1 tbd Sun Microsystems, Inc.
4 2.6.16.60-0.37 (SLES 10),
5 2.6.18-128.1.6.el5 (RHEL 5),
6 2.6.22.14 vanilla (kernel.org)
7 * Client support for unpatched kernels:
8 (see http://wiki.lustre.org/index.php?title=Patchless_Client)
9 2.6.16 - 2.6.27 vanilla (kernel.org)
10 * Recommended e2fsprogs version: 1.41.5.sun2
11 * File join has been disabled in this release, refer to Bugzilla 16929.
12 * A new Lustre ADIO driver is available for MPICH2-1.0.7.
13 * NFS export disabled when stack size < 8192. Since the NFSv4 export
14 of Lustre filesystem with 4K stack may cause a stack overflow. For
15 more information, please refer to bugzilla 17630.
19 Description: abort bulk too early if client is reconnected
22 Frequency : rare, if used wide striped file and one ost in down.
24 Descriptoin: don't return error if we created a subset of objects for file.
25 Details : lov_update_create_set() uses set->set_success as index for created
26 objects, so if some requests failed, they will have hole at end of
27 array and we can use qos_shrink_lsm for allocate correct lsm.
29 Severity : enhancement
31 Description: implement async create (obd_async_create) method for osc, to avoid
32 too long waiting new ost objects with holding ldlm lock.
37 Description: fix lqs' reference which won't be put in some situations
38 Details : This patch fixes:
39 1. In quota_check_common(), this function will check quota
40 for user and group, but only send one return via "pending".
41 In most cases, the pendings should be same. But that is not
43 2. If quotaoff runs between lquota_chkquota() and
44 lquota_pending_commit(), the same thing will happen too.
46 - if (!ll_sb_any_quota_active(qctxt->lqc_sb))
51 Description: Rare Client crash on resend if the file was deleted.
52 Details : When file is opened, but open reply is lost and file is
53 subsequently deleted before resend, resend processing logic
54 breaks trying to open the file again, should not try to open.
56 ------------------------------------------------------------------------------
57 tbd Sun Microsystems, Inc.
59 * Support for kernels:
60 2.6.16.60-0.37 (SLES 10),
61 2.6.18-128.1.6.el5 (RHEL 5),
62 2.6.22.14 vanilla (kernel.org)
63 * Client support for unpatched kernels:
64 (see http://wiki.lustre.org/index.php?title=Patchless_Client)
65 2.6.16 - 2.6.27 vanilla (kernel.org)
66 * Recommended e2fsprogs version: 1.41.5-sun2
67 * File join has been disabled in this release, refer to Bugzilla 16929.
68 * A new Lustre ADIO driver is available for MPICH2-1.0.7.
69 * NFS export disabled when stack size < 8192. Since the NFSv4 export
70 of Lustre filesystem with 4K stack may cause a stack overflow. For
71 more information, please refer to bugzilla 17630.
75 Description: resolve race between obd_disconnect and class_disconnect_exports
76 Details : if obd_disconnect will be called to already disconnected export he
77 forget release one reference and osc module can't unloaded.
79 Severity : enhancement
81 Description: move AT tunable parameters for more consistent usage
82 Details : add AT tunables under /proc/sys/lustre, add to conf_param parsing
84 Severity : enhancement
86 Description: Update kernel to RHEL5.3 2.6.18-128.1.6.el5.
88 Severity : enhancement
90 Description: Update kernel to SLES10 SP2 2.6.16.60-0.37.
92 Severity : enhancement
94 Description: Caching OSS
95 Details : introduce data caching on the OSS. The OSS now relies on the linux
96 kernel page cache to keep recently accessed data in memory.
97 It is worth noting that all write requests are still flushed
98 synchronously as in lustre 1.6.
100 Severity : enhancement
102 Description: version based recovery
103 Details : introduce finer grained recovery able to detect transaction
104 dependencies and can deal with transaction gaps.
106 Severity : enhancement
108 Description: Enable adaptive timeouts by default
112 Descriptoin: correctly skip time estimate if in recovery
113 Details : rq_send_state insn't bitmask so using bitwise ops is forbid.
117 Descriptoin: fix goal inodes
118 Details : Allocate inodes for llog in last inode group for avoid broken
123 Descriptoin: OSS DeadLock
124 Details : Use trylock to prevent deadlock when shrink icache.
126 Severity : enhancement
128 Description: Allow tuning service thread via /proc
129 Details : For each service a new
130 /proc/fs/lustre/{service}/*/thread_{min,max,started} entry is
131 created that can be used to set min/max thread counts, and get the
132 current number of running threads.
134 Severity : enhancement
136 Description: Add state history info file, enhance import info file
137 Details : Track import connection state changes in a new osc/mdc proc file;
138 add overview-type data to the osc/mdc import proc file.
142 Description: Reduce small size read RPC
143 Details : Set read-ahead limite for every file and only do read-ahead when
144 available read-ahead pages are bigger than 1M to avoid small size
149 Description: free_entry erroneously used groups_free instead of put_group_info
151 Severity : enhancement
153 Description: Make read-ahead stripe size aligned.
155 Severity : enhancement
157 Description: MDS create should not wait for statfs RPC while holding DLM lock.
162 Descriptoin: OST grant too much space to client even there are not enough space.
163 Details : Client will shrink its grant cache to OST if there are no write
164 activity over 6 mins (GRANT_SHRINK_INTERVAL), and OST will retrieve
165 this grant cache if there are already not enough avaible space
166 (left_space < total_clients * 32M).
168 Severity : enhancement
170 Description: Update client support to vanilla kernels up to 2.6.27.
173 Frequency : rare, connect and disconnect target at same time
175 Descriptoin: ASSERTION(atomic_read(&imp->imp_inflight) == 0
176 Details : Don't call obd_disconnect under lov_lock. this long time
177 operation and can block ptlrpcd which answer to connect request.
179 Severity : enhancement
181 Description: Update to sles9 kernel-2.6.5-7.315.
184 Frequency : start MDS on uncleanly shutdowned MDS device
186 Descriptoin: ll_sync thread stay in waiting mds<>ost recovery finished
187 Details : Stay in waiting mds<>ost recovery finished produce random bugs
188 due race between two ll_sync thread for one lov target. send
189 ACTIVATE event only if connect realy finished and import have
193 Frequency : start MDS on uncleanly shutdowned MDS device
195 Descriptoin: aborting recovery hang on MDS
196 Details : Don't throttle destroy RPCs for the MDT.
198 Severity : enhancement
200 Descriptoin: Don't sync journal after every i/o
201 Details : Implement write RPC replay to allow server replies for write RPCs
202 before data is on disk. However, this feature is disabled by
203 default since some issues leading to data corruptions have been
204 found during recovery (e.g. bug 19128). This feature can be
205 enabled by running the following command on the OSSs:
206 lctl set_param obdfilter.*.sync_journal=0
210 Description: Slow reads beyond 8Tb offsets.
211 Details : Page index integer overflow in ll_read_ahead_page
215 Description: Soft lockup on OSS after MDS failover
216 Details : MSG_CONNECT_INITIAL is not set on the initial MDS->OST connect.
217 As a conseqence, the patch from bug 18224 is not operational
218 and the MDS export cannot be reused on the OSTs until it gets
222 Frequency : rare, only if using MMP with Linux RAID
224 Description: MMP doesn't work with Linux RAID
225 Details : While using HA for Lustre servers with Linux RAID, it is possible
226 that MMP will not detect multiple mounts. To make this work we
227 need to unplug the device queue in RAID when the MMP block is being
228 written. Also while reading the MMP block, we should read it from
229 disk and not the cached one.
232 Frequency : rare, during recovery
234 Description: Assertion failure in ldlm_lock_put
235 Details : Do not put cancelled locks into replay list, hold references on
240 Description: 1.6.5 mdsrate performance is slower than 1.4.11 (MDS not cpu bound)
241 Details : create_count always drops to the min value (=32) because grow_count
242 is being changed before the precreate RPC completes.
245 Frequency : Only in RHEL5 when mounting multiple ext3 filesystems
248 Description: "kmem_cache_create: duplicate cache jbd_4k" error message
249 Details : add proper locking for creation of jbd_4k slab cache
253 Description: MMP check in ext3_remount() fails without displaying any error
254 Details : When multiple mount protection fails during remount, proper error
257 -------------------------------------------------------------------------------
258 2008-12-31 Sun Microsystems, Inc.
260 * Support for kernels:
261 2.6.5-7.314 (SLES 9),
262 2.6.9-67.0.22.EL (RHEL 4),
263 2.6.16.60-0.31 (SLES 10),
264 2.6.18-92.1.17.el5 (RHEL 5),
265 2.6.22.14 vanilla (kernel.org)
266 * Client support for unpatched kernels:
267 (see http://wiki.lustre.org/index.php?title=Patchless_Client)
268 2.6.16 - 2.6.22 vanilla (kernel.org)
269 * Client support for unpatched kernels:
270 we do not recommend using patchless RHEL4 clients with kernels
271 prior to 2.6.9-55EL (RHEL4U5).
272 * Recommended e2fsprogs version: 1.40.11-sun1
273 * Note that reiserfs quotas are disabled on SLES 10 in this kernel.
274 * RHEL 4 and RHEL 5/SLES 10 clients behaves differently on 'cd' to a
275 removed cwd "./" (refer to Bugzilla 14399).
276 * A new quota file format has been introduced in 1.6.5.
277 The format conversion from prior releases is handled transparently,
278 but releases older than 1.4.12/1.6.5 don't understand this new
279 format. The automatic format conversion can be avoided by running
280 the following command on the MDS:
281 'tunefs.lustre --param="mdt.quota_type=ug1" $MDTDEV'.
282 For more information, please refer to bugzilla 13904.
283 * A new quota file format was introduced in 1.6.6/1.8.0 (kernel 2.6.16+)
284 The format conversion from prior releases is handled transparently,
285 but releases older than 1.6.6/1.8.0 don't understand this new
286 format. The automatic format conversion can be avoided by running
287 the following commands on the MDS and OSS servers (for
288 pre 1.4.12-1.6.5 quota files):
289 'tunefs.lustre --param="mdt.quota_type=ug1" $MDTDEV',
290 'tunefs.lustre --param="ost.quota_type=ug1" $MDTDEV'
291 or (for 1.4.12/1.6.5 quota files)
292 'tunefs.lustre --param="mdt.quota_type=ug2" $MDTDEV',
293 'tunefs.lustre --param="ost.quota_type=ug2" $MDTDEV'
294 For more information, please refer to bugzilla 13904.
295 * Output of lfs quota has been made less detailed by default,
296 old (verbose) output can be obtained by using -v option.
297 * File join has been disabled in this release, refer to Bugzilla 16929.
298 * A new Lustre ADIO driver is available for MPICH2-1.0.7.
299 * NFS export disabled when stack size < 8192. Since the NFSv4 export of
300 Lustre filesystem with 4K stack may cause a stack overflow. For more
301 information, please refer to bugzilla 17630.
303 Severity : enhancement
305 Description: add lazystatfs mount option to allow statfs(2) to skip down OSTs
306 Details : Allow skip disconnected ost for send statfs request and hide error
310 Frequency : rare, on llog test 6
312 Descriptoin: don't allow connect to already connected import
313 Details : Connect to already connected import hides connection problem.
316 Frequency : rare, on failed llog setup
318 Descriptoin: don't leak obd reference on failed llog setup
319 Details : For failed llog setup - mgc forget call class_destroy_import for
320 client import, move destroy import to more generic place.
325 Descriptoin: allow kill process which wait statahead result
326 Details : For some reasons 'ls' can stick in waiting result from statahead,
327 in this case need way for kill this process.
330 Frequency : rare, at shutdown
332 Descriptoin: panic at umount
333 Details : llap_shrinker can be raced with killing super block from list and
334 this produce panic with access to already freeded pointer
339 Descriptoin: don't lose wakeup for imp_recovery_waitq
340 Details : Recover_import_no_retry or invalidate_import and import_close can
341 both sleep on imp_recovery_waitq, but we was send only one wakeup
347 Descriptoin: panic in mds_open
348 Details : don't confuse mds_finish_transno() with PTR_ERR(-ENOENT)
353 Descriptoin: stuck in cache_remove_extent() or panic with accessing to already
355 Details : Release lock refernce only after add page to pages list.
358 Frequency : always with long access acl
360 Descriptoin: mds can't pack reply with long acl.
361 Details : mds don't control size of acl but they limited by reint/getattr
366 Frequency : on remount
368 Description: external journal device not working after the remount
369 Details : clear dev_rdonly flag for external journal devices in
375 Description: shutdown vs evict race
376 Details : client_disconnect_export vs connect request race.
377 if client will evicted at this time - we start invalidate
378 thread without referece to import and import can be freed
384 Description: shrink LOV EAs before replying
385 Details : correctly adjust LOV EA buffer for reply.
390 Description: don't skip ost target if they assigned to file
391 Details : Drop slow OSCs if we can, but not for requested start idx.
392 This means "if OSC is slow and it is not the requested
393 start OST, then it can be skipped, otherwise skip it only
394 if it is inactive/recovering/out-of-space.
396 Severity : enhancement
398 Description: Update to RHEL5 kernel-2.6.18-92.1.17.el5.
401 Frequency : rare, need acl's on inode.
403 Description: client can't handle ost additional correctly
404 Details : if ost was added after client connected to mds client can have
405 hit lnet_try_match_md ... to big messages to wide striped files.
406 in this case need teach client to handle config events about add
407 lov target and update client max ea size at that event.
409 Severity : enhancement
411 Description: Update to sles9 kernel-2.6.5-7.314.
413 Severity : enhancement
415 Description: Update to SLES10 SP2 kernel-2.6.16.60-0.31.
418 Frequency : Create a symlink file with a very long name
420 Description: ldlm_cancel_pack()) ASSERTION(max >= dlm->lock_count + count)
421 Details : If there is no extra space in the request for early cancels,
422 ldlm_req_handles_avail() returns 0 instead of a negative value.
427 Description: mds is deadlocked
428 Details : In rare cases, inode in catalog can have i_no less than have parent
429 i_no, this produce wrong order for locking during open, and
430 parallel unlink can be lock open. this need teach mds_open to grab
431 locks in resource id order, not at parent -> child order.
433 Severity : enhancement
435 Description: Add /proc entry for import status
436 Details : The mdc, osc, and mgc import directories now have
437 an import directory that contains useful import data for debugging
440 Severity : enhancement
442 Description: Re-disable certain /proc logging
443 Details : Enable and disable client's offset_stats, extents_stats and
444 extents_stats_per_process stats logging on the fly.
447 Frequency : Only on FC kernels 2.6.22+
449 Description: oops in statahead
450 Details : Do not drop reference count for the dentry from VFS when lookup,
451 VFS will do that by itself.
453 Severity : enhancement
455 Description: Generic /proc file permissions
456 Details : Set /Proc file permissions in a more generic way to enable non-
457 root users operate on some /proc files.
461 Description: Hitting mdc_commit_close() ASSERTION
462 Details : Properly handle request reference release in
463 ll_release_openhandle().
467 Frequency : only patchless client
468 Description: add workaround for race between add/remove dentry from hash
470 Severity : enhancement
472 Description: Allow OST glimpses to return PW locks
476 Description: LBUG when llog conf file is full
477 Details : When llog bitmap is full, ENOSPC should be returned for plain
482 Description: Prevent import from entering FULL state when server in recovery
486 Description: service mount cannot take device name with ":"
487 Details : Only when device name contains ":/" will mount treat it as
493 Description: Replace ptlrpcd with the statahead thread to interpret the async
494 statahead RPC callback
498 Frequency : on recovery
499 Description: I/O failures after umount during fail back
500 Details : If client reconnected to restarted server we need join to recovery
501 instead of find server handler is changed and process self eviction
502 with cancel all locks.
504 Severity : enhancement
506 Description: Update to RHEL5 kernel-2.6.18-92.1.10.el5.
510 Description: Kernel BUG tries to release flock
511 Details : Lustre does not destroy flock lock before last reference goes
512 away. So always drop flock locks when client is evicted and
513 perform unlock regardless of successfulness of speaking to MDS.
515 Severity : enhancement
517 Description: Update to SLES10 SP2 kernel-2.6.16.60-0.27.
519 Severity : enhancement
521 Description: Upcall on Lustre log has been dumped
522 Details : Allow for a user mode script to be called once a Lustre log has
523 been dumped. It passes the filename of the dumped log to the
524 script, the location of the script can be specified via
525 /proc/sys/lnet/debug_log_upcall.
530 Description: avoid messages about idr_remove called for id that is not allocated
531 Details : Move assigment s_dev for clustered nfs to end of initialization,
532 for avoid problem with error handling.
537 Description: avoid Already found the key in hash [CONN_UNUSED_HASH] messages
538 Details : When connection is reused this not moved from CONN_UNUSED_HASH into
539 CONN_USED_HASH and this prodice warning when put connection again
545 Description: avoid ASSERTION(client_stat->nid_exp_ref_count == 0) failed
546 Details : Release reference to stats when client disconnected, not
547 when export destroyed for avoid races when client destroyed
548 after main ost export.
552 Description: more cleanup in mds_lov
553 Details : Add workaround for get valid ost count for avoid warnings about
554 drop too big messages, not init llog cat under semphore which
555 can be blocked on reconnect and break normal replay, fix access
558 Severity : enhancement
560 Description: File striping can now be set to use an arbitrary pool of OSTs.
562 Severity : enhancement
564 Description: Export bytes_read/bytes_write count on OSC/OST.
568 Description: Early reply size mismatch, MGC loses connection
569 Details : Apply the MGS_CONNECT_SUPPORTED mask at reconnect time so
570 the connect flags are properly negotiated.
574 Description: Properly propagate oinfo flags from lov to osc for statfs
575 Details : Restore missing copy oi_flags to lov requests.
579 Description: exports in /proc are broken
580 Details : recreate /proc entries for clients when they reconnect.
582 Severity : enhancement
584 Description: Add man pages for llobdstat(8), llstat(8), plot-llstat(8),
585 : l_getgroups(8), lst(8), routerstat(8)
586 Details : Included man pages for llobdstat(8), llstat(8),
587 : plot-llstat(8), l_getgroups(8), lst(8), routerstat(8)
589 Severity : enhancement
591 Description: Implement lustre ll_show_options method.
593 Severity : enhancement
595 Description: Update to SLES9 kernel-2.6.5-7.312.
597 Severity : enhancement
599 Description: Update to RHEL4 kernel-2.6.9-67.0.22.EL.
603 Description: exports in /proc are broken
604 Details : Recreate /proc entries for clients when they reconnect.
608 Description: don't fail open with -ERANGE
609 Details : If client connected until mds will be know about real ost count
610 get LOV EA can be fail because mds not allocate enougth buffer
615 Description: Resolve device initialization race
616 Details : Prevent proc handler from accessing devices added to the
617 obd_devs array but yet be intialized.
619 Severity : enhancement
621 Description: configure's --enable-quota should check the
622 kernel .config for CONFIG_QUOTA
623 Details : Configure is terminated if --enable-quota is passed but
624 no quota support is in kernel
626 Severity : enhancement
628 Description: Update to SLES10 SP2 kernel-2.6.16.60-0.23.
630 Severity : enhancement
632 Description: Update to RHEL5 kernel-2.6.18-92.1.6.el5.
636 Frequency : rare, on PPC clients
637 Description: don't swab ost objects in response about directory, because
639 Details : Bug similar bug 14856, but in different function.
641 Severity : enhancement
643 Description: lfs quota tool enhancement
644 Details : Added units specifiers support for setquota, default to
645 current uid/gid for quota report, short quota stats by
646 default, nonpositional parameters for setquota, added
647 llapi_quotactl manual page.
649 Severity : enhancement
651 Description: *optional* service tags registration
652 Details : If the "service tags" package is installed on a Lustre node
653 When the filesystem is mounted, a local-node service tag will
654 be created. See http://inventory.sun.com/ for more information
655 about the Service Tags asset management system.
659 Description: Client runs out of low memory
660 Details : Consider only lowmem when counting initial number of llap pages
663 Description: Kernel BUG tries to release flock
664 Details : Lustre does not destroy flock lock before last reference goes
665 away. So always drop flock locks when client is evicted and
666 perform unlock regardless of successfulness of speaking to MDS.
669 Frequency : occasional
671 Description: add refcount for osc callbacks, so avoid panic on shutdown
673 Severity : enhancement
675 Description: Update to RHEL4 kernel-2.6.9-67.0.20.
678 Frequency : testing only
680 Description: sanity test 65a fails if stripecount of -1 is set
681 Details : Handle -1 striping on filesystem in ll_dirstripe_verify
684 Frequency : only in unusual configurations
686 Description: Kernel panic with find ost index.
687 Details : lov_obd have panic if some OST's have sparse indexes.
689 Severity : enhancement
691 Description: Update to RHEL5 kernel-2.6.18-53.1.21.el5.
694 Frequency : rarely, if filesystem is mounted with -o flock
696 Description: do not process already freed flock
697 Details : Flock can possibly be freed by another thread before it reaches
698 to ldlm_flock_completion_ast.
701 Frequency : rarely, if filesystem is mounted with -o flock
703 Description: LBUG during stress test
704 Details : Need properly lock accesses the flock deadlock detection list.
707 Frequency : rarely, if binaries are being run from Lustre
709 Description: oops in page fault handler
710 Details : Kernel page fault handler can return two special 'pages' in
711 error case, don't try dereference NOPAGE_SIGBUS and NOPAGE_OMM.
714 Frequency : rarely, during shutdown
716 Description: timeout with invalidate import.
717 Details : ptlrpcd_check call obd_zombie_impexp_cull and wait request which
718 should be handled by ptlrpcd. This produce long age waiting and
719 -ETIMEOUT ptlrpc_invalidate_import and as result LASSERT.
725 Description: ASSERTION(CheckWriteback(page,cmd)) failed
726 Details : Badly clear PG_Writeback bit in ll_ap_completion can produce false
730 Frequency : only with broken builds/installations
732 Description: no LBUG if lquota.ko and fsfilt_ldiskfs.ko are different versions
733 Details : Just return an error to a user, put a console error message.
735 Severity : enhancement
737 Description: Update to RHEL5 kernel-2.6.18-53.1.19.el5.
739 Severity : enhancement
741 Description: Update to RHEL4 kernel-2.6.9-67.0.15.
743 Severity : enhancement
745 Description: enable MGS and MDT services start separately
746 Details : add a 'nomgs' option in mount.lustre to enable start a MDT with
747 a co-located MGS without starting the MGS, which is a complement
748 to 'nosvc' mount option.
751 Frequency : always, on big-endian systems
753 Description: cleanup in ptlrpc code, related to PPC platform
754 Details : store magic in native order avoid panic's in recovery on PPC
755 node and forbid from this error in future. Also fix posibily
756 of twice swab data. Fix get lov striping to userpace.
759 Frequency : rarely, if replay get lost on server
761 Description: server incorrectly drop resent replays lead to recovery failure.
762 Details : do not drop replay according to msg flags, instead we check the
763 per-export recovery request queue for duplication of transno.
766 Frequency : after recovery
768 Description: precreate to many object's after del orphan.
769 Details : del orphan st in oscc last_id == next_id and this triger growing
770 count of precreated objects. Set flag LOW to skip increase count
771 of precreated objects.
774 Frequency : rare, on clear nid stats
776 Description: ASSERTION(client_stat->nid_exp_ref_count == 0)
777 Details : when clean nid stats sometimes try destroy live entry,
778 and this produce panic in free.
781 Frequency : occasionally since 1.6.4
783 Description: Stack overflow during MDS log replay
784 Details : ease stack pressure by using a thread dealing llog_process.
787 Frequency : very rare
789 Description: MDT cannot be unmounted, reporting "Mount still busy"
790 Details : Mountpoint references were being leaked during open reply
791 reconstruction after an MDS restart. Drop mountpoint reference
792 in reconstruct_open() and free dentry reference also.
797 Description: wait until IO finished before start new when do lock cancel.
798 Details : VM protocol want old IO finished before start new, in this case
799 need wait until PG_writeback is cleared until check dirty flag
800 and call writepages in lock cancel callback.
805 Description: mds_mfd_close() ASSERTION(rc == 0)
806 Details : In mds_mfd_close(), we need protect inode's writecount change
807 within its orphan write semaphore to prevent possible races.
810 Frequency : rare, on shutdown ost
812 Description: don't hit live lock with umount ost.
813 Details : shrink_dcache_parent can be in long loop with destroy dentries,
814 use shrink_dcache_sb instead.
817 Frequency : only when echo_client is used
819 Description: don't panic with use echo_client
820 Details : echo client pass NULL as client nid pointer and this produce NULL
824 Frequency : Always on 32-bit PowerPC systems
826 Description: fix build on PPC32
827 Details : compile code with -m64 flag produce wrong object file for PPC32.
832 Description: MDS LBUG: ASSERTION(!IS_ERR(dchild))
833 Details : In reconstruct_* functions, LASSERTs on both the data supplied
834 by a client, and the data on disk are dangerous and incorrect.
835 Change them with client eviction.
837 Severity : enhancement
839 Description: skiplist implementation simplification
840 Details : skiplists are used to group compatible locks on granted list
841 that was implemented as tracking first and last lock of each lock
842 group the patch changes that to using doubly linked lists
846 Description: delete compatibility for 32bit qdata
847 Details : as planned, when lustre is beyond b1_8, lquota won't support 32bit
848 qunit. That means servers of b1_4 and servers of b1_8 can't be
849 used together if users want to use quota.
852 Frequency : only with administrator action
854 Description: mount failure if config log has invalid conf_param setting
855 Details : If administrator specified an incorrect configuration parameter
856 with "lctl conf_param" this would cause an error during future
857 client mounts. Instead, ignore the bad configuration parameter.
860 Frequency : blocks per group < blocksize*8 and uninit_groups is enabled
862 Description: ldiskfs error: XXX blocks in bitmap, YYY in gd
863 Details : If blocks per group is less than blocksize*8, set rest of the
867 Frequency : Application do stride read on lustre
869 Description: The read performance will drop a lot if the application does
871 Details : Because the stride_start_offset are missing in stride read-ahead,
872 it will cause clients read a lot of unused pages in read-ahead,
873 then the read-performance drops.
877 Description: more ldlm soft lockups
878 Details : In ldlm_resource_add_lock(), call to ldlm_resource_dump()
879 starve other threads from the resource lock for a long time in
880 case of long waiting queue, so change the debug level from
881 D_OTHER to the less frequently used D_INFO.
883 Severity : enhancement
885 Description: add -gid, -group, -uid, -user options to lfs find
887 Severity : enhancement
889 Description: ll_recover_lost_found_objs - recover objects in lost+found
890 Details : OST corruption and subsequent e2fsck can leave objects in the
891 lost+found directory. Using the "ll_recover_lost_found_objs"
892 tool, these objects can be retrieved and data can be salvaged
893 by using the object ID saved in the fid EA on each object.
898 Description: this bug _only_ happens when inode quota limitation is very low
899 (less than 12), so that inode quota unit is 1 at initialization.
900 Details : if remaining quota equates 1, it is a sign to demonstate that quota
901 is effective now. So least quota qunit should be 2.
905 Description: Hung threads in invalidate_inode_pages2_range
906 Details : The direct IO path doesn't call check_rpcs to submit a new RPC once
907 one is completed. As a result, some RPCs are stuck in the queue
912 Description: Procfs and llog threads access destoryed import sometimes.
913 Details : Sync the import destoryed process with procfs and llog threads by
914 the import refcount and semaphore.
918 Description: mds fails to respond, threads stuck in ldlm_completion_ast
919 Details : Sort source/child resource pair after updating child resource.
924 Description: kernel BUG at ldiskfs2_ext_new_extent_cb
925 Details : If insertion of an extent fails, then discard the inode
926 preallocation and free data blocks else it can lead to duplicate
931 Description: don't always update ctime in ext3_xattr_set_handle()
932 Details : Current xattr code updates inode ctime in ext3_xattr_set_handle()
933 In some cases the ctime should not be updated, for example for
934 2.0->1.8 compatibility it is necessary to delete an xattr and it
935 should not update the ctime.
939 Description: add quota statistics
940 Details : 1. sort out quota proc entries and proc code.
941 2. add quota statistics
946 Description: quotas are not honored with O_DIRECT
947 Details : all writes with the flag O_DIRECT will use grants which leads to
948 this problem. Now using OBD_BRW_SYNC to guard this.
952 Bugzilla : 15713/16362
953 Description: Assertion in iopen_connect_dentry in 1.6.3
954 Details : looking up an inode via iopen with the wrong generation number can
955 populate the dcache with a disconneced dentry while the inode
956 number is in the process of being reallocated. This causes an
957 assertion failure in iopen since the inode's dentry list contains
958 both a connected and disconnected dentry.
962 Description: assertion failure in ldlm_handle2lock()
963 Details : fix a race between class_handle_unhash() and class_handle2object()
964 introduced in lustre 1.6.5 by bug 13622.
966 Severity : enhancement
968 Description: superblock lock contention with many SMP cores on one client
969 Details : several client filesystem locks were highly contended on SMP
970 NUMA systems with 8 or more cores. Per-CPU datastructures
971 and more efficient locking implemented to reduce contention.
976 Description: Kernel BUG: sd_iostats_bump: unexpected disk index
977 Details : remove the limit of 256 scsi disks in the sd_iostat patch
982 Description: oops in sd_iostats_seq_show()
983 Details : unloading/reloading the scsi low level driver triggers a kernel
984 bug when trying to access the sd iostat file.
989 Description: Kernel panics during QLogic driver reload
990 Details : REQ_BLOCK_PC requests are not handled properly in the sd iostat
991 patch, causing memory corruption.
996 Description: journal_dev option does not work in b1_6
997 Details : pass mount option during pre-mount.
999 Severity : enhancement
1001 Description: Add a FIEMAP(FIle Extent MAP) ioctl for ldiskfs
1002 Details : FIEMAP ioctl will allow an application to efficiently fetch the
1003 extent information of a file. It can be used to map logical blocks
1004 in a file to physical blocks in the block device.
1007 Frequency : only with adaptive timeout enabled
1009 Description: DEBUG_REQ() bad paging request
1010 Details : ptlrpc_at_recv_early_reply() should not modify req->rq_repmsg
1011 because it can be accessed by reply_in_callback() without the
1015 Frequency : only on Cray X2
1017 Description: X2 build failures
1018 Details : fix build failures on Cray X2.
1022 Description: xid & resent requests
1023 Details : Initialize RPC XID from clock at startup (randomly if clock is
1028 Description: quota recovery deadlock during mds failover
1029 Details : This patch includes att18982, att18236, att18237 in bz14840.
1031 1. fix osts hang when mds does failover with quotaon
1032 2. prevent watchdog storm when osts threads wait for the
1037 Description: kernel panic on racer
1038 Details : Do not access dchild->d_inode when IS_ERR(dchild) is true.
1040 Severity : enhancement
1042 Description: Add lustre_start utility to start or stop multiple Lustre servers
1047 Description: Lustre GPF in {:ptlrpc:ptlrpc_server_free_request+373}
1048 Details : In case of memory pressure, list_del() can be called twice on
1049 req->rq_history_list, causing a kernel oops.
1053 Description: (ptllnd_peer.c:557:kptllnd_peer_check_sends()) ASSERTION(!in_interrupt()) failed
1054 Details : fix stack overflow in the distributed lock manager by defering
1055 export eviction after a failed ast to the elt thread instead of
1056 handling it in the dlm interpret routine.
1058 Severity : enhancement
1060 Description: More exported tunables for mballoc
1061 Details : Add support for tunable preallocation window and new tunables for
1062 large/small requests
1066 Description: Detect corruption of block bitmap and checking for preallocations
1067 Details : Checks validity of on-disk block bitmap. Also it does better
1068 checking of number of applied preallocations. When corruption is
1069 found, it turns filesystem readonly to prevent further corruptions.
1073 Frequency : only for big-endian servers
1074 Description: Check for big-endian system while mounting fs with extents feature
1075 Details : Mounting a filesystem with extents feature will fail on big-endian
1076 systems since ext3-based ldiskfs is not supported on big-endian
1077 systems. Can be over-ridden with "bigendian_extents" mount option.
1081 Description: Excessive recovery window
1082 Details : With AT enabled, the recovery window can be excessively long (6000+
1083 seconds). To address this problem, we no longer use
1084 OBD_RECOVERY_FACTOR when extending the recovery window (the connect
1085 timeout no longer depends on the service time, it is set to
1086 INITIAL_CONNECT_TIMEOUT now) and clients report the old service
1087 time via pb_service_time.
1091 Description: Watchdog triggered on MDS failover
1092 Details : enable OBD_CONNECT_MDT flag when connecting from the MDS so that
1093 the OSTs know that the MDS "UUID" can be reused for the same export
1094 from a different NID, so we do not need to wait for the export to
1099 Descriptoin: Lustre detected file system corruption with inode out of bounds
1100 Details : don't update i_size on MDS_CLOSE for directories. This causes
1101 directory corruptions on the MDT.
1103 -------------------------------------------------------------------------------
1105 2008-05-26 Sun Microsystems, Inc.
1107 * Support for kernels:
1108 2.6.5-7.311 (SLES 9),
1109 2.6.9-67.0.7.EL (RHEL 4),
1110 2.6.16.54-0.2.5 (SLES 10),
1111 2.6.18-53.1.14.el5 (RHEL 5),
1112 2.6.22.14 vanilla (kernel.org)
1113 * Client support for unpatched kernels:
1114 (see http://wiki.lustre.org/index.php?title=Patchless_Client)
1115 2.6.16 - 2.6.22 vanilla (kernel.org)
1116 * Due to problems with nested symlinks and FMODE_EXEC (bug 12652),
1117 we do not recommend using patchless RHEL4 clients with kernels
1118 prior to 2.6.9-55EL (RHEL4U5).
1119 * Recommended e2fsprogs version: 1.40.7-sun1
1120 * Note that reiserfs quotas are disabled on SLES 10 in this kernel.
1121 * RHEL 4 and RHEL 5/SLES 10 clients behaves differently on 'cd' to a
1122 removed cwd "./" (refer to Bugzilla 14399).
1123 * A new quota file format has been introduced in 1.6.5.
1124 The format conversion from prior releases is handled transparently,
1125 but releases older than 1.4.12/1.6.5 will not understand this new
1126 format. The automatic format conversion can be avoided by running
1127 the following command on the MDS before upgrading:
1128 'tunefs.lustre --param="mdt.quota_type=ug1" $MDTDEV'.
1129 For more information, please refer to bugzilla 13904.
1133 Description: quota performance fix
1134 Details : quota data is written in journalled mode instead of ordered to
1135 increase performance
1139 Description: lfs support for human-readable quota grace time strings
1140 Details : lfs setquota -t and lfs quota -t represent quota grace times
1141 in "XXwXXdXXhXXmXXs" format instead of large values in seconds
1144 Frequency : always with o2ib 1.3 and sles10
1146 Description: fix build with SLES10 and o2ib v3.
1147 Details : sles10 uses diffrent name for Module.symver file but configure
1148 assume this file has same name on RHEL/SLES/vanila kernels.
1151 Frequency : very rare, if additional xattrs are used on kernels >= 2.6.12
1153 Description: MDS may lose file striping (and hence file data) in some cases
1154 Details : If there are additional extended attributes stored on the MDS,
1155 in particular ACLs, SELinux, or user attributes (if user_xattr
1156 is specified for the client mount options) then there is a risk
1157 of attribute loss. Additionally, the Lustre file striping
1158 needs to be larger than default (e.g. striped over all OSTs),
1159 and an additional attribute must be stored initially in the
1160 inode and then increase in size enough to be moved to the
1161 external attribute block (e.g. ACL growing in size) for file
1164 Severity : enhancement
1166 Description: add message levels for liblustreapi
1171 Description: MDT cannot be unmounted, reporting "Mount still busy"
1172 Details : Mountpoint references were being leaked during open reply
1173 reconstruction after an MDS restart. Drop mountpoint reference
1174 in reconstruct_open() and free dentry reference also.
1179 Description: fix for occasional failure case of -ENOSPC in recovery-small tests
1180 Details : Move the 'good_osts' check before the 'total_bavail' check. This
1181 will result in an -EAGAIN and in the exit call path we call
1182 alloc_rr() which will with increasing aggressiveness attempt to
1183 aquire precreated objects on the minimum number of required OSCs.
1187 Description: Use old size assignment to avoid deadlock
1188 Details : This reverts the changes in bugs 2369 and bug 14138 that introduced
1189 the scheduling while holding a spinlock. We do not need locking
1190 for size in ll_update_inode() because size is only updated from
1191 the MDS for directories or files without objects, so there is no
1192 other place to do the update, and concurrent access to such inodes
1193 are protected by the inode lock.
1197 Description: Use __u64 instead of int for valid bits
1201 Description: resolve "_IOWR redefined" build error on SLES10
1205 Description: dump the memory debugging after all modules are unloaded to
1206 suppress false negative in conf_sanity test 39
1210 Description: the recovery timer never expires
1211 Details : for new client connect request, the recovery timer should not be
1212 reset, otherwise recovery timer will never expired, if the old
1213 client never come. Only old client connect and first connection
1214 req should trigger recovery timer reset.
1218 Description: the min numbers of lproc stats are wrong
1219 Details : adding a new constant LC_MIN_INIT and use it for initialization
1223 Frequency : always with interactive lfs
1225 Description: Reinitialize optind to 0 so that interactive lfs works in all cases
1228 Frequency : with multiple concurrent readdir processes in same directory
1229 Bugzilla : 15406, 15169, 15175
1230 Description: misc fixes for directory readahead.
1231 Details : prevent previous statahead async RPC callback from processing the
1232 current "statahead_info", race condition between async RPC callback
1233 add dentry into dentry hash table and "ls" thread revalidate such
1234 dentry, statahead his/miss control for hidden items, and so on.
1236 Severity : enhancement
1238 Description: build kernel-ib packages for OFED 1.3 in our release cycle
1242 Description: incore types cleaning in quota code (with respect to 64-bit limits)
1243 Details : several u32 variables declarations are replaced with u64 declarations
1248 Description: fix SLES kernel versioning
1249 Details : the kernel version for our SLES 10 kernel did not include a "-"
1250 before the "smp" at the end. while this was not a problem in
1251 general, it did mean that software trying to use the kernel
1252 version to try to detect a vendor specific kernel would fail.
1253 this was most evident by the OFED build scripts.
1258 Description: Don't update lov_desc members until making sure they are valid
1259 Details : When updating lov_desc members via proc fs, need fix their
1260 validities before doing the real update.
1263 Frequency : very rare
1265 Description: don't put request into delay list while invalidate in flight.
1266 Details : ptlrpc_delay_request sometimes put in delay list while invalidate
1267 import in flight. this produce timeout for invalidate and sometimes
1268 can cause stale data.
1270 Severity : enhancement
1272 Description: Update kernel to SLES9 2.6.5-7.311.
1274 Severity : enhancement
1276 Description: Update kernel to RHEL4 2.6.9-67.0.7.
1281 Frequency : on PPC only
1282 Description: not convert ost objects for directory because it's not exist.
1283 Details : ll_dir_getstripe assume dirrectory has ost objects but this wrong.
1285 Severity : enhancement
1287 Description: Fix warnings with compile liblustre at sles10/rhel5 which have
1288 __u64 as usingied long long type.
1291 Frequency : rare, on shutdown
1293 Description: race process ast vs remove callback
1294 Details : removing callback before disconnect import open race with
1295 processing callback.
1297 Severity : enhancement
1299 Description: Update kernel to SLES9 2.6.5-7.311.
1301 Severity : enhancement
1303 Description: Files open for execute are not marked busy on SLES10
1304 Details : Add FMODE_EXEC to SLES10 SP1 server kernel series.
1306 Severity : enhancement
1308 Description: Add server support for vanilla-2.6.22.14.
1311 Frequency : occasional
1313 Description: Avoid lov_create() getting stuck in obd_statfs_rqset()
1314 Details : If an OST is down the MDS will hang indefinitely in
1315 obd_statfs_rqset() waiting for the statfs data. While for
1316 MDS QOS usage of statfs, it should not stuck in waiting.
1318 Severity : enhancement
1320 Description: Disable adaptive timeouts by default
1323 Frequency : on network error
1325 Description: panic with double free request if network error
1326 Details : mdc_finish_enqueue is finish request if any network error occuring,
1327 but it's true only for synchronus enqueue, for async enqueue
1328 (via ptlrpcd) this incorrect and ptlrpcd want finish request
1332 Frequency : rare, on recovery
1334 Description: read procfs can produce deadlock in some situation
1335 Details : Holding lprocfs lock with send rpc can produce block for destroy
1336 obd objects and this also block reconnect with -EALREADY.
1337 This isn't fix all lprocfs bugs - but make it rare.
1339 Severity : enhancement
1341 Description: Update kernel to RHEL5 2.6.18-53.1.14.el5.
1344 Frequency : frequent on X2 node
1346 Description: mdc_set_open_replay_data LBUG
1347 Details : Set replay data for requests that are eligible for replay.
1352 Description: lustre_mgs: operation 101 on unconnected MGS
1353 Details : When MGC is disconnected from MGS long enough, MGS will evict the
1354 MGC, and late on MGC cannot successfully connect to MGS and a lot
1355 of the error messages complaining that MGS is not connected.
1358 Frequency : rare, depends on device drivers and load
1360 Description: MDS or OSS nodes crash due to stack overflow
1361 Details : Code changes in 1.6.4 increased the stack usage of some functions.
1362 In some cases, in conjunction with device drivers that use a lot
1363 of stack, the MDS (or possibly OSS) service threads could overflow
1364 the stack. One change which was identified to consume additional
1365 stack has been reworked to avoid the extra stack usage.
1367 Severity : enhancement
1369 Description: Update to RHEL5 latest kernel-2.6.18-53.1.13.el5.
1371 Severity : enhancement
1373 Description: Update to SLES10 SP1 latest kernel-2.6.16.54-0.2.5.
1375 Severity : enhancement
1377 Description: Update to RHEL5 latest kernel-2.6.18-53.1.6.el5.
1379 Serverity : enhancement
1381 Description: Update RHEL4 kernel to 2.6.9-67.0.4.
1384 Frequency : rare on shutdown OST
1386 Description: Don't allow skipping OSTs if index has been specified.
1387 Details : Don't allow skipping OSTs if index has been specified, make locking
1388 in internal create lots better.
1393 Description: ASSERTION(!PageDirty(page)) failed
1394 Details : Wrong check could lead to an assertion failure under specific
1400 Description: LBUG in ptlrpc_check_set() bad phase ebc0de00
1401 Details : access to bitfield in structure is always rounded to long
1402 and this produce problem with not atomic change any bit.
1407 Description: Lustre make rpms failed.
1408 Details : Remove ldiskfs spec file to avoids rpmbuild be confused when
1409 builds Lustre rpms from tarball.
1411 Severity : enhancement
1413 Description: Update to SLES9 SP4 kernel-2.6.5-7.308.
1416 Frequency : rare on shutdown OST
1418 Description: If llog cancel was not send before clean_exports phase, this can
1419 produce deadlock in llog code.
1420 Details : If llog thread has last reference to obd and call class_import_put
1421 this produce deadlock because llog_cleanup_commit_master wait when
1422 last llog_commit_thread exited, but this never success because was
1423 called from llog_commit_thread.
1426 Frequency : only if OST index is skipped
1428 Description: NULL lov_tgts causing MDS oops
1429 Details : more safe checks for NULL lov_tgts for avoid oops.
1431 Severity : enhancement
1433 Description: Update to RHEL4 latest kernel-2.6.9-67.0.1.EL.
1435 Severity : enhancement
1437 Description: Update to RHEL5 latest kernel-2.6.18-53.1.4.el5.
1442 Description: make mgs_setparam() handle fsname containing dash
1443 Details : fsname containing a dash does not work with lctl conf_param
1445 Severity : enhancement
1447 Description: Update to RHEL4 Update-6 kernel-2.6.9-67.EL.
1450 Frequency : rare, in recovery and (or) destroy lovobjid file.
1452 Description: rewrite lov objid code.
1453 Details : Cleanup for lov objid code, remove scability problems and wrong
1454 locking. Fix sending last_id into ost.
1456 Severity : enhancement
1458 Description: Update to SLES10 SP1 latest kernel-2.6.16.54-0.2.3.
1460 Severity : enhancement
1462 Description: Update to RHEL5 Update-1 kernel 2.6.18-53.el5.
1463 Details : Use d_move_locked instead of __d_move.
1466 Frequency : rare, at shutdown
1468 Description: access already free / zero obd_namespace.
1469 Details : if client_disconnect_export was called without force flag set,
1470 and exist connect request in flight, this can produce access to
1471 NULL pointer (or already free pointer) when connect_interpret
1472 store ocd flags in obd_namespace.
1475 Frequency : only at startup
1477 Description: not alloc memory with spinlock held.
1478 Details : allocation memory with GFP_KERNEL can produce sleep deadlock,
1479 if any spinlock held.
1484 Description: lfs find does not continue on file error
1485 Details : Continue other files processing when a file/dir is absent.
1489 Description: Inconsistent usage of lustre_pack_reply()
1490 Details : Standardize the usage of lustre_pack_reply() such that it
1491 always generate a CERROR on failure.
1494 Frequency : very rare
1496 Description: Fix replay if there is an un-replied request and open
1497 Details : In some cases, older replay request will revert the
1498 mcd->mcd_last_xid on MDS which is used to record the client's
1499 latest sent request.
1501 Severity : enhancement
1503 Description: Update to RHEL5 kernel 2.6.18-8.1.15.el5.
1505 Severity : enhancement
1507 Description: Update to SLES10 SP1 kernel 2.6.16.53-0.16
1509 Severity : enhancement
1511 Description: Update to SLES9 kernel-2.6.5-7.287.3.
1513 Severity : enhancement
1515 Description: Update to RHEL4 kernel-2.6.9-55.0.12.EL.
1517 Severity : enhancement
1519 Description: Build SLES10 patchless client fails
1520 Details : The configure was broken by run ./configure with
1521 --with-linux-obj=.... argument for patchless client. When the
1522 configure use --with-linux-obj, the LINUXINCLUDE= -Iinclude
1523 can't search header adequately. Use absolute path such as
1524 -I($LINUX)/include instead.
1526 Severity : enhancement
1528 Description: Lustre Page Accounting
1529 Details : New macros for page alloc and free which enable accounting
1530 of page allocation of Lustre. Use percpu counters to store memory
1531 and page statistics.
1534 Frequency : only if debugging is disabled
1536 Description: LASSERT_{REQ,REP}SWAB macros are buggy
1537 Details : If SWAB_PARANOIA is disabled, the LASSERT_REQSWAB and
1538 LASSERT_REPSWAB macros become no-ops, which is incorrect. Drop
1539 these macros and replace them with their definitions instead.
1544 Description: interrupt oig_wait produce painc on resend.
1545 Details : brw_redo_request can be used for resend requests from ptlrpcd and
1546 private set, and this produce situation when rq_ptlrpcd_data not
1547 copyed to new allocated request and triggered LBUG on assert
1548 req->rq_ptlrpcd_data != NULL. But this member used only for wakeup
1549 ptlrpcd set if request is changed and can be safety changed to use
1552 Severity : enhancement
1554 Description: organize the server-side client stats on per-nid basis
1555 Details : Change the structure of stats under obdfilter and mds to
1563 The "uuid"s file would list the uuids of _active_ exports.
1564 And the clear entry is to clear all stats and stale nids.
1569 Description: Processes looping in ll_readdir() on Lustre clients finally causing
1570 a full node pseudo-hang
1571 Details : Concurrent access to the same directory from multiple clients with
1572 intensive file creation/removal can cause a client node to spin in
1573 ll_readdir(). i_version must be increased every time the lock is
1574 cancelled to ensure a revalidate is done.
1579 Description: touch file failed when fs is not full
1580 Details : OST in recovery should not be discarded by MDS in alloc_qos(),
1581 otherwise we can get ENOSP while fs is not full.
1584 Frequency : only for Cray XT3
1585 Bugzilla : 12829/13455
1586 Description: Changing primary group doesn't change the group lustre assigns to
1588 Details : When CRAY_XT3 is defined, the fsgid supplied by the client is
1589 overridden with the primary group provided by the group upcall,
1590 whereas the supplied fsgid can be trusted if it is in the list of
1591 supplementary groups returned by the group upcall.
1593 Severity : enhancement
1595 Description: Root Squash Functionality
1596 Details : Implementation of NFS-like root squash capability. Specifically,
1597 don't allow someone with root access on a client node to be able
1598 to manipulate files owned by root on a server node.
1600 Severity : enhancement
1602 Description: Slow trucate/writes to huge files at high offsets.
1603 Details : Directly associate cached pages to lock that protect those pages,
1604 this allows us to quickly find what pages to write and remove
1605 once lock callback is received.
1610 Description: Too many locks accumulating on client during NFS usage
1611 Details : mds_open improperly used accmode to find out access mode to a
1612 file. Also mdc_intent_lock logic to find out if we already have
1613 lock similar to just received was flawed since introduction of
1614 skiplists - locks are now added to the front of the granted
1619 Description: Hit ASSERTION(obd->obd_stopping == 1) failed in some setup failed
1621 Details : In obd setup failure handler, obd_stopping will not necessarily to
1622 be 1, and obd_set_up should also be checked to make sure whether
1623 obd is completely setup.
1625 Severity : enhancement
1627 Description: Allow masking D_WARNING, D_ERROR messages from console
1628 Details : Console messages can now be disabled via lnet.printk.
1633 Description: User code with malformed file open parameter crashes client node
1634 Details : Before packing join_file req, all the related reference should be
1635 checked carefully in case some malformed flags cause fake join_file
1641 Description: shrink/enlarge qunit size when needed; fix the problem of coarse
1642 grain of quota doing harm to quota's accuracy
1643 Details : qunit size will be changed when quota limitation is too low/high;
1644 record the pending quota write in order to get more accureate
1645 quota; delete the patch for bug12588, which is unnecessary when
1646 this patch is landed. This bug also contains fixes for bug 14526,
1647 14299, 14601 and 13794.
1651 Description: LDLM_ENQUEUE races with LDLM_CP_CALLBACK
1652 Details : ldlm_completion_ast() assumes that a lock is granted when the req
1653 mode is equal to the granted mode. However, it should also check
1654 that LDLM_FL_CP_REQD is not set.
1658 Description: Heavy nfs access might result in deadlocks
1659 Details : After ELC code landed, it is now improper to enqueue any mds
1660 locks under och_sem, because enqueue might want to decide to
1661 cancel open locks for same inode we are holding och_sem for.
1665 Description: 35% write performance drop with ldiskfs2 when quotas are on
1666 Details : Enable ext3 journalled quota by default to improve performance
1667 when quotas are turned on.
1671 Description: Client eviction while running blogbench
1672 Details : A lot of unlink operations with concurrent I/O can lead to a
1673 deadlock causing evictions. To address the problem, the number of
1674 oustanding OST_DESTROY requests is now throttled to
1675 max_rpcs_in_flight per OSC and LDLM_FL_DISCARD_DATA blocking
1676 callbacks are processed in priority.
1679 Frequency : RHEL4 only
1681 Description: mkfs is very slow on IA64/RHEL4
1682 Details : A performance regression has been discovered in the MPT Fusion
1683 driver between versions 3.02.73rh and 3.02.99.00rh. As a
1684 consequence, we have downgraded the MPT Fusion driver in the RHEL4
1685 kernel from 3.02.99.00 to 3.02.73 until this problem is fixed.
1688 Frequency : PPC/PPC64 only
1690 Description: conflicts between asm-ppc64/types.h and lustre_types.h
1691 Details : fix duplicated definitions between asm-ppc64/types.h and
1692 lustre_types.h on PPC.
1695 Frequency : PPC/PPC64 only
1697 Description: asm-ppc/segment.h does not exist
1698 Details : fix compile issue on PPC.
1702 Description: data checksumming impacts single node performance
1703 Details : add support for several checksum algorithms. Currently, CRC32 and
1704 Adler-32 are supported. The checksum type can be changed on the fly
1705 through /proc/fs/lustre/osc/*/checksum_type.
1709 Description: use adler32 for page checksums
1710 Details : when available, use the Adler-32 algorithm instead of CRC32 for
1715 Description: better handle error messages in extents code
1717 Severity : enhancement
1719 Description: SNMP support enhancement
1720 Details : Adding total number of sampled request for an MDS node in snmp
1723 Severity : enhancement
1725 Description: Optimize ldlm waiting list processing for PR extent locks
1726 Details : When processing waiting list for read extent lock and meeting read
1727 lock that is same or wider to it that is not contended, skip
1728 processing rest of the list and immediatelly return current
1729 status of conflictness, since we are guaranteed there are no
1730 conflicting locks in the rest of the list.
1734 Description: Time out and refuse to reconnect
1735 Details : When the failover node is the primary node, it is possible
1736 to have two identical connections in imp_conn_list. We must
1737 compare not conn's pointers but NIDs, otherwise we
1738 can defeat connection throttling.
1742 Description: Client not clear own cache if answer to reconnect is lost.
1743 Details : Client gets evicted from server. Now client also thinks it is
1744 disconnected (or gets ENOTCONN on its operation) and decides to
1745 reconnect. Server receives reconnect message, but cannot find
1746 export. New export is created that is fully valid (new cookie!),
1747 but reply is lost and not reported to client. Client reconnects
1748 again and gets back a just-created connection, but it is not new
1749 so client thinks it was not evicted and does not do recovery.
1753 Description: Detect stride IO mode in read-ahead
1754 Details : When a client does stride read, read-ahead should detect that and
1755 read-ahead pages according to the detected stride pattern.
1759 Description: build for x2 fails
1760 Details : fix compile issue on Cray systems.
1762 Severity : enhancement
1764 Description: implement readv/writev APIs(aio_read/aio_writes in newer kernels)
1765 Details : This greatly improves speed of NFS writes on 2.6 kernels.
1768 Frequency : only on PPC/SLES10
1770 Description: "BITS_PER_LONG is not 32 or 64" in linux/idr.h
1771 Details : On SLES10/PPC, fs.h includes idr.h which requires BITS_PER_LONG to
1772 be defined. Add a hack in mkfs_lustre.c to work around this compile
1777 Description: LASSERT on MDS when client holding flock lock dies
1778 Details : ldlm pool logic depends on number of granted locks equal to
1779 number of released locks which is not true for flock locks, so
1780 just exclude such locks from consideration.
1784 Description: MDS deadlock with many ll_sync_lov threads and I/O stalled
1785 Details : Use fsfilt_sync() for both the whole filesystem sync and
1786 individual file sync to eliminate dangerous inode locking
1787 with I_LOCK that can lead to a deadlock.
1791 Description: Update an obsolete wirecheck.c generator
1792 Details : Update wirecheck.c/wirehdr.c and regenerate wiretest.c
1796 Description: Client can panic on open sometimes
1797 Details : It is possible that we try to free already freed request in
1798 ll_file_open in some error cases when we send request from
1803 Description: performance in 1.6.3
1804 Details : Force q->max_phys_segments to MAX_PHYS_SEGMENTS on SLES10 to be
1805 sure that 1MB requests are not fragmented by the block layer.
1809 Description: LDLM soft lockups - improvement
1810 Details : It is be possible to send the lock handle along with each read
1811 or write request because the client is already doing a lock match
1812 itself so there isn't any reason the OST should have to re-do that
1818 Description: lfs quota fails with deactivated OSTS
1819 Details : With this patch, three improvements are included:
1820 1. detete the softlimit in mds and osts when use "lfs quota".
1821 2. display the inaccurate data in the output of "lfs quota".
1822 3. try to get quota info when "lfs quota" is executed.
1827 Description: Extent locks not granted with no conflicts sometimes.
1828 Details : When race occurs in glimpse handler and nothing is returned,
1829 we do not reprocess the queue after lock cancel, and that leads
1830 to a stall until next activity on a resource
1833 Frequency : failover with quotaon
1835 Description: during mds failovers with quota on, OSTs got into deadlock state
1836 and causing dumpstack.
1837 Details : for every quota slave, at any time, there is only one quota req
1838 is sent to quota master for every uid/gid. Before that quota req
1839 returns, all the thread relative to the same uid/gid will wait.
1840 So if the quota req is lost because mds failovers or any other
1841 reasons, this bug will be hit. Now, dqacq_interpret() will handle
1842 quota reqs who time out.
1844 Severity : enhancement
1847 Description: when quota slave checks if quota is enough, there is an unnecessary
1849 Details : place this wait on necessary place instead of always waiting.
1851 --------------------------------------------------------------------------------
1853 2007-12-07 Cluster File Systems, Inc. <info@clusterfs.com>
1855 * Support for kernels:
1856 2.6.5-7.286 (SLES 9),
1857 2.6.9-55.0.9.EL (RHEL 4),
1858 2.6.16.53-0.8 (SLES 10),
1859 2.6.18-8.1.14.el5 (RHEL 5),
1860 2.6.18.8 vanilla (kernel.org)
1861 * Client support for unpatched kernels:
1862 (see http://wiki.lustre.org/index.php?title=Patchless_Client)
1863 2.6.16 - 2.6.22 vanilla (kernel.org)
1864 * Due to recently discovered recovery problems, we do not recommend
1865 using patchless RHEL 4 clients with this or any earlier release.
1866 * Recommended e2fsprogs version: 1.40.2-cfs1
1867 * Note that reiserfs quotas are disabled on SLES 10 in this kernel.
1870 Frequency : occasional
1872 Description: MDS hang or stay in waiting lock
1873 Details : If client receive lock with CBPENDING flag ldlm need send lock
1874 cancel as separate rpc, to avoid situation when cancel request
1875 can't processed due all i/o threads stay in wait lock.
1878 Frequency : occasional
1880 Description: Do not fail import if osc_interpret_create gets -EAGAIN
1881 Details : If osc_interpret_create got -EAGAIN it immediately exits and
1882 wakeup oscc_waitq. After wakeup oscc_wait_for_objects call
1883 oscc_has_objects and see OSC has no objests and call
1884 oscc_internal_create to resend create request.
1886 Severity : enhancement
1888 Description: Update kernel patches for SLES10 2.6.16.53-0.8.
1889 Details : Update which_patch & target file for SLES10 latest kernel.
1891 Severity : enhancement
1893 Description: add --type and --size parameters to lfs find
1894 Details : Enhance lfs find by adding filetype and filesize parameters. Also
1895 multiple OBDs can now be specified for the --obd option.
1897 Severity : enhancement
1899 Description: eliminate client locks in face of contention
1900 Details : file contention detection and lockless i/o implementation
1901 for contended files.
1903 Severity : enhancement
1905 Description: Remove client patches from SLES 10 kernel.
1906 Details : This causes SLES 10 clients to behave as patchless clients
1907 even on a Lustre-patched (server) kernel.
1909 Severity : enhancement
1911 Description: use i_size_read and i_size_write in 2.6 port
1912 Details : replace inode->i_size access with i_size_read/write()
1915 Frequency : when removing large files
1917 Description: scheduling issue during removal of large Lustre files
1918 Details : Don't take the BKL in fsfilt_ext3_setattr() for 2.6 kernels.
1919 It causes scheduling issues when removing large files (17TB in the
1925 Description: 1.4.11 Can't handle directories with stripe set and extended ACLs
1926 Details : Impossible (EPROTO is returned) to access a directory that has a
1927 non-default striping and ACLs.
1930 Frequency : only on ppc
1932 Description: /proc/fs/lustre/devices broken on ppc
1933 Details : The patch as applied to 1.6.2 doesn't look correct for all arches.
1934 We should make sure the type of 'index' is loff_t and then cast
1935 explicitly as needed below. Do not assign an explicitly cast
1939 Frequency : only for rhel5
1941 Description: Kernel patches update for RHEL5 2.6.18-8.1.10.el5.
1942 Details : Modify the target file & which_kernel.
1945 Frequency : if the uninit_groups feature is enabled on ldiskfs
1947 Description: e2fsck reports "invalid unused inodes count"
1948 Details : If a new ldiskfs filesystem is created with the "uninit_groups"
1949 feature and only a single inode is created in a group then the
1950 "bg_unused_inodes" count is incorrectly updated. Creating a
1951 second inode in that group would update it correctly.
1956 Description: buffer overruns could theoretically occur
1957 Details : llapi_semantic_traverse() modifies the "path" argument by
1958 appending values to the end of the origin string, and a buffer
1959 overrun may occur. Adding buffer overrun check in liblustreapi.
1961 Severity : enhancement
1963 Description: Add jbd statistics patch for RHEL5 and 2.6.18-vanilla.
1966 Frequency : only if filesystem is inconsistent
1968 Description: handle "serious error: objid * already exists" more gracefully
1969 Details : If LAST_ID value on disk is smaller than the objects existing in
1970 the O/0/d* directories, it indicates disk corruption and causes an
1971 LBUG(). If the object is 0-length, then we should use the existing
1972 object. This will help to avoid a full fsck in most cases.
1974 Severity : enhancement
1976 Description: Kernel patches update for RHEL4 2.6.9-55.0.6.
1977 Details : Modify vm-tunables-rhel4.patch.
1979 Severity : enhancement
1981 Description: Kernel config for 2.6.18-vanilla.
1982 Details : Modify targets/2.6-vanilla.target.in.
1983 Add config file kernel-2.6.18-2.6-vanilla-i686.config.
1984 Add config file kernel-2.6.18-2.6-vanilla-i686-smp.config.
1985 Add config file kernel-2.6.18-2.6-vanilla-x86_64.config.
1986 Add config file kernel-2.6.18-2.6-vanilla-x86_64-smp.config.
1989 Frequency : occasional
1991 Description: improve handling recoverable errors
1992 Details : If request processed with error which can be recoverable on server
1993 request should be resend, otherwise page released from cache and
1999 Description: Kernel patches update for RHEL5 2.6.18-8.1.14.el5.
2000 Details : Modify target file & which_patch.
2001 A flaw was found in the IA32 system call emulation provided
2002 on AMD64 and Intel 64 platforms. An improperly validated 64-bit
2003 value could be stored in the %RAX register, which could trigger an
2004 out-of-bounds system call table access. An untrusted local user
2005 could exploit this flaw to run code in the kernel
2006 (ie a root privilege escalation). (CVE-2007-4573).
2010 Description: change order of libsysio includes
2011 Details : '#include sysio.h' should always come before '#include xtio.h'
2013 Severity : enhancement
2015 Description: adapt the lustre_config script to support the upgrade case
2016 Details : Add "-u" option for lustre_config script to support upgrading 1.4
2017 server targets to 1.6 in parallel.
2022 Description: To avoid grant space > avaible space when the disk is almost
2023 full. Without this patch you might see the error "grant XXXX >
2024 available" or some LBUG about grant, when the disk is almost
2026 Details : In filter_check_grant, for non_grant cache write, we should
2027 check the left space by if (*left > ungranted + bytes), instead
2028 of (*left > ungranted), because only we are sure the left space
2029 is enough for another "bytes", then the ungrant space should be
2030 increase. In client, we should update cl_avail_grant only there is
2031 OBD_MD_FLGRANT in the reply.
2036 Description: Update RHEL 4 kernel to fix local root privilege escalation.
2037 Details : Update to the latest RHEL 4 kernel to fix the vulnerability
2038 described in CVE-2007-4573. This problem could allow untrusted
2039 local users to gain root access.
2042 Frequency : when using O_DIRECT and quotas
2044 Description: Incorrect file ownership on O_DIRECT output files
2045 Details : block usage reported by 'lfs quota' does not take into account
2046 files that have been written with O_DIRECT.
2050 Description: (rw.c:1323:ll_read_ahead_pages()) ASSERTION(page_idx > ria->ria_stoff) failed
2051 Details : Once the unmatched stride IO mode is detected, shrink the stride-ahead
2052 window to 0. If it does hit cache miss, and read-pattern is still
2053 stride-io mode, does not reset the stride window, but also does not
2054 increase the stride window length in this case.
2056 --------------------------------------------------------------------------------
2058 2007-09-27 Cluster File Systems, Inc. <info@clusterfs.com>
2060 * Support for kernels:
2061 2.6.5-7.286 (SLES 9),
2062 2.6.9-55.0.2.EL (RHEL 4),
2063 2.6.16.46-0.14 (SLES 10),
2064 2.6.18-8.1.8.el5 (RHEL 5),
2065 2.6.18.8 vanilla (kernel.org)
2066 * Client support for unpatched kernels:
2067 (see http://wiki.lustre.org/index.php?title=Patchless_Client)
2068 2.6.16 - 2.6.21 vanilla (kernel.org)
2069 * Due to recently discovered recovery problems, we do not recommend
2070 using patchless RHEL 4 clients with this or any earlier release.
2071 * Recommended e2fsprogs version: 1.40.2-cfs1
2072 * Note that reiserfs quotas are disabled on SLES 10 in this kernel.
2076 Description: Fix errors in lfs documentation
2077 Details : Fixes man pages
2079 Severity : enhancement
2081 Description: Adaptive timeouts
2082 Details : RPC timeouts adapt to changing server load and network
2083 conditions to reduce resend attempts and improve recovery time.
2085 Severity : enhancement
2087 Description: llapi_file_create() does not allow some changes
2088 Details : add llapi_file_open() that allows specifying the file creation
2089 mode and open flags, and also returns an open file handle.
2091 Severity : enhancement
2093 Description: df doesn't work properly if diskfs blocksize != 4K
2094 Details : Choose biggest blocksize of OST's as the LOV's blocksize.
2097 Frequency : always on directories with default striping set
2099 Description: lfs find on -1 stripe looping in lsm_lmm_verify_common()
2100 Details : Avoid lov_verify_lmm_common() on directory with -1 stripe count.
2102 Severity : enhancement
2104 Description: merge and cleanup kernel patches.
2105 Details : Remove mnt_lustre_list in vfs_intent-2.6-rhel4.patch.
2107 Severity : enhancement
2109 Description: RedHat Update kernel for RHEL5
2110 Details : Kernel config file for RHEL5.
2112 Severity : enhancement
2114 Description: OSS needs mutliple precreate threads
2115 Details : Add ability to start more than one create thread per OSS.
2118 Frequency : only with quota on the root user
2120 Description: mds_obd_create error creating tmp object
2121 Details : When the user sets quota on root, llog will be affected and can't
2122 create files and write files.
2125 Frequency : Always on ia64 patchless client, and possibly others.
2127 Description: Add EXPORT_SYMBOL check for node_to_cpumask symbol.
2128 Details : This allows the patchless client to be loaded on architectures
2129 without this export.
2134 Description: disorder of journal start and llog_add cause deadlock.
2135 Details : in llog_origin_connect, journal start should happen before
2136 llog_add keep the same order as other functions to avoid
2139 Severity : enhancement
2141 Description: RedHat Update kernel for RHEL5
2142 Details : Modify the kernel config file more closer RHEL5.
2144 Severity : enhancement
2146 Description: Build failure against Centos5 (RHEL5)
2147 Details : Define PAGE_SIZE when it isn't present.
2150 Frequency : occasionally when using NFS
2152 Description: "ll_intent_file_open()) lock enqueue: err: -13" with nfs
2153 Details : with NFS, the anon dentry's parent was set to itself in
2154 d_alloc_anon(), so in MDS, we use rec->ur_fid1 to find the
2155 corresponding dentry other than use rec->ur_name.
2157 Severity : enhancement
2159 Description: enable data checksumming by default
2160 Details : enable checksum by default, allow --disable-checksum
2161 configure option and "-o nochecksum" mount option. Checksums
2162 can also be disabled at runtime via $LPROC/osc/*/checksum_pages.
2165 Frequency : Occasionally with failover
2167 Description: Client eviction due to failover config
2168 Details : after a connection loss, the lustre client should attempt to
2169 reconnect to the last active server first before trying the
2170 other potential connections.
2172 Severity : enhancement
2174 Description: client-side metadata stat-ahead during readdir(directory readahead)
2175 Details : perform client-side metadata stat-ahead when the client detects
2176 readdir and sequential stat of dir entries therein
2181 Description: when mds and osts use different quota unit(32bit and 64bit),
2182 quota will be released repeatly.
2183 Details : void sending multiple quota reqs to mds, which will keep the
2184 status between the reqs.
2187 Frequency : only with liblustre clients on XT3
2189 Description: evictions taking too long
2190 Details : allow llrd to evict clients directly on OSTs
2194 Description: osts not allocated evenly to files
2195 Details : change the condition to increase offset_idx
2197 Severity : enhancement
2199 Description: self-adjustable client's lru lists
2200 Details : use adaptive algorithm for managing client cached locks lru
2201 lists according to current server load, other client's work
2202 pattern, memory activities, etc. Both, server and client
2203 side namespaces provide number of proc tunables for controlling
2208 Description: rewrite ext2-derived code in llite/dir.c and obdclass/uuid.c
2209 Details : rewrite inherited code (uuid parsing code from ext2 utils and
2210 readdir code from ext3) from scratch preserving functionality.
2214 Description: Only those disconnect error should be returned by rq_status.
2215 Details : In open/enqueue processs, Some errors, which will cause client
2216 disconnected, should be returned by rq_status, while other
2217 errors should still be returned by intent, then mdc or llite will
2220 Severity : enhancement
2222 Description: Tune the kernel for good SCSI performance.
2223 Details : Set the value of /sys/block/{dev}/queue/max_sectors_kb
2224 to the value of /sys/block/{dev}/queue/max_hw_sectors_kb
2228 Frequency : Always for filesystems larger than 2TB on 32-bit systems.
2229 Bugzilla : 13547, 13627
2230 Description: Data corruption for OSTs that are formatted larger than 2TB
2232 Details : When generating the bio request for lustre file writes the
2233 sector number would overflow a temporary variable before being
2234 used for the IO. The data reads correctly from Lustre (which
2235 will overflow in a similar manner) but other file data or
2236 filesystem metadata may be corrupted in some cases.
2240 Description: "lfs find -obd UUID" prints directories
2241 Details : "lfs find -obd UUID" will return all directory names instead
2242 of just file names. It is incorrect because the directories
2243 do not reside on the OSTs.
2246 Bugzilla : 13596, 13614
2247 Description: MDS hang after unclean shutdown of lots of clients
2248 Details : Never resend AST requests.
2252 Frequency : Always, for kernels after 2.6.16
2253 Description: Fix warning idr_remove called for id=.. which is not allocated.
2254 Details : Last kernels save old s_dev before kill super and not allow
2255 to restore from callback - restore it before call kill_anon_super.
2259 Description: Fix error on 'ls .' at the top of the Lustre mount.
2260 Details : Don't revalidate dentry if it is a root dentry.
2262 --------------------------------------------------------------------------------
2264 2007-08-27 Cluster File Systems, Inc. <info@clusterfs.com>
2266 * Support for kernels:
2267 2.6.5-7.286 (SLES 9),
2268 2.6.9-55.0.2.EL (RHEL 4),
2269 2.6.16.46-0.14 (SLES 10),
2270 2.6.18-8.1.8.el5 (RHEL 5),
2271 2.6.18.8 vanilla (kernel.org)
2272 * Client support for unpatched kernels:
2273 (see http://wiki.lustre.org/index.php?title=Patchless_Client)
2274 2.6.16 - 2.6.22 vanilla (kernel.org)
2275 * Due to recently discovered recovery problems, we do not recommend
2276 using patchless RHEL 4 clients with this or any earlier release.
2277 * Recommended e2fsprogs version: 1.39.cfs8
2278 * Note that reiserfs quotas are disabled on SLES 10 in this kernel.
2283 Description: block reactivating mgc import until all deactivates complete
2284 Details : Fix race when failing back MDT/MGS to itself (testing)
2286 Severity : enhancement
2288 Description: lfs setstripe enhancement
2289 Details : Make lfs setstripe understand 'k', 'm' and 'g' for stripe size.
2292 Frequency : mds/oss recovery
2294 Description: llog ctxt is refrenced after it has been freed.
2295 Details : llog ctxt refcount was added to avoide the race between ctxt free
2296 and llog recovery process. Each llog user must hold ctxt refcount
2297 before it access the llog. And the llog ctxt can only be freed
2298 when its refcount is zero.
2300 Severity : enhancement
2302 Description: randomly memory allocation failure util
2303 Details : Make lustre randomly failed allocating memory for testing purpose.
2306 Frequency : only for SLES10
2308 Description: Update kernel patch for SLES10 SP1
2309 Details : Add patch blkdev_tunables-2.6-sles10.patch to 2.6-sles10.series.
2311 Severity : enhancement
2313 Description: omit set fsid for export NFS
2314 Details : fix set/restore device id for avoid EMFILE error and mark lustre fs
2315 as FS_REQUIRES_DEV for avoid problems with generate fsid.
2318 Frequency : after network failures
2320 Description: Add sync option to mount_lustre.c
2321 Details : Client loses data written to lustre after a network interruption.
2323 Severity : enhancement
2325 Description: Error message improvement.
2326 Details : Merging of two LCONSOLE_ERROR_MSG into one.
2328 Severity : enhancement
2330 Description: don't use GFP_* in generic Lustre code.
2331 Details : Use cfs_alloc_* functions and CFS_* flags for code portability.
2333 Severity : enhancement
2335 Description: obdclass is limited by single OBD_ALLOC(idarray)
2336 Details : replace OBD_ALLOC/OBD_FREE with OBD_VMALLOC/OBD_VFREE
2338 Severity : enhancement
2340 Description: updated patchess for new RHEL4 kernel
2341 Details : Fixed ext3-unlink-race.patch per Kalpak's comment.
2343 Severity : enhancement
2345 Description: warnings with build patchless client with vanila 2.6.19 and up
2346 Details : change old ctl_table style and replace ctl_table/ctl_table_header
2347 with cfs_sysctl_table_t/cfs_sysctl_table_header_t
2349 Severity : enhancement
2351 Description: O_DIRECT bypasses client statistics.
2352 Details : When running with O_DIRECT I/O, neither the client rpc_stats nor
2353 read_ahead_stats were updated. Copied stats section from
2354 osc_send_oap_rpc() into async_internal().
2357 Frequency : only for Cray XT3
2359 Description: peer credits not enough on many OST per OSS systems.
2360 Details : Use new lnet way to add credits as we need those for pings and ASTs
2363 Frequency : only with liblustre
2365 Description: Liblustre is not releasing flock locks on file close.
2366 Details : Release flock locks on file close.
2369 Frequency : only for RHEL4
2371 Description: Update kernel patches for kernel-2.6.9-55.0.2.EL
2372 Details : Remove inode-nr_unused-2.6.9-rhel4.patch from 2.6-rhel4.series
2373 Update target file and kernel config.
2377 Description: lustre support for RHEL5
2378 Details : Add support for RHEL5.
2383 Description: ASSERTION(export != NULL) failed in target_handle_connect
2384 Details : Assetion hit is result of rare race between disconnect and connet
2385 to same nid. target_handle_connect found old connect cockie and
2386 tried to reconnect, but can't find export for this cockie.
2391 Description: umount blocks forever on error
2392 Details : In result of wrong using obd_no_recov and obd_force flags client
2393 can hand if cancel or some other requests is lost.
2396 Frequency : Only for SLES
2398 Description: sanity_quota fail test_1
2399 Details : There are multiple occurences of $TSTUSR in SLES's /etc/group
2400 file, which makes TSTID[2] inunique.
2402 Severity : enhancement
2404 Description: Kernel patches for SLES9 2.6.5-7.286 kernel
2405 Details : Update target/ChangeLog/which_patch .
2407 Severity : enhancement
2409 Description: jbd statistics
2410 Details : Port older jbd statistics patch for sles10
2415 Description: lvbo_init failed for resource with missing objects.
2416 Details : Fix returning error if we do stat for file with missing/corrupted
2417 objects and i_size set to all sum of size of all avaible objects.
2418 if we to truncate/write to missing object - it is recreated.
2423 Description: Oops in read and write path when failing to allocate lock.
2424 Details : Check if lock allocation failed and return error back.
2427 Frequency : When flocks are used.
2429 Description: assertion failure in ldlm_cli_enquque_fini for non NULL lock.
2430 Details : Flock locks might destroy just granted lock if it could be merged
2431 with another existing flock, this is done in completion handler,
2432 so teach ldlm_cli_enquque_fini that this is a valid case for
2438 Description: reply_lock_interpret crash due to race with it and lock cancel.
2439 Details : Do not replay locks that are being cancelled. Do not reference
2440 locks by their address during replay, just by their handle.
2443 Frequency : only with deactivated OSTs
2445 Description: lstripe command fails for valid OST index
2446 Details : The stripe offset is compared to 'lov->desc.ld_tgt_count'
2447 instead of lov->desc.ld_active_tgt_count.
2449 Severity : enhancement
2451 Description: Build failure against Centos5 (RHEL5)
2452 Details : Use getpagesize() instead of PAGE_SIZE.
2454 --------------------------------------------------------------------------------
2456 2007-07-30 Cluster File Systems, Inc. <info@clusterfs.com>
2458 * Support for kernels:
2459 2.6.5-7.283 (SLES 9),
2460 2.6.9-55.EL (RHEL 4),
2461 2.6.16.46-0.14 (SLES 10),
2462 2.6.18.8 vanilla (kernel.org)
2463 * Client support for unpatched kernels:
2464 (see http://wiki.lustre.org/index.php?title=Patchless_Client)
2465 2.6.16 - 2.6.22 vanilla (kernel.org)
2466 * Due to recently discovered recovery problems, we do not recommend
2467 using patchless RHEL 4 clients with this or any earlier release.
2468 * Recommended e2fsprogs version: 1.39.cfs8
2469 * Note that reiserfs quotas are disabled on SLES 10 in this kernel.
2470 * Starting with this release, the ldiskfs backing filesystem required
2471 by Lustre is now in its own package, lustre-ldiskfs. This package
2472 should be installed. It is versioned separately from Lustre and
2473 may be released separately in future.
2478 Description: server LBUG when shutting down
2479 Details : Block umount forever until the mount refcount is zero rather
2480 than giving up after an arbitrary timeout.
2482 Severity : enhancement
2484 Description: add optional extra BUILD_VERSION info
2485 Details : add a new environment variable (namely LUSTRE_VERS) which allows
2486 to override the lustre version.
2489 Frequency : 2.6.18 servers only
2491 Description: ll_kern_mount() doesn't release the module reference
2492 Details : The ldiskfs module reference count never drops down to 0
2493 because ll_kern_mount() doesn't release the module reference.
2498 Description: server LBUG when using old ost_num_threads parameter
2499 Details : Accept the old ost_num_threads parameter but warn that it
2500 is deprecated, and fix an off-by-one error that caused an LBUG.
2505 Description: Transient SCSI error results in persistent IO issue
2506 Details : iobuf->dr_error is not reinitialized to 0 between two
2510 Frequency : sometimes when underlying device returns I/O errors
2512 Description: OSTs not going read-only during write failures
2513 Details : OSTs are not remounted read-only when the journal commit threads
2514 get I/O errors because fsfilt_ext3 calls journal_start/stop()
2515 instead of the ext3 wrappers.
2518 Frequency : only with 10000 clients or more
2520 Description: poor connect scaling with increasing client count
2521 Details : Don't run filter_grant_sanity_check for more than 100 exports
2522 to improve scaling for large numbers of clients.
2525 Frequency : SLES10 only
2527 Description: sanity-quota.sh quotacheck failed: rc = -22
2528 Details : Quotas cannot be enabled on SLES10.
2531 Frequency : liblustre clients only
2533 Description: getdirentries does not give error when run on compute nodes
2534 Details : getdirentries does not fail when the size specified as an argument
2535 is too small to contain at least one entry
2537 Severity : enhancement
2539 Description: Add LNET router traceability for debug purposes
2540 Details : If a checksum failure occurs with a router as part of the
2541 IO path, the NID of the last router that forwarded the bulk data
2542 is printed so it can be identified.
2547 Description: OST "spontaneously" evicts client; client has imp_pingable == 0
2548 Details : Due to a race condition, liblustre clients were occasionally
2549 evicted incorrectly.
2551 Severity : enhancement
2553 Description: lfs setstripe use optional parameters instead of postional
2556 Severity : enhancement
2558 Description: Nanosecond timestamp support for ldiskfs
2559 Details : The on-disk ldiskfs filesystem has added support for nanosecond
2560 resolution timestamps. There is not yet support for this at
2561 the Lustre filesystem level.
2564 Frequency : during server recovery
2566 Description: MDS failing to send precreate requests due to OSCC_FLAG_RECOVERING
2567 Details : request with rq_no_resend flag not awake l_wait_event if they get a
2571 Frequency : nfs export on patchless client
2573 Description: connectathon hang when test nfs export over patchless client
2574 Details : Disconnected dentry cannot be found with lookup, so we do not need
2575 to unhash it or make it invalid
2578 Description: fix llapi_lov_get_uuids() to allow many OSTs to be returned
2579 Details: : Change llapi_lov_get_uuids() to read the UUIDs from /proc instead
2580 of using an ioctl. This allows lfsck for > 160 OSTs to succeed.
2585 Description: open req refcounting wrong on reconnect
2586 Details : If reconnect happened between getting open reply from server and
2587 call to mdc_set_replay_data in ll_file_open, we will schedule
2588 replay for unreferenced request that we are about to free.
2589 Subsequent close will crash in variety of ways.
2590 Check that request is still eligible for replay in
2591 mdc_set_replay_data().
2596 Description: disable writes to filesystem when reading health_check file
2597 Details : the default for reading the health_check proc file has changed
2598 to NOT do a journal transaction and write to disk, because this
2599 can cause reads of the /proc file to hang and block HA state
2600 checking on a healthy but otherwise heavily loaded system. It
2601 is possible to return to the previous behaviour during configure
2602 with --enable-health-write.
2604 Severity : enhancement
2606 Description: 64-bit inode version
2607 Details: : Add a on-disk 64-bit inode version for ext3 to track changes made
2608 to the inode. This will be required for version-based recovery.
2613 Description: MDS fails to start if a duplicate client export is detected
2614 Details : in some rare cases it was possible for a client to connect to
2615 an MDS multiple times. Upon recovery the MDS would detect this
2616 and fail during startup. Handle this more gracefully.
2618 Severity : enhancement
2620 Description: Add -o localflock option to simulate old noflock behaviour.
2621 Details : This will achieve local-only flock/fcntl locks coherentness.
2626 Description: log_commit_thread vs filter_destroy race leads to crash
2627 Details : Take import reference before releasing llog record semaphore
2632 Description: Wrong request locking in request set processing
2633 Details : ptlrpc_check_set wrongly uses req->rq_lock for proctect add to
2634 imp_delayed_list, in this place should be used imp_lock.
2637 Frequency : when reconnecting
2639 Description: Grant leak when OSC reconnect to OST
2640 Details : When osc reconnect ost, OST(filter) should check whether it
2641 should grant more space to client by comparing fed_grant and
2642 cl_avail_grant, and return the granted space to client instead
2643 of "new granted" space, because client will call osc_init_grant
2644 to update the client grant space info.
2647 Frequency : when client reconnects to OST
2649 Description: Grant leak when OSC does a resend and replays bulk write
2650 Details : When osc reconnect to OST, OST(filter) should clear grant info of
2651 bulk write request, because the grant info while be sync between
2652 OSC and OST when reconnect, and we should ignore the grant info
2653 these of resend/replay write req.
2658 Description: Grant space more than avaiable space sometimes.
2659 Details : When then OST is about to be full, if two bulk writing from
2660 different clients came to OST. Accord the avaliable space of the
2661 OST, the first req should be permitted, and the second one
2662 should be denied by ENOSPC. But if the seconde arrived before
2663 the first one is commited. The OST might wrongly permit second
2664 writing, which will cause grant space > avaiable space.
2667 Frequency : when client is evicted
2669 Description: Grant might be wrongly erased when osc is evicted by OST
2670 Details : when the import is evicted by server, it will fork another
2671 thread ptlrpc_invalidate_import_thread to invalidate the
2672 import, where the grant will be set to 0. While the original
2673 thread will update the grant it got when connecting. So if
2674 the former happened latter, the grant will be wrongly errased
2675 because of this race.
2680 Description: Checking Stale with correct fid
2681 Details : ll_revalidate_it should uses de_inode instead of op_data.fid2
2682 to check whether it is stale, because sometimes, we want the
2683 enqueue happened anyway, and op_data.fid2 will not be initialized.
2685 Severity : enhancement
2687 Description: update patchless client
2688 Details : Add support for patchless client with 2.6.20, 2.6.21 and RHEL 5
2691 Frequency : only with 2.4 kernel
2693 Description: random memory corruption
2694 Details : size of struct ll_inode_info is to big for union inode.u and this
2695 can be cause of random memory corruption.
2700 Description: Memory leak in recovery
2701 Details : Lov_mds_md was not free in an error handler in mds_create_object.
2702 It should also check obd_fail before fsfilt_start, otherwise if
2703 fsfilt_start return -EROFS,(failover mds during mds recovery).
2704 then the req will return with repmsg->transno = 0 and rc = EROFS.
2705 and we met hit the assert LASSERT(req->rq_reqmsg->transno ==
2706 req->rq_repmsg->transno) in ptlrpc_replay_interpret. Fcc should
2707 be freed no matter whether fsfilt_commit success or not.
2710 Frequency : only with large numbers of cores on a single node
2712 Description: Prevents from taking the superblock lock in llap_from_page for
2714 Details : using LL_ORIGIN_REMOVEPAGE origin flag instead of LL_ORIGIN_UNKNOW
2715 for llap_from_page call in ll_removepage() prevents from taking
2716 the superblock lock for a soon died page.
2721 Description: Not check open intent error before release open handle
2722 Details : in some rare cases, the open intent error is not checked before
2723 release open handle, which may cause
2724 ASSERTION(open_req->rq_transno != 0), because it tries to release
2725 the failed open handle.
2730 Description: Set cat log bitmap only after create log success.
2731 Details : in some rare cases, the cat log bitmap is set too early. and it
2732 should be set only after create log success.
2736 Description: Accessing a block bevice can re-enable I/O when Lustre is
2737 tearing down a device.
2738 Details : dev_clear_rdonly(bdev) must be called in kill_bdev() instead of
2743 Description: service threads may hog cpus when there are a lot of requests
2744 Details : Insert cond_resched to give other threads a chance to use some CPU
2749 Description: the cat log was not initialized in recovery
2750 Details : When mds(mgs) do recovery, the tgt_count might be zero, so the
2751 unlink log on mds will not be initialized until mds post
2752 recovery. And also in mds post recovery, the unlink log will
2753 initialization will be done asynchronausly, so there will be race
2754 between add unlink log and unlink log initialization.
2758 Description: brw_stats were being printed incorrectly
2759 Details : brw_stats were being printed as log2 but all of them were not
2760 recorded as log2. Also remove some code duplication arising from
2761 filter_tally_{read,write}.
2765 Frequency : rare, only in recovery.
2766 Description: ASSERTION(req->rq_type != LI_POISON) failed
2767 Details : imp_lock should be held while iterating over imp_sending_list for
2768 prevent destroy request after get timeout in ptlrpc_queue_wait.
2772 Description: replay-single.sh test 52 fails
2773 Details : A lock's skiplist need to be cleanup when it being unlinked
2774 from its resource list.
2779 Description: Short directio read returns full requested size rather than
2781 Details : Direct I/O operations should return actual amount of bytes
2782 transferred rather than requested size.
2784 Severity : enhancement
2786 Description: metadata RPC reduction (e.g. for rm performance)
2787 Details : decrease the amount of synchronous RPC between clients and servers
2788 by canceling conflicing lock before the operation on the client
2789 and packing thier handles into the main operation RPC to server.
2791 Severity : enhancement
2793 Description: add #ifdef HAVE_KERNEL_CONFIG_H
2794 Details : kernels from 2.6.19 not need include linux/config.h, but add
2795 include linux/autoconf.h in commpiler command line.
2797 Severity : enhancement
2799 Description: patchless client support for 2.6.22 kernel
2800 Details : 2.6.22 has only one visble change, SLAB_CTOR_* constants is
2801 removed. In this case we need drop using os depended interface to
2802 kmem_cache and use cfs_mem_cache API.
2807 Description: fix mal-formatted messages
2808 Details : fix some mal-formatted DEBUG_REQ and LCONSOLE_ERROR_MSG messages
2812 Frequency : always in liblustre
2813 Description: wrong IS_ERR implementation in liblustre.h
2814 Details : fix IS_ERR implementation in liblustre.h for right detect errors.
2819 Description: Correct condition for output debug message.
2820 Details : inode i_nlink equal zero is not enough for output message about
2821 disk corruption, i_ctime and i_mode should be also checked.
2825 Frequency : always in patchless client
2826 Description: add configure check for truncate_complete_page
2827 Details : improve checks for exported symbols. This allow run check without
2828 sources, but with Module.symvers shipped with kernel distribution.
2829 add check for truncate_complete_page used by patchless client.
2833 Description: sanity.sh test_77h fails with "test_77h file compare failed"
2834 Details : test_77h uses a file which was messed by other test case.
2838 Description: Not Check whether lov_tgts is NULL in some lov functions
2839 Details : Checking whether lov_tgts is NULL in some functions.
2843 Description: replace obdo_alloc() with OBDO_ALLOC macro
2844 Details : nothing special is done in obdo_alloc() function, and for
2845 debugging purpose, it needs to be replaced with macros.
2849 Description: bad return value and errno from fcntl call
2850 Details : In liblustre API, errno should be a negative value if error
2855 Description: ptlrpc_check_set() LBUG
2856 Details : In case of positive reply from server and failed client bulk
2857 callback after bulk transfer shouldn't LBUG, but process this
2858 request as erroneous.
2860 Severity : enhancement
2862 Description: VFS operations stats tool.
2863 Details : Tool which collects stats by tracking value written in pid,
2864 ppid, gid and uses llstat to generate output to plot graph using
2866 Updated lustre/utils/Makefile.am
2867 Added lustre/utils/ltrack_stats.c
2869 Severity : enhancement
2871 Description: 2.6.18 server support (lustre 1.6.1)
2872 Details : Support for 2.6.18 kernels on the server side.
2877 Description: ASSERTION(imp->imp_conn_current) failed
2878 Details : an assertion failure is hit if a client node boots and attempts to
2879 mount a lustre filesystem faster than RECONNECT_INTERVAL seconds.
2882 Frequency : only for i686
2884 Description: 1.4.11 RC1 build fails for RHEL 4, i686
2885 Details : Fixed config variable for build.
2890 Description: Updated patchess for new RHEL4 kernel
2891 Details : Updated patch inode-nr_unused-2.6.9-rhel4.patch
2892 Updated patch jbd-stats-2.6.9.patch
2893 Updated patch qsnet-rhel4-2.6.patch
2894 Updated patch quota-deadlock-on-pagelock-core.patch
2895 Updated patch vfs_intent-2.6-rhel4.patch
2896 Updated patch vfs_races-2.6-rhel4.patch
2897 Updated series file 2.6-rhel4-titech.series
2898 Updated series file 2.6-rhel4.series
2899 Updated kernel config files
2904 Description: lquota slave complains LBUG when reconnecting with mds
2906 Details : quota slave depends qctxt->lqc_import to send its quota request.
2907 This pointer will be invalid if mds did failover or broke its
2908 connect to osts, which leads to LBUG.
2911 Frequency : when qunit size is too small(less than 20M)
2913 Description: write is stopped by improper -EDQUOT
2914 Details : If the master is busy and qunit size is small enough(let's say 1M),
2915 the slave can not get quota from master on time, which will lead
2916 slave to trigger a -EQUOTA to client.
2921 Description: Deadlock during metadata tests
2922 Details : in prune_dir_dentries(), shrink_dcache_parent() should not be
2923 called with the per-dentry lock held.
2926 Frequency : SLES9 only
2928 Description: Lustre patched kernel for SLES9 SP3 has NR_CPUS set to 8
2929 Details : set CONFIG_NR_CPUS to 128 instead of 8.
2931 Severity : enhancement
2933 Description: remove fs_prep_san_write operation and related patches
2934 Details : remove the ext3-san-jdike patches which are no longer useful.
2939 Description: LDISKFS-fs error (device sdc): ldiskfs_free_blocks
2940 Details : a disk corruption can cause the mballoc code to assert on a
2941 double free or other extent corruptions. Handle these with
2942 ext3_error() instead of assertions.
2945 Frequency : only with mballoc3 code and deep extent trees
2947 Description: ldiskfs_ext_search_right: bad header in inode: unexpected eh_depth
2948 Details : a wrong check of extent headers in ldiskfs_ext_search_right()
2949 can cause the filesystem to be remounted read-only.
2954 Description: lfsck built against 1.4.x cannot run against 1.6.0 lustre
2955 Details : the definition for OBD_IOC_GETNAME changed in 1.6.0. One of the
2956 few external users of this ioctl number is lfsck's call to
2957 llapi_lov_get_uuids() and this caused lfsck to fail at startup.
2958 Add the old ioctl number to the handler so both old and new
2963 Description: parallel lock callbacks
2964 Details : Instead of sending blocking and completion callbacks as separated
2965 requests, adding them to a set and sending in parallel.
2969 Description: Disable most debugging by default
2970 Details : To improve performance, disable most logging (for debug purposes)
2971 by default. VFSTRACE, RPCTRACE, and DLMTRACE are now off by
2972 default, and HA includes fewer messages.
2976 Frequency : only run on patchless client.
2977 Description: use do_facet on sanity.sh for test handling recoverables errors
2978 Details : use do_facet instead of direct use sysctl for set fail_loc on OST
2982 Description: hash tables for lists of nids, connections and uuids
2983 Details : Hash tables noticeably help when a lot of clients connect to a
2984 server, to faster identify duplicate connections or reconnects,
2985 also to faster find export to evict in manual eviction case.
2989 Description: Sometimes, when the server evict a client, and the client will
2990 not be evicted as soon as possible.
2991 Details : In enqueue req, the error was returned by intent, instead of
2992 rq_status which make ptlrpc layer not detect this error, and
2993 does not evict the client. So enqueue error should be returned
2997 Frequency : only at startup
2999 Description: Delay client connections to MDT until fist MDT->OST connect
3000 Details : If a client tried to create a new file before the MDT had
3001 connected to any OSTs, the create would return EIO. Now
3002 the client will simply block until the MDT connects to the
3003 first OST and the create can succeed.
3006 Frequency : only for SLES9
3008 Description: Routinely utilize latest Quadrics drivers in CFS releases
3009 Details : Update patch qsnet-suse-2.6.patch.
3012 Frequency : only for sles10
3014 Description: Update patches for SLES 10 SP1 kernel.
3015 Details : Update the patch vfs_intent-2.6-sles10.patch.
3020 Description: Routinely utilize latest Quadrics drivers in CFS releases
3021 Details : Update patch qsnet-rhel4-2.6.patch.
3024 Frequency : at statup only
3026 Description: mds_lov_synchronize race leads to various problems
3027 Details : simultaneous MDT->OST connections at startup can cause the
3028 sync to abort, leaving the OSC in a bad state.
3033 Description: Using wrong pointer in osc_brw_prep_request
3034 Details : Access to array[-1] can produce panic if kernel compiled with
3035 CONFIG_PAGE_ALLOC enabled
3037 Severity : enhancement
3039 Description: Async OSC create to avoid the blocking unnecessarily.
3040 Details : If a OST has no remain object, system will block on the creating
3041 when need to create a new object on this OST. Now, ways use
3042 pre-created objects when available, instead of blocking on an
3043 empty osc while others are not empty. If we must block, we block
3044 for the shortest possible period of time.
3048 Frequency : only in recovery
3049 Description: Mark OST as early accessible if his start SYNC.
3050 Details : osc_precreate return flag early accessible if oscc marked as
3051 OSCC_FLAG_SYNC_IN_PROGRESS.
3056 Description: Sometimes precreate code can triger create object on wrong ost
3057 Details : Wrong protected or not not restored variables aftre precreate loop
3058 can produce creation object on wrong ost.
3061 Frequency : oss recovery
3063 Description: llog_commit_thread cleanup should sync with llog_commit_thread
3065 Details : llog_commit_thread_count should be synced between llog_commit
3066 start and cleanup, so new llog_commit thread should not be started
3067 when llog_commit threads being stopped to avoid accessing some
3070 Severity : enhancement
3072 Description: Add printing inode info into message about error in writepage.
3074 --------------------------------------------------------------------------------
3076 2007-05-03 Cluster File Systems, Inc. <info@clusterfs.com>
3081 Frequency : on some architectures
3083 Description: 1.6 client sometimes fails to mount from a 1.4 MDT
3084 Details : Uninitialized flags sometimes cause configuration commands to
3088 Frequency : patchless clients only
3090 Description: missing __iget() symbol export
3091 Details : The __iget() symbol export is missing. To avoid the need for
3092 this on patchless clients the deathrow inode reaper is turned
3093 off, and we depend on the VM to clean up old inodes. This
3094 dependency was during via the fix for bug 12181.
3099 Description: sanity.sh fail: test_52b
3100 Details : The ll_inode_to_ext_flags() has a glitch which makes MDS return
3101 incorrect inode's flags to client.
3103 --------------------------------------------------------------------------------
3105 2007-04-19 Cluster File Systems, Inc. <info@clusterfs.com>
3107 * CONFIGURATION CHANGE. This version of Lustre WILL NOT
3108 INTEROPERATE with older versions automatically. In many cases a
3109 special upgrade step is needed. Please read the
3110 user documentation before upgrading any part of a 1.4.x system.
3111 * WARNING: Lustre configuration and startup changes are required with
3112 this release. See https://mail.clusterfs.com/wikis/lustre/MountConf
3114 * Support for kernels:
3115 2.4.21-47.0.1.EL (RHEL 3),
3116 2.6.5-7.283 (SLES 9),
3117 2.6.9-42.0.10.EL (RHEL 4),
3118 2.6.12.6 vanilla (kernel.org),
3119 2.6.16.27-0.9 (SLES10)
3120 * Client support for unpatched kernels:
3121 (see https://mail.clusterfs.com/wikis/lustre/PatchlessClient)
3122 2.6.16 - 2.6.19 vanilla (kernel.org),
3123 2.6.9-42.0.8EL (RHEL 4)
3124 * Recommended e2fsprogs version: 1.39.cfs6
3125 * Note that reiserfs quotas are disabled on SLES 10 in this kernel
3128 Severity : enhancement
3130 Description: Async OSC create to avoid the blocking unnecessarily.
3131 Details : If a OST has no remain object, system will block on the creating
3132 when need to create a new object on this OST. Now, ways use
3133 pre-created objects when available, instead of blocking on an
3134 empty osc while others are not empty. If we must block, we block
3135 for the shortest possible period of time.
3137 Severity : enhancement
3139 Description: MountConf
3140 Details : Lustre configuration is now managed via mkfs and mount
3141 commands instead of lmc and lconf. New obd types (MGS, MGC)
3142 are added for dynamic configuration management. See
3143 https://mail.clusterfs.com/wikis/lustre/MountConf for
3146 Severity : enhancement
3148 Description: dynamic OST addition
3149 Details : OSTs can now be added to a live filesystem
3151 Severity : enhancement
3153 Description: startup order invariance
3154 Details : MDTs and OSTs can be started in any order. Clients only
3155 require the MDT to complete startup.
3157 Severity : enhancement
3159 Description: parallel, asynchronous orphan cleanup
3160 Details : orphan cleanup is now performed in separate threads for each
3161 OST, allowing parallel non-blocking operation.
3163 Severity : enhancement
3165 Description: optimized stripe assignment
3166 Details : stripe assignments are now made based on ost space available,
3167 ost previous usage, and OSS previous usage, in order to try
3168 to optimize storage space and networking resources.
3170 Severity : enhancement
3172 Description: Permanently set tunables
3173 Details : All writable /proc/fs/lustre tunables can now be permanently
3174 set on a per-server basis, at mkfs time or on a live
3177 Severity : enhancement
3179 Description: Lustre message v2
3180 Details : Add lustre message format v2.
3182 Severity : enhancement
3184 Description: client OST exclusion list
3185 Details : Clients can be started with a list of OSTs that should be
3186 declared "inactive" for known non-responsive OSTs.
3190 Description: ENOENT returned for valid filehandle during dbench.
3191 Details : Check if a directory has children when invalidating dentries
3192 associated with an inode during lock cancellation. This fixes
3193 an incorrect ENOENT sometimes seen for valid filehandles during
3194 testing with dbench.
3197 Frequency : SFS test only (otherwise harmless)
3199 Description: SPEC SFS validation failure on NFS v2 over lustre.
3200 Details : Changes the blocksize for regular files to be 2x RPC size,
3201 and not depend on stripe size.
3203 Severity : enhancement
3205 Description: fine-grained SMP locking inside DLM
3206 Details : Improve DLM performance on SMP systems by removing the single
3207 per-namespace lock and replace it with per-resource locks.
3209 Severity : enhancement
3211 Description: don't hold multiple extent locks at one time
3212 Details : To avoid client eviction during large writes, locks are not
3213 held on multiple stripes at one time or for very large writes.
3214 Otherwise, clients can block waiting for a lock on a failed OST
3215 while holding locks on other OSTs and be evicted.
3217 Severity : enhancement
3219 Description: Multiple MD RPCs in flight.
3220 Details : Further unserialise some read-only MDT RPCs - learn about intents.
3221 To avoid overly-overloading MDT, introduce a limit on number of
3222 MDT RPCs in flight for a single client and add /proc controls
3223 to adjust this limit.
3225 Severity : enhancement
3227 Description: client read/write statistics
3228 Details : Add client read/write call usage stats for performance
3229 analysis of user processes.
3230 /proc/fs/lustre/llite/*/offset_stats shows non-sequential
3231 file access. extents_stats shows chunk size distribution.
3232 extents_stats_per_process show chunk size distribution per
3235 Severity : enhancement
3237 Description: per-client statistics on server
3238 Details : Add ldlm and operations statistics for each client in
3239 /proc/fs/lustre/mds|obdfilter/*/exports/
3241 Severity : enhancement
3243 Description: improved MDT statistics
3244 Details : Add detailed MDT operations statistics in
3245 /proc/fs/lustre/mds/*/stats
3247 Severity : enhancement
3249 Description: VFS operations stats
3250 Details : Add client VFS call stats, trackable by pid, ppid, or gid
3251 /proc/fs/lustre/llite/*/stats_track_[pid|ppid|gid]
3256 Description: Fix client-side osc byte counters
3257 Details : The osc read/write byte counters in
3258 /proc/fs/lustre/osc/*/stats are now working
3261 Frequency : always as root on SLES
3263 Description: Failure of copying files with lustre special EAs.
3264 Details : Client side always return success for setxattr call for lustre
3265 special xattr (currently only "trusted.lov").
3270 Description: Refcount LNET uuids
3271 Details : The global LNET uuid list grew linearly with every startup;
3272 refcount repeated list entries instead of always adding to
3275 Severity : enhancement
3277 Description: Dynamic service threads
3278 Details : Within a small range, start extra service threads
3279 automatically when the request queue builds up.
3282 Frequency : mixed-endian client/server environments
3284 Description: mixed-endian crashes
3285 Details : The new msg_v2 system had some failures in mixed-endian
3288 Severity : enhancement
3290 Description: Easy OST removal
3291 Details : OSTs can be permanently deactivated with e.g. 'lctl
3292 conf_param lustre-OST0001.osc.active=0'
3294 Severity : enhancement
3296 Description: MGS proc entries
3297 Details : Added basic proc entries for the MGS showing what filesystems
3300 Severity : enhancement
3302 Description: provide MGS failover
3303 Details : Added config lock reacquisition after MGS server failover.
3305 Severity : enhancement
3307 Description: add Linux 2.4 support
3308 Details : Added support for RHEL 2.4.21 kernel for 1.6 servers and clients
3312 Description: a large application tries to do I/O to the same resource and dies
3313 in the middle of it.
3314 Details : Check the req->rq_arrival time after the call to
3315 ost_brw_lock_get(), but before we do anything about
3316 processing it & sending the BULK transfer request. This
3317 should help move old stale pending locks off the queue as
3318 quickly as obd_timeout.
3321 Frequency : when an incorrect nid is specified during startup
3323 Description: ptlrpc connect to non-existant node causes kernel crash
3324 Details : LNET can't be re-entered from an event callback, which
3325 happened when we expire a message after the export has been
3326 cleaned up. Instead, hand the zombie cleanup off to another
3329 Severity : enhancement
3331 Description: plain/inodebits lock performance improvement
3332 Details : Grouping plain/inodebits in granted list by their request modes
3333 and bits policy, thus improving the performance of search through
3337 Frequency : only if OST filesystem is corrupted
3339 Description: client incorrectly hits assertion in ptlrpc_replay_req()
3340 Details : for a short time RPCs with bulk IO are in the replay list,
3341 but replay of bulk IOs is unimplemented. If the OST filesystem
3342 is corrupted due to disk cache incoherency and then replay is
3343 started it is possible to trip an assertion. Avoid putting
3344 committed RPCs into the replay list at all to avoid this issue.
3347 Frequency : liblustre (e.g. catamount) on a large cluster with >= 8 OSTs/OSS
3349 Description: System hang on startup
3350 Details : This bug allowed the liblustre (e.g. catamount) client to
3351 return to the app before handling all startup RPCs. This
3352 could leave the node unresponsive to lustre network traffic
3353 and manifested as a server ptllnd timeout.
3355 Severity : enhancement
3357 Description: Add "/proc/sys/lustre/debug_peer_on_timeout"
3358 Details : liblustre envirable: LIBLUSTRE_DEBUG_PEER_ON_TIMEOUT
3359 boolean to control whether to print peer debug info when a
3360 client's RPC times out.
3363 Frequency : only for kernels with patches from Lustre below 1.4.3
3365 Description: Remove old rdonly API
3366 Details : Remove old rdonly API which unused from at least lustre 1.4.3
3369 Frequency : only for devices with external journals
3371 Description: Set external device read-only also
3372 Details : During a commanded failover stop, we set the disk device
3373 read-only while the server shuts down. We now also set any
3374 external journal device read-only at the same time.
3377 Frequency : when upgrading from 1.4 while trying to change parameters
3379 Description: The wrong (new) MDC name was used when setting parameters for
3380 upgraded MDT's. Also allows changing of OSC (and MDC)
3381 parameters if --writeconf is specified at tunefs upgrade time.
3384 Frequency : when setting specific ost indicies
3386 Description: QOS code breaks on skipped indicies
3387 Details : Add checks for missing OST indicies in the QOS code, so OSTs
3388 created with --index need not be sequential.
3390 Severity : enhancement
3392 Description: Add uninit_groups feature to ldiskfs2 to speed up e2fsck
3393 Details : The uninit_groups feature works in conjunction with the kernel
3394 filesystem code (ldiskfs2 only) and e2fsprogs-1.39-cfs6 to speed
3395 up the pass1 processing of e2fsck. This is a read-only feature
3396 in ldiskfs2 only, so older kernels and current ldiskfs cannot
3397 mount filesystems that have had this feature enabled.
3399 Severity : enhancement
3401 Description: Improve multi-block allocation algorithm to avoid fragmentation
3402 Details : The mballoc3 code (ldiskfs2 only) adds new mechanisms to improve
3403 allocation locality and avoid filesystem fragmentation.
3405 ------------------------------------------------------------------------------
3407 2007-04-01 Cluster File Systems, Inc. <info@clusterfs.com>
3409 * Support for kernels:
3410 2.4.21-47.0.1.EL (RHEL 3)
3411 2.6.5-7.283 (SLES 9)
3412 2.6.9-42.0.10.EL (RHEL 4)
3413 2.6.12.6 vanilla (kernel.org)
3414 2.6.16.27-0.9 (SLES 10)
3415 * Recommended e2fsprogs version: 1.39.cfs5
3417 * Note that reiserfs quotas are disabled on SLES 10 in this kernel
3421 Frequency : occasional, depends on client load and configuration
3422 Bugzilla : 12181, 12203
3423 Description: data loss for recently-modified files
3425 Details : In some cases it is possible that recently written or created
3426 files may not be written to disk in a timely manner (this should
3427 normally be within 30s unless client IO load is very high).
3428 The problem appears as zero-length files or files that are a
3429 multiple of 1MB in size after a client crash or client eviction
3430 that are missing data at the end of the file.
3432 This problem is more likely to be hit on clients where files are
3433 repeatedly created and unlinked in the same directory, clients
3434 have a large amount of RAM, have many CPUs, the filesystem has
3435 many OSTs, the clients are rebooted frequently, and/or the files
3436 are not accessed by other nodes after being written.
3438 The presence of the problem can be detected by looking at
3439 /proc/sys/fs/inode-state. If the first number (nr_inodes) is
3440 smaller than the second (nr_unused) then dirty files will not
3441 be flushed automatically to disk. "sync; sleep 10" should be
3442 run several times on the node before unmounting it to update
3443 Lustre (this is also safe to run on nodes without this problem).
3445 There is also a related kernel bug in the RHEL4 4 2.6.9 kernel
3446 that can cause this same problem, so customers using that kernel
3447 also need to update the kernel in addition to Lustre. In order
3448 to properly fix this bug, the RHEL3 2.4.21 kernel is also updated.
3450 It is normal that files written just before a client crash (less
3451 than 30s) may not yet have been flushed to disk, even for local
3455 Frequency : frequent on thin XT3 nodes
3457 Description: UUID collision on thin XT3 Linux nodes
3458 Details : UUIDs on Compute Node Linux XT3 nodes were not generated
3459 randomly, since we relied on an insufficiently-seeded PRNG.
3464 Description: OSS hangs after "All ost request buffers busy"
3465 Details : A deadlock between quota and journal operations caused OSS
3466 hangs after printing "All ost request buffers busy."
3469 Frequency : always on liblustre builds
3471 Description: Cleanup compiler warnings on liblustre
3474 Frequency : always on liblustre builds on XT3
3476 Description: LC_CONFIG_CDEBUG don't run while build liblustre on XT3.
3480 Description: Addition of EXT3_FEATURE_RO_COMPAT_DIR_NLINKS flag for
3481 > 32000 subdirectories
3482 Details : Add EXT3_FEATURE_RO_COMPAT_DIR_NLINK flag to
3483 EXT3_FEATURE_RO_COMPAT_SUPP. This flag will be set whenever
3484 subdirectory count crosses 32000. This will aid e2fsck to
3485 correctly handle more than 32000 subdirectories.
3488 Frequency : liblustre (e.g. catamount) on a large cluster with >= 8 OSTs/OSS
3490 Description: System hang on startup
3491 Details : This bug allowed the liblustre (e.g. catamount) client to
3492 return to the app before handling all startup RPCs. This
3493 could leave the node unresponsive to lustre network traffic
3494 and manifested as a server ptllnd timeout.
3496 Severity : enhancement
3498 Description: Add "/proc/sys/lustre/debug_peer_on_timeout"
3499 (liblustre envirable: LIBLUSTRE_DEBUG_PEER_ON_TIMEOUT)
3500 boolean to control whether to print peer debug info when a
3501 client's RPC times out.
3506 Description: make O_SYNC working on 2.6 kernels
3507 Details : 2.6 kernels use different method for mark pages for write,
3508 so need add a code to lustre for O_SYNC work.
3513 Description: Failure to close file and release space on NFS
3514 Details : Put inode details into lock acquired in ll_intent_file_open.
3515 Use mdc_intent_lock in ll_intent_open to properly
3516 detect all kind of errors unhandled by mdc_enqueue.
3521 Description: proc file read during shutdown sometimes raced obd removal,
3523 Details : Add lock to prevent obd access after proc file removal.
3526 Frequency : Only for files larger than 4GB on 32-bit clients.
3528 Description: improperly doing page alignment of locks
3529 Details : Modify lustre core code to use CFS_PAGE_* defines instead of
3530 PAGE_*. Make CFS_PAGE_MASK a 64-bit mask.
3535 Description: RPCs being resent when they shouldn't be
3536 Details : Some RPCs that should not be resent are being resent. This
3537 can cause inconsistencies in the RPC state machine. Do not
3538 resend such requests.
3541 Frequency : rare, only with NFS export
3543 Description: Crash on NFS re-export node
3544 Details : under very unusual load conditions an assertion is hit in
3545 ll_intent_file_open()
3548 Frequency : only if OST filesystem is corrupted
3550 Description: client incorrectly hits assertion in ptlrpc_replay_req()
3551 Details : for a short time RPCs with bulk IO are in the replay list,
3552 but replay of bulk IOs is unimplemented. If the OST filesystem
3553 is corrupted due to disk cache incoherency and then replay is
3554 started it is possible to trip an assertion. Avoid putting
3555 committed RPCs into the replay list at all to avoid this issue.
3560 Description: large O_DIRECT requests fail under memory pressure/fragmentation
3561 Details : Large single O_DIRECT read and write calls can fail to allocate
3562 a sufficiently large buffer to process the request. In case of
3563 allocation failure the allocation is retried with a smaller
3564 buffer and broken into smaller requests.
3566 Severity : enhancement
3568 Description: Add -o localflock option to simulate old noflock behaviour.
3569 Details : This will achieve local-only flock/fcntl locks coherentness.
3574 Description: versioning check is incomplete
3575 Details : Checking the version difference of client vs. server, report
3576 error if the gap is too big.
3581 Description: add support PG_writeback bit
3582 Details : add support for PG_writeback bit for Lustre, for more carefull
3583 work with page cache in 2.6 kernel. This also fix some deadlocks
3584 and remove hack for work O_SYNC with 2.6 kernel.
3586 Severity : enhancement
3588 Description: Add uninit_groups feature to ldiskfs2 to speed up e2fsck
3589 Details : The uninit_groups feature works in conjunction with the kernel
3590 filesystem code (ldiskfs2 only) and e2fsprogs-1.39-cfs6 to speed
3591 up the pass1 processing of e2fsck. This is a read-only feature
3592 in ldiskfs2 only, so older kernels and current ldiskfs cannot
3593 mount filesystems that have had this feature enabled.
3595 Severity : enhancement
3597 Description: Improve multi-block allocation algorithm to avoid fragmentation
3598 Details : The mballoc3 code (ldiskfs2 only) adds new mechanisms to improve
3599 allocation locality and avoid filesystem fragmentation.
3601 ------------------------------------------------------------------------------
3603 2007-02-09 Cluster File Systems, Inc. <info@clusterfs.com>
3605 * Support for kernels:
3606 2.6.9-42.0.3.EL (RHEL 4)
3607 2.6.5-7.276 (SLES 9)
3608 2.4.21-47.0.1.EL (RHEL 3)
3609 2.6.12.6 vanilla (kernel.org)
3610 2.6.16.21-0.8 (SLES10)
3611 * Recommended e2fsprogs version: 1.39.cfs2-0
3613 * The backwards-compatible /proc/sys/portals symlink has been removed
3614 in this release. Before upgrading, please ensure that you change
3615 any configuration scripts or /etc/sysctl.conf files that access
3616 /proc/sys/portals/* or sysctl portals.* to use the corresponding
3617 entry in /proc/sys/lnet or sysctl lnet.*. This change can be made
3618 in advance of the upgrade on any system running Lustre 1.4.6 or
3619 newer, since /proc/sys/lnet was added in that version.
3620 * Note that reiserfs quotas are disabled on SLES 10 in this kernel
3624 Frequency : only when quota is used
3626 Description: avoid scanning export list for quota master
3627 Details : Change the algorithms to avoid scanning export list in order
3628 to improve the efficiency.
3631 Frequency : MDS failover only, very rarely
3633 Description: "went back in time" messages on mds failover
3634 Details : The greatest transno may be lost when the current operation
3635 finishes with an error (transno==0) and the client's last_rcvd
3636 record is over-written. Save the greatest transno in the
3637 mds_last_transno for this case.
3640 Frequency : always for specific kernels and striping counts
3642 Description: client may get "Matching packet too big" without ACL support
3643 Details : Clients compiled without CONFIG_FS_POSIX_ACL get an error message
3644 when trying to access files in certain configurations. The
3645 clients should in fact be denied when mounting because they do
3646 not understand ACLs.
3649 Frequency : Cray XT3 with more than 4000 clients and multiple jobs
3651 Description: many clients connecting with IO in progress causes connect timeouts
3652 Details : Avoid synchronous journal commits to avoid delays caused by many
3653 clients connecting/disconnecting when bulk IO is in progress.
3654 Queue liblustre connect requests on OST_REQUEST_PORTAL instead of
3655 OST_IO_PORTAL to avoid delays behind potentially many pending
3659 Frequency : occasionally with multiple writers to a single file
3661 Description: shared writes to file may result in wrong size reported by stat()
3662 Details : Allow growing of kms when extent lock is cancelled
3665 Frequency : always with random mmap IO to multi-striped file
3667 Description: mmap write might be lost if we are writing to a 'hole' in stripe
3668 Details : Only if the hole is at the end of OST object so that kms is too
3669 small. Fix is to increase kms accordingly in ll_nopage.
3672 Frequency : rare, only if OST filesystem is inconsistent with MDS filesystem
3674 Description: writes to a missing object would leak memory on the OST
3675 Details : If there is an inconsistency between the MDS and OST filesystems,
3676 such that the MDS references an object that doesn't exist, writes
3677 to that object will leak memory due to incorrect cleanup in the
3678 error handling path, eventually running out of memory on the OST.
3683 Description: Creating too long symlink causes lustre errors
3684 Details : Check symlink and name lengths before sending requests to MDS.
3687 Frequency : only if flock is enabled (not on by default)
3689 Description: posix locks not released on fd closure on 2.6.9+
3690 Details : We failed to add posix locks to list of inode locks on 2.6.9+
3691 kernels, this caused such locks not to be released on fd close and
3692 then assertions on fs unmount about still used locks.
3695 Frequency : MDS failover only, very rarely
3697 Description: clients may get ASSERTION(granted_lock != NULL)
3698 Details : When request was taking a long time, and a client was resending
3699 a getattr by name lock request. The were multiple lock requests
3700 with the same client lock handle and
3701 mds_getattr_name->fixup_handle_for_resent_request found one of the
3702 lock handles but later failed with ASSERTION(granted_lock != NULL).
3707 Description: handle->h_buffer_credits > 0, assertion failure
3708 Details : h_buffer_credits is zero after truncate, causing assertion
3709 failure. This patch extends the transaction or creates a new
3713 Frequency : NFS re-export or patchless client
3714 Bugzilla : 11179, 10796
3715 Description: Crash on NFS re-export node (__d_move)
3716 Details : We do not want to hash the dentry if we don't have a lock.
3717 But if this dentry is later used in d_move, we'd hit uninitialised
3718 list head d_hash, so we just do this to init d_hash field but
3719 leave dentry unhashed.
3722 Frequency : NFS re-export or patchless client
3724 Description: NFS exports has problem with symbolic link
3725 Details : lustre client didn't properly install dentry when re-exported
3726 to NFS or running patchless client.
3729 Frequency : NFS re-export or patchless client
3731 Description: Various nfs/patchless fixes.
3732 Details : fixes reuse disconected alias for lookup process - this fixes
3733 warning "find_exported_dentry: npd != pd",
3734 fix permission error with open files at nfs.
3735 fix apply umask when do revalidate.
3738 Frequency : occasional
3740 Description: Crash on NFS re-export node
3741 Details : calling clear_page() on the wrong pointer triggered oops in
3742 generic_mapping_read().
3745 Frequency : rarely, using O_DIRECT IO
3747 Description: unaligned directio crashes client with LASSERT
3748 Details : check for unaligned buffers before trying any requests.
3751 Frequency : rarely, using CFS RAID5 patches in non-standard kernel series
3753 Description: stale data returned from RAID cache
3754 Details : If only a small amount of IO is done to the RAID device before
3755 reading it again it is possible to get stale data from the RAID
3756 cache instead of reading it from disk.
3759 Frequency : always for sles10 kernel
3761 Description: sles10 support
3762 Details : ll_follow_link: compile fixes and using of nd_set_link
3763 under newer kernels.
3766 Frequency : depends on arch, kernel and compiler version, always on sles10
3769 Description: recursive or deep enough symlinks cause stack overflow
3770 Details : getting rid of large stack-allocated variable in
3774 Frequency : depends on hardware
3776 Description: lustre write performance loss in the SLES10 kernel
3777 Details : the performance loss is caused by using of write barriers in the
3778 ext3 code. The SLES10 kernel turns barrier support on by
3779 default. The fix is to undo that change for ldiskfs.
3781 ------------------------------------------------------------------------------
3783 2006-12-09 Cluster File Systems, Inc. <info@clusterfs.com>
3785 * Support for kernels:
3786 2.6.9-42.0.3EL (RHEL 4)
3787 2.6.5-7.276 (SLES 9)
3788 2.4.21-47.0.1.EL (RHEL 3)
3789 2.6.12.6 vanilla (kernel.org)
3793 Frequency : quota enabled and large files being deleted
3795 Description: releasing more than 4GB of quota at once hangs OST
3796 Details : If a user deletes more than 4GB of files on a single OST it
3797 will cause the OST to spin in an infinite loop. Release
3798 quota in < 4GB chunks, or use a 64-bit value for 1.4.7.1+.
3803 Description: statfs data retrieved from /proc may be stale or zero
3804 Details : When reading per-device statfs data from /proc, in the
3805 {kbytes,files}_{total,free,avail} files, it may appear
3806 as zero or be out of date.
3809 Frequency : systems with MD RAID1 external journal devices
3811 Description: lconf's call to blkid is confused by RAID1 journal devices
3812 Details : Use the "blkid -l" flag to locate the MD RAID device instead
3813 of returning all block devices that match the journal UUID.
3816 Frequency : always, for aggregate stripe size over 4GB
3818 Description: "lfs setstripe" fails assertion when setting 4GB+ stripe width
3819 Details : Using "lfs setstripe" to set stripe size * stripe count over 4GB
3820 will fail the kernel with "ASSERTION(lsm->lsm_xfersize != 0)"
3823 Frequency : always if "lfs find" used on a local file/directory
3825 Description: "lfs find" segfaults if used on a local file/directory
3826 Details : The case where a directory component was not specified wasn't
3827 handled correctly. Handle this properly.
3830 Frequency : always on ppc64
3832 Description: the write to an ext3 filesystem mounted with mballoc got stuck
3833 Details : ext3_mb_generate_buddy() uses find_next_bit() which does not
3834 perform endianness conversion.
3837 Frequency : rarely (truncate to non-zero file size after write under load)
3838 Bugzilla : 10730, 10687
3839 Description: Files padded with zeros to next 4K multiple
3840 Details : With filesystems mounted using the "extents" option (2.6 kernels)
3841 it is possible that files that are truncated to a non-zero size
3842 immediately after being written are filled with zero bytes beyond
3843 the truncated size. No file data is lost.
3845 Severity : enhancement
3847 Description: Allow recovery/failover for liblustre clients.
3848 Details : liblustre clients were unaware of failover configurations until
3851 Severity : enhancement
3853 Description: user file locks should fail when not mounting with flock option
3854 Details : Set up an error-returning stub in ll_file_operations.lock field
3855 to prevent incorrect behaviour when client is mounted without
3856 flock option. Also, set up properly f_op->flock field for
3860 Frequency : always on ia64
3862 Description: "lfs df" loops on printing out MDS statfs information
3863 Details : The obd_ioctl_data was not initialized and in some systems
3864 this caused a failure during the ioctl that did not return
3865 an error. Initialize the struct and return an error on failure.
3868 Frequency : SLES 9 only
3870 Description: Error of copying files with lustre special EAs as root
3871 Details : Client side always return success for setxattr call for lustre
3872 special xattr (currently only "trusted.lov").
3875 Frequency : rarely on clusters with both ia64+i386 clients
3877 Description: ia64+i686 clients doing shared IO on the same file may LBUG
3878 Details : In rare cases when both ia64+i686 (or other mixed-PAGE_SIZE)
3879 clients are doing concurrent writes to the same file it is
3880 possible that the ia64 clients may LASSERT because the OST
3881 extent locks are not PAGE_SIZE aligned. Ensure that grown
3882 locks are always aligned on the request boundary.
3885 Frequency : specific use, occasional
3887 Description: Overwriting in use executable truncates on-disk binary image
3888 Details : If one node attempts to overwrite an executable in use by
3889 another node, we now correctly return ETXTBSY instead of
3890 truncating the file.
3892 Severity : enhancement
3894 Description: Async OSC create to avoid the blocking unnecessarily.
3895 Details : If a OST has no remain object, system will block on the creating
3896 when need to create a new object on this OST. Now, ways use
3897 pre-created objects when available, instead of blocking on an
3898 empty osc while others are not empty. If we must block, we block
3899 for the shortest possible period of time.
3904 Description: chmod on Lustre root is propagated to other clients
3905 Details : Re-validate root's dentry in ll_lookup_it to avoid having it
3906 invalid by the follow_mount time.
3909 Frequency : liblustre clients only
3911 Description: Race in 'instant cancel' lock handling could lead to such locks
3912 never to be granted in case of SMP MDS
3913 Details : Do not destroy not yet granted but cbpending locks in
3917 Frequency : replay/resend of open
3919 Description: non null lock assetion failure in mds_intent_policy
3920 Details : Trying to replay/resend lockless open requests resulted in
3921 mds_open() returning 0 with no lock. Now it sets a flag if
3922 a lock is going to be returned.
3924 Severity : enhancement
3926 Description: Checksum enhancements
3927 Details : New checksum enhancements allow for resending RPCs that failed
3930 Severity : enhancement
3932 Description: Tunables on number of dirty pages in cacche
3933 Details : Allow to set limit on number of dirty pages cached.
3938 Description: client crash on unmount - lock still has references
3939 Details : In some error handling cases it was possible to leak a lock
3940 reference on a client while accessing a file. This was not
3941 harmful to the client during operation, but would cause the
3942 client to crash when the filesystem is unmounted.
3945 Frequency : specific case, rare
3947 Description: ETXTBSY on mds though file not in use by client
3948 Details : ETXTBSY is no longer incorrectly returned when attempting to
3949 chmod or chown a directory that the user previously tried to
3950 execute or a currently-executing binary.
3953 Frequency : extremely rare except on liblustre-based clients
3955 Description: Lustre space not freed when files are deleted
3956 Details : Clean up open-unlinked files after client eviction. Previously
3957 the unlink was skipped and the files remained as orphans.
3962 Description: OST failure "would be an LBUG" in waiting_locks_callback()
3963 Details : In some cases it was possible to send a blocking callback to
3964 a client doing a glimpse, even though that client didn't get
3965 a lock granted. When the glimpse lock is cancelled on the OST
3966 the freed lock is left on the waiting list and corrupted the list.
3969 Frequency : all core dumps
3971 Description: Broke core dumps to lustre
3972 Details : Negative dentry may be unhashed if parent does not have UPDATE
3973 lock, but some callers, e.g. do_coredump, expect dentry to be
3974 hashed after successful create, hash it in ll_create_it.
3976 ------------------------------------------------------------------------------
3978 2006-09-13 Cluster File Systems, Inc. <info@clusterfs.com>
3980 * Support for kernels:
3981 2.6.9-42.0.2.EL (RHEL 4)
3982 2.6.5-7.276 (SLES 9)
3983 2.4.21-40.EL (RHEL 3)
3984 2.6.12.6 vanilla (kernel.org)
3988 Frequency : always on RHEL 3
3990 Description: Number of open files grows over time
3991 Details : The number of open files grows over time, whether or not
3992 Lustre is started. This was due to a filp leak introduced
3993 by one of our kernel patches.
3995 ------------------------------------------------------------------------------
3997 2006-08-20 Cluster File Systems, Inc. <info@clusterfs.com>
3999 * Support for kernels:
4000 2.6.9-42.EL (RHEL 4)
4001 2.6.5-7.267 (SLES 9)
4002 2.4.21-40.EL (RHEL 3)
4003 2.6.12.6 vanilla (kernel.org)
4008 Bugzilla : 5719, 9635, 9792, 9684
4009 Description: OST (or MDS) trips assertions in (re)connection under heavy load
4010 Details : If a server is under heavy load and cannot reply to new
4011 connection requests before the client resends the (re)connect,
4012 the connection handling code can behave badly if two service
4013 threads are concurrently handing separate (re)connections from
4014 the same client. Add better locking to the connection handling
4015 code, and ensure that only a single connection will be processed
4016 for a given client UUID, even if the lock is dropped.
4018 Severity : enhancement
4020 Description: add TCP zero-copy support to kernel
4021 Details : Add support to the kernel TCP stack to allow zero-copy bulk
4022 sends if the hardware supports scatter-gather and checksumming.
4023 This allows socklnd to do client-write and server-read more
4024 efficiently and reduce CPU utilization from skbuf copying.
4027 Frequency : only if NFS exporting from client
4029 Description: NULL pointer deref in ll_iocontrol() if chattr mknod file
4030 Details : If setting attributes on a file created under NFS that had
4031 never been opened it would be possible to oops the client
4032 if the file had no objects.
4036 Bugzilla : 9326, 10402, 10897
4037 Description: client crash in ptlrpcd_wake() thread when sending async RPC
4038 Details : It is possible that ptlrpcd_wake() dereferences a freed async
4039 RPC. In rare cases the ptlrpcd thread alread processed the RPC
4040 before ptlrpcd_wake() was called and the request was freed.
4043 Frequency : always for liblustre
4045 Description: liblustre client does MDS+OSTs setattr RPC for each write
4046 Details : When doing a write from a liblustre client, the client
4047 incorrectly issued an RPC to the MDS and each OST the file was
4048 striped over in order to update the timestamps. When writing
4049 with small chunks and many clients this could overwhelm the MDS
4050 with RPCs. In all cases it would slow down the write because
4051 these RPCs are unnecessary.
4053 Severity : enhancement
4055 Description: allow number of MDS service threads to be changed at module load
4056 Details : It is now possible to change the number of MDS service threads
4057 running. Adding "options mds mds_num_threads={N}" to the MDS's
4058 /etc/modprobe.conf will set the number of threads for the next
4059 time Lustre is restarted (assuming the "mds" module is also
4060 reloaded at that time). The default number of threads will
4061 stay the same, 32 for most systems.
4066 Description: OST crash if filesystem is unformatted or corrupt
4067 Details : If an OST is started on a device that has never been formatted
4068 or if the filesystem is corrupt and cannot even mount then the
4069 error handling cleanup routines would dereference a NULL pointer.
4074 Description: NULL pointer deref in llap_from_page.
4075 Details : get_cache_page_nowait can return a page with NULL (or otherwise
4076 incorrect) mapping if the page was truncated/reclaimed while it was
4077 searched for. Check for this condition and skip such pages when
4078 doing readahead. Introduce extra check to llap_from_page() to
4079 verify page->mapping->host is non-NULL (so page is not anonymous).
4082 Frequency : Sometimes when using sys_sendfile
4084 Description: "page not covered by a lock" warnings from ll_readpage
4085 Details : sendfile called ll_readpage without right page locks present.
4086 Now we introduced ll_file_sendfile that does necessary locking
4087 around call to generic_file_sendfile() much like we do in
4091 Frequency : with certain MDS communication failures at client mount time
4093 Description: NULL pointer deref after failed client mount
4094 Details : a client connection request may delayed by the network layer
4095 and not be sent until after the PTLRPC layer has timed out the
4096 request. If the client fails the mount immediately it will try
4097 to clean up before the network times out the request. Add a
4098 reference from the request import to the obd device and delay
4099 the cleanup until the network drops the request.
4102 Frequency : occasionally during client (re)connect
4104 Description: assertion failure during client (re)connect
4105 Details : processing a client connection request may be delayed by the
4106 client or server longer than the client connect timeout. This
4107 causes the client to resend the connection request. If the
4108 original connection request is replied in this interval, the
4109 client may trip an assertion failure in ptlrpc_connect_interpret()
4110 which thought it would be the only running connect process.
4113 Frequency : only with obd_echo servers and clients that are rebooted
4115 Description: kernel BUG accessing uninitialized data structure
4116 Details : When running an obd_echo server it did not start the ping_evictor
4117 thread, and when a client was evicted an uninitialized data
4118 structure was accessed. Start the ping_evictor in the RPC
4119 service startup instead of the OBD startup.
4121 Severity : enhancement
4122 Bugzilla : 10193 (patchless)
4123 Description: Remove dependency on various unexported kernel interfaces.
4124 Details : No longer need reparent_to_init, exit_mm, exit_files,
4125 sock_getsockopt, filemap_populate, FMODE_EXEC, put_filp.
4128 Frequency : rare (only users of deprecated and unsupported LDAP config)
4130 Description: write_conf for zeroconf mount queried LDAP incorrectly for client
4131 Details : LDAP apparently contains 'lustreName' attributes instead of
4132 'name'. A simple remapping of the name is sufficient.
4135 Frequency : rare (only with non-default dump_on_timeout debug enabled)
4137 Description: waiting_locks_callback trips kernel BUG if client is evicted
4138 Details : Running with the dump_on_timeout debug flag turned on makes
4139 it possible that the waiting_locks_callback() can try to dump
4140 the Lustre kernel debug logs from an interrupt handler. Defer
4141 this log dumping to the expired_lock_main() thread.
4143 Severity : enhancement
4145 Description: Support NFS exporting on 2.6 kernels.
4146 Details : Implement non-rawops metadata methods for NFS server to use without
4147 changing NFS server code.
4150 Frequency : very rare (synthetic metadata workload only)
4152 Description: two racing renames might cause an MDS thread to deadlock
4153 Details : Running the "racer" program may cause one MDS thread to rename
4154 a file from being the source of a rename to being the target of
4155 a rename at exactly the same time that another thread is doing
4156 so, and the second thread has already enqueued these locks after
4157 doing a lookup of the target and is trying to relock them in
4158 order. Ensure that we don't try to re-lock the same resource.
4161 Frequency : only very large systems with liblustre clients
4163 Description: slow eviction of liblustre clients with the "evict_by_nid" RPC
4164 Details : Use asynchronous set_info RPCs to send the "evict_by_nid" to
4165 all OSTs in parallel. This allows the eviction of stale liblustre
4166 clients to proceed much faster than if they were done in series,
4167 and also offers similar improvements for other set_info RPCs.
4172 Description: excessive CPU usage during initial read phase on client
4173 Details : During the initial read phase on a client, it would agressively
4174 retry readahead on the file, consuming too much CPU and impacting
4175 performance (since 1.4.5.8). Improve the readahead algorithm
4176 to avoid this, and also improve some other common cases (read
4177 of small files in particular, where "small" is files smaller than
4178 /proc/fs/lustre/llite/*/max_read_ahead_whole_mb, 2MB by default).
4183 Description: MDS crash when receiving packet with unknown intent.
4184 Details : Do not LBUG in unknown intent case, just return -EFAULT
4186 Severity : enhancement
4187 Bugzilla : 9293, 9385
4188 Description: MDS RPCs are serialised on client. This is unnecessary for some.
4189 Details : Do not serialize getattr (non-intent version) and statfs.
4192 Frequency : occasional, when OST network is overloaded/intermittent
4194 Description: client evicted by OST after bulk IO timeout
4195 Details : If a client sends a bulk IO request (read or write) the OST
4196 may evict the client if it is unresposive to its data GET/PUT
4197 request. This is incorrect if the network is overloaded (takes
4198 too long to transfer the RPC data) or dropped the OST GET/PUT
4199 request. There is no need to evict the client at all, since
4200 the pinger and/or lock callbacks will handle this, and the
4201 client can restart the bulk request.
4204 Frequency : Always when mmapping file with no objects
4206 Description: client crashes when mmapping file with no objects
4207 Details : Check that we actually have objects in a file before doing any
4208 operations on objects in ll_vm_open, ll_vm_close and
4214 Description: Request leak when working with deleted CWD
4215 Details : Introduce advanced request refcount tracking for requests
4216 referenced from lustre intent.
4218 Severity : Enhancement
4220 Description: Cache open file handles on client.
4221 Details : MDS now will return special lock along with openhandle, if
4222 requested and client is allowed to hold openhandle, even if unused,
4223 until such a lock is revoked. Helps NFS a lot, since NFS is opening
4224 closing files for every read/write openration.
4226 Severity : Enhancement
4228 Description: Cache open negative dentries on client when possible.
4229 Details : Guard negative dentries with UPDATE lock on parent dir, drop
4230 negative dentries on lock revocation.
4235 Description: Remounting a client read-only wasn't possible with a zconf mount
4236 Details : It wasn't possible to remount a client read-only with llmount.
4238 Severity : enhancement
4239 Description: Include MPICH 1.2.6 Lustre ADIO interface patch
4240 Details : In lustre/contrib/ or /usr/share/lustre in RPM a patch for
4241 MPICH is included to add Lustre-specific ADIO interfaces.
4242 This is based closely on the UFS ADIO layer and only differs
4243 in file creation, in order to allow the OST striping to be set.
4244 This is user-contributed code and not supported by CFS.
4249 Description: extended inode attributes (immutable, append-only) work improperly
4250 when 2.4 and 2.6 kernels are used on client/server or vice versa
4251 Details : Introduce kernel-independent values for these flags.
4253 Severity : enhancement
4256 Description: Allow fractional MB tunings for lustre in /proc/ filesystem.
4257 Details : Many of the /proc/ tunables can only be tuned at a megabyte
4258 granularity. Now, Fractional MB granularity is be supported,
4259 this is very useful for low memory system.
4261 Severity : enhancement
4263 Description: Getattr by fid
4264 Details : Getting a file attributes by its fid, obtaining UPDATE|LOOKUP
4265 locks, avoids extra getattr rpc requests to MDS, allows '/' to
4266 have locks and avoids getattr rpc requests for it on every stat.
4269 Frequency : Always, for filesystems larger than 2TB
4271 Description: ldiskfs crash at mount for filesystem larger than 2TB with mballoc
4272 Details : Kenrel kmalloc limits allocations to 128kB and this prevents
4273 filesystems larger than 2TB to be mounted with mballoc enabled.
4276 Frequency : Always, for 32-bit kernel without CONFIG_LBD and filesystem > 2TB
4278 Description: filesystem corruption for non-standard kernels and very large OSTs
4279 Details : If a 32-bit kernel is compiled without CONFIG_LBD enabled and a
4280 filesystems larger than 2TB is mounted then the kernel will
4281 silently corrupt the start of the filesystem. CONFIG_LBD is
4282 enabled for all CFS-supported kernels, but the possibility of
4283 this happening with a modified kernel config exists.
4285 Severity : enhancement
4287 Description: add client O_DIRECT support for 2.6 kernels
4288 Details : It is now possible to do O_DIRECT reads and writes to files
4289 in the Lustre client mountpoint on 2.6 kernel clients.
4291 Severity : enhancement
4293 Description: parallel glimpse, setattr, statfs, punch, destroy requests
4294 Details : Sends glimpse, setattr, statfs, punch, destroy requests to OSTs in
4295 parallel, not waiting for response from every OST before sending
4296 a rpc to the next OST.
4301 Description: setattr vs write race when updating file timestamps
4302 Details : Client processes that update a file timestamp into the past
4303 right after writing to the file (e.g. tar) it is possible that
4304 the updated file modification time can be reset to the current
4305 time due to a race between processing the setattr and write RPC.
4307 Severity : enhancement
4309 Description: Bring 'lfs find' closer in line with regular Linux find.
4310 Details : lfs find util supports -atime, -mtime, -ctime, -maxdepth, -print,
4311 -print0 options and obtains all the needed info through the lustre
4314 Severity : enhancement
4316 Description: support up to 1024 configured devices on one node
4317 Details : change obd_dev array from statically allocated to dynamically
4318 allocated structs as they are first used to reduce memory usage
4323 Description: Flush dirty partially truncated pages during truncate
4324 Details : Immediatelly flush partially truncated pages in filter_setattr,
4325 this way we completely avoid having any pages in page cache on OST
4326 and can retire ugly workarounds during writes to flush such pages.
4331 Description: i_sem vs transaction deadlock in mds_obd_destroy during unlink.
4332 Details : protect inode from truncation within vfs_unlink() context
4333 just take a reference before calling vfs_unlink() and release it
4334 when parent's i_sem is free.
4337 Frequency : always, if extents are used on OSTs
4339 Description: index ei_leaf_hi (48-bit extension) is not zeroed in extent index
4340 Details : OSTs using the extents format would not zero the high 16 bits of
4341 the index physical block number. This is not a problem for any
4342 OST filesystems smaller than 16TB, and no kernels support ext3
4343 filesystems larger than 16TB yet. This is fixed in 1.4.7 (all
4344 new/modified files) and can be fixed for existing filesystems
4345 with e2fsprogs-1.39-cfs1.
4350 Description: import connection selection may be incorrect if timer wraps
4351 Details : Using a 32-bit jiffies timer with HZ=1000 may cause backup
4352 import connections to be ignored if the 32-bit jiffies counter
4353 wraps. Use a 64-bit jiffies counter.
4356 Frequency : during server recovery
4358 Description: crash after server is denying duplicate export
4359 Details : If clients are resending connect requests to the server, the
4360 server refuses to allow a client to connect multiple times.
4361 Fixed a bug in the handling of this case.
4364 Frequency : very large clusters immediately after boot
4366 Description: LNET request buffers exhausted under heavy short-term load
4367 Details : If a large number of client requests are generated on a service
4368 that has previously never seen so many requests it is possible
4369 that the request buffer growth cannot keep up with the spike in
4370 demand. Instead of dropping incoming requests, they are held in
4371 the LND until the RPC service can accept more requests.
4374 Frequency : Sometimes during replay
4376 Description: Assertion failure in ll_local_open after replay.
4377 Details : If replay happened on an open request reply before we were able
4378 to set replay handler, reply will become not swabbed tripping the
4379 assertion in ll_local_open. Now we set the handler right after
4380 recognising of open request
4383 Frequency : very rare
4385 Description: kernel reports "badness in vsnprintf"
4386 Details : Reading from the "recovery_status" /proc file in small chunks
4387 may cause a negative length in lprocfs_obd_rd_recovery_status()
4388 call to vsnprintf() (which is otherwise harmless). Exit early
4389 if there is no more space in the output buffer.
4391 Severity : enhancement
4393 Description: clear OBD RPC statistics by writing to them
4394 Details : It is now possible to clear the OBD RPC statistics by writing
4395 to the "stats" file.
4400 Description: Client mtime is not the same on different clients after utimes
4401 Details : In some cases, the client was using the utimes() syscall on
4402 a file cached on another node. The clients now validate the
4403 ctime from the MDS + OSTs to determine which one is right.
4408 Description: Inability to activate failout mode
4409 Details : lconf script incorrectly assumed that in python string's numeric
4410 value is used in comparisons.
4413 Frequency : always with multiple stripes per file
4415 Description: Inefficient object allocation for mutli-stripe files
4416 Details : When selecting which OSTs to stripe files over, for files with
4417 a stripe count that divides evenly into the number of OSTs,
4418 the MDS is always picking the same starting OST for each file.
4419 Return the OST selection heuristic to the original design.
4424 Description: mount failures may take full timeout to return an error
4425 Details : Under some heavy load conditions it is possible that a
4426 failed mount can wait for the full obd_timeout interval,
4427 possibly several minutes, before reporting an error.
4428 Instead return an error as soon as the status is known.
4430 ------------------------------------------------------------------------------
4432 2006-02-14 Cluster File Systems, Inc. <info@clusterfs.com>
4434 * WIRE PROTOCOL CHANGE. This version of Lustre networking WILL NOT
4435 INTEROPERATE with older versions automatically. Please read the
4436 user documentation before upgrading any part of a live system.
4437 * WARNING: Lustre networking configuration changes are required with
4438 this release. See https://bugzilla.clusterfs.com/show_bug.cgi?id=10052
4441 * Support for kernels:
4442 2.6.9-22.0.2.EL (RHEL 4)
4443 2.6.5-7.244 (SLES 9)
4444 2.6.12.6 vanilla (kernel.org)
4447 Severity : enhancement
4448 Bugzilla : 7981/8208
4449 Description: Introduced Lustre Networking (LNET)
4450 Details : LNET is new networking infrastructure for Lustre, it includes
4451 a reorganized network configuration mode (see the user
4452 documentation for full details) as well as support for routing
4453 between different network fabrics. Lustre Networking Devices
4454 (LNDs) for the supported network fabrics have also been
4455 created for this new infrastructure.
4457 Severity : enhancement
4458 Description: Introduced Access control lists
4459 Details : clients can set ACLs on files and directories in order to have
4460 more fine-grained permissions than the standard Unix UGO+RWX.
4461 The MDS must be started with the "-o acl" mount option.
4463 Severity : enhancement
4464 Description: Introduced filesystem quotas
4465 Details : Administrators may now establish per-user quotas on the
4468 Severity : enhancement
4470 Description: Configuration change for the XT3
4471 The PTLLND is now used to run Lustre over Portals on the XT3
4472 The configure option(s) --with-cray-portals are no longer used.
4473 Rather --with-portals=<path-to-portals-includes> is used to
4474 enable building on the XT3. In addition to enable XT3 specific
4475 features the option --enable-cray-xt3 must be used.
4480 Description: Running on many-way SMP OSTs can trigger oops in llcd_send()
4481 Details : A race between allocating a new llcd and re-getting the llcd_lock
4482 allowed another thread to grab newly-allocated llcd.
4484 Severity : enhancement
4486 Description: 2.6 OST async journal commit and locking fix to improve performance
4487 Details : The filter_direct_io()+filter_commitrw_write() journal commits for
4488 2.6 kernels are now async as they already were in 2.4 kernels so
4489 that they can commit concurrently with the network bulk transfer.
4490 For block-allocated files the filter allocation semaphore is held
4491 to avoid filesystem fragmentation during allocation. BKL lock
4492 removed for 2.6 xattr operations where it is no longer needed.
4497 Description: lconf incorrectly determined whether two IP networks could talk
4498 Details : In some more complicated routing and multiple-network
4499 configurations, lconf will avoid trying to make a network
4500 connection to a disjoint part of the IP space. It was doing the
4501 math incorrectly for one set of cases.
4506 Description: Fix for potential infinite loop processing records in an llog.
4507 Details : If an llog record is corrupted/zeroed, it is possible to loop
4508 forever in llog_process(). Validate the llog record length
4509 and skip the remainder of the block on error.
4512 Frequency : occasional (liblustre only)
4514 Description: liblustre could not open files whose last component is a symlink
4515 Details : sysio_path_walk() would incorrectly pass the open intent to
4516 intermediate path components.
4519 Frequency : rare (liblustre only with non-standard tuning)
4520 Bugzilla : 7201 (7350)
4521 Description: Tuning the MDC DLM LRU size to zero triggers client LASSERT
4522 Details : llu_lookup_finish_locks() tries to set lock data on a lock
4523 after it has been released, only do this for referenced locks
4525 Severity : enhancement
4527 Description: specifying an (invalid) directory default stripe_size of -1
4528 would reset the directory default striping
4529 Details : stripe_size -1 was used internally to signal directory stripe
4530 removal, now use "all default" to signal dir stripe removal
4531 as a directory striping of "all default" is not useful
4534 Frequency : common for large clusters running liblustre clients
4536 Description: doing an ls when liblustre clients are running is slow
4537 Details : sending a glimpse AST to a liblustre client waits for every AST
4538 to time out, as liblustre clients will not respond. Since they
4539 cannot cache data we refresh the OST lock LVB from disk instead.
4541 Severity : enhancement
4543 Description: doing an ls at the same time as file IO can be slow
4544 Details : enqueue and other "small" requests can be blocked behind many
4545 large IO requests. Create a new OST IO portal for non-IO
4546 requests so they can be processed faster.
4549 Frequency : rare (only HPUX clients mounting unsupported re-exported NFS vol)
4551 Description: an HPUX NFS client would get -EACCESS when ftruncate()ing a newly
4552 created file with mode 000
4553 Details : the Linux NFS server relies on an MDS_OPEN_OWNEROVERRIDE hack to
4554 allow an ftruncate() as a non-root user to a file with mode 000.
4555 Lustre now respects this flag to disable mode checks when
4556 truncating a file owned by the user
4559 Frequency : liblustre-only, when liblustre client dies unexpectedly or becomes
4562 Description: Revoking locks from clients that went dead or catatonic might take
4564 Details : New lock flags FL_CANCEL_ON_BLOCK used by liblustre makes
4565 cancellation of such locks instant on servers without waiting for
4566 any reply from clients. Clients drops these locks when cancel
4567 notification from server is received without replying.
4570 Frequency : liblustre-only, when liblustre client dies or becomes busy
4572 Description: Doing ls on Linux clients can take a long time with active
4574 Details : Liblustre client cannot handle ASTs in timely manner, so avoid
4575 granting such locks to it in the first place if possible. Locks
4576 are taken by proxy on the OST during the read or write and
4577 dropped immediately afterward. Add connect flags handling, do
4578 not grant locks to liblustre clients for glimpse ASTs.
4580 Severity : enhancement
4582 Description: Improve read-ahead algorithm to avoid excessive IO for random reads
4583 Details : Existing read-ahead algorithm is tuned for the case of streamlined
4584 sequential reads and behaves badly with applications doing random
4585 reads. Improve it by reading ahead at least read region, and
4586 avoiding excessive large RPC for small reads.
4588 Severity : enhancement
4590 Description: Creating more than 1000 files for a single job may cause a load
4591 imbalance on the OSTs if there are also a large number of OSTs.
4592 Details : qos_prep_create() uses an OST index reseed value that is an
4593 even multiple of the number of available OSTs so that if the
4594 reseed happens in the middle of the object allocation it will
4595 still utilize the OSTs as uniformly as possible.
4600 Description: OST or MDS may oops in ping_evictor_main()
4601 Details : ping_evictor_main() drops obd_dev_lock if deleting a stale export
4602 but doesn't restart at beginning of obd_exports_timed list
4605 Severity : enhancement
4607 Description: improve by-nid export eviction on the MDS and OST
4608 Details : allow multiple exports with the same NID to be evicted at one
4609 time without re-searching the exports list.
4612 Frequency : rare, only with supplementary groups enabled on SMP 2.6 kernels
4614 Description: MDS may oops in groups_free()
4615 Details : in rare race conditions a newly allocated group_info struct is
4616 freed again, and this can be NULL. The 2.4 compatibility code
4617 for groups_free() checked for a NULL pointer, but 2.6 did not.
4620 Frequency : common for liblustre clients doing little filesystem IO
4621 Bugzilla : 9352, 7313
4622 Description: server may evict liblustre clients accessing contended locks
4623 Details : if a client is granted a lock or receives a completion AST
4624 with a blocking AST already set it would not reply to the AST
4625 for LDLM_FL_CANCEL_ON_BLOCK locks. It now replies to such ASTs.
4628 Frequency : lfs setstripe, only systems with more than 160 OSTs
4630 Description: unable to set striping with a starting offset beyond OST 160
4631 Details : llapi_create_file() incorrectly limited the starting stripe
4632 index to the maximum single-file stripe count.
4635 Frequency : LDAP users only
4637 Description: lconf did not handle in-kernel recovery with LDAP properly
4638 Details : lconf/LustreDB get_refs() is searching the wrong namespace
4640 Severity : enhancement
4642 Description: bind OST threads to NUMA nodes to improve performance
4643 Details : all OST threads are uniformly bound to CPUs on a single NUMA
4644 node and do their allocations there to localize memory access
4646 Severity : enhancement
4648 Description: llmount can determine client NID directly from Myrinet (GM)
4649 Details : the client NID code from gmnalnid was moved directly into
4650 llmount, removing the need to use this or specifying the
4651 client NID explicitly when mounting GM clients with zeroconf
4654 Frequency : if client is started with down MDS
4656 Description: if client is started with down MDS mount hangs in ptlrpc_queue_wait
4657 Details : Having an LWI_INTR() wait event (interruptible, but no timeout)
4658 will wait indefinitely in ptlrpc_queue_wait->l_wait_event() after
4659 ptlrpc_import_delayed_req() because we didn't check if the
4660 request was interrupted, and we also didn't break out of the
4661 event loop if there was no timeout
4666 Description: data loss during non-page-aligned writes to a single file from
4667 both multiple nodes and multiple threads on one node at same time
4668 Details : updates to KMS and lsm weren't protected by common lock. Resulting
4669 inconsistency led to false short-reads, that were cached and later
4670 used by ->prepare_write() to fill in partially written page,
4671 leading to data loss.
4674 Frequency : always, if lconf --abort_recovery used
4676 Description: lconf --abort_recovery fails with 'Operation not supported'
4677 Details : lconf was attempting to abort recovery on the MDT device and not
4680 Severity : enhancement
4682 Description: remove cleanup logs
4683 Details : replace lconf-generated cleanup logs with lustre internal
4684 cleanup routines. Eliminates the need for client-cleanup and
4687 Severity : enhancement
4689 Description: add support for EAs (user and system) on lustre filesystems
4690 Details : it is now possible to store extended attributes in the Lustre
4691 client filesystem, and with the user_xattr mount option it
4692 is possible to allow users to store EAs on their files also
4694 Severity : enhancement
4696 Description: Add possibility (config option) to show minimal available OST free
4698 Details : When compiled with --enable-mindf configure option, statfs(2)
4699 (and so, df) will return least minimal free space available from
4700 all OSTs as amount of free space on FS, instead of summary of
4701 free spaces of all OSTs.
4703 Severity : enhancement
4705 Description: do not expand extent locks acquired on OST-side
4706 Details : Modify ldlm_extent_policy() to not expand local locks, acquired
4707 by server: they are not cached anyway.
4710 Frequency : when mmap is used/binaries executed from Lustre
4712 Description: Unmmap pages before throwing them away from read cache.
4713 Details : llap_shrink cache now attempts to unmap pages before discarding
4714 them (if unmapping failed - do not discard). SLES9 kernel has
4715 extra checks that trigger if this unmapping is not done first.
4720 Description: lconf didn't resolve symlinks before checking to see whether a
4721 given mountpoint was already in use
4724 Frequency : when migrating failover services
4725 Bugzilla : 6395, 9514
4726 Description: When migrating a subset of services from a node (e.g. failback
4727 from a failover service node) the remaining services would
4728 time out and evict clients.
4729 Details : lconf --force (implied by --failover) sets the global obd_timeout
4730 to 5 seconds in order to quickly disconnect, but this caused
4731 other RPCs to time out too quickly. Do not change the global
4732 obd_timeout for force cleanup, only set it for DISCONNECT RPCs.
4734 Severity : enhancement
4735 Frequency : if MDS is started with down OST
4736 Bugzilla : 9439,5706
4737 Description: Allow startup/shutdown of an MDS without depending on the
4738 availability of the OSTs.
4739 Details : Asynchronously call mds_lov_synchronize during MDS startup.
4740 Add appropriate locking and lov-osc refcounts for safe
4741 cleaning. Add osc abort_inflight calls in case the
4742 synchronize never started.
4745 Frequency : occasional (Cray XT3 only)
4747 Description: root not authorized to access files in CRAY_PORTALS environment
4748 Details : The client process capabilities were not honoured on the MDS in
4749 a CRAY_PORTALS/CRAY_XT3 environment. If the file had previously
4750 been accessed by an authorized user then root was able to access
4751 the file on the local client also. The root user capabilities
4752 are now allowed on the MDS, as this environment has secure UID.
4755 Frequency : occasional
4757 Description: ldiskfs "too long searching" message happens too often
4758 Details : A debugging message (otherwise harmless) prints too often on
4759 the OST console. This has been reduced to only happen when
4760 there are fragmentation problems on the filesystem.
4765 Description: Division by zero in statfs when all OSCs are inactive
4766 Details : lov_get_stripecnt() returns zero due to incorrect order of checks,
4767 lov_statfs divides by value returned by lov_get_stripecnt().
4771 Bugzilla : 9489, 3273
4772 Description: First write from each client to each OST was only 4kB in size,
4773 to initialize client writeback cache, which caused sub-optimal
4774 RPCs and poor layout on disk for the first writen file.
4775 Details : Clients now request an initial cache grant at (re)connect time
4776 and so that they can start streaming writes to the cache right
4777 away and always do full-sized RPCs if there is enough data.
4778 If the OST is rebooted the client also re-establishes its grant
4779 so that client cached writes will be honoured under the grant.
4784 Description: Slow ls (and stat(2) syscall) on files residing on IO-loaded OSTs
4785 Details : Now I/O RPCs go to different portal number and (presumably) fast
4786 lock requests (and glimses) and other RPCs get their own service
4787 threads pool that should be able to service those RPCs
4790 Severity : enhancement
4792 Description: Ability to exchange lustre version between client and servers and
4793 issue warnings at client side if client is too old. Also for
4794 liblustre clients there is ability to refuse connection of too old
4796 Details : New 'version' field is added to connect data structure that is
4797 filled with version info. That info is later checked by server and
4801 Frequency : rare, liblustre only.
4802 Bugzilla : 9296, 9581
4803 Description: Two simultaneous writes from liblustre at offset within same page
4804 might proceed at the same time overwriting eachother with stale
4806 Details : I/O lock withing llu_file_prwv was released too early, before data
4807 actually was hitting the wire. Extended lock-holding time until
4808 server acknowledges receiving data.
4811 Frequency : extremely rare. Never observed in practice.
4813 Description: avoid generating lustre_handle cookie of 0.
4814 Details : class_handle_hash() generates handle cookies by incrementing
4815 global counter, and can hit 0 occasionaly (this is unlikely, but
4816 not impossible, because initial value of cookie counter is
4817 selected randonly). Value of 0 is used as a sentinel meaning
4818 "unassigned handle" --- avoid it. Also coalesce two critical
4819 sections in this function into one.
4821 Severity : enhancement
4823 Description: allow liblustre clients to delegate truncate locking to OST
4824 Details : To avoid overhead of locking, liblustre client instructs OST to
4825 take extent lock in ost_punch() on client's behalf. New connection
4826 flag is added to handle backward compatibility.
4828 Severity : enhancement
4829 Bugzilla : 4928, 7341, 9758
4830 Description: allow number of OST service threads to be specified
4831 Details : a module parameter allows the number of OST service threads
4832 to be specified via "options ost ost_num_threads={N}" in the
4833 OSS's /etc/modules.conf or /etc/modprobe.conf.
4837 Bugzilla : 6146, 9635, 9895
4838 Description: servers crash with bad pointer in target_handle_connect()
4839 Details : In rare cases when a client is reconnecting it was possible that
4840 the connection request was the last reference for that export.
4841 We would temporarily drop the export reference and get a new
4842 one, but this may have been the last reference and the export
4843 was just destroyed. Get new reference before dropping old one.
4845 Severity : enhancement
4846 Frequency : if client is started with failover MDS
4848 Description: Allow multiple MDS hostnames in the mount command
4849 Details : Try to read the configuration from all specified MDS
4850 hostnames during a client mount in case the "primary"
4853 Severity : enhancement
4855 Description: Stop sending data to evicted clients as soon as possible.
4856 Details : Check if the client we are about to send or are sending data to
4857 was evicted already. (Check is done every second of waiting,
4858 for which l_wait_event interface was extended to allow checking
4859 of exit condition at specified intervals).
4862 Frequency : rare, normally only when NFS exporting is done from client
4864 Description: 'bad disk LOV MAGIC: 0x00000000' error when chown'ing files
4866 Details : Make mds_get_md() recognise empty md case and set lmm size to 0.
4869 Frequency : always, if srand() is called before liblustre initialization
4871 Description: Liblustre uses system PRNG disturbing its usage by user application
4872 Details : Introduce internal to lustre fast and high-quality PRNG for
4873 lustre usage and make liblustre and some other places in generic
4874 lustre code to use it.
4876 Severity : enhancement
4877 Bugzilla : 9477, 9557, 9870
4878 Description: Verify that the MDS configuration logs are updated when xml is
4879 Details : Check if the .xml configuration logs are newer than the config
4880 logs stored on the MDS and report an error if this is the case.
4881 Request --write-conf, or allow starting with --old_conf.
4883 Severity : enhancement
4885 Description: Handle symlinks in the path when checking if Lustre is mounted.
4886 Details : Resolve intermediate symlinks when checking if a client has
4887 mounted a filesystem to avoid duplicate client mounts.
4892 Description: lconf can hit an error exception but still return success.
4893 Details : The lconf command catches the Command error exception at the top
4894 level script context and will exit with the associated exit
4895 status, but doesn't ensure that this exit status is non-zero.
4900 Description: failure of ptlrpc thread startup can cause oops
4901 Details : Starting a ptlrpc service thread can fail if there are a large
4902 number of threads or the server memory is very fragmented.
4903 Handle this without oopsing.
4906 Frequency : always, only if liblustre and non-default acceptor port was used
4908 Description: liblustre cannot connect to servers with non-default acceptor port
4909 Details : tcpnal_set_default_params() was not called and was therefore
4910 ignoring the environment varaible TCPNAL_PORT, as well as other
4911 TCPNAL_ environment variables
4916 Description: two objects could be created on the same OST for a single file
4917 Details : If an OST is down, in some cases it was possible to create two
4918 objects on a single OST for a single file. No problems other
4919 than potential performance impact and spurious error messages.
4923 Bugzilla : 5681, 9562
4924 Description: Client may oops in ll_unhash_aliases
4925 Details : Client dcache may become inconsistent in race condition.
4926 In some cases "getcwd" can fail if the current directory is
4932 Description: Inode refcounting problems in NFS export code
4933 Details : link_raw functions used to call d_instantiate without obtaining
4934 extra inode reference first.
4938 Bugzilla : 9942, 9903
4939 Description: Referencing freed requests leading to crash, memleaks with NFS.
4940 Details : We used to require that call to ll_revalidate_it was always
4941 followed by ll_lookup_it. Also with revalidate_special() it is
4942 possible to call ll_revalidate_it() twice for the same dentry
4943 even if first occurence returned success. This fix changes semantic
4944 between DISP_ENQ_COMPLETE disposition flag to mean there is extra
4945 reference on a request referred from the intent.
4946 ll_intent_release() then releases such a request.
4949 Frequency : rare, normally benchmark loads only
4951 Description: unlinked inodes were kept in memory on the client
4952 Details : If a client is repeatedly creating and unlinking files it
4953 can accumulate a lot of stale inodes in the inode slab cache.
4954 If there is no other client load running this can cause the
4955 client node to run out of memory. Instead flush old inodes
4956 from client cache that have the same inode number as a new inode.
4959 Frequency : SLES9 2.6.5 kernel and long filenames only
4960 Bugzilla : 9969, 10379
4961 Description: utime reports stale NFS file handle
4962 Details : SLES9 uses out-of-dentry names in some cases, which confused
4963 the lustre dentry revalidation. Change it to always use the
4967 Frequency : rare, unless heavy write-truncate concurrency is continuous
4968 Bugzilla : 4180, 6984, 7171, 9963, 9331
4969 Description: OST becomes very slow and/or deadlocked during object unlink
4970 Details : filter_destroy() was holding onto the parent directory lock
4971 while truncating+unlinking objects. For very large objects this
4972 may block other threads for a long time and slow overall OST
4973 responsiveness. It may also be possible to get a lock ordering
4974 deadlock in this case, or run out of journal credits because of
4975 the combined truncate+unlink. Solution is to do object truncate
4976 first in one transaction without parent lock, and then do the
4977 final unlink in a new transaction with the parent lock. This
4978 reduces the lock hold time dramatically.
4981 Frequency : rare, 2.4 kernels only
4983 Description: MDS or OST cleanup may trip kernel BUG when dropping kernel lock
4984 Details : mds_cleanup() and filter_cleanup() need to drop the kernel lock
4985 before unmounting their filesystem in order to avoid deadlock.
4986 The kernel_locked() function in 2.4 kernels only checks whether
4987 the kernel lock is held, not whether it is this process that is
4988 holding it as 2.6 kernels do.
4993 Description: MDS or OST may oops/LBUG if a client is connecting multiple times
4994 Details : The client ptlrpc code may be trying to reconnect to a down
4995 server before a previous connection attempt has timed out.
4996 Increase the reconnect interval to be longer than the connection
4997 timeout interval to avoid sending duplicate connections to
5001 Frequency : echo_client brw_test command
5003 Description: fix echo_client to work with OST preallocated code
5004 Details : OST preallocation code (5137) didn't take echo_client IO path
5005 into account: echo_client calls filter methods outside of any
5006 OST thread and, hence, there is no per-thread preallocated
5007 pages and buffers to use. Solution: hijack pga pages for IO. As
5008 a byproduct, this avoids unnecessary data copying.
5012 Bugzilla : 3555, 5962, 6025, 6155, 6296, 9574
5013 Description: Client can oops in mdc_commit_close() after open replay
5014 Details : It was possible for the MDS to return an open request with no
5015 transaction number in mds_finish_transno() if the client was
5016 evicted, but without actually returning an error. Clients
5017 would later try to replay that open and may trip an assertion
5018 Simplify the client close codepath, and always return an error
5019 from the MDS in case the open is not successful.
5022 Frequency : rare, 2.6 OSTs only
5024 Description: OST may deadlock under high load on fragmented files
5025 Details : If there was a heavy load and highly-fragmented OST filesystems
5026 it was possible to have all the OST threads deadlock waiting on
5027 allocation of biovecs, because the biovecs were not released
5028 until the entire RPC IO was completed. Instead, release biovecs
5029 as soon as they are complete to ensure forward IO progress.
5031 Severity : enhancement
5033 Description: Support for specifying external journal device at mount
5034 Details : If an OST or MDS device is formatted with an external journal
5035 device, this device major/minor is stored in the ext3 superblock
5036 and may not be valid for failover. Allow detecting and
5037 specifying the external journal at mount time.
5042 Description: Mounting an MDS with pending unlinked files may cause oops
5043 Details : target_finish_recovery() calls mds_postrecov() which returned
5044 the number of orphans unlinked. mds_lov_connect->mds_postsetup()
5045 considers this an error and immediately begins cleaning up the
5046 lov, just after starting the mds_lov process
5048 Severity : enhancement
5050 Description: Implement 'lfs df' to report actual free space on per-OST basis
5051 Details : Add sub-command 'df' on 'lfs' to report the disk space usage of
5052 MDS/OSDs. Usage: lfs df [-i][-h]. Command Options: '-i' to report
5053 usage of objects; '-h' to report in human readable format.
5055 ------------------------------------------------------------------------------
5057 2005-08-26 Cluster File Systems, Inc. <info@clusterfs.com>
5064 Description: Mounting an ldiskfs file system with mballoc may crash OST node.
5065 Details : ldiskfs mballoc code may reference an uninitialized buddy struct
5066 at startup during orphan unlinking. Instead, skip buddy update
5067 before setup, as it will be regenerated after recovery is complete.
5072 Description: If an OST is inactive, its locks might reference stale inodes.
5073 Details : lov_change_cbdata() must iterate over all namespaces, even if
5074 they are inactive to clear inode references from the lock.
5076 Severity : enhancement
5077 Frequency : occasional, if non-standard max_dirty_mb used
5079 Description: Client will block write RPCs if not enough grant
5080 Details : If a client has max_dirty_mb smaller than max_rpcs_in_flight,
5081 then the client will block writes while waiting for another RPC
5082 to complete instead of consuming its dirty limit. With change
5083 we get improved performance when max_dirty_mb is small.
5085 Severity : enhancement
5086 Bugzilla : 3389, 6253
5087 Description: Add support for supplementary groups on the MDS.
5088 Details : The MDS has an upcall /proc/fs/lustre/mds/{mds}/group_upcall
5089 (set to /usr/sbin/l_getgroups if enabled) which will do MDS-side
5090 lookups for user supplementary groups into a cache.
5094 Description: O_CREAT|O_EXCL open flags in liblustre always return -EEXIST
5095 Details : Make libsysio to not enforce O_EXCL by clearing the flag,
5096 for liblustre O_EXCL is enforced by MDS.
5100 Description: readdir never returns NULL in liblustre.
5101 Details : Corrected llu_iop_getdirentries logic, to return offset of next
5102 dentry in struct dirent.
5106 Frequency : liblustre only, depends on application IO pattern
5107 Description: liblustre clients evicted if not contacting servers
5108 Details : Don't put liblustre clients into the ping_evictor list, so
5109 they will not be evicted by the pinger ever.
5111 Severity : enhancement
5113 Description: Add ability to evict clients by NID from MDS.
5114 Details : By echoing "nid:$NID" string into
5115 /proc/fs/lustre/mds/.../evict_client client with nid that equals to
5116 $NID would be instantly evicted from this MDS and from all active
5117 OSTs connected to it.
5121 Description: Do not query file size twice, somewhat slowing stat(2) calls.
5122 Details : lookup_it_finish() used to query file size from OSTs that was not
5127 Description: service threads change working directory to that of init
5128 Details : Starting lustre service threads may pin the working directory
5129 of the parent thread, making that filesystem busy. Threads
5130 now change to the working directory of init to avoid this.
5134 Frequency : during shutdown only
5135 Description: shutdown with a failed MDS or OST can cause unmount to hang
5136 Details : Don't resend DISCONNECT messages in ptlrpc_disconnect_import()
5141 Frequency : 2.6 only
5142 Description: chmod/chown may include an extra supplementary group
5143 Details : ll{,u}_mdc_pack_op_data() does not properly initialize the
5144 supplementary group and if none is specified this is used.
5147 Bugzilla : 5479 (6816)
5149 Description: Racing open + rm can assert client in mdc_set_open_replay_data()
5150 Details : If lookup is in progress on a file that is unlinked we might try
5151 to revalidate the inode and fail in revalidate after lookup is
5152 complete and ll_file_open() enqueues the open again but
5153 it_open_error() was not checking DISP_OPEN_OPEN errors correctly.
5156 Frequency : always, if lconf --abort_recovery used
5158 Description: lconf --abort_recovery fails with 'Operation not supported'
5159 Details : lconf was attempting to abort recovery on the MDT device and not
5162 ------------------------------------------------------------------------------
5164 2005-08-08 Cluster File Systems, Inc. <info@clusterfs.com>
5169 Frequency : rare (only unsupported configurations with a node running as an
5171 Bugzilla : 6514, 5137
5172 Description: Mounting a Lustre file system on a node running as an OST could
5174 Details : OSTs now preallocates memory needed to write out data at
5175 startup, instead of when needed, to avoid having to
5176 allocate memory in possibly low memory situations.
5177 Specifically, if the file system is mounted on on OST,
5178 memory pressure could force it to try to write out data,
5179 which it needed to allocate memory to do. Due to the low
5180 memory, it would be unable to do so and the node would
5181 become unresponsive.
5183 Severity : enhancement
5185 Description: Addition of lconf --service command line option
5186 Details : lconf now accepts a '--service <arg>' option, which is
5187 shorthand for 'lconf --group <arg> --select <arg>=<hostname>'
5189 Severity : enhancement
5191 Description: Failover mode is now the default for OSTs.
5192 Details : By default, OSTs will now run in failover mode. To return to
5193 the old behaviour, add '--failout' to the lmc line for OSTs.
5195 Severity : enhancement
5197 Description: Health checks are now provided for MDS and OSTs
5198 Details : Additional detailed health check information on MSD and OSTs
5199 is now provided through the procfs health_check value.
5202 Frequency : occasional, depends on IO load
5204 Description: Disk fragmentation on the OSTs could eventually cause slowdowns
5205 after numerous create/delete cycles
5206 Details : The ext3 inode allocation policy would not allocate new inodes
5207 very well on the OSTs because there are no new directories
5208 being created. Instead we look for groups with free space if
5209 the parent directories are nearly full.
5214 Description: Network or server problems during mount may cause partially
5215 mounted clients instead of returning an error.
5216 Details : The config llog parsing code may overwrite the error return
5217 code during mount error handling, returning success instead
5223 Description: MDS can fail to allocate large reply buffers
5224 Details : After long uptimes the MDS can fail to allocate large reply
5225 buffers (e.g. zconf client mount config records) due to memory
5226 fragmentation or consumption by the buffer cache. Preallocate
5227 some large reply buffers so that these replies can be sent even
5228 under memory pressure.
5232 Frequency : rare (liblustre)
5233 Description: fsx running with liblustre complained that using truncate() to
5234 extend the file doesn't work. This patch corrects that issue.
5235 Details : This is the liblustre equivalent of the fix for bug 6196. Fixes
5236 ATTR_SIZE and lsm use in llu_setattr_raw.
5240 Frequency : rare, only 2.6 kernels
5241 Description: Unusual file access patterns on the MDS may result in inode
5242 data being lost in very rare circumstances.
5243 Details : Bad interaction between the ea-in-inode patch and the "no-read"
5244 code in the 2.6 kernel caused the inode and/or EA data not to
5245 be read from disk, causing single-file corruption.
5249 Frequency : rare, only 2.6 filesystems using extents
5250 Description: Heavy concurrent write and delete load may cause data corruption.
5251 Details : It was possible under high-load situations to have an extent
5252 metadata block in the block device cache from a just-unlinked
5253 file overwrite a newly-allocated data block. We now unmap any
5254 metadata buffers that alias just-allocated data blocks.
5258 Frequency : filesystems with default stripe_count larger than 77
5259 Description: lconf+mke2fs fail when formatting filesystem with > 77 stripes
5260 Details : lconf specifies an inode size of 4096 bytes when the default
5261 stripe_count is larger than 77. This conflicts with the default
5262 inode density of 1 per 4096 bytes. Allocate smaller inodes in
5263 this case to avoid pinning too much memory for large EAs.
5265 ------------------------------------------------------------------------------
5267 2005-07-07 Cluster File Systems, Inc. <info@clusterfs.com>
5272 Frequency : rare (extremely heavy IO load with hundreds of clients)
5274 Description: Client is evicted, gets IO error writing to file
5275 Details : lock ordering changes for bug 5492 reintroduced bug 3267 and
5276 caused clients to be evicted for AST timeouts. The fixes in
5277 bug 5192 mean we no longer need to have such short AST timeouts
5278 so ldlm_timeout has been increased.
5281 Frequency : occasional during --force or --failover shutdown under load
5282 Bugzilla : 5949, 4834
5283 Description: Server oops/LBUG if stopped with --force or --failover under load
5284 Details : a collection of import/export refcount and cleanup ordering
5285 issues fixed for safer force cleanup
5288 Frequency : only filesystems larger than 120 OSTs
5289 Bugzilla : 5990, 6223
5290 Description: lfs getstripe would oops on a very large filesystem
5291 Details : lov_getconfig used kfree on vmalloc'd memory
5294 Frequency : only filesystems exporting via NFS to Solaris 10 clients
5295 Bugzilla : 6242, 6243
5296 Description: reading from files that had been truncated to a non-zero size
5297 but never opened returned no data
5298 Details : ll_file_read() reads zeros from no-object files to EOF
5303 Description: A bug in MDS/OSS recovery could cause the OSS to fail an assertion
5304 Details : There's little harm in aborting MDS/OSS recovery and letting it
5305 try again, so I removed the LASSERT and return an error instead.
5307 Severity : enhancement
5309 Description: New debugging infrastructure for tracking down data corruption
5310 Details : The I/O checksum code was replaced to: (a) control it at runtime,
5311 (b) cover more of the client-side code path, and (c) try to narrow
5312 down where problems occurred
5316 Bugzilla : 3819, 4364, 4397, 6313
5317 Description: Racing close and eviction MDS could cause assertion in mds_close
5318 Details : It was possible to get multiple mfd references during close and
5319 client eviction, leading to one thread referencing a freed mfd.
5321 Severity: : enhancement
5322 Bugzilla : 3262, 6359
5323 Description: Attempts to reconnect to servers are now more aggressive.
5324 Details : This builds on the enhanced upcall-less recovery that was added
5325 in 1.4.2. When trying to reconnect to servers, clients will
5326 now try each server in the failover group every 10 seconds. By
5327 default, clients would previously try one server every 25 seconds.
5332 Description: After recovery, certain operations trigger a failed
5333 assertion on a client.
5334 Details : Failing over an mds, using lconf -d --failover, while a
5335 client was doing a readdir() call would cause the client to
5336 LBUG after recovery completed and the readdir() was resent.
5338 Severity : enhancement
5340 Description: Default groups are now added by lconf
5341 Details : You can now run lconf --group <servicename> without having to
5342 manually add groups with lmc.
5345 Frequency : occasional
5347 Description: Nodes with an elan id of 0 trigger a failed assertion
5350 Frequency : always when accessing e.g. tty/console device nodes
5352 Description: tty and some other devices nodes cannot be used on lustre
5353 Details : file's private_data field is used by device data and lustre
5354 values in there got lost. New field was added to struct file to
5355 store fs-specific private data.
5358 Frequency : when exporting Lustre via NFS
5360 Description: NFSD failed occasionally when looking up a path component
5361 Details : NFSD is looking up ".." which was broken in ext3 directories
5362 that had grown large enough to become hashed.
5365 Frequency : Clusters with multiple interfaces not on the same subnet
5367 Description: Nodes will repeatedly try to reconnect to an interface which it
5368 cannot reach and report an error to the log.
5369 Details : Extra peer list entries will be created by lconf with some peers
5370 unreachable. lconf now validates the peer before adding it.
5373 Frequency : Only if a default stripe is set on the filesystem root.
5375 Description: Setting a default stripe on the filesystem root prevented the
5376 filesystem from being remounted.
5377 Details : The client was sending extra request flags in the root getattr
5378 request and did not allocate a reply buffer for the dir EA.
5381 Frequency : occasional, higher if lots of files are accessed by one client
5382 Bugzilla : 6159, 6097
5383 Description: Client trips assertion regarding lsm mismatch/magic
5384 Details : While revalidating inodes the VFS looks up inodes with ifind()
5385 and in rare cases can find an inode that is being freed.
5386 The ll_test_inode() code will free the lsm during ifind()
5387 when it finds an existing inode and then the VFS later attaches
5388 this free lsm to a new inode.
5392 Bugzilla : 6422, 7030
5393 Description: MDS deadlock between mkdir and client eviction
5394 Details : Creating a new file via mkdir or mknod (starting a transaction
5395 and getting the ns lock) can deadlock with client eviction
5396 (gets ns lock and trying to finish a synchronous transaction).
5399 Frequency : occasional
5400 Description: While starting a server, the fsfilt_ext3 module could not be
5402 Details : CFS's improved ext3 filesystem is named ldiskfs for 2.6
5403 kernels. Previously, lconf would still use the ext3 name
5404 when trying to load modules. Now, it will correctly use
5405 ext3 on 2.4 and ldiskfs on 2.6.
5407 Severity : enhancement
5408 Description: The default stripe count has been changed to 1
5409 Details : The interpretation of the default stripe count (0, to lfs
5410 or lmc) has been changed to mean striping across a single
5411 OST, rather than all available. For general usage we have
5412 found a stripe count of 1 or 2 works best.
5414 Severity : enhancement
5415 Description: Add support for compiling against Cray portals.
5416 Details : Conditional compiling for some areas that are different
5420 Frequency : occasional
5421 Bugzilla : 6409, 6834
5422 Description: Creating files with an explicit stripe count may lead to
5423 a failed assertion on the MDS
5424 Details : If some OSTs are full or unavailable, creating files may
5425 trigger a failed assertion on the MDS. Now, Lustre will
5426 try to use other servers or return an error to the
5430 Frequency : occasional
5432 Description: Multiple concurrent overlapping read+write on multiple SMP nodes
5433 caused lock timeout during readahead (since 1.4.2).
5434 Details : Processes doing readahead might match a lock that hasn't been
5435 granted yet if there are overlapping and conflicting lock
5436 requests. The readahead process waits on ungranted lock
5437 (original lock is CBPENDING), while OST waits for that process
5438 to cancel CBPENDING read lock and eventually evicts client.
5440 Severity : enhancement
5442 Description: Initial enabling of flock support for clients
5443 Details : Implements fcntl advisory locking and file status functions.
5444 This feature is provided as an optional mount flag (default
5445 off), and is NOT CURRENTLY SUPPORTED. Not all types of record
5446 locking are implemented yet, and those that are are not guaranteed
5447 to be completely correct in production environments.
5448 mount -t lustre -o [flock|noflock] ...
5451 Frequency : occasional
5453 Description: OSTs running 2.4 kernels but with extents enabled might trip an
5454 assertion in the ext3 JBD (journaling) layer.
5455 Details : The b_committed_data struct is protected by the big kernel lock
5456 in 2.4 kernels, serializing journal_commit_transaction() and
5457 ext3_get_block_handle->ext3_new_block->find_next_usable_block()
5458 access to this struct. In 2.6 kernels there is finer grained
5459 locking to improve SMP performance of the JBD layer.
5463 Description: Changes the "SCSI I/O Stats" kernel patch to default to "enabled"
5465 -----------------------------------------------------------------------------
5467 2005-05-05 Cluster File Systems, Inc. <info@clusterfs.com>
5469 NOTE: Lustre 1.4.2 uses an incompatible network protocol than previous
5470 versions of Lustre. Please update all servers and clients to
5471 version 1.4.2 or later at the same time. You must also run
5472 "lconf --write-conf {config}.xml" on the MDS while it is stopped
5473 to update the configuration logs.
5475 - fix for HPUX NFS client breakage when NFS exporting Lustre (5781)
5476 - mdc_enqueue does not need max_mds_easize request buffer on send (5707)
5477 - swab llog records of type '0' so we get proper header size/idx (5861)
5478 - send llog cancel req to DLM cancel portal instead of cb portal (5515)
5479 - fix rename of one directory over another leaking an inode (5953)
5480 - avoid SetPageDirty on 2.6 (5981)
5481 - don't re-add just-being-destroyed locks to the waiting list (5653)
5482 - when creating new directories, inherit the parent's custom
5483 striping settings if present parent (3048)
5484 - flush buffers from cache before direct IO in 2.6 obdfilter (4982)
5485 - don't hold i_size_sem in ll_nopage() and ll_ap_refresh_count (6077)
5486 - don't hold client locks on temporary worklist from l_lru (5666)
5487 - handle IO errors in 2.6 obdfilter bio completion routine (6046)
5488 - automatically evict dead clients (5921)
5489 - Update file size properly in create+truncate+fstat case (6196)
5490 - Do not unhash mountpoint dentries, do not allow removal of
5492 - Avoid lock ordering deadlock issue with write/truncate (6203,5654)
5493 - reserve enough journal credits in fsfilt_start_log for setattr (4554)
5494 - ldlm_enqueue freed-export error path would always LBUG (6149,6184)
5495 - don't reference lr_lvb_data until after we hold lr_lvb_sem (6170)
5496 - don't overwrite last_rcvd if there is a *_client_add() error (6086)
5497 - Correctly handle reads of files with no objects (6243)
5498 - lctl recover will also mark a device active if deactivate used (5933)
5500 - by default create 1 inode per 4kB space on MDS, per 16kB on OSTs
5501 - allow --write-conf on an MDS with different nettype than client (5619)
5502 - don't write config llogs to MDS for mounts not from that MDS (5617)
5503 - lconf should create multiple TCP connections from a client (5201)
5504 - init scripts are now turned off by default; run chkconfig --on
5505 lustre and chkconfig --on lustrefs to use them
5506 - upcalls are no longer needed for clients to recover to failover
5508 - add --abort-recovery option to lconf to abort recovery on device
5510 - add support for an arbitrary number of OSTs (3026)
5511 - Quota support protocol changes.
5512 - forward compatibility changes to wire structs (6007)
5513 - rmmod NALs that might be loaded because of /etc/modules.conf (6133)
5514 - support for mountfsoptions and clientoptions to the Lustre LDAP (5873)
5515 - improved "lustre status" script
5516 - initialize blocksize for non-regular files (6062)
5517 - added --disable-server and --disable-client configure options (5782)
5518 - introduce a lookup cache for lconf to avoid repeated DB scans (6204)
5519 - Vanilla 2.4.29 support
5520 - increase maximum number of obd devices to 520 (6242)
5521 - remove the tcp-zero-copy patch from the suse-2.4 series (5902)
5522 - Quadrics Elan drivers are now included for the RHEL 3 2.4.21 and
5523 SLES 9 2.6.5 kernels
5524 - limit stripes per file to 160 (the maximum EA size) (6093)
5526 2005-03-22 Cluster File Systems, Inc. <info@clusterfs.com>
5529 - don't LASSERT in ll_release on NULL lld with NFS export (4655, 5760)
5530 - hold NS lock when calling handle_ast_error->del_waiting_lock (5746)
5531 - fix setattr mtime regression from lovcleanup merge (4829, 5669)
5532 - workaround for 2.6 crash in ll_unhash_aliases (5687, 5210)
5533 - small ext3 extents cleanups and fixes (5733)
5534 - improved mballoc code, several small races and bugs fixed (5733, 5638)
5535 - kernel version 43 - fix remove_suid bugs in both 2.4 and 2.6 (5695)
5536 - avoid needless client->OST connect, fix handle mismatch (5317)
5537 - fix DLM error path that led to out-of-sync client, long delays (5779)
5538 - support common vfs-enforced mount options (nodev,nosuid,noexec) (5637)
5539 - fix several locking issues related to i_size (5492,5624,5654,5672)
5540 - don't move pending lock onto export if it is already evicted (5683)
5541 - fix kernel oops when creating .foo in unlinked directory (5548)
5542 - fix deadlock in obdfilter statistics vs. object create (5811)
5543 - use time_{before,after} to avoid timer jiffies wrap (5882)
5544 - shutdown --force/--failover stability (3607,3651,4797,5203,4834)
5545 - Do not leak request if server was not able to process it (5154)
5546 - If mds_open unable to find parent dir, make that negative lookup(5154)
5547 - don't create new directories with extent-mapping (5909, 5936)
5549 - fix lustre/lustrefs init scripts for SuSE (patch from Scali, 5702)
5550 - don't hold the pinger_sem in ptlrpc_pinger_sending_on_import
5551 - change obd_increase_kms to obd_adjust_kms (up or down) (5654)
5552 - lconf, lmc search both /usr/lib and /usr/lib64 for Python libs (5800)
5553 - support for RHEL4 kernel on i686 (5773)
5554 - provide error messages when incompatible logs are encountered (5898)
5556 2005-02-18 Cluster File Systems, Inc. <info@clusterfs.com>
5557 * version 1.4.0.10 (1.4.1 release candidate 1)
5559 - don't keep a lock reference when lock is not granted (4238)
5560 - unsafe list practices (rarely) led to infinite eviction loop (4908)
5561 - add per-fs limit of Lustre pages in page cache, avoid OOM (4699)
5562 - drop import inflight refcount on signal_completed_replay error (5255)
5563 - unlock page after async write error during send (3677)
5564 - handle missing objects in filter_preprw_read properly (5265)
5565 - no transno return for symlink open, don't save no-trasno open (3440)
5566 - don't try to complete elan receive that already failed (4012)
5567 - free RPC server reply state on error (5406)
5568 - clean up thread from ptlrpc_start_thread() on error (5160)
5569 - readahead could read extra page into cache that wasn't ejected (5388)
5570 - prevent races in class_attach/setup/cleanup/detach (5260)
5571 - don't dereference de->d_inode after l_dput of de (5458)
5572 - use "int" for stripe value returned from lock_to_stripe (5544)
5573 - mballoc allocation and error-checking fixes in 2.6 (5504)
5574 - block device patches to fix I/O request sizes in 2.6 (5482)
5575 - look up hostnames for IB nals (5602)
5576 - 2.6 changed lock ordering of 2 semaphores, caused deadlock (5654)
5577 - don't start multiple acceptors for the same port (5277)
5578 - fix incorrect LASSERT in mds_getattr_name (5635)
5579 - export a proc file for general "ping" checking (5628)
5580 - fix "lfs check" to not block when the MDS is down (5628)
5582 - service request history (4965)
5583 - put {ll,lov,osc}_async_page structs in a single slab (4699)
5584 - create an "evict_client" /proc entry on OSTs, like the MDS has
5585 - fix mount usage message, return errors per mount(8) (5168)
5586 - change grep [] to grep "[]" in tests so they work in more UMLs
5587 - fix ppc64/x86_64 spec to use %{_libdir} instead of /usr/lib (5389)
5588 - remove ancient LOV_MAGIC_V0 EA support (5047)
5589 - add "disk I/Os in flight" and "I/O req time" stats in obdfilter
5590 - align r/w RPCs to PTLRPC_MAX_BRW_SIZE boundary for performance (3451)
5591 - allow readahead allocations to fail when low on memory (5383)
5592 - mmap locking landed again, after considerable improvement (2828)
5593 - add get_hostaddr() to lustreDB.py for LDAP support (5459)
5595 2004-11-23 Cluster File Systems, Inc. <info@clusterfs.com>
5598 - send OST transaction number in read/write reply to free req (4966)
5599 - don't ASSERT in ptl_send_rpc() if we run out of memory (5119)
5600 - lock /proc/sys/portals/routes internal state, avoiding oops (4827)
5601 - the watchdog thread now runs as interruptible (5246)
5602 - flock/lockf fixes (but it's still disabled, pending 5135)
5603 - don't use EXT3 constants in llite code (5094)
5604 - memory shortage at startup could cause assertion (5176)
5606 - reorganization of lov code
5607 - single portals codebase
5609 - add extents/mballoc support (5025)
5610 - direct I/O reads in the obdfilter (4048)
5611 - kernel patches from LNXI for 2.6 (bluesmoke, perfctr, mtd, kexec)
5613 tbd Cluster File Systems, Inc. <info@clusterfs.com>
5616 - send OST transaction number in read/write reply to free req (4966)
5617 - don't ASSERT in ptl_send_rpc() if we run out of memory (5119)
5618 - lock /proc/sys/portals/routes internal state, avoiding oops (4827)
5619 - the watchdog thread now runs as interruptible (5246)
5620 - handle missing objects in filter_preprw_read properly (5265)
5621 - unsafe list practices (rarely) led to infinite eviction loop (4908)
5622 - drop import inflight refcount on signal_completed_replay error (5255)
5623 - unlock page after async write error during send (3677)
5624 - return original error code on reconstructed replies (3761)
5625 - no transno return for symlink open, don't save no-trasno open (3440)
5627 - add pid to ldlm debugging output (4922)
5628 - bump the watchdog timeouts -- we can't handle 30sec yet
5629 - extra debugging for orphan dentry/inode bug (5259)
5631 2004-11-16 Cluster File Systems, Inc. <info@clusterfs.com>
5634 - fix TCP_NODELAY bug, which caused extreme perf regression (5134)
5635 - allocate qswnal tx descriptors singly to avoid fragmentation (4504)
5636 - don't LBUG on obdo_alloc() failure, use OBD_SLAB_ALLOC() (4800)
5637 - fix NULL dereference in /proc/sys/portals/routes (4827)
5638 - allow failed mdc_close() operations to be interrupted (4561)
5639 - stop precreate on OST before MDS would time out on it (4778)
5640 - don't send partial-page writes before EOF from client (4410)
5641 - discard client grant for sub-page writes on large-page clients (4520)
5642 - don't free dentries not owned by NFS code, check generation (4806)
5643 - fix lsm leak if mds_create_objects() fails (4801)
5644 - limit debug_daemon file size, always print CERROR messages (4789)
5645 - use transno after validating reply (3892)
5646 - process timed out requests if import state changes (3754)
5647 - update mtime on OST during writes, return in glimpse (4829)
5648 - add mkfsoptions to LDAP (4679)
5649 - use ->max_readahead method instead of zapping global ra (5039)
5650 - don't interrupt __l_wait_event() during strace
5652 - add software watchdogs to catch hung threads quickly (4941)
5653 - make lustrefs init script start after nfs is mounted
5654 - fix CWARN/ERROR duplication (4930)
5655 - return async write errors to application if possible (2248)
5656 - add /proc/sys/portal/memused (bytes allocated by PORTALS_ALLOC)
5657 - print NAL number in %x format (4645)
5658 - update barely-supported suse-2.4.21-171 series (4842)
5659 - support for sles 9 %post scripts
5660 - support for building 2.6 kernel-source packages
5661 - support for sles km_* packages
5663 2004-10-07 Cluster File Systems, Inc. <info@clusterfs.com>
5666 - ignore -ENOENT errors in osc_destroy (3639)
5667 - notify osc create thread that OSC is being cleaned up (4600)
5668 - add nettype argument for llmount in #5d in conf-sanity.sh (3936)
5669 - reconstruct ost_handle() like mds_handle() (4657)
5670 - create a new thread to do import eviction to avoid deadlock (3969)
5671 - let lconf resolve symlinked-to devices (4629)
5672 - don't unlink "objects" from directory with default EA (4554)
5673 - hold socknal file ref over connect in case target is down (4394)
5674 - allow more than 32000 subdirectories in a single directory (3244)
5675 - fix blocks count for O_DIRECT writes (3751)
5676 - OST returns ENOSPC from object create when no space left (4539)
5677 - don't send truncate RPC if file size isn't changing (4410)
5678 - limit OSC precreate to 1/2 of value OST considers bogus (4778)
5679 - bind to privileged port in socknal and tcpnal (3689)
5681 - rate limit CERROR/CWARN console message to avoid overload (4519)
5682 - GETFILEINFO dir ioctl returns LOV EA + MDS stat in 1 call (3327)
5683 - basic mmap support (3918)
5684 - kernel patch series update from b1_4 (4711)
5686 2004-09-16 Cluster File Systems, Inc. <info@clusterfs.com>
5689 - avoid crash during MDS cleanup with OST shut down (2775)
5690 - fix loi_list_lock/oig_lock inversion on interrupted IO (4136)
5691 - don't use bad inodes on the MDS (3744)
5692 - dynamic object preallocation to improve recovery speed (4236)
5693 - don't hold spinlock over lock dumping or change debug flags (4401)
5694 - don't zero obd_dev when it is force cleaned (3651)
5695 - print grants to console if they go negative (4431)
5696 - "lctl deactivate" will stop automatic recovery attempts (3406)
5697 - look for existing locks in ldlm_handle_enqueue() (3764)
5698 - don't resolve lock handle twice in recovery avoiding race (4401)
5699 - revalidate should check working dir is a directory (4134)
5701 - don't always mark "slow" obdfilter messages as errors (4418)
5703 2004-08-24 Cluster File Systems, Inc. <info@clusterfs.com>
5706 - don't close LustreDB during write_conf until it is done (3860)
5707 - fix typo in lconf for_each_profile (3821)
5708 - allow dumping logs from multiple threads at one time (3820)
5709 - don't allow multiple threads in OSC recovery (3812)
5710 - fix debug_size parameters (3864)
5711 - fix mds_postrecov to initialize import for llog ctxt (3121)
5712 - replace config semaphore with spinlock (3306)
5713 - be sure to send a reply for a CANCEL rpc with bad export (3863)
5714 - don't allow enqueue to complete on a destroyed export (3822)
5715 - down write_lock before checking llog header bitmap (3825)
5716 - recover from lock replay timeout (3764)
5717 - up llog sem before sending rpc (3652)
5718 - reduce ns lock hold times when setting kms (3267)
5719 - change a dlm LBUG to LASSERTF, to maybe learn something (4228)
5720 - fix NULL deref and obd_dev leak on setup error (3312)
5721 - replace some LBUG about llog ops with error handling (3841)
5722 - don't match INVALID dentries from d_lookup and spin (3784)
5723 - hold dcache_lock while marking dentries INVALID and hashing (4255)
5724 - fix invalid assertion in ptlrpc_set_wait (3880)
5726 - add libwrap support for the TCP acceptor (3996)
5727 - add /proc/sys/portals/routes for non-root route listing (3994)
5728 - allow setting MDS UUID in .xml (2580)
5729 - print the stack of a process that LBUGs (4228)
5731 2004-07-14 Cluster File Systems, Inc. <info@clusterfs.com>
5734 - don't cleanup request in ll_file_open() on failed MDS open (3430)
5735 - make sure to unset replay flag from failed open requests (3440)
5736 - if default stripe count is 0, use OST count for inode size (3636)
5737 - update parent mtime/ctime on client for create/unlink (2611)
5738 - drop dentry ref in ext3_add_link from open_connect_dentry (3266)
5739 - free recovery state on server during a forced cleanup (3571)
5740 - unregister_reply for resent reqs (3063)
5741 - loop back devices mounting and status check on 2.6 (3563)
5742 - fix resource-creation race that can provoke i_size == 0 (3513)
5743 - don't try to use bad inodes returned from MDS/OST fs lookup (3688)
5744 - more debugging for page-accounting assertion (3746)
5745 - return -ENOENT instead of asserting if ost getattr+unlink race (3558)
5746 - avoid deadlock after precreation failure (3758)
5747 - fix race and lock order deadlock in orphan handling (3450, 3750)
5748 - add validity checks when grabbing inodes from l_ast_data (3599)
5750 - add /proc/.../recovery_status to obdfilter (3428)
5751 - lightweight CDEBUG infrastructure, debug daemon (3668)
5752 - change default OSC RPC parameters to be better on small clusters
5753 - turn off OST read cache for files smaller than 32MB
5754 - install man pages and include them in rpms (3100)
5755 - add new init script for (un)mounting lustre filesystems (2593)
5756 - run chkconfig in %post for init scripts (3701)
5757 - drop scimac NAL (unmaintained)
5759 2004-06-17 Cluster File Systems, Inc. <info@clusterfs.com>
5762 - clean kiobufs before and after use (3485)
5763 - strip trailing '/'s before comparing paths with /proc/mounts (3486)
5764 - remove assertions to work around "in-flight rpcs" recovery bug (3063)
5765 - change init script to fail more clearly if not run as root (1528)
5766 - allow clients to reconnect during replay (1742)
5767 - fix ns_lock/i_sem lock ordering deadlock for kms update (3477)
5768 - don't do DNS lookups on NIDs too small for IP addresses (3442)
5769 - re-awaken ptlrpcd if new requests arrive during check_set (3554)
5770 - fix cond_resched (3554)
5771 - only evict unfinished clients after recovery (3515)
5772 - allow bulk resend, prevent data loss (3570)
5773 - dynamic ptlrpc request buffer allocation (2102)
5774 - don't allow unlinking open directory if it isn't empty (2904)
5775 - set MDS/OST threads to umask 0 to not clobber client modes (3359)
5776 - remove extraneous obd dereference causing LASSERT failure (3334)
5777 - don't use get_cycles() when creating temp. files on the mds (3156)
5778 - hold i_sem when setting i_size in ll_extent_lock() (3564)
5779 - handle EEXIST for set-stripe, set proper directory name (3336)
5781 - servers can dump a log evicting a client - lustre.dump_on_timeout=1
5782 - fix ksocknal_fmb_callback() error messages (2918)
5784 2004-05-27 Cluster File Systems, Inc. <info@clusterfs.com>
5787 - don't copy lvb into (possibly NULL) reply on error (2983)
5788 - don't deref dentry after dput, don't free lvb on error (2922)
5789 - use the kms to determine writeback rpc length (2947)
5790 - increment oti_logcookies when osc is inactive (2948)
5791 - update client's i_blocks count via lvb messages (2543)
5792 - handle intent open/close of special files properly (1557)
5793 - mount MDS with errors=remount-ro, like obdfilter (2009)
5794 - initialize lock handle to avoid ASSERT on error cleanup (3057)
5795 - don't use cancelling-locks' kms values (2947)
5796 - use highest lock extent for kms, not last one (2925)
5797 - don't dereference ERR_PTR() dentry in error handling path (3107)
5798 - fix thread race in portals_debug_dumplog() (3122)
5799 - create lprocfs device entries at setup instead of at attach (1519)
5800 - common AST error handler, don't evict client on completion race (3145)
5801 - zero nameidata in detach_mnt in 2.6 (3118)
5802 - verify d_inode after revalidate_special is valid in 2.6 (3116)
5803 - use lustre_put_super() to handle zconf unmounts in 2.6 (3064)
5804 - initialize RPC timeout timer earlier for 2.6 (3219)
5805 - don't dereference NULL reply buffer if mdc_close was never sent (2410)
5806 - print nal/nid for unknown nid (3258)
5807 - additional checks for oscc recovery before doing precreate (3284)
5808 - fix ll_extent_lock() error return code for 64-bit systems (3043)
5809 - don't crash in mdc_close for bad permissions on open (3285)
5810 - zero i_rdev for non-device files (3147)
5811 - clear page->private before handing to FS, better assertion (3119)
5812 - tune the read pipeline (3236)
5813 - fix incorrect decref of invalidated dentry (2350)
5814 - provide read-ahead stats and refine rpc in flight stats (3328)
5815 - don't hold journal transaction open across create RPC (3313)
5816 - update atime on MDS at close time (3265)
5817 - close LDAP connection when recovering to avoid server load (3315)
5818 - update iopen-2.6 patch with fixes from 2399,2517,2904 (3301)
5819 - don't leak open file on MDS after open resend (3325)
5820 - serialize filter_precreate and filter_destroy_precreated (3329)
5821 - loop device shouldn't call sync_dev() for nul device (3092)
5822 - clear page cache after eviction (2766)
5823 - resynchronize MDS->OST in background (2824)
5824 - refuse to mount the same filesystem twice on same mountpoint (3394)
5825 - allow llmount to create routes for mounting behind routers (3320)
5826 - push lock cancellation to blocking thread for glimpse ASTs (3409)
5827 - don't call osc_set_data_with_check() for TEST_LOCK matches (3159)
5828 - fix rare problem with rename on htree directories (3417)
5830 - allow default OST striping configuration per directory (1414)
5831 - fix compilation for qswnal for 2.6 kernels (3125)
5832 - increase maximum number of MDS request buffers for large systems
5833 - change liblustreapi to be useful for external progs like lfsck (3098)
5834 - increase local configuration timeout for slow disks (3353)
5835 - allow configuring ldlm AST timeout - lustre.ldlm_timeout=<seconds>
5837 2004-03-22 Cluster File Systems, Inc. <info@clusterfs.com>
5840 - fixes for glimpse AST timeouts / incorrectly 0-sized files (2818)
5841 - don't overwrite extent policy data in reply if lock was blocked (2901)
5842 - drop filter export grants atomically with removal from device (2663)
5843 - del obd_self_export from work_list in class_disconnect_exports (2908)
5844 - don't LBUG if MDS recovery times out during orphan cleanup (2530)
5845 - swab reply message in mdc_close, other PPC fixes (2464)
5846 - fix destroying of named logs (2325)
5847 - overwrite old logs when running lconf --write_conf (2264)
5848 - bump LLOG_CHUNKSIZE to 8k to allow for larger clusters (2306)
5849 - fix race in target_handle_connect (2898)
5850 - mds_reint_create() should take same inode create lock (2926)
5851 - correct journal credits calculated for CANCEL_UNLINK_LOG (2931)
5852 - don't close files for self_export to avoid uninitialized obd (2936)
5853 - allow MDS with the same name as client node (2939)
5854 - hold dentry reference for closed log files for unlink (2325)
5855 - reserve space for all logs during transactions (2059)
5856 - don't evict page beyond end of stripe extent (2925)
5857 - don't oops on a deleted current working directory (2399)
5858 - handle hard links to targets without a parent properly (2517)
5859 - don't dereference NULL lock when racing during eviction (2867)
5860 - don't grow lock extents when lots of conflicting locks (2919)
5862 2004-03-04 Cluster File Systems, Inc. <info@clusterfs.com>
5865 - account for cache space usage on clients to avoid data loss (974)
5866 - lfsck support in lustre kernel code (2349)
5867 - reduce journal credits needed for BRW writes (2370)
5868 - orphan handling to avoid losing space on client/server crashes
5869 - ptlrpcd can be blocked, stopping ALL progress (2477)
5870 - use lock value blocks to assist in proper KMS, faster stat (1021)
5871 - takes i_sem instead of DLM locks internally on obdfilter (2720)
5872 - recovery for initial connections (2355)
5873 - fixes for mds_cleanup_orphans (1934)
5874 - abort_recovery crashes MDS in b_eq (mds_unlink_orphan) (2584)
5875 - block all file creations until orphan recovery completes (1901)
5876 - client remove rq_connection from request struct (2423)
5877 - conf-sanity test_5, proper cleanup in umount log not availale (2640)
5878 - recovery timer race (2670)
5879 - mdc_close recovey bug (2532)
5880 - ptlrpc cleanup bug (2710)
5881 - mds timeout on local locks (2588)
5882 - namespace lock held during RPCs (2431)
5883 - handle interrupted sync write properly (2503)
5884 - don't try to handle a message that hasn't been replied to (2699)
5885 - client assert failure during cleanup after abort recovery (2701)
5886 - leak mdc device after failed mount (2712)
5887 - ptlrpc_check_set allows timedout requests to complete (2714)
5888 - wait for inflight reqs when ptlrpcd finishes (2710)
5889 - make sure unregistered services are removed from the srv_list
5890 - reset bulk XID's when resending them (caught by 1138 test)
5891 - unregister_bulk after timeout
5892 - fix lconf error (2694)
5893 - handle write after unfinished setstripe, stripe-only getstripe (2388)
5894 - readahead locks pages, leaves pending causing memory pressure (2673)
5895 - increase OST request buffers to 4096 on large machines (2729)
5896 - fix up permission of existing directories in simple_mkdir (2661)
5897 - init deleted item, add assertions ptlrpc_abort_inflight() (2725)
5898 - don't assign transno to errored transactions (2742)
5899 - don't delete objects on OST if given a bogus objid from MDS (2751)
5900 - handle large client PAGE_SIZE readdir on small PAGE_SIZE MDS (2777)
5901 - if rq_no_resend, then timeout request after recovery (2432)
5902 - fix MDS llog_logid record size, 64-bit array alignment (2733)
5903 - don't call usermode_helper from ptlrpcd, DEFAULT upcall (2773)
5904 - put magic in mount.lustre data, check for bad/NULL mount data (2529)
5905 - MDS recovery shouldn't delete objects that it has given out (2730)
5906 - if enqueue arrives after completion, don't clobber LVB (2819)
5907 - don't unlock pages twice when trigger_group_io returns error (2814)
5908 - don't deref NULL rq_repmsg if ldlm_handle_enqueue failed (2822)
5909 - don't write pages to disk if there was an error (1450)
5910 - don't ping imports that have recovery disabled (2676)
5911 - take buffered bytes into account when balancing socknal conn (2817)
5912 - hold a DLM lock over readdir always, use truncate_inode_pages (2706)
5913 - reconnect unlink llog connection after MDS reconnects to OST (2816)
5914 - remove little-endian swabbing of llog records (1987)
5915 - set/limit i_blksize to LL_MAX_BLKSIZE on client (2884)
5916 - retry reposting request buffers if they fail (1191)
5917 - grow extent at grant time to avoid granting a revoked lock (2809)
5918 - lock revoke doesn't evict page if covered by a second lock (2765)
5919 - disable VM readahead to avoid reading outside lock extents (2805)
5921 - return LL_SUPER_MAGIC from statfs for the filesystem type (1972)
5922 - updated kernel patches for hp-2.4.20 kernel (2681)
5924 2004-02-07 Cluster File Systems, Inc. <info@clusterfs.com>
5927 - fix truncated write corruption (2366)
5928 - fix for failed assertion in iopen_connect_dentry (1792,2517)
5930 - don't flag the ptlrpcd thread with PF_MEMALLOC (2636)
5931 - ensure len(uuid) < 37 in lmc (1171)
5932 - fix ia64 OOPS in llog_test (2255)
5933 - zero end of page at obdfilter for partial page writes (2648)
5934 - don't leave stale dentries around after renames (bug 2428)
5935 - fix timeouts when evicting a client with a single lock held (2642)
5936 - set deadline for the initial HELLO message to drain (2634)
5937 - print out dotted-quad IP addresses in the socknal (2302)
5939 - additional debugging for MDS client eviction problem (2443)
5940 - fix mkfsoptions support for osts (2603, 2604)
5942 2004-01-27 Cluster File Systems, Inc. <info@clusterfs.com>
5945 - add series for the vanilla 2.6.0 kernel
5946 - add series for the vanilla 2.4.24 kernel
5947 - add series for a cray x86/64 UL kernel drop
5948 - fix xattr patches for the vanilla 2.4.19 series
5950 - generate true UUIDs in lmc (1171)
5951 - have portals stack dumping break in UML (2466)
5952 - avoid bad dchild deref; avoid inum lock w/o creation (2362)
5953 - allocate with _NOFS in ldlm to avoid deadlock (1933)
5954 - wake callback waiting threads on client eviction (2460)
5955 - Add --ptldebug and --subsystem to lmc (1719)
5956 - update assertion to allow safe interrupt allocation
5957 - set rq_no_resend for cancel requests (2432)
5958 - recalculate ptlrpcd timeout after resend (2494)
5959 - call vfs_rmdir when removing pending directories (2368)
5960 - fix renaming a file to itself (2429)
5961 - lmc creates a default one-stripe lov (2454)
5962 - expand procfs space to handle large clusters (2326)
5963 - increase UML stack to avoid overflow
5964 - update lconf's list of debug and subsystem masks
5965 - fix lfs find --obd (2510)
5966 - /proc tunable for disabling filter read caching (2591)
5967 - stop rpm packages from altering slapd.conf (2301)
5968 - disable nagle in the socknal under 0conf (2578)
5969 - choose mds inode size based on stripe count (2572)
5970 - fix kernel-source rpm problems (2516)
5972 - add --disable-doc to avoid pdf generation (2421)
5973 - update documentation, tests, type-os, comments
5974 - avoid format warnings on ia64
5975 - remove the TOE NAL
5976 - tiny code cleanups by removing unused fields
5978 2004-01-07 Cluster File Systems, Inc. <info@clusterfs.com>
5981 - fix obvious semaphore misuse in as-yet-unused setattr path (2348)
5982 - remove the most blatant lies from BUILDING file (2371)
5983 - change default debug level to reasonable production setting
5984 - reduce client side cache size to reduce cache flush time
5985 - reduce max RPCs in flight to avoid unnecessary file fragmentation
5986 - make TCP zerocopy and pinger support enabled by default (2476)
5987 - sync writes completed after process exits caused crashes (2319)
5988 - maintain correct mount count on the MDS (2356)
5989 - backout 1557, because 2316 wasn't really fixed
5990 - better file I/O statistics gathering in /proc
5991 - don't take unnecessary, deadlock-inducing bug in readpage (2383)
5992 - another kernel patch to fix zero-copy TCP function export
5993 - don't take duplicate lock when processing re-sent getattr (2420)
5994 - lctl uses obd_self_export instead of creating new conn (2353)
5995 - MDS/OST recovery case which requires object creation asserted (2425)
5996 - move lfs from /usr/sbin to /usr/bin in packages
5997 - fix race between mds_client_add and mds_client_free (2417)
5998 - use kmalloc instead of slabs in portals (2430)
5999 - don't create duplicate records when a failover MDS is present (2442)
6000 - remove unnecessary mount age check (2332)
6001 - don't remove directory inodes from locks prematurely (2451)
6002 - don't break if MDS service name is the same as hostname (2103)
6003 - fix races in client write RPC generation when cache full (2482)
6005 2003-12-13 Cluster File Systems, Inc. <info@clusterfs.com>
6008 - remove now-unused request->rq_obd (278)
6009 - if an allocation fails, print out how much memory we've used (1933)
6010 - use PORTAL_SLAB_ALLOC for structures, to get GFP_MEMALLOC (1933)
6011 - add the "configurable stack size" patch to most series files (1256)
6012 - ability to write large log records, for 100+ OST configs (2306)
6013 - fix NULL deref when filter_prep fails (2314)
6014 - fix operator precedence error in filter_sync
6015 - dynamic allocation of socknal TX descriptors (2315)
6016 - fix a missed case in the GFP_MEMALLOC patch, can cause deadlock (2310)
6017 - fix gcc 2.96 compilation problem in xattr kernel patch (2294)
6018 - ensure that CWARN messages in Portals always get to the syslog
6019 - __init/__exit are not for prototype decls (ldlm_init/exit)
6020 - x86-64 compile warning fixes
6021 - fix gateway LMC keyword conflict (2318)
6022 - fix MDS lock inversions in getattr/reint paths (1844)
6023 - fix a rare lock re-ordering bug, which caused deadlock (2322)
6024 - fix i_sem/journal inversion in fsfilt_ext3_write_record (2306)
6025 - DLM race condition prevented some lock evictions (2328)
6026 - ENOMEM detection and retry on socknal sends (2230)
6027 - use GFP_NOFS throughout Lustre, to combat ENOMEM (2230)
6028 - move osc_rpcd into ptlrpc, for use in MDC and others (2329)
6029 - protect MDS inode fsdata with stronger locking; fixes assertion (2313)
6030 - better error messages when a client is rejected during recovery (1505)
6031 - avoid cancelling locks which were never granted, after failure (2330)
6032 - fix i_sem/journal inversion in mds_client_add (2333)
6033 - fix truncate/getattr lock cycle deadlock (2334)
6034 - use rpcd to send close; allows resend after timeout, avoid leak (1897)
6035 - fix two rare exit paths which could leak an l_lock() ref (2321)
6036 - fencepost error in MDS/OST orphan recovery (2226)
6037 - make log record alignment 8 bytes (1988)
6038 - lstripe now fails when requested offset > ost_count (2237)
6039 - ensure that all kernel series have a complete list.h (1607)
6040 - fix crashes in special-file operations (2316)
6041 - lctl create/brw OID mismatch, caused by obsolete filter loop (2339)
6043 - allow configurable automake binary, for testing new versions
6044 - small update to the lfs documentation
6046 2003-12-03 Cluster File Systems, Inc. <info@clusterfs.com>
6048 * fix negative export reference count in fsfilt_sync (2312)
6050 2003-12-01 Cluster File Systems, Inc. <info@clusterfs.com>
6051 * release candidate 0.9.1
6053 - orphans are moved into the PENDING directory for possible recovery
6054 - replayed opens now open by fid for orphan/rename safety (1042)
6055 - last close of an orphan inode generates a transno (683)
6056 - chdir() and mount() now pin the directory entry (1020)
6057 - avoid CERROR in normal ll_setattr_raw() error case (1500)
6058 - discard very old requests without processing them (1502)
6059 - remove some common, well-understood CERRORs (1505)
6060 - require O_DIRECT I/O to be page-sized to workaround IA64 crash (1609)
6061 - clear "grant" flags in OST replies until OST grant code lands (1644)
6062 - fix read performance by not clobbering i_blksize on client (1598)
6063 - fix __ldlm_handle2lock oops by not dereferencing lock after PUT (1625)
6064 - make LRU size a /proc tunable, clears locks when reduced (707)
6065 - fix some lprocfs rot that prevented ptlbd from loading (1732)
6066 - server locks take references on exports now (1558)
6067 - build fixes for 2.4.20-rh trees (1663)
6068 - return an error from lov_create if all OSCs are inactive (1751)
6069 - fix import levels when a reconnect happens without a timeout (1597)
6070 - exit early from mds_open if we get a lookup error (1749)
6071 - partial page read at EOF wouldn't wait for disk before sending (1642)
6072 - avoid NULL deref in obdfilter when reading page past EOF (1592)
6073 - avoid LASSERT in ll_intent_lock if server failed very early (1090)
6074 - fix LBUG in ll_it_open_error with rc = -2 (1861)
6075 - write/truncate lock inversion (1639)
6076 - Don't auto-load obdclass, portals modules during cleanup (1495)
6077 - fix timestamps from jumping to "now" (1763)
6078 - extra journal assertions (1648)
6079 - add an extra multiunlink test (1771)
6080 - fix read_record/write_record API (1776)
6081 - fix leak of offset_extent, possible incorrect i_size later (1772)
6082 - fix lasserts in mis-matched transnos during open-unlink testing (1541)
6083 - Debugging for the kqswnal_get_idle_tx problems (1820)
6084 - Allow recovery to be attempted multiple times (1536)
6085 - Write out MDS last_rcvd file after it is first created (1600)
6086 - Fix tx_descriptor leak in failed transmit situations (1827)
6087 - ext3 journaling fixes for assertion failure after IO error (1871)
6088 - class_export_put() on freed export after completion AST error (1896)
6089 - Fix revalidate looping in VFS (1322)
6090 - Don't access a freed export during MDS_REINT timeout (1521)
6091 - Add open-unlink recovery support on the MDS (1673,1764)
6092 - Return an error if no MDS data was read from last_rcvd (1946)
6093 - Fix for lookup "." or ".." crash on error (1932,1931,1935)
6094 - Don't setup a disk device that doesn't match exported UUID (317)
6095 - Reduce bulk RPC timeout to avoid cascading client/OST failures (1845)
6096 - avoid committing NULL handle in force close
6097 - local.sh is now a one-stripe LOV configuration
6098 - POSIX utime.4 -EPERM on FIFO not owned by user (56)
6099 - fix ext3 htree duplicate directory entry corruption (1516)
6100 - POSIX creat.13, fstat.1, open.18, stat.3 new file atime/mtime (2020)
6101 - update to new LOV EA format (2097)
6102 - interoperability for different PAGE_SIZE/wordsize (686,1821,1343,2042)
6104 2003-06-15 Phil Schwan <phil@clusterfs.com>
6107 - imports and exports cleanup too early, need refcounts (349, 879, 1045)
6108 - per-import/export recovery handling (958, 931, 959)
6109 - multiple last-rcvd slots, for serving multiple FSes (949)
6110 - connections are again shared between multiple imp/exports (963, 964)
6111 - "umount -f" would hang if any requests needed to be sent (393, 978)
6112 - avoid pinning large req buffer by copying for queued messages (989)
6113 - add "uuid" to "lctl device" command to help upcalls (991)
6114 - "open" RPCs with transnos would confuse recovery counters (1037)
6115 - do proper endian conversion of all wire messages (288, 340, 891)
6116 - remove OST bulk get LBUGs, fix ost_brw_write cleanup (1126)
6117 - call waiting locks callback from LDLM recovery thread (1127, 1151)
6118 - fix ptlrpc_connection leak in target_handle_connect (1174)
6119 - fix import refcounting bug in OST and MDS cleanup (1134)
6120 - if an invalid-at-open-time OSC returned before close(), LBUG (1150)
6121 - fix very unlikely obd_types race condition (501)
6122 - remove osc_open hack for echo_client (1187)
6123 - we leaked exports/dlmimps for forcibly disconnected clients (1143)
6124 - a failure in read_inode2 leads to deadlock (1139)
6125 - cancel ack-locks as soon as transaction is committed (1072)
6126 - fix major leaks and crashes in the bulk I/O path (937, 1057)
6127 - make sure to commitrw after any preprw to avoid deadlock (1162)
6128 - failing to execute a file in a lustre FS would lock inode (1203)
6129 - small DEBUG_REQ fix to avoid dereferencing a NULL (1227)
6130 - don't ASSERT while cleaning up an incompletely-setup obd (1248)
6131 - obd_uuid2tgt would walk off the end of the list (1255)
6132 - on IA64 the osc would give portals incorrect bulk size (1258)
6133 - fix debug daemon ioctl interface; allows daemon on ia64 (1274)
6134 - fix lock inversion caused by new llite matching code (1282)
6135 - limit the number of dirty pages on a client to 10MB (1286)
6136 - timed out locks were not being corrected cancelled (1289)
6137 - fix O_DIRECT above 4GB on IA-32 (1292)
6138 * major user-visible changes
6139 - fail out/fail over policy now controlled by the upcall (993)
6141 - add OBD_PING to check server availability and failure (954)
6142 - lustre messages are now sent in sending host order (288, 340, 891)
6143 - add eadatalen to MDS getattr reply (340)
6144 - OST read replies may contain second buffer, with per-page status (593)
6146 2003-03-11 Phil Schwan <phil@clusterfs.com>
6149 - LDLM_DEBUG macro fix, for gcc 3.2 (850)
6150 - failed open()s could cause deadlock; fixed (867, 869)
6151 - stop cancelling OST locks when files are closed (481)
6152 - overlapping XID spaces caused network corruption (851, 853)
6153 - fix unsafe fsfilt counter arithmetic; change to atomic_t
6154 - setattr_raw added, to do single-RPC, server-side setattrs
6155 - lmc/lconf syntax change for OST UUIDs
6156 - fix crashy race condition between ptlrpc_free_req and osc_close
6157 - don't use request in mdc_enqueue if we hit a timeout (889)
6158 - don't set the inode i_size for regular files from the MDS (896)
6159 - handle out of order completion AST (842)
6160 - don't LBUG if a lock request times out after receiving AST (913)
6161 - avoid d_rehash race in ll_find_alias by rehashing inside dcache_lock
6162 - if a bad lock AST arrives, send an error instead of dropping entirely
6163 - return 0 from revalidate2 if ll_intent_lock returns -EINTR (912)
6164 - fix leak in bulk IO when only partially completed (899, 900, 926)
6165 - fix O_DIRECT for ia64 (55)
6166 - (almost) eliminate Lustre-kernel-thread effects on load average (722)
6167 - C-z after timeout could hang a process forever; fixed (977)
6169 - client-side I/O cache (678, 924, 929, 941, 970)
6171 - READPAGE and SETATTRs which don't take server-side locks get
6174 2003-02-11 Phil Schwan <phil@clusterfs.com>
6177 - Fix ldlm_lock_match on the MDS to avoid matching remote locks (592)
6178 - Fix fsfilt_extN_readpage() to read a full page of directory
6179 entries, or fake the remainder if PAGE_SIZE != blocksize (500)
6180 - Avoid extra mdc_getattr() in ll_intent_lock when possible (534, 604)
6181 - Fix imbalanced LOV object allocation and out-of-bound access (469)
6182 - Most intent operations were removed, in favour of a new RPC mode
6183 that does a single RPC to the server and bypasses most of the VFS
6184 - All LDLM resource ID arrays were removed in favour of ldlm_res_id
6185 - Aggressively cancel local locks on DLM servers
6186 - mds_reint_unlink sends EA to the client if it's the last nlink.
6187 client uses that EA to unlink OST objects.
6188 - mds_reint_{rename,unlink,link} were rewritten to take ordered locks
6189 - recursive symlinks were fixed (439)
6190 - fixed NULL deref in DEBUG_REQ
6191 - filter_update_lastobjid no longer calls sync, which annoyed extN
6192 - fixed multi-client small-writes to a single file problem (445)
6193 - fixed mtime updates during file writes (607)
6194 - fixed vector writes on obdfilter causing problems when ENOSPC (670)
6195 - fixed bug in obd_brw_read/write() (under guise of testing 367)
6196 - fixed Linux OST size reporting problem (444, 656)
6197 - OST now updates object mtime with writes or setattr (607, 619)
6198 - client verifies file size before zeroing page past EOF (445)
6199 - OST now writes last allocated objid to disk with allocation (108)
6200 - LOV on echo now works (409)
6202 - mds_reint_unlink sends a new buffer, with the EA included. this
6203 buffer is only valid if body->valid & OBD_MD_FLEASIZE, which is only
6204 set if a regular file was being unlinked, and it was the last link
6205 - use PtlGet from the target for bulk writes (315)
6206 - OST now updates object mtime with writes or setattr (607, 619)
6207 - LDLM now has a grant-time callback to revalidate locked items, if
6209 - Many MDS operations were reorganized to combat race conditions
6211 - Merge b_intel branch (updated lprocfs code) - now at /proc/fs/lustre
6212 - configure check to avoid gcc version 2.96 20000731-2.96-98) (606)
6214 2003-01-06 Andreas Dilger <adilger@clusterfs.com>
6217 - Fully reactivate OST imports after reconnection (512, others)
6218 - Make sure client sees our -ENOTCONN from mds_handle (513 - partial)
6219 - More graceful error handling for truncating on dead OST (515)
6220 - Don't error out unless we're actually accessing dead stripes (474)
6221 - Fix garbage sizes when stripes are missing (410)
6222 - LRU counters were broken, causing constant lock purge (433, 432)
6223 - garbage on read from stripes with failed OSTs (441)
6224 - mark OSCs as active before reconnecting during recovery (438)
6225 - lov_enqueue and lov_cancel need to handle inactive OSTs (403)
6226 - lfind did not preserve OST order in output (443)
6227 - symlinks cause hung clients, incorrect data (439)
6228 - stop dereferencing request after dropping refcount (457)
6229 - don't LASSERT(spin_is_locked) on non-SMP (455)
6230 - fixes for many rename() bugs
6231 - fstat didn't correctly synchronize attributes (399)
6232 - server must handle lock cancellation during blocking AST prep (487)
6233 - bulk descriptors were free()d too soon (511)
6234 - fix paths in lconf, which would load incorrect modules (451, 507)
6235 - fix confusing lconf 'host not found' error message (386)
6236 - fix lock order deadlock on OST (O/R i_sem before journal ops, 478)
6237 - fix race condition in mdc_blocking_ast() for inode access (526)
6238 - fix lov_unpackmd() unpacking wrong number of stripes (537)
6239 - fix lov_set_osc_active() marking wrong OSC inactive (440)
6240 - fix bad lstripe lov_unpackmd() assertion (fix layering too) (527)
6241 - fix multiple writes of stripe MD to MDS (358, maybe 519)
6242 - fix lstripe in several ways (kernel side) (527)
6243 - fix request leak in ldlm_cli_enqueue (262)
6244 - incorrect OSC was marked inactive after OST failure
6245 - call mds_fs_cleanup before unmounting filesystem (524)
6246 - fix races between taking ns_lock and ldlm_lock_change_resource
6247 - fix races updating LOV export open file list
6248 - fix lov_enqueue error path, avoid decref-ing bad lock handle
6249 - fix recovery NULL deref in ldlm_cli_cancel_unused
6250 - fix some DLM races by using new hash table for lock handles (419)
6251 - permit the client to specify desired inodes, at replay
6252 - duplicate requests when we queue them for replay reintegration
6253 - fix last_rcvd offset calculation
6254 - sync after each recovered transaction, so we always make progress
6255 - never, not always, ERESTART requests without transnos
6256 - store the lov_desc in the MDS, so we don't depend on getlovinfo to
6258 - skip replay if the MDS says that the client is already connected
6259 - don't check for a recovery-enabled export to match lctl's UUID
6260 - don't INC_USE_COUNT for phantom exports
6261 - don't crash when cleaning up phantom exports (567)
6262 - don't double-finish or set replay data for errored mdc_open requests
6263 - abort requests when they time out, so we don't get old replies
6264 - send/receive replies for AST messages again
6265 - if the client says that it doesn't have the lock, cancel it on the
6267 - if we timeout during I/O, don't try to cancel an in-use lock; instead
6268 mark it as destroyed, it will all work out when decref is called
6269 - fix module use counts (22, 581)
6271 - ASTs now expect a reply (server cancels lock on error reply)
6273 2002-12-02 Andreas Dilger <adilger@clusterfs.com>
6276 - fix many simultaneous client startup (392)
6277 - fix dentry->d_it clobbering
6278 - credentials weren't being shipped for readdir/getattr operations
6279 - remove invalid assertions triggered during some concurrent MD
6281 - proper Lustre versions added (336, 389)
6282 - fix memory leak for create error case (398)
6283 - fix LOV locking bug that would get cli/srv out of sync
6284 - fix echo client over LOV (409)
6285 - fix dbench 2, extN refcount problem (170, 258, 356, 418)
6286 - fix double-O_EXCL intent crash (424)
6287 - avoid sending multiple lock CANCELs (352)
6289 - MDS can do multi-client recovery (modulo bugs in new code)
6291 - many updates, edits, cleanups
6293 2002-11-18 Phil Schwan <phil@clusterfs.com>
6296 - fix null d_it dereference (346)
6297 - fix full OST/dbench hang (333)
6298 - fix permission problem with file removal (286)
6299 - fix removal of OSCs from LOV when they fail
6300 - fix NULL deref during bulk timeout (214)
6301 - fix problems related to multiple filesystems on one MDS (241)
6302 - fixed serious subtle metadata locking bugs
6303 - free locks on clients when inodes are removed due to memory
6305 - fix inode pointer in lock data (285)
6306 - partial support for multiple MDS on a single host (241)
6307 - data locks weren't cancelled at clear_inode time (290, 311)
6308 - intent locks could lead to unbounded lock growth (205)
6309 - added a maximum lock count, an LRU list, and a flusher
6310 - fix multiple rename (365)
6311 - properly abstracted the echo client
6312 - OSC locked 1 byte too many; fixed
6313 - rewrote brw callback code:
6314 - fixed recovery bugs related to LOVs (306)
6315 - fixed too-many-pages-in-one-write crash (191)
6316 - fixed (again) crash in sync_io_timeout (214)
6317 - probably fixed callback-related race (385)
6319 - Add capability to MDS protocol
6320 - LDLM cancellations and callbacks on different portals
6322 2002-10-28 Andreas Dilger <adilger@clusterfs.com>
6325 - limit client IOV size to PTL_MD_MAX_IOV (611336, 191)
6326 - defer open object destruction to close time (601981, 138)
6327 - open/close OST file handle in obdo (OBD_MD_FLHANDLE) (601981, 138)
6328 - move LDLM_ENQUEUE/CONVERT back to MDS portal (625069)
6329 - abstract ll_lookup2, fix ll_revalidate2 to use abstraction (256)
6330 - don't call obd_setattr in ll_file_release for destroyed objects
6331 * protocol change to lustre_msg: move |version| and add |flags|
6332 * protocol change to osc_punch: "start" in "o_size", "end" in "o_blocks"
6333 * lock replay: for LDLM_FL_REPLAY trust client to do right thing
6334 * added replay of create, unlink, link and rename operations during
6335 MDS failover; recovery should be much more robust now
6336 * remove failed OSCs from LOVs (only lov_create uses this so far)
6337 * the lustre-HOWTO was brought (more) up to date (582544)
6339 2002-10-23 Phil Schwan <phil@clusterfs.com>
6342 - in-use dentries weren't being reused properly (617851)
6343 - prevent multiple LDLM setup (599178)
6344 - fix LOV size calculations for truncate (617853)
6345 - fix client handling of MDS intent errors (POSIX)
6346 - fix permission bug in lovstripe.c test (624321)
6347 - fix MDS thread deadlock - move LDLM handler to DLM portal (625069)
6348 - truncate past end of file could corrupt data
6349 - proper cleanup after timeouts, crashes, etc (592524, 550815)
6350 - a race in recovery could return ETIMEDOUT to apps (623947)
6351 - building outside the source directory was fixed
6352 * the lustre-HOWTO was brought (more) up to date (582544)
6353 * major progress was made on recovery functionality
6355 2002-10-10 Phil Schwan <phil@clusterfs.com>
6358 - recovery deadlock fix
6359 - rm -rf causes LBUG fix (617817)
6360 - file open by multiple tasks fix (618962)
6361 - directory permissions bugs (602707 and 620007)
6362 - journal_stop fixed with locking (611313)
6363 - O_APPEND failures resolved (618273, perhaps 614459)
6364 - lconf PATH fix (619770)
6365 - IA64 build fix (621450)
6366 - RPC buffer sizes scale with amount of memory
6368 2002-10-01 Phil Schwan <phil@clusterfs.com>
6371 - locks would be cancelled without throwing away data pages,
6372 resulting in inconsistent data (605627)
6373 - inode attributes were not always being refreshed (605627, 612449)
6374 - lconf now continues to cleanup after lctl reports an error
6375 - MDS now enforces user permissions (602707)
6376 - lprocfs cleanup fixed, but not yet enabled (614157)
6377 - fixed infinite server hang, should a client not respond to an AST
6378 - avoid going into recovery if user calls readlink() with a buffer
6379 that's too small (613941)
6380 - AST RPCs no longer require replies (614867) -- this may be changed
6381 - don't crash server if client sends an IOV that's too big (611336)
6382 - fixed lock conversion deadlock (611892)
6383 - fixed the following of symlinks (614622)
6384 * recovery: the server can remove locks from a client that dies, other
6385 clients can make progress
6386 * more extN patch fixes
6387 * compile-time configurable ptlrpc buffer allocations
6389 - collaborative read cache document
6390 - Lustre Lite Performance CDR document-in-progress
6392 2002-09-20 Andreas Dilger <adilger@clusterfs.com>
6395 - fix typo in patch-2.4.18
6397 2002-09-20 Andreas Dilger <adilger@clusterfs.com>
6400 - clear ptlrpc request each time in handle_incoming_request()
6401 - unlink of files now destroys the object on the OST
6403 2002-09-19 Peter Braam <braam@clusterfs.com>
6405 * add hard link support
6406 * change obdfile creation method
6407 * kernel patch changed
6409 2002-09-19 Peter Braam <braam@clusterfs.com>
6412 - stack overflow bug in extN fixed
6414 2002-09-18 Andreas Dilger <adilger@clusterfs.com>
6416 * documentation updates
6417 - add man pages for config tools
6418 - update tests/README to describe testing with new config tools
6419 - finish metadata API descriptions
6420 * bug fixes and cleanups
6421 - statfs workaround for 16TB limit
6422 - LOV stripe allocation improved, can stripe on subset of OSTs
6423 - LOV file size/IO offset was wrong for files > 4GB in size
6424 - object EA data was being dropped, caused files to be unreadable
6425 - memory overflow with non-LOV OST caused memory corruption
6426 - fixed regression tests to work with new config tools, obdfilter
6427 - fixed bug when directory size became larger than 1 block
6428 - fixed bug (for single client case) when PWD was deleted
6429 - invalidate local directory pages when doing intent-based ops
6430 - avoid LDLM oops when lock callback contained bad data
6432 2002-09-09 Andreas Dilger <adilger@clusterfs.com>
6434 * documentation updates
6435 * bug fixes and cleanups
6436 - configuration tools
6439 - 64-bit compile warnings
6440 - 64-bit internal statfs data
6442 * test_brw on persistent OST devices
6444 * lprocfs (disabled)
6446 2002-09-04 Andreas Dilger <adilger@clusterfs.com>
6448 * documentation updates
6449 * bug fixes and cleanups
6450 * configuration tools
6452 2002-08-30 Peter J. Braam <braam@clusterfs.com>
6455 * many small fixes to 0_5_4
6456 * io/network handling
6457 * thinkos in MDS operations
6459 2002-08-24 Peter J. Braam <braam@clusterfs.com>
6462 * crucial basic fixes to 0.5.3
6463 * IOR, Iozone work over Elan
6466 2002-08-07 Phil Schwan <phil@clusterfs.com>
6467 * version 0_5_3, our first alpha
6468 * we use the new Portals iovs
6469 * documentation updates
6470 * bug fixes and cleanups
6471 * small changes in the DLM wire protocol
6473 2002-07-25 Peter J. Braam <braam@clusterfs.com>
6474 * version 0_5_1 with some initial stability,
6475 * locking on MD and file I/O.
6476 * documentation updates
6477 * several bug fixes since 0.5.0
6478 * small changes in wire protocol
6480 2002-07-18 Phil Schwan <phil@clusterfs.com>
6482 * delivered as Lustre Light Alpha
6483 * fixed a crash after handling invalid MDS requests
6484 * fixed directory pages for architectures with non-4k pages sizes
6486 2002-07-11 Andreas Dilger <adilger@clusterfs.com>
6487 * release version v0_4_4
6488 * Moves TCP acceptor to be on port 2432 (unused Coda port) instead
6490 * Fixes a number of interruption problems with OST operations.
6491 * Update documentation for portals header changes
6492 * Move all wire protocol structs/defines to lustre_idl.h
6493 * Fixes symlink length bug.
6494 * Add tcpdump to repository.
6496 2002-07-05 Andreas Dilger <adilger@clusterfs.com>
6497 * release version v0_4_3
6498 * Fixes statfs for inodes on extN.
6499 * Fixes bug in runtests which would delete /etc/hosts.
6500 * Use 64-bit object IDs wherever possible (not into VFS though)
6501 Remove ost_get_info, which is unused by lustre, and out of date.
6503 2002-07-03 Peter Braam <braam@clusterfs.com>
6504 * release version v0_4_2 Fixes a lookup error (type not passed)
6505 * move forward to head of Portals
6506 * move forward to latest Lustre kernel
6508 2002-06-25 Peter Braam <braam@clusterfs.com>
6509 * release version v0_4_1. Hopefully stable on single node use.