set_wait (3880)
5746 - add libwrap support for the TCP acceptor (3996)
5747 - add /proc/sys/portals/routes for non-root route listing (3994)
5748 - allow setting MDS UUID in .xml (2580)
5749 - print the stack of a process that LBUGs (4228)
5751 2004-07-14 Cluster File Systems, Inc. <info@clusterfs.com>
5754 - don't cleanup request in ll_file_open() on failed MDS open (3430)
5755 - make sure to unset replay flag from failed open requests (3440)
5756 - if default stripe count is 0, use OST count for inode size (3636)
5757 - update parent mtime/ctime on client for create/unlink (2611)
5758 - drop dentry ref in ext3_add_link from open_connect_dentry (3266)
5759 - free recovery state on server during a forced cleanup (3571)
5760 - unregister_reply for resent reqs (3063)
5761 - loop back devices mounting and status check on 2.6 (3563)
5762 - fix resource-creation race that can provoke i_size == 0 (3513)
5763 - don't try to use bad inodes returned from MDS/OST fs lookup (3688)
5764 - more debugging for page-accounting assertion (3746)
5765 - return -ENOENT instead of asserting if ost getattr+unlink race (3558)
5766 - avoid deadlock after precreation failure (3758)
5767 - fix race and lock order deadlock in orphan handling (3450, 3750)
5768 - add validity checks when grabbing inodes from l_ast_data (3599)
5770 - add /proc/.../recovery_status to obdfilter (3428)
5771 - lightweight CDEBUG infrastructure, debug daemon (3668)
5772 - change default OSC RPC parameters to be better on small clusters
5773 - turn off OST read cache for files smaller than 32MB
5774 - install man pages and include them in rpms (3100)
5775 - add new init script for (un)mounting lustre filesystems (2593)
5776 - run chkconfig in %post for init scripts (3701)
5777 - drop scimac NAL (unmaintained)
5779 2004-06-17 Cluster File Systems, Inc. <info@clusterfs.com>
5782 - clean kiobufs before and after use (3485)
5783 - strip trailing '/'s before comparing paths with /proc/mounts (3486)
5784 - remove assertions to work around "in-flight rpcs" recovery bug (3063)
5785 - change init script to fail more clearly if not run as root (1528)
5786 - allow clients to reconnect during replay (1742)
5787 - fix ns_lock/i_sem lock ordering deadlock for kms update (3477)
5788 - don't do DNS lookups on NIDs too small for IP addresses (3442)
5789 - re-awaken ptlrpcd if new requests arrive during check_set (3554)
5790 - fix cond_resched (3554)
5791 - only evict unfinished clients after recovery (3515)
5792 - allow bulk resend, prevent data loss (3570)
5793 - dynamic ptlrpc request buffer allocation (2102)
5794 - don't allow unlinking open directory if it isn't empty (2904)
5795 - set MDS/OST threads to umask 0 to not clobber client modes (3359)
5796 - remove extraneous obd dereference causing LASSERT failure (3334)
5797 - don't use get_cycles() when creating temp. files on the mds (3156)
5798 - hold i_sem when setting i_size in ll_extent_lock() (3564)
5799 - handle EEXIST for set-stripe, set proper directory name (3336)
5801 - servers can dump a log evicting a client - lustre.dump_on_timeout=1
5802 - fix ksocknal_fmb_callback() error messages (2918)
5804 2004-05-27 Cluster File Systems, Inc. <info@clusterfs.com>
5807 - don't copy lvb into (possibly NULL) reply on error (2983)
5808 - don't deref dentry after dput, don't free lvb on error (2922)
5809 - use the kms to determine writeback rpc length (2947)
5810 - increment oti_logcookies when osc is inactive (2948)
5811 - update client's i_blocks count via lvb messages (2543)
5812 - handle intent open/close of special files properly (1557)
5813 - mount MDS with errors=remount-ro, like obdfilter (2009)
5814 - initialize lock handle to avoid ASSERT on error cleanup (3057)
5815 - don't use cancelling-locks' kms values (2947)
5816 - use highest lock extent for kms, not last one (2925)
5817 - don't dereference ERR_PTR() dentry in error handling path (3107)
5818 - fix thread race in portals_debug_dumplog() (3122)
5819 - create lprocfs device entries at setup instead of at attach (1519)
5820 - common AST error handler, don't evict client on completion race (3145)
5821 - zero nameidata in detach_mnt in 2.6 (3118)
5822 - verify d_inode after revalidate_special is valid in 2.6 (3116)
5823 - use lustre_put_super() to handle zconf unmounts in 2.6 (3064)
5824 - initialize RPC timeout timer earlier for 2.6 (3219)
5825 - don't dereference NULL reply buffer if mdc_close was never sent (2410)
5826 - print nal/nid for unknown nid (3258)
5827 - additional checks for oscc recovery before doing precreate (3284)
5828 - fix ll_extent_lock() error return code for 64-bit systems (3043)
5829 - don't crash in mdc_close for bad permissions on open (3285)
5830 - zero i_rdev for non-device files (3147)
5831 - clear page->private before handing to FS, better assertion (3119)
5832 - tune the read pipeline (3236)
5833 - fix incorrect decref of invalidated dentry (2350)
5834 - provide read-ahead stats and refine rpc in flight stats (3328)
5835 - don't hold journal transaction open across create RPC (3313)
5836 - update atime on MDS at close time (3265)
5837 - close LDAP connection when recovering to avoid server load (3315)
5838 - update iopen-2.6 patch with fixes from 2399,2517,2904 (3301)
5839 - don't leak open file on MDS after open resend (3325)
5840 - serialize filter_precreate and filter_destroy_precreated (3329)
5841 - loop device shouldn't call sync_dev() for nul device (3092)
5842 - clear page cache after eviction (2766)
5843 - resynchronize MDS->OST in background (2824)
5844 - refuse to mount the same filesystem twice on same mountpoint (3394)
5845 - allow llmount to create routes for mounting behind routers (3320)
5846 - push lock cancellation to blocking thread for glimpse ASTs (3409)
5847 - don't call osc_set_data_with_check() for TEST_LOCK matches (3159)
5848 - fix rare problem with rename on htree directories (3417)
5850 - allow default OST striping configuration per directory (1414)
5851 - fix compilation for qswnal for 2.6 kernels (3125)
5852 - increase maximum number of MDS request buffers for large systems
5853 - change liblustreapi to be useful for external progs like lfsck (3098)
5854 - increase local configuration timeout for slow disks (3353)
5855 - allow configuring ldlm AST timeout - lustre.ldlm_timeout=<seconds>
5857 2004-03-22 Cluster File Systems, Inc. <info@clusterfs.com>
5860 - fixes for glimpse AST timeouts / incorrectly 0-sized files (2818)
5861 - don't overwrite extent policy data in reply if lock was blocked (2901)
5862 - drop filter export grants atomically with removal from device (2663)
5863 - del obd_self_export from work_list in class_disconnect_exports (2908)
5864 - don't LBUG if MDS recovery times out during orphan cleanup (2530)
5865 - swab reply message in mdc_close, other PPC fixes (2464)
5866 - fix destroying of named logs (2325)
5867 - overwrite old logs when running lconf --write_conf (2264)
5868 - bump LLOG_CHUNKSIZE to 8k to allow for larger clusters (2306)
5869 - fix race in target_handle_connect (2898)
5870 - mds_reint_create() should take same inode create lock (2926)
5871 - correct journal credits calculated for CANCEL_UNLINK_LOG (2931)
5872 - don't close files for self_export to avoid uninitialized obd (2936)
5873 - allow MDS with the same name as client node (2939)
5874 - hold dentry reference for closed log files for unlink (2325)
5875 - reserve space for all logs during transactions (2059)
5876 - don't evict page beyond end of stripe extent (2925)
5877 - don't oops on a deleted current working directory (2399)
5878 - handle hard links to targets without a parent properly (2517)
5879 - don't dereference NULL lock when racing during eviction (2867)
5880 - don't grow lock extents when lots of conflicting locks (2919)
5882 2004-03-04 Cluster File Systems, Inc. <info@clusterfs.com>
5885 - account for cache space usage on clients to avoid data loss (974)
5886 - lfsck support in lustre kernel code (2349)
5887 - reduce journal credits needed for BRW writes (2370)
5888 - orphan handling to avoid losing space on client/server crashes
5889 - ptlrpcd can be blocked, stopping ALL progress (2477)
5890 - use lock value blocks to assist in proper KMS, faster stat (1021)
5891 - takes i_sem instead of DLM locks internally on obdfilter (2720)
5892 - recovery for initial connections (2355)
5893 - fixes for mds_cleanup_orphans (1934)
5894 - abort_recovery crashes MDS in b_eq (mds_unlink_orphan) (2584)
5895 - block all file creations until orphan recovery completes (1901)
5896 - client remove rq_connection from request struct (2423)
5897 - conf-sanity test_5, proper cleanup in umount log not availale (2640)
5898 - recovery timer race (2670)
5899 - mdc_close recovey bug (2532)
5900 - ptlrpc cleanup bug (2710)
5901 - mds timeout on local locks (2588)
5902 - namespace lock held during RPCs (2431)
5903 - handle interrupted sync write properly (2503)
5904 - don't try to handle a message that hasn't been replied to (2699)
5905 - client assert failure during cleanup after abort recovery (2701)
5906 - leak mdc device after failed mount (2712)
5907 - ptlrpc_check_set allows timedout requests to complete (2714)
5908 - wait for inflight reqs when ptlrpcd finishes (2710)
5909 - make sure unregistered services are removed from the srv_list
5910 - reset bulk XID's when resending them (caught by 1138 test)
5911 - unregister_bulk after timeout
5912 - fix lconf error (2694)
5913 - handle write after unfinished setstripe, stripe-only getstripe (2388)
5914 - readahead locks pages, leaves pending causing memory pressure (2673)
5915 - increase OST request buffers to 4096 on large machines (2729)
5916 - fix up permission of existing directories in simple_mkdir (2661)
5917 - init deleted item, add assertions ptlrpc_abort_inflight() (2725)
5918 - don't assign transno to errored transactions (2742)
5919 - don't delete objects on OST if given a bogus objid from MDS (2751)
5920 - handle large client PAGE_SIZE readdir on small PAGE_SIZE MDS (2777)
5921 - if rq_no_resend, then timeout request after recovery (2432)
5922 - fix MDS llog_logid record size, 64-bit array alignment (2733)
5923 - don't call usermode_helper from ptlrpcd, DEFAULT upcall (2773)
5924 - put magic in mount.lustre data, check for bad/NULL mount data (2529)
5925 - MDS recovery shouldn't delete objects that it has given out (2730)
5926 - if enqueue arrives after completion, don't clobber LVB (2819)
5927 - don't unlock pages twice when trigger_group_io returns error (2814)
5928 - don't deref NULL rq_repmsg if ldlm_handle_enqueue failed (2822)
5929 - don't write pages to disk if there was an error (1450)
5930 - don't ping imports that have recovery disabled (2676)
5931 - take buffered bytes into account when balancing socknal conn (2817)
5932 - hold a DLM lock over readdir always, use truncate_inode_pages (2706)
5933 - reconnect unlink llog connection after MDS reconnects to OST (2816)
5934 - remove little-endian swabbing of llog records (1987)
5935 - set/limit i_blksize to LL_MAX_BLKSIZE on client (2884)
5936 - retry reposting request buffers if they fail (1191)
5937 - grow extent at grant time to avoid granting a revoked lock (2809)
5938 - lock revoke doesn't evict page if covered by a second lock (2765)
5939 - disable VM readahead to avoid reading outside lock extents (2805)
5941 - return LL_SUPER_MAGIC from statfs for the filesystem type (1972)
5942 - updated kernel patches for hp-2.4.20 kernel (2681)
5944 2004-02-07 Cluster File Systems, Inc. <info@clusterfs.com>
5947 - fix truncated write corruption (2366)
5948 - fix for failed assertion in iopen_connect_dentry (1792,2517)
5950 - don't flag the ptlrpcd thread with PF_MEMALLOC (2636)
5951 - ensure len(uuid) < 37 in lmc (1171)
5952 - fix ia64 OOPS in llog_test (2255)
5953 - zero end of page at obdfilter for partial page writes (2648)
5954 - don't leave stale dentries around after renames (bug 2428)
5955 - fix timeouts when evicting a client with a single lock held (2642)
5956 - set deadline for the initial HELLO message to drain (2634)
5957 - print out dotted-quad IP addresses in the socknal (2302)
5959 - additional debugging for MDS client eviction problem (2443)
5960 - fix mkfsoptions support for osts (2603, 2604)
5962 2004-01-27 Cluster File Systems, Inc. <info@clusterfs.com>
5965 - add series for the vanilla 2.6.0 kernel
5966 - add series for the vanilla 2.4.24 kernel
5967 - add series for a cray x86/64 UL kernel drop
5968 - fix xattr patches for the vanilla 2.4.19 series
5970 - generate true UUIDs in lmc (1171)
5971 - have portals stack dumping break in UML (2466)
5972 - avoid bad dchild deref; avoid inum lock w/o creation (2362)
5973 - allocate with _NOFS in ldlm to avoid deadlock (1933)
5974 - wake callback waiting threads on client eviction (2460)
5975 - Add --ptldebug and --subsystem to lmc (1719)
5976 - update assertion to allow safe interrupt allocation
5977 - set rq_no_resend for cancel requests (2432)
5978 - recalculate ptlrpcd timeout after resend (2494)
5979 - call vfs_rmdir when removing pending directories (2368)
5980 - fix renaming a file to itself (2429)
5981 - lmc creates a default one-stripe lov (2454)
5982 - expand procfs space to handle large clusters (2326)
5983 - increase UML stack to avoid overflow
5984 - update lconf's list of debug and subsystem masks
5985 - fix lfs find --obd (2510)
5986 - /proc tunable for disabling filter read caching (2591)
5987 - stop rpm packages from altering slapd.conf (2301)
5988 - disable nagle in the socknal under 0conf (2578)
5989 - choose mds inode size based on stripe count (2572)
5990 - fix kernel-source rpm problems (2516)
5992 - add --disable-doc to avoid pdf generation (2421)
5993 - update documentation, tests, type-os, comments
5994 - avoid format warnings on ia64
5995 - remove the TOE NAL
5996 - tiny code cleanups by removing unused fields
5998 2004-01-07 Cluster File Systems, Inc. <info@clusterfs.com>
6001 - fix obvious semaphore misuse in as-yet-unused setattr path (2348)
6002 - remove the most blatant lies from BUILDING file (2371)
6003 - change default debug level to reasonable production setting
6004 - reduce client side cache size to reduce cache flush time
6005 - reduce max RPCs in flight to avoid unnecessary file fragmentation
6006 - make TCP zerocopy and pinger support enabled by default (2476)
6007 - sync writes completed after process exits caused crashes (2319)
6008 - maintain correct mount count on the MDS (2356)
6009 - backout 1557, because 2316 wasn't really fixed
6010 - better file I/O statistics gathering in /proc
6011 - don't take unnecessary, deadlock-inducing bug in readpage (2383)
6012 - another kernel patch to fix zero-copy TCP function export
6013 - don't take duplicate lock when processing re-sent getattr (2420)
6014 - lctl uses obd_self_export instead of creating new conn (2353)
6015 - MDS/OST recovery case which requires object creation asserted (2425)
6016 - move lfs from /usr/sbin to /usr/bin in packages
6017 - fix race between mds_client_add and mds_client_free (2417)
6018 - use kmalloc instead of slabs in portals (2430)
6019 - don't create duplicate records when a failover MDS is present (2442)
6020 - remove unnecessary mount age check (2332)
6021 - don't remove directory inodes from locks prematurely (2451)
6022 - don't break if MDS service name is the same as hostname (2103)
6023 - fix races in client write RPC generation when cache full (2482)
6025 2003-12-13 Cluster File Systems, Inc. <info@clusterfs.com>
6028 - remove now-unused request->rq_obd (278)
6029 - if an allocation fails, print out how much memory we've used (1933)
6030 - use PORTAL_SLAB_ALLOC for structures, to get GFP_MEMALLOC (1933)
6031 - add the "configurable stack size" patch to most series files (1256)
6032 - ability to write large log records, for 100+ OST configs (2306)
6033 - fix NULL deref when filter_prep fails (2314)
6034 - fix operator precedence error in filter_sync
6035 - dynamic allocation of socknal TX descriptors (2315)
6036 - fix a missed case in the GFP_MEMALLOC patch, can cause deadlock (2310)
6037 - fix gcc 2.96 compilation problem in xattr kernel patch (2294)
6038 - ensure that CWARN messages in Portals always get to the syslog
6039 - __init/__exit are not for prototype decls (ldlm_init/exit)
6040 - x86-64 compile warning fixes
6041 - fix gateway LMC keyword conflict (2318)
6042 - fix MDS lock inversions in getattr/reint paths (1844)
6043 - fix a rare lock re-ordering bug, which caused deadlock (2322)
6044 - fix i_sem/journal inversion in fsfilt_ext3_write_record (2306)
6045 - DLM race condition prevented some lock evictions (2328)
6046 - ENOMEM detection and retry on socknal sends (2230)
6047 - use GFP_NOFS throughout Lustre, to combat ENOMEM (2230)
6048 - move osc_rpcd into ptlrpc, for use in MDC and others (2329)
6049 - protect MDS inode fsdata with stronger locking; fixes assertion (2313)
6050 - better error messages when a client is rejected during recovery (1505)
6051 - avoid cancelling locks which were never granted, after failure (2330)
6052 - fix i_sem/journal inversion in mds_client_add (2333)
6053 - fix truncate/getattr lock cycle deadlock (2334)
6054 - use rpcd to send close; allows resend after timeout, avoid leak (1897)
6055 - fix two rare exit paths which could leak an l_lock() ref (2321)
6056 - fencepost error in MDS/OST orphan recovery (2226)
6057 - make log record alignment 8 bytes (1988)
6058 - lstripe now fails when requested offset > ost_count (2237)
6059 - ensure that all kernel series have a complete list.h (1607)
6060 - fix crashes in special-file operations (2316)
6061 - lctl create/brw OID mismatch, caused by obsolete filter loop (2339)
6063 - allow configurable automake binary, for testing new versions
6064 - small update to the lfs documentation
6066 2003-12-03 Cluster File Systems, Inc. <info@clusterfs.com>
6068 * fix negative export reference count in fsfilt_sync (2312)
6070 2003-12-01 Cluster File Systems, Inc. <info@clusterfs.com>
6071 * release candidate 0.9.1
6073 - orphans are moved into the PENDING directory for possible recovery
6074 - replayed opens now open by fid for orphan/rename safety (1042)
6075 - last close of an orphan inode generates a transno (683)
6076 - chdir() and mount() now pin the directory entry (1020)
6077 - avoid CERROR in normal ll_setattr_raw() error case (1500)
6078 - discard very old requests without processing them (1502)
6079 - remove some common, well-understood CERRORs (1505)
6080 - require O_DIRECT I/O to be page-sized to workaround IA64 crash (1609)
6081 - clear "grant" flags in OST replies until OST grant code lands (1644)
6082 - fix read performance by not clobbering i_blksize on client (1598)
6083 - fix __ldlm_handle2lock oops by not dereferencing lock after PUT (1625)
6084 - make LRU size a /proc tunable, clears locks when reduced (707)
6085 - fix some lprocfs rot that prevented ptlbd from loading (1732)
6086 - server locks take references on exports now (1558)
6087 - build fixes for 2.4.20-rh trees (1663)
6088 - return an error from lov_create if all OSCs are inactive (1751)
6089 - fix import levels when a reconnect happens without a timeout (1597)
6090 - exit early from mds_open if we get a lookup error (1749)
6091 - partial page read at EOF wouldn't wait for disk before sending (1642)
6092 - avoid NULL deref in obdfilter when reading page past EOF (1592)
6093 - avoid LASSERT in ll_intent_lock if server failed very early (1090)
6094 - fix LBUG in ll_it_open_error with rc = -2 (1861)
6095 - write/truncate lock inversion (1639)
6096 - Don't auto-load obdclass, portals modules during cleanup (1495)
6097 - fix timestamps from jumping to "now" (1763)
6098 - extra journal assertions (1648)
6099 - add an extra multiunlink test (1771)
6100 - fix read_record/write_record API (1776)
6101 - fix leak of offset_extent, possible incorrect i_size later (1772)
6102 - fix lasserts in mis-matched transnos during open-unlink testing (1541)
6103 - Debugging for the kqswnal_get_idle_tx problems (1820)
6104 - Allow recovery to be attempted multiple times (1536)
6105 - Write out MDS last_rcvd file after it is first created (1600)
6106 - Fix tx_descriptor leak in failed transmit situations (1827)
6107 - ext3 journaling fixes for assertion failure after IO error (1871)
6108 - class_export_put() on freed export after completion AST error (1896)
6109 - Fix revalidate looping in VFS (1322)
6110 - Don't access a freed export during MDS_REINT timeout (1521)
6111 - Add open-unlink recovery support on the MDS (1673,1764)
6112 - Return an error if no MDS data was read from last_rcvd (1946)
6113 - Fix for lookup "." or ".." crash on error (1932,1931,1935)
6114 - Don't setup a disk device that doesn't match exported UUID (317)
6115 - Reduce bulk RPC timeout to avoid cascading client/OST failures (1845)
6116 - avoid committing NULL handle in force close
6117 - local.sh is now a one-stripe LOV configuration
6118 - POSIX utime.4 -EPERM on FIFO not owned by user (56)
6119 - fix ext3 htree duplicate directory entry corruption (1516)
6120 - POSIX creat.13, fstat.1, open.18, stat.3 new file atime/mtime (2020)
6121 - update to new LOV EA format (2097)
6122 - interoperability for different PAGE_SIZE/wordsize (686,1821,1343,2042)
6124 2003-06-15 Phil Schwan <phil@clusterfs.com>
6127 - imports and exports cleanup too early, need refcounts (349, 879, 1045)
6128 - per-import/export recovery handling (958, 931, 959)
6129 - multiple last-rcvd slots, for serving multiple FSes (949)
6130 - connections are again shared between multiple imp/exports (963, 964)
6131 - "umount -f" would hang if any requests needed to be sent (393, 978)
6132 - avoid pinning large req buffer by copying for queued messages (989)
6133 - add "uuid" to "lctl device" command to help upcalls (991)
6134 - "open" RPCs with transnos would confuse recovery counters (1037)
6135 - do proper endian conversion of all wire messages (288, 340, 891)
6136 - remove OST bulk get LBUGs, fix ost_brw_write cleanup (1126)
6137 - call waiting locks callback from LDLM recovery thread (1127, 1151)
6138 - fix ptlrpc_connection leak in target_handle_connect (1174)
6139 - fix import refcounting bug in OST and MDS cleanup (1134)
6140 - if an invalid-at-open-time OSC returned before close(), LBUG (1150)
6141 - fix very unlikely obd_types race condition (501)
6142 - remove osc_open hack for echo_client (1187)
6143 - we leaked exports/dlmimps for forcibly disconnected clients (1143)
6144 - a failure in read_inode2 leads to deadlock (1139)
6145 - cancel ack-locks as soon as transaction is committed (1072)
6146 - fix major leaks and crashes in the bulk I/O path (937, 1057)
6147 - make sure to commitrw after any preprw to avoid deadlock (1162)
6148 - failing to execute a file in a lustre FS would lock inode (1203)
6149 - small DEBUG_REQ fix to avoid dereferencing a NULL (1227)
6150 - don't ASSERT while cleaning up an incompletely-setup obd (1248)
6151 - obd_uuid2tgt would walk off the end of the list (1255)
6152 - on IA64 the osc would give portals incorrect bulk size (1258)
6153 - fix debug daemon ioctl interface; allows daemon on ia64 (1274)
6154 - fix lock inversion caused by new llite matching code (1282)
6155 - limit the number of dirty pages on a client to 10MB (1286)
6156 - timed out locks were not being corrected cancelled (1289)
6157 - fix O_DIRECT above 4GB on IA-32 (1292)
6158 * major user-visible changes
6159 - fail out/fail over policy now controlled by the upcall (993)
6161 - add OBD_PING to check server availability and failure (954)
6162 - lustre messages are now sent in sending host order (288, 340, 891)
6163 - add eadatalen to MDS getattr reply (340)
6164 - OST read replies may contain second buffer, with per-page status (593)
6166 2003-03-11 Phil Schwan <phil@clusterfs.com>
6169 - LDLM_DEBUG macro fix, for gcc 3.2 (850)
6170 - failed open()s could cause deadlock; fixed (867, 869)
6171 - stop cancelling OST locks when files are closed (481)
6172 - overlapping XID spaces caused network corruption (851, 853)
6173 - fix unsafe fsfilt counter arithmetic; change to atomic_t
6174 - setattr_raw added, to do single-RPC, server-side setattrs
6175 - lmc/lconf syntax change for OST UUIDs
6176 - fix crashy race condition between ptlrpc_free_req and osc_close
6177 - don't use request in mdc_enqueue if we hit a timeout (889)
6178 - don't set the inode i_size for regular files from the MDS (896)
6179 - handle out of order completion AST (842)
6180 - don't LBUG if a lock request times out after receiving AST (913)
6181 - avoid d_rehash race in ll_find_alias by rehashing inside dcache_lock
6182 - if a bad lock AST arrives, send an error instead of dropping entirely
6183 - return 0 from revalidate2 if ll_intent_lock returns -EINTR (912)
6184 - fix leak in bulk IO when only partially completed (899, 900, 926)
6185 - fix O_DIRECT for ia64 (55)
6186 - (almost) eliminate Lustre-kernel-thread effects on load average (722)
6187 - C-z after timeout could hang a process forever; fixed (977)
6189 - client-side I/O cache (678, 924, 929, 941, 970)
6191 - READPAGE and SETATTRs which don't take server-side locks get
6194 2003-02-11 Phil Schwan <phil@clusterfs.com>
6197 - Fix ldlm_lock_match on the MDS to avoid matching remote locks (592)
6198 - Fix fsfilt_extN_readpage() to read a full page of directory
6199 entries, or fake the remainder if PAGE_SIZE != blocksize (500)
6200 - Avoid extra mdc_getattr() in ll_intent_lock when possible (534, 604)
6201 - Fix imbalanced LOV object allocation and out-of-bound access (469)
6202 - Most intent operations were removed, in favour of a new RPC mode
6203 that does a single RPC to the server and bypasses most of the VFS
6204 - All LDLM resource ID arrays were removed in favour of ldlm_res_id
6205 - Aggressively cancel local locks on DLM servers
6206 - mds_reint_unlink sends EA to the client if it's the last nlink.
6207 client uses that EA to unlink OST objects.
6208 - mds_reint_{rename,unlink,link} were rewritten to take ordered locks
6209 - recursive symlinks were fixed (439)
6210 - fixed NULL deref in DEBUG_REQ
6211 - filter_update_lastobjid no longer calls sync, which annoyed extN
6212 - fixed multi-client small-writes to a single file problem (445)
6213 - fixed mtime updates during file writes (607)
6214 - fixed vector writes on obdfilter causing problems when ENOSPC (670)
6215 - fixed bug in obd_brw_read/write() (under guise of testing 367)
6216 - fixed Linux OST size reporting problem (444, 656)
6217 - OST now updates object mtime with writes or setattr (607, 619)
6218 - client verifies file size before zeroing page past EOF (445)
6219 - OST now writes last allocated objid to disk with allocation (108)
6220 - LOV on echo now works (409)
6222 - mds_reint_unlink sends a new buffer, with the EA included. this
6223 buffer is only valid if body->valid & OBD_MD_FLEASIZE, which is only
6224 set if a regular file was being unlinked, and it was the last link
6225 - use PtlGet from the target for bulk writes (315)
6226 - OST now updates object mtime with writes or setattr (607, 619)
6227 - LDLM now has a grant-time callback to revalidate locked items, if
6229 - Many MDS operations were reorganized to combat race conditions
6231 - Merge b_intel branch (updated lprocfs code) - now at /proc/fs/lustre
6232 - configure check to avoid gcc version 2.96 20000731-2.96-98) (606)
6234 2003-01-06 Andreas Dilger <adilger@clusterfs.com>
6237 - Fully reactivate OST imports after reconnection (512, others)
6238 - Make sure client sees our -ENOTCONN from mds_handle (513 - partial)
6239 - More graceful error handling for truncating on dead OST (515)
6240 - Don't error out unless we're actually accessing dead stripes (474)
6241 - Fix garbage sizes when stripes are missing (410)
6242 - LRU counters were broken, causing constant lock purge (433, 432)
6243 - garbage on read from stripes with failed OSTs (441)
6244 - mark OSCs as active before reconnecting during recovery (438)
6245 - lov_enqueue and lov_cancel need to handle inactive OSTs (403)
6246 - lfind did not preserve OST order in output (443)
6247 - symlinks cause hung clients, incorrect data (439)
6248 - stop dereferencing request after dropping refcount (457)
6249 - don't LASSERT(spin_is_locked) on non-SMP (455)
6250 - fixes for many rename() bugs
6251 - fstat didn't correctly synchronize attributes (399)
6252 - server must handle lock cancellation during blocking AST prep (487)
6253 - bulk descriptors were free()d too soon (511)
6254 - fix paths in lconf, which would load incorrect modules (451, 507)
6255 - fix confusing lconf 'host not found' error message (386)
6256 - fix lock order deadlock on OST (O/R i_sem before journal ops, 478)
6257 - fix race condition in mdc_blocking_ast() for inode access (526)
6258 - fix lov_unpackmd() unpacking wrong number of stripes (537)
6259 - fix lov_set_osc_active() marking wrong OSC inactive (440)
6260 - fix bad lstripe lov_unpackmd() assertion (fix layering too) (527)
6261 - fix multiple writes of stripe MD to MDS (358, maybe 519)
6262 - fix lstripe in several ways (kernel side) (527)
6263 - fix request leak in ldlm_cli_enqueue (262)
6264 - incorrect OSC was marked inactive after OST failure
6265 - call mds_fs_cleanup before unmounting filesystem (524)
6266 - fix races between taking ns_lock and ldlm_lock_change_resource
6267 - fix races updating LOV export open file list
6268 - fix lov_enqueue error path, avoid decref-ing bad lock handle
6269 - fix recovery NULL deref in ldlm_cli_cancel_unused
6270 - fix some DLM races by using new hash table for lock handles (419)
6271 - permit the client to specify desired inodes, at replay
6272 - duplicate requests when we queue them for replay reintegration
6273 - fix last_rcvd offset calculation
6274 - sync after each recovered transaction, so we always make progress
6275 - never, not always, ERESTART requests without transnos
6276 - store the lov_desc in the MDS, so we don't depend on getlovinfo to
6278 - skip replay if the MDS says that the client is already connected
6279 - don't check for a recovery-enabled export to match lctl's UUID
6280 - don't INC_USE_COUNT for phantom exports
6281 - don't crash when cleaning up phantom exports (567)
6282 - don't double-finish or set replay data for errored mdc_open requests
6283 - abort requests when they time out, so we don't get old replies
6284 - send/receive replies for AST messages again
6285 - if the client says that it doesn't have the lock, cancel it on the
6287 - if we timeout during I/O, don't try to cancel an in-use lock; instead
6288 mark it as destroyed, it will all work out when decref is called
6289 - fix module use counts (22, 581)
6291 - ASTs now expect a reply (server cancels lock on error reply)
6293 2002-12-02 Andreas Dilger <adilger@clusterfs.com>
6296 - fix many simultaneous client startup (392)
6297 - fix dentry->d_it clobbering
6298 - credentials weren't being shipped for readdir/getattr operations
6299 - remove invalid assertions triggered during some concurrent MD
6301 - proper Lustre versions added (336, 389)
6302 - fix memory leak for create error case (398)
6303 - fix LOV locking bug that would get cli/srv out of sync
6304 - fix echo client over LOV (409)
6305 - fix dbench 2, extN refcount problem (170, 258, 356, 418)
6306 - fix double-O_EXCL intent crash (424)
6307 - avoid sending multiple lock CANCELs (352)
6309 - MDS can do multi-client recovery (modulo bugs in new code)
6311 - many updates, edits, cleanups
6313 2002-11-18 Phil Schwan <phil@clusterfs.com>
6316 - fix null d_it dereference (346)
6317 - fix full OST/dbench hang (333)
6318 - fix permission problem with file removal (286)
6319 - fix removal of OSCs from LOV when they fail
6320 - fix NULL deref during bulk timeout (214)
6321 - fix problems related to multiple filesystems on one MDS (241)
6322 - fixed serious subtle metadata locking bugs
6323 - free locks on clients when inodes are removed due to memory
6325 - fix inode pointer in lock data (285)
6326 - partial support for multiple MDS on a single host (241)
6327 - data locks weren't cancelled at clear_inode time (290, 311)
6328 - intent locks could lead to unbounded lock growth (205)
6329 - added a maximum lock count, an LRU list, and a flusher
6330 - fix multiple rename (365)
6331 - properly abstracted the echo client
6332 - OSC locked 1 byte too many; fixed
6333 - rewrote brw callback code:
6334 - fixed recovery bugs related to LOVs (306)
6335 - fixed too-many-pages-in-one-write crash (191)
6336 - fixed (again) crash in sync_io_timeout (214)
6337 - probably fixed callback-related race (385)
6339 - Add capability to MDS protocol
6340 - LDLM cancellations and callbacks on different portals
6342 2002-10-28 Andreas Dilger <adilger@clusterfs.com>
6345 - limit client IOV size to PTL_MD_MAX_IOV (611336, 191)
6346 - defer open object destruction to close time (601981, 138)
6347 - open/close OST file handle in obdo (OBD_MD_FLHANDLE) (601981, 138)
6348 - move LDLM_ENQUEUE/CONVERT back to MDS portal (625069)
6349 - abstract ll_lookup2, fix ll_revalidate2 to use abstraction (256)
6350 - don't call obd_setattr in ll_file_release for destroyed objects
6351 * protocol change to lustre_msg: move |version| and add |flags|
6352 * protocol change to osc_punch: "start" in "o_size", "end" in "o_blocks"
6353 * lock replay: for LDLM_FL_REPLAY trust client to do right thing
6354 * added replay of create, unlink, link and rename operations during
6355 MDS failover; recovery should be much more robust now
6356 * remove failed OSCs from LOVs (only lov_create uses this so far)
6357 * the lustre-HOWTO was brought (more) up to date (582544)
6359 2002-10-23 Phil Schwan <phil@clusterfs.com>
6362 - in-use dentries weren't being reused properly (617851)
6363 - prevent multiple LDLM setup (599178)
6364 - fix LOV size calculations for truncate (617853)
6365 - fix client handling of MDS intent errors (POSIX)
6366 - fix permission bug in lovstripe.c test (624321)
6367 - fix MDS thread deadlock - move LDLM handler to DLM portal (625069)
6368 - truncate past end of file could corrupt data
6369 - proper cleanup after timeouts, crashes, etc (592524, 550815)
6370 - a race in recovery could return ETIMEDOUT to apps (623947)
6371 - building outside the source directory was fixed
6372 * the lustre-HOWTO was brought (more) up to date (582544)
6373 * major progress was made on recovery functionality
6375 2002-10-10 Phil Schwan <phil@clusterfs.com>
6378 - recovery deadlock fix
6379 - rm -rf causes LBUG fix (617817)
6380 - file open by multiple tasks fix (618962)
6381 - directory permissions bugs (602707 and 620007)
6382 - journal_stop fixed with locking (611313)
6383 - O_APPEND failures resolved (618273, perhaps 614459)
6384 - lconf PATH fix (619770)
6385 - IA64 build fix (621450)
6386 - RPC buffer sizes scale with amount of memory
6388 2002-10-01 Phil Schwan <phil@clusterfs.com>
6391 - locks would be cancelled without throwing away data pages,
6392 resulting in inconsistent data (605627)
6393 - inode attributes were not always being refreshed (605627, 612449)
6394 - lconf now continues to cleanup after lctl reports an error
6395 - MDS now enforces user permissions (602707)
6396 - lprocfs cleanup fixed, but not yet enabled (614157)
6397 - fixed infinite server hang, should a client not respond to an AST
6398 - avoid going into recovery if user calls readlink() with a buffer
6399 that's too small (613941)
6400 - AST RPCs no longer require replies (614867) -- this may be changed
6401 - don't crash server if client sends an IOV that's too big (611336)
6402 - fixed lock conversion deadlock (611892)
6403 - fixed the following of symlinks (614622)
6404 * recovery: the server can remove locks from a client that dies, other
6405 clients can make progress
6406 * more extN patch fixes
6407 * compile-time configurable ptlrpc buffer allocations
6409 - collaborative read cache document
6410 - Lustre Lite Performance CDR document-in-progress
6412 2002-09-20 Andreas Dilger <adilger@clusterfs.com>
6415 - fix typo in patch-2.4.18
6417 2002-09-20 Andreas Dilger <adilger@clusterfs.com>
6420 - clear ptlrpc request each time in handle_incoming_request()
6421 - unlink of files now destroys the object on the OST
6423 2002-09-19 Peter Braam <braam@clusterfs.com>
6425 * add hard link support
6426 * change obdfile creation method
6427 * kernel patch changed
6429 2002-09-19 Peter Braam <braam@clusterfs.com>
6432 - stack overflow bug in extN fixed
6434 2002-09-18 Andreas Dilger <adilger@clusterfs.com>
6436 * documentation updates
6437 - add man pages for config tools
6438 - update tests/README to describe testing with new config tools
6439 - finish metadata API descriptions
6440 * bug fixes and cleanups
6441 - statfs workaround for 16TB limit
6442 - LOV stripe allocation improved, can stripe on subset of OSTs
6443 - LOV file size/IO offset was wrong for files > 4GB in size
6444 - object EA data was being dropped, caused files to be unreadable
6445 - memory overflow with non-LOV OST caused memory corruption
6446 - fixed regression tests to work with new config tools, obdfilter
6447 - fixed bug when directory size became larger than 1 block
6448 - fixed bug (for single client case) when PWD was deleted
6449 - invalidate local directory pages when doing intent-based ops
6450 - avoid LDLM oops when lock callback contained bad data
6452 2002-09-09 Andreas Dilger <adilger@clusterfs.com>
6454 * documentation updates
6455 * bug fixes and cleanups
6456 - configuration tools
6459 - 64-bit compile warnings
6460 - 64-bit internal statfs data
6462 * test_brw on persistent OST devices
6464 * lprocfs (disabled)
6466 2002-09-04 Andreas Dilger <adilger@clusterfs.com>
6468 * documentation updates
6469 * bug fixes and cleanups
6470 * configuration tools
6472 2002-08-30 Peter J. Braam <braam@clusterfs.com>
6475 * many small fixes to 0_5_4
6476 * io/network handling
6477 * thinkos in MDS operations
6479 2002-08-24 Peter J. Braam <braam@clusterfs.com>
6482 * crucial basic fixes to 0.5.3
6483 * IOR, Iozone work over Elan
6486 2002-08-07 Phil Schwan <phil@clusterfs.com>
6487 * version 0_5_3, our first alpha
6488 * we use the new Portals iovs
6489 * documentation updates
6490 * bug fixes and cleanups
6491 * small changes in the DLM wire protocol
6493 2002-07-25 Peter J. Braam <braam@clusterfs.com>
6494 * version 0_5_1 with some initial stability,
6495 * locking on MD and file I/O.
6496 * documentation updates
6497 * several bug fixes since 0.5.0
6498 * small changes in wire protocol
6500 2002-07-18 Phil Schwan <phil@clusterfs.com>
6502 * delivered as Lustre Light Alpha
6503 * fixed a crash after handling invalid MDS requests
6504 * fixed directory pages for architectures with non-4k pages sizes
6506 2002-07-11 Andreas Dilger <adilger@clusterfs.com>
6507 * release version v0_4_4
6508 * Moves TCP acceptor to be on port 2432 (unused Coda port) instead
6510 * Fixes a number of interruption problems with OST operations.
6511 * Update documentation for portals header changes
6512 * Move all wire protocol structs/defines to lustre_idl.h
6513 * Fixes symlink length bug.
6514 * Add tcpdump to repository.
6516 2002-07-05 Andreas Dilger <adilger@clusterfs.com>
6517 * release version v0_4_3
6518 * Fixes statfs for inodes on extN.
6519 * Fixes bug in runtests which would delete /etc/hosts.
6520 * Use 64-bit object IDs wherever possible (not into VFS though)
6521 Remove ost_get_info, which is unused by lustre, and out of date.
6523 2002-07-03 Peter Braam <braam@clusterfs.com>
6524 * release version v0_4_2 Fixes a lookup error (type not passed)
6525 * move forward to head of Portals
6526 * move forward to latest Lustre kernel
6528 2002-06-25 Peter Braam <braam@clusterfs.com>
6529 * release version v0_4_1. Hopefully stable on single node use.