1 tbd Sun Microsystems, Inc.
5 2.6.9-67.0.22.EL (RHEL 4),
6 2.6.16.60-0.33 (SLES 10),
7 2.6.18-128.1.1.el5 (RHEL 5),
8 2.6.22.14 vanilla (kernel.org)
9 * Client support for unpatched kernels:
10 (see http://wiki.lustre.org/index.php?title=Patchless_Client)
11 2.6.16 - 2.6.22 vanilla (kernel.org)
12 * Client support for unpatched kernels:
13 we do not recommend using patchless RHEL4 clients with kernels
14 prior to 2.6.9-55EL (RHEL4U5).
15 * Recommended e2fsprogs version: 1.40.11-sun1
16 * Note that reiserfs quotas are disabled on SLES 10 in this kernel.
17 * RHEL 4 and RHEL 5/SLES 10 clients behaves differently on 'cd' to a
18 removed cwd "./" (refer to Bugzilla 14399).
19 * A new quota file format has been introduced in 1.6.5.
20 The format conversion from prior releases is handled transparently,
21 but releases older than 1.4.12/1.6.5 don't understand this new
22 format. The automatic format conversion can be avoided by running
23 the following command on the MDS:
24 'tunefs.lustre --param="mdt.quota_type=ug1" $MDTDEV'.
25 For more information, please refer to bugzilla 13904.
26 * Output of lfs quota has been made less detailed by default,
27 old (verbose) output can be obtained by using -v option.
28 * File join has been disabled in this release, refer to Bugzilla 16929.
29 * A new Lustre ADIO driver is available for MPICH2-1.0.7.
30 * NFS export disabled when stack size < 8192. Since the NFSv4 export
31 of Lustre filesystem with 4K stack may cause a stack overflow. For
32 more information, please refer to bugzilla 17630.
36 Description: free_entry erroneously used groups_free instead of put_group_info
38 Severity : enhancement
40 Description: Make read-ahead stripe size aligned.
42 Severity : enhancement
44 Description: MDS create should not wait for statfs RPC while holding DLM lock.
49 Descriptoin: OST grant too much space to client even there are not enough space.
50 Details : Client will shrink its grant cache to OST if there are no write
51 activity over 6 mins (GRANT_SHRINK_INTERVAL), and OST will retrieve
52 this grant cache if there are already not enough avaible space
53 (left_space < total_clients * 32M).
55 Severity : enhancement
57 Description: Update to RHEL5U3 kernel-2.6.18-128.1.1.el5.
59 Severity : enhancement
61 Description: Update client support to vanila kernels up to 2.6.27.
64 Frequency : rare, connect and disconnect target at same time
66 Descriptoin: ASSERTION(atomic_read(&imp->imp_inflight) == 0
67 Details : don't call obd_disconnect under lov_lock. this long time
68 operation and can block ptlrpcd which answer to connect request.
70 Severity : enhancement
72 Description: Update to SLES10 kernel-2.6.16.60-0.33.
74 Severity : enhancement
76 Description: Update to sles9 kernel-2.6.5-7.315.
79 Frequency : start MDS on uncleanly shutdowned MDS device
81 Descriptoin: ll_sync thread stay in waiting mds<>ost recovery finished
82 Details : stay in waiting mds<>ost recovery finished produce random bugs
83 due race between two ll_sync thread for one lov target. send
84 ACTIVATE event only if connect realy finished and import have
87 Severity : enhancement
89 Description: Update to RHEL5 kernel-2.6.18-92.1.22.el5.
92 Frequency : start MDS on uncleanly shutdowned MDS device
94 Descriptoin: aborting recovery hang on MDS
95 Details : don't throttle destroy RPCs for the MDT.
97 Severity : enhancement
99 Descriptoin: Don't sync journal after every i/o
100 Details : Implement write RPC replay to allow server replies for write RPCs
101 before data is on disk.
105 Description: Slow reads beyond 8Tb offsets.
106 Details : Page index integer overflow in ll_read_ahead_page
110 Description: Soft lockup on OSS after MDS failover
111 Details : MSG_CONNECT_INITIAL is not set on the initial MDS->OST connect.
112 As a conseqence, the patch from bug 18224 is not operational
113 and the MDS export cannot be reused on the OSTs until it gets
117 Frequency : rare, only if using MMP with Linux RAID
119 Description: MMP doesn't work with Linux RAID
120 Details : While using HA for Lustre servers with Linux RAID, it is possible
121 that MMP will not detect multiple mounts. To make this work we
122 need to unplug the device queue in RAID when the MMP block is being
123 written. Also while reading the MMP block, we should read it from
124 disk and not the cached one.
127 Frequency : rare, during recovery
129 Description: Assertion failure in ldlm_lock_put
130 Details : Do not put cancelled locks into replay list, hold references on
134 -------------------------------------------------------------------------------
135 2008-12-31 Sun Microsystems, Inc.
137 * Support for kernels:
138 2.6.5-7.314 (SLES 9),
139 2.6.9-67.0.22.EL (RHEL 4),
140 2.6.16.60-0.31 (SLES 10),
141 2.6.18-92.1.17.el5 (RHEL 5),
142 2.6.22.14 vanilla (kernel.org)
143 * Client support for unpatched kernels:
144 (see http://wiki.lustre.org/index.php?title=Patchless_Client)
145 2.6.16 - 2.6.22 vanilla (kernel.org)
146 * Client support for unpatched kernels:
147 we do not recommend using patchless RHEL4 clients with kernels
148 prior to 2.6.9-55EL (RHEL4U5).
149 * Recommended e2fsprogs version: 1.40.11-sun1
150 * Note that reiserfs quotas are disabled on SLES 10 in this kernel.
151 * RHEL 4 and RHEL 5/SLES 10 clients behaves differently on 'cd' to a
152 removed cwd "./" (refer to Bugzilla 14399).
153 * A new quota file format has been introduced in 1.6.5.
154 The format conversion from prior releases is handled transparently,
155 but releases older than 1.4.12/1.6.5 don't understand this new
156 format. The automatic format conversion can be avoided by running
157 the following command on the MDS:
158 'tunefs.lustre --param="mdt.quota_type=ug1" $MDTDEV'.
159 For more information, please refer to bugzilla 13904.
160 * A new quota file format was introduced in 1.6.6/1.8.0 (kernels 2.6.16+).
161 The format conversion from prior releases is handled transparently,
162 but releases older than 1.6.6/1.8.0 don't understand this new
163 format. The automatic format conversion can be avoided by running
164 the following commands on the MDS and OSS servers (for
165 pre 1.4.12-1.6.5 quota files):
166 'tunefs.lustre --param="mdt.quota_type=ug1" $MDTDEV',
167 'tunefs.lustre --param="ost.quota_type=ug1" $MDTDEV'
168 or (for 1.4.12/1.6.5 quota files)
169 'tunefs.lustre --param="mdt.quota_type=ug2" $MDTDEV',
170 'tunefs.lustre --param="ost.quota_type=ug2" $MDTDEV'
171 For more information, please refer to bugzilla 13904.
172 * Output of lfs quota has been made less detailed by default,
173 old (verbose) output can be obtained by using -v option.
174 * File join has been disabled in this release, refer to Bugzilla 16929.
175 * A new Lustre ADIO driver is available for MPICH2-1.0.7.
176 * NFS export disabled when stack size < 8192. Since the NFSv4 export of
177 Lustre filesystem with 4K stack may cause a stack overflow. For more
178 information, please refer to bugzilla 17630.
183 Descriptoin: don't lose wakeup for imp_recovery_waitq
184 Details : recover_import_no_retry or invalidate_import and import_close can
185 both sleep on imp_recovery_waitq, but we was send only one wakeup
191 Descriptoin: panic in mds_open
192 Details : don't confuse mds_finish_transno() with PTR_ERR(-ENOENT)
197 Descriptoin: stuck in cache_remove_extent() or panic with accessing to already
199 Details : release lock refernce only after add page to pages list.
202 Frequency : always with long access acl
204 Descriptoin: mds can't pack reply with long acl.
205 Details : mds don't control size of acl but they limited by reint/getattr
210 Frequency : on remount
212 Description: external journal device not working after the remount
213 Details : clear dev_rdonly flag for external journal devices in
219 Description: shutdown vs evict race
220 Details : client_disconnect_export vs connect request race.
221 if client will evicted at this time - we start invalidate
222 thread without referece to import and import can be freed
228 Description: shrink LOV EAs before replying
229 Details : correctly adjust LOV EA buffer for reply.
234 Description: don't skip ost target if they assigned to file
235 Details : Drop slow OSCs if we can, but not for requested start idx.
236 This means "if OSC is slow and it is not the requested
237 start OST, then it can be skipped, otherwise skip it only
238 if it is inactive/recovering/out-of-space.
240 Severity : enhancement
242 Description: Update to RHEL5 kernel-2.6.18-92.1.17.el5.
245 Frequency : rare, need acl's on inode.
247 Description: client can't handle ost additional correctly
248 Details : if ost was added after client connected to mds client can have
249 hit lnet_try_match_md ... to big messages to wide striped files.
250 in this case need teach client to handle config events about add
251 lov target and update client max ea size at that event.
253 Severity : enhancement
255 Description: Update to sles9 kernel-2.6.5-7.314.
257 Severity : enhancement
259 Description: Update to SLES10 SP2 kernel-2.6.16.60-0.31.
262 Frequency : Create a symlink file with a very long name
264 Description: ldlm_cancel_pack()) ASSERTION(max >= dlm->lock_count + count)
265 Details : If there is no extra space in the request for early cancels,
266 ldlm_req_handles_avail() returns 0 instead of a negative value.
271 Description: mds is deadlocked
272 Details : in rare cases, inode in catalog can have i_no less than have parent
273 i_no, this produce wrong order for locking during open, and
274 parallel unlink can be lock open. this need teach mds_open to grab
275 locks in resource id order, not at parent -> child order.
277 Severity : enhancement
279 Description: Add /proc entry for import status
280 Details : The mdc, osc, and mgc import directories now have
281 an import directory that contains useful import data for debugging
284 Severity : enhancement
286 Description: Re-disable certain /proc logging
287 Details : Enable and disable client's offset_stats, extents_stats and
288 extents_stats_per_process stats logging on the fly.
291 Frequency : Only on FC kernels 2.6.22+
293 Description: oops in statahead
294 Details : Do not drop reference count for the dentry from VFS when lookup,
295 VFS will do that by itself.
297 Severity : enhancement
299 Description: Generic /proc file permissions
300 Details : Set /Proc file permissions in a more generic way to enable non-
301 root users operate on some /proc files.
305 Description: Hitting mdc_commit_close() ASSERTION
306 Details : Properly handle request reference release in
307 ll_release_openhandle().
311 Frequency : only patchless client
312 Description: add workaround for race between add/remove dentry from hash
314 Severity : enhancement
316 Description: Allow OST glimpses to return PW locks
320 Description: LBUG when llog conf file is full
321 Details : When llog bitmap is full, ENOSPC should be returned for plain
326 Description: Prevent import from entering FULL state when server in recovery
330 Description: service mount cannot take device name with ":"
331 Details : Only when device name contains ":/" will mount treat it as
337 Description: replace ptlrpcd with the statahead thread to interpret the async
338 statahead RPC callback
342 Frequency : on recovery
343 Description: I/O failures after umount during fail back
344 Details : if client reconnected to restarted server we need join to recovery
345 instead of find server handler is changed and process self eviction
346 with cancel all locks.
348 Severity : enhancement
350 Description: Update to RHEL5 kernel-2.6.18-92.1.10.el5.
354 Description: Kernel BUG tries to release flock
355 Details : Lustre does not destroy flock lock before last reference goes
356 away. So always drop flock locks when client is evicted and
357 perform unlock regardless of successfulness of speaking to MDS.
359 Severity : enhancement
361 Description: Update to SLES10 SP2 kernel-2.6.16.60-0.27.
363 Severity : enhancement
365 Description: Upcall on Lustre log has been dumped
366 Details : Allow for a user mode script to be called once a Lustre log has
367 been dumped. It passes the filename of the dumped log to the
368 script, the location of the script can be specified via
369 /proc/sys/lnet/debug_log_upcall.
374 Description: avoid messages about idr_remove called for id that is not allocated
375 Details : Move assigment s_dev for clustered nfs to end of initialization,
376 for avoid problem with error handling.
381 Description: avoid Already found the key in hash [CONN_UNUSED_HASH] messages
382 Details : When connection is reused this not moved from CONN_UNUSED_HASH into
383 CONN_USED_HASH and this prodice warning when put connection again
389 Description: avoid ASSERTION(client_stat->nid_exp_ref_count == 0) failed
390 Details : release reference to stats when client disconnected, not
391 when export destroyed for avoid races when client destroyed
392 after main ost export.
396 Description: more cleanup in mds_lov
397 Details : add workaround for get valid ost count for avoid warnings about
398 drop too big messages, not init llog cat under semphore which
399 can be blocked on reconnect and break normal replay, fix access
402 Severity : enhancement
404 Description: File striping can now be set to use an arbitrary pool of OSTs.
406 Severity : enhancement
408 Description: Export bytes_read/bytes_write count on OSC/OST.
412 Description: Early reply size mismatch, MGC loses connection
413 Details : Apply the MGS_CONNECT_SUPPORTED mask at reconnect time so
414 the connect flags are properly negotiated.
418 Description: Properly propagate oinfo flags from lov to osc for statfs
419 Details : restore missing copy oi_flags to lov requests.
423 Description: exports in /proc are broken
424 Details : recreate /proc entries for clients when they reconnect.
426 Severity : enhancement
428 Description: Add man pages for llobdstat(8), llstat(8), plot-llstat(8),
429 : l_getgroups(8), lst(8), routerstat(8)
430 Details : included man pages for llobdstat(8), llstat(8),
431 : plot-llstat(8), l_getgroups(8), lst(8), routerstat(8)
433 Severity : enhancement
435 Description: Implement lustre ll_show_options method.
437 Severity : enhancement
439 Description: Update to SLES9 kernel-2.6.5-7.312.
441 Severity : enhancement
443 Description: Update to RHEL4 kernel-2.6.9-67.0.22.EL.
447 Description: exports in /proc are broken
448 Details : recreate /proc entries for clients when they reconnect.
452 Description: don't fail open with -ERANGE
453 Details : if client connected until mds will be know about real ost count
454 get LOV EA can be fail because mds not allocate enougth buffer
459 Description: Resolve device initialization race
460 Details : Prevent proc handler from accessing devices added to the
461 obd_devs array but yet be intialized.
463 Severity : enhancement
465 Description: configure's --enable-quota should check the
466 : kernel .config for CONFIG_QUOTA
467 Details : configure is terminated if --enable-quota is passed but
468 : no quota support is in kernel
470 Severity : enhancement
472 Description: Update to SLES10 SP2 kernel-2.6.16.60-0.23.
474 Severity : enhancement
476 Description: Update to RHEL5 kernel-2.6.18-92.1.6.el5.
480 Frequency : rare, on PPC clients
481 Description: don't swab ost objects in response about directory, because
483 Details : bug similar bug 14856, but in different function.
485 Severity : enhancement
487 Description: lfs quota tool enhancement
488 Details : added units specifiers support for setquota, default to
489 current uid/gid for quota report, short quota stats by
490 default, nonpositional parameters for setquota, added
491 llapi_quotactl manual page.
493 Severity : enhancement
495 Description: *optional* service tags registration
496 Details : if the "service tags" package is installed on a Lustre node
497 When the filesystem is mounted, a local-node service tag will
498 be created. See http://inventory.sun.com/ for more information
499 about the Service Tags asset management system.
503 Description: Client runs out of low memory
504 Details : Consider only lowmem when counting initial number of llap pages
507 Description: Kernel BUG tries to release flock
508 Details : Lustre does not destroy flock lock before last reference goes
509 away. So always drop flock locks when client is evicted and
510 perform unlock regardless of successfulness of speaking to MDS.
513 Frequency : occasional
515 Description: add refcount for osc callbacks, so avoid panic on shutdown
517 Severity : enhancement
519 Description: Update to RHEL4 kernel-2.6.9-67.0.20.
522 Frequency : testing only
524 Description: sanity test 65a fails if stripecount of -1 is set
525 Details : handle -1 striping on filesystem in ll_dirstripe_verify
528 Frequency : only in unusual configurations
530 Description: Kernel panic with find ost index.
531 Details : lov_obd have panic if some OST's have sparse indexes.
533 Severity : enhancement
535 Description: Update to RHEL5 kernel-2.6.18-53.1.21.el5.
538 Frequency : rarely, if filesystem is mounted with -o flock
540 Description: do not process already freed flock
541 Details : flock can possibly be freed by another thread before it reaches
542 to ldlm_flock_completion_ast.
545 Frequency : rarely, if filesystem is mounted with -o flock
547 Description: LBUG during stress test
548 Details : Need properly lock accesses the flock deadlock detection list.
551 Frequency : rarely, if binaries are being run from Lustre
553 Description: oops in page fault handler
554 Details : kernel page fault handler can return two special 'pages' in
555 error case, don't try dereference NOPAGE_SIGBUS and NOPAGE_OMM.
558 Frequency : rarely, during shutdown
560 Description: timeout with invalidate import.
561 Details : ptlrpcd_check call obd_zombie_impexp_cull and wait request which
562 should be handled by ptlrpcd. This produce long age waiting and
563 -ETIMEOUT ptlrpc_invalidate_import and as result LASSERT.
569 Description: ASSERTION(CheckWriteback(page,cmd)) failed
570 Details : badly clear PG_Writeback bit in ll_ap_completion can produce false
574 Frequency : only with broken builds/installations
576 Description: no LBUG if lquota.ko and fsfilt_ldiskfs.ko are different versions
577 Details : just return an error to a user, put a console error message
579 Severity : enhancement
581 Description: Update to RHEL5 kernel-2.6.18-53.1.19.el5.
583 Severity : enhancement
585 Description: Update to RHEL4 kernel-2.6.9-67.0.15.
587 Severity : enhancement
589 Description: enable MGS and MDT services start separately
590 Details : add a 'nomgs' option in mount.lustre to enable start a MDT with
591 a co-located MGS without starting the MGS, which is a complement
592 to 'nosvc' mount option.
595 Frequency : always, on big-endian systems
597 Description: cleanup in ptlrpc code, related to PPC platform
598 Details : store magic in native order avoid panic's in recovery on PPC
599 node and forbid from this error in future. Also fix posibily
600 of twice swab data. Fix get lov striping to userpace.
603 Frequency : rarely, if replay get lost on server
605 Description: server incorrectly drop resent replays lead to recovery failure.
606 Details : do not drop replay according to msg flags, instead we check the
607 per-export recovery request queue for duplication of transno.
610 Frequency : after recovery
612 Description: precreate to many object's after del orphan.
613 Details : del orphan st in oscc last_id == next_id and this triger growing
614 count of precreated objects. Set flag LOW to skip increase count
615 of precreated objects.
618 Frequency : rare, on clear nid stats
620 Description: ASSERTION(client_stat->nid_exp_ref_count == 0)
621 Details : when clean nid stats sometimes try destroy live entry,
622 and this produce panic in free.
625 Frequency : occasionally since 1.6.4
627 Description: Stack overflow during MDS log replay
628 Details : ease stack pressure by using a thread dealing llog_process.
631 Frequency : very rare
633 Description: MDT cannot be unmounted, reporting "Mount still busy"
634 Details : Mountpoint references were being leaked during open reply
635 reconstruction after an MDS restart. Drop mountpoint reference
636 in reconstruct_open() and free dentry reference also.
641 Description: wait until IO finished before start new when do lock cancel.
642 Details : VM protocol want old IO finished before start new, in this case
643 need wait until PG_writeback is cleared until check dirty flag
644 and call writepages in lock cancel callback.
649 Description: mds_mfd_close() ASSERTION(rc == 0)
650 Details : In mds_mfd_close(), we need protect inode's writecount change
651 within its orphan write semaphore to prevent possible races.
654 Frequency : rare, on shutdown ost
656 Description: don't hit live lock with umount ost.
657 Details : shrink_dcache_parent can be in long loop with destroy dentries,
658 use shrink_dcache_sb instead.
661 Frequency : only when echo_client is used
663 Description: don't panic with use echo_client
664 Details : echo client pass NULL as client nid pointer and this produce NULL
668 Frequency : Always on 32-bit PowerPC systems
670 Description: fix build on PPC32
671 Details : compile code with -m64 flag produce wrong object file for PPC32.
676 Description: MDS LBUG: ASSERTION(!IS_ERR(dchild))
677 Details : In reconstruct_* functions, LASSERTs on both the data supplied
678 by a client, and the data on disk are dangerous and incorrect.
679 Change them with client eviction.
681 Severity : enhancement
683 Description: skiplist implementation simplification
684 Details : skiplists are used to group compatible locks on granted list
685 that was implemented as tracking first and last lock of each lock
686 group the patch changes that to using doubly linked lists
690 Description: delete compatibility for 32bit qdata
691 Details : as planned, when lustre is beyond b1_8, lquota won't support 32bit
692 qunit. That means servers of b1_4 and servers of b1_8 can't be
693 used together if users want to use quota.
696 Frequency : only with administrator action
698 Description: mount failure if config log has invalid conf_param setting
699 Details : If administrator specified an incorrect configuration parameter
700 with "lctl conf_param" this would cause an error during future
701 client mounts. Instead, ignore the bad configuration parameter.
704 Frequency : blocks per group < blocksize*8 and uninit_groups is enabled
706 Description: ldiskfs error: XXX blocks in bitmap, YYY in gd
707 Details : If blocks per group is less than blocksize*8, set rest of the
711 Frequency : Application do stride read on lustre
713 Description: The read performance will drop a lot if the application does
715 Details : Because the stride_start_offset are missing in stride read-ahead,
716 it will cause clients read a lot of unused pages in read-ahead,
717 then the read-performance drops.
721 Description: more ldlm soft lockups
722 Details : In ldlm_resource_add_lock(), call to ldlm_resource_dump()
723 starve other threads from the resource lock for a long time in
724 case of long waiting queue, so change the debug level from
725 D_OTHER to the less frequently used D_INFO.
727 Severity : enhancement
729 Description: add -gid, -group, -uid, -user options to lfs find
731 Severity : enhancement
733 Description: ll_recover_lost_found_objs - recover objects in lost+found
734 Details : OST corruption and subsequent e2fsck can leave objects in the
735 lost+found directory. Using the "ll_recover_lost_found_objs"
736 tool, these objects can be retrieved and data can be salvaged
737 by using the object ID saved in the fid EA on each object.
742 Description: this bug _only_ happens when inode quota limitation is very low
743 (less than 12), so that inode quota unit is 1 at initialization.
744 Details : if remaining quota equates 1, it is a sign to demonstate that quota
745 is effective now. So least quota qunit should be 2.
749 Description: Hung threads in invalidate_inode_pages2_range
750 Details : The direct IO path doesn't call check_rpcs to submit a new RPC once
751 one is completed. As a result, some RPCs are stuck in the queue
756 Description: Procfs and llog threads access destoryed import sometimes.
757 Details : Sync the import destoryed process with procfs and llog threads by
758 the import refcount and semaphore.
762 Description: mds fails to respond, threads stuck in ldlm_completion_ast
763 Details : Sort source/child resource pair after updating child resource.
768 Description: kernel BUG at ldiskfs2_ext_new_extent_cb
769 Details : If insertion of an extent fails, then discard the inode
770 preallocation and free data blocks else it can lead to duplicate
775 Description: don't always update ctime in ext3_xattr_set_handle()
776 Details : Current xattr code updates inode ctime in ext3_xattr_set_handle()
777 In some cases the ctime should not be updated, for example for
778 2.0->1.8 compatibility it is necessary to delete an xattr and it
779 should not update the ctime.
783 Description: add quota statistics
784 Details : 1. sort out quota proc entries and proc code.
785 2. add quota statistics
790 Description: quotas are not honored with O_DIRECT
791 Details : all writes with the flag O_DIRECT will use grants which leads to
792 this problem. Now using OBD_BRW_SYNC to guard this.
796 Bugzilla : 15713/16362
797 Description: Assertion in iopen_connect_dentry in 1.6.3
798 Details : looking up an inode via iopen with the wrong generation number can
799 populate the dcache with a disconneced dentry while the inode
800 number is in the process of being reallocated. This causes an
801 assertion failure in iopen since the inode's dentry list contains
802 both a connected and disconnected dentry.
806 Description: assertion failure in ldlm_handle2lock()
807 Details : fix a race between class_handle_unhash() and class_handle2object()
808 introduced in lustre 1.6.5 by bug 13622.
810 Severity : enhancement
812 Description: superblock lock contention with many SMP cores on one client
813 Details : several client filesystem locks were highly contended on SMP
814 NUMA systems with 8 or more cores. Per-CPU datastructures
815 and more efficient locking implemented to reduce contention.
820 Description: Kernel BUG: sd_iostats_bump: unexpected disk index
821 Details : remove the limit of 256 scsi disks in the sd_iostat patch
826 Description: oops in sd_iostats_seq_show()
827 Details : unloading/reloading the scsi low level driver triggers a kernel
828 bug when trying to access the sd iostat file.
833 Description: Kernel panics during QLogic driver reload
834 Details : REQ_BLOCK_PC requests are not handled properly in the sd iostat
835 patch, causing memory corruption.
840 Description: journal_dev option does not work in b1_6
841 Details : pass mount option during pre-mount.
843 Severity : enhancement
845 Description: Add a FIEMAP(FIle Extent MAP) ioctl for ldiskfs
846 Details : FIEMAP ioctl will allow an application to efficiently fetch the
847 extent information of a file. It can be used to map logical blocks
848 in a file to physical blocks in the block device.
851 Frequency : only with adaptive timeout enabled
853 Description: DEBUG_REQ() bad paging request
854 Details : ptlrpc_at_recv_early_reply() should not modify req->rq_repmsg
855 because it can be accessed by reply_in_callback() without the
859 Frequency : only on Cray X2
861 Description: X2 build failures
862 Details : fix build failures on Cray X2.
866 Description: xid & resent requests
867 Details : Initialize RPC XID from clock at startup (randomly if clock is
872 Description: quota recovery deadlock during mds failover
873 Details : This patch includes att18982, att18236, att18237 in bz14840.
875 1. fix osts hang when mds does failover with quotaon
876 2. prevent watchdog storm when osts threads wait for the
881 Description: kernel panic on racer
882 Details : Do not access dchild->d_inode when IS_ERR(dchild) is true.
884 Severity : enhancement
886 Description: Add lustre_start utility to start or stop multiple Lustre servers
891 Description: Lustre GPF in {:ptlrpc:ptlrpc_server_free_request+373}
892 Details : In case of memory pressure, list_del() can be called twice on
893 req->rq_history_list, causing a kernel oops.
897 Description: (ptllnd_peer.c:557:kptllnd_peer_check_sends()) ASSERTION(!in_interrupt()) failed
898 Details : fix stack overflow in the distributed lock manager by defering
899 export eviction after a failed ast to the elt thread instead of
900 handling it in the dlm interpret routine.
902 Severity : enhancement
904 Description: More exported tunables for mballoc
905 Details : Add support for tunable preallocation window and new tunables for
910 Description: Detect corruption of block bitmap and checking for preallocations
911 Details : Checks validity of on-disk block bitmap. Also it does better
912 checking of number of applied preallocations. When corruption is
913 found, it turns filesystem readonly to prevent further corruptions.
917 Frequency : only for big-endian servers
918 Description: Check for big-endian system while mounting fs with extents feature
919 Details : Mounting a filesystem with extents feature will fail on big-endian
920 systems since ext3-based ldiskfs is not supported on big-endian
921 systems. Can be over-ridden with "bigendian_extents" mount option.
925 Description: Excessive recovery window
926 Details : With AT enabled, the recovery window can be excessively long (6000+
927 seconds). To address this problem, we no longer use
928 OBD_RECOVERY_FACTOR when extending the recovery window (the connect
929 timeout no longer depends on the service time, it is set to
930 INITIAL_CONNECT_TIMEOUT now) and clients report the old service
931 time via pb_service_time.
935 Description: Watchdog triggered on MDS failover
936 Details : enable OBD_CONNECT_MDT flag when connecting from the MDS so that
937 the OSTs know that the MDS "UUID" can be reused for the same export
938 from a different NID, so we do not need to wait for the export to
941 -------------------------------------------------------------------------------
943 2008-05-26 Sun Microsystems, Inc.
945 * Support for kernels:
946 2.6.5-7.311 (SLES 9),
947 2.6.9-67.0.7.EL (RHEL 4),
948 2.6.16.54-0.2.5 (SLES 10),
949 2.6.18-53.1.14.el5 (RHEL 5),
950 2.6.22.14 vanilla (kernel.org)
951 * Client support for unpatched kernels:
952 (see http://wiki.lustre.org/index.php?title=Patchless_Client)
953 2.6.16 - 2.6.22 vanilla (kernel.org)
954 * Due to problems with nested symlinks and FMODE_EXEC (bug 12652),
955 we do not recommend using patchless RHEL4 clients with kernels
956 prior to 2.6.9-55EL (RHEL4U5).
957 * Recommended e2fsprogs version: 1.40.7-sun1
958 * Note that reiserfs quotas are disabled on SLES 10 in this kernel.
959 * RHEL 4 and RHEL 5/SLES 10 clients behaves differently on 'cd' to a
960 removed cwd "./" (refer to Bugzilla 14399).
961 * A new quota file format has been introduced in 1.6.5.
962 The format conversion from prior releases is handled transparently,
963 but releases older than 1.4.12/1.6.5 will not understand this new
964 format. The automatic format conversion can be avoided by running
965 the following command on the MDS before upgrading:
966 'tunefs.lustre --param="mdt.quota_type=ug1" $MDTDEV'.
967 For more information, please refer to bugzilla 13904.
971 Description: quota performance fix
972 Details : quota data is written in journalled mode instead of ordered to
977 Description: lfs support for human-readable quota grace time strings
978 Details : lfs setquota -t and lfs quota -t represent quota grace times
979 in "XXwXXdXXhXXmXXs" format instead of large values in seconds
982 Frequency : always with o2ib 1.3 and sles10
984 Description: fix build with SLES10 and o2ib v3.
985 Details : sles10 uses diffrent name for Module.symver file but configure
986 assume this file has same name on RHEL/SLES/vanila kernels.
989 Frequency : very rare, if additional xattrs are used on kernels >= 2.6.12
991 Description: MDS may lose file striping (and hence file data) in some cases
992 Details : If there are additional extended attributes stored on the MDS,
993 in particular ACLs, SELinux, or user attributes (if user_xattr
994 is specified for the client mount options) then there is a risk
995 of attribute loss. Additionally, the Lustre file striping
996 needs to be larger than default (e.g. striped over all OSTs),
997 and an additional attribute must be stored initially in the
998 inode and then increase in size enough to be moved to the
999 external attribute block (e.g. ACL growing in size) for file
1002 Severity : enhancement
1004 Description: add message levels for liblustreapi
1009 Description: MDT cannot be unmounted, reporting "Mount still busy"
1010 Details : Mountpoint references were being leaked during open reply
1011 reconstruction after an MDS restart. Drop mountpoint reference
1012 in reconstruct_open() and free dentry reference also.
1017 Description: fix for occasional failure case of -ENOSPC in recovery-small tests
1018 Details : Move the 'good_osts' check before the 'total_bavail' check. This
1019 will result in an -EAGAIN and in the exit call path we call
1020 alloc_rr() which will with increasing aggressiveness attempt to
1021 aquire precreated objects on the minimum number of required OSCs.
1025 Description: Use old size assignment to avoid deadlock
1026 Details : This reverts the changes in bugs 2369 and bug 14138 that introduced
1027 the scheduling while holding a spinlock. We do not need locking
1028 for size in ll_update_inode() because size is only updated from
1029 the MDS for directories or files without objects, so there is no
1030 other place to do the update, and concurrent access to such inodes
1031 are protected by the inode lock.
1035 Description: Use __u64 instead of int for valid bits
1039 Description: resolve "_IOWR redefined" build error on SLES10
1043 Description: dump the memory debugging after all modules are unloaded to
1044 suppress false negative in conf_sanity test 39
1048 Description: the recovery timer never expires
1049 Details : for new client connect request, the recovery timer should not be
1050 reset, otherwise recovery timer will never expired, if the old
1051 client never come. Only old client connect and first connection
1052 req should trigger recovery timer reset.
1056 Description: the min numbers of lproc stats are wrong
1057 Details : adding a new constant LC_MIN_INIT and use it for initialization
1061 Frequency : always with interactive lfs
1063 Description: Reinitialize optind to 0 so that interactive lfs works in all cases
1066 Frequency : with multiple concurrent readdir processes in same directory
1067 Bugzilla : 15406, 15169, 15175
1068 Description: misc fixes for directory readahead.
1069 Details : prevent previous statahead async RPC callback from processing the
1070 current "statahead_info", race condition between async RPC callback
1071 add dentry into dentry hash table and "ls" thread revalidate such
1072 dentry, statahead his/miss control for hidden items, and so on.
1074 Severity : enhancement
1076 Description: build kernel-ib packages for OFED 1.3 in our release cycle
1080 Description: incore types cleaning in quota code (with respect to 64-bit limits)
1081 Details : several u32 variables declarations are replaced with u64 declarations
1086 Description: fix SLES kernel versioning
1087 Details : the kernel version for our SLES 10 kernel did not include a "-"
1088 before the "smp" at the end. while this was not a problem in
1089 general, it did mean that software trying to use the kernel
1090 version to try to detect a vendor specific kernel would fail.
1091 this was most evident by the OFED build scripts.
1096 Description: Don't update lov_desc members until making sure they are valid
1097 Details : When updating lov_desc members via proc fs, need fix their
1098 validities before doing the real update.
1101 Frequency : very rare
1103 Description: don't put request into delay list while invalidate in flight.
1104 Details : ptlrpc_delay_request sometimes put in delay list while invalidate
1105 import in flight. this produce timeout for invalidate and sometimes
1106 can cause stale data.
1108 Severity : enhancement
1110 Description: Update kernel to SLES9 2.6.5-7.311.
1112 Severity : enhancement
1114 Description: Update kernel to RHEL4 2.6.9-67.0.7.
1119 Frequency : on PPC only
1120 Description: not convert ost objects for directory because it's not exist.
1121 Details : ll_dir_getstripe assume dirrectory has ost objects but this wrong.
1123 Severity : enhancement
1125 Description: Fix warnings with compile liblustre at sles10/rhel5 which have
1126 __u64 as usingied long long type.
1129 Frequency : rare, on shutdown
1131 Description: race process ast vs remove callback
1132 Details : removing callback before disconnect import open race with
1133 processing callback.
1135 Severity : enhancement
1137 Description: Update kernel to SLES9 2.6.5-7.311.
1139 Severity : enhancement
1141 Description: Files open for execute are not marked busy on SLES10
1142 Details : Add FMODE_EXEC to SLES10 SP1 server kernel series.
1144 Severity : enhancement
1146 Description: Add server support for vanilla-2.6.22.14.
1149 Frequency : occasional
1151 Description: Avoid lov_create() getting stuck in obd_statfs_rqset()
1152 Details : If an OST is down the MDS will hang indefinitely in
1153 obd_statfs_rqset() waiting for the statfs data. While for
1154 MDS QOS usage of statfs, it should not stuck in waiting.
1156 Severity : enhancement
1158 Description: Disable adaptive timeouts by default
1161 Frequency : on network error
1163 Description: panic with double free request if network error
1164 Details : mdc_finish_enqueue is finish request if any network error occuring,
1165 but it's true only for synchronus enqueue, for async enqueue
1166 (via ptlrpcd) this incorrect and ptlrpcd want finish request
1170 Frequency : rare, on recovery
1172 Description: read procfs can produce deadlock in some situation
1173 Details : Holding lprocfs lock with send rpc can produce block for destroy
1174 obd objects and this also block reconnect with -EALREADY.
1175 This isn't fix all lprocfs bugs - but make it rare.
1177 Severity : enhancement
1179 Description: Update kernel to RHEL5 2.6.18-53.1.14.el5.
1182 Frequency : frequent on X2 node
1184 Description: mdc_set_open_replay_data LBUG
1185 Details : Set replay data for requests that are eligible for replay.
1190 Description: lustre_mgs: operation 101 on unconnected MGS
1191 Details : When MGC is disconnected from MGS long enough, MGS will evict the
1192 MGC, and late on MGC cannot successfully connect to MGS and a lot
1193 of the error messages complaining that MGS is not connected.
1196 Frequency : rare, depends on device drivers and load
1198 Description: MDS or OSS nodes crash due to stack overflow
1199 Details : Code changes in 1.6.4 increased the stack usage of some functions.
1200 In some cases, in conjunction with device drivers that use a lot
1201 of stack, the MDS (or possibly OSS) service threads could overflow
1202 the stack. One change which was identified to consume additional
1203 stack has been reworked to avoid the extra stack usage.
1205 Severity : enhancement
1207 Description: Update to RHEL5 latest kernel-2.6.18-53.1.13.el5.
1209 Severity : enhancement
1211 Description: Update to SLES10 SP1 latest kernel-2.6.16.54-0.2.5.
1213 Severity : enhancement
1215 Description: Update to RHEL5 latest kernel-2.6.18-53.1.6.el5.
1217 Serverity : enhancement
1219 Description: Update RHEL4 kernel to 2.6.9-67.0.4.
1222 Frequency : rare on shutdown OST
1224 Description: Don't allow skipping OSTs if index has been specified.
1225 Details : Don't allow skipping OSTs if index has been specified, make locking
1226 in internal create lots better.
1231 Description: ASSERTION(!PageDirty(page)) failed
1232 Details : Wrong check could lead to an assertion failure under specific
1238 Description: LBUG in ptlrpc_check_set() bad phase ebc0de00
1239 Details : access to bitfield in structure is always rounded to long
1240 and this produce problem with not atomic change any bit.
1245 Description: Lustre make rpms failed.
1246 Details : Remove ldiskfs spec file to avoids rpmbuild be confused when
1247 builds Lustre rpms from tarball.
1249 Severity : enhancement
1251 Description: Update to SLES9 SP4 kernel-2.6.5-7.308.
1254 Frequency : rare on shutdown OST
1256 Description: If llog cancel was not send before clean_exports phase, this can
1257 produce deadlock in llog code.
1258 Details : If llog thread has last reference to obd and call class_import_put
1259 this produce deadlock because llog_cleanup_commit_master wait when
1260 last llog_commit_thread exited, but this never success because was
1261 called from llog_commit_thread.
1264 Frequency : only if OST index is skipped
1266 Description: NULL lov_tgts causing MDS oops
1267 Details : more safe checks for NULL lov_tgts for avoid oops.
1269 Severity : enhancement
1271 Description: Update to RHEL4 latest kernel-2.6.9-67.0.1.EL.
1273 Severity : enhancement
1275 Description: Update to RHEL5 latest kernel-2.6.18-53.1.4.el5.
1280 Description: make mgs_setparam() handle fsname containing dash
1281 Details : fsname containing a dash does not work with lctl conf_param
1283 Severity : enhancement
1285 Description: Update to RHEL4 Update-6 kernel-2.6.9-67.EL.
1288 Frequency : rare, in recovery and (or) destroy lovobjid file.
1290 Description: rewrite lov objid code.
1291 Details : Cleanup for lov objid code, remove scability problems and wrong
1292 locking. Fix sending last_id into ost.
1294 Severity : enhancement
1296 Description: Update to SLES10 SP1 latest kernel-2.6.16.54-0.2.3.
1298 Severity : enhancement
1300 Description: Update to RHEL5 Update-1 kernel 2.6.18-53.el5.
1301 Details : Use d_move_locked instead of __d_move.
1304 Frequency : rare, at shutdown
1306 Description: access already free / zero obd_namespace.
1307 Details : if client_disconnect_export was called without force flag set,
1308 and exist connect request in flight, this can produce access to
1309 NULL pointer (or already free pointer) when connect_interpret
1310 store ocd flags in obd_namespace.
1313 Frequency : only at startup
1315 Description: not alloc memory with spinlock held.
1316 Details : allocation memory with GFP_KERNEL can produce sleep deadlock,
1317 if any spinlock held.
1322 Description: lfs find does not continue on file error
1323 Details : Continue other files processing when a file/dir is absent.
1327 Description: Inconsistent usage of lustre_pack_reply()
1328 Details : Standardize the usage of lustre_pack_reply() such that it
1329 always generate a CERROR on failure.
1332 Frequency : very rare
1334 Description: Fix replay if there is an un-replied request and open
1335 Details : In some cases, older replay request will revert the
1336 mcd->mcd_last_xid on MDS which is used to record the client's
1337 latest sent request.
1339 Severity : enhancement
1341 Description: Update to RHEL5 kernel 2.6.18-8.1.15.el5.
1343 Severity : enhancement
1345 Description: Update to SLES10 SP1 kernel 2.6.16.53-0.16
1347 Severity : enhancement
1349 Description: Update to SLES9 kernel-2.6.5-7.287.3.
1351 Severity : enhancement
1353 Description: Update to RHEL4 kernel-2.6.9-55.0.12.EL.
1355 Severity : enhancement
1357 Description: Build SLES10 patchless client fails
1358 Details : The configure was broken by run ./configure with
1359 --with-linux-obj=.... argument for patchless client. When the
1360 configure use --with-linux-obj, the LINUXINCLUDE= -Iinclude
1361 can't search header adequately. Use absolute path such as
1362 -I($LINUX)/include instead.
1364 Severity : enhancement
1366 Description: Lustre Page Accounting
1367 Details : New macros for page alloc and free which enable accounting
1368 of page allocation of Lustre. Use percpu counters to store memory
1369 and page statistics.
1372 Frequency : only if debugging is disabled
1374 Description: LASSERT_{REQ,REP}SWAB macros are buggy
1375 Details : If SWAB_PARANOIA is disabled, the LASSERT_REQSWAB and
1376 LASSERT_REPSWAB macros become no-ops, which is incorrect. Drop
1377 these macros and replace them with their definitions instead.
1382 Description: interrupt oig_wait produce painc on resend.
1383 Details : brw_redo_request can be used for resend requests from ptlrpcd and
1384 private set, and this produce situation when rq_ptlrpcd_data not
1385 copyed to new allocated request and triggered LBUG on assert
1386 req->rq_ptlrpcd_data != NULL. But this member used only for wakeup
1387 ptlrpcd set if request is changed and can be safety changed to use
1390 Severity : enhancement
1392 Description: organize the server-side client stats on per-nid basis
1393 Details : Change the structure of stats under obdfilter and mds to
1401 The "uuid"s file would list the uuids of _active_ exports.
1402 And the clear entry is to clear all stats and stale nids.
1407 Description: Processes looping in ll_readdir() on Lustre clients finally causing
1408 a full node pseudo-hang
1409 Details : Concurrent access to the same directory from multiple clients with
1410 intensive file creation/removal can cause a client node to spin in
1411 ll_readdir(). i_version must be increased every time the lock is
1412 cancelled to ensure a revalidate is done.
1417 Description: touch file failed when fs is not full
1418 Details : OST in recovery should not be discarded by MDS in alloc_qos(),
1419 otherwise we can get ENOSP while fs is not full.
1422 Frequency : only for Cray XT3
1423 Bugzilla : 12829/13455
1424 Description: Changing primary group doesn't change the group lustre assigns to
1426 Details : When CRAY_XT3 is defined, the fsgid supplied by the client is
1427 overridden with the primary group provided by the group upcall,
1428 whereas the supplied fsgid can be trusted if it is in the list of
1429 supplementary groups returned by the group upcall.
1431 Severity : enhancement
1433 Description: Root Squash Functionality
1434 Details : Implementation of NFS-like root squash capability. Specifically,
1435 don't allow someone with root access on a client node to be able
1436 to manipulate files owned by root on a server node.
1438 Severity : enhancement
1440 Description: Slow trucate/writes to huge files at high offsets.
1441 Details : Directly associate cached pages to lock that protect those pages,
1442 this allows us to quickly find what pages to write and remove
1443 once lock callback is received.
1448 Description: Too many locks accumulating on client during NFS usage
1449 Details : mds_open improperly used accmode to find out access mode to a
1450 file. Also mdc_intent_lock logic to find out if we already have
1451 lock similar to just received was flawed since introduction of
1452 skiplists - locks are now added to the front of the granted
1457 Description: Hit ASSERTION(obd->obd_stopping == 1) failed in some setup failed
1459 Details : In obd setup failure handler, obd_stopping will not necessarily to
1460 be 1, and obd_set_up should also be checked to make sure whether
1461 obd is completely setup.
1463 Severity : enhancement
1465 Description: Allow masking D_WARNING, D_ERROR messages from console
1466 Details : Console messages can now be disabled via lnet.printk.
1471 Description: User code with malformed file open parameter crashes client node
1472 Details : Before packing join_file req, all the related reference should be
1473 checked carefully in case some malformed flags cause fake join_file
1479 Description: shrink/enlarge qunit size when needed; fix the problem of coarse
1480 grain of quota doing harm to quota's accuracy
1481 Details : qunit size will be changed when quota limitation is too low/high;
1482 record the pending quota write in order to get more accureate
1483 quota; delete the patch for bug12588, which is unnecessary when
1484 this patch is landed. This bug also contains fixes for bug 14526,
1485 14299, 14601 and 13794.
1489 Description: LDLM_ENQUEUE races with LDLM_CP_CALLBACK
1490 Details : ldlm_completion_ast() assumes that a lock is granted when the req
1491 mode is equal to the granted mode. However, it should also check
1492 that LDLM_FL_CP_REQD is not set.
1496 Description: Heavy nfs access might result in deadlocks
1497 Details : After ELC code landed, it is now improper to enqueue any mds
1498 locks under och_sem, because enqueue might want to decide to
1499 cancel open locks for same inode we are holding och_sem for.
1503 Description: 35% write performance drop with ldiskfs2 when quotas are on
1504 Details : Enable ext3 journalled quota by default to improve performance
1505 when quotas are turned on.
1509 Description: Client eviction while running blogbench
1510 Details : A lot of unlink operations with concurrent I/O can lead to a
1511 deadlock causing evictions. To address the problem, the number of
1512 oustanding OST_DESTROY requests is now throttled to
1513 max_rpcs_in_flight per OSC and LDLM_FL_DISCARD_DATA blocking
1514 callbacks are processed in priority.
1517 Frequency : RHEL4 only
1519 Description: mkfs is very slow on IA64/RHEL4
1520 Details : A performance regression has been discovered in the MPT Fusion
1521 driver between versions 3.02.73rh and 3.02.99.00rh. As a
1522 consequence, we have downgraded the MPT Fusion driver in the RHEL4
1523 kernel from 3.02.99.00 to 3.02.73 until this problem is fixed.
1526 Frequency : PPC/PPC64 only
1528 Description: conflicts between asm-ppc64/types.h and lustre_types.h
1529 Details : fix duplicated definitions between asm-ppc64/types.h and
1530 lustre_types.h on PPC.
1533 Frequency : PPC/PPC64 only
1535 Description: asm-ppc/segment.h does not exist
1536 Details : fix compile issue on PPC.
1540 Description: data checksumming impacts single node performance
1541 Details : add support for several checksum algorithms. Currently, CRC32 and
1542 Adler-32 are supported. The checksum type can be changed on the fly
1543 through /proc/fs/lustre/osc/*/checksum_type.
1547 Description: use adler32 for page checksums
1548 Details : when available, use the Adler-32 algorithm instead of CRC32 for
1553 Description: better handle error messages in extents code
1555 Severity : enhancement
1557 Description: SNMP support enhancement
1558 Details : Adding total number of sampled request for an MDS node in snmp
1561 Severity : enhancement
1563 Description: Optimize ldlm waiting list processing for PR extent locks
1564 Details : When processing waiting list for read extent lock and meeting read
1565 lock that is same or wider to it that is not contended, skip
1566 processing rest of the list and immediatelly return current
1567 status of conflictness, since we are guaranteed there are no
1568 conflicting locks in the rest of the list.
1572 Description: Time out and refuse to reconnect
1573 Details : When the failover node is the primary node, it is possible
1574 to have two identical connections in imp_conn_list. We must
1575 compare not conn's pointers but NIDs, otherwise we
1576 can defeat connection throttling.
1580 Description: Client not clear own cache if answer to reconnect is lost.
1581 Details : Client gets evicted from server. Now client also thinks it is
1582 disconnected (or gets ENOTCONN on its operation) and decides to
1583 reconnect. Server receives reconnect message, but cannot find
1584 export. New export is created that is fully valid (new cookie!),
1585 but reply is lost and not reported to client. Client reconnects
1586 again and gets back a just-created connection, but it is not new
1587 so client thinks it was not evicted and does not do recovery.
1591 Description: Detect stride IO mode in read-ahead
1592 Details : When a client does stride read, read-ahead should detect that and
1593 read-ahead pages according to the detected stride pattern.
1597 Description: build for x2 fails
1598 Details : fix compile issue on Cray systems.
1600 Severity : enhancement
1602 Description: implement readv/writev APIs(aio_read/aio_writes in newer kernels)
1603 Details : This greatly improves speed of NFS writes on 2.6 kernels.
1606 Frequency : only on PPC/SLES10
1608 Description: "BITS_PER_LONG is not 32 or 64" in linux/idr.h
1609 Details : On SLES10/PPC, fs.h includes idr.h which requires BITS_PER_LONG to
1610 be defined. Add a hack in mkfs_lustre.c to work around this compile
1615 Description: LASSERT on MDS when client holding flock lock dies
1616 Details : ldlm pool logic depends on number of granted locks equal to
1617 number of released locks which is not true for flock locks, so
1618 just exclude such locks from consideration.
1622 Description: MDS deadlock with many ll_sync_lov threads and I/O stalled
1623 Details : Use fsfilt_sync() for both the whole filesystem sync and
1624 individual file sync to eliminate dangerous inode locking
1625 with I_LOCK that can lead to a deadlock.
1629 Description: Update an obsolete wirecheck.c generator
1630 Details : Update wirecheck.c/wirehdr.c and regenerate wiretest.c
1634 Description: Client can panic on open sometimes
1635 Details : It is possible that we try to free already freed request in
1636 ll_file_open in some error cases when we send request from
1641 Description: performance in 1.6.3
1642 Details : Force q->max_phys_segments to MAX_PHYS_SEGMENTS on SLES10 to be
1643 sure that 1MB requests are not fragmented by the block layer.
1647 Description: LDLM soft lockups - improvement
1648 Details : It is be possible to send the lock handle along with each read
1649 or write request because the client is already doing a lock match
1650 itself so there isn't any reason the OST should have to re-do that
1656 Description: lfs quota fails with deactivated OSTS
1657 Details : With this patch, three improvements are included:
1658 1. detete the softlimit in mds and osts when use "lfs quota".
1659 2. display the inaccurate data in the output of "lfs quota".
1660 3. try to get quota info when "lfs quota" is executed.
1665 Description: Extent locks not granted with no conflicts sometimes.
1666 Details : When race occurs in glimpse handler and nothing is returned,
1667 we do not reprocess the queue after lock cancel, and that leads
1668 to a stall until next activity on a resource
1671 Frequency : failover with quotaon
1673 Description: during mds failovers with quota on, OSTs got into deadlock state
1674 and causing dumpstack.
1675 Details : for every quota slave, at any time, there is only one quota req
1676 is sent to quota master for every uid/gid. Before that quota req
1677 returns, all the thread relative to the same uid/gid will wait.
1678 So if the quota req is lost because mds failovers or any other
1679 reasons, this bug will be hit. Now, dqacq_interpret() will handle
1680 quota reqs who time out.
1682 Severity : enhancement
1685 Description: when quota slave checks if quota is enough, there is an unnecessary
1687 Details : place this wait on necessary place instead of always waiting.
1689 --------------------------------------------------------------------------------
1691 2007-12-07 Cluster File Systems, Inc. <info@clusterfs.com>
1693 * Support for kernels:
1694 2.6.5-7.286 (SLES 9),
1695 2.6.9-55.0.9.EL (RHEL 4),
1696 2.6.16.53-0.8 (SLES 10),
1697 2.6.18-8.1.14.el5 (RHEL 5),
1698 2.6.18.8 vanilla (kernel.org)
1699 * Client support for unpatched kernels:
1700 (see http://wiki.lustre.org/index.php?title=Patchless_Client)
1701 2.6.16 - 2.6.22 vanilla (kernel.org)
1702 * Due to recently discovered recovery problems, we do not recommend
1703 using patchless RHEL 4 clients with this or any earlier release.
1704 * Recommended e2fsprogs version: 1.40.2-cfs1
1705 * Note that reiserfs quotas are disabled on SLES 10 in this kernel.
1708 Frequency : occasional
1710 Description: MDS hang or stay in waiting lock
1711 Details : If client receive lock with CBPENDING flag ldlm need send lock
1712 cancel as separate rpc, to avoid situation when cancel request
1713 can't processed due all i/o threads stay in wait lock.
1716 Frequency : occasional
1718 Description: Do not fail import if osc_interpret_create gets -EAGAIN
1719 Details : If osc_interpret_create got -EAGAIN it immediately exits and
1720 wakeup oscc_waitq. After wakeup oscc_wait_for_objects call
1721 oscc_has_objects and see OSC has no objests and call
1722 oscc_internal_create to resend create request.
1724 Severity : enhancement
1726 Description: Update kernel patches for SLES10 2.6.16.53-0.8.
1727 Details : Update which_patch & target file for SLES10 latest kernel.
1729 Severity : enhancement
1731 Description: add --type and --size parameters to lfs find
1732 Details : Enhance lfs find by adding filetype and filesize parameters. Also
1733 multiple OBDs can now be specified for the --obd option.
1735 Severity : enhancement
1737 Description: eliminate client locks in face of contention
1738 Details : file contention detection and lockless i/o implementation
1739 for contended files.
1741 Severity : enhancement
1743 Description: Remove client patches from SLES 10 kernel.
1744 Details : This causes SLES 10 clients to behave as patchless clients
1745 even on a Lustre-patched (server) kernel.
1747 Severity : enhancement
1749 Description: use i_size_read and i_size_write in 2.6 port
1750 Details : replace inode->i_size access with i_size_read/write()
1753 Frequency : when removing large files
1755 Description: scheduling issue during removal of large Lustre files
1756 Details : Don't take the BKL in fsfilt_ext3_setattr() for 2.6 kernels.
1757 It causes scheduling issues when removing large files (17TB in the
1763 Description: 1.4.11 Can't handle directories with stripe set and extended ACLs
1764 Details : Impossible (EPROTO is returned) to access a directory that has a
1765 non-default striping and ACLs.
1768 Frequency : only on ppc
1770 Description: /proc/fs/lustre/devices broken on ppc
1771 Details : The patch as applied to 1.6.2 doesn't look correct for all arches.
1772 We should make sure the type of 'index' is loff_t and then cast
1773 explicitly as needed below. Do not assign an explicitly cast
1777 Frequency : only for rhel5
1779 Description: Kernel patches update for RHEL5 2.6.18-8.1.10.el5.
1780 Details : Modify the target file & which_kernel.
1783 Frequency : if the uninit_groups feature is enabled on ldiskfs
1785 Description: e2fsck reports "invalid unused inodes count"
1786 Details : If a new ldiskfs filesystem is created with the "uninit_groups"
1787 feature and only a single inode is created in a group then the
1788 "bg_unused_inodes" count is incorrectly updated. Creating a
1789 second inode in that group would update it correctly.
1794 Description: buffer overruns could theoretically occur
1795 Details : llapi_semantic_traverse() modifies the "path" argument by
1796 appending values to the end of the origin string, and a buffer
1797 overrun may occur. Adding buffer overrun check in liblustreapi.
1799 Severity : enhancement
1801 Description: Add jbd statistics patch for RHEL5 and 2.6.18-vanilla.
1804 Frequency : only if filesystem is inconsistent
1806 Description: handle "serious error: objid * already exists" more gracefully
1807 Details : If LAST_ID value on disk is smaller than the objects existing in
1808 the O/0/d* directories, it indicates disk corruption and causes an
1809 LBUG(). If the object is 0-length, then we should use the existing
1810 object. This will help to avoid a full fsck in most cases.
1812 Severity : enhancement
1814 Description: Kernel patches update for RHEL4 2.6.9-55.0.6.
1815 Details : Modify vm-tunables-rhel4.patch.
1817 Severity : enhancement
1819 Description: Kernel config for 2.6.18-vanilla.
1820 Details : Modify targets/2.6-vanilla.target.in.
1821 Add config file kernel-2.6.18-2.6-vanilla-i686.config.
1822 Add config file kernel-2.6.18-2.6-vanilla-i686-smp.config.
1823 Add config file kernel-2.6.18-2.6-vanilla-x86_64.config.
1824 Add config file kernel-2.6.18-2.6-vanilla-x86_64-smp.config.
1827 Frequency : occasional
1829 Description: improve handling recoverable errors
1830 Details : If request processed with error which can be recoverable on server
1831 request should be resend, otherwise page released from cache and
1837 Description: Kernel patches update for RHEL5 2.6.18-8.1.14.el5.
1838 Details : Modify target file & which_patch.
1839 A flaw was found in the IA32 system call emulation provided
1840 on AMD64 and Intel 64 platforms. An improperly validated 64-bit
1841 value could be stored in the %RAX register, which could trigger an
1842 out-of-bounds system call table access. An untrusted local user
1843 could exploit this flaw to run code in the kernel
1844 (ie a root privilege escalation). (CVE-2007-4573).
1848 Description: change order of libsysio includes
1849 Details : '#include sysio.h' should always come before '#include xtio.h'
1851 Severity : enhancement
1853 Description: adapt the lustre_config script to support the upgrade case
1854 Details : Add "-u" option for lustre_config script to support upgrading 1.4
1855 server targets to 1.6 in parallel.
1860 Description: To avoid grant space > avaible space when the disk is almost
1861 full. Without this patch you might see the error "grant XXXX >
1862 available" or some LBUG about grant, when the disk is almost
1864 Details : In filter_check_grant, for non_grant cache write, we should
1865 check the left space by if (*left > ungranted + bytes), instead
1866 of (*left > ungranted), because only we are sure the left space
1867 is enough for another "bytes", then the ungrant space should be
1868 increase. In client, we should update cl_avail_grant only there is
1869 OBD_MD_FLGRANT in the reply.
1874 Description: Update RHEL 4 kernel to fix local root privilege escalation.
1875 Details : Update to the latest RHEL 4 kernel to fix the vulnerability
1876 described in CVE-2007-4573. This problem could allow untrusted
1877 local users to gain root access.
1880 Frequency : when using O_DIRECT and quotas
1882 Description: Incorrect file ownership on O_DIRECT output files
1883 Details : block usage reported by 'lfs quota' does not take into account
1884 files that have been written with O_DIRECT.
1888 Description: (rw.c:1323:ll_read_ahead_pages()) ASSERTION(page_idx > ria->ria_stoff) failed
1889 Details : Once the unmatched stride IO mode is detected, shrink the stride-ahead
1890 window to 0. If it does hit cache miss, and read-pattern is still
1891 stride-io mode, does not reset the stride window, but also does not
1892 increase the stride window length in this case.
1894 --------------------------------------------------------------------------------
1896 2007-09-27 Cluster File Systems, Inc. <info@clusterfs.com>
1898 * Support for kernels:
1899 2.6.5-7.286 (SLES 9),
1900 2.6.9-55.0.2.EL (RHEL 4),
1901 2.6.16.46-0.14 (SLES 10),
1902 2.6.18-8.1.8.el5 (RHEL 5),
1903 2.6.18.8 vanilla (kernel.org)
1904 * Client support for unpatched kernels:
1905 (see http://wiki.lustre.org/index.php?title=Patchless_Client)
1906 2.6.16 - 2.6.21 vanilla (kernel.org)
1907 * Due to recently discovered recovery problems, we do not recommend
1908 using patchless RHEL 4 clients with this or any earlier release.
1909 * Recommended e2fsprogs version: 1.40.2-cfs1
1910 * Note that reiserfs quotas are disabled on SLES 10 in this kernel.
1914 Description: Fix errors in lfs documentation
1915 Details : Fixes man pages
1917 Severity : enhancement
1919 Description: Adaptive timeouts
1920 Details : RPC timeouts adapt to changing server load and network
1921 conditions to reduce resend attempts and improve recovery time.
1923 Severity : enhancement
1925 Description: llapi_file_create() does not allow some changes
1926 Details : add llapi_file_open() that allows specifying the file creation
1927 mode and open flags, and also returns an open file handle.
1929 Severity : enhancement
1931 Description: df doesn't work properly if diskfs blocksize != 4K
1932 Details : Choose biggest blocksize of OST's as the LOV's blocksize.
1935 Frequency : always on directories with default striping set
1937 Description: lfs find on -1 stripe looping in lsm_lmm_verify_common()
1938 Details : Avoid lov_verify_lmm_common() on directory with -1 stripe count.
1940 Severity : enhancement
1942 Description: merge and cleanup kernel patches.
1943 Details : Remove mnt_lustre_list in vfs_intent-2.6-rhel4.patch.
1945 Severity : enhancement
1947 Description: RedHat Update kernel for RHEL5
1948 Details : Kernel config file for RHEL5.
1950 Severity : enhancement
1952 Description: OSS needs mutliple precreate threads
1953 Details : Add ability to start more than one create thread per OSS.
1956 Frequency : only with quota on the root user
1958 Description: mds_obd_create error creating tmp object
1959 Details : When the user sets quota on root, llog will be affected and can't
1960 create files and write files.
1963 Frequency : Always on ia64 patchless client, and possibly others.
1965 Description: Add EXPORT_SYMBOL check for node_to_cpumask symbol.
1966 Details : This allows the patchless client to be loaded on architectures
1967 without this export.
1972 Description: disorder of journal start and llog_add cause deadlock.
1973 Details : in llog_origin_connect, journal start should happen before
1974 llog_add keep the same order as other functions to avoid
1977 Severity : enhancement
1979 Description: RedHat Update kernel for RHEL5
1980 Details : Modify the kernel config file more closer RHEL5.
1982 Severity : enhancement
1984 Description: Build failure against Centos5 (RHEL5)
1985 Details : Define PAGE_SIZE when it isn't present.
1988 Frequency : occasionally when using NFS
1990 Description: "ll_intent_file_open()) lock enqueue: err: -13" with nfs
1991 Details : with NFS, the anon dentry's parent was set to itself in
1992 d_alloc_anon(), so in MDS, we use rec->ur_fid1 to find the
1993 corresponding dentry other than use rec->ur_name.
1995 Severity : enhancement
1997 Description: enable data checksumming by default
1998 Details : enable checksum by default, allow --disable-checksum
1999 configure option and "-o nochecksum" mount option. Checksums
2000 can also be disabled at runtime via $LPROC/osc/*/checksum_pages.
2003 Frequency : Occasionally with failover
2005 Description: Client eviction due to failover config
2006 Details : after a connection loss, the lustre client should attempt to
2007 reconnect to the last active server first before trying the
2008 other potential connections.
2010 Severity : enhancement
2012 Description: client-side metadata stat-ahead during readdir(directory readahead)
2013 Details : perform client-side metadata stat-ahead when the client detects
2014 readdir and sequential stat of dir entries therein
2019 Description: when mds and osts use different quota unit(32bit and 64bit),
2020 quota will be released repeatly.
2021 Details : void sending multiple quota reqs to mds, which will keep the
2022 status between the reqs.
2025 Frequency : only with liblustre clients on XT3
2027 Description: evictions taking too long
2028 Details : allow llrd to evict clients directly on OSTs
2032 Description: osts not allocated evenly to files
2033 Details : change the condition to increase offset_idx
2035 Severity : enhancement
2037 Description: self-adjustable client's lru lists
2038 Details : use adaptive algorithm for managing client cached locks lru
2039 lists according to current server load, other client's work
2040 pattern, memory activities, etc. Both, server and client
2041 side namespaces provide number of proc tunables for controlling
2046 Description: rewrite ext2-derived code in llite/dir.c and obdclass/uuid.c
2047 Details : rewrite inherited code (uuid parsing code from ext2 utils and
2048 readdir code from ext3) from scratch preserving functionality.
2052 Description: Only those disconnect error should be returned by rq_status.
2053 Details : In open/enqueue processs, Some errors, which will cause client
2054 disconnected, should be returned by rq_status, while other
2055 errors should still be returned by intent, then mdc or llite will
2058 Severity : enhancement
2060 Description: Tune the kernel for good SCSI performance.
2061 Details : Set the value of /sys/block/{dev}/queue/max_sectors_kb
2062 to the value of /sys/block/{dev}/queue/max_hw_sectors_kb
2066 Frequency : Always for filesystems larger than 2TB on 32-bit systems.
2067 Bugzilla : 13547, 13627
2068 Description: Data corruption for OSTs that are formatted larger than 2TB
2070 Details : When generating the bio request for lustre file writes the
2071 sector number would overflow a temporary variable before being
2072 used for the IO. The data reads correctly from Lustre (which
2073 will overflow in a similar manner) but other file data or
2074 filesystem metadata may be corrupted in some cases.
2078 Description: "lfs find -obd UUID" prints directories
2079 Details : "lfs find -obd UUID" will return all directory names instead
2080 of just file names. It is incorrect because the directories
2081 do not reside on the OSTs.
2084 Bugzilla : 13596, 13614
2085 Description: MDS hang after unclean shutdown of lots of clients
2086 Details : Never resend AST requests.
2090 Frequency : Always, for kernels after 2.6.16
2091 Description: Fix warning idr_remove called for id=.. which is not allocated.
2092 Details : Last kernels save old s_dev before kill super and not allow
2093 to restore from callback - restore it before call kill_anon_super.
2097 Description: Fix error on 'ls .' at the top of the Lustre mount.
2098 Details : Don't revalidate dentry if it is a root dentry.
2100 --------------------------------------------------------------------------------
2102 2007-08-27 Cluster File Systems, Inc. <info@clusterfs.com>
2104 * Support for kernels:
2105 2.6.5-7.286 (SLES 9),
2106 2.6.9-55.0.2.EL (RHEL 4),
2107 2.6.16.46-0.14 (SLES 10),
2108 2.6.18-8.1.8.el5 (RHEL 5),
2109 2.6.18.8 vanilla (kernel.org)
2110 * Client support for unpatched kernels:
2111 (see http://wiki.lustre.org/index.php?title=Patchless_Client)
2112 2.6.16 - 2.6.22 vanilla (kernel.org)
2113 * Due to recently discovered recovery problems, we do not recommend
2114 using patchless RHEL 4 clients with this or any earlier release.
2115 * Recommended e2fsprogs version: 1.39.cfs8
2116 * Note that reiserfs quotas are disabled on SLES 10 in this kernel.
2121 Description: block reactivating mgc import until all deactivates complete
2122 Details : Fix race when failing back MDT/MGS to itself (testing)
2124 Severity : enhancement
2126 Description: lfs setstripe enhancement
2127 Details : Make lfs setstripe understand 'k', 'm' and 'g' for stripe size.
2130 Frequency : mds/oss recovery
2132 Description: llog ctxt is refrenced after it has been freed.
2133 Details : llog ctxt refcount was added to avoide the race between ctxt free
2134 and llog recovery process. Each llog user must hold ctxt refcount
2135 before it access the llog. And the llog ctxt can only be freed
2136 when its refcount is zero.
2138 Severity : enhancement
2140 Description: randomly memory allocation failure util
2141 Details : Make lustre randomly failed allocating memory for testing purpose.
2144 Frequency : only for SLES10
2146 Description: Update kernel patch for SLES10 SP1
2147 Details : Add patch blkdev_tunables-2.6-sles10.patch to 2.6-sles10.series.
2149 Severity : enhancement
2151 Description: omit set fsid for export NFS
2152 Details : fix set/restore device id for avoid EMFILE error and mark lustre fs
2153 as FS_REQUIRES_DEV for avoid problems with generate fsid.
2156 Frequency : after network failures
2158 Description: Add sync option to mount_lustre.c
2159 Details : Client loses data written to lustre after a network interruption.
2161 Severity : enhancement
2163 Description: Error message improvement.
2164 Details : Merging of two LCONSOLE_ERROR_MSG into one.
2166 Severity : enhancement
2168 Description: don't use GFP_* in generic Lustre code.
2169 Details : Use cfs_alloc_* functions and CFS_* flags for code portability.
2171 Severity : enhancement
2173 Description: obdclass is limited by single OBD_ALLOC(idarray)
2174 Details : replace OBD_ALLOC/OBD_FREE with OBD_VMALLOC/OBD_VFREE
2176 Severity : enhancement
2178 Description: updated patchess for new RHEL4 kernel
2179 Details : Fixed ext3-unlink-race.patch per Kalpak's comment.
2181 Severity : enhancement
2183 Description: warnings with build patchless client with vanila 2.6.19 and up
2184 Details : change old ctl_table style and replace ctl_table/ctl_table_header
2185 with cfs_sysctl_table_t/cfs_sysctl_table_header_t
2187 Severity : enhancement
2189 Description: O_DIRECT bypasses client statistics.
2190 Details : When running with O_DIRECT I/O, neither the client rpc_stats nor
2191 read_ahead_stats were updated. Copied stats section from
2192 osc_send_oap_rpc() into async_internal().
2195 Frequency : only for Cray XT3
2197 Description: peer credits not enough on many OST per OSS systems.
2198 Details : Use new lnet way to add credits as we need those for pings and ASTs
2201 Frequency : only with liblustre
2203 Description: Liblustre is not releasing flock locks on file close.
2204 Details : Release flock locks on file close.
2207 Frequency : only for RHEL4
2209 Description: Update kernel patches for kernel-2.6.9-55.0.2.EL
2210 Details : Remove inode-nr_unused-2.6.9-rhel4.patch from 2.6-rhel4.series
2211 Update target file and kernel config.
2215 Description: lustre support for RHEL5
2216 Details : Add support for RHEL5.
2221 Description: ASSERTION(export != NULL) failed in target_handle_connect
2222 Details : Assetion hit is result of rare race between disconnect and connet
2223 to same nid. target_handle_connect found old connect cockie and
2224 tried to reconnect, but can't find export for this cockie.
2229 Description: umount blocks forever on error
2230 Details : In result of wrong using obd_no_recov and obd_force flags client
2231 can hand if cancel or some other requests is lost.
2234 Frequency : Only for SLES
2236 Description: sanity_quota fail test_1
2237 Details : There are multiple occurences of $TSTUSR in SLES's /etc/group
2238 file, which makes TSTID[2] inunique.
2240 Severity : enhancement
2242 Description: Kernel patches for SLES9 2.6.5-7.286 kernel
2243 Details : Update target/ChangeLog/which_patch .
2245 Severity : enhancement
2247 Description: jbd statistics
2248 Details : Port older jbd statistics patch for sles10
2253 Description: lvbo_init failed for resource with missing objects.
2254 Details : Fix returning error if we do stat for file with missing/corrupted
2255 objects and i_size set to all sum of size of all avaible objects.
2256 if we to truncate/write to missing object - it is recreated.
2261 Description: Oops in read and write path when failing to allocate lock.
2262 Details : Check if lock allocation failed and return error back.
2265 Frequency : When flocks are used.
2267 Description: assertion failure in ldlm_cli_enquque_fini for non NULL lock.
2268 Details : Flock locks might destroy just granted lock if it could be merged
2269 with another existing flock, this is done in completion handler,
2270 so teach ldlm_cli_enquque_fini that this is a valid case for
2276 Description: reply_lock_interpret crash due to race with it and lock cancel.
2277 Details : Do not replay locks that are being cancelled. Do not reference
2278 locks by their address during replay, just by their handle.
2281 Frequency : only with deactivated OSTs
2283 Description: lstripe command fails for valid OST index
2284 Details : The stripe offset is compared to 'lov->desc.ld_tgt_count'
2285 instead of lov->desc.ld_active_tgt_count.
2287 Severity : enhancement
2289 Description: Build failure against Centos5 (RHEL5)
2290 Details : Use getpagesize() instead of PAGE_SIZE.
2292 --------------------------------------------------------------------------------
2294 2007-07-30 Cluster File Systems, Inc. <info@clusterfs.com>
2296 * Support for kernels:
2297 2.6.5-7.283 (SLES 9),
2298 2.6.9-55.EL (RHEL 4),
2299 2.6.16.46-0.14 (SLES 10),
2300 2.6.18.8 vanilla (kernel.org)
2301 * Client support for unpatched kernels:
2302 (see http://wiki.lustre.org/index.php?title=Patchless_Client)
2303 2.6.16 - 2.6.22 vanilla (kernel.org)
2304 * Due to recently discovered recovery problems, we do not recommend
2305 using patchless RHEL 4 clients with this or any earlier release.
2306 * Recommended e2fsprogs version: 1.39.cfs8
2307 * Note that reiserfs quotas are disabled on SLES 10 in this kernel.
2308 * Starting with this release, the ldiskfs backing filesystem required
2309 by Lustre is now in its own package, lustre-ldiskfs. This package
2310 should be installed. It is versioned separately from Lustre and
2311 may be released separately in future.
2316 Description: server LBUG when shutting down
2317 Details : Block umount forever until the mount refcount is zero rather
2318 than giving up after an arbitrary timeout.
2320 Severity : enhancement
2322 Description: add optional extra BUILD_VERSION info
2323 Details : add a new environment variable (namely LUSTRE_VERS) which allows
2324 to override the lustre version.
2327 Frequency : 2.6.18 servers only
2329 Description: ll_kern_mount() doesn't release the module reference
2330 Details : The ldiskfs module reference count never drops down to 0
2331 because ll_kern_mount() doesn't release the module reference.
2336 Description: server LBUG when using old ost_num_threads parameter
2337 Details : Accept the old ost_num_threads parameter but warn that it
2338 is deprecated, and fix an off-by-one error that caused an LBUG.
2343 Description: Transient SCSI error results in persistent IO issue
2344 Details : iobuf->dr_error is not reinitialized to 0 between two
2348 Frequency : sometimes when underlying device returns I/O errors
2350 Description: OSTs not going read-only during write failures
2351 Details : OSTs are not remounted read-only when the journal commit threads
2352 get I/O errors because fsfilt_ext3 calls journal_start/stop()
2353 instead of the ext3 wrappers.
2356 Frequency : only with 10000 clients or more
2358 Description: poor connect scaling with increasing client count
2359 Details : Don't run filter_grant_sanity_check for more than 100 exports
2360 to improve scaling for large numbers of clients.
2363 Frequency : SLES10 only
2365 Description: sanity-quota.sh quotacheck failed: rc = -22
2366 Details : Quotas cannot be enabled on SLES10.
2369 Frequency : liblustre clients only
2371 Description: getdirentries does not give error when run on compute nodes
2372 Details : getdirentries does not fail when the size specified as an argument
2373 is too small to contain at least one entry
2375 Severity : enhancement
2377 Description: Add LNET router traceability for debug purposes
2378 Details : If a checksum failure occurs with a router as part of the
2379 IO path, the NID of the last router that forwarded the bulk data
2380 is printed so it can be identified.
2385 Description: OST "spontaneously" evicts client; client has imp_pingable == 0
2386 Details : Due to a race condition, liblustre clients were occasionally
2387 evicted incorrectly.
2389 Severity : enhancement
2391 Description: lfs setstripe use optional parameters instead of postional
2394 Severity : enhancement
2396 Description: Nanosecond timestamp support for ldiskfs
2397 Details : The on-disk ldiskfs filesystem has added support for nanosecond
2398 resolution timestamps. There is not yet support for this at
2399 the Lustre filesystem level.
2402 Frequency : during server recovery
2404 Description: MDS failing to send precreate requests due to OSCC_FLAG_RECOVERING
2405 Details : request with rq_no_resend flag not awake l_wait_event if they get a
2409 Frequency : nfs export on patchless client
2411 Description: connectathon hang when test nfs export over patchless client
2412 Details : Disconnected dentry cannot be found with lookup, so we do not need
2413 to unhash it or make it invalid
2416 Description: fix llapi_lov_get_uuids() to allow many OSTs to be returned
2417 Details: : Change llapi_lov_get_uuids() to read the UUIDs from /proc instead
2418 of using an ioctl. This allows lfsck for > 160 OSTs to succeed.
2423 Description: open req refcounting wrong on reconnect
2424 Details : If reconnect happened between getting open reply from server and
2425 call to mdc_set_replay_data in ll_file_open, we will schedule
2426 replay for unreferenced request that we are about to free.
2427 Subsequent close will crash in variety of ways.
2428 Check that request is still eligible for replay in
2429 mdc_set_replay_data().
2434 Description: disable writes to filesystem when reading health_check file
2435 Details : the default for reading the health_check proc file has changed
2436 to NOT do a journal transaction and write to disk, because this
2437 can cause reads of the /proc file to hang and block HA state
2438 checking on a healthy but otherwise heavily loaded system. It
2439 is possible to return to the previous behaviour during configure
2440 with --enable-health-write.
2442 Severity : enhancement
2444 Description: 64-bit inode version
2445 Details: : Add a on-disk 64-bit inode version for ext3 to track changes made
2446 to the inode. This will be required for version-based recovery.
2451 Description: MDS fails to start if a duplicate client export is detected
2452 Details : in some rare cases it was possible for a client to connect to
2453 an MDS multiple times. Upon recovery the MDS would detect this
2454 and fail during startup. Handle this more gracefully.
2456 Severity : enhancement
2458 Description: Add -o localflock option to simulate old noflock behaviour.
2459 Details : This will achieve local-only flock/fcntl locks coherentness.
2464 Description: log_commit_thread vs filter_destroy race leads to crash
2465 Details : Take import reference before releasing llog record semaphore
2470 Description: Wrong request locking in request set processing
2471 Details : ptlrpc_check_set wrongly uses req->rq_lock for proctect add to
2472 imp_delayed_list, in this place should be used imp_lock.
2475 Frequency : when reconnecting
2477 Description: Grant leak when OSC reconnect to OST
2478 Details : When osc reconnect ost, OST(filter) should check whether it
2479 should grant more space to client by comparing fed_grant and
2480 cl_avail_grant, and return the granted space to client instead
2481 of "new granted" space, because client will call osc_init_grant
2482 to update the client grant space info.
2485 Frequency : when client reconnects to OST
2487 Description: Grant leak when OSC does a resend and replays bulk write
2488 Details : When osc reconnect to OST, OST(filter) should clear grant info of
2489 bulk write request, because the grant info while be sync between
2490 OSC and OST when reconnect, and we should ignore the grant info
2491 these of resend/replay write req.
2496 Description: Grant space more than avaiable space sometimes.
2497 Details : When then OST is about to be full, if two bulk writing from
2498 different clients came to OST. Accord the avaliable space of the
2499 OST, the first req should be permitted, and the second one
2500 should be denied by ENOSPC. But if the seconde arrived before
2501 the first one is commited. The OST might wrongly permit second
2502 writing, which will cause grant space > avaiable space.
2505 Frequency : when client is evicted
2507 Description: Grant might be wrongly erased when osc is evicted by OST
2508 Details : when the import is evicted by server, it will fork another
2509 thread ptlrpc_invalidate_import_thread to invalidate the
2510 import, where the grant will be set to 0. While the original
2511 thread will update the grant it got when connecting. So if
2512 the former happened latter, the grant will be wrongly errased
2513 because of this race.
2518 Description: Checking Stale with correct fid
2519 Details : ll_revalidate_it should uses de_inode instead of op_data.fid2
2520 to check whether it is stale, because sometimes, we want the
2521 enqueue happened anyway, and op_data.fid2 will not be initialized.
2523 Severity : enhancement
2525 Description: update patchless client
2526 Details : Add support for patchless client with 2.6.20, 2.6.21 and RHEL 5
2529 Frequency : only with 2.4 kernel
2531 Description: random memory corruption
2532 Details : size of struct ll_inode_info is to big for union inode.u and this
2533 can be cause of random memory corruption.
2538 Description: Memory leak in recovery
2539 Details : Lov_mds_md was not free in an error handler in mds_create_object.
2540 It should also check obd_fail before fsfilt_start, otherwise if
2541 fsfilt_start return -EROFS,(failover mds during mds recovery).
2542 then the req will return with repmsg->transno = 0 and rc = EROFS.
2543 and we met hit the assert LASSERT(req->rq_reqmsg->transno ==
2544 req->rq_repmsg->transno) in ptlrpc_replay_interpret. Fcc should
2545 be freed no matter whether fsfilt_commit success or not.
2548 Frequency : only with large numbers of cores on a single node
2550 Description: Prevents from taking the superblock lock in llap_from_page for
2552 Details : using LL_ORIGIN_REMOVEPAGE origin flag instead of LL_ORIGIN_UNKNOW
2553 for llap_from_page call in ll_removepage() prevents from taking
2554 the superblock lock for a soon died page.
2559 Description: Not check open intent error before release open handle
2560 Details : in some rare cases, the open intent error is not checked before
2561 release open handle, which may cause
2562 ASSERTION(open_req->rq_transno != 0), because it tries to release
2563 the failed open handle.
2568 Description: Set cat log bitmap only after create log success.
2569 Details : in some rare cases, the cat log bitmap is set too early. and it
2570 should be set only after create log success.
2574 Description: Accessing a block bevice can re-enable I/O when Lustre is
2575 tearing down a device.
2576 Details : dev_clear_rdonly(bdev) must be called in kill_bdev() instead of
2581 Description: service threads may hog cpus when there are a lot of requests
2582 Details : Insert cond_resched to give other threads a chance to use some CPU
2587 Description: the cat log was not initialized in recovery
2588 Details : When mds(mgs) do recovery, the tgt_count might be zero, so the
2589 unlink log on mds will not be initialized until mds post
2590 recovery. And also in mds post recovery, the unlink log will
2591 initialization will be done asynchronausly, so there will be race
2592 between add unlink log and unlink log initialization.
2596 Description: brw_stats were being printed incorrectly
2597 Details : brw_stats were being printed as log2 but all of them were not
2598 recorded as log2. Also remove some code duplication arising from
2599 filter_tally_{read,write}.
2603 Frequency : rare, only in recovery.
2604 Description: ASSERTION(req->rq_type != LI_POISON) failed
2605 Details : imp_lock should be held while iterating over imp_sending_list for
2606 prevent destroy request after get timeout in ptlrpc_queue_wait.
2610 Description: replay-single.sh test 52 fails
2611 Details : A lock's skiplist need to be cleanup when it being unlinked
2612 from its resource list.
2617 Description: Short directio read returns full requested size rather than
2619 Details : Direct I/O operations should return actual amount of bytes
2620 transferred rather than requested size.
2622 Severity : enhancement
2624 Description: metadata RPC reduction (e.g. for rm performance)
2625 Details : decrease the amount of synchronous RPC between clients and servers
2626 by canceling conflicing lock before the operation on the client
2627 and packing thier handles into the main operation RPC to server.
2629 Severity : enhancement
2631 Description: add #ifdef HAVE_KERNEL_CONFIG_H
2632 Details : kernels from 2.6.19 not need include linux/config.h, but add
2633 include linux/autoconf.h in commpiler command line.
2635 Severity : enhancement
2637 Description: patchless client support for 2.6.22 kernel
2638 Details : 2.6.22 has only one visble change, SLAB_CTOR_* constants is
2639 removed. In this case we need drop using os depended interface to
2640 kmem_cache and use cfs_mem_cache API.
2645 Description: fix mal-formatted messages
2646 Details : fix some mal-formatted DEBUG_REQ and LCONSOLE_ERROR_MSG messages
2650 Frequency : always in liblustre
2651 Description: wrong IS_ERR implementation in liblustre.h
2652 Details : fix IS_ERR implementation in liblustre.h for right detect errors.
2657 Description: Correct condition for output debug message.
2658 Details : inode i_nlink equal zero is not enough for output message about
2659 disk corruption, i_ctime and i_mode should be also checked.
2663 Frequency : always in patchless client
2664 Description: add configure check for truncate_complete_page
2665 Details : improve checks for exported symbols. This allow run check without
2666 sources, but with Module.symvers shipped with kernel distribution.
2667 add check for truncate_complete_page used by patchless client.
2671 Description: sanity.sh test_77h fails with "test_77h file compare failed"
2672 Details : test_77h uses a file which was messed by other test case.
2676 Description: Not Check whether lov_tgts is NULL in some lov functions
2677 Details : Checking whether lov_tgts is NULL in some functions.
2681 Description: replace obdo_alloc() with OBDO_ALLOC macro
2682 Details : nothing special is done in obdo_alloc() function, and for
2683 debugging purpose, it needs to be replaced with macros.
2687 Description: bad return value and errno from fcntl call
2688 Details : In liblustre API, errno should be a negative value if error
2693 Description: ptlrpc_check_set() LBUG
2694 Details : In case of positive reply from server and failed client bulk
2695 callback after bulk transfer shouldn't LBUG, but process this
2696 request as erroneous.
2698 Severity : enhancement
2700 Description: VFS operations stats tool.
2701 Details : Tool which collects stats by tracking value written in pid,
2702 ppid, gid and uses llstat to generate output to plot graph using
2704 Updated lustre/utils/Makefile.am
2705 Added lustre/utils/ltrack_stats.c
2707 Severity : enhancement
2709 Description: 2.6.18 server support (lustre 1.6.1)
2710 Details : Support for 2.6.18 kernels on the server side.
2715 Description: ASSERTION(imp->imp_conn_current) failed
2716 Details : an assertion failure is hit if a client node boots and attempts to
2717 mount a lustre filesystem faster than RECONNECT_INTERVAL seconds.
2720 Frequency : only for i686
2722 Description: 1.4.11 RC1 build fails for RHEL 4, i686
2723 Details : Fixed config variable for build.
2728 Description: Updated patchess for new RHEL4 kernel
2729 Details : Updated patch inode-nr_unused-2.6.9-rhel4.patch
2730 Updated patch jbd-stats-2.6.9.patch
2731 Updated patch qsnet-rhel4-2.6.patch
2732 Updated patch quota-deadlock-on-pagelock-core.patch
2733 Updated patch vfs_intent-2.6-rhel4.patch
2734 Updated patch vfs_races-2.6-rhel4.patch
2735 Updated series file 2.6-rhel4-titech.series
2736 Updated series file 2.6-rhel4.series
2737 Updated kernel config files
2742 Description: lquota slave complains LBUG when reconnecting with mds
2744 Details : quota slave depends qctxt->lqc_import to send its quota request.
2745 This pointer will be invalid if mds did failover or broke its
2746 connect to osts, which leads to LBUG.
2749 Frequency : when qunit size is too small(less than 20M)
2751 Description: write is stopped by improper -EDQUOT
2752 Details : If the master is busy and qunit size is small enough(let's say 1M),
2753 the slave can not get quota from master on time, which will lead
2754 slave to trigger a -EQUOTA to client.
2759 Description: Deadlock during metadata tests
2760 Details : in prune_dir_dentries(), shrink_dcache_parent() should not be
2761 called with the per-dentry lock held.
2764 Frequency : SLES9 only
2766 Description: Lustre patched kernel for SLES9 SP3 has NR_CPUS set to 8
2767 Details : set CONFIG_NR_CPUS to 128 instead of 8.
2769 Severity : enhancement
2771 Description: remove fs_prep_san_write operation and related patches
2772 Details : remove the ext3-san-jdike patches which are no longer useful.
2777 Description: LDISKFS-fs error (device sdc): ldiskfs_free_blocks
2778 Details : a disk corruption can cause the mballoc code to assert on a
2779 double free or other extent corruptions. Handle these with
2780 ext3_error() instead of assertions.
2783 Frequency : only with mballoc3 code and deep extent trees
2785 Description: ldiskfs_ext_search_right: bad header in inode: unexpected eh_depth
2786 Details : a wrong check of extent headers in ldiskfs_ext_search_right()
2787 can cause the filesystem to be remounted read-only.
2792 Description: lfsck built against 1.4.x cannot run against 1.6.0 lustre
2793 Details : the definition for OBD_IOC_GETNAME changed in 1.6.0. One of the
2794 few external users of this ioctl number is lfsck's call to
2795 llapi_lov_get_uuids() and this caused lfsck to fail at startup.
2796 Add the old ioctl number to the handler so both old and new
2801 Description: parallel lock callbacks
2802 Details : Instead of sending blocking and completion callbacks as separated
2803 requests, adding them to a set and sending in parallel.
2807 Description: Disable most debugging by default
2808 Details : To improve performance, disable most logging (for debug purposes)
2809 by default. VFSTRACE, RPCTRACE, and DLMTRACE are now off by
2810 default, and HA includes fewer messages.
2814 Frequency : only run on patchless client.
2815 Description: use do_facet on sanity.sh for test handling recoverables errors
2816 Details : use do_facet instead of direct use sysctl for set fail_loc on OST
2820 Description: hash tables for lists of nids, connections and uuids
2821 Details : Hash tables noticeably help when a lot of clients connect to a
2822 server, to faster identify duplicate connections or reconnects,
2823 also to faster find export to evict in manual eviction case.
2827 Description: Sometimes, when the server evict a client, and the client will
2828 not be evicted as soon as possible.
2829 Details : In enqueue req, the error was returned by intent, instead of
2830 rq_status which make ptlrpc layer not detect this error, and
2831 does not evict the client. So enqueue error should be returned
2835 Frequency : only at startup
2837 Description: Delay client connections to MDT until fist MDT->OST connect
2838 Details : If a client tried to create a new file before the MDT had
2839 connected to any OSTs, the create would return EIO. Now
2840 the client will simply block until the MDT connects to the
2841 first OST and the create can succeed.
2844 Frequency : only for SLES9
2846 Description: Routinely utilize latest Quadrics drivers in CFS releases
2847 Details : Update patch qsnet-suse-2.6.patch.
2850 Frequency : only for sles10
2852 Description: Update patches for SLES 10 SP1 kernel.
2853 Details : Update the patch vfs_intent-2.6-sles10.patch.
2858 Description: Routinely utilize latest Quadrics drivers in CFS releases
2859 Details : Update patch qsnet-rhel4-2.6.patch.
2862 Frequency : at statup only
2864 Description: mds_lov_synchronize race leads to various problems
2865 Details : simultaneous MDT->OST connections at startup can cause the
2866 sync to abort, leaving the OSC in a bad state.
2871 Description: Using wrong pointer in osc_brw_prep_request
2872 Details : Access to array[-1] can produce panic if kernel compiled with
2873 CONFIG_PAGE_ALLOC enabled
2875 Severity : enhancement
2877 Description: Async OSC create to avoid the blocking unnecessarily.
2878 Details : If a OST has no remain object, system will block on the creating
2879 when need to create a new object on this OST. Now, ways use
2880 pre-created objects when available, instead of blocking on an
2881 empty osc while others are not empty. If we must block, we block
2882 for the shortest possible period of time.
2886 Frequency : only in recovery
2887 Description: Mark OST as early accessible if his start SYNC.
2888 Details : osc_precreate return flag early accessible if oscc marked as
2889 OSCC_FLAG_SYNC_IN_PROGRESS.
2894 Description: Sometimes precreate code can triger create object on wrong ost
2895 Details : Wrong protected or not not restored variables aftre precreate loop
2896 can produce creation object on wrong ost.
2899 Frequency : oss recovery
2901 Description: llog_commit_thread cleanup should sync with llog_commit_thread
2903 Details : llog_commit_thread_count should be synced between llog_commit
2904 start and cleanup, so new llog_commit thread should not be started
2905 when llog_commit threads being stopped to avoid accessing some
2908 Severity : enhancement
2910 Description: Add printing inode info into message about error in writepage.
2912 --------------------------------------------------------------------------------
2914 2007-05-03 Cluster File Systems, Inc. <info@clusterfs.com>
2919 Frequency : on some architectures
2921 Description: 1.6 client sometimes fails to mount from a 1.4 MDT
2922 Details : Uninitialized flags sometimes cause configuration commands to
2926 Frequency : patchless clients only
2928 Description: missing __iget() symbol export
2929 Details : The __iget() symbol export is missing. To avoid the need for
2930 this on patchless clients the deathrow inode reaper is turned
2931 off, and we depend on the VM to clean up old inodes. This
2932 dependency was during via the fix for bug 12181.
2937 Description: sanity.sh fail: test_52b
2938 Details : The ll_inode_to_ext_flags() has a glitch which makes MDS return
2939 incorrect inode's flags to client.
2941 --------------------------------------------------------------------------------
2943 2007-04-19 Cluster File Systems, Inc. <info@clusterfs.com>
2945 * CONFIGURATION CHANGE. This version of Lustre WILL NOT
2946 INTEROPERATE with older versions automatically. In many cases a
2947 special upgrade step is needed. Please read the
2948 user documentation before upgrading any part of a 1.4.x system.
2949 * WARNING: Lustre configuration and startup changes are required with
2950 this release. See https://mail.clusterfs.com/wikis/lustre/MountConf
2952 * Support for kernels:
2953 2.4.21-47.0.1.EL (RHEL 3),
2954 2.6.5-7.283 (SLES 9),
2955 2.6.9-42.0.10.EL (RHEL 4),
2956 2.6.12.6 vanilla (kernel.org),
2957 2.6.16.27-0.9 (SLES10)
2958 * Client support for unpatched kernels:
2959 (see https://mail.clusterfs.com/wikis/lustre/PatchlessClient)
2960 2.6.16 - 2.6.19 vanilla (kernel.org),
2961 2.6.9-42.0.8EL (RHEL 4)
2962 * Recommended e2fsprogs version: 1.39.cfs6
2963 * Note that reiserfs quotas are disabled on SLES 10 in this kernel
2966 Severity : enhancement
2968 Description: Async OSC create to avoid the blocking unnecessarily.
2969 Details : If a OST has no remain object, system will block on the creating
2970 when need to create a new object on this OST. Now, ways use
2971 pre-created objects when available, instead of blocking on an
2972 empty osc while others are not empty. If we must block, we block
2973 for the shortest possible period of time.
2975 Severity : enhancement
2977 Description: MountConf
2978 Details : Lustre configuration is now managed via mkfs and mount
2979 commands instead of lmc and lconf. New obd types (MGS, MGC)
2980 are added for dynamic configuration management. See
2981 https://mail.clusterfs.com/wikis/lustre/MountConf for
2984 Severity : enhancement
2986 Description: dynamic OST addition
2987 Details : OSTs can now be added to a live filesystem
2989 Severity : enhancement
2991 Description: startup order invariance
2992 Details : MDTs and OSTs can be started in any order. Clients only
2993 require the MDT to complete startup.
2995 Severity : enhancement
2997 Description: parallel, asynchronous orphan cleanup
2998 Details : orphan cleanup is now performed in separate threads for each
2999 OST, allowing parallel non-blocking operation.
3001 Severity : enhancement
3003 Description: optimized stripe assignment
3004 Details : stripe assignments are now made based on ost space available,
3005 ost previous usage, and OSS previous usage, in order to try
3006 to optimize storage space and networking resources.
3008 Severity : enhancement
3010 Description: Permanently set tunables
3011 Details : All writable /proc/fs/lustre tunables can now be permanently
3012 set on a per-server basis, at mkfs time or on a live
3015 Severity : enhancement
3017 Description: Lustre message v2
3018 Details : Add lustre message format v2.
3020 Severity : enhancement
3022 Description: client OST exclusion list
3023 Details : Clients can be started with a list of OSTs that should be
3024 declared "inactive" for known non-responsive OSTs.
3028 Description: ENOENT returned for valid filehandle during dbench.
3029 Details : Check if a directory has children when invalidating dentries
3030 associated with an inode during lock cancellation. This fixes
3031 an incorrect ENOENT sometimes seen for valid filehandles during
3032 testing with dbench.
3035 Frequency : SFS test only (otherwise harmless)
3037 Description: SPEC SFS validation failure on NFS v2 over lustre.
3038 Details : Changes the blocksize for regular files to be 2x RPC size,
3039 and not depend on stripe size.
3041 Severity : enhancement
3043 Description: fine-grained SMP locking inside DLM
3044 Details : Improve DLM performance on SMP systems by removing the single
3045 per-namespace lock and replace it with per-resource locks.
3047 Severity : enhancement
3049 Description: don't hold multiple extent locks at one time
3050 Details : To avoid client eviction during large writes, locks are not
3051 held on multiple stripes at one time or for very large writes.
3052 Otherwise, clients can block waiting for a lock on a failed OST
3053 while holding locks on other OSTs and be evicted.
3055 Severity : enhancement
3057 Description: Multiple MD RPCs in flight.
3058 Details : Further unserialise some read-only MDT RPCs - learn about intents.
3059 To avoid overly-overloading MDT, introduce a limit on number of
3060 MDT RPCs in flight for a single client and add /proc controls
3061 to adjust this limit.
3063 Severity : enhancement
3065 Description: client read/write statistics
3066 Details : Add client read/write call usage stats for performance
3067 analysis of user processes.
3068 /proc/fs/lustre/llite/*/offset_stats shows non-sequential
3069 file access. extents_stats shows chunk size distribution.
3070 extents_stats_per_process show chunk size distribution per
3073 Severity : enhancement
3075 Description: per-client statistics on server
3076 Details : Add ldlm and operations statistics for each client in
3077 /proc/fs/lustre/mds|obdfilter/*/exports/
3079 Severity : enhancement
3081 Description: improved MDT statistics
3082 Details : Add detailed MDT operations statistics in
3083 /proc/fs/lustre/mds/*/stats
3085 Severity : enhancement
3087 Description: VFS operations stats
3088 Details : Add client VFS call stats, trackable by pid, ppid, or gid
3089 /proc/fs/lustre/llite/*/stats_track_[pid|ppid|gid]
3094 Description: Fix client-side osc byte counters
3095 Details : The osc read/write byte counters in
3096 /proc/fs/lustre/osc/*/stats are now working
3099 Frequency : always as root on SLES
3101 Description: Failure of copying files with lustre special EAs.
3102 Details : Client side always return success for setxattr call for lustre
3103 special xattr (currently only "trusted.lov").
3108 Description: Refcount LNET uuids
3109 Details : The global LNET uuid list grew linearly with every startup;
3110 refcount repeated list entries instead of always adding to
3113 Severity : enhancement
3115 Description: Dynamic service threads
3116 Details : Within a small range, start extra service threads
3117 automatically when the request queue builds up.
3120 Frequency : mixed-endian client/server environments
3122 Description: mixed-endian crashes
3123 Details : The new msg_v2 system had some failures in mixed-endian
3126 Severity : enhancement
3128 Description: Easy OST removal
3129 Details : OSTs can be permanently deactivated with e.g. 'lctl
3130 conf_param lustre-OST0001.osc.active=0'
3132 Severity : enhancement
3134 Description: MGS proc entries
3135 Details : Added basic proc entries for the MGS showing what filesystems
3138 Severity : enhancement
3140 Description: provide MGS failover
3141 Details : Added config lock reacquisition after MGS server failover.
3143 Severity : enhancement
3145 Description: add Linux 2.4 support
3146 Details : Added support for RHEL 2.4.21 kernel for 1.6 servers and clients
3150 Description: a large application tries to do I/O to the same resource and dies
3151 in the middle of it.
3152 Details : Check the req->rq_arrival time after the call to
3153 ost_brw_lock_get(), but before we do anything about
3154 processing it & sending the BULK transfer request. This
3155 should help move old stale pending locks off the queue as
3156 quickly as obd_timeout.
3159 Frequency : when an incorrect nid is specified during startup
3161 Description: ptlrpc connect to non-existant node causes kernel crash
3162 Details : LNET can't be re-entered from an event callback, which
3163 happened when we expire a message after the export has been
3164 cleaned up. Instead, hand the zombie cleanup off to another
3167 Severity : enhancement
3169 Description: plain/inodebits lock performance improvement
3170 Details : Grouping plain/inodebits in granted list by their request modes
3171 and bits policy, thus improving the performance of search through
3175 Frequency : only if OST filesystem is corrupted
3177 Description: client incorrectly hits assertion in ptlrpc_replay_req()
3178 Details : for a short time RPCs with bulk IO are in the replay list,
3179 but replay of bulk IOs is unimplemented. If the OST filesystem
3180 is corrupted due to disk cache incoherency and then replay is
3181 started it is possible to trip an assertion. Avoid putting
3182 committed RPCs into the replay list at all to avoid this issue.
3185 Frequency : liblustre (e.g. catamount) on a large cluster with >= 8 OSTs/OSS
3187 Description: System hang on startup
3188 Details : This bug allowed the liblustre (e.g. catamount) client to
3189 return to the app before handling all startup RPCs. This
3190 could leave the node unresponsive to lustre network traffic
3191 and manifested as a server ptllnd timeout.
3193 Severity : enhancement
3195 Description: Add "/proc/sys/lustre/debug_peer_on_timeout"
3196 Details : liblustre envirable: LIBLUSTRE_DEBUG_PEER_ON_TIMEOUT
3197 boolean to control whether to print peer debug info when a
3198 client's RPC times out.
3201 Frequency : only for kernels with patches from Lustre below 1.4.3
3203 Description: Remove old rdonly API
3204 Details : Remove old rdonly API which unused from at least lustre 1.4.3
3207 Frequency : only for devices with external journals
3209 Description: Set external device read-only also
3210 Details : During a commanded failover stop, we set the disk device
3211 read-only while the server shuts down. We now also set any
3212 external journal device read-only at the same time.
3215 Frequency : when upgrading from 1.4 while trying to change parameters
3217 Description: The wrong (new) MDC name was used when setting parameters for
3218 upgraded MDT's. Also allows changing of OSC (and MDC)
3219 parameters if --writeconf is specified at tunefs upgrade time.
3222 Frequency : when setting specific ost indicies
3224 Description: QOS code breaks on skipped indicies
3225 Details : Add checks for missing OST indicies in the QOS code, so OSTs
3226 created with --index need not be sequential.
3228 Severity : enhancement
3230 Description: Add uninit_groups feature to ldiskfs2 to speed up e2fsck
3231 Details : The uninit_groups feature works in conjunction with the kernel
3232 filesystem code (ldiskfs2 only) and e2fsprogs-1.39-cfs6 to speed
3233 up the pass1 processing of e2fsck. This is a read-only feature
3234 in ldiskfs2 only, so older kernels and current ldiskfs cannot
3235 mount filesystems that have had this feature enabled.
3237 Severity : enhancement
3239 Description: Improve multi-block allocation algorithm to avoid fragmentation
3240 Details : The mballoc3 code (ldiskfs2 only) adds new mechanisms to improve
3241 allocation locality and avoid filesystem fragmentation.
3243 ------------------------------------------------------------------------------
3245 2007-04-01 Cluster File Systems, Inc. <info@clusterfs.com>
3247 * Support for kernels:
3248 2.4.21-47.0.1.EL (RHEL 3)
3249 2.6.5-7.283 (SLES 9)
3250 2.6.9-42.0.10.EL (RHEL 4)
3251 2.6.12.6 vanilla (kernel.org)
3252 2.6.16.27-0.9 (SLES 10)
3253 * Recommended e2fsprogs version: 1.39.cfs5
3255 * Note that reiserfs quotas are disabled on SLES 10 in this kernel
3259 Frequency : occasional, depends on client load and configuration
3260 Bugzilla : 12181, 12203
3261 Description: data loss for recently-modified files
3263 Details : In some cases it is possible that recently written or created
3264 files may not be written to disk in a timely manner (this should
3265 normally be within 30s unless client IO load is very high).
3266 The problem appears as zero-length files or files that are a
3267 multiple of 1MB in size after a client crash or client eviction
3268 that are missing data at the end of the file.
3270 This problem is more likely to be hit on clients where files are
3271 repeatedly created and unlinked in the same directory, clients
3272 have a large amount of RAM, have many CPUs, the filesystem has
3273 many OSTs, the clients are rebooted frequently, and/or the files
3274 are not accessed by other nodes after being written.
3276 The presence of the problem can be detected by looking at
3277 /proc/sys/fs/inode-state. If the first number (nr_inodes) is
3278 smaller than the second (nr_unused) then dirty files will not
3279 be flushed automatically to disk. "sync; sleep 10" should be
3280 run several times on the node before unmounting it to update
3281 Lustre (this is also safe to run on nodes without this problem).
3283 There is also a related kernel bug in the RHEL4 4 2.6.9 kernel
3284 that can cause this same problem, so customers using that kernel
3285 also need to update the kernel in addition to Lustre. In order
3286 to properly fix this bug, the RHEL3 2.4.21 kernel is also updated.
3288 It is normal that files written just before a client crash (less
3289 than 30s) may not yet have been flushed to disk, even for local
3293 Frequency : frequent on thin XT3 nodes
3295 Description: UUID collision on thin XT3 Linux nodes
3296 Details : UUIDs on Compute Node Linux XT3 nodes were not generated
3297 randomly, since we relied on an insufficiently-seeded PRNG.
3302 Description: OSS hangs after "All ost request buffers busy"
3303 Details : A deadlock between quota and journal operations caused OSS
3304 hangs after printing "All ost request buffers busy."
3307 Frequency : always on liblustre builds
3309 Description: Cleanup compiler warnings on liblustre
3312 Frequency : always on liblustre builds on XT3
3314 Description: LC_CONFIG_CDEBUG don't run while build liblustre on XT3.
3318 Description: Addition of EXT3_FEATURE_RO_COMPAT_DIR_NLINKS flag for
3319 > 32000 subdirectories
3320 Details : Add EXT3_FEATURE_RO_COMPAT_DIR_NLINK flag to
3321 EXT3_FEATURE_RO_COMPAT_SUPP. This flag will be set whenever
3322 subdirectory count crosses 32000. This will aid e2fsck to
3323 correctly handle more than 32000 subdirectories.
3326 Frequency : liblustre (e.g. catamount) on a large cluster with >= 8 OSTs/OSS
3328 Description: System hang on startup
3329 Details : This bug allowed the liblustre (e.g. catamount) client to
3330 return to the app before handling all startup RPCs. This
3331 could leave the node unresponsive to lustre network traffic
3332 and manifested as a server ptllnd timeout.
3334 Severity : enhancement
3336 Description: Add "/proc/sys/lustre/debug_peer_on_timeout"
3337 (liblustre envirable: LIBLUSTRE_DEBUG_PEER_ON_TIMEOUT)
3338 boolean to control whether to print peer debug info when a
3339 client's RPC times out.
3344 Description: make O_SYNC working on 2.6 kernels
3345 Details : 2.6 kernels use different method for mark pages for write,
3346 so need add a code to lustre for O_SYNC work.
3351 Description: Failure to close file and release space on NFS
3352 Details : Put inode details into lock acquired in ll_intent_file_open.
3353 Use mdc_intent_lock in ll_intent_open to properly
3354 detect all kind of errors unhandled by mdc_enqueue.
3359 Description: proc file read during shutdown sometimes raced obd removal,
3361 Details : Add lock to prevent obd access after proc file removal.
3364 Frequency : Only for files larger than 4GB on 32-bit clients.
3366 Description: improperly doing page alignment of locks
3367 Details : Modify lustre core code to use CFS_PAGE_* defines instead of
3368 PAGE_*. Make CFS_PAGE_MASK a 64-bit mask.
3373 Description: RPCs being resent when they shouldn't be
3374 Details : Some RPCs that should not be resent are being resent. This
3375 can cause inconsistencies in the RPC state machine. Do not
3376 resend such requests.
3379 Frequency : rare, only with NFS export
3381 Description: Crash on NFS re-export node
3382 Details : under very unusual load conditions an assertion is hit in
3383 ll_intent_file_open()
3386 Frequency : only if OST filesystem is corrupted
3388 Description: client incorrectly hits assertion in ptlrpc_replay_req()
3389 Details : for a short time RPCs with bulk IO are in the replay list,
3390 but replay of bulk IOs is unimplemented. If the OST filesystem
3391 is corrupted due to disk cache incoherency and then replay is
3392 started it is possible to trip an assertion. Avoid putting
3393 committed RPCs into the replay list at all to avoid this issue.
3398 Description: large O_DIRECT requests fail under memory pressure/fragmentation
3399 Details : Large single O_DIRECT read and write calls can fail to allocate
3400 a sufficiently large buffer to process the request. In case of
3401 allocation failure the allocation is retried with a smaller
3402 buffer and broken into smaller requests.
3404 Severity : enhancement
3406 Description: Add -o localflock option to simulate old noflock behaviour.
3407 Details : This will achieve local-only flock/fcntl locks coherentness.
3412 Description: versioning check is incomplete
3413 Details : Checking the version difference of client vs. server, report
3414 error if the gap is too big.
3419 Description: add support PG_writeback bit
3420 Details : add support for PG_writeback bit for Lustre, for more carefull
3421 work with page cache in 2.6 kernel. This also fix some deadlocks
3422 and remove hack for work O_SYNC with 2.6 kernel.
3424 Severity : enhancement
3426 Description: Add uninit_groups feature to ldiskfs2 to speed up e2fsck
3427 Details : The uninit_groups feature works in conjunction with the kernel
3428 filesystem code (ldiskfs2 only) and e2fsprogs-1.39-cfs6 to speed
3429 up the pass1 processing of e2fsck. This is a read-only feature
3430 in ldiskfs2 only, so older kernels and current ldiskfs cannot
3431 mount filesystems that have had this feature enabled.
3433 Severity : enhancement
3435 Description: Improve multi-block allocation algorithm to avoid fragmentation
3436 Details : The mballoc3 code (ldiskfs2 only) adds new mechanisms to improve
3437 allocation locality and avoid filesystem fragmentation.
3439 ------------------------------------------------------------------------------
3441 2007-02-09 Cluster File Systems, Inc. <info@clusterfs.com>
3443 * Support for kernels:
3444 2.6.9-42.0.3.EL (RHEL 4)
3445 2.6.5-7.276 (SLES 9)
3446 2.4.21-47.0.1.EL (RHEL 3)
3447 2.6.12.6 vanilla (kernel.org)
3448 2.6.16.21-0.8 (SLES10)
3449 * Recommended e2fsprogs version: 1.39.cfs2-0
3451 * The backwards-compatible /proc/sys/portals symlink has been removed
3452 in this release. Before upgrading, please ensure that you change
3453 any configuration scripts or /etc/sysctl.conf files that access
3454 /proc/sys/portals/* or sysctl portals.* to use the corresponding
3455 entry in /proc/sys/lnet or sysctl lnet.*. This change can be made
3456 in advance of the upgrade on any system running Lustre 1.4.6 or
3457 newer, since /proc/sys/lnet was added in that version.
3458 * Note that reiserfs quotas are disabled on SLES 10 in this kernel
3462 Frequency : only when quota is used
3464 Description: avoid scanning export list for quota master
3465 Details : Change the algorithms to avoid scanning export list in order
3466 to improve the efficiency.
3469 Frequency : MDS failover only, very rarely
3471 Description: "went back in time" messages on mds failover
3472 Details : The greatest transno may be lost when the current operation
3473 finishes with an error (transno==0) and the client's last_rcvd
3474 record is over-written. Save the greatest transno in the
3475 mds_last_transno for this case.
3478 Frequency : always for specific kernels and striping counts
3480 Description: client may get "Matching packet too big" without ACL support
3481 Details : Clients compiled without CONFIG_FS_POSIX_ACL get an error message
3482 when trying to access files in certain configurations. The
3483 clients should in fact be denied when mounting because they do
3484 not understand ACLs.
3487 Frequency : Cray XT3 with more than 4000 clients and multiple jobs
3489 Description: many clients connecting with IO in progress causes connect timeouts
3490 Details : Avoid synchronous journal commits to avoid delays caused by many
3491 clients connecting/disconnecting when bulk IO is in progress.
3492 Queue liblustre connect requests on OST_REQUEST_PORTAL instead of
3493 OST_IO_PORTAL to avoid delays behind potentially many pending
3497 Frequency : occasionally with multiple writers to a single file
3499 Description: shared writes to file may result in wrong size reported by stat()
3500 Details : Allow growing of kms when extent lock is cancelled
3503 Frequency : always with random mmap IO to multi-striped file
3505 Description: mmap write might be lost if we are writing to a 'hole' in stripe
3506 Details : Only if the hole is at the end of OST object so that kms is too
3507 small. Fix is to increase kms accordingly in ll_nopage.
3510 Frequency : rare, only if OST filesystem is inconsistent with MDS filesystem
3512 Description: writes to a missing object would leak memory on the OST
3513 Details : If there is an inconsistency between the MDS and OST filesystems,
3514 such that the MDS references an object that doesn't exist, writes
3515 to that object will leak memory due to incorrect cleanup in the
3516 error handling path, eventually running out of memory on the OST.
3521 Description: Creating too long symlink causes lustre errors
3522 Details : Check symlink and name lengths before sending requests to MDS.
3525 Frequency : only if flock is enabled (not on by default)
3527 Description: posix locks not released on fd closure on 2.6.9+
3528 Details : We failed to add posix locks to list of inode locks on 2.6.9+
3529 kernels, this caused such locks not to be released on fd close and
3530 then assertions on fs unmount about still used locks.
3533 Frequency : MDS failover only, very rarely
3535 Description: clients may get ASSERTION(granted_lock != NULL)
3536 Details : When request was taking a long time, and a client was resending
3537 a getattr by name lock request. The were multiple lock requests
3538 with the same client lock handle and
3539 mds_getattr_name->fixup_handle_for_resent_request found one of the
3540 lock handles but later failed with ASSERTION(granted_lock != NULL).
3545 Description: handle->h_buffer_credits > 0, assertion failure
3546 Details : h_buffer_credits is zero after truncate, causing assertion
3547 failure. This patch extends the transaction or creates a new
3551 Frequency : NFS re-export or patchless client
3552 Bugzilla : 11179, 10796
3553 Description: Crash on NFS re-export node (__d_move)
3554 Details : We do not want to hash the dentry if we don't have a lock.
3555 But if this dentry is later used in d_move, we'd hit uninitialised
3556 list head d_hash, so we just do this to init d_hash field but
3557 leave dentry unhashed.
3560 Frequency : NFS re-export or patchless client
3562 Description: NFS exports has problem with symbolic link
3563 Details : lustre client didn't properly install dentry when re-exported
3564 to NFS or running patchless client.
3567 Frequency : NFS re-export or patchless client
3569 Description: Various nfs/patchless fixes.
3570 Details : fixes reuse disconected alias for lookup process - this fixes
3571 warning "find_exported_dentry: npd != pd",
3572 fix permission error with open files at nfs.
3573 fix apply umask when do revalidate.
3576 Frequency : occasional
3578 Description: Crash on NFS re-export node
3579 Details : calling clear_page() on the wrong pointer triggered oops in
3580 generic_mapping_read().
3583 Frequency : rarely, using O_DIRECT IO
3585 Description: unaligned directio crashes client with LASSERT
3586 Details : check for unaligned buffers before trying any requests.
3589 Frequency : rarely, using CFS RAID5 patches in non-standard kernel series
3591 Description: stale data returned from RAID cache
3592 Details : If only a small amount of IO is done to the RAID device before
3593 reading it again it is possible to get stale data from the RAID
3594 cache instead of reading it from disk.
3597 Frequency : always for sles10 kernel
3599 Description: sles10 support
3600 Details : ll_follow_link: compile fixes and using of nd_set_link
3601 under newer kernels.
3604 Frequency : depends on arch, kernel and compiler version, always on sles10
3607 Description: recursive or deep enough symlinks cause stack overflow
3608 Details : getting rid of large stack-allocated variable in
3612 Frequency : depends on hardware
3614 Description: lustre write performance loss in the SLES10 kernel
3615 Details : the performance loss is caused by using of write barriers in the
3616 ext3 code. The SLES10 kernel turns barrier support on by
3617 default. The fix is to undo that change for ldiskfs.
3619 ------------------------------------------------------------------------------
3621 2006-12-09 Cluster File Systems, Inc. <info@clusterfs.com>
3623 * Support for kernels:
3624 2.6.9-42.0.3EL (RHEL 4)
3625 2.6.5-7.276 (SLES 9)
3626 2.4.21-47.0.1.EL (RHEL 3)
3627 2.6.12.6 vanilla (kernel.org)
3631 Frequency : quota enabled and large files being deleted
3633 Description: releasing more than 4GB of quota at once hangs OST
3634 Details : If a user deletes more than 4GB of files on a single OST it
3635 will cause the OST to spin in an infinite loop. Release
3636 quota in < 4GB chunks, or use a 64-bit value for 1.4.7.1+.
3641 Description: statfs data retrieved from /proc may be stale or zero
3642 Details : When reading per-device statfs data from /proc, in the
3643 {kbytes,files}_{total,free,avail} files, it may appear
3644 as zero or be out of date.
3647 Frequency : systems with MD RAID1 external journal devices
3649 Description: lconf's call to blkid is confused by RAID1 journal devices
3650 Details : Use the "blkid -l" flag to locate the MD RAID device instead
3651 of returning all block devices that match the journal UUID.
3654 Frequency : always, for aggregate stripe size over 4GB
3656 Description: "lfs setstripe" fails assertion when setting 4GB+ stripe width
3657 Details : Using "lfs setstripe" to set stripe size * stripe count over 4GB
3658 will fail the kernel with "ASSERTION(lsm->lsm_xfersize != 0)"
3661 Frequency : always if "lfs find" used on a local file/directory
3663 Description: "lfs find" segfaults if used on a local file/directory
3664 Details : The case where a directory component was not specified wasn't
3665 handled correctly. Handle this properly.
3668 Frequency : always on ppc64
3670 Description: the write to an ext3 filesystem mounted with mballoc got stuck
3671 Details : ext3_mb_generate_buddy() uses find_next_bit() which does not
3672 perform endianness conversion.
3675 Frequency : rarely (truncate to non-zero file size after write under load)
3676 Bugzilla : 10730, 10687
3677 Description: Files padded with zeros to next 4K multiple
3678 Details : With filesystems mounted using the "extents" option (2.6 kernels)
3679 it is possible that files that are truncated to a non-zero size
3680 immediately after being written are filled with zero bytes beyond
3681 the truncated size. No file data is lost.
3683 Severity : enhancement
3685 Description: Allow recovery/failover for liblustre clients.
3686 Details : liblustre clients were unaware of failover configurations until
3689 Severity : enhancement
3691 Description: user file locks should fail when not mounting with flock option
3692 Details : Set up an error-returning stub in ll_file_operations.lock field
3693 to prevent incorrect behaviour when client is mounted without
3694 flock option. Also, set up properly f_op->flock field for
3698 Frequency : always on ia64
3700 Description: "lfs df" loops on printing out MDS statfs information
3701 Details : The obd_ioctl_data was not initialized and in some systems
3702 this caused a failure during the ioctl that did not return
3703 an error. Initialize the struct and return an error on failure.
3706 Frequency : SLES 9 only
3708 Description: Error of copying files with lustre special EAs as root
3709 Details : Client side always return success for setxattr call for lustre
3710 special xattr (currently only "trusted.lov").
3713 Frequency : rarely on clusters with both ia64+i386 clients
3715 Description: ia64+i686 clients doing shared IO on the same file may LBUG
3716 Details : In rare cases when both ia64+i686 (or other mixed-PAGE_SIZE)
3717 clients are doing concurrent writes to the same file it is
3718 possible that the ia64 clients may LASSERT because the OST
3719 extent locks are not PAGE_SIZE aligned. Ensure that grown
3720 locks are always aligned on the request boundary.
3723 Frequency : specific use, occasional
3725 Description: Overwriting in use executable truncates on-disk binary image
3726 Details : If one node attempts to overwrite an executable in use by
3727 another node, we now correctly return ETXTBSY instead of
3728 truncating the file.
3730 Severity : enhancement
3732 Description: Async OSC create to avoid the blocking unnecessarily.
3733 Details : If a OST has no remain object, system will block on the creating
3734 when need to create a new object on this OST. Now, ways use
3735 pre-created objects when available, instead of blocking on an
3736 empty osc while others are not empty. If we must block, we block
3737 for the shortest possible period of time.
3742 Description: chmod on Lustre root is propagated to other clients
3743 Details : Re-validate root's dentry in ll_lookup_it to avoid having it
3744 invalid by the follow_mount time.
3747 Frequency : liblustre clients only
3749 Description: Race in 'instant cancel' lock handling could lead to such locks
3750 never to be granted in case of SMP MDS
3751 Details : Do not destroy not yet granted but cbpending locks in
3755 Frequency : replay/resend of open
3757 Description: non null lock assetion failure in mds_intent_policy
3758 Details : Trying to replay/resend lockless open requests resulted in
3759 mds_open() returning 0 with no lock. Now it sets a flag if
3760 a lock is going to be returned.
3762 Severity : enhancement
3764 Description: Checksum enhancements
3765 Details : New checksum enhancements allow for resending RPCs that failed
3768 Severity : enhancement
3770 Description: Tunables on number of dirty pages in cacche
3771 Details : Allow to set limit on number of dirty pages cached.
3776 Description: client crash on unmount - lock still has references
3777 Details : In some error handling cases it was possible to leak a lock
3778 reference on a client while accessing a file. This was not
3779 harmful to the client during operation, but would cause the
3780 client to crash when the filesystem is unmounted.
3783 Frequency : specific case, rare
3785 Description: ETXTBSY on mds though file not in use by client
3786 Details : ETXTBSY is no longer incorrectly returned when attempting to
3787 chmod or chown a directory that the user previously tried to
3788 execute or a currently-executing binary.
3791 Frequency : extremely rare except on liblustre-based clients
3793 Description: Lustre space not freed when files are deleted
3794 Details : Clean up open-unlinked files after client eviction. Previously
3795 the unlink was skipped and the files remained as orphans.
3800 Description: OST failure "would be an LBUG" in waiting_locks_callback()
3801 Details : In some cases it was possible to send a blocking callback to
3802 a client doing a glimpse, even though that client didn't get
3803 a lock granted. When the glimpse lock is cancelled on the OST
3804 the freed lock is left on the waiting list and corrupted the list.
3807 Frequency : all core dumps
3809 Description: Broke core dumps to lustre
3810 Details : Negative dentry may be unhashed if parent does not have UPDATE
3811 lock, but some callers, e.g. do_coredump, expect dentry to be
3812 hashed after successful create, hash it in ll_create_it.
3814 ------------------------------------------------------------------------------
3816 2006-09-13 Cluster File Systems, Inc. <info@clusterfs.com>
3818 * Support for kernels:
3819 2.6.9-42.0.2.EL (RHEL 4)
3820 2.6.5-7.276 (SLES 9)
3821 2.4.21-40.EL (RHEL 3)
3822 2.6.12.6 vanilla (kernel.org)
3826 Frequency : always on RHEL 3
3828 Description: Number of open files grows over time
3829 Details : The number of open files grows over time, whether or not
3830 Lustre is started. This was due to a filp leak introduced
3831 by one of our kernel patches.
3833 ------------------------------------------------------------------------------
3835 2006-08-20 Cluster File Systems, Inc. <info@clusterfs.com>
3837 * Support for kernels:
3838 2.6.9-42.EL (RHEL 4)
3839 2.6.5-7.267 (SLES 9)
3840 2.4.21-40.EL (RHEL 3)
3841 2.6.12.6 vanilla (kernel.org)
3846 Bugzilla : 5719, 9635, 9792, 9684
3847 Description: OST (or MDS) trips assertions in (re)connection under heavy load
3848 Details : If a server is under heavy load and cannot reply to new
3849 connection requests before the client resends the (re)connect,
3850 the connection handling code can behave badly if two service
3851 threads are concurrently handing separate (re)connections from
3852 the same client. Add better locking to the connection handling
3853 code, and ensure that only a single connection will be processed
3854 for a given client UUID, even if the lock is dropped.
3856 Severity : enhancement
3858 Description: add TCP zero-copy support to kernel
3859 Details : Add support to the kernel TCP stack to allow zero-copy bulk
3860 sends if the hardware supports scatter-gather and checksumming.
3861 This allows socklnd to do client-write and server-read more
3862 efficiently and reduce CPU utilization from skbuf copying.
3865 Frequency : only if NFS exporting from client
3867 Description: NULL pointer deref in ll_iocontrol() if chattr mknod file
3868 Details : If setting attributes on a file created under NFS that had
3869 never been opened it would be possible to oops the client
3870 if the file had no objects.
3874 Bugzilla : 9326, 10402, 10897
3875 Description: client crash in ptlrpcd_wake() thread when sending async RPC
3876 Details : It is possible that ptlrpcd_wake() dereferences a freed async
3877 RPC. In rare cases the ptlrpcd thread alread processed the RPC
3878 before ptlrpcd_wake() was called and the request was freed.
3881 Frequency : always for liblustre
3883 Description: liblustre client does MDS+OSTs setattr RPC for each write
3884 Details : When doing a write from a liblustre client, the client
3885 incorrectly issued an RPC to the MDS and each OST the file was
3886 striped over in order to update the timestamps. When writing
3887 with small chunks and many clients this could overwhelm the MDS
3888 with RPCs. In all cases it would slow down the write because
3889 these RPCs are unnecessary.
3891 Severity : enhancement
3893 Description: allow number of MDS service threads to be changed at module load
3894 Details : It is now possible to change the number of MDS service threads
3895 running. Adding "options mds mds_num_threads={N}" to the MDS's
3896 /etc/modprobe.conf will set the number of threads for the next
3897 time Lustre is restarted (assuming the "mds" module is also
3898 reloaded at that time). The default number of threads will
3899 stay the same, 32 for most systems.
3904 Description: OST crash if filesystem is unformatted or corrupt
3905 Details : If an OST is started on a device that has never been formatted
3906 or if the filesystem is corrupt and cannot even mount then the
3907 error handling cleanup routines would dereference a NULL pointer.
3912 Description: NULL pointer deref in llap_from_page.
3913 Details : get_cache_page_nowait can return a page with NULL (or otherwise
3914 incorrect) mapping if the page was truncated/reclaimed while it was
3915 searched for. Check for this condition and skip such pages when
3916 doing readahead. Introduce extra check to llap_from_page() to
3917 verify page->mapping->host is non-NULL (so page is not anonymous).
3920 Frequency : Sometimes when using sys_sendfile
3922 Description: "page not covered by a lock" warnings from ll_readpage
3923 Details : sendfile called ll_readpage without right page locks present.
3924 Now we introduced ll_file_sendfile that does necessary locking
3925 around call to generic_file_sendfile() much like we do in
3929 Frequency : with certain MDS communication failures at client mount time
3931 Description: NULL pointer deref after failed client mount
3932 Details : a client connection request may delayed by the network layer
3933 and not be sent until after the PTLRPC layer has timed out the
3934 request. If the client fails the mount immediately it will try
3935 to clean up before the network times out the request. Add a
3936 reference from the request import to the obd device and delay
3937 the cleanup until the network drops the request.
3940 Frequency : occasionally during client (re)connect
3942 Description: assertion failure during client (re)connect
3943 Details : processing a client connection request may be delayed by the
3944 client or server longer than the client connect timeout. This
3945 causes the client to resend the connection request. If the
3946 original connection request is replied in this interval, the
3947 client may trip an assertion failure in ptlrpc_connect_interpret()
3948 which thought it would be the only running connect process.
3951 Frequency : only with obd_echo servers and clients that are rebooted
3953 Description: kernel BUG accessing uninitialized data structure
3954 Details : When running an obd_echo server it did not start the ping_evictor
3955 thread, and when a client was evicted an uninitialized data
3956 structure was accessed. Start the ping_evictor in the RPC
3957 service startup instead of the OBD startup.
3959 Severity : enhancement
3960 Bugzilla : 10193 (patchless)
3961 Description: Remove dependency on various unexported kernel interfaces.
3962 Details : No longer need reparent_to_init, exit_mm, exit_files,
3963 sock_getsockopt, filemap_populate, FMODE_EXEC, put_filp.
3966 Frequency : rare (only users of deprecated and unsupported LDAP config)
3968 Description: write_conf for zeroconf mount queried LDAP incorrectly for client
3969 Details : LDAP apparently contains 'lustreName' attributes instead of
3970 'name'. A simple remapping of the name is sufficient.
3973 Frequency : rare (only with non-default dump_on_timeout debug enabled)
3975 Description: waiting_locks_callback trips kernel BUG if client is evicted
3976 Details : Running with the dump_on_timeout debug flag turned on makes
3977 it possible that the waiting_locks_callback() can try to dump
3978 the Lustre kernel debug logs from an interrupt handler. Defer
3979 this log dumping to the expired_lock_main() thread.
3981 Severity : enhancement
3983 Description: Support NFS exporting on 2.6 kernels.
3984 Details : Implement non-rawops metadata methods for NFS server to use without
3985 changing NFS server code.
3988 Frequency : very rare (synthetic metadata workload only)
3990 Description: two racing renames might cause an MDS thread to deadlock
3991 Details : Running the "racer" program may cause one MDS thread to rename
3992 a file from being the source of a rename to being the target of
3993 a rename at exactly the same time that another thread is doing
3994 so, and the second thread has already enqueued these locks after
3995 doing a lookup of the target and is trying to relock them in
3996 order. Ensure that we don't try to re-lock the same resource.
3999 Frequency : only very large systems with liblustre clients
4001 Description: slow eviction of liblustre clients with the "evict_by_nid" RPC
4002 Details : Use asynchronous set_info RPCs to send the "evict_by_nid" to
4003 all OSTs in parallel. This allows the eviction of stale liblustre
4004 clients to proceed much faster than if they were done in series,
4005 and also offers similar improvements for other set_info RPCs.
4010 Description: excessive CPU usage during initial read phase on client
4011 Details : During the initial read phase on a client, it would agressively
4012 retry readahead on the file, consuming too much CPU and impacting
4013 performance (since 1.4.5.8). Improve the readahead algorithm
4014 to avoid this, and also improve some other common cases (read
4015 of small files in particular, where "small" is files smaller than
4016 /proc/fs/lustre/llite/*/max_read_ahead_whole_mb, 2MB by default).
4021 Description: MDS crash when receiving packet with unknown intent.
4022 Details : Do not LBUG in unknown intent case, just return -EFAULT
4024 Severity : enhancement
4025 Bugzilla : 9293, 9385
4026 Description: MDS RPCs are serialised on client. This is unnecessary for some.
4027 Details : Do not serialize getattr (non-intent version) and statfs.
4030 Frequency : occasional, when OST network is overloaded/intermittent
4032 Description: client evicted by OST after bulk IO timeout
4033 Details : If a client sends a bulk IO request (read or write) the OST
4034 may evict the client if it is unresposive to its data GET/PUT
4035 request. This is incorrect if the network is overloaded (takes
4036 too long to transfer the RPC data) or dropped the OST GET/PUT
4037 request. There is no need to evict the client at all, since
4038 the pinger and/or lock callbacks will handle this, and the
4039 client can restart the bulk request.
4042 Frequency : Always when mmapping file with no objects
4044 Description: client crashes when mmapping file with no objects
4045 Details : Check that we actually have objects in a file before doing any
4046 operations on objects in ll_vm_open, ll_vm_close and
4052 Description: Request leak when working with deleted CWD
4053 Details : Introduce advanced request refcount tracking for requests
4054 referenced from lustre intent.
4056 Severity : Enhancement
4058 Description: Cache open file handles on client.
4059 Details : MDS now will return special lock along with openhandle, if
4060 requested and client is allowed to hold openhandle, even if unused,
4061 until such a lock is revoked. Helps NFS a lot, since NFS is opening
4062 closing files for every read/write openration.
4064 Severity : Enhancement
4066 Description: Cache open negative dentries on client when possible.
4067 Details : Guard negative dentries with UPDATE lock on parent dir, drop
4068 negative dentries on lock revocation.
4073 Description: Remounting a client read-only wasn't possible with a zconf mount
4074 Details : It wasn't possible to remount a client read-only with llmount.
4076 Severity : enhancement
4077 Description: Include MPICH 1.2.6 Lustre ADIO interface patch
4078 Details : In lustre/contrib/ or /usr/share/lustre in RPM a patch for
4079 MPICH is included to add Lustre-specific ADIO interfaces.
4080 This is based closely on the UFS ADIO layer and only differs
4081 in file creation, in order to allow the OST striping to be set.
4082 This is user-contributed code and not supported by CFS.
4087 Description: extended inode attributes (immutable, append-only) work improperly
4088 when 2.4 and 2.6 kernels are used on client/server or vice versa
4089 Details : Introduce kernel-independent values for these flags.
4091 Severity : enhancement
4094 Description: Allow fractional MB tunings for lustre in /proc/ filesystem.
4095 Details : Many of the /proc/ tunables can only be tuned at a megabyte
4096 granularity. Now, Fractional MB granularity is be supported,
4097 this is very useful for low memory system.
4099 Severity : enhancement
4101 Description: Getattr by fid
4102 Details : Getting a file attributes by its fid, obtaining UPDATE|LOOKUP
4103 locks, avoids extra getattr rpc requests to MDS, allows '/' to
4104 have locks and avoids getattr rpc requests for it on every stat.
4107 Frequency : Always, for filesystems larger than 2TB
4109 Description: ldiskfs crash at mount for filesystem larger than 2TB with mballoc
4110 Details : Kenrel kmalloc limits allocations to 128kB and this prevents
4111 filesystems larger than 2TB to be mounted with mballoc enabled.
4114 Frequency : Always, for 32-bit kernel without CONFIG_LBD and filesystem > 2TB
4116 Description: filesystem corruption for non-standard kernels and very large OSTs
4117 Details : If a 32-bit kernel is compiled without CONFIG_LBD enabled and a
4118 filesystems larger than 2TB is mounted then the kernel will
4119 silently corrupt the start of the filesystem. CONFIG_LBD is
4120 enabled for all CFS-supported kernels, but the possibility of
4121 this happening with a modified kernel config exists.
4123 Severity : enhancement
4125 Description: add client O_DIRECT support for 2.6 kernels
4126 Details : It is now possible to do O_DIRECT reads and writes to files
4127 in the Lustre client mountpoint on 2.6 kernel clients.
4129 Severity : enhancement
4131 Description: parallel glimpse, setattr, statfs, punch, destroy requests
4132 Details : Sends glimpse, setattr, statfs, punch, destroy requests to OSTs in
4133 parallel, not waiting for response from every OST before sending
4134 a rpc to the next OST.
4139 Description: setattr vs write race when updating file timestamps
4140 Details : Client processes that update a file timestamp into the past
4141 right after writing to the file (e.g. tar) it is possible that
4142 the updated file modification time can be reset to the current
4143 time due to a race between processing the setattr and write RPC.
4145 Severity : enhancement
4147 Description: Bring 'lfs find' closer in line with regular Linux find.
4148 Details : lfs find util supports -atime, -mtime, -ctime, -maxdepth, -print,
4149 -print0 options and obtains all the needed info through the lustre
4152 Severity : enhancement
4154 Description: support up to 1024 configured devices on one node
4155 Details : change obd_dev array from statically allocated to dynamically
4156 allocated structs as they are first used to reduce memory usage
4161 Description: Flush dirty partially truncated pages during truncate
4162 Details : Immediatelly flush partially truncated pages in filter_setattr,
4163 this way we completely avoid having any pages in page cache on OST
4164 and can retire ugly workarounds during writes to flush such pages.
4169 Description: i_sem vs transaction deadlock in mds_obd_destroy during unlink.
4170 Details : protect inode from truncation within vfs_unlink() context
4171 just take a reference before calling vfs_unlink() and release it
4172 when parent's i_sem is free.
4175 Frequency : always, if extents are used on OSTs
4177 Description: index ei_leaf_hi (48-bit extension) is not zeroed in extent index
4178 Details : OSTs using the extents format would not zero the high 16 bits of
4179 the index physical block number. This is not a problem for any
4180 OST filesystems smaller than 16TB, and no kernels support ext3
4181 filesystems larger than 16TB yet. This is fixed in 1.4.7 (all
4182 new/modified files) and can be fixed for existing filesystems
4183 with e2fsprogs-1.39-cfs1.
4188 Description: import connection selection may be incorrect if timer wraps
4189 Details : Using a 32-bit jiffies timer with HZ=1000 may cause backup
4190 import connections to be ignored if the 32-bit jiffies counter
4191 wraps. Use a 64-bit jiffies counter.
4194 Frequency : during server recovery
4196 Description: crash after server is denying duplicate export
4197 Details : If clients are resending connect requests to the server, the
4198 server refuses to allow a client to connect multiple times.
4199 Fixed a bug in the handling of this case.
4202 Frequency : very large clusters immediately after boot
4204 Description: LNET request buffers exhausted under heavy short-term load
4205 Details : If a large number of client requests are generated on a service
4206 that has previously never seen so many requests it is possible
4207 that the request buffer growth cannot keep up with the spike in
4208 demand. Instead of dropping incoming requests, they are held in
4209 the LND until the RPC service can accept more requests.
4212 Frequency : Sometimes during replay
4214 Description: Assertion failure in ll_local_open after replay.
4215 Details : If replay happened on an open request reply before we were able
4216 to set replay handler, reply will become not swabbed tripping the
4217 assertion in ll_local_open. Now we set the handler right after
4218 recognising of open request
4221 Frequency : very rare
4223 Description: kernel reports "badness in vsnprintf"
4224 Details : Reading from the "recovery_status" /proc file in small chunks
4225 may cause a negative length in lprocfs_obd_rd_recovery_status()
4226 call to vsnprintf() (which is otherwise harmless). Exit early
4227 if there is no more space in the output buffer.
4229 Severity : enhancement
4231 Description: clear OBD RPC statistics by writing to them
4232 Details : It is now possible to clear the OBD RPC statistics by writing
4233 to the "stats" file.
4238 Description: Client mtime is not the same on different clients after utimes
4239 Details : In some cases, the client was using the utimes() syscall on
4240 a file cached on another node. The clients now validate the
4241 ctime from the MDS + OSTs to determine which one is right.
4246 Description: Inability to activate failout mode
4247 Details : lconf script incorrectly assumed that in python string's numeric
4248 value is used in comparisons.
4251 Frequency : always with multiple stripes per file
4253 Description: Inefficient object allocation for mutli-stripe files
4254 Details : When selecting which OSTs to stripe files over, for files with
4255 a stripe count that divides evenly into the number of OSTs,
4256 the MDS is always picking the same starting OST for each file.
4257 Return the OST selection heuristic to the original design.
4262 Description: mount failures may take full timeout to return an error
4263 Details : Under some heavy load conditions it is possible that a
4264 failed mount can wait for the full obd_timeout interval,
4265 possibly several minutes, before reporting an error.
4266 Instead return an error as soon as the status is known.
4268 ------------------------------------------------------------------------------
4270 2006-02-14 Cluster File Systems, Inc. <info@clusterfs.com>
4272 * WIRE PROTOCOL CHANGE. This version of Lustre networking WILL NOT
4273 INTEROPERATE with older versions automatically. Please read the
4274 user documentation before upgrading any part of a live system.
4275 * WARNING: Lustre networking configuration changes are required with
4276 this release. See https://bugzilla.clusterfs.com/show_bug.cgi?id=10052
4279 * Support for kernels:
4280 2.6.9-22.0.2.EL (RHEL 4)
4281 2.6.5-7.244 (SLES 9)
4282 2.6.12.6 vanilla (kernel.org)
4285 Severity : enhancement
4286 Bugzilla : 7981/8208
4287 Description: Introduced Lustre Networking (LNET)
4288 Details : LNET is new networking infrastructure for Lustre, it includes
4289 a reorganized network configuration mode (see the user
4290 documentation for full details) as well as support for routing
4291 between different network fabrics. Lustre Networking Devices
4292 (LNDs) for the supported network fabrics have also been
4293 created for this new infrastructure.
4295 Severity : enhancement
4296 Description: Introduced Access control lists
4297 Details : clients can set ACLs on files and directories in order to have
4298 more fine-grained permissions than the standard Unix UGO+RWX.
4299 The MDS must be started with the "-o acl" mount option.
4301 Severity : enhancement
4302 Description: Introduced filesystem quotas
4303 Details : Administrators may now establish per-user quotas on the
4306 Severity : enhancement
4308 Description: Configuration change for the XT3
4309 The PTLLND is now used to run Lustre over Portals on the XT3
4310 The configure option(s) --with-cray-portals are no longer used.
4311 Rather --with-portals=<path-to-portals-includes> is used to
4312 enable building on the XT3. In addition to enable XT3 specific
4313 features the option --enable-cray-xt3 must be used.
4318 Description: Running on many-way SMP OSTs can trigger oops in llcd_send()
4319 Details : A race between allocating a new llcd and re-getting the llcd_lock
4320 allowed another thread to grab newly-allocated llcd.
4322 Severity : enhancement
4324 Description: 2.6 OST async journal commit and locking fix to improve performance
4325 Details : The filter_direct_io()+filter_commitrw_write() journal commits for
4326 2.6 kernels are now async as they already were in 2.4 kernels so
4327 that they can commit concurrently with the network bulk transfer.
4328 For block-allocated files the filter allocation semaphore is held
4329 to avoid filesystem fragmentation during allocation. BKL lock
4330 removed for 2.6 xattr operations where it is no longer needed.
4335 Description: lconf incorrectly determined whether two IP networks could talk
4336 Details : In some more complicated routing and multiple-network
4337 configurations, lconf will avoid trying to make a network
4338 connection to a disjoint part of the IP space. It was doing the
4339 math incorrectly for one set of cases.
4344 Description: Fix for potential infinite loop processing records in an llog.
4345 Details : If an llog record is corrupted/zeroed, it is possible to loop
4346 forever in llog_process(). Validate the llog record length
4347 and skip the remainder of the block on error.
4350 Frequency : occasional (liblustre only)
4352 Description: liblustre could not open files whose last component is a symlink
4353 Details : sysio_path_walk() would incorrectly pass the open intent to
4354 intermediate path components.
4357 Frequency : rare (liblustre only with non-standard tuning)
4358 Bugzilla : 7201 (7350)
4359 Description: Tuning the MDC DLM LRU size to zero triggers client LASSERT
4360 Details : llu_lookup_finish_locks() tries to set lock data on a lock
4361 after it has been released, only do this for referenced locks
4363 Severity : enhancement
4365 Description: specifying an (invalid) directory default stripe_size of -1
4366 would reset the directory default striping
4367 Details : stripe_size -1 was used internally to signal directory stripe
4368 removal, now use "all default" to signal dir stripe removal
4369 as a directory striping of "all default" is not useful
4372 Frequency : common for large clusters running liblustre clients
4374 Description: doing an ls when liblustre clients are running is slow
4375 Details : sending a glimpse AST to a liblustre client waits for every AST
4376 to time out, as liblustre clients will not respond. Since they
4377 cannot cache data we refresh the OST lock LVB from disk instead.
4379 Severity : enhancement
4381 Description: doing an ls at the same time as file IO can be slow
4382 Details : enqueue and other "small" requests can be blocked behind many
4383 large IO requests. Create a new OST IO portal for non-IO
4384 requests so they can be processed faster.
4387 Frequency : rare (only HPUX clients mounting unsupported re-exported NFS vol)
4389 Description: an HPUX NFS client would get -EACCESS when ftruncate()ing a newly
4390 created file with mode 000
4391 Details : the Linux NFS server relies on an MDS_OPEN_OWNEROVERRIDE hack to
4392 allow an ftruncate() as a non-root user to a file with mode 000.
4393 Lustre now respects this flag to disable mode checks when
4394 truncating a file owned by the user
4397 Frequency : liblustre-only, when liblustre client dies unexpectedly or becomes
4400 Description: Revoking locks from clients that went dead or catatonic might take
4402 Details : New lock flags FL_CANCEL_ON_BLOCK used by liblustre makes
4403 cancellation of such locks instant on servers without waiting for
4404 any reply from clients. Clients drops these locks when cancel
4405 notification from server is received without replying.
4408 Frequency : liblustre-only, when liblustre client dies or becomes busy
4410 Description: Doing ls on Linux clients can take a long time with active
4412 Details : Liblustre client cannot handle ASTs in timely manner, so avoid
4413 granting such locks to it in the first place if possible. Locks
4414 are taken by proxy on the OST during the read or write and
4415 dropped immediately afterward. Add connect flags handling, do
4416 not grant locks to liblustre clients for glimpse ASTs.
4418 Severity : enhancement
4420 Description: Improve read-ahead algorithm to avoid excessive IO for random reads
4421 Details : Existing read-ahead algorithm is tuned for the case of streamlined
4422 sequential reads and behaves badly with applications doing random
4423 reads. Improve it by reading ahead at least read region, and
4424 avoiding excessive large RPC for small reads.
4426 Severity : enhancement
4428 Description: Creating more than 1000 files for a single job may cause a load
4429 imbalance on the OSTs if there are also a large number of OSTs.
4430 Details : qos_prep_create() uses an OST index reseed value that is an
4431 even multiple of the number of available OSTs so that if the
4432 reseed happens in the middle of the object allocation it will
4433 still utilize the OSTs as uniformly as possible.
4438 Description: OST or MDS may oops in ping_evictor_main()
4439 Details : ping_evictor_main() drops obd_dev_lock if deleting a stale export
4440 but doesn't restart at beginning of obd_exports_timed list
4443 Severity : enhancement
4445 Description: improve by-nid export eviction on the MDS and OST
4446 Details : allow multiple exports with the same NID to be evicted at one
4447 time without re-searching the exports list.
4450 Frequency : rare, only with supplementary groups enabled on SMP 2.6 kernels
4452 Description: MDS may oops in groups_free()
4453 Details : in rare race conditions a newly allocated group_info struct is
4454 freed again, and this can be NULL. The 2.4 compatibility code
4455 for groups_free() checked for a NULL pointer, but 2.6 did not.
4458 Frequency : common for liblustre clients doing little filesystem IO
4459 Bugzilla : 9352, 7313
4460 Description: server may evict liblustre clients accessing contended locks
4461 Details : if a client is granted a lock or receives a completion AST
4462 with a blocking AST already set it would not reply to the AST
4463 for LDLM_FL_CANCEL_ON_BLOCK locks. It now replies to such ASTs.
4466 Frequency : lfs setstripe, only systems with more than 160 OSTs
4468 Description: unable to set striping with a starting offset beyond OST 160
4469 Details : llapi_create_file() incorrectly limited the starting stripe
4470 index to the maximum single-file stripe count.
4473 Frequency : LDAP users only
4475 Description: lconf did not handle in-kernel recovery with LDAP properly
4476 Details : lconf/LustreDB get_refs() is searching the wrong namespace
4478 Severity : enhancement
4480 Description: bind OST threads to NUMA nodes to improve performance
4481 Details : all OST threads are uniformly bound to CPUs on a single NUMA
4482 node and do their allocations there to localize memory access
4484 Severity : enhancement
4486 Description: llmount can determine client NID directly from Myrinet (GM)
4487 Details : the client NID code from gmnalnid was moved directly into
4488 llmount, removing the need to use this or specifying the
4489 client NID explicitly when mounting GM clients with zeroconf
4492 Frequency : if client is started with down MDS
4494 Description: if client is started with down MDS mount hangs in ptlrpc_queue_wait
4495 Details : Having an LWI_INTR() wait event (interruptible, but no timeout)
4496 will wait indefinitely in ptlrpc_queue_wait->l_wait_event() after
4497 ptlrpc_import_delayed_req() because we didn't check if the
4498 request was interrupted, and we also didn't break out of the
4499 event loop if there was no timeout
4504 Description: data loss during non-page-aligned writes to a single file from
4505 both multiple nodes and multiple threads on one node at same time
4506 Details : updates to KMS and lsm weren't protected by common lock. Resulting
4507 inconsistency led to false short-reads, that were cached and later
4508 used by ->prepare_write() to fill in partially written page,
4509 leading to data loss.
4512 Frequency : always, if lconf --abort_recovery used
4514 Description: lconf --abort_recovery fails with 'Operation not supported'
4515 Details : lconf was attempting to abort recovery on the MDT device and not
4518 Severity : enhancement
4520 Description: remove cleanup logs
4521 Details : replace lconf-generated cleanup logs with lustre internal
4522 cleanup routines. Eliminates the need for client-cleanup and
4525 Severity : enhancement
4527 Description: add support for EAs (user and system) on lustre filesystems
4528 Details : it is now possible to store extended attributes in the Lustre
4529 client filesystem, and with the user_xattr mount option it
4530 is possible to allow users to store EAs on their files also
4532 Severity : enhancement
4534 Description: Add possibility (config option) to show minimal available OST free
4536 Details : When compiled with --enable-mindf configure option, statfs(2)
4537 (and so, df) will return least minimal free space available from
4538 all OSTs as amount of free space on FS, instead of summary of
4539 free spaces of all OSTs.
4541 Severity : enhancement
4543 Description: do not expand extent locks acquired on OST-side
4544 Details : Modify ldlm_extent_policy() to not expand local locks, acquired
4545 by server: they are not cached anyway.
4548 Frequency : when mmap is used/binaries executed from Lustre
4550 Description: Unmmap pages before throwing them away from read cache.
4551 Details : llap_shrink cache now attempts to unmap pages before discarding
4552 them (if unmapping failed - do not discard). SLES9 kernel has
4553 extra checks that trigger if this unmapping is not done first.
4558 Description: lconf didn't resolve symlinks before checking to see whether a
4559 given mountpoint was already in use
4562 Frequency : when migrating failover services
4563 Bugzilla : 6395, 9514
4564 Description: When migrating a subset of services from a node (e.g. failback
4565 from a failover service node) the remaining services would
4566 time out and evict clients.
4567 Details : lconf --force (implied by --failover) sets the global obd_timeout
4568 to 5 seconds in order to quickly disconnect, but this caused
4569 other RPCs to time out too quickly. Do not change the global
4570 obd_timeout for force cleanup, only set it for DISCONNECT RPCs.
4572 Severity : enhancement
4573 Frequency : if MDS is started with down OST
4574 Bugzilla : 9439,5706
4575 Description: Allow startup/shutdown of an MDS without depending on the
4576 availability of the OSTs.
4577 Details : Asynchronously call mds_lov_synchronize during MDS startup.
4578 Add appropriate locking and lov-osc refcounts for safe
4579 cleaning. Add osc abort_inflight calls in case the
4580 synchronize never started.
4583 Frequency : occasional (Cray XT3 only)
4585 Description: root not authorized to access files in CRAY_PORTALS environment
4586 Details : The client process capabilities were not honoured on the MDS in
4587 a CRAY_PORTALS/CRAY_XT3 environment. If the file had previously
4588 been accessed by an authorized user then root was able to access
4589 the file on the local client also. The root user capabilities
4590 are now allowed on the MDS, as this environment has secure UID.
4593 Frequency : occasional
4595 Description: ldiskfs "too long searching" message happens too often
4596 Details : A debugging message (otherwise harmless) prints too often on
4597 the OST console. This has been reduced to only happen when
4598 there are fragmentation problems on the filesystem.
4603 Description: Division by zero in statfs when all OSCs are inactive
4604 Details : lov_get_stripecnt() returns zero due to incorrect order of checks,
4605 lov_statfs divides by value returned by lov_get_stripecnt().
4609 Bugzilla : 9489, 3273
4610 Description: First write from each client to each OST was only 4kB in size,
4611 to initialize client writeback cache, which caused sub-optimal
4612 RPCs and poor layout on disk for the first writen file.
4613 Details : Clients now request an initial cache grant at (re)connect time
4614 and so that they can start streaming writes to the cache right
4615 away and always do full-sized RPCs if there is enough data.
4616 If the OST is rebooted the client also re-establishes its grant
4617 so that client cached writes will be honoured under the grant.
4622 Description: Slow ls (and stat(2) syscall) on files residing on IO-loaded OSTs
4623 Details : Now I/O RPCs go to different portal number and (presumably) fast
4624 lock requests (and glimses) and other RPCs get their own service
4625 threads pool that should be able to service those RPCs
4628 Severity : enhancement
4630 Description: Ability to exchange lustre version between client and servers and
4631 issue warnings at client side if client is too old. Also for
4632 liblustre clients there is ability to refuse connection of too old
4634 Details : New 'version' field is added to connect data structure that is
4635 filled with version info. That info is later checked by server and
4639 Frequency : rare, liblustre only.
4640 Bugzilla : 9296, 9581
4641 Description: Two simultaneous writes from liblustre at offset within same page
4642 might proceed at the same time overwriting eachother with stale
4644 Details : I/O lock withing llu_file_prwv was released too early, before data
4645 actually was hitting the wire. Extended lock-holding time until
4646 server acknowledges receiving data.
4649 Frequency : extremely rare. Never observed in practice.
4651 Description: avoid generating lustre_handle cookie of 0.
4652 Details : class_handle_hash() generates handle cookies by incrementing
4653 global counter, and can hit 0 occasionaly (this is unlikely, but
4654 not impossible, because initial value of cookie counter is
4655 selected randonly). Value of 0 is used as a sentinel meaning
4656 "unassigned handle" --- avoid it. Also coalesce two critical
4657 sections in this function into one.
4659 Severity : enhancement
4661 Description: allow liblustre clients to delegate truncate locking to OST
4662 Details : To avoid overhead of locking, liblustre client instructs OST to
4663 take extent lock in ost_punch() on client's behalf. New connection
4664 flag is added to handle backward compatibility.
4666 Severity : enhancement
4667 Bugzilla : 4928, 7341, 9758
4668 Description: allow number of OST service threads to be specified
4669 Details : a module parameter allows the number of OST service threads
4670 to be specified via "options ost ost_num_threads={N}" in the
4671 OSS's /etc/modules.conf or /etc/modprobe.conf.
4675 Bugzilla : 6146, 9635, 9895
4676 Description: servers crash with bad pointer in target_handle_connect()
4677 Details : In rare cases when a client is reconnecting it was possible that
4678 the connection request was the last reference for that export.
4679 We would temporarily drop the export reference and get a new
4680 one, but this may have been the last reference and the export
4681 was just destroyed. Get new reference before dropping old one.
4683 Severity : enhancement
4684 Frequency : if client is started with failover MDS
4686 Description: Allow multiple MDS hostnames in the mount command
4687 Details : Try to read the configuration from all specified MDS
4688 hostnames during a client mount in case the "primary"
4691 Severity : enhancement
4693 Description: Stop sending data to evicted clients as soon as possible.
4694 Details : Check if the client we are about to send or are sending data to
4695 was evicted already. (Check is done every second of waiting,
4696 for which l_wait_event interface was extended to allow checking
4697 of exit condition at specified intervals).
4700 Frequency : rare, normally only when NFS exporting is done from client
4702 Description: 'bad disk LOV MAGIC: 0x00000000' error when chown'ing files
4704 Details : Make mds_get_md() recognise empty md case and set lmm size to 0.
4707 Frequency : always, if srand() is called before liblustre initialization
4709 Description: Liblustre uses system PRNG disturbing its usage by user application
4710 Details : Introduce internal to lustre fast and high-quality PRNG for
4711 lustre usage and make liblustre and some other places in generic
4712 lustre code to use it.
4714 Severity : enhancement
4715 Bugzilla : 9477, 9557, 9870
4716 Description: Verify that the MDS configuration logs are updated when xml is
4717 Details : Check if the .xml configuration logs are newer than the config
4718 logs stored on the MDS and report an error if this is the case.
4719 Request --write-conf, or allow starting with --old_conf.
4721 Severity : enhancement
4723 Description: Handle symlinks in the path when checking if Lustre is mounted.
4724 Details : Resolve intermediate symlinks when checking if a client has
4725 mounted a filesystem to avoid duplicate client mounts.
4730 Description: lconf can hit an error exception but still return success.
4731 Details : The lconf command catches the Command error exception at the top
4732 level script context and will exit with the associated exit
4733 status, but doesn't ensure that this exit status is non-zero.
4738 Description: failure of ptlrpc thread startup can cause oops
4739 Details : Starting a ptlrpc service thread can fail if there are a large
4740 number of threads or the server memory is very fragmented.
4741 Handle this without oopsing.
4744 Frequency : always, only if liblustre and non-default acceptor port was used
4746 Description: liblustre cannot connect to servers with non-default acceptor port
4747 Details : tcpnal_set_default_params() was not called and was therefore
4748 ignoring the environment varaible TCPNAL_PORT, as well as other
4749 TCPNAL_ environment variables
4754 Description: two objects could be created on the same OST for a single file
4755 Details : If an OST is down, in some cases it was possible to create two
4756 objects on a single OST for a single file. No problems other
4757 than potential performance impact and spurious error messages.
4761 Bugzilla : 5681, 9562
4762 Description: Client may oops in ll_unhash_aliases
4763 Details : Client dcache may become inconsistent in race condition.
4764 In some cases "getcwd" can fail if the current directory is
4770 Description: Inode refcounting problems in NFS export code
4771 Details : link_raw functions used to call d_instantiate without obtaining
4772 extra inode reference first.
4776 Bugzilla : 9942, 9903
4777 Description: Referencing freed requests leading to crash, memleaks with NFS.
4778 Details : We used to require that call to ll_revalidate_it was always
4779 followed by ll_lookup_it. Also with revalidate_special() it is
4780 possible to call ll_revalidate_it() twice for the same dentry
4781 even if first occurence returned success. This fix changes semantic
4782 between DISP_ENQ_COMPLETE disposition flag to mean there is extra
4783 reference on a request referred from the intent.
4784 ll_intent_release() then releases such a request.
4787 Frequency : rare, normally benchmark loads only
4789 Description: unlinked inodes were kept in memory on the client
4790 Details : If a client is repeatedly creating and unlinking files it
4791 can accumulate a lot of stale inodes in the inode slab cache.
4792 If there is no other client load running this can cause the
4793 client node to run out of memory. Instead flush old inodes
4794 from client cache that have the same inode number as a new inode.
4797 Frequency : SLES9 2.6.5 kernel and long filenames only
4798 Bugzilla : 9969, 10379
4799 Description: utime reports stale NFS file handle
4800 Details : SLES9 uses out-of-dentry names in some cases, which confused
4801 the lustre dentry revalidation. Change it to always use the
4805 Frequency : rare, unless heavy write-truncate concurrency is continuous
4806 Bugzilla : 4180, 6984, 7171, 9963, 9331
4807 Description: OST becomes very slow and/or deadlocked during object unlink
4808 Details : filter_destroy() was holding onto the parent directory lock
4809 while truncating+unlinking objects. For very large objects this
4810 may block other threads for a long time and slow overall OST
4811 responsiveness. It may also be possible to get a lock ordering
4812 deadlock in this case, or run out of journal credits because of
4813 the combined truncate+unlink. Solution is to do object truncate
4814 first in one transaction without parent lock, and then do the
4815 final unlink in a new transaction with the parent lock. This
4816 reduces the lock hold time dramatically.
4819 Frequency : rare, 2.4 kernels only
4821 Description: MDS or OST cleanup may trip kernel BUG when dropping kernel lock
4822 Details : mds_cleanup() and filter_cleanup() need to drop the kernel lock
4823 before unmounting their filesystem in order to avoid deadlock.
4824 The kernel_locked() function in 2.4 kernels only checks whether
4825 the kernel lock is held, not whether it is this process that is
4826 holding it as 2.6 kernels do.
4831 Description: MDS or OST may oops/LBUG if a client is connecting multiple times
4832 Details : The client ptlrpc code may be trying to reconnect to a down
4833 server before a previous connection attempt has timed out.
4834 Increase the reconnect interval to be longer than the connection
4835 timeout interval to avoid sending duplicate connections to
4839 Frequency : echo_client brw_test command
4841 Description: fix echo_client to work with OST preallocated code
4842 Details : OST preallocation code (5137) didn't take echo_client IO path
4843 into account: echo_client calls filter methods outside of any
4844 OST thread and, hence, there is no per-thread preallocated
4845 pages and buffers to use. Solution: hijack pga pages for IO. As
4846 a byproduct, this avoids unnecessary data copying.
4850 Bugzilla : 3555, 5962, 6025, 6155, 6296, 9574
4851 Description: Client can oops in mdc_commit_close() after open replay
4852 Details : It was possible for the MDS to return an open request with no
4853 transaction number in mds_finish_transno() if the client was
4854 evicted, but without actually returning an error. Clients
4855 would later try to replay that open and may trip an assertion
4856 Simplify the client close codepath, and always return an error
4857 from the MDS in case the open is not successful.
4860 Frequency : rare, 2.6 OSTs only
4862 Description: OST may deadlock under high load on fragmented files
4863 Details : If there was a heavy load and highly-fragmented OST filesystems
4864 it was possible to have all the OST threads deadlock waiting on
4865 allocation of biovecs, because the biovecs were not released
4866 until the entire RPC IO was completed. Instead, release biovecs
4867 as soon as they are complete to ensure forward IO progress.
4869 Severity : enhancement
4871 Description: Support for specifying external journal device at mount
4872 Details : If an OST or MDS device is formatted with an external journal
4873 device, this device major/minor is stored in the ext3 superblock
4874 and may not be valid for failover. Allow detecting and
4875 specifying the external journal at mount time.
4880 Description: Mounting an MDS with pending unlinked files may cause oops
4881 Details : target_finish_recovery() calls mds_postrecov() which returned
4882 the number of orphans unlinked. mds_lov_connect->mds_postsetup()
4883 considers this an error and immediately begins cleaning up the
4884 lov, just after starting the mds_lov process
4886 Severity : enhancement
4888 Description: Implement 'lfs df' to report actual free space on per-OST basis
4889 Details : Add sub-command 'df' on 'lfs' to report the disk space usage of
4890 MDS/OSDs. Usage: lfs df [-i][-h]. Command Options: '-i' to report
4891 usage of objects; '-h' to report in human readable format.
4893 ------------------------------------------------------------------------------
4895 2005-08-26 Cluster File Systems, Inc. <info@clusterfs.com>
4902 Description: Mounting an ldiskfs file system with mballoc may crash OST node.
4903 Details : ldiskfs mballoc code may reference an uninitialized buddy struct
4904 at startup during orphan unlinking. Instead, skip buddy update
4905 before setup, as it will be regenerated after recovery is complete.
4910 Description: If an OST is inactive, its locks might reference stale inodes.
4911 Details : lov_change_cbdata() must iterate over all namespaces, even if
4912 they are inactive to clear inode references from the lock.
4914 Severity : enhancement
4915 Frequency : occasional, if non-standard max_dirty_mb used
4917 Description: Client will block write RPCs if not enough grant
4918 Details : If a client has max_dirty_mb smaller than max_rpcs_in_flight,
4919 then the client will block writes while waiting for another RPC
4920 to complete instead of consuming its dirty limit. With change
4921 we get improved performance when max_dirty_mb is small.
4923 Severity : enhancement
4924 Bugzilla : 3389, 6253
4925 Description: Add support for supplementary groups on the MDS.
4926 Details : The MDS has an upcall /proc/fs/lustre/mds/{mds}/group_upcall
4927 (set to /usr/sbin/l_getgroups if enabled) which will do MDS-side
4928 lookups for user supplementary groups into a cache.
4932 Description: O_CREAT|O_EXCL open flags in liblustre always return -EEXIST
4933 Details : Make libsysio to not enforce O_EXCL by clearing the flag,
4934 for liblustre O_EXCL is enforced by MDS.
4938 Description: readdir never returns NULL in liblustre.
4939 Details : Corrected llu_iop_getdirentries logic, to return offset of next
4940 dentry in struct dirent.
4944 Frequency : liblustre only, depends on application IO pattern
4945 Description: liblustre clients evicted if not contacting servers
4946 Details : Don't put liblustre clients into the ping_evictor list, so
4947 they will not be evicted by the pinger ever.
4949 Severity : enhancement
4951 Description: Add ability to evict clients by NID from MDS.
4952 Details : By echoing "nid:$NID" string into
4953 /proc/fs/lustre/mds/.../evict_client client with nid that equals to
4954 $NID would be instantly evicted from this MDS and from all active
4955 OSTs connected to it.
4959 Description: Do not query file size twice, somewhat slowing stat(2) calls.
4960 Details : lookup_it_finish() used to query file size from OSTs that was not
4965 Description: service threads change working directory to that of init
4966 Details : Starting lustre service threads may pin the working directory
4967 of the parent thread, making that filesystem busy. Threads
4968 now change to the working directory of init to avoid this.
4972 Frequency : during shutdown only
4973 Description: shutdown with a failed MDS or OST can cause unmount to hang
4974 Details : Don't resend DISCONNECT messages in ptlrpc_disconnect_import()
4979 Frequency : 2.6 only
4980 Description: chmod/chown may include an extra supplementary group
4981 Details : ll{,u}_mdc_pack_op_data() does not properly initialize the
4982 supplementary group and if none is specified this is used.
4985 Bugzilla : 5479 (6816)
4987 Description: Racing open + rm can assert client in mdc_set_open_replay_data()
4988 Details : If lookup is in progress on a file that is unlinked we might try
4989 to revalidate the inode and fail in revalidate after lookup is
4990 complete and ll_file_open() enqueues the open again but
4991 it_open_error() was not checking DISP_OPEN_OPEN errors correctly.
4994 Frequency : always, if lconf --abort_recovery used
4996 Description: lconf --abort_recovery fails with 'Operation not supported'
4997 Details : lconf was attempting to abort recovery on the MDT device and not
5000 ------------------------------------------------------------------------------
5002 2005-08-08 Cluster File Systems, Inc. <info@clusterfs.com>
5007 Frequency : rare (only unsupported configurations with a node running as an
5009 Bugzilla : 6514, 5137
5010 Description: Mounting a Lustre file system on a node running as an OST could
5012 Details : OSTs now preallocates memory needed to write out data at
5013 startup, instead of when needed, to avoid having to
5014 allocate memory in possibly low memory situations.
5015 Specifically, if the file system is mounted on on OST,
5016 memory pressure could force it to try to write out data,
5017 which it needed to allocate memory to do. Due to the low
5018 memory, it would be unable to do so and the node would
5019 become unresponsive.
5021 Severity : enhancement
5023 Description: Addition of lconf --service command line option
5024 Details : lconf now accepts a '--service <arg>' option, which is
5025 shorthand for 'lconf --group <arg> --select <arg>=<hostname>'
5027 Severity : enhancement
5029 Description: Failover mode is now the default for OSTs.
5030 Details : By default, OSTs will now run in failover mode. To return to
5031 the old behaviour, add '--failout' to the lmc line for OSTs.
5033 Severity : enhancement
5035 Description: Health checks are now provided for MDS and OSTs
5036 Details : Additional detailed health check information on MSD and OSTs
5037 is now provided through the procfs health_check value.
5040 Frequency : occasional, depends on IO load
5042 Description: Disk fragmentation on the OSTs could eventually cause slowdowns
5043 after numerous create/delete cycles
5044 Details : The ext3 inode allocation policy would not allocate new inodes
5045 very well on the OSTs because there are no new directories
5046 being created. Instead we look for groups with free space if
5047 the parent directories are nearly full.
5052 Description: Network or server problems during mount may cause partially
5053 mounted clients instead of returning an error.
5054 Details : The config llog parsing code may overwrite the error return
5055 code during mount error handling, returning success instead
5061 Description: MDS can fail to allocate large reply buffers
5062 Details : After long uptimes the MDS can fail to allocate large reply
5063 buffers (e.g. zconf client mount config records) due to memory
5064 fragmentation or consumption by the buffer cache. Preallocate
5065 some large reply buffers so that these replies can be sent even
5066 under memory pressure.
5070 Frequency : rare (liblustre)
5071 Description: fsx running with liblustre complained that using truncate() to
5072 extend the file doesn't work. This patch corrects that issue.
5073 Details : This is the liblustre equivalent of the fix for bug 6196. Fixes
5074 ATTR_SIZE and lsm use in llu_setattr_raw.
5078 Frequency : rare, only 2.6 kernels
5079 Description: Unusual file access patterns on the MDS may result in inode
5080 data being lost in very rare circumstances.
5081 Details : Bad interaction between the ea-in-inode patch and the "no-read"
5082 code in the 2.6 kernel caused the inode and/or EA data not to
5083 be read from disk, causing single-file corruption.
5087 Frequency : rare, only 2.6 filesystems using extents
5088 Description: Heavy concurrent write and delete load may cause data corruption.
5089 Details : It was possible under high-load situations to have an extent
5090 metadata block in the block device cache from a just-unlinked
5091 file overwrite a newly-allocated data block. We now unmap any
5092 metadata buffers that alias just-allocated data blocks.
5096 Frequency : filesystems with default stripe_count larger than 77
5097 Description: lconf+mke2fs fail when formatting filesystem with > 77 stripes
5098 Details : lconf specifies an inode size of 4096 bytes when the default
5099 stripe_count is larger than 77. This conflicts with the default
5100 inode density of 1 per 4096 bytes. Allocate smaller inodes in
5101 this case to avoid pinning too much memory for large EAs.
5103 ------------------------------------------------------------------------------
5105 2005-07-07 Cluster File Systems, Inc. <info@clusterfs.com>
5110 Frequency : rare (extremely heavy IO load with hundreds of clients)
5112 Description: Client is evicted, gets IO error writing to file
5113 Details : lock ordering changes for bug 5492 reintroduced bug 3267 and
5114 caused clients to be evicted for AST timeouts. The fixes in
5115 bug 5192 mean we no longer need to have such short AST timeouts
5116 so ldlm_timeout has been increased.
5119 Frequency : occasional during --force or --failover shutdown under load
5120 Bugzilla : 5949, 4834
5121 Description: Server oops/LBUG if stopped with --force or --failover under load
5122 Details : a collection of import/export refcount and cleanup ordering
5123 issues fixed for safer force cleanup
5126 Frequency : only filesystems larger than 120 OSTs
5127 Bugzilla : 5990, 6223
5128 Description: lfs getstripe would oops on a very large filesystem
5129 Details : lov_getconfig used kfree on vmalloc'd memory
5132 Frequency : only filesystems exporting via NFS to Solaris 10 clients
5133 Bugzilla : 6242, 6243
5134 Description: reading from files that had been truncated to a non-zero size
5135 but never opened returned no data
5136 Details : ll_file_read() reads zeros from no-object files to EOF
5141 Description: A bug in MDS/OSS recovery could cause the OSS to fail an assertion
5142 Details : There's little harm in aborting MDS/OSS recovery and letting it
5143 try again, so I removed the LASSERT and return an error instead.
5145 Severity : enhancement
5147 Description: New debugging infrastructure for tracking down data corruption
5148 Details : The I/O checksum code was replaced to: (a) control it at runtime,
5149 (b) cover more of the client-side code path, and (c) try to narrow
5150 down where problems occurred
5154 Bugzilla : 3819, 4364, 4397, 6313
5155 Description: Racing close and eviction MDS could cause assertion in mds_close
5156 Details : It was possible to get multiple mfd references during close and
5157 client eviction, leading to one thread referencing a freed mfd.
5159 Severity: : enhancement
5160 Bugzilla : 3262, 6359
5161 Description: Attempts to reconnect to servers are now more aggressive.
5162 Details : This builds on the enhanced upcall-less recovery that was added
5163 in 1.4.2. When trying to reconnect to servers, clients will
5164 now try each server in the failover group every 10 seconds. By
5165 default, clients would previously try one server every 25 seconds.
5170 Description: After recovery, certain operations trigger a failed
5171 assertion on a client.
5172 Details : Failing over an mds, using lconf -d --failover, while a
5173 client was doing a readdir() call would cause the client to
5174 LBUG after recovery completed and the readdir() was resent.
5176 Severity : enhancement
5178 Description: Default groups are now added by lconf
5179 Details : You can now run lconf --group <servicename> without having to
5180 manually add groups with lmc.
5183 Frequency : occasional
5185 Description: Nodes with an elan id of 0 trigger a failed assertion
5188 Frequency : always when accessing e.g. tty/console device nodes
5190 Description: tty and some other devices nodes cannot be used on lustre
5191 Details : file's private_data field is used by device data and lustre
5192 values in there got lost. New field was added to struct file to
5193 store fs-specific private data.
5196 Frequency : when exporting Lustre via NFS
5198 Description: NFSD failed occasionally when looking up a path component
5199 Details : NFSD is looking up ".." which was broken in ext3 directories
5200 that had grown large enough to become hashed.
5203 Frequency : Clusters with multiple interfaces not on the same subnet
5205 Description: Nodes will repeatedly try to reconnect to an interface which it
5206 cannot reach and report an error to the log.
5207 Details : Extra peer list entries will be created by lconf with some peers
5208 unreachable. lconf now validates the peer before adding it.
5211 Frequency : Only if a default stripe is set on the filesystem root.
5213 Description: Setting a default stripe on the filesystem root prevented the
5214 filesystem from being remounted.
5215 Details : The client was sending extra request flags in the root getattr
5216 request and did not allocate a reply buffer for the dir EA.
5219 Frequency : occasional, higher if lots of files are accessed by one client
5220 Bugzilla : 6159, 6097
5221 Description: Client trips assertion regarding lsm mismatch/magic
5222 Details : While revalidating inodes the VFS looks up inodes with ifind()
5223 and in rare cases can find an inode that is being freed.
5224 The ll_test_inode() code will free the lsm during ifind()
5225 when it finds an existing inode and then the VFS later attaches
5226 this free lsm to a new inode.
5230 Bugzilla : 6422, 7030
5231 Description: MDS deadlock between mkdir and client eviction
5232 Details : Creating a new file via mkdir or mknod (starting a transaction
5233 and getting the ns lock) can deadlock with client eviction
5234 (gets ns lock and trying to finish a synchronous transaction).
5237 Frequency : occasional
5238 Description: While starting a server, the fsfilt_ext3 module could not be
5240 Details : CFS's improved ext3 filesystem is named ldiskfs for 2.6
5241 kernels. Previously, lconf would still use the ext3 name
5242 when trying to load modules. Now, it will correctly use
5243 ext3 on 2.4 and ldiskfs on 2.6.
5245 Severity : enhancement
5246 Description: The default stripe count has been changed to 1
5247 Details : The interpretation of the default stripe count (0, to lfs
5248 or lmc) has been changed to mean striping across a single
5249 OST, rather than all available. For general usage we have
5250 found a stripe count of 1 or 2 works best.
5252 Severity : enhancement
5253 Description: Add support for compiling against Cray portals.
5254 Details : Conditional compiling for some areas that are different
5258 Frequency : occasional
5259 Bugzilla : 6409, 6834
5260 Description: Creating files with an explicit stripe count may lead to
5261 a failed assertion on the MDS
5262 Details : If some OSTs are full or unavailable, creating files may
5263 trigger a failed assertion on the MDS. Now, Lustre will
5264 try to use other servers or return an error to the
5268 Frequency : occasional
5270 Description: Multiple concurrent overlapping read+write on multiple SMP nodes
5271 caused lock timeout during readahead (since 1.4.2).
5272 Details : Processes doing readahead might match a lock that hasn't been
5273 granted yet if there are overlapping and conflicting lock
5274 requests. The readahead process waits on ungranted lock
5275 (original lock is CBPENDING), while OST waits for that process
5276 to cancel CBPENDING read lock and eventually evicts client.
5278 Severity : enhancement
5280 Description: Initial enabling of flock support for clients
5281 Details : Implements fcntl advisory locking and file status functions.
5282 This feature is provided as an optional mount flag (default
5283 off), and is NOT CURRENTLY SUPPORTED. Not all types of record
5284 locking are implemented yet, and those that are are not guaranteed
5285 to be completely correct in production environments.
5286 mount -t lustre -o [flock|noflock] ...
5289 Frequency : occasional
5291 Description: OSTs running 2.4 kernels but with extents enabled might trip an
5292 assertion in the ext3 JBD (journaling) layer.
5293 Details : The b_committed_data struct is protected by the big kernel lock
5294 in 2.4 kernels, serializing journal_commit_transaction() and
5295 ext3_get_block_handle->ext3_new_block->find_next_usable_block()
5296 access to this struct. In 2.6 kernels there is finer grained
5297 locking to improve SMP performance of the JBD layer.
5301 Description: Changes the "SCSI I/O Stats" kernel patch to default to "enabled"
5303 -----------------------------------------------------------------------------
5305 2005-05-05 Cluster File Systems, Inc. <info@clusterfs.com>
5307 NOTE: Lustre 1.4.2 uses an incompatible network protocol than previous
5308 versions of Lustre. Please update all servers and clients to
5309 version 1.4.2 or later at the same time. You must also run
5310 "lconf --write-conf {config}.xml" on the MDS while it is stopped
5311 to update the configuration logs.
5313 - fix for HPUX NFS client breakage when NFS exporting Lustre (5781)
5314 - mdc_enqueue does not need max_mds_easize request buffer on send (5707)
5315 - swab llog records of type '0' so we get proper header size/idx (5861)
5316 - send llog cancel req to DLM cancel portal instead of cb portal (5515)
5317 - fix rename of one directory over another leaking an inode (5953)
5318 - avoid SetPageDirty on 2.6 (5981)
5319 - don't re-add just-being-destroyed locks to the waiting list (5653)
5320 - when creating new directories, inherit the parent's custom
5321 striping settings if present parent (3048)
5322 - flush buffers from cache before direct IO in 2.6 obdfilter (4982)
5323 - don't hold i_size_sem in ll_nopage() and ll_ap_refresh_count (6077)
5324 - don't hold client locks on temporary worklist from l_lru (5666)
5325 - handle IO errors in 2.6 obdfilter bio completion routine (6046)
5326 - automatically evict dead clients (5921)
5327 - Update file size properly in create+truncate+fstat case (6196)
5328 - Do not unhash mountpoint dentries, do not allow removal of
5330 - Avoid lock ordering deadlock issue with write/truncate (6203,5654)
5331 - reserve enough journal credits in fsfilt_start_log for setattr (4554)
5332 - ldlm_enqueue freed-export error path would always LBUG (6149,6184)
5333 - don't reference lr_lvb_data until after we hold lr_lvb_sem (6170)
5334 - don't overwrite last_rcvd if there is a *_client_add() error (6086)
5335 - Correctly handle reads of files with no objects (6243)
5336 - lctl recover will also mark a device active if deactivate used (5933)
5338 - by default create 1 inode per 4kB space on MDS, per 16kB on OSTs
5339 - allow --write-conf on an MDS with different nettype than client (5619)
5340 - don't write config llogs to MDS for mounts not from that MDS (5617)
5341 - lconf should create multiple TCP connections from a client (5201)
5342 - init scripts are now turned off by default; run chkconfig --on
5343 lustre and chkconfig --on lustrefs to use them
5344 - upcalls are no longer needed for clients to recover to failover
5346 - add --abort-recovery option to lconf to abort recovery on device
5348 - add support for an arbitrary number of OSTs (3026)
5349 - Quota support protocol changes.
5350 - forward compatibility changes to wire structs (6007)
5351 - rmmod NALs that might be loaded because of /etc/modules.conf (6133)
5352 - support for mountfsoptions and clientoptions to the Lustre LDAP (5873)
5353 - improved "lustre status" script
5354 - initialize blocksize for non-regular files (6062)
5355 - added --disable-server and --disable-client configure options (5782)
5356 - introduce a lookup cache for lconf to avoid repeated DB scans (6204)
5357 - Vanilla 2.4.29 support
5358 - increase maximum number of obd devices to 520 (6242)
5359 - remove the tcp-zero-copy patch from the suse-2.4 series (5902)
5360 - Quadrics Elan drivers are now included for the RHEL 3 2.4.21 and
5361 SLES 9 2.6.5 kernels
5362 - limit stripes per file to 160 (the maximum EA size) (6093)
5364 2005-03-22 Cluster File Systems, Inc. <info@clusterfs.com>
5367 - don't LASSERT in ll_release on NULL lld with NFS export (4655, 5760)
5368 - hold NS lock when calling handle_ast_error->del_waiting_lock (5746)
5369 - fix setattr mtime regression from lovcleanup merge (4829, 5669)
5370 - workaround for 2.6 crash in ll_unhash_aliases (5687, 5210)
5371 - small ext3 extents cleanups and fixes (5733)
5372 - improved mballoc code, several small races and bugs fixed (5733, 5638)
5373 - kernel version 43 - fix remove_suid bugs in both 2.4 and 2.6 (5695)
5374 - avoid needless client->OST connect, fix handle mismatch (5317)
5375 - fix DLM error path that led to out-of-sync client, long delays (5779)
5376 - support common vfs-enforced mount options (nodev,nosuid,noexec) (5637)
5377 - fix several locking issues related to i_size (5492,5624,5654,5672)
5378 - don't move pending lock onto export if it is already evicted (5683)
5379 - fix kernel oops when creating .foo in unlinked directory (5548)
5380 - fix deadlock in obdfilter statistics vs. object create (5811)
5381 - use time_{before,after} to avoid timer jiffies wrap (5882)
5382 - shutdown --force/--failover stability (3607,3651,4797,5203,4834)
5383 - Do not leak request if server was not able to process it (5154)
5384 - If mds_open unable to find parent dir, make that negative lookup(5154)
5385 - don't create new directories with extent-mapping (5909, 5936)
5387 - fix lustre/lustrefs init scripts for SuSE (patch from Scali, 5702)
5388 - don't hold the pinger_sem in ptlrpc_pinger_sending_on_import
5389 - change obd_increase_kms to obd_adjust_kms (up or down) (5654)
5390 - lconf, lmc search both /usr/lib and /usr/lib64 for Python libs (5800)
5391 - support for RHEL4 kernel on i686 (5773)
5392 - provide error messages when incompatible logs are encountered (5898)
5394 2005-02-18 Cluster File Systems, Inc. <info@clusterfs.com>
5395 * version 1.4.0.10 (1.4.1 release candidate 1)
5397 - don't keep a lock reference when lock is not granted (4238)
5398 - unsafe list practices (rarely) led to infinite eviction loop (4908)
5399 - add per-fs limit of Lustre pages in page cache, avoid OOM (4699)
5400 - drop import inflight refcount on signal_completed_replay error (5255)
5401 - unlock page after async write error during send (3677)
5402 - handle missing objects in filter_preprw_read properly (5265)
5403 - no transno return for symlink open, don't save no-trasno open (3440)
5404 - don't try to complete elan receive that already failed (4012)
5405 - free RPC server reply state on error (5406)
5406 - clean up thread from ptlrpc_start_thread() on error (5160)
5407 - readahead could read extra page into cache that wasn't ejected (5388)
5408 - prevent races in class_attach/setup/cleanup/detach (5260)
5409 - don't dereference de->d_inode after l_dput of de (5458)
5410 - use "int" for stripe value returned from lock_to_stripe (5544)
5411 - mballoc allocation and error-checking fixes in 2.6 (5504)
5412 - block device patches to fix I/O request sizes in 2.6 (5482)
5413 - look up hostnames for IB nals (5602)
5414 - 2.6 changed lock ordering of 2 semaphores, caused deadlock (5654)
5415 - don't start multiple acceptors for the same port (5277)
5416 - fix incorrect LASSERT in mds_getattr_name (5635)
5417 - export a proc file for general "ping" checking (5628)
5418 - fix "lfs check" to not block when the MDS is down (5628)
5420 - service request history (4965)
5421 - put {ll,lov,osc}_async_page structs in a single slab (4699)
5422 - create an "evict_client" /proc entry on OSTs, like the MDS has
5423 - fix mount usage message, return errors per mount(8) (5168)
5424 - change grep [] to grep "[]" in tests so they work in more UMLs
5425 - fix ppc64/x86_64 spec to use %{_libdir} instead of /usr/lib (5389)
5426 - remove ancient LOV_MAGIC_V0 EA support (5047)
5427 - add "disk I/Os in flight" and "I/O req time" stats in obdfilter
5428 - align r/w RPCs to PTLRPC_MAX_BRW_SIZE boundary for performance (3451)
5429 - allow readahead allocations to fail when low on memory (5383)
5430 - mmap locking landed again, after considerable improvement (2828)
5431 - add get_hostaddr() to lustreDB.py for LDAP support (5459)
5433 2004-11-23 Cluster File Systems, Inc. <info@clusterfs.com>
5436 - send OST transaction number in read/write reply to free req (4966)
5437 - don't ASSERT in ptl_send_rpc() if we run out of memory (5119)
5438 - lock /proc/sys/portals/routes internal state, avoiding oops (4827)
5439 - the watchdog thread now runs as interruptible (5246)
5440 - flock/lockf fixes (but it's still disabled, pending 5135)
5441 - don't use EXT3 constants in llite code (5094)
5442 - memory shortage at startup could cause assertion (5176)
5444 - reorganization of lov code
5445 - single portals codebase
5447 - add extents/mballoc support (5025)
5448 - direct I/O reads in the obdfilter (4048)
5449 - kernel patches from LNXI for 2.6 (bluesmoke, perfctr, mtd, kexec)
5451 tbd Cluster File Systems, Inc. <info@clusterfs.com>
5454 - send OST transaction number in read/write reply to free req (4966)
5455 - don't ASSERT in ptl_send_rpc() if we run out of memory (5119)
5456 - lock /proc/sys/portals/routes internal state, avoiding oops (4827)
5457 - the watchdog thread now runs as interruptible (5246)
5458 - handle missing objects in filter_preprw_read properly (5265)
5459 - unsafe list practices (rarely) led to infinite eviction loop (4908)
5460 - drop import inflight refcount on signal_completed_replay error (5255)
5461 - unlock page after async write error during send (3677)
5462 - return original error code on reconstructed replies (3761)
5463 - no transno return for symlink open, don't save no-trasno open (3440)
5465 - add pid to ldlm debugging output (4922)
5466 - bump the watchdog timeouts -- we can't handle 30sec yet
5467 - extra debugging for orphan dentry/inode bug (5259)
5469 2004-11-16 Cluster File Systems, Inc. <info@clusterfs.com>
5472 - fix TCP_NODELAY bug, which caused extreme perf regression (5134)
5473 - allocate qswnal tx descriptors singly to avoid fragmentation (4504)
5474 - don't LBUG on obdo_alloc() failure, use OBD_SLAB_ALLOC() (4800)
5475 - fix NULL dereference in /proc/sys/portals/routes (4827)
5476 - allow failed mdc_close() operations to be interrupted (4561)
5477 - stop precreate on OST before MDS would time out on it (4778)
5478 - don't send partial-page writes before EOF from client (4410)
5479 - discard client grant for sub-page writes on large-page clients (4520)
5480 - don't free dentries not owned by NFS code, check generation (4806)
5481 - fix lsm leak if mds_create_objects() fails (4801)
5482 - limit debug_daemon file size, always print CERROR messages (4789)
5483 - use transno after validating reply (3892)
5484 - process timed out requests if import state changes (3754)
5485 - update mtime on OST during writes, return in glimpse (4829)
5486 - add mkfsoptions to LDAP (4679)
5487 - use ->max_readahead method instead of zapping global ra (5039)
5488 - don't interrupt __l_wait_event() during strace
5490 - add software watchdogs to catch hung threads quickly (4941)
5491 - make lustrefs init script start after nfs is mounted
5492 - fix CWARN/ERROR duplication (4930)
5493 - return async write errors to application if possible (2248)
5494 - add /proc/sys/portal/memused (bytes allocated by PORTALS_ALLOC)
5495 - print NAL number in %x format (4645)
5496 - update barely-supported suse-2.4.21-171 series (4842)
5497 - support for sles 9 %post scripts
5498 - support for building 2.6 kernel-source packages
5499 - support for sles km_* packages
5501 2004-10-07 Cluster File Systems, Inc. <info@clusterfs.com>
5504 - ignore -ENOENT errors in osc_destroy (3639)
5505 - notify osc create thread that OSC is being cleaned up (4600)
5506 - add nettype argument for llmount in #5d in conf-sanity.sh (3936)
5507 - reconstruct ost_handle() like mds_handle() (4657)
5508 - create a new thread to do import eviction to avoid deadlock (3969)
5509 - let lconf resolve symlinked-to devices (4629)
5510 - don't unlink "objects" from directory with default EA (4554)
5511 - hold socknal file ref over connect in case target is down (4394)
5512 - allow more than 32000 subdirectories in a single directory (3244)
5513 - fix blocks count for O_DIRECT writes (3751)
5514 - OST returns ENOSPC from object create when no space left (4539)
5515 - don't send truncate RPC if file size isn't changing (4410)
5516 - limit OSC precreate to 1/2 of value OST considers bogus (4778)
5517 - bind to privileged port in socknal and tcpnal (3689)
5519 - rate limit CERROR/CWARN console message to avoid overload (4519)
5520 - GETFILEINFO dir ioctl returns LOV EA + MDS stat in 1 call (3327)
5521 - basic mmap support (3918)
5522 - kernel patch series update from b1_4 (4711)
5524 2004-09-16 Cluster File Systems, Inc. <info@clusterfs.com>
5527 - avoid crash during MDS cleanup with OST shut down (2775)
5528 - fix loi_list_lock/oig_lock inversion on interrupted IO (4136)
5529 - don't use bad inodes on the MDS (3744)
5530 - dynamic object preallocation to improve recovery speed (4236)
5531 - don't hold spinlock over lock dumping or change debug flags (4401)
5532 - don't zero obd_dev when it is force cleaned (3651)
5533 - print grants to console if they go negative (4431)
5534 - "lctl deactivate" will stop automatic recovery attempts (3406)
5535 - look for existing locks in ldlm_handle_enqueue() (3764)
5536 - don't resolve lock handle twice in recovery avoiding race (4401)
5537 - revalidate should check working dir is a directory (4134)
5539 - don't always mark "slow" obdfilter messages as errors (4418)
5541 2004-08-24 Cluster File Systems, Inc. <info@clusterfs.com>
5544 - don't close LustreDB during write_conf until it is done (3860)
5545 - fix typo in lconf for_each_profile (3821)
5546 - allow dumping logs from multiple threads at one time (3820)
5547 - don't allow multiple threads in OSC recovery (3812)
5548 - fix debug_size parameters (3864)
5549 - fix mds_postrecov to initialize import for llog ctxt (3121)
5550 - replace config semaphore with spinlock (3306)
5551 - be sure to send a reply for a CANCEL rpc with bad export (3863)
5552 - don't allow enqueue to complete on a destroyed export (3822)
5553 - down write_lock before checking llog header bitmap (3825)
5554 - recover from lock replay timeout (3764)
5555 - up llog sem before sending rpc (3652)
5556 - reduce ns lock hold times when setting kms (3267)
5557 - change a dlm LBUG to LASSERTF, to maybe learn something (4228)
5558 - fix NULL deref and obd_dev leak on setup error (3312)
5559 - replace some LBUG about llog ops with error handling (3841)
5560 - don't match INVALID dentries from d_lookup and spin (3784)
5561 - hold dcache_lock while marking dentries INVALID and hashing (4255)
5562 - fix invalid assertion in ptlrpc_set_wait (3880)
5564 - add libwrap support for the TCP acceptor (3996)
5565 - add /proc/sys/portals/routes for non-root route listing (3994)
5566 - allow setting MDS UUID in .xml (2580)
5567 - print the stack of a process that LBUGs (4228)
5569 2004-07-14 Cluster File Systems, Inc. <info@clusterfs.com>
5572 - don't cleanup request in ll_file_open() on failed MDS open (3430)
5573 - make sure to unset replay flag from failed open requests (3440)
5574 - if default stripe count is 0, use OST count for inode size (3636)
5575 - update parent mtime/ctime on client for create/unlink (2611)
5576 - drop dentry ref in ext3_add_link from open_connect_dentry (3266)
5577 - free recovery state on server during a forced cleanup (3571)
5578 - unregister_reply for resent reqs (3063)
5579 - loop back devices mounting and status check on 2.6 (3563)
5580 - fix resource-creation race that can provoke i_size == 0 (3513)
5581 - don't try to use bad inodes returned from MDS/OST fs lookup (3688)
5582 - more debugging for page-accounting assertion (3746)
5583 - return -ENOENT instead of asserting if ost getattr+unlink race (3558)
5584 - avoid deadlock after precreation failure (3758)
5585 - fix race and lock order deadlock in orphan handling (3450, 3750)
5586 - add validity checks when grabbing inodes from l_ast_data (3599)
5588 - add /proc/.../recovery_status to obdfilter (3428)
5589 - lightweight CDEBUG infrastructure, debug daemon (3668)
5590 - change default OSC RPC parameters to be better on small clusters
5591 - turn off OST read cache for files smaller than 32MB
5592 - install man pages and include them in rpms (3100)
5593 - add new init script for (un)mounting lustre filesystems (2593)
5594 - run chkconfig in %post for init scripts (3701)
5595 - drop scimac NAL (unmaintained)
5597 2004-06-17 Cluster File Systems, Inc. <info@clusterfs.com>
5600 - clean kiobufs before and after use (3485)
5601 - strip trailing '/'s before comparing paths with /proc/mounts (3486)
5602 - remove assertions to work around "in-flight rpcs" recovery bug (3063)
5603 - change init script to fail more clearly if not run as root (1528)
5604 - allow clients to reconnect during replay (1742)
5605 - fix ns_lock/i_sem lock ordering deadlock for kms update (3477)
5606 - don't do DNS lookups on NIDs too small for IP addresses (3442)
5607 - re-awaken ptlrpcd if new requests arrive during check_set (3554)
5608 - fix cond_resched (3554)
5609 - only evict unfinished clients after recovery (3515)
5610 - allow bulk resend, prevent data loss (3570)
5611 - dynamic ptlrpc request buffer allocation (2102)
5612 - don't allow unlinking open directory if it isn't empty (2904)
5613 - set MDS/OST threads to umask 0 to not clobber client modes (3359)
5614 - remove extraneous obd dereference causing LASSERT failure (3334)
5615 - don't use get_cycles() when creating temp. files on the mds (3156)
5616 - hold i_sem when setting i_size in ll_extent_lock() (3564)
5617 - handle EEXIST for set-stripe, set proper directory name (3336)
5619 - servers can dump a log evicting a client - lustre.dump_on_timeout=1
5620 - fix ksocknal_fmb_callback() error messages (2918)
5622 2004-05-27 Cluster File Systems, Inc. <info@clusterfs.com>
5625 - don't copy lvb into (possibly NULL) reply on error (2983)
5626 - don't deref dentry after dput, don't free lvb on error (2922)
5627 - use the kms to determine writeback rpc length (2947)
5628 - increment oti_logcookies when osc is inactive (2948)
5629 - update client's i_blocks count via lvb messages (2543)
5630 - handle intent open/close of special files properly (1557)
5631 - mount MDS with errors=remount-ro, like obdfilter (2009)
5632 - initialize lock handle to avoid ASSERT on error cleanup (3057)
5633 - don't use cancelling-locks' kms values (2947)
5634 - use highest lock extent for kms, not last one (2925)
5635 - don't dereference ERR_PTR() dentry in error handling path (3107)
5636 - fix thread race in portals_debug_dumplog() (3122)
5637 - create lprocfs device entries at setup instead of at attach (1519)
5638 - common AST error handler, don't evict client on completion race (3145)
5639 - zero nameidata in detach_mnt in 2.6 (3118)
5640 - verify d_inode after revalidate_special is valid in 2.6 (3116)
5641 - use lustre_put_super() to handle zconf unmounts in 2.6 (3064)
5642 - initialize RPC timeout timer earlier for 2.6 (3219)
5643 - don't dereference NULL reply buffer if mdc_close was never sent (2410)
5644 - print nal/nid for unknown nid (3258)
5645 - additional checks for oscc recovery before doing precreate (3284)
5646 - fix ll_extent_lock() error return code for 64-bit systems (3043)
5647 - don't crash in mdc_close for bad permissions on open (3285)
5648 - zero i_rdev for non-device files (3147)
5649 - clear page->private before handing to FS, better assertion (3119)
5650 - tune the read pipeline (3236)
5651 - fix incorrect decref of invalidated dentry (2350)
5652 - provide read-ahead stats and refine rpc in flight stats (3328)
5653 - don't hold journal transaction open across create RPC (3313)
5654 - update atime on MDS at close time (3265)
5655 - close LDAP connection when recovering to avoid server load (3315)
5656 - update iopen-2.6 patch with fixes from 2399,2517,2904 (3301)
5657 - don't leak open file on MDS after open resend (3325)
5658 - serialize filter_precreate and filter_destroy_precreated (3329)
5659 - loop device shouldn't call sync_dev() for nul device (3092)
5660 - clear page cache after eviction (2766)
5661 - resynchronize MDS->OST in background (2824)
5662 - refuse to mount the same filesystem twice on same mountpoint (3394)
5663 - allow llmount to create routes for mounting behind routers (3320)
5664 - push lock cancellation to blocking thread for glimpse ASTs (3409)
5665 - don't call osc_set_data_with_check() for TEST_LOCK matches (3159)
5666 - fix rare problem with rename on htree directories (3417)
5668 - allow default OST striping configuration per directory (1414)
5669 - fix compilation for qswnal for 2.6 kernels (3125)
5670 - increase maximum number of MDS request buffers for large systems
5671 - change liblustreapi to be useful for external progs like lfsck (3098)
5672 - increase local configuration timeout for slow disks (3353)
5673 - allow configuring ldlm AST timeout - lustre.ldlm_timeout=<seconds>
5675 2004-03-22 Cluster File Systems, Inc. <info@clusterfs.com>
5678 - fixes for glimpse AST timeouts / incorrectly 0-sized files (2818)
5679 - don't overwrite extent policy data in reply if lock was blocked (2901)
5680 - drop filter export grants atomically with removal from device (2663)
5681 - del obd_self_export from work_list in class_disconnect_exports (2908)
5682 - don't LBUG if MDS recovery times out during orphan cleanup (2530)
5683 - swab reply message in mdc_close, other PPC fixes (2464)
5684 - fix destroying of named logs (2325)
5685 - overwrite old logs when running lconf --write_conf (2264)
5686 - bump LLOG_CHUNKSIZE to 8k to allow for larger clusters (2306)
5687 - fix race in target_handle_connect (2898)
5688 - mds_reint_create() should take same inode create lock (2926)
5689 - correct journal credits calculated for CANCEL_UNLINK_LOG (2931)
5690 - don't close files for self_export to avoid uninitialized obd (2936)
5691 - allow MDS with the same name as client node (2939)
5692 - hold dentry reference for closed log files for unlink (2325)
5693 - reserve space for all logs during transactions (2059)
5694 - don't evict page beyond end of stripe extent (2925)
5695 - don't oops on a deleted current working directory (2399)
5696 - handle hard links to targets without a parent properly (2517)
5697 - don't dereference NULL lock when racing during eviction (2867)
5698 - don't grow lock extents when lots of conflicting locks (2919)
5700 2004-03-04 Cluster File Systems, Inc. <info@clusterfs.com>
5703 - account for cache space usage on clients to avoid data loss (974)
5704 - lfsck support in lustre kernel code (2349)
5705 - reduce journal credits needed for BRW writes (2370)
5706 - orphan handling to avoid losing space on client/server crashes
5707 - ptlrpcd can be blocked, stopping ALL progress (2477)
5708 - use lock value blocks to assist in proper KMS, faster stat (1021)
5709 - takes i_sem instead of DLM locks internally on obdfilter (2720)
5710 - recovery for initial connections (2355)
5711 - fixes for mds_cleanup_orphans (1934)
5712 - abort_recovery crashes MDS in b_eq (mds_unlink_orphan) (2584)
5713 - block all file creations until orphan recovery completes (1901)
5714 - client remove rq_connection from request struct (2423)
5715 - conf-sanity test_5, proper cleanup in umount log not availale (2640)
5716 - recovery timer race (2670)
5717 - mdc_close recovey bug (2532)
5718 - ptlrpc cleanup bug (2710)
5719 - mds timeout on local locks (2588)
5720 - namespace lock held during RPCs (2431)
5721 - handle interrupted sync write properly (2503)
5722 - don't try to handle a message that hasn't been replied to (2699)
5723 - client assert failure during cleanup after abort recovery (2701)
5724 - leak mdc device after failed mount (2712)
5725 - ptlrpc_check_set allows timedout requests to complete (2714)
5726 - wait for inflight reqs when ptlrpcd finishes (2710)
5727 - make sure unregistered services are removed from the srv_list
5728 - reset bulk XID's when resending them (caught by 1138 test)
5729 - unregister_bulk after timeout
5730 - fix lconf error (2694)
5731 - handle write after unfinished setstripe, stripe-only getstripe (2388)
5732 - readahead locks pages, leaves pending causing memory pressure (2673)
5733 - increase OST request buffers to 4096 on large machines (2729)
5734 - fix up permission of existing directories in simple_mkdir (2661)
5735 - init deleted item, add assertions ptlrpc_abort_inflight() (2725)
5736 - don't assign transno to errored transactions (2742)
5737 - don't delete objects on OST if given a bogus objid from MDS (2751)
5738 - handle large client PAGE_SIZE readdir on small PAGE_SIZE MDS (2777)
5739 - if rq_no_resend, then timeout request after recovery (2432)
5740 - fix MDS llog_logid record size, 64-bit array alignment (2733)
5741 - don't call usermode_helper from ptlrpcd, DEFAULT upcall (2773)
5742 - put magic in mount.lustre data, check for bad/NULL mount data (2529)
5743 - MDS recovery shouldn't delete objects that it has given out (2730)
5744 - if enqueue arrives after completion, don't clobber LVB (2819)
5745 - don't unlock pages twice when trigger_group_io returns error (2814)
5746 - don't deref NULL rq_repmsg if ldlm_handle_enqueue failed (2822)
5747 - don't write pages to disk if there was an error (1450)
5748 - don't ping imports that have recovery disabled (2676)
5749 - take buffered bytes into account when balancing socknal conn (2817)
5750 - hold a DLM lock over readdir always, use truncate_inode_pages (2706)
5751 - reconnect unlink llog connection after MDS reconnects to OST (2816)
5752 - remove little-endian swabbing of llog records (1987)
5753 - set/limit i_blksize to LL_MAX_BLKSIZE on client (2884)
5754 - retry reposting request buffers if they fail (1191)
5755 - grow extent at grant time to avoid granting a revoked lock (2809)
5756 - lock revoke doesn't evict page if covered by a second lock (2765)
5757 - disable VM readahead to avoid reading outside lock extents (2805)
5759 - return LL_SUPER_MAGIC from statfs for the filesystem type (1972)
5760 - updated kernel patches for hp-2.4.20 kernel (2681)
5762 2004-02-07 Cluster File Systems, Inc. <info@clusterfs.com>
5765 - fix truncated write corruption (2366)
5766 - fix for failed assertion in iopen_connect_dentry (1792,2517)
5768 - don't flag the ptlrpcd thread with PF_MEMALLOC (2636)
5769 - ensure len(uuid) < 37 in lmc (1171)
5770 - fix ia64 OOPS in llog_test (2255)
5771 - zero end of page at obdfilter for partial page writes (2648)
5772 - don't leave stale dentries around after renames (bug 2428)
5773 - fix timeouts when evicting a client with a single lock held (2642)
5774 - set deadline for the initial HELLO message to drain (2634)
5775 - print out dotted-quad IP addresses in the socknal (2302)
5777 - additional debugging for MDS client eviction problem (2443)
5778 - fix mkfsoptions support for osts (2603, 2604)
5780 2004-01-27 Cluster File Systems, Inc. <info@clusterfs.com>
5783 - add series for the vanilla 2.6.0 kernel
5784 - add series for the vanilla 2.4.24 kernel
5785 - add series for a cray x86/64 UL kernel drop
5786 - fix xattr patches for the vanilla 2.4.19 series
5788 - generate true UUIDs in lmc (1171)
5789 - have portals stack dumping break in UML (2466)
5790 - avoid bad dchild deref; avoid inum lock w/o creation (2362)
5791 - allocate with _NOFS in ldlm to avoid deadlock (1933)
5792 - wake callback waiting threads on client eviction (2460)
5793 - Add --ptldebug and --subsystem to lmc (1719)
5794 - update assertion to allow safe interrupt allocation
5795 - set rq_no_resend for cancel requests (2432)
5796 - recalculate ptlrpcd timeout after resend (2494)
5797 - call vfs_rmdir when removing pending directories (2368)
5798 - fix renaming a file to itself (2429)
5799 - lmc creates a default one-stripe lov (2454)
5800 - expand procfs space to handle large clusters (2326)
5801 - increase UML stack to avoid overflow
5802 - update lconf's list of debug and subsystem masks
5803 - fix lfs find --obd (2510)
5804 - /proc tunable for disabling filter read caching (2591)
5805 - stop rpm packages from altering slapd.conf (2301)
5806 - disable nagle in the socknal under 0conf (2578)
5807 - choose mds inode size based on stripe count (2572)
5808 - fix kernel-source rpm problems (2516)
5810 - add --disable-doc to avoid pdf generation (2421)
5811 - update documentation, tests, type-os, comments
5812 - avoid format warnings on ia64
5813 - remove the TOE NAL
5814 - tiny code cleanups by removing unused fields
5816 2004-01-07 Cluster File Systems, Inc. <info@clusterfs.com>
5819 - fix obvious semaphore misuse in as-yet-unused setattr path (2348)
5820 - remove the most blatant lies from BUILDING file (2371)
5821 - change default debug level to reasonable production setting
5822 - reduce client side cache size to reduce cache flush time
5823 - reduce max RPCs in flight to avoid unnecessary file fragmentation
5824 - make TCP zerocopy and pinger support enabled by default (2476)
5825 - sync writes completed after process exits caused crashes (2319)
5826 - maintain correct mount count on the MDS (2356)
5827 - backout 1557, because 2316 wasn't really fixed
5828 - better file I/O statistics gathering in /proc
5829 - don't take unnecessary, deadlock-inducing bug in readpage (2383)
5830 - another kernel patch to fix zero-copy TCP function export
5831 - don't take duplicate lock when processing re-sent getattr (2420)
5832 - lctl uses obd_self_export instead of creating new conn (2353)
5833 - MDS/OST recovery case which requires object creation asserted (2425)
5834 - move lfs from /usr/sbin to /usr/bin in packages
5835 - fix race between mds_client_add and mds_client_free (2417)
5836 - use kmalloc instead of slabs in portals (2430)
5837 - don't create duplicate records when a failover MDS is present (2442)
5838 - remove unnecessary mount age check (2332)
5839 - don't remove directory inodes from locks prematurely (2451)
5840 - don't break if MDS service name is the same as hostname (2103)
5841 - fix races in client write RPC generation when cache full (2482)
5843 2003-12-13 Cluster File Systems, Inc. <info@clusterfs.com>
5846 - remove now-unused request->rq_obd (278)
5847 - if an allocation fails, print out how much memory we've used (1933)
5848 - use PORTAL_SLAB_ALLOC for structures, to get GFP_MEMALLOC (1933)
5849 - add the "configurable stack size" patch to most series files (1256)
5850 - ability to write large log records, for 100+ OST configs (2306)
5851 - fix NULL deref when filter_prep fails (2314)
5852 - fix operator precedence error in filter_sync
5853 - dynamic allocation of socknal TX descriptors (2315)
5854 - fix a missed case in the GFP_MEMALLOC patch, can cause deadlock (2310)
5855 - fix gcc 2.96 compilation problem in xattr kernel patch (2294)
5856 - ensure that CWARN messages in Portals always get to the syslog
5857 - __init/__exit are not for prototype decls (ldlm_init/exit)
5858 - x86-64 compile warning fixes
5859 - fix gateway LMC keyword conflict (2318)
5860 - fix MDS lock inversions in getattr/reint paths (1844)
5861 - fix a rare lock re-ordering bug, which caused deadlock (2322)
5862 - fix i_sem/journal inversion in fsfilt_ext3_write_record (2306)
5863 - DLM race condition prevented some lock evictions (2328)
5864 - ENOMEM detection and retry on socknal sends (2230)
5865 - use GFP_NOFS throughout Lustre, to combat ENOMEM (2230)
5866 - move osc_rpcd into ptlrpc, for use in MDC and others (2329)
5867 - protect MDS inode fsdata with stronger locking; fixes assertion (2313)
5868 - better error messages when a client is rejected during recovery (1505)
5869 - avoid cancelling locks which were never granted, after failure (2330)
5870 - fix i_sem/journal inversion in mds_client_add (2333)
5871 - fix truncate/getattr lock cycle deadlock (2334)
5872 - use rpcd to send close; allows resend after timeout, avoid leak (1897)
5873 - fix two rare exit paths which could leak an l_lock() ref (2321)
5874 - fencepost error in MDS/OST orphan recovery (2226)
5875 - make log record alignment 8 bytes (1988)
5876 - lstripe now fails when requested offset > ost_count (2237)
5877 - ensure that all kernel series have a complete list.h (1607)
5878 - fix crashes in special-file operations (2316)
5879 - lctl create/brw OID mismatch, caused by obsolete filter loop (2339)
5881 - allow configurable automake binary, for testing new versions
5882 - small update to the lfs documentation
5884 2003-12-03 Cluster File Systems, Inc. <info@clusterfs.com>
5886 * fix negative export reference count in fsfilt_sync (2312)
5888 2003-12-01 Cluster File Systems, Inc. <info@clusterfs.com>
5889 * release candidate 0.9.1
5891 - orphans are moved into the PENDING directory for possible recovery
5892 - replayed opens now open by fid for orphan/rename safety (1042)
5893 - last close of an orphan inode generates a transno (683)
5894 - chdir() and mount() now pin the directory entry (1020)
5895 - avoid CERROR in normal ll_setattr_raw() error case (1500)
5896 - discard very old requests without processing them (1502)
5897 - remove some common, well-understood CERRORs (1505)
5898 - require O_DIRECT I/O to be page-sized to workaround IA64 crash (1609)
5899 - clear "grant" flags in OST replies until OST grant code lands (1644)
5900 - fix read performance by not clobbering i_blksize on client (1598)
5901 - fix __ldlm_handle2lock oops by not dereferencing lock after PUT (1625)
5902 - make LRU size a /proc tunable, clears locks when reduced (707)
5903 - fix some lprocfs rot that prevented ptlbd from loading (1732)
5904 - server locks take references on exports now (1558)
5905 - build fixes for 2.4.20-rh trees (1663)
5906 - return an error from lov_create if all OSCs are inactive (1751)
5907 - fix import levels when a reconnect happens without a timeout (1597)
5908 - exit early from mds_open if we get a lookup error (1749)
5909 - partial page read at EOF wouldn't wait for disk before sending (1642)
5910 - avoid NULL deref in obdfilter when reading page past EOF (1592)
5911 - avoid LASSERT in ll_intent_lock if server failed very early (1090)
5912 - fix LBUG in ll_it_open_error with rc = -2 (1861)
5913 - write/truncate lock inversion (1639)
5914 - Don't auto-load obdclass, portals modules during cleanup (1495)
5915 - fix timestamps from jumping to "now" (1763)
5916 - extra journal assertions (1648)
5917 - add an extra multiunlink test (1771)
5918 - fix read_record/write_record API (1776)
5919 - fix leak of offset_extent, possible incorrect i_size later (1772)
5920 - fix lasserts in mis-matched transnos during open-unlink testing (1541)
5921 - Debugging for the kqswnal_get_idle_tx problems (1820)
5922 - Allow recovery to be attempted multiple times (1536)
5923 - Write out MDS last_rcvd file after it is first created (1600)
5924 - Fix tx_descriptor leak in failed transmit situations (1827)
5925 - ext3 journaling fixes for assertion failure after IO error (1871)
5926 - class_export_put() on freed export after completion AST error (1896)
5927 - Fix revalidate looping in VFS (1322)
5928 - Don't access a freed export during MDS_REINT timeout (1521)
5929 - Add open-unlink recovery support on the MDS (1673,1764)
5930 - Return an error if no MDS data was read from last_rcvd (1946)
5931 - Fix for lookup "." or ".." crash on error (1932,1931,1935)
5932 - Don't setup a disk device that doesn't match exported UUID (317)
5933 - Reduce bulk RPC timeout to avoid cascading client/OST failures (1845)
5934 - avoid committing NULL handle in force close
5935 - local.sh is now a one-stripe LOV configuration
5936 - POSIX utime.4 -EPERM on FIFO not owned by user (56)
5937 - fix ext3 htree duplicate directory entry corruption (1516)
5938 - POSIX creat.13, fstat.1, open.18, stat.3 new file atime/mtime (2020)
5939 - update to new LOV EA format (2097)
5940 - interoperability for different PAGE_SIZE/wordsize (686,1821,1343,2042)
5942 2003-06-15 Phil Schwan <phil@clusterfs.com>
5945 - imports and exports cleanup too early, need refcounts (349, 879, 1045)
5946 - per-import/export recovery handling (958, 931, 959)
5947 - multiple last-rcvd slots, for serving multiple FSes (949)
5948 - connections are again shared between multiple imp/exports (963, 964)
5949 - "umount -f" would hang if any requests needed to be sent (393, 978)
5950 - avoid pinning large req buffer by copying for queued messages (989)
5951 - add "uuid" to "lctl device" command to help upcalls (991)
5952 - "open" RPCs with transnos would confuse recovery counters (1037)
5953 - do proper endian conversion of all wire messages (288, 340, 891)
5954 - remove OST bulk get LBUGs, fix ost_brw_write cleanup (1126)
5955 - call waiting locks callback from LDLM recovery thread (1127, 1151)
5956 - fix ptlrpc_connection leak in target_handle_connect (1174)
5957 - fix import refcounting bug in OST and MDS cleanup (1134)
5958 - if an invalid-at-open-time OSC returned before close(), LBUG (1150)
5959 - fix very unlikely obd_types race condition (501)
5960 - remove osc_open hack for echo_client (1187)
5961 - we leaked exports/dlmimps for forcibly disconnected clients (1143)
5962 - a failure in read_inode2 leads to deadlock (1139)
5963 - cancel ack-locks as soon as transaction is committed (1072)
5964 - fix major leaks and crashes in the bulk I/O path (937, 1057)
5965 - make sure to commitrw after any preprw to avoid deadlock (1162)
5966 - failing to execute a file in a lustre FS would lock inode (1203)
5967 - small DEBUG_REQ fix to avoid dereferencing a NULL (1227)
5968 - don't ASSERT while cleaning up an incompletely-setup obd (1248)
5969 - obd_uuid2tgt would walk off the end of the list (1255)
5970 - on IA64 the osc would give portals incorrect bulk size (1258)
5971 - fix debug daemon ioctl interface; allows daemon on ia64 (1274)
5972 - fix lock inversion caused by new llite matching code (1282)
5973 - limit the number of dirty pages on a client to 10MB (1286)
5974 - timed out locks were not being corrected cancelled (1289)
5975 - fix O_DIRECT above 4GB on IA-32 (1292)
5976 * major user-visible changes
5977 - fail out/fail over policy now controlled by the upcall (993)
5979 - add OBD_PING to check server availability and failure (954)
5980 - lustre messages are now sent in sending host order (288, 340, 891)
5981 - add eadatalen to MDS getattr reply (340)
5982 - OST read replies may contain second buffer, with per-page status (593)
5984 2003-03-11 Phil Schwan <phil@clusterfs.com>
5987 - LDLM_DEBUG macro fix, for gcc 3.2 (850)
5988 - failed open()s could cause deadlock; fixed (867, 869)
5989 - stop cancelling OST locks when files are closed (481)
5990 - overlapping XID spaces caused network corruption (851, 853)
5991 - fix unsafe fsfilt counter arithmetic; change to atomic_t
5992 - setattr_raw added, to do single-RPC, server-side setattrs
5993 - lmc/lconf syntax change for OST UUIDs
5994 - fix crashy race condition between ptlrpc_free_req and osc_close
5995 - don't use request in mdc_enqueue if we hit a timeout (889)
5996 - don't set the inode i_size for regular files from the MDS (896)
5997 - handle out of order completion AST (842)
5998 - don't LBUG if a lock request times out after receiving AST (913)
5999 - avoid d_rehash race in ll_find_alias by rehashing inside dcache_lock
6000 - if a bad lock AST arrives, send an error instead of dropping entirely
6001 - return 0 from revalidate2 if ll_intent_lock returns -EINTR (912)
6002 - fix leak in bulk IO when only partially completed (899, 900, 926)
6003 - fix O_DIRECT for ia64 (55)
6004 - (almost) eliminate Lustre-kernel-thread effects on load average (722)
6005 - C-z after timeout could hang a process forever; fixed (977)
6007 - client-side I/O cache (678, 924, 929, 941, 970)
6009 - READPAGE and SETATTRs which don't take server-side locks get
6012 2003-02-11 Phil Schwan <phil@clusterfs.com>
6015 - Fix ldlm_lock_match on the MDS to avoid matching remote locks (592)
6016 - Fix fsfilt_extN_readpage() to read a full page of directory
6017 entries, or fake the remainder if PAGE_SIZE != blocksize (500)
6018 - Avoid extra mdc_getattr() in ll_intent_lock when possible (534, 604)
6019 - Fix imbalanced LOV object allocation and out-of-bound access (469)
6020 - Most intent operations were removed, in favour of a new RPC mode
6021 that does a single RPC to the server and bypasses most of the VFS
6022 - All LDLM resource ID arrays were removed in favour of ldlm_res_id
6023 - Aggressively cancel local locks on DLM servers
6024 - mds_reint_unlink sends EA to the client if it's the last nlink.
6025 client uses that EA to unlink OST objects.
6026 - mds_reint_{rename,unlink,link} were rewritten to take ordered locks
6027 - recursive symlinks were fixed (439)
6028 - fixed NULL deref in DEBUG_REQ
6029 - filter_update_lastobjid no longer calls sync, which annoyed extN
6030 - fixed multi-client small-writes to a single file problem (445)
6031 - fixed mtime updates during file writes (607)
6032 - fixed vector writes on obdfilter causing problems when ENOSPC (670)
6033 - fixed bug in obd_brw_read/write() (under guise of testing 367)
6034 - fixed Linux OST size reporting problem (444, 656)
6035 - OST now updates object mtime with writes or setattr (607, 619)
6036 - client verifies file size before zeroing page past EOF (445)
6037 - OST now writes last allocated objid to disk with allocation (108)
6038 - LOV on echo now works (409)
6040 - mds_reint_unlink sends a new buffer, with the EA included. this
6041 buffer is only valid if body->valid & OBD_MD_FLEASIZE, which is only
6042 set if a regular file was being unlinked, and it was the last link
6043 - use PtlGet from the target for bulk writes (315)
6044 - OST now updates object mtime with writes or setattr (607, 619)
6045 - LDLM now has a grant-time callback to revalidate locked items, if
6047 - Many MDS operations were reorganized to combat race conditions
6049 - Merge b_intel branch (updated lprocfs code) - now at /proc/fs/lustre
6050 - configure check to avoid gcc version 2.96 20000731-2.96-98) (606)
6052 2003-01-06 Andreas Dilger <adilger@clusterfs.com>
6055 - Fully reactivate OST imports after reconnection (512, others)
6056 - Make sure client sees our -ENOTCONN from mds_handle (513 - partial)
6057 - More graceful error handling for truncating on dead OST (515)
6058 - Don't error out unless we're actually accessing dead stripes (474)
6059 - Fix garbage sizes when stripes are missing (410)
6060 - LRU counters were broken, causing constant lock purge (433, 432)
6061 - garbage on read from stripes with failed OSTs (441)
6062 - mark OSCs as active before reconnecting during recovery (438)
6063 - lov_enqueue and lov_cancel need to handle inactive OSTs (403)
6064 - lfind did not preserve OST order in output (443)
6065 - symlinks cause hung clients, incorrect data (439)
6066 - stop dereferencing request after dropping refcount (457)
6067 - don't LASSERT(spin_is_locked) on non-SMP (455)
6068 - fixes for many rename() bugs
6069 - fstat didn't correctly synchronize attributes (399)
6070 - server must handle lock cancellation during blocking AST prep (487)
6071 - bulk descriptors were free()d too soon (511)
6072 - fix paths in lconf, which would load incorrect modules (451, 507)
6073 - fix confusing lconf 'host not found' error message (386)
6074 - fix lock order deadlock on OST (O/R i_sem before journal ops, 478)
6075 - fix race condition in mdc_blocking_ast() for inode access (526)
6076 - fix lov_unpackmd() unpacking wrong number of stripes (537)
6077 - fix lov_set_osc_active() marking wrong OSC inactive (440)
6078 - fix bad lstripe lov_unpackmd() assertion (fix layering too) (527)
6079 - fix multiple writes of stripe MD to MDS (358, maybe 519)
6080 - fix lstripe in several ways (kernel side) (527)
6081 - fix request leak in ldlm_cli_enqueue (262)
6082 - incorrect OSC was marked inactive after OST failure
6083 - call mds_fs_cleanup before unmounting filesystem (524)
6084 - fix races between taking ns_lock and ldlm_lock_change_resource
6085 - fix races updating LOV export open file list
6086 - fix lov_enqueue error path, avoid decref-ing bad lock handle
6087 - fix recovery NULL deref in ldlm_cli_cancel_unused
6088 - fix some DLM races by using new hash table for lock handles (419)
6089 - permit the client to specify desired inodes, at replay
6090 - duplicate requests when we queue them for replay reintegration
6091 - fix last_rcvd offset calculation
6092 - sync after each recovered transaction, so we always make progress
6093 - never, not always, ERESTART requests without transnos
6094 - store the lov_desc in the MDS, so we don't depend on getlovinfo to
6096 - skip replay if the MDS says that the client is already connected
6097 - don't check for a recovery-enabled export to match lctl's UUID
6098 - don't INC_USE_COUNT for phantom exports
6099 - don't crash when cleaning up phantom exports (567)
6100 - don't double-finish or set replay data for errored mdc_open requests
6101 - abort requests when they time out, so we don't get old replies
6102 - send/receive replies for AST messages again
6103 - if the client says that it doesn't have the lock, cancel it on the
6105 - if we timeout during I/O, don't try to cancel an in-use lock; instead
6106 mark it as destroyed, it will all work out when decref is called
6107 - fix module use counts (22, 581)
6109 - ASTs now expect a reply (server cancels lock on error reply)
6111 2002-12-02 Andreas Dilger <adilger@clusterfs.com>
6114 - fix many simultaneous client startup (392)
6115 - fix dentry->d_it clobbering
6116 - credentials weren't being shipped for readdir/getattr operations
6117 - remove invalid assertions triggered during some concurrent MD
6119 - proper Lustre versions added (336, 389)
6120 - fix memory leak for create error case (398)
6121 - fix LOV locking bug that would get cli/srv out of sync
6122 - fix echo client over LOV (409)
6123 - fix dbench 2, extN refcount problem (170, 258, 356, 418)
6124 - fix double-O_EXCL intent crash (424)
6125 - avoid sending multiple lock CANCELs (352)
6127 - MDS can do multi-client recovery (modulo bugs in new code)
6129 - many updates, edits, cleanups
6131 2002-11-18 Phil Schwan <phil@clusterfs.com>
6134 - fix null d_it dereference (346)
6135 - fix full OST/dbench hang (333)
6136 - fix permission problem with file removal (286)
6137 - fix removal of OSCs from LOV when they fail
6138 - fix NULL deref during bulk timeout (214)
6139 - fix problems related to multiple filesystems on one MDS (241)
6140 - fixed serious subtle metadata locking bugs
6141 - free locks on clients when inodes are removed due to memory
6143 - fix inode pointer in lock data (285)
6144 - partial support for multiple MDS on a single host (241)
6145 - data locks weren't cancelled at clear_inode time (290, 311)
6146 - intent locks could lead to unbounded lock growth (205)
6147 - added a maximum lock count, an LRU list, and a flusher
6148 - fix multiple rename (365)
6149 - properly abstracted the echo client
6150 - OSC locked 1 byte too many; fixed
6151 - rewrote brw callback code:
6152 - fixed recovery bugs related to LOVs (306)
6153 - fixed too-many-pages-in-one-write crash (191)
6154 - fixed (again) crash in sync_io_timeout (214)
6155 - probably fixed callback-related race (385)
6157 - Add capability to MDS protocol
6158 - LDLM cancellations and callbacks on different portals
6160 2002-10-28 Andreas Dilger <adilger@clusterfs.com>
6163 - limit client IOV size to PTL_MD_MAX_IOV (611336, 191)
6164 - defer open object destruction to close time (601981, 138)
6165 - open/close OST file handle in obdo (OBD_MD_FLHANDLE) (601981, 138)
6166 - move LDLM_ENQUEUE/CONVERT back to MDS portal (625069)
6167 - abstract ll_lookup2, fix ll_revalidate2 to use abstraction (256)
6168 - don't call obd_setattr in ll_file_release for destroyed objects
6169 * protocol change to lustre_msg: move |version| and add |flags|
6170 * protocol change to osc_punch: "start" in "o_size", "end" in "o_blocks"
6171 * lock replay: for LDLM_FL_REPLAY trust client to do right thing
6172 * added replay of create, unlink, link and rename operations during
6173 MDS failover; recovery should be much more robust now
6174 * remove failed OSCs from LOVs (only lov_create uses this so far)
6175 * the lustre-HOWTO was brought (more) up to date (582544)
6177 2002-10-23 Phil Schwan <phil@clusterfs.com>
6180 - in-use dentries weren't being reused properly (617851)
6181 - prevent multiple LDLM setup (599178)
6182 - fix LOV size calculations for truncate (617853)
6183 - fix client handling of MDS intent errors (POSIX)
6184 - fix permission bug in lovstripe.c test (624321)
6185 - fix MDS thread deadlock - move LDLM handler to DLM portal (625069)
6186 - truncate past end of file could corrupt data
6187 - proper cleanup after timeouts, crashes, etc (592524, 550815)
6188 - a race in recovery could return ETIMEDOUT to apps (623947)
6189 - building outside the source directory was fixed
6190 * the lustre-HOWTO was brought (more) up to date (582544)
6191 * major progress was made on recovery functionality
6193 2002-10-10 Phil Schwan <phil@clusterfs.com>
6196 - recovery deadlock fix
6197 - rm -rf causes LBUG fix (617817)
6198 - file open by multiple tasks fix (618962)
6199 - directory permissions bugs (602707 and 620007)
6200 - journal_stop fixed with locking (611313)
6201 - O_APPEND failures resolved (618273, perhaps 614459)
6202 - lconf PATH fix (619770)
6203 - IA64 build fix (621450)
6204 - RPC buffer sizes scale with amount of memory
6206 2002-10-01 Phil Schwan <phil@clusterfs.com>
6209 - locks would be cancelled without throwing away data pages,
6210 resulting in inconsistent data (605627)
6211 - inode attributes were not always being refreshed (605627, 612449)
6212 - lconf now continues to cleanup after lctl reports an error
6213 - MDS now enforces user permissions (602707)
6214 - lprocfs cleanup fixed, but not yet enabled (614157)
6215 - fixed infinite server hang, should a client not respond to an AST
6216 - avoid going into recovery if user calls readlink() with a buffer
6217 that's too small (613941)
6218 - AST RPCs no longer require replies (614867) -- this may be changed
6219 - don't crash server if client sends an IOV that's too big (611336)
6220 - fixed lock conversion deadlock (611892)
6221 - fixed the following of symlinks (614622)
6222 * recovery: the server can remove locks from a client that dies, other
6223 clients can make progress
6224 * more extN patch fixes
6225 * compile-time configurable ptlrpc buffer allocations
6227 - collaborative read cache document
6228 - Lustre Lite Performance CDR document-in-progress
6230 2002-09-20 Andreas Dilger <adilger@clusterfs.com>
6233 - fix typo in patch-2.4.18
6235 2002-09-20 Andreas Dilger <adilger@clusterfs.com>
6238 - clear ptlrpc request each time in handle_incoming_request()
6239 - unlink of files now destroys the object on the OST
6241 2002-09-19 Peter Braam <braam@clusterfs.com>
6243 * add hard link support
6244 * change obdfile creation method
6245 * kernel patch changed
6247 2002-09-19 Peter Braam <braam@clusterfs.com>
6250 - stack overflow bug in extN fixed
6252 2002-09-18 Andreas Dilger <adilger@clusterfs.com>
6254 * documentation updates
6255 - add man pages for config tools
6256 - update tests/README to describe testing with new config tools
6257 - finish metadata API descriptions
6258 * bug fixes and cleanups
6259 - statfs workaround for 16TB limit
6260 - LOV stripe allocation improved, can stripe on subset of OSTs
6261 - LOV file size/IO offset was wrong for files > 4GB in size
6262 - object EA data was being dropped, caused files to be unreadable
6263 - memory overflow with non-LOV OST caused memory corruption
6264 - fixed regression tests to work with new config tools, obdfilter
6265 - fixed bug when directory size became larger than 1 block
6266 - fixed bug (for single client case) when PWD was deleted
6267 - invalidate local directory pages when doing intent-based ops
6268 - avoid LDLM oops when lock callback contained bad data
6270 2002-09-09 Andreas Dilger <adilger@clusterfs.com>
6272 * documentation updates
6273 * bug fixes and cleanups
6274 - configuration tools
6277 - 64-bit compile warnings
6278 - 64-bit internal statfs data
6280 * test_brw on persistent OST devices
6282 * lprocfs (disabled)
6284 2002-09-04 Andreas Dilger <adilger@clusterfs.com>
6286 * documentation updates
6287 * bug fixes and cleanups
6288 * configuration tools
6290 2002-08-30 Peter J. Braam <braam@clusterfs.com>
6293 * many small fixes to 0_5_4
6294 * io/network handling
6295 * thinkos in MDS operations
6297 2002-08-24 Peter J. Braam <braam@clusterfs.com>
6300 * crucial basic fixes to 0.5.3
6301 * IOR, Iozone work over Elan
6304 2002-08-07 Phil Schwan <phil@clusterfs.com>
6305 * version 0_5_3, our first alpha
6306 * we use the new Portals iovs
6307 * documentation updates
6308 * bug fixes and cleanups
6309 * small changes in the DLM wire protocol
6311 2002-07-25 Peter J. Braam <braam@clusterfs.com>
6312 * version 0_5_1 with some initial stability,
6313 * locking on MD and file I/O.
6314 * documentation updates
6315 * several bug fixes since 0.5.0
6316 * small changes in wire protocol
6318 2002-07-18 Phil Schwan <phil@clusterfs.com>
6320 * delivered as Lustre Light Alpha
6321 * fixed a crash after handling invalid MDS requests
6322 * fixed directory pages for architectures with non-4k pages sizes
6324 2002-07-11 Andreas Dilger <adilger@clusterfs.com>
6325 * release version v0_4_4
6326 * Moves TCP acceptor to be on port 2432 (unused Coda port) instead
6328 * Fixes a number of interruption problems with OST operations.
6329 * Update documentation for portals header changes
6330 * Move all wire protocol structs/defines to lustre_idl.h
6331 * Fixes symlink length bug.
6332 * Add tcpdump to repository.
6334 2002-07-05 Andreas Dilger <adilger@clusterfs.com>
6335 * release version v0_4_3
6336 * Fixes statfs for inodes on extN.
6337 * Fixes bug in runtests which would delete /etc/hosts.
6338 * Use 64-bit object IDs wherever possible (not into VFS though)
6339 Remove ost_get_info, which is unused by lustre, and out of date.
6341 2002-07-03 Peter Braam <braam@clusterfs.com>
6342 * release version v0_4_2 Fixes a lookup error (type not passed)
6343 * move forward to head of Portals
6344 * move forward to latest Lustre kernel
6346 2002-06-25 Peter Braam <braam@clusterfs.com>
6347 * release version v0_4_1. Hopefully stable on single node use.