1 tbd Sun Microsystems, Inc.
4 2.6.16.60-0.37 (SLES 10),
5 2.6.27.23-0.1 (SLES11),
6 2.6.18-128.1.14.el5 (RHEL 5)
7 * Client support for unpatched kernels:
8 (see http://wiki.lustre.org/index.php?title=Patchless_Client)
9 2.6.16 - 2.6.27 vanilla (kernel.org)
10 * Recommended e2fsprogs version: 1.41.6.sun1
11 * File join has been disabled in this release, refer to Bugzilla 16929.
12 * A new Lustre ADIO driver is available for MPICH2-1.0.7.
13 * NFS export disabled when stack size < 8192. Since the NFSv4 export
14 of Lustre filesystem with 4K stack may cause a stack overflow. For
15 more information, please refer to bugzilla 17630.
19 Description: Avoid deadlock for local client writes
20 Details : Use new OBD_BRW_MEMALLOC flag to notify OST about writes in the
21 memory freeing context. This allows OST threads to set the
22 PF_MEMALLOC flag on task structures in order to allocate memory
23 from reserved pools and complete IO.
24 Use GFP_HIGHUSER for OST allocations for non-local client writes,
25 so that the OST threads generate memory pressure and allow
26 inactive pages to be reclaimed.
31 Descriptoin: lock ordering violation between &cli->cl_sem and _lprocfs_lock
32 Details : move ldlm namespace creation in setup phase to avoid grab
33 _lprocfs_lock with cli_sem held.
35 Severity : enhancement
37 Description: Update kernel to RHEL5 2.6.18-128.1.14.el5.
39 Severity : enhancement
41 Description: Update kernel to SLES11 2.6.27.23-0.1.
45 Description: Unable to run several mkfs.lustre on loop devices at the same
47 Details : mkfs.lustre returns error 256 on the concurrent loop devices
48 formatting. The solution is to proper handle the error.
50 Severity : enhancement
52 Description: implement async create (obd_async_create) method for osc, to avoid
53 too long waiting new ost objects with holding ldlm lock.
55 Severity : enhancement
57 Descriptoin: OST grant too much space to client even there are not enough space.
58 Details : Client will shrink its grant cache to OST if there are no write
59 activity over 6 mins (GRANT_SHRINK_INTERVAL), and OST will retrieve
60 this grant cache if there are already not enough avaible space
61 (left_space < total_clients * 32M).
64 Frequency : rare, if clients are evicted due to failure to return a lock
66 Description: don't increase ldlm timeout if previous client was evicted
67 Details : if a client doesn't respond to a blocking callback within the
68 adaptive ldlm enqueue timeout, don't adjust the adaptive estimate
69 when the lock is next granted.
73 Description: client not allowed to reconnect to OST because of active request
74 Details : abort bulk requests received by the OST once the client has timed
75 out since the client will resend the request anyway. The client
76 also now retries to reconnect to the same server if a connect
77 request failed with EBUSY or -EAGAIN.
80 Frequency : rare, if used wide striped file and one ost in down.
82 Descriptoin: don't return error if we created a subset of objects for file.
83 Details : lov_update_create_set() uses set->set_success as index for created
84 objects, so if some requests failed, they will have hole at end of
85 array and we can use qos_shrink_lsm for allocate correct lsm.
90 Description: fix lqs' reference which won't be put in some situations
91 Details : This patch fixes:
92 1. In quota_check_common(), this function will check quota
93 for user and group, but only send one return via "pending".
94 In most cases, the pendings should be same. But that is not
96 2. If quotaoff runs between lquota_chkquota() and
97 lquota_pending_commit(), the same thing will happen too.
99 - if (!ll_sb_any_quota_active(qctxt->lqc_sb))
104 Description: Rare Client crash on resend if the file was deleted.
105 Details : When file is opened, but open reply is lost and file is
106 subsequently deleted before resend, resend processing logic
107 breaks trying to open the file again, should not try to open.
111 Description: Rare dentry leakage on MDS on forceful shutdown and
113 Details : Sometimes on transaction errors mds_mfd_close did not release
114 mfd and dentry pointed to it. reconstruct_open did not release
115 parent dentry if no matching mfd was found and child lookup was
120 Description: add check for >8TB ldiskfs filesystems
121 Details : ext3-based ldiskfs does not support greater than 8TB LUNs.
122 Don't allow >8TB ldiskfs filesystems to be mounted without
123 force_over_8tb mount option
125 ------------------------------------------------------------------------------
126 tbd Sun Microsystems, Inc.
128 * Support for kernels:
129 2.6.16.60-0.37 (SLES 10),
130 2.6.27.21-0.1 (SLES11),
131 2.6.18-128.1.6.el5 (RHEL 5)
132 * Client support for unpatched kernels:
133 (see http://wiki.lustre.org/index.php?title=Patchless_Client)
134 2.6.16 - 2.6.27 vanilla (kernel.org)
135 * Recommended e2fsprogs version: 1.41.5-sun2
136 * File join has been disabled in this release, refer to Bugzilla 16929.
137 * A new Lustre ADIO driver is available for MPICH2-1.0.7.
138 * NFS export disabled when stack size < 8192. Since the NFSv4 export
139 of Lustre filesystem with 4K stack may cause a stack overflow. For
140 more information, please refer to bugzilla 17630.
144 Description: resolve race between obd_disconnect and class_disconnect_exports
145 Details : if obd_disconnect will be called to already disconnected export he
146 forget release one reference and osc module can't unloaded.
148 Severity : enhancement
150 Description: move AT tunable parameters for more consistent usage
151 Details : add AT tunables under /proc/sys/lustre, add to conf_param parsing
153 Severity : enhancement
155 Description: Update kernel to RHEL5.3 2.6.18-128.1.6.el5.
157 Severity : enhancement
159 Description: Update kernel to SLES10 SP2 2.6.16.60-0.37.
161 Severity : enhancement
163 Description: Caching OSS
164 Details : introduce data caching on the OSS. The OSS now relies on the linux
165 kernel page cache to keep recently accessed data in memory.
166 It is worth noting that all write requests are still flushed
167 synchronously as in lustre 1.6.
169 Severity : enhancement
171 Description: version based recovery
172 Details : introduce finer grained recovery able to detect transaction
173 dependencies and can deal with transaction gaps.
175 Severity : enhancement
177 Description: Enable adaptive timeouts by default
181 Descriptoin: correctly skip time estimate if in recovery
182 Details : rq_send_state insn't bitmask so using bitwise ops is forbid.
186 Descriptoin: fix goal inodes
187 Details : Allocate inodes for llog in last inode group for avoid broken
192 Descriptoin: OSS DeadLock
193 Details : Use trylock to prevent deadlock when shrink icache.
195 Severity : enhancement
197 Description: Allow tuning service thread via /proc
198 Details : For each service a new
199 /proc/fs/lustre/{service}/*/thread_{min,max,started} entry is
200 created that can be used to set min/max thread counts, and get the
201 current number of running threads.
203 Severity : enhancement
205 Description: Add state history info file, enhance import info file
206 Details : Track import connection state changes in a new osc/mdc proc file;
207 add overview-type data to the osc/mdc import proc file.
211 Description: Reduce small size read RPC
212 Details : Set read-ahead limite for every file and only do read-ahead when
213 available read-ahead pages are bigger than 1M to avoid small size
218 Description: free_entry erroneously used groups_free instead of put_group_info
220 Severity : enhancement
222 Description: Make read-ahead stripe size aligned.
224 Severity : enhancement
226 Description: MDS create should not wait for statfs RPC while holding DLM lock.
228 Severity : enhancement
230 Description: Update client support to vanilla kernels up to 2.6.27.
233 Frequency : rare, connect and disconnect target at same time
235 Descriptoin: ASSERTION(atomic_read(&imp->imp_inflight) == 0
236 Details : Don't call obd_disconnect under lov_lock. this long time
237 operation and can block ptlrpcd which answer to connect request.
239 Severity : enhancement
241 Description: Update to sles9 kernel-2.6.5-7.315.
244 Frequency : start MDS on uncleanly shutdowned MDS device
246 Descriptoin: ll_sync thread stay in waiting mds<>ost recovery finished
247 Details : Stay in waiting mds<>ost recovery finished produce random bugs
248 due race between two ll_sync thread for one lov target. send
249 ACTIVATE event only if connect realy finished and import have
253 Frequency : start MDS on uncleanly shutdowned MDS device
255 Descriptoin: aborting recovery hang on MDS
256 Details : Don't throttle destroy RPCs for the MDT.
258 Severity : enhancement
260 Descriptoin: Don't sync journal after every i/o
261 Details : Implement write RPC replay to allow server replies for write RPCs
262 before data is on disk. However, this feature is disabled by
263 default since some issues leading to data corruptions have been
264 found during recovery (e.g. bug 19128). This feature can be
265 enabled by running the following command on the OSSs:
266 lctl set_param obdfilter.*.sync_journal=0
270 Description: Slow reads beyond 8Tb offsets.
271 Details : Page index integer overflow in ll_read_ahead_page
275 Description: Soft lockup on OSS after MDS failover
276 Details : MSG_CONNECT_INITIAL is not set on the initial MDS->OST connect.
277 As a conseqence, the patch from bug 18224 is not operational
278 and the MDS export cannot be reused on the OSTs until it gets
282 Frequency : rare, only if using MMP with Linux RAID
284 Description: MMP doesn't work with Linux RAID
285 Details : While using HA for Lustre servers with Linux RAID, it is possible
286 that MMP will not detect multiple mounts. To make this work we
287 need to unplug the device queue in RAID when the MMP block is being
288 written. Also while reading the MMP block, we should read it from
289 disk and not the cached one.
292 Frequency : rare, during recovery
294 Description: Assertion failure in ldlm_lock_put
295 Details : Do not put cancelled locks into replay list, hold references on
300 Description: 1.6.5 mdsrate performance is slower than 1.4.11 (MDS not cpu bound)
301 Details : create_count always drops to the min value (=32) because grow_count
302 is being changed before the precreate RPC completes.
305 Frequency : Only in RHEL5 when mounting multiple ext3 filesystems
308 Description: "kmem_cache_create: duplicate cache jbd_4k" error message
309 Details : add proper locking for creation of jbd_4k slab cache
313 Description: MMP check in ext3_remount() fails without displaying any error
314 Details : When multiple mount protection fails during remount, proper error
319 Description: Client locked up when running multiple instances of an app. on
320 multiple mount points
321 Details : ll_shrink_cache() can sleep while holding the ll_sb_lock.
322 Convert ll_sb_lock to a read/write semaphore to fix the problem.
326 Description: Cannot acces an NFS-mounted Lustre filesystem
327 Details : An NFS client cannot access the Lustre filesystem NFS-mounted
328 from a Lustre-client exporting the Lustre filesystem via NFS.
330 -------------------------------------------------------------------------------
331 2008-12-31 Sun Microsystems, Inc.
333 * Support for kernels:
334 2.6.5-7.314 (SLES 9),
335 2.6.9-67.0.22.EL (RHEL 4),
336 2.6.16.60-0.31 (SLES 10),
337 2.6.18-92.1.17.el5 (RHEL 5),
338 2.6.22.14 vanilla (kernel.org)
339 * Client support for unpatched kernels:
340 (see http://wiki.lustre.org/index.php?title=Patchless_Client)
341 2.6.16 - 2.6.22 vanilla (kernel.org)
342 * Client support for unpatched kernels:
343 we do not recommend using patchless RHEL4 clients with kernels
344 prior to 2.6.9-55EL (RHEL4U5).
345 * Recommended e2fsprogs version: 1.40.11-sun1
346 * Note that reiserfs quotas are disabled on SLES 10 in this kernel.
347 * RHEL 4 and RHEL 5/SLES 10 clients behaves differently on 'cd' to a
348 removed cwd "./" (refer to Bugzilla 14399).
349 * A new quota file format has been introduced in 1.6.5.
350 The format conversion from prior releases is handled transparently,
351 but releases older than 1.4.12/1.6.5 don't understand this new
352 format. The automatic format conversion can be avoided by running
353 the following command on the MDS:
354 'tunefs.lustre --param="mdt.quota_type=ug1" $MDTDEV'.
355 For more information, please refer to bugzilla 13904.
356 * A new quota file format was introduced in 1.6.6/1.8.0 (kernel 2.6.16+)
357 The format conversion from prior releases is handled transparently,
358 but releases older than 1.6.6/1.8.0 don't understand this new
359 format. The automatic format conversion can be avoided by running
360 the following commands on the MDS and OSS servers (for
361 pre 1.4.12-1.6.5 quota files):
362 'tunefs.lustre --param="mdt.quota_type=ug1" $MDTDEV',
363 'tunefs.lustre --param="ost.quota_type=ug1" $MDTDEV'
364 or (for 1.4.12/1.6.5 quota files)
365 'tunefs.lustre --param="mdt.quota_type=ug2" $MDTDEV',
366 'tunefs.lustre --param="ost.quota_type=ug2" $MDTDEV'
367 For more information, please refer to bugzilla 13904.
368 * Output of lfs quota has been made less detailed by default,
369 old (verbose) output can be obtained by using -v option.
370 * File join has been disabled in this release, refer to Bugzilla 16929.
371 * A new Lustre ADIO driver is available for MPICH2-1.0.7.
372 * NFS export disabled when stack size < 8192. Since the NFSv4 export of
373 Lustre filesystem with 4K stack may cause a stack overflow. For more
374 information, please refer to bugzilla 17630.
376 Severity : enhancement
378 Description: add lazystatfs mount option to allow statfs(2) to skip down OSTs
379 Details : Allow skip disconnected ost for send statfs request and hide error
383 Frequency : rare, on llog test 6
385 Descriptoin: don't allow connect to already connected import
386 Details : Connect to already connected import hides connection problem.
389 Frequency : rare, on failed llog setup
391 Descriptoin: don't leak obd reference on failed llog setup
392 Details : For failed llog setup - mgc forget call class_destroy_import for
393 client import, move destroy import to more generic place.
398 Descriptoin: allow kill process which wait statahead result
399 Details : For some reasons 'ls' can stick in waiting result from statahead,
400 in this case need way for kill this process.
403 Frequency : rare, at shutdown
405 Descriptoin: panic at umount
406 Details : llap_shrinker can be raced with killing super block from list and
407 this produce panic with access to already freeded pointer
412 Descriptoin: don't lose wakeup for imp_recovery_waitq
413 Details : Recover_import_no_retry or invalidate_import and import_close can
414 both sleep on imp_recovery_waitq, but we was send only one wakeup
420 Descriptoin: panic in mds_open
421 Details : don't confuse mds_finish_transno() with PTR_ERR(-ENOENT)
426 Descriptoin: stuck in cache_remove_extent() or panic with accessing to already
428 Details : Release lock refernce only after add page to pages list.
431 Frequency : always with long access acl
433 Descriptoin: mds can't pack reply with long acl.
434 Details : mds don't control size of acl but they limited by reint/getattr
439 Frequency : on remount
441 Description: external journal device not working after the remount
442 Details : clear dev_rdonly flag for external journal devices in
448 Description: shutdown vs evict race
449 Details : client_disconnect_export vs connect request race.
450 if client will evicted at this time - we start invalidate
451 thread without referece to import and import can be freed
457 Description: shrink LOV EAs before replying
458 Details : correctly adjust LOV EA buffer for reply.
463 Description: don't skip ost target if they assigned to file
464 Details : Drop slow OSCs if we can, but not for requested start idx.
465 This means "if OSC is slow and it is not the requested
466 start OST, then it can be skipped, otherwise skip it only
467 if it is inactive/recovering/out-of-space.
469 Severity : enhancement
471 Description: Update to RHEL5 kernel-2.6.18-92.1.17.el5.
474 Frequency : rare, need acl's on inode.
476 Description: client can't handle ost additional correctly
477 Details : if ost was added after client connected to mds client can have
478 hit lnet_try_match_md ... to big messages to wide striped files.
479 in this case need teach client to handle config events about add
480 lov target and update client max ea size at that event.
482 Severity : enhancement
484 Description: Update to sles9 kernel-2.6.5-7.314.
486 Severity : enhancement
488 Description: Update to SLES10 SP2 kernel-2.6.16.60-0.31.
491 Frequency : Create a symlink file with a very long name
493 Description: ldlm_cancel_pack()) ASSERTION(max >= dlm->lock_count + count)
494 Details : If there is no extra space in the request for early cancels,
495 ldlm_req_handles_avail() returns 0 instead of a negative value.
500 Description: mds is deadlocked
501 Details : In rare cases, inode in catalog can have i_no less than have parent
502 i_no, this produce wrong order for locking during open, and
503 parallel unlink can be lock open. this need teach mds_open to grab
504 locks in resource id order, not at parent -> child order.
506 Severity : enhancement
508 Description: Add /proc entry for import status
509 Details : The mdc, osc, and mgc import directories now have
510 an import directory that contains useful import data for debugging
513 Severity : enhancement
515 Description: Re-disable certain /proc logging
516 Details : Enable and disable client's offset_stats, extents_stats and
517 extents_stats_per_process stats logging on the fly.
520 Frequency : Only on FC kernels 2.6.22+
522 Description: oops in statahead
523 Details : Do not drop reference count for the dentry from VFS when lookup,
524 VFS will do that by itself.
526 Severity : enhancement
528 Description: Generic /proc file permissions
529 Details : Set /Proc file permissions in a more generic way to enable non-
530 root users operate on some /proc files.
534 Description: Hitting mdc_commit_close() ASSERTION
535 Details : Properly handle request reference release in
536 ll_release_openhandle().
540 Frequency : only patchless client
541 Description: add workaround for race between add/remove dentry from hash
543 Severity : enhancement
545 Description: Allow OST glimpses to return PW locks
549 Description: LBUG when llog conf file is full
550 Details : When llog bitmap is full, ENOSPC should be returned for plain
555 Description: Prevent import from entering FULL state when server in recovery
559 Description: service mount cannot take device name with ":"
560 Details : Only when device name contains ":/" will mount treat it as
566 Description: Replace ptlrpcd with the statahead thread to interpret the async
567 statahead RPC callback
571 Frequency : on recovery
572 Description: I/O failures after umount during fail back
573 Details : If client reconnected to restarted server we need join to recovery
574 instead of find server handler is changed and process self eviction
575 with cancel all locks.
577 Severity : enhancement
579 Description: Update to RHEL5 kernel-2.6.18-92.1.10.el5.
583 Description: Kernel BUG tries to release flock
584 Details : Lustre does not destroy flock lock before last reference goes
585 away. So always drop flock locks when client is evicted and
586 perform unlock regardless of successfulness of speaking to MDS.
588 Severity : enhancement
590 Description: Update to SLES10 SP2 kernel-2.6.16.60-0.27.
592 Severity : enhancement
594 Description: Upcall on Lustre log has been dumped
595 Details : Allow for a user mode script to be called once a Lustre log has
596 been dumped. It passes the filename of the dumped log to the
597 script, the location of the script can be specified via
598 /proc/sys/lnet/debug_log_upcall.
603 Description: avoid messages about idr_remove called for id that is not allocated
604 Details : Move assigment s_dev for clustered nfs to end of initialization,
605 for avoid problem with error handling.
610 Description: avoid Already found the key in hash [CONN_UNUSED_HASH] messages
611 Details : When connection is reused this not moved from CONN_UNUSED_HASH into
612 CONN_USED_HASH and this prodice warning when put connection again
618 Description: avoid ASSERTION(client_stat->nid_exp_ref_count == 0) failed
619 Details : Release reference to stats when client disconnected, not
620 when export destroyed for avoid races when client destroyed
621 after main ost export.
625 Description: more cleanup in mds_lov
626 Details : Add workaround for get valid ost count for avoid warnings about
627 drop too big messages, not init llog cat under semphore which
628 can be blocked on reconnect and break normal replay, fix access
631 Severity : enhancement
633 Description: File striping can now be set to use an arbitrary pool of OSTs.
635 Severity : enhancement
637 Description: Export bytes_read/bytes_write count on OSC/OST.
641 Description: Early reply size mismatch, MGC loses connection
642 Details : Apply the MGS_CONNECT_SUPPORTED mask at reconnect time so
643 the connect flags are properly negotiated.
647 Description: Properly propagate oinfo flags from lov to osc for statfs
648 Details : Restore missing copy oi_flags to lov requests.
652 Description: exports in /proc are broken
653 Details : recreate /proc entries for clients when they reconnect.
655 Severity : enhancement
657 Description: Add man pages for llobdstat(8), llstat(8), plot-llstat(8),
658 : l_getgroups(8), lst(8), routerstat(8)
659 Details : Included man pages for llobdstat(8), llstat(8),
660 : plot-llstat(8), l_getgroups(8), lst(8), routerstat(8)
662 Severity : enhancement
664 Description: Implement lustre ll_show_options method.
666 Severity : enhancement
668 Description: Update to SLES9 kernel-2.6.5-7.312.
670 Severity : enhancement
672 Description: Update to RHEL4 kernel-2.6.9-67.0.22.EL.
676 Description: exports in /proc are broken
677 Details : Recreate /proc entries for clients when they reconnect.
681 Description: don't fail open with -ERANGE
682 Details : If client connected until mds will be know about real ost count
683 get LOV EA can be fail because mds not allocate enougth buffer
688 Description: Resolve device initialization race
689 Details : Prevent proc handler from accessing devices added to the
690 obd_devs array but yet be intialized.
692 Severity : enhancement
694 Description: configure's --enable-quota should check the
695 kernel .config for CONFIG_QUOTA
696 Details : Configure is terminated if --enable-quota is passed but
697 no quota support is in kernel
699 Severity : enhancement
701 Description: Update to SLES10 SP2 kernel-2.6.16.60-0.23.
703 Severity : enhancement
705 Description: Update to RHEL5 kernel-2.6.18-92.1.6.el5.
709 Frequency : rare, on PPC clients
710 Description: don't swab ost objects in response about directory, because
712 Details : Bug similar bug 14856, but in different function.
714 Severity : enhancement
716 Description: lfs quota tool enhancement
717 Details : Added units specifiers support for setquota, default to
718 current uid/gid for quota report, short quota stats by
719 default, nonpositional parameters for setquota, added
720 llapi_quotactl manual page.
722 Severity : enhancement
724 Description: *optional* service tags registration
725 Details : If the "service tags" package is installed on a Lustre node
726 When the filesystem is mounted, a local-node service tag will
727 be created. See http://inventory.sun.com/ for more information
728 about the Service Tags asset management system.
732 Description: Client runs out of low memory
733 Details : Consider only lowmem when counting initial number of llap pages
736 Description: Kernel BUG tries to release flock
737 Details : Lustre does not destroy flock lock before last reference goes
738 away. So always drop flock locks when client is evicted and
739 perform unlock regardless of successfulness of speaking to MDS.
742 Frequency : occasional
744 Description: add refcount for osc callbacks, so avoid panic on shutdown
746 Severity : enhancement
748 Description: Update to RHEL4 kernel-2.6.9-67.0.20.
751 Frequency : testing only
753 Description: sanity test 65a fails if stripecount of -1 is set
754 Details : Handle -1 striping on filesystem in ll_dirstripe_verify
757 Frequency : only in unusual configurations
759 Description: Kernel panic with find ost index.
760 Details : lov_obd have panic if some OST's have sparse indexes.
762 Severity : enhancement
764 Description: Update to RHEL5 kernel-2.6.18-53.1.21.el5.
767 Frequency : rarely, if filesystem is mounted with -o flock
769 Description: do not process already freed flock
770 Details : Flock can possibly be freed by another thread before it reaches
771 to ldlm_flock_completion_ast.
774 Frequency : rarely, if filesystem is mounted with -o flock
776 Description: LBUG during stress test
777 Details : Need properly lock accesses the flock deadlock detection list.
780 Frequency : rarely, if binaries are being run from Lustre
782 Description: oops in page fault handler
783 Details : Kernel page fault handler can return two special 'pages' in
784 error case, don't try dereference NOPAGE_SIGBUS and NOPAGE_OMM.
787 Frequency : rarely, during shutdown
789 Description: timeout with invalidate import.
790 Details : ptlrpcd_check call obd_zombie_impexp_cull and wait request which
791 should be handled by ptlrpcd. This produce long age waiting and
792 -ETIMEOUT ptlrpc_invalidate_import and as result LASSERT.
798 Description: ASSERTION(CheckWriteback(page,cmd)) failed
799 Details : Badly clear PG_Writeback bit in ll_ap_completion can produce false
803 Frequency : only with broken builds/installations
805 Description: no LBUG if lquota.ko and fsfilt_ldiskfs.ko are different versions
806 Details : Just return an error to a user, put a console error message.
808 Severity : enhancement
810 Description: Update to RHEL5 kernel-2.6.18-53.1.19.el5.
812 Severity : enhancement
814 Description: Update to RHEL4 kernel-2.6.9-67.0.15.
816 Severity : enhancement
818 Description: enable MGS and MDT services start separately
819 Details : add a 'nomgs' option in mount.lustre to enable start a MDT with
820 a co-located MGS without starting the MGS, which is a complement
821 to 'nosvc' mount option.
824 Frequency : always, on big-endian systems
826 Description: cleanup in ptlrpc code, related to PPC platform
827 Details : store magic in native order avoid panic's in recovery on PPC
828 node and forbid from this error in future. Also fix posibily
829 of twice swab data. Fix get lov striping to userpace.
832 Frequency : rarely, if replay get lost on server
834 Description: server incorrectly drop resent replays lead to recovery failure.
835 Details : do not drop replay according to msg flags, instead we check the
836 per-export recovery request queue for duplication of transno.
839 Frequency : after recovery
841 Description: precreate to many object's after del orphan.
842 Details : del orphan st in oscc last_id == next_id and this triger growing
843 count of precreated objects. Set flag LOW to skip increase count
844 of precreated objects.
847 Frequency : rare, on clear nid stats
849 Description: ASSERTION(client_stat->nid_exp_ref_count == 0)
850 Details : when clean nid stats sometimes try destroy live entry,
851 and this produce panic in free.
854 Frequency : occasionally since 1.6.4
856 Description: Stack overflow during MDS log replay
857 Details : ease stack pressure by using a thread dealing llog_process.
860 Frequency : very rare
862 Description: MDT cannot be unmounted, reporting "Mount still busy"
863 Details : Mountpoint references were being leaked during open reply
864 reconstruction after an MDS restart. Drop mountpoint reference
865 in reconstruct_open() and free dentry reference also.
870 Description: wait until IO finished before start new when do lock cancel.
871 Details : VM protocol want old IO finished before start new, in this case
872 need wait until PG_writeback is cleared until check dirty flag
873 and call writepages in lock cancel callback.
878 Description: mds_mfd_close() ASSERTION(rc == 0)
879 Details : In mds_mfd_close(), we need protect inode's writecount change
880 within its orphan write semaphore to prevent possible races.
883 Frequency : rare, on shutdown ost
885 Description: don't hit live lock with umount ost.
886 Details : shrink_dcache_parent can be in long loop with destroy dentries,
887 use shrink_dcache_sb instead.
890 Frequency : only when echo_client is used
892 Description: don't panic with use echo_client
893 Details : echo client pass NULL as client nid pointer and this produce NULL
897 Frequency : Always on 32-bit PowerPC systems
899 Description: fix build on PPC32
900 Details : compile code with -m64 flag produce wrong object file for PPC32.
905 Description: MDS LBUG: ASSERTION(!IS_ERR(dchild))
906 Details : In reconstruct_* functions, LASSERTs on both the data supplied
907 by a client, and the data on disk are dangerous and incorrect.
908 Change them with client eviction.
910 Severity : enhancement
912 Description: skiplist implementation simplification
913 Details : skiplists are used to group compatible locks on granted list
914 that was implemented as tracking first and last lock of each lock
915 group the patch changes that to using doubly linked lists
919 Description: delete compatibility for 32bit qdata
920 Details : as planned, when lustre is beyond b1_8, lquota won't support 32bit
921 qunit. That means servers of b1_4 and servers of b1_8 can't be
922 used together if users want to use quota.
925 Frequency : only with administrator action
927 Description: mount failure if config log has invalid conf_param setting
928 Details : If administrator specified an incorrect configuration parameter
929 with "lctl conf_param" this would cause an error during future
930 client mounts. Instead, ignore the bad configuration parameter.
933 Frequency : blocks per group < blocksize*8 and uninit_groups is enabled
935 Description: ldiskfs error: XXX blocks in bitmap, YYY in gd
936 Details : If blocks per group is less than blocksize*8, set rest of the
940 Frequency : Application do stride read on lustre
942 Description: The read performance will drop a lot if the application does
944 Details : Because the stride_start_offset are missing in stride read-ahead,
945 it will cause clients read a lot of unused pages in read-ahead,
946 then the read-performance drops.
950 Description: more ldlm soft lockups
951 Details : In ldlm_resource_add_lock(), call to ldlm_resource_dump()
952 starve other threads from the resource lock for a long time in
953 case of long waiting queue, so change the debug level from
954 D_OTHER to the less frequently used D_INFO.
956 Severity : enhancement
958 Description: add -gid, -group, -uid, -user options to lfs find
960 Severity : enhancement
962 Description: ll_recover_lost_found_objs - recover objects in lost+found
963 Details : OST corruption and subsequent e2fsck can leave objects in the
964 lost+found directory. Using the "ll_recover_lost_found_objs"
965 tool, these objects can be retrieved and data can be salvaged
966 by using the object ID saved in the fid EA on each object.
971 Description: this bug _only_ happens when inode quota limitation is very low
972 (less than 12), so that inode quota unit is 1 at initialization.
973 Details : if remaining quota equates 1, it is a sign to demonstate that quota
974 is effective now. So least quota qunit should be 2.
978 Description: Hung threads in invalidate_inode_pages2_range
979 Details : The direct IO path doesn't call check_rpcs to submit a new RPC once
980 one is completed. As a result, some RPCs are stuck in the queue
985 Description: Procfs and llog threads access destoryed import sometimes.
986 Details : Sync the import destoryed process with procfs and llog threads by
987 the import refcount and semaphore.
991 Description: mds fails to respond, threads stuck in ldlm_completion_ast
992 Details : Sort source/child resource pair after updating child resource.
997 Description: kernel BUG at ldiskfs2_ext_new_extent_cb
998 Details : If insertion of an extent fails, then discard the inode
999 preallocation and free data blocks else it can lead to duplicate
1004 Description: don't always update ctime in ext3_xattr_set_handle()
1005 Details : Current xattr code updates inode ctime in ext3_xattr_set_handle()
1006 In some cases the ctime should not be updated, for example for
1007 2.0->1.8 compatibility it is necessary to delete an xattr and it
1008 should not update the ctime.
1012 Description: add quota statistics
1013 Details : 1. sort out quota proc entries and proc code.
1014 2. add quota statistics
1019 Description: quotas are not honored with O_DIRECT
1020 Details : all writes with the flag O_DIRECT will use grants which leads to
1021 this problem. Now using OBD_BRW_SYNC to guard this.
1025 Bugzilla : 15713/16362
1026 Description: Assertion in iopen_connect_dentry in 1.6.3
1027 Details : looking up an inode via iopen with the wrong generation number can
1028 populate the dcache with a disconneced dentry while the inode
1029 number is in the process of being reallocated. This causes an
1030 assertion failure in iopen since the inode's dentry list contains
1031 both a connected and disconnected dentry.
1035 Description: assertion failure in ldlm_handle2lock()
1036 Details : fix a race between class_handle_unhash() and class_handle2object()
1037 introduced in lustre 1.6.5 by bug 13622.
1039 Severity : enhancement
1041 Description: superblock lock contention with many SMP cores on one client
1042 Details : several client filesystem locks were highly contended on SMP
1043 NUMA systems with 8 or more cores. Per-CPU datastructures
1044 and more efficient locking implemented to reduce contention.
1049 Description: Kernel BUG: sd_iostats_bump: unexpected disk index
1050 Details : remove the limit of 256 scsi disks in the sd_iostat patch
1055 Description: oops in sd_iostats_seq_show()
1056 Details : unloading/reloading the scsi low level driver triggers a kernel
1057 bug when trying to access the sd iostat file.
1062 Description: Kernel panics during QLogic driver reload
1063 Details : REQ_BLOCK_PC requests are not handled properly in the sd iostat
1064 patch, causing memory corruption.
1069 Description: journal_dev option does not work in b1_6
1070 Details : pass mount option during pre-mount.
1072 Severity : enhancement
1074 Description: Add a FIEMAP(FIle Extent MAP) ioctl for ldiskfs
1075 Details : FIEMAP ioctl will allow an application to efficiently fetch the
1076 extent information of a file. It can be used to map logical blocks
1077 in a file to physical blocks in the block device.
1080 Frequency : only with adaptive timeout enabled
1082 Description: DEBUG_REQ() bad paging request
1083 Details : ptlrpc_at_recv_early_reply() should not modify req->rq_repmsg
1084 because it can be accessed by reply_in_callback() without the
1088 Frequency : only on Cray X2
1090 Description: X2 build failures
1091 Details : fix build failures on Cray X2.
1095 Description: xid & resent requests
1096 Details : Initialize RPC XID from clock at startup (randomly if clock is
1101 Description: quota recovery deadlock during mds failover
1102 Details : This patch includes att18982, att18236, att18237 in bz14840.
1104 1. fix osts hang when mds does failover with quotaon
1105 2. prevent watchdog storm when osts threads wait for the
1110 Description: kernel panic on racer
1111 Details : Do not access dchild->d_inode when IS_ERR(dchild) is true.
1113 Severity : enhancement
1115 Description: Add lustre_start utility to start or stop multiple Lustre servers
1120 Description: Lustre GPF in {:ptlrpc:ptlrpc_server_free_request+373}
1121 Details : In case of memory pressure, list_del() can be called twice on
1122 req->rq_history_list, causing a kernel oops.
1126 Description: (ptllnd_peer.c:557:kptllnd_peer_check_sends()) ASSERTION(!in_interrupt()) failed
1127 Details : fix stack overflow in the distributed lock manager by defering
1128 export eviction after a failed ast to the elt thread instead of
1129 handling it in the dlm interpret routine.
1131 Severity : enhancement
1133 Description: More exported tunables for mballoc
1134 Details : Add support for tunable preallocation window and new tunables for
1135 large/small requests
1139 Description: Detect corruption of block bitmap and checking for preallocations
1140 Details : Checks validity of on-disk block bitmap. Also it does better
1141 checking of number of applied preallocations. When corruption is
1142 found, it turns filesystem readonly to prevent further corruptions.
1146 Frequency : only for big-endian servers
1147 Description: Check for big-endian system while mounting fs with extents feature
1148 Details : Mounting a filesystem with extents feature will fail on big-endian
1149 systems since ext3-based ldiskfs is not supported on big-endian
1150 systems. Can be over-ridden with "bigendian_extents" mount option.
1154 Description: Excessive recovery window
1155 Details : With AT enabled, the recovery window can be excessively long (6000+
1156 seconds). To address this problem, we no longer use
1157 OBD_RECOVERY_FACTOR when extending the recovery window (the connect
1158 timeout no longer depends on the service time, it is set to
1159 INITIAL_CONNECT_TIMEOUT now) and clients report the old service
1160 time via pb_service_time.
1164 Description: Watchdog triggered on MDS failover
1165 Details : enable OBD_CONNECT_MDT flag when connecting from the MDS so that
1166 the OSTs know that the MDS "UUID" can be reused for the same export
1167 from a different NID, so we do not need to wait for the export to
1172 Descriptoin: Lustre detected file system corruption with inode out of bounds
1173 Details : don't update i_size on MDS_CLOSE for directories. This causes
1174 directory corruptions on the MDT.
1176 -------------------------------------------------------------------------------
1178 2008-05-26 Sun Microsystems, Inc.
1180 * Support for kernels:
1181 2.6.5-7.311 (SLES 9),
1182 2.6.9-67.0.7.EL (RHEL 4),
1183 2.6.16.54-0.2.5 (SLES 10),
1184 2.6.18-53.1.14.el5 (RHEL 5),
1185 2.6.22.14 vanilla (kernel.org)
1186 * Client support for unpatched kernels:
1187 (see http://wiki.lustre.org/index.php?title=Patchless_Client)
1188 2.6.16 - 2.6.22 vanilla (kernel.org)
1189 * Due to problems with nested symlinks and FMODE_EXEC (bug 12652),
1190 we do not recommend using patchless RHEL4 clients with kernels
1191 prior to 2.6.9-55EL (RHEL4U5).
1192 * Recommended e2fsprogs version: 1.40.7-sun1
1193 * Note that reiserfs quotas are disabled on SLES 10 in this kernel.
1194 * RHEL 4 and RHEL 5/SLES 10 clients behaves differently on 'cd' to a
1195 removed cwd "./" (refer to Bugzilla 14399).
1196 * A new quota file format has been introduced in 1.6.5.
1197 The format conversion from prior releases is handled transparently,
1198 but releases older than 1.4.12/1.6.5 will not understand this new
1199 format. The automatic format conversion can be avoided by running
1200 the following command on the MDS before upgrading:
1201 'tunefs.lustre --param="mdt.quota_type=ug1" $MDTDEV'.
1202 For more information, please refer to bugzilla 13904.
1206 Description: quota performance fix
1207 Details : quota data is written in journalled mode instead of ordered to
1208 increase performance
1212 Description: lfs support for human-readable quota grace time strings
1213 Details : lfs setquota -t and lfs quota -t represent quota grace times
1214 in "XXwXXdXXhXXmXXs" format instead of large values in seconds
1217 Frequency : always with o2ib 1.3 and sles10
1219 Description: fix build with SLES10 and o2ib v3.
1220 Details : sles10 uses diffrent name for Module.symver file but configure
1221 assume this file has same name on RHEL/SLES/vanila kernels.
1224 Frequency : very rare, if additional xattrs are used on kernels >= 2.6.12
1226 Description: MDS may lose file striping (and hence file data) in some cases
1227 Details : If there are additional extended attributes stored on the MDS,
1228 in particular ACLs, SELinux, or user attributes (if user_xattr
1229 is specified for the client mount options) then there is a risk
1230 of attribute loss. Additionally, the Lustre file striping
1231 needs to be larger than default (e.g. striped over all OSTs),
1232 and an additional attribute must be stored initially in the
1233 inode and then increase in size enough to be moved to the
1234 external attribute block (e.g. ACL growing in size) for file
1237 Severity : enhancement
1239 Description: add message levels for liblustreapi
1244 Description: MDT cannot be unmounted, reporting "Mount still busy"
1245 Details : Mountpoint references were being leaked during open reply
1246 reconstruction after an MDS restart. Drop mountpoint reference
1247 in reconstruct_open() and free dentry reference also.
1252 Description: fix for occasional failure case of -ENOSPC in recovery-small tests
1253 Details : Move the 'good_osts' check before the 'total_bavail' check. This
1254 will result in an -EAGAIN and in the exit call path we call
1255 alloc_rr() which will with increasing aggressiveness attempt to
1256 aquire precreated objects on the minimum number of required OSCs.
1260 Description: Use old size assignment to avoid deadlock
1261 Details : This reverts the changes in bugs 2369 and bug 14138 that introduced
1262 the scheduling while holding a spinlock. We do not need locking
1263 for size in ll_update_inode() because size is only updated from
1264 the MDS for directories or files without objects, so there is no
1265 other place to do the update, and concurrent access to such inodes
1266 are protected by the inode lock.
1270 Description: Use __u64 instead of int for valid bits
1274 Description: resolve "_IOWR redefined" build error on SLES10
1278 Description: dump the memory debugging after all modules are unloaded to
1279 suppress false negative in conf_sanity test 39
1283 Description: the recovery timer never expires
1284 Details : for new client connect request, the recovery timer should not be
1285 reset, otherwise recovery timer will never expired, if the old
1286 client never come. Only old client connect and first connection
1287 req should trigger recovery timer reset.
1291 Description: the min numbers of lproc stats are wrong
1292 Details : adding a new constant LC_MIN_INIT and use it for initialization
1296 Frequency : always with interactive lfs
1298 Description: Reinitialize optind to 0 so that interactive lfs works in all cases
1301 Frequency : with multiple concurrent readdir processes in same directory
1302 Bugzilla : 15406, 15169, 15175
1303 Description: misc fixes for directory readahead.
1304 Details : prevent previous statahead async RPC callback from processing the
1305 current "statahead_info", race condition between async RPC callback
1306 add dentry into dentry hash table and "ls" thread revalidate such
1307 dentry, statahead his/miss control for hidden items, and so on.
1309 Severity : enhancement
1311 Description: build kernel-ib packages for OFED 1.3 in our release cycle
1315 Description: incore types cleaning in quota code (with respect to 64-bit limits)
1316 Details : several u32 variables declarations are replaced with u64 declarations
1321 Description: fix SLES kernel versioning
1322 Details : the kernel version for our SLES 10 kernel did not include a "-"
1323 before the "smp" at the end. while this was not a problem in
1324 general, it did mean that software trying to use the kernel
1325 version to try to detect a vendor specific kernel would fail.
1326 this was most evident by the OFED build scripts.
1331 Description: Don't update lov_desc members until making sure they are valid
1332 Details : When updating lov_desc members via proc fs, need fix their
1333 validities before doing the real update.
1336 Frequency : very rare
1338 Description: don't put request into delay list while invalidate in flight.
1339 Details : ptlrpc_delay_request sometimes put in delay list while invalidate
1340 import in flight. this produce timeout for invalidate and sometimes
1341 can cause stale data.
1343 Severity : enhancement
1345 Description: Update kernel to SLES9 2.6.5-7.311.
1347 Severity : enhancement
1349 Description: Update kernel to RHEL4 2.6.9-67.0.7.
1354 Frequency : on PPC only
1355 Description: not convert ost objects for directory because it's not exist.
1356 Details : ll_dir_getstripe assume dirrectory has ost objects but this wrong.
1358 Severity : enhancement
1360 Description: Fix warnings with compile liblustre at sles10/rhel5 which have
1361 __u64 as usingied long long type.
1364 Frequency : rare, on shutdown
1366 Description: race process ast vs remove callback
1367 Details : removing callback before disconnect import open race with
1368 processing callback.
1370 Severity : enhancement
1372 Description: Update kernel to SLES9 2.6.5-7.311.
1374 Severity : enhancement
1376 Description: Files open for execute are not marked busy on SLES10
1377 Details : Add FMODE_EXEC to SLES10 SP1 server kernel series.
1379 Severity : enhancement
1381 Description: Add server support for vanilla-2.6.22.14.
1384 Frequency : occasional
1386 Description: Avoid lov_create() getting stuck in obd_statfs_rqset()
1387 Details : If an OST is down the MDS will hang indefinitely in
1388 obd_statfs_rqset() waiting for the statfs data. While for
1389 MDS QOS usage of statfs, it should not stuck in waiting.
1391 Severity : enhancement
1393 Description: Disable adaptive timeouts by default
1396 Frequency : on network error
1398 Description: panic with double free request if network error
1399 Details : mdc_finish_enqueue is finish request if any network error occuring,
1400 but it's true only for synchronus enqueue, for async enqueue
1401 (via ptlrpcd) this incorrect and ptlrpcd want finish request
1405 Frequency : rare, on recovery
1407 Description: read procfs can produce deadlock in some situation
1408 Details : Holding lprocfs lock with send rpc can produce block for destroy
1409 obd objects and this also block reconnect with -EALREADY.
1410 This isn't fix all lprocfs bugs - but make it rare.
1412 Severity : enhancement
1414 Description: Update kernel to RHEL5 2.6.18-53.1.14.el5.
1417 Frequency : frequent on X2 node
1419 Description: mdc_set_open_replay_data LBUG
1420 Details : Set replay data for requests that are eligible for replay.
1425 Description: lustre_mgs: operation 101 on unconnected MGS
1426 Details : When MGC is disconnected from MGS long enough, MGS will evict the
1427 MGC, and late on MGC cannot successfully connect to MGS and a lot
1428 of the error messages complaining that MGS is not connected.
1431 Frequency : rare, depends on device drivers and load
1433 Description: MDS or OSS nodes crash due to stack overflow
1434 Details : Code changes in 1.6.4 increased the stack usage of some functions.
1435 In some cases, in conjunction with device drivers that use a lot
1436 of stack, the MDS (or possibly OSS) service threads could overflow
1437 the stack. One change which was identified to consume additional
1438 stack has been reworked to avoid the extra stack usage.
1440 Severity : enhancement
1442 Description: Update to RHEL5 latest kernel-2.6.18-53.1.13.el5.
1444 Severity : enhancement
1446 Description: Update to SLES10 SP1 latest kernel-2.6.16.54-0.2.5.
1448 Severity : enhancement
1450 Description: Update to RHEL5 latest kernel-2.6.18-53.1.6.el5.
1452 Serverity : enhancement
1454 Description: Update RHEL4 kernel to 2.6.9-67.0.4.
1457 Frequency : rare on shutdown OST
1459 Description: Don't allow skipping OSTs if index has been specified.
1460 Details : Don't allow skipping OSTs if index has been specified, make locking
1461 in internal create lots better.
1466 Description: ASSERTION(!PageDirty(page)) failed
1467 Details : Wrong check could lead to an assertion failure under specific
1473 Description: LBUG in ptlrpc_check_set() bad phase ebc0de00
1474 Details : access to bitfield in structure is always rounded to long
1475 and this produce problem with not atomic change any bit.
1480 Description: Lustre make rpms failed.
1481 Details : Remove ldiskfs spec file to avoids rpmbuild be confused when
1482 builds Lustre rpms from tarball.
1484 Severity : enhancement
1486 Description: Update to SLES9 SP4 kernel-2.6.5-7.308.
1489 Frequency : rare on shutdown OST
1491 Description: If llog cancel was not send before clean_exports phase, this can
1492 produce deadlock in llog code.
1493 Details : If llog thread has last reference to obd and call class_import_put
1494 this produce deadlock because llog_cleanup_commit_master wait when
1495 last llog_commit_thread exited, but this never success because was
1496 called from llog_commit_thread.
1499 Frequency : only if OST index is skipped
1501 Description: NULL lov_tgts causing MDS oops
1502 Details : more safe checks for NULL lov_tgts for avoid oops.
1504 Severity : enhancement
1506 Description: Update to RHEL4 latest kernel-2.6.9-67.0.1.EL.
1508 Severity : enhancement
1510 Description: Update to RHEL5 latest kernel-2.6.18-53.1.4.el5.
1515 Description: make mgs_setparam() handle fsname containing dash
1516 Details : fsname containing a dash does not work with lctl conf_param
1518 Severity : enhancement
1520 Description: Update to RHEL4 Update-6 kernel-2.6.9-67.EL.
1523 Frequency : rare, in recovery and (or) destroy lovobjid file.
1525 Description: rewrite lov objid code.
1526 Details : Cleanup for lov objid code, remove scability problems and wrong
1527 locking. Fix sending last_id into ost.
1529 Severity : enhancement
1531 Description: Update to SLES10 SP1 latest kernel-2.6.16.54-0.2.3.
1533 Severity : enhancement
1535 Description: Update to RHEL5 Update-1 kernel 2.6.18-53.el5.
1536 Details : Use d_move_locked instead of __d_move.
1539 Frequency : rare, at shutdown
1541 Description: access already free / zero obd_namespace.
1542 Details : if client_disconnect_export was called without force flag set,
1543 and exist connect request in flight, this can produce access to
1544 NULL pointer (or already free pointer) when connect_interpret
1545 store ocd flags in obd_namespace.
1548 Frequency : only at startup
1550 Description: not alloc memory with spinlock held.
1551 Details : allocation memory with GFP_KERNEL can produce sleep deadlock,
1552 if any spinlock held.
1557 Description: lfs find does not continue on file error
1558 Details : Continue other files processing when a file/dir is absent.
1562 Description: Inconsistent usage of lustre_pack_reply()
1563 Details : Standardize the usage of lustre_pack_reply() such that it
1564 always generate a CERROR on failure.
1567 Frequency : very rare
1569 Description: Fix replay if there is an un-replied request and open
1570 Details : In some cases, older replay request will revert the
1571 mcd->mcd_last_xid on MDS which is used to record the client's
1572 latest sent request.
1574 Severity : enhancement
1576 Description: Update to RHEL5 kernel 2.6.18-8.1.15.el5.
1578 Severity : enhancement
1580 Description: Update to SLES10 SP1 kernel 2.6.16.53-0.16
1582 Severity : enhancement
1584 Description: Update to SLES9 kernel-2.6.5-7.287.3.
1586 Severity : enhancement
1588 Description: Update to RHEL4 kernel-2.6.9-55.0.12.EL.
1590 Severity : enhancement
1592 Description: Build SLES10 patchless client fails
1593 Details : The configure was broken by run ./configure with
1594 --with-linux-obj=.... argument for patchless client. When the
1595 configure use --with-linux-obj, the LINUXINCLUDE= -Iinclude
1596 can't search header adequately. Use absolute path such as
1597 -I($LINUX)/include instead.
1599 Severity : enhancement
1601 Description: Lustre Page Accounting
1602 Details : New macros for page alloc and free which enable accounting
1603 of page allocation of Lustre. Use percpu counters to store memory
1604 and page statistics.
1607 Frequency : only if debugging is disabled
1609 Description: LASSERT_{REQ,REP}SWAB macros are buggy
1610 Details : If SWAB_PARANOIA is disabled, the LASSERT_REQSWAB and
1611 LASSERT_REPSWAB macros become no-ops, which is incorrect. Drop
1612 these macros and replace them with their definitions instead.
1617 Description: interrupt oig_wait produce painc on resend.
1618 Details : brw_redo_request can be used for resend requests from ptlrpcd and
1619 private set, and this produce situation when rq_ptlrpcd_data not
1620 copyed to new allocated request and triggered LBUG on assert
1621 req->rq_ptlrpcd_data != NULL. But this member used only for wakeup
1622 ptlrpcd set if request is changed and can be safety changed to use
1625 Severity : enhancement
1627 Description: organize the server-side client stats on per-nid basis
1628 Details : Change the structure of stats under obdfilter and mds to
1636 The "uuid"s file would list the uuids of _active_ exports.
1637 And the clear entry is to clear all stats and stale nids.
1642 Description: Processes looping in ll_readdir() on Lustre clients finally causing
1643 a full node pseudo-hang
1644 Details : Concurrent access to the same directory from multiple clients with
1645 intensive file creation/removal can cause a client node to spin in
1646 ll_readdir(). i_version must be increased every time the lock is
1647 cancelled to ensure a revalidate is done.
1652 Description: touch file failed when fs is not full
1653 Details : OST in recovery should not be discarded by MDS in alloc_qos(),
1654 otherwise we can get ENOSP while fs is not full.
1657 Frequency : only for Cray XT3
1658 Bugzilla : 12829/13455
1659 Description: Changing primary group doesn't change the group lustre assigns to
1661 Details : When CRAY_XT3 is defined, the fsgid supplied by the client is
1662 overridden with the primary group provided by the group upcall,
1663 whereas the supplied fsgid can be trusted if it is in the list of
1664 supplementary groups returned by the group upcall.
1666 Severity : enhancement
1668 Description: Root Squash Functionality
1669 Details : Implementation of NFS-like root squash capability. Specifically,
1670 don't allow someone with root access on a client node to be able
1671 to manipulate files owned by root on a server node.
1673 Severity : enhancement
1675 Description: Slow trucate/writes to huge files at high offsets.
1676 Details : Directly associate cached pages to lock that protect those pages,
1677 this allows us to quickly find what pages to write and remove
1678 once lock callback is received.
1683 Description: Too many locks accumulating on client during NFS usage
1684 Details : mds_open improperly used accmode to find out access mode to a
1685 file. Also mdc_intent_lock logic to find out if we already have
1686 lock similar to just received was flawed since introduction of
1687 skiplists - locks are now added to the front of the granted
1692 Description: Hit ASSERTION(obd->obd_stopping == 1) failed in some setup failed
1694 Details : In obd setup failure handler, obd_stopping will not necessarily to
1695 be 1, and obd_set_up should also be checked to make sure whether
1696 obd is completely setup.
1698 Severity : enhancement
1700 Description: Allow masking D_WARNING, D_ERROR messages from console
1701 Details : Console messages can now be disabled via lnet.printk.
1706 Description: User code with malformed file open parameter crashes client node
1707 Details : Before packing join_file req, all the related reference should be
1708 checked carefully in case some malformed flags cause fake join_file
1714 Description: shrink/enlarge qunit size when needed; fix the problem of coarse
1715 grain of quota doing harm to quota's accuracy
1716 Details : qunit size will be changed when quota limitation is too low/high;
1717 record the pending quota write in order to get more accureate
1718 quota; delete the patch for bug12588, which is unnecessary when
1719 this patch is landed. This bug also contains fixes for bug 14526,
1720 14299, 14601 and 13794.
1724 Description: LDLM_ENQUEUE races with LDLM_CP_CALLBACK
1725 Details : ldlm_completion_ast() assumes that a lock is granted when the req
1726 mode is equal to the granted mode. However, it should also check
1727 that LDLM_FL_CP_REQD is not set.
1731 Description: Heavy nfs access might result in deadlocks
1732 Details : After ELC code landed, it is now improper to enqueue any mds
1733 locks under och_sem, because enqueue might want to decide to
1734 cancel open locks for same inode we are holding och_sem for.
1738 Description: 35% write performance drop with ldiskfs2 when quotas are on
1739 Details : Enable ext3 journalled quota by default to improve performance
1740 when quotas are turned on.
1744 Description: Client eviction while running blogbench
1745 Details : A lot of unlink operations with concurrent I/O can lead to a
1746 deadlock causing evictions. To address the problem, the number of
1747 oustanding OST_DESTROY requests is now throttled to
1748 max_rpcs_in_flight per OSC and LDLM_FL_DISCARD_DATA blocking
1749 callbacks are processed in priority.
1752 Frequency : RHEL4 only
1754 Description: mkfs is very slow on IA64/RHEL4
1755 Details : A performance regression has been discovered in the MPT Fusion
1756 driver between versions 3.02.73rh and 3.02.99.00rh. As a
1757 consequence, we have downgraded the MPT Fusion driver in the RHEL4
1758 kernel from 3.02.99.00 to 3.02.73 until this problem is fixed.
1761 Frequency : PPC/PPC64 only
1763 Description: conflicts between asm-ppc64/types.h and lustre_types.h
1764 Details : fix duplicated definitions between asm-ppc64/types.h and
1765 lustre_types.h on PPC.
1768 Frequency : PPC/PPC64 only
1770 Description: asm-ppc/segment.h does not exist
1771 Details : fix compile issue on PPC.
1775 Description: data checksumming impacts single node performance
1776 Details : add support for several checksum algorithms. Currently, CRC32 and
1777 Adler-32 are supported. The checksum type can be changed on the fly
1778 through /proc/fs/lustre/osc/*/checksum_type.
1782 Description: use adler32 for page checksums
1783 Details : when available, use the Adler-32 algorithm instead of CRC32 for
1788 Description: better handle error messages in extents code
1790 Severity : enhancement
1792 Description: SNMP support enhancement
1793 Details : Adding total number of sampled request for an MDS node in snmp
1796 Severity : enhancement
1798 Description: Optimize ldlm waiting list processing for PR extent locks
1799 Details : When processing waiting list for read extent lock and meeting read
1800 lock that is same or wider to it that is not contended, skip
1801 processing rest of the list and immediatelly return current
1802 status of conflictness, since we are guaranteed there are no
1803 conflicting locks in the rest of the list.
1807 Description: Time out and refuse to reconnect
1808 Details : When the failover node is the primary node, it is possible
1809 to have two identical connections in imp_conn_list. We must
1810 compare not conn's pointers but NIDs, otherwise we
1811 can defeat connection throttling.
1815 Description: Client not clear own cache if answer to reconnect is lost.
1816 Details : Client gets evicted from server. Now client also thinks it is
1817 disconnected (or gets ENOTCONN on its operation) and decides to
1818 reconnect. Server receives reconnect message, but cannot find
1819 export. New export is created that is fully valid (new cookie!),
1820 but reply is lost and not reported to client. Client reconnects
1821 again and gets back a just-created connection, but it is not new
1822 so client thinks it was not evicted and does not do recovery.
1826 Description: Detect stride IO mode in read-ahead
1827 Details : When a client does stride read, read-ahead should detect that and
1828 read-ahead pages according to the detected stride pattern.
1832 Description: build for x2 fails
1833 Details : fix compile issue on Cray systems.
1835 Severity : enhancement
1837 Description: implement readv/writev APIs(aio_read/aio_writes in newer kernels)
1838 Details : This greatly improves speed of NFS writes on 2.6 kernels.
1841 Frequency : only on PPC/SLES10
1843 Description: "BITS_PER_LONG is not 32 or 64" in linux/idr.h
1844 Details : On SLES10/PPC, fs.h includes idr.h which requires BITS_PER_LONG to
1845 be defined. Add a hack in mkfs_lustre.c to work around this compile
1850 Description: LASSERT on MDS when client holding flock lock dies
1851 Details : ldlm pool logic depends on number of granted locks equal to
1852 number of released locks which is not true for flock locks, so
1853 just exclude such locks from consideration.
1857 Description: MDS deadlock with many ll_sync_lov threads and I/O stalled
1858 Details : Use fsfilt_sync() for both the whole filesystem sync and
1859 individual file sync to eliminate dangerous inode locking
1860 with I_LOCK that can lead to a deadlock.
1864 Description: Update an obsolete wirecheck.c generator
1865 Details : Update wirecheck.c/wirehdr.c and regenerate wiretest.c
1869 Description: Client can panic on open sometimes
1870 Details : It is possible that we try to free already freed request in
1871 ll_file_open in some error cases when we send request from
1876 Description: performance in 1.6.3
1877 Details : Force q->max_phys_segments to MAX_PHYS_SEGMENTS on SLES10 to be
1878 sure that 1MB requests are not fragmented by the block layer.
1882 Description: LDLM soft lockups - improvement
1883 Details : It is be possible to send the lock handle along with each read
1884 or write request because the client is already doing a lock match
1885 itself so there isn't any reason the OST should have to re-do that
1891 Description: lfs quota fails with deactivated OSTS
1892 Details : With this patch, three improvements are included:
1893 1. detete the softlimit in mds and osts when use "lfs quota".
1894 2. display the inaccurate data in the output of "lfs quota".
1895 3. try to get quota info when "lfs quota" is executed.
1900 Description: Extent locks not granted with no conflicts sometimes.
1901 Details : When race occurs in glimpse handler and nothing is returned,
1902 we do not reprocess the queue after lock cancel, and that leads
1903 to a stall until next activity on a resource
1906 Frequency : failover with quotaon
1908 Description: during mds failovers with quota on, OSTs got into deadlock state
1909 and causing dumpstack.
1910 Details : for every quota slave, at any time, there is only one quota req
1911 is sent to quota master for every uid/gid. Before that quota req
1912 returns, all the thread relative to the same uid/gid will wait.
1913 So if the quota req is lost because mds failovers or any other
1914 reasons, this bug will be hit. Now, dqacq_interpret() will handle
1915 quota reqs who time out.
1917 Severity : enhancement
1920 Description: when quota slave checks if quota is enough, there is an unnecessary
1922 Details : place this wait on necessary place instead of always waiting.
1924 --------------------------------------------------------------------------------
1926 2007-12-07 Cluster File Systems, Inc. <info@clusterfs.com>
1928 * Support for kernels:
1929 2.6.5-7.286 (SLES 9),
1930 2.6.9-55.0.9.EL (RHEL 4),
1931 2.6.16.53-0.8 (SLES 10),
1932 2.6.18-8.1.14.el5 (RHEL 5),
1933 2.6.18.8 vanilla (kernel.org)
1934 * Client support for unpatched kernels:
1935 (see http://wiki.lustre.org/index.php?title=Patchless_Client)
1936 2.6.16 - 2.6.22 vanilla (kernel.org)
1937 * Due to recently discovered recovery problems, we do not recommend
1938 using patchless RHEL 4 clients with this or any earlier release.
1939 * Recommended e2fsprogs version: 1.40.2-cfs1
1940 * Note that reiserfs quotas are disabled on SLES 10 in this kernel.
1943 Frequency : occasional
1945 Description: MDS hang or stay in waiting lock
1946 Details : If client receive lock with CBPENDING flag ldlm need send lock
1947 cancel as separate rpc, to avoid situation when cancel request
1948 can't processed due all i/o threads stay in wait lock.
1951 Frequency : occasional
1953 Description: Do not fail import if osc_interpret_create gets -EAGAIN
1954 Details : If osc_interpret_create got -EAGAIN it immediately exits and
1955 wakeup oscc_waitq. After wakeup oscc_wait_for_objects call
1956 oscc_has_objects and see OSC has no objests and call
1957 oscc_internal_create to resend create request.
1959 Severity : enhancement
1961 Description: Update kernel patches for SLES10 2.6.16.53-0.8.
1962 Details : Update which_patch & target file for SLES10 latest kernel.
1964 Severity : enhancement
1966 Description: add --type and --size parameters to lfs find
1967 Details : Enhance lfs find by adding filetype and filesize parameters. Also
1968 multiple OBDs can now be specified for the --obd option.
1970 Severity : enhancement
1972 Description: eliminate client locks in face of contention
1973 Details : file contention detection and lockless i/o implementation
1974 for contended files.
1976 Severity : enhancement
1978 Description: Remove client patches from SLES 10 kernel.
1979 Details : This causes SLES 10 clients to behave as patchless clients
1980 even on a Lustre-patched (server) kernel.
1982 Severity : enhancement
1984 Description: use i_size_read and i_size_write in 2.6 port
1985 Details : replace inode->i_size access with i_size_read/write()
1988 Frequency : when removing large files
1990 Description: scheduling issue during removal of large Lustre files
1991 Details : Don't take the BKL in fsfilt_ext3_setattr() for 2.6 kernels.
1992 It causes scheduling issues when removing large files (17TB in the
1998 Description: 1.4.11 Can't handle directories with stripe set and extended ACLs
1999 Details : Impossible (EPROTO is returned) to access a directory that has a
2000 non-default striping and ACLs.
2003 Frequency : only on ppc
2005 Description: /proc/fs/lustre/devices broken on ppc
2006 Details : The patch as applied to 1.6.2 doesn't look correct for all arches.
2007 We should make sure the type of 'index' is loff_t and then cast
2008 explicitly as needed below. Do not assign an explicitly cast
2012 Frequency : only for rhel5
2014 Description: Kernel patches update for RHEL5 2.6.18-8.1.10.el5.
2015 Details : Modify the target file & which_kernel.
2018 Frequency : if the uninit_groups feature is enabled on ldiskfs
2020 Description: e2fsck reports "invalid unused inodes count"
2021 Details : If a new ldiskfs filesystem is created with the "uninit_groups"
2022 feature and only a single inode is created in a group then the
2023 "bg_unused_inodes" count is incorrectly updated. Creating a
2024 second inode in that group would update it correctly.
2029 Description: buffer overruns could theoretically occur
2030 Details : llapi_semantic_traverse() modifies the "path" argument by
2031 appending values to the end of the origin string, and a buffer
2032 overrun may occur. Adding buffer overrun check in liblustreapi.
2034 Severity : enhancement
2036 Description: Add jbd statistics patch for RHEL5 and 2.6.18-vanilla.
2039 Frequency : only if filesystem is inconsistent
2041 Description: handle "serious error: objid * already exists" more gracefully
2042 Details : If LAST_ID value on disk is smaller than the objects existing in
2043 the O/0/d* directories, it indicates disk corruption and causes an
2044 LBUG(). If the object is 0-length, then we should use the existing
2045 object. This will help to avoid a full fsck in most cases.
2047 Severity : enhancement
2049 Description: Kernel patches update for RHEL4 2.6.9-55.0.6.
2050 Details : Modify vm-tunables-rhel4.patch.
2052 Severity : enhancement
2054 Description: Kernel config for 2.6.18-vanilla.
2055 Details : Modify targets/2.6-vanilla.target.in.
2056 Add config file kernel-2.6.18-2.6-vanilla-i686.config.
2057 Add config file kernel-2.6.18-2.6-vanilla-i686-smp.config.
2058 Add config file kernel-2.6.18-2.6-vanilla-x86_64.config.
2059 Add config file kernel-2.6.18-2.6-vanilla-x86_64-smp.config.
2062 Frequency : occasional
2064 Description: improve handling recoverable errors
2065 Details : If request processed with error which can be recoverable on server
2066 request should be resend, otherwise page released from cache and
2072 Description: Kernel patches update for RHEL5 2.6.18-8.1.14.el5.
2073 Details : Modify target file & which_patch.
2074 A flaw was found in the IA32 system call emulation provided
2075 on AMD64 and Intel 64 platforms. An improperly validated 64-bit
2076 value could be stored in the %RAX register, which could trigger an
2077 out-of-bounds system call table access. An untrusted local user
2078 could exploit this flaw to run code in the kernel
2079 (ie a root privilege escalation). (CVE-2007-4573).
2083 Description: change order of libsysio includes
2084 Details : '#include sysio.h' should always come before '#include xtio.h'
2086 Severity : enhancement
2088 Description: adapt the lustre_config script to support the upgrade case
2089 Details : Add "-u" option for lustre_config script to support upgrading 1.4
2090 server targets to 1.6 in parallel.
2095 Description: To avoid grant space > avaible space when the disk is almost
2096 full. Without this patch you might see the error "grant XXXX >
2097 available" or some LBUG about grant, when the disk is almost
2099 Details : In filter_check_grant, for non_grant cache write, we should
2100 check the left space by if (*left > ungranted + bytes), instead
2101 of (*left > ungranted), because only we are sure the left space
2102 is enough for another "bytes", then the ungrant space should be
2103 increase. In client, we should update cl_avail_grant only there is
2104 OBD_MD_FLGRANT in the reply.
2109 Description: Update RHEL 4 kernel to fix local root privilege escalation.
2110 Details : Update to the latest RHEL 4 kernel to fix the vulnerability
2111 described in CVE-2007-4573. This problem could allow untrusted
2112 local users to gain root access.
2115 Frequency : when using O_DIRECT and quotas
2117 Description: Incorrect file ownership on O_DIRECT output files
2118 Details : block usage reported by 'lfs quota' does not take into account
2119 files that have been written with O_DIRECT.
2123 Description: (rw.c:1323:ll_read_ahead_pages()) ASSERTION(page_idx > ria->ria_stoff) failed
2124 Details : Once the unmatched stride IO mode is detected, shrink the stride-ahead
2125 window to 0. If it does hit cache miss, and read-pattern is still
2126 stride-io mode, does not reset the stride window, but also does not
2127 increase the stride window length in this case.
2129 --------------------------------------------------------------------------------
2131 2007-09-27 Cluster File Systems, Inc. <info@clusterfs.com>
2133 * Support for kernels:
2134 2.6.5-7.286 (SLES 9),
2135 2.6.9-55.0.2.EL (RHEL 4),
2136 2.6.16.46-0.14 (SLES 10),
2137 2.6.18-8.1.8.el5 (RHEL 5),
2138 2.6.18.8 vanilla (kernel.org)
2139 * Client support for unpatched kernels:
2140 (see http://wiki.lustre.org/index.php?title=Patchless_Client)
2141 2.6.16 - 2.6.21 vanilla (kernel.org)
2142 * Due to recently discovered recovery problems, we do not recommend
2143 using patchless RHEL 4 clients with this or any earlier release.
2144 * Recommended e2fsprogs version: 1.40.2-cfs1
2145 * Note that reiserfs quotas are disabled on SLES 10 in this kernel.
2149 Description: Fix errors in lfs documentation
2150 Details : Fixes man pages
2152 Severity : enhancement
2154 Description: Adaptive timeouts
2155 Details : RPC timeouts adapt to changing server load and network
2156 conditions to reduce resend attempts and improve recovery time.
2158 Severity : enhancement
2160 Description: llapi_file_create() does not allow some changes
2161 Details : add llapi_file_open() that allows specifying the file creation
2162 mode and open flags, and also returns an open file handle.
2164 Severity : enhancement
2166 Description: df doesn't work properly if diskfs blocksize != 4K
2167 Details : Choose biggest blocksize of OST's as the LOV's blocksize.
2170 Frequency : always on directories with default striping set
2172 Description: lfs find on -1 stripe looping in lsm_lmm_verify_common()
2173 Details : Avoid lov_verify_lmm_common() on directory with -1 stripe count.
2175 Severity : enhancement
2177 Description: merge and cleanup kernel patches.
2178 Details : Remove mnt_lustre_list in vfs_intent-2.6-rhel4.patch.
2180 Severity : enhancement
2182 Description: RedHat Update kernel for RHEL5
2183 Details : Kernel config file for RHEL5.
2185 Severity : enhancement
2187 Description: OSS needs mutliple precreate threads
2188 Details : Add ability to start more than one create thread per OSS.
2191 Frequency : only with quota on the root user
2193 Description: mds_obd_create error creating tmp object
2194 Details : When the user sets quota on root, llog will be affected and can't
2195 create files and write files.
2198 Frequency : Always on ia64 patchless client, and possibly others.
2200 Description: Add EXPORT_SYMBOL check for node_to_cpumask symbol.
2201 Details : This allows the patchless client to be loaded on architectures
2202 without this export.
2207 Description: disorder of journal start and llog_add cause deadlock.
2208 Details : in llog_origin_connect, journal start should happen before
2209 llog_add keep the same order as other functions to avoid
2212 Severity : enhancement
2214 Description: RedHat Update kernel for RHEL5
2215 Details : Modify the kernel config file more closer RHEL5.
2217 Severity : enhancement
2219 Description: Build failure against Centos5 (RHEL5)
2220 Details : Define PAGE_SIZE when it isn't present.
2223 Frequency : occasionally when using NFS
2225 Description: "ll_intent_file_open()) lock enqueue: err: -13" with nfs
2226 Details : with NFS, the anon dentry's parent was set to itself in
2227 d_alloc_anon(), so in MDS, we use rec->ur_fid1 to find the
2228 corresponding dentry other than use rec->ur_name.
2230 Severity : enhancement
2232 Description: enable data checksumming by default
2233 Details : enable checksum by default, allow --disable-checksum
2234 configure option and "-o nochecksum" mount option. Checksums
2235 can also be disabled at runtime via $LPROC/osc/*/checksum_pages.
2238 Frequency : Occasionally with failover
2240 Description: Client eviction due to failover config
2241 Details : after a connection loss, the lustre client should attempt to
2242 reconnect to the last active server first before trying the
2243 other potential connections.
2245 Severity : enhancement
2247 Description: client-side metadata stat-ahead during readdir(directory readahead)
2248 Details : perform client-side metadata stat-ahead when the client detects
2249 readdir and sequential stat of dir entries therein
2254 Description: when mds and osts use different quota unit(32bit and 64bit),
2255 quota will be released repeatly.
2256 Details : void sending multiple quota reqs to mds, which will keep the
2257 status between the reqs.
2260 Frequency : only with liblustre clients on XT3
2262 Description: evictions taking too long
2263 Details : allow llrd to evict clients directly on OSTs
2267 Description: osts not allocated evenly to files
2268 Details : change the condition to increase offset_idx
2270 Severity : enhancement
2272 Description: self-adjustable client's lru lists
2273 Details : use adaptive algorithm for managing client cached locks lru
2274 lists according to current server load, other client's work
2275 pattern, memory activities, etc. Both, server and client
2276 side namespaces provide number of proc tunables for controlling
2281 Description: rewrite ext2-derived code in llite/dir.c and obdclass/uuid.c
2282 Details : rewrite inherited code (uuid parsing code from ext2 utils and
2283 readdir code from ext3) from scratch preserving functionality.
2287 Description: Only those disconnect error should be returned by rq_status.
2288 Details : In open/enqueue processs, Some errors, which will cause client
2289 disconnected, should be returned by rq_status, while other
2290 errors should still be returned by intent, then mdc or llite will
2293 Severity : enhancement
2295 Description: Tune the kernel for good SCSI performance.
2296 Details : Set the value of /sys/block/{dev}/queue/max_sectors_kb
2297 to the value of /sys/block/{dev}/queue/max_hw_sectors_kb
2301 Frequency : Always for filesystems larger than 2TB on 32-bit systems.
2302 Bugzilla : 13547, 13627
2303 Description: Data corruption for OSTs that are formatted larger than 2TB
2305 Details : When generating the bio request for lustre file writes the
2306 sector number would overflow a temporary variable before being
2307 used for the IO. The data reads correctly from Lustre (which
2308 will overflow in a similar manner) but other file data or
2309 filesystem metadata may be corrupted in some cases.
2313 Description: "lfs find -obd UUID" prints directories
2314 Details : "lfs find -obd UUID" will return all directory names instead
2315 of just file names. It is incorrect because the directories
2316 do not reside on the OSTs.
2319 Bugzilla : 13596, 13614
2320 Description: MDS hang after unclean shutdown of lots of clients
2321 Details : Never resend AST requests.
2325 Frequency : Always, for kernels after 2.6.16
2326 Description: Fix warning idr_remove called for id=.. which is not allocated.
2327 Details : Last kernels save old s_dev before kill super and not allow
2328 to restore from callback - restore it before call kill_anon_super.
2332 Description: Fix error on 'ls .' at the top of the Lustre mount.
2333 Details : Don't revalidate dentry if it is a root dentry.
2335 --------------------------------------------------------------------------------
2337 2007-08-27 Cluster File Systems, Inc. <info@clusterfs.com>
2339 * Support for kernels:
2340 2.6.5-7.286 (SLES 9),
2341 2.6.9-55.0.2.EL (RHEL 4),
2342 2.6.16.46-0.14 (SLES 10),
2343 2.6.18-8.1.8.el5 (RHEL 5),
2344 2.6.18.8 vanilla (kernel.org)
2345 * Client support for unpatched kernels:
2346 (see http://wiki.lustre.org/index.php?title=Patchless_Client)
2347 2.6.16 - 2.6.22 vanilla (kernel.org)
2348 * Due to recently discovered recovery problems, we do not recommend
2349 using patchless RHEL 4 clients with this or any earlier release.
2350 * Recommended e2fsprogs version: 1.39.cfs8
2351 * Note that reiserfs quotas are disabled on SLES 10 in this kernel.
2356 Description: block reactivating mgc import until all deactivates complete
2357 Details : Fix race when failing back MDT/MGS to itself (testing)
2359 Severity : enhancement
2361 Description: lfs setstripe enhancement
2362 Details : Make lfs setstripe understand 'k', 'm' and 'g' for stripe size.
2365 Frequency : mds/oss recovery
2367 Description: llog ctxt is refrenced after it has been freed.
2368 Details : llog ctxt refcount was added to avoide the race between ctxt free
2369 and llog recovery process. Each llog user must hold ctxt refcount
2370 before it access the llog. And the llog ctxt can only be freed
2371 when its refcount is zero.
2373 Severity : enhancement
2375 Description: randomly memory allocation failure util
2376 Details : Make lustre randomly failed allocating memory for testing purpose.
2379 Frequency : only for SLES10
2381 Description: Update kernel patch for SLES10 SP1
2382 Details : Add patch blkdev_tunables-2.6-sles10.patch to 2.6-sles10.series.
2384 Severity : enhancement
2386 Description: omit set fsid for export NFS
2387 Details : fix set/restore device id for avoid EMFILE error and mark lustre fs
2388 as FS_REQUIRES_DEV for avoid problems with generate fsid.
2391 Frequency : after network failures
2393 Description: Add sync option to mount_lustre.c
2394 Details : Client loses data written to lustre after a network interruption.
2396 Severity : enhancement
2398 Description: Error message improvement.
2399 Details : Merging of two LCONSOLE_ERROR_MSG into one.
2401 Severity : enhancement
2403 Description: don't use GFP_* in generic Lustre code.
2404 Details : Use cfs_alloc_* functions and CFS_* flags for code portability.
2406 Severity : enhancement
2408 Description: obdclass is limited by single OBD_ALLOC(idarray)
2409 Details : replace OBD_ALLOC/OBD_FREE with OBD_VMALLOC/OBD_VFREE
2411 Severity : enhancement
2413 Description: updated patchess for new RHEL4 kernel
2414 Details : Fixed ext3-unlink-race.patch per Kalpak's comment.
2416 Severity : enhancement
2418 Description: warnings with build patchless client with vanila 2.6.19 and up
2419 Details : change old ctl_table style and replace ctl_table/ctl_table_header
2420 with cfs_sysctl_table_t/cfs_sysctl_table_header_t
2422 Severity : enhancement
2424 Description: O_DIRECT bypasses client statistics.
2425 Details : When running with O_DIRECT I/O, neither the client rpc_stats nor
2426 read_ahead_stats were updated. Copied stats section from
2427 osc_send_oap_rpc() into async_internal().
2430 Frequency : only for Cray XT3
2432 Description: peer credits not enough on many OST per OSS systems.
2433 Details : Use new lnet way to add credits as we need those for pings and ASTs
2436 Frequency : only with liblustre
2438 Description: Liblustre is not releasing flock locks on file close.
2439 Details : Release flock locks on file close.
2442 Frequency : only for RHEL4
2444 Description: Update kernel patches for kernel-2.6.9-55.0.2.EL
2445 Details : Remove inode-nr_unused-2.6.9-rhel4.patch from 2.6-rhel4.series
2446 Update target file and kernel config.
2450 Description: lustre support for RHEL5
2451 Details : Add support for RHEL5.
2456 Description: ASSERTION(export != NULL) failed in target_handle_connect
2457 Details : Assetion hit is result of rare race between disconnect and connet
2458 to same nid. target_handle_connect found old connect cockie and
2459 tried to reconnect, but can't find export for this cockie.
2464 Description: umount blocks forever on error
2465 Details : In result of wrong using obd_no_recov and obd_force flags client
2466 can hand if cancel or some other requests is lost.
2469 Frequency : Only for SLES
2471 Description: sanity_quota fail test_1
2472 Details : There are multiple occurences of $TSTUSR in SLES's /etc/group
2473 file, which makes TSTID[2] inunique.
2475 Severity : enhancement
2477 Description: Kernel patches for SLES9 2.6.5-7.286 kernel
2478 Details : Update target/ChangeLog/which_patch .
2480 Severity : enhancement
2482 Description: jbd statistics
2483 Details : Port older jbd statistics patch for sles10
2488 Description: lvbo_init failed for resource with missing objects.
2489 Details : Fix returning error if we do stat for file with missing/corrupted
2490 objects and i_size set to all sum of size of all avaible objects.
2491 if we to truncate/write to missing object - it is recreated.
2496 Description: Oops in read and write path when failing to allocate lock.
2497 Details : Check if lock allocation failed and return error back.
2500 Frequency : When flocks are used.
2502 Description: assertion failure in ldlm_cli_enquque_fini for non NULL lock.
2503 Details : Flock locks might destroy just granted lock if it could be merged
2504 with another existing flock, this is done in completion handler,
2505 so teach ldlm_cli_enquque_fini that this is a valid case for
2511 Description: reply_lock_interpret crash due to race with it and lock cancel.
2512 Details : Do not replay locks that are being cancelled. Do not reference
2513 locks by their address during replay, just by their handle.
2516 Frequency : only with deactivated OSTs
2518 Description: lstripe command fails for valid OST index
2519 Details : The stripe offset is compared to 'lov->desc.ld_tgt_count'
2520 instead of lov->desc.ld_active_tgt_count.
2522 Severity : enhancement
2524 Description: Build failure against Centos5 (RHEL5)
2525 Details : Use getpagesize() instead of PAGE_SIZE.
2527 --------------------------------------------------------------------------------
2529 2007-07-30 Cluster File Systems, Inc. <info@clusterfs.com>
2531 * Support for kernels:
2532 2.6.5-7.283 (SLES 9),
2533 2.6.9-55.EL (RHEL 4),
2534 2.6.16.46-0.14 (SLES 10),
2535 2.6.18.8 vanilla (kernel.org)
2536 * Client support for unpatched kernels:
2537 (see http://wiki.lustre.org/index.php?title=Patchless_Client)
2538 2.6.16 - 2.6.22 vanilla (kernel.org)
2539 * Due to recently discovered recovery problems, we do not recommend
2540 using patchless RHEL 4 clients with this or any earlier release.
2541 * Recommended e2fsprogs version: 1.39.cfs8
2542 * Note that reiserfs quotas are disabled on SLES 10 in this kernel.
2543 * Starting with this release, the ldiskfs backing filesystem required
2544 by Lustre is now in its own package, lustre-ldiskfs. This package
2545 should be installed. It is versioned separately from Lustre and
2546 may be released separately in future.
2551 Description: server LBUG when shutting down
2552 Details : Block umount forever until the mount refcount is zero rather
2553 than giving up after an arbitrary timeout.
2555 Severity : enhancement
2557 Description: add optional extra BUILD_VERSION info
2558 Details : add a new environment variable (namely LUSTRE_VERS) which allows
2559 to override the lustre version.
2562 Frequency : 2.6.18 servers only
2564 Description: ll_kern_mount() doesn't release the module reference
2565 Details : The ldiskfs module reference count never drops down to 0
2566 because ll_kern_mount() doesn't release the module reference.
2571 Description: server LBUG when using old ost_num_threads parameter
2572 Details : Accept the old ost_num_threads parameter but warn that it
2573 is deprecated, and fix an off-by-one error that caused an LBUG.
2578 Description: Transient SCSI error results in persistent IO issue
2579 Details : iobuf->dr_error is not reinitialized to 0 between two
2583 Frequency : sometimes when underlying device returns I/O errors
2585 Description: OSTs not going read-only during write failures
2586 Details : OSTs are not remounted read-only when the journal commit threads
2587 get I/O errors because fsfilt_ext3 calls journal_start/stop()
2588 instead of the ext3 wrappers.
2591 Frequency : only with 10000 clients or more
2593 Description: poor connect scaling with increasing client count
2594 Details : Don't run filter_grant_sanity_check for more than 100 exports
2595 to improve scaling for large numbers of clients.
2598 Frequency : SLES10 only
2600 Description: sanity-quota.sh quotacheck failed: rc = -22
2601 Details : Quotas cannot be enabled on SLES10.
2604 Frequency : liblustre clients only
2606 Description: getdirentries does not give error when run on compute nodes
2607 Details : getdirentries does not fail when the size specified as an argument
2608 is too small to contain at least one entry
2610 Severity : enhancement
2612 Description: Add LNET router traceability for debug purposes
2613 Details : If a checksum failure occurs with a router as part of the
2614 IO path, the NID of the last router that forwarded the bulk data
2615 is printed so it can be identified.
2620 Description: OST "spontaneously" evicts client; client has imp_pingable == 0
2621 Details : Due to a race condition, liblustre clients were occasionally
2622 evicted incorrectly.
2624 Severity : enhancement
2626 Description: lfs setstripe use optional parameters instead of postional
2629 Severity : enhancement
2631 Description: Nanosecond timestamp support for ldiskfs
2632 Details : The on-disk ldiskfs filesystem has added support for nanosecond
2633 resolution timestamps. There is not yet support for this at
2634 the Lustre filesystem level.
2637 Frequency : during server recovery
2639 Description: MDS failing to send precreate requests due to OSCC_FLAG_RECOVERING
2640 Details : request with rq_no_resend flag not awake l_wait_event if they get a
2644 Frequency : nfs export on patchless client
2646 Description: connectathon hang when test nfs export over patchless client
2647 Details : Disconnected dentry cannot be found with lookup, so we do not need
2648 to unhash it or make it invalid
2651 Description: fix llapi_lov_get_uuids() to allow many OSTs to be returned
2652 Details: : Change llapi_lov_get_uuids() to read the UUIDs from /proc instead
2653 of using an ioctl. This allows lfsck for > 160 OSTs to succeed.
2658 Description: open req refcounting wrong on reconnect
2659 Details : If reconnect happened between getting open reply from server and
2660 call to mdc_set_replay_data in ll_file_open, we will schedule
2661 replay for unreferenced request that we are about to free.
2662 Subsequent close will crash in variety of ways.
2663 Check that request is still eligible for replay in
2664 mdc_set_replay_data().
2669 Description: disable writes to filesystem when reading health_check file
2670 Details : the default for reading the health_check proc file has changed
2671 to NOT do a journal transaction and write to disk, because this
2672 can cause reads of the /proc file to hang and block HA state
2673 checking on a healthy but otherwise heavily loaded system. It
2674 is possible to return to the previous behaviour during configure
2675 with --enable-health-write.
2677 Severity : enhancement
2679 Description: 64-bit inode version
2680 Details: : Add a on-disk 64-bit inode version for ext3 to track changes made
2681 to the inode. This will be required for version-based recovery.
2686 Description: MDS fails to start if a duplicate client export is detected
2687 Details : in some rare cases it was possible for a client to connect to
2688 an MDS multiple times. Upon recovery the MDS would detect this
2689 and fail during startup. Handle this more gracefully.
2691 Severity : enhancement
2693 Description: Add -o localflock option to simulate old noflock behaviour.
2694 Details : This will achieve local-only flock/fcntl locks coherentness.
2699 Description: log_commit_thread vs filter_destroy race leads to crash
2700 Details : Take import reference before releasing llog record semaphore
2705 Description: Wrong request locking in request set processing
2706 Details : ptlrpc_check_set wrongly uses req->rq_lock for proctect add to
2707 imp_delayed_list, in this place should be used imp_lock.
2710 Frequency : when reconnecting
2712 Description: Grant leak when OSC reconnect to OST
2713 Details : When osc reconnect ost, OST(filter) should check whether it
2714 should grant more space to client by comparing fed_grant and
2715 cl_avail_grant, and return the granted space to client instead
2716 of "new granted" space, because client will call osc_init_grant
2717 to update the client grant space info.
2720 Frequency : when client reconnects to OST
2722 Description: Grant leak when OSC does a resend and replays bulk write
2723 Details : When osc reconnect to OST, OST(filter) should clear grant info of
2724 bulk write request, because the grant info while be sync between
2725 OSC and OST when reconnect, and we should ignore the grant info
2726 these of resend/replay write req.
2731 Description: Grant space more than avaiable space sometimes.
2732 Details : When then OST is about to be full, if two bulk writing from
2733 different clients came to OST. Accord the avaliable space of the
2734 OST, the first req should be permitted, and the second one
2735 should be denied by ENOSPC. But if the seconde arrived before
2736 the first one is commited. The OST might wrongly permit second
2737 writing, which will cause grant space > avaiable space.
2740 Frequency : when client is evicted
2742 Description: Grant might be wrongly erased when osc is evicted by OST
2743 Details : when the import is evicted by server, it will fork another
2744 thread ptlrpc_invalidate_import_thread to invalidate the
2745 import, where the grant will be set to 0. While the original
2746 thread will update the grant it got when connecting. So if
2747 the former happened latter, the grant will be wrongly errased
2748 because of this race.
2753 Description: Checking Stale with correct fid
2754 Details : ll_revalidate_it should uses de_inode instead of op_data.fid2
2755 to check whether it is stale, because sometimes, we want the
2756 enqueue happened anyway, and op_data.fid2 will not be initialized.
2758 Severity : enhancement
2760 Description: update patchless client
2761 Details : Add support for patchless client with 2.6.20, 2.6.21 and RHEL 5
2764 Frequency : only with 2.4 kernel
2766 Description: random memory corruption
2767 Details : size of struct ll_inode_info is to big for union inode.u and this
2768 can be cause of random memory corruption.
2773 Description: Memory leak in recovery
2774 Details : Lov_mds_md was not free in an error handler in mds_create_object.
2775 It should also check obd_fail before fsfilt_start, otherwise if
2776 fsfilt_start return -EROFS,(failover mds during mds recovery).
2777 then the req will return with repmsg->transno = 0 and rc = EROFS.
2778 and we met hit the assert LASSERT(req->rq_reqmsg->transno ==
2779 req->rq_repmsg->transno) in ptlrpc_replay_interpret. Fcc should
2780 be freed no matter whether fsfilt_commit success or not.
2783 Frequency : only with large numbers of cores on a single node
2785 Description: Prevents from taking the superblock lock in llap_from_page for
2787 Details : using LL_ORIGIN_REMOVEPAGE origin flag instead of LL_ORIGIN_UNKNOW
2788 for llap_from_page call in ll_removepage() prevents from taking
2789 the superblock lock for a soon died page.
2794 Description: Not check open intent error before release open handle
2795 Details : in some rare cases, the open intent error is not checked before
2796 release open handle, which may cause
2797 ASSERTION(open_req->rq_transno != 0), because it tries to release
2798 the failed open handle.
2803 Description: Set cat log bitmap only after create log success.
2804 Details : in some rare cases, the cat log bitmap is set too early. and it
2805 should be set only after create log success.
2809 Description: Accessing a block bevice can re-enable I/O when Lustre is
2810 tearing down a device.
2811 Details : dev_clear_rdonly(bdev) must be called in kill_bdev() instead of
2816 Description: service threads may hog cpus when there are a lot of requests
2817 Details : Insert cond_resched to give other threads a chance to use some CPU
2822 Description: the cat log was not initialized in recovery
2823 Details : When mds(mgs) do recovery, the tgt_count might be zero, so the
2824 unlink log on mds will not be initialized until mds post
2825 recovery. And also in mds post recovery, the unlink log will
2826 initialization will be done asynchronausly, so there will be race
2827 between add unlink log and unlink log initialization.
2831 Description: brw_stats were being printed incorrectly
2832 Details : brw_stats were being printed as log2 but all of them were not
2833 recorded as log2. Also remove some code duplication arising from
2834 filter_tally_{read,write}.
2838 Frequency : rare, only in recovery.
2839 Description: ASSERTION(req->rq_type != LI_POISON) failed
2840 Details : imp_lock should be held while iterating over imp_sending_list for
2841 prevent destroy request after get timeout in ptlrpc_queue_wait.
2845 Description: replay-single.sh test 52 fails
2846 Details : A lock's skiplist need to be cleanup when it being unlinked
2847 from its resource list.
2852 Description: Short directio read returns full requested size rather than
2854 Details : Direct I/O operations should return actual amount of bytes
2855 transferred rather than requested size.
2857 Severity : enhancement
2859 Description: metadata RPC reduction (e.g. for rm performance)
2860 Details : decrease the amount of synchronous RPC between clients and servers
2861 by canceling conflicing lock before the operation on the client
2862 and packing thier handles into the main operation RPC to server.
2864 Severity : enhancement
2866 Description: add #ifdef HAVE_KERNEL_CONFIG_H
2867 Details : kernels from 2.6.19 not need include linux/config.h, but add
2868 include linux/autoconf.h in commpiler command line.
2870 Severity : enhancement
2872 Description: patchless client support for 2.6.22 kernel
2873 Details : 2.6.22 has only one visble change, SLAB_CTOR_* constants is
2874 removed. In this case we need drop using os depended interface to
2875 kmem_cache and use cfs_mem_cache API.
2880 Description: fix mal-formatted messages
2881 Details : fix some mal-formatted DEBUG_REQ and LCONSOLE_ERROR_MSG messages
2885 Frequency : always in liblustre
2886 Description: wrong IS_ERR implementation in liblustre.h
2887 Details : fix IS_ERR implementation in liblustre.h for right detect errors.
2892 Description: Correct condition for output debug message.
2893 Details : inode i_nlink equal zero is not enough for output message about
2894 disk corruption, i_ctime and i_mode should be also checked.
2898 Frequency : always in patchless client
2899 Description: add configure check for truncate_complete_page
2900 Details : improve checks for exported symbols. This allow run check without
2901 sources, but with Module.symvers shipped with kernel distribution.
2902 add check for truncate_complete_page used by patchless client.
2906 Description: sanity.sh test_77h fails with "test_77h file compare failed"
2907 Details : test_77h uses a file which was messed by other test case.
2911 Description: Not Check whether lov_tgts is NULL in some lov functions
2912 Details : Checking whether lov_tgts is NULL in some functions.
2916 Description: replace obdo_alloc() with OBDO_ALLOC macro
2917 Details : nothing special is done in obdo_alloc() function, and for
2918 debugging purpose, it needs to be replaced with macros.
2922 Description: bad return value and errno from fcntl call
2923 Details : In liblustre API, errno should be a negative value if error
2928 Description: ptlrpc_check_set() LBUG
2929 Details : In case of positive reply from server and failed client bulk
2930 callback after bulk transfer shouldn't LBUG, but process this
2931 request as erroneous.
2933 Severity : enhancement
2935 Description: VFS operations stats tool.
2936 Details : Tool which collects stats by tracking value written in pid,
2937 ppid, gid and uses llstat to generate output to plot graph using
2939 Updated lustre/utils/Makefile.am
2940 Added lustre/utils/ltrack_stats.c
2942 Severity : enhancement
2944 Description: 2.6.18 server support (lustre 1.6.1)
2945 Details : Support for 2.6.18 kernels on the server side.
2950 Description: ASSERTION(imp->imp_conn_current) failed
2951 Details : an assertion failure is hit if a client node boots and attempts to
2952 mount a lustre filesystem faster than RECONNECT_INTERVAL seconds.
2955 Frequency : only for i686
2957 Description: 1.4.11 RC1 build fails for RHEL 4, i686
2958 Details : Fixed config variable for build.
2963 Description: Updated patchess for new RHEL4 kernel
2964 Details : Updated patch inode-nr_unused-2.6.9-rhel4.patch
2965 Updated patch jbd-stats-2.6.9.patch
2966 Updated patch qsnet-rhel4-2.6.patch
2967 Updated patch quota-deadlock-on-pagelock-core.patch
2968 Updated patch vfs_intent-2.6-rhel4.patch
2969 Updated patch vfs_races-2.6-rhel4.patch
2970 Updated series file 2.6-rhel4-titech.series
2971 Updated series file 2.6-rhel4.series
2972 Updated kernel config files
2977 Description: lquota slave complains LBUG when reconnecting with mds
2979 Details : quota slave depends qctxt->lqc_import to send its quota request.
2980 This pointer will be invalid if mds did failover or broke its
2981 connect to osts, which leads to LBUG.
2984 Frequency : when qunit size is too small(less than 20M)
2986 Description: write is stopped by improper -EDQUOT
2987 Details : If the master is busy and qunit size is small enough(let's say 1M),
2988 the slave can not get quota from master on time, which will lead
2989 slave to trigger a -EQUOTA to client.
2994 Description: Deadlock during metadata tests
2995 Details : in prune_dir_dentries(), shrink_dcache_parent() should not be
2996 called with the per-dentry lock held.
2999 Frequency : SLES9 only
3001 Description: Lustre patched kernel for SLES9 SP3 has NR_CPUS set to 8
3002 Details : set CONFIG_NR_CPUS to 128 instead of 8.
3004 Severity : enhancement
3006 Description: remove fs_prep_san_write operation and related patches
3007 Details : remove the ext3-san-jdike patches which are no longer useful.
3012 Description: LDISKFS-fs error (device sdc): ldiskfs_free_blocks
3013 Details : a disk corruption can cause the mballoc code to assert on a
3014 double free or other extent corruptions. Handle these with
3015 ext3_error() instead of assertions.
3018 Frequency : only with mballoc3 code and deep extent trees
3020 Description: ldiskfs_ext_search_right: bad header in inode: unexpected eh_depth
3021 Details : a wrong check of extent headers in ldiskfs_ext_search_right()
3022 can cause the filesystem to be remounted read-only.
3027 Description: lfsck built against 1.4.x cannot run against 1.6.0 lustre
3028 Details : the definition for OBD_IOC_GETNAME changed in 1.6.0. One of the
3029 few external users of this ioctl number is lfsck's call to
3030 llapi_lov_get_uuids() and this caused lfsck to fail at startup.
3031 Add the old ioctl number to the handler so both old and new
3036 Description: parallel lock callbacks
3037 Details : Instead of sending blocking and completion callbacks as separated
3038 requests, adding them to a set and sending in parallel.
3042 Description: Disable most debugging by default
3043 Details : To improve performance, disable most logging (for debug purposes)
3044 by default. VFSTRACE, RPCTRACE, and DLMTRACE are now off by
3045 default, and HA includes fewer messages.
3049 Frequency : only run on patchless client.
3050 Description: use do_facet on sanity.sh for test handling recoverables errors
3051 Details : use do_facet instead of direct use sysctl for set fail_loc on OST
3055 Description: hash tables for lists of nids, connections and uuids
3056 Details : Hash tables noticeably help when a lot of clients connect to a
3057 server, to faster identify duplicate connections or reconnects,
3058 also to faster find export to evict in manual eviction case.
3062 Description: Sometimes, when the server evict a client, and the client will
3063 not be evicted as soon as possible.
3064 Details : In enqueue req, the error was returned by intent, instead of
3065 rq_status which make ptlrpc layer not detect this error, and
3066 does not evict the client. So enqueue error should be returned
3070 Frequency : only at startup
3072 Description: Delay client connections to MDT until fist MDT->OST connect
3073 Details : If a client tried to create a new file before the MDT had
3074 connected to any OSTs, the create would return EIO. Now
3075 the client will simply block until the MDT connects to the
3076 first OST and the create can succeed.
3079 Frequency : only for SLES9
3081 Description: Routinely utilize latest Quadrics drivers in CFS releases
3082 Details : Update patch qsnet-suse-2.6.patch.
3085 Frequency : only for sles10
3087 Description: Update patches for SLES 10 SP1 kernel.
3088 Details : Update the patch vfs_intent-2.6-sles10.patch.
3093 Description: Routinely utilize latest Quadrics drivers in CFS releases
3094 Details : Update patch qsnet-rhel4-2.6.patch.
3097 Frequency : at statup only
3099 Description: mds_lov_synchronize race leads to various problems
3100 Details : simultaneous MDT->OST connections at startup can cause the
3101 sync to abort, leaving the OSC in a bad state.
3106 Description: Using wrong pointer in osc_brw_prep_request
3107 Details : Access to array[-1] can produce panic if kernel compiled with
3108 CONFIG_PAGE_ALLOC enabled
3110 Severity : enhancement
3112 Description: Async OSC create to avoid the blocking unnecessarily.
3113 Details : If a OST has no remain object, system will block on the creating
3114 when need to create a new object on this OST. Now, ways use
3115 pre-created objects when available, instead of blocking on an
3116 empty osc while others are not empty. If we must block, we block
3117 for the shortest possible period of time.
3121 Frequency : only in recovery
3122 Description: Mark OST as early accessible if his start SYNC.
3123 Details : osc_precreate return flag early accessible if oscc marked as
3124 OSCC_FLAG_SYNC_IN_PROGRESS.
3129 Description: Sometimes precreate code can triger create object on wrong ost
3130 Details : Wrong protected or not not restored variables aftre precreate loop
3131 can produce creation object on wrong ost.
3134 Frequency : oss recovery
3136 Description: llog_commit_thread cleanup should sync with llog_commit_thread
3138 Details : llog_commit_thread_count should be synced between llog_commit
3139 start and cleanup, so new llog_commit thread should not be started
3140 when llog_commit threads being stopped to avoid accessing some
3143 Severity : enhancement
3145 Description: Add printing inode info into message about error in writepage.
3147 --------------------------------------------------------------------------------
3149 2007-05-03 Cluster File Systems, Inc. <info@clusterfs.com>
3154 Frequency : on some architectures
3156 Description: 1.6 client sometimes fails to mount from a 1.4 MDT
3157 Details : Uninitialized flags sometimes cause configuration commands to
3161 Frequency : patchless clients only
3163 Description: missing __iget() symbol export
3164 Details : The __iget() symbol export is missing. To avoid the need for
3165 this on patchless clients the deathrow inode reaper is turned
3166 off, and we depend on the VM to clean up old inodes. This
3167 dependency was during via the fix for bug 12181.
3172 Description: sanity.sh fail: test_52b
3173 Details : The ll_inode_to_ext_flags() has a glitch which makes MDS return
3174 incorrect inode's flags to client.
3176 --------------------------------------------------------------------------------
3178 2007-04-19 Cluster File Systems, Inc. <info@clusterfs.com>
3180 * CONFIGURATION CHANGE. This version of Lustre WILL NOT
3181 INTEROPERATE with older versions automatically. In many cases a
3182 special upgrade step is needed. Please read the
3183 user documentation before upgrading any part of a 1.4.x system.
3184 * WARNING: Lustre configuration and startup changes are required with
3185 this release. See https://mail.clusterfs.com/wikis/lustre/MountConf
3187 * Support for kernels:
3188 2.4.21-47.0.1.EL (RHEL 3),
3189 2.6.5-7.283 (SLES 9),
3190 2.6.9-42.0.10.EL (RHEL 4),
3191 2.6.12.6 vanilla (kernel.org),
3192 2.6.16.27-0.9 (SLES10)
3193 * Client support for unpatched kernels:
3194 (see https://mail.clusterfs.com/wikis/lustre/PatchlessClient)
3195 2.6.16 - 2.6.19 vanilla (kernel.org),
3196 2.6.9-42.0.8EL (RHEL 4)
3197 * Recommended e2fsprogs version: 1.39.cfs6
3198 * Note that reiserfs quotas are disabled on SLES 10 in this kernel
3201 Severity : enhancement
3203 Description: Async OSC create to avoid the blocking unnecessarily.
3204 Details : If a OST has no remain object, system will block on the creating
3205 when need to create a new object on this OST. Now, ways use
3206 pre-created objects when available, instead of blocking on an
3207 empty osc while others are not empty. If we must block, we block
3208 for the shortest possible period of time.
3210 Severity : enhancement
3212 Description: MountConf
3213 Details : Lustre configuration is now managed via mkfs and mount
3214 commands instead of lmc and lconf. New obd types (MGS, MGC)
3215 are added for dynamic configuration management. See
3216 https://mail.clusterfs.com/wikis/lustre/MountConf for
3219 Severity : enhancement
3221 Description: dynamic OST addition
3222 Details : OSTs can now be added to a live filesystem
3224 Severity : enhancement
3226 Description: startup order invariance
3227 Details : MDTs and OSTs can be started in any order. Clients only
3228 require the MDT to complete startup.
3230 Severity : enhancement
3232 Description: parallel, asynchronous orphan cleanup
3233 Details : orphan cleanup is now performed in separate threads for each
3234 OST, allowing parallel non-blocking operation.
3236 Severity : enhancement
3238 Description: optimized stripe assignment
3239 Details : stripe assignments are now made based on ost space available,
3240 ost previous usage, and OSS previous usage, in order to try
3241 to optimize storage space and networking resources.
3243 Severity : enhancement
3245 Description: Permanently set tunables
3246 Details : All writable /proc/fs/lustre tunables can now be permanently
3247 set on a per-server basis, at mkfs time or on a live
3250 Severity : enhancement
3252 Description: Lustre message v2
3253 Details : Add lustre message format v2.
3255 Severity : enhancement
3257 Description: client OST exclusion list
3258 Details : Clients can be started with a list of OSTs that should be
3259 declared "inactive" for known non-responsive OSTs.
3263 Description: ENOENT returned for valid filehandle during dbench.
3264 Details : Check if a directory has children when invalidating dentries
3265 associated with an inode during lock cancellation. This fixes
3266 an incorrect ENOENT sometimes seen for valid filehandles during
3267 testing with dbench.
3270 Frequency : SFS test only (otherwise harmless)
3272 Description: SPEC SFS validation failure on NFS v2 over lustre.
3273 Details : Changes the blocksize for regular files to be 2x RPC size,
3274 and not depend on stripe size.
3276 Severity : enhancement
3278 Description: fine-grained SMP locking inside DLM
3279 Details : Improve DLM performance on SMP systems by removing the single
3280 per-namespace lock and replace it with per-resource locks.
3282 Severity : enhancement
3284 Description: don't hold multiple extent locks at one time
3285 Details : To avoid client eviction during large writes, locks are not
3286 held on multiple stripes at one time or for very large writes.
3287 Otherwise, clients can block waiting for a lock on a failed OST
3288 while holding locks on other OSTs and be evicted.
3290 Severity : enhancement
3292 Description: Multiple MD RPCs in flight.
3293 Details : Further unserialise some read-only MDT RPCs - learn about intents.
3294 To avoid overly-overloading MDT, introduce a limit on number of
3295 MDT RPCs in flight for a single client and add /proc controls
3296 to adjust this limit.
3298 Severity : enhancement
3300 Description: client read/write statistics
3301 Details : Add client read/write call usage stats for performance
3302 analysis of user processes.
3303 /proc/fs/lustre/llite/*/offset_stats shows non-sequential
3304 file access. extents_stats shows chunk size distribution.
3305 extents_stats_per_process show chunk size distribution per
3308 Severity : enhancement
3310 Description: per-client statistics on server
3311 Details : Add ldlm and operations statistics for each client in
3312 /proc/fs/lustre/mds|obdfilter/*/exports/
3314 Severity : enhancement
3316 Description: improved MDT statistics
3317 Details : Add detailed MDT operations statistics in
3318 /proc/fs/lustre/mds/*/stats
3320 Severity : enhancement
3322 Description: VFS operations stats
3323 Details : Add client VFS call stats, trackable by pid, ppid, or gid
3324 /proc/fs/lustre/llite/*/stats_track_[pid|ppid|gid]
3329 Description: Fix client-side osc byte counters
3330 Details : The osc read/write byte counters in
3331 /proc/fs/lustre/osc/*/stats are now working
3334 Frequency : always as root on SLES
3336 Description: Failure of copying files with lustre special EAs.
3337 Details : Client side always return success for setxattr call for lustre
3338 special xattr (currently only "trusted.lov").
3343 Description: Refcount LNET uuids
3344 Details : The global LNET uuid list grew linearly with every startup;
3345 refcount repeated list entries instead of always adding to
3348 Severity : enhancement
3350 Description: Dynamic service threads
3351 Details : Within a small range, start extra service threads
3352 automatically when the request queue builds up.
3355 Frequency : mixed-endian client/server environments
3357 Description: mixed-endian crashes
3358 Details : The new msg_v2 system had some failures in mixed-endian
3361 Severity : enhancement
3363 Description: Easy OST removal
3364 Details : OSTs can be permanently deactivated with e.g. 'lctl
3365 conf_param lustre-OST0001.osc.active=0'
3367 Severity : enhancement
3369 Description: MGS proc entries
3370 Details : Added basic proc entries for the MGS showing what filesystems
3373 Severity : enhancement
3375 Description: provide MGS failover
3376 Details : Added config lock reacquisition after MGS server failover.
3378 Severity : enhancement
3380 Description: add Linux 2.4 support
3381 Details : Added support for RHEL 2.4.21 kernel for 1.6 servers and clients
3385 Description: a large application tries to do I/O to the same resource and dies
3386 in the middle of it.
3387 Details : Check the req->rq_arrival time after the call to
3388 ost_brw_lock_get(), but before we do anything about
3389 processing it & sending the BULK transfer request. This
3390 should help move old stale pending locks off the queue as
3391 quickly as obd_timeout.
3394 Frequency : when an incorrect nid is specified during startup
3396 Description: ptlrpc connect to non-existant node causes kernel crash
3397 Details : LNET can't be re-entered from an event callback, which
3398 happened when we expire a message after the export has been
3399 cleaned up. Instead, hand the zombie cleanup off to another
3402 Severity : enhancement
3404 Description: plain/inodebits lock performance improvement
3405 Details : Grouping plain/inodebits in granted list by their request modes
3406 and bits policy, thus improving the performance of search through
3410 Frequency : only if OST filesystem is corrupted
3412 Description: client incorrectly hits assertion in ptlrpc_replay_req()
3413 Details : for a short time RPCs with bulk IO are in the replay list,
3414 but replay of bulk IOs is unimplemented. If the OST filesystem
3415 is corrupted due to disk cache incoherency and then replay is
3416 started it is possible to trip an assertion. Avoid putting
3417 committed RPCs into the replay list at all to avoid this issue.
3420 Frequency : liblustre (e.g. catamount) on a large cluster with >= 8 OSTs/OSS
3422 Description: System hang on startup
3423 Details : This bug allowed the liblustre (e.g. catamount) client to
3424 return to the app before handling all startup RPCs. This
3425 could leave the node unresponsive to lustre network traffic
3426 and manifested as a server ptllnd timeout.
3428 Severity : enhancement
3430 Description: Add "/proc/sys/lustre/debug_peer_on_timeout"
3431 Details : liblustre envirable: LIBLUSTRE_DEBUG_PEER_ON_TIMEOUT
3432 boolean to control whether to print peer debug info when a
3433 client's RPC times out.
3436 Frequency : only for kernels with patches from Lustre below 1.4.3
3438 Description: Remove old rdonly API
3439 Details : Remove old rdonly API which unused from at least lustre 1.4.3
3442 Frequency : only for devices with external journals
3444 Description: Set external device read-only also
3445 Details : During a commanded failover stop, we set the disk device
3446 read-only while the server shuts down. We now also set any
3447 external journal device read-only at the same time.
3450 Frequency : when upgrading from 1.4 while trying to change parameters
3452 Description: The wrong (new) MDC name was used when setting parameters for
3453 upgraded MDT's. Also allows changing of OSC (and MDC)
3454 parameters if --writeconf is specified at tunefs upgrade time.
3457 Frequency : when setting specific ost indicies
3459 Description: QOS code breaks on skipped indicies
3460 Details : Add checks for missing OST indicies in the QOS code, so OSTs
3461 created with --index need not be sequential.
3463 Severity : enhancement
3465 Description: Add uninit_groups feature to ldiskfs2 to speed up e2fsck
3466 Details : The uninit_groups feature works in conjunction with the kernel
3467 filesystem code (ldiskfs2 only) and e2fsprogs-1.39-cfs6 to speed
3468 up the pass1 processing of e2fsck. This is a read-only feature
3469 in ldiskfs2 only, so older kernels and current ldiskfs cannot
3470 mount filesystems that have had this feature enabled.
3472 Severity : enhancement
3474 Description: Improve multi-block allocation algorithm to avoid fragmentation
3475 Details : The mballoc3 code (ldiskfs2 only) adds new mechanisms to improve
3476 allocation locality and avoid filesystem fragmentation.
3478 ------------------------------------------------------------------------------
3480 2007-04-01 Cluster File Systems, Inc. <info@clusterfs.com>
3482 * Support for kernels:
3483 2.4.21-47.0.1.EL (RHEL 3)
3484 2.6.5-7.283 (SLES 9)
3485 2.6.9-42.0.10.EL (RHEL 4)
3486 2.6.12.6 vanilla (kernel.org)
3487 2.6.16.27-0.9 (SLES 10)
3488 * Recommended e2fsprogs version: 1.39.cfs5
3490 * Note that reiserfs quotas are disabled on SLES 10 in this kernel
3494 Frequency : occasional, depends on client load and configuration
3495 Bugzilla : 12181, 12203
3496 Description: data loss for recently-modified files
3498 Details : In some cases it is possible that recently written or created
3499 files may not be written to disk in a timely manner (this should
3500 normally be within 30s unless client IO load is very high).
3501 The problem appears as zero-length files or files that are a
3502 multiple of 1MB in size after a client crash or client eviction
3503 that are missing data at the end of the file.
3505 This problem is more likely to be hit on clients where files are
3506 repeatedly created and unlinked in the same directory, clients
3507 have a large amount of RAM, have many CPUs, the filesystem has
3508 many OSTs, the clients are rebooted frequently, and/or the files
3509 are not accessed by other nodes after being written.
3511 The presence of the problem can be detected by looking at
3512 /proc/sys/fs/inode-state. If the first number (nr_inodes) is
3513 smaller than the second (nr_unused) then dirty files will not
3514 be flushed automatically to disk. "sync; sleep 10" should be
3515 run several times on the node before unmounting it to update
3516 Lustre (this is also safe to run on nodes without this problem).
3518 There is also a related kernel bug in the RHEL4 4 2.6.9 kernel
3519 that can cause this same problem, so customers using that kernel
3520 also need to update the kernel in addition to Lustre. In order
3521 to properly fix this bug, the RHEL3 2.4.21 kernel is also updated.
3523 It is normal that files written just before a client crash (less
3524 than 30s) may not yet have been flushed to disk, even for local
3528 Frequency : frequent on thin XT3 nodes
3530 Description: UUID collision on thin XT3 Linux nodes
3531 Details : UUIDs on Compute Node Linux XT3 nodes were not generated
3532 randomly, since we relied on an insufficiently-seeded PRNG.
3537 Description: OSS hangs after "All ost request buffers busy"
3538 Details : A deadlock between quota and journal operations caused OSS
3539 hangs after printing "All ost request buffers busy."
3542 Frequency : always on liblustre builds
3544 Description: Cleanup compiler warnings on liblustre
3547 Frequency : always on liblustre builds on XT3
3549 Description: LC_CONFIG_CDEBUG don't run while build liblustre on XT3.
3553 Description: Addition of EXT3_FEATURE_RO_COMPAT_DIR_NLINKS flag for
3554 > 32000 subdirectories
3555 Details : Add EXT3_FEATURE_RO_COMPAT_DIR_NLINK flag to
3556 EXT3_FEATURE_RO_COMPAT_SUPP. This flag will be set whenever
3557 subdirectory count crosses 32000. This will aid e2fsck to
3558 correctly handle more than 32000 subdirectories.
3561 Frequency : liblustre (e.g. catamount) on a large cluster with >= 8 OSTs/OSS
3563 Description: System hang on startup
3564 Details : This bug allowed the liblustre (e.g. catamount) client to
3565 return to the app before handling all startup RPCs. This
3566 could leave the node unresponsive to lustre network traffic
3567 and manifested as a server ptllnd timeout.
3569 Severity : enhancement
3571 Description: Add "/proc/sys/lustre/debug_peer_on_timeout"
3572 (liblustre envirable: LIBLUSTRE_DEBUG_PEER_ON_TIMEOUT)
3573 boolean to control whether to print peer debug info when a
3574 client's RPC times out.
3579 Description: make O_SYNC working on 2.6 kernels
3580 Details : 2.6 kernels use different method for mark pages for write,
3581 so need add a code to lustre for O_SYNC work.
3586 Description: Failure to close file and release space on NFS
3587 Details : Put inode details into lock acquired in ll_intent_file_open.
3588 Use mdc_intent_lock in ll_intent_open to properly
3589 detect all kind of errors unhandled by mdc_enqueue.
3594 Description: proc file read during shutdown sometimes raced obd removal,
3596 Details : Add lock to prevent obd access after proc file removal.
3599 Frequency : Only for files larger than 4GB on 32-bit clients.
3601 Description: improperly doing page alignment of locks
3602 Details : Modify lustre core code to use CFS_PAGE_* defines instead of
3603 PAGE_*. Make CFS_PAGE_MASK a 64-bit mask.
3608 Description: RPCs being resent when they shouldn't be
3609 Details : Some RPCs that should not be resent are being resent. This
3610 can cause inconsistencies in the RPC state machine. Do not
3611 resend such requests.
3614 Frequency : rare, only with NFS export
3616 Description: Crash on NFS re-export node
3617 Details : under very unusual load conditions an assertion is hit in
3618 ll_intent_file_open()
3621 Frequency : only if OST filesystem is corrupted
3623 Description: client incorrectly hits assertion in ptlrpc_replay_req()
3624 Details : for a short time RPCs with bulk IO are in the replay list,
3625 but replay of bulk IOs is unimplemented. If the OST filesystem
3626 is corrupted due to disk cache incoherency and then replay is
3627 started it is possible to trip an assertion. Avoid putting
3628 committed RPCs into the replay list at all to avoid this issue.
3633 Description: large O_DIRECT requests fail under memory pressure/fragmentation
3634 Details : Large single O_DIRECT read and write calls can fail to allocate
3635 a sufficiently large buffer to process the request. In case of
3636 allocation failure the allocation is retried with a smaller
3637 buffer and broken into smaller requests.
3639 Severity : enhancement
3641 Description: Add -o localflock option to simulate old noflock behaviour.
3642 Details : This will achieve local-only flock/fcntl locks coherentness.
3647 Description: versioning check is incomplete
3648 Details : Checking the version difference of client vs. server, report
3649 error if the gap is too big.
3654 Description: add support PG_writeback bit
3655 Details : add support for PG_writeback bit for Lustre, for more carefull
3656 work with page cache in 2.6 kernel. This also fix some deadlocks
3657 and remove hack for work O_SYNC with 2.6 kernel.
3659 Severity : enhancement
3661 Description: Add uninit_groups feature to ldiskfs2 to speed up e2fsck
3662 Details : The uninit_groups feature works in conjunction with the kernel
3663 filesystem code (ldiskfs2 only) and e2fsprogs-1.39-cfs6 to speed
3664 up the pass1 processing of e2fsck. This is a read-only feature
3665 in ldiskfs2 only, so older kernels and current ldiskfs cannot
3666 mount filesystems that have had this feature enabled.
3668 Severity : enhancement
3670 Description: Improve multi-block allocation algorithm to avoid fragmentation
3671 Details : The mballoc3 code (ldiskfs2 only) adds new mechanisms to improve
3672 allocation locality and avoid filesystem fragmentation.
3674 ------------------------------------------------------------------------------
3676 2007-02-09 Cluster File Systems, Inc. <info@clusterfs.com>
3678 * Support for kernels:
3679 2.6.9-42.0.3.EL (RHEL 4)
3680 2.6.5-7.276 (SLES 9)
3681 2.4.21-47.0.1.EL (RHEL 3)
3682 2.6.12.6 vanilla (kernel.org)
3683 2.6.16.21-0.8 (SLES10)
3684 * Recommended e2fsprogs version: 1.39.cfs2-0
3686 * The backwards-compatible /proc/sys/portals symlink has been removed
3687 in this release. Before upgrading, please ensure that you change
3688 any configuration scripts or /etc/sysctl.conf files that access
3689 /proc/sys/portals/* or sysctl portals.* to use the corresponding
3690 entry in /proc/sys/lnet or sysctl lnet.*. This change can be made
3691 in advance of the upgrade on any system running Lustre 1.4.6 or
3692 newer, since /proc/sys/lnet was added in that version.
3693 * Note that reiserfs quotas are disabled on SLES 10 in this kernel
3697 Frequency : only when quota is used
3699 Description: avoid scanning export list for quota master
3700 Details : Change the algorithms to avoid scanning export list in order
3701 to improve the efficiency.
3704 Frequency : MDS failover only, very rarely
3706 Description: "went back in time" messages on mds failover
3707 Details : The greatest transno may be lost when the current operation
3708 finishes with an error (transno==0) and the client's last_rcvd
3709 record is over-written. Save the greatest transno in the
3710 mds_last_transno for this case.
3713 Frequency : always for specific kernels and striping counts
3715 Description: client may get "Matching packet too big" without ACL support
3716 Details : Clients compiled without CONFIG_FS_POSIX_ACL get an error message
3717 when trying to access files in certain configurations. The
3718 clients should in fact be denied when mounting because they do
3719 not understand ACLs.
3722 Frequency : Cray XT3 with more than 4000 clients and multiple jobs
3724 Description: many clients connecting with IO in progress causes connect timeouts
3725 Details : Avoid synchronous journal commits to avoid delays caused by many
3726 clients connecting/disconnecting when bulk IO is in progress.
3727 Queue liblustre connect requests on OST_REQUEST_PORTAL instead of
3728 OST_IO_PORTAL to avoid delays behind potentially many pending
3732 Frequency : occasionally with multiple writers to a single file
3734 Description: shared writes to file may result in wrong size reported by stat()
3735 Details : Allow growing of kms when extent lock is cancelled
3738 Frequency : always with random mmap IO to multi-striped file
3740 Description: mmap write might be lost if we are writing to a 'hole' in stripe
3741 Details : Only if the hole is at the end of OST object so that kms is too
3742 small. Fix is to increase kms accordingly in ll_nopage.
3745 Frequency : rare, only if OST filesystem is inconsistent with MDS filesystem
3747 Description: writes to a missing object would leak memory on the OST
3748 Details : If there is an inconsistency between the MDS and OST filesystems,
3749 such that the MDS references an object that doesn't exist, writes
3750 to that object will leak memory due to incorrect cleanup in the
3751 error handling path, eventually running out of memory on the OST.
3756 Description: Creating too long symlink causes lustre errors
3757 Details : Check symlink and name lengths before sending requests to MDS.
3760 Frequency : only if flock is enabled (not on by default)
3762 Description: posix locks not released on fd closure on 2.6.9+
3763 Details : We failed to add posix locks to list of inode locks on 2.6.9+
3764 kernels, this caused such locks not to be released on fd close and
3765 then assertions on fs unmount about still used locks.
3768 Frequency : MDS failover only, very rarely
3770 Description: clients may get ASSERTION(granted_lock != NULL)
3771 Details : When request was taking a long time, and a client was resending
3772 a getattr by name lock request. The were multiple lock requests
3773 with the same client lock handle and
3774 mds_getattr_name->fixup_handle_for_resent_request found one of the
3775 lock handles but later failed with ASSERTION(granted_lock != NULL).
3780 Description: handle->h_buffer_credits > 0, assertion failure
3781 Details : h_buffer_credits is zero after truncate, causing assertion
3782 failure. This patch extends the transaction or creates a new
3786 Frequency : NFS re-export or patchless client
3787 Bugzilla : 11179, 10796
3788 Description: Crash on NFS re-export node (__d_move)
3789 Details : We do not want to hash the dentry if we don't have a lock.
3790 But if this dentry is later used in d_move, we'd hit uninitialised
3791 list head d_hash, so we just do this to init d_hash field but
3792 leave dentry unhashed.
3795 Frequency : NFS re-export or patchless client
3797 Description: NFS exports has problem with symbolic link
3798 Details : lustre client didn't properly install dentry when re-exported
3799 to NFS or running patchless client.
3802 Frequency : NFS re-export or patchless client
3804 Description: Various nfs/patchless fixes.
3805 Details : fixes reuse disconected alias for lookup process - this fixes
3806 warning "find_exported_dentry: npd != pd",
3807 fix permission error with open files at nfs.
3808 fix apply umask when do revalidate.
3811 Frequency : occasional
3813 Description: Crash on NFS re-export node
3814 Details : calling clear_page() on the wrong pointer triggered oops in
3815 generic_mapping_read().
3818 Frequency : rarely, using O_DIRECT IO
3820 Description: unaligned directio crashes client with LASSERT
3821 Details : check for unaligned buffers before trying any requests.
3824 Frequency : rarely, using CFS RAID5 patches in non-standard kernel series
3826 Description: stale data returned from RAID cache
3827 Details : If only a small amount of IO is done to the RAID device before
3828 reading it again it is possible to get stale data from the RAID
3829 cache instead of reading it from disk.
3832 Frequency : always for sles10 kernel
3834 Description: sles10 support
3835 Details : ll_follow_link: compile fixes and using of nd_set_link
3836 under newer kernels.
3839 Frequency : depends on arch, kernel and compiler version, always on sles10
3842 Description: recursive or deep enough symlinks cause stack overflow
3843 Details : getting rid of large stack-allocated variable in
3847 Frequency : depends on hardware
3849 Description: lustre write performance loss in the SLES10 kernel
3850 Details : the performance loss is caused by using of write barriers in the
3851 ext3 code. The SLES10 kernel turns barrier support on by
3852 default. The fix is to undo that change for ldiskfs.
3854 ------------------------------------------------------------------------------
3856 2006-12-09 Cluster File Systems, Inc. <info@clusterfs.com>
3858 * Support for kernels:
3859 2.6.9-42.0.3EL (RHEL 4)
3860 2.6.5-7.276 (SLES 9)
3861 2.4.21-47.0.1.EL (RHEL 3)
3862 2.6.12.6 vanilla (kernel.org)
3866 Frequency : quota enabled and large files being deleted
3868 Description: releasing more than 4GB of quota at once hangs OST
3869 Details : If a user deletes more than 4GB of files on a single OST it
3870 will cause the OST to spin in an infinite loop. Release
3871 quota in < 4GB chunks, or use a 64-bit value for 1.4.7.1+.
3876 Description: statfs data retrieved from /proc may be stale or zero
3877 Details : When reading per-device statfs data from /proc, in the
3878 {kbytes,files}_{total,free,avail} files, it may appear
3879 as zero or be out of date.
3882 Frequency : systems with MD RAID1 external journal devices
3884 Description: lconf's call to blkid is confused by RAID1 journal devices
3885 Details : Use the "blkid -l" flag to locate the MD RAID device instead
3886 of returning all block devices that match the journal UUID.
3889 Frequency : always, for aggregate stripe size over 4GB
3891 Description: "lfs setstripe" fails assertion when setting 4GB+ stripe width
3892 Details : Using "lfs setstripe" to set stripe size * stripe count over 4GB
3893 will fail the kernel with "ASSERTION(lsm->lsm_xfersize != 0)"
3896 Frequency : always if "lfs find" used on a local file/directory
3898 Description: "lfs find" segfaults if used on a local file/directory
3899 Details : The case where a directory component was not specified wasn't
3900 handled correctly. Handle this properly.
3903 Frequency : always on ppc64
3905 Description: the write to an ext3 filesystem mounted with mballoc got stuck
3906 Details : ext3_mb_generate_buddy() uses find_next_bit() which does not
3907 perform endianness conversion.
3910 Frequency : rarely (truncate to non-zero file size after write under load)
3911 Bugzilla : 10730, 10687
3912 Description: Files padded with zeros to next 4K multiple
3913 Details : With filesystems mounted using the "extents" option (2.6 kernels)
3914 it is possible that files that are truncated to a non-zero size
3915 immediately after being written are filled with zero bytes beyond
3916 the truncated size. No file data is lost.
3918 Severity : enhancement
3920 Description: Allow recovery/failover for liblustre clients.
3921 Details : liblustre clients were unaware of failover configurations until
3924 Severity : enhancement
3926 Description: user file locks should fail when not mounting with flock option
3927 Details : Set up an error-returning stub in ll_file_operations.lock field
3928 to prevent incorrect behaviour when client is mounted without
3929 flock option. Also, set up properly f_op->flock field for
3933 Frequency : always on ia64
3935 Description: "lfs df" loops on printing out MDS statfs information
3936 Details : The obd_ioctl_data was not initialized and in some systems
3937 this caused a failure during the ioctl that did not return
3938 an error. Initialize the struct and return an error on failure.
3941 Frequency : SLES 9 only
3943 Description: Error of copying files with lustre special EAs as root
3944 Details : Client side always return success for setxattr call for lustre
3945 special xattr (currently only "trusted.lov").
3948 Frequency : rarely on clusters with both ia64+i386 clients
3950 Description: ia64+i686 clients doing shared IO on the same file may LBUG
3951 Details : In rare cases when both ia64+i686 (or other mixed-PAGE_SIZE)
3952 clients are doing concurrent writes to the same file it is
3953 possible that the ia64 clients may LASSERT because the OST
3954 extent locks are not PAGE_SIZE aligned. Ensure that grown
3955 locks are always aligned on the request boundary.
3958 Frequency : specific use, occasional
3960 Description: Overwriting in use executable truncates on-disk binary image
3961 Details : If one node attempts to overwrite an executable in use by
3962 another node, we now correctly return ETXTBSY instead of
3963 truncating the file.
3965 Severity : enhancement
3967 Description: Async OSC create to avoid the blocking unnecessarily.
3968 Details : If a OST has no remain object, system will block on the creating
3969 when need to create a new object on this OST. Now, ways use
3970 pre-created objects when available, instead of blocking on an
3971 empty osc while others are not empty. If we must block, we block
3972 for the shortest possible period of time.
3977 Description: chmod on Lustre root is propagated to other clients
3978 Details : Re-validate root's dentry in ll_lookup_it to avoid having it
3979 invalid by the follow_mount time.
3982 Frequency : liblustre clients only
3984 Description: Race in 'instant cancel' lock handling could lead to such locks
3985 never to be granted in case of SMP MDS
3986 Details : Do not destroy not yet granted but cbpending locks in
3990 Frequency : replay/resend of open
3992 Description: non null lock assetion failure in mds_intent_policy
3993 Details : Trying to replay/resend lockless open requests resulted in
3994 mds_open() returning 0 with no lock. Now it sets a flag if
3995 a lock is going to be returned.
3997 Severity : enhancement
3999 Description: Checksum enhancements
4000 Details : New checksum enhancements allow for resending RPCs that failed
4003 Severity : enhancement
4005 Description: Tunables on number of dirty pages in cacche
4006 Details : Allow to set limit on number of dirty pages cached.
4011 Description: client crash on unmount - lock still has references
4012 Details : In some error handling cases it was possible to leak a lock
4013 reference on a client while accessing a file. This was not
4014 harmful to the client during operation, but would cause the
4015 client to crash when the filesystem is unmounted.
4018 Frequency : specific case, rare
4020 Description: ETXTBSY on mds though file not in use by client
4021 Details : ETXTBSY is no longer incorrectly returned when attempting to
4022 chmod or chown a directory that the user previously tried to
4023 execute or a currently-executing binary.
4026 Frequency : extremely rare except on liblustre-based clients
4028 Description: Lustre space not freed when files are deleted
4029 Details : Clean up open-unlinked files after client eviction. Previously
4030 the unlink was skipped and the files remained as orphans.
4035 Description: OST failure "would be an LBUG" in waiting_locks_callback()
4036 Details : In some cases it was possible to send a blocking callback to
4037 a client doing a glimpse, even though that client didn't get
4038 a lock granted. When the glimpse lock is cancelled on the OST
4039 the freed lock is left on the waiting list and corrupted the list.
4042 Frequency : all core dumps
4044 Description: Broke core dumps to lustre
4045 Details : Negative dentry may be unhashed if parent does not have UPDATE
4046 lock, but some callers, e.g. do_coredump, expect dentry to be
4047 hashed after successful create, hash it in ll_create_it.
4049 ------------------------------------------------------------------------------
4051 2006-09-13 Cluster File Systems, Inc. <info@clusterfs.com>
4053 * Support for kernels:
4054 2.6.9-42.0.2.EL (RHEL 4)
4055 2.6.5-7.276 (SLES 9)
4056 2.4.21-40.EL (RHEL 3)
4057 2.6.12.6 vanilla (kernel.org)
4061 Frequency : always on RHEL 3
4063 Description: Number of open files grows over time
4064 Details : The number of open files grows over time, whether or not
4065 Lustre is started. This was due to a filp leak introduced
4066 by one of our kernel patches.
4068 ------------------------------------------------------------------------------
4070 2006-08-20 Cluster File Systems, Inc. <info@clusterfs.com>
4072 * Support for kernels:
4073 2.6.9-42.EL (RHEL 4)
4074 2.6.5-7.267 (SLES 9)
4075 2.4.21-40.EL (RHEL 3)
4076 2.6.12.6 vanilla (kernel.org)
4081 Bugzilla : 5719, 9635, 9792, 9684
4082 Description: OST (or MDS) trips assertions in (re)connection under heavy load
4083 Details : If a server is under heavy load and cannot reply to new
4084 connection requests before the client resends the (re)connect,
4085 the connection handling code can behave badly if two service
4086 threads are concurrently handing separate (re)connections from
4087 the same client. Add better locking to the connection handling
4088 code, and ensure that only a single connection will be processed
4089 for a given client UUID, even if the lock is dropped.
4091 Severity : enhancement
4093 Description: add TCP zero-copy support to kernel
4094 Details : Add support to the kernel TCP stack to allow zero-copy bulk
4095 sends if the hardware supports scatter-gather and checksumming.
4096 This allows socklnd to do client-write and server-read more
4097 efficiently and reduce CPU utilization from skbuf copying.
4100 Frequency : only if NFS exporting from client
4102 Description: NULL pointer deref in ll_iocontrol() if chattr mknod file
4103 Details : If setting attributes on a file created under NFS that had
4104 never been opened it would be possible to oops the client
4105 if the file had no objects.
4109 Bugzilla : 9326, 10402, 10897
4110 Description: client crash in ptlrpcd_wake() thread when sending async RPC
4111 Details : It is possible that ptlrpcd_wake() dereferences a freed async
4112 RPC. In rare cases the ptlrpcd thread alread processed the RPC
4113 before ptlrpcd_wake() was called and the request was freed.
4116 Frequency : always for liblustre
4118 Description: liblustre client does MDS+OSTs setattr RPC for each write
4119 Details : When doing a write from a liblustre client, the client
4120 incorrectly issued an RPC to the MDS and each OST the file was
4121 striped over in order to update the timestamps. When writing
4122 with small chunks and many clients this could overwhelm the MDS
4123 with RPCs. In all cases it would slow down the write because
4124 these RPCs are unnecessary.
4126 Severity : enhancement
4128 Description: allow number of MDS service threads to be changed at module load
4129 Details : It is now possible to change the number of MDS service threads
4130 running. Adding "options mds mds_num_threads={N}" to the MDS's
4131 /etc/modprobe.conf will set the number of threads for the next
4132 time Lustre is restarted (assuming the "mds" module is also
4133 reloaded at that time). The default number of threads will
4134 stay the same, 32 for most systems.
4139 Description: OST crash if filesystem is unformatted or corrupt
4140 Details : If an OST is started on a device that has never been formatted
4141 or if the filesystem is corrupt and cannot even mount then the
4142 error handling cleanup routines would dereference a NULL pointer.
4147 Description: NULL pointer deref in llap_from_page.
4148 Details : get_cache_page_nowait can return a page with NULL (or otherwise
4149 incorrect) mapping if the page was truncated/reclaimed while it was
4150 searched for. Check for this condition and skip such pages when
4151 doing readahead. Introduce extra check to llap_from_page() to
4152 verify page->mapping->host is non-NULL (so page is not anonymous).
4155 Frequency : Sometimes when using sys_sendfile
4157 Description: "page not covered by a lock" warnings from ll_readpage
4158 Details : sendfile called ll_readpage without right page locks present.
4159 Now we introduced ll_file_sendfile that does necessary locking
4160 around call to generic_file_sendfile() much like we do in
4164 Frequency : with certain MDS communication failures at client mount time
4166 Description: NULL pointer deref after failed client mount
4167 Details : a client connection request may delayed by the network layer
4168 and not be sent until after the PTLRPC layer has timed out the
4169 request. If the client fails the mount immediately it will try
4170 to clean up before the network times out the request. Add a
4171 reference from the request import to the obd device and delay
4172 the cleanup until the network drops the request.
4175 Frequency : occasionally during client (re)connect
4177 Description: assertion failure during client (re)connect
4178 Details : processing a client connection request may be delayed by the
4179 client or server longer than the client connect timeout. This
4180 causes the client to resend the connection request. If the
4181 original connection request is replied in this interval, the
4182 client may trip an assertion failure in ptlrpc_connect_interpret()
4183 which thought it would be the only running connect process.
4186 Frequency : only with obd_echo servers and clients that are rebooted
4188 Description: kernel BUG accessing uninitialized data structure
4189 Details : When running an obd_echo server it did not start the ping_evictor
4190 thread, and when a client was evicted an uninitialized data
4191 structure was accessed. Start the ping_evictor in the RPC
4192 service startup instead of the OBD startup.
4194 Severity : enhancement
4195 Bugzilla : 10193 (patchless)
4196 Description: Remove dependency on various unexported kernel interfaces.
4197 Details : No longer need reparent_to_init, exit_mm, exit_files,
4198 sock_getsockopt, filemap_populate, FMODE_EXEC, put_filp.
4201 Frequency : rare (only users of deprecated and unsupported LDAP config)
4203 Description: write_conf for zeroconf mount queried LDAP incorrectly for client
4204 Details : LDAP apparently contains 'lustreName' attributes instead of
4205 'name'. A simple remapping of the name is sufficient.
4208 Frequency : rare (only with non-default dump_on_timeout debug enabled)
4210 Description: waiting_locks_callback trips kernel BUG if client is evicted
4211 Details : Running with the dump_on_timeout debug flag turned on makes
4212 it possible that the waiting_locks_callback() can try to dump
4213 the Lustre kernel debug logs from an interrupt handler. Defer
4214 this log dumping to the expired_lock_main() thread.
4216 Severity : enhancement
4218 Description: Support NFS exporting on 2.6 kernels.
4219 Details : Implement non-rawops metadata methods for NFS server to use without
4220 changing NFS server code.
4223 Frequency : very rare (synthetic metadata workload only)
4225 Description: two racing renames might cause an MDS thread to deadlock
4226 Details : Running the "racer" program may cause one MDS thread to rename
4227 a file from being the source of a rename to being the target of
4228 a rename at exactly the same time that another thread is doing
4229 so, and the second thread has already enqueued these locks after
4230 doing a lookup of the target and is trying to relock them in
4231 order. Ensure that we don't try to re-lock the same resource.
4234 Frequency : only very large systems with liblustre clients
4236 Description: slow eviction of liblustre clients with the "evict_by_nid" RPC
4237 Details : Use asynchronous set_info RPCs to send the "evict_by_nid" to
4238 all OSTs in parallel. This allows the eviction of stale liblustre
4239 clients to proceed much faster than if they were done in series,
4240 and also offers similar improvements for other set_info RPCs.
4245 Description: excessive CPU usage during initial read phase on client
4246 Details : During the initial read phase on a client, it would agressively
4247 retry readahead on the file, consuming too much CPU and impacting
4248 performance (since 1.4.5.8). Improve the readahead algorithm
4249 to avoid this, and also improve some other common cases (read
4250 of small files in particular, where "small" is files smaller than
4251 /proc/fs/lustre/llite/*/max_read_ahead_whole_mb, 2MB by default).
4256 Description: MDS crash when receiving packet with unknown intent.
4257 Details : Do not LBUG in unknown intent case, just return -EFAULT
4259 Severity : enhancement
4260 Bugzilla : 9293, 9385
4261 Description: MDS RPCs are serialised on client. This is unnecessary for some.
4262 Details : Do not serialize getattr (non-intent version) and statfs.
4265 Frequency : occasional, when OST network is overloaded/intermittent
4267 Description: client evicted by OST after bulk IO timeout
4268 Details : If a client sends a bulk IO request (read or write) the OST
4269 may evict the client if it is unresposive to its data GET/PUT
4270 request. This is incorrect if the network is overloaded (takes
4271 too long to transfer the RPC data) or dropped the OST GET/PUT
4272 request. There is no need to evict the client at all, since
4273 the pinger and/or lock callbacks will handle this, and the
4274 client can restart the bulk request.
4277 Frequency : Always when mmapping file with no objects
4279 Description: client crashes when mmapping file with no objects
4280 Details : Check that we actually have objects in a file before doing any
4281 operations on objects in ll_vm_open, ll_vm_close and
4287 Description: Request leak when working with deleted CWD
4288 Details : Introduce advanced request refcount tracking for requests
4289 referenced from lustre intent.
4291 Severity : Enhancement
4293 Description: Cache open file handles on client.
4294 Details : MDS now will return special lock along with openhandle, if
4295 requested and client is allowed to hold openhandle, even if unused,
4296 until such a lock is revoked. Helps NFS a lot, since NFS is opening
4297 closing files for every read/write openration.
4299 Severity : Enhancement
4301 Description: Cache open negative dentries on client when possible.
4302 Details : Guard negative dentries with UPDATE lock on parent dir, drop
4303 negative dentries on lock revocation.
4308 Description: Remounting a client read-only wasn't possible with a zconf mount
4309 Details : It wasn't possible to remount a client read-only with llmount.
4311 Severity : enhancement
4312 Description: Include MPICH 1.2.6 Lustre ADIO interface patch
4313 Details : In lustre/contrib/ or /usr/share/lustre in RPM a patch for
4314 MPICH is included to add Lustre-specific ADIO interfaces.
4315 This is based closely on the UFS ADIO layer and only differs
4316 in file creation, in order to allow the OST striping to be set.
4317 This is user-contributed code and not supported by CFS.
4322 Description: extended inode attributes (immutable, append-only) work improperly
4323 when 2.4 and 2.6 kernels are used on client/server or vice versa
4324 Details : Introduce kernel-independent values for these flags.
4326 Severity : enhancement
4329 Description: Allow fractional MB tunings for lustre in /proc/ filesystem.
4330 Details : Many of the /proc/ tunables can only be tuned at a megabyte
4331 granularity. Now, Fractional MB granularity is be supported,
4332 this is very useful for low memory system.
4334 Severity : enhancement
4336 Description: Getattr by fid
4337 Details : Getting a file attributes by its fid, obtaining UPDATE|LOOKUP
4338 locks, avoids extra getattr rpc requests to MDS, allows '/' to
4339 have locks and avoids getattr rpc requests for it on every stat.
4342 Frequency : Always, for filesystems larger than 2TB
4344 Description: ldiskfs crash at mount for filesystem larger than 2TB with mballoc
4345 Details : Kenrel kmalloc limits allocations to 128kB and this prevents
4346 filesystems larger than 2TB to be mounted with mballoc enabled.
4349 Frequency : Always, for 32-bit kernel without CONFIG_LBD and filesystem > 2TB
4351 Description: filesystem corruption for non-standard kernels and very large OSTs
4352 Details : If a 32-bit kernel is compiled without CONFIG_LBD enabled and a
4353 filesystems larger than 2TB is mounted then the kernel will
4354 silently corrupt the start of the filesystem. CONFIG_LBD is
4355 enabled for all CFS-supported kernels, but the possibility of
4356 this happening with a modified kernel config exists.
4358 Severity : enhancement
4360 Description: add client O_DIRECT support for 2.6 kernels
4361 Details : It is now possible to do O_DIRECT reads and writes to files
4362 in the Lustre client mountpoint on 2.6 kernel clients.
4364 Severity : enhancement
4366 Description: parallel glimpse, setattr, statfs, punch, destroy requests
4367 Details : Sends glimpse, setattr, statfs, punch, destroy requests to OSTs in
4368 parallel, not waiting for response from every OST before sending
4369 a rpc to the next OST.
4374 Description: setattr vs write race when updating file timestamps
4375 Details : Client processes that update a file timestamp into the past
4376 right after writing to the file (e.g. tar) it is possible that
4377 the updated file modification time can be reset to the current
4378 time due to a race between processing the setattr and write RPC.
4380 Severity : enhancement
4382 Description: Bring 'lfs find' closer in line with regular Linux find.
4383 Details : lfs find util supports -atime, -mtime, -ctime, -maxdepth, -print,
4384 -print0 options and obtains all the needed info through the lustre
4387 Severity : enhancement
4389 Description: support up to 1024 configured devices on one node
4390 Details : change obd_dev array from statically allocated to dynamically
4391 allocated structs as they are first used to reduce memory usage
4396 Description: Flush dirty partially truncated pages during truncate
4397 Details : Immediatelly flush partially truncated pages in filter_setattr,
4398 this way we completely avoid having any pages in page cache on OST
4399 and can retire ugly workarounds during writes to flush such pages.
4404 Description: i_sem vs transaction deadlock in mds_obd_destroy during unlink.
4405 Details : protect inode from truncation within vfs_unlink() context
4406 just take a reference before calling vfs_unlink() and release it
4407 when parent's i_sem is free.
4410 Frequency : always, if extents are used on OSTs
4412 Description: index ei_leaf_hi (48-bit extension) is not zeroed in extent index
4413 Details : OSTs using the extents format would not zero the high 16 bits of
4414 the index physical block number. This is not a problem for any
4415 OST filesystems smaller than 16TB, and no kernels support ext3
4416 filesystems larger than 16TB yet. This is fixed in 1.4.7 (all
4417 new/modified files) and can be fixed for existing filesystems
4418 with e2fsprogs-1.39-cfs1.
4423 Description: import connection selection may be incorrect if timer wraps
4424 Details : Using a 32-bit jiffies timer with HZ=1000 may cause backup
4425 import connections to be ignored if the 32-bit jiffies counter
4426 wraps. Use a 64-bit jiffies counter.
4429 Frequency : during server recovery
4431 Description: crash after server is denying duplicate export
4432 Details : If clients are resending connect requests to the server, the
4433 server refuses to allow a client to connect multiple times.
4434 Fixed a bug in the handling of this case.
4437 Frequency : very large clusters immediately after boot
4439 Description: LNET request buffers exhausted under heavy short-term load
4440 Details : If a large number of client requests are generated on a service
4441 that has previously never seen so many requests it is possible
4442 that the request buffer growth cannot keep up with the spike in
4443 demand. Instead of dropping incoming requests, they are held in
4444 the LND until the RPC service can accept more requests.
4447 Frequency : Sometimes during replay
4449 Description: Assertion failure in ll_local_open after replay.
4450 Details : If replay happened on an open request reply before we were able
4451 to set replay handler, reply will become not swabbed tripping the
4452 assertion in ll_local_open. Now we set the handler right after
4453 recognising of open request
4456 Frequency : very rare
4458 Description: kernel reports "badness in vsnprintf"
4459 Details : Reading from the "recovery_status" /proc file in small chunks
4460 may cause a negative length in lprocfs_obd_rd_recovery_status()
4461 call to vsnprintf() (which is otherwise harmless). Exit early
4462 if there is no more space in the output buffer.
4464 Severity : enhancement
4466 Description: clear OBD RPC statistics by writing to them
4467 Details : It is now possible to clear the OBD RPC statistics by writing
4468 to the "stats" file.
4473 Description: Client mtime is not the same on different clients after utimes
4474 Details : In some cases, the client was using the utimes() syscall on
4475 a file cached on another node. The clients now validate the
4476 ctime from the MDS + OSTs to determine which one is right.
4481 Description: Inability to activate failout mode
4482 Details : lconf script incorrectly assumed that in python string's numeric
4483 value is used in comparisons.
4486 Frequency : always with multiple stripes per file
4488 Description: Inefficient object allocation for mutli-stripe files
4489 Details : When selecting which OSTs to stripe files over, for files with
4490 a stripe count that divides evenly into the number of OSTs,
4491 the MDS is always picking the same starting OST for each file.
4492 Return the OST selection heuristic to the original design.
4497 Description: mount failures may take full timeout to return an error
4498 Details : Under some heavy load conditions it is possible that a
4499 failed mount can wait for the full obd_timeout interval,
4500 possibly several minutes, before reporting an error.
4501 Instead return an error as soon as the status is known.
4503 ------------------------------------------------------------------------------
4505 2006-02-14 Cluster File Systems, Inc. <info@clusterfs.com>
4507 * WIRE PROTOCOL CHANGE. This version of Lustre networking WILL NOT
4508 INTEROPERATE with older versions automatically. Please read the
4509 user documentation before upgrading any part of a live system.
4510 * WARNING: Lustre networking configuration changes are required with
4511 this release. See https://bugzilla.clusterfs.com/show_bug.cgi?id=10052
4514 * Support for kernels:
4515 2.6.9-22.0.2.EL (RHEL 4)
4516 2.6.5-7.244 (SLES 9)
4517 2.6.12.6 vanilla (kernel.org)
4520 Severity : enhancement
4521 Bugzilla : 7981/8208
4522 Description: Introduced Lustre Networking (LNET)
4523 Details : LNET is new networking infrastructure for Lustre, it includes
4524 a reorganized network configuration mode (see the user
4525 documentation for full details) as well as support for routing
4526 between different network fabrics. Lustre Networking Devices
4527 (LNDs) for the supported network fabrics have also been
4528 created for this new infrastructure.
4530 Severity : enhancement
4531 Description: Introduced Access control lists
4532 Details : clients can set ACLs on files and directories in order to have
4533 more fine-grained permissions than the standard Unix UGO+RWX.
4534 The MDS must be started with the "-o acl" mount option.
4536 Severity : enhancement
4537 Description: Introduced filesystem quotas
4538 Details : Administrators may now establish per-user quotas on the
4541 Severity : enhancement
4543 Description: Configuration change for the XT3
4544 The PTLLND is now used to run Lustre over Portals on the XT3
4545 The configure option(s) --with-cray-portals are no longer used.
4546 Rather --with-portals=<path-to-portals-includes> is used to
4547 enable building on the XT3. In addition to enable XT3 specific
4548 features the option --enable-cray-xt3 must be used.
4553 Description: Running on many-way SMP OSTs can trigger oops in llcd_send()
4554 Details : A race between allocating a new llcd and re-getting the llcd_lock
4555 allowed another thread to grab newly-allocated llcd.
4557 Severity : enhancement
4559 Description: 2.6 OST async journal commit and locking fix to improve performance
4560 Details : The filter_direct_io()+filter_commitrw_write() journal commits for
4561 2.6 kernels are now async as they already were in 2.4 kernels so
4562 that they can commit concurrently with the network bulk transfer.
4563 For block-allocated files the filter allocation semaphore is held
4564 to avoid filesystem fragmentation during allocation. BKL lock
4565 removed for 2.6 xattr operations where it is no longer needed.
4570 Description: lconf incorrectly determined whether two IP networks could talk
4571 Details : In some more complicated routing and multiple-network
4572 configurations, lconf will avoid trying to make a network
4573 connection to a disjoint part of the IP space. It was doing the
4574 math incorrectly for one set of cases.
4579 Description: Fix for potential infinite loop processing records in an llog.
4580 Details : If an llog record is corrupted/zeroed, it is possible to loop
4581 forever in llog_process(). Validate the llog record length
4582 and skip the remainder of the block on error.
4585 Frequency : occasional (liblustre only)
4587 Description: liblustre could not open files whose last component is a symlink
4588 Details : sysio_path_walk() would incorrectly pass the open intent to
4589 intermediate path components.
4592 Frequency : rare (liblustre only with non-standard tuning)
4593 Bugzilla : 7201 (7350)
4594 Description: Tuning the MDC DLM LRU size to zero triggers client LASSERT
4595 Details : llu_lookup_finish_locks() tries to set lock data on a lock
4596 after it has been released, only do this for referenced locks
4598 Severity : enhancement
4600 Description: specifying an (invalid) directory default stripe_size of -1
4601 would reset the directory default striping
4602 Details : stripe_size -1 was used internally to signal directory stripe
4603 removal, now use "all default" to signal dir stripe removal
4604 as a directory striping of "all default" is not useful
4607 Frequency : common for large clusters running liblustre clients
4609 Description: doing an ls when liblustre clients are running is slow
4610 Details : sending a glimpse AST to a liblustre client waits for every AST
4611 to time out, as liblustre clients will not respond. Since they
4612 cannot cache data we refresh the OST lock LVB from disk instead.
4614 Severity : enhancement
4616 Description: doing an ls at the same time as file IO can be slow
4617 Details : enqueue and other "small" requests can be blocked behind many
4618 large IO requests. Create a new OST IO portal for non-IO
4619 requests so they can be processed faster.
4622 Frequency : rare (only HPUX clients mounting unsupported re-exported NFS vol)
4624 Description: an HPUX NFS client would get -EACCESS when ftruncate()ing a newly
4625 created file with mode 000
4626 Details : the Linux NFS server relies on an MDS_OPEN_OWNEROVERRIDE hack to
4627 allow an ftruncate() as a non-root user to a file with mode 000.
4628 Lustre now respects this flag to disable mode checks when
4629 truncating a file owned by the user
4632 Frequency : liblustre-only, when liblustre client dies unexpectedly or becomes
4635 Description: Revoking locks from clients that went dead or catatonic might take
4637 Details : New lock flags FL_CANCEL_ON_BLOCK used by liblustre makes
4638 cancellation of such locks instant on servers without waiting for
4639 any reply from clients. Clients drops these locks when cancel
4640 notification from server is received without replying.
4643 Frequency : liblustre-only, when liblustre client dies or becomes busy
4645 Description: Doing ls on Linux clients can take a long time with active
4647 Details : Liblustre client cannot handle ASTs in timely manner, so avoid
4648 granting such locks to it in the first place if possible. Locks
4649 are taken by proxy on the OST during the read or write and
4650 dropped immediately afterward. Add connect flags handling, do
4651 not grant locks to liblustre clients for glimpse ASTs.
4653 Severity : enhancement
4655 Description: Improve read-ahead algorithm to avoid excessive IO for random reads
4656 Details : Existing read-ahead algorithm is tuned for the case of streamlined
4657 sequential reads and behaves badly with applications doing random
4658 reads. Improve it by reading ahead at least read region, and
4659 avoiding excessive large RPC for small reads.
4661 Severity : enhancement
4663 Description: Creating more than 1000 files for a single job may cause a load
4664 imbalance on the OSTs if there are also a large number of OSTs.
4665 Details : qos_prep_create() uses an OST index reseed value that is an
4666 even multiple of the number of available OSTs so that if the
4667 reseed happens in the middle of the object allocation it will
4668 still utilize the OSTs as uniformly as possible.
4673 Description: OST or MDS may oops in ping_evictor_main()
4674 Details : ping_evictor_main() drops obd_dev_lock if deleting a stale export
4675 but doesn't restart at beginning of obd_exports_timed list
4678 Severity : enhancement
4680 Description: improve by-nid export eviction on the MDS and OST
4681 Details : allow multiple exports with the same NID to be evicted at one
4682 time without re-searching the exports list.
4685 Frequency : rare, only with supplementary groups enabled on SMP 2.6 kernels
4687 Description: MDS may oops in groups_free()
4688 Details : in rare race conditions a newly allocated group_info struct is
4689 freed again, and this can be NULL. The 2.4 compatibility code
4690 for groups_free() checked for a NULL pointer, but 2.6 did not.
4693 Frequency : common for liblustre clients doing little filesystem IO
4694 Bugzilla : 9352, 7313
4695 Description: server may evict liblustre clients accessing contended locks
4696 Details : if a client is granted a lock or receives a completion AST
4697 with a blocking AST already set it would not reply to the AST
4698 for LDLM_FL_CANCEL_ON_BLOCK locks. It now replies to such ASTs.
4701 Frequency : lfs setstripe, only systems with more than 160 OSTs
4703 Description: unable to set striping with a starting offset beyond OST 160
4704 Details : llapi_create_file() incorrectly limited the starting stripe
4705 index to the maximum single-file stripe count.
4708 Frequency : LDAP users only
4710 Description: lconf did not handle in-kernel recovery with LDAP properly
4711 Details : lconf/LustreDB get_refs() is searching the wrong namespace
4713 Severity : enhancement
4715 Description: bind OST threads to NUMA nodes to improve performance
4716 Details : all OST threads are uniformly bound to CPUs on a single NUMA
4717 node and do their allocations there to localize memory access
4719 Severity : enhancement
4721 Description: llmount can determine client NID directly from Myrinet (GM)
4722 Details : the client NID code from gmnalnid was moved directly into
4723 llmount, removing the need to use this or specifying the
4724 client NID explicitly when mounting GM clients with zeroconf
4727 Frequency : if client is started with down MDS
4729 Description: if client is started with down MDS mount hangs in ptlrpc_queue_wait
4730 Details : Having an LWI_INTR() wait event (interruptible, but no timeout)
4731 will wait indefinitely in ptlrpc_queue_wait->l_wait_event() after
4732 ptlrpc_import_delayed_req() because we didn't check if the
4733 request was interrupted, and we also didn't break out of the
4734 event loop if there was no timeout
4739 Description: data loss during non-page-aligned writes to a single file from
4740 both multiple nodes and multiple threads on one node at same time
4741 Details : updates to KMS and lsm weren't protected by common lock. Resulting
4742 inconsistency led to false short-reads, that were cached and later
4743 used by ->prepare_write() to fill in partially written page,
4744 leading to data loss.
4747 Frequency : always, if lconf --abort_recovery used
4749 Description: lconf --abort_recovery fails with 'Operation not supported'
4750 Details : lconf was attempting to abort recovery on the MDT device and not
4753 Severity : enhancement
4755 Description: remove cleanup logs
4756 Details : replace lconf-generated cleanup logs with lustre internal
4757 cleanup routines. Eliminates the need for client-cleanup and
4760 Severity : enhancement
4762 Description: add support for EAs (user and system) on lustre filesystems
4763 Details : it is now possible to store extended attributes in the Lustre
4764 client filesystem, and with the user_xattr mount option it
4765 is possible to allow users to store EAs on their files also
4767 Severity : enhancement
4769 Description: Add possibility (config option) to show minimal available OST free
4771 Details : When compiled with --enable-mindf configure option, statfs(2)
4772 (and so, df) will return least minimal free space available from
4773 all OSTs as amount of free space on FS, instead of summary of
4774 free spaces of all OSTs.
4776 Severity : enhancement
4778 Description: do not expand extent locks acquired on OST-side
4779 Details : Modify ldlm_extent_policy() to not expand local locks, acquired
4780 by server: they are not cached anyway.
4783 Frequency : when mmap is used/binaries executed from Lustre
4785 Description: Unmmap pages before throwing them away from read cache.
4786 Details : llap_shrink cache now attempts to unmap pages before discarding
4787 them (if unmapping failed - do not discard). SLES9 kernel has
4788 extra checks that trigger if this unmapping is not done first.
4793 Description: lconf didn't resolve symlinks before checking to see whether a
4794 given mountpoint was already in use
4797 Frequency : when migrating failover services
4798 Bugzilla : 6395, 9514
4799 Description: When migrating a subset of services from a node (e.g. failback
4800 from a failover service node) the remaining services would
4801 time out and evict clients.
4802 Details : lconf --force (implied by --failover) sets the global obd_timeout
4803 to 5 seconds in order to quickly disconnect, but this caused
4804 other RPCs to time out too quickly. Do not change the global
4805 obd_timeout for force cleanup, only set it for DISCONNECT RPCs.
4807 Severity : enhancement
4808 Frequency : if MDS is started with down OST
4809 Bugzilla : 9439,5706
4810 Description: Allow startup/shutdown of an MDS without depending on the
4811 availability of the OSTs.
4812 Details : Asynchronously call mds_lov_synchronize during MDS startup.
4813 Add appropriate locking and lov-osc refcounts for safe
4814 cleaning. Add osc abort_inflight calls in case the
4815 synchronize never started.
4818 Frequency : occasional (Cray XT3 only)
4820 Description: root not authorized to access files in CRAY_PORTALS environment
4821 Details : The client process capabilities were not honoured on the MDS in
4822 a CRAY_PORTALS/CRAY_XT3 environment. If the file had previously
4823 been accessed by an authorized user then root was able to access
4824 the file on the local client also. The root user capabilities
4825 are now allowed on the MDS, as this environment has secure UID.
4828 Frequency : occasional
4830 Description: ldiskfs "too long searching" message happens too often
4831 Details : A debugging message (otherwise harmless) prints too often on
4832 the OST console. This has been reduced to only happen when
4833 there are fragmentation problems on the filesystem.
4838 Description: Division by zero in statfs when all OSCs are inactive
4839 Details : lov_get_stripecnt() returns zero due to incorrect order of checks,
4840 lov_statfs divides by value returned by lov_get_stripecnt().
4844 Bugzilla : 9489, 3273
4845 Description: First write from each client to each OST was only 4kB in size,
4846 to initialize client writeback cache, which caused sub-optimal
4847 RPCs and poor layout on disk for the first writen file.
4848 Details : Clients now request an initial cache grant at (re)connect time
4849 and so that they can start streaming writes to the cache right
4850 away and always do full-sized RPCs if there is enough data.
4851 If the OST is rebooted the client also re-establishes its grant
4852 so that client cached writes will be honoured under the grant.
4857 Description: Slow ls (and stat(2) syscall) on files residing on IO-loaded OSTs
4858 Details : Now I/O RPCs go to different portal number and (presumably) fast
4859 lock requests (and glimses) and other RPCs get their own service
4860 threads pool that should be able to service those RPCs
4863 Severity : enhancement
4865 Description: Ability to exchange lustre version between client and servers and
4866 issue warnings at client side if client is too old. Also for
4867 liblustre clients there is ability to refuse connection of too old
4869 Details : New 'version' field is added to connect data structure that is
4870 filled with version info. That info is later checked by server and
4874 Frequency : rare, liblustre only.
4875 Bugzilla : 9296, 9581
4876 Description: Two simultaneous writes from liblustre at offset within same page
4877 might proceed at the same time overwriting eachother with stale
4879 Details : I/O lock withing llu_file_prwv was released too early, before data
4880 actually was hitting the wire. Extended lock-holding time until
4881 server acknowledges receiving data.
4884 Frequency : extremely rare. Never observed in practice.
4886 Description: avoid generating lustre_handle cookie of 0.
4887 Details : class_handle_hash() generates handle cookies by incrementing
4888 global counter, and can hit 0 occasionaly (this is unlikely, but
4889 not impossible, because initial value of cookie counter is
4890 selected randonly). Value of 0 is used as a sentinel meaning
4891 "unassigned handle" --- avoid it. Also coalesce two critical
4892 sections in this function into one.
4894 Severity : enhancement
4896 Description: allow liblustre clients to delegate truncate locking to OST
4897 Details : To avoid overhead of locking, liblustre client instructs OST to
4898 take extent lock in ost_punch() on client's behalf. New connection
4899 flag is added to handle backward compatibility.
4901 Severity : enhancement
4902 Bugzilla : 4928, 7341, 9758
4903 Description: allow number of OST service threads to be specified
4904 Details : a module parameter allows the number of OST service threads
4905 to be specified via "options ost ost_num_threads={N}" in the
4906 OSS's /etc/modules.conf or /etc/modprobe.conf.
4910 Bugzilla : 6146, 9635, 9895
4911 Description: servers crash with bad pointer in target_handle_connect()
4912 Details : In rare cases when a client is reconnecting it was possible that
4913 the connection request was the last reference for that export.
4914 We would temporarily drop the export reference and get a new
4915 one, but this may have been the last reference and the export
4916 was just destroyed. Get new reference before dropping old one.
4918 Severity : enhancement
4919 Frequency : if client is started with failover MDS
4921 Description: Allow multiple MDS hostnames in the mount command
4922 Details : Try to read the configuration from all specified MDS
4923 hostnames during a client mount in case the "primary"
4926 Severity : enhancement
4928 Description: Stop sending data to evicted clients as soon as possible.
4929 Details : Check if the client we are about to send or are sending data to
4930 was evicted already. (Check is done every second of waiting,
4931 for which l_wait_event interface was extended to allow checking
4932 of exit condition at specified intervals).
4935 Frequency : rare, normally only when NFS exporting is done from client
4937 Description: 'bad disk LOV MAGIC: 0x00000000' error when chown'ing files
4939 Details : Make mds_get_md() recognise empty md case and set lmm size to 0.
4942 Frequency : always, if srand() is called before liblustre initialization
4944 Description: Liblustre uses system PRNG disturbing its usage by user application
4945 Details : Introduce internal to lustre fast and high-quality PRNG for
4946 lustre usage and make liblustre and some other places in generic
4947 lustre code to use it.
4949 Severity : enhancement
4950 Bugzilla : 9477, 9557, 9870
4951 Description: Verify that the MDS configuration logs are updated when xml is
4952 Details : Check if the .xml configuration logs are newer than the config
4953 logs stored on the MDS and report an error if this is the case.
4954 Request --write-conf, or allow starting with --old_conf.
4956 Severity : enhancement
4958 Description: Handle symlinks in the path when checking if Lustre is mounted.
4959 Details : Resolve intermediate symlinks when checking if a client has
4960 mounted a filesystem to avoid duplicate client mounts.
4965 Description: lconf can hit an error exception but still return success.
4966 Details : The lconf command catches the Command error exception at the top
4967 level script context and will exit with the associated exit
4968 status, but doesn't ensure that this exit status is non-zero.
4973 Description: failure of ptlrpc thread startup can cause oops
4974 Details : Starting a ptlrpc service thread can fail if there are a large
4975 number of threads or the server memory is very fragmented.
4976 Handle this without oopsing.
4979 Frequency : always, only if liblustre and non-default acceptor port was used
4981 Description: liblustre cannot connect to servers with non-default acceptor port
4982 Details : tcpnal_set_default_params() was not called and was therefore
4983 ignoring the environment varaible TCPNAL_PORT, as well as other
4984 TCPNAL_ environment variables
4989 Description: two objects could be created on the same OST for a single file
4990 Details : If an OST is down, in some cases it was possible to create two
4991 objects on a single OST for a single file. No problems other
4992 than potential performance impact and spurious error messages.
4996 Bugzilla : 5681, 9562
4997 Description: Client may oops in ll_unhash_aliases
4998 Details : Client dcache may become inconsistent in race condition.
4999 In some cases "getcwd" can fail if the current directory is
5005 Description: Inode refcounting problems in NFS export code
5006 Details : link_raw functions used to call d_instantiate without obtaining
5007 extra inode reference first.
5011 Bugzilla : 9942, 9903
5012 Description: Referencing freed requests leading to crash, memleaks with NFS.
5013 Details : We used to require that call to ll_revalidate_it was always
5014 followed by ll_lookup_it. Also with revalidate_special() it is
5015 possible to call ll_revalidate_it() twice for the same dentry
5016 even if first occurence returned success. This fix changes semantic
5017 between DISP_ENQ_COMPLETE disposition flag to mean there is extra
5018 reference on a request referred from the intent.
5019 ll_intent_release() then releases such a request.
5022 Frequency : rare, normally benchmark loads only
5024 Description: unlinked inodes were kept in memory on the client
5025 Details : If a client is repeatedly creating and unlinking files it
5026 can accumulate a lot of stale inodes in the inode slab cache.
5027 If there is no other client load running this can cause the
5028 client node to run out of memory. Instead flush old inodes
5029 from client cache that have the same inode number as a new inode.
5032 Frequency : SLES9 2.6.5 kernel and long filenames only
5033 Bugzilla : 9969, 10379
5034 Description: utime reports stale NFS file handle
5035 Details : SLES9 uses out-of-dentry names in some cases, which confused
5036 the lustre dentry revalidation. Change it to always use the
5040 Frequency : rare, unless heavy write-truncate concurrency is continuous
5041 Bugzilla : 4180, 6984, 7171, 9963, 9331
5042 Description: OST becomes very slow and/or deadlocked during object unlink
5043 Details : filter_destroy() was holding onto the parent directory lock
5044 while truncating+unlinking objects. For very large objects this
5045 may block other threads for a long time and slow overall OST
5046 responsiveness. It may also be possible to get a lock ordering
5047 deadlock in this case, or run out of journal credits because of
5048 the combined truncate+unlink. Solution is to do object truncate
5049 first in one transaction without parent lock, and then do the
5050 final unlink in a new transaction with the parent lock. This
5051 reduces the lock hold time dramatically.
5054 Frequency : rare, 2.4 kernels only
5056 Description: MDS or OST cleanup may trip kernel BUG when dropping kernel lock
5057 Details : mds_cleanup() and filter_cleanup() need to drop the kernel lock
5058 before unmounting their filesystem in order to avoid deadlock.
5059 The kernel_locked() function in 2.4 kernels only checks whether
5060 the kernel lock is held, not whether it is this process that is
5061 holding it as 2.6 kernels do.
5066 Description: MDS or OST may oops/LBUG if a client is connecting multiple times
5067 Details : The client ptlrpc code may be trying to reconnect to a down
5068 server before a previous connection attempt has timed out.
5069 Increase the reconnect interval to be longer than the connection
5070 timeout interval to avoid sending duplicate connections to
5074 Frequency : echo_client brw_test command
5076 Description: fix echo_client to work with OST preallocated code
5077 Details : OST preallocation code (5137) didn't take echo_client IO path
5078 into account: echo_client calls filter methods outside of any
5079 OST thread and, hence, there is no per-thread preallocated
5080 pages and buffers to use. Solution: hijack pga pages for IO. As
5081 a byproduct, this avoids unnecessary data copying.
5085 Bugzilla : 3555, 5962, 6025, 6155, 6296, 9574
5086 Description: Client can oops in mdc_commit_close() after open replay
5087 Details : It was possible for the MDS to return an open request with no
5088 transaction number in mds_finish_transno() if the client was
5089 evicted, but without actually returning an error. Clients
5090 would later try to replay that open and may trip an assertion
5091 Simplify the client close codepath, and always return an error
5092 from the MDS in case the open is not successful.
5095 Frequency : rare, 2.6 OSTs only
5097 Description: OST may deadlock under high load on fragmented files
5098 Details : If there was a heavy load and highly-fragmented OST filesystems
5099 it was possible to have all the OST threads deadlock waiting on
5100 allocation of biovecs, because the biovecs were not released
5101 until the entire RPC IO was completed. Instead, release biovecs
5102 as soon as they are complete to ensure forward IO progress.
5104 Severity : enhancement
5106 Description: Support for specifying external journal device at mount
5107 Details : If an OST or MDS device is formatted with an external journal
5108 device, this device major/minor is stored in the ext3 superblock
5109 and may not be valid for failover. Allow detecting and
5110 specifying the external journal at mount time.
5115 Description: Mounting an MDS with pending unlinked files may cause oops
5116 Details : target_finish_recovery() calls mds_postrecov() which returned
5117 the number of orphans unlinked. mds_lov_connect->mds_postsetup()
5118 considers this an error and immediately begins cleaning up the
5119 lov, just after starting the mds_lov process
5121 Severity : enhancement
5123 Description: Implement 'lfs df' to report actual free space on per-OST basis
5124 Details : Add sub-command 'df' on 'lfs' to report the disk space usage of
5125 MDS/OSDs. Usage: lfs df [-i][-h]. Command Options: '-i' to report
5126 usage of objects; '-h' to report in human readable format.
5128 ------------------------------------------------------------------------------
5130 2005-08-26 Cluster File Systems, Inc. <info@clusterfs.com>
5137 Description: Mounting an ldiskfs file system with mballoc may crash OST node.
5138 Details : ldiskfs mballoc code may reference an uninitialized buddy struct
5139 at startup during orphan unlinking. Instead, skip buddy update
5140 before setup, as it will be regenerated after recovery is complete.
5145 Description: If an OST is inactive, its locks might reference stale inodes.
5146 Details : lov_change_cbdata() must iterate over all namespaces, even if
5147 they are inactive to clear inode references from the lock.
5149 Severity : enhancement
5150 Frequency : occasional, if non-standard max_dirty_mb used
5152 Description: Client will block write RPCs if not enough grant
5153 Details : If a client has max_dirty_mb smaller than max_rpcs_in_flight,
5154 then the client will block writes while waiting for another RPC
5155 to complete instead of consuming its dirty limit. With change
5156 we get improved performance when max_dirty_mb is small.
5158 Severity : enhancement
5159 Bugzilla : 3389, 6253
5160 Description: Add support for supplementary groups on the MDS.
5161 Details : The MDS has an upcall /proc/fs/lustre/mds/{mds}/group_upcall
5162 (set to /usr/sbin/l_getgroups if enabled) which will do MDS-side
5163 lookups for user supplementary groups into a cache.
5167 Description: O_CREAT|O_EXCL open flags in liblustre always return -EEXIST
5168 Details : Make libsysio to not enforce O_EXCL by clearing the flag,
5169 for liblustre O_EXCL is enforced by MDS.
5173 Description: readdir never returns NULL in liblustre.
5174 Details : Corrected llu_iop_getdirentries logic, to return offset of next
5175 dentry in struct dirent.
5179 Frequency : liblustre only, depends on application IO pattern
5180 Description: liblustre clients evicted if not contacting servers
5181 Details : Don't put liblustre clients into the ping_evictor list, so
5182 they will not be evicted by the pinger ever.
5184 Severity : enhancement
5186 Description: Add ability to evict clients by NID from MDS.
5187 Details : By echoing "nid:$NID" string into
5188 /proc/fs/lustre/mds/.../evict_client client with nid that equals to
5189 $NID would be instantly evicted from this MDS and from all active
5190 OSTs connected to it.
5194 Description: Do not query file size twice, somewhat slowing stat(2) calls.
5195 Details : lookup_it_finish() used to query file size from OSTs that was not
5200 Description: service threads change working directory to that of init
5201 Details : Starting lustre service threads may pin the working directory
5202 of the parent thread, making that filesystem busy. Threads
5203 now change to the working directory of init to avoid this.
5207 Frequency : during shutdown only
5208 Description: shutdown with a failed MDS or OST can cause unmount to hang
5209 Details : Don't resend DISCONNECT messages in ptlrpc_disconnect_import()
5214 Frequency : 2.6 only
5215 Description: chmod/chown may include an extra supplementary group
5216 Details : ll{,u}_mdc_pack_op_data() does not properly initialize the
5217 supplementary group and if none is specified this is used.
5220 Bugzilla : 5479 (6816)
5222 Description: Racing open + rm can assert client in mdc_set_open_replay_data()
5223 Details : If lookup is in progress on a file that is unlinked we might try
5224 to revalidate the inode and fail in revalidate after lookup is
5225 complete and ll_file_open() enqueues the open again but
5226 it_open_error() was not checking DISP_OPEN_OPEN errors correctly.
5229 Frequency : always, if lconf --abort_recovery used
5231 Description: lconf --abort_recovery fails with 'Operation not supported'
5232 Details : lconf was attempting to abort recovery on the MDT device and not
5235 ------------------------------------------------------------------------------
5237 2005-08-08 Cluster File Systems, Inc. <info@clusterfs.com>
5242 Frequency : rare (only unsupported configurations with a node running as an
5244 Bugzilla : 6514, 5137
5245 Description: Mounting a Lustre file system on a node running as an OST could
5247 Details : OSTs now preallocates memory needed to write out data at
5248 startup, instead of when needed, to avoid having to
5249 allocate memory in possibly low memory situations.
5250 Specifically, if the file system is mounted on on OST,
5251 memory pressure could force it to try to write out data,
5252 which it needed to allocate memory to do. Due to the low
5253 memory, it would be unable to do so and the node would
5254 become unresponsive.
5256 Severity : enhancement
5258 Description: Addition of lconf --service command line option
5259 Details : lconf now accepts a '--service <arg>' option, which is
5260 shorthand for 'lconf --group <arg> --select <arg>=<hostname>'
5262 Severity : enhancement
5264 Description: Failover mode is now the default for OSTs.
5265 Details : By default, OSTs will now run in failover mode. To return to
5266 the old behaviour, add '--failout' to the lmc line for OSTs.
5268 Severity : enhancement
5270 Description: Health checks are now provided for MDS and OSTs
5271 Details : Additional detailed health check information on MSD and OSTs
5272 is now provided through the procfs health_check value.
5275 Frequency : occasional, depends on IO load
5277 Description: Disk fragmentation on the OSTs could eventually cause slowdowns
5278 after numerous create/delete cycles
5279 Details : The ext3 inode allocation policy would not allocate new inodes
5280 very well on the OSTs because there are no new directories
5281 being created. Instead we look for groups with free space if
5282 the parent directories are nearly full.
5287 Description: Network or server problems during mount may cause partially
5288 mounted clients instead of returning an error.
5289 Details : The config llog parsing code may overwrite the error return
5290 code during mount error handling, returning success instead
5296 Description: MDS can fail to allocate large reply buffers
5297 Details : After long uptimes the MDS can fail to allocate large reply
5298 buffers (e.g. zconf client mount config records) due to memory
5299 fragmentation or consumption by the buffer cache. Preallocate
5300 some large reply buffers so that these replies can be sent even
5301 under memory pressure.
5305 Frequency : rare (liblustre)
5306 Description: fsx running with liblustre complained that using truncate() to
5307 extend the file doesn't work. This patch corrects that issue.
5308 Details : This is the liblustre equivalent of the fix for bug 6196. Fixes
5309 ATTR_SIZE and lsm use in llu_setattr_raw.
5313 Frequency : rare, only 2.6 kernels
5314 Description: Unusual file access patterns on the MDS may result in inode
5315 data being lost in very rare circumstances.
5316 Details : Bad interaction between the ea-in-inode patch and the "no-read"
5317 code in the 2.6 kernel caused the inode and/or EA data not to
5318 be read from disk, causing single-file corruption.
5322 Frequency : rare, only 2.6 filesystems using extents
5323 Description: Heavy concurrent write and delete load may cause data corruption.
5324 Details : It was possible under high-load situations to have an extent
5325 metadata block in the block device cache from a just-unlinked
5326 file overwrite a newly-allocated data block. We now unmap any
5327 metadata buffers that alias just-allocated data blocks.
5331 Frequency : filesystems with default stripe_count larger than 77
5332 Description: lconf+mke2fs fail when formatting filesystem with > 77 stripes
5333 Details : lconf specifies an inode size of 4096 bytes when the default
5334 stripe_count is larger than 77. This conflicts with the default
5335 inode density of 1 per 4096 bytes. Allocate smaller inodes in
5336 this case to avoid pinning too much memory for large EAs.
5338 ------------------------------------------------------------------------------
5340 2005-07-07 Cluster File Systems, Inc. <info@clusterfs.com>
5345 Frequency : rare (extremely heavy IO load with hundreds of clients)
5347 Description: Client is evicted, gets IO error writing to file
5348 Details : lock ordering changes for bug 5492 reintroduced bug 3267 and
5349 caused clients to be evicted for AST timeouts. The fixes in
5350 bug 5192 mean we no longer need to have such short AST timeouts
5351 so ldlm_timeout has been increased.
5354 Frequency : occasional during --force or --failover shutdown under load
5355 Bugzilla : 5949, 4834
5356 Description: Server oops/LBUG if stopped with --force or --failover under load
5357 Details : a collection of import/export refcount and cleanup ordering
5358 issues fixed for safer force cleanup
5361 Frequency : only filesystems larger than 120 OSTs
5362 Bugzilla : 5990, 6223
5363 Description: lfs getstripe would oops on a very large filesystem
5364 Details : lov_getconfig used kfree on vmalloc'd memory
5367 Frequency : only filesystems exporting via NFS to Solaris 10 clients
5368 Bugzilla : 6242, 6243
5369 Description: reading from files that had been truncated to a non-zero size
5370 but never opened returned no data
5371 Details : ll_file_read() reads zeros from no-object files to EOF
5376 Description: A bug in MDS/OSS recovery could cause the OSS to fail an assertion
5377 Details : There's little harm in aborting MDS/OSS recovery and letting it
5378 try again, so I removed the LASSERT and return an error instead.
5380 Severity : enhancement
5382 Description: New debugging infrastructure for tracking down data corruption
5383 Details : The I/O checksum code was replaced to: (a) control it at runtime,
5384 (b) cover more of the client-side code path, and (c) try to narrow
5385 down where problems occurred
5389 Bugzilla : 3819, 4364, 4397, 6313
5390 Description: Racing close and eviction MDS could cause assertion in mds_close
5391 Details : It was possible to get multiple mfd references during close and
5392 client eviction, leading to one thread referencing a freed mfd.
5394 Severity: : enhancement
5395 Bugzilla : 3262, 6359
5396 Description: Attempts to reconnect to servers are now more aggressive.
5397 Details : This builds on the enhanced upcall-less recovery that was added
5398 in 1.4.2. When trying to reconnect to servers, clients will
5399 now try each server in the failover group every 10 seconds. By
5400 default, clients would previously try one server every 25 seconds.
5405 Description: After recovery, certain operations trigger a failed
5406 assertion on a client.
5407 Details : Failing over an mds, using lconf -d --failover, while a
5408 client was doing a readdir() call would cause the client to
5409 LBUG after recovery completed and the readdir() was resent.
5411 Severity : enhancement
5413 Description: Default groups are now added by lconf
5414 Details : You can now run lconf --group <servicename> without having to
5415 manually add groups with lmc.
5418 Frequency : occasional
5420 Description: Nodes with an elan id of 0 trigger a failed assertion
5423 Frequency : always when accessing e.g. tty/console device nodes
5425 Description: tty and some other devices nodes cannot be used on lustre
5426 Details : file's private_data field is used by device data and lustre
5427 values in there got lost. New field was added to struct file to
5428 store fs-specific private data.
5431 Frequency : when exporting Lustre via NFS
5433 Description: NFSD failed occasionally when looking up a path component
5434 Details : NFSD is looking up ".." which was broken in ext3 directories
5435 that had grown large enough to become hashed.
5438 Frequency : Clusters with multiple interfaces not on the same subnet
5440 Description: Nodes will repeatedly try to reconnect to an interface which it
5441 cannot reach and report an error to the log.
5442 Details : Extra peer list entries will be created by lconf with some peers
5443 unreachable. lconf now validates the peer before adding it.
5446 Frequency : Only if a default stripe is set on the filesystem root.
5448 Description: Setting a default stripe on the filesystem root prevented the
5449 filesystem from being remounted.
5450 Details : The client was sending extra request flags in the root getattr
5451 request and did not allocate a reply buffer for the dir EA.
5454 Frequency : occasional, higher if lots of files are accessed by one client
5455 Bugzilla : 6159, 6097
5456 Description: Client trips assertion regarding lsm mismatch/magic
5457 Details : While revalidating inodes the VFS looks up inodes with ifind()
5458 and in rare cases can find an inode that is being freed.
5459 The ll_test_inode() code will free the lsm during ifind()
5460 when it finds an existing inode and then the VFS later attaches
5461 this free lsm to a new inode.
5465 Bugzilla : 6422, 7030
5466 Description: MDS deadlock between mkdir and client eviction
5467 Details : Creating a new file via mkdir or mknod (starting a transaction
5468 and getting the ns lock) can deadlock with client eviction
5469 (gets ns lock and trying to finish a synchronous transaction).
5472 Frequency : occasional
5473 Description: While starting a server, the fsfilt_ext3 module could not be
5475 Details : CFS's improved ext3 filesystem is named ldiskfs for 2.6
5476 kernels. Previously, lconf would still use the ext3 name
5477 when trying to load modules. Now, it will correctly use
5478 ext3 on 2.4 and ldiskfs on 2.6.
5480 Severity : enhancement
5481 Description: The default stripe count has been changed to 1
5482 Details : The interpretation of the default stripe count (0, to lfs
5483 or lmc) has been changed to mean striping across a single
5484 OST, rather than all available. For general usage we have
5485 found a stripe count of 1 or 2 works best.
5487 Severity : enhancement
5488 Description: Add support for compiling against Cray portals.
5489 Details : Conditional compiling for some areas that are different
5493 Frequency : occasional
5494 Bugzilla : 6409, 6834
5495 Description: Creating files with an explicit stripe count may lead to
5496 a failed assertion on the MDS
5497 Details : If some OSTs are full or unavailable, creating files may
5498 trigger a failed assertion on the MDS. Now, Lustre will
5499 try to use other servers or return an error to the
5503 Frequency : occasional
5505 Description: Multiple concurrent overlapping read+write on multiple SMP nodes
5506 caused lock timeout during readahead (since 1.4.2).
5507 Details : Processes doing readahead might match a lock that hasn't been
5508 granted yet if there are overlapping and conflicting lock
5509 requests. The readahead process waits on ungranted lock
5510 (original lock is CBPENDING), while OST waits for that process
5511 to cancel CBPENDING read lock and eventually evicts client.
5513 Severity : enhancement
5515 Description: Initial enabling of flock support for clients
5516 Details : Implements fcntl advisory locking and file status functions.
5517 This feature is provided as an optional mount flag (default
5518 off), and is NOT CURRENTLY SUPPORTED. Not all types of record
5519 locking are implemented yet, and those that are are not guaranteed
5520 to be completely correct in production environments.
5521 mount -t lustre -o [flock|noflock] ...
5524 Frequency : occasional
5526 Description: OSTs running 2.4 kernels but with extents enabled might trip an
5527 assertion in the ext3 JBD (journaling) layer.
5528 Details : The b_committed_data struct is protected by the big kernel lock
5529 in 2.4 kernels, serializing journal_commit_transaction() and
5530 ext3_get_block_handle->ext3_new_block->find_next_usable_block()
5531 access to this struct. In 2.6 kernels there is finer grained
5532 locking to improve SMP performance of the JBD layer.
5536 Description: Changes the "SCSI I/O Stats" kernel patch to default to "enabled"
5538 -----------------------------------------------------------------------------
5540 2005-05-05 Cluster File Systems, Inc. <info@clusterfs.com>
5542 NOTE: Lustre 1.4.2 uses an incompatible network protocol than previous
5543 versions of Lustre. Please update all servers and clients to
5544 version 1.4.2 or later at the same time. You must also run
5545 "lconf --write-conf {config}.xml" on the MDS while it is stopped
5546 to update the configuration logs.
5548 - fix for HPUX NFS client breakage when NFS exporting Lustre (5781)
5549 - mdc_enqueue does not need max_mds_easize request buffer on send (5707)
5550 - swab llog records of type '0' so we get proper header size/idx (5861)
5551 - send llog cancel req to DLM cancel portal instead of cb portal (5515)
5552 - fix rename of one directory over another leaking an inode (5953)
5553 - avoid SetPageDirty on 2.6 (5981)
5554 - don't re-add just-being-destroyed locks to the waiting list (5653)
5555 - when creating new directories, inherit the parent's custom
5556 striping settings if present parent (3048)
5557 - flush buffers from cache before direct IO in 2.6 obdfilter (4982)
5558 - don't hold i_size_sem in ll_nopage() and ll_ap_refresh_count (6077)
5559 - don't hold client locks on temporary worklist from l_lru (5666)
5560 - handle IO errors in 2.6 obdfilter bio completion routine (6046)
5561 - automatically evict dead clients (5921)
5562 - Update file size properly in create+truncate+fstat case (6196)
5563 - Do not unhash mountpoint dentries, do not allow removal of
5565 - Avoid lock ordering deadlock issue with write/truncate (6203,5654)
5566 - reserve enough journal credits in fsfilt_start_log for setattr (4554)
5567 - ldlm_enqueue freed-export error path would always LBUG (6149,6184)
5568 - don't reference lr_lvb_data until after we hold lr_lvb_sem (6170)
5569 - don't overwrite last_rcvd if there is a *_client_add() error (6086)
5570 - Correctly handle reads of files with no objects (6243)
5571 - lctl recover will also mark a device active if deactivate used (5933)
5573 - by default create 1 inode per 4kB space on MDS, per 16kB on OSTs
5574 - allow --write-conf on an MDS with different nettype than client (5619)
5575 - don't write config llogs to MDS for mounts not from that MDS (5617)
5576 - lconf should create multiple TCP connections from a client (5201)
5577 - init scripts are now turned off by default; run chkconfig --on
5578 lustre and chkconfig --on lustrefs to use them
5579 - upcalls are no longer needed for clients to recover to failover
5581 - add --abort-recovery option to lconf to abort recovery on device
5583 - add support for an arbitrary number of OSTs (3026)
5584 - Quota support protocol changes.
5585 - forward compatibility changes to wire structs (6007)
5586 - rmmod NALs that might be loaded because of /etc/modules.conf (6133)
5587 - support for mountfsoptions and clientoptions to the Lustre LDAP (5873)
5588 - improved "lustre status" script
5589 - initialize blocksize for non-regular files (6062)
5590 - added --disable-server and --disable-client configure options (5782)
5591 - introduce a lookup cache for lconf to avoid repeated DB scans (6204)
5592 - Vanilla 2.4.29 support
5593 - increase maximum number of obd devices to 520 (6242)
5594 - remove the tcp-zero-copy patch from the suse-2.4 series (5902)
5595 - Quadrics Elan drivers are now included for the RHEL 3 2.4.21 and
5596 SLES 9 2.6.5 kernels
5597 - limit stripes per file to 160 (the maximum EA size) (6093)
5599 2005-03-22 Cluster File Systems, Inc. <info@clusterfs.com>
5602 - don't LASSERT in ll_release on NULL lld with NFS export (4655, 5760)
5603 - hold NS lock when calling handle_ast_error->del_waiting_lock (5746)
5604 - fix setattr mtime regression from lovcleanup merge (4829, 5669)
5605 - workaround for 2.6 crash in ll_unhash_aliases (5687, 5210)
5606 - small ext3 extents cleanups and fixes (5733)
5607 - improved mballoc code, several small races and bugs fixed (5733, 5638)
5608 - kernel version 43 - fix remove_suid bugs in both 2.4 and 2.6 (5695)
5609 - avoid needless client->OST connect, fix handle mismatch (5317)
5610 - fix DLM error path that led to out-of-sync client, long delays (5779)
5611 - support common vfs-enforced mount options (nodev,nosuid,noexec) (5637)
5612 - fix several locking issues related to i_size (5492,5624,5654,5672)
5613 - don't move pending lock onto export if it is already evicted (5683)
5614 - fix kernel oops when creating .foo in unlinked directory (5548)
5615 - fix deadlock in obdfilter statistics vs. object create (5811)
5616 - use time_{before,after} to avoid timer jiffies wrap (5882)
5617 - shutdown --force/--failover stability (3607,3651,4797,5203,4834)
5618 - Do not leak request if server was not able to process it (5154)
5619 - If mds_open unable to find parent dir, make that negative lookup(5154)
5620 - don't create new directories with extent-mapping (5909, 5936)
5622 - fix lustre/lustrefs init scripts for SuSE (patch from Scali, 5702)
5623 - don't hold the pinger_sem in ptlrpc_pinger_sending_on_import
5624 - change obd_increase_kms to obd_adjust_kms (up or down) (5654)
5625 - lconf, lmc search both /usr/lib and /usr/lib64 for Python libs (5800)
5626 - support for RHEL4 kernel on i686 (5773)
5627 - provide error messages when incompatible logs are encountered (5898)
5629 2005-02-18 Cluster File Systems, Inc. <info@clusterfs.com>
5630 * version 1.4.0.10 (1.4.1 release candidate 1)
5632 - don't keep a lock reference when lock is not granted (4238)
5633 - unsafe list practices (rarely) led to infinite eviction loop (4908)
5634 - add per-fs limit of Lustre pages in page cache, avoid OOM (4699)
5635 - drop import inflight refcount on signal_completed_replay error (5255)
5636 - unlock page after async write error during send (3677)
5637 - handle missing objects in filter_preprw_read properly (5265)
5638 - no transno return for symlink open, don't save no-trasno open (3440)
5639 - don't try to complete elan receive that already failed (4012)
5640 - free RPC server reply state on error (5406)
5641 - clean up thread from ptlrpc_start_thread() on error (5160)
5642 - readahead could read extra page into cache that wasn't ejected (5388)
5643 - prevent races in class_attach/setup/cleanup/detach (5260)
5644 - don't dereference de->d_inode after l_dput of de (5458)
5645 - use "int" for stripe value returned from lock_to_stripe (5544)
5646 - mballoc allocation and error-checking fixes in 2.6 (5504)
5647 - block device patches to fix I/O request sizes in 2.6 (5482)
5648 - look up hostnames for IB nals (5602)
5649 - 2.6 changed lock ordering of 2 semaphores, caused deadlock (5654)
5650 - don't start multiple acceptors for the same port (5277)
5651 - fix incorrect LASSERT in mds_getattr_name (5635)
5652 - export a proc file for general "ping" checking (5628)
5653 - fix "lfs check" to not block when the MDS is down (5628)
5655 - service request history (4965)
5656 - put {ll,lov,osc}_async_page structs in a single slab (4699)
5657 - create an "evict_client" /proc entry on OSTs, like the MDS has
5658 - fix mount usage message, return errors per mount(8) (5168)
5659 - change grep [] to grep "[]" in tests so they work in more UMLs
5660 - fix ppc64/x86_64 spec to use %{_libdir} instead of /usr/lib (5389)
5661 - remove ancient LOV_MAGIC_V0 EA support (5047)
5662 - add "disk I/Os in flight" and "I/O req time" stats in obdfilter
5663 - align r/w RPCs to PTLRPC_MAX_BRW_SIZE boundary for performance (3451)
5664 - allow readahead allocations to fail when low on memory (5383)
5665 - mmap locking landed again, after considerable improvement (2828)
5666 - add get_hostaddr() to lustreDB.py for LDAP support (5459)
5668 2004-11-23 Cluster File Systems, Inc. <info@clusterfs.com>
5671 - send OST transaction number in read/write reply to free req (4966)
5672 - don't ASSERT in ptl_send_rpc() if we run out of memory (5119)
5673 - lock /proc/sys/portals/routes internal state, avoiding oops (4827)
5674 - the watchdog thread now runs as interruptible (5246)
5675 - flock/lockf fixes (but it's still disabled, pending 5135)
5676 - don't use EXT3 constants in llite code (5094)
5677 - memory shortage at startup could cause assertion (5176)
5679 - reorganization of lov code
5680 - single portals codebase
5682 - add extents/mballoc support (5025)
5683 - direct I/O reads in the obdfilter (4048)
5684 - kernel patches from LNXI for 2.6 (bluesmoke, perfctr, mtd, kexec)
5686 tbd Cluster File Systems, Inc. <info@clusterfs.com>
5689 - send OST transaction number in read/write reply to free req (4966)
5690 - don't ASSERT in ptl_send_rpc() if we run out of memory (5119)
5691 - lock /proc/sys/portals/routes internal state, avoiding oops (4827)
5692 - the watchdog thread now runs as interruptible (5246)
5693 - handle missing objects in filter_preprw_read properly (5265)
5694 - unsafe list practices (rarely) led to infinite eviction loop (4908)
5695 - drop import inflight refcount on signal_completed_replay error (5255)
5696 - unlock page after async write error during send (3677)
5697 - return original error code on reconstructed replies (3761)
5698 - no transno return for symlink open, don't save no-trasno open (3440)
5700 - add pid to ldlm debugging output (4922)
5701 - bump the watchdog timeouts -- we can't handle 30sec yet
5702 - extra debugging for orphan dentry/inode bug (5259)
5704 2004-11-16 Cluster File Systems, Inc. <info@clusterfs.com>
5707 - fix TCP_NODELAY bug, which caused extreme perf regression (5134)
5708 - allocate qswnal tx descriptors singly to avoid fragmentation (4504)
5709 - don't LBUG on obdo_alloc() failure, use OBD_SLAB_ALLOC() (4800)
5710 - fix NULL dereference in /proc/sys/portals/routes (4827)
5711 - allow failed mdc_close() operations to be interrupted (4561)
5712 - stop precreate on OST before MDS would time out on it (4778)
5713 - don't send partial-page writes before EOF from client (4410)
5714 - discard client grant for sub-page writes on large-page clients (4520)
5715 - don't free dentries not owned by NFS code, check generation (4806)
5716 - fix lsm leak if mds_create_objects() fails (4801)
5717 - limit debug_daemon file size, always print CERROR messages (4789)
5718 - use transno after validating reply (3892)
5719 - process timed out requests if import state changes (3754)
5720 - update mtime on OST during writes, return in glimpse (4829)
5721 - add mkfsoptions to LDAP (4679)
5722 - use ->max_readahead method instead of zapping global ra (5039)
5723 - don't interrupt __l_wait_event() during strace
5725 - add software watchdogs to catch hung threads quickly (4941)
5726 - make lustrefs init script start after nfs is mounted
5727 - fix CWARN/ERROR duplication (4930)
5728 - return async write errors to application if possible (2248)
5729 - add /proc/sys/portal/memused (bytes allocated by PORTALS_ALLOC)
5730 - print NAL number in %x format (4645)
5731 - update barely-supported suse-2.4.21-171 series (4842)
5732 - support for sles 9 %post scripts
5733 - support for building 2.6 kernel-source packages
5734 - support for sles km_* packages
5736 2004-10-07 Cluster File Systems, Inc. <info@clusterfs.com>
5739 - ignore -ENOENT errors in osc_destroy (3639)
5740 - notify osc create thread that OSC is being cleaned up (4600)
5741 - add nettype argument for llmount in #5d in conf-sanity.sh (3936)
5742 - reconstruct ost_handle() like mds_handle() (4657)
5743 - create a new thread to do import eviction to avoid deadlock (3969)
5744 - let lconf resolve symlinked-to devices (4629)
5745 - don't unlink "objects" from directory with default EA (4554)
5746 - hold socknal file ref over connect in case target is down (4394)
5747 - allow more than 32000 subdirectories in a single directory (3244)
5748 - fix blocks count for O_DIRECT writes (3751)
5749 - OST returns ENOSPC from object create when no space left (4539)
5750 - don't send truncate RPC if file size isn't changing (4410)
5751 - limit OSC precreate to 1/2 of value OST considers bogus (4778)
5752 - bind to privileged port in socknal and tcpnal (3689)
5754 - rate limit CERROR/CWARN console message to avoid overload (4519)
5755 - GETFILEINFO dir ioctl returns LOV EA + MDS stat in 1 call (3327)
5756 - basic mmap support (3918)
5757 - kernel patch series update from b1_4 (4711)
5759 2004-09-16 Cluster File Systems, Inc. <info@clusterfs.com>
5762 - avoid crash during MDS cleanup with OST shut down (2775)
5763 - fix loi_list_lock/oig_lock inversion on interrupted IO (4136)
5764 - don't use bad inodes on the MDS (3744)
5765 - dynamic object preallocation to improve recovery speed (4236)
5766 - don't hold spinlock over lock dumping or change debug flags (4401)
5767 - don't zero obd_dev when it is force cleaned (3651)
5768 - print grants to console if they go negative (4431)
5769 - "lctl deactivate" will stop automatic recovery attempts (3406)
5770 - look for existing locks in ldlm_handle_enqueue() (3764)
5771 - don't resolve lock handle twice in recovery avoiding race (4401)
5772 - revalidate should check working dir is a directory (4134)
5774 - don't always mark "slow" obdfilter messages as errors (4418)
5776 2004-08-24 Cluster File Systems, Inc. <info@clusterfs.com>
5779 - don't close LustreDB during write_conf until it is done (3860)
5780 - fix typo in lconf for_each_profile (3821)
5781 - allow dumping logs from multiple threads at one time (3820)
5782 - don't allow multiple threads in OSC recovery (3812)
5783 - fix debug_size parameters (3864)
5784 - fix mds_postrecov to initialize import for llog ctxt (3121)
5785 - replace config semaphore with spinlock (3306)
5786 - be sure to send a reply for a CANCEL rpc with bad export (3863)
5787 - don't allow enqueue to complete on a destroyed export (3822)
5788 - down write_lock before checking llog header bitmap (3825)
5789 - recover from lock replay timeout (3764)
5790 - up llog sem before sending rpc (3652)
5791 - reduce ns lock hold times when setting kms (3267)
5792 - change a dlm LBUG to LASSERTF, to maybe learn something (4228)
5793 - fix NULL deref and obd_dev leak on setup error (3312)
5794 - replace some LBUG about llog ops with error handling (3841)
5795 - don't match INVALID dentries from d_lookup and spin (3784)
5796 - hold dcache_lock while marking dentries INVALID and hashing (4255)
5797 - fix invalid assertion in ptlrpc_set_wait (3880)
5799 - add libwrap support for the TCP acceptor (3996)
5800 - add /proc/sys/portals/routes for non-root route listing (3994)
5801 - allow setting MDS UUID in .xml (2580)
5802 - print the stack of a process that LBUGs (4228)
5804 2004-07-14 Cluster File Systems, Inc. <info@clusterfs.com>
5807 - don't cleanup request in ll_file_open() on failed MDS open (3430)
5808 - make sure to unset replay flag from failed open requests (3440)
5809 - if default stripe count is 0, use OST count for inode size (3636)
5810 - update parent mtime/ctime on client for create/unlink (2611)
5811 - drop dentry ref in ext3_add_link from open_connect_dentry (3266)
5812 - free recovery state on server during a forced cleanup (3571)
5813 - unregister_reply for resent reqs (3063)
5814 - loop back devices mounting and status check on 2.6 (3563)
5815 - fix resource-creation race that can provoke i_size == 0 (3513)
5816 - don't try to use bad inodes returned from MDS/OST fs lookup (3688)
5817 - more debugging for page-accounting assertion (3746)
5818 - return -ENOENT instead of asserting if ost getattr+unlink race (3558)
5819 - avoid deadlock after precreation failure (3758)
5820 - fix race and lock order deadlock in orphan handling (3450, 3750)
5821 - add validity checks when grabbing inodes from l_ast_data (3599)
5823 - add /proc/.../recovery_status to obdfilter (3428)
5824 - lightweight CDEBUG infrastructure, debug daemon (3668)
5825 - change default OSC RPC parameters to be better on small clusters
5826 - turn off OST read cache for files smaller than 32MB
5827 - install man pages and include them in rpms (3100)
5828 - add new init script for (un)mounting lustre filesystems (2593)
5829 - run chkconfig in %post for init scripts (3701)
5830 - drop scimac NAL (unmaintained)
5832 2004-06-17 Cluster File Systems, Inc. <info@clusterfs.com>
5835 - clean kiobufs before and after use (3485)
5836 - strip trailing '/'s before comparing paths with /proc/mounts (3486)
5837 - remove assertions to work around "in-flight rpcs" recovery bug (3063)
5838 - change init script to fail more clearly if not run as root (1528)
5839 - allow clients to reconnect during replay (1742)
5840 - fix ns_lock/i_sem lock ordering deadlock for kms update (3477)
5841 - don't do DNS lookups on NIDs too small for IP addresses (3442)
5842 - re-awaken ptlrpcd if new requests arrive during check_set (3554)
5843 - fix cond_resched (3554)
5844 - only evict unfinished clients after recovery (3515)
5845 - allow bulk resend, prevent data loss (3570)
5846 - dynamic ptlrpc request buffer allocation (2102)
5847 - don't allow unlinking open directory if it isn't empty (2904)
5848 - set MDS/OST threads to umask 0 to not clobber client modes (3359)
5849 - remove extraneous obd dereference causing LASSERT failure (3334)
5850 - don't use get_cycles() when creating temp. files on the mds (3156)
5851 - hold i_sem when setting i_size in ll_extent_lock() (3564)
5852 - handle EEXIST for set-stripe, set proper directory name (3336)
5854 - servers can dump a log evicting a client - lustre.dump_on_timeout=1
5855 - fix ksocknal_fmb_callback() error messages (2918)
5857 2004-05-27 Cluster File Systems, Inc. <info@clusterfs.com>
5860 - don't copy lvb into (possibly NULL) reply on error (2983)
5861 - don't deref dentry after dput, don't free lvb on error (2922)
5862 - use the kms to determine writeback rpc length (2947)
5863 - increment oti_logcookies when osc is inactive (2948)
5864 - update client's i_blocks count via lvb messages (2543)
5865 - handle intent open/close of special files properly (1557)
5866 - mount MDS with errors=remount-ro, like obdfilter (2009)
5867 - initialize lock handle to avoid ASSERT on error cleanup (3057)
5868 - don't use cancelling-locks' kms values (2947)
5869 - use highest lock extent for kms, not last one (2925)
5870 - don't dereference ERR_PTR() dentry in error handling path (3107)
5871 - fix thread race in portals_debug_dumplog() (3122)
5872 - create lprocfs device entries at setup instead of at attach (1519)
5873 - common AST error handler, don't evict client on completion race (3145)
5874 - zero nameidata in detach_mnt in 2.6 (3118)
5875 - verify d_inode after revalidate_special is valid in 2.6 (3116)
5876 - use lustre_put_super() to handle zconf unmounts in 2.6 (3064)
5877 - initialize RPC timeout timer earlier for 2.6 (3219)
5878 - don't dereference NULL reply buffer if mdc_close was never sent (2410)
5879 - print nal/nid for unknown nid (3258)
5880 - additional checks for oscc recovery before doing precreate (3284)
5881 - fix ll_extent_lock() error return code for 64-bit systems (3043)
5882 - don't crash in mdc_close for bad permissions on open (3285)
5883 - zero i_rdev for non-device files (3147)
5884 - clear page->private before handing to FS, better assertion (3119)
5885 - tune the read pipeline (3236)
5886 - fix incorrect decref of invalidated dentry (2350)
5887 - provide read-ahead stats and refine rpc in flight stats (3328)
5888 - don't hold journal transaction open across create RPC (3313)
5889 - update atime on MDS at close time (3265)
5890 - close LDAP connection when recovering to avoid server load (3315)
5891 - update iopen-2.6 patch with fixes from 2399,2517,2904 (3301)
5892 - don't leak open file on MDS after open resend (3325)
5893 - serialize filter_precreate and filter_destroy_precreated (3329)
5894 - loop device shouldn't call sync_dev() for nul device (3092)
5895 - clear page cache after eviction (2766)
5896 - resynchronize MDS->OST in background (2824)
5897 - refuse to mount the same filesystem twice on same mountpoint (3394)
5898 - allow llmount to create routes for mounting behind routers (3320)
5899 - push lock cancellation to blocking thread for glimpse ASTs (3409)
5900 - don't call osc_set_data_with_check() for TEST_LOCK matches (3159)
5901 - fix rare problem with rename on htree directories (3417)
5903 - allow default OST striping configuration per directory (1414)
5904 - fix compilation for qswnal for 2.6 kernels (3125)
5905 - increase maximum number of MDS request buffers for large systems
5906 - change liblustreapi to be useful for external progs like lfsck (3098)
5907 - increase local configuration timeout for slow disks (3353)
5908 - allow configuring ldlm AST timeout - lustre.ldlm_timeout=<seconds>
5910 2004-03-22 Cluster File Systems, Inc. <info@clusterfs.com>
5913 - fixes for glimpse AST timeouts / incorrectly 0-sized files (2818)
5914 - don't overwrite extent policy data in reply if lock was blocked (2901)
5915 - drop filter export grants atomically with removal from device (2663)
5916 - del obd_self_export from work_list in class_disconnect_exports (2908)
5917 - don't LBUG if MDS recovery times out during orphan cleanup (2530)
5918 - swab reply message in mdc_close, other PPC fixes (2464)
5919 - fix destroying of named logs (2325)
5920 - overwrite old logs when running lconf --write_conf (2264)
5921 - bump LLOG_CHUNKSIZE to 8k to allow for larger clusters (2306)
5922 - fix race in target_handle_connect (2898)
5923 - mds_reint_create() should take same inode create lock (2926)
5924 - correct journal credits calculated for CANCEL_UNLINK_LOG (2931)
5925 - don't close files for self_export to avoid uninitialized obd (2936)
5926 - allow MDS with the same name as client node (2939)
5927 - hold dentry reference for closed log files for unlink (2325)
5928 - reserve space for all logs during transactions (2059)
5929 - don't evict page beyond end of stripe extent (2925)
5930 - don't oops on a deleted current working directory (2399)
5931 - handle hard links to targets without a parent properly (2517)
5932 - don't dereference NULL lock when racing during eviction (2867)
5933 - don't grow lock extents when lots of conflicting locks (2919)
5935 2004-03-04 Cluster File Systems, Inc. <info@clusterfs.com>
5938 - account for cache space usage on clients to avoid data loss (974)
5939 - lfsck support in lustre kernel code (2349)
5940 - reduce journal credits needed for BRW writes (2370)
5941 - orphan handling to avoid losing space on client/server crashes
5942 - ptlrpcd can be blocked, stopping ALL progress (2477)
5943 - use lock value blocks to assist in proper KMS, faster stat (1021)
5944 - takes i_sem instead of DLM locks internally on obdfilter (2720)
5945 - recovery for initial connections (2355)
5946 - fixes for mds_cleanup_orphans (1934)
5947 - abort_recovery crashes MDS in b_eq (mds_unlink_orphan) (2584)
5948 - block all file creations until orphan recovery completes (1901)
5949 - client remove rq_connection from request struct (2423)
5950 - conf-sanity test_5, proper cleanup in umount log not availale (2640)
5951 - recovery timer race (2670)
5952 - mdc_close recovey bug (2532)
5953 - ptlrpc cleanup bug (2710)
5954 - mds timeout on local locks (2588)
5955 - namespace lock held during RPCs (2431)
5956 - handle interrupted sync write properly (2503)
5957 - don't try to handle a message that hasn't been replied to (2699)
5958 - client assert failure during cleanup after abort recovery (2701)
5959 - leak mdc device after failed mount (2712)
5960 - ptlrpc_check_set allows timedout requests to complete (2714)
5961 - wait for inflight reqs when ptlrpcd finishes (2710)
5962 - make sure unregistered services are removed from the srv_list
5963 - reset bulk XID's when resending them (caught by 1138 test)
5964 - unregister_bulk after timeout
5965 - fix lconf error (2694)
5966 - handle write after unfinished setstripe, stripe-only getstripe (2388)
5967 - readahead locks pages, leaves pending causing memory pressure (2673)
5968 - increase OST request buffers to 4096 on large machines (2729)
5969 - fix up permission of existing directories in simple_mkdir (2661)
5970 - init deleted item, add assertions ptlrpc_abort_inflight() (2725)
5971 - don't assign transno to errored transactions (2742)
5972 - don't delete objects on OST if given a bogus objid from MDS (2751)
5973 - handle large client PAGE_SIZE readdir on small PAGE_SIZE MDS (2777)
5974 - if rq_no_resend, then timeout request after recovery (2432)
5975 - fix MDS llog_logid record size, 64-bit array alignment (2733)
5976 - don't call usermode_helper from ptlrpcd, DEFAULT upcall (2773)
5977 - put magic in mount.lustre data, check for bad/NULL mount data (2529)
5978 - MDS recovery shouldn't delete objects that it has given out (2730)
5979 - if enqueue arrives after completion, don't clobber LVB (2819)
5980 - don't unlock pages twice when trigger_group_io returns error (2814)
5981 - don't deref NULL rq_repmsg if ldlm_handle_enqueue failed (2822)
5982 - don't write pages to disk if there was an error (1450)
5983 - don't ping imports that have recovery disabled (2676)
5984 - take buffered bytes into account when balancing socknal conn (2817)
5985 - hold a DLM lock over readdir always, use truncate_inode_pages (2706)
5986 - reconnect unlink llog connection after MDS reconnects to OST (2816)
5987 - remove little-endian swabbing of llog records (1987)
5988 - set/limit i_blksize to LL_MAX_BLKSIZE on client (2884)
5989 - retry reposting request buffers if they fail (1191)
5990 - grow extent at grant time to avoid granting a revoked lock (2809)
5991 - lock revoke doesn't evict page if covered by a second lock (2765)
5992 - disable VM readahead to avoid reading outside lock extents (2805)
5994 - return LL_SUPER_MAGIC from statfs for the filesystem type (1972)
5995 - updated kernel patches for hp-2.4.20 kernel (2681)
5997 2004-02-07 Cluster File Systems, Inc. <info@clusterfs.com>
6000 - fix truncated write corruption (2366)
6001 - fix for failed assertion in iopen_connect_dentry (1792,2517)
6003 - don't flag the ptlrpcd thread with PF_MEMALLOC (2636)
6004 - ensure len(uuid) < 37 in lmc (1171)
6005 - fix ia64 OOPS in llog_test (2255)
6006 - zero end of page at obdfilter for partial page writes (2648)
6007 - don't leave stale dentries around after renames (bug 2428)
6008 - fix timeouts when evicting a client with a single lock held (2642)
6009 - set deadline for the initial HELLO message to drain (2634)
6010 - print out dotted-quad IP addresses in the socknal (2302)
6012 - additional debugging for MDS client eviction problem (2443)
6013 - fix mkfsoptions support for osts (2603, 2604)
6015 2004-01-27 Cluster File Systems, Inc. <info@clusterfs.com>
6018 - add series for the vanilla 2.6.0 kernel
6019 - add series for the vanilla 2.4.24 kernel
6020 - add series for a cray x86/64 UL kernel drop
6021 - fix xattr patches for the vanilla 2.4.19 series
6023 - generate true UUIDs in lmc (1171)
6024 - have portals stack dumping break in UML (2466)
6025 - avoid bad dchild deref; avoid inum lock w/o creation (2362)
6026 - allocate with _NOFS in ldlm to avoid deadlock (1933)
6027 - wake callback waiting threads on client eviction (2460)
6028 - Add --ptldebug and --subsystem to lmc (1719)
6029 - update assertion to allow safe interrupt allocation
6030 - set rq_no_resend for cancel requests (2432)
6031 - recalculate ptlrpcd timeout after resend (2494)
6032 - call vfs_rmdir when removing pending directories (2368)
6033 - fix renaming a file to itself (2429)
6034 - lmc creates a default one-stripe lov (2454)
6035 - expand procfs space to handle large clusters (2326)
6036 - increase UML stack to avoid overflow
6037 - update lconf's list of debug and subsystem masks
6038 - fix lfs find --obd (2510)
6039 - /proc tunable for disabling filter read caching (2591)
6040 - stop rpm packages from altering slapd.conf (2301)
6041 - disable nagle in the socknal under 0conf (2578)
6042 - choose mds inode size based on stripe count (2572)
6043 - fix kernel-source rpm problems (2516)
6045 - add --disable-doc to avoid pdf generation (2421)
6046 - update documentation, tests, type-os, comments
6047 - avoid format warnings on ia64
6048 - remove the TOE NAL
6049 - tiny code cleanups by removing unused fields
6051 2004-01-07 Cluster File Systems, Inc. <info@clusterfs.com>
6054 - fix obvious semaphore misuse in as-yet-unused setattr path (2348)
6055 - remove the most blatant lies from BUILDING file (2371)
6056 - change default debug level to reasonable production setting
6057 - reduce client side cache size to reduce cache flush time
6058 - reduce max RPCs in flight to avoid unnecessary file fragmentation
6059 - make TCP zerocopy and pinger support enabled by default (2476)
6060 - sync writes completed after process exits caused crashes (2319)
6061 - maintain correct mount count on the MDS (2356)
6062 - backout 1557, because 2316 wasn't really fixed
6063 - better file I/O statistics gathering in /proc
6064 - don't take unnecessary, deadlock-inducing bug in readpage (2383)
6065 - another kernel patch to fix zero-copy TCP function export
6066 - don't take duplicate lock when processing re-sent getattr (2420)
6067 - lctl uses obd_self_export instead of creating new conn (2353)
6068 - MDS/OST recovery case which requires object creation asserted (2425)
6069 - move lfs from /usr/sbin to /usr/bin in packages
6070 - fix race between mds_client_add and mds_client_free (2417)
6071 - use kmalloc instead of slabs in portals (2430)
6072 - don't create duplicate records when a failover MDS is present (2442)
6073 - remove unnecessary mount age check (2332)
6074 - don't remove directory inodes from locks prematurely (2451)
6075 - don't break if MDS service name is the same as hostname (2103)
6076 - fix races in client write RPC generation when cache full (2482)
6078 2003-12-13 Cluster File Systems, Inc. <info@clusterfs.com>
6081 - remove now-unused request->rq_obd (278)
6082 - if an allocation fails, print out how much memory we've used (1933)
6083 - use PORTAL_SLAB_ALLOC for structures, to get GFP_MEMALLOC (1933)
6084 - add the "configurable stack size" patch to most series files (1256)
6085 - ability to write large log records, for 100+ OST configs (2306)
6086 - fix NULL deref when filter_prep fails (2314)
6087 - fix operator precedence error in filter_sync
6088 - dynamic allocation of socknal TX descriptors (2315)
6089 - fix a missed case in the GFP_MEMALLOC patch, can cause deadlock (2310)
6090 - fix gcc 2.96 compilation problem in xattr kernel patch (2294)
6091 - ensure that CWARN messages in Portals always get to the syslog
6092 - __init/__exit are not for prototype decls (ldlm_init/exit)
6093 - x86-64 compile warning fixes
6094 - fix gateway LMC keyword conflict (2318)
6095 - fix MDS lock inversions in getattr/reint paths (1844)
6096 - fix a rare lock re-ordering bug, which caused deadlock (2322)
6097 - fix i_sem/journal inversion in fsfilt_ext3_write_record (2306)
6098 - DLM race condition prevented some lock evictions (2328)
6099 - ENOMEM detection and retry on socknal sends (2230)
6100 - use GFP_NOFS throughout Lustre, to combat ENOMEM (2230)
6101 - move osc_rpcd into ptlrpc, for use in MDC and others (2329)
6102 - protect MDS inode fsdata with stronger locking; fixes assertion (2313)
6103 - better error messages when a client is rejected during recovery (1505)
6104 - avoid cancelling locks which were never granted, after failure (2330)
6105 - fix i_sem/journal inversion in mds_client_add (2333)
6106 - fix truncate/getattr lock cycle deadlock (2334)
6107 - use rpcd to send close; allows resend after timeout, avoid leak (1897)
6108 - fix two rare exit paths which could leak an l_lock() ref (2321)
6109 - fencepost error in MDS/OST orphan recovery (2226)
6110 - make log record alignment 8 bytes (1988)
6111 - lstripe now fails when requested offset > ost_count (2237)
6112 - ensure that all kernel series have a complete list.h (1607)
6113 - fix crashes in special-file operations (2316)
6114 - lctl create/brw OID mismatch, caused by obsolete filter loop (2339)
6116 - allow configurable automake binary, for testing new versions
6117 - small update to the lfs documentation
6119 2003-12-03 Cluster File Systems, Inc. <info@clusterfs.com>
6121 * fix negative export reference count in fsfilt_sync (2312)
6123 2003-12-01 Cluster File Systems, Inc. <info@clusterfs.com>
6124 * release candidate 0.9.1
6126 - orphans are moved into the PENDING directory for possible recovery
6127 - replayed opens now open by fid for orphan/rename safety (1042)
6128 - last close of an orphan inode generates a transno (683)
6129 - chdir() and mount() now pin the directory entry (1020)
6130 - avoid CERROR in normal ll_setattr_raw() error case (1500)
6131 - discard very old requests without processing them (1502)
6132 - remove some common, well-understood CERRORs (1505)
6133 - require O_DIRECT I/O to be page-sized to workaround IA64 crash (1609)
6134 - clear "grant" flags in OST replies until OST grant code lands (1644)
6135 - fix read performance by not clobbering i_blksize on client (1598)
6136 - fix __ldlm_handle2lock oops by not dereferencing lock after PUT (1625)
6137 - make LRU size a /proc tunable, clears locks when reduced (707)
6138 - fix some lprocfs rot that prevented ptlbd from loading (1732)
6139 - server locks take references on exports now (1558)
6140 - build fixes for 2.4.20-rh trees (1663)
6141 - return an error from lov_create if all OSCs are inactive (1751)
6142 - fix import levels when a reconnect happens without a timeout (1597)
6143 - exit early from mds_open if we get a lookup error (1749)
6144 - partial page read at EOF wouldn't wait for disk before sending (1642)
6145 - avoid NULL deref in obdfilter when reading page past EOF (1592)
6146 - avoid LASSERT in ll_intent_lock if server failed very early (1090)
6147 - fix LBUG in ll_it_open_error with rc = -2 (1861)
6148 - write/truncate lock inversion (1639)
6149 - Don't auto-load obdclass, portals modules during cleanup (1495)
6150 - fix timestamps from jumping to "now" (1763)
6151 - extra journal assertions (1648)
6152 - add an extra multiunlink test (1771)
6153 - fix read_record/write_record API (1776)
6154 - fix leak of offset_extent, possible incorrect i_size later (1772)
6155 - fix lasserts in mis-matched transnos during open-unlink testing (1541)
6156 - Debugging for the kqswnal_get_idle_tx problems (1820)
6157 - Allow recovery to be attempted multiple times (1536)
6158 - Write out MDS last_rcvd file after it is first created (1600)
6159 - Fix tx_descriptor leak in failed transmit situations (1827)
6160 - ext3 journaling fixes for assertion failure after IO error (1871)
6161 - class_export_put() on freed export after completion AST error (1896)
6162 - Fix revalidate looping in VFS (1322)
6163 - Don't access a freed export during MDS_REINT timeout (1521)
6164 - Add open-unlink recovery support on the MDS (1673,1764)
6165 - Return an error if no MDS data was read from last_rcvd (1946)
6166 - Fix for lookup "." or ".." crash on error (1932,1931,1935)
6167 - Don't setup a disk device that doesn't match exported UUID (317)
6168 - Reduce bulk RPC timeout to avoid cascading client/OST failures (1845)
6169 - avoid committing NULL handle in force close
6170 - local.sh is now a one-stripe LOV configuration
6171 - POSIX utime.4 -EPERM on FIFO not owned by user (56)
6172 - fix ext3 htree duplicate directory entry corruption (1516)
6173 - POSIX creat.13, fstat.1, open.18, stat.3 new file atime/mtime (2020)
6174 - update to new LOV EA format (2097)
6175 - interoperability for different PAGE_SIZE/wordsize (686,1821,1343,2042)
6177 2003-06-15 Phil Schwan <phil@clusterfs.com>
6180 - imports and exports cleanup too early, need refcounts (349, 879, 1045)
6181 - per-import/export recovery handling (958, 931, 959)
6182 - multiple last-rcvd slots, for serving multiple FSes (949)
6183 - connections are again shared between multiple imp/exports (963, 964)
6184 - "umount -f" would hang if any requests needed to be sent (393, 978)
6185 - avoid pinning large req buffer by copying for queued messages (989)
6186 - add "uuid" to "lctl device" command to help upcalls (991)
6187 - "open" RPCs with transnos would confuse recovery counters (1037)
6188 - do proper endian conversion of all wire messages (288, 340, 891)
6189 - remove OST bulk get LBUGs, fix ost_brw_write cleanup (1126)
6190 - call waiting locks callback from LDLM recovery thread (1127, 1151)
6191 - fix ptlrpc_connection leak in target_handle_connect (1174)
6192 - fix import refcounting bug in OST and MDS cleanup (1134)
6193 - if an invalid-at-open-time OSC returned before close(), LBUG (1150)
6194 - fix very unlikely obd_types race condition (501)
6195 - remove osc_open hack for echo_client (1187)
6196 - we leaked exports/dlmimps for forcibly disconnected clients (1143)
6197 - a failure in read_inode2 leads to deadlock (1139)
6198 - cancel ack-locks as soon as transaction is committed (1072)
6199 - fix major leaks and crashes in the bulk I/O path (937, 1057)
6200 - make sure to commitrw after any preprw to avoid deadlock (1162)
6201 - failing to execute a file in a lustre FS would lock inode (1203)
6202 - small DEBUG_REQ fix to avoid dereferencing a NULL (1227)
6203 - don't ASSERT while cleaning up an incompletely-setup obd (1248)
6204 - obd_uuid2tgt would walk off the end of the list (1255)
6205 - on IA64 the osc would give portals incorrect bulk size (1258)
6206 - fix debug daemon ioctl interface; allows daemon on ia64 (1274)
6207 - fix lock inversion caused by new llite matching code (1282)
6208 - limit the number of dirty pages on a client to 10MB (1286)
6209 - timed out locks were not being corrected cancelled (1289)
6210 - fix O_DIRECT above 4GB on IA-32 (1292)
6211 * major user-visible changes
6212 - fail out/fail over policy now controlled by the upcall (993)
6214 - add OBD_PING to check server availability and failure (954)
6215 - lustre messages are now sent in sending host order (288, 340, 891)
6216 - add eadatalen to MDS getattr reply (340)
6217 - OST read replies may contain second buffer, with per-page status (593)
6219 2003-03-11 Phil Schwan <phil@clusterfs.com>
6222 - LDLM_DEBUG macro fix, for gcc 3.2 (850)
6223 - failed open()s could cause deadlock; fixed (867, 869)
6224 - stop cancelling OST locks when files are closed (481)
6225 - overlapping XID spaces caused network corruption (851, 853)
6226 - fix unsafe fsfilt counter arithmetic; change to atomic_t
6227 - setattr_raw added, to do single-RPC, server-side setattrs
6228 - lmc/lconf syntax change for OST UUIDs
6229 - fix crashy race condition between ptlrpc_free_req and osc_close
6230 - don't use request in mdc_enqueue if we hit a timeout (889)
6231 - don't set the inode i_size for regular files from the MDS (896)
6232 - handle out of order completion AST (842)
6233 - don't LBUG if a lock request times out after receiving AST (913)
6234 - avoid d_rehash race in ll_find_alias by rehashing inside dcache_lock
6235 - if a bad lock AST arrives, send an error instead of dropping entirely
6236 - return 0 from revalidate2 if ll_intent_lock returns -EINTR (912)
6237 - fix leak in bulk IO when only partially completed (899, 900, 926)
6238 - fix O_DIRECT for ia64 (55)
6239 - (almost) eliminate Lustre-kernel-thread effects on load average (722)
6240 - C-z after timeout could hang a process forever; fixed (977)
6242 - client-side I/O cache (678, 924, 929, 941, 970)
6244 - READPAGE and SETATTRs which don't take server-side locks get
6247 2003-02-11 Phil Schwan <phil@clusterfs.com>
6250 - Fix ldlm_lock_match on the MDS to avoid matching remote locks (592)
6251 - Fix fsfilt_extN_readpage() to read a full page of directory
6252 entries, or fake the remainder if PAGE_SIZE != blocksize (500)
6253 - Avoid extra mdc_getattr() in ll_intent_lock when possible (534, 604)
6254 - Fix imbalanced LOV object allocation and out-of-bound access (469)
6255 - Most intent operations were removed, in favour of a new RPC mode
6256 that does a single RPC to the server and bypasses most of the VFS
6257 - All LDLM resource ID arrays were removed in favour of ldlm_res_id
6258 - Aggressively cancel local locks on DLM servers
6259 - mds_reint_unlink sends EA to the client if it's the last nlink.
6260 client uses that EA to unlink OST objects.
6261 - mds_reint_{rename,unlink,link} were rewritten to take ordered locks
6262 - recursive symlinks were fixed (439)
6263 - fixed NULL deref in DEBUG_REQ
6264 - filter_update_lastobjid no longer calls sync, which annoyed extN
6265 - fixed multi-client small-writes to a single file problem (445)
6266 - fixed mtime updates during file writes (607)
6267 - fixed vector writes on obdfilter causing problems when ENOSPC (670)
6268 - fixed bug in obd_brw_read/write() (under guise of testing 367)
6269 - fixed Linux OST size reporting problem (444, 656)
6270 - OST now updates object mtime with writes or setattr (607, 619)
6271 - client verifies file size before zeroing page past EOF (445)
6272 - OST now writes last allocated objid to disk with allocation (108)
6273 - LOV on echo now works (409)
6275 - mds_reint_unlink sends a new buffer, with the EA included. this
6276 buffer is only valid if body->valid & OBD_MD_FLEASIZE, which is only
6277 set if a regular file was being unlinked, and it was the last link
6278 - use PtlGet from the target for bulk writes (315)
6279 - OST now updates object mtime with writes or setattr (607, 619)
6280 - LDLM now has a grant-time callback to revalidate locked items, if
6282 - Many MDS operations were reorganized to combat race conditions
6284 - Merge b_intel branch (updated lprocfs code) - now at /proc/fs/lustre
6285 - configure check to avoid gcc version 2.96 20000731-2.96-98) (606)
6287 2003-01-06 Andreas Dilger <adilger@clusterfs.com>
6290 - Fully reactivate OST imports after reconnection (512, others)
6291 - Make sure client sees our -ENOTCONN from mds_handle (513 - partial)
6292 - More graceful error handling for truncating on dead OST (515)
6293 - Don't error out unless we're actually accessing dead stripes (474)
6294 - Fix garbage sizes when stripes are missing (410)
6295 - LRU counters were broken, causing constant lock purge (433, 432)
6296 - garbage on read from stripes with failed OSTs (441)
6297 - mark OSCs as active before reconnecting during recovery (438)
6298 - lov_enqueue and lov_cancel need to handle inactive OSTs (403)
6299 - lfind did not preserve OST order in output (443)
6300 - symlinks cause hung clients, incorrect data (439)
6301 - stop dereferencing request after dropping refcount (457)
6302 - don't LASSERT(spin_is_locked) on non-SMP (455)
6303 - fixes for many rename() bugs
6304 - fstat didn't correctly synchronize attributes (399)
6305 - server must handle lock cancellation during blocking AST prep (487)
6306 - bulk descriptors were free()d too soon (511)
6307 - fix paths in lconf, which would load incorrect modules (451, 507)
6308 - fix confusing lconf 'host not found' error message (386)
6309 - fix lock order deadlock on OST (O/R i_sem before journal ops, 478)
6310 - fix race condition in mdc_blocking_ast() for inode access (526)
6311 - fix lov_unpackmd() unpacking wrong number of stripes (537)
6312 - fix lov_set_osc_active() marking wrong OSC inactive (440)
6313 - fix bad lstripe lov_unpackmd() assertion (fix layering too) (527)
6314 - fix multiple writes of stripe MD to MDS (358, maybe 519)
6315 - fix lstripe in several ways (kernel side) (527)
6316 - fix request leak in ldlm_cli_enqueue (262)
6317 - incorrect OSC was marked inactive after OST failure
6318 - call mds_fs_cleanup before unmounting filesystem (524)
6319 - fix races between taking ns_lock and ldlm_lock_change_resource
6320 - fix races updating LOV export open file list
6321 - fix lov_enqueue error path, avoid decref-ing bad lock handle
6322 - fix recovery NULL deref in ldlm_cli_cancel_unused
6323 - fix some DLM races by using new hash table for lock handles (419)
6324 - permit the client to specify desired inodes, at replay
6325 - duplicate requests when we queue them for replay reintegration
6326 - fix last_rcvd offset calculation
6327 - sync after each recovered transaction, so we always make progress
6328 - never, not always, ERESTART requests without transnos
6329 - store the lov_desc in the MDS, so we don't depend on getlovinfo to
6331 - skip replay if the MDS says that the client is already connected
6332 - don't check for a recovery-enabled export to match lctl's UUID
6333 - don't INC_USE_COUNT for phantom exports
6334 - don't crash when cleaning up phantom exports (567)
6335 - don't double-finish or set replay data for errored mdc_open requests
6336 - abort requests when they time out, so we don't get old replies
6337 - send/receive replies for AST messages again
6338 - if the client says that it doesn't have the lock, cancel it on the
6340 - if we timeout during I/O, don't try to cancel an in-use lock; instead
6341 mark it as destroyed, it will all work out when decref is called
6342 - fix module use counts (22, 581)
6344 - ASTs now expect a reply (server cancels lock on error reply)
6346 2002-12-02 Andreas Dilger <adilger@clusterfs.com>
6349 - fix many simultaneous client startup (392)
6350 - fix dentry->d_it clobbering
6351 - credentials weren't being shipped for readdir/getattr operations
6352 - remove invalid assertions triggered during some concurrent MD
6354 - proper Lustre versions added (336, 389)
6355 - fix memory leak for create error case (398)
6356 - fix LOV locking bug that would get cli/srv out of sync
6357 - fix echo client over LOV (409)
6358 - fix dbench 2, extN refcount problem (170, 258, 356, 418)
6359 - fix double-O_EXCL intent crash (424)
6360 - avoid sending multiple lock CANCELs (352)
6362 - MDS can do multi-client recovery (modulo bugs in new code)
6364 - many updates, edits, cleanups
6366 2002-11-18 Phil Schwan <phil@clusterfs.com>
6369 - fix null d_it dereference (346)
6370 - fix full OST/dbench hang (333)
6371 - fix permission problem with file removal (286)
6372 - fix removal of OSCs from LOV when they fail
6373 - fix NULL deref during bulk timeout (214)
6374 - fix problems related to multiple filesystems on one MDS (241)
6375 - fixed serious subtle metadata locking bugs
6376 - free locks on clients when inodes are removed due to memory
6378 - fix inode pointer in lock data (285)
6379 - partial support for multiple MDS on a single host (241)
6380 - data locks weren't cancelled at clear_inode time (290, 311)
6381 - intent locks could lead to unbounded lock growth (205)
6382 - added a maximum lock count, an LRU list, and a flusher
6383 - fix multiple rename (365)
6384 - properly abstracted the echo client
6385 - OSC locked 1 byte too many; fixed
6386 - rewrote brw callback code:
6387 - fixed recovery bugs related to LOVs (306)
6388 - fixed too-many-pages-in-one-write crash (191)
6389 - fixed (again) crash in sync_io_timeout (214)
6390 - probably fixed callback-related race (385)
6392 - Add capability to MDS protocol
6393 - LDLM cancellations and callbacks on different portals
6395 2002-10-28 Andreas Dilger <adilger@clusterfs.com>
6398 - limit client IOV size to PTL_MD_MAX_IOV (611336, 191)
6399 - defer open object destruction to close time (601981, 138)
6400 - open/close OST file handle in obdo (OBD_MD_FLHANDLE) (601981, 138)
6401 - move LDLM_ENQUEUE/CONVERT back to MDS portal (625069)
6402 - abstract ll_lookup2, fix ll_revalidate2 to use abstraction (256)
6403 - don't call obd_setattr in ll_file_release for destroyed objects
6404 * protocol change to lustre_msg: move |version| and add |flags|
6405 * protocol change to osc_punch: "start" in "o_size", "end" in "o_blocks"
6406 * lock replay: for LDLM_FL_REPLAY trust client to do right thing
6407 * added replay of create, unlink, link and rename operations during
6408 MDS failover; recovery should be much more robust now
6409 * remove failed OSCs from LOVs (only lov_create uses this so far)
6410 * the lustre-HOWTO was brought (more) up to date (582544)
6412 2002-10-23 Phil Schwan <phil@clusterfs.com>
6415 - in-use dentries weren't being reused properly (617851)
6416 - prevent multiple LDLM setup (599178)
6417 - fix LOV size calculations for truncate (617853)
6418 - fix client handling of MDS intent errors (POSIX)
6419 - fix permission bug in lovstripe.c test (624321)
6420 - fix MDS thread deadlock - move LDLM handler to DLM portal (625069)
6421 - truncate past end of file could corrupt data
6422 - proper cleanup after timeouts, crashes, etc (592524, 550815)
6423 - a race in recovery could return ETIMEDOUT to apps (623947)
6424 - building outside the source directory was fixed
6425 * the lustre-HOWTO was brought (more) up to date (582544)
6426 * major progress was made on recovery functionality
6428 2002-10-10 Phil Schwan <phil@clusterfs.com>
6431 - recovery deadlock fix
6432 - rm -rf causes LBUG fix (617817)
6433 - file open by multiple tasks fix (618962)
6434 - directory permissions bugs (602707 and 620007)
6435 - journal_stop fixed with locking (611313)
6436 - O_APPEND failures resolved (618273, perhaps 614459)
6437 - lconf PATH fix (619770)
6438 - IA64 build fix (621450)
6439 - RPC buffer sizes scale with amount of memory
6441 2002-10-01 Phil Schwan <phil@clusterfs.com>
6444 - locks would be cancelled without throwing away data pages,
6445 resulting in inconsistent data (605627)
6446 - inode attributes were not always being refreshed (605627, 612449)
6447 - lconf now continues to cleanup after lctl reports an error
6448 - MDS now enforces user permissions (602707)
6449 - lprocfs cleanup fixed, but not yet enabled (614157)
6450 - fixed infinite server hang, should a client not respond to an AST
6451 - avoid going into recovery if user calls readlink() with a buffer
6452 that's too small (613941)
6453 - AST RPCs no longer require replies (614867) -- this may be changed
6454 - don't crash server if client sends an IOV that's too big (611336)
6455 - fixed lock conversion deadlock (611892)
6456 - fixed the following of symlinks (614622)
6457 * recovery: the server can remove locks from a client that dies, other
6458 clients can make progress
6459 * more extN patch fixes
6460 * compile-time configurable ptlrpc buffer allocations
6462 - collaborative read cache document
6463 - Lustre Lite Performance CDR document-in-progress
6465 2002-09-20 Andreas Dilger <adilger@clusterfs.com>
6468 - fix typo in patch-2.4.18
6470 2002-09-20 Andreas Dilger <adilger@clusterfs.com>
6473 - clear ptlrpc request each time in handle_incoming_request()
6474 - unlink of files now destroys the object on the OST
6476 2002-09-19 Peter Braam <braam@clusterfs.com>
6478 * add hard link support
6479 * change obdfile creation method
6480 * kernel patch changed
6482 2002-09-19 Peter Braam <braam@clusterfs.com>
6485 - stack overflow bug in extN fixed
6487 2002-09-18 Andreas Dilger <adilger@clusterfs.com>
6489 * documentation updates
6490 - add man pages for config tools
6491 - update tests/README to describe testing with new config tools
6492 - finish metadata API descriptions
6493 * bug fixes and cleanups
6494 - statfs workaround for 16TB limit
6495 - LOV stripe allocation improved, can stripe on subset of OSTs
6496 - LOV file size/IO offset was wrong for files > 4GB in size
6497 - object EA data was being dropped, caused files to be unreadable
6498 - memory overflow with non-LOV OST caused memory corruption
6499 - fixed regression tests to work with new config tools, obdfilter
6500 - fixed bug when directory size became larger than 1 block
6501 - fixed bug (for single client case) when PWD was deleted
6502 - invalidate local directory pages when doing intent-based ops
6503 - avoid LDLM oops when lock callback contained bad data
6505 2002-09-09 Andreas Dilger <adilger@clusterfs.com>
6507 * documentation updates
6508 * bug fixes and cleanups
6509 - configuration tools
6512 - 64-bit compile warnings
6513 - 64-bit internal statfs data
6515 * test_brw on persistent OST devices
6517 * lprocfs (disabled)
6519 2002-09-04 Andreas Dilger <adilger@clusterfs.com>
6521 * documentation updates
6522 * bug fixes and cleanups
6523 * configuration tools
6525 2002-08-30 Peter J. Braam <braam@clusterfs.com>
6528 * many small fixes to 0_5_4
6529 * io/network handling
6530 * thinkos in MDS operations
6532 2002-08-24 Peter J. Braam <braam@clusterfs.com>
6535 * crucial basic fixes to 0.5.3
6536 * IOR, Iozone work over Elan
6539 2002-08-07 Phil Schwan <phil@clusterfs.com>
6540 * version 0_5_3, our first alpha
6541 * we use the new Portals iovs
6542 * documentation updates
6543 * bug fixes and cleanups
6544 * small changes in the DLM wire protocol
6546 2002-07-25 Peter J. Braam <braam@clusterfs.com>
6547 * version 0_5_1 with some initial stability,
6548 * locking on MD and file I/O.
6549 * documentation updates
6550 * several bug fixes since 0.5.0
6551 * small changes in wire protocol
6553 2002-07-18 Phil Schwan <phil@clusterfs.com>
6555 * delivered as Lustre Light Alpha
6556 * fixed a crash after handling invalid MDS requests
6557 * fixed directory pages for architectures with non-4k pages sizes
6559 2002-07-11 Andreas Dilger <adilger@clusterfs.com>
6560 * release version v0_4_4
6561 * Moves TCP acceptor to be on port 2432 (unused Coda port) instead
6563 * Fixes a number of interruption problems with OST operations.
6564 * Update documentation for portals header changes
6565 * Move all wire protocol structs/defines to lustre_idl.h
6566 * Fixes symlink length bug.
6567 * Add tcpdump to repository.
6569 2002-07-05 Andreas Dilger <adilger@clusterfs.com>
6570 * release version v0_4_3
6571 * Fixes statfs for inodes on extN.
6572 * Fixes bug in runtests which would delete /etc/hosts.
6573 * Use 64-bit object IDs wherever possible (not into VFS though)
6574 Remove ost_get_info, which is unused by lustre, and out of date.
6576 2002-07-03 Peter Braam <braam@clusterfs.com>
6577 * release version v0_4_2 Fixes a lookup error (type not passed)
6578 * move forward to head of Portals
6579 * move forward to latest Lustre kernel
6581 2002-06-25 Peter Braam <braam@clusterfs.com>
6582 * release version v0_4_1. Hopefully stable on single node use.