1 tbd Sun Microsystems, Inc.
4 2.6.16.60-0.33 (SLES 10),
5 2.6.18-128.1.1.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: 1.40.11-sun1
11 * File join has been disabled in this release, refer to Bugzilla 16929.
12 * A new Lustre ADIO driver is available for MPICH2-1.0.7.
13 * NFS export disabled when stack size < 8192. Since the NFSv4 export
14 of Lustre filesystem with 4K stack may cause a stack overflow. For
15 more information, please refer to bugzilla 17630.
17 Severity : enhancement
19 Description: Caching OSS
20 Details : introduce data caching on the OSS. The OSS now relies on the linux
21 kernel page cache to keep recently accessed data in memory.
22 It is worth noting that all write requests are still flushed
23 synchronously as in lustre 1.6.
25 Severity : enhancement
27 Description: version based recovery
28 Details : introduce finer grained recovery able to detect transaction
29 dependencies and can deal with transaction gaps.
31 Severity : enhancement
33 Description: Enable adaptive timeouts by default
37 Descriptoin: fix goal inodes
38 Details : Allocate inodes for llog in last inode group for avoid broke
43 Descriptoin: OSS DeadLock
44 Details : Use trylock to prevent deadlock when shrink icache.
46 Severity : enhancement
48 Description: Allow tuning service thread via /proc
49 Details : For each service a new
50 /proc/fs/lustre/{service}/*/thread_{min,max,started} entry is
51 created that can be used to set min/max thread counts, and get the
52 current number of running threads.
54 Severity : enhancement
56 Description: Add state history info file, enhance import info file
57 Details : Track import connection state changes in a new osc/mdc proc file;
58 add overview-type data to the osc/mdc import proc file.
62 Description: Reduce small size read RPC
63 Details : Set read-ahead limite for every file and only do read-ahead when
64 available read-ahead pages are bigger than 1M to avoid small size
69 Description: free_entry erroneously used groups_free instead of put_group_info
71 Severity : enhancement
73 Description: Make read-ahead stripe size aligned.
75 Severity : enhancement
77 Description: MDS create should not wait for statfs RPC while holding DLM lock.
82 Descriptoin: OST grant too much space to client even there are not enough space.
83 Details : Client will shrink its grant cache to OST if there are no write
84 activity over 6 mins (GRANT_SHRINK_INTERVAL), and OST will retrieve
85 this grant cache if there are already not enough avaible space
86 (left_space < total_clients * 32M).
88 Severity : enhancement
90 Description: Update to RHEL5U3 kernel-2.6.18-128.1.1.el5.
92 Severity : enhancement
94 Description: Update client support to vanila kernels up to 2.6.27.
97 Frequency : rare, connect and disconnect target at same time
99 Descriptoin: ASSERTION(atomic_read(&imp->imp_inflight) == 0
100 Details : don't call obd_disconnect under lov_lock. this long time
101 operation and can block ptlrpcd which answer to connect request.
103 Severity : enhancement
105 Description: Update to SLES10 kernel-2.6.16.60-0.33.
107 Severity : enhancement
109 Description: Update to sles9 kernel-2.6.5-7.315.
112 Frequency : start MDS on uncleanly shutdowned MDS device
114 Descriptoin: ll_sync thread stay in waiting mds<>ost recovery finished
115 Details : stay in waiting mds<>ost recovery finished produce random bugs
116 due race between two ll_sync thread for one lov target. send
117 ACTIVATE event only if connect realy finished and import have
120 Severity : enhancement
122 Description: Update to RHEL5 kernel-2.6.18-92.1.22.el5.
125 Frequency : start MDS on uncleanly shutdowned MDS device
127 Descriptoin: aborting recovery hang on MDS
128 Details : don't throttle destroy RPCs for the MDT.
130 Severity : enhancement
132 Descriptoin: Don't sync journal after every i/o
133 Details : Implement write RPC replay to allow server replies for write RPCs
134 before data is on disk.
138 Description: Slow reads beyond 8Tb offsets.
139 Details : Page index integer overflow in ll_read_ahead_page
143 Description: Soft lockup on OSS after MDS failover
144 Details : MSG_CONNECT_INITIAL is not set on the initial MDS->OST connect.
145 As a conseqence, the patch from bug 18224 is not operational
146 and the MDS export cannot be reused on the OSTs until it gets
150 Frequency : rare, only if using MMP with Linux RAID
152 Description: MMP doesn't work with Linux RAID
153 Details : While using HA for Lustre servers with Linux RAID, it is possible
154 that MMP will not detect multiple mounts. To make this work we
155 need to unplug the device queue in RAID when the MMP block is being
156 written. Also while reading the MMP block, we should read it from
157 disk and not the cached one.
160 Frequency : rare, during recovery
162 Description: Assertion failure in ldlm_lock_put
163 Details : Do not put cancelled locks into replay list, hold references on
168 Description: 1.6.5 mdsrate performance is slower than 1.4.11/12 (MDS is not cpu bound!)
169 Details : create_count always drops to the min value (=32) because grow_count
170 is being changed before the precreate RPC completes.
172 -------------------------------------------------------------------------------
173 2008-12-31 Sun Microsystems, Inc.
175 * Support for kernels:
176 2.6.5-7.314 (SLES 9),
177 2.6.9-67.0.22.EL (RHEL 4),
178 2.6.16.60-0.31 (SLES 10),
179 2.6.18-92.1.17.el5 (RHEL 5),
180 2.6.22.14 vanilla (kernel.org)
181 * Client support for unpatched kernels:
182 (see http://wiki.lustre.org/index.php?title=Patchless_Client)
183 2.6.16 - 2.6.22 vanilla (kernel.org)
184 * Client support for unpatched kernels:
185 we do not recommend using patchless RHEL4 clients with kernels
186 prior to 2.6.9-55EL (RHEL4U5).
187 * Recommended e2fsprogs version: 1.40.11-sun1
188 * Note that reiserfs quotas are disabled on SLES 10 in this kernel.
189 * RHEL 4 and RHEL 5/SLES 10 clients behaves differently on 'cd' to a
190 removed cwd "./" (refer to Bugzilla 14399).
191 * A new quota file format has been introduced in 1.6.5.
192 The format conversion from prior releases is handled transparently,
193 but releases older than 1.4.12/1.6.5 don't understand this new
194 format. The automatic format conversion can be avoided by running
195 the following command on the MDS:
196 'tunefs.lustre --param="mdt.quota_type=ug1" $MDTDEV'.
197 For more information, please refer to bugzilla 13904.
198 * A new quota file format was introduced in 1.6.6/1.8.0 (kernels 2.6.16+).
199 The format conversion from prior releases is handled transparently,
200 but releases older than 1.6.6/1.8.0 don't understand this new
201 format. The automatic format conversion can be avoided by running
202 the following commands on the MDS and OSS servers (for
203 pre 1.4.12-1.6.5 quota files):
204 'tunefs.lustre --param="mdt.quota_type=ug1" $MDTDEV',
205 'tunefs.lustre --param="ost.quota_type=ug1" $MDTDEV'
206 or (for 1.4.12/1.6.5 quota files)
207 'tunefs.lustre --param="mdt.quota_type=ug2" $MDTDEV',
208 'tunefs.lustre --param="ost.quota_type=ug2" $MDTDEV'
209 For more information, please refer to bugzilla 13904.
210 * Output of lfs quota has been made less detailed by default,
211 old (verbose) output can be obtained by using -v option.
212 * File join has been disabled in this release, refer to Bugzilla 16929.
213 * A new Lustre ADIO driver is available for MPICH2-1.0.7.
214 * NFS export disabled when stack size < 8192. Since the NFSv4 export of
215 Lustre filesystem with 4K stack may cause a stack overflow. For more
216 information, please refer to bugzilla 17630.
218 Severity : enhancement
220 Description: add lazystatfs mount option to allow statfs(2) to skip down OSTs
221 Details : allow skip disconnected ost for send statfs request and hide error
225 Frequency : rare, on llog test 6
227 Descriptoin: don't allow connect to already connected import
228 Details : allowing connect to already connected import is hide connecting problem.
231 Frequency : rare, on failed llog setup
233 Descriptoin: don't leak obd reference on failed llog setup
234 Details : for failed llog setup - mgc forget call class_destroy_import for
235 client import, move destroy import to more generic place.
240 Descriptoin: allow kill process which wait statahead result
241 Details : for some reasons 'ls' can stick in waiting result from statahead,
242 in this case need way for kill this process.
245 Frequency : rare, at shutdown
247 Descriptoin: panic at umount
248 Details : llap_shrinker can be raced with killing super block from list and
249 this produce panic with access to already freeded pointer
254 Descriptoin: don't lose wakeup for imp_recovery_waitq
255 Details : recover_import_no_retry or invalidate_import and import_close can
256 both sleep on imp_recovery_waitq, but we was send only one wakeup
262 Descriptoin: panic in mds_open
263 Details : don't confuse mds_finish_transno() with PTR_ERR(-ENOENT)
268 Descriptoin: stuck in cache_remove_extent() or panic with accessing to already
270 Details : release lock refernce only after add page to pages list.
273 Frequency : always with long access acl
275 Descriptoin: mds can't pack reply with long acl.
276 Details : mds don't control size of acl but they limited by reint/getattr
281 Frequency : on remount
283 Description: external journal device not working after the remount
284 Details : clear dev_rdonly flag for external journal devices in
290 Description: shutdown vs evict race
291 Details : client_disconnect_export vs connect request race.
292 if client will evicted at this time - we start invalidate
293 thread without referece to import and import can be freed
299 Description: shrink LOV EAs before replying
300 Details : correctly adjust LOV EA buffer for reply.
305 Description: don't skip ost target if they assigned to file
306 Details : Drop slow OSCs if we can, but not for requested start idx.
307 This means "if OSC is slow and it is not the requested
308 start OST, then it can be skipped, otherwise skip it only
309 if it is inactive/recovering/out-of-space.
311 Severity : enhancement
313 Description: Update to RHEL5 kernel-2.6.18-92.1.17.el5.
316 Frequency : rare, need acl's on inode.
318 Description: client can't handle ost additional correctly
319 Details : if ost was added after client connected to mds client can have
320 hit lnet_try_match_md ... to big messages to wide striped files.
321 in this case need teach client to handle config events about add
322 lov target and update client max ea size at that event.
324 Severity : enhancement
326 Description: Update to sles9 kernel-2.6.5-7.314.
328 Severity : enhancement
330 Description: Update to SLES10 SP2 kernel-2.6.16.60-0.31.
333 Frequency : Create a symlink file with a very long name
335 Description: ldlm_cancel_pack()) ASSERTION(max >= dlm->lock_count + count)
336 Details : If there is no extra space in the request for early cancels,
337 ldlm_req_handles_avail() returns 0 instead of a negative value.
342 Description: mds is deadlocked
343 Details : in rare cases, inode in catalog can have i_no less than have parent
344 i_no, this produce wrong order for locking during open, and
345 parallel unlink can be lock open. this need teach mds_open to grab
346 locks in resource id order, not at parent -> child order.
348 Severity : enhancement
350 Description: Add /proc entry for import status
351 Details : The mdc, osc, and mgc import directories now have
352 an import directory that contains useful import data for debugging
355 Severity : enhancement
357 Description: Re-disable certain /proc logging
358 Details : Enable and disable client's offset_stats, extents_stats and
359 extents_stats_per_process stats logging on the fly.
362 Frequency : Only on FC kernels 2.6.22+
364 Description: oops in statahead
365 Details : Do not drop reference count for the dentry from VFS when lookup,
366 VFS will do that by itself.
368 Severity : enhancement
370 Description: Generic /proc file permissions
371 Details : Set /Proc file permissions in a more generic way to enable non-
372 root users operate on some /proc files.
376 Description: Hitting mdc_commit_close() ASSERTION
377 Details : Properly handle request reference release in
378 ll_release_openhandle().
382 Frequency : only patchless client
383 Description: add workaround for race between add/remove dentry from hash
385 Severity : enhancement
387 Description: Allow OST glimpses to return PW locks
391 Description: LBUG when llog conf file is full
392 Details : When llog bitmap is full, ENOSPC should be returned for plain
397 Description: Prevent import from entering FULL state when server in recovery
401 Description: service mount cannot take device name with ":"
402 Details : Only when device name contains ":/" will mount treat it as
408 Description: replace ptlrpcd with the statahead thread to interpret the async
409 statahead RPC callback
413 Frequency : on recovery
414 Description: I/O failures after umount during fail back
415 Details : if client reconnected to restarted server we need join to recovery
416 instead of find server handler is changed and process self eviction
417 with cancel all locks.
419 Severity : enhancement
421 Description: Update to RHEL5 kernel-2.6.18-92.1.10.el5.
425 Description: Kernel BUG tries to release flock
426 Details : Lustre does not destroy flock lock before last reference goes
427 away. So always drop flock locks when client is evicted and
428 perform unlock regardless of successfulness of speaking to MDS.
430 Severity : enhancement
432 Description: Update to SLES10 SP2 kernel-2.6.16.60-0.27.
434 Severity : enhancement
436 Description: Upcall on Lustre log has been dumped
437 Details : Allow for a user mode script to be called once a Lustre log has
438 been dumped. It passes the filename of the dumped log to the
439 script, the location of the script can be specified via
440 /proc/sys/lnet/debug_log_upcall.
445 Description: avoid messages about idr_remove called for id that is not allocated
446 Details : Move assigment s_dev for clustered nfs to end of initialization,
447 for avoid problem with error handling.
452 Description: avoid Already found the key in hash [CONN_UNUSED_HASH] messages
453 Details : When connection is reused this not moved from CONN_UNUSED_HASH into
454 CONN_USED_HASH and this prodice warning when put connection again
460 Description: avoid ASSERTION(client_stat->nid_exp_ref_count == 0) failed
461 Details : release reference to stats when client disconnected, not
462 when export destroyed for avoid races when client destroyed
463 after main ost export.
467 Description: more cleanup in mds_lov
468 Details : add workaround for get valid ost count for avoid warnings about
469 drop too big messages, not init llog cat under semphore which
470 can be blocked on reconnect and break normal replay, fix access
473 Severity : enhancement
475 Description: File striping can now be set to use an arbitrary pool of OSTs.
477 Severity : enhancement
479 Description: Export bytes_read/bytes_write count on OSC/OST.
483 Description: Early reply size mismatch, MGC loses connection
484 Details : Apply the MGS_CONNECT_SUPPORTED mask at reconnect time so
485 the connect flags are properly negotiated.
489 Description: Properly propagate oinfo flags from lov to osc for statfs
490 Details : restore missing copy oi_flags to lov requests.
494 Description: exports in /proc are broken
495 Details : recreate /proc entries for clients when they reconnect.
497 Severity : enhancement
499 Description: Add man pages for llobdstat(8), llstat(8), plot-llstat(8),
500 : l_getgroups(8), lst(8), routerstat(8)
501 Details : included man pages for llobdstat(8), llstat(8),
502 : plot-llstat(8), l_getgroups(8), lst(8), routerstat(8)
504 Severity : enhancement
506 Description: Implement lustre ll_show_options method.
508 Severity : enhancement
510 Description: Update to SLES9 kernel-2.6.5-7.312.
512 Severity : enhancement
514 Description: Update to RHEL4 kernel-2.6.9-67.0.22.EL.
518 Description: exports in /proc are broken
519 Details : recreate /proc entries for clients when they reconnect.
523 Description: don't fail open with -ERANGE
524 Details : if client connected until mds will be know about real ost count
525 get LOV EA can be fail because mds not allocate enougth buffer
530 Description: Resolve device initialization race
531 Details : Prevent proc handler from accessing devices added to the
532 obd_devs array but yet be intialized.
534 Severity : enhancement
536 Description: configure's --enable-quota should check the
537 : kernel .config for CONFIG_QUOTA
538 Details : configure is terminated if --enable-quota is passed but
539 : no quota support is in kernel
541 Severity : enhancement
543 Description: Update to SLES10 SP2 kernel-2.6.16.60-0.23.
545 Severity : enhancement
547 Description: Update to RHEL5 kernel-2.6.18-92.1.6.el5.
551 Frequency : rare, on PPC clients
552 Description: don't swab ost objects in response about directory, because
554 Details : bug similar bug 14856, but in different function.
556 Severity : enhancement
558 Description: lfs quota tool enhancement
559 Details : added units specifiers support for setquota, default to
560 current uid/gid for quota report, short quota stats by
561 default, nonpositional parameters for setquota, added
562 llapi_quotactl manual page.
564 Severity : enhancement
566 Description: *optional* service tags registration
567 Details : if the "service tags" package is installed on a Lustre node
568 When the filesystem is mounted, a local-node service tag will
569 be created. See http://inventory.sun.com/ for more information
570 about the Service Tags asset management system.
574 Description: Client runs out of low memory
575 Details : Consider only lowmem when counting initial number of llap pages
578 Description: Kernel BUG tries to release flock
579 Details : Lustre does not destroy flock lock before last reference goes
580 away. So always drop flock locks when client is evicted and
581 perform unlock regardless of successfulness of speaking to MDS.
584 Frequency : occasional
586 Description: add refcount for osc callbacks, so avoid panic on shutdown
588 Severity : enhancement
590 Description: Update to RHEL4 kernel-2.6.9-67.0.20.
593 Frequency : testing only
595 Description: sanity test 65a fails if stripecount of -1 is set
596 Details : handle -1 striping on filesystem in ll_dirstripe_verify
599 Frequency : only in unusual configurations
601 Description: Kernel panic with find ost index.
602 Details : lov_obd have panic if some OST's have sparse indexes.
604 Severity : enhancement
606 Description: Update to RHEL5 kernel-2.6.18-53.1.21.el5.
609 Frequency : rarely, if filesystem is mounted with -o flock
611 Description: do not process already freed flock
612 Details : flock can possibly be freed by another thread before it reaches
613 to ldlm_flock_completion_ast.
616 Frequency : rarely, if filesystem is mounted with -o flock
618 Description: LBUG during stress test
619 Details : Need properly lock accesses the flock deadlock detection list.
622 Frequency : rarely, if binaries are being run from Lustre
624 Description: oops in page fault handler
625 Details : kernel page fault handler can return two special 'pages' in
626 error case, don't try dereference NOPAGE_SIGBUS and NOPAGE_OMM.
629 Frequency : rarely, during shutdown
631 Description: timeout with invalidate import.
632 Details : ptlrpcd_check call obd_zombie_impexp_cull and wait request which
633 should be handled by ptlrpcd. This produce long age waiting and
634 -ETIMEOUT ptlrpc_invalidate_import and as result LASSERT.
640 Description: ASSERTION(CheckWriteback(page,cmd)) failed
641 Details : badly clear PG_Writeback bit in ll_ap_completion can produce false
645 Frequency : only with broken builds/installations
647 Description: no LBUG if lquota.ko and fsfilt_ldiskfs.ko are different versions
648 Details : just return an error to a user, put a console error message
650 Severity : enhancement
652 Description: Update to RHEL5 kernel-2.6.18-53.1.19.el5.
654 Severity : enhancement
656 Description: Update to RHEL4 kernel-2.6.9-67.0.15.
658 Severity : enhancement
660 Description: enable MGS and MDT services start separately
661 Details : add a 'nomgs' option in mount.lustre to enable start a MDT with
662 a co-located MGS without starting the MGS, which is a complement
663 to 'nosvc' mount option.
666 Frequency : always, on big-endian systems
668 Description: cleanup in ptlrpc code, related to PPC platform
669 Details : store magic in native order avoid panic's in recovery on PPC
670 node and forbid from this error in future. Also fix posibily
671 of twice swab data. Fix get lov striping to userpace.
674 Frequency : rarely, if replay get lost on server
676 Description: server incorrectly drop resent replays lead to recovery failure.
677 Details : do not drop replay according to msg flags, instead we check the
678 per-export recovery request queue for duplication of transno.
681 Frequency : after recovery
683 Description: precreate to many object's after del orphan.
684 Details : del orphan st in oscc last_id == next_id and this triger growing
685 count of precreated objects. Set flag LOW to skip increase count
686 of precreated objects.
689 Frequency : rare, on clear nid stats
691 Description: ASSERTION(client_stat->nid_exp_ref_count == 0)
692 Details : when clean nid stats sometimes try destroy live entry,
693 and this produce panic in free.
696 Frequency : occasionally since 1.6.4
698 Description: Stack overflow during MDS log replay
699 Details : ease stack pressure by using a thread dealing llog_process.
702 Frequency : very rare
704 Description: MDT cannot be unmounted, reporting "Mount still busy"
705 Details : Mountpoint references were being leaked during open reply
706 reconstruction after an MDS restart. Drop mountpoint reference
707 in reconstruct_open() and free dentry reference also.
712 Description: wait until IO finished before start new when do lock cancel.
713 Details : VM protocol want old IO finished before start new, in this case
714 need wait until PG_writeback is cleared until check dirty flag
715 and call writepages in lock cancel callback.
720 Description: mds_mfd_close() ASSERTION(rc == 0)
721 Details : In mds_mfd_close(), we need protect inode's writecount change
722 within its orphan write semaphore to prevent possible races.
725 Frequency : rare, on shutdown ost
727 Description: don't hit live lock with umount ost.
728 Details : shrink_dcache_parent can be in long loop with destroy dentries,
729 use shrink_dcache_sb instead.
732 Frequency : only when echo_client is used
734 Description: don't panic with use echo_client
735 Details : echo client pass NULL as client nid pointer and this produce NULL
739 Frequency : Always on 32-bit PowerPC systems
741 Description: fix build on PPC32
742 Details : compile code with -m64 flag produce wrong object file for PPC32.
747 Description: MDS LBUG: ASSERTION(!IS_ERR(dchild))
748 Details : In reconstruct_* functions, LASSERTs on both the data supplied
749 by a client, and the data on disk are dangerous and incorrect.
750 Change them with client eviction.
752 Severity : enhancement
754 Description: skiplist implementation simplification
755 Details : skiplists are used to group compatible locks on granted list
756 that was implemented as tracking first and last lock of each lock
757 group the patch changes that to using doubly linked lists
761 Description: delete compatibility for 32bit qdata
762 Details : as planned, when lustre is beyond b1_8, lquota won't support 32bit
763 qunit. That means servers of b1_4 and servers of b1_8 can't be
764 used together if users want to use quota.
767 Frequency : only with administrator action
769 Description: mount failure if config log has invalid conf_param setting
770 Details : If administrator specified an incorrect configuration parameter
771 with "lctl conf_param" this would cause an error during future
772 client mounts. Instead, ignore the bad configuration parameter.
775 Frequency : blocks per group < blocksize*8 and uninit_groups is enabled
777 Description: ldiskfs error: XXX blocks in bitmap, YYY in gd
778 Details : If blocks per group is less than blocksize*8, set rest of the
782 Frequency : Application do stride read on lustre
784 Description: The read performance will drop a lot if the application does
786 Details : Because the stride_start_offset are missing in stride read-ahead,
787 it will cause clients read a lot of unused pages in read-ahead,
788 then the read-performance drops.
792 Description: more ldlm soft lockups
793 Details : In ldlm_resource_add_lock(), call to ldlm_resource_dump()
794 starve other threads from the resource lock for a long time in
795 case of long waiting queue, so change the debug level from
796 D_OTHER to the less frequently used D_INFO.
798 Severity : enhancement
800 Description: add -gid, -group, -uid, -user options to lfs find
802 Severity : enhancement
804 Description: ll_recover_lost_found_objs - recover objects in lost+found
805 Details : OST corruption and subsequent e2fsck can leave objects in the
806 lost+found directory. Using the "ll_recover_lost_found_objs"
807 tool, these objects can be retrieved and data can be salvaged
808 by using the object ID saved in the fid EA on each object.
813 Description: this bug _only_ happens when inode quota limitation is very low
814 (less than 12), so that inode quota unit is 1 at initialization.
815 Details : if remaining quota equates 1, it is a sign to demonstate that quota
816 is effective now. So least quota qunit should be 2.
820 Description: Hung threads in invalidate_inode_pages2_range
821 Details : The direct IO path doesn't call check_rpcs to submit a new RPC once
822 one is completed. As a result, some RPCs are stuck in the queue
827 Description: Procfs and llog threads access destoryed import sometimes.
828 Details : Sync the import destoryed process with procfs and llog threads by
829 the import refcount and semaphore.
833 Description: mds fails to respond, threads stuck in ldlm_completion_ast
834 Details : Sort source/child resource pair after updating child resource.
839 Description: kernel BUG at ldiskfs2_ext_new_extent_cb
840 Details : If insertion of an extent fails, then discard the inode
841 preallocation and free data blocks else it can lead to duplicate
846 Description: don't always update ctime in ext3_xattr_set_handle()
847 Details : Current xattr code updates inode ctime in ext3_xattr_set_handle()
848 In some cases the ctime should not be updated, for example for
849 2.0->1.8 compatibility it is necessary to delete an xattr and it
850 should not update the ctime.
854 Description: add quota statistics
855 Details : 1. sort out quota proc entries and proc code.
856 2. add quota statistics
861 Description: quotas are not honored with O_DIRECT
862 Details : all writes with the flag O_DIRECT will use grants which leads to
863 this problem. Now using OBD_BRW_SYNC to guard this.
867 Bugzilla : 15713/16362
868 Description: Assertion in iopen_connect_dentry in 1.6.3
869 Details : looking up an inode via iopen with the wrong generation number can
870 populate the dcache with a disconneced dentry while the inode
871 number is in the process of being reallocated. This causes an
872 assertion failure in iopen since the inode's dentry list contains
873 both a connected and disconnected dentry.
877 Description: assertion failure in ldlm_handle2lock()
878 Details : fix a race between class_handle_unhash() and class_handle2object()
879 introduced in lustre 1.6.5 by bug 13622.
881 Severity : enhancement
883 Description: superblock lock contention with many SMP cores on one client
884 Details : several client filesystem locks were highly contended on SMP
885 NUMA systems with 8 or more cores. Per-CPU datastructures
886 and more efficient locking implemented to reduce contention.
891 Description: Kernel BUG: sd_iostats_bump: unexpected disk index
892 Details : remove the limit of 256 scsi disks in the sd_iostat patch
897 Description: oops in sd_iostats_seq_show()
898 Details : unloading/reloading the scsi low level driver triggers a kernel
899 bug when trying to access the sd iostat file.
904 Description: Kernel panics during QLogic driver reload
905 Details : REQ_BLOCK_PC requests are not handled properly in the sd iostat
906 patch, causing memory corruption.
911 Description: journal_dev option does not work in b1_6
912 Details : pass mount option during pre-mount.
914 Severity : enhancement
916 Description: Add a FIEMAP(FIle Extent MAP) ioctl for ldiskfs
917 Details : FIEMAP ioctl will allow an application to efficiently fetch the
918 extent information of a file. It can be used to map logical blocks
919 in a file to physical blocks in the block device.
922 Frequency : only with adaptive timeout enabled
924 Description: DEBUG_REQ() bad paging request
925 Details : ptlrpc_at_recv_early_reply() should not modify req->rq_repmsg
926 because it can be accessed by reply_in_callback() without the
930 Frequency : only on Cray X2
932 Description: X2 build failures
933 Details : fix build failures on Cray X2.
937 Description: xid & resent requests
938 Details : Initialize RPC XID from clock at startup (randomly if clock is
943 Description: quota recovery deadlock during mds failover
944 Details : This patch includes att18982, att18236, att18237 in bz14840.
946 1. fix osts hang when mds does failover with quotaon
947 2. prevent watchdog storm when osts threads wait for the
952 Description: kernel panic on racer
953 Details : Do not access dchild->d_inode when IS_ERR(dchild) is true.
955 Severity : enhancement
957 Description: Add lustre_start utility to start or stop multiple Lustre servers
962 Description: Lustre GPF in {:ptlrpc:ptlrpc_server_free_request+373}
963 Details : In case of memory pressure, list_del() can be called twice on
964 req->rq_history_list, causing a kernel oops.
968 Description: (ptllnd_peer.c:557:kptllnd_peer_check_sends()) ASSERTION(!in_interrupt()) failed
969 Details : fix stack overflow in the distributed lock manager by defering
970 export eviction after a failed ast to the elt thread instead of
971 handling it in the dlm interpret routine.
973 Severity : enhancement
975 Description: More exported tunables for mballoc
976 Details : Add support for tunable preallocation window and new tunables for
981 Description: Detect corruption of block bitmap and checking for preallocations
982 Details : Checks validity of on-disk block bitmap. Also it does better
983 checking of number of applied preallocations. When corruption is
984 found, it turns filesystem readonly to prevent further corruptions.
988 Frequency : only for big-endian servers
989 Description: Check for big-endian system while mounting fs with extents feature
990 Details : Mounting a filesystem with extents feature will fail on big-endian
991 systems since ext3-based ldiskfs is not supported on big-endian
992 systems. Can be over-ridden with "bigendian_extents" mount option.
996 Description: Excessive recovery window
997 Details : With AT enabled, the recovery window can be excessively long (6000+
998 seconds). To address this problem, we no longer use
999 OBD_RECOVERY_FACTOR when extending the recovery window (the connect
1000 timeout no longer depends on the service time, it is set to
1001 INITIAL_CONNECT_TIMEOUT now) and clients report the old service
1002 time via pb_service_time.
1006 Description: Watchdog triggered on MDS failover
1007 Details : enable OBD_CONNECT_MDT flag when connecting from the MDS so that
1008 the OSTs know that the MDS "UUID" can be reused for the same export
1009 from a different NID, so we do not need to wait for the export to
1014 Descriptoin: Lustre detected file system corruption with inode out of bounds
1015 Details : don't update i_size on MDS_CLOSE for directories. This causes
1016 directory corruptions on the MDT.
1018 -------------------------------------------------------------------------------
1020 2008-05-26 Sun Microsystems, Inc.
1022 * Support for kernels:
1023 2.6.5-7.311 (SLES 9),
1024 2.6.9-67.0.7.EL (RHEL 4),
1025 2.6.16.54-0.2.5 (SLES 10),
1026 2.6.18-53.1.14.el5 (RHEL 5),
1027 2.6.22.14 vanilla (kernel.org)
1028 * Client support for unpatched kernels:
1029 (see http://wiki.lustre.org/index.php?title=Patchless_Client)
1030 2.6.16 - 2.6.22 vanilla (kernel.org)
1031 * Due to problems with nested symlinks and FMODE_EXEC (bug 12652),
1032 we do not recommend using patchless RHEL4 clients with kernels
1033 prior to 2.6.9-55EL (RHEL4U5).
1034 * Recommended e2fsprogs version: 1.40.7-sun1
1035 * Note that reiserfs quotas are disabled on SLES 10 in this kernel.
1036 * RHEL 4 and RHEL 5/SLES 10 clients behaves differently on 'cd' to a
1037 removed cwd "./" (refer to Bugzilla 14399).
1038 * A new quota file format has been introduced in 1.6.5.
1039 The format conversion from prior releases is handled transparently,
1040 but releases older than 1.4.12/1.6.5 will not understand this new
1041 format. The automatic format conversion can be avoided by running
1042 the following command on the MDS before upgrading:
1043 'tunefs.lustre --param="mdt.quota_type=ug1" $MDTDEV'.
1044 For more information, please refer to bugzilla 13904.
1048 Description: quota performance fix
1049 Details : quota data is written in journalled mode instead of ordered to
1050 increase performance
1054 Description: lfs support for human-readable quota grace time strings
1055 Details : lfs setquota -t and lfs quota -t represent quota grace times
1056 in "XXwXXdXXhXXmXXs" format instead of large values in seconds
1059 Frequency : always with o2ib 1.3 and sles10
1061 Description: fix build with SLES10 and o2ib v3.
1062 Details : sles10 uses diffrent name for Module.symver file but configure
1063 assume this file has same name on RHEL/SLES/vanila kernels.
1066 Frequency : very rare, if additional xattrs are used on kernels >= 2.6.12
1068 Description: MDS may lose file striping (and hence file data) in some cases
1069 Details : If there are additional extended attributes stored on the MDS,
1070 in particular ACLs, SELinux, or user attributes (if user_xattr
1071 is specified for the client mount options) then there is a risk
1072 of attribute loss. Additionally, the Lustre file striping
1073 needs to be larger than default (e.g. striped over all OSTs),
1074 and an additional attribute must be stored initially in the
1075 inode and then increase in size enough to be moved to the
1076 external attribute block (e.g. ACL growing in size) for file
1079 Severity : enhancement
1081 Description: add message levels for liblustreapi
1086 Description: MDT cannot be unmounted, reporting "Mount still busy"
1087 Details : Mountpoint references were being leaked during open reply
1088 reconstruction after an MDS restart. Drop mountpoint reference
1089 in reconstruct_open() and free dentry reference also.
1094 Description: fix for occasional failure case of -ENOSPC in recovery-small tests
1095 Details : Move the 'good_osts' check before the 'total_bavail' check. This
1096 will result in an -EAGAIN and in the exit call path we call
1097 alloc_rr() which will with increasing aggressiveness attempt to
1098 aquire precreated objects on the minimum number of required OSCs.
1102 Description: Use old size assignment to avoid deadlock
1103 Details : This reverts the changes in bugs 2369 and bug 14138 that introduced
1104 the scheduling while holding a spinlock. We do not need locking
1105 for size in ll_update_inode() because size is only updated from
1106 the MDS for directories or files without objects, so there is no
1107 other place to do the update, and concurrent access to such inodes
1108 are protected by the inode lock.
1112 Description: Use __u64 instead of int for valid bits
1116 Description: resolve "_IOWR redefined" build error on SLES10
1120 Description: dump the memory debugging after all modules are unloaded to
1121 suppress false negative in conf_sanity test 39
1125 Description: the recovery timer never expires
1126 Details : for new client connect request, the recovery timer should not be
1127 reset, otherwise recovery timer will never expired, if the old
1128 client never come. Only old client connect and first connection
1129 req should trigger recovery timer reset.
1133 Description: the min numbers of lproc stats are wrong
1134 Details : adding a new constant LC_MIN_INIT and use it for initialization
1138 Frequency : always with interactive lfs
1140 Description: Reinitialize optind to 0 so that interactive lfs works in all cases
1143 Frequency : with multiple concurrent readdir processes in same directory
1144 Bugzilla : 15406, 15169, 15175
1145 Description: misc fixes for directory readahead.
1146 Details : prevent previous statahead async RPC callback from processing the
1147 current "statahead_info", race condition between async RPC callback
1148 add dentry into dentry hash table and "ls" thread revalidate such
1149 dentry, statahead his/miss control for hidden items, and so on.
1151 Severity : enhancement
1153 Description: build kernel-ib packages for OFED 1.3 in our release cycle
1157 Description: incore types cleaning in quota code (with respect to 64-bit limits)
1158 Details : several u32 variables declarations are replaced with u64 declarations
1163 Description: fix SLES kernel versioning
1164 Details : the kernel version for our SLES 10 kernel did not include a "-"
1165 before the "smp" at the end. while this was not a problem in
1166 general, it did mean that software trying to use the kernel
1167 version to try to detect a vendor specific kernel would fail.
1168 this was most evident by the OFED build scripts.
1173 Description: Don't update lov_desc members until making sure they are valid
1174 Details : When updating lov_desc members via proc fs, need fix their
1175 validities before doing the real update.
1178 Frequency : very rare
1180 Description: don't put request into delay list while invalidate in flight.
1181 Details : ptlrpc_delay_request sometimes put in delay list while invalidate
1182 import in flight. this produce timeout for invalidate and sometimes
1183 can cause stale data.
1185 Severity : enhancement
1187 Description: Update kernel to SLES9 2.6.5-7.311.
1189 Severity : enhancement
1191 Description: Update kernel to RHEL4 2.6.9-67.0.7.
1196 Frequency : on PPC only
1197 Description: not convert ost objects for directory because it's not exist.
1198 Details : ll_dir_getstripe assume dirrectory has ost objects but this wrong.
1200 Severity : enhancement
1202 Description: Fix warnings with compile liblustre at sles10/rhel5 which have
1203 __u64 as usingied long long type.
1206 Frequency : rare, on shutdown
1208 Description: race process ast vs remove callback
1209 Details : removing callback before disconnect import open race with
1210 processing callback.
1212 Severity : enhancement
1214 Description: Update kernel to SLES9 2.6.5-7.311.
1216 Severity : enhancement
1218 Description: Files open for execute are not marked busy on SLES10
1219 Details : Add FMODE_EXEC to SLES10 SP1 server kernel series.
1221 Severity : enhancement
1223 Description: Add server support for vanilla-2.6.22.14.
1226 Frequency : occasional
1228 Description: Avoid lov_create() getting stuck in obd_statfs_rqset()
1229 Details : If an OST is down the MDS will hang indefinitely in
1230 obd_statfs_rqset() waiting for the statfs data. While for
1231 MDS QOS usage of statfs, it should not stuck in waiting.
1233 Severity : enhancement
1235 Description: Disable adaptive timeouts by default
1238 Frequency : on network error
1240 Description: panic with double free request if network error
1241 Details : mdc_finish_enqueue is finish request if any network error occuring,
1242 but it's true only for synchronus enqueue, for async enqueue
1243 (via ptlrpcd) this incorrect and ptlrpcd want finish request
1247 Frequency : rare, on recovery
1249 Description: read procfs can produce deadlock in some situation
1250 Details : Holding lprocfs lock with send rpc can produce block for destroy
1251 obd objects and this also block reconnect with -EALREADY.
1252 This isn't fix all lprocfs bugs - but make it rare.
1254 Severity : enhancement
1256 Description: Update kernel to RHEL5 2.6.18-53.1.14.el5.
1259 Frequency : frequent on X2 node
1261 Description: mdc_set_open_replay_data LBUG
1262 Details : Set replay data for requests that are eligible for replay.
1267 Description: lustre_mgs: operation 101 on unconnected MGS
1268 Details : When MGC is disconnected from MGS long enough, MGS will evict the
1269 MGC, and late on MGC cannot successfully connect to MGS and a lot
1270 of the error messages complaining that MGS is not connected.
1273 Frequency : rare, depends on device drivers and load
1275 Description: MDS or OSS nodes crash due to stack overflow
1276 Details : Code changes in 1.6.4 increased the stack usage of some functions.
1277 In some cases, in conjunction with device drivers that use a lot
1278 of stack, the MDS (or possibly OSS) service threads could overflow
1279 the stack. One change which was identified to consume additional
1280 stack has been reworked to avoid the extra stack usage.
1282 Severity : enhancement
1284 Description: Update to RHEL5 latest kernel-2.6.18-53.1.13.el5.
1286 Severity : enhancement
1288 Description: Update to SLES10 SP1 latest kernel-2.6.16.54-0.2.5.
1290 Severity : enhancement
1292 Description: Update to RHEL5 latest kernel-2.6.18-53.1.6.el5.
1294 Serverity : enhancement
1296 Description: Update RHEL4 kernel to 2.6.9-67.0.4.
1299 Frequency : rare on shutdown OST
1301 Description: Don't allow skipping OSTs if index has been specified.
1302 Details : Don't allow skipping OSTs if index has been specified, make locking
1303 in internal create lots better.
1308 Description: ASSERTION(!PageDirty(page)) failed
1309 Details : Wrong check could lead to an assertion failure under specific
1315 Description: LBUG in ptlrpc_check_set() bad phase ebc0de00
1316 Details : access to bitfield in structure is always rounded to long
1317 and this produce problem with not atomic change any bit.
1322 Description: Lustre make rpms failed.
1323 Details : Remove ldiskfs spec file to avoids rpmbuild be confused when
1324 builds Lustre rpms from tarball.
1326 Severity : enhancement
1328 Description: Update to SLES9 SP4 kernel-2.6.5-7.308.
1331 Frequency : rare on shutdown OST
1333 Description: If llog cancel was not send before clean_exports phase, this can
1334 produce deadlock in llog code.
1335 Details : If llog thread has last reference to obd and call class_import_put
1336 this produce deadlock because llog_cleanup_commit_master wait when
1337 last llog_commit_thread exited, but this never success because was
1338 called from llog_commit_thread.
1341 Frequency : only if OST index is skipped
1343 Description: NULL lov_tgts causing MDS oops
1344 Details : more safe checks for NULL lov_tgts for avoid oops.
1346 Severity : enhancement
1348 Description: Update to RHEL4 latest kernel-2.6.9-67.0.1.EL.
1350 Severity : enhancement
1352 Description: Update to RHEL5 latest kernel-2.6.18-53.1.4.el5.
1357 Description: make mgs_setparam() handle fsname containing dash
1358 Details : fsname containing a dash does not work with lctl conf_param
1360 Severity : enhancement
1362 Description: Update to RHEL4 Update-6 kernel-2.6.9-67.EL.
1365 Frequency : rare, in recovery and (or) destroy lovobjid file.
1367 Description: rewrite lov objid code.
1368 Details : Cleanup for lov objid code, remove scability problems and wrong
1369 locking. Fix sending last_id into ost.
1371 Severity : enhancement
1373 Description: Update to SLES10 SP1 latest kernel-2.6.16.54-0.2.3.
1375 Severity : enhancement
1377 Description: Update to RHEL5 Update-1 kernel 2.6.18-53.el5.
1378 Details : Use d_move_locked instead of __d_move.
1381 Frequency : rare, at shutdown
1383 Description: access already free / zero obd_namespace.
1384 Details : if client_disconnect_export was called without force flag set,
1385 and exist connect request in flight, this can produce access to
1386 NULL pointer (or already free pointer) when connect_interpret
1387 store ocd flags in obd_namespace.
1390 Frequency : only at startup
1392 Description: not alloc memory with spinlock held.
1393 Details : allocation memory with GFP_KERNEL can produce sleep deadlock,
1394 if any spinlock held.
1399 Description: lfs find does not continue on file error
1400 Details : Continue other files processing when a file/dir is absent.
1404 Description: Inconsistent usage of lustre_pack_reply()
1405 Details : Standardize the usage of lustre_pack_reply() such that it
1406 always generate a CERROR on failure.
1409 Frequency : very rare
1411 Description: Fix replay if there is an un-replied request and open
1412 Details : In some cases, older replay request will revert the
1413 mcd->mcd_last_xid on MDS which is used to record the client's
1414 latest sent request.
1416 Severity : enhancement
1418 Description: Update to RHEL5 kernel 2.6.18-8.1.15.el5.
1420 Severity : enhancement
1422 Description: Update to SLES10 SP1 kernel 2.6.16.53-0.16
1424 Severity : enhancement
1426 Description: Update to SLES9 kernel-2.6.5-7.287.3.
1428 Severity : enhancement
1430 Description: Update to RHEL4 kernel-2.6.9-55.0.12.EL.
1432 Severity : enhancement
1434 Description: Build SLES10 patchless client fails
1435 Details : The configure was broken by run ./configure with
1436 --with-linux-obj=.... argument for patchless client. When the
1437 configure use --with-linux-obj, the LINUXINCLUDE= -Iinclude
1438 can't search header adequately. Use absolute path such as
1439 -I($LINUX)/include instead.
1441 Severity : enhancement
1443 Description: Lustre Page Accounting
1444 Details : New macros for page alloc and free which enable accounting
1445 of page allocation of Lustre. Use percpu counters to store memory
1446 and page statistics.
1449 Frequency : only if debugging is disabled
1451 Description: LASSERT_{REQ,REP}SWAB macros are buggy
1452 Details : If SWAB_PARANOIA is disabled, the LASSERT_REQSWAB and
1453 LASSERT_REPSWAB macros become no-ops, which is incorrect. Drop
1454 these macros and replace them with their definitions instead.
1459 Description: interrupt oig_wait produce painc on resend.
1460 Details : brw_redo_request can be used for resend requests from ptlrpcd and
1461 private set, and this produce situation when rq_ptlrpcd_data not
1462 copyed to new allocated request and triggered LBUG on assert
1463 req->rq_ptlrpcd_data != NULL. But this member used only for wakeup
1464 ptlrpcd set if request is changed and can be safety changed to use
1467 Severity : enhancement
1469 Description: organize the server-side client stats on per-nid basis
1470 Details : Change the structure of stats under obdfilter and mds to
1478 The "uuid"s file would list the uuids of _active_ exports.
1479 And the clear entry is to clear all stats and stale nids.
1484 Description: Processes looping in ll_readdir() on Lustre clients finally causing
1485 a full node pseudo-hang
1486 Details : Concurrent access to the same directory from multiple clients with
1487 intensive file creation/removal can cause a client node to spin in
1488 ll_readdir(). i_version must be increased every time the lock is
1489 cancelled to ensure a revalidate is done.
1494 Description: touch file failed when fs is not full
1495 Details : OST in recovery should not be discarded by MDS in alloc_qos(),
1496 otherwise we can get ENOSP while fs is not full.
1499 Frequency : only for Cray XT3
1500 Bugzilla : 12829/13455
1501 Description: Changing primary group doesn't change the group lustre assigns to
1503 Details : When CRAY_XT3 is defined, the fsgid supplied by the client is
1504 overridden with the primary group provided by the group upcall,
1505 whereas the supplied fsgid can be trusted if it is in the list of
1506 supplementary groups returned by the group upcall.
1508 Severity : enhancement
1510 Description: Root Squash Functionality
1511 Details : Implementation of NFS-like root squash capability. Specifically,
1512 don't allow someone with root access on a client node to be able
1513 to manipulate files owned by root on a server node.
1515 Severity : enhancement
1517 Description: Slow trucate/writes to huge files at high offsets.
1518 Details : Directly associate cached pages to lock that protect those pages,
1519 this allows us to quickly find what pages to write and remove
1520 once lock callback is received.
1525 Description: Too many locks accumulating on client during NFS usage
1526 Details : mds_open improperly used accmode to find out access mode to a
1527 file. Also mdc_intent_lock logic to find out if we already have
1528 lock similar to just received was flawed since introduction of
1529 skiplists - locks are now added to the front of the granted
1534 Description: Hit ASSERTION(obd->obd_stopping == 1) failed in some setup failed
1536 Details : In obd setup failure handler, obd_stopping will not necessarily to
1537 be 1, and obd_set_up should also be checked to make sure whether
1538 obd is completely setup.
1540 Severity : enhancement
1542 Description: Allow masking D_WARNING, D_ERROR messages from console
1543 Details : Console messages can now be disabled via lnet.printk.
1548 Description: User code with malformed file open parameter crashes client node
1549 Details : Before packing join_file req, all the related reference should be
1550 checked carefully in case some malformed flags cause fake join_file
1556 Description: shrink/enlarge qunit size when needed; fix the problem of coarse
1557 grain of quota doing harm to quota's accuracy
1558 Details : qunit size will be changed when quota limitation is too low/high;
1559 record the pending quota write in order to get more accureate
1560 quota; delete the patch for bug12588, which is unnecessary when
1561 this patch is landed. This bug also contains fixes for bug 14526,
1562 14299, 14601 and 13794.
1566 Description: LDLM_ENQUEUE races with LDLM_CP_CALLBACK
1567 Details : ldlm_completion_ast() assumes that a lock is granted when the req
1568 mode is equal to the granted mode. However, it should also check
1569 that LDLM_FL_CP_REQD is not set.
1573 Description: Heavy nfs access might result in deadlocks
1574 Details : After ELC code landed, it is now improper to enqueue any mds
1575 locks under och_sem, because enqueue might want to decide to
1576 cancel open locks for same inode we are holding och_sem for.
1580 Description: 35% write performance drop with ldiskfs2 when quotas are on
1581 Details : Enable ext3 journalled quota by default to improve performance
1582 when quotas are turned on.
1586 Description: Client eviction while running blogbench
1587 Details : A lot of unlink operations with concurrent I/O can lead to a
1588 deadlock causing evictions. To address the problem, the number of
1589 oustanding OST_DESTROY requests is now throttled to
1590 max_rpcs_in_flight per OSC and LDLM_FL_DISCARD_DATA blocking
1591 callbacks are processed in priority.
1594 Frequency : RHEL4 only
1596 Description: mkfs is very slow on IA64/RHEL4
1597 Details : A performance regression has been discovered in the MPT Fusion
1598 driver between versions 3.02.73rh and 3.02.99.00rh. As a
1599 consequence, we have downgraded the MPT Fusion driver in the RHEL4
1600 kernel from 3.02.99.00 to 3.02.73 until this problem is fixed.
1603 Frequency : PPC/PPC64 only
1605 Description: conflicts between asm-ppc64/types.h and lustre_types.h
1606 Details : fix duplicated definitions between asm-ppc64/types.h and
1607 lustre_types.h on PPC.
1610 Frequency : PPC/PPC64 only
1612 Description: asm-ppc/segment.h does not exist
1613 Details : fix compile issue on PPC.
1617 Description: data checksumming impacts single node performance
1618 Details : add support for several checksum algorithms. Currently, CRC32 and
1619 Adler-32 are supported. The checksum type can be changed on the fly
1620 through /proc/fs/lustre/osc/*/checksum_type.
1624 Description: use adler32 for page checksums
1625 Details : when available, use the Adler-32 algorithm instead of CRC32 for
1630 Description: better handle error messages in extents code
1632 Severity : enhancement
1634 Description: SNMP support enhancement
1635 Details : Adding total number of sampled request for an MDS node in snmp
1638 Severity : enhancement
1640 Description: Optimize ldlm waiting list processing for PR extent locks
1641 Details : When processing waiting list for read extent lock and meeting read
1642 lock that is same or wider to it that is not contended, skip
1643 processing rest of the list and immediatelly return current
1644 status of conflictness, since we are guaranteed there are no
1645 conflicting locks in the rest of the list.
1649 Description: Time out and refuse to reconnect
1650 Details : When the failover node is the primary node, it is possible
1651 to have two identical connections in imp_conn_list. We must
1652 compare not conn's pointers but NIDs, otherwise we
1653 can defeat connection throttling.
1657 Description: Client not clear own cache if answer to reconnect is lost.
1658 Details : Client gets evicted from server. Now client also thinks it is
1659 disconnected (or gets ENOTCONN on its operation) and decides to
1660 reconnect. Server receives reconnect message, but cannot find
1661 export. New export is created that is fully valid (new cookie!),
1662 but reply is lost and not reported to client. Client reconnects
1663 again and gets back a just-created connection, but it is not new
1664 so client thinks it was not evicted and does not do recovery.
1668 Description: Detect stride IO mode in read-ahead
1669 Details : When a client does stride read, read-ahead should detect that and
1670 read-ahead pages according to the detected stride pattern.
1674 Description: build for x2 fails
1675 Details : fix compile issue on Cray systems.
1677 Severity : enhancement
1679 Description: implement readv/writev APIs(aio_read/aio_writes in newer kernels)
1680 Details : This greatly improves speed of NFS writes on 2.6 kernels.
1683 Frequency : only on PPC/SLES10
1685 Description: "BITS_PER_LONG is not 32 or 64" in linux/idr.h
1686 Details : On SLES10/PPC, fs.h includes idr.h which requires BITS_PER_LONG to
1687 be defined. Add a hack in mkfs_lustre.c to work around this compile
1692 Description: LASSERT on MDS when client holding flock lock dies
1693 Details : ldlm pool logic depends on number of granted locks equal to
1694 number of released locks which is not true for flock locks, so
1695 just exclude such locks from consideration.
1699 Description: MDS deadlock with many ll_sync_lov threads and I/O stalled
1700 Details : Use fsfilt_sync() for both the whole filesystem sync and
1701 individual file sync to eliminate dangerous inode locking
1702 with I_LOCK that can lead to a deadlock.
1706 Description: Update an obsolete wirecheck.c generator
1707 Details : Update wirecheck.c/wirehdr.c and regenerate wiretest.c
1711 Description: Client can panic on open sometimes
1712 Details : It is possible that we try to free already freed request in
1713 ll_file_open in some error cases when we send request from
1718 Description: performance in 1.6.3
1719 Details : Force q->max_phys_segments to MAX_PHYS_SEGMENTS on SLES10 to be
1720 sure that 1MB requests are not fragmented by the block layer.
1724 Description: LDLM soft lockups - improvement
1725 Details : It is be possible to send the lock handle along with each read
1726 or write request because the client is already doing a lock match
1727 itself so there isn't any reason the OST should have to re-do that
1733 Description: lfs quota fails with deactivated OSTS
1734 Details : With this patch, three improvements are included:
1735 1. detete the softlimit in mds and osts when use "lfs quota".
1736 2. display the inaccurate data in the output of "lfs quota".
1737 3. try to get quota info when "lfs quota" is executed.
1742 Description: Extent locks not granted with no conflicts sometimes.
1743 Details : When race occurs in glimpse handler and nothing is returned,
1744 we do not reprocess the queue after lock cancel, and that leads
1745 to a stall until next activity on a resource
1748 Frequency : failover with quotaon
1750 Description: during mds failovers with quota on, OSTs got into deadlock state
1751 and causing dumpstack.
1752 Details : for every quota slave, at any time, there is only one quota req
1753 is sent to quota master for every uid/gid. Before that quota req
1754 returns, all the thread relative to the same uid/gid will wait.
1755 So if the quota req is lost because mds failovers or any other
1756 reasons, this bug will be hit. Now, dqacq_interpret() will handle
1757 quota reqs who time out.
1759 Severity : enhancement
1762 Description: when quota slave checks if quota is enough, there is an unnecessary
1764 Details : place this wait on necessary place instead of always waiting.
1766 --------------------------------------------------------------------------------
1768 2007-12-07 Cluster File Systems, Inc. <info@clusterfs.com>
1770 * Support for kernels:
1771 2.6.5-7.286 (SLES 9),
1772 2.6.9-55.0.9.EL (RHEL 4),
1773 2.6.16.53-0.8 (SLES 10),
1774 2.6.18-8.1.14.el5 (RHEL 5),
1775 2.6.18.8 vanilla (kernel.org)
1776 * Client support for unpatched kernels:
1777 (see http://wiki.lustre.org/index.php?title=Patchless_Client)
1778 2.6.16 - 2.6.22 vanilla (kernel.org)
1779 * Due to recently discovered recovery problems, we do not recommend
1780 using patchless RHEL 4 clients with this or any earlier release.
1781 * Recommended e2fsprogs version: 1.40.2-cfs1
1782 * Note that reiserfs quotas are disabled on SLES 10 in this kernel.
1785 Frequency : occasional
1787 Description: MDS hang or stay in waiting lock
1788 Details : If client receive lock with CBPENDING flag ldlm need send lock
1789 cancel as separate rpc, to avoid situation when cancel request
1790 can't processed due all i/o threads stay in wait lock.
1793 Frequency : occasional
1795 Description: Do not fail import if osc_interpret_create gets -EAGAIN
1796 Details : If osc_interpret_create got -EAGAIN it immediately exits and
1797 wakeup oscc_waitq. After wakeup oscc_wait_for_objects call
1798 oscc_has_objects and see OSC has no objests and call
1799 oscc_internal_create to resend create request.
1801 Severity : enhancement
1803 Description: Update kernel patches for SLES10 2.6.16.53-0.8.
1804 Details : Update which_patch & target file for SLES10 latest kernel.
1806 Severity : enhancement
1808 Description: add --type and --size parameters to lfs find
1809 Details : Enhance lfs find by adding filetype and filesize parameters. Also
1810 multiple OBDs can now be specified for the --obd option.
1812 Severity : enhancement
1814 Description: eliminate client locks in face of contention
1815 Details : file contention detection and lockless i/o implementation
1816 for contended files.
1818 Severity : enhancement
1820 Description: Remove client patches from SLES 10 kernel.
1821 Details : This causes SLES 10 clients to behave as patchless clients
1822 even on a Lustre-patched (server) kernel.
1824 Severity : enhancement
1826 Description: use i_size_read and i_size_write in 2.6 port
1827 Details : replace inode->i_size access with i_size_read/write()
1830 Frequency : when removing large files
1832 Description: scheduling issue during removal of large Lustre files
1833 Details : Don't take the BKL in fsfilt_ext3_setattr() for 2.6 kernels.
1834 It causes scheduling issues when removing large files (17TB in the
1840 Description: 1.4.11 Can't handle directories with stripe set and extended ACLs
1841 Details : Impossible (EPROTO is returned) to access a directory that has a
1842 non-default striping and ACLs.
1845 Frequency : only on ppc
1847 Description: /proc/fs/lustre/devices broken on ppc
1848 Details : The patch as applied to 1.6.2 doesn't look correct for all arches.
1849 We should make sure the type of 'index' is loff_t and then cast
1850 explicitly as needed below. Do not assign an explicitly cast
1854 Frequency : only for rhel5
1856 Description: Kernel patches update for RHEL5 2.6.18-8.1.10.el5.
1857 Details : Modify the target file & which_kernel.
1860 Frequency : if the uninit_groups feature is enabled on ldiskfs
1862 Description: e2fsck reports "invalid unused inodes count"
1863 Details : If a new ldiskfs filesystem is created with the "uninit_groups"
1864 feature and only a single inode is created in a group then the
1865 "bg_unused_inodes" count is incorrectly updated. Creating a
1866 second inode in that group would update it correctly.
1871 Description: buffer overruns could theoretically occur
1872 Details : llapi_semantic_traverse() modifies the "path" argument by
1873 appending values to the end of the origin string, and a buffer
1874 overrun may occur. Adding buffer overrun check in liblustreapi.
1876 Severity : enhancement
1878 Description: Add jbd statistics patch for RHEL5 and 2.6.18-vanilla.
1881 Frequency : only if filesystem is inconsistent
1883 Description: handle "serious error: objid * already exists" more gracefully
1884 Details : If LAST_ID value on disk is smaller than the objects existing in
1885 the O/0/d* directories, it indicates disk corruption and causes an
1886 LBUG(). If the object is 0-length, then we should use the existing
1887 object. This will help to avoid a full fsck in most cases.
1889 Severity : enhancement
1891 Description: Kernel patches update for RHEL4 2.6.9-55.0.6.
1892 Details : Modify vm-tunables-rhel4.patch.
1894 Severity : enhancement
1896 Description: Kernel config for 2.6.18-vanilla.
1897 Details : Modify targets/2.6-vanilla.target.in.
1898 Add config file kernel-2.6.18-2.6-vanilla-i686.config.
1899 Add config file kernel-2.6.18-2.6-vanilla-i686-smp.config.
1900 Add config file kernel-2.6.18-2.6-vanilla-x86_64.config.
1901 Add config file kernel-2.6.18-2.6-vanilla-x86_64-smp.config.
1904 Frequency : occasional
1906 Description: improve handling recoverable errors
1907 Details : If request processed with error which can be recoverable on server
1908 request should be resend, otherwise page released from cache and
1914 Description: Kernel patches update for RHEL5 2.6.18-8.1.14.el5.
1915 Details : Modify target file & which_patch.
1916 A flaw was found in the IA32 system call emulation provided
1917 on AMD64 and Intel 64 platforms. An improperly validated 64-bit
1918 value could be stored in the %RAX register, which could trigger an
1919 out-of-bounds system call table access. An untrusted local user
1920 could exploit this flaw to run code in the kernel
1921 (ie a root privilege escalation). (CVE-2007-4573).
1925 Description: change order of libsysio includes
1926 Details : '#include sysio.h' should always come before '#include xtio.h'
1928 Severity : enhancement
1930 Description: adapt the lustre_config script to support the upgrade case
1931 Details : Add "-u" option for lustre_config script to support upgrading 1.4
1932 server targets to 1.6 in parallel.
1937 Description: To avoid grant space > avaible space when the disk is almost
1938 full. Without this patch you might see the error "grant XXXX >
1939 available" or some LBUG about grant, when the disk is almost
1941 Details : In filter_check_grant, for non_grant cache write, we should
1942 check the left space by if (*left > ungranted + bytes), instead
1943 of (*left > ungranted), because only we are sure the left space
1944 is enough for another "bytes", then the ungrant space should be
1945 increase. In client, we should update cl_avail_grant only there is
1946 OBD_MD_FLGRANT in the reply.
1951 Description: Update RHEL 4 kernel to fix local root privilege escalation.
1952 Details : Update to the latest RHEL 4 kernel to fix the vulnerability
1953 described in CVE-2007-4573. This problem could allow untrusted
1954 local users to gain root access.
1957 Frequency : when using O_DIRECT and quotas
1959 Description: Incorrect file ownership on O_DIRECT output files
1960 Details : block usage reported by 'lfs quota' does not take into account
1961 files that have been written with O_DIRECT.
1965 Description: (rw.c:1323:ll_read_ahead_pages()) ASSERTION(page_idx > ria->ria_stoff) failed
1966 Details : Once the unmatched stride IO mode is detected, shrink the stride-ahead
1967 window to 0. If it does hit cache miss, and read-pattern is still
1968 stride-io mode, does not reset the stride window, but also does not
1969 increase the stride window length in this case.
1971 --------------------------------------------------------------------------------
1973 2007-09-27 Cluster File Systems, Inc. <info@clusterfs.com>
1975 * Support for kernels:
1976 2.6.5-7.286 (SLES 9),
1977 2.6.9-55.0.2.EL (RHEL 4),
1978 2.6.16.46-0.14 (SLES 10),
1979 2.6.18-8.1.8.el5 (RHEL 5),
1980 2.6.18.8 vanilla (kernel.org)
1981 * Client support for unpatched kernels:
1982 (see http://wiki.lustre.org/index.php?title=Patchless_Client)
1983 2.6.16 - 2.6.21 vanilla (kernel.org)
1984 * Due to recently discovered recovery problems, we do not recommend
1985 using patchless RHEL 4 clients with this or any earlier release.
1986 * Recommended e2fsprogs version: 1.40.2-cfs1
1987 * Note that reiserfs quotas are disabled on SLES 10 in this kernel.
1991 Description: Fix errors in lfs documentation
1992 Details : Fixes man pages
1994 Severity : enhancement
1996 Description: Adaptive timeouts
1997 Details : RPC timeouts adapt to changing server load and network
1998 conditions to reduce resend attempts and improve recovery time.
2000 Severity : enhancement
2002 Description: llapi_file_create() does not allow some changes
2003 Details : add llapi_file_open() that allows specifying the file creation
2004 mode and open flags, and also returns an open file handle.
2006 Severity : enhancement
2008 Description: df doesn't work properly if diskfs blocksize != 4K
2009 Details : Choose biggest blocksize of OST's as the LOV's blocksize.
2012 Frequency : always on directories with default striping set
2014 Description: lfs find on -1 stripe looping in lsm_lmm_verify_common()
2015 Details : Avoid lov_verify_lmm_common() on directory with -1 stripe count.
2017 Severity : enhancement
2019 Description: merge and cleanup kernel patches.
2020 Details : Remove mnt_lustre_list in vfs_intent-2.6-rhel4.patch.
2022 Severity : enhancement
2024 Description: RedHat Update kernel for RHEL5
2025 Details : Kernel config file for RHEL5.
2027 Severity : enhancement
2029 Description: OSS needs mutliple precreate threads
2030 Details : Add ability to start more than one create thread per OSS.
2033 Frequency : only with quota on the root user
2035 Description: mds_obd_create error creating tmp object
2036 Details : When the user sets quota on root, llog will be affected and can't
2037 create files and write files.
2040 Frequency : Always on ia64 patchless client, and possibly others.
2042 Description: Add EXPORT_SYMBOL check for node_to_cpumask symbol.
2043 Details : This allows the patchless client to be loaded on architectures
2044 without this export.
2049 Description: disorder of journal start and llog_add cause deadlock.
2050 Details : in llog_origin_connect, journal start should happen before
2051 llog_add keep the same order as other functions to avoid
2054 Severity : enhancement
2056 Description: RedHat Update kernel for RHEL5
2057 Details : Modify the kernel config file more closer RHEL5.
2059 Severity : enhancement
2061 Description: Build failure against Centos5 (RHEL5)
2062 Details : Define PAGE_SIZE when it isn't present.
2065 Frequency : occasionally when using NFS
2067 Description: "ll_intent_file_open()) lock enqueue: err: -13" with nfs
2068 Details : with NFS, the anon dentry's parent was set to itself in
2069 d_alloc_anon(), so in MDS, we use rec->ur_fid1 to find the
2070 corresponding dentry other than use rec->ur_name.
2072 Severity : enhancement
2074 Description: enable data checksumming by default
2075 Details : enable checksum by default, allow --disable-checksum
2076 configure option and "-o nochecksum" mount option. Checksums
2077 can also be disabled at runtime via $LPROC/osc/*/checksum_pages.
2080 Frequency : Occasionally with failover
2082 Description: Client eviction due to failover config
2083 Details : after a connection loss, the lustre client should attempt to
2084 reconnect to the last active server first before trying the
2085 other potential connections.
2087 Severity : enhancement
2089 Description: client-side metadata stat-ahead during readdir(directory readahead)
2090 Details : perform client-side metadata stat-ahead when the client detects
2091 readdir and sequential stat of dir entries therein
2096 Description: when mds and osts use different quota unit(32bit and 64bit),
2097 quota will be released repeatly.
2098 Details : void sending multiple quota reqs to mds, which will keep the
2099 status between the reqs.
2102 Frequency : only with liblustre clients on XT3
2104 Description: evictions taking too long
2105 Details : allow llrd to evict clients directly on OSTs
2109 Description: osts not allocated evenly to files
2110 Details : change the condition to increase offset_idx
2112 Severity : enhancement
2114 Description: self-adjustable client's lru lists
2115 Details : use adaptive algorithm for managing client cached locks lru
2116 lists according to current server load, other client's work
2117 pattern, memory activities, etc. Both, server and client
2118 side namespaces provide number of proc tunables for controlling
2123 Description: rewrite ext2-derived code in llite/dir.c and obdclass/uuid.c
2124 Details : rewrite inherited code (uuid parsing code from ext2 utils and
2125 readdir code from ext3) from scratch preserving functionality.
2129 Description: Only those disconnect error should be returned by rq_status.
2130 Details : In open/enqueue processs, Some errors, which will cause client
2131 disconnected, should be returned by rq_status, while other
2132 errors should still be returned by intent, then mdc or llite will
2135 Severity : enhancement
2137 Description: Tune the kernel for good SCSI performance.
2138 Details : Set the value of /sys/block/{dev}/queue/max_sectors_kb
2139 to the value of /sys/block/{dev}/queue/max_hw_sectors_kb
2143 Frequency : Always for filesystems larger than 2TB on 32-bit systems.
2144 Bugzilla : 13547, 13627
2145 Description: Data corruption for OSTs that are formatted larger than 2TB
2147 Details : When generating the bio request for lustre file writes the
2148 sector number would overflow a temporary variable before being
2149 used for the IO. The data reads correctly from Lustre (which
2150 will overflow in a similar manner) but other file data or
2151 filesystem metadata may be corrupted in some cases.
2155 Description: "lfs find -obd UUID" prints directories
2156 Details : "lfs find -obd UUID" will return all directory names instead
2157 of just file names. It is incorrect because the directories
2158 do not reside on the OSTs.
2161 Bugzilla : 13596, 13614
2162 Description: MDS hang after unclean shutdown of lots of clients
2163 Details : Never resend AST requests.
2167 Frequency : Always, for kernels after 2.6.16
2168 Description: Fix warning idr_remove called for id=.. which is not allocated.
2169 Details : Last kernels save old s_dev before kill super and not allow
2170 to restore from callback - restore it before call kill_anon_super.
2174 Description: Fix error on 'ls .' at the top of the Lustre mount.
2175 Details : Don't revalidate dentry if it is a root dentry.
2177 --------------------------------------------------------------------------------
2179 2007-08-27 Cluster File Systems, Inc. <info@clusterfs.com>
2181 * Support for kernels:
2182 2.6.5-7.286 (SLES 9),
2183 2.6.9-55.0.2.EL (RHEL 4),
2184 2.6.16.46-0.14 (SLES 10),
2185 2.6.18-8.1.8.el5 (RHEL 5),
2186 2.6.18.8 vanilla (kernel.org)
2187 * Client support for unpatched kernels:
2188 (see http://wiki.lustre.org/index.php?title=Patchless_Client)
2189 2.6.16 - 2.6.22 vanilla (kernel.org)
2190 * Due to recently discovered recovery problems, we do not recommend
2191 using patchless RHEL 4 clients with this or any earlier release.
2192 * Recommended e2fsprogs version: 1.39.cfs8
2193 * Note that reiserfs quotas are disabled on SLES 10 in this kernel.
2198 Description: block reactivating mgc import until all deactivates complete
2199 Details : Fix race when failing back MDT/MGS to itself (testing)
2201 Severity : enhancement
2203 Description: lfs setstripe enhancement
2204 Details : Make lfs setstripe understand 'k', 'm' and 'g' for stripe size.
2207 Frequency : mds/oss recovery
2209 Description: llog ctxt is refrenced after it has been freed.
2210 Details : llog ctxt refcount was added to avoide the race between ctxt free
2211 and llog recovery process. Each llog user must hold ctxt refcount
2212 before it access the llog. And the llog ctxt can only be freed
2213 when its refcount is zero.
2215 Severity : enhancement
2217 Description: randomly memory allocation failure util
2218 Details : Make lustre randomly failed allocating memory for testing purpose.
2221 Frequency : only for SLES10
2223 Description: Update kernel patch for SLES10 SP1
2224 Details : Add patch blkdev_tunables-2.6-sles10.patch to 2.6-sles10.series.
2226 Severity : enhancement
2228 Description: omit set fsid for export NFS
2229 Details : fix set/restore device id for avoid EMFILE error and mark lustre fs
2230 as FS_REQUIRES_DEV for avoid problems with generate fsid.
2233 Frequency : after network failures
2235 Description: Add sync option to mount_lustre.c
2236 Details : Client loses data written to lustre after a network interruption.
2238 Severity : enhancement
2240 Description: Error message improvement.
2241 Details : Merging of two LCONSOLE_ERROR_MSG into one.
2243 Severity : enhancement
2245 Description: don't use GFP_* in generic Lustre code.
2246 Details : Use cfs_alloc_* functions and CFS_* flags for code portability.
2248 Severity : enhancement
2250 Description: obdclass is limited by single OBD_ALLOC(idarray)
2251 Details : replace OBD_ALLOC/OBD_FREE with OBD_VMALLOC/OBD_VFREE
2253 Severity : enhancement
2255 Description: updated patchess for new RHEL4 kernel
2256 Details : Fixed ext3-unlink-race.patch per Kalpak's comment.
2258 Severity : enhancement
2260 Description: warnings with build patchless client with vanila 2.6.19 and up
2261 Details : change old ctl_table style and replace ctl_table/ctl_table_header
2262 with cfs_sysctl_table_t/cfs_sysctl_table_header_t
2264 Severity : enhancement
2266 Description: O_DIRECT bypasses client statistics.
2267 Details : When running with O_DIRECT I/O, neither the client rpc_stats nor
2268 read_ahead_stats were updated. Copied stats section from
2269 osc_send_oap_rpc() into async_internal().
2272 Frequency : only for Cray XT3
2274 Description: peer credits not enough on many OST per OSS systems.
2275 Details : Use new lnet way to add credits as we need those for pings and ASTs
2278 Frequency : only with liblustre
2280 Description: Liblustre is not releasing flock locks on file close.
2281 Details : Release flock locks on file close.
2284 Frequency : only for RHEL4
2286 Description: Update kernel patches for kernel-2.6.9-55.0.2.EL
2287 Details : Remove inode-nr_unused-2.6.9-rhel4.patch from 2.6-rhel4.series
2288 Update target file and kernel config.
2292 Description: lustre support for RHEL5
2293 Details : Add support for RHEL5.
2298 Description: ASSERTION(export != NULL) failed in target_handle_connect
2299 Details : Assetion hit is result of rare race between disconnect and connet
2300 to same nid. target_handle_connect found old connect cockie and
2301 tried to reconnect, but can't find export for this cockie.
2306 Description: umount blocks forever on error
2307 Details : In result of wrong using obd_no_recov and obd_force flags client
2308 can hand if cancel or some other requests is lost.
2311 Frequency : Only for SLES
2313 Description: sanity_quota fail test_1
2314 Details : There are multiple occurences of $TSTUSR in SLES's /etc/group
2315 file, which makes TSTID[2] inunique.
2317 Severity : enhancement
2319 Description: Kernel patches for SLES9 2.6.5-7.286 kernel
2320 Details : Update target/ChangeLog/which_patch .
2322 Severity : enhancement
2324 Description: jbd statistics
2325 Details : Port older jbd statistics patch for sles10
2330 Description: lvbo_init failed for resource with missing objects.
2331 Details : Fix returning error if we do stat for file with missing/corrupted
2332 objects and i_size set to all sum of size of all avaible objects.
2333 if we to truncate/write to missing object - it is recreated.
2338 Description: Oops in read and write path when failing to allocate lock.
2339 Details : Check if lock allocation failed and return error back.
2342 Frequency : When flocks are used.
2344 Description: assertion failure in ldlm_cli_enquque_fini for non NULL lock.
2345 Details : Flock locks might destroy just granted lock if it could be merged
2346 with another existing flock, this is done in completion handler,
2347 so teach ldlm_cli_enquque_fini that this is a valid case for
2353 Description: reply_lock_interpret crash due to race with it and lock cancel.
2354 Details : Do not replay locks that are being cancelled. Do not reference
2355 locks by their address during replay, just by their handle.
2358 Frequency : only with deactivated OSTs
2360 Description: lstripe command fails for valid OST index
2361 Details : The stripe offset is compared to 'lov->desc.ld_tgt_count'
2362 instead of lov->desc.ld_active_tgt_count.
2364 Severity : enhancement
2366 Description: Build failure against Centos5 (RHEL5)
2367 Details : Use getpagesize() instead of PAGE_SIZE.
2369 --------------------------------------------------------------------------------
2371 2007-07-30 Cluster File Systems, Inc. <info@clusterfs.com>
2373 * Support for kernels:
2374 2.6.5-7.283 (SLES 9),
2375 2.6.9-55.EL (RHEL 4),
2376 2.6.16.46-0.14 (SLES 10),
2377 2.6.18.8 vanilla (kernel.org)
2378 * Client support for unpatched kernels:
2379 (see http://wiki.lustre.org/index.php?title=Patchless_Client)
2380 2.6.16 - 2.6.22 vanilla (kernel.org)
2381 * Due to recently discovered recovery problems, we do not recommend
2382 using patchless RHEL 4 clients with this or any earlier release.
2383 * Recommended e2fsprogs version: 1.39.cfs8
2384 * Note that reiserfs quotas are disabled on SLES 10 in this kernel.
2385 * Starting with this release, the ldiskfs backing filesystem required
2386 by Lustre is now in its own package, lustre-ldiskfs. This package
2387 should be installed. It is versioned separately from Lustre and
2388 may be released separately in future.
2393 Description: server LBUG when shutting down
2394 Details : Block umount forever until the mount refcount is zero rather
2395 than giving up after an arbitrary timeout.
2397 Severity : enhancement
2399 Description: add optional extra BUILD_VERSION info
2400 Details : add a new environment variable (namely LUSTRE_VERS) which allows
2401 to override the lustre version.
2404 Frequency : 2.6.18 servers only
2406 Description: ll_kern_mount() doesn't release the module reference
2407 Details : The ldiskfs module reference count never drops down to 0
2408 because ll_kern_mount() doesn't release the module reference.
2413 Description: server LBUG when using old ost_num_threads parameter
2414 Details : Accept the old ost_num_threads parameter but warn that it
2415 is deprecated, and fix an off-by-one error that caused an LBUG.
2420 Description: Transient SCSI error results in persistent IO issue
2421 Details : iobuf->dr_error is not reinitialized to 0 between two
2425 Frequency : sometimes when underlying device returns I/O errors
2427 Description: OSTs not going read-only during write failures
2428 Details : OSTs are not remounted read-only when the journal commit threads
2429 get I/O errors because fsfilt_ext3 calls journal_start/stop()
2430 instead of the ext3 wrappers.
2433 Frequency : only with 10000 clients or more
2435 Description: poor connect scaling with increasing client count
2436 Details : Don't run filter_grant_sanity_check for more than 100 exports
2437 to improve scaling for large numbers of clients.
2440 Frequency : SLES10 only
2442 Description: sanity-quota.sh quotacheck failed: rc = -22
2443 Details : Quotas cannot be enabled on SLES10.
2446 Frequency : liblustre clients only
2448 Description: getdirentries does not give error when run on compute nodes
2449 Details : getdirentries does not fail when the size specified as an argument
2450 is too small to contain at least one entry
2452 Severity : enhancement
2454 Description: Add LNET router traceability for debug purposes
2455 Details : If a checksum failure occurs with a router as part of the
2456 IO path, the NID of the last router that forwarded the bulk data
2457 is printed so it can be identified.
2462 Description: OST "spontaneously" evicts client; client has imp_pingable == 0
2463 Details : Due to a race condition, liblustre clients were occasionally
2464 evicted incorrectly.
2466 Severity : enhancement
2468 Description: lfs setstripe use optional parameters instead of postional
2471 Severity : enhancement
2473 Description: Nanosecond timestamp support for ldiskfs
2474 Details : The on-disk ldiskfs filesystem has added support for nanosecond
2475 resolution timestamps. There is not yet support for this at
2476 the Lustre filesystem level.
2479 Frequency : during server recovery
2481 Description: MDS failing to send precreate requests due to OSCC_FLAG_RECOVERING
2482 Details : request with rq_no_resend flag not awake l_wait_event if they get a
2486 Frequency : nfs export on patchless client
2488 Description: connectathon hang when test nfs export over patchless client
2489 Details : Disconnected dentry cannot be found with lookup, so we do not need
2490 to unhash it or make it invalid
2493 Description: fix llapi_lov_get_uuids() to allow many OSTs to be returned
2494 Details: : Change llapi_lov_get_uuids() to read the UUIDs from /proc instead
2495 of using an ioctl. This allows lfsck for > 160 OSTs to succeed.
2500 Description: open req refcounting wrong on reconnect
2501 Details : If reconnect happened between getting open reply from server and
2502 call to mdc_set_replay_data in ll_file_open, we will schedule
2503 replay for unreferenced request that we are about to free.
2504 Subsequent close will crash in variety of ways.
2505 Check that request is still eligible for replay in
2506 mdc_set_replay_data().
2511 Description: disable writes to filesystem when reading health_check file
2512 Details : the default for reading the health_check proc file has changed
2513 to NOT do a journal transaction and write to disk, because this
2514 can cause reads of the /proc file to hang and block HA state
2515 checking on a healthy but otherwise heavily loaded system. It
2516 is possible to return to the previous behaviour during configure
2517 with --enable-health-write.
2519 Severity : enhancement
2521 Description: 64-bit inode version
2522 Details: : Add a on-disk 64-bit inode version for ext3 to track changes made
2523 to the inode. This will be required for version-based recovery.
2528 Description: MDS fails to start if a duplicate client export is detected
2529 Details : in some rare cases it was possible for a client to connect to
2530 an MDS multiple times. Upon recovery the MDS would detect this
2531 and fail during startup. Handle this more gracefully.
2533 Severity : enhancement
2535 Description: Add -o localflock option to simulate old noflock behaviour.
2536 Details : This will achieve local-only flock/fcntl locks coherentness.
2541 Description: log_commit_thread vs filter_destroy race leads to crash
2542 Details : Take import reference before releasing llog record semaphore
2547 Description: Wrong request locking in request set processing
2548 Details : ptlrpc_check_set wrongly uses req->rq_lock for proctect add to
2549 imp_delayed_list, in this place should be used imp_lock.
2552 Frequency : when reconnecting
2554 Description: Grant leak when OSC reconnect to OST
2555 Details : When osc reconnect ost, OST(filter) should check whether it
2556 should grant more space to client by comparing fed_grant and
2557 cl_avail_grant, and return the granted space to client instead
2558 of "new granted" space, because client will call osc_init_grant
2559 to update the client grant space info.
2562 Frequency : when client reconnects to OST
2564 Description: Grant leak when OSC does a resend and replays bulk write
2565 Details : When osc reconnect to OST, OST(filter) should clear grant info of
2566 bulk write request, because the grant info while be sync between
2567 OSC and OST when reconnect, and we should ignore the grant info
2568 these of resend/replay write req.
2573 Description: Grant space more than avaiable space sometimes.
2574 Details : When then OST is about to be full, if two bulk writing from
2575 different clients came to OST. Accord the avaliable space of the
2576 OST, the first req should be permitted, and the second one
2577 should be denied by ENOSPC. But if the seconde arrived before
2578 the first one is commited. The OST might wrongly permit second
2579 writing, which will cause grant space > avaiable space.
2582 Frequency : when client is evicted
2584 Description: Grant might be wrongly erased when osc is evicted by OST
2585 Details : when the import is evicted by server, it will fork another
2586 thread ptlrpc_invalidate_import_thread to invalidate the
2587 import, where the grant will be set to 0. While the original
2588 thread will update the grant it got when connecting. So if
2589 the former happened latter, the grant will be wrongly errased
2590 because of this race.
2595 Description: Checking Stale with correct fid
2596 Details : ll_revalidate_it should uses de_inode instead of op_data.fid2
2597 to check whether it is stale, because sometimes, we want the
2598 enqueue happened anyway, and op_data.fid2 will not be initialized.
2600 Severity : enhancement
2602 Description: update patchless client
2603 Details : Add support for patchless client with 2.6.20, 2.6.21 and RHEL 5
2606 Frequency : only with 2.4 kernel
2608 Description: random memory corruption
2609 Details : size of struct ll_inode_info is to big for union inode.u and this
2610 can be cause of random memory corruption.
2615 Description: Memory leak in recovery
2616 Details : Lov_mds_md was not free in an error handler in mds_create_object.
2617 It should also check obd_fail before fsfilt_start, otherwise if
2618 fsfilt_start return -EROFS,(failover mds during mds recovery).
2619 then the req will return with repmsg->transno = 0 and rc = EROFS.
2620 and we met hit the assert LASSERT(req->rq_reqmsg->transno ==
2621 req->rq_repmsg->transno) in ptlrpc_replay_interpret. Fcc should
2622 be freed no matter whether fsfilt_commit success or not.
2625 Frequency : only with large numbers of cores on a single node
2627 Description: Prevents from taking the superblock lock in llap_from_page for
2629 Details : using LL_ORIGIN_REMOVEPAGE origin flag instead of LL_ORIGIN_UNKNOW
2630 for llap_from_page call in ll_removepage() prevents from taking
2631 the superblock lock for a soon died page.
2636 Description: Not check open intent error before release open handle
2637 Details : in some rare cases, the open intent error is not checked before
2638 release open handle, which may cause
2639 ASSERTION(open_req->rq_transno != 0), because it tries to release
2640 the failed open handle.
2645 Description: Set cat log bitmap only after create log success.
2646 Details : in some rare cases, the cat log bitmap is set too early. and it
2647 should be set only after create log success.
2651 Description: Accessing a block bevice can re-enable I/O when Lustre is
2652 tearing down a device.
2653 Details : dev_clear_rdonly(bdev) must be called in kill_bdev() instead of
2658 Description: service threads may hog cpus when there are a lot of requests
2659 Details : Insert cond_resched to give other threads a chance to use some CPU
2664 Description: the cat log was not initialized in recovery
2665 Details : When mds(mgs) do recovery, the tgt_count might be zero, so the
2666 unlink log on mds will not be initialized until mds post
2667 recovery. And also in mds post recovery, the unlink log will
2668 initialization will be done asynchronausly, so there will be race
2669 between add unlink log and unlink log initialization.
2673 Description: brw_stats were being printed incorrectly
2674 Details : brw_stats were being printed as log2 but all of them were not
2675 recorded as log2. Also remove some code duplication arising from
2676 filter_tally_{read,write}.
2680 Frequency : rare, only in recovery.
2681 Description: ASSERTION(req->rq_type != LI_POISON) failed
2682 Details : imp_lock should be held while iterating over imp_sending_list for
2683 prevent destroy request after get timeout in ptlrpc_queue_wait.
2687 Description: replay-single.sh test 52 fails
2688 Details : A lock's skiplist need to be cleanup when it being unlinked
2689 from its resource list.
2694 Description: Short directio read returns full requested size rather than
2696 Details : Direct I/O operations should return actual amount of bytes
2697 transferred rather than requested size.
2699 Severity : enhancement
2701 Description: metadata RPC reduction (e.g. for rm performance)
2702 Details : decrease the amount of synchronous RPC between clients and servers
2703 by canceling conflicing lock before the operation on the client
2704 and packing thier handles into the main operation RPC to server.
2706 Severity : enhancement
2708 Description: add #ifdef HAVE_KERNEL_CONFIG_H
2709 Details : kernels from 2.6.19 not need include linux/config.h, but add
2710 include linux/autoconf.h in commpiler command line.
2712 Severity : enhancement
2714 Description: patchless client support for 2.6.22 kernel
2715 Details : 2.6.22 has only one visble change, SLAB_CTOR_* constants is
2716 removed. In this case we need drop using os depended interface to
2717 kmem_cache and use cfs_mem_cache API.
2722 Description: fix mal-formatted messages
2723 Details : fix some mal-formatted DEBUG_REQ and LCONSOLE_ERROR_MSG messages
2727 Frequency : always in liblustre
2728 Description: wrong IS_ERR implementation in liblustre.h
2729 Details : fix IS_ERR implementation in liblustre.h for right detect errors.
2734 Description: Correct condition for output debug message.
2735 Details : inode i_nlink equal zero is not enough for output message about
2736 disk corruption, i_ctime and i_mode should be also checked.
2740 Frequency : always in patchless client
2741 Description: add configure check for truncate_complete_page
2742 Details : improve checks for exported symbols. This allow run check without
2743 sources, but with Module.symvers shipped with kernel distribution.
2744 add check for truncate_complete_page used by patchless client.
2748 Description: sanity.sh test_77h fails with "test_77h file compare failed"
2749 Details : test_77h uses a file which was messed by other test case.
2753 Description: Not Check whether lov_tgts is NULL in some lov functions
2754 Details : Checking whether lov_tgts is NULL in some functions.
2758 Description: replace obdo_alloc() with OBDO_ALLOC macro
2759 Details : nothing special is done in obdo_alloc() function, and for
2760 debugging purpose, it needs to be replaced with macros.
2764 Description: bad return value and errno from fcntl call
2765 Details : In liblustre API, errno should be a negative value if error
2770 Description: ptlrpc_check_set() LBUG
2771 Details : In case of positive reply from server and failed client bulk
2772 callback after bulk transfer shouldn't LBUG, but process this
2773 request as erroneous.
2775 Severity : enhancement
2777 Description: VFS operations stats tool.
2778 Details : Tool which collects stats by tracking value written in pid,
2779 ppid, gid and uses llstat to generate output to plot graph using
2781 Updated lustre/utils/Makefile.am
2782 Added lustre/utils/ltrack_stats.c
2784 Severity : enhancement
2786 Description: 2.6.18 server support (lustre 1.6.1)
2787 Details : Support for 2.6.18 kernels on the server side.
2792 Description: ASSERTION(imp->imp_conn_current) failed
2793 Details : an assertion failure is hit if a client node boots and attempts to
2794 mount a lustre filesystem faster than RECONNECT_INTERVAL seconds.
2797 Frequency : only for i686
2799 Description: 1.4.11 RC1 build fails for RHEL 4, i686
2800 Details : Fixed config variable for build.
2805 Description: Updated patchess for new RHEL4 kernel
2806 Details : Updated patch inode-nr_unused-2.6.9-rhel4.patch
2807 Updated patch jbd-stats-2.6.9.patch
2808 Updated patch qsnet-rhel4-2.6.patch
2809 Updated patch quota-deadlock-on-pagelock-core.patch
2810 Updated patch vfs_intent-2.6-rhel4.patch
2811 Updated patch vfs_races-2.6-rhel4.patch
2812 Updated series file 2.6-rhel4-titech.series
2813 Updated series file 2.6-rhel4.series
2814 Updated kernel config files
2819 Description: lquota slave complains LBUG when reconnecting with mds
2821 Details : quota slave depends qctxt->lqc_import to send its quota request.
2822 This pointer will be invalid if mds did failover or broke its
2823 connect to osts, which leads to LBUG.
2826 Frequency : when qunit size is too small(less than 20M)
2828 Description: write is stopped by improper -EDQUOT
2829 Details : If the master is busy and qunit size is small enough(let's say 1M),
2830 the slave can not get quota from master on time, which will lead
2831 slave to trigger a -EQUOTA to client.
2836 Description: Deadlock during metadata tests
2837 Details : in prune_dir_dentries(), shrink_dcache_parent() should not be
2838 called with the per-dentry lock held.
2841 Frequency : SLES9 only
2843 Description: Lustre patched kernel for SLES9 SP3 has NR_CPUS set to 8
2844 Details : set CONFIG_NR_CPUS to 128 instead of 8.
2846 Severity : enhancement
2848 Description: remove fs_prep_san_write operation and related patches
2849 Details : remove the ext3-san-jdike patches which are no longer useful.
2854 Description: LDISKFS-fs error (device sdc): ldiskfs_free_blocks
2855 Details : a disk corruption can cause the mballoc code to assert on a
2856 double free or other extent corruptions. Handle these with
2857 ext3_error() instead of assertions.
2860 Frequency : only with mballoc3 code and deep extent trees
2862 Description: ldiskfs_ext_search_right: bad header in inode: unexpected eh_depth
2863 Details : a wrong check of extent headers in ldiskfs_ext_search_right()
2864 can cause the filesystem to be remounted read-only.
2869 Description: lfsck built against 1.4.x cannot run against 1.6.0 lustre
2870 Details : the definition for OBD_IOC_GETNAME changed in 1.6.0. One of the
2871 few external users of this ioctl number is lfsck's call to
2872 llapi_lov_get_uuids() and this caused lfsck to fail at startup.
2873 Add the old ioctl number to the handler so both old and new
2878 Description: parallel lock callbacks
2879 Details : Instead of sending blocking and completion callbacks as separated
2880 requests, adding them to a set and sending in parallel.
2884 Description: Disable most debugging by default
2885 Details : To improve performance, disable most logging (for debug purposes)
2886 by default. VFSTRACE, RPCTRACE, and DLMTRACE are now off by
2887 default, and HA includes fewer messages.
2891 Frequency : only run on patchless client.
2892 Description: use do_facet on sanity.sh for test handling recoverables errors
2893 Details : use do_facet instead of direct use sysctl for set fail_loc on OST
2897 Description: hash tables for lists of nids, connections and uuids
2898 Details : Hash tables noticeably help when a lot of clients connect to a
2899 server, to faster identify duplicate connections or reconnects,
2900 also to faster find export to evict in manual eviction case.
2904 Description: Sometimes, when the server evict a client, and the client will
2905 not be evicted as soon as possible.
2906 Details : In enqueue req, the error was returned by intent, instead of
2907 rq_status which make ptlrpc layer not detect this error, and
2908 does not evict the client. So enqueue error should be returned
2912 Frequency : only at startup
2914 Description: Delay client connections to MDT until fist MDT->OST connect
2915 Details : If a client tried to create a new file before the MDT had
2916 connected to any OSTs, the create would return EIO. Now
2917 the client will simply block until the MDT connects to the
2918 first OST and the create can succeed.
2921 Frequency : only for SLES9
2923 Description: Routinely utilize latest Quadrics drivers in CFS releases
2924 Details : Update patch qsnet-suse-2.6.patch.
2927 Frequency : only for sles10
2929 Description: Update patches for SLES 10 SP1 kernel.
2930 Details : Update the patch vfs_intent-2.6-sles10.patch.
2935 Description: Routinely utilize latest Quadrics drivers in CFS releases
2936 Details : Update patch qsnet-rhel4-2.6.patch.
2939 Frequency : at statup only
2941 Description: mds_lov_synchronize race leads to various problems
2942 Details : simultaneous MDT->OST connections at startup can cause the
2943 sync to abort, leaving the OSC in a bad state.
2948 Description: Using wrong pointer in osc_brw_prep_request
2949 Details : Access to array[-1] can produce panic if kernel compiled with
2950 CONFIG_PAGE_ALLOC enabled
2952 Severity : enhancement
2954 Description: Async OSC create to avoid the blocking unnecessarily.
2955 Details : If a OST has no remain object, system will block on the creating
2956 when need to create a new object on this OST. Now, ways use
2957 pre-created objects when available, instead of blocking on an
2958 empty osc while others are not empty. If we must block, we block
2959 for the shortest possible period of time.
2963 Frequency : only in recovery
2964 Description: Mark OST as early accessible if his start SYNC.
2965 Details : osc_precreate return flag early accessible if oscc marked as
2966 OSCC_FLAG_SYNC_IN_PROGRESS.
2971 Description: Sometimes precreate code can triger create object on wrong ost
2972 Details : Wrong protected or not not restored variables aftre precreate loop
2973 can produce creation object on wrong ost.
2976 Frequency : oss recovery
2978 Description: llog_commit_thread cleanup should sync with llog_commit_thread
2980 Details : llog_commit_thread_count should be synced between llog_commit
2981 start and cleanup, so new llog_commit thread should not be started
2982 when llog_commit threads being stopped to avoid accessing some
2985 Severity : enhancement
2987 Description: Add printing inode info into message about error in writepage.
2989 --------------------------------------------------------------------------------
2991 2007-05-03 Cluster File Systems, Inc. <info@clusterfs.com>
2996 Frequency : on some architectures
2998 Description: 1.6 client sometimes fails to mount from a 1.4 MDT
2999 Details : Uninitialized flags sometimes cause configuration commands to
3003 Frequency : patchless clients only
3005 Description: missing __iget() symbol export
3006 Details : The __iget() symbol export is missing. To avoid the need for
3007 this on patchless clients the deathrow inode reaper is turned
3008 off, and we depend on the VM to clean up old inodes. This
3009 dependency was during via the fix for bug 12181.
3014 Description: sanity.sh fail: test_52b
3015 Details : The ll_inode_to_ext_flags() has a glitch which makes MDS return
3016 incorrect inode's flags to client.
3018 --------------------------------------------------------------------------------
3020 2007-04-19 Cluster File Systems, Inc. <info@clusterfs.com>
3022 * CONFIGURATION CHANGE. This version of Lustre WILL NOT
3023 INTEROPERATE with older versions automatically. In many cases a
3024 special upgrade step is needed. Please read the
3025 user documentation before upgrading any part of a 1.4.x system.
3026 * WARNING: Lustre configuration and startup changes are required with
3027 this release. See https://mail.clusterfs.com/wikis/lustre/MountConf
3029 * Support for kernels:
3030 2.4.21-47.0.1.EL (RHEL 3),
3031 2.6.5-7.283 (SLES 9),
3032 2.6.9-42.0.10.EL (RHEL 4),
3033 2.6.12.6 vanilla (kernel.org),
3034 2.6.16.27-0.9 (SLES10)
3035 * Client support for unpatched kernels:
3036 (see https://mail.clusterfs.com/wikis/lustre/PatchlessClient)
3037 2.6.16 - 2.6.19 vanilla (kernel.org),
3038 2.6.9-42.0.8EL (RHEL 4)
3039 * Recommended e2fsprogs version: 1.39.cfs6
3040 * Note that reiserfs quotas are disabled on SLES 10 in this kernel
3043 Severity : enhancement
3045 Description: Async OSC create to avoid the blocking unnecessarily.
3046 Details : If a OST has no remain object, system will block on the creating
3047 when need to create a new object on this OST. Now, ways use
3048 pre-created objects when available, instead of blocking on an
3049 empty osc while others are not empty. If we must block, we block
3050 for the shortest possible period of time.
3052 Severity : enhancement
3054 Description: MountConf
3055 Details : Lustre configuration is now managed via mkfs and mount
3056 commands instead of lmc and lconf. New obd types (MGS, MGC)
3057 are added for dynamic configuration management. See
3058 https://mail.clusterfs.com/wikis/lustre/MountConf for
3061 Severity : enhancement
3063 Description: dynamic OST addition
3064 Details : OSTs can now be added to a live filesystem
3066 Severity : enhancement
3068 Description: startup order invariance
3069 Details : MDTs and OSTs can be started in any order. Clients only
3070 require the MDT to complete startup.
3072 Severity : enhancement
3074 Description: parallel, asynchronous orphan cleanup
3075 Details : orphan cleanup is now performed in separate threads for each
3076 OST, allowing parallel non-blocking operation.
3078 Severity : enhancement
3080 Description: optimized stripe assignment
3081 Details : stripe assignments are now made based on ost space available,
3082 ost previous usage, and OSS previous usage, in order to try
3083 to optimize storage space and networking resources.
3085 Severity : enhancement
3087 Description: Permanently set tunables
3088 Details : All writable /proc/fs/lustre tunables can now be permanently
3089 set on a per-server basis, at mkfs time or on a live
3092 Severity : enhancement
3094 Description: Lustre message v2
3095 Details : Add lustre message format v2.
3097 Severity : enhancement
3099 Description: client OST exclusion list
3100 Details : Clients can be started with a list of OSTs that should be
3101 declared "inactive" for known non-responsive OSTs.
3105 Description: ENOENT returned for valid filehandle during dbench.
3106 Details : Check if a directory has children when invalidating dentries
3107 associated with an inode during lock cancellation. This fixes
3108 an incorrect ENOENT sometimes seen for valid filehandles during
3109 testing with dbench.
3112 Frequency : SFS test only (otherwise harmless)
3114 Description: SPEC SFS validation failure on NFS v2 over lustre.
3115 Details : Changes the blocksize for regular files to be 2x RPC size,
3116 and not depend on stripe size.
3118 Severity : enhancement
3120 Description: fine-grained SMP locking inside DLM
3121 Details : Improve DLM performance on SMP systems by removing the single
3122 per-namespace lock and replace it with per-resource locks.
3124 Severity : enhancement
3126 Description: don't hold multiple extent locks at one time
3127 Details : To avoid client eviction during large writes, locks are not
3128 held on multiple stripes at one time or for very large writes.
3129 Otherwise, clients can block waiting for a lock on a failed OST
3130 while holding locks on other OSTs and be evicted.
3132 Severity : enhancement
3134 Description: Multiple MD RPCs in flight.
3135 Details : Further unserialise some read-only MDT RPCs - learn about intents.
3136 To avoid overly-overloading MDT, introduce a limit on number of
3137 MDT RPCs in flight for a single client and add /proc controls
3138 to adjust this limit.
3140 Severity : enhancement
3142 Description: client read/write statistics
3143 Details : Add client read/write call usage stats for performance
3144 analysis of user processes.
3145 /proc/fs/lustre/llite/*/offset_stats shows non-sequential
3146 file access. extents_stats shows chunk size distribution.
3147 extents_stats_per_process show chunk size distribution per
3150 Severity : enhancement
3152 Description: per-client statistics on server
3153 Details : Add ldlm and operations statistics for each client in
3154 /proc/fs/lustre/mds|obdfilter/*/exports/
3156 Severity : enhancement
3158 Description: improved MDT statistics
3159 Details : Add detailed MDT operations statistics in
3160 /proc/fs/lustre/mds/*/stats
3162 Severity : enhancement
3164 Description: VFS operations stats
3165 Details : Add client VFS call stats, trackable by pid, ppid, or gid
3166 /proc/fs/lustre/llite/*/stats_track_[pid|ppid|gid]
3171 Description: Fix client-side osc byte counters
3172 Details : The osc read/write byte counters in
3173 /proc/fs/lustre/osc/*/stats are now working
3176 Frequency : always as root on SLES
3178 Description: Failure of copying files with lustre special EAs.
3179 Details : Client side always return success for setxattr call for lustre
3180 special xattr (currently only "trusted.lov").
3185 Description: Refcount LNET uuids
3186 Details : The global LNET uuid list grew linearly with every startup;
3187 refcount repeated list entries instead of always adding to
3190 Severity : enhancement
3192 Description: Dynamic service threads
3193 Details : Within a small range, start extra service threads
3194 automatically when the request queue builds up.
3197 Frequency : mixed-endian client/server environments
3199 Description: mixed-endian crashes
3200 Details : The new msg_v2 system had some failures in mixed-endian
3203 Severity : enhancement
3205 Description: Easy OST removal
3206 Details : OSTs can be permanently deactivated with e.g. 'lctl
3207 conf_param lustre-OST0001.osc.active=0'
3209 Severity : enhancement
3211 Description: MGS proc entries
3212 Details : Added basic proc entries for the MGS showing what filesystems
3215 Severity : enhancement
3217 Description: provide MGS failover
3218 Details : Added config lock reacquisition after MGS server failover.
3220 Severity : enhancement
3222 Description: add Linux 2.4 support
3223 Details : Added support for RHEL 2.4.21 kernel for 1.6 servers and clients
3227 Description: a large application tries to do I/O to the same resource and dies
3228 in the middle of it.
3229 Details : Check the req->rq_arrival time after the call to
3230 ost_brw_lock_get(), but before we do anything about
3231 processing it & sending the BULK transfer request. This
3232 should help move old stale pending locks off the queue as
3233 quickly as obd_timeout.
3236 Frequency : when an incorrect nid is specified during startup
3238 Description: ptlrpc connect to non-existant node causes kernel crash
3239 Details : LNET can't be re-entered from an event callback, which
3240 happened when we expire a message after the export has been
3241 cleaned up. Instead, hand the zombie cleanup off to another
3244 Severity : enhancement
3246 Description: plain/inodebits lock performance improvement
3247 Details : Grouping plain/inodebits in granted list by their request modes
3248 and bits policy, thus improving the performance of search through
3252 Frequency : only if OST filesystem is corrupted
3254 Description: client incorrectly hits assertion in ptlrpc_replay_req()
3255 Details : for a short time RPCs with bulk IO are in the replay list,
3256 but replay of bulk IOs is unimplemented. If the OST filesystem
3257 is corrupted due to disk cache incoherency and then replay is
3258 started it is possible to trip an assertion. Avoid putting
3259 committed RPCs into the replay list at all to avoid this issue.
3262 Frequency : liblustre (e.g. catamount) on a large cluster with >= 8 OSTs/OSS
3264 Description: System hang on startup
3265 Details : This bug allowed the liblustre (e.g. catamount) client to
3266 return to the app before handling all startup RPCs. This
3267 could leave the node unresponsive to lustre network traffic
3268 and manifested as a server ptllnd timeout.
3270 Severity : enhancement
3272 Description: Add "/proc/sys/lustre/debug_peer_on_timeout"
3273 Details : liblustre envirable: LIBLUSTRE_DEBUG_PEER_ON_TIMEOUT
3274 boolean to control whether to print peer debug info when a
3275 client's RPC times out.
3278 Frequency : only for kernels with patches from Lustre below 1.4.3
3280 Description: Remove old rdonly API
3281 Details : Remove old rdonly API which unused from at least lustre 1.4.3
3284 Frequency : only for devices with external journals
3286 Description: Set external device read-only also
3287 Details : During a commanded failover stop, we set the disk device
3288 read-only while the server shuts down. We now also set any
3289 external journal device read-only at the same time.
3292 Frequency : when upgrading from 1.4 while trying to change parameters
3294 Description: The wrong (new) MDC name was used when setting parameters for
3295 upgraded MDT's. Also allows changing of OSC (and MDC)
3296 parameters if --writeconf is specified at tunefs upgrade time.
3299 Frequency : when setting specific ost indicies
3301 Description: QOS code breaks on skipped indicies
3302 Details : Add checks for missing OST indicies in the QOS code, so OSTs
3303 created with --index need not be sequential.
3305 Severity : enhancement
3307 Description: Add uninit_groups feature to ldiskfs2 to speed up e2fsck
3308 Details : The uninit_groups feature works in conjunction with the kernel
3309 filesystem code (ldiskfs2 only) and e2fsprogs-1.39-cfs6 to speed
3310 up the pass1 processing of e2fsck. This is a read-only feature
3311 in ldiskfs2 only, so older kernels and current ldiskfs cannot
3312 mount filesystems that have had this feature enabled.
3314 Severity : enhancement
3316 Description: Improve multi-block allocation algorithm to avoid fragmentation
3317 Details : The mballoc3 code (ldiskfs2 only) adds new mechanisms to improve
3318 allocation locality and avoid filesystem fragmentation.
3320 ------------------------------------------------------------------------------
3322 2007-04-01 Cluster File Systems, Inc. <info@clusterfs.com>
3324 * Support for kernels:
3325 2.4.21-47.0.1.EL (RHEL 3)
3326 2.6.5-7.283 (SLES 9)
3327 2.6.9-42.0.10.EL (RHEL 4)
3328 2.6.12.6 vanilla (kernel.org)
3329 2.6.16.27-0.9 (SLES 10)
3330 * Recommended e2fsprogs version: 1.39.cfs5
3332 * Note that reiserfs quotas are disabled on SLES 10 in this kernel
3336 Frequency : occasional, depends on client load and configuration
3337 Bugzilla : 12181, 12203
3338 Description: data loss for recently-modified files
3340 Details : In some cases it is possible that recently written or created
3341 files may not be written to disk in a timely manner (this should
3342 normally be within 30s unless client IO load is very high).
3343 The problem appears as zero-length files or files that are a
3344 multiple of 1MB in size after a client crash or client eviction
3345 that are missing data at the end of the file.
3347 This problem is more likely to be hit on clients where files are
3348 repeatedly created and unlinked in the same directory, clients
3349 have a large amount of RAM, have many CPUs, the filesystem has
3350 many OSTs, the clients are rebooted frequently, and/or the files
3351 are not accessed by other nodes after being written.
3353 The presence of the problem can be detected by looking at
3354 /proc/sys/fs/inode-state. If the first number (nr_inodes) is
3355 smaller than the second (nr_unused) then dirty files will not
3356 be flushed automatically to disk. "sync; sleep 10" should be
3357 run several times on the node before unmounting it to update
3358 Lustre (this is also safe to run on nodes without this problem).
3360 There is also a related kernel bug in the RHEL4 4 2.6.9 kernel
3361 that can cause this same problem, so customers using that kernel
3362 also need to update the kernel in addition to Lustre. In order
3363 to properly fix this bug, the RHEL3 2.4.21 kernel is also updated.
3365 It is normal that files written just before a client crash (less
3366 than 30s) may not yet have been flushed to disk, even for local
3370 Frequency : frequent on thin XT3 nodes
3372 Description: UUID collision on thin XT3 Linux nodes
3373 Details : UUIDs on Compute Node Linux XT3 nodes were not generated
3374 randomly, since we relied on an insufficiently-seeded PRNG.
3379 Description: OSS hangs after "All ost request buffers busy"
3380 Details : A deadlock between quota and journal operations caused OSS
3381 hangs after printing "All ost request buffers busy."
3384 Frequency : always on liblustre builds
3386 Description: Cleanup compiler warnings on liblustre
3389 Frequency : always on liblustre builds on XT3
3391 Description: LC_CONFIG_CDEBUG don't run while build liblustre on XT3.
3395 Description: Addition of EXT3_FEATURE_RO_COMPAT_DIR_NLINKS flag for
3396 > 32000 subdirectories
3397 Details : Add EXT3_FEATURE_RO_COMPAT_DIR_NLINK flag to
3398 EXT3_FEATURE_RO_COMPAT_SUPP. This flag will be set whenever
3399 subdirectory count crosses 32000. This will aid e2fsck to
3400 correctly handle more than 32000 subdirectories.
3403 Frequency : liblustre (e.g. catamount) on a large cluster with >= 8 OSTs/OSS
3405 Description: System hang on startup
3406 Details : This bug allowed the liblustre (e.g. catamount) client to
3407 return to the app before handling all startup RPCs. This
3408 could leave the node unresponsive to lustre network traffic
3409 and manifested as a server ptllnd timeout.
3411 Severity : enhancement
3413 Description: Add "/proc/sys/lustre/debug_peer_on_timeout"
3414 (liblustre envirable: LIBLUSTRE_DEBUG_PEER_ON_TIMEOUT)
3415 boolean to control whether to print peer debug info when a
3416 client's RPC times out.
3421 Description: make O_SYNC working on 2.6 kernels
3422 Details : 2.6 kernels use different method for mark pages for write,
3423 so need add a code to lustre for O_SYNC work.
3428 Description: Failure to close file and release space on NFS
3429 Details : Put inode details into lock acquired in ll_intent_file_open.
3430 Use mdc_intent_lock in ll_intent_open to properly
3431 detect all kind of errors unhandled by mdc_enqueue.
3436 Description: proc file read during shutdown sometimes raced obd removal,
3438 Details : Add lock to prevent obd access after proc file removal.
3441 Frequency : Only for files larger than 4GB on 32-bit clients.
3443 Description: improperly doing page alignment of locks
3444 Details : Modify lustre core code to use CFS_PAGE_* defines instead of
3445 PAGE_*. Make CFS_PAGE_MASK a 64-bit mask.
3450 Description: RPCs being resent when they shouldn't be
3451 Details : Some RPCs that should not be resent are being resent. This
3452 can cause inconsistencies in the RPC state machine. Do not
3453 resend such requests.
3456 Frequency : rare, only with NFS export
3458 Description: Crash on NFS re-export node
3459 Details : under very unusual load conditions an assertion is hit in
3460 ll_intent_file_open()
3463 Frequency : only if OST filesystem is corrupted
3465 Description: client incorrectly hits assertion in ptlrpc_replay_req()
3466 Details : for a short time RPCs with bulk IO are in the replay list,
3467 but replay of bulk IOs is unimplemented. If the OST filesystem
3468 is corrupted due to disk cache incoherency and then replay is
3469 started it is possible to trip an assertion. Avoid putting
3470 committed RPCs into the replay list at all to avoid this issue.
3475 Description: large O_DIRECT requests fail under memory pressure/fragmentation
3476 Details : Large single O_DIRECT read and write calls can fail to allocate
3477 a sufficiently large buffer to process the request. In case of
3478 allocation failure the allocation is retried with a smaller
3479 buffer and broken into smaller requests.
3481 Severity : enhancement
3483 Description: Add -o localflock option to simulate old noflock behaviour.
3484 Details : This will achieve local-only flock/fcntl locks coherentness.
3489 Description: versioning check is incomplete
3490 Details : Checking the version difference of client vs. server, report
3491 error if the gap is too big.
3496 Description: add support PG_writeback bit
3497 Details : add support for PG_writeback bit for Lustre, for more carefull
3498 work with page cache in 2.6 kernel. This also fix some deadlocks
3499 and remove hack for work O_SYNC with 2.6 kernel.
3501 Severity : enhancement
3503 Description: Add uninit_groups feature to ldiskfs2 to speed up e2fsck
3504 Details : The uninit_groups feature works in conjunction with the kernel
3505 filesystem code (ldiskfs2 only) and e2fsprogs-1.39-cfs6 to speed
3506 up the pass1 processing of e2fsck. This is a read-only feature
3507 in ldiskfs2 only, so older kernels and current ldiskfs cannot
3508 mount filesystems that have had this feature enabled.
3510 Severity : enhancement
3512 Description: Improve multi-block allocation algorithm to avoid fragmentation
3513 Details : The mballoc3 code (ldiskfs2 only) adds new mechanisms to improve
3514 allocation locality and avoid filesystem fragmentation.
3516 ------------------------------------------------------------------------------
3518 2007-02-09 Cluster File Systems, Inc. <info@clusterfs.com>
3520 * Support for kernels:
3521 2.6.9-42.0.3.EL (RHEL 4)
3522 2.6.5-7.276 (SLES 9)
3523 2.4.21-47.0.1.EL (RHEL 3)
3524 2.6.12.6 vanilla (kernel.org)
3525 2.6.16.21-0.8 (SLES10)
3526 * Recommended e2fsprogs version: 1.39.cfs2-0
3528 * The backwards-compatible /proc/sys/portals symlink has been removed
3529 in this release. Before upgrading, please ensure that you change
3530 any configuration scripts or /etc/sysctl.conf files that access
3531 /proc/sys/portals/* or sysctl portals.* to use the corresponding
3532 entry in /proc/sys/lnet or sysctl lnet.*. This change can be made
3533 in advance of the upgrade on any system running Lustre 1.4.6 or
3534 newer, since /proc/sys/lnet was added in that version.
3535 * Note that reiserfs quotas are disabled on SLES 10 in this kernel
3539 Frequency : only when quota is used
3541 Description: avoid scanning export list for quota master
3542 Details : Change the algorithms to avoid scanning export list in order
3543 to improve the efficiency.
3546 Frequency : MDS failover only, very rarely
3548 Description: "went back in time" messages on mds failover
3549 Details : The greatest transno may be lost when the current operation
3550 finishes with an error (transno==0) and the client's last_rcvd
3551 record is over-written. Save the greatest transno in the
3552 mds_last_transno for this case.
3555 Frequency : always for specific kernels and striping counts
3557 Description: client may get "Matching packet too big" without ACL support
3558 Details : Clients compiled without CONFIG_FS_POSIX_ACL get an error message
3559 when trying to access files in certain configurations. The
3560 clients should in fact be denied when mounting because they do
3561 not understand ACLs.
3564 Frequency : Cray XT3 with more than 4000 clients and multiple jobs
3566 Description: many clients connecting with IO in progress causes connect timeouts
3567 Details : Avoid synchronous journal commits to avoid delays caused by many
3568 clients connecting/disconnecting when bulk IO is in progress.
3569 Queue liblustre connect requests on OST_REQUEST_PORTAL instead of
3570 OST_IO_PORTAL to avoid delays behind potentially many pending
3574 Frequency : occasionally with multiple writers to a single file
3576 Description: shared writes to file may result in wrong size reported by stat()
3577 Details : Allow growing of kms when extent lock is cancelled
3580 Frequency : always with random mmap IO to multi-striped file
3582 Description: mmap write might be lost if we are writing to a 'hole' in stripe
3583 Details : Only if the hole is at the end of OST object so that kms is too
3584 small. Fix is to increase kms accordingly in ll_nopage.
3587 Frequency : rare, only if OST filesystem is inconsistent with MDS filesystem
3589 Description: writes to a missing object would leak memory on the OST
3590 Details : If there is an inconsistency between the MDS and OST filesystems,
3591 such that the MDS references an object that doesn't exist, writes
3592 to that object will leak memory due to incorrect cleanup in the
3593 error handling path, eventually running out of memory on the OST.
3598 Description: Creating too long symlink causes lustre errors
3599 Details : Check symlink and name lengths before sending requests to MDS.
3602 Frequency : only if flock is enabled (not on by default)
3604 Description: posix locks not released on fd closure on 2.6.9+
3605 Details : We failed to add posix locks to list of inode locks on 2.6.9+
3606 kernels, this caused such locks not to be released on fd close and
3607 then assertions on fs unmount about still used locks.
3610 Frequency : MDS failover only, very rarely
3612 Description: clients may get ASSERTION(granted_lock != NULL)
3613 Details : When request was taking a long time, and a client was resending
3614 a getattr by name lock request. The were multiple lock requests
3615 with the same client lock handle and
3616 mds_getattr_name->fixup_handle_for_resent_request found one of the
3617 lock handles but later failed with ASSERTION(granted_lock != NULL).
3622 Description: handle->h_buffer_credits > 0, assertion failure
3623 Details : h_buffer_credits is zero after truncate, causing assertion
3624 failure. This patch extends the transaction or creates a new
3628 Frequency : NFS re-export or patchless client
3629 Bugzilla : 11179, 10796
3630 Description: Crash on NFS re-export node (__d_move)
3631 Details : We do not want to hash the dentry if we don't have a lock.
3632 But if this dentry is later used in d_move, we'd hit uninitialised
3633 list head d_hash, so we just do this to init d_hash field but
3634 leave dentry unhashed.
3637 Frequency : NFS re-export or patchless client
3639 Description: NFS exports has problem with symbolic link
3640 Details : lustre client didn't properly install dentry when re-exported
3641 to NFS or running patchless client.
3644 Frequency : NFS re-export or patchless client
3646 Description: Various nfs/patchless fixes.
3647 Details : fixes reuse disconected alias for lookup process - this fixes
3648 warning "find_exported_dentry: npd != pd",
3649 fix permission error with open files at nfs.
3650 fix apply umask when do revalidate.
3653 Frequency : occasional
3655 Description: Crash on NFS re-export node
3656 Details : calling clear_page() on the wrong pointer triggered oops in
3657 generic_mapping_read().
3660 Frequency : rarely, using O_DIRECT IO
3662 Description: unaligned directio crashes client with LASSERT
3663 Details : check for unaligned buffers before trying any requests.
3666 Frequency : rarely, using CFS RAID5 patches in non-standard kernel series
3668 Description: stale data returned from RAID cache
3669 Details : If only a small amount of IO is done to the RAID device before
3670 reading it again it is possible to get stale data from the RAID
3671 cache instead of reading it from disk.
3674 Frequency : always for sles10 kernel
3676 Description: sles10 support
3677 Details : ll_follow_link: compile fixes and using of nd_set_link
3678 under newer kernels.
3681 Frequency : depends on arch, kernel and compiler version, always on sles10
3684 Description: recursive or deep enough symlinks cause stack overflow
3685 Details : getting rid of large stack-allocated variable in
3689 Frequency : depends on hardware
3691 Description: lustre write performance loss in the SLES10 kernel
3692 Details : the performance loss is caused by using of write barriers in the
3693 ext3 code. The SLES10 kernel turns barrier support on by
3694 default. The fix is to undo that change for ldiskfs.
3696 ------------------------------------------------------------------------------
3698 2006-12-09 Cluster File Systems, Inc. <info@clusterfs.com>
3700 * Support for kernels:
3701 2.6.9-42.0.3EL (RHEL 4)
3702 2.6.5-7.276 (SLES 9)
3703 2.4.21-47.0.1.EL (RHEL 3)
3704 2.6.12.6 vanilla (kernel.org)
3708 Frequency : quota enabled and large files being deleted
3710 Description: releasing more than 4GB of quota at once hangs OST
3711 Details : If a user deletes more than 4GB of files on a single OST it
3712 will cause the OST to spin in an infinite loop. Release
3713 quota in < 4GB chunks, or use a 64-bit value for 1.4.7.1+.
3718 Description: statfs data retrieved from /proc may be stale or zero
3719 Details : When reading per-device statfs data from /proc, in the
3720 {kbytes,files}_{total,free,avail} files, it may appear
3721 as zero or be out of date.
3724 Frequency : systems with MD RAID1 external journal devices
3726 Description: lconf's call to blkid is confused by RAID1 journal devices
3727 Details : Use the "blkid -l" flag to locate the MD RAID device instead
3728 of returning all block devices that match the journal UUID.
3731 Frequency : always, for aggregate stripe size over 4GB
3733 Description: "lfs setstripe" fails assertion when setting 4GB+ stripe width
3734 Details : Using "lfs setstripe" to set stripe size * stripe count over 4GB
3735 will fail the kernel with "ASSERTION(lsm->lsm_xfersize != 0)"
3738 Frequency : always if "lfs find" used on a local file/directory
3740 Description: "lfs find" segfaults if used on a local file/directory
3741 Details : The case where a directory component was not specified wasn't
3742 handled correctly. Handle this properly.
3745 Frequency : always on ppc64
3747 Description: the write to an ext3 filesystem mounted with mballoc got stuck
3748 Details : ext3_mb_generate_buddy() uses find_next_bit() which does not
3749 perform endianness conversion.
3752 Frequency : rarely (truncate to non-zero file size after write under load)
3753 Bugzilla : 10730, 10687
3754 Description: Files padded with zeros to next 4K multiple
3755 Details : With filesystems mounted using the "extents" option (2.6 kernels)
3756 it is possible that files that are truncated to a non-zero size
3757 immediately after being written are filled with zero bytes beyond
3758 the truncated size. No file data is lost.
3760 Severity : enhancement
3762 Description: Allow recovery/failover for liblustre clients.
3763 Details : liblustre clients were unaware of failover configurations until
3766 Severity : enhancement
3768 Description: user file locks should fail when not mounting with flock option
3769 Details : Set up an error-returning stub in ll_file_operations.lock field
3770 to prevent incorrect behaviour when client is mounted without
3771 flock option. Also, set up properly f_op->flock field for
3775 Frequency : always on ia64
3777 Description: "lfs df" loops on printing out MDS statfs information
3778 Details : The obd_ioctl_data was not initialized and in some systems
3779 this caused a failure during the ioctl that did not return
3780 an error. Initialize the struct and return an error on failure.
3783 Frequency : SLES 9 only
3785 Description: Error of copying files with lustre special EAs as root
3786 Details : Client side always return success for setxattr call for lustre
3787 special xattr (currently only "trusted.lov").
3790 Frequency : rarely on clusters with both ia64+i386 clients
3792 Description: ia64+i686 clients doing shared IO on the same file may LBUG
3793 Details : In rare cases when both ia64+i686 (or other mixed-PAGE_SIZE)
3794 clients are doing concurrent writes to the same file it is
3795 possible that the ia64 clients may LASSERT because the OST
3796 extent locks are not PAGE_SIZE aligned. Ensure that grown
3797 locks are always aligned on the request boundary.
3800 Frequency : specific use, occasional
3802 Description: Overwriting in use executable truncates on-disk binary image
3803 Details : If one node attempts to overwrite an executable in use by
3804 another node, we now correctly return ETXTBSY instead of
3805 truncating the file.
3807 Severity : enhancement
3809 Description: Async OSC create to avoid the blocking unnecessarily.
3810 Details : If a OST has no remain object, system will block on the creating
3811 when need to create a new object on this OST. Now, ways use
3812 pre-created objects when available, instead of blocking on an
3813 empty osc while others are not empty. If we must block, we block
3814 for the shortest possible period of time.
3819 Description: chmod on Lustre root is propagated to other clients
3820 Details : Re-validate root's dentry in ll_lookup_it to avoid having it
3821 invalid by the follow_mount time.
3824 Frequency : liblustre clients only
3826 Description: Race in 'instant cancel' lock handling could lead to such locks
3827 never to be granted in case of SMP MDS
3828 Details : Do not destroy not yet granted but cbpending locks in
3832 Frequency : replay/resend of open
3834 Description: non null lock assetion failure in mds_intent_policy
3835 Details : Trying to replay/resend lockless open requests resulted in
3836 mds_open() returning 0 with no lock. Now it sets a flag if
3837 a lock is going to be returned.
3839 Severity : enhancement
3841 Description: Checksum enhancements
3842 Details : New checksum enhancements allow for resending RPCs that failed
3845 Severity : enhancement
3847 Description: Tunables on number of dirty pages in cacche
3848 Details : Allow to set limit on number of dirty pages cached.
3853 Description: client crash on unmount - lock still has references
3854 Details : In some error handling cases it was possible to leak a lock
3855 reference on a client while accessing a file. This was not
3856 harmful to the client during operation, but would cause the
3857 client to crash when the filesystem is unmounted.
3860 Frequency : specific case, rare
3862 Description: ETXTBSY on mds though file not in use by client
3863 Details : ETXTBSY is no longer incorrectly returned when attempting to
3864 chmod or chown a directory that the user previously tried to
3865 execute or a currently-executing binary.
3868 Frequency : extremely rare except on liblustre-based clients
3870 Description: Lustre space not freed when files are deleted
3871 Details : Clean up open-unlinked files after client eviction. Previously
3872 the unlink was skipped and the files remained as orphans.
3877 Description: OST failure "would be an LBUG" in waiting_locks_callback()
3878 Details : In some cases it was possible to send a blocking callback to
3879 a client doing a glimpse, even though that client didn't get
3880 a lock granted. When the glimpse lock is cancelled on the OST
3881 the freed lock is left on the waiting list and corrupted the list.
3884 Frequency : all core dumps
3886 Description: Broke core dumps to lustre
3887 Details : Negative dentry may be unhashed if parent does not have UPDATE
3888 lock, but some callers, e.g. do_coredump, expect dentry to be
3889 hashed after successful create, hash it in ll_create_it.
3891 ------------------------------------------------------------------------------
3893 2006-09-13 Cluster File Systems, Inc. <info@clusterfs.com>
3895 * Support for kernels:
3896 2.6.9-42.0.2.EL (RHEL 4)
3897 2.6.5-7.276 (SLES 9)
3898 2.4.21-40.EL (RHEL 3)
3899 2.6.12.6 vanilla (kernel.org)
3903 Frequency : always on RHEL 3
3905 Description: Number of open files grows over time
3906 Details : The number of open files grows over time, whether or not
3907 Lustre is started. This was due to a filp leak introduced
3908 by one of our kernel patches.
3910 ------------------------------------------------------------------------------
3912 2006-08-20 Cluster File Systems, Inc. <info@clusterfs.com>
3914 * Support for kernels:
3915 2.6.9-42.EL (RHEL 4)
3916 2.6.5-7.267 (SLES 9)
3917 2.4.21-40.EL (RHEL 3)
3918 2.6.12.6 vanilla (kernel.org)
3923 Bugzilla : 5719, 9635, 9792, 9684
3924 Description: OST (or MDS) trips assertions in (re)connection under heavy load
3925 Details : If a server is under heavy load and cannot reply to new
3926 connection requests before the client resends the (re)connect,
3927 the connection handling code can behave badly if two service
3928 threads are concurrently handing separate (re)connections from
3929 the same client. Add better locking to the connection handling
3930 code, and ensure that only a single connection will be processed
3931 for a given client UUID, even if the lock is dropped.
3933 Severity : enhancement
3935 Description: add TCP zero-copy support to kernel
3936 Details : Add support to the kernel TCP stack to allow zero-copy bulk
3937 sends if the hardware supports scatter-gather and checksumming.
3938 This allows socklnd to do client-write and server-read more
3939 efficiently and reduce CPU utilization from skbuf copying.
3942 Frequency : only if NFS exporting from client
3944 Description: NULL pointer deref in ll_iocontrol() if chattr mknod file
3945 Details : If setting attributes on a file created under NFS that had
3946 never been opened it would be possible to oops the client
3947 if the file had no objects.
3951 Bugzilla : 9326, 10402, 10897
3952 Description: client crash in ptlrpcd_wake() thread when sending async RPC
3953 Details : It is possible that ptlrpcd_wake() dereferences a freed async
3954 RPC. In rare cases the ptlrpcd thread alread processed the RPC
3955 before ptlrpcd_wake() was called and the request was freed.
3958 Frequency : always for liblustre
3960 Description: liblustre client does MDS+OSTs setattr RPC for each write
3961 Details : When doing a write from a liblustre client, the client
3962 incorrectly issued an RPC to the MDS and each OST the file was
3963 striped over in order to update the timestamps. When writing
3964 with small chunks and many clients this could overwhelm the MDS
3965 with RPCs. In all cases it would slow down the write because
3966 these RPCs are unnecessary.
3968 Severity : enhancement
3970 Description: allow number of MDS service threads to be changed at module load
3971 Details : It is now possible to change the number of MDS service threads
3972 running. Adding "options mds mds_num_threads={N}" to the MDS's
3973 /etc/modprobe.conf will set the number of threads for the next
3974 time Lustre is restarted (assuming the "mds" module is also
3975 reloaded at that time). The default number of threads will
3976 stay the same, 32 for most systems.
3981 Description: OST crash if filesystem is unformatted or corrupt
3982 Details : If an OST is started on a device that has never been formatted
3983 or if the filesystem is corrupt and cannot even mount then the
3984 error handling cleanup routines would dereference a NULL pointer.
3989 Description: NULL pointer deref in llap_from_page.
3990 Details : get_cache_page_nowait can return a page with NULL (or otherwise
3991 incorrect) mapping if the page was truncated/reclaimed while it was
3992 searched for. Check for this condition and skip such pages when
3993 doing readahead. Introduce extra check to llap_from_page() to
3994 verify page->mapping->host is non-NULL (so page is not anonymous).
3997 Frequency : Sometimes when using sys_sendfile
3999 Description: "page not covered by a lock" warnings from ll_readpage
4000 Details : sendfile called ll_readpage without right page locks present.
4001 Now we introduced ll_file_sendfile that does necessary locking
4002 around call to generic_file_sendfile() much like we do in
4006 Frequency : with certain MDS communication failures at client mount time
4008 Description: NULL pointer deref after failed client mount
4009 Details : a client connection request may delayed by the network layer
4010 and not be sent until after the PTLRPC layer has timed out the
4011 request. If the client fails the mount immediately it will try
4012 to clean up before the network times out the request. Add a
4013 reference from the request import to the obd device and delay
4014 the cleanup until the network drops the request.
4017 Frequency : occasionally during client (re)connect
4019 Description: assertion failure during client (re)connect
4020 Details : processing a client connection request may be delayed by the
4021 client or server longer than the client connect timeout. This
4022 causes the client to resend the connection request. If the
4023 original connection request is replied in this interval, the
4024 client may trip an assertion failure in ptlrpc_connect_interpret()
4025 which thought it would be the only running connect process.
4028 Frequency : only with obd_echo servers and clients that are rebooted
4030 Description: kernel BUG accessing uninitialized data structure
4031 Details : When running an obd_echo server it did not start the ping_evictor
4032 thread, and when a client was evicted an uninitialized data
4033 structure was accessed. Start the ping_evictor in the RPC
4034 service startup instead of the OBD startup.
4036 Severity : enhancement
4037 Bugzilla : 10193 (patchless)
4038 Description: Remove dependency on various unexported kernel interfaces.
4039 Details : No longer need reparent_to_init, exit_mm, exit_files,
4040 sock_getsockopt, filemap_populate, FMODE_EXEC, put_filp.
4043 Frequency : rare (only users of deprecated and unsupported LDAP config)
4045 Description: write_conf for zeroconf mount queried LDAP incorrectly for client
4046 Details : LDAP apparently contains 'lustreName' attributes instead of
4047 'name'. A simple remapping of the name is sufficient.
4050 Frequency : rare (only with non-default dump_on_timeout debug enabled)
4052 Description: waiting_locks_callback trips kernel BUG if client is evicted
4053 Details : Running with the dump_on_timeout debug flag turned on makes
4054 it possible that the waiting_locks_callback() can try to dump
4055 the Lustre kernel debug logs from an interrupt handler. Defer
4056 this log dumping to the expired_lock_main() thread.
4058 Severity : enhancement
4060 Description: Support NFS exporting on 2.6 kernels.
4061 Details : Implement non-rawops metadata methods for NFS server to use without
4062 changing NFS server code.
4065 Frequency : very rare (synthetic metadata workload only)
4067 Description: two racing renames might cause an MDS thread to deadlock
4068 Details : Running the "racer" program may cause one MDS thread to rename
4069 a file from being the source of a rename to being the target of
4070 a rename at exactly the same time that another thread is doing
4071 so, and the second thread has already enqueued these locks after
4072 doing a lookup of the target and is trying to relock them in
4073 order. Ensure that we don't try to re-lock the same resource.
4076 Frequency : only very large systems with liblustre clients
4078 Description: slow eviction of liblustre clients with the "evict_by_nid" RPC
4079 Details : Use asynchronous set_info RPCs to send the "evict_by_nid" to
4080 all OSTs in parallel. This allows the eviction of stale liblustre
4081 clients to proceed much faster than if they were done in series,
4082 and also offers similar improvements for other set_info RPCs.
4087 Description: excessive CPU usage during initial read phase on client
4088 Details : During the initial read phase on a client, it would agressively
4089 retry readahead on the file, consuming too much CPU and impacting
4090 performance (since 1.4.5.8). Improve the readahead algorithm
4091 to avoid this, and also improve some other common cases (read
4092 of small files in particular, where "small" is files smaller than
4093 /proc/fs/lustre/llite/*/max_read_ahead_whole_mb, 2MB by default).
4098 Description: MDS crash when receiving packet with unknown intent.
4099 Details : Do not LBUG in unknown intent case, just return -EFAULT
4101 Severity : enhancement
4102 Bugzilla : 9293, 9385
4103 Description: MDS RPCs are serialised on client. This is unnecessary for some.
4104 Details : Do not serialize getattr (non-intent version) and statfs.
4107 Frequency : occasional, when OST network is overloaded/intermittent
4109 Description: client evicted by OST after bulk IO timeout
4110 Details : If a client sends a bulk IO request (read or write) the OST
4111 may evict the client if it is unresposive to its data GET/PUT
4112 request. This is incorrect if the network is overloaded (takes
4113 too long to transfer the RPC data) or dropped the OST GET/PUT
4114 request. There is no need to evict the client at all, since
4115 the pinger and/or lock callbacks will handle this, and the
4116 client can restart the bulk request.
4119 Frequency : Always when mmapping file with no objects
4121 Description: client crashes when mmapping file with no objects
4122 Details : Check that we actually have objects in a file before doing any
4123 operations on objects in ll_vm_open, ll_vm_close and
4129 Description: Request leak when working with deleted CWD
4130 Details : Introduce advanced request refcount tracking for requests
4131 referenced from lustre intent.
4133 Severity : Enhancement
4135 Description: Cache open file handles on client.
4136 Details : MDS now will return special lock along with openhandle, if
4137 requested and client is allowed to hold openhandle, even if unused,
4138 until such a lock is revoked. Helps NFS a lot, since NFS is opening
4139 closing files for every read/write openration.
4141 Severity : Enhancement
4143 Description: Cache open negative dentries on client when possible.
4144 Details : Guard negative dentries with UPDATE lock on parent dir, drop
4145 negative dentries on lock revocation.
4150 Description: Remounting a client read-only wasn't possible with a zconf mount
4151 Details : It wasn't possible to remount a client read-only with llmount.
4153 Severity : enhancement
4154 Description: Include MPICH 1.2.6 Lustre ADIO interface patch
4155 Details : In lustre/contrib/ or /usr/share/lustre in RPM a patch for
4156 MPICH is included to add Lustre-specific ADIO interfaces.
4157 This is based closely on the UFS ADIO layer and only differs
4158 in file creation, in order to allow the OST striping to be set.
4159 This is user-contributed code and not supported by CFS.
4164 Description: extended inode attributes (immutable, append-only) work improperly
4165 when 2.4 and 2.6 kernels are used on client/server or vice versa
4166 Details : Introduce kernel-independent values for these flags.
4168 Severity : enhancement
4171 Description: Allow fractional MB tunings for lustre in /proc/ filesystem.
4172 Details : Many of the /proc/ tunables can only be tuned at a megabyte
4173 granularity. Now, Fractional MB granularity is be supported,
4174 this is very useful for low memory system.
4176 Severity : enhancement
4178 Description: Getattr by fid
4179 Details : Getting a file attributes by its fid, obtaining UPDATE|LOOKUP
4180 locks, avoids extra getattr rpc requests to MDS, allows '/' to
4181 have locks and avoids getattr rpc requests for it on every stat.
4184 Frequency : Always, for filesystems larger than 2TB
4186 Description: ldiskfs crash at mount for filesystem larger than 2TB with mballoc
4187 Details : Kenrel kmalloc limits allocations to 128kB and this prevents
4188 filesystems larger than 2TB to be mounted with mballoc enabled.
4191 Frequency : Always, for 32-bit kernel without CONFIG_LBD and filesystem > 2TB
4193 Description: filesystem corruption for non-standard kernels and very large OSTs
4194 Details : If a 32-bit kernel is compiled without CONFIG_LBD enabled and a
4195 filesystems larger than 2TB is mounted then the kernel will
4196 silently corrupt the start of the filesystem. CONFIG_LBD is
4197 enabled for all CFS-supported kernels, but the possibility of
4198 this happening with a modified kernel config exists.
4200 Severity : enhancement
4202 Description: add client O_DIRECT support for 2.6 kernels
4203 Details : It is now possible to do O_DIRECT reads and writes to files
4204 in the Lustre client mountpoint on 2.6 kernel clients.
4206 Severity : enhancement
4208 Description: parallel glimpse, setattr, statfs, punch, destroy requests
4209 Details : Sends glimpse, setattr, statfs, punch, destroy requests to OSTs in
4210 parallel, not waiting for response from every OST before sending
4211 a rpc to the next OST.
4216 Description: setattr vs write race when updating file timestamps
4217 Details : Client processes that update a file timestamp into the past
4218 right after writing to the file (e.g. tar) it is possible that
4219 the updated file modification time can be reset to the current
4220 time due to a race between processing the setattr and write RPC.
4222 Severity : enhancement
4224 Description: Bring 'lfs find' closer in line with regular Linux find.
4225 Details : lfs find util supports -atime, -mtime, -ctime, -maxdepth, -print,
4226 -print0 options and obtains all the needed info through the lustre
4229 Severity : enhancement
4231 Description: support up to 1024 configured devices on one node
4232 Details : change obd_dev array from statically allocated to dynamically
4233 allocated structs as they are first used to reduce memory usage
4238 Description: Flush dirty partially truncated pages during truncate
4239 Details : Immediatelly flush partially truncated pages in filter_setattr,
4240 this way we completely avoid having any pages in page cache on OST
4241 and can retire ugly workarounds during writes to flush such pages.
4246 Description: i_sem vs transaction deadlock in mds_obd_destroy during unlink.
4247 Details : protect inode from truncation within vfs_unlink() context
4248 just take a reference before calling vfs_unlink() and release it
4249 when parent's i_sem is free.
4252 Frequency : always, if extents are used on OSTs
4254 Description: index ei_leaf_hi (48-bit extension) is not zeroed in extent index
4255 Details : OSTs using the extents format would not zero the high 16 bits of
4256 the index physical block number. This is not a problem for any
4257 OST filesystems smaller than 16TB, and no kernels support ext3
4258 filesystems larger than 16TB yet. This is fixed in 1.4.7 (all
4259 new/modified files) and can be fixed for existing filesystems
4260 with e2fsprogs-1.39-cfs1.
4265 Description: import connection selection may be incorrect if timer wraps
4266 Details : Using a 32-bit jiffies timer with HZ=1000 may cause backup
4267 import connections to be ignored if the 32-bit jiffies counter
4268 wraps. Use a 64-bit jiffies counter.
4271 Frequency : during server recovery
4273 Description: crash after server is denying duplicate export
4274 Details : If clients are resending connect requests to the server, the
4275 server refuses to allow a client to connect multiple times.
4276 Fixed a bug in the handling of this case.
4279 Frequency : very large clusters immediately after boot
4281 Description: LNET request buffers exhausted under heavy short-term load
4282 Details : If a large number of client requests are generated on a service
4283 that has previously never seen so many requests it is possible
4284 that the request buffer growth cannot keep up with the spike in
4285 demand. Instead of dropping incoming requests, they are held in
4286 the LND until the RPC service can accept more requests.
4289 Frequency : Sometimes during replay
4291 Description: Assertion failure in ll_local_open after replay.
4292 Details : If replay happened on an open request reply before we were able
4293 to set replay handler, reply will become not swabbed tripping the
4294 assertion in ll_local_open. Now we set the handler right after
4295 recognising of open request
4298 Frequency : very rare
4300 Description: kernel reports "badness in vsnprintf"
4301 Details : Reading from the "recovery_status" /proc file in small chunks
4302 may cause a negative length in lprocfs_obd_rd_recovery_status()
4303 call to vsnprintf() (which is otherwise harmless). Exit early
4304 if there is no more space in the output buffer.
4306 Severity : enhancement
4308 Description: clear OBD RPC statistics by writing to them
4309 Details : It is now possible to clear the OBD RPC statistics by writing
4310 to the "stats" file.
4315 Description: Client mtime is not the same on different clients after utimes
4316 Details : In some cases, the client was using the utimes() syscall on
4317 a file cached on another node. The clients now validate the
4318 ctime from the MDS + OSTs to determine which one is right.
4323 Description: Inability to activate failout mode
4324 Details : lconf script incorrectly assumed that in python string's numeric
4325 value is used in comparisons.
4328 Frequency : always with multiple stripes per file
4330 Description: Inefficient object allocation for mutli-stripe files
4331 Details : When selecting which OSTs to stripe files over, for files with
4332 a stripe count that divides evenly into the number of OSTs,
4333 the MDS is always picking the same starting OST for each file.
4334 Return the OST selection heuristic to the original design.
4339 Description: mount failures may take full timeout to return an error
4340 Details : Under some heavy load conditions it is possible that a
4341 failed mount can wait for the full obd_timeout interval,
4342 possibly several minutes, before reporting an error.
4343 Instead return an error as soon as the status is known.
4345 ------------------------------------------------------------------------------
4347 2006-02-14 Cluster File Systems, Inc. <info@clusterfs.com>
4349 * WIRE PROTOCOL CHANGE. This version of Lustre networking WILL NOT
4350 INTEROPERATE with older versions automatically. Please read the
4351 user documentation before upgrading any part of a live system.
4352 * WARNING: Lustre networking configuration changes are required with
4353 this release. See https://bugzilla.clusterfs.com/show_bug.cgi?id=10052
4356 * Support for kernels:
4357 2.6.9-22.0.2.EL (RHEL 4)
4358 2.6.5-7.244 (SLES 9)
4359 2.6.12.6 vanilla (kernel.org)
4362 Severity : enhancement
4363 Bugzilla : 7981/8208
4364 Description: Introduced Lustre Networking (LNET)
4365 Details : LNET is new networking infrastructure for Lustre, it includes
4366 a reorganized network configuration mode (see the user
4367 documentation for full details) as well as support for routing
4368 between different network fabrics. Lustre Networking Devices
4369 (LNDs) for the supported network fabrics have also been
4370 created for this new infrastructure.
4372 Severity : enhancement
4373 Description: Introduced Access control lists
4374 Details : clients can set ACLs on files and directories in order to have
4375 more fine-grained permissions than the standard Unix UGO+RWX.
4376 The MDS must be started with the "-o acl" mount option.
4378 Severity : enhancement
4379 Description: Introduced filesystem quotas
4380 Details : Administrators may now establish per-user quotas on the
4383 Severity : enhancement
4385 Description: Configuration change for the XT3
4386 The PTLLND is now used to run Lustre over Portals on the XT3
4387 The configure option(s) --with-cray-portals are no longer used.
4388 Rather --with-portals=<path-to-portals-includes> is used to
4389 enable building on the XT3. In addition to enable XT3 specific
4390 features the option --enable-cray-xt3 must be used.
4395 Description: Running on many-way SMP OSTs can trigger oops in llcd_send()
4396 Details : A race between allocating a new llcd and re-getting the llcd_lock
4397 allowed another thread to grab newly-allocated llcd.
4399 Severity : enhancement
4401 Description: 2.6 OST async journal commit and locking fix to improve performance
4402 Details : The filter_direct_io()+filter_commitrw_write() journal commits for
4403 2.6 kernels are now async as they already were in 2.4 kernels so
4404 that they can commit concurrently with the network bulk transfer.
4405 For block-allocated files the filter allocation semaphore is held
4406 to avoid filesystem fragmentation during allocation. BKL lock
4407 removed for 2.6 xattr operations where it is no longer needed.
4412 Description: lconf incorrectly determined whether two IP networks could talk
4413 Details : In some more complicated routing and multiple-network
4414 configurations, lconf will avoid trying to make a network
4415 connection to a disjoint part of the IP space. It was doing the
4416 math incorrectly for one set of cases.
4421 Description: Fix for potential infinite loop processing records in an llog.
4422 Details : If an llog record is corrupted/zeroed, it is possible to loop
4423 forever in llog_process(). Validate the llog record length
4424 and skip the remainder of the block on error.
4427 Frequency : occasional (liblustre only)
4429 Description: liblustre could not open files whose last component is a symlink
4430 Details : sysio_path_walk() would incorrectly pass the open intent to
4431 intermediate path components.
4434 Frequency : rare (liblustre only with non-standard tuning)
4435 Bugzilla : 7201 (7350)
4436 Description: Tuning the MDC DLM LRU size to zero triggers client LASSERT
4437 Details : llu_lookup_finish_locks() tries to set lock data on a lock
4438 after it has been released, only do this for referenced locks
4440 Severity : enhancement
4442 Description: specifying an (invalid) directory default stripe_size of -1
4443 would reset the directory default striping
4444 Details : stripe_size -1 was used internally to signal directory stripe
4445 removal, now use "all default" to signal dir stripe removal
4446 as a directory striping of "all default" is not useful
4449 Frequency : common for large clusters running liblustre clients
4451 Description: doing an ls when liblustre clients are running is slow
4452 Details : sending a glimpse AST to a liblustre client waits for every AST
4453 to time out, as liblustre clients will not respond. Since they
4454 cannot cache data we refresh the OST lock LVB from disk instead.
4456 Severity : enhancement
4458 Description: doing an ls at the same time as file IO can be slow
4459 Details : enqueue and other "small" requests can be blocked behind many
4460 large IO requests. Create a new OST IO portal for non-IO
4461 requests so they can be processed faster.
4464 Frequency : rare (only HPUX clients mounting unsupported re-exported NFS vol)
4466 Description: an HPUX NFS client would get -EACCESS when ftruncate()ing a newly
4467 created file with mode 000
4468 Details : the Linux NFS server relies on an MDS_OPEN_OWNEROVERRIDE hack to
4469 allow an ftruncate() as a non-root user to a file with mode 000.
4470 Lustre now respects this flag to disable mode checks when
4471 truncating a file owned by the user
4474 Frequency : liblustre-only, when liblustre client dies unexpectedly or becomes
4477 Description: Revoking locks from clients that went dead or catatonic might take
4479 Details : New lock flags FL_CANCEL_ON_BLOCK used by liblustre makes
4480 cancellation of such locks instant on servers without waiting for
4481 any reply from clients. Clients drops these locks when cancel
4482 notification from server is received without replying.
4485 Frequency : liblustre-only, when liblustre client dies or becomes busy
4487 Description: Doing ls on Linux clients can take a long time with active
4489 Details : Liblustre client cannot handle ASTs in timely manner, so avoid
4490 granting such locks to it in the first place if possible. Locks
4491 are taken by proxy on the OST during the read or write and
4492 dropped immediately afterward. Add connect flags handling, do
4493 not grant locks to liblustre clients for glimpse ASTs.
4495 Severity : enhancement
4497 Description: Improve read-ahead algorithm to avoid excessive IO for random reads
4498 Details : Existing read-ahead algorithm is tuned for the case of streamlined
4499 sequential reads and behaves badly with applications doing random
4500 reads. Improve it by reading ahead at least read region, and
4501 avoiding excessive large RPC for small reads.
4503 Severity : enhancement
4505 Description: Creating more than 1000 files for a single job may cause a load
4506 imbalance on the OSTs if there are also a large number of OSTs.
4507 Details : qos_prep_create() uses an OST index reseed value that is an
4508 even multiple of the number of available OSTs so that if the
4509 reseed happens in the middle of the object allocation it will
4510 still utilize the OSTs as uniformly as possible.
4515 Description: OST or MDS may oops in ping_evictor_main()
4516 Details : ping_evictor_main() drops obd_dev_lock if deleting a stale export
4517 but doesn't restart at beginning of obd_exports_timed list
4520 Severity : enhancement
4522 Description: improve by-nid export eviction on the MDS and OST
4523 Details : allow multiple exports with the same NID to be evicted at one
4524 time without re-searching the exports list.
4527 Frequency : rare, only with supplementary groups enabled on SMP 2.6 kernels
4529 Description: MDS may oops in groups_free()
4530 Details : in rare race conditions a newly allocated group_info struct is
4531 freed again, and this can be NULL. The 2.4 compatibility code
4532 for groups_free() checked for a NULL pointer, but 2.6 did not.
4535 Frequency : common for liblustre clients doing little filesystem IO
4536 Bugzilla : 9352, 7313
4537 Description: server may evict liblustre clients accessing contended locks
4538 Details : if a client is granted a lock or receives a completion AST
4539 with a blocking AST already set it would not reply to the AST
4540 for LDLM_FL_CANCEL_ON_BLOCK locks. It now replies to such ASTs.
4543 Frequency : lfs setstripe, only systems with more than 160 OSTs
4545 Description: unable to set striping with a starting offset beyond OST 160
4546 Details : llapi_create_file() incorrectly limited the starting stripe
4547 index to the maximum single-file stripe count.
4550 Frequency : LDAP users only
4552 Description: lconf did not handle in-kernel recovery with LDAP properly
4553 Details : lconf/LustreDB get_refs() is searching the wrong namespace
4555 Severity : enhancement
4557 Description: bind OST threads to NUMA nodes to improve performance
4558 Details : all OST threads are uniformly bound to CPUs on a single NUMA
4559 node and do their allocations there to localize memory access
4561 Severity : enhancement
4563 Description: llmount can determine client NID directly from Myrinet (GM)
4564 Details : the client NID code from gmnalnid was moved directly into
4565 llmount, removing the need to use this or specifying the
4566 client NID explicitly when mounting GM clients with zeroconf
4569 Frequency : if client is started with down MDS
4571 Description: if client is started with down MDS mount hangs in ptlrpc_queue_wait
4572 Details : Having an LWI_INTR() wait event (interruptible, but no timeout)
4573 will wait indefinitely in ptlrpc_queue_wait->l_wait_event() after
4574 ptlrpc_import_delayed_req() because we didn't check if the
4575 request was interrupted, and we also didn't break out of the
4576 event loop if there was no timeout
4581 Description: data loss during non-page-aligned writes to a single file from
4582 both multiple nodes and multiple threads on one node at same time
4583 Details : updates to KMS and lsm weren't protected by common lock. Resulting
4584 inconsistency led to false short-reads, that were cached and later
4585 used by ->prepare_write() to fill in partially written page,
4586 leading to data loss.
4589 Frequency : always, if lconf --abort_recovery used
4591 Description: lconf --abort_recovery fails with 'Operation not supported'
4592 Details : lconf was attempting to abort recovery on the MDT device and not
4595 Severity : enhancement
4597 Description: remove cleanup logs
4598 Details : replace lconf-generated cleanup logs with lustre internal
4599 cleanup routines. Eliminates the need for client-cleanup and
4602 Severity : enhancement
4604 Description: add support for EAs (user and system) on lustre filesystems
4605 Details : it is now possible to store extended attributes in the Lustre
4606 client filesystem, and with the user_xattr mount option it
4607 is possible to allow users to store EAs on their files also
4609 Severity : enhancement
4611 Description: Add possibility (config option) to show minimal available OST free
4613 Details : When compiled with --enable-mindf configure option, statfs(2)
4614 (and so, df) will return least minimal free space available from
4615 all OSTs as amount of free space on FS, instead of summary of
4616 free spaces of all OSTs.
4618 Severity : enhancement
4620 Description: do not expand extent locks acquired on OST-side
4621 Details : Modify ldlm_extent_policy() to not expand local locks, acquired
4622 by server: they are not cached anyway.
4625 Frequency : when mmap is used/binaries executed from Lustre
4627 Description: Unmmap pages before throwing them away from read cache.
4628 Details : llap_shrink cache now attempts to unmap pages before discarding
4629 them (if unmapping failed - do not discard). SLES9 kernel has
4630 extra checks that trigger if this unmapping is not done first.
4635 Description: lconf didn't resolve symlinks before checking to see whether a
4636 given mountpoint was already in use
4639 Frequency : when migrating failover services
4640 Bugzilla : 6395, 9514
4641 Description: When migrating a subset of services from a node (e.g. failback
4642 from a failover service node) the remaining services would
4643 time out and evict clients.
4644 Details : lconf --force (implied by --failover) sets the global obd_timeout
4645 to 5 seconds in order to quickly disconnect, but this caused
4646 other RPCs to time out too quickly. Do not change the global
4647 obd_timeout for force cleanup, only set it for DISCONNECT RPCs.
4649 Severity : enhancement
4650 Frequency : if MDS is started with down OST
4651 Bugzilla : 9439,5706
4652 Description: Allow startup/shutdown of an MDS without depending on the
4653 availability of the OSTs.
4654 Details : Asynchronously call mds_lov_synchronize during MDS startup.
4655 Add appropriate locking and lov-osc refcounts for safe
4656 cleaning. Add osc abort_inflight calls in case the
4657 synchronize never started.
4660 Frequency : occasional (Cray XT3 only)
4662 Description: root not authorized to access files in CRAY_PORTALS environment
4663 Details : The client process capabilities were not honoured on the MDS in
4664 a CRAY_PORTALS/CRAY_XT3 environment. If the file had previously
4665 been accessed by an authorized user then root was able to access
4666 the file on the local client also. The root user capabilities
4667 are now allowed on the MDS, as this environment has secure UID.
4670 Frequency : occasional
4672 Description: ldiskfs "too long searching" message happens too often
4673 Details : A debugging message (otherwise harmless) prints too often on
4674 the OST console. This has been reduced to only happen when
4675 there are fragmentation problems on the filesystem.
4680 Description: Division by zero in statfs when all OSCs are inactive
4681 Details : lov_get_stripecnt() returns zero due to incorrect order of checks,
4682 lov_statfs divides by value returned by lov_get_stripecnt().
4686 Bugzilla : 9489, 3273
4687 Description: First write from each client to each OST was only 4kB in size,
4688 to initialize client writeback cache, which caused sub-optimal
4689 RPCs and poor layout on disk for the first writen file.
4690 Details : Clients now request an initial cache grant at (re)connect time
4691 and so that they can start streaming writes to the cache right
4692 away and always do full-sized RPCs if there is enough data.
4693 If the OST is rebooted the client also re-establishes its grant
4694 so that client cached writes will be honoured under the grant.
4699 Description: Slow ls (and stat(2) syscall) on files residing on IO-loaded OSTs
4700 Details : Now I/O RPCs go to different portal number and (presumably) fast
4701 lock requests (and glimses) and other RPCs get their own service
4702 threads pool that should be able to service those RPCs
4705 Severity : enhancement
4707 Description: Ability to exchange lustre version between client and servers and
4708 issue warnings at client side if client is too old. Also for
4709 liblustre clients there is ability to refuse connection of too old
4711 Details : New 'version' field is added to connect data structure that is
4712 filled with version info. That info is later checked by server and
4716 Frequency : rare, liblustre only.
4717 Bugzilla : 9296, 9581
4718 Description: Two simultaneous writes from liblustre at offset within same page
4719 might proceed at the same time overwriting eachother with stale
4721 Details : I/O lock withing llu_file_prwv was released too early, before data
4722 actually was hitting the wire. Extended lock-holding time until
4723 server acknowledges receiving data.
4726 Frequency : extremely rare. Never observed in practice.
4728 Description: avoid generating lustre_handle cookie of 0.
4729 Details : class_handle_hash() generates handle cookies by incrementing
4730 global counter, and can hit 0 occasionaly (this is unlikely, but
4731 not impossible, because initial value of cookie counter is
4732 selected randonly). Value of 0 is used as a sentinel meaning
4733 "unassigned handle" --- avoid it. Also coalesce two critical
4734 sections in this function into one.
4736 Severity : enhancement
4738 Description: allow liblustre clients to delegate truncate locking to OST
4739 Details : To avoid overhead of locking, liblustre client instructs OST to
4740 take extent lock in ost_punch() on client's behalf. New connection
4741 flag is added to handle backward compatibility.
4743 Severity : enhancement
4744 Bugzilla : 4928, 7341, 9758
4745 Description: allow number of OST service threads to be specified
4746 Details : a module parameter allows the number of OST service threads
4747 to be specified via "options ost ost_num_threads={N}" in the
4748 OSS's /etc/modules.conf or /etc/modprobe.conf.
4752 Bugzilla : 6146, 9635, 9895
4753 Description: servers crash with bad pointer in target_handle_connect()
4754 Details : In rare cases when a client is reconnecting it was possible that
4755 the connection request was the last reference for that export.
4756 We would temporarily drop the export reference and get a new
4757 one, but this may have been the last reference and the export
4758 was just destroyed. Get new reference before dropping old one.
4760 Severity : enhancement
4761 Frequency : if client is started with failover MDS
4763 Description: Allow multiple MDS hostnames in the mount command
4764 Details : Try to read the configuration from all specified MDS
4765 hostnames during a client mount in case the "primary"
4768 Severity : enhancement
4770 Description: Stop sending data to evicted clients as soon as possible.
4771 Details : Check if the client we are about to send or are sending data to
4772 was evicted already. (Check is done every second of waiting,
4773 for which l_wait_event interface was extended to allow checking
4774 of exit condition at specified intervals).
4777 Frequency : rare, normally only when NFS exporting is done from client
4779 Description: 'bad disk LOV MAGIC: 0x00000000' error when chown'ing files
4781 Details : Make mds_get_md() recognise empty md case and set lmm size to 0.
4784 Frequency : always, if srand() is called before liblustre initialization
4786 Description: Liblustre uses system PRNG disturbing its usage by user application
4787 Details : Introduce internal to lustre fast and high-quality PRNG for
4788 lustre usage and make liblustre and some other places in generic
4789 lustre code to use it.
4791 Severity : enhancement
4792 Bugzilla : 9477, 9557, 9870
4793 Description: Verify that the MDS configuration logs are updated when xml is
4794 Details : Check if the .xml configuration logs are newer than the config
4795 logs stored on the MDS and report an error if this is the case.
4796 Request --write-conf, or allow starting with --old_conf.
4798 Severity : enhancement
4800 Description: Handle symlinks in the path when checking if Lustre is mounted.
4801 Details : Resolve intermediate symlinks when checking if a client has
4802 mounted a filesystem to avoid duplicate client mounts.
4807 Description: lconf can hit an error exception but still return success.
4808 Details : The lconf command catches the Command error exception at the top
4809 level script context and will exit with the associated exit
4810 status, but doesn't ensure that this exit status is non-zero.
4815 Description: failure of ptlrpc thread startup can cause oops
4816 Details : Starting a ptlrpc service thread can fail if there are a large
4817 number of threads or the server memory is very fragmented.
4818 Handle this without oopsing.
4821 Frequency : always, only if liblustre and non-default acceptor port was used
4823 Description: liblustre cannot connect to servers with non-default acceptor port
4824 Details : tcpnal_set_default_params() was not called and was therefore
4825 ignoring the environment varaible TCPNAL_PORT, as well as other
4826 TCPNAL_ environment variables
4831 Description: two objects could be created on the same OST for a single file
4832 Details : If an OST is down, in some cases it was possible to create two
4833 objects on a single OST for a single file. No problems other
4834 than potential performance impact and spurious error messages.
4838 Bugzilla : 5681, 9562
4839 Description: Client may oops in ll_unhash_aliases
4840 Details : Client dcache may become inconsistent in race condition.
4841 In some cases "getcwd" can fail if the current directory is
4847 Description: Inode refcounting problems in NFS export code
4848 Details : link_raw functions used to call d_instantiate without obtaining
4849 extra inode reference first.
4853 Bugzilla : 9942, 9903
4854 Description: Referencing freed requests leading to crash, memleaks with NFS.
4855 Details : We used to require that call to ll_revalidate_it was always
4856 followed by ll_lookup_it. Also with revalidate_special() it is
4857 possible to call ll_revalidate_it() twice for the same dentry
4858 even if first occurence returned success. This fix changes semantic
4859 between DISP_ENQ_COMPLETE disposition flag to mean there is extra
4860 reference on a request referred from the intent.
4861 ll_intent_release() then releases such a request.
4864 Frequency : rare, normally benchmark loads only
4866 Description: unlinked inodes were kept in memory on the client
4867 Details : If a client is repeatedly creating and unlinking files it
4868 can accumulate a lot of stale inodes in the inode slab cache.
4869 If there is no other client load running this can cause the
4870 client node to run out of memory. Instead flush old inodes
4871 from client cache that have the same inode number as a new inode.
4874 Frequency : SLES9 2.6.5 kernel and long filenames only
4875 Bugzilla : 9969, 10379
4876 Description: utime reports stale NFS file handle
4877 Details : SLES9 uses out-of-dentry names in some cases, which confused
4878 the lustre dentry revalidation. Change it to always use the
4882 Frequency : rare, unless heavy write-truncate concurrency is continuous
4883 Bugzilla : 4180, 6984, 7171, 9963, 9331
4884 Description: OST becomes very slow and/or deadlocked during object unlink
4885 Details : filter_destroy() was holding onto the parent directory lock
4886 while truncating+unlinking objects. For very large objects this
4887 may block other threads for a long time and slow overall OST
4888 responsiveness. It may also be possible to get a lock ordering
4889 deadlock in this case, or run out of journal credits because of
4890 the combined truncate+unlink. Solution is to do object truncate
4891 first in one transaction without parent lock, and then do the
4892 final unlink in a new transaction with the parent lock. This
4893 reduces the lock hold time dramatically.
4896 Frequency : rare, 2.4 kernels only
4898 Description: MDS or OST cleanup may trip kernel BUG when dropping kernel lock
4899 Details : mds_cleanup() and filter_cleanup() need to drop the kernel lock
4900 before unmounting their filesystem in order to avoid deadlock.
4901 The kernel_locked() function in 2.4 kernels only checks whether
4902 the kernel lock is held, not whether it is this process that is
4903 holding it as 2.6 kernels do.
4908 Description: MDS or OST may oops/LBUG if a client is connecting multiple times
4909 Details : The client ptlrpc code may be trying to reconnect to a down
4910 server before a previous connection attempt has timed out.
4911 Increase the reconnect interval to be longer than the connection
4912 timeout interval to avoid sending duplicate connections to
4916 Frequency : echo_client brw_test command
4918 Description: fix echo_client to work with OST preallocated code
4919 Details : OST preallocation code (5137) didn't take echo_client IO path
4920 into account: echo_client calls filter methods outside of any
4921 OST thread and, hence, there is no per-thread preallocated
4922 pages and buffers to use. Solution: hijack pga pages for IO. As
4923 a byproduct, this avoids unnecessary data copying.
4927 Bugzilla : 3555, 5962, 6025, 6155, 6296, 9574
4928 Description: Client can oops in mdc_commit_close() after open replay
4929 Details : It was possible for the MDS to return an open request with no
4930 transaction number in mds_finish_transno() if the client was
4931 evicted, but without actually returning an error. Clients
4932 would later try to replay that open and may trip an assertion
4933 Simplify the client close codepath, and always return an error
4934 from the MDS in case the open is not successful.
4937 Frequency : rare, 2.6 OSTs only
4939 Description: OST may deadlock under high load on fragmented files
4940 Details : If there was a heavy load and highly-fragmented OST filesystems
4941 it was possible to have all the OST threads deadlock waiting on
4942 allocation of biovecs, because the biovecs were not released
4943 until the entire RPC IO was completed. Instead, release biovecs
4944 as soon as they are complete to ensure forward IO progress.
4946 Severity : enhancement
4948 Description: Support for specifying external journal device at mount
4949 Details : If an OST or MDS device is formatted with an external journal
4950 device, this device major/minor is stored in the ext3 superblock
4951 and may not be valid for failover. Allow detecting and
4952 specifying the external journal at mount time.
4957 Description: Mounting an MDS with pending unlinked files may cause oops
4958 Details : target_finish_recovery() calls mds_postrecov() which returned
4959 the number of orphans unlinked. mds_lov_connect->mds_postsetup()
4960 considers this an error and immediately begins cleaning up the
4961 lov, just after starting the mds_lov process
4963 Severity : enhancement
4965 Description: Implement 'lfs df' to report actual free space on per-OST basis
4966 Details : Add sub-command 'df' on 'lfs' to report the disk space usage of
4967 MDS/OSDs. Usage: lfs df [-i][-h]. Command Options: '-i' to report
4968 usage of objects; '-h' to report in human readable format.
4970 ------------------------------------------------------------------------------
4972 2005-08-26 Cluster File Systems, Inc. <info@clusterfs.com>
4979 Description: Mounting an ldiskfs file system with mballoc may crash OST node.
4980 Details : ldiskfs mballoc code may reference an uninitialized buddy struct
4981 at startup during orphan unlinking. Instead, skip buddy update
4982 before setup, as it will be regenerated after recovery is complete.
4987 Description: If an OST is inactive, its locks might reference stale inodes.
4988 Details : lov_change_cbdata() must iterate over all namespaces, even if
4989 they are inactive to clear inode references from the lock.
4991 Severity : enhancement
4992 Frequency : occasional, if non-standard max_dirty_mb used
4994 Description: Client will block write RPCs if not enough grant
4995 Details : If a client has max_dirty_mb smaller than max_rpcs_in_flight,
4996 then the client will block writes while waiting for another RPC
4997 to complete instead of consuming its dirty limit. With change
4998 we get improved performance when max_dirty_mb is small.
5000 Severity : enhancement
5001 Bugzilla : 3389, 6253
5002 Description: Add support for supplementary groups on the MDS.
5003 Details : The MDS has an upcall /proc/fs/lustre/mds/{mds}/group_upcall
5004 (set to /usr/sbin/l_getgroups if enabled) which will do MDS-side
5005 lookups for user supplementary groups into a cache.
5009 Description: O_CREAT|O_EXCL open flags in liblustre always return -EEXIST
5010 Details : Make libsysio to not enforce O_EXCL by clearing the flag,
5011 for liblustre O_EXCL is enforced by MDS.
5015 Description: readdir never returns NULL in liblustre.
5016 Details : Corrected llu_iop_getdirentries logic, to return offset of next
5017 dentry in struct dirent.
5021 Frequency : liblustre only, depends on application IO pattern
5022 Description: liblustre clients evicted if not contacting servers
5023 Details : Don't put liblustre clients into the ping_evictor list, so
5024 they will not be evicted by the pinger ever.
5026 Severity : enhancement
5028 Description: Add ability to evict clients by NID from MDS.
5029 Details : By echoing "nid:$NID" string into
5030 /proc/fs/lustre/mds/.../evict_client client with nid that equals to
5031 $NID would be instantly evicted from this MDS and from all active
5032 OSTs connected to it.
5036 Description: Do not query file size twice, somewhat slowing stat(2) calls.
5037 Details : lookup_it_finish() used to query file size from OSTs that was not
5042 Description: service threads change working directory to that of init
5043 Details : Starting lustre service threads may pin the working directory
5044 of the parent thread, making that filesystem busy. Threads
5045 now change to the working directory of init to avoid this.
5049 Frequency : during shutdown only
5050 Description: shutdown with a failed MDS or OST can cause unmount to hang
5051 Details : Don't resend DISCONNECT messages in ptlrpc_disconnect_import()
5056 Frequency : 2.6 only
5057 Description: chmod/chown may include an extra supplementary group
5058 Details : ll{,u}_mdc_pack_op_data() does not properly initialize the
5059 supplementary group and if none is specified this is used.
5062 Bugzilla : 5479 (6816)
5064 Description: Racing open + rm can assert client in mdc_set_open_replay_data()
5065 Details : If lookup is in progress on a file that is unlinked we might try
5066 to revalidate the inode and fail in revalidate after lookup is
5067 complete and ll_file_open() enqueues the open again but
5068 it_open_error() was not checking DISP_OPEN_OPEN errors correctly.
5071 Frequency : always, if lconf --abort_recovery used
5073 Description: lconf --abort_recovery fails with 'Operation not supported'
5074 Details : lconf was attempting to abort recovery on the MDT device and not
5077 ------------------------------------------------------------------------------
5079 2005-08-08 Cluster File Systems, Inc. <info@clusterfs.com>
5084 Frequency : rare (only unsupported configurations with a node running as an
5086 Bugzilla : 6514, 5137
5087 Description: Mounting a Lustre file system on a node running as an OST could
5089 Details : OSTs now preallocates memory needed to write out data at
5090 startup, instead of when needed, to avoid having to
5091 allocate memory in possibly low memory situations.
5092 Specifically, if the file system is mounted on on OST,
5093 memory pressure could force it to try to write out data,
5094 which it needed to allocate memory to do. Due to the low
5095 memory, it would be unable to do so and the node would
5096 become unresponsive.
5098 Severity : enhancement
5100 Description: Addition of lconf --service command line option
5101 Details : lconf now accepts a '--service <arg>' option, which is
5102 shorthand for 'lconf --group <arg> --select <arg>=<hostname>'
5104 Severity : enhancement
5106 Description: Failover mode is now the default for OSTs.
5107 Details : By default, OSTs will now run in failover mode. To return to
5108 the old behaviour, add '--failout' to the lmc line for OSTs.
5110 Severity : enhancement
5112 Description: Health checks are now provided for MDS and OSTs
5113 Details : Additional detailed health check information on MSD and OSTs
5114 is now provided through the procfs health_check value.
5117 Frequency : occasional, depends on IO load
5119 Description: Disk fragmentation on the OSTs could eventually cause slowdowns
5120 after numerous create/delete cycles
5121 Details : The ext3 inode allocation policy would not allocate new inodes
5122 very well on the OSTs because there are no new directories
5123 being created. Instead we look for groups with free space if
5124 the parent directories are nearly full.
5129 Description: Network or server problems during mount may cause partially
5130 mounted clients instead of returning an error.
5131 Details : The config llog parsing code may overwrite the error return
5132 code during mount error handling, returning success instead
5138 Description: MDS can fail to allocate large reply buffers
5139 Details : After long uptimes the MDS can fail to allocate large reply
5140 buffers (e.g. zconf client mount config records) due to memory
5141 fragmentation or consumption by the buffer cache. Preallocate
5142 some large reply buffers so that these replies can be sent even
5143 under memory pressure.
5147 Frequency : rare (liblustre)
5148 Description: fsx running with liblustre complained that using truncate() to
5149 extend the file doesn't work. This patch corrects that issue.
5150 Details : This is the liblustre equivalent of the fix for bug 6196. Fixes
5151 ATTR_SIZE and lsm use in llu_setattr_raw.
5155 Frequency : rare, only 2.6 kernels
5156 Description: Unusual file access patterns on the MDS may result in inode
5157 data being lost in very rare circumstances.
5158 Details : Bad interaction between the ea-in-inode patch and the "no-read"
5159 code in the 2.6 kernel caused the inode and/or EA data not to
5160 be read from disk, causing single-file corruption.
5164 Frequency : rare, only 2.6 filesystems using extents
5165 Description: Heavy concurrent write and delete load may cause data corruption.
5166 Details : It was possible under high-load situations to have an extent
5167 metadata block in the block device cache from a just-unlinked
5168 file overwrite a newly-allocated data block. We now unmap any
5169 metadata buffers that alias just-allocated data blocks.
5173 Frequency : filesystems with default stripe_count larger than 77
5174 Description: lconf+mke2fs fail when formatting filesystem with > 77 stripes
5175 Details : lconf specifies an inode size of 4096 bytes when the default
5176 stripe_count is larger than 77. This conflicts with the default
5177 inode density of 1 per 4096 bytes. Allocate smaller inodes in
5178 this case to avoid pinning too much memory for large EAs.
5180 ------------------------------------------------------------------------------
5182 2005-07-07 Cluster File Systems, Inc. <info@clusterfs.com>
5187 Frequency : rare (extremely heavy IO load with hundreds of clients)
5189 Description: Client is evicted, gets IO error writing to file
5190 Details : lock ordering changes for bug 5492 reintroduced bug 3267 and
5191 caused clients to be evicted for AST timeouts. The fixes in
5192 bug 5192 mean we no longer need to have such short AST timeouts
5193 so ldlm_timeout has been increased.
5196 Frequency : occasional during --force or --failover shutdown under load
5197 Bugzilla : 5949, 4834
5198 Description: Server oops/LBUG if stopped with --force or --failover under load
5199 Details : a collection of import/export refcount and cleanup ordering
5200 issues fixed for safer force cleanup
5203 Frequency : only filesystems larger than 120 OSTs
5204 Bugzilla : 5990, 6223
5205 Description: lfs getstripe would oops on a very large filesystem
5206 Details : lov_getconfig used kfree on vmalloc'd memory
5209 Frequency : only filesystems exporting via NFS to Solaris 10 clients
5210 Bugzilla : 6242, 6243
5211 Description: reading from files that had been truncated to a non-zero size
5212 but never opened returned no data
5213 Details : ll_file_read() reads zeros from no-object files to EOF
5218 Description: A bug in MDS/OSS recovery could cause the OSS to fail an assertion
5219 Details : There's little harm in aborting MDS/OSS recovery and letting it
5220 try again, so I removed the LASSERT and return an error instead.
5222 Severity : enhancement
5224 Description: New debugging infrastructure for tracking down data corruption
5225 Details : The I/O checksum code was replaced to: (a) control it at runtime,
5226 (b) cover more of the client-side code path, and (c) try to narrow
5227 down where problems occurred
5231 Bugzilla : 3819, 4364, 4397, 6313
5232 Description: Racing close and eviction MDS could cause assertion in mds_close
5233 Details : It was possible to get multiple mfd references during close and
5234 client eviction, leading to one thread referencing a freed mfd.
5236 Severity: : enhancement
5237 Bugzilla : 3262, 6359
5238 Description: Attempts to reconnect to servers are now more aggressive.
5239 Details : This builds on the enhanced upcall-less recovery that was added
5240 in 1.4.2. When trying to reconnect to servers, clients will
5241 now try each server in the failover group every 10 seconds. By
5242 default, clients would previously try one server every 25 seconds.
5247 Description: After recovery, certain operations trigger a failed
5248 assertion on a client.
5249 Details : Failing over an mds, using lconf -d --failover, while a
5250 client was doing a readdir() call would cause the client to
5251 LBUG after recovery completed and the readdir() was resent.
5253 Severity : enhancement
5255 Description: Default groups are now added by lconf
5256 Details : You can now run lconf --group <servicename> without having to
5257 manually add groups with lmc.
5260 Frequency : occasional
5262 Description: Nodes with an elan id of 0 trigger a failed assertion
5265 Frequency : always when accessing e.g. tty/console device nodes
5267 Description: tty and some other devices nodes cannot be used on lustre
5268 Details : file's private_data field is used by device data and lustre
5269 values in there got lost. New field was added to struct file to
5270 store fs-specific private data.
5273 Frequency : when exporting Lustre via NFS
5275 Description: NFSD failed occasionally when looking up a path component
5276 Details : NFSD is looking up ".." which was broken in ext3 directories
5277 that had grown large enough to become hashed.
5280 Frequency : Clusters with multiple interfaces not on the same subnet
5282 Description: Nodes will repeatedly try to reconnect to an interface which it
5283 cannot reach and report an error to the log.
5284 Details : Extra peer list entries will be created by lconf with some peers
5285 unreachable. lconf now validates the peer before adding it.
5288 Frequency : Only if a default stripe is set on the filesystem root.
5290 Description: Setting a default stripe on the filesystem root prevented the
5291 filesystem from being remounted.
5292 Details : The client was sending extra request flags in the root getattr
5293 request and did not allocate a reply buffer for the dir EA.
5296 Frequency : occasional, higher if lots of files are accessed by one client
5297 Bugzilla : 6159, 6097
5298 Description: Client trips assertion regarding lsm mismatch/magic
5299 Details : While revalidating inodes the VFS looks up inodes with ifind()
5300 and in rare cases can find an inode that is being freed.
5301 The ll_test_inode() code will free the lsm during ifind()
5302 when it finds an existing inode and then the VFS later attaches
5303 this free lsm to a new inode.
5307 Bugzilla : 6422, 7030
5308 Description: MDS deadlock between mkdir and client eviction
5309 Details : Creating a new file via mkdir or mknod (starting a transaction
5310 and getting the ns lock) can deadlock with client eviction
5311 (gets ns lock and trying to finish a synchronous transaction).
5314 Frequency : occasional
5315 Description: While starting a server, the fsfilt_ext3 module could not be
5317 Details : CFS's improved ext3 filesystem is named ldiskfs for 2.6
5318 kernels. Previously, lconf would still use the ext3 name
5319 when trying to load modules. Now, it will correctly use
5320 ext3 on 2.4 and ldiskfs on 2.6.
5322 Severity : enhancement
5323 Description: The default stripe count has been changed to 1
5324 Details : The interpretation of the default stripe count (0, to lfs
5325 or lmc) has been changed to mean striping across a single
5326 OST, rather than all available. For general usage we have
5327 found a stripe count of 1 or 2 works best.
5329 Severity : enhancement
5330 Description: Add support for compiling against Cray portals.
5331 Details : Conditional compiling for some areas that are different
5335 Frequency : occasional
5336 Bugzilla : 6409, 6834
5337 Description: Creating files with an explicit stripe count may lead to
5338 a failed assertion on the MDS
5339 Details : If some OSTs are full or unavailable, creating files may
5340 trigger a failed assertion on the MDS. Now, Lustre will
5341 try to use other servers or return an error to the
5345 Frequency : occasional
5347 Description: Multiple concurrent overlapping read+write on multiple SMP nodes
5348 caused lock timeout during readahead (since 1.4.2).
5349 Details : Processes doing readahead might match a lock that hasn't been
5350 granted yet if there are overlapping and conflicting lock
5351 requests. The readahead process waits on ungranted lock
5352 (original lock is CBPENDING), while OST waits for that process
5353 to cancel CBPENDING read lock and eventually evicts client.
5355 Severity : enhancement
5357 Description: Initial enabling of flock support for clients
5358 Details : Implements fcntl advisory locking and file status functions.
5359 This feature is provided as an optional mount flag (default
5360 off), and is NOT CURRENTLY SUPPORTED. Not all types of record
5361 locking are implemented yet, and those that are are not guaranteed
5362 to be completely correct in production environments.
5363 mount -t lustre -o [flock|noflock] ...
5366 Frequency : occasional
5368 Description: OSTs running 2.4 kernels but with extents enabled might trip an
5369 assertion in the ext3 JBD (journaling) layer.
5370 Details : The b_committed_data struct is protected by the big kernel lock
5371 in 2.4 kernels, serializing journal_commit_transaction() and
5372 ext3_get_block_handle->ext3_new_block->find_next_usable_block()
5373 access to this struct. In 2.6 kernels there is finer grained
5374 locking to improve SMP performance of the JBD layer.
5378 Description: Changes the "SCSI I/O Stats" kernel patch to default to "enabled"
5380 -----------------------------------------------------------------------------
5382 2005-05-05 Cluster File Systems, Inc. <info@clusterfs.com>
5384 NOTE: Lustre 1.4.2 uses an incompatible network protocol than previous
5385 versions of Lustre. Please update all servers and clients to
5386 version 1.4.2 or later at the same time. You must also run
5387 "lconf --write-conf {config}.xml" on the MDS while it is stopped
5388 to update the configuration logs.
5390 - fix for HPUX NFS client breakage when NFS exporting Lustre (5781)
5391 - mdc_enqueue does not need max_mds_easize request buffer on send (5707)
5392 - swab llog records of type '0' so we get proper header size/idx (5861)
5393 - send llog cancel req to DLM cancel portal instead of cb portal (5515)
5394 - fix rename of one directory over another leaking an inode (5953)
5395 - avoid SetPageDirty on 2.6 (5981)
5396 - don't re-add just-being-destroyed locks to the waiting list (5653)
5397 - when creating new directories, inherit the parent's custom
5398 striping settings if present parent (3048)
5399 - flush buffers from cache before direct IO in 2.6 obdfilter (4982)
5400 - don't hold i_size_sem in ll_nopage() and ll_ap_refresh_count (6077)
5401 - don't hold client locks on temporary worklist from l_lru (5666)
5402 - handle IO errors in 2.6 obdfilter bio completion routine (6046)
5403 - automatically evict dead clients (5921)
5404 - Update file size properly in create+truncate+fstat case (6196)
5405 - Do not unhash mountpoint dentries, do not allow removal of
5407 - Avoid lock ordering deadlock issue with write/truncate (6203,5654)
5408 - reserve enough journal credits in fsfilt_start_log for setattr (4554)
5409 - ldlm_enqueue freed-export error path would always LBUG (6149,6184)
5410 - don't reference lr_lvb_data until after we hold lr_lvb_sem (6170)
5411 - don't overwrite last_rcvd if there is a *_client_add() error (6086)
5412 - Correctly handle reads of files with no objects (6243)
5413 - lctl recover will also mark a device active if deactivate used (5933)
5415 - by default create 1 inode per 4kB space on MDS, per 16kB on OSTs
5416 - allow --write-conf on an MDS with different nettype than client (5619)
5417 - don't write config llogs to MDS for mounts not from that MDS (5617)
5418 - lconf should create multiple TCP connections from a client (5201)
5419 - init scripts are now turned off by default; run chkconfig --on
5420 lustre and chkconfig --on lustrefs to use them
5421 - upcalls are no longer needed for clients to recover to failover
5423 - add --abort-recovery option to lconf to abort recovery on device
5425 - add support for an arbitrary number of OSTs (3026)
5426 - Quota support protocol changes.
5427 - forward compatibility changes to wire structs (6007)
5428 - rmmod NALs that might be loaded because of /etc/modules.conf (6133)
5429 - support for mountfsoptions and clientoptions to the Lustre LDAP (5873)
5430 - improved "lustre status" script
5431 - initialize blocksize for non-regular files (6062)
5432 - added --disable-server and --disable-client configure options (5782)
5433 - introduce a lookup cache for lconf to avoid repeated DB scans (6204)
5434 - Vanilla 2.4.29 support
5435 - increase maximum number of obd devices to 520 (6242)
5436 - remove the tcp-zero-copy patch from the suse-2.4 series (5902)
5437 - Quadrics Elan drivers are now included for the RHEL 3 2.4.21 and
5438 SLES 9 2.6.5 kernels
5439 - limit stripes per file to 160 (the maximum EA size) (6093)
5441 2005-03-22 Cluster File Systems, Inc. <info@clusterfs.com>
5444 - don't LASSERT in ll_release on NULL lld with NFS export (4655, 5760)
5445 - hold NS lock when calling handle_ast_error->del_waiting_lock (5746)
5446 - fix setattr mtime regression from lovcleanup merge (4829, 5669)
5447 - workaround for 2.6 crash in ll_unhash_aliases (5687, 5210)
5448 - small ext3 extents cleanups and fixes (5733)
5449 - improved mballoc code, several small races and bugs fixed (5733, 5638)
5450 - kernel version 43 - fix remove_suid bugs in both 2.4 and 2.6 (5695)
5451 - avoid needless client->OST connect, fix handle mismatch (5317)
5452 - fix DLM error path that led to out-of-sync client, long delays (5779)
5453 - support common vfs-enforced mount options (nodev,nosuid,noexec) (5637)
5454 - fix several locking issues related to i_size (5492,5624,5654,5672)
5455 - don't move pending lock onto export if it is already evicted (5683)
5456 - fix kernel oops when creating .foo in unlinked directory (5548)
5457 - fix deadlock in obdfilter statistics vs. object create (5811)
5458 - use time_{before,after} to avoid timer jiffies wrap (5882)
5459 - shutdown --force/--failover stability (3607,3651,4797,5203,4834)
5460 - Do not leak request if server was not able to process it (5154)
5461 - If mds_open unable to find parent dir, make that negative lookup(5154)
5462 - don't create new directories with extent-mapping (5909, 5936)
5464 - fix lustre/lustrefs init scripts for SuSE (patch from Scali, 5702)
5465 - don't hold the pinger_sem in ptlrpc_pinger_sending_on_import
5466 - change obd_increase_kms to obd_adjust_kms (up or down) (5654)
5467 - lconf, lmc search both /usr/lib and /usr/lib64 for Python libs (5800)
5468 - support for RHEL4 kernel on i686 (5773)
5469 - provide error messages when incompatible logs are encountered (5898)
5471 2005-02-18 Cluster File Systems, Inc. <info@clusterfs.com>
5472 * version 1.4.0.10 (1.4.1 release candidate 1)
5474 - don't keep a lock reference when lock is not granted (4238)
5475 - unsafe list practices (rarely) led to infinite eviction loop (4908)
5476 - add per-fs limit of Lustre pages in page cache, avoid OOM (4699)
5477 - drop import inflight refcount on signal_completed_replay error (5255)
5478 - unlock page after async write error during send (3677)
5479 - handle missing objects in filter_preprw_read properly (5265)
5480 - no transno return for symlink open, don't save no-trasno open (3440)
5481 - don't try to complete elan receive that already failed (4012)
5482 - free RPC server reply state on error (5406)
5483 - clean up thread from ptlrpc_start_thread() on error (5160)
5484 - readahead could read extra page into cache that wasn't ejected (5388)
5485 - prevent races in class_attach/setup/cleanup/detach (5260)
5486 - don't dereference de->d_inode after l_dput of de (5458)
5487 - use "int" for stripe value returned from lock_to_stripe (5544)
5488 - mballoc allocation and error-checking fixes in 2.6 (5504)
5489 - block device patches to fix I/O request sizes in 2.6 (5482)
5490 - look up hostnames for IB nals (5602)
5491 - 2.6 changed lock ordering of 2 semaphores, caused deadlock (5654)
5492 - don't start multiple acceptors for the same port (5277)
5493 - fix incorrect LASSERT in mds_getattr_name (5635)
5494 - export a proc file for general "ping" checking (5628)
5495 - fix "lfs check" to not block when the MDS is down (5628)
5497 - service request history (4965)
5498 - put {ll,lov,osc}_async_page structs in a single slab (4699)
5499 - create an "evict_client" /proc entry on OSTs, like the MDS has
5500 - fix mount usage message, return errors per mount(8) (5168)
5501 - change grep [] to grep "[]" in tests so they work in more UMLs
5502 - fix ppc64/x86_64 spec to use %{_libdir} instead of /usr/lib (5389)
5503 - remove ancient LOV_MAGIC_V0 EA support (5047)
5504 - add "disk I/Os in flight" and "I/O req time" stats in obdfilter
5505 - align r/w RPCs to PTLRPC_MAX_BRW_SIZE boundary for performance (3451)
5506 - allow readahead allocations to fail when low on memory (5383)
5507 - mmap locking landed again, after considerable improvement (2828)
5508 - add get_hostaddr() to lustreDB.py for LDAP support (5459)
5510 2004-11-23 Cluster File Systems, Inc. <info@clusterfs.com>
5513 - send OST transaction number in read/write reply to free req (4966)
5514 - don't ASSERT in ptl_send_rpc() if we run out of memory (5119)
5515 - lock /proc/sys/portals/routes internal state, avoiding oops (4827)
5516 - the watchdog thread now runs as interruptible (5246)
5517 - flock/lockf fixes (but it's still disabled, pending 5135)
5518 - don't use EXT3 constants in llite code (5094)
5519 - memory shortage at startup could cause assertion (5176)
5521 - reorganization of lov code
5522 - single portals codebase
5524 - add extents/mballoc support (5025)
5525 - direct I/O reads in the obdfilter (4048)
5526 - kernel patches from LNXI for 2.6 (bluesmoke, perfctr, mtd, kexec)
5528 tbd Cluster File Systems, Inc. <info@clusterfs.com>
5531 - send OST transaction number in read/write reply to free req (4966)
5532 - don't ASSERT in ptl_send_rpc() if we run out of memory (5119)
5533 - lock /proc/sys/portals/routes internal state, avoiding oops (4827)
5534 - the watchdog thread now runs as interruptible (5246)
5535 - handle missing objects in filter_preprw_read properly (5265)
5536 - unsafe list practices (rarely) led to infinite eviction loop (4908)
5537 - drop import inflight refcount on signal_completed_replay error (5255)
5538 - unlock page after async write error during send (3677)
5539 - return original error code on reconstructed replies (3761)
5540 - no transno return for symlink open, don't save no-trasno open (3440)
5542 - add pid to ldlm debugging output (4922)
5543 - bump the watchdog timeouts -- we can't handle 30sec yet
5544 - extra debugging for orphan dentry/inode bug (5259)
5546 2004-11-16 Cluster File Systems, Inc. <info@clusterfs.com>
5549 - fix TCP_NODELAY bug, which caused extreme perf regression (5134)
5550 - allocate qswnal tx descriptors singly to avoid fragmentation (4504)
5551 - don't LBUG on obdo_alloc() failure, use OBD_SLAB_ALLOC() (4800)
5552 - fix NULL dereference in /proc/sys/portals/routes (4827)
5553 - allow failed mdc_close() operations to be interrupted (4561)
5554 - stop precreate on OST before MDS would time out on it (4778)
5555 - don't send partial-page writes before EOF from client (4410)
5556 - discard client grant for sub-page writes on large-page clients (4520)
5557 - don't free dentries not owned by NFS code, check generation (4806)
5558 - fix lsm leak if mds_create_objects() fails (4801)
5559 - limit debug_daemon file size, always print CERROR messages (4789)
5560 - use transno after validating reply (3892)
5561 - process timed out requests if import state changes (3754)
5562 - update mtime on OST during writes, return in glimpse (4829)
5563 - add mkfsoptions to LDAP (4679)
5564 - use ->max_readahead method instead of zapping global ra (5039)
5565 - don't interrupt __l_wait_event() during strace
5567 - add software watchdogs to catch hung threads quickly (4941)
5568 - make lustrefs init script start after nfs is mounted
5569 - fix CWARN/ERROR duplication (4930)
5570 - return async write errors to application if possible (2248)
5571 - add /proc/sys/portal/memused (bytes allocated by PORTALS_ALLOC)
5572 - print NAL number in %x format (4645)
5573 - update barely-supported suse-2.4.21-171 series (4842)
5574 - support for sles 9 %post scripts
5575 - support for building 2.6 kernel-source packages
5576 - support for sles km_* packages
5578 2004-10-07 Cluster File Systems, Inc. <info@clusterfs.com>
5581 - ignore -ENOENT errors in osc_destroy (3639)
5582 - notify osc create thread that OSC is being cleaned up (4600)
5583 - add nettype argument for llmount in #5d in conf-sanity.sh (3936)
5584 - reconstruct ost_handle() like mds_handle() (4657)
5585 - create a new thread to do import eviction to avoid deadlock (3969)
5586 - let lconf resolve symlinked-to devices (4629)
5587 - don't unlink "objects" from directory with default EA (4554)
5588 - hold socknal file ref over connect in case target is down (4394)
5589 - allow more than 32000 subdirectories in a single directory (3244)
5590 - fix blocks count for O_DIRECT writes (3751)
5591 - OST returns ENOSPC from object create when no space left (4539)
5592 - don't send truncate RPC if file size isn't changing (4410)
5593 - limit OSC precreate to 1/2 of value OST considers bogus (4778)
5594 - bind to privileged port in socknal and tcpnal (3689)
5596 - rate limit CERROR/CWARN console message to avoid overload (4519)
5597 - GETFILEINFO dir ioctl returns LOV EA + MDS stat in 1 call (3327)
5598 - basic mmap support (3918)
5599 - kernel patch series update from b1_4 (4711)
5601 2004-09-16 Cluster File Systems, Inc. <info@clusterfs.com>
5604 - avoid crash during MDS cleanup with OST shut down (2775)
5605 - fix loi_list_lock/oig_lock inversion on interrupted IO (4136)
5606 - don't use bad inodes on the MDS (3744)
5607 - dynamic object preallocation to improve recovery speed (4236)
5608 - don't hold spinlock over lock dumping or change debug flags (4401)
5609 - don't zero obd_dev when it is force cleaned (3651)
5610 - print grants to console if they go negative (4431)
5611 - "lctl deactivate" will stop automatic recovery attempts (3406)
5612 - look for existing locks in ldlm_handle_enqueue() (3764)
5613 - don't resolve lock handle twice in recovery avoiding race (4401)
5614 - revalidate should check working dir is a directory (4134)
5616 - don't always mark "slow" obdfilter messages as errors (4418)
5618 2004-08-24 Cluster File Systems, Inc. <info@clusterfs.com>
5621 - don't close LustreDB during write_conf until it is done (3860)
5622 - fix typo in lconf for_each_profile (3821)
5623 - allow dumping logs from multiple threads at one time (3820)
5624 - don't allow multiple threads in OSC recovery (3812)
5625 - fix debug_size parameters (3864)
5626 - fix mds_postrecov to initialize import for llog ctxt (3121)
5627 - replace config semaphore with spinlock (3306)
5628 - be sure to send a reply for a CANCEL rpc with bad export (3863)
5629 - don't allow enqueue to complete on a destroyed export (3822)
5630 - down write_lock before checking llog header bitmap (3825)
5631 - recover from lock replay timeout (3764)
5632 - up llog sem before sending rpc (3652)
5633 - reduce ns lock hold times when setting kms (3267)
5634 - change a dlm LBUG to LASSERTF, to maybe learn something (4228)
5635 - fix NULL deref and obd_dev leak on setup error (3312)
5636 - replace some LBUG about llog ops with error handling (3841)
5637 - don't match INVALID dentries from d_lookup and spin (3784)
5638 - hold dcache_lock while marking dentries INVALID and hashing (4255)
5639 - fix invalid assertion in ptlrpc_set_wait (3880)
5641 - add libwrap support for the TCP acceptor (3996)
5642 - add /proc/sys/portals/routes for non-root route listing (3994)
5643 - allow setting MDS UUID in .xml (2580)
5644 - print the stack of a process that LBUGs (4228)
5646 2004-07-14 Cluster File Systems, Inc. <info@clusterfs.com>
5649 - don't cleanup request in ll_file_open() on failed MDS open (3430)
5650 - make sure to unset replay flag from failed open requests (3440)
5651 - if default stripe count is 0, use OST count for inode size (3636)
5652 - update parent mtime/ctime on client for create/unlink (2611)
5653 - drop dentry ref in ext3_add_link from open_connect_dentry (3266)
5654 - free recovery state on server during a forced cleanup (3571)
5655 - unregister_reply for resent reqs (3063)
5656 - loop back devices mounting and status check on 2.6 (3563)
5657 - fix resource-creation race that can provoke i_size == 0 (3513)
5658 - don't try to use bad inodes returned from MDS/OST fs lookup (3688)
5659 - more debugging for page-accounting assertion (3746)
5660 - return -ENOENT instead of asserting if ost getattr+unlink race (3558)
5661 - avoid deadlock after precreation failure (3758)
5662 - fix race and lock order deadlock in orphan handling (3450, 3750)
5663 - add validity checks when grabbing inodes from l_ast_data (3599)
5665 - add /proc/.../recovery_status to obdfilter (3428)
5666 - lightweight CDEBUG infrastructure, debug daemon (3668)
5667 - change default OSC RPC parameters to be better on small clusters
5668 - turn off OST read cache for files smaller than 32MB
5669 - install man pages and include them in rpms (3100)
5670 - add new init script for (un)mounting lustre filesystems (2593)
5671 - run chkconfig in %post for init scripts (3701)
5672 - drop scimac NAL (unmaintained)
5674 2004-06-17 Cluster File Systems, Inc. <info@clusterfs.com>
5677 - clean kiobufs before and after use (3485)
5678 - strip trailing '/'s before comparing paths with /proc/mounts (3486)
5679 - remove assertions to work around "in-flight rpcs" recovery bug (3063)
5680 - change init script to fail more clearly if not run as root (1528)
5681 - allow clients to reconnect during replay (1742)
5682 - fix ns_lock/i_sem lock ordering deadlock for kms update (3477)
5683 - don't do DNS lookups on NIDs too small for IP addresses (3442)
5684 - re-awaken ptlrpcd if new requests arrive during check_set (3554)
5685 - fix cond_resched (3554)
5686 - only evict unfinished clients after recovery (3515)
5687 - allow bulk resend, prevent data loss (3570)
5688 - dynamic ptlrpc request buffer allocation (2102)
5689 - don't allow unlinking open directory if it isn't empty (2904)
5690 - set MDS/OST threads to umask 0 to not clobber client modes (3359)
5691 - remove extraneous obd dereference causing LASSERT failure (3334)
5692 - don't use get_cycles() when creating temp. files on the mds (3156)
5693 - hold i_sem when setting i_size in ll_extent_lock() (3564)
5694 - handle EEXIST for set-stripe, set proper directory name (3336)
5696 - servers can dump a log evicting a client - lustre.dump_on_timeout=1
5697 - fix ksocknal_fmb_callback() error messages (2918)
5699 2004-05-27 Cluster File Systems, Inc. <info@clusterfs.com>
5702 - don't copy lvb into (possibly NULL) reply on error (2983)
5703 - don't deref dentry after dput, don't free lvb on error (2922)
5704 - use the kms to determine writeback rpc length (2947)
5705 - increment oti_logcookies when osc is inactive (2948)
5706 - update client's i_blocks count via lvb messages (2543)
5707 - handle intent open/close of special files properly (1557)
5708 - mount MDS with errors=remount-ro, like obdfilter (2009)
5709 - initialize lock handle to avoid ASSERT on error cleanup (3057)
5710 - don't use cancelling-locks' kms values (2947)
5711 - use highest lock extent for kms, not last one (2925)
5712 - don't dereference ERR_PTR() dentry in error handling path (3107)
5713 - fix thread race in portals_debug_dumplog() (3122)
5714 - create lprocfs device entries at setup instead of at attach (1519)
5715 - common AST error handler, don't evict client on completion race (3145)
5716 - zero nameidata in detach_mnt in 2.6 (3118)
5717 - verify d_inode after revalidate_special is valid in 2.6 (3116)
5718 - use lustre_put_super() to handle zconf unmounts in 2.6 (3064)
5719 - initialize RPC timeout timer earlier for 2.6 (3219)
5720 - don't dereference NULL reply buffer if mdc_close was never sent (2410)
5721 - print nal/nid for unknown nid (3258)
5722 - additional checks for oscc recovery before doing precreate (3284)
5723 - fix ll_extent_lock() error return code for 64-bit systems (3043)
5724 - don't crash in mdc_close for bad permissions on open (3285)
5725 - zero i_rdev for non-device files (3147)
5726 - clear page->private before handing to FS, better assertion (3119)
5727 - tune the read pipeline (3236)
5728 - fix incorrect decref of invalidated dentry (2350)
5729 - provide read-ahead stats and refine rpc in flight stats (3328)
5730 - don't hold journal transaction open across create RPC (3313)
5731 - update atime on MDS at close time (3265)
5732 - close LDAP connection when recovering to avoid server load (3315)
5733 - update iopen-2.6 patch with fixes from 2399,2517,2904 (3301)
5734 - don't leak open file on MDS after open resend (3325)
5735 - serialize filter_precreate and filter_destroy_precreated (3329)
5736 - loop device shouldn't call sync_dev() for nul device (3092)
5737 - clear page cache after eviction (2766)
5738 - resynchronize MDS->OST in background (2824)
5739 - refuse to mount the same filesystem twice on same mountpoint (3394)
5740 - allow llmount to create routes for mounting behind routers (3320)
5741 - push lock cancellation to blocking thread for glimpse ASTs (3409)
5742 - don't call osc_set_data_with_check() for TEST_LOCK matches (3159)
5743 - fix rare problem with rename on htree directories (3417)
5745 - allow default OST striping configuration per directory (1414)
5746 - fix compilation for qswnal for 2.6 kernels (3125)
5747 - increase maximum number of MDS request buffers for large systems
5748 - change liblustreapi to be useful for external progs like lfsck (3098)
5749 - increase local configuration timeout for slow disks (3353)
5750 - allow configuring ldlm AST timeout - lustre.ldlm_timeout=<seconds>
5752 2004-03-22 Cluster File Systems, Inc. <info@clusterfs.com>
5755 - fixes for glimpse AST timeouts / incorrectly 0-sized files (2818)
5756 - don't overwrite extent policy data in reply if lock was blocked (2901)
5757 - drop filter export grants atomically with removal from device (2663)
5758 - del obd_self_export from work_list in class_disconnect_exports (2908)
5759 - don't LBUG if MDS recovery times out during orphan cleanup (2530)
5760 - swab reply message in mdc_close, other PPC fixes (2464)
5761 - fix destroying of named logs (2325)
5762 - overwrite old logs when running lconf --write_conf (2264)
5763 - bump LLOG_CHUNKSIZE to 8k to allow for larger clusters (2306)
5764 - fix race in target_handle_connect (2898)
5765 - mds_reint_create() should take same inode create lock (2926)
5766 - correct journal credits calculated for CANCEL_UNLINK_LOG (2931)
5767 - don't close files for self_export to avoid uninitialized obd (2936)
5768 - allow MDS with the same name as client node (2939)
5769 - hold dentry reference for closed log files for unlink (2325)
5770 - reserve space for all logs during transactions (2059)
5771 - don't evict page beyond end of stripe extent (2925)
5772 - don't oops on a deleted current working directory (2399)
5773 - handle hard links to targets without a parent properly (2517)
5774 - don't dereference NULL lock when racing during eviction (2867)
5775 - don't grow lock extents when lots of conflicting locks (2919)
5777 2004-03-04 Cluster File Systems, Inc. <info@clusterfs.com>
5780 - account for cache space usage on clients to avoid data loss (974)
5781 - lfsck support in lustre kernel code (2349)
5782 - reduce journal credits needed for BRW writes (2370)
5783 - orphan handling to avoid losing space on client/server crashes
5784 - ptlrpcd can be blocked, stopping ALL progress (2477)
5785 - use lock value blocks to assist in proper KMS, faster stat (1021)
5786 - takes i_sem instead of DLM locks internally on obdfilter (2720)
5787 - recovery for initial connections (2355)
5788 - fixes for mds_cleanup_orphans (1934)
5789 - abort_recovery crashes MDS in b_eq (mds_unlink_orphan) (2584)
5790 - block all file creations until orphan recovery completes (1901)
5791 - client remove rq_connection from request struct (2423)
5792 - conf-sanity test_5, proper cleanup in umount log not availale (2640)
5793 - recovery timer race (2670)
5794 - mdc_close recovey bug (2532)
5795 - ptlrpc cleanup bug (2710)
5796 - mds timeout on local locks (2588)
5797 - namespace lock held during RPCs (2431)
5798 - handle interrupted sync write properly (2503)
5799 - don't try to handle a message that hasn't been replied to (2699)
5800 - client assert failure during cleanup after abort recovery (2701)
5801 - leak mdc device after failed mount (2712)
5802 - ptlrpc_check_set allows timedout requests to complete (2714)
5803 - wait for inflight reqs when ptlrpcd finishes (2710)
5804 - make sure unregistered services are removed from the srv_list
5805 - reset bulk XID's when resending them (caught by 1138 test)
5806 - unregister_bulk after timeout
5807 - fix lconf error (2694)
5808 - handle write after unfinished setstripe, stripe-only getstripe (2388)
5809 - readahead locks pages, leaves pending causing memory pressure (2673)
5810 - increase OST request buffers to 4096 on large machines (2729)
5811 - fix up permission of existing directories in simple_mkdir (2661)
5812 - init deleted item, add assertions ptlrpc_abort_inflight() (2725)
5813 - don't assign transno to errored transactions (2742)
5814 - don't delete objects on OST if given a bogus objid from MDS (2751)
5815 - handle large client PAGE_SIZE readdir on small PAGE_SIZE MDS (2777)
5816 - if rq_no_resend, then timeout request after recovery (2432)
5817 - fix MDS llog_logid record size, 64-bit array alignment (2733)
5818 - don't call usermode_helper from ptlrpcd, DEFAULT upcall (2773)
5819 - put magic in mount.lustre data, check for bad/NULL mount data (2529)
5820 - MDS recovery shouldn't delete objects that it has given out (2730)
5821 - if enqueue arrives after completion, don't clobber LVB (2819)
5822 - don't unlock pages twice when trigger_group_io returns error (2814)
5823 - don't deref NULL rq_repmsg if ldlm_handle_enqueue failed (2822)
5824 - don't write pages to disk if there was an error (1450)
5825 - don't ping imports that have recovery disabled (2676)
5826 - take buffered bytes into account when balancing socknal conn (2817)
5827 - hold a DLM lock over readdir always, use truncate_inode_pages (2706)
5828 - reconnect unlink llog connection after MDS reconnects to OST (2816)
5829 - remove little-endian swabbing of llog records (1987)
5830 - set/limit i_blksize to LL_MAX_BLKSIZE on client (2884)
5831 - retry reposting request buffers if they fail (1191)
5832 - grow extent at grant time to avoid granting a revoked lock (2809)
5833 - lock revoke doesn't evict page if covered by a second lock (2765)
5834 - disable VM readahead to avoid reading outside lock extents (2805)
5836 - return LL_SUPER_MAGIC from statfs for the filesystem type (1972)
5837 - updated kernel patches for hp-2.4.20 kernel (2681)
5839 2004-02-07 Cluster File Systems, Inc. <info@clusterfs.com>
5842 - fix truncated write corruption (2366)
5843 - fix for failed assertion in iopen_connect_dentry (1792,2517)
5845 - don't flag the ptlrpcd thread with PF_MEMALLOC (2636)
5846 - ensure len(uuid) < 37 in lmc (1171)
5847 - fix ia64 OOPS in llog_test (2255)
5848 - zero end of page at obdfilter for partial page writes (2648)
5849 - don't leave stale dentries around after renames (bug 2428)
5850 - fix timeouts when evicting a client with a single lock held (2642)
5851 - set deadline for the initial HELLO message to drain (2634)
5852 - print out dotted-quad IP addresses in the socknal (2302)
5854 - additional debugging for MDS client eviction problem (2443)
5855 - fix mkfsoptions support for osts (2603, 2604)
5857 2004-01-27 Cluster File Systems, Inc. <info@clusterfs.com>
5860 - add series for the vanilla 2.6.0 kernel
5861 - add series for the vanilla 2.4.24 kernel
5862 - add series for a cray x86/64 UL kernel drop
5863 - fix xattr patches for the vanilla 2.4.19 series
5865 - generate true UUIDs in lmc (1171)
5866 - have portals stack dumping break in UML (2466)
5867 - avoid bad dchild deref; avoid inum lock w/o creation (2362)
5868 - allocate with _NOFS in ldlm to avoid deadlock (1933)
5869 - wake callback waiting threads on client eviction (2460)
5870 - Add --ptldebug and --subsystem to lmc (1719)
5871 - update assertion to allow safe interrupt allocation
5872 - set rq_no_resend for cancel requests (2432)
5873 - recalculate ptlrpcd timeout after resend (2494)
5874 - call vfs_rmdir when removing pending directories (2368)
5875 - fix renaming a file to itself (2429)
5876 - lmc creates a default one-stripe lov (2454)
5877 - expand procfs space to handle large clusters (2326)
5878 - increase UML stack to avoid overflow
5879 - update lconf's list of debug and subsystem masks
5880 - fix lfs find --obd (2510)
5881 - /proc tunable for disabling filter read caching (2591)
5882 - stop rpm packages from altering slapd.conf (2301)
5883 - disable nagle in the socknal under 0conf (2578)
5884 - choose mds inode size based on stripe count (2572)
5885 - fix kernel-source rpm problems (2516)
5887 - add --disable-doc to avoid pdf generation (2421)
5888 - update documentation, tests, type-os, comments
5889 - avoid format warnings on ia64
5890 - remove the TOE NAL
5891 - tiny code cleanups by removing unused fields
5893 2004-01-07 Cluster File Systems, Inc. <info@clusterfs.com>
5896 - fix obvious semaphore misuse in as-yet-unused setattr path (2348)
5897 - remove the most blatant lies from BUILDING file (2371)
5898 - change default debug level to reasonable production setting
5899 - reduce client side cache size to reduce cache flush time
5900 - reduce max RPCs in flight to avoid unnecessary file fragmentation
5901 - make TCP zerocopy and pinger support enabled by default (2476)
5902 - sync writes completed after process exits caused crashes (2319)
5903 - maintain correct mount count on the MDS (2356)
5904 - backout 1557, because 2316 wasn't really fixed
5905 - better file I/O statistics gathering in /proc
5906 - don't take unnecessary, deadlock-inducing bug in readpage (2383)
5907 - another kernel patch to fix zero-copy TCP function export
5908 - don't take duplicate lock when processing re-sent getattr (2420)
5909 - lctl uses obd_self_export instead of creating new conn (2353)
5910 - MDS/OST recovery case which requires object creation asserted (2425)
5911 - move lfs from /usr/sbin to /usr/bin in packages
5912 - fix race between mds_client_add and mds_client_free (2417)
5913 - use kmalloc instead of slabs in portals (2430)
5914 - don't create duplicate records when a failover MDS is present (2442)
5915 - remove unnecessary mount age check (2332)
5916 - don't remove directory inodes from locks prematurely (2451)
5917 - don't break if MDS service name is the same as hostname (2103)
5918 - fix races in client write RPC generation when cache full (2482)
5920 2003-12-13 Cluster File Systems, Inc. <info@clusterfs.com>
5923 - remove now-unused request->rq_obd (278)
5924 - if an allocation fails, print out how much memory we've used (1933)
5925 - use PORTAL_SLAB_ALLOC for structures, to get GFP_MEMALLOC (1933)
5926 - add the "configurable stack size" patch to most series files (1256)
5927 - ability to write large log records, for 100+ OST configs (2306)
5928 - fix NULL deref when filter_prep fails (2314)
5929 - fix operator precedence error in filter_sync
5930 - dynamic allocation of socknal TX descriptors (2315)
5931 - fix a missed case in the GFP_MEMALLOC patch, can cause deadlock (2310)
5932 - fix gcc 2.96 compilation problem in xattr kernel patch (2294)
5933 - ensure that CWARN messages in Portals always get to the syslog
5934 - __init/__exit are not for prototype decls (ldlm_init/exit)
5935 - x86-64 compile warning fixes
5936 - fix gateway LMC keyword conflict (2318)
5937 - fix MDS lock inversions in getattr/reint paths (1844)
5938 - fix a rare lock re-ordering bug, which caused deadlock (2322)
5939 - fix i_sem/journal inversion in fsfilt_ext3_write_record (2306)
5940 - DLM race condition prevented some lock evictions (2328)
5941 - ENOMEM detection and retry on socknal sends (2230)
5942 - use GFP_NOFS throughout Lustre, to combat ENOMEM (2230)
5943 - move osc_rpcd into ptlrpc, for use in MDC and others (2329)
5944 - protect MDS inode fsdata with stronger locking; fixes assertion (2313)
5945 - better error messages when a client is rejected during recovery (1505)
5946 - avoid cancelling locks which were never granted, after failure (2330)
5947 - fix i_sem/journal inversion in mds_client_add (2333)
5948 - fix truncate/getattr lock cycle deadlock (2334)
5949 - use rpcd to send close; allows resend after timeout, avoid leak (1897)
5950 - fix two rare exit paths which could leak an l_lock() ref (2321)
5951 - fencepost error in MDS/OST orphan recovery (2226)
5952 - make log record alignment 8 bytes (1988)
5953 - lstripe now fails when requested offset > ost_count (2237)
5954 - ensure that all kernel series have a complete list.h (1607)
5955 - fix crashes in special-file operations (2316)
5956 - lctl create/brw OID mismatch, caused by obsolete filter loop (2339)
5958 - allow configurable automake binary, for testing new versions
5959 - small update to the lfs documentation
5961 2003-12-03 Cluster File Systems, Inc. <info@clusterfs.com>
5963 * fix negative export reference count in fsfilt_sync (2312)
5965 2003-12-01 Cluster File Systems, Inc. <info@clusterfs.com>
5966 * release candidate 0.9.1
5968 - orphans are moved into the PENDING directory for possible recovery
5969 - replayed opens now open by fid for orphan/rename safety (1042)
5970 - last close of an orphan inode generates a transno (683)
5971 - chdir() and mount() now pin the directory entry (1020)
5972 - avoid CERROR in normal ll_setattr_raw() error case (1500)
5973 - discard very old requests without processing them (1502)
5974 - remove some common, well-understood CERRORs (1505)
5975 - require O_DIRECT I/O to be page-sized to workaround IA64 crash (1609)
5976 - clear "grant" flags in OST replies until OST grant code lands (1644)
5977 - fix read performance by not clobbering i_blksize on client (1598)
5978 - fix __ldlm_handle2lock oops by not dereferencing lock after PUT (1625)
5979 - make LRU size a /proc tunable, clears locks when reduced (707)
5980 - fix some lprocfs rot that prevented ptlbd from loading (1732)
5981 - server locks take references on exports now (1558)
5982 - build fixes for 2.4.20-rh trees (1663)
5983 - return an error from lov_create if all OSCs are inactive (1751)
5984 - fix import levels when a reconnect happens without a timeout (1597)
5985 - exit early from mds_open if we get a lookup error (1749)
5986 - partial page read at EOF wouldn't wait for disk before sending (1642)
5987 - avoid NULL deref in obdfilter when reading page past EOF (1592)
5988 - avoid LASSERT in ll_intent_lock if server failed very early (1090)
5989 - fix LBUG in ll_it_open_error with rc = -2 (1861)
5990 - write/truncate lock inversion (1639)
5991 - Don't auto-load obdclass, portals modules during cleanup (1495)
5992 - fix timestamps from jumping to "now" (1763)
5993 - extra journal assertions (1648)
5994 - add an extra multiunlink test (1771)
5995 - fix read_record/write_record API (1776)
5996 - fix leak of offset_extent, possible incorrect i_size later (1772)
5997 - fix lasserts in mis-matched transnos during open-unlink testing (1541)
5998 - Debugging for the kqswnal_get_idle_tx problems (1820)
5999 - Allow recovery to be attempted multiple times (1536)
6000 - Write out MDS last_rcvd file after it is first created (1600)
6001 - Fix tx_descriptor leak in failed transmit situations (1827)
6002 - ext3 journaling fixes for assertion failure after IO error (1871)
6003 - class_export_put() on freed export after completion AST error (1896)
6004 - Fix revalidate looping in VFS (1322)
6005 - Don't access a freed export during MDS_REINT timeout (1521)
6006 - Add open-unlink recovery support on the MDS (1673,1764)
6007 - Return an error if no MDS data was read from last_rcvd (1946)
6008 - Fix for lookup "." or ".." crash on error (1932,1931,1935)
6009 - Don't setup a disk device that doesn't match exported UUID (317)
6010 - Reduce bulk RPC timeout to avoid cascading client/OST failures (1845)
6011 - avoid committing NULL handle in force close
6012 - local.sh is now a one-stripe LOV configuration
6013 - POSIX utime.4 -EPERM on FIFO not owned by user (56)
6014 - fix ext3 htree duplicate directory entry corruption (1516)
6015 - POSIX creat.13, fstat.1, open.18, stat.3 new file atime/mtime (2020)
6016 - update to new LOV EA format (2097)
6017 - interoperability for different PAGE_SIZE/wordsize (686,1821,1343,2042)
6019 2003-06-15 Phil Schwan <phil@clusterfs.com>
6022 - imports and exports cleanup too early, need refcounts (349, 879, 1045)
6023 - per-import/export recovery handling (958, 931, 959)
6024 - multiple last-rcvd slots, for serving multiple FSes (949)
6025 - connections are again shared between multiple imp/exports (963, 964)
6026 - "umount -f" would hang if any requests needed to be sent (393, 978)
6027 - avoid pinning large req buffer by copying for queued messages (989)
6028 - add "uuid" to "lctl device" command to help upcalls (991)
6029 - "open" RPCs with transnos would confuse recovery counters (1037)
6030 - do proper endian conversion of all wire messages (288, 340, 891)
6031 - remove OST bulk get LBUGs, fix ost_brw_write cleanup (1126)
6032 - call waiting locks callback from LDLM recovery thread (1127, 1151)
6033 - fix ptlrpc_connection leak in target_handle_connect (1174)
6034 - fix import refcounting bug in OST and MDS cleanup (1134)
6035 - if an invalid-at-open-time OSC returned before close(), LBUG (1150)
6036 - fix very unlikely obd_types race condition (501)
6037 - remove osc_open hack for echo_client (1187)
6038 - we leaked exports/dlmimps for forcibly disconnected clients (1143)
6039 - a failure in read_inode2 leads to deadlock (1139)
6040 - cancel ack-locks as soon as transaction is committed (1072)
6041 - fix major leaks and crashes in the bulk I/O path (937, 1057)
6042 - make sure to commitrw after any preprw to avoid deadlock (1162)
6043 - failing to execute a file in a lustre FS would lock inode (1203)
6044 - small DEBUG_REQ fix to avoid dereferencing a NULL (1227)
6045 - don't ASSERT while cleaning up an incompletely-setup obd (1248)
6046 - obd_uuid2tgt would walk off the end of the list (1255)
6047 - on IA64 the osc would give portals incorrect bulk size (1258)
6048 - fix debug daemon ioctl interface; allows daemon on ia64 (1274)
6049 - fix lock inversion caused by new llite matching code (1282)
6050 - limit the number of dirty pages on a client to 10MB (1286)
6051 - timed out locks were not being corrected cancelled (1289)
6052 - fix O_DIRECT above 4GB on IA-32 (1292)
6053 * major user-visible changes
6054 - fail out/fail over policy now controlled by the upcall (993)
6056 - add OBD_PING to check server availability and failure (954)
6057 - lustre messages are now sent in sending host order (288, 340, 891)
6058 - add eadatalen to MDS getattr reply (340)
6059 - OST read replies may contain second buffer, with per-page status (593)
6061 2003-03-11 Phil Schwan <phil@clusterfs.com>
6064 - LDLM_DEBUG macro fix, for gcc 3.2 (850)
6065 - failed open()s could cause deadlock; fixed (867, 869)
6066 - stop cancelling OST locks when files are closed (481)
6067 - overlapping XID spaces caused network corruption (851, 853)
6068 - fix unsafe fsfilt counter arithmetic; change to atomic_t
6069 - setattr_raw added, to do single-RPC, server-side setattrs
6070 - lmc/lconf syntax change for OST UUIDs
6071 - fix crashy race condition between ptlrpc_free_req and osc_close
6072 - don't use request in mdc_enqueue if we hit a timeout (889)
6073 - don't set the inode i_size for regular files from the MDS (896)
6074 - handle out of order completion AST (842)
6075 - don't LBUG if a lock request times out after receiving AST (913)
6076 - avoid d_rehash race in ll_find_alias by rehashing inside dcache_lock
6077 - if a bad lock AST arrives, send an error instead of dropping entirely
6078 - return 0 from revalidate2 if ll_intent_lock returns -EINTR (912)
6079 - fix leak in bulk IO when only partially completed (899, 900, 926)
6080 - fix O_DIRECT for ia64 (55)
6081 - (almost) eliminate Lustre-kernel-thread effects on load average (722)
6082 - C-z after timeout could hang a process forever; fixed (977)
6084 - client-side I/O cache (678, 924, 929, 941, 970)
6086 - READPAGE and SETATTRs which don't take server-side locks get
6089 2003-02-11 Phil Schwan <phil@clusterfs.com>
6092 - Fix ldlm_lock_match on the MDS to avoid matching remote locks (592)
6093 - Fix fsfilt_extN_readpage() to read a full page of directory
6094 entries, or fake the remainder if PAGE_SIZE != blocksize (500)
6095 - Avoid extra mdc_getattr() in ll_intent_lock when possible (534, 604)
6096 - Fix imbalanced LOV object allocation and out-of-bound access (469)
6097 - Most intent operations were removed, in favour of a new RPC mode
6098 that does a single RPC to the server and bypasses most of the VFS
6099 - All LDLM resource ID arrays were removed in favour of ldlm_res_id
6100 - Aggressively cancel local locks on DLM servers
6101 - mds_reint_unlink sends EA to the client if it's the last nlink.
6102 client uses that EA to unlink OST objects.
6103 - mds_reint_{rename,unlink,link} were rewritten to take ordered locks
6104 - recursive symlinks were fixed (439)
6105 - fixed NULL deref in DEBUG_REQ
6106 - filter_update_lastobjid no longer calls sync, which annoyed extN
6107 - fixed multi-client small-writes to a single file problem (445)
6108 - fixed mtime updates during file writes (607)
6109 - fixed vector writes on obdfilter causing problems when ENOSPC (670)
6110 - fixed bug in obd_brw_read/write() (under guise of testing 367)
6111 - fixed Linux OST size reporting problem (444, 656)
6112 - OST now updates object mtime with writes or setattr (607, 619)
6113 - client verifies file size before zeroing page past EOF (445)
6114 - OST now writes last allocated objid to disk with allocation (108)
6115 - LOV on echo now works (409)
6117 - mds_reint_unlink sends a new buffer, with the EA included. this
6118 buffer is only valid if body->valid & OBD_MD_FLEASIZE, which is only
6119 set if a regular file was being unlinked, and it was the last link
6120 - use PtlGet from the target for bulk writes (315)
6121 - OST now updates object mtime with writes or setattr (607, 619)
6122 - LDLM now has a grant-time callback to revalidate locked items, if
6124 - Many MDS operations were reorganized to combat race conditions
6126 - Merge b_intel branch (updated lprocfs code) - now at /proc/fs/lustre
6127 - configure check to avoid gcc version 2.96 20000731-2.96-98) (606)
6129 2003-01-06 Andreas Dilger <adilger@clusterfs.com>
6132 - Fully reactivate OST imports after reconnection (512, others)
6133 - Make sure client sees our -ENOTCONN from mds_handle (513 - partial)
6134 - More graceful error handling for truncating on dead OST (515)
6135 - Don't error out unless we're actually accessing dead stripes (474)
6136 - Fix garbage sizes when stripes are missing (410)
6137 - LRU counters were broken, causing constant lock purge (433, 432)
6138 - garbage on read from stripes with failed OSTs (441)
6139 - mark OSCs as active before reconnecting during recovery (438)
6140 - lov_enqueue and lov_cancel need to handle inactive OSTs (403)
6141 - lfind did not preserve OST order in output (443)
6142 - symlinks cause hung clients, incorrect data (439)
6143 - stop dereferencing request after dropping refcount (457)
6144 - don't LASSERT(spin_is_locked) on non-SMP (455)
6145 - fixes for many rename() bugs
6146 - fstat didn't correctly synchronize attributes (399)
6147 - server must handle lock cancellation during blocking AST prep (487)
6148 - bulk descriptors were free()d too soon (511)
6149 - fix paths in lconf, which would load incorrect modules (451, 507)
6150 - fix confusing lconf 'host not found' error message (386)
6151 - fix lock order deadlock on OST (O/R i_sem before journal ops, 478)
6152 - fix race condition in mdc_blocking_ast() for inode access (526)
6153 - fix lov_unpackmd() unpacking wrong number of stripes (537)
6154 - fix lov_set_osc_active() marking wrong OSC inactive (440)
6155 - fix bad lstripe lov_unpackmd() assertion (fix layering too) (527)
6156 - fix multiple writes of stripe MD to MDS (358, maybe 519)
6157 - fix lstripe in several ways (kernel side) (527)
6158 - fix request leak in ldlm_cli_enqueue (262)
6159 - incorrect OSC was marked inactive after OST failure
6160 - call mds_fs_cleanup before unmounting filesystem (524)
6161 - fix races between taking ns_lock and ldlm_lock_change_resource
6162 - fix races updating LOV export open file list
6163 - fix lov_enqueue error path, avoid decref-ing bad lock handle
6164 - fix recovery NULL deref in ldlm_cli_cancel_unused
6165 - fix some DLM races by using new hash table for lock handles (419)
6166 - permit the client to specify desired inodes, at replay
6167 - duplicate requests when we queue them for replay reintegration
6168 - fix last_rcvd offset calculation
6169 - sync after each recovered transaction, so we always make progress
6170 - never, not always, ERESTART requests without transnos
6171 - store the lov_desc in the MDS, so we don't depend on getlovinfo to
6173 - skip replay if the MDS says that the client is already connected
6174 - don't check for a recovery-enabled export to match lctl's UUID
6175 - don't INC_USE_COUNT for phantom exports
6176 - don't crash when cleaning up phantom exports (567)
6177 - don't double-finish or set replay data for errored mdc_open requests
6178 - abort requests when they time out, so we don't get old replies
6179 - send/receive replies for AST messages again
6180 - if the client says that it doesn't have the lock, cancel it on the
6182 - if we timeout during I/O, don't try to cancel an in-use lock; instead
6183 mark it as destroyed, it will all work out when decref is called
6184 - fix module use counts (22, 581)
6186 - ASTs now expect a reply (server cancels lock on error reply)
6188 2002-12-02 Andreas Dilger <adilger@clusterfs.com>
6191 - fix many simultaneous client startup (392)
6192 - fix dentry->d_it clobbering
6193 - credentials weren't being shipped for readdir/getattr operations
6194 - remove invalid assertions triggered during some concurrent MD
6196 - proper Lustre versions added (336, 389)
6197 - fix memory leak for create error case (398)
6198 - fix LOV locking bug that would get cli/srv out of sync
6199 - fix echo client over LOV (409)
6200 - fix dbench 2, extN refcount problem (170, 258, 356, 418)
6201 - fix double-O_EXCL intent crash (424)
6202 - avoid sending multiple lock CANCELs (352)
6204 - MDS can do multi-client recovery (modulo bugs in new code)
6206 - many updates, edits, cleanups
6208 2002-11-18 Phil Schwan <phil@clusterfs.com>
6211 - fix null d_it dereference (346)
6212 - fix full OST/dbench hang (333)
6213 - fix permission problem with file removal (286)
6214 - fix removal of OSCs from LOV when they fail
6215 - fix NULL deref during bulk timeout (214)
6216 - fix problems related to multiple filesystems on one MDS (241)
6217 - fixed serious subtle metadata locking bugs
6218 - free locks on clients when inodes are removed due to memory
6220 - fix inode pointer in lock data (285)
6221 - partial support for multiple MDS on a single host (241)
6222 - data locks weren't cancelled at clear_inode time (290, 311)
6223 - intent locks could lead to unbounded lock growth (205)
6224 - added a maximum lock count, an LRU list, and a flusher
6225 - fix multiple rename (365)
6226 - properly abstracted the echo client
6227 - OSC locked 1 byte too many; fixed
6228 - rewrote brw callback code:
6229 - fixed recovery bugs related to LOVs (306)
6230 - fixed too-many-pages-in-one-write crash (191)
6231 - fixed (again) crash in sync_io_timeout (214)
6232 - probably fixed callback-related race (385)
6234 - Add capability to MDS protocol
6235 - LDLM cancellations and callbacks on different portals
6237 2002-10-28 Andreas Dilger <adilger@clusterfs.com>
6240 - limit client IOV size to PTL_MD_MAX_IOV (611336, 191)
6241 - defer open object destruction to close time (601981, 138)
6242 - open/close OST file handle in obdo (OBD_MD_FLHANDLE) (601981, 138)
6243 - move LDLM_ENQUEUE/CONVERT back to MDS portal (625069)
6244 - abstract ll_lookup2, fix ll_revalidate2 to use abstraction (256)
6245 - don't call obd_setattr in ll_file_release for destroyed objects
6246 * protocol change to lustre_msg: move |version| and add |flags|
6247 * protocol change to osc_punch: "start" in "o_size", "end" in "o_blocks"
6248 * lock replay: for LDLM_FL_REPLAY trust client to do right thing
6249 * added replay of create, unlink, link and rename operations during
6250 MDS failover; recovery should be much more robust now
6251 * remove failed OSCs from LOVs (only lov_create uses this so far)
6252 * the lustre-HOWTO was brought (more) up to date (582544)
6254 2002-10-23 Phil Schwan <phil@clusterfs.com>
6257 - in-use dentries weren't being reused properly (617851)
6258 - prevent multiple LDLM setup (599178)
6259 - fix LOV size calculations for truncate (617853)
6260 - fix client handling of MDS intent errors (POSIX)
6261 - fix permission bug in lovstripe.c test (624321)
6262 - fix MDS thread deadlock - move LDLM handler to DLM portal (625069)
6263 - truncate past end of file could corrupt data
6264 - proper cleanup after timeouts, crashes, etc (592524, 550815)
6265 - a race in recovery could return ETIMEDOUT to apps (623947)
6266 - building outside the source directory was fixed
6267 * the lustre-HOWTO was brought (more) up to date (582544)
6268 * major progress was made on recovery functionality
6270 2002-10-10 Phil Schwan <phil@clusterfs.com>
6273 - recovery deadlock fix
6274 - rm -rf causes LBUG fix (617817)
6275 - file open by multiple tasks fix (618962)
6276 - directory permissions bugs (602707 and 620007)
6277 - journal_stop fixed with locking (611313)
6278 - O_APPEND failures resolved (618273, perhaps 614459)
6279 - lconf PATH fix (619770)
6280 - IA64 build fix (621450)
6281 - RPC buffer sizes scale with amount of memory
6283 2002-10-01 Phil Schwan <phil@clusterfs.com>
6286 - locks would be cancelled without throwing away data pages,
6287 resulting in inconsistent data (605627)
6288 - inode attributes were not always being refreshed (605627, 612449)
6289 - lconf now continues to cleanup after lctl reports an error
6290 - MDS now enforces user permissions (602707)
6291 - lprocfs cleanup fixed, but not yet enabled (614157)
6292 - fixed infinite server hang, should a client not respond to an AST
6293 - avoid going into recovery if user calls readlink() with a buffer
6294 that's too small (613941)
6295 - AST RPCs no longer require replies (614867) -- this may be changed
6296 - don't crash server if client sends an IOV that's too big (611336)
6297 - fixed lock conversion deadlock (611892)
6298 - fixed the following of symlinks (614622)
6299 * recovery: the server can remove locks from a client that dies, other
6300 clients can make progress
6301 * more extN patch fixes
6302 * compile-time configurable ptlrpc buffer allocations
6304 - collaborative read cache document
6305 - Lustre Lite Performance CDR document-in-progress
6307 2002-09-20 Andreas Dilger <adilger@clusterfs.com>
6310 - fix typo in patch-2.4.18
6312 2002-09-20 Andreas Dilger <adilger@clusterfs.com>
6315 - clear ptlrpc request each time in handle_incoming_request()
6316 - unlink of files now destroys the object on the OST
6318 2002-09-19 Peter Braam <braam@clusterfs.com>
6320 * add hard link support
6321 * change obdfile creation method
6322 * kernel patch changed
6324 2002-09-19 Peter Braam <braam@clusterfs.com>
6327 - stack overflow bug in extN fixed
6329 2002-09-18 Andreas Dilger <adilger@clusterfs.com>
6331 * documentation updates
6332 - add man pages for config tools
6333 - update tests/README to describe testing with new config tools
6334 - finish metadata API descriptions
6335 * bug fixes and cleanups
6336 - statfs workaround for 16TB limit
6337 - LOV stripe allocation improved, can stripe on subset of OSTs
6338 - LOV file size/IO offset was wrong for files > 4GB in size
6339 - object EA data was being dropped, caused files to be unreadable
6340 - memory overflow with non-LOV OST caused memory corruption
6341 - fixed regression tests to work with new config tools, obdfilter
6342 - fixed bug when directory size became larger than 1 block
6343 - fixed bug (for single client case) when PWD was deleted
6344 - invalidate local directory pages when doing intent-based ops
6345 - avoid LDLM oops when lock callback contained bad data
6347 2002-09-09 Andreas Dilger <adilger@clusterfs.com>
6349 * documentation updates
6350 * bug fixes and cleanups
6351 - configuration tools
6354 - 64-bit compile warnings
6355 - 64-bit internal statfs data
6357 * test_brw on persistent OST devices
6359 * lprocfs (disabled)
6361 2002-09-04 Andreas Dilger <adilger@clusterfs.com>
6363 * documentation updates
6364 * bug fixes and cleanups
6365 * configuration tools
6367 2002-08-30 Peter J. Braam <braam@clusterfs.com>
6370 * many small fixes to 0_5_4
6371 * io/network handling
6372 * thinkos in MDS operations
6374 2002-08-24 Peter J. Braam <braam@clusterfs.com>
6377 * crucial basic fixes to 0.5.3
6378 * IOR, Iozone work over Elan
6381 2002-08-07 Phil Schwan <phil@clusterfs.com>
6382 * version 0_5_3, our first alpha
6383 * we use the new Portals iovs
6384 * documentation updates
6385 * bug fixes and cleanups
6386 * small changes in the DLM wire protocol
6388 2002-07-25 Peter J. Braam <braam@clusterfs.com>
6389 * version 0_5_1 with some initial stability,
6390 * locking on MD and file I/O.
6391 * documentation updates
6392 * several bug fixes since 0.5.0
6393 * small changes in wire protocol
6395 2002-07-18 Phil Schwan <phil@clusterfs.com>
6397 * delivered as Lustre Light Alpha
6398 * fixed a crash after handling invalid MDS requests
6399 * fixed directory pages for architectures with non-4k pages sizes
6401 2002-07-11 Andreas Dilger <adilger@clusterfs.com>
6402 * release version v0_4_4
6403 * Moves TCP acceptor to be on port 2432 (unused Coda port) instead
6405 * Fixes a number of interruption problems with OST operations.
6406 * Update documentation for portals header changes
6407 * Move all wire protocol structs/defines to lustre_idl.h
6408 * Fixes symlink length bug.
6409 * Add tcpdump to repository.
6411 2002-07-05 Andreas Dilger <adilger@clusterfs.com>
6412 * release version v0_4_3
6413 * Fixes statfs for inodes on extN.
6414 * Fixes bug in runtests which would delete /etc/hosts.
6415 * Use 64-bit object IDs wherever possible (not into VFS though)
6416 Remove ost_get_info, which is unused by lustre, and out of date.
6418 2002-07-03 Peter Braam <braam@clusterfs.com>
6419 * release version v0_4_2 Fixes a lookup error (type not passed)
6420 * move forward to head of Portals
6421 * move forward to latest Lustre kernel
6423 2002-06-25 Peter Braam <braam@clusterfs.com>
6424 * release version v0_4_1. Hopefully stable on single node use.