1 tbd Sun Microsystems, Inc.
4 2.6.16.60-0.37 (SLES 10),
5 2.6.18-128.1.6.el5 (RHEL 5),
6 2.6.22.14 vanilla (kernel.org)
7 * Client support for unpatched kernels:
8 (see http://wiki.lustre.org/index.php?title=Patchless_Client)
9 2.6.16 - 2.6.22 vanilla (kernel.org)
10 * Recommended e2fsprogs version: 1.41.5.sun2
11 * File join has been disabled in this release, refer to Bugzilla 16929.
12 * A new Lustre ADIO driver is available for MPICH2-1.0.7.
13 * NFS export disabled when stack size < 8192. Since the NFSv4 export
14 of Lustre filesystem with 4K stack may cause a stack overflow. For
15 more information, please refer to bugzilla 17630.
17 Severity : enhancement
19 Description: implement async create (obd_async_create) method for osc, to avoid
20 too long waiting new ost objects with holding ldlm lock.
25 Description: fix lqs' reference which won't be put in some situations
26 Details : This patch fixes:
27 1. In quota_check_common(), this function will check quota
28 for user and group, but only send one return via "pending".
29 In most cases, the pendings should be same. But that is not
31 2. If quotaoff runs between lquota_chkquota() and
32 lquota_pending_commit(), the same thing will happen too.
34 - if (!ll_sb_any_quota_active(qctxt->lqc_sb))
37 ------------------------------------------------------------------------------
38 tbd Sun Microsystems, Inc.
40 * Support for kernels:
41 2.6.16.60-0.37 (SLES 10),
42 2.6.18-128.1.6.el5 (RHEL 5),
43 2.6.22.14 vanilla (kernel.org)
44 * Client support for unpatched kernels:
45 (see http://wiki.lustre.org/index.php?title=Patchless_Client)
46 2.6.16 - 2.6.22 vanilla (kernel.org)
47 * Recommended e2fsprogs version: 1.40.11-sun1
48 * File join has been disabled in this release, refer to Bugzilla 16929.
49 * A new Lustre ADIO driver is available for MPICH2-1.0.7.
50 * NFS export disabled when stack size < 8192. Since the NFSv4 export
51 of Lustre filesystem with 4K stack may cause a stack overflow. For
52 more information, please refer to bugzilla 17630.
54 Severity : enhancement
56 Description: Update kernel to RHEL5.3 2.6.18-128.1.6.el5.
58 Severity : enhancement
60 Description: Update kernel to SLES10 SP2 2.6.16.60-0.37.
62 Severity : enhancement
64 Description: Caching OSS
65 Details : introduce data caching on the OSS. The OSS now relies on the linux
66 kernel page cache to keep recently accessed data in memory.
67 It is worth noting that all write requests are still flushed
68 synchronously as in lustre 1.6.
70 Severity : enhancement
72 Description: version based recovery
73 Details : introduce finer grained recovery able to detect transaction
74 dependencies and can deal with transaction gaps.
76 Severity : enhancement
78 Description: Enable adaptive timeouts by default
82 Descriptoin: correctly skip time estimate if in recovery
83 Details : rq_send_state insn't bitmask so using bitwise ops is forbid.
87 Descriptoin: fix goal inodes
88 Details : Allocate inodes for llog in last inode group for avoid broke
93 Descriptoin: OSS DeadLock
94 Details : Use trylock to prevent deadlock when shrink icache.
96 Severity : enhancement
98 Description: Allow tuning service thread via /proc
99 Details : For each service a new
100 /proc/fs/lustre/{service}/*/thread_{min,max,started} entry is
101 created that can be used to set min/max thread counts, and get the
102 current number of running threads.
104 Severity : enhancement
106 Description: Add state history info file, enhance import info file
107 Details : Track import connection state changes in a new osc/mdc proc file;
108 add overview-type data to the osc/mdc import proc file.
112 Description: Reduce small size read RPC
113 Details : Set read-ahead limite for every file and only do read-ahead when
114 available read-ahead pages are bigger than 1M to avoid small size
119 Description: free_entry erroneously used groups_free instead of put_group_info
121 Severity : enhancement
123 Description: Make read-ahead stripe size aligned.
125 Severity : enhancement
127 Description: MDS create should not wait for statfs RPC while holding DLM lock.
132 Descriptoin: OST grant too much space to client even there are not enough space.
133 Details : Client will shrink its grant cache to OST if there are no write
134 activity over 6 mins (GRANT_SHRINK_INTERVAL), and OST will retrieve
135 this grant cache if there are already not enough avaible space
136 (left_space < total_clients * 32M).
138 Severity : enhancement
140 Description: Update to RHEL5U3 kernel-2.6.18-128.1.1.el5.
142 Severity : enhancement
144 Description: Update client support to vanila kernels up to 2.6.27.
147 Frequency : rare, connect and disconnect target at same time
149 Descriptoin: ASSERTION(atomic_read(&imp->imp_inflight) == 0
150 Details : don't call obd_disconnect under lov_lock. this long time
151 operation and can block ptlrpcd which answer to connect request.
153 Severity : enhancement
155 Description: Update to SLES10 kernel-2.6.16.60-0.33.
157 Severity : enhancement
159 Description: Update to sles9 kernel-2.6.5-7.315.
162 Frequency : start MDS on uncleanly shutdowned MDS device
164 Descriptoin: ll_sync thread stay in waiting mds<>ost recovery finished
165 Details : stay in waiting mds<>ost recovery finished produce random bugs
166 due race between two ll_sync thread for one lov target. send
167 ACTIVATE event only if connect realy finished and import have
170 Severity : enhancement
172 Description: Update to RHEL5 kernel-2.6.18-92.1.22.el5.
175 Frequency : start MDS on uncleanly shutdowned MDS device
177 Descriptoin: aborting recovery hang on MDS
178 Details : don't throttle destroy RPCs for the MDT.
180 Severity : enhancement
182 Descriptoin: Don't sync journal after every i/o
183 Details : Implement write RPC replay to allow server replies for write RPCs
184 before data is on disk. However, this feature is disabled by
185 default since some issues leading to data corruptions have been
186 found during recovery (e.g. bug 19128). This feature can be enabled
187 by running the following command on the OSSs:
188 lctl set_param obdfilter.*.sync_journal=0
192 Description: Slow reads beyond 8Tb offsets.
193 Details : Page index integer overflow in ll_read_ahead_page
197 Description: Soft lockup on OSS after MDS failover
198 Details : MSG_CONNECT_INITIAL is not set on the initial MDS->OST connect.
199 As a conseqence, the patch from bug 18224 is not operational
200 and the MDS export cannot be reused on the OSTs until it gets
204 Frequency : rare, only if using MMP with Linux RAID
206 Description: MMP doesn't work with Linux RAID
207 Details : While using HA for Lustre servers with Linux RAID, it is possible
208 that MMP will not detect multiple mounts. To make this work we
209 need to unplug the device queue in RAID when the MMP block is being
210 written. Also while reading the MMP block, we should read it from
211 disk and not the cached one.
214 Frequency : rare, during recovery
216 Description: Assertion failure in ldlm_lock_put
217 Details : Do not put cancelled locks into replay list, hold references on
222 Description: 1.6.5 mdsrate performance is slower than 1.4.11/12 (MDS is not cpu bound!)
223 Details : create_count always drops to the min value (=32) because grow_count
224 is being changed before the precreate RPC completes.
227 Frequency : Only in RHEL5 when mounting multiple ext3 filesystems
230 Description: "kmem_cache_create: duplicate cache jbd_4k" error message
231 Details : add proper locking for creation of jbd_4k slab cache
235 Description: MMP check in ext3_remount() fails without displaying any error
236 Details : When multiple mount protection fails during remount, proper error
239 -------------------------------------------------------------------------------
240 2008-12-31 Sun Microsystems, Inc.
242 * Support for kernels:
243 2.6.5-7.314 (SLES 9),
244 2.6.9-67.0.22.EL (RHEL 4),
245 2.6.16.60-0.31 (SLES 10),
246 2.6.18-92.1.17.el5 (RHEL 5),
247 2.6.22.14 vanilla (kernel.org)
248 * Client support for unpatched kernels:
249 (see http://wiki.lustre.org/index.php?title=Patchless_Client)
250 2.6.16 - 2.6.22 vanilla (kernel.org)
251 * Client support for unpatched kernels:
252 we do not recommend using patchless RHEL4 clients with kernels
253 prior to 2.6.9-55EL (RHEL4U5).
254 * Recommended e2fsprogs version: 1.40.11-sun1
255 * Note that reiserfs quotas are disabled on SLES 10 in this kernel.
256 * RHEL 4 and RHEL 5/SLES 10 clients behaves differently on 'cd' to a
257 removed cwd "./" (refer to Bugzilla 14399).
258 * A new quota file format has been introduced in 1.6.5.
259 The format conversion from prior releases is handled transparently,
260 but releases older than 1.4.12/1.6.5 don't understand this new
261 format. The automatic format conversion can be avoided by running
262 the following command on the MDS:
263 'tunefs.lustre --param="mdt.quota_type=ug1" $MDTDEV'.
264 For more information, please refer to bugzilla 13904.
265 * A new quota file format was introduced in 1.6.6/1.8.0 (kernels 2.6.16+).
266 The format conversion from prior releases is handled transparently,
267 but releases older than 1.6.6/1.8.0 don't understand this new
268 format. The automatic format conversion can be avoided by running
269 the following commands on the MDS and OSS servers (for
270 pre 1.4.12-1.6.5 quota files):
271 'tunefs.lustre --param="mdt.quota_type=ug1" $MDTDEV',
272 'tunefs.lustre --param="ost.quota_type=ug1" $MDTDEV'
273 or (for 1.4.12/1.6.5 quota files)
274 'tunefs.lustre --param="mdt.quota_type=ug2" $MDTDEV',
275 'tunefs.lustre --param="ost.quota_type=ug2" $MDTDEV'
276 For more information, please refer to bugzilla 13904.
277 * Output of lfs quota has been made less detailed by default,
278 old (verbose) output can be obtained by using -v option.
279 * File join has been disabled in this release, refer to Bugzilla 16929.
280 * A new Lustre ADIO driver is available for MPICH2-1.0.7.
281 * NFS export disabled when stack size < 8192. Since the NFSv4 export of
282 Lustre filesystem with 4K stack may cause a stack overflow. For more
283 information, please refer to bugzilla 17630.
285 Severity : enhancement
287 Description: add lazystatfs mount option to allow statfs(2) to skip down OSTs
288 Details : allow skip disconnected ost for send statfs request and hide error
292 Frequency : rare, on llog test 6
294 Descriptoin: don't allow connect to already connected import
295 Details : allowing connect to already connected import is hide connecting problem.
298 Frequency : rare, on failed llog setup
300 Descriptoin: don't leak obd reference on failed llog setup
301 Details : for failed llog setup - mgc forget call class_destroy_import for
302 client import, move destroy import to more generic place.
307 Descriptoin: allow kill process which wait statahead result
308 Details : for some reasons 'ls' can stick in waiting result from statahead,
309 in this case need way for kill this process.
312 Frequency : rare, at shutdown
314 Descriptoin: panic at umount
315 Details : llap_shrinker can be raced with killing super block from list and
316 this produce panic with access to already freeded pointer
321 Descriptoin: don't lose wakeup for imp_recovery_waitq
322 Details : recover_import_no_retry or invalidate_import and import_close can
323 both sleep on imp_recovery_waitq, but we was send only one wakeup
329 Descriptoin: panic in mds_open
330 Details : don't confuse mds_finish_transno() with PTR_ERR(-ENOENT)
335 Descriptoin: stuck in cache_remove_extent() or panic with accessing to already
337 Details : release lock refernce only after add page to pages list.
340 Frequency : always with long access acl
342 Descriptoin: mds can't pack reply with long acl.
343 Details : mds don't control size of acl but they limited by reint/getattr
348 Frequency : on remount
350 Description: external journal device not working after the remount
351 Details : clear dev_rdonly flag for external journal devices in
357 Description: shutdown vs evict race
358 Details : client_disconnect_export vs connect request race.
359 if client will evicted at this time - we start invalidate
360 thread without referece to import and import can be freed
366 Description: shrink LOV EAs before replying
367 Details : correctly adjust LOV EA buffer for reply.
372 Description: don't skip ost target if they assigned to file
373 Details : Drop slow OSCs if we can, but not for requested start idx.
374 This means "if OSC is slow and it is not the requested
375 start OST, then it can be skipped, otherwise skip it only
376 if it is inactive/recovering/out-of-space.
378 Severity : enhancement
380 Description: Update to RHEL5 kernel-2.6.18-92.1.17.el5.
383 Frequency : rare, need acl's on inode.
385 Description: client can't handle ost additional correctly
386 Details : if ost was added after client connected to mds client can have
387 hit lnet_try_match_md ... to big messages to wide striped files.
388 in this case need teach client to handle config events about add
389 lov target and update client max ea size at that event.
391 Severity : enhancement
393 Description: Update to sles9 kernel-2.6.5-7.314.
395 Severity : enhancement
397 Description: Update to SLES10 SP2 kernel-2.6.16.60-0.31.
400 Frequency : Create a symlink file with a very long name
402 Description: ldlm_cancel_pack()) ASSERTION(max >= dlm->lock_count + count)
403 Details : If there is no extra space in the request for early cancels,
404 ldlm_req_handles_avail() returns 0 instead of a negative value.
409 Description: mds is deadlocked
410 Details : in rare cases, inode in catalog can have i_no less than have parent
411 i_no, this produce wrong order for locking during open, and
412 parallel unlink can be lock open. this need teach mds_open to grab
413 locks in resource id order, not at parent -> child order.
415 Severity : enhancement
417 Description: Add /proc entry for import status
418 Details : The mdc, osc, and mgc import directories now have
419 an import directory that contains useful import data for debugging
422 Severity : enhancement
424 Description: Re-disable certain /proc logging
425 Details : Enable and disable client's offset_stats, extents_stats and
426 extents_stats_per_process stats logging on the fly.
429 Frequency : Only on FC kernels 2.6.22+
431 Description: oops in statahead
432 Details : Do not drop reference count for the dentry from VFS when lookup,
433 VFS will do that by itself.
435 Severity : enhancement
437 Description: Generic /proc file permissions
438 Details : Set /Proc file permissions in a more generic way to enable non-
439 root users operate on some /proc files.
443 Description: Hitting mdc_commit_close() ASSERTION
444 Details : Properly handle request reference release in
445 ll_release_openhandle().
449 Frequency : only patchless client
450 Description: add workaround for race between add/remove dentry from hash
452 Severity : enhancement
454 Description: Allow OST glimpses to return PW locks
458 Description: LBUG when llog conf file is full
459 Details : When llog bitmap is full, ENOSPC should be returned for plain
464 Description: Prevent import from entering FULL state when server in recovery
468 Description: service mount cannot take device name with ":"
469 Details : Only when device name contains ":/" will mount treat it as
475 Description: replace ptlrpcd with the statahead thread to interpret the async
476 statahead RPC callback
480 Frequency : on recovery
481 Description: I/O failures after umount during fail back
482 Details : if client reconnected to restarted server we need join to recovery
483 instead of find server handler is changed and process self eviction
484 with cancel all locks.
486 Severity : enhancement
488 Description: Update to RHEL5 kernel-2.6.18-92.1.10.el5.
492 Description: Kernel BUG tries to release flock
493 Details : Lustre does not destroy flock lock before last reference goes
494 away. So always drop flock locks when client is evicted and
495 perform unlock regardless of successfulness of speaking to MDS.
497 Severity : enhancement
499 Description: Update to SLES10 SP2 kernel-2.6.16.60-0.27.
501 Severity : enhancement
503 Description: Upcall on Lustre log has been dumped
504 Details : Allow for a user mode script to be called once a Lustre log has
505 been dumped. It passes the filename of the dumped log to the
506 script, the location of the script can be specified via
507 /proc/sys/lnet/debug_log_upcall.
512 Description: avoid messages about idr_remove called for id that is not allocated
513 Details : Move assigment s_dev for clustered nfs to end of initialization,
514 for avoid problem with error handling.
519 Description: avoid Already found the key in hash [CONN_UNUSED_HASH] messages
520 Details : When connection is reused this not moved from CONN_UNUSED_HASH into
521 CONN_USED_HASH and this prodice warning when put connection again
527 Description: avoid ASSERTION(client_stat->nid_exp_ref_count == 0) failed
528 Details : release reference to stats when client disconnected, not
529 when export destroyed for avoid races when client destroyed
530 after main ost export.
534 Description: more cleanup in mds_lov
535 Details : add workaround for get valid ost count for avoid warnings about
536 drop too big messages, not init llog cat under semphore which
537 can be blocked on reconnect and break normal replay, fix access
540 Severity : enhancement
542 Description: File striping can now be set to use an arbitrary pool of OSTs.
544 Severity : enhancement
546 Description: Export bytes_read/bytes_write count on OSC/OST.
550 Description: Early reply size mismatch, MGC loses connection
551 Details : Apply the MGS_CONNECT_SUPPORTED mask at reconnect time so
552 the connect flags are properly negotiated.
556 Description: Properly propagate oinfo flags from lov to osc for statfs
557 Details : restore missing copy oi_flags to lov requests.
561 Description: exports in /proc are broken
562 Details : recreate /proc entries for clients when they reconnect.
564 Severity : enhancement
566 Description: Add man pages for llobdstat(8), llstat(8), plot-llstat(8),
567 : l_getgroups(8), lst(8), routerstat(8)
568 Details : included man pages for llobdstat(8), llstat(8),
569 : plot-llstat(8), l_getgroups(8), lst(8), routerstat(8)
571 Severity : enhancement
573 Description: Implement lustre ll_show_options method.
575 Severity : enhancement
577 Description: Update to SLES9 kernel-2.6.5-7.312.
579 Severity : enhancement
581 Description: Update to RHEL4 kernel-2.6.9-67.0.22.EL.
585 Description: exports in /proc are broken
586 Details : recreate /proc entries for clients when they reconnect.
590 Description: don't fail open with -ERANGE
591 Details : if client connected until mds will be know about real ost count
592 get LOV EA can be fail because mds not allocate enougth buffer
597 Description: Resolve device initialization race
598 Details : Prevent proc handler from accessing devices added to the
599 obd_devs array but yet be intialized.
601 Severity : enhancement
603 Description: configure's --enable-quota should check the
604 : kernel .config for CONFIG_QUOTA
605 Details : configure is terminated if --enable-quota is passed but
606 : no quota support is in kernel
608 Severity : enhancement
610 Description: Update to SLES10 SP2 kernel-2.6.16.60-0.23.
612 Severity : enhancement
614 Description: Update to RHEL5 kernel-2.6.18-92.1.6.el5.
618 Frequency : rare, on PPC clients
619 Description: don't swab ost objects in response about directory, because
621 Details : bug similar bug 14856, but in different function.
623 Severity : enhancement
625 Description: lfs quota tool enhancement
626 Details : added units specifiers support for setquota, default to
627 current uid/gid for quota report, short quota stats by
628 default, nonpositional parameters for setquota, added
629 llapi_quotactl manual page.
631 Severity : enhancement
633 Description: *optional* service tags registration
634 Details : if the "service tags" package is installed on a Lustre node
635 When the filesystem is mounted, a local-node service tag will
636 be created. See http://inventory.sun.com/ for more information
637 about the Service Tags asset management system.
641 Description: Client runs out of low memory
642 Details : Consider only lowmem when counting initial number of llap pages
645 Description: Kernel BUG tries to release flock
646 Details : Lustre does not destroy flock lock before last reference goes
647 away. So always drop flock locks when client is evicted and
648 perform unlock regardless of successfulness of speaking to MDS.
651 Frequency : occasional
653 Description: add refcount for osc callbacks, so avoid panic on shutdown
655 Severity : enhancement
657 Description: Update to RHEL4 kernel-2.6.9-67.0.20.
660 Frequency : testing only
662 Description: sanity test 65a fails if stripecount of -1 is set
663 Details : handle -1 striping on filesystem in ll_dirstripe_verify
666 Frequency : only in unusual configurations
668 Description: Kernel panic with find ost index.
669 Details : lov_obd have panic if some OST's have sparse indexes.
671 Severity : enhancement
673 Description: Update to RHEL5 kernel-2.6.18-53.1.21.el5.
676 Frequency : rarely, if filesystem is mounted with -o flock
678 Description: do not process already freed flock
679 Details : flock can possibly be freed by another thread before it reaches
680 to ldlm_flock_completion_ast.
683 Frequency : rarely, if filesystem is mounted with -o flock
685 Description: LBUG during stress test
686 Details : Need properly lock accesses the flock deadlock detection list.
689 Frequency : rarely, if binaries are being run from Lustre
691 Description: oops in page fault handler
692 Details : kernel page fault handler can return two special 'pages' in
693 error case, don't try dereference NOPAGE_SIGBUS and NOPAGE_OMM.
696 Frequency : rarely, during shutdown
698 Description: timeout with invalidate import.
699 Details : ptlrpcd_check call obd_zombie_impexp_cull and wait request which
700 should be handled by ptlrpcd. This produce long age waiting and
701 -ETIMEOUT ptlrpc_invalidate_import and as result LASSERT.
707 Description: ASSERTION(CheckWriteback(page,cmd)) failed
708 Details : badly clear PG_Writeback bit in ll_ap_completion can produce false
712 Frequency : only with broken builds/installations
714 Description: no LBUG if lquota.ko and fsfilt_ldiskfs.ko are different versions
715 Details : just return an error to a user, put a console error message
717 Severity : enhancement
719 Description: Update to RHEL5 kernel-2.6.18-53.1.19.el5.
721 Severity : enhancement
723 Description: Update to RHEL4 kernel-2.6.9-67.0.15.
725 Severity : enhancement
727 Description: enable MGS and MDT services start separately
728 Details : add a 'nomgs' option in mount.lustre to enable start a MDT with
729 a co-located MGS without starting the MGS, which is a complement
730 to 'nosvc' mount option.
733 Frequency : always, on big-endian systems
735 Description: cleanup in ptlrpc code, related to PPC platform
736 Details : store magic in native order avoid panic's in recovery on PPC
737 node and forbid from this error in future. Also fix posibily
738 of twice swab data. Fix get lov striping to userpace.
741 Frequency : rarely, if replay get lost on server
743 Description: server incorrectly drop resent replays lead to recovery failure.
744 Details : do not drop replay according to msg flags, instead we check the
745 per-export recovery request queue for duplication of transno.
748 Frequency : after recovery
750 Description: precreate to many object's after del orphan.
751 Details : del orphan st in oscc last_id == next_id and this triger growing
752 count of precreated objects. Set flag LOW to skip increase count
753 of precreated objects.
756 Frequency : rare, on clear nid stats
758 Description: ASSERTION(client_stat->nid_exp_ref_count == 0)
759 Details : when clean nid stats sometimes try destroy live entry,
760 and this produce panic in free.
763 Frequency : occasionally since 1.6.4
765 Description: Stack overflow during MDS log replay
766 Details : ease stack pressure by using a thread dealing llog_process.
769 Frequency : very rare
771 Description: MDT cannot be unmounted, reporting "Mount still busy"
772 Details : Mountpoint references were being leaked during open reply
773 reconstruction after an MDS restart. Drop mountpoint reference
774 in reconstruct_open() and free dentry reference also.
779 Description: wait until IO finished before start new when do lock cancel.
780 Details : VM protocol want old IO finished before start new, in this case
781 need wait until PG_writeback is cleared until check dirty flag
782 and call writepages in lock cancel callback.
787 Description: mds_mfd_close() ASSERTION(rc == 0)
788 Details : In mds_mfd_close(), we need protect inode's writecount change
789 within its orphan write semaphore to prevent possible races.
792 Frequency : rare, on shutdown ost
794 Description: don't hit live lock with umount ost.
795 Details : shrink_dcache_parent can be in long loop with destroy dentries,
796 use shrink_dcache_sb instead.
799 Frequency : only when echo_client is used
801 Description: don't panic with use echo_client
802 Details : echo client pass NULL as client nid pointer and this produce NULL
806 Frequency : Always on 32-bit PowerPC systems
808 Description: fix build on PPC32
809 Details : compile code with -m64 flag produce wrong object file for PPC32.
814 Description: MDS LBUG: ASSERTION(!IS_ERR(dchild))
815 Details : In reconstruct_* functions, LASSERTs on both the data supplied
816 by a client, and the data on disk are dangerous and incorrect.
817 Change them with client eviction.
819 Severity : enhancement
821 Description: skiplist implementation simplification
822 Details : skiplists are used to group compatible locks on granted list
823 that was implemented as tracking first and last lock of each lock
824 group the patch changes that to using doubly linked lists
828 Description: delete compatibility for 32bit qdata
829 Details : as planned, when lustre is beyond b1_8, lquota won't support 32bit
830 qunit. That means servers of b1_4 and servers of b1_8 can't be
831 used together if users want to use quota.
834 Frequency : only with administrator action
836 Description: mount failure if config log has invalid conf_param setting
837 Details : If administrator specified an incorrect configuration parameter
838 with "lctl conf_param" this would cause an error during future
839 client mounts. Instead, ignore the bad configuration parameter.
842 Frequency : blocks per group < blocksize*8 and uninit_groups is enabled
844 Description: ldiskfs error: XXX blocks in bitmap, YYY in gd
845 Details : If blocks per group is less than blocksize*8, set rest of the
849 Frequency : Application do stride read on lustre
851 Description: The read performance will drop a lot if the application does
853 Details : Because the stride_start_offset are missing in stride read-ahead,
854 it will cause clients read a lot of unused pages in read-ahead,
855 then the read-performance drops.
859 Description: more ldlm soft lockups
860 Details : In ldlm_resource_add_lock(), call to ldlm_resource_dump()
861 starve other threads from the resource lock for a long time in
862 case of long waiting queue, so change the debug level from
863 D_OTHER to the less frequently used D_INFO.
865 Severity : enhancement
867 Description: add -gid, -group, -uid, -user options to lfs find
869 Severity : enhancement
871 Description: ll_recover_lost_found_objs - recover objects in lost+found
872 Details : OST corruption and subsequent e2fsck can leave objects in the
873 lost+found directory. Using the "ll_recover_lost_found_objs"
874 tool, these objects can be retrieved and data can be salvaged
875 by using the object ID saved in the fid EA on each object.
880 Description: this bug _only_ happens when inode quota limitation is very low
881 (less than 12), so that inode quota unit is 1 at initialization.
882 Details : if remaining quota equates 1, it is a sign to demonstate that quota
883 is effective now. So least quota qunit should be 2.
887 Description: Hung threads in invalidate_inode_pages2_range
888 Details : The direct IO path doesn't call check_rpcs to submit a new RPC once
889 one is completed. As a result, some RPCs are stuck in the queue
894 Description: Procfs and llog threads access destoryed import sometimes.
895 Details : Sync the import destoryed process with procfs and llog threads by
896 the import refcount and semaphore.
900 Description: mds fails to respond, threads stuck in ldlm_completion_ast
901 Details : Sort source/child resource pair after updating child resource.
906 Description: kernel BUG at ldiskfs2_ext_new_extent_cb
907 Details : If insertion of an extent fails, then discard the inode
908 preallocation and free data blocks else it can lead to duplicate
913 Description: don't always update ctime in ext3_xattr_set_handle()
914 Details : Current xattr code updates inode ctime in ext3_xattr_set_handle()
915 In some cases the ctime should not be updated, for example for
916 2.0->1.8 compatibility it is necessary to delete an xattr and it
917 should not update the ctime.
921 Description: add quota statistics
922 Details : 1. sort out quota proc entries and proc code.
923 2. add quota statistics
928 Description: quotas are not honored with O_DIRECT
929 Details : all writes with the flag O_DIRECT will use grants which leads to
930 this problem. Now using OBD_BRW_SYNC to guard this.
934 Bugzilla : 15713/16362
935 Description: Assertion in iopen_connect_dentry in 1.6.3
936 Details : looking up an inode via iopen with the wrong generation number can
937 populate the dcache with a disconneced dentry while the inode
938 number is in the process of being reallocated. This causes an
939 assertion failure in iopen since the inode's dentry list contains
940 both a connected and disconnected dentry.
944 Description: assertion failure in ldlm_handle2lock()
945 Details : fix a race between class_handle_unhash() and class_handle2object()
946 introduced in lustre 1.6.5 by bug 13622.
948 Severity : enhancement
950 Description: superblock lock contention with many SMP cores on one client
951 Details : several client filesystem locks were highly contended on SMP
952 NUMA systems with 8 or more cores. Per-CPU datastructures
953 and more efficient locking implemented to reduce contention.
958 Description: Kernel BUG: sd_iostats_bump: unexpected disk index
959 Details : remove the limit of 256 scsi disks in the sd_iostat patch
964 Description: oops in sd_iostats_seq_show()
965 Details : unloading/reloading the scsi low level driver triggers a kernel
966 bug when trying to access the sd iostat file.
971 Description: Kernel panics during QLogic driver reload
972 Details : REQ_BLOCK_PC requests are not handled properly in the sd iostat
973 patch, causing memory corruption.
978 Description: journal_dev option does not work in b1_6
979 Details : pass mount option during pre-mount.
981 Severity : enhancement
983 Description: Add a FIEMAP(FIle Extent MAP) ioctl for ldiskfs
984 Details : FIEMAP ioctl will allow an application to efficiently fetch the
985 extent information of a file. It can be used to map logical blocks
986 in a file to physical blocks in the block device.
989 Frequency : only with adaptive timeout enabled
991 Description: DEBUG_REQ() bad paging request
992 Details : ptlrpc_at_recv_early_reply() should not modify req->rq_repmsg
993 because it can be accessed by reply_in_callback() without the
997 Frequency : only on Cray X2
999 Description: X2 build failures
1000 Details : fix build failures on Cray X2.
1004 Description: xid & resent requests
1005 Details : Initialize RPC XID from clock at startup (randomly if clock is
1010 Description: quota recovery deadlock during mds failover
1011 Details : This patch includes att18982, att18236, att18237 in bz14840.
1013 1. fix osts hang when mds does failover with quotaon
1014 2. prevent watchdog storm when osts threads wait for the
1019 Description: kernel panic on racer
1020 Details : Do not access dchild->d_inode when IS_ERR(dchild) is true.
1022 Severity : enhancement
1024 Description: Add lustre_start utility to start or stop multiple Lustre servers
1029 Description: Lustre GPF in {:ptlrpc:ptlrpc_server_free_request+373}
1030 Details : In case of memory pressure, list_del() can be called twice on
1031 req->rq_history_list, causing a kernel oops.
1035 Description: (ptllnd_peer.c:557:kptllnd_peer_check_sends()) ASSERTION(!in_interrupt()) failed
1036 Details : fix stack overflow in the distributed lock manager by defering
1037 export eviction after a failed ast to the elt thread instead of
1038 handling it in the dlm interpret routine.
1040 Severity : enhancement
1042 Description: More exported tunables for mballoc
1043 Details : Add support for tunable preallocation window and new tunables for
1044 large/small requests
1048 Description: Detect corruption of block bitmap and checking for preallocations
1049 Details : Checks validity of on-disk block bitmap. Also it does better
1050 checking of number of applied preallocations. When corruption is
1051 found, it turns filesystem readonly to prevent further corruptions.
1055 Frequency : only for big-endian servers
1056 Description: Check for big-endian system while mounting fs with extents feature
1057 Details : Mounting a filesystem with extents feature will fail on big-endian
1058 systems since ext3-based ldiskfs is not supported on big-endian
1059 systems. Can be over-ridden with "bigendian_extents" mount option.
1063 Description: Excessive recovery window
1064 Details : With AT enabled, the recovery window can be excessively long (6000+
1065 seconds). To address this problem, we no longer use
1066 OBD_RECOVERY_FACTOR when extending the recovery window (the connect
1067 timeout no longer depends on the service time, it is set to
1068 INITIAL_CONNECT_TIMEOUT now) and clients report the old service
1069 time via pb_service_time.
1073 Description: Watchdog triggered on MDS failover
1074 Details : enable OBD_CONNECT_MDT flag when connecting from the MDS so that
1075 the OSTs know that the MDS "UUID" can be reused for the same export
1076 from a different NID, so we do not need to wait for the export to
1081 Descriptoin: Lustre detected file system corruption with inode out of bounds
1082 Details : don't update i_size on MDS_CLOSE for directories. This causes
1083 directory corruptions on the MDT.
1085 -------------------------------------------------------------------------------
1087 2008-05-26 Sun Microsystems, Inc.
1089 * Support for kernels:
1090 2.6.5-7.311 (SLES 9),
1091 2.6.9-67.0.7.EL (RHEL 4),
1092 2.6.16.54-0.2.5 (SLES 10),
1093 2.6.18-53.1.14.el5 (RHEL 5),
1094 2.6.22.14 vanilla (kernel.org)
1095 * Client support for unpatched kernels:
1096 (see http://wiki.lustre.org/index.php?title=Patchless_Client)
1097 2.6.16 - 2.6.22 vanilla (kernel.org)
1098 * Due to problems with nested symlinks and FMODE_EXEC (bug 12652),
1099 we do not recommend using patchless RHEL4 clients with kernels
1100 prior to 2.6.9-55EL (RHEL4U5).
1101 * Recommended e2fsprogs version: 1.40.7-sun1
1102 * Note that reiserfs quotas are disabled on SLES 10 in this kernel.
1103 * RHEL 4 and RHEL 5/SLES 10 clients behaves differently on 'cd' to a
1104 removed cwd "./" (refer to Bugzilla 14399).
1105 * A new quota file format has been introduced in 1.6.5.
1106 The format conversion from prior releases is handled transparently,
1107 but releases older than 1.4.12/1.6.5 will not understand this new
1108 format. The automatic format conversion can be avoided by running
1109 the following command on the MDS before upgrading:
1110 'tunefs.lustre --param="mdt.quota_type=ug1" $MDTDEV'.
1111 For more information, please refer to bugzilla 13904.
1115 Description: quota performance fix
1116 Details : quota data is written in journalled mode instead of ordered to
1117 increase performance
1121 Description: lfs support for human-readable quota grace time strings
1122 Details : lfs setquota -t and lfs quota -t represent quota grace times
1123 in "XXwXXdXXhXXmXXs" format instead of large values in seconds
1126 Frequency : always with o2ib 1.3 and sles10
1128 Description: fix build with SLES10 and o2ib v3.
1129 Details : sles10 uses diffrent name for Module.symver file but configure
1130 assume this file has same name on RHEL/SLES/vanila kernels.
1133 Frequency : very rare, if additional xattrs are used on kernels >= 2.6.12
1135 Description: MDS may lose file striping (and hence file data) in some cases
1136 Details : If there are additional extended attributes stored on the MDS,
1137 in particular ACLs, SELinux, or user attributes (if user_xattr
1138 is specified for the client mount options) then there is a risk
1139 of attribute loss. Additionally, the Lustre file striping
1140 needs to be larger than default (e.g. striped over all OSTs),
1141 and an additional attribute must be stored initially in the
1142 inode and then increase in size enough to be moved to the
1143 external attribute block (e.g. ACL growing in size) for file
1146 Severity : enhancement
1148 Description: add message levels for liblustreapi
1153 Description: MDT cannot be unmounted, reporting "Mount still busy"
1154 Details : Mountpoint references were being leaked during open reply
1155 reconstruction after an MDS restart. Drop mountpoint reference
1156 in reconstruct_open() and free dentry reference also.
1161 Description: fix for occasional failure case of -ENOSPC in recovery-small tests
1162 Details : Move the 'good_osts' check before the 'total_bavail' check. This
1163 will result in an -EAGAIN and in the exit call path we call
1164 alloc_rr() which will with increasing aggressiveness attempt to
1165 aquire precreated objects on the minimum number of required OSCs.
1169 Description: Use old size assignment to avoid deadlock
1170 Details : This reverts the changes in bugs 2369 and bug 14138 that introduced
1171 the scheduling while holding a spinlock. We do not need locking
1172 for size in ll_update_inode() because size is only updated from
1173 the MDS for directories or files without objects, so there is no
1174 other place to do the update, and concurrent access to such inodes
1175 are protected by the inode lock.
1179 Description: Use __u64 instead of int for valid bits
1183 Description: resolve "_IOWR redefined" build error on SLES10
1187 Description: dump the memory debugging after all modules are unloaded to
1188 suppress false negative in conf_sanity test 39
1192 Description: the recovery timer never expires
1193 Details : for new client connect request, the recovery timer should not be
1194 reset, otherwise recovery timer will never expired, if the old
1195 client never come. Only old client connect and first connection
1196 req should trigger recovery timer reset.
1200 Description: the min numbers of lproc stats are wrong
1201 Details : adding a new constant LC_MIN_INIT and use it for initialization
1205 Frequency : always with interactive lfs
1207 Description: Reinitialize optind to 0 so that interactive lfs works in all cases
1210 Frequency : with multiple concurrent readdir processes in same directory
1211 Bugzilla : 15406, 15169, 15175
1212 Description: misc fixes for directory readahead.
1213 Details : prevent previous statahead async RPC callback from processing the
1214 current "statahead_info", race condition between async RPC callback
1215 add dentry into dentry hash table and "ls" thread revalidate such
1216 dentry, statahead his/miss control for hidden items, and so on.
1218 Severity : enhancement
1220 Description: build kernel-ib packages for OFED 1.3 in our release cycle
1224 Description: incore types cleaning in quota code (with respect to 64-bit limits)
1225 Details : several u32 variables declarations are replaced with u64 declarations
1230 Description: fix SLES kernel versioning
1231 Details : the kernel version for our SLES 10 kernel did not include a "-"
1232 before the "smp" at the end. while this was not a problem in
1233 general, it did mean that software trying to use the kernel
1234 version to try to detect a vendor specific kernel would fail.
1235 this was most evident by the OFED build scripts.
1240 Description: Don't update lov_desc members until making sure they are valid
1241 Details : When updating lov_desc members via proc fs, need fix their
1242 validities before doing the real update.
1245 Frequency : very rare
1247 Description: don't put request into delay list while invalidate in flight.
1248 Details : ptlrpc_delay_request sometimes put in delay list while invalidate
1249 import in flight. this produce timeout for invalidate and sometimes
1250 can cause stale data.
1252 Severity : enhancement
1254 Description: Update kernel to SLES9 2.6.5-7.311.
1256 Severity : enhancement
1258 Description: Update kernel to RHEL4 2.6.9-67.0.7.
1263 Frequency : on PPC only
1264 Description: not convert ost objects for directory because it's not exist.
1265 Details : ll_dir_getstripe assume dirrectory has ost objects but this wrong.
1267 Severity : enhancement
1269 Description: Fix warnings with compile liblustre at sles10/rhel5 which have
1270 __u64 as usingied long long type.
1273 Frequency : rare, on shutdown
1275 Description: race process ast vs remove callback
1276 Details : removing callback before disconnect import open race with
1277 processing callback.
1279 Severity : enhancement
1281 Description: Update kernel to SLES9 2.6.5-7.311.
1283 Severity : enhancement
1285 Description: Files open for execute are not marked busy on SLES10
1286 Details : Add FMODE_EXEC to SLES10 SP1 server kernel series.
1288 Severity : enhancement
1290 Description: Add server support for vanilla-2.6.22.14.
1293 Frequency : occasional
1295 Description: Avoid lov_create() getting stuck in obd_statfs_rqset()
1296 Details : If an OST is down the MDS will hang indefinitely in
1297 obd_statfs_rqset() waiting for the statfs data. While for
1298 MDS QOS usage of statfs, it should not stuck in waiting.
1300 Severity : enhancement
1302 Description: Disable adaptive timeouts by default
1305 Frequency : on network error
1307 Description: panic with double free request if network error
1308 Details : mdc_finish_enqueue is finish request if any network error occuring,
1309 but it's true only for synchronus enqueue, for async enqueue
1310 (via ptlrpcd) this incorrect and ptlrpcd want finish request
1314 Frequency : rare, on recovery
1316 Description: read procfs can produce deadlock in some situation
1317 Details : Holding lprocfs lock with send rpc can produce block for destroy
1318 obd objects and this also block reconnect with -EALREADY.
1319 This isn't fix all lprocfs bugs - but make it rare.
1321 Severity : enhancement
1323 Description: Update kernel to RHEL5 2.6.18-53.1.14.el5.
1326 Frequency : frequent on X2 node
1328 Description: mdc_set_open_replay_data LBUG
1329 Details : Set replay data for requests that are eligible for replay.
1334 Description: lustre_mgs: operation 101 on unconnected MGS
1335 Details : When MGC is disconnected from MGS long enough, MGS will evict the
1336 MGC, and late on MGC cannot successfully connect to MGS and a lot
1337 of the error messages complaining that MGS is not connected.
1340 Frequency : rare, depends on device drivers and load
1342 Description: MDS or OSS nodes crash due to stack overflow
1343 Details : Code changes in 1.6.4 increased the stack usage of some functions.
1344 In some cases, in conjunction with device drivers that use a lot
1345 of stack, the MDS (or possibly OSS) service threads could overflow
1346 the stack. One change which was identified to consume additional
1347 stack has been reworked to avoid the extra stack usage.
1349 Severity : enhancement
1351 Description: Update to RHEL5 latest kernel-2.6.18-53.1.13.el5.
1353 Severity : enhancement
1355 Description: Update to SLES10 SP1 latest kernel-2.6.16.54-0.2.5.
1357 Severity : enhancement
1359 Description: Update to RHEL5 latest kernel-2.6.18-53.1.6.el5.
1361 Serverity : enhancement
1363 Description: Update RHEL4 kernel to 2.6.9-67.0.4.
1366 Frequency : rare on shutdown OST
1368 Description: Don't allow skipping OSTs if index has been specified.
1369 Details : Don't allow skipping OSTs if index has been specified, make locking
1370 in internal create lots better.
1375 Description: ASSERTION(!PageDirty(page)) failed
1376 Details : Wrong check could lead to an assertion failure under specific
1382 Description: LBUG in ptlrpc_check_set() bad phase ebc0de00
1383 Details : access to bitfield in structure is always rounded to long
1384 and this produce problem with not atomic change any bit.
1389 Description: Lustre make rpms failed.
1390 Details : Remove ldiskfs spec file to avoids rpmbuild be confused when
1391 builds Lustre rpms from tarball.
1393 Severity : enhancement
1395 Description: Update to SLES9 SP4 kernel-2.6.5-7.308.
1398 Frequency : rare on shutdown OST
1400 Description: If llog cancel was not send before clean_exports phase, this can
1401 produce deadlock in llog code.
1402 Details : If llog thread has last reference to obd and call class_import_put
1403 this produce deadlock because llog_cleanup_commit_master wait when
1404 last llog_commit_thread exited, but this never success because was
1405 called from llog_commit_thread.
1408 Frequency : only if OST index is skipped
1410 Description: NULL lov_tgts causing MDS oops
1411 Details : more safe checks for NULL lov_tgts for avoid oops.
1413 Severity : enhancement
1415 Description: Update to RHEL4 latest kernel-2.6.9-67.0.1.EL.
1417 Severity : enhancement
1419 Description: Update to RHEL5 latest kernel-2.6.18-53.1.4.el5.
1424 Description: make mgs_setparam() handle fsname containing dash
1425 Details : fsname containing a dash does not work with lctl conf_param
1427 Severity : enhancement
1429 Description: Update to RHEL4 Update-6 kernel-2.6.9-67.EL.
1432 Frequency : rare, in recovery and (or) destroy lovobjid file.
1434 Description: rewrite lov objid code.
1435 Details : Cleanup for lov objid code, remove scability problems and wrong
1436 locking. Fix sending last_id into ost.
1438 Severity : enhancement
1440 Description: Update to SLES10 SP1 latest kernel-2.6.16.54-0.2.3.
1442 Severity : enhancement
1444 Description: Update to RHEL5 Update-1 kernel 2.6.18-53.el5.
1445 Details : Use d_move_locked instead of __d_move.
1448 Frequency : rare, at shutdown
1450 Description: access already free / zero obd_namespace.
1451 Details : if client_disconnect_export was called without force flag set,
1452 and exist connect request in flight, this can produce access to
1453 NULL pointer (or already free pointer) when connect_interpret
1454 store ocd flags in obd_namespace.
1457 Frequency : only at startup
1459 Description: not alloc memory with spinlock held.
1460 Details : allocation memory with GFP_KERNEL can produce sleep deadlock,
1461 if any spinlock held.
1466 Description: lfs find does not continue on file error
1467 Details : Continue other files processing when a file/dir is absent.
1471 Description: Inconsistent usage of lustre_pack_reply()
1472 Details : Standardize the usage of lustre_pack_reply() such that it
1473 always generate a CERROR on failure.
1476 Frequency : very rare
1478 Description: Fix replay if there is an un-replied request and open
1479 Details : In some cases, older replay request will revert the
1480 mcd->mcd_last_xid on MDS which is used to record the client's
1481 latest sent request.
1483 Severity : enhancement
1485 Description: Update to RHEL5 kernel 2.6.18-8.1.15.el5.
1487 Severity : enhancement
1489 Description: Update to SLES10 SP1 kernel 2.6.16.53-0.16
1491 Severity : enhancement
1493 Description: Update to SLES9 kernel-2.6.5-7.287.3.
1495 Severity : enhancement
1497 Description: Update to RHEL4 kernel-2.6.9-55.0.12.EL.
1499 Severity : enhancement
1501 Description: Build SLES10 patchless client fails
1502 Details : The configure was broken by run ./configure with
1503 --with-linux-obj=.... argument for patchless client. When the
1504 configure use --with-linux-obj, the LINUXINCLUDE= -Iinclude
1505 can't search header adequately. Use absolute path such as
1506 -I($LINUX)/include instead.
1508 Severity : enhancement
1510 Description: Lustre Page Accounting
1511 Details : New macros for page alloc and free which enable accounting
1512 of page allocation of Lustre. Use percpu counters to store memory
1513 and page statistics.
1516 Frequency : only if debugging is disabled
1518 Description: LASSERT_{REQ,REP}SWAB macros are buggy
1519 Details : If SWAB_PARANOIA is disabled, the LASSERT_REQSWAB and
1520 LASSERT_REPSWAB macros become no-ops, which is incorrect. Drop
1521 these macros and replace them with their definitions instead.
1526 Description: interrupt oig_wait produce painc on resend.
1527 Details : brw_redo_request can be used for resend requests from ptlrpcd and
1528 private set, and this produce situation when rq_ptlrpcd_data not
1529 copyed to new allocated request and triggered LBUG on assert
1530 req->rq_ptlrpcd_data != NULL. But this member used only for wakeup
1531 ptlrpcd set if request is changed and can be safety changed to use
1534 Severity : enhancement
1536 Description: organize the server-side client stats on per-nid basis
1537 Details : Change the structure of stats under obdfilter and mds to
1545 The "uuid"s file would list the uuids of _active_ exports.
1546 And the clear entry is to clear all stats and stale nids.
1551 Description: Processes looping in ll_readdir() on Lustre clients finally causing
1552 a full node pseudo-hang
1553 Details : Concurrent access to the same directory from multiple clients with
1554 intensive file creation/removal can cause a client node to spin in
1555 ll_readdir(). i_version must be increased every time the lock is
1556 cancelled to ensure a revalidate is done.
1561 Description: touch file failed when fs is not full
1562 Details : OST in recovery should not be discarded by MDS in alloc_qos(),
1563 otherwise we can get ENOSP while fs is not full.
1566 Frequency : only for Cray XT3
1567 Bugzilla : 12829/13455
1568 Description: Changing primary group doesn't change the group lustre assigns to
1570 Details : When CRAY_XT3 is defined, the fsgid supplied by the client is
1571 overridden with the primary group provided by the group upcall,
1572 whereas the supplied fsgid can be trusted if it is in the list of
1573 supplementary groups returned by the group upcall.
1575 Severity : enhancement
1577 Description: Root Squash Functionality
1578 Details : Implementation of NFS-like root squash capability. Specifically,
1579 don't allow someone with root access on a client node to be able
1580 to manipulate files owned by root on a server node.
1582 Severity : enhancement
1584 Description: Slow trucate/writes to huge files at high offsets.
1585 Details : Directly associate cached pages to lock that protect those pages,
1586 this allows us to quickly find what pages to write and remove
1587 once lock callback is received.
1592 Description: Too many locks accumulating on client during NFS usage
1593 Details : mds_open improperly used accmode to find out access mode to a
1594 file. Also mdc_intent_lock logic to find out if we already have
1595 lock similar to just received was flawed since introduction of
1596 skiplists - locks are now added to the front of the granted
1601 Description: Hit ASSERTION(obd->obd_stopping == 1) failed in some setup failed
1603 Details : In obd setup failure handler, obd_stopping will not necessarily to
1604 be 1, and obd_set_up should also be checked to make sure whether
1605 obd is completely setup.
1607 Severity : enhancement
1609 Description: Allow masking D_WARNING, D_ERROR messages from console
1610 Details : Console messages can now be disabled via lnet.printk.
1615 Description: User code with malformed file open parameter crashes client node
1616 Details : Before packing join_file req, all the related reference should be
1617 checked carefully in case some malformed flags cause fake join_file
1623 Description: shrink/enlarge qunit size when needed; fix the problem of coarse
1624 grain of quota doing harm to quota's accuracy
1625 Details : qunit size will be changed when quota limitation is too low/high;
1626 record the pending quota write in order to get more accureate
1627 quota; delete the patch for bug12588, which is unnecessary when
1628 this patch is landed. This bug also contains fixes for bug 14526,
1629 14299, 14601 and 13794.
1633 Description: LDLM_ENQUEUE races with LDLM_CP_CALLBACK
1634 Details : ldlm_completion_ast() assumes that a lock is granted when the req
1635 mode is equal to the granted mode. However, it should also check
1636 that LDLM_FL_CP_REQD is not set.
1640 Description: Heavy nfs access might result in deadlocks
1641 Details : After ELC code landed, it is now improper to enqueue any mds
1642 locks under och_sem, because enqueue might want to decide to
1643 cancel open locks for same inode we are holding och_sem for.
1647 Description: 35% write performance drop with ldiskfs2 when quotas are on
1648 Details : Enable ext3 journalled quota by default to improve performance
1649 when quotas are turned on.
1653 Description: Client eviction while running blogbench
1654 Details : A lot of unlink operations with concurrent I/O can lead to a
1655 deadlock causing evictions. To address the problem, the number of
1656 oustanding OST_DESTROY requests is now throttled to
1657 max_rpcs_in_flight per OSC and LDLM_FL_DISCARD_DATA blocking
1658 callbacks are processed in priority.
1661 Frequency : RHEL4 only
1663 Description: mkfs is very slow on IA64/RHEL4
1664 Details : A performance regression has been discovered in the MPT Fusion
1665 driver between versions 3.02.73rh and 3.02.99.00rh. As a
1666 consequence, we have downgraded the MPT Fusion driver in the RHEL4
1667 kernel from 3.02.99.00 to 3.02.73 until this problem is fixed.
1670 Frequency : PPC/PPC64 only
1672 Description: conflicts between asm-ppc64/types.h and lustre_types.h
1673 Details : fix duplicated definitions between asm-ppc64/types.h and
1674 lustre_types.h on PPC.
1677 Frequency : PPC/PPC64 only
1679 Description: asm-ppc/segment.h does not exist
1680 Details : fix compile issue on PPC.
1684 Description: data checksumming impacts single node performance
1685 Details : add support for several checksum algorithms. Currently, CRC32 and
1686 Adler-32 are supported. The checksum type can be changed on the fly
1687 through /proc/fs/lustre/osc/*/checksum_type.
1691 Description: use adler32 for page checksums
1692 Details : when available, use the Adler-32 algorithm instead of CRC32 for
1697 Description: better handle error messages in extents code
1699 Severity : enhancement
1701 Description: SNMP support enhancement
1702 Details : Adding total number of sampled request for an MDS node in snmp
1705 Severity : enhancement
1707 Description: Optimize ldlm waiting list processing for PR extent locks
1708 Details : When processing waiting list for read extent lock and meeting read
1709 lock that is same or wider to it that is not contended, skip
1710 processing rest of the list and immediatelly return current
1711 status of conflictness, since we are guaranteed there are no
1712 conflicting locks in the rest of the list.
1716 Description: Time out and refuse to reconnect
1717 Details : When the failover node is the primary node, it is possible
1718 to have two identical connections in imp_conn_list. We must
1719 compare not conn's pointers but NIDs, otherwise we
1720 can defeat connection throttling.
1724 Description: Client not clear own cache if answer to reconnect is lost.
1725 Details : Client gets evicted from server. Now client also thinks it is
1726 disconnected (or gets ENOTCONN on its operation) and decides to
1727 reconnect. Server receives reconnect message, but cannot find
1728 export. New export is created that is fully valid (new cookie!),
1729 but reply is lost and not reported to client. Client reconnects
1730 again and gets back a just-created connection, but it is not new
1731 so client thinks it was not evicted and does not do recovery.
1735 Description: Detect stride IO mode in read-ahead
1736 Details : When a client does stride read, read-ahead should detect that and
1737 read-ahead pages according to the detected stride pattern.
1741 Description: build for x2 fails
1742 Details : fix compile issue on Cray systems.
1744 Severity : enhancement
1746 Description: implement readv/writev APIs(aio_read/aio_writes in newer kernels)
1747 Details : This greatly improves speed of NFS writes on 2.6 kernels.
1750 Frequency : only on PPC/SLES10
1752 Description: "BITS_PER_LONG is not 32 or 64" in linux/idr.h
1753 Details : On SLES10/PPC, fs.h includes idr.h which requires BITS_PER_LONG to
1754 be defined. Add a hack in mkfs_lustre.c to work around this compile
1759 Description: LASSERT on MDS when client holding flock lock dies
1760 Details : ldlm pool logic depends on number of granted locks equal to
1761 number of released locks which is not true for flock locks, so
1762 just exclude such locks from consideration.
1766 Description: MDS deadlock with many ll_sync_lov threads and I/O stalled
1767 Details : Use fsfilt_sync() for both the whole filesystem sync and
1768 individual file sync to eliminate dangerous inode locking
1769 with I_LOCK that can lead to a deadlock.
1773 Description: Update an obsolete wirecheck.c generator
1774 Details : Update wirecheck.c/wirehdr.c and regenerate wiretest.c
1778 Description: Client can panic on open sometimes
1779 Details : It is possible that we try to free already freed request in
1780 ll_file_open in some error cases when we send request from
1785 Description: performance in 1.6.3
1786 Details : Force q->max_phys_segments to MAX_PHYS_SEGMENTS on SLES10 to be
1787 sure that 1MB requests are not fragmented by the block layer.
1791 Description: LDLM soft lockups - improvement
1792 Details : It is be possible to send the lock handle along with each read
1793 or write request because the client is already doing a lock match
1794 itself so there isn't any reason the OST should have to re-do that
1800 Description: lfs quota fails with deactivated OSTS
1801 Details : With this patch, three improvements are included:
1802 1. detete the softlimit in mds and osts when use "lfs quota".
1803 2. display the inaccurate data in the output of "lfs quota".
1804 3. try to get quota info when "lfs quota" is executed.
1809 Description: Extent locks not granted with no conflicts sometimes.
1810 Details : When race occurs in glimpse handler and nothing is returned,
1811 we do not reprocess the queue after lock cancel, and that leads
1812 to a stall until next activity on a resource
1815 Frequency : failover with quotaon
1817 Description: during mds failovers with quota on, OSTs got into deadlock state
1818 and causing dumpstack.
1819 Details : for every quota slave, at any time, there is only one quota req
1820 is sent to quota master for every uid/gid. Before that quota req
1821 returns, all the thread relative to the same uid/gid will wait.
1822 So if the quota req is lost because mds failovers or any other
1823 reasons, this bug will be hit. Now, dqacq_interpret() will handle
1824 quota reqs who time out.
1826 Severity : enhancement
1829 Description: when quota slave checks if quota is enough, there is an unnecessary
1831 Details : place this wait on necessary place instead of always waiting.
1833 --------------------------------------------------------------------------------
1835 2007-12-07 Cluster File Systems, Inc. <info@clusterfs.com>
1837 * Support for kernels:
1838 2.6.5-7.286 (SLES 9),
1839 2.6.9-55.0.9.EL (RHEL 4),
1840 2.6.16.53-0.8 (SLES 10),
1841 2.6.18-8.1.14.el5 (RHEL 5),
1842 2.6.18.8 vanilla (kernel.org)
1843 * Client support for unpatched kernels:
1844 (see http://wiki.lustre.org/index.php?title=Patchless_Client)
1845 2.6.16 - 2.6.22 vanilla (kernel.org)
1846 * Due to recently discovered recovery problems, we do not recommend
1847 using patchless RHEL 4 clients with this or any earlier release.
1848 * Recommended e2fsprogs version: 1.40.2-cfs1
1849 * Note that reiserfs quotas are disabled on SLES 10 in this kernel.
1852 Frequency : occasional
1854 Description: MDS hang or stay in waiting lock
1855 Details : If client receive lock with CBPENDING flag ldlm need send lock
1856 cancel as separate rpc, to avoid situation when cancel request
1857 can't processed due all i/o threads stay in wait lock.
1860 Frequency : occasional
1862 Description: Do not fail import if osc_interpret_create gets -EAGAIN
1863 Details : If osc_interpret_create got -EAGAIN it immediately exits and
1864 wakeup oscc_waitq. After wakeup oscc_wait_for_objects call
1865 oscc_has_objects and see OSC has no objests and call
1866 oscc_internal_create to resend create request.
1868 Severity : enhancement
1870 Description: Update kernel patches for SLES10 2.6.16.53-0.8.
1871 Details : Update which_patch & target file for SLES10 latest kernel.
1873 Severity : enhancement
1875 Description: add --type and --size parameters to lfs find
1876 Details : Enhance lfs find by adding filetype and filesize parameters. Also
1877 multiple OBDs can now be specified for the --obd option.
1879 Severity : enhancement
1881 Description: eliminate client locks in face of contention
1882 Details : file contention detection and lockless i/o implementation
1883 for contended files.
1885 Severity : enhancement
1887 Description: Remove client patches from SLES 10 kernel.
1888 Details : This causes SLES 10 clients to behave as patchless clients
1889 even on a Lustre-patched (server) kernel.
1891 Severity : enhancement
1893 Description: use i_size_read and i_size_write in 2.6 port
1894 Details : replace inode->i_size access with i_size_read/write()
1897 Frequency : when removing large files
1899 Description: scheduling issue during removal of large Lustre files
1900 Details : Don't take the BKL in fsfilt_ext3_setattr() for 2.6 kernels.
1901 It causes scheduling issues when removing large files (17TB in the
1907 Description: 1.4.11 Can't handle directories with stripe set and extended ACLs
1908 Details : Impossible (EPROTO is returned) to access a directory that has a
1909 non-default striping and ACLs.
1912 Frequency : only on ppc
1914 Description: /proc/fs/lustre/devices broken on ppc
1915 Details : The patch as applied to 1.6.2 doesn't look correct for all arches.
1916 We should make sure the type of 'index' is loff_t and then cast
1917 explicitly as needed below. Do not assign an explicitly cast
1921 Frequency : only for rhel5
1923 Description: Kernel patches update for RHEL5 2.6.18-8.1.10.el5.
1924 Details : Modify the target file & which_kernel.
1927 Frequency : if the uninit_groups feature is enabled on ldiskfs
1929 Description: e2fsck reports "invalid unused inodes count"
1930 Details : If a new ldiskfs filesystem is created with the "uninit_groups"
1931 feature and only a single inode is created in a group then the
1932 "bg_unused_inodes" count is incorrectly updated. Creating a
1933 second inode in that group would update it correctly.
1938 Description: buffer overruns could theoretically occur
1939 Details : llapi_semantic_traverse() modifies the "path" argument by
1940 appending values to the end of the origin string, and a buffer
1941 overrun may occur. Adding buffer overrun check in liblustreapi.
1943 Severity : enhancement
1945 Description: Add jbd statistics patch for RHEL5 and 2.6.18-vanilla.
1948 Frequency : only if filesystem is inconsistent
1950 Description: handle "serious error: objid * already exists" more gracefully
1951 Details : If LAST_ID value on disk is smaller than the objects existing in
1952 the O/0/d* directories, it indicates disk corruption and causes an
1953 LBUG(). If the object is 0-length, then we should use the existing
1954 object. This will help to avoid a full fsck in most cases.
1956 Severity : enhancement
1958 Description: Kernel patches update for RHEL4 2.6.9-55.0.6.
1959 Details : Modify vm-tunables-rhel4.patch.
1961 Severity : enhancement
1963 Description: Kernel config for 2.6.18-vanilla.
1964 Details : Modify targets/2.6-vanilla.target.in.
1965 Add config file kernel-2.6.18-2.6-vanilla-i686.config.
1966 Add config file kernel-2.6.18-2.6-vanilla-i686-smp.config.
1967 Add config file kernel-2.6.18-2.6-vanilla-x86_64.config.
1968 Add config file kernel-2.6.18-2.6-vanilla-x86_64-smp.config.
1971 Frequency : occasional
1973 Description: improve handling recoverable errors
1974 Details : If request processed with error which can be recoverable on server
1975 request should be resend, otherwise page released from cache and
1981 Description: Kernel patches update for RHEL5 2.6.18-8.1.14.el5.
1982 Details : Modify target file & which_patch.
1983 A flaw was found in the IA32 system call emulation provided
1984 on AMD64 and Intel 64 platforms. An improperly validated 64-bit
1985 value could be stored in the %RAX register, which could trigger an
1986 out-of-bounds system call table access. An untrusted local user
1987 could exploit this flaw to run code in the kernel
1988 (ie a root privilege escalation). (CVE-2007-4573).
1992 Description: change order of libsysio includes
1993 Details : '#include sysio.h' should always come before '#include xtio.h'
1995 Severity : enhancement
1997 Description: adapt the lustre_config script to support the upgrade case
1998 Details : Add "-u" option for lustre_config script to support upgrading 1.4
1999 server targets to 1.6 in parallel.
2004 Description: To avoid grant space > avaible space when the disk is almost
2005 full. Without this patch you might see the error "grant XXXX >
2006 available" or some LBUG about grant, when the disk is almost
2008 Details : In filter_check_grant, for non_grant cache write, we should
2009 check the left space by if (*left > ungranted + bytes), instead
2010 of (*left > ungranted), because only we are sure the left space
2011 is enough for another "bytes", then the ungrant space should be
2012 increase. In client, we should update cl_avail_grant only there is
2013 OBD_MD_FLGRANT in the reply.
2018 Description: Update RHEL 4 kernel to fix local root privilege escalation.
2019 Details : Update to the latest RHEL 4 kernel to fix the vulnerability
2020 described in CVE-2007-4573. This problem could allow untrusted
2021 local users to gain root access.
2024 Frequency : when using O_DIRECT and quotas
2026 Description: Incorrect file ownership on O_DIRECT output files
2027 Details : block usage reported by 'lfs quota' does not take into account
2028 files that have been written with O_DIRECT.
2032 Description: (rw.c:1323:ll_read_ahead_pages()) ASSERTION(page_idx > ria->ria_stoff) failed
2033 Details : Once the unmatched stride IO mode is detected, shrink the stride-ahead
2034 window to 0. If it does hit cache miss, and read-pattern is still
2035 stride-io mode, does not reset the stride window, but also does not
2036 increase the stride window length in this case.
2038 --------------------------------------------------------------------------------
2040 2007-09-27 Cluster File Systems, Inc. <info@clusterfs.com>
2042 * Support for kernels:
2043 2.6.5-7.286 (SLES 9),
2044 2.6.9-55.0.2.EL (RHEL 4),
2045 2.6.16.46-0.14 (SLES 10),
2046 2.6.18-8.1.8.el5 (RHEL 5),
2047 2.6.18.8 vanilla (kernel.org)
2048 * Client support for unpatched kernels:
2049 (see http://wiki.lustre.org/index.php?title=Patchless_Client)
2050 2.6.16 - 2.6.21 vanilla (kernel.org)
2051 * Due to recently discovered recovery problems, we do not recommend
2052 using patchless RHEL 4 clients with this or any earlier release.
2053 * Recommended e2fsprogs version: 1.40.2-cfs1
2054 * Note that reiserfs quotas are disabled on SLES 10 in this kernel.
2058 Description: Fix errors in lfs documentation
2059 Details : Fixes man pages
2061 Severity : enhancement
2063 Description: Adaptive timeouts
2064 Details : RPC timeouts adapt to changing server load and network
2065 conditions to reduce resend attempts and improve recovery time.
2067 Severity : enhancement
2069 Description: llapi_file_create() does not allow some changes
2070 Details : add llapi_file_open() that allows specifying the file creation
2071 mode and open flags, and also returns an open file handle.
2073 Severity : enhancement
2075 Description: df doesn't work properly if diskfs blocksize != 4K
2076 Details : Choose biggest blocksize of OST's as the LOV's blocksize.
2079 Frequency : always on directories with default striping set
2081 Description: lfs find on -1 stripe looping in lsm_lmm_verify_common()
2082 Details : Avoid lov_verify_lmm_common() on directory with -1 stripe count.
2084 Severity : enhancement
2086 Description: merge and cleanup kernel patches.
2087 Details : Remove mnt_lustre_list in vfs_intent-2.6-rhel4.patch.
2089 Severity : enhancement
2091 Description: RedHat Update kernel for RHEL5
2092 Details : Kernel config file for RHEL5.
2094 Severity : enhancement
2096 Description: OSS needs mutliple precreate threads
2097 Details : Add ability to start more than one create thread per OSS.
2100 Frequency : only with quota on the root user
2102 Description: mds_obd_create error creating tmp object
2103 Details : When the user sets quota on root, llog will be affected and can't
2104 create files and write files.
2107 Frequency : Always on ia64 patchless client, and possibly others.
2109 Description: Add EXPORT_SYMBOL check for node_to_cpumask symbol.
2110 Details : This allows the patchless client to be loaded on architectures
2111 without this export.
2116 Description: disorder of journal start and llog_add cause deadlock.
2117 Details : in llog_origin_connect, journal start should happen before
2118 llog_add keep the same order as other functions to avoid
2121 Severity : enhancement
2123 Description: RedHat Update kernel for RHEL5
2124 Details : Modify the kernel config file more closer RHEL5.
2126 Severity : enhancement
2128 Description: Build failure against Centos5 (RHEL5)
2129 Details : Define PAGE_SIZE when it isn't present.
2132 Frequency : occasionally when using NFS
2134 Description: "ll_intent_file_open()) lock enqueue: err: -13" with nfs
2135 Details : with NFS, the anon dentry's parent was set to itself in
2136 d_alloc_anon(), so in MDS, we use rec->ur_fid1 to find the
2137 corresponding dentry other than use rec->ur_name.
2139 Severity : enhancement
2141 Description: enable data checksumming by default
2142 Details : enable checksum by default, allow --disable-checksum
2143 configure option and "-o nochecksum" mount option. Checksums
2144 can also be disabled at runtime via $LPROC/osc/*/checksum_pages.
2147 Frequency : Occasionally with failover
2149 Description: Client eviction due to failover config
2150 Details : after a connection loss, the lustre client should attempt to
2151 reconnect to the last active server first before trying the
2152 other potential connections.
2154 Severity : enhancement
2156 Description: client-side metadata stat-ahead during readdir(directory readahead)
2157 Details : perform client-side metadata stat-ahead when the client detects
2158 readdir and sequential stat of dir entries therein
2163 Description: when mds and osts use different quota unit(32bit and 64bit),
2164 quota will be released repeatly.
2165 Details : void sending multiple quota reqs to mds, which will keep the
2166 status between the reqs.
2169 Frequency : only with liblustre clients on XT3
2171 Description: evictions taking too long
2172 Details : allow llrd to evict clients directly on OSTs
2176 Description: osts not allocated evenly to files
2177 Details : change the condition to increase offset_idx
2179 Severity : enhancement
2181 Description: self-adjustable client's lru lists
2182 Details : use adaptive algorithm for managing client cached locks lru
2183 lists according to current server load, other client's work
2184 pattern, memory activities, etc. Both, server and client
2185 side namespaces provide number of proc tunables for controlling
2190 Description: rewrite ext2-derived code in llite/dir.c and obdclass/uuid.c
2191 Details : rewrite inherited code (uuid parsing code from ext2 utils and
2192 readdir code from ext3) from scratch preserving functionality.
2196 Description: Only those disconnect error should be returned by rq_status.
2197 Details : In open/enqueue processs, Some errors, which will cause client
2198 disconnected, should be returned by rq_status, while other
2199 errors should still be returned by intent, then mdc or llite will
2202 Severity : enhancement
2204 Description: Tune the kernel for good SCSI performance.
2205 Details : Set the value of /sys/block/{dev}/queue/max_sectors_kb
2206 to the value of /sys/block/{dev}/queue/max_hw_sectors_kb
2210 Frequency : Always for filesystems larger than 2TB on 32-bit systems.
2211 Bugzilla : 13547, 13627
2212 Description: Data corruption for OSTs that are formatted larger than 2TB
2214 Details : When generating the bio request for lustre file writes the
2215 sector number would overflow a temporary variable before being
2216 used for the IO. The data reads correctly from Lustre (which
2217 will overflow in a similar manner) but other file data or
2218 filesystem metadata may be corrupted in some cases.
2222 Description: "lfs find -obd UUID" prints directories
2223 Details : "lfs find -obd UUID" will return all directory names instead
2224 of just file names. It is incorrect because the directories
2225 do not reside on the OSTs.
2228 Bugzilla : 13596, 13614
2229 Description: MDS hang after unclean shutdown of lots of clients
2230 Details : Never resend AST requests.
2234 Frequency : Always, for kernels after 2.6.16
2235 Description: Fix warning idr_remove called for id=.. which is not allocated.
2236 Details : Last kernels save old s_dev before kill super and not allow
2237 to restore from callback - restore it before call kill_anon_super.
2241 Description: Fix error on 'ls .' at the top of the Lustre mount.
2242 Details : Don't revalidate dentry if it is a root dentry.
2244 --------------------------------------------------------------------------------
2246 2007-08-27 Cluster File Systems, Inc. <info@clusterfs.com>
2248 * Support for kernels:
2249 2.6.5-7.286 (SLES 9),
2250 2.6.9-55.0.2.EL (RHEL 4),
2251 2.6.16.46-0.14 (SLES 10),
2252 2.6.18-8.1.8.el5 (RHEL 5),
2253 2.6.18.8 vanilla (kernel.org)
2254 * Client support for unpatched kernels:
2255 (see http://wiki.lustre.org/index.php?title=Patchless_Client)
2256 2.6.16 - 2.6.22 vanilla (kernel.org)
2257 * Due to recently discovered recovery problems, we do not recommend
2258 using patchless RHEL 4 clients with this or any earlier release.
2259 * Recommended e2fsprogs version: 1.39.cfs8
2260 * Note that reiserfs quotas are disabled on SLES 10 in this kernel.
2265 Description: block reactivating mgc import until all deactivates complete
2266 Details : Fix race when failing back MDT/MGS to itself (testing)
2268 Severity : enhancement
2270 Description: lfs setstripe enhancement
2271 Details : Make lfs setstripe understand 'k', 'm' and 'g' for stripe size.
2274 Frequency : mds/oss recovery
2276 Description: llog ctxt is refrenced after it has been freed.
2277 Details : llog ctxt refcount was added to avoide the race between ctxt free
2278 and llog recovery process. Each llog user must hold ctxt refcount
2279 before it access the llog. And the llog ctxt can only be freed
2280 when its refcount is zero.
2282 Severity : enhancement
2284 Description: randomly memory allocation failure util
2285 Details : Make lustre randomly failed allocating memory for testing purpose.
2288 Frequency : only for SLES10
2290 Description: Update kernel patch for SLES10 SP1
2291 Details : Add patch blkdev_tunables-2.6-sles10.patch to 2.6-sles10.series.
2293 Severity : enhancement
2295 Description: omit set fsid for export NFS
2296 Details : fix set/restore device id for avoid EMFILE error and mark lustre fs
2297 as FS_REQUIRES_DEV for avoid problems with generate fsid.
2300 Frequency : after network failures
2302 Description: Add sync option to mount_lustre.c
2303 Details : Client loses data written to lustre after a network interruption.
2305 Severity : enhancement
2307 Description: Error message improvement.
2308 Details : Merging of two LCONSOLE_ERROR_MSG into one.
2310 Severity : enhancement
2312 Description: don't use GFP_* in generic Lustre code.
2313 Details : Use cfs_alloc_* functions and CFS_* flags for code portability.
2315 Severity : enhancement
2317 Description: obdclass is limited by single OBD_ALLOC(idarray)
2318 Details : replace OBD_ALLOC/OBD_FREE with OBD_VMALLOC/OBD_VFREE
2320 Severity : enhancement
2322 Description: updated patchess for new RHEL4 kernel
2323 Details : Fixed ext3-unlink-race.patch per Kalpak's comment.
2325 Severity : enhancement
2327 Description: warnings with build patchless client with vanila 2.6.19 and up
2328 Details : change old ctl_table style and replace ctl_table/ctl_table_header
2329 with cfs_sysctl_table_t/cfs_sysctl_table_header_t
2331 Severity : enhancement
2333 Description: O_DIRECT bypasses client statistics.
2334 Details : When running with O_DIRECT I/O, neither the client rpc_stats nor
2335 read_ahead_stats were updated. Copied stats section from
2336 osc_send_oap_rpc() into async_internal().
2339 Frequency : only for Cray XT3
2341 Description: peer credits not enough on many OST per OSS systems.
2342 Details : Use new lnet way to add credits as we need those for pings and ASTs
2345 Frequency : only with liblustre
2347 Description: Liblustre is not releasing flock locks on file close.
2348 Details : Release flock locks on file close.
2351 Frequency : only for RHEL4
2353 Description: Update kernel patches for kernel-2.6.9-55.0.2.EL
2354 Details : Remove inode-nr_unused-2.6.9-rhel4.patch from 2.6-rhel4.series
2355 Update target file and kernel config.
2359 Description: lustre support for RHEL5
2360 Details : Add support for RHEL5.
2365 Description: ASSERTION(export != NULL) failed in target_handle_connect
2366 Details : Assetion hit is result of rare race between disconnect and connet
2367 to same nid. target_handle_connect found old connect cockie and
2368 tried to reconnect, but can't find export for this cockie.
2373 Description: umount blocks forever on error
2374 Details : In result of wrong using obd_no_recov and obd_force flags client
2375 can hand if cancel or some other requests is lost.
2378 Frequency : Only for SLES
2380 Description: sanity_quota fail test_1
2381 Details : There are multiple occurences of $TSTUSR in SLES's /etc/group
2382 file, which makes TSTID[2] inunique.
2384 Severity : enhancement
2386 Description: Kernel patches for SLES9 2.6.5-7.286 kernel
2387 Details : Update target/ChangeLog/which_patch .
2389 Severity : enhancement
2391 Description: jbd statistics
2392 Details : Port older jbd statistics patch for sles10
2397 Description: lvbo_init failed for resource with missing objects.
2398 Details : Fix returning error if we do stat for file with missing/corrupted
2399 objects and i_size set to all sum of size of all avaible objects.
2400 if we to truncate/write to missing object - it is recreated.
2405 Description: Oops in read and write path when failing to allocate lock.
2406 Details : Check if lock allocation failed and return error back.
2409 Frequency : When flocks are used.
2411 Description: assertion failure in ldlm_cli_enquque_fini for non NULL lock.
2412 Details : Flock locks might destroy just granted lock if it could be merged
2413 with another existing flock, this is done in completion handler,
2414 so teach ldlm_cli_enquque_fini that this is a valid case for
2420 Description: reply_lock_interpret crash due to race with it and lock cancel.
2421 Details : Do not replay locks that are being cancelled. Do not reference
2422 locks by their address during replay, just by their handle.
2425 Frequency : only with deactivated OSTs
2427 Description: lstripe command fails for valid OST index
2428 Details : The stripe offset is compared to 'lov->desc.ld_tgt_count'
2429 instead of lov->desc.ld_active_tgt_count.
2431 Severity : enhancement
2433 Description: Build failure against Centos5 (RHEL5)
2434 Details : Use getpagesize() instead of PAGE_SIZE.
2436 --------------------------------------------------------------------------------
2438 2007-07-30 Cluster File Systems, Inc. <info@clusterfs.com>
2440 * Support for kernels:
2441 2.6.5-7.283 (SLES 9),
2442 2.6.9-55.EL (RHEL 4),
2443 2.6.16.46-0.14 (SLES 10),
2444 2.6.18.8 vanilla (kernel.org)
2445 * Client support for unpatched kernels:
2446 (see http://wiki.lustre.org/index.php?title=Patchless_Client)
2447 2.6.16 - 2.6.22 vanilla (kernel.org)
2448 * Due to recently discovered recovery problems, we do not recommend
2449 using patchless RHEL 4 clients with this or any earlier release.
2450 * Recommended e2fsprogs version: 1.39.cfs8
2451 * Note that reiserfs quotas are disabled on SLES 10 in this kernel.
2452 * Starting with this release, the ldiskfs backing filesystem required
2453 by Lustre is now in its own package, lustre-ldiskfs. This package
2454 should be installed. It is versioned separately from Lustre and
2455 may be released separately in future.
2460 Description: server LBUG when shutting down
2461 Details : Block umount forever until the mount refcount is zero rather
2462 than giving up after an arbitrary timeout.
2464 Severity : enhancement
2466 Description: add optional extra BUILD_VERSION info
2467 Details : add a new environment variable (namely LUSTRE_VERS) which allows
2468 to override the lustre version.
2471 Frequency : 2.6.18 servers only
2473 Description: ll_kern_mount() doesn't release the module reference
2474 Details : The ldiskfs module reference count never drops down to 0
2475 because ll_kern_mount() doesn't release the module reference.
2480 Description: server LBUG when using old ost_num_threads parameter
2481 Details : Accept the old ost_num_threads parameter but warn that it
2482 is deprecated, and fix an off-by-one error that caused an LBUG.
2487 Description: Transient SCSI error results in persistent IO issue
2488 Details : iobuf->dr_error is not reinitialized to 0 between two
2492 Frequency : sometimes when underlying device returns I/O errors
2494 Description: OSTs not going read-only during write failures
2495 Details : OSTs are not remounted read-only when the journal commit threads
2496 get I/O errors because fsfilt_ext3 calls journal_start/stop()
2497 instead of the ext3 wrappers.
2500 Frequency : only with 10000 clients or more
2502 Description: poor connect scaling with increasing client count
2503 Details : Don't run filter_grant_sanity_check for more than 100 exports
2504 to improve scaling for large numbers of clients.
2507 Frequency : SLES10 only
2509 Description: sanity-quota.sh quotacheck failed: rc = -22
2510 Details : Quotas cannot be enabled on SLES10.
2513 Frequency : liblustre clients only
2515 Description: getdirentries does not give error when run on compute nodes
2516 Details : getdirentries does not fail when the size specified as an argument
2517 is too small to contain at least one entry
2519 Severity : enhancement
2521 Description: Add LNET router traceability for debug purposes
2522 Details : If a checksum failure occurs with a router as part of the
2523 IO path, the NID of the last router that forwarded the bulk data
2524 is printed so it can be identified.
2529 Description: OST "spontaneously" evicts client; client has imp_pingable == 0
2530 Details : Due to a race condition, liblustre clients were occasionally
2531 evicted incorrectly.
2533 Severity : enhancement
2535 Description: lfs setstripe use optional parameters instead of postional
2538 Severity : enhancement
2540 Description: Nanosecond timestamp support for ldiskfs
2541 Details : The on-disk ldiskfs filesystem has added support for nanosecond
2542 resolution timestamps. There is not yet support for this at
2543 the Lustre filesystem level.
2546 Frequency : during server recovery
2548 Description: MDS failing to send precreate requests due to OSCC_FLAG_RECOVERING
2549 Details : request with rq_no_resend flag not awake l_wait_event if they get a
2553 Frequency : nfs export on patchless client
2555 Description: connectathon hang when test nfs export over patchless client
2556 Details : Disconnected dentry cannot be found with lookup, so we do not need
2557 to unhash it or make it invalid
2560 Description: fix llapi_lov_get_uuids() to allow many OSTs to be returned
2561 Details: : Change llapi_lov_get_uuids() to read the UUIDs from /proc instead
2562 of using an ioctl. This allows lfsck for > 160 OSTs to succeed.
2567 Description: open req refcounting wrong on reconnect
2568 Details : If reconnect happened between getting open reply from server and
2569 call to mdc_set_replay_data in ll_file_open, we will schedule
2570 replay for unreferenced request that we are about to free.
2571 Subsequent close will crash in variety of ways.
2572 Check that request is still eligible for replay in
2573 mdc_set_replay_data().
2578 Description: disable writes to filesystem when reading health_check file
2579 Details : the default for reading the health_check proc file has changed
2580 to NOT do a journal transaction and write to disk, because this
2581 can cause reads of the /proc file to hang and block HA state
2582 checking on a healthy but otherwise heavily loaded system. It
2583 is possible to return to the previous behaviour during configure
2584 with --enable-health-write.
2586 Severity : enhancement
2588 Description: 64-bit inode version
2589 Details: : Add a on-disk 64-bit inode version for ext3 to track changes made
2590 to the inode. This will be required for version-based recovery.
2595 Description: MDS fails to start if a duplicate client export is detected
2596 Details : in some rare cases it was possible for a client to connect to
2597 an MDS multiple times. Upon recovery the MDS would detect this
2598 and fail during startup. Handle this more gracefully.
2600 Severity : enhancement
2602 Description: Add -o localflock option to simulate old noflock behaviour.
2603 Details : This will achieve local-only flock/fcntl locks coherentness.
2608 Description: log_commit_thread vs filter_destroy race leads to crash
2609 Details : Take import reference before releasing llog record semaphore
2614 Description: Wrong request locking in request set processing
2615 Details : ptlrpc_check_set wrongly uses req->rq_lock for proctect add to
2616 imp_delayed_list, in this place should be used imp_lock.
2619 Frequency : when reconnecting
2621 Description: Grant leak when OSC reconnect to OST
2622 Details : When osc reconnect ost, OST(filter) should check whether it
2623 should grant more space to client by comparing fed_grant and
2624 cl_avail_grant, and return the granted space to client instead
2625 of "new granted" space, because client will call osc_init_grant
2626 to update the client grant space info.
2629 Frequency : when client reconnects to OST
2631 Description: Grant leak when OSC does a resend and replays bulk write
2632 Details : When osc reconnect to OST, OST(filter) should clear grant info of
2633 bulk write request, because the grant info while be sync between
2634 OSC and OST when reconnect, and we should ignore the grant info
2635 these of resend/replay write req.
2640 Description: Grant space more than avaiable space sometimes.
2641 Details : When then OST is about to be full, if two bulk writing from
2642 different clients came to OST. Accord the avaliable space of the
2643 OST, the first req should be permitted, and the second one
2644 should be denied by ENOSPC. But if the seconde arrived before
2645 the first one is commited. The OST might wrongly permit second
2646 writing, which will cause grant space > avaiable space.
2649 Frequency : when client is evicted
2651 Description: Grant might be wrongly erased when osc is evicted by OST
2652 Details : when the import is evicted by server, it will fork another
2653 thread ptlrpc_invalidate_import_thread to invalidate the
2654 import, where the grant will be set to 0. While the original
2655 thread will update the grant it got when connecting. So if
2656 the former happened latter, the grant will be wrongly errased
2657 because of this race.
2662 Description: Checking Stale with correct fid
2663 Details : ll_revalidate_it should uses de_inode instead of op_data.fid2
2664 to check whether it is stale, because sometimes, we want the
2665 enqueue happened anyway, and op_data.fid2 will not be initialized.
2667 Severity : enhancement
2669 Description: update patchless client
2670 Details : Add support for patchless client with 2.6.20, 2.6.21 and RHEL 5
2673 Frequency : only with 2.4 kernel
2675 Description: random memory corruption
2676 Details : size of struct ll_inode_info is to big for union inode.u and this
2677 can be cause of random memory corruption.
2682 Description: Memory leak in recovery
2683 Details : Lov_mds_md was not free in an error handler in mds_create_object.
2684 It should also check obd_fail before fsfilt_start, otherwise if
2685 fsfilt_start return -EROFS,(failover mds during mds recovery).
2686 then the req will return with repmsg->transno = 0 and rc = EROFS.
2687 and we met hit the assert LASSERT(req->rq_reqmsg->transno ==
2688 req->rq_repmsg->transno) in ptlrpc_replay_interpret. Fcc should
2689 be freed no matter whether fsfilt_commit success or not.
2692 Frequency : only with large numbers of cores on a single node
2694 Description: Prevents from taking the superblock lock in llap_from_page for
2696 Details : using LL_ORIGIN_REMOVEPAGE origin flag instead of LL_ORIGIN_UNKNOW
2697 for llap_from_page call in ll_removepage() prevents from taking
2698 the superblock lock for a soon died page.
2703 Description: Not check open intent error before release open handle
2704 Details : in some rare cases, the open intent error is not checked before
2705 release open handle, which may cause
2706 ASSERTION(open_req->rq_transno != 0), because it tries to release
2707 the failed open handle.
2712 Description: Set cat log bitmap only after create log success.
2713 Details : in some rare cases, the cat log bitmap is set too early. and it
2714 should be set only after create log success.
2718 Description: Accessing a block bevice can re-enable I/O when Lustre is
2719 tearing down a device.
2720 Details : dev_clear_rdonly(bdev) must be called in kill_bdev() instead of
2725 Description: service threads may hog cpus when there are a lot of requests
2726 Details : Insert cond_resched to give other threads a chance to use some CPU
2731 Description: the cat log was not initialized in recovery
2732 Details : When mds(mgs) do recovery, the tgt_count might be zero, so the
2733 unlink log on mds will not be initialized until mds post
2734 recovery. And also in mds post recovery, the unlink log will
2735 initialization will be done asynchronausly, so there will be race
2736 between add unlink log and unlink log initialization.
2740 Description: brw_stats were being printed incorrectly
2741 Details : brw_stats were being printed as log2 but all of them were not
2742 recorded as log2. Also remove some code duplication arising from
2743 filter_tally_{read,write}.
2747 Frequency : rare, only in recovery.
2748 Description: ASSERTION(req->rq_type != LI_POISON) failed
2749 Details : imp_lock should be held while iterating over imp_sending_list for
2750 prevent destroy request after get timeout in ptlrpc_queue_wait.
2754 Description: replay-single.sh test 52 fails
2755 Details : A lock's skiplist need to be cleanup when it being unlinked
2756 from its resource list.
2761 Description: Short directio read returns full requested size rather than
2763 Details : Direct I/O operations should return actual amount of bytes
2764 transferred rather than requested size.
2766 Severity : enhancement
2768 Description: metadata RPC reduction (e.g. for rm performance)
2769 Details : decrease the amount of synchronous RPC between clients and servers
2770 by canceling conflicing lock before the operation on the client
2771 and packing thier handles into the main operation RPC to server.
2773 Severity : enhancement
2775 Description: add #ifdef HAVE_KERNEL_CONFIG_H
2776 Details : kernels from 2.6.19 not need include linux/config.h, but add
2777 include linux/autoconf.h in commpiler command line.
2779 Severity : enhancement
2781 Description: patchless client support for 2.6.22 kernel
2782 Details : 2.6.22 has only one visble change, SLAB_CTOR_* constants is
2783 removed. In this case we need drop using os depended interface to
2784 kmem_cache and use cfs_mem_cache API.
2789 Description: fix mal-formatted messages
2790 Details : fix some mal-formatted DEBUG_REQ and LCONSOLE_ERROR_MSG messages
2794 Frequency : always in liblustre
2795 Description: wrong IS_ERR implementation in liblustre.h
2796 Details : fix IS_ERR implementation in liblustre.h for right detect errors.
2801 Description: Correct condition for output debug message.
2802 Details : inode i_nlink equal zero is not enough for output message about
2803 disk corruption, i_ctime and i_mode should be also checked.
2807 Frequency : always in patchless client
2808 Description: add configure check for truncate_complete_page
2809 Details : improve checks for exported symbols. This allow run check without
2810 sources, but with Module.symvers shipped with kernel distribution.
2811 add check for truncate_complete_page used by patchless client.
2815 Description: sanity.sh test_77h fails with "test_77h file compare failed"
2816 Details : test_77h uses a file which was messed by other test case.
2820 Description: Not Check whether lov_tgts is NULL in some lov functions
2821 Details : Checking whether lov_tgts is NULL in some functions.
2825 Description: replace obdo_alloc() with OBDO_ALLOC macro
2826 Details : nothing special is done in obdo_alloc() function, and for
2827 debugging purpose, it needs to be replaced with macros.
2831 Description: bad return value and errno from fcntl call
2832 Details : In liblustre API, errno should be a negative value if error
2837 Description: ptlrpc_check_set() LBUG
2838 Details : In case of positive reply from server and failed client bulk
2839 callback after bulk transfer shouldn't LBUG, but process this
2840 request as erroneous.
2842 Severity : enhancement
2844 Description: VFS operations stats tool.
2845 Details : Tool which collects stats by tracking value written in pid,
2846 ppid, gid and uses llstat to generate output to plot graph using
2848 Updated lustre/utils/Makefile.am
2849 Added lustre/utils/ltrack_stats.c
2851 Severity : enhancement
2853 Description: 2.6.18 server support (lustre 1.6.1)
2854 Details : Support for 2.6.18 kernels on the server side.
2859 Description: ASSERTION(imp->imp_conn_current) failed
2860 Details : an assertion failure is hit if a client node boots and attempts to
2861 mount a lustre filesystem faster than RECONNECT_INTERVAL seconds.
2864 Frequency : only for i686
2866 Description: 1.4.11 RC1 build fails for RHEL 4, i686
2867 Details : Fixed config variable for build.
2872 Description: Updated patchess for new RHEL4 kernel
2873 Details : Updated patch inode-nr_unused-2.6.9-rhel4.patch
2874 Updated patch jbd-stats-2.6.9.patch
2875 Updated patch qsnet-rhel4-2.6.patch
2876 Updated patch quota-deadlock-on-pagelock-core.patch
2877 Updated patch vfs_intent-2.6-rhel4.patch
2878 Updated patch vfs_races-2.6-rhel4.patch
2879 Updated series file 2.6-rhel4-titech.series
2880 Updated series file 2.6-rhel4.series
2881 Updated kernel config files
2886 Description: lquota slave complains LBUG when reconnecting with mds
2888 Details : quota slave depends qctxt->lqc_import to send its quota request.
2889 This pointer will be invalid if mds did failover or broke its
2890 connect to osts, which leads to LBUG.
2893 Frequency : when qunit size is too small(less than 20M)
2895 Description: write is stopped by improper -EDQUOT
2896 Details : If the master is busy and qunit size is small enough(let's say 1M),
2897 the slave can not get quota from master on time, which will lead
2898 slave to trigger a -EQUOTA to client.
2903 Description: Deadlock during metadata tests
2904 Details : in prune_dir_dentries(), shrink_dcache_parent() should not be
2905 called with the per-dentry lock held.
2908 Frequency : SLES9 only
2910 Description: Lustre patched kernel for SLES9 SP3 has NR_CPUS set to 8
2911 Details : set CONFIG_NR_CPUS to 128 instead of 8.
2913 Severity : enhancement
2915 Description: remove fs_prep_san_write operation and related patches
2916 Details : remove the ext3-san-jdike patches which are no longer useful.
2921 Description: LDISKFS-fs error (device sdc): ldiskfs_free_blocks
2922 Details : a disk corruption can cause the mballoc code to assert on a
2923 double free or other extent corruptions. Handle these with
2924 ext3_error() instead of assertions.
2927 Frequency : only with mballoc3 code and deep extent trees
2929 Description: ldiskfs_ext_search_right: bad header in inode: unexpected eh_depth
2930 Details : a wrong check of extent headers in ldiskfs_ext_search_right()
2931 can cause the filesystem to be remounted read-only.
2936 Description: lfsck built against 1.4.x cannot run against 1.6.0 lustre
2937 Details : the definition for OBD_IOC_GETNAME changed in 1.6.0. One of the
2938 few external users of this ioctl number is lfsck's call to
2939 llapi_lov_get_uuids() and this caused lfsck to fail at startup.
2940 Add the old ioctl number to the handler so both old and new
2945 Description: parallel lock callbacks
2946 Details : Instead of sending blocking and completion callbacks as separated
2947 requests, adding them to a set and sending in parallel.
2951 Description: Disable most debugging by default
2952 Details : To improve performance, disable most logging (for debug purposes)
2953 by default. VFSTRACE, RPCTRACE, and DLMTRACE are now off by
2954 default, and HA includes fewer messages.
2958 Frequency : only run on patchless client.
2959 Description: use do_facet on sanity.sh for test handling recoverables errors
2960 Details : use do_facet instead of direct use sysctl for set fail_loc on OST
2964 Description: hash tables for lists of nids, connections and uuids
2965 Details : Hash tables noticeably help when a lot of clients connect to a
2966 server, to faster identify duplicate connections or reconnects,
2967 also to faster find export to evict in manual eviction case.
2971 Description: Sometimes, when the server evict a client, and the client will
2972 not be evicted as soon as possible.
2973 Details : In enqueue req, the error was returned by intent, instead of
2974 rq_status which make ptlrpc layer not detect this error, and
2975 does not evict the client. So enqueue error should be returned
2979 Frequency : only at startup
2981 Description: Delay client connections to MDT until fist MDT->OST connect
2982 Details : If a client tried to create a new file before the MDT had
2983 connected to any OSTs, the create would return EIO. Now
2984 the client will simply block until the MDT connects to the
2985 first OST and the create can succeed.
2988 Frequency : only for SLES9
2990 Description: Routinely utilize latest Quadrics drivers in CFS releases
2991 Details : Update patch qsnet-suse-2.6.patch.
2994 Frequency : only for sles10
2996 Description: Update patches for SLES 10 SP1 kernel.
2997 Details : Update the patch vfs_intent-2.6-sles10.patch.
3002 Description: Routinely utilize latest Quadrics drivers in CFS releases
3003 Details : Update patch qsnet-rhel4-2.6.patch.
3006 Frequency : at statup only
3008 Description: mds_lov_synchronize race leads to various problems
3009 Details : simultaneous MDT->OST connections at startup can cause the
3010 sync to abort, leaving the OSC in a bad state.
3015 Description: Using wrong pointer in osc_brw_prep_request
3016 Details : Access to array[-1] can produce panic if kernel compiled with
3017 CONFIG_PAGE_ALLOC enabled
3019 Severity : enhancement
3021 Description: Async OSC create to avoid the blocking unnecessarily.
3022 Details : If a OST has no remain object, system will block on the creating
3023 when need to create a new object on this OST. Now, ways use
3024 pre-created objects when available, instead of blocking on an
3025 empty osc while others are not empty. If we must block, we block
3026 for the shortest possible period of time.
3030 Frequency : only in recovery
3031 Description: Mark OST as early accessible if his start SYNC.
3032 Details : osc_precreate return flag early accessible if oscc marked as
3033 OSCC_FLAG_SYNC_IN_PROGRESS.
3038 Description: Sometimes precreate code can triger create object on wrong ost
3039 Details : Wrong protected or not not restored variables aftre precreate loop
3040 can produce creation object on wrong ost.
3043 Frequency : oss recovery
3045 Description: llog_commit_thread cleanup should sync with llog_commit_thread
3047 Details : llog_commit_thread_count should be synced between llog_commit
3048 start and cleanup, so new llog_commit thread should not be started
3049 when llog_commit threads being stopped to avoid accessing some
3052 Severity : enhancement
3054 Description: Add printing inode info into message about error in writepage.
3056 --------------------------------------------------------------------------------
3058 2007-05-03 Cluster File Systems, Inc. <info@clusterfs.com>
3063 Frequency : on some architectures
3065 Description: 1.6 client sometimes fails to mount from a 1.4 MDT
3066 Details : Uninitialized flags sometimes cause configuration commands to
3070 Frequency : patchless clients only
3072 Description: missing __iget() symbol export
3073 Details : The __iget() symbol export is missing. To avoid the need for
3074 this on patchless clients the deathrow inode reaper is turned
3075 off, and we depend on the VM to clean up old inodes. This
3076 dependency was during via the fix for bug 12181.
3081 Description: sanity.sh fail: test_52b
3082 Details : The ll_inode_to_ext_flags() has a glitch which makes MDS return
3083 incorrect inode's flags to client.
3085 --------------------------------------------------------------------------------
3087 2007-04-19 Cluster File Systems, Inc. <info@clusterfs.com>
3089 * CONFIGURATION CHANGE. This version of Lustre WILL NOT
3090 INTEROPERATE with older versions automatically. In many cases a
3091 special upgrade step is needed. Please read the
3092 user documentation before upgrading any part of a 1.4.x system.
3093 * WARNING: Lustre configuration and startup changes are required with
3094 this release. See https://mail.clusterfs.com/wikis/lustre/MountConf
3096 * Support for kernels:
3097 2.4.21-47.0.1.EL (RHEL 3),
3098 2.6.5-7.283 (SLES 9),
3099 2.6.9-42.0.10.EL (RHEL 4),
3100 2.6.12.6 vanilla (kernel.org),
3101 2.6.16.27-0.9 (SLES10)
3102 * Client support for unpatched kernels:
3103 (see https://mail.clusterfs.com/wikis/lustre/PatchlessClient)
3104 2.6.16 - 2.6.19 vanilla (kernel.org),
3105 2.6.9-42.0.8EL (RHEL 4)
3106 * Recommended e2fsprogs version: 1.39.cfs6
3107 * Note that reiserfs quotas are disabled on SLES 10 in this kernel
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.
3119 Severity : enhancement
3121 Description: MountConf
3122 Details : Lustre configuration is now managed via mkfs and mount
3123 commands instead of lmc and lconf. New obd types (MGS, MGC)
3124 are added for dynamic configuration management. See
3125 https://mail.clusterfs.com/wikis/lustre/MountConf for
3128 Severity : enhancement
3130 Description: dynamic OST addition
3131 Details : OSTs can now be added to a live filesystem
3133 Severity : enhancement
3135 Description: startup order invariance
3136 Details : MDTs and OSTs can be started in any order. Clients only
3137 require the MDT to complete startup.
3139 Severity : enhancement
3141 Description: parallel, asynchronous orphan cleanup
3142 Details : orphan cleanup is now performed in separate threads for each
3143 OST, allowing parallel non-blocking operation.
3145 Severity : enhancement
3147 Description: optimized stripe assignment
3148 Details : stripe assignments are now made based on ost space available,
3149 ost previous usage, and OSS previous usage, in order to try
3150 to optimize storage space and networking resources.
3152 Severity : enhancement
3154 Description: Permanently set tunables
3155 Details : All writable /proc/fs/lustre tunables can now be permanently
3156 set on a per-server basis, at mkfs time or on a live
3159 Severity : enhancement
3161 Description: Lustre message v2
3162 Details : Add lustre message format v2.
3164 Severity : enhancement
3166 Description: client OST exclusion list
3167 Details : Clients can be started with a list of OSTs that should be
3168 declared "inactive" for known non-responsive OSTs.
3172 Description: ENOENT returned for valid filehandle during dbench.
3173 Details : Check if a directory has children when invalidating dentries
3174 associated with an inode during lock cancellation. This fixes
3175 an incorrect ENOENT sometimes seen for valid filehandles during
3176 testing with dbench.
3179 Frequency : SFS test only (otherwise harmless)
3181 Description: SPEC SFS validation failure on NFS v2 over lustre.
3182 Details : Changes the blocksize for regular files to be 2x RPC size,
3183 and not depend on stripe size.
3185 Severity : enhancement
3187 Description: fine-grained SMP locking inside DLM
3188 Details : Improve DLM performance on SMP systems by removing the single
3189 per-namespace lock and replace it with per-resource locks.
3191 Severity : enhancement
3193 Description: don't hold multiple extent locks at one time
3194 Details : To avoid client eviction during large writes, locks are not
3195 held on multiple stripes at one time or for very large writes.
3196 Otherwise, clients can block waiting for a lock on a failed OST
3197 while holding locks on other OSTs and be evicted.
3199 Severity : enhancement
3201 Description: Multiple MD RPCs in flight.
3202 Details : Further unserialise some read-only MDT RPCs - learn about intents.
3203 To avoid overly-overloading MDT, introduce a limit on number of
3204 MDT RPCs in flight for a single client and add /proc controls
3205 to adjust this limit.
3207 Severity : enhancement
3209 Description: client read/write statistics
3210 Details : Add client read/write call usage stats for performance
3211 analysis of user processes.
3212 /proc/fs/lustre/llite/*/offset_stats shows non-sequential
3213 file access. extents_stats shows chunk size distribution.
3214 extents_stats_per_process show chunk size distribution per
3217 Severity : enhancement
3219 Description: per-client statistics on server
3220 Details : Add ldlm and operations statistics for each client in
3221 /proc/fs/lustre/mds|obdfilter/*/exports/
3223 Severity : enhancement
3225 Description: improved MDT statistics
3226 Details : Add detailed MDT operations statistics in
3227 /proc/fs/lustre/mds/*/stats
3229 Severity : enhancement
3231 Description: VFS operations stats
3232 Details : Add client VFS call stats, trackable by pid, ppid, or gid
3233 /proc/fs/lustre/llite/*/stats_track_[pid|ppid|gid]
3238 Description: Fix client-side osc byte counters
3239 Details : The osc read/write byte counters in
3240 /proc/fs/lustre/osc/*/stats are now working
3243 Frequency : always as root on SLES
3245 Description: Failure of copying files with lustre special EAs.
3246 Details : Client side always return success for setxattr call for lustre
3247 special xattr (currently only "trusted.lov").
3252 Description: Refcount LNET uuids
3253 Details : The global LNET uuid list grew linearly with every startup;
3254 refcount repeated list entries instead of always adding to
3257 Severity : enhancement
3259 Description: Dynamic service threads
3260 Details : Within a small range, start extra service threads
3261 automatically when the request queue builds up.
3264 Frequency : mixed-endian client/server environments
3266 Description: mixed-endian crashes
3267 Details : The new msg_v2 system had some failures in mixed-endian
3270 Severity : enhancement
3272 Description: Easy OST removal
3273 Details : OSTs can be permanently deactivated with e.g. 'lctl
3274 conf_param lustre-OST0001.osc.active=0'
3276 Severity : enhancement
3278 Description: MGS proc entries
3279 Details : Added basic proc entries for the MGS showing what filesystems
3282 Severity : enhancement
3284 Description: provide MGS failover
3285 Details : Added config lock reacquisition after MGS server failover.
3287 Severity : enhancement
3289 Description: add Linux 2.4 support
3290 Details : Added support for RHEL 2.4.21 kernel for 1.6 servers and clients
3294 Description: a large application tries to do I/O to the same resource and dies
3295 in the middle of it.
3296 Details : Check the req->rq_arrival time after the call to
3297 ost_brw_lock_get(), but before we do anything about
3298 processing it & sending the BULK transfer request. This
3299 should help move old stale pending locks off the queue as
3300 quickly as obd_timeout.
3303 Frequency : when an incorrect nid is specified during startup
3305 Description: ptlrpc connect to non-existant node causes kernel crash
3306 Details : LNET can't be re-entered from an event callback, which
3307 happened when we expire a message after the export has been
3308 cleaned up. Instead, hand the zombie cleanup off to another
3311 Severity : enhancement
3313 Description: plain/inodebits lock performance improvement
3314 Details : Grouping plain/inodebits in granted list by their request modes
3315 and bits policy, thus improving the performance of search through
3319 Frequency : only if OST filesystem is corrupted
3321 Description: client incorrectly hits assertion in ptlrpc_replay_req()
3322 Details : for a short time RPCs with bulk IO are in the replay list,
3323 but replay of bulk IOs is unimplemented. If the OST filesystem
3324 is corrupted due to disk cache incoherency and then replay is
3325 started it is possible to trip an assertion. Avoid putting
3326 committed RPCs into the replay list at all to avoid this issue.
3329 Frequency : liblustre (e.g. catamount) on a large cluster with >= 8 OSTs/OSS
3331 Description: System hang on startup
3332 Details : This bug allowed the liblustre (e.g. catamount) client to
3333 return to the app before handling all startup RPCs. This
3334 could leave the node unresponsive to lustre network traffic
3335 and manifested as a server ptllnd timeout.
3337 Severity : enhancement
3339 Description: Add "/proc/sys/lustre/debug_peer_on_timeout"
3340 Details : liblustre envirable: LIBLUSTRE_DEBUG_PEER_ON_TIMEOUT
3341 boolean to control whether to print peer debug info when a
3342 client's RPC times out.
3345 Frequency : only for kernels with patches from Lustre below 1.4.3
3347 Description: Remove old rdonly API
3348 Details : Remove old rdonly API which unused from at least lustre 1.4.3
3351 Frequency : only for devices with external journals
3353 Description: Set external device read-only also
3354 Details : During a commanded failover stop, we set the disk device
3355 read-only while the server shuts down. We now also set any
3356 external journal device read-only at the same time.
3359 Frequency : when upgrading from 1.4 while trying to change parameters
3361 Description: The wrong (new) MDC name was used when setting parameters for
3362 upgraded MDT's. Also allows changing of OSC (and MDC)
3363 parameters if --writeconf is specified at tunefs upgrade time.
3366 Frequency : when setting specific ost indicies
3368 Description: QOS code breaks on skipped indicies
3369 Details : Add checks for missing OST indicies in the QOS code, so OSTs
3370 created with --index need not be sequential.
3372 Severity : enhancement
3374 Description: Add uninit_groups feature to ldiskfs2 to speed up e2fsck
3375 Details : The uninit_groups feature works in conjunction with the kernel
3376 filesystem code (ldiskfs2 only) and e2fsprogs-1.39-cfs6 to speed
3377 up the pass1 processing of e2fsck. This is a read-only feature
3378 in ldiskfs2 only, so older kernels and current ldiskfs cannot
3379 mount filesystems that have had this feature enabled.
3381 Severity : enhancement
3383 Description: Improve multi-block allocation algorithm to avoid fragmentation
3384 Details : The mballoc3 code (ldiskfs2 only) adds new mechanisms to improve
3385 allocation locality and avoid filesystem fragmentation.
3387 ------------------------------------------------------------------------------
3389 2007-04-01 Cluster File Systems, Inc. <info@clusterfs.com>
3391 * Support for kernels:
3392 2.4.21-47.0.1.EL (RHEL 3)
3393 2.6.5-7.283 (SLES 9)
3394 2.6.9-42.0.10.EL (RHEL 4)
3395 2.6.12.6 vanilla (kernel.org)
3396 2.6.16.27-0.9 (SLES 10)
3397 * Recommended e2fsprogs version: 1.39.cfs5
3399 * Note that reiserfs quotas are disabled on SLES 10 in this kernel
3403 Frequency : occasional, depends on client load and configuration
3404 Bugzilla : 12181, 12203
3405 Description: data loss for recently-modified files
3407 Details : In some cases it is possible that recently written or created
3408 files may not be written to disk in a timely manner (this should
3409 normally be within 30s unless client IO load is very high).
3410 The problem appears as zero-length files or files that are a
3411 multiple of 1MB in size after a client crash or client eviction
3412 that are missing data at the end of the file.
3414 This problem is more likely to be hit on clients where files are
3415 repeatedly created and unlinked in the same directory, clients
3416 have a large amount of RAM, have many CPUs, the filesystem has
3417 many OSTs, the clients are rebooted frequently, and/or the files
3418 are not accessed by other nodes after being written.
3420 The presence of the problem can be detected by looking at
3421 /proc/sys/fs/inode-state. If the first number (nr_inodes) is
3422 smaller than the second (nr_unused) then dirty files will not
3423 be flushed automatically to disk. "sync; sleep 10" should be
3424 run several times on the node before unmounting it to update
3425 Lustre (this is also safe to run on nodes without this problem).
3427 There is also a related kernel bug in the RHEL4 4 2.6.9 kernel
3428 that can cause this same problem, so customers using that kernel
3429 also need to update the kernel in addition to Lustre. In order
3430 to properly fix this bug, the RHEL3 2.4.21 kernel is also updated.
3432 It is normal that files written just before a client crash (less
3433 than 30s) may not yet have been flushed to disk, even for local
3437 Frequency : frequent on thin XT3 nodes
3439 Description: UUID collision on thin XT3 Linux nodes
3440 Details : UUIDs on Compute Node Linux XT3 nodes were not generated
3441 randomly, since we relied on an insufficiently-seeded PRNG.
3446 Description: OSS hangs after "All ost request buffers busy"
3447 Details : A deadlock between quota and journal operations caused OSS
3448 hangs after printing "All ost request buffers busy."
3451 Frequency : always on liblustre builds
3453 Description: Cleanup compiler warnings on liblustre
3456 Frequency : always on liblustre builds on XT3
3458 Description: LC_CONFIG_CDEBUG don't run while build liblustre on XT3.
3462 Description: Addition of EXT3_FEATURE_RO_COMPAT_DIR_NLINKS flag for
3463 > 32000 subdirectories
3464 Details : Add EXT3_FEATURE_RO_COMPAT_DIR_NLINK flag to
3465 EXT3_FEATURE_RO_COMPAT_SUPP. This flag will be set whenever
3466 subdirectory count crosses 32000. This will aid e2fsck to
3467 correctly handle more than 32000 subdirectories.
3470 Frequency : liblustre (e.g. catamount) on a large cluster with >= 8 OSTs/OSS
3472 Description: System hang on startup
3473 Details : This bug allowed the liblustre (e.g. catamount) client to
3474 return to the app before handling all startup RPCs. This
3475 could leave the node unresponsive to lustre network traffic
3476 and manifested as a server ptllnd timeout.
3478 Severity : enhancement
3480 Description: Add "/proc/sys/lustre/debug_peer_on_timeout"
3481 (liblustre envirable: LIBLUSTRE_DEBUG_PEER_ON_TIMEOUT)
3482 boolean to control whether to print peer debug info when a
3483 client's RPC times out.
3488 Description: make O_SYNC working on 2.6 kernels
3489 Details : 2.6 kernels use different method for mark pages for write,
3490 so need add a code to lustre for O_SYNC work.
3495 Description: Failure to close file and release space on NFS
3496 Details : Put inode details into lock acquired in ll_intent_file_open.
3497 Use mdc_intent_lock in ll_intent_open to properly
3498 detect all kind of errors unhandled by mdc_enqueue.
3503 Description: proc file read during shutdown sometimes raced obd removal,
3505 Details : Add lock to prevent obd access after proc file removal.
3508 Frequency : Only for files larger than 4GB on 32-bit clients.
3510 Description: improperly doing page alignment of locks
3511 Details : Modify lustre core code to use CFS_PAGE_* defines instead of
3512 PAGE_*. Make CFS_PAGE_MASK a 64-bit mask.
3517 Description: RPCs being resent when they shouldn't be
3518 Details : Some RPCs that should not be resent are being resent. This
3519 can cause inconsistencies in the RPC state machine. Do not
3520 resend such requests.
3523 Frequency : rare, only with NFS export
3525 Description: Crash on NFS re-export node
3526 Details : under very unusual load conditions an assertion is hit in
3527 ll_intent_file_open()
3530 Frequency : only if OST filesystem is corrupted
3532 Description: client incorrectly hits assertion in ptlrpc_replay_req()
3533 Details : for a short time RPCs with bulk IO are in the replay list,
3534 but replay of bulk IOs is unimplemented. If the OST filesystem
3535 is corrupted due to disk cache incoherency and then replay is
3536 started it is possible to trip an assertion. Avoid putting
3537 committed RPCs into the replay list at all to avoid this issue.
3542 Description: large O_DIRECT requests fail under memory pressure/fragmentation
3543 Details : Large single O_DIRECT read and write calls can fail to allocate
3544 a sufficiently large buffer to process the request. In case of
3545 allocation failure the allocation is retried with a smaller
3546 buffer and broken into smaller requests.
3548 Severity : enhancement
3550 Description: Add -o localflock option to simulate old noflock behaviour.
3551 Details : This will achieve local-only flock/fcntl locks coherentness.
3556 Description: versioning check is incomplete
3557 Details : Checking the version difference of client vs. server, report
3558 error if the gap is too big.
3563 Description: add support PG_writeback bit
3564 Details : add support for PG_writeback bit for Lustre, for more carefull
3565 work with page cache in 2.6 kernel. This also fix some deadlocks
3566 and remove hack for work O_SYNC with 2.6 kernel.
3568 Severity : enhancement
3570 Description: Add uninit_groups feature to ldiskfs2 to speed up e2fsck
3571 Details : The uninit_groups feature works in conjunction with the kernel
3572 filesystem code (ldiskfs2 only) and e2fsprogs-1.39-cfs6 to speed
3573 up the pass1 processing of e2fsck. This is a read-only feature
3574 in ldiskfs2 only, so older kernels and current ldiskfs cannot
3575 mount filesystems that have had this feature enabled.
3577 Severity : enhancement
3579 Description: Improve multi-block allocation algorithm to avoid fragmentation
3580 Details : The mballoc3 code (ldiskfs2 only) adds new mechanisms to improve
3581 allocation locality and avoid filesystem fragmentation.
3583 ------------------------------------------------------------------------------
3585 2007-02-09 Cluster File Systems, Inc. <info@clusterfs.com>
3587 * Support for kernels:
3588 2.6.9-42.0.3.EL (RHEL 4)
3589 2.6.5-7.276 (SLES 9)
3590 2.4.21-47.0.1.EL (RHEL 3)
3591 2.6.12.6 vanilla (kernel.org)
3592 2.6.16.21-0.8 (SLES10)
3593 * Recommended e2fsprogs version: 1.39.cfs2-0
3595 * The backwards-compatible /proc/sys/portals symlink has been removed
3596 in this release. Before upgrading, please ensure that you change
3597 any configuration scripts or /etc/sysctl.conf files that access
3598 /proc/sys/portals/* or sysctl portals.* to use the corresponding
3599 entry in /proc/sys/lnet or sysctl lnet.*. This change can be made
3600 in advance of the upgrade on any system running Lustre 1.4.6 or
3601 newer, since /proc/sys/lnet was added in that version.
3602 * Note that reiserfs quotas are disabled on SLES 10 in this kernel
3606 Frequency : only when quota is used
3608 Description: avoid scanning export list for quota master
3609 Details : Change the algorithms to avoid scanning export list in order
3610 to improve the efficiency.
3613 Frequency : MDS failover only, very rarely
3615 Description: "went back in time" messages on mds failover
3616 Details : The greatest transno may be lost when the current operation
3617 finishes with an error (transno==0) and the client's last_rcvd
3618 record is over-written. Save the greatest transno in the
3619 mds_last_transno for this case.
3622 Frequency : always for specific kernels and striping counts
3624 Description: client may get "Matching packet too big" without ACL support
3625 Details : Clients compiled without CONFIG_FS_POSIX_ACL get an error message
3626 when trying to access files in certain configurations. The
3627 clients should in fact be denied when mounting because they do
3628 not understand ACLs.
3631 Frequency : Cray XT3 with more than 4000 clients and multiple jobs
3633 Description: many clients connecting with IO in progress causes connect timeouts
3634 Details : Avoid synchronous journal commits to avoid delays caused by many
3635 clients connecting/disconnecting when bulk IO is in progress.
3636 Queue liblustre connect requests on OST_REQUEST_PORTAL instead of
3637 OST_IO_PORTAL to avoid delays behind potentially many pending
3641 Frequency : occasionally with multiple writers to a single file
3643 Description: shared writes to file may result in wrong size reported by stat()
3644 Details : Allow growing of kms when extent lock is cancelled
3647 Frequency : always with random mmap IO to multi-striped file
3649 Description: mmap write might be lost if we are writing to a 'hole' in stripe
3650 Details : Only if the hole is at the end of OST object so that kms is too
3651 small. Fix is to increase kms accordingly in ll_nopage.
3654 Frequency : rare, only if OST filesystem is inconsistent with MDS filesystem
3656 Description: writes to a missing object would leak memory on the OST
3657 Details : If there is an inconsistency between the MDS and OST filesystems,
3658 such that the MDS references an object that doesn't exist, writes
3659 to that object will leak memory due to incorrect cleanup in the
3660 error handling path, eventually running out of memory on the OST.
3665 Description: Creating too long symlink causes lustre errors
3666 Details : Check symlink and name lengths before sending requests to MDS.
3669 Frequency : only if flock is enabled (not on by default)
3671 Description: posix locks not released on fd closure on 2.6.9+
3672 Details : We failed to add posix locks to list of inode locks on 2.6.9+
3673 kernels, this caused such locks not to be released on fd close and
3674 then assertions on fs unmount about still used locks.
3677 Frequency : MDS failover only, very rarely
3679 Description: clients may get ASSERTION(granted_lock != NULL)
3680 Details : When request was taking a long time, and a client was resending
3681 a getattr by name lock request. The were multiple lock requests
3682 with the same client lock handle and
3683 mds_getattr_name->fixup_handle_for_resent_request found one of the
3684 lock handles but later failed with ASSERTION(granted_lock != NULL).
3689 Description: handle->h_buffer_credits > 0, assertion failure
3690 Details : h_buffer_credits is zero after truncate, causing assertion
3691 failure. This patch extends the transaction or creates a new
3695 Frequency : NFS re-export or patchless client
3696 Bugzilla : 11179, 10796
3697 Description: Crash on NFS re-export node (__d_move)
3698 Details : We do not want to hash the dentry if we don't have a lock.
3699 But if this dentry is later used in d_move, we'd hit uninitialised
3700 list head d_hash, so we just do this to init d_hash field but
3701 leave dentry unhashed.
3704 Frequency : NFS re-export or patchless client
3706 Description: NFS exports has problem with symbolic link
3707 Details : lustre client didn't properly install dentry when re-exported
3708 to NFS or running patchless client.
3711 Frequency : NFS re-export or patchless client
3713 Description: Various nfs/patchless fixes.
3714 Details : fixes reuse disconected alias for lookup process - this fixes
3715 warning "find_exported_dentry: npd != pd",
3716 fix permission error with open files at nfs.
3717 fix apply umask when do revalidate.
3720 Frequency : occasional
3722 Description: Crash on NFS re-export node
3723 Details : calling clear_page() on the wrong pointer triggered oops in
3724 generic_mapping_read().
3727 Frequency : rarely, using O_DIRECT IO
3729 Description: unaligned directio crashes client with LASSERT
3730 Details : check for unaligned buffers before trying any requests.
3733 Frequency : rarely, using CFS RAID5 patches in non-standard kernel series
3735 Description: stale data returned from RAID cache
3736 Details : If only a small amount of IO is done to the RAID device before
3737 reading it again it is possible to get stale data from the RAID
3738 cache instead of reading it from disk.
3741 Frequency : always for sles10 kernel
3743 Description: sles10 support
3744 Details : ll_follow_link: compile fixes and using of nd_set_link
3745 under newer kernels.
3748 Frequency : depends on arch, kernel and compiler version, always on sles10
3751 Description: recursive or deep enough symlinks cause stack overflow
3752 Details : getting rid of large stack-allocated variable in
3756 Frequency : depends on hardware
3758 Description: lustre write performance loss in the SLES10 kernel
3759 Details : the performance loss is caused by using of write barriers in the
3760 ext3 code. The SLES10 kernel turns barrier support on by
3761 default. The fix is to undo that change for ldiskfs.
3763 ------------------------------------------------------------------------------
3765 2006-12-09 Cluster File Systems, Inc. <info@clusterfs.com>
3767 * Support for kernels:
3768 2.6.9-42.0.3EL (RHEL 4)
3769 2.6.5-7.276 (SLES 9)
3770 2.4.21-47.0.1.EL (RHEL 3)
3771 2.6.12.6 vanilla (kernel.org)
3775 Frequency : quota enabled and large files being deleted
3777 Description: releasing more than 4GB of quota at once hangs OST
3778 Details : If a user deletes more than 4GB of files on a single OST it
3779 will cause the OST to spin in an infinite loop. Release
3780 quota in < 4GB chunks, or use a 64-bit value for 1.4.7.1+.
3785 Description: statfs data retrieved from /proc may be stale or zero
3786 Details : When reading per-device statfs data from /proc, in the
3787 {kbytes,files}_{total,free,avail} files, it may appear
3788 as zero or be out of date.
3791 Frequency : systems with MD RAID1 external journal devices
3793 Description: lconf's call to blkid is confused by RAID1 journal devices
3794 Details : Use the "blkid -l" flag to locate the MD RAID device instead
3795 of returning all block devices that match the journal UUID.
3798 Frequency : always, for aggregate stripe size over 4GB
3800 Description: "lfs setstripe" fails assertion when setting 4GB+ stripe width
3801 Details : Using "lfs setstripe" to set stripe size * stripe count over 4GB
3802 will fail the kernel with "ASSERTION(lsm->lsm_xfersize != 0)"
3805 Frequency : always if "lfs find" used on a local file/directory
3807 Description: "lfs find" segfaults if used on a local file/directory
3808 Details : The case where a directory component was not specified wasn't
3809 handled correctly. Handle this properly.
3812 Frequency : always on ppc64
3814 Description: the write to an ext3 filesystem mounted with mballoc got stuck
3815 Details : ext3_mb_generate_buddy() uses find_next_bit() which does not
3816 perform endianness conversion.
3819 Frequency : rarely (truncate to non-zero file size after write under load)
3820 Bugzilla : 10730, 10687
3821 Description: Files padded with zeros to next 4K multiple
3822 Details : With filesystems mounted using the "extents" option (2.6 kernels)
3823 it is possible that files that are truncated to a non-zero size
3824 immediately after being written are filled with zero bytes beyond
3825 the truncated size. No file data is lost.
3827 Severity : enhancement
3829 Description: Allow recovery/failover for liblustre clients.
3830 Details : liblustre clients were unaware of failover configurations until
3833 Severity : enhancement
3835 Description: user file locks should fail when not mounting with flock option
3836 Details : Set up an error-returning stub in ll_file_operations.lock field
3837 to prevent incorrect behaviour when client is mounted without
3838 flock option. Also, set up properly f_op->flock field for
3842 Frequency : always on ia64
3844 Description: "lfs df" loops on printing out MDS statfs information
3845 Details : The obd_ioctl_data was not initialized and in some systems
3846 this caused a failure during the ioctl that did not return
3847 an error. Initialize the struct and return an error on failure.
3850 Frequency : SLES 9 only
3852 Description: Error of copying files with lustre special EAs as root
3853 Details : Client side always return success for setxattr call for lustre
3854 special xattr (currently only "trusted.lov").
3857 Frequency : rarely on clusters with both ia64+i386 clients
3859 Description: ia64+i686 clients doing shared IO on the same file may LBUG
3860 Details : In rare cases when both ia64+i686 (or other mixed-PAGE_SIZE)
3861 clients are doing concurrent writes to the same file it is
3862 possible that the ia64 clients may LASSERT because the OST
3863 extent locks are not PAGE_SIZE aligned. Ensure that grown
3864 locks are always aligned on the request boundary.
3867 Frequency : specific use, occasional
3869 Description: Overwriting in use executable truncates on-disk binary image
3870 Details : If one node attempts to overwrite an executable in use by
3871 another node, we now correctly return ETXTBSY instead of
3872 truncating the file.
3874 Severity : enhancement
3876 Description: Async OSC create to avoid the blocking unnecessarily.
3877 Details : If a OST has no remain object, system will block on the creating
3878 when need to create a new object on this OST. Now, ways use
3879 pre-created objects when available, instead of blocking on an
3880 empty osc while others are not empty. If we must block, we block
3881 for the shortest possible period of time.
3886 Description: chmod on Lustre root is propagated to other clients
3887 Details : Re-validate root's dentry in ll_lookup_it to avoid having it
3888 invalid by the follow_mount time.
3891 Frequency : liblustre clients only
3893 Description: Race in 'instant cancel' lock handling could lead to such locks
3894 never to be granted in case of SMP MDS
3895 Details : Do not destroy not yet granted but cbpending locks in
3899 Frequency : replay/resend of open
3901 Description: non null lock assetion failure in mds_intent_policy
3902 Details : Trying to replay/resend lockless open requests resulted in
3903 mds_open() returning 0 with no lock. Now it sets a flag if
3904 a lock is going to be returned.
3906 Severity : enhancement
3908 Description: Checksum enhancements
3909 Details : New checksum enhancements allow for resending RPCs that failed
3912 Severity : enhancement
3914 Description: Tunables on number of dirty pages in cacche
3915 Details : Allow to set limit on number of dirty pages cached.
3920 Description: client crash on unmount - lock still has references
3921 Details : In some error handling cases it was possible to leak a lock
3922 reference on a client while accessing a file. This was not
3923 harmful to the client during operation, but would cause the
3924 client to crash when the filesystem is unmounted.
3927 Frequency : specific case, rare
3929 Description: ETXTBSY on mds though file not in use by client
3930 Details : ETXTBSY is no longer incorrectly returned when attempting to
3931 chmod or chown a directory that the user previously tried to
3932 execute or a currently-executing binary.
3935 Frequency : extremely rare except on liblustre-based clients
3937 Description: Lustre space not freed when files are deleted
3938 Details : Clean up open-unlinked files after client eviction. Previously
3939 the unlink was skipped and the files remained as orphans.
3944 Description: OST failure "would be an LBUG" in waiting_locks_callback()
3945 Details : In some cases it was possible to send a blocking callback to
3946 a client doing a glimpse, even though that client didn't get
3947 a lock granted. When the glimpse lock is cancelled on the OST
3948 the freed lock is left on the waiting list and corrupted the list.
3951 Frequency : all core dumps
3953 Description: Broke core dumps to lustre
3954 Details : Negative dentry may be unhashed if parent does not have UPDATE
3955 lock, but some callers, e.g. do_coredump, expect dentry to be
3956 hashed after successful create, hash it in ll_create_it.
3958 ------------------------------------------------------------------------------
3960 2006-09-13 Cluster File Systems, Inc. <info@clusterfs.com>
3962 * Support for kernels:
3963 2.6.9-42.0.2.EL (RHEL 4)
3964 2.6.5-7.276 (SLES 9)
3965 2.4.21-40.EL (RHEL 3)
3966 2.6.12.6 vanilla (kernel.org)
3970 Frequency : always on RHEL 3
3972 Description: Number of open files grows over time
3973 Details : The number of open files grows over time, whether or not
3974 Lustre is started. This was due to a filp leak introduced
3975 by one of our kernel patches.
3977 ------------------------------------------------------------------------------
3979 2006-08-20 Cluster File Systems, Inc. <info@clusterfs.com>
3981 * Support for kernels:
3982 2.6.9-42.EL (RHEL 4)
3983 2.6.5-7.267 (SLES 9)
3984 2.4.21-40.EL (RHEL 3)
3985 2.6.12.6 vanilla (kernel.org)
3990 Bugzilla : 5719, 9635, 9792, 9684
3991 Description: OST (or MDS) trips assertions in (re)connection under heavy load
3992 Details : If a server is under heavy load and cannot reply to new
3993 connection requests before the client resends the (re)connect,
3994 the connection handling code can behave badly if two service
3995 threads are concurrently handing separate (re)connections from
3996 the same client. Add better locking to the connection handling
3997 code, and ensure that only a single connection will be processed
3998 for a given client UUID, even if the lock is dropped.
4000 Severity : enhancement
4002 Description: add TCP zero-copy support to kernel
4003 Details : Add support to the kernel TCP stack to allow zero-copy bulk
4004 sends if the hardware supports scatter-gather and checksumming.
4005 This allows socklnd to do client-write and server-read more
4006 efficiently and reduce CPU utilization from skbuf copying.
4009 Frequency : only if NFS exporting from client
4011 Description: NULL pointer deref in ll_iocontrol() if chattr mknod file
4012 Details : If setting attributes on a file created under NFS that had
4013 never been opened it would be possible to oops the client
4014 if the file had no objects.
4018 Bugzilla : 9326, 10402, 10897
4019 Description: client crash in ptlrpcd_wake() thread when sending async RPC
4020 Details : It is possible that ptlrpcd_wake() dereferences a freed async
4021 RPC. In rare cases the ptlrpcd thread alread processed the RPC
4022 before ptlrpcd_wake() was called and the request was freed.
4025 Frequency : always for liblustre
4027 Description: liblustre client does MDS+OSTs setattr RPC for each write
4028 Details : When doing a write from a liblustre client, the client
4029 incorrectly issued an RPC to the MDS and each OST the file was
4030 striped over in order to update the timestamps. When writing
4031 with small chunks and many clients this could overwhelm the MDS
4032 with RPCs. In all cases it would slow down the write because
4033 these RPCs are unnecessary.
4035 Severity : enhancement
4037 Description: allow number of MDS service threads to be changed at module load
4038 Details : It is now possible to change the number of MDS service threads
4039 running. Adding "options mds mds_num_threads={N}" to the MDS's
4040 /etc/modprobe.conf will set the number of threads for the next
4041 time Lustre is restarted (assuming the "mds" module is also
4042 reloaded at that time). The default number of threads will
4043 stay the same, 32 for most systems.
4048 Description: OST crash if filesystem is unformatted or corrupt
4049 Details : If an OST is started on a device that has never been formatted
4050 or if the filesystem is corrupt and cannot even mount then the
4051 error handling cleanup routines would dereference a NULL pointer.
4056 Description: NULL pointer deref in llap_from_page.
4057 Details : get_cache_page_nowait can return a page with NULL (or otherwise
4058 incorrect) mapping if the page was truncated/reclaimed while it was
4059 searched for. Check for this condition and skip such pages when
4060 doing readahead. Introduce extra check to llap_from_page() to
4061 verify page->mapping->host is non-NULL (so page is not anonymous).
4064 Frequency : Sometimes when using sys_sendfile
4066 Description: "page not covered by a lock" warnings from ll_readpage
4067 Details : sendfile called ll_readpage without right page locks present.
4068 Now we introduced ll_file_sendfile that does necessary locking
4069 around call to generic_file_sendfile() much like we do in
4073 Frequency : with certain MDS communication failures at client mount time
4075 Description: NULL pointer deref after failed client mount
4076 Details : a client connection request may delayed by the network layer
4077 and not be sent until after the PTLRPC layer has timed out the
4078 request. If the client fails the mount immediately it will try
4079 to clean up before the network times out the request. Add a
4080 reference from the request import to the obd device and delay
4081 the cleanup until the network drops the request.
4084 Frequency : occasionally during client (re)connect
4086 Description: assertion failure during client (re)connect
4087 Details : processing a client connection request may be delayed by the
4088 client or server longer than the client connect timeout. This
4089 causes the client to resend the connection request. If the
4090 original connection request is replied in this interval, the
4091 client may trip an assertion failure in ptlrpc_connect_interpret()
4092 which thought it would be the only running connect process.
4095 Frequency : only with obd_echo servers and clients that are rebooted
4097 Description: kernel BUG accessing uninitialized data structure
4098 Details : When running an obd_echo server it did not start the ping_evictor
4099 thread, and when a client was evicted an uninitialized data
4100 structure was accessed. Start the ping_evictor in the RPC
4101 service startup instead of the OBD startup.
4103 Severity : enhancement
4104 Bugzilla : 10193 (patchless)
4105 Description: Remove dependency on various unexported kernel interfaces.
4106 Details : No longer need reparent_to_init, exit_mm, exit_files,
4107 sock_getsockopt, filemap_populate, FMODE_EXEC, put_filp.
4110 Frequency : rare (only users of deprecated and unsupported LDAP config)
4112 Description: write_conf for zeroconf mount queried LDAP incorrectly for client
4113 Details : LDAP apparently contains 'lustreName' attributes instead of
4114 'name'. A simple remapping of the name is sufficient.
4117 Frequency : rare (only with non-default dump_on_timeout debug enabled)
4119 Description: waiting_locks_callback trips kernel BUG if client is evicted
4120 Details : Running with the dump_on_timeout debug flag turned on makes
4121 it possible that the waiting_locks_callback() can try to dump
4122 the Lustre kernel debug logs from an interrupt handler. Defer
4123 this log dumping to the expired_lock_main() thread.
4125 Severity : enhancement
4127 Description: Support NFS exporting on 2.6 kernels.
4128 Details : Implement non-rawops metadata methods for NFS server to use without
4129 changing NFS server code.
4132 Frequency : very rare (synthetic metadata workload only)
4134 Description: two racing renames might cause an MDS thread to deadlock
4135 Details : Running the "racer" program may cause one MDS thread to rename
4136 a file from being the source of a rename to being the target of
4137 a rename at exactly the same time that another thread is doing
4138 so, and the second thread has already enqueued these locks after
4139 doing a lookup of the target and is trying to relock them in
4140 order. Ensure that we don't try to re-lock the same resource.
4143 Frequency : only very large systems with liblustre clients
4145 Description: slow eviction of liblustre clients with the "evict_by_nid" RPC
4146 Details : Use asynchronous set_info RPCs to send the "evict_by_nid" to
4147 all OSTs in parallel. This allows the eviction of stale liblustre
4148 clients to proceed much faster than if they were done in series,
4149 and also offers similar improvements for other set_info RPCs.
4154 Description: excessive CPU usage during initial read phase on client
4155 Details : During the initial read phase on a client, it would agressively
4156 retry readahead on the file, consuming too much CPU and impacting
4157 performance (since 1.4.5.8). Improve the readahead algorithm
4158 to avoid this, and also improve some other common cases (read
4159 of small files in particular, where "small" is files smaller than
4160 /proc/fs/lustre/llite/*/max_read_ahead_whole_mb, 2MB by default).
4165 Description: MDS crash when receiving packet with unknown intent.
4166 Details : Do not LBUG in unknown intent case, just return -EFAULT
4168 Severity : enhancement
4169 Bugzilla : 9293, 9385
4170 Description: MDS RPCs are serialised on client. This is unnecessary for some.
4171 Details : Do not serialize getattr (non-intent version) and statfs.
4174 Frequency : occasional, when OST network is overloaded/intermittent
4176 Description: client evicted by OST after bulk IO timeout
4177 Details : If a client sends a bulk IO request (read or write) the OST
4178 may evict the client if it is unresposive to its data GET/PUT
4179 request. This is incorrect if the network is overloaded (takes
4180 too long to transfer the RPC data) or dropped the OST GET/PUT
4181 request. There is no need to evict the client at all, since
4182 the pinger and/or lock callbacks will handle this, and the
4183 client can restart the bulk request.
4186 Frequency : Always when mmapping file with no objects
4188 Description: client crashes when mmapping file with no objects
4189 Details : Check that we actually have objects in a file before doing any
4190 operations on objects in ll_vm_open, ll_vm_close and
4196 Description: Request leak when working with deleted CWD
4197 Details : Introduce advanced request refcount tracking for requests
4198 referenced from lustre intent.
4200 Severity : Enhancement
4202 Description: Cache open file handles on client.
4203 Details : MDS now will return special lock along with openhandle, if
4204 requested and client is allowed to hold openhandle, even if unused,
4205 until such a lock is revoked. Helps NFS a lot, since NFS is opening
4206 closing files for every read/write openration.
4208 Severity : Enhancement
4210 Description: Cache open negative dentries on client when possible.
4211 Details : Guard negative dentries with UPDATE lock on parent dir, drop
4212 negative dentries on lock revocation.
4217 Description: Remounting a client read-only wasn't possible with a zconf mount
4218 Details : It wasn't possible to remount a client read-only with llmount.
4220 Severity : enhancement
4221 Description: Include MPICH 1.2.6 Lustre ADIO interface patch
4222 Details : In lustre/contrib/ or /usr/share/lustre in RPM a patch for
4223 MPICH is included to add Lustre-specific ADIO interfaces.
4224 This is based closely on the UFS ADIO layer and only differs
4225 in file creation, in order to allow the OST striping to be set.
4226 This is user-contributed code and not supported by CFS.
4231 Description: extended inode attributes (immutable, append-only) work improperly
4232 when 2.4 and 2.6 kernels are used on client/server or vice versa
4233 Details : Introduce kernel-independent values for these flags.
4235 Severity : enhancement
4238 Description: Allow fractional MB tunings for lustre in /proc/ filesystem.
4239 Details : Many of the /proc/ tunables can only be tuned at a megabyte
4240 granularity. Now, Fractional MB granularity is be supported,
4241 this is very useful for low memory system.
4243 Severity : enhancement
4245 Description: Getattr by fid
4246 Details : Getting a file attributes by its fid, obtaining UPDATE|LOOKUP
4247 locks, avoids extra getattr rpc requests to MDS, allows '/' to
4248 have locks and avoids getattr rpc requests for it on every stat.
4251 Frequency : Always, for filesystems larger than 2TB
4253 Description: ldiskfs crash at mount for filesystem larger than 2TB with mballoc
4254 Details : Kenrel kmalloc limits allocations to 128kB and this prevents
4255 filesystems larger than 2TB to be mounted with mballoc enabled.
4258 Frequency : Always, for 32-bit kernel without CONFIG_LBD and filesystem > 2TB
4260 Description: filesystem corruption for non-standard kernels and very large OSTs
4261 Details : If a 32-bit kernel is compiled without CONFIG_LBD enabled and a
4262 filesystems larger than 2TB is mounted then the kernel will
4263 silently corrupt the start of the filesystem. CONFIG_LBD is
4264 enabled for all CFS-supported kernels, but the possibility of
4265 this happening with a modified kernel config exists.
4267 Severity : enhancement
4269 Description: add client O_DIRECT support for 2.6 kernels
4270 Details : It is now possible to do O_DIRECT reads and writes to files
4271 in the Lustre client mountpoint on 2.6 kernel clients.
4273 Severity : enhancement
4275 Description: parallel glimpse, setattr, statfs, punch, destroy requests
4276 Details : Sends glimpse, setattr, statfs, punch, destroy requests to OSTs in
4277 parallel, not waiting for response from every OST before sending
4278 a rpc to the next OST.
4283 Description: setattr vs write race when updating file timestamps
4284 Details : Client processes that update a file timestamp into the past
4285 right after writing to the file (e.g. tar) it is possible that
4286 the updated file modification time can be reset to the current
4287 time due to a race between processing the setattr and write RPC.
4289 Severity : enhancement
4291 Description: Bring 'lfs find' closer in line with regular Linux find.
4292 Details : lfs find util supports -atime, -mtime, -ctime, -maxdepth, -print,
4293 -print0 options and obtains all the needed info through the lustre
4296 Severity : enhancement
4298 Description: support up to 1024 configured devices on one node
4299 Details : change obd_dev array from statically allocated to dynamically
4300 allocated structs as they are first used to reduce memory usage
4305 Description: Flush dirty partially truncated pages during truncate
4306 Details : Immediatelly flush partially truncated pages in filter_setattr,
4307 this way we completely avoid having any pages in page cache on OST
4308 and can retire ugly workarounds during writes to flush such pages.
4313 Description: i_sem vs transaction deadlock in mds_obd_destroy during unlink.
4314 Details : protect inode from truncation within vfs_unlink() context
4315 just take a reference before calling vfs_unlink() and release it
4316 when parent's i_sem is free.
4319 Frequency : always, if extents are used on OSTs
4321 Description: index ei_leaf_hi (48-bit extension) is not zeroed in extent index
4322 Details : OSTs using the extents format would not zero the high 16 bits of
4323 the index physical block number. This is not a problem for any
4324 OST filesystems smaller than 16TB, and no kernels support ext3
4325 filesystems larger than 16TB yet. This is fixed in 1.4.7 (all
4326 new/modified files) and can be fixed for existing filesystems
4327 with e2fsprogs-1.39-cfs1.
4332 Description: import connection selection may be incorrect if timer wraps
4333 Details : Using a 32-bit jiffies timer with HZ=1000 may cause backup
4334 import connections to be ignored if the 32-bit jiffies counter
4335 wraps. Use a 64-bit jiffies counter.
4338 Frequency : during server recovery
4340 Description: crash after server is denying duplicate export
4341 Details : If clients are resending connect requests to the server, the
4342 server refuses to allow a client to connect multiple times.
4343 Fixed a bug in the handling of this case.
4346 Frequency : very large clusters immediately after boot
4348 Description: LNET request buffers exhausted under heavy short-term load
4349 Details : If a large number of client requests are generated on a service
4350 that has previously never seen so many requests it is possible
4351 that the request buffer growth cannot keep up with the spike in
4352 demand. Instead of dropping incoming requests, they are held in
4353 the LND until the RPC service can accept more requests.
4356 Frequency : Sometimes during replay
4358 Description: Assertion failure in ll_local_open after replay.
4359 Details : If replay happened on an open request reply before we were able
4360 to set replay handler, reply will become not swabbed tripping the
4361 assertion in ll_local_open. Now we set the handler right after
4362 recognising of open request
4365 Frequency : very rare
4367 Description: kernel reports "badness in vsnprintf"
4368 Details : Reading from the "recovery_status" /proc file in small chunks
4369 may cause a negative length in lprocfs_obd_rd_recovery_status()
4370 call to vsnprintf() (which is otherwise harmless). Exit early
4371 if there is no more space in the output buffer.
4373 Severity : enhancement
4375 Description: clear OBD RPC statistics by writing to them
4376 Details : It is now possible to clear the OBD RPC statistics by writing
4377 to the "stats" file.
4382 Description: Client mtime is not the same on different clients after utimes
4383 Details : In some cases, the client was using the utimes() syscall on
4384 a file cached on another node. The clients now validate the
4385 ctime from the MDS + OSTs to determine which one is right.
4390 Description: Inability to activate failout mode
4391 Details : lconf script incorrectly assumed that in python string's numeric
4392 value is used in comparisons.
4395 Frequency : always with multiple stripes per file
4397 Description: Inefficient object allocation for mutli-stripe files
4398 Details : When selecting which OSTs to stripe files over, for files with
4399 a stripe count that divides evenly into the number of OSTs,
4400 the MDS is always picking the same starting OST for each file.
4401 Return the OST selection heuristic to the original design.
4406 Description: mount failures may take full timeout to return an error
4407 Details : Under some heavy load conditions it is possible that a
4408 failed mount can wait for the full obd_timeout interval,
4409 possibly several minutes, before reporting an error.
4410 Instead return an error as soon as the status is known.
4412 ------------------------------------------------------------------------------
4414 2006-02-14 Cluster File Systems, Inc. <info@clusterfs.com>
4416 * WIRE PROTOCOL CHANGE. This version of Lustre networking WILL NOT
4417 INTEROPERATE with older versions automatically. Please read the
4418 user documentation before upgrading any part of a live system.
4419 * WARNING: Lustre networking configuration changes are required with
4420 this release. See https://bugzilla.clusterfs.com/show_bug.cgi?id=10052
4423 * Support for kernels:
4424 2.6.9-22.0.2.EL (RHEL 4)
4425 2.6.5-7.244 (SLES 9)
4426 2.6.12.6 vanilla (kernel.org)
4429 Severity : enhancement
4430 Bugzilla : 7981/8208
4431 Description: Introduced Lustre Networking (LNET)
4432 Details : LNET is new networking infrastructure for Lustre, it includes
4433 a reorganized network configuration mode (see the user
4434 documentation for full details) as well as support for routing
4435 between different network fabrics. Lustre Networking Devices
4436 (LNDs) for the supported network fabrics have also been
4437 created for this new infrastructure.
4439 Severity : enhancement
4440 Description: Introduced Access control lists
4441 Details : clients can set ACLs on files and directories in order to have
4442 more fine-grained permissions than the standard Unix UGO+RWX.
4443 The MDS must be started with the "-o acl" mount option.
4445 Severity : enhancement
4446 Description: Introduced filesystem quotas
4447 Details : Administrators may now establish per-user quotas on the
4450 Severity : enhancement
4452 Description: Configuration change for the XT3
4453 The PTLLND is now used to run Lustre over Portals on the XT3
4454 The configure option(s) --with-cray-portals are no longer used.
4455 Rather --with-portals=<path-to-portals-includes> is used to
4456 enable building on the XT3. In addition to enable XT3 specific
4457 features the option --enable-cray-xt3 must be used.
4462 Description: Running on many-way SMP OSTs can trigger oops in llcd_send()
4463 Details : A race between allocating a new llcd and re-getting the llcd_lock
4464 allowed another thread to grab newly-allocated llcd.
4466 Severity : enhancement
4468 Description: 2.6 OST async journal commit and locking fix to improve performance
4469 Details : The filter_direct_io()+filter_commitrw_write() journal commits for
4470 2.6 kernels are now async as they already were in 2.4 kernels so
4471 that they can commit concurrently with the network bulk transfer.
4472 For block-allocated files the filter allocation semaphore is held
4473 to avoid filesystem fragmentation during allocation. BKL lock
4474 removed for 2.6 xattr operations where it is no longer needed.
4479 Description: lconf incorrectly determined whether two IP networks could talk
4480 Details : In some more complicated routing and multiple-network
4481 configurations, lconf will avoid trying to make a network
4482 connection to a disjoint part of the IP space. It was doing the
4483 math incorrectly for one set of cases.
4488 Description: Fix for potential infinite loop processing records in an llog.
4489 Details : If an llog record is corrupted/zeroed, it is possible to loop
4490 forever in llog_process(). Validate the llog record length
4491 and skip the remainder of the block on error.
4494 Frequency : occasional (liblustre only)
4496 Description: liblustre could not open files whose last component is a symlink
4497 Details : sysio_path_walk() would incorrectly pass the open intent to
4498 intermediate path components.
4501 Frequency : rare (liblustre only with non-standard tuning)
4502 Bugzilla : 7201 (7350)
4503 Description: Tuning the MDC DLM LRU size to zero triggers client LASSERT
4504 Details : llu_lookup_finish_locks() tries to set lock data on a lock
4505 after it has been released, only do this for referenced locks
4507 Severity : enhancement
4509 Description: specifying an (invalid) directory default stripe_size of -1
4510 would reset the directory default striping
4511 Details : stripe_size -1 was used internally to signal directory stripe
4512 removal, now use "all default" to signal dir stripe removal
4513 as a directory striping of "all default" is not useful
4516 Frequency : common for large clusters running liblustre clients
4518 Description: doing an ls when liblustre clients are running is slow
4519 Details : sending a glimpse AST to a liblustre client waits for every AST
4520 to time out, as liblustre clients will not respond. Since they
4521 cannot cache data we refresh the OST lock LVB from disk instead.
4523 Severity : enhancement
4525 Description: doing an ls at the same time as file IO can be slow
4526 Details : enqueue and other "small" requests can be blocked behind many
4527 large IO requests. Create a new OST IO portal for non-IO
4528 requests so they can be processed faster.
4531 Frequency : rare (only HPUX clients mounting unsupported re-exported NFS vol)
4533 Description: an HPUX NFS client would get -EACCESS when ftruncate()ing a newly
4534 created file with mode 000
4535 Details : the Linux NFS server relies on an MDS_OPEN_OWNEROVERRIDE hack to
4536 allow an ftruncate() as a non-root user to a file with mode 000.
4537 Lustre now respects this flag to disable mode checks when
4538 truncating a file owned by the user
4541 Frequency : liblustre-only, when liblustre client dies unexpectedly or becomes
4544 Description: Revoking locks from clients that went dead or catatonic might take
4546 Details : New lock flags FL_CANCEL_ON_BLOCK used by liblustre makes
4547 cancellation of such locks instant on servers without waiting for
4548 any reply from clients. Clients drops these locks when cancel
4549 notification from server is received without replying.
4552 Frequency : liblustre-only, when liblustre client dies or becomes busy
4554 Description: Doing ls on Linux clients can take a long time with active
4556 Details : Liblustre client cannot handle ASTs in timely manner, so avoid
4557 granting such locks to it in the first place if possible. Locks
4558 are taken by proxy on the OST during the read or write and
4559 dropped immediately afterward. Add connect flags handling, do
4560 not grant locks to liblustre clients for glimpse ASTs.
4562 Severity : enhancement
4564 Description: Improve read-ahead algorithm to avoid excessive IO for random reads
4565 Details : Existing read-ahead algorithm is tuned for the case of streamlined
4566 sequential reads and behaves badly with applications doing random
4567 reads. Improve it by reading ahead at least read region, and
4568 avoiding excessive large RPC for small reads.
4570 Severity : enhancement
4572 Description: Creating more than 1000 files for a single job may cause a load
4573 imbalance on the OSTs if there are also a large number of OSTs.
4574 Details : qos_prep_create() uses an OST index reseed value that is an
4575 even multiple of the number of available OSTs so that if the
4576 reseed happens in the middle of the object allocation it will
4577 still utilize the OSTs as uniformly as possible.
4582 Description: OST or MDS may oops in ping_evictor_main()
4583 Details : ping_evictor_main() drops obd_dev_lock if deleting a stale export
4584 but doesn't restart at beginning of obd_exports_timed list
4587 Severity : enhancement
4589 Description: improve by-nid export eviction on the MDS and OST
4590 Details : allow multiple exports with the same NID to be evicted at one
4591 time without re-searching the exports list.
4594 Frequency : rare, only with supplementary groups enabled on SMP 2.6 kernels
4596 Description: MDS may oops in groups_free()
4597 Details : in rare race conditions a newly allocated group_info struct is
4598 freed again, and this can be NULL. The 2.4 compatibility code
4599 for groups_free() checked for a NULL pointer, but 2.6 did not.
4602 Frequency : common for liblustre clients doing little filesystem IO
4603 Bugzilla : 9352, 7313
4604 Description: server may evict liblustre clients accessing contended locks
4605 Details : if a client is granted a lock or receives a completion AST
4606 with a blocking AST already set it would not reply to the AST
4607 for LDLM_FL_CANCEL_ON_BLOCK locks. It now replies to such ASTs.
4610 Frequency : lfs setstripe, only systems with more than 160 OSTs
4612 Description: unable to set striping with a starting offset beyond OST 160
4613 Details : llapi_create_file() incorrectly limited the starting stripe
4614 index to the maximum single-file stripe count.
4617 Frequency : LDAP users only
4619 Description: lconf did not handle in-kernel recovery with LDAP properly
4620 Details : lconf/LustreDB get_refs() is searching the wrong namespace
4622 Severity : enhancement
4624 Description: bind OST threads to NUMA nodes to improve performance
4625 Details : all OST threads are uniformly bound to CPUs on a single NUMA
4626 node and do their allocations there to localize memory access
4628 Severity : enhancement
4630 Description: llmount can determine client NID directly from Myrinet (GM)
4631 Details : the client NID code from gmnalnid was moved directly into
4632 llmount, removing the need to use this or specifying the
4633 client NID explicitly when mounting GM clients with zeroconf
4636 Frequency : if client is started with down MDS
4638 Description: if client is started with down MDS mount hangs in ptlrpc_queue_wait
4639 Details : Having an LWI_INTR() wait event (interruptible, but no timeout)
4640 will wait indefinitely in ptlrpc_queue_wait->l_wait_event() after
4641 ptlrpc_import_delayed_req() because we didn't check if the
4642 request was interrupted, and we also didn't break out of the
4643 event loop if there was no timeout
4648 Description: data loss during non-page-aligned writes to a single file from
4649 both multiple nodes and multiple threads on one node at same time
4650 Details : updates to KMS and lsm weren't protected by common lock. Resulting
4651 inconsistency led to false short-reads, that were cached and later
4652 used by ->prepare_write() to fill in partially written page,
4653 leading to data loss.
4656 Frequency : always, if lconf --abort_recovery used
4658 Description: lconf --abort_recovery fails with 'Operation not supported'
4659 Details : lconf was attempting to abort recovery on the MDT device and not
4662 Severity : enhancement
4664 Description: remove cleanup logs
4665 Details : replace lconf-generated cleanup logs with lustre internal
4666 cleanup routines. Eliminates the need for client-cleanup and
4669 Severity : enhancement
4671 Description: add support for EAs (user and system) on lustre filesystems
4672 Details : it is now possible to store extended attributes in the Lustre
4673 client filesystem, and with the user_xattr mount option it
4674 is possible to allow users to store EAs on their files also
4676 Severity : enhancement
4678 Description: Add possibility (config option) to show minimal available OST free
4680 Details : When compiled with --enable-mindf configure option, statfs(2)
4681 (and so, df) will return least minimal free space available from
4682 all OSTs as amount of free space on FS, instead of summary of
4683 free spaces of all OSTs.
4685 Severity : enhancement
4687 Description: do not expand extent locks acquired on OST-side
4688 Details : Modify ldlm_extent_policy() to not expand local locks, acquired
4689 by server: they are not cached anyway.
4692 Frequency : when mmap is used/binaries executed from Lustre
4694 Description: Unmmap pages before throwing them away from read cache.
4695 Details : llap_shrink cache now attempts to unmap pages before discarding
4696 them (if unmapping failed - do not discard). SLES9 kernel has
4697 extra checks that trigger if this unmapping is not done first.
4702 Description: lconf didn't resolve symlinks before checking to see whether a
4703 given mountpoint was already in use
4706 Frequency : when migrating failover services
4707 Bugzilla : 6395, 9514
4708 Description: When migrating a subset of services from a node (e.g. failback
4709 from a failover service node) the remaining services would
4710 time out and evict clients.
4711 Details : lconf --force (implied by --failover) sets the global obd_timeout
4712 to 5 seconds in order to quickly disconnect, but this caused
4713 other RPCs to time out too quickly. Do not change the global
4714 obd_timeout for force cleanup, only set it for DISCONNECT RPCs.
4716 Severity : enhancement
4717 Frequency : if MDS is started with down OST
4718 Bugzilla : 9439,5706
4719 Description: Allow startup/shutdown of an MDS without depending on the
4720 availability of the OSTs.
4721 Details : Asynchronously call mds_lov_synchronize during MDS startup.
4722 Add appropriate locking and lov-osc refcounts for safe
4723 cleaning. Add osc abort_inflight calls in case the
4724 synchronize never started.
4727 Frequency : occasional (Cray XT3 only)
4729 Description: root not authorized to access files in CRAY_PORTALS environment
4730 Details : The client process capabilities were not honoured on the MDS in
4731 a CRAY_PORTALS/CRAY_XT3 environment. If the file had previously
4732 been accessed by an authorized user then root was able to access
4733 the file on the local client also. The root user capabilities
4734 are now allowed on the MDS, as this environment has secure UID.
4737 Frequency : occasional
4739 Description: ldiskfs "too long searching" message happens too often
4740 Details : A debugging message (otherwise harmless) prints too often on
4741 the OST console. This has been reduced to only happen when
4742 there are fragmentation problems on the filesystem.
4747 Description: Division by zero in statfs when all OSCs are inactive
4748 Details : lov_get_stripecnt() returns zero due to incorrect order of checks,
4749 lov_statfs divides by value returned by lov_get_stripecnt().
4753 Bugzilla : 9489, 3273
4754 Description: First write from each client to each OST was only 4kB in size,
4755 to initialize client writeback cache, which caused sub-optimal
4756 RPCs and poor layout on disk for the first writen file.
4757 Details : Clients now request an initial cache grant at (re)connect time
4758 and so that they can start streaming writes to the cache right
4759 away and always do full-sized RPCs if there is enough data.
4760 If the OST is rebooted the client also re-establishes its grant
4761 so that client cached writes will be honoured under the grant.
4766 Description: Slow ls (and stat(2) syscall) on files residing on IO-loaded OSTs
4767 Details : Now I/O RPCs go to different portal number and (presumably) fast
4768 lock requests (and glimses) and other RPCs get their own service
4769 threads pool that should be able to service those RPCs
4772 Severity : enhancement
4774 Description: Ability to exchange lustre version between client and servers and
4775 issue warnings at client side if client is too old. Also for
4776 liblustre clients there is ability to refuse connection of too old
4778 Details : New 'version' field is added to connect data structure that is
4779 filled with version info. That info is later checked by server and
4783 Frequency : rare, liblustre only.
4784 Bugzilla : 9296, 9581
4785 Description: Two simultaneous writes from liblustre at offset within same page
4786 might proceed at the same time overwriting eachother with stale
4788 Details : I/O lock withing llu_file_prwv was released too early, before data
4789 actually was hitting the wire. Extended lock-holding time until
4790 server acknowledges receiving data.
4793 Frequency : extremely rare. Never observed in practice.
4795 Description: avoid generating lustre_handle cookie of 0.
4796 Details : class_handle_hash() generates handle cookies by incrementing
4797 global counter, and can hit 0 occasionaly (this is unlikely, but
4798 not impossible, because initial value of cookie counter is
4799 selected randonly). Value of 0 is used as a sentinel meaning
4800 "unassigned handle" --- avoid it. Also coalesce two critical
4801 sections in this function into one.
4803 Severity : enhancement
4805 Description: allow liblustre clients to delegate truncate locking to OST
4806 Details : To avoid overhead of locking, liblustre client instructs OST to
4807 take extent lock in ost_punch() on client's behalf. New connection
4808 flag is added to handle backward compatibility.
4810 Severity : enhancement
4811 Bugzilla : 4928, 7341, 9758
4812 Description: allow number of OST service threads to be specified
4813 Details : a module parameter allows the number of OST service threads
4814 to be specified via "options ost ost_num_threads={N}" in the
4815 OSS's /etc/modules.conf or /etc/modprobe.conf.
4819 Bugzilla : 6146, 9635, 9895
4820 Description: servers crash with bad pointer in target_handle_connect()
4821 Details : In rare cases when a client is reconnecting it was possible that
4822 the connection request was the last reference for that export.
4823 We would temporarily drop the export reference and get a new
4824 one, but this may have been the last reference and the export
4825 was just destroyed. Get new reference before dropping old one.
4827 Severity : enhancement
4828 Frequency : if client is started with failover MDS
4830 Description: Allow multiple MDS hostnames in the mount command
4831 Details : Try to read the configuration from all specified MDS
4832 hostnames during a client mount in case the "primary"
4835 Severity : enhancement
4837 Description: Stop sending data to evicted clients as soon as possible.
4838 Details : Check if the client we are about to send or are sending data to
4839 was evicted already. (Check is done every second of waiting,
4840 for which l_wait_event interface was extended to allow checking
4841 of exit condition at specified intervals).
4844 Frequency : rare, normally only when NFS exporting is done from client
4846 Description: 'bad disk LOV MAGIC: 0x00000000' error when chown'ing files
4848 Details : Make mds_get_md() recognise empty md case and set lmm size to 0.
4851 Frequency : always, if srand() is called before liblustre initialization
4853 Description: Liblustre uses system PRNG disturbing its usage by user application
4854 Details : Introduce internal to lustre fast and high-quality PRNG for
4855 lustre usage and make liblustre and some other places in generic
4856 lustre code to use it.
4858 Severity : enhancement
4859 Bugzilla : 9477, 9557, 9870
4860 Description: Verify that the MDS configuration logs are updated when xml is
4861 Details : Check if the .xml configuration logs are newer than the config
4862 logs stored on the MDS and report an error if this is the case.
4863 Request --write-conf, or allow starting with --old_conf.
4865 Severity : enhancement
4867 Description: Handle symlinks in the path when checking if Lustre is mounted.
4868 Details : Resolve intermediate symlinks when checking if a client has
4869 mounted a filesystem to avoid duplicate client mounts.
4874 Description: lconf can hit an error exception but still return success.
4875 Details : The lconf command catches the Command error exception at the top
4876 level script context and will exit with the associated exit
4877 status, but doesn't ensure that this exit status is non-zero.
4882 Description: failure of ptlrpc thread startup can cause oops
4883 Details : Starting a ptlrpc service thread can fail if there are a large
4884 number of threads or the server memory is very fragmented.
4885 Handle this without oopsing.
4888 Frequency : always, only if liblustre and non-default acceptor port was used
4890 Description: liblustre cannot connect to servers with non-default acceptor port
4891 Details : tcpnal_set_default_params() was not called and was therefore
4892 ignoring the environment varaible TCPNAL_PORT, as well as other
4893 TCPNAL_ environment variables
4898 Description: two objects could be created on the same OST for a single file
4899 Details : If an OST is down, in some cases it was possible to create two
4900 objects on a single OST for a single file. No problems other
4901 than potential performance impact and spurious error messages.
4905 Bugzilla : 5681, 9562
4906 Description: Client may oops in ll_unhash_aliases
4907 Details : Client dcache may become inconsistent in race condition.
4908 In some cases "getcwd" can fail if the current directory is
4914 Description: Inode refcounting problems in NFS export code
4915 Details : link_raw functions used to call d_instantiate without obtaining
4916 extra inode reference first.
4920 Bugzilla : 9942, 9903
4921 Description: Referencing freed requests leading to crash, memleaks with NFS.
4922 Details : We used to require that call to ll_revalidate_it was always
4923 followed by ll_lookup_it. Also with revalidate_special() it is
4924 possible to call ll_revalidate_it() twice for the same dentry
4925 even if first occurence returned success. This fix changes semantic
4926 between DISP_ENQ_COMPLETE disposition flag to mean there is extra
4927 reference on a request referred from the intent.
4928 ll_intent_release() then releases such a request.
4931 Frequency : rare, normally benchmark loads only
4933 Description: unlinked inodes were kept in memory on the client
4934 Details : If a client is repeatedly creating and unlinking files it
4935 can accumulate a lot of stale inodes in the inode slab cache.
4936 If there is no other client load running this can cause the
4937 client node to run out of memory. Instead flush old inodes
4938 from client cache that have the same inode number as a new inode.
4941 Frequency : SLES9 2.6.5 kernel and long filenames only
4942 Bugzilla : 9969, 10379
4943 Description: utime reports stale NFS file handle
4944 Details : SLES9 uses out-of-dentry names in some cases, which confused
4945 the lustre dentry revalidation. Change it to always use the
4949 Frequency : rare, unless heavy write-truncate concurrency is continuous
4950 Bugzilla : 4180, 6984, 7171, 9963, 9331
4951 Description: OST becomes very slow and/or deadlocked during object unlink
4952 Details : filter_destroy() was holding onto the parent directory lock
4953 while truncating+unlinking objects. For very large objects this
4954 may block other threads for a long time and slow overall OST
4955 responsiveness. It may also be possible to get a lock ordering
4956 deadlock in this case, or run out of journal credits because of
4957 the combined truncate+unlink. Solution is to do object truncate
4958 first in one transaction without parent lock, and then do the
4959 final unlink in a new transaction with the parent lock. This
4960 reduces the lock hold time dramatically.
4963 Frequency : rare, 2.4 kernels only
4965 Description: MDS or OST cleanup may trip kernel BUG when dropping kernel lock
4966 Details : mds_cleanup() and filter_cleanup() need to drop the kernel lock
4967 before unmounting their filesystem in order to avoid deadlock.
4968 The kernel_locked() function in 2.4 kernels only checks whether
4969 the kernel lock is held, not whether it is this process that is
4970 holding it as 2.6 kernels do.
4975 Description: MDS or OST may oops/LBUG if a client is connecting multiple times
4976 Details : The client ptlrpc code may be trying to reconnect to a down
4977 server before a previous connection attempt has timed out.
4978 Increase the reconnect interval to be longer than the connection
4979 timeout interval to avoid sending duplicate connections to
4983 Frequency : echo_client brw_test command
4985 Description: fix echo_client to work with OST preallocated code
4986 Details : OST preallocation code (5137) didn't take echo_client IO path
4987 into account: echo_client calls filter methods outside of any
4988 OST thread and, hence, there is no per-thread preallocated
4989 pages and buffers to use. Solution: hijack pga pages for IO. As
4990 a byproduct, this avoids unnecessary data copying.
4994 Bugzilla : 3555, 5962, 6025, 6155, 6296, 9574
4995 Description: Client can oops in mdc_commit_close() after open replay
4996 Details : It was possible for the MDS to return an open request with no
4997 transaction number in mds_finish_transno() if the client was
4998 evicted, but without actually returning an error. Clients
4999 would later try to replay that open and may trip an assertion
5000 Simplify the client close codepath, and always return an error
5001 from the MDS in case the open is not successful.
5004 Frequency : rare, 2.6 OSTs only
5006 Description: OST may deadlock under high load on fragmented files
5007 Details : If there was a heavy load and highly-fragmented OST filesystems
5008 it was possible to have all the OST threads deadlock waiting on
5009 allocation of biovecs, because the biovecs were not released
5010 until the entire RPC IO was completed. Instead, release biovecs
5011 as soon as they are complete to ensure forward IO progress.
5013 Severity : enhancement
5015 Description: Support for specifying external journal device at mount
5016 Details : If an OST or MDS device is formatted with an external journal
5017 device, this device major/minor is stored in the ext3 superblock
5018 and may not be valid for failover. Allow detecting and
5019 specifying the external journal at mount time.
5024 Description: Mounting an MDS with pending unlinked files may cause oops
5025 Details : target_finish_recovery() calls mds_postrecov() which returned
5026 the number of orphans unlinked. mds_lov_connect->mds_postsetup()
5027 considers this an error and immediately begins cleaning up the
5028 lov, just after starting the mds_lov process
5030 Severity : enhancement
5032 Description: Implement 'lfs df' to report actual free space on per-OST basis
5033 Details : Add sub-command 'df' on 'lfs' to report the disk space usage of
5034 MDS/OSDs. Usage: lfs df [-i][-h]. Command Options: '-i' to report
5035 usage of objects; '-h' to report in human readable format.
5037 ------------------------------------------------------------------------------
5039 2005-08-26 Cluster File Systems, Inc. <info@clusterfs.com>
5046 Description: Mounting an ldiskfs file system with mballoc may crash OST node.
5047 Details : ldiskfs mballoc code may reference an uninitialized buddy struct
5048 at startup during orphan unlinking. Instead, skip buddy update
5049 before setup, as it will be regenerated after recovery is complete.
5054 Description: If an OST is inactive, its locks might reference stale inodes.
5055 Details : lov_change_cbdata() must iterate over all namespaces, even if
5056 they are inactive to clear inode references from the lock.
5058 Severity : enhancement
5059 Frequency : occasional, if non-standard max_dirty_mb used
5061 Description: Client will block write RPCs if not enough grant
5062 Details : If a client has max_dirty_mb smaller than max_rpcs_in_flight,
5063 then the client will block writes while waiting for another RPC
5064 to complete instead of consuming its dirty limit. With change
5065 we get improved performance when max_dirty_mb is small.
5067 Severity : enhancement
5068 Bugzilla : 3389, 6253
5069 Description: Add support for supplementary groups on the MDS.
5070 Details : The MDS has an upcall /proc/fs/lustre/mds/{mds}/group_upcall
5071 (set to /usr/sbin/l_getgroups if enabled) which will do MDS-side
5072 lookups for user supplementary groups into a cache.
5076 Description: O_CREAT|O_EXCL open flags in liblustre always return -EEXIST
5077 Details : Make libsysio to not enforce O_EXCL by clearing the flag,
5078 for liblustre O_EXCL is enforced by MDS.
5082 Description: readdir never returns NULL in liblustre.
5083 Details : Corrected llu_iop_getdirentries logic, to return offset of next
5084 dentry in struct dirent.
5088 Frequency : liblustre only, depends on application IO pattern
5089 Description: liblustre clients evicted if not contacting servers
5090 Details : Don't put liblustre clients into the ping_evictor list, so
5091 they will not be evicted by the pinger ever.
5093 Severity : enhancement
5095 Description: Add ability to evict clients by NID from MDS.
5096 Details : By echoing "nid:$NID" string into
5097 /proc/fs/lustre/mds/.../evict_client client with nid that equals to
5098 $NID would be instantly evicted from this MDS and from all active
5099 OSTs connected to it.
5103 Description: Do not query file size twice, somewhat slowing stat(2) calls.
5104 Details : lookup_it_finish() used to query file size from OSTs that was not
5109 Description: service threads change working directory to that of init
5110 Details : Starting lustre service threads may pin the working directory
5111 of the parent thread, making that filesystem busy. Threads
5112 now change to the working directory of init to avoid this.
5116 Frequency : during shutdown only
5117 Description: shutdown with a failed MDS or OST can cause unmount to hang
5118 Details : Don't resend DISCONNECT messages in ptlrpc_disconnect_import()
5123 Frequency : 2.6 only
5124 Description: chmod/chown may include an extra supplementary group
5125 Details : ll{,u}_mdc_pack_op_data() does not properly initialize the
5126 supplementary group and if none is specified this is used.
5129 Bugzilla : 5479 (6816)
5131 Description: Racing open + rm can assert client in mdc_set_open_replay_data()
5132 Details : If lookup is in progress on a file that is unlinked we might try
5133 to revalidate the inode and fail in revalidate after lookup is
5134 complete and ll_file_open() enqueues the open again but
5135 it_open_error() was not checking DISP_OPEN_OPEN errors correctly.
5138 Frequency : always, if lconf --abort_recovery used
5140 Description: lconf --abort_recovery fails with 'Operation not supported'
5141 Details : lconf was attempting to abort recovery on the MDT device and not
5144 ------------------------------------------------------------------------------
5146 2005-08-08 Cluster File Systems, Inc. <info@clusterfs.com>
5151 Frequency : rare (only unsupported configurations with a node running as an
5153 Bugzilla : 6514, 5137
5154 Description: Mounting a Lustre file system on a node running as an OST could
5156 Details : OSTs now preallocates memory needed to write out data at
5157 startup, instead of when needed, to avoid having to
5158 allocate memory in possibly low memory situations.
5159 Specifically, if the file system is mounted on on OST,
5160 memory pressure could force it to try to write out data,
5161 which it needed to allocate memory to do. Due to the low
5162 memory, it would be unable to do so and the node would
5163 become unresponsive.
5165 Severity : enhancement
5167 Description: Addition of lconf --service command line option
5168 Details : lconf now accepts a '--service <arg>' option, which is
5169 shorthand for 'lconf --group <arg> --select <arg>=<hostname>'
5171 Severity : enhancement
5173 Description: Failover mode is now the default for OSTs.
5174 Details : By default, OSTs will now run in failover mode. To return to
5175 the old behaviour, add '--failout' to the lmc line for OSTs.
5177 Severity : enhancement
5179 Description: Health checks are now provided for MDS and OSTs
5180 Details : Additional detailed health check information on MSD and OSTs
5181 is now provided through the procfs health_check value.
5184 Frequency : occasional, depends on IO load
5186 Description: Disk fragmentation on the OSTs could eventually cause slowdowns
5187 after numerous create/delete cycles
5188 Details : The ext3 inode allocation policy would not allocate new inodes
5189 very well on the OSTs because there are no new directories
5190 being created. Instead we look for groups with free space if
5191 the parent directories are nearly full.
5196 Description: Network or server problems during mount may cause partially
5197 mounted clients instead of returning an error.
5198 Details : The config llog parsing code may overwrite the error return
5199 code during mount error handling, returning success instead
5205 Description: MDS can fail to allocate large reply buffers
5206 Details : After long uptimes the MDS can fail to allocate large reply
5207 buffers (e.g. zconf client mount config records) due to memory
5208 fragmentation or consumption by the buffer cache. Preallocate
5209 some large reply buffers so that these replies can be sent even
5210 under memory pressure.
5214 Frequency : rare (liblustre)
5215 Description: fsx running with liblustre complained that using truncate() to
5216 extend the file doesn't work. This patch corrects that issue.
5217 Details : This is the liblustre equivalent of the fix for bug 6196. Fixes
5218 ATTR_SIZE and lsm use in llu_setattr_raw.
5222 Frequency : rare, only 2.6 kernels
5223 Description: Unusual file access patterns on the MDS may result in inode
5224 data being lost in very rare circumstances.
5225 Details : Bad interaction between the ea-in-inode patch and the "no-read"
5226 code in the 2.6 kernel caused the inode and/or EA data not to
5227 be read from disk, causing single-file corruption.
5231 Frequency : rare, only 2.6 filesystems using extents
5232 Description: Heavy concurrent write and delete load may cause data corruption.
5233 Details : It was possible under high-load situations to have an extent
5234 metadata block in the block device cache from a just-unlinked
5235 file overwrite a newly-allocated data block. We now unmap any
5236 metadata buffers that alias just-allocated data blocks.
5240 Frequency : filesystems with default stripe_count larger than 77
5241 Description: lconf+mke2fs fail when formatting filesystem with > 77 stripes
5242 Details : lconf specifies an inode size of 4096 bytes when the default
5243 stripe_count is larger than 77. This conflicts with the default
5244 inode density of 1 per 4096 bytes. Allocate smaller inodes in
5245 this case to avoid pinning too much memory for large EAs.
5247 ------------------------------------------------------------------------------
5249 2005-07-07 Cluster File Systems, Inc. <info@clusterfs.com>
5254 Frequency : rare (extremely heavy IO load with hundreds of clients)
5256 Description: Client is evicted, gets IO error writing to file
5257 Details : lock ordering changes for bug 5492 reintroduced bug 3267 and
5258 caused clients to be evicted for AST timeouts. The fixes in
5259 bug 5192 mean we no longer need to have such short AST timeouts
5260 so ldlm_timeout has been increased.
5263 Frequency : occasional during --force or --failover shutdown under load
5264 Bugzilla : 5949, 4834
5265 Description: Server oops/LBUG if stopped with --force or --failover under load
5266 Details : a collection of import/export refcount and cleanup ordering
5267 issues fixed for safer force cleanup
5270 Frequency : only filesystems larger than 120 OSTs
5271 Bugzilla : 5990, 6223
5272 Description: lfs getstripe would oops on a very large filesystem
5273 Details : lov_getconfig used kfree on vmalloc'd memory
5276 Frequency : only filesystems exporting via NFS to Solaris 10 clients
5277 Bugzilla : 6242, 6243
5278 Description: reading from files that had been truncated to a non-zero size
5279 but never opened returned no data
5280 Details : ll_file_read() reads zeros from no-object files to EOF
5285 Description: A bug in MDS/OSS recovery could cause the OSS to fail an assertion
5286 Details : There's little harm in aborting MDS/OSS recovery and letting it
5287 try again, so I removed the LASSERT and return an error instead.
5289 Severity : enhancement
5291 Description: New debugging infrastructure for tracking down data corruption
5292 Details : The I/O checksum code was replaced to: (a) control it at runtime,
5293 (b) cover more of the client-side code path, and (c) try to narrow
5294 down where problems occurred
5298 Bugzilla : 3819, 4364, 4397, 6313
5299 Description: Racing close and eviction MDS could cause assertion in mds_close
5300 Details : It was possible to get multiple mfd references during close and
5301 client eviction, leading to one thread referencing a freed mfd.
5303 Severity: : enhancement
5304 Bugzilla : 3262, 6359
5305 Description: Attempts to reconnect to servers are now more aggressive.
5306 Details : This builds on the enhanced upcall-less recovery that was added
5307 in 1.4.2. When trying to reconnect to servers, clients will
5308 now try each server in the failover group every 10 seconds. By
5309 default, clients would previously try one server every 25 seconds.
5314 Description: After recovery, certain operations trigger a failed
5315 assertion on a client.
5316 Details : Failing over an mds, using lconf -d --failover, while a
5317 client was doing a readdir() call would cause the client to
5318 LBUG after recovery completed and the readdir() was resent.
5320 Severity : enhancement
5322 Description: Default groups are now added by lconf
5323 Details : You can now run lconf --group <servicename> without having to
5324 manually add groups with lmc.
5327 Frequency : occasional
5329 Description: Nodes with an elan id of 0 trigger a failed assertion
5332 Frequency : always when accessing e.g. tty/console device nodes
5334 Description: tty and some other devices nodes cannot be used on lustre
5335 Details : file's private_data field is used by device data and lustre
5336 values in there got lost. New field was added to struct file to
5337 store fs-specific private data.
5340 Frequency : when exporting Lustre via NFS
5342 Description: NFSD failed occasionally when looking up a path component
5343 Details : NFSD is looking up ".." which was broken in ext3 directories
5344 that had grown large enough to become hashed.
5347 Frequency : Clusters with multiple interfaces not on the same subnet
5349 Description: Nodes will repeatedly try to reconnect to an interface which it
5350 cannot reach and report an error to the log.
5351 Details : Extra peer list entries will be created by lconf with some peers
5352 unreachable. lconf now validates the peer before adding it.
5355 Frequency : Only if a default stripe is set on the filesystem root.
5357 Description: Setting a default stripe on the filesystem root prevented the
5358 filesystem from being remounted.
5359 Details : The client was sending extra request flags in the root getattr
5360 request and did not allocate a reply buffer for the dir EA.
5363 Frequency : occasional, higher if lots of files are accessed by one client
5364 Bugzilla : 6159, 6097
5365 Description: Client trips assertion regarding lsm mismatch/magic
5366 Details : While revalidating inodes the VFS looks up inodes with ifind()
5367 and in rare cases can find an inode that is being freed.
5368 The ll_test_inode() code will free the lsm during ifind()
5369 when it finds an existing inode and then the VFS later attaches
5370 this free lsm to a new inode.
5374 Bugzilla : 6422, 7030
5375 Description: MDS deadlock between mkdir and client eviction
5376 Details : Creating a new file via mkdir or mknod (starting a transaction
5377 and getting the ns lock) can deadlock with client eviction
5378 (gets ns lock and trying to finish a synchronous transaction).
5381 Frequency : occasional
5382 Description: While starting a server, the fsfilt_ext3 module could not be
5384 Details : CFS's improved ext3 filesystem is named ldiskfs for 2.6
5385 kernels. Previously, lconf would still use the ext3 name
5386 when trying to load modules. Now, it will correctly use
5387 ext3 on 2.4 and ldiskfs on 2.6.
5389 Severity : enhancement
5390 Description: The default stripe count has been changed to 1
5391 Details : The interpretation of the default stripe count (0, to lfs
5392 or lmc) has been changed to mean striping across a single
5393 OST, rather than all available. For general usage we have
5394 found a stripe count of 1 or 2 works best.
5396 Severity : enhancement
5397 Description: Add support for compiling against Cray portals.
5398 Details : Conditional compiling for some areas that are different
5402 Frequency : occasional
5403 Bugzilla : 6409, 6834
5404 Description: Creating files with an explicit stripe count may lead to
5405 a failed assertion on the MDS
5406 Details : If some OSTs are full or unavailable, creating files may
5407 trigger a failed assertion on the MDS. Now, Lustre will
5408 try to use other servers or return an error to the
5412 Frequency : occasional
5414 Description: Multiple concurrent overlapping read+write on multiple SMP nodes
5415 caused lock timeout during readahead (since 1.4.2).
5416 Details : Processes doing readahead might match a lock that hasn't been
5417 granted yet if there are overlapping and conflicting lock
5418 requests. The readahead process waits on ungranted lock
5419 (original lock is CBPENDING), while OST waits for that process
5420 to cancel CBPENDING read lock and eventually evicts client.
5422 Severity : enhancement
5424 Description: Initial enabling of flock support for clients
5425 Details : Implements fcntl advisory locking and file status functions.
5426 This feature is provided as an optional mount flag (default
5427 off), and is NOT CURRENTLY SUPPORTED. Not all types of record
5428 locking are implemented yet, and those that are are not guaranteed
5429 to be completely correct in production environments.
5430 mount -t lustre -o [flock|noflock] ...
5433 Frequency : occasional
5435 Description: OSTs running 2.4 kernels but with extents enabled might trip an
5436 assertion in the ext3 JBD (journaling) layer.
5437 Details : The b_committed_data struct is protected by the big kernel lock
5438 in 2.4 kernels, serializing journal_commit_transaction() and
5439 ext3_get_block_handle->ext3_new_block->find_next_usable_block()
5440 access to this struct. In 2.6 kernels there is finer grained
5441 locking to improve SMP performance of the JBD layer.
5445 Description: Changes the "SCSI I/O Stats" kernel patch to default to "enabled"
5447 -----------------------------------------------------------------------------
5449 2005-05-05 Cluster File Systems, Inc. <info@clusterfs.com>
5451 NOTE: Lustre 1.4.2 uses an incompatible network protocol than previous
5452 versions of Lustre. Please update all servers and clients to
5453 version 1.4.2 or later at the same time. You must also run
5454 "lconf --write-conf {config}.xml" on the MDS while it is stopped
5455 to update the configuration logs.
5457 - fix for HPUX NFS client breakage when NFS exporting Lustre (5781)
5458 - mdc_enqueue does not need max_mds_easize request buffer on send (5707)
5459 - swab llog records of type '0' so we get proper header size/idx (5861)
5460 - send llog cancel req to DLM cancel portal instead of cb portal (5515)
5461 - fix rename of one directory over another leaking an inode (5953)
5462 - avoid SetPageDirty on 2.6 (5981)
5463 - don't re-add just-being-destroyed locks to the waiting list (5653)
5464 - when creating new directories, inherit the parent's custom
5465 striping settings if present parent (3048)
5466 - flush buffers from cache before direct IO in 2.6 obdfilter (4982)
5467 - don't hold i_size_sem in ll_nopage() and ll_ap_refresh_count (6077)
5468 - don't hold client locks on temporary worklist from l_lru (5666)
5469 - handle IO errors in 2.6 obdfilter bio completion routine (6046)
5470 - automatically evict dead clients (5921)
5471 - Update file size properly in create+truncate+fstat case (6196)
5472 - Do not unhash mountpoint dentries, do not allow removal of
5474 - Avoid lock ordering deadlock issue with write/truncate (6203,5654)
5475 - reserve enough journal credits in fsfilt_start_log for setattr (4554)
5476 - ldlm_enqueue freed-export error path would always LBUG (6149,6184)
5477 - don't reference lr_lvb_data until after we hold lr_lvb_sem (6170)
5478 - don't overwrite last_rcvd if there is a *_client_add() error (6086)
5479 - Correctly handle reads of files with no objects (6243)
5480 - lctl recover will also mark a device active if deactivate used (5933)
5482 - by default create 1 inode per 4kB space on MDS, per 16kB on OSTs
5483 - allow --write-conf on an MDS with different nettype than client (5619)
5484 - don't write config llogs to MDS for mounts not from that MDS (5617)
5485 - lconf should create multiple TCP connections from a client (5201)
5486 - init scripts are now turned off by default; run chkconfig --on
5487 lustre and chkconfig --on lustrefs to use them
5488 - upcalls are no longer needed for clients to recover to failover
5490 - add --abort-recovery option to lconf to abort recovery on device
5492 - add support for an arbitrary number of OSTs (3026)
5493 - Quota support protocol changes.
5494 - forward compatibility changes to wire structs (6007)
5495 - rmmod NALs that might be loaded because of /etc/modules.conf (6133)
5496 - support for mountfsoptions and clientoptions to the Lustre LDAP (5873)
5497 - improved "lustre status" script
5498 - initialize blocksize for non-regular files (6062)
5499 - added --disable-server and --disable-client configure options (5782)
5500 - introduce a lookup cache for lconf to avoid repeated DB scans (6204)
5501 - Vanilla 2.4.29 support
5502 - increase maximum number of obd devices to 520 (6242)
5503 - remove the tcp-zero-copy patch from the suse-2.4 series (5902)
5504 - Quadrics Elan drivers are now included for the RHEL 3 2.4.21 and
5505 SLES 9 2.6.5 kernels
5506 - limit stripes per file to 160 (the maximum EA size) (6093)
5508 2005-03-22 Cluster File Systems, Inc. <info@clusterfs.com>
5511 - don't LASSERT in ll_release on NULL lld with NFS export (4655, 5760)
5512 - hold NS lock when calling handle_ast_error->del_waiting_lock (5746)
5513 - fix setattr mtime regression from lovcleanup merge (4829, 5669)
5514 - workaround for 2.6 crash in ll_unhash_aliases (5687, 5210)
5515 - small ext3 extents cleanups and fixes (5733)
5516 - improved mballoc code, several small races and bugs fixed (5733, 5638)
5517 - kernel version 43 - fix remove_suid bugs in both 2.4 and 2.6 (5695)
5518 - avoid needless client->OST connect, fix handle mismatch (5317)
5519 - fix DLM error path that led to out-of-sync client, long delays (5779)
5520 - support common vfs-enforced mount options (nodev,nosuid,noexec) (5637)
5521 - fix several locking issues related to i_size (5492,5624,5654,5672)
5522 - don't move pending lock onto export if it is already evicted (5683)
5523 - fix kernel oops when creating .foo in unlinked directory (5548)
5524 - fix deadlock in obdfilter statistics vs. object create (5811)
5525 - use time_{before,after} to avoid timer jiffies wrap (5882)
5526 - shutdown --force/--failover stability (3607,3651,4797,5203,4834)
5527 - Do not leak request if server was not able to process it (5154)
5528 - If mds_open unable to find parent dir, make that negative lookup(5154)
5529 - don't create new directories with extent-mapping (5909, 5936)
5531 - fix lustre/lustrefs init scripts for SuSE (patch from Scali, 5702)
5532 - don't hold the pinger_sem in ptlrpc_pinger_sending_on_import
5533 - change obd_increase_kms to obd_adjust_kms (up or down) (5654)
5534 - lconf, lmc search both /usr/lib and /usr/lib64 for Python libs (5800)
5535 - support for RHEL4 kernel on i686 (5773)
5536 - provide error messages when incompatible logs are encountered (5898)
5538 2005-02-18 Cluster File Systems, Inc. <info@clusterfs.com>
5539 * version 1.4.0.10 (1.4.1 release candidate 1)
5541 - don't keep a lock reference when lock is not granted (4238)
5542 - unsafe list practices (rarely) led to infinite eviction loop (4908)
5543 - add per-fs limit of Lustre pages in page cache, avoid OOM (4699)
5544 - drop import inflight refcount on signal_completed_replay error (5255)
5545 - unlock page after async write error during send (3677)
5546 - handle missing objects in filter_preprw_read properly (5265)
5547 - no transno return for symlink open, don't save no-trasno open (3440)
5548 - don't try to complete elan receive that already failed (4012)
5549 - free RPC server reply state on error (5406)
5550 - clean up thread from ptlrpc_start_thread() on error (5160)
5551 - readahead could read extra page into cache that wasn't ejected (5388)
5552 - prevent races in class_attach/setup/cleanup/detach (5260)
5553 - don't dereference de->d_inode after l_dput of de (5458)
5554 - use "int" for stripe value returned from lock_to_stripe (5544)
5555 - mballoc allocation and error-checking fixes in 2.6 (5504)
5556 - block device patches to fix I/O request sizes in 2.6 (5482)
5557 - look up hostnames for IB nals (5602)
5558 - 2.6 changed lock ordering of 2 semaphores, caused deadlock (5654)
5559 - don't start multiple acceptors for the same port (5277)
5560 - fix incorrect LASSERT in mds_getattr_name (5635)
5561 - export a proc file for general "ping" checking (5628)
5562 - fix "lfs check" to not block when the MDS is down (5628)
5564 - service request history (4965)
5565 - put {ll,lov,osc}_async_page structs in a single slab (4699)
5566 - create an "evict_client" /proc entry on OSTs, like the MDS has
5567 - fix mount usage message, return errors per mount(8) (5168)
5568 - change grep [] to grep "[]" in tests so they work in more UMLs
5569 - fix ppc64/x86_64 spec to use %{_libdir} instead of /usr/lib (5389)
5570 - remove ancient LOV_MAGIC_V0 EA support (5047)
5571 - add "disk I/Os in flight" and "I/O req time" stats in obdfilter
5572 - align r/w RPCs to PTLRPC_MAX_BRW_SIZE boundary for performance (3451)
5573 - allow readahead allocations to fail when low on memory (5383)
5574 - mmap locking landed again, after considerable improvement (2828)
5575 - add get_hostaddr() to lustreDB.py for LDAP support (5459)
5577 2004-11-23 Cluster File Systems, Inc. <info@clusterfs.com>
5580 - send OST transaction number in read/write reply to free req (4966)
5581 - don't ASSERT in ptl_send_rpc() if we run out of memory (5119)
5582 - lock /proc/sys/portals/routes internal state, avoiding oops (4827)
5583 - the watchdog thread now runs as interruptible (5246)
5584 - flock/lockf fixes (but it's still disabled, pending 5135)
5585 - don't use EXT3 constants in llite code (5094)
5586 - memory shortage at startup could cause assertion (5176)
5588 - reorganization of lov code
5589 - single portals codebase
5591 - add extents/mballoc support (5025)
5592 - direct I/O reads in the obdfilter (4048)
5593 - kernel patches from LNXI for 2.6 (bluesmoke, perfctr, mtd, kexec)
5595 tbd Cluster File Systems, Inc. <info@clusterfs.com>
5598 - send OST transaction number in read/write reply to free req (4966)
5599 - don't ASSERT in ptl_send_rpc() if we run out of memory (5119)
5600 - lock /proc/sys/portals/routes internal state, avoiding oops (4827)
5601 - the watchdog thread now runs as interruptible (5246)
5602 - handle missing objects in filter_preprw_read properly (5265)
5603 - unsafe list practices (rarely) led to infinite eviction loop (4908)
5604 - drop import inflight refcount on signal_completed_replay error (5255)
5605 - unlock page after async write error during send (3677)
5606 - return original error code on reconstructed replies (3761)
5607 - no transno return for symlink open, don't save no-trasno open (3440)
5609 - add pid to ldlm debugging output (4922)
5610 - bump the watchdog timeouts -- we can't handle 30sec yet
5611 - extra debugging for orphan dentry/inode bug (5259)
5613 2004-11-16 Cluster File Systems, Inc. <info@clusterfs.com>
5616 - fix TCP_NODELAY bug, which caused extreme perf regression (5134)
5617 - allocate qswnal tx descriptors singly to avoid fragmentation (4504)
5618 - don't LBUG on obdo_alloc() failure, use OBD_SLAB_ALLOC() (4800)
5619 - fix NULL dereference in /proc/sys/portals/routes (4827)
5620 - allow failed mdc_close() operations to be interrupted (4561)
5621 - stop precreate on OST before MDS would time out on it (4778)
5622 - don't send partial-page writes before EOF from client (4410)
5623 - discard client grant for sub-page writes on large-page clients (4520)
5624 - don't free dentries not owned by NFS code, check generation (4806)
5625 - fix lsm leak if mds_create_objects() fails (4801)
5626 - limit debug_daemon file size, always print CERROR messages (4789)
5627 - use transno after validating reply (3892)
5628 - process timed out requests if import state changes (3754)
5629 - update mtime on OST during writes, return in glimpse (4829)
5630 - add mkfsoptions to LDAP (4679)
5631 - use ->max_readahead method instead of zapping global ra (5039)
5632 - don't interrupt __l_wait_event() during strace
5634 - add software watchdogs to catch hung threads quickly (4941)
5635 - make lustrefs init script start after nfs is mounted
5636 - fix CWARN/ERROR duplication (4930)
5637 - return async write errors to application if possible (2248)
5638 - add /proc/sys/portal/memused (bytes allocated by PORTALS_ALLOC)
5639 - print NAL number in %x format (4645)
5640 - update barely-supported suse-2.4.21-171 series (4842)
5641 - support for sles 9 %post scripts
5642 - support for building 2.6 kernel-source packages
5643 - support for sles km_* packages
5645 2004-10-07 Cluster File Systems, Inc. <info@clusterfs.com>
5648 - ignore -ENOENT errors in osc_destroy (3639)
5649 - notify osc create thread that OSC is being cleaned up (4600)
5650 - add nettype argument for llmount in #5d in conf-sanity.sh (3936)
5651 - reconstruct ost_handle() like mds_handle() (4657)
5652 - create a new thread to do import eviction to avoid deadlock (3969)
5653 - let lconf resolve symlinked-to devices (4629)
5654 - don't unlink "objects" from directory with default EA (4554)
5655 - hold socknal file ref over connect in case target is down (4394)
5656 - allow more than 32000 subdirectories in a single directory (3244)
5657 - fix blocks count for O_DIRECT writes (3751)
5658 - OST returns ENOSPC from object create when no space left (4539)
5659 - don't send truncate RPC if file size isn't changing (4410)
5660 - limit OSC precreate to 1/2 of value OST considers bogus (4778)
5661 - bind to privileged port in socknal and tcpnal (3689)
5663 - rate limit CERROR/CWARN console message to avoid overload (4519)
5664 - GETFILEINFO dir ioctl returns LOV EA + MDS stat in 1 call (3327)
5665 - basic mmap support (3918)
5666 - kernel patch series update from b1_4 (4711)
5668 2004-09-16 Cluster File Systems, Inc. <info@clusterfs.com>
5671 - avoid crash during MDS cleanup with OST shut down (2775)
5672 - fix loi_list_lock/oig_lock inversion on interrupted IO (4136)
5673 - don't use bad inodes on the MDS (3744)
5674 - dynamic object preallocation to improve recovery speed (4236)
5675 - don't hold spinlock over lock dumping or change debug flags (4401)
5676 - don't zero obd_dev when it is force cleaned (3651)
5677 - print grants to console if they go negative (4431)
5678 - "lctl deactivate" will stop automatic recovery attempts (3406)
5679 - look for existing locks in ldlm_handle_enqueue() (3764)
5680 - don't resolve lock handle twice in recovery avoiding race (4401)
5681 - revalidate should check working dir is a directory (4134)
5683 - don't always mark "slow" obdfilter messages as errors (4418)
5685 2004-08-24 Cluster File Systems, Inc. <info@clusterfs.com>
5688 - don't close LustreDB during write_conf until it is done (3860)
5689 - fix typo in lconf for_each_profile (3821)
5690 - allow dumping logs from multiple threads at one time (3820)
5691 - don't allow multiple threads in OSC recovery (3812)
5692 - fix debug_size parameters (3864)
5693 - fix mds_postrecov to initialize import for llog ctxt (3121)
5694 - replace config semaphore with spinlock (3306)
5695 - be sure to send a reply for a CANCEL rpc with bad export (3863)
5696 - don't allow enqueue to complete on a destroyed export (3822)
5697 - down write_lock before checking llog header bitmap (3825)
5698 - recover from lock replay timeout (3764)
5699 - up llog sem before sending rpc (3652)
5700 - reduce ns lock hold times when setting kms (3267)
5701 - change a dlm LBUG to LASSERTF, to maybe learn something (4228)
5702 - fix NULL deref and obd_dev leak on setup error (3312)
5703 - replace some LBUG about llog ops with error handling (3841)
5704 - don't match INVALID dentries from d_lookup and spin (3784)
5705 - hold dcache_lock while marking dentries INVALID and hashing (4255)
5706 - fix invalid assertion in ptlrpc_set_wait (3880)
5708 - add libwrap support for the TCP acceptor (3996)
5709 - add /proc/sys/portals/routes for non-root route listing (3994)
5710 - allow setting MDS UUID in .xml (2580)
5711 - print the stack of a process that LBUGs (4228)
5713 2004-07-14 Cluster File Systems, Inc. <info@clusterfs.com>
5716 - don't cleanup request in ll_file_open() on failed MDS open (3430)
5717 - make sure to unset replay flag from failed open requests (3440)
5718 - if default stripe count is 0, use OST count for inode size (3636)
5719 - update parent mtime/ctime on client for create/unlink (2611)
5720 - drop dentry ref in ext3_add_link from open_connect_dentry (3266)
5721 - free recovery state on server during a forced cleanup (3571)
5722 - unregister_reply for resent reqs (3063)
5723 - loop back devices mounting and status check on 2.6 (3563)
5724 - fix resource-creation race that can provoke i_size == 0 (3513)
5725 - don't try to use bad inodes returned from MDS/OST fs lookup (3688)
5726 - more debugging for page-accounting assertion (3746)
5727 - return -ENOENT instead of asserting if ost getattr+unlink race (3558)
5728 - avoid deadlock after precreation failure (3758)
5729 - fix race and lock order deadlock in orphan handling (3450, 3750)
5730 - add validity checks when grabbing inodes from l_ast_data (3599)
5732 - add /proc/.../recovery_status to obdfilter (3428)
5733 - lightweight CDEBUG infrastructure, debug daemon (3668)
5734 - change default OSC RPC parameters to be better on small clusters
5735 - turn off OST read cache for files smaller than 32MB
5736 - install man pages and include them in rpms (3100)
5737 - add new init script for (un)mounting lustre filesystems (2593)
5738 - run chkconfig in %post for init scripts (3701)
5739 - drop scimac NAL (unmaintained)
5741 2004-06-17 Cluster File Systems, Inc. <info@clusterfs.com>
5744 - clean kiobufs before and after use (3485)
5745 - strip trailing '/'s before comparing paths with /proc/mounts (3486)
5746 - remove assertions to work around "in-flight rpcs" recovery bug (3063)
5747 - change init script to fail more clearly if not run as root (1528)
5748 - allow clients to reconnect during replay (1742)
5749 - fix ns_lock/i_sem lock ordering deadlock for kms update (3477)
5750 - don't do DNS lookups on NIDs too small for IP addresses (3442)
5751 - re-awaken ptlrpcd if new requests arrive during check_set (3554)
5752 - fix cond_resched (3554)
5753 - only evict unfinished clients after recovery (3515)
5754 - allow bulk resend, prevent data loss (3570)
5755 - dynamic ptlrpc request buffer allocation (2102)
5756 - don't allow unlinking open directory if it isn't empty (2904)
5757 - set MDS/OST threads to umask 0 to not clobber client modes (3359)
5758 - remove extraneous obd dereference causing LASSERT failure (3334)
5759 - don't use get_cycles() when creating temp. files on the mds (3156)
5760 - hold i_sem when setting i_size in ll_extent_lock() (3564)
5761 - handle EEXIST for set-stripe, set proper directory name (3336)
5763 - servers can dump a log evicting a client - lustre.dump_on_timeout=1
5764 - fix ksocknal_fmb_callback() error messages (2918)
5766 2004-05-27 Cluster File Systems, Inc. <info@clusterfs.com>
5769 - don't copy lvb into (possibly NULL) reply on error (2983)
5770 - don't deref dentry after dput, don't free lvb on error (2922)
5771 - use the kms to determine writeback rpc length (2947)
5772 - increment oti_logcookies when osc is inactive (2948)
5773 - update client's i_blocks count via lvb messages (2543)
5774 - handle intent open/close of special files properly (1557)
5775 - mount MDS with errors=remount-ro, like obdfilter (2009)
5776 - initialize lock handle to avoid ASSERT on error cleanup (3057)
5777 - don't use cancelling-locks' kms values (2947)
5778 - use highest lock extent for kms, not last one (2925)
5779 - don't dereference ERR_PTR() dentry in error handling path (3107)
5780 - fix thread race in portals_debug_dumplog() (3122)
5781 - create lprocfs device entries at setup instead of at attach (1519)
5782 - common AST error handler, don't evict client on completion race (3145)
5783 - zero nameidata in detach_mnt in 2.6 (3118)
5784 - verify d_inode after revalidate_special is valid in 2.6 (3116)
5785 - use lustre_put_super() to handle zconf unmounts in 2.6 (3064)
5786 - initialize RPC timeout timer earlier for 2.6 (3219)
5787 - don't dereference NULL reply buffer if mdc_close was never sent (2410)
5788 - print nal/nid for unknown nid (3258)
5789 - additional checks for oscc recovery before doing precreate (3284)
5790 - fix ll_extent_lock() error return code for 64-bit systems (3043)
5791 - don't crash in mdc_close for bad permissions on open (3285)
5792 - zero i_rdev for non-device files (3147)
5793 - clear page->private before handing to FS, better assertion (3119)
5794 - tune the read pipeline (3236)
5795 - fix incorrect decref of invalidated dentry (2350)
5796 - provide read-ahead stats and refine rpc in flight stats (3328)
5797 - don't hold journal transaction open across create RPC (3313)
5798 - update atime on MDS at close time (3265)
5799 - close LDAP connection when recovering to avoid server load (3315)
5800 - update iopen-2.6 patch with fixes from 2399,2517,2904 (3301)
5801 - don't leak open file on MDS after open resend (3325)
5802 - serialize filter_precreate and filter_destroy_precreated (3329)
5803 - loop device shouldn't call sync_dev() for nul device (3092)
5804 - clear page cache after eviction (2766)
5805 - resynchronize MDS->OST in background (2824)
5806 - refuse to mount the same filesystem twice on same mountpoint (3394)
5807 - allow llmount to create routes for mounting behind routers (3320)
5808 - push lock cancellation to blocking thread for glimpse ASTs (3409)
5809 - don't call osc_set_data_with_check() for TEST_LOCK matches (3159)
5810 - fix rare problem with rename on htree directories (3417)
5812 - allow default OST striping configuration per directory (1414)
5813 - fix compilation for qswnal for 2.6 kernels (3125)
5814 - increase maximum number of MDS request buffers for large systems
5815 - change liblustreapi to be useful for external progs like lfsck (3098)
5816 - increase local configuration timeout for slow disks (3353)
5817 - allow configuring ldlm AST timeout - lustre.ldlm_timeout=<seconds>
5819 2004-03-22 Cluster File Systems, Inc. <info@clusterfs.com>
5822 - fixes for glimpse AST timeouts / incorrectly 0-sized files (2818)
5823 - don't overwrite extent policy data in reply if lock was blocked (2901)
5824 - drop filter export grants atomically with removal from device (2663)
5825 - del obd_self_export from work_list in class_disconnect_exports (2908)
5826 - don't LBUG if MDS recovery times out during orphan cleanup (2530)
5827 - swab reply message in mdc_close, other PPC fixes (2464)
5828 - fix destroying of named logs (2325)
5829 - overwrite old logs when running lconf --write_conf (2264)
5830 - bump LLOG_CHUNKSIZE to 8k to allow for larger clusters (2306)
5831 - fix race in target_handle_connect (2898)
5832 - mds_reint_create() should take same inode create lock (2926)
5833 - correct journal credits calculated for CANCEL_UNLINK_LOG (2931)
5834 - don't close files for self_export to avoid uninitialized obd (2936)
5835 - allow MDS with the same name as client node (2939)
5836 - hold dentry reference for closed log files for unlink (2325)
5837 - reserve space for all logs during transactions (2059)
5838 - don't evict page beyond end of stripe extent (2925)
5839 - don't oops on a deleted current working directory (2399)
5840 - handle hard links to targets without a parent properly (2517)
5841 - don't dereference NULL lock when racing during eviction (2867)
5842 - don't grow lock extents when lots of conflicting locks (2919)
5844 2004-03-04 Cluster File Systems, Inc. <info@clusterfs.com>
5847 - account for cache space usage on clients to avoid data loss (974)
5848 - lfsck support in lustre kernel code (2349)
5849 - reduce journal credits needed for BRW writes (2370)
5850 - orphan handling to avoid losing space on client/server crashes
5851 - ptlrpcd can be blocked, stopping ALL progress (2477)
5852 - use lock value blocks to assist in proper KMS, faster stat (1021)
5853 - takes i_sem instead of DLM locks internally on obdfilter (2720)
5854 - recovery for initial connections (2355)
5855 - fixes for mds_cleanup_orphans (1934)
5856 - abort_recovery crashes MDS in b_eq (mds_unlink_orphan) (2584)
5857 - block all file creations until orphan recovery completes (1901)
5858 - client remove rq_connection from request struct (2423)
5859 - conf-sanity test_5, proper cleanup in umount log not availale (2640)
5860 - recovery timer race (2670)
5861 - mdc_close recovey bug (2532)
5862 - ptlrpc cleanup bug (2710)
5863 - mds timeout on local locks (2588)
5864 - namespace lock held during RPCs (2431)
5865 - handle interrupted sync write properly (2503)
5866 - don't try to handle a message that hasn't been replied to (2699)
5867 - client assert failure during cleanup after abort recovery (2701)
5868 - leak mdc device after failed mount (2712)
5869 - ptlrpc_check_set allows timedout requests to complete (2714)
5870 - wait for inflight reqs when ptlrpcd finishes (2710)
5871 - make sure unregistered services are removed from the srv_list
5872 - reset bulk XID's when resending them (caught by 1138 test)
5873 - unregister_bulk after timeout
5874 - fix lconf error (2694)
5875 - handle write after unfinished setstripe, stripe-only getstripe (2388)
5876 - readahead locks pages, leaves pending causing memory pressure (2673)
5877 - increase OST request buffers to 4096 on large machines (2729)
5878 - fix up permission of existing directories in simple_mkdir (2661)
5879 - init deleted item, add assertions ptlrpc_abort_inflight() (2725)
5880 - don't assign transno to errored transactions (2742)
5881 - don't delete objects on OST if given a bogus objid from MDS (2751)
5882 - handle large client PAGE_SIZE readdir on small PAGE_SIZE MDS (2777)
5883 - if rq_no_resend, then timeout request after recovery (2432)
5884 - fix MDS llog_logid record size, 64-bit array alignment (2733)
5885 - don't call usermode_helper from ptlrpcd, DEFAULT upcall (2773)
5886 - put magic in mount.lustre data, check for bad/NULL mount data (2529)
5887 - MDS recovery shouldn't delete objects that it has given out (2730)
5888 - if enqueue arrives after completion, don't clobber LVB (2819)
5889 - don't unlock pages twice when trigger_group_io returns error (2814)
5890 - don't deref NULL rq_repmsg if ldlm_handle_enqueue failed (2822)
5891 - don't write pages to disk if there was an error (1450)
5892 - don't ping imports that have recovery disabled (2676)
5893 - take buffered bytes into account when balancing socknal conn (2817)
5894 - hold a DLM lock over readdir always, use truncate_inode_pages (2706)
5895 - reconnect unlink llog connection after MDS reconnects to OST (2816)
5896 - remove little-endian swabbing of llog records (1987)
5897 - set/limit i_blksize to LL_MAX_BLKSIZE on client (2884)
5898 - retry reposting request buffers if they fail (1191)
5899 - grow extent at grant time to avoid granting a revoked lock (2809)
5900 - lock revoke doesn't evict page if covered by a second lock (2765)
5901 - disable VM readahead to avoid reading outside lock extents (2805)
5903 - return LL_SUPER_MAGIC from statfs for the filesystem type (1972)
5904 - updated kernel patches for hp-2.4.20 kernel (2681)
5906 2004-02-07 Cluster File Systems, Inc. <info@clusterfs.com>
5909 - fix truncated write corruption (2366)
5910 - fix for failed assertion in iopen_connect_dentry (1792,2517)
5912 - don't flag the ptlrpcd thread with PF_MEMALLOC (2636)
5913 - ensure len(uuid) < 37 in lmc (1171)
5914 - fix ia64 OOPS in llog_test (2255)
5915 - zero end of page at obdfilter for partial page writes (2648)
5916 - don't leave stale dentries around after renames (bug 2428)
5917 - fix timeouts when evicting a client with a single lock held (2642)
5918 - set deadline for the initial HELLO message to drain (2634)
5919 - print out dotted-quad IP addresses in the socknal (2302)
5921 - additional debugging for MDS client eviction problem (2443)
5922 - fix mkfsoptions support for osts (2603, 2604)
5924 2004-01-27 Cluster File Systems, Inc. <info@clusterfs.com>
5927 - add series for the vanilla 2.6.0 kernel
5928 - add series for the vanilla 2.4.24 kernel
5929 - add series for a cray x86/64 UL kernel drop
5930 - fix xattr patches for the vanilla 2.4.19 series
5932 - generate true UUIDs in lmc (1171)
5933 - have portals stack dumping break in UML (2466)
5934 - avoid bad dchild deref; avoid inum lock w/o creation (2362)
5935 - allocate with _NOFS in ldlm to avoid deadlock (1933)
5936 - wake callback waiting threads on client eviction (2460)
5937 - Add --ptldebug and --subsystem to lmc (1719)
5938 - update assertion to allow safe interrupt allocation
5939 - set rq_no_resend for cancel requests (2432)
5940 - recalculate ptlrpcd timeout after resend (2494)
5941 - call vfs_rmdir when removing pending directories (2368)
5942 - fix renaming a file to itself (2429)
5943 - lmc creates a default one-stripe lov (2454)
5944 - expand procfs space to handle large clusters (2326)
5945 - increase UML stack to avoid overflow
5946 - update lconf's list of debug and subsystem masks
5947 - fix lfs find --obd (2510)
5948 - /proc tunable for disabling filter read caching (2591)
5949 - stop rpm packages from altering slapd.conf (2301)
5950 - disable nagle in the socknal under 0conf (2578)
5951 - choose mds inode size based on stripe count (2572)
5952 - fix kernel-source rpm problems (2516)
5954 - add --disable-doc to avoid pdf generation (2421)
5955 - update documentation, tests, type-os, comments
5956 - avoid format warnings on ia64
5957 - remove the TOE NAL
5958 - tiny code cleanups by removing unused fields
5960 2004-01-07 Cluster File Systems, Inc. <info@clusterfs.com>
5963 - fix obvious semaphore misuse in as-yet-unused setattr path (2348)
5964 - remove the most blatant lies from BUILDING file (2371)
5965 - change default debug level to reasonable production setting
5966 - reduce client side cache size to reduce cache flush time
5967 - reduce max RPCs in flight to avoid unnecessary file fragmentation
5968 - make TCP zerocopy and pinger support enabled by default (2476)
5969 - sync writes completed after process exits caused crashes (2319)
5970 - maintain correct mount count on the MDS (2356)
5971 - backout 1557, because 2316 wasn't really fixed
5972 - better file I/O statistics gathering in /proc
5973 - don't take unnecessary, deadlock-inducing bug in readpage (2383)
5974 - another kernel patch to fix zero-copy TCP function export
5975 - don't take duplicate lock when processing re-sent getattr (2420)
5976 - lctl uses obd_self_export instead of creating new conn (2353)
5977 - MDS/OST recovery case which requires object creation asserted (2425)
5978 - move lfs from /usr/sbin to /usr/bin in packages
5979 - fix race between mds_client_add and mds_client_free (2417)
5980 - use kmalloc instead of slabs in portals (2430)
5981 - don't create duplicate records when a failover MDS is present (2442)
5982 - remove unnecessary mount age check (2332)
5983 - don't remove directory inodes from locks prematurely (2451)
5984 - don't break if MDS service name is the same as hostname (2103)
5985 - fix races in client write RPC generation when cache full (2482)
5987 2003-12-13 Cluster File Systems, Inc. <info@clusterfs.com>
5990 - remove now-unused request->rq_obd (278)
5991 - if an allocation fails, print out how much memory we've used (1933)
5992 - use PORTAL_SLAB_ALLOC for structures, to get GFP_MEMALLOC (1933)
5993 - add the "configurable stack size" patch to most series files (1256)
5994 - ability to write large log records, for 100+ OST configs (2306)
5995 - fix NULL deref when filter_prep fails (2314)
5996 - fix operator precedence error in filter_sync
5997 - dynamic allocation of socknal TX descriptors (2315)
5998 - fix a missed case in the GFP_MEMALLOC patch, can cause deadlock (2310)
5999 - fix gcc 2.96 compilation problem in xattr kernel patch (2294)
6000 - ensure that CWARN messages in Portals always get to the syslog
6001 - __init/__exit are not for prototype decls (ldlm_init/exit)
6002 - x86-64 compile warning fixes
6003 - fix gateway LMC keyword conflict (2318)
6004 - fix MDS lock inversions in getattr/reint paths (1844)
6005 - fix a rare lock re-ordering bug, which caused deadlock (2322)
6006 - fix i_sem/journal inversion in fsfilt_ext3_write_record (2306)
6007 - DLM race condition prevented some lock evictions (2328)
6008 - ENOMEM detection and retry on socknal sends (2230)
6009 - use GFP_NOFS throughout Lustre, to combat ENOMEM (2230)
6010 - move osc_rpcd into ptlrpc, for use in MDC and others (2329)
6011 - protect MDS inode fsdata with stronger locking; fixes assertion (2313)
6012 - better error messages when a client is rejected during recovery (1505)
6013 - avoid cancelling locks which were never granted, after failure (2330)
6014 - fix i_sem/journal inversion in mds_client_add (2333)
6015 - fix truncate/getattr lock cycle deadlock (2334)
6016 - use rpcd to send close; allows resend after timeout, avoid leak (1897)
6017 - fix two rare exit paths which could leak an l_lock() ref (2321)
6018 - fencepost error in MDS/OST orphan recovery (2226)
6019 - make log record alignment 8 bytes (1988)
6020 - lstripe now fails when requested offset > ost_count (2237)
6021 - ensure that all kernel series have a complete list.h (1607)
6022 - fix crashes in special-file operations (2316)
6023 - lctl create/brw OID mismatch, caused by obsolete filter loop (2339)
6025 - allow configurable automake binary, for testing new versions
6026 - small update to the lfs documentation
6028 2003-12-03 Cluster File Systems, Inc. <info@clusterfs.com>
6030 * fix negative export reference count in fsfilt_sync (2312)
6032 2003-12-01 Cluster File Systems, Inc. <info@clusterfs.com>
6033 * release candidate 0.9.1
6035 - orphans are moved into the PENDING directory for possible recovery
6036 - replayed opens now open by fid for orphan/rename safety (1042)
6037 - last close of an orphan inode generates a transno (683)
6038 - chdir() and mount() now pin the directory entry (1020)
6039 - avoid CERROR in normal ll_setattr_raw() error case (1500)
6040 - discard very old requests without processing them (1502)
6041 - remove some common, well-understood CERRORs (1505)
6042 - require O_DIRECT I/O to be page-sized to workaround IA64 crash (1609)
6043 - clear "grant" flags in OST replies until OST grant code lands (1644)
6044 - fix read performance by not clobbering i_blksize on client (1598)
6045 - fix __ldlm_handle2lock oops by not dereferencing lock after PUT (1625)
6046 - make LRU size a /proc tunable, clears locks when reduced (707)
6047 - fix some lprocfs rot that prevented ptlbd from loading (1732)
6048 - server locks take references on exports now (1558)
6049 - build fixes for 2.4.20-rh trees (1663)
6050 - return an error from lov_create if all OSCs are inactive (1751)
6051 - fix import levels when a reconnect happens without a timeout (1597)
6052 - exit early from mds_open if we get a lookup error (1749)
6053 - partial page read at EOF wouldn't wait for disk before sending (1642)
6054 - avoid NULL deref in obdfilter when reading page past EOF (1592)
6055 - avoid LASSERT in ll_intent_lock if server failed very early (1090)
6056 - fix LBUG in ll_it_open_error with rc = -2 (1861)
6057 - write/truncate lock inversion (1639)
6058 - Don't auto-load obdclass, portals modules during cleanup (1495)
6059 - fix timestamps from jumping to "now" (1763)
6060 - extra journal assertions (1648)
6061 - add an extra multiunlink test (1771)
6062 - fix read_record/write_record API (1776)
6063 - fix leak of offset_extent, possible incorrect i_size later (1772)
6064 - fix lasserts in mis-matched transnos during open-unlink testing (1541)
6065 - Debugging for the kqswnal_get_idle_tx problems (1820)
6066 - Allow recovery to be attempted multiple times (1536)
6067 - Write out MDS last_rcvd file after it is first created (1600)
6068 - Fix tx_descriptor leak in failed transmit situations (1827)
6069 - ext3 journaling fixes for assertion failure after IO error (1871)
6070 - class_export_put() on freed export after completion AST error (1896)
6071 - Fix revalidate looping in VFS (1322)
6072 - Don't access a freed export during MDS_REINT timeout (1521)
6073 - Add open-unlink recovery support on the MDS (1673,1764)
6074 - Return an error if no MDS data was read from last_rcvd (1946)
6075 - Fix for lookup "." or ".." crash on error (1932,1931,1935)
6076 - Don't setup a disk device that doesn't match exported UUID (317)
6077 - Reduce bulk RPC timeout to avoid cascading client/OST failures (1845)
6078 - avoid committing NULL handle in force close
6079 - local.sh is now a one-stripe LOV configuration
6080 - POSIX utime.4 -EPERM on FIFO not owned by user (56)
6081 - fix ext3 htree duplicate directory entry corruption (1516)
6082 - POSIX creat.13, fstat.1, open.18, stat.3 new file atime/mtime (2020)
6083 - update to new LOV EA format (2097)
6084 - interoperability for different PAGE_SIZE/wordsize (686,1821,1343,2042)
6086 2003-06-15 Phil Schwan <phil@clusterfs.com>
6089 - imports and exports cleanup too early, need refcounts (349, 879, 1045)
6090 - per-import/export recovery handling (958, 931, 959)
6091 - multiple last-rcvd slots, for serving multiple FSes (949)
6092 - connections are again shared between multiple imp/exports (963, 964)
6093 - "umount -f" would hang if any requests needed to be sent (393, 978)
6094 - avoid pinning large req buffer by copying for queued messages (989)
6095 - add "uuid" to "lctl device" command to help upcalls (991)
6096 - "open" RPCs with transnos would confuse recovery counters (1037)
6097 - do proper endian conversion of all wire messages (288, 340, 891)
6098 - remove OST bulk get LBUGs, fix ost_brw_write cleanup (1126)
6099 - call waiting locks callback from LDLM recovery thread (1127, 1151)
6100 - fix ptlrpc_connection leak in target_handle_connect (1174)
6101 - fix import refcounting bug in OST and MDS cleanup (1134)
6102 - if an invalid-at-open-time OSC returned before close(), LBUG (1150)
6103 - fix very unlikely obd_types race condition (501)
6104 - remove osc_open hack for echo_client (1187)
6105 - we leaked exports/dlmimps for forcibly disconnected clients (1143)
6106 - a failure in read_inode2 leads to deadlock (1139)
6107 - cancel ack-locks as soon as transaction is committed (1072)
6108 - fix major leaks and crashes in the bulk I/O path (937, 1057)
6109 - make sure to commitrw after any preprw to avoid deadlock (1162)
6110 - failing to execute a file in a lustre FS would lock inode (1203)
6111 - small DEBUG_REQ fix to avoid dereferencing a NULL (1227)
6112 - don't ASSERT while cleaning up an incompletely-setup obd (1248)
6113 - obd_uuid2tgt would walk off the end of the list (1255)
6114 - on IA64 the osc would give portals incorrect bulk size (1258)
6115 - fix debug daemon ioctl interface; allows daemon on ia64 (1274)
6116 - fix lock inversion caused by new llite matching code (1282)
6117 - limit the number of dirty pages on a client to 10MB (1286)
6118 - timed out locks were not being corrected cancelled (1289)
6119 - fix O_DIRECT above 4GB on IA-32 (1292)
6120 * major user-visible changes
6121 - fail out/fail over policy now controlled by the upcall (993)
6123 - add OBD_PING to check server availability and failure (954)
6124 - lustre messages are now sent in sending host order (288, 340, 891)
6125 - add eadatalen to MDS getattr reply (340)
6126 - OST read replies may contain second buffer, with per-page status (593)
6128 2003-03-11 Phil Schwan <phil@clusterfs.com>
6131 - LDLM_DEBUG macro fix, for gcc 3.2 (850)
6132 - failed open()s could cause deadlock; fixed (867, 869)
6133 - stop cancelling OST locks when files are closed (481)
6134 - overlapping XID spaces caused network corruption (851, 853)
6135 - fix unsafe fsfilt counter arithmetic; change to atomic_t
6136 - setattr_raw added, to do single-RPC, server-side setattrs
6137 - lmc/lconf syntax change for OST UUIDs
6138 - fix crashy race condition between ptlrpc_free_req and osc_close
6139 - don't use request in mdc_enqueue if we hit a timeout (889)
6140 - don't set the inode i_size for regular files from the MDS (896)
6141 - handle out of order completion AST (842)
6142 - don't LBUG if a lock request times out after receiving AST (913)
6143 - avoid d_rehash race in ll_find_alias by rehashing inside dcache_lock
6144 - if a bad lock AST arrives, send an error instead of dropping entirely
6145 - return 0 from revalidate2 if ll_intent_lock returns -EINTR (912)
6146 - fix leak in bulk IO when only partially completed (899, 900, 926)
6147 - fix O_DIRECT for ia64 (55)
6148 - (almost) eliminate Lustre-kernel-thread effects on load average (722)
6149 - C-z after timeout could hang a process forever; fixed (977)
6151 - client-side I/O cache (678, 924, 929, 941, 970)
6153 - READPAGE and SETATTRs which don't take server-side locks get
6156 2003-02-11 Phil Schwan <phil@clusterfs.com>
6159 - Fix ldlm_lock_match on the MDS to avoid matching remote locks (592)
6160 - Fix fsfilt_extN_readpage() to read a full page of directory
6161 entries, or fake the remainder if PAGE_SIZE != blocksize (500)
6162 - Avoid extra mdc_getattr() in ll_intent_lock when possible (534, 604)
6163 - Fix imbalanced LOV object allocation and out-of-bound access (469)
6164 - Most intent operations were removed, in favour of a new RPC mode
6165 that does a single RPC to the server and bypasses most of the VFS
6166 - All LDLM resource ID arrays were removed in favour of ldlm_res_id
6167 - Aggressively cancel local locks on DLM servers
6168 - mds_reint_unlink sends EA to the client if it's the last nlink.
6169 client uses that EA to unlink OST objects.
6170 - mds_reint_{rename,unlink,link} were rewritten to take ordered locks
6171 - recursive symlinks were fixed (439)
6172 - fixed NULL deref in DEBUG_REQ
6173 - filter_update_lastobjid no longer calls sync, which annoyed extN
6174 - fixed multi-client small-writes to a single file problem (445)
6175 - fixed mtime updates during file writes (607)
6176 - fixed vector writes on obdfilter causing problems when ENOSPC (670)
6177 - fixed bug in obd_brw_read/write() (under guise of testing 367)
6178 - fixed Linux OST size reporting problem (444, 656)
6179 - OST now updates object mtime with writes or setattr (607, 619)
6180 - client verifies file size before zeroing page past EOF (445)
6181 - OST now writes last allocated objid to disk with allocation (108)
6182 - LOV on echo now works (409)
6184 - mds_reint_unlink sends a new buffer, with the EA included. this
6185 buffer is only valid if body->valid & OBD_MD_FLEASIZE, which is only
6186 set if a regular file was being unlinked, and it was the last link
6187 - use PtlGet from the target for bulk writes (315)
6188 - OST now updates object mtime with writes or setattr (607, 619)
6189 - LDLM now has a grant-time callback to revalidate locked items, if
6191 - Many MDS operations were reorganized to combat race conditions
6193 - Merge b_intel branch (updated lprocfs code) - now at /proc/fs/lustre
6194 - configure check to avoid gcc version 2.96 20000731-2.96-98) (606)
6196 2003-01-06 Andreas Dilger <adilger@clusterfs.com>
6199 - Fully reactivate OST imports after reconnection (512, others)
6200 - Make sure client sees our -ENOTCONN from mds_handle (513 - partial)
6201 - More graceful error handling for truncating on dead OST (515)
6202 - Don't error out unless we're actually accessing dead stripes (474)
6203 - Fix garbage sizes when stripes are missing (410)
6204 - LRU counters were broken, causing constant lock purge (433, 432)
6205 - garbage on read from stripes with failed OSTs (441)
6206 - mark OSCs as active before reconnecting during recovery (438)
6207 - lov_enqueue and lov_cancel need to handle inactive OSTs (403)
6208 - lfind did not preserve OST order in output (443)
6209 - symlinks cause hung clients, incorrect data (439)
6210 - stop dereferencing request after dropping refcount (457)
6211 - don't LASSERT(spin_is_locked) on non-SMP (455)
6212 - fixes for many rename() bugs
6213 - fstat didn't correctly synchronize attributes (399)
6214 - server must handle lock cancellation during blocking AST prep (487)
6215 - bulk descriptors were free()d too soon (511)
6216 - fix paths in lconf, which would load incorrect modules (451, 507)
6217 - fix confusing lconf 'host not found' error message (386)
6218 - fix lock order deadlock on OST (O/R i_sem before journal ops, 478)
6219 - fix race condition in mdc_blocking_ast() for inode access (526)
6220 - fix lov_unpackmd() unpacking wrong number of stripes (537)
6221 - fix lov_set_osc_active() marking wrong OSC inactive (440)
6222 - fix bad lstripe lov_unpackmd() assertion (fix layering too) (527)
6223 - fix multiple writes of stripe MD to MDS (358, maybe 519)
6224 - fix lstripe in several ways (kernel side) (527)
6225 - fix request leak in ldlm_cli_enqueue (262)
6226 - incorrect OSC was marked inactive after OST failure
6227 - call mds_fs_cleanup before unmounting filesystem (524)
6228 - fix races between taking ns_lock and ldlm_lock_change_resource
6229 - fix races updating LOV export open file list
6230 - fix lov_enqueue error path, avoid decref-ing bad lock handle
6231 - fix recovery NULL deref in ldlm_cli_cancel_unused
6232 - fix some DLM races by using new hash table for lock handles (419)
6233 - permit the client to specify desired inodes, at replay
6234 - duplicate requests when we queue them for replay reintegration
6235 - fix last_rcvd offset calculation
6236 - sync after each recovered transaction, so we always make progress
6237 - never, not always, ERESTART requests without transnos
6238 - store the lov_desc in the MDS, so we don't depend on getlovinfo to
6240 - skip replay if the MDS says that the client is already connected
6241 - don't check for a recovery-enabled export to match lctl's UUID
6242 - don't INC_USE_COUNT for phantom exports
6243 - don't crash when cleaning up phantom exports (567)
6244 - don't double-finish or set replay data for errored mdc_open requests
6245 - abort requests when they time out, so we don't get old replies
6246 - send/receive replies for AST messages again
6247 - if the client says that it doesn't have the lock, cancel it on the
6249 - if we timeout during I/O, don't try to cancel an in-use lock; instead
6250 mark it as destroyed, it will all work out when decref is called
6251 - fix module use counts (22, 581)
6253 - ASTs now expect a reply (server cancels lock on error reply)
6255 2002-12-02 Andreas Dilger <adilger@clusterfs.com>
6258 - fix many simultaneous client startup (392)
6259 - fix dentry->d_it clobbering
6260 - credentials weren't being shipped for readdir/getattr operations
6261 - remove invalid assertions triggered during some concurrent MD
6263 - proper Lustre versions added (336, 389)
6264 - fix memory leak for create error case (398)
6265 - fix LOV locking bug that would get cli/srv out of sync
6266 - fix echo client over LOV (409)
6267 - fix dbench 2, extN refcount problem (170, 258, 356, 418)
6268 - fix double-O_EXCL intent crash (424)
6269 - avoid sending multiple lock CANCELs (352)
6271 - MDS can do multi-client recovery (modulo bugs in new code)
6273 - many updates, edits, cleanups
6275 2002-11-18 Phil Schwan <phil@clusterfs.com>
6278 - fix null d_it dereference (346)
6279 - fix full OST/dbench hang (333)
6280 - fix permission problem with file removal (286)
6281 - fix removal of OSCs from LOV when they fail
6282 - fix NULL deref during bulk timeout (214)
6283 - fix problems related to multiple filesystems on one MDS (241)
6284 - fixed serious subtle metadata locking bugs
6285 - free locks on clients when inodes are removed due to memory
6287 - fix inode pointer in lock data (285)
6288 - partial support for multiple MDS on a single host (241)
6289 - data locks weren't cancelled at clear_inode time (290, 311)
6290 - intent locks could lead to unbounded lock growth (205)
6291 - added a maximum lock count, an LRU list, and a flusher
6292 - fix multiple rename (365)
6293 - properly abstracted the echo client
6294 - OSC locked 1 byte too many; fixed
6295 - rewrote brw callback code:
6296 - fixed recovery bugs related to LOVs (306)
6297 - fixed too-many-pages-in-one-write crash (191)
6298 - fixed (again) crash in sync_io_timeout (214)
6299 - probably fixed callback-related race (385)
6301 - Add capability to MDS protocol
6302 - LDLM cancellations and callbacks on different portals
6304 2002-10-28 Andreas Dilger <adilger@clusterfs.com>
6307 - limit client IOV size to PTL_MD_MAX_IOV (611336, 191)
6308 - defer open object destruction to close time (601981, 138)
6309 - open/close OST file handle in obdo (OBD_MD_FLHANDLE) (601981, 138)
6310 - move LDLM_ENQUEUE/CONVERT back to MDS portal (625069)
6311 - abstract ll_lookup2, fix ll_revalidate2 to use abstraction (256)
6312 - don't call obd_setattr in ll_file_release for destroyed objects
6313 * protocol change to lustre_msg: move |version| and add |flags|
6314 * protocol change to osc_punch: "start" in "o_size", "end" in "o_blocks"
6315 * lock replay: for LDLM_FL_REPLAY trust client to do right thing
6316 * added replay of create, unlink, link and rename operations during
6317 MDS failover; recovery should be much more robust now
6318 * remove failed OSCs from LOVs (only lov_create uses this so far)
6319 * the lustre-HOWTO was brought (more) up to date (582544)
6321 2002-10-23 Phil Schwan <phil@clusterfs.com>
6324 - in-use dentries weren't being reused properly (617851)
6325 - prevent multiple LDLM setup (599178)
6326 - fix LOV size calculations for truncate (617853)
6327 - fix client handling of MDS intent errors (POSIX)
6328 - fix permission bug in lovstripe.c test (624321)
6329 - fix MDS thread deadlock - move LDLM handler to DLM portal (625069)
6330 - truncate past end of file could corrupt data
6331 - proper cleanup after timeouts, crashes, etc (592524, 550815)
6332 - a race in recovery could return ETIMEDOUT to apps (623947)
6333 - building outside the source directory was fixed
6334 * the lustre-HOWTO was brought (more) up to date (582544)
6335 * major progress was made on recovery functionality
6337 2002-10-10 Phil Schwan <phil@clusterfs.com>
6340 - recovery deadlock fix
6341 - rm -rf causes LBUG fix (617817)
6342 - file open by multiple tasks fix (618962)
6343 - directory permissions bugs (602707 and 620007)
6344 - journal_stop fixed with locking (611313)
6345 - O_APPEND failures resolved (618273, perhaps 614459)
6346 - lconf PATH fix (619770)
6347 - IA64 build fix (621450)
6348 - RPC buffer sizes scale with amount of memory
6350 2002-10-01 Phil Schwan <phil@clusterfs.com>
6353 - locks would be cancelled without throwing away data pages,
6354 resulting in inconsistent data (605627)
6355 - inode attributes were not always being refreshed (605627, 612449)
6356 - lconf now continues to cleanup after lctl reports an error
6357 - MDS now enforces user permissions (602707)
6358 - lprocfs cleanup fixed, but not yet enabled (614157)
6359 - fixed infinite server hang, should a client not respond to an AST
6360 - avoid going into recovery if user calls readlink() with a buffer
6361 that's too small (613941)
6362 - AST RPCs no longer require replies (614867) -- this may be changed
6363 - don't crash server if client sends an IOV that's too big (611336)
6364 - fixed lock conversion deadlock (611892)
6365 - fixed the following of symlinks (614622)
6366 * recovery: the server can remove locks from a client that dies, other
6367 clients can make progress
6368 * more extN patch fixes
6369 * compile-time configurable ptlrpc buffer allocations
6371 - collaborative read cache document
6372 - Lustre Lite Performance CDR document-in-progress
6374 2002-09-20 Andreas Dilger <adilger@clusterfs.com>
6377 - fix typo in patch-2.4.18
6379 2002-09-20 Andreas Dilger <adilger@clusterfs.com>
6382 - clear ptlrpc request each time in handle_incoming_request()
6383 - unlink of files now destroys the object on the OST
6385 2002-09-19 Peter Braam <braam@clusterfs.com>
6387 * add hard link support
6388 * change obdfile creation method
6389 * kernel patch changed
6391 2002-09-19 Peter Braam <braam@clusterfs.com>
6394 - stack overflow bug in extN fixed
6396 2002-09-18 Andreas Dilger <adilger@clusterfs.com>
6398 * documentation updates
6399 - add man pages for config tools
6400 - update tests/README to describe testing with new config tools
6401 - finish metadata API descriptions
6402 * bug fixes and cleanups
6403 - statfs workaround for 16TB limit
6404 - LOV stripe allocation improved, can stripe on subset of OSTs
6405 - LOV file size/IO offset was wrong for files > 4GB in size
6406 - object EA data was being dropped, caused files to be unreadable
6407 - memory overflow with non-LOV OST caused memory corruption
6408 - fixed regression tests to work with new config tools, obdfilter
6409 - fixed bug when directory size became larger than 1 block
6410 - fixed bug (for single client case) when PWD was deleted
6411 - invalidate local directory pages when doing intent-based ops
6412 - avoid LDLM oops when lock callback contained bad data
6414 2002-09-09 Andreas Dilger <adilger@clusterfs.com>
6416 * documentation updates
6417 * bug fixes and cleanups
6418 - configuration tools
6421 - 64-bit compile warnings
6422 - 64-bit internal statfs data
6424 * test_brw on persistent OST devices
6426 * lprocfs (disabled)
6428 2002-09-04 Andreas Dilger <adilger@clusterfs.com>
6430 * documentation updates
6431 * bug fixes and cleanups
6432 * configuration tools
6434 2002-08-30 Peter J. Braam <braam@clusterfs.com>
6437 * many small fixes to 0_5_4
6438 * io/network handling
6439 * thinkos in MDS operations
6441 2002-08-24 Peter J. Braam <braam@clusterfs.com>
6444 * crucial basic fixes to 0.5.3
6445 * IOR, Iozone work over Elan
6448 2002-08-07 Phil Schwan <phil@clusterfs.com>
6449 * version 0_5_3, our first alpha
6450 * we use the new Portals iovs
6451 * documentation updates
6452 * bug fixes and cleanups
6453 * small changes in the DLM wire protocol
6455 2002-07-25 Peter J. Braam <braam@clusterfs.com>
6456 * version 0_5_1 with some initial stability,
6457 * locking on MD and file I/O.
6458 * documentation updates
6459 * several bug fixes since 0.5.0
6460 * small changes in wire protocol
6462 2002-07-18 Phil Schwan <phil@clusterfs.com>
6464 * delivered as Lustre Light Alpha
6465 * fixed a crash after handling invalid MDS requests
6466 * fixed directory pages for architectures with non-4k pages sizes
6468 2002-07-11 Andreas Dilger <adilger@clusterfs.com>
6469 * release version v0_4_4
6470 * Moves TCP acceptor to be on port 2432 (unused Coda port) instead
6472 * Fixes a number of interruption problems with OST operations.
6473 * Update documentation for portals header changes
6474 * Move all wire protocol structs/defines to lustre_idl.h
6475 * Fixes symlink length bug.
6476 * Add tcpdump to repository.
6478 2002-07-05 Andreas Dilger <adilger@clusterfs.com>
6479 * release version v0_4_3
6480 * Fixes statfs for inodes on extN.
6481 * Fixes bug in runtests which would delete /etc/hosts.
6482 * Use 64-bit object IDs wherever possible (not into VFS though)
6483 Remove ost_get_info, which is unused by lustre, and out of date.
6485 2002-07-03 Peter Braam <braam@clusterfs.com>
6486 * release version v0_4_2 Fixes a lookup error (type not passed)
6487 * move forward to head of Portals
6488 * move forward to latest Lustre kernel
6490 2002-06-25 Peter Braam <braam@clusterfs.com>
6491 * release version v0_4_1. Hopefully stable on single node use.