1 tbd Sun Microsystems, Inc.
5 2.6.9-67.0.22.EL (RHEL 4),
6 2.6.16.60-0.31 (SLES 10),
7 2.6.18-92.1.17.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.
35 Frequency : race on file read and write
37 Description: Lustre doesn't delete files
38 Details : Clients drop lock reference and release openhandle when they find
42 -------------------------------------------------------------------------------
43 2008-12-31 Sun Microsystems, Inc.
45 * Support for kernels:
47 2.6.9-67.0.22.EL (RHEL 4),
48 2.6.16.60-0.31 (SLES 10),
49 2.6.18-92.1.17.el5 (RHEL 5),
50 2.6.22.14 vanilla (kernel.org)
51 * Client support for unpatched kernels:
52 (see http://wiki.lustre.org/index.php?title=Patchless_Client)
53 2.6.16 - 2.6.22 vanilla (kernel.org)
54 * Client support for unpatched kernels:
55 we do not recommend using patchless RHEL4 clients with kernels
56 prior to 2.6.9-55EL (RHEL4U5).
57 * Recommended e2fsprogs version: 1.40.11-sun1
58 * Note that reiserfs quotas are disabled on SLES 10 in this kernel.
59 * RHEL 4 and RHEL 5/SLES 10 clients behaves differently on 'cd' to a
60 removed cwd "./" (refer to Bugzilla 14399).
61 * A new quota file format has been introduced in 1.6.5.
62 The format conversion from prior releases is handled transparently,
63 but releases older than 1.4.12/1.6.5 don't understand this new
64 format. The automatic format conversion can be avoided by running
65 the following command on the MDS:
66 'tunefs.lustre --param="mdt.quota_type=ug1" $MDTDEV'.
67 For more information, please refer to bugzilla 13904.
68 * A new quota file format was introduced in 1.6.6/1.8.0.
69 The format conversion from prior releases is handled transparently,
70 but releases older than 1.6.6/1.8.0 don't understand this new
71 format. The automatic format conversion can be avoided by running
72 the following commands on the MDS and OSS servers (for
73 pre 1.4.12-1.6.5 quota files):
74 'tunefs.lustre --param="mdt.quota_type=ug1" $MDTDEV',
75 'tunefs.lustre --param="ost.quota_type=ug1" $MDTDEV'
76 or (for 1.4.12/1.6.5 quota files)
77 'tunefs.lustre --param="mdt.quota_type=ug2" $MDTDEV',
78 'tunefs.lustre --param="ost.quota_type=ug2" $MDTDEV'
79 For more information, please refer to bugzilla 13904.
80 * Output of lfs quota has been made less detailed by default,
81 old (verbose) output can be obtained by using -v option.
82 * File join has been disabled in this release, refer to Bugzilla 16929.
83 * A new Lustre ADIO driver is available for MPICH2-1.0.7.
84 * NFS export disabled when stack size < 8192. Since the NFSv4 export of
85 Lustre filesystem with 4K stack may cause a stack overflow. For more
86 information, please refer to bugzilla 17630.
89 Frequency : on remount
91 Description: external journal device not working after the remount
92 Details : clear dev_rdonly flag for external journal devices in
98 Description: shutdown vs evict race
99 Details : client_disconnect_export vs connect request race.
100 if client will evicted at this time - we start invalidate
101 thread without referece to import and import can be freed
107 Description: shrink LOV EAs before replying
108 Details : correctly adjust LOV EA buffer for reply.
113 Description: don't skip ost target if they assigned to file
114 Details : Drop slow OSCs if we can, but not for requested start idx.
115 This means "if OSC is slow and it is not the requested
116 start OST, then it can be skipped, otherwise skip it only
117 if it is inactive/recovering/out-of-space.
119 Severity : enhancement
121 Description: Update to RHEL5 kernel-2.6.18-92.1.17.el5.
124 Frequency : rare, need acl's on inode.
126 Description: client can't handle ost additional correctly
127 Details : if ost was added after client connected to mds client can have
128 hit lnet_try_match_md ... to big messages to wide striped files.
129 in this case need teach client to handle config events about add
130 lov target and update client max ea size at that event.
132 Severity : enhancement
134 Description: Update to sles9 kernel-2.6.5-7.314.
136 Severity : enhancement
138 Description: Update to SLES10 SP2 kernel-2.6.16.60-0.31.
141 Frequency : Create a symlink file with a very long name
143 Description: ldlm_cancel_pack()) ASSERTION(max >= dlm->lock_count + count)
144 Details : If there is no extra space in the request for early cancels,
145 ldlm_req_handles_avail() returns 0 instead of a negative value.
150 Description: mds is deadlocked
151 Details : in rare cases, inode in catalog can have i_no less than have parent
152 i_no, this produce wrong order for locking during open, and
153 parallel unlink can be lock open. this need teach mds_open to grab
154 locks in resource id order, not at parent -> child order.
156 Severity : enhancement
158 Description: Add /proc entry for import status
159 Details : The mdc, osc, and mgc import directories now have
160 an import directory that contains useful import data for debugging
163 Severity : enhancement
165 Description: Re-disable certain /proc logging
166 Details : Enable and disable client's offset_stats, extents_stats and
167 extents_stats_per_process stats logging on the fly.
170 Frequency : Only on FC kernels 2.6.22+
172 Description: oops in statahead
173 Details : Do not drop reference count for the dentry from VFS when lookup,
174 VFS will do that by itself.
176 Severity : enhancement
178 Description: Generic /proc file permissions
179 Details : Set /Proc file permissions in a more generic way to enable non-
180 root users operate on some /proc files.
184 Description: Hitting mdc_commit_close() ASSERTION
185 Details : Properly handle request reference release in
186 ll_release_openhandle().
190 Frequency : only patchless client
191 Description: add workaround for race between add/remove dentry from hash
193 Severity : enhancement
195 Description: Allow OST glimpses to return PW locks
199 Description: LBUG when llog conf file is full
200 Details : When llog bitmap is full, ENOSPC should be returned for plain
205 Description: Prevent import from entering FULL state when server in recovery
209 Description: service mount cannot take device name with ":"
210 Details : Only when device name contains ":/" will mount treat it as
216 Description: replace ptlrpcd with the statahead thread to interpret the async
217 statahead RPC callback
221 Frequency : on recovery
222 Description: I/O failures after umount during fail back
223 Details : if client reconnected to restarted server we need join to recovery
224 instead of find server handler is changed and process self eviction
225 with cancel all locks.
227 Severity : enhancement
229 Description: Update to RHEL5 kernel-2.6.18-92.1.10.el5.
233 Description: Kernel BUG tries to release flock
234 Details : Lustre does not destroy flock lock before last reference goes
235 away. So always drop flock locks when client is evicted and
236 perform unlock regardless of successfulness of speaking to MDS.
238 Severity : enhancement
240 Description: Update to SLES10 SP2 kernel-2.6.16.60-0.27.
242 Severity : enhancement
244 Description: Upcall on Lustre log has been dumped
245 Details : Allow for a user mode script to be called once a Lustre log has
246 been dumped. It passes the filename of the dumped log to the
247 script, the location of the script can be specified via
248 /proc/sys/lnet/debug_log_upcall.
253 Description: avoid messages about idr_remove called for id that is not allocated
254 Details : Move assigment s_dev for clustered nfs to end of initialization,
255 for avoid problem with error handling.
260 Description: avoid Already found the key in hash [CONN_UNUSED_HASH] messages
261 Details : When connection is reused this not moved from CONN_UNUSED_HASH into
262 CONN_USED_HASH and this prodice warning when put connection again
268 Description: avoid ASSERTION(client_stat->nid_exp_ref_count == 0) failed
269 Details : release reference to stats when client disconnected, not
270 when export destroyed for avoid races when client destroyed
271 after main ost export.
275 Description: more cleanup in mds_lov
276 Details : add workaround for get valid ost count for avoid warnings about
277 drop too big messages, not init llog cat under semphore which
278 can be blocked on reconnect and break normal replay, fix access
281 Severity : enhancement
283 Description: File striping can now be set to use an arbitrary pool of OSTs.
285 Severity : enhancement
287 Description: Export bytes_read/bytes_write count on OSC/OST.
291 Description: Early reply size mismatch, MGC loses connection
292 Details : Apply the MGS_CONNECT_SUPPORTED mask at reconnect time so
293 the connect flags are properly negotiated.
297 Description: Properly propagate oinfo flags from lov to osc for statfs
298 Details : restore missing copy oi_flags to lov requests.
302 Description: exports in /proc are broken
303 Details : recreate /proc entries for clients when they reconnect.
305 Severity : enhancement
307 Description: Add man pages for llobdstat(8), llstat(8), plot-llstat(8),
308 : l_getgroups(8), lst(8), routerstat(8)
309 Details : included man pages for llobdstat(8), llstat(8),
310 : plot-llstat(8), l_getgroups(8), lst(8), routerstat(8)
312 Severity : enhancement
314 Description: Implement lustre ll_show_options method.
316 Severity : enhancement
318 Description: Update to SLES9 kernel-2.6.5-7.312.
320 Severity : enhancement
322 Description: Update to RHEL4 kernel-2.6.9-67.0.22.EL.
326 Description: exports in /proc are broken
327 Details : recreate /proc entries for clients when they reconnect.
331 Description: don't fail open with -ERANGE
332 Details : if client connected until mds will be know about real ost count
333 get LOV EA can be fail because mds not allocate enougth buffer
338 Description: Resolve device initialization race
339 Details : Prevent proc handler from accessing devices added to the
340 obd_devs array but yet be intialized.
342 Severity : enhancement
344 Description: configure's --enable-quota should check the
345 : kernel .config for CONFIG_QUOTA
346 Details : configure is terminated if --enable-quota is passed but
347 : no quota support is in kernel
349 Severity : enhancement
351 Description: Update to SLES10 SP2 kernel-2.6.16.60-0.23.
353 Severity : enhancement
355 Description: Update to RHEL5 kernel-2.6.18-92.1.6.el5.
359 Frequency : rare, on PPC clients
360 Description: don't swab ost objects in response about directory, because
362 Details : bug similar bug 14856, but in different function.
364 Severity : enhancement
366 Description: lfs quota tool enhancement
367 Details : added units specifiers support for setquota, default to
368 current uid/gid for quota report, short quota stats by
369 default, nonpositional parameters for setquota, added
370 llapi_quotactl manual page.
372 Severity : enhancement
374 Description: *optional* service tags registration
375 Details : if the "service tags" package is installed on a Lustre node
376 When the filesystem is mounted, a local-node service tag will
377 be created. See http://inventory.sun.com/ for more information
378 about the Service Tags asset management system.
382 Description: Client runs out of low memory
383 Details : Consider only lowmem when counting initial number of llap pages
386 Description: Kernel BUG tries to release flock
387 Details : Lustre does not destroy flock lock before last reference goes
388 away. So always drop flock locks when client is evicted and
389 perform unlock regardless of successfulness of speaking to MDS.
392 Frequency : occasional
394 Description: add refcount for osc callbacks, so avoid panic on shutdown
396 Severity : enhancement
398 Description: Update to RHEL4 kernel-2.6.9-67.0.20.
401 Frequency : testing only
403 Description: sanity test 65a fails if stripecount of -1 is set
404 Details : handle -1 striping on filesystem in ll_dirstripe_verify
407 Frequency : only in unusual configurations
409 Description: Kernel panic with find ost index.
410 Details : lov_obd have panic if some OST's have sparse indexes.
412 Severity : enhancement
414 Description: Update to RHEL5 kernel-2.6.18-53.1.21.el5.
417 Frequency : rarely, if filesystem is mounted with -o flock
419 Description: do not process already freed flock
420 Details : flock can possibly be freed by another thread before it reaches
421 to ldlm_flock_completion_ast.
424 Frequency : rarely, if filesystem is mounted with -o flock
426 Description: LBUG during stress test
427 Details : Need properly lock accesses the flock deadlock detection list.
430 Frequency : rarely, if binaries are being run from Lustre
432 Description: oops in page fault handler
433 Details : kernel page fault handler can return two special 'pages' in
434 error case, don't try dereference NOPAGE_SIGBUS and NOPAGE_OMM.
437 Frequency : rarely, during shutdown
439 Description: timeout with invalidate import.
440 Details : ptlrpcd_check call obd_zombie_impexp_cull and wait request which
441 should be handled by ptlrpcd. This produce long age waiting and
442 -ETIMEOUT ptlrpc_invalidate_import and as result LASSERT.
448 Description: ASSERTION(CheckWriteback(page,cmd)) failed
449 Details : badly clear PG_Writeback bit in ll_ap_completion can produce false
453 Frequency : only with broken builds/installations
455 Description: no LBUG if lquota.ko and fsfilt_ldiskfs.ko are different versions
456 Details : just return an error to a user, put a console error message
458 Severity : enhancement
460 Description: Update to RHEL5 kernel-2.6.18-53.1.19.el5.
462 Severity : enhancement
464 Description: Update to RHEL4 kernel-2.6.9-67.0.15.
466 Severity : enhancement
468 Description: enable MGS and MDT services start separately
469 Details : add a 'nomgs' option in mount.lustre to enable start a MDT with
470 a co-located MGS without starting the MGS, which is a complement
471 to 'nosvc' mount option.
474 Frequency : always, on big-endian systems
476 Description: cleanup in ptlrpc code, related to PPC platform
477 Details : store magic in native order avoid panic's in recovery on PPC
478 node and forbid from this error in future. Also fix posibily
479 of twice swab data. Fix get lov striping to userpace.
482 Frequency : rarely, if replay get lost on server
484 Description: server incorrectly drop resent replays lead to recovery failure.
485 Details : do not drop replay according to msg flags, instead we check the
486 per-export recovery request queue for duplication of transno.
489 Frequency : after recovery
491 Description: precreate to many object's after del orphan.
492 Details : del orphan st in oscc last_id == next_id and this triger growing
493 count of precreated objects. Set flag LOW to skip increase count
494 of precreated objects.
497 Frequency : rare, on clear nid stats
499 Description: ASSERTION(client_stat->nid_exp_ref_count == 0)
500 Details : when clean nid stats sometimes try destroy live entry,
501 and this produce panic in free.
504 Frequency : occasionally since 1.6.4
506 Description: Stack overflow during MDS log replay
507 Details : ease stack pressure by using a thread dealing llog_process.
510 Frequency : very rare
512 Description: MDT cannot be unmounted, reporting "Mount still busy"
513 Details : Mountpoint references were being leaked during open reply
514 reconstruction after an MDS restart. Drop mountpoint reference
515 in reconstruct_open() and free dentry reference also.
520 Description: wait until IO finished before start new when do lock cancel.
521 Details : VM protocol want old IO finished before start new, in this case
522 need wait until PG_writeback is cleared until check dirty flag
523 and call writepages in lock cancel callback.
528 Description: mds_mfd_close() ASSERTION(rc == 0)
529 Details : In mds_mfd_close(), we need protect inode's writecount change
530 within its orphan write semaphore to prevent possible races.
533 Frequency : rare, on shutdown ost
535 Description: don't hit live lock with umount ost.
536 Details : shrink_dcache_parent can be in long loop with destroy dentries,
537 use shrink_dcache_sb instead.
540 Frequency : only when echo_client is used
542 Description: don't panic with use echo_client
543 Details : echo client pass NULL as client nid pointer and this produce NULL
547 Frequency : Always on 32-bit PowerPC systems
549 Description: fix build on PPC32
550 Details : compile code with -m64 flag produce wrong object file for PPC32.
555 Description: MDS LBUG: ASSERTION(!IS_ERR(dchild))
556 Details : In reconstruct_* functions, LASSERTs on both the data supplied
557 by a client, and the data on disk are dangerous and incorrect.
558 Change them with client eviction.
560 Severity : enhancement
562 Description: skiplist implementation simplification
563 Details : skiplists are used to group compatible locks on granted list
564 that was implemented as tracking first and last lock of each lock
565 group the patch changes that to using doubly linked lists
569 Description: delete compatibility for 32bit qdata
570 Details : as planned, when lustre is beyond b1_8, lquota won't support 32bit
571 qunit. That means servers of b1_4 and servers of b1_8 can't be
572 used together if users want to use quota.
575 Frequency : only with administrator action
577 Description: mount failure if config log has invalid conf_param setting
578 Details : If administrator specified an incorrect configuration parameter
579 with "lctl conf_param" this would cause an error during future
580 client mounts. Instead, ignore the bad configuration parameter.
583 Frequency : blocks per group < blocksize*8 and uninit_groups is enabled
585 Description: ldiskfs error: XXX blocks in bitmap, YYY in gd
586 Details : If blocks per group is less than blocksize*8, set rest of the
590 Frequency : Application do stride read on lustre
592 Description: The read performance will drop a lot if the application does
594 Details : Because the stride_start_offset are missing in stride read-ahead,
595 it will cause clients read a lot of unused pages in read-ahead,
596 then the read-performance drops.
600 Description: more ldlm soft lockups
601 Details : In ldlm_resource_add_lock(), call to ldlm_resource_dump()
602 starve other threads from the resource lock for a long time in
603 case of long waiting queue, so change the debug level from
604 D_OTHER to the less frequently used D_INFO.
606 Severity : enhancement
608 Description: add -gid, -group, -uid, -user options to lfs find
610 Severity : enhancement
612 Description: ll_recover_lost_found_objs - recover objects in lost+found
613 Details : OST corruption and subsequent e2fsck can leave objects in the
614 lost+found directory. Using the "ll_recover_lost_found_objs"
615 tool, these objects can be retrieved and data can be salvaged
616 by using the object ID saved in the fid EA on each object.
621 Description: this bug _only_ happens when inode quota limitation is very low
622 (less than 12), so that inode quota unit is 1 at initialization.
623 Details : if remaining quota equates 1, it is a sign to demonstate that quota
624 is effective now. So least quota qunit should be 2.
628 Description: Hung threads in invalidate_inode_pages2_range
629 Details : The direct IO path doesn't call check_rpcs to submit a new RPC once
630 one is completed. As a result, some RPCs are stuck in the queue
635 Description: Procfs and llog threads access destoryed import sometimes.
636 Details : Sync the import destoryed process with procfs and llog threads by
637 the import refcount and semaphore.
641 Description: mds fails to respond, threads stuck in ldlm_completion_ast
642 Details : Sort source/child resource pair after updating child resource.
647 Description: kernel BUG at ldiskfs2_ext_new_extent_cb
648 Details : If insertion of an extent fails, then discard the inode
649 preallocation and free data blocks else it can lead to duplicate
654 Description: don't always update ctime in ext3_xattr_set_handle()
655 Details : Current xattr code updates inode ctime in ext3_xattr_set_handle()
656 In some cases the ctime should not be updated, for example for
657 2.0->1.8 compatibility it is necessary to delete an xattr and it
658 should not update the ctime.
662 Description: add quota statistics
663 Details : 1. sort out quota proc entries and proc code.
664 2. add quota statistics
669 Description: quotas are not honored with O_DIRECT
670 Details : all writes with the flag O_DIRECT will use grants which leads to
671 this problem. Now using OBD_BRW_SYNC to guard this.
675 Bugzilla : 15713/16362
676 Description: Assertion in iopen_connect_dentry in 1.6.3
677 Details : looking up an inode via iopen with the wrong generation number can
678 populate the dcache with a disconneced dentry while the inode
679 number is in the process of being reallocated. This causes an
680 assertion failure in iopen since the inode's dentry list contains
681 both a connected and disconnected dentry.
685 Description: assertion failure in ldlm_handle2lock()
686 Details : fix a race between class_handle_unhash() and class_handle2object()
687 introduced in lustre 1.6.5 by bug 13622.
689 Severity : enhancement
691 Description: superblock lock contention with many SMP cores on one client
692 Details : several client filesystem locks were highly contended on SMP
693 NUMA systems with 8 or more cores. Per-CPU datastructures
694 and more efficient locking implemented to reduce contention.
699 Description: Kernel BUG: sd_iostats_bump: unexpected disk index
700 Details : remove the limit of 256 scsi disks in the sd_iostat patch
705 Description: oops in sd_iostats_seq_show()
706 Details : unloading/reloading the scsi low level driver triggers a kernel
707 bug when trying to access the sd iostat file.
712 Description: Kernel panics during QLogic driver reload
713 Details : REQ_BLOCK_PC requests are not handled properly in the sd iostat
714 patch, causing memory corruption.
719 Description: journal_dev option does not work in b1_6
720 Details : pass mount option during pre-mount.
722 Severity : enhancement
724 Description: Add a FIEMAP(FIle Extent MAP) ioctl for ldiskfs
725 Details : FIEMAP ioctl will allow an application to efficiently fetch the
726 extent information of a file. It can be used to map logical blocks
727 in a file to physical blocks in the block device.
730 Frequency : only with adaptive timeout enabled
732 Description: DEBUG_REQ() bad paging request
733 Details : ptlrpc_at_recv_early_reply() should not modify req->rq_repmsg
734 because it can be accessed by reply_in_callback() without the
738 Frequency : only on Cray X2
740 Description: X2 build failures
741 Details : fix build failures on Cray X2.
745 Description: xid & resent requests
746 Details : Initialize RPC XID from clock at startup (randomly if clock is
751 Description: quota recovery deadlock during mds failover
752 Details : This patch includes att18982, att18236, att18237 in bz14840.
754 1. fix osts hang when mds does failover with quotaon
755 2. prevent watchdog storm when osts threads wait for the
760 Description: kernel panic on racer
761 Details : Do not access dchild->d_inode when IS_ERR(dchild) is true.
763 Severity : enhancement
765 Description: Add lustre_start utility to start or stop multiple Lustre servers
770 Description: Lustre GPF in {:ptlrpc:ptlrpc_server_free_request+373}
771 Details : In case of memory pressure, list_del() can be called twice on
772 req->rq_history_list, causing a kernel oops.
776 Description: (ptllnd_peer.c:557:kptllnd_peer_check_sends()) ASSERTION(!in_interrupt()) failed
777 Details : fix stack overflow in the distributed lock manager by defering
778 export eviction after a failed ast to the elt thread instead of
779 handling it in the dlm interpret routine.
781 Severity : enhancement
783 Description: More exported tunables for mballoc
784 Details : Add support for tunable preallocation window and new tunables for
789 Description: Detect corruption of block bitmap and checking for preallocations
790 Details : Checks validity of on-disk block bitmap. Also it does better
791 checking of number of applied preallocations. When corruption is
792 found, it turns filesystem readonly to prevent further corruptions.
796 Frequency : only for big-endian servers
797 Description: Check for big-endian system while mounting fs with extents feature
798 Details : Mounting a filesystem with extents feature will fail on big-endian
799 systems since ext3-based ldiskfs is not supported on big-endian
800 systems. Can be over-ridden with "bigendian_extents" mount option.
804 Description: Excessive recovery window
805 Details : With AT enabled, the recovery window can be excessively long (6000+
806 seconds). To address this problem, we no longer use
807 OBD_RECOVERY_FACTOR when extending the recovery window (the connect
808 timeout no longer depends on the service time, it is set to
809 INITIAL_CONNECT_TIMEOUT now) and clients report the old service
810 time via pb_service_time.
814 Description: Watchdog triggered on MDS failover
815 Details : enable OBD_CONNECT_MDT flag when connecting from the MDS so that
816 the OSTs know that the MDS "UUID" can be reused for the same export
817 from a different NID, so we do not need to wait for the export to
820 -------------------------------------------------------------------------------
822 2008-05-26 Sun Microsystems, Inc.
824 * Support for kernels:
825 2.6.5-7.311 (SLES 9),
826 2.6.9-67.0.7.EL (RHEL 4),
827 2.6.16.54-0.2.5 (SLES 10),
828 2.6.18-53.1.14.el5 (RHEL 5),
829 2.6.22.14 vanilla (kernel.org)
830 * Client support for unpatched kernels:
831 (see http://wiki.lustre.org/index.php?title=Patchless_Client)
832 2.6.16 - 2.6.22 vanilla (kernel.org)
833 * Due to problems with nested symlinks and FMODE_EXEC (bug 12652),
834 we do not recommend using patchless RHEL4 clients with kernels
835 prior to 2.6.9-55EL (RHEL4U5).
836 * Recommended e2fsprogs version: 1.40.7-sun1
837 * Note that reiserfs quotas are disabled on SLES 10 in this kernel.
838 * RHEL 4 and RHEL 5/SLES 10 clients behaves differently on 'cd' to a
839 removed cwd "./" (refer to Bugzilla 14399).
840 * A new quota file format has been introduced in 1.6.5.
841 The format conversion from prior releases is handled transparently,
842 but releases older than 1.4.12/1.6.5 will not understand this new
843 format. The automatic format conversion can be avoided by running
844 the following command on the MDS before upgrading:
845 'tunefs.lustre --param="mdt.quota_type=ug1" $MDTDEV'.
846 For more information, please refer to bugzilla 13904.
850 Description: quota performance fix
851 Details : quota data is written in journalled mode instead of ordered to
856 Description: lfs support for human-readable quota grace time strings
857 Details : lfs setquota -t and lfs quota -t represent quota grace times
858 in "XXwXXdXXhXXmXXs" format instead of large values in seconds
861 Frequency : always with o2ib 1.3 and sles10
863 Description: fix build with SLES10 and o2ib v3.
864 Details : sles10 uses diffrent name for Module.symver file but configure
865 assume this file has same name on RHEL/SLES/vanila kernels.
868 Frequency : very rare, if additional xattrs are used on kernels >= 2.6.12
870 Description: MDS may lose file striping (and hence file data) in some cases
871 Details : If there are additional extended attributes stored on the MDS,
872 in particular ACLs, SELinux, or user attributes (if user_xattr
873 is specified for the client mount options) then there is a risk
874 of attribute loss. Additionally, the Lustre file striping
875 needs to be larger than default (e.g. striped over all OSTs),
876 and an additional attribute must be stored initially in the
877 inode and then increase in size enough to be moved to the
878 external attribute block (e.g. ACL growing in size) for file
881 Severity : enhancement
883 Description: add message levels for liblustreapi
888 Description: MDT cannot be unmounted, reporting "Mount still busy"
889 Details : Mountpoint references were being leaked during open reply
890 reconstruction after an MDS restart. Drop mountpoint reference
891 in reconstruct_open() and free dentry reference also.
896 Description: fix for occasional failure case of -ENOSPC in recovery-small tests
897 Details : Move the 'good_osts' check before the 'total_bavail' check. This
898 will result in an -EAGAIN and in the exit call path we call
899 alloc_rr() which will with increasing aggressiveness attempt to
900 aquire precreated objects on the minimum number of required OSCs.
904 Description: Use old size assignment to avoid deadlock
905 Details : This reverts the changes in bugs 2369 and bug 14138 that introduced
906 the scheduling while holding a spinlock. We do not need locking
907 for size in ll_update_inode() because size is only updated from
908 the MDS for directories or files without objects, so there is no
909 other place to do the update, and concurrent access to such inodes
910 are protected by the inode lock.
914 Description: Use __u64 instead of int for valid bits
918 Description: resolve "_IOWR redefined" build error on SLES10
922 Description: dump the memory debugging after all modules are unloaded to
923 suppress false negative in conf_sanity test 39
927 Description: the recovery timer never expires
928 Details : for new client connect request, the recovery timer should not be
929 reset, otherwise recovery timer will never expired, if the old
930 client never come. Only old client connect and first connection
931 req should trigger recovery timer reset.
935 Description: the min numbers of lproc stats are wrong
936 Details : adding a new constant LC_MIN_INIT and use it for initialization
940 Frequency : always with interactive lfs
942 Description: Reinitialize optind to 0 so that interactive lfs works in all cases
945 Frequency : with multiple concurrent readdir processes in same directory
946 Bugzilla : 15406, 15169, 15175
947 Description: misc fixes for directory readahead.
948 Details : prevent previous statahead async RPC callback from processing the
949 current "statahead_info", race condition between async RPC callback
950 add dentry into dentry hash table and "ls" thread revalidate such
951 dentry, statahead his/miss control for hidden items, and so on.
953 Severity : enhancement
955 Description: build kernel-ib packages for OFED 1.3 in our release cycle
959 Description: incore types cleaning in quota code (with respect to 64-bit limits)
960 Details : several u32 variables declarations are replaced with u64 declarations
965 Description: fix SLES kernel versioning
966 Details : the kernel version for our SLES 10 kernel did not include a "-"
967 before the "smp" at the end. while this was not a problem in
968 general, it did mean that software trying to use the kernel
969 version to try to detect a vendor specific kernel would fail.
970 this was most evident by the OFED build scripts.
975 Description: Don't update lov_desc members until making sure they are valid
976 Details : When updating lov_desc members via proc fs, need fix their
977 validities before doing the real update.
980 Frequency : very rare
982 Description: don't put request into delay list while invalidate in flight.
983 Details : ptlrpc_delay_request sometimes put in delay list while invalidate
984 import in flight. this produce timeout for invalidate and sometimes
985 can cause stale data.
987 Severity : enhancement
989 Description: Update kernel to SLES9 2.6.5-7.311.
991 Severity : enhancement
993 Description: Update kernel to RHEL4 2.6.9-67.0.7.
998 Frequency : on PPC only
999 Description: not convert ost objects for directory because it's not exist.
1000 Details : ll_dir_getstripe assume dirrectory has ost objects but this wrong.
1002 Severity : enhancement
1004 Description: Fix warnings with compile liblustre at sles10/rhel5 which have
1005 __u64 as usingied long long type.
1008 Frequency : rare, on shutdown
1010 Description: race process ast vs remove callback
1011 Details : removing callback before disconnect import open race with
1012 processing callback.
1014 Severity : enhancement
1016 Description: Update kernel to SLES9 2.6.5-7.311.
1018 Severity : enhancement
1020 Description: Files open for execute are not marked busy on SLES10
1021 Details : Add FMODE_EXEC to SLES10 SP1 server kernel series.
1023 Severity : enhancement
1025 Description: Add server support for vanilla-2.6.22.14.
1028 Frequency : occasional
1030 Description: Avoid lov_create() getting stuck in obd_statfs_rqset()
1031 Details : If an OST is down the MDS will hang indefinitely in
1032 obd_statfs_rqset() waiting for the statfs data. While for
1033 MDS QOS usage of statfs, it should not stuck in waiting.
1035 Severity : enhancement
1037 Description: Disable adaptive timeouts by default
1040 Frequency : on network error
1042 Description: panic with double free request if network error
1043 Details : mdc_finish_enqueue is finish request if any network error occuring,
1044 but it's true only for synchronus enqueue, for async enqueue
1045 (via ptlrpcd) this incorrect and ptlrpcd want finish request
1049 Frequency : rare, on recovery
1051 Description: read procfs can produce deadlock in some situation
1052 Details : Holding lprocfs lock with send rpc can produce block for destroy
1053 obd objects and this also block reconnect with -EALREADY.
1054 This isn't fix all lprocfs bugs - but make it rare.
1056 Severity : enhancement
1058 Description: Update kernel to RHEL5 2.6.18-53.1.14.el5.
1061 Frequency : frequent on X2 node
1063 Description: mdc_set_open_replay_data LBUG
1064 Details : Set replay data for requests that are eligible for replay.
1069 Description: lustre_mgs: operation 101 on unconnected MGS
1070 Details : When MGC is disconnected from MGS long enough, MGS will evict the
1071 MGC, and late on MGC cannot successfully connect to MGS and a lot
1072 of the error messages complaining that MGS is not connected.
1075 Frequency : rare, depends on device drivers and load
1077 Description: MDS or OSS nodes crash due to stack overflow
1078 Details : Code changes in 1.6.4 increased the stack usage of some functions.
1079 In some cases, in conjunction with device drivers that use a lot
1080 of stack, the MDS (or possibly OSS) service threads could overflow
1081 the stack. One change which was identified to consume additional
1082 stack has been reworked to avoid the extra stack usage.
1084 Severity : enhancement
1086 Description: Update to RHEL5 latest kernel-2.6.18-53.1.13.el5.
1088 Severity : enhancement
1090 Description: Update to SLES10 SP1 latest kernel-2.6.16.54-0.2.5.
1092 Severity : enhancement
1094 Description: Update to RHEL5 latest kernel-2.6.18-53.1.6.el5.
1096 Serverity : enhancement
1098 Description: Update RHEL4 kernel to 2.6.9-67.0.4.
1101 Frequency : rare on shutdown OST
1103 Description: Don't allow skipping OSTs if index has been specified.
1104 Details : Don't allow skipping OSTs if index has been specified, make locking
1105 in internal create lots better.
1110 Description: ASSERTION(!PageDirty(page)) failed
1111 Details : Wrong check could lead to an assertion failure under specific
1117 Description: LBUG in ptlrpc_check_set() bad phase ebc0de00
1118 Details : access to bitfield in structure is always rounded to long
1119 and this produce problem with not atomic change any bit.
1124 Description: Lustre make rpms failed.
1125 Details : Remove ldiskfs spec file to avoids rpmbuild be confused when
1126 builds Lustre rpms from tarball.
1128 Severity : enhancement
1130 Description: Update to SLES9 SP4 kernel-2.6.5-7.308.
1133 Frequency : rare on shutdown OST
1135 Description: If llog cancel was not send before clean_exports phase, this can
1136 produce deadlock in llog code.
1137 Details : If llog thread has last reference to obd and call class_import_put
1138 this produce deadlock because llog_cleanup_commit_master wait when
1139 last llog_commit_thread exited, but this never success because was
1140 called from llog_commit_thread.
1143 Frequency : only if OST index is skipped
1145 Description: NULL lov_tgts causing MDS oops
1146 Details : more safe checks for NULL lov_tgts for avoid oops.
1148 Severity : enhancement
1150 Description: Update to RHEL4 latest kernel-2.6.9-67.0.1.EL.
1152 Severity : enhancement
1154 Description: Update to RHEL5 latest kernel-2.6.18-53.1.4.el5.
1159 Description: make mgs_setparam() handle fsname containing dash
1160 Details : fsname containing a dash does not work with lctl conf_param
1162 Severity : enhancement
1164 Description: Update to RHEL4 Update-6 kernel-2.6.9-67.EL.
1167 Frequency : rare, in recovery and (or) destroy lovobjid file.
1169 Description: rewrite lov objid code.
1170 Details : Cleanup for lov objid code, remove scability problems and wrong
1171 locking. Fix sending last_id into ost.
1173 Severity : enhancement
1175 Description: Update to SLES10 SP1 latest kernel-2.6.16.54-0.2.3.
1177 Severity : enhancement
1179 Description: Update to RHEL5 Update-1 kernel 2.6.18-53.el5.
1180 Details : Use d_move_locked instead of __d_move.
1183 Frequency : rare, at shutdown
1185 Description: access already free / zero obd_namespace.
1186 Details : if client_disconnect_export was called without force flag set,
1187 and exist connect request in flight, this can produce access to
1188 NULL pointer (or already free pointer) when connect_interpret
1189 store ocd flags in obd_namespace.
1192 Frequency : only at startup
1194 Description: not alloc memory with spinlock held.
1195 Details : allocation memory with GFP_KERNEL can produce sleep deadlock,
1196 if any spinlock held.
1201 Description: lfs find does not continue on file error
1202 Details : Continue other files processing when a file/dir is absent.
1206 Description: Inconsistent usage of lustre_pack_reply()
1207 Details : Standardize the usage of lustre_pack_reply() such that it
1208 always generate a CERROR on failure.
1211 Frequency : very rare
1213 Description: Fix replay if there is an un-replied request and open
1214 Details : In some cases, older replay request will revert the
1215 mcd->mcd_last_xid on MDS which is used to record the client's
1216 latest sent request.
1218 Severity : enhancement
1220 Description: Update to RHEL5 kernel 2.6.18-8.1.15.el5.
1222 Severity : enhancement
1224 Description: Update to SLES10 SP1 kernel 2.6.16.53-0.16
1226 Severity : enhancement
1228 Description: Update to SLES9 kernel-2.6.5-7.287.3.
1230 Severity : enhancement
1232 Description: Update to RHEL4 kernel-2.6.9-55.0.12.EL.
1234 Severity : enhancement
1236 Description: Build SLES10 patchless client fails
1237 Details : The configure was broken by run ./configure with
1238 --with-linux-obj=.... argument for patchless client. When the
1239 configure use --with-linux-obj, the LINUXINCLUDE= -Iinclude
1240 can't search header adequately. Use absolute path such as
1241 -I($LINUX)/include instead.
1243 Severity : enhancement
1245 Description: Lustre Page Accounting
1246 Details : New macros for page alloc and free which enable accounting
1247 of page allocation of Lustre. Use percpu counters to store memory
1248 and page statistics.
1251 Frequency : only if debugging is disabled
1253 Description: LASSERT_{REQ,REP}SWAB macros are buggy
1254 Details : If SWAB_PARANOIA is disabled, the LASSERT_REQSWAB and
1255 LASSERT_REPSWAB macros become no-ops, which is incorrect. Drop
1256 these macros and replace them with their definitions instead.
1261 Description: interrupt oig_wait produce painc on resend.
1262 Details : brw_redo_request can be used for resend requests from ptlrpcd and
1263 private set, and this produce situation when rq_ptlrpcd_data not
1264 copyed to new allocated request and triggered LBUG on assert
1265 req->rq_ptlrpcd_data != NULL. But this member used only for wakeup
1266 ptlrpcd set if request is changed and can be safety changed to use
1269 Severity : enhancement
1271 Description: organize the server-side client stats on per-nid basis
1272 Details : Change the structure of stats under obdfilter and mds to
1280 The "uuid"s file would list the uuids of _active_ exports.
1281 And the clear entry is to clear all stats and stale nids.
1286 Description: Processes looping in ll_readdir() on Lustre clients finally causing
1287 a full node pseudo-hang
1288 Details : Concurrent access to the same directory from multiple clients with
1289 intensive file creation/removal can cause a client node to spin in
1290 ll_readdir(). i_version must be increased every time the lock is
1291 cancelled to ensure a revalidate is done.
1296 Description: touch file failed when fs is not full
1297 Details : OST in recovery should not be discarded by MDS in alloc_qos(),
1298 otherwise we can get ENOSP while fs is not full.
1301 Frequency : only for Cray XT3
1302 Bugzilla : 12829/13455
1303 Description: Changing primary group doesn't change the group lustre assigns to
1305 Details : When CRAY_XT3 is defined, the fsgid supplied by the client is
1306 overridden with the primary group provided by the group upcall,
1307 whereas the supplied fsgid can be trusted if it is in the list of
1308 supplementary groups returned by the group upcall.
1310 Severity : enhancement
1312 Description: Root Squash Functionality
1313 Details : Implementation of NFS-like root squash capability. Specifically,
1314 don't allow someone with root access on a client node to be able
1315 to manipulate files owned by root on a server node.
1317 Severity : enhancement
1319 Description: Slow trucate/writes to huge files at high offsets.
1320 Details : Directly associate cached pages to lock that protect those pages,
1321 this allows us to quickly find what pages to write and remove
1322 once lock callback is received.
1327 Description: Too many locks accumulating on client during NFS usage
1328 Details : mds_open improperly used accmode to find out access mode to a
1329 file. Also mdc_intent_lock logic to find out if we already have
1330 lock similar to just received was flawed since introduction of
1331 skiplists - locks are now added to the front of the granted
1336 Description: Hit ASSERTION(obd->obd_stopping == 1) failed in some setup failed
1338 Details : In obd setup failure handler, obd_stopping will not necessarily to
1339 be 1, and obd_set_up should also be checked to make sure whether
1340 obd is completely setup.
1342 Severity : enhancement
1344 Description: Allow masking D_WARNING, D_ERROR messages from console
1345 Details : Console messages can now be disabled via lnet.printk.
1350 Description: User code with malformed file open parameter crashes client node
1351 Details : Before packing join_file req, all the related reference should be
1352 checked carefully in case some malformed flags cause fake join_file
1358 Description: shrink/enlarge qunit size when needed; fix the problem of coarse
1359 grain of quota doing harm to quota's accuracy
1360 Details : qunit size will be changed when quota limitation is too low/high;
1361 record the pending quota write in order to get more accureate
1362 quota; delete the patch for bug12588, which is unnecessary when
1363 this patch is landed. This bug also contains fixes for bug 14526,
1364 14299, 14601 and 13794.
1368 Description: LDLM_ENQUEUE races with LDLM_CP_CALLBACK
1369 Details : ldlm_completion_ast() assumes that a lock is granted when the req
1370 mode is equal to the granted mode. However, it should also check
1371 that LDLM_FL_CP_REQD is not set.
1375 Description: Heavy nfs access might result in deadlocks
1376 Details : After ELC code landed, it is now improper to enqueue any mds
1377 locks under och_sem, because enqueue might want to decide to
1378 cancel open locks for same inode we are holding och_sem for.
1382 Description: 35% write performance drop with ldiskfs2 when quotas are on
1383 Details : Enable ext3 journalled quota by default to improve performance
1384 when quotas are turned on.
1388 Description: Client eviction while running blogbench
1389 Details : A lot of unlink operations with concurrent I/O can lead to a
1390 deadlock causing evictions. To address the problem, the number of
1391 oustanding OST_DESTROY requests is now throttled to
1392 max_rpcs_in_flight per OSC and LDLM_FL_DISCARD_DATA blocking
1393 callbacks are processed in priority.
1396 Frequency : RHEL4 only
1398 Description: mkfs is very slow on IA64/RHEL4
1399 Details : A performance regression has been discovered in the MPT Fusion
1400 driver between versions 3.02.73rh and 3.02.99.00rh. As a
1401 consequence, we have downgraded the MPT Fusion driver in the RHEL4
1402 kernel from 3.02.99.00 to 3.02.73 until this problem is fixed.
1405 Frequency : PPC/PPC64 only
1407 Description: conflicts between asm-ppc64/types.h and lustre_types.h
1408 Details : fix duplicated definitions between asm-ppc64/types.h and
1409 lustre_types.h on PPC.
1412 Frequency : PPC/PPC64 only
1414 Description: asm-ppc/segment.h does not exist
1415 Details : fix compile issue on PPC.
1419 Description: data checksumming impacts single node performance
1420 Details : add support for several checksum algorithms. Currently, CRC32 and
1421 Adler-32 are supported. The checksum type can be changed on the fly
1422 through /proc/fs/lustre/osc/*/checksum_type.
1426 Description: use adler32 for page checksums
1427 Details : when available, use the Adler-32 algorithm instead of CRC32 for
1432 Description: better handle error messages in extents code
1434 Severity : enhancement
1436 Description: SNMP support enhancement
1437 Details : Adding total number of sampled request for an MDS node in snmp
1440 Severity : enhancement
1442 Description: Optimize ldlm waiting list processing for PR extent locks
1443 Details : When processing waiting list for read extent lock and meeting read
1444 lock that is same or wider to it that is not contended, skip
1445 processing rest of the list and immediatelly return current
1446 status of conflictness, since we are guaranteed there are no
1447 conflicting locks in the rest of the list.
1451 Description: Time out and refuse to reconnect
1452 Details : When the failover node is the primary node, it is possible
1453 to have two identical connections in imp_conn_list. We must
1454 compare not conn's pointers but NIDs, otherwise we
1455 can defeat connection throttling.
1459 Description: Client not clear own cache if answer to reconnect is lost.
1460 Details : Client gets evicted from server. Now client also thinks it is
1461 disconnected (or gets ENOTCONN on its operation) and decides to
1462 reconnect. Server receives reconnect message, but cannot find
1463 export. New export is created that is fully valid (new cookie!),
1464 but reply is lost and not reported to client. Client reconnects
1465 again and gets back a just-created connection, but it is not new
1466 so client thinks it was not evicted and does not do recovery.
1470 Description: Detect stride IO mode in read-ahead
1471 Details : When a client does stride read, read-ahead should detect that and
1472 read-ahead pages according to the detected stride pattern.
1476 Description: build for x2 fails
1477 Details : fix compile issue on Cray systems.
1479 Severity : enhancement
1481 Description: implement readv/writev APIs(aio_read/aio_writes in newer kernels)
1482 Details : This greatly improves speed of NFS writes on 2.6 kernels.
1485 Frequency : only on PPC/SLES10
1487 Description: "BITS_PER_LONG is not 32 or 64" in linux/idr.h
1488 Details : On SLES10/PPC, fs.h includes idr.h which requires BITS_PER_LONG to
1489 be defined. Add a hack in mkfs_lustre.c to work around this compile
1494 Description: LASSERT on MDS when client holding flock lock dies
1495 Details : ldlm pool logic depends on number of granted locks equal to
1496 number of released locks which is not true for flock locks, so
1497 just exclude such locks from consideration.
1501 Description: MDS deadlock with many ll_sync_lov threads and I/O stalled
1502 Details : Use fsfilt_sync() for both the whole filesystem sync and
1503 individual file sync to eliminate dangerous inode locking
1504 with I_LOCK that can lead to a deadlock.
1508 Description: Update an obsolete wirecheck.c generator
1509 Details : Update wirecheck.c/wirehdr.c and regenerate wiretest.c
1513 Description: Client can panic on open sometimes
1514 Details : It is possible that we try to free already freed request in
1515 ll_file_open in some error cases when we send request from
1520 Description: performance in 1.6.3
1521 Details : Force q->max_phys_segments to MAX_PHYS_SEGMENTS on SLES10 to be
1522 sure that 1MB requests are not fragmented by the block layer.
1526 Description: LDLM soft lockups - improvement
1527 Details : It is be possible to send the lock handle along with each read
1528 or write request because the client is already doing a lock match
1529 itself so there isn't any reason the OST should have to re-do that
1535 Description: lfs quota fails with deactivated OSTS
1536 Details : With this patch, three improvements are included:
1537 1. detete the softlimit in mds and osts when use "lfs quota".
1538 2. display the inaccurate data in the output of "lfs quota".
1539 3. try to get quota info when "lfs quota" is executed.
1544 Description: Extent locks not granted with no conflicts sometimes.
1545 Details : When race occurs in glimpse handler and nothing is returned,
1546 we do not reprocess the queue after lock cancel, and that leads
1547 to a stall until next activity on a resource
1550 Frequency : failover with quotaon
1552 Description: during mds failovers with quota on, OSTs got into deadlock state
1553 and causing dumpstack.
1554 Details : for every quota slave, at any time, there is only one quota req
1555 is sent to quota master for every uid/gid. Before that quota req
1556 returns, all the thread relative to the same uid/gid will wait.
1557 So if the quota req is lost because mds failovers or any other
1558 reasons, this bug will be hit. Now, dqacq_interpret() will handle
1559 quota reqs who time out.
1561 Severity : enhancement
1564 Description: when quota slave checks if quota is enough, there is an unnecessary
1566 Details : place this wait on necessary place instead of always waiting.
1568 --------------------------------------------------------------------------------
1570 2007-12-07 Cluster File Systems, Inc. <info@clusterfs.com>
1572 * Support for kernels:
1573 2.6.5-7.286 (SLES 9),
1574 2.6.9-55.0.9.EL (RHEL 4),
1575 2.6.16.53-0.8 (SLES 10),
1576 2.6.18-8.1.14.el5 (RHEL 5),
1577 2.6.18.8 vanilla (kernel.org)
1578 * Client support for unpatched kernels:
1579 (see http://wiki.lustre.org/index.php?title=Patchless_Client)
1580 2.6.16 - 2.6.22 vanilla (kernel.org)
1581 * Due to recently discovered recovery problems, we do not recommend
1582 using patchless RHEL 4 clients with this or any earlier release.
1583 * Recommended e2fsprogs version: 1.40.2-cfs1
1584 * Note that reiserfs quotas are disabled on SLES 10 in this kernel.
1587 Frequency : occasional
1589 Description: MDS hang or stay in waiting lock
1590 Details : If client receive lock with CBPENDING flag ldlm need send lock
1591 cancel as separate rpc, to avoid situation when cancel request
1592 can't processed due all i/o threads stay in wait lock.
1595 Frequency : occasional
1597 Description: Do not fail import if osc_interpret_create gets -EAGAIN
1598 Details : If osc_interpret_create got -EAGAIN it immediately exits and
1599 wakeup oscc_waitq. After wakeup oscc_wait_for_objects call
1600 oscc_has_objects and see OSC has no objests and call
1601 oscc_internal_create to resend create request.
1603 Severity : enhancement
1605 Description: Update kernel patches for SLES10 2.6.16.53-0.8.
1606 Details : Update which_patch & target file for SLES10 latest kernel.
1608 Severity : enhancement
1610 Description: add --type and --size parameters to lfs find
1611 Details : Enhance lfs find by adding filetype and filesize parameters. Also
1612 multiple OBDs can now be specified for the --obd option.
1614 Severity : enhancement
1616 Description: eliminate client locks in face of contention
1617 Details : file contention detection and lockless i/o implementation
1618 for contended files.
1620 Severity : enhancement
1622 Description: Remove client patches from SLES 10 kernel.
1623 Details : This causes SLES 10 clients to behave as patchless clients
1624 even on a Lustre-patched (server) kernel.
1626 Severity : enhancement
1628 Description: use i_size_read and i_size_write in 2.6 port
1629 Details : replace inode->i_size access with i_size_read/write()
1632 Frequency : when removing large files
1634 Description: scheduling issue during removal of large Lustre files
1635 Details : Don't take the BKL in fsfilt_ext3_setattr() for 2.6 kernels.
1636 It causes scheduling issues when removing large files (17TB in the
1642 Description: 1.4.11 Can't handle directories with stripe set and extended ACLs
1643 Details : Impossible (EPROTO is returned) to access a directory that has a
1644 non-default striping and ACLs.
1647 Frequency : only on ppc
1649 Description: /proc/fs/lustre/devices broken on ppc
1650 Details : The patch as applied to 1.6.2 doesn't look correct for all arches.
1651 We should make sure the type of 'index' is loff_t and then cast
1652 explicitly as needed below. Do not assign an explicitly cast
1656 Frequency : only for rhel5
1658 Description: Kernel patches update for RHEL5 2.6.18-8.1.10.el5.
1659 Details : Modify the target file & which_kernel.
1662 Frequency : if the uninit_groups feature is enabled on ldiskfs
1664 Description: e2fsck reports "invalid unused inodes count"
1665 Details : If a new ldiskfs filesystem is created with the "uninit_groups"
1666 feature and only a single inode is created in a group then the
1667 "bg_unused_inodes" count is incorrectly updated. Creating a
1668 second inode in that group would update it correctly.
1673 Description: buffer overruns could theoretically occur
1674 Details : llapi_semantic_traverse() modifies the "path" argument by
1675 appending values to the end of the origin string, and a buffer
1676 overrun may occur. Adding buffer overrun check in liblustreapi.
1678 Severity : enhancement
1680 Description: Add jbd statistics patch for RHEL5 and 2.6.18-vanilla.
1683 Frequency : only if filesystem is inconsistent
1685 Description: handle "serious error: objid * already exists" more gracefully
1686 Details : If LAST_ID value on disk is smaller than the objects existing in
1687 the O/0/d* directories, it indicates disk corruption and causes an
1688 LBUG(). If the object is 0-length, then we should use the existing
1689 object. This will help to avoid a full fsck in most cases.
1691 Severity : enhancement
1693 Description: Kernel patches update for RHEL4 2.6.9-55.0.6.
1694 Details : Modify vm-tunables-rhel4.patch.
1696 Severity : enhancement
1698 Description: Kernel config for 2.6.18-vanilla.
1699 Details : Modify targets/2.6-vanilla.target.in.
1700 Add config file kernel-2.6.18-2.6-vanilla-i686.config.
1701 Add config file kernel-2.6.18-2.6-vanilla-i686-smp.config.
1702 Add config file kernel-2.6.18-2.6-vanilla-x86_64.config.
1703 Add config file kernel-2.6.18-2.6-vanilla-x86_64-smp.config.
1706 Frequency : occasional
1708 Description: improve handling recoverable errors
1709 Details : If request processed with error which can be recoverable on server
1710 request should be resend, otherwise page released from cache and
1716 Description: Kernel patches update for RHEL5 2.6.18-8.1.14.el5.
1717 Details : Modify target file & which_patch.
1718 A flaw was found in the IA32 system call emulation provided
1719 on AMD64 and Intel 64 platforms. An improperly validated 64-bit
1720 value could be stored in the %RAX register, which could trigger an
1721 out-of-bounds system call table access. An untrusted local user
1722 could exploit this flaw to run code in the kernel
1723 (ie a root privilege escalation). (CVE-2007-4573).
1727 Description: change order of libsysio includes
1728 Details : '#include sysio.h' should always come before '#include xtio.h'
1730 Severity : enhancement
1732 Description: adapt the lustre_config script to support the upgrade case
1733 Details : Add "-u" option for lustre_config script to support upgrading 1.4
1734 server targets to 1.6 in parallel.
1739 Description: To avoid grant space > avaible space when the disk is almost
1740 full. Without this patch you might see the error "grant XXXX >
1741 available" or some LBUG about grant, when the disk is almost
1743 Details : In filter_check_grant, for non_grant cache write, we should
1744 check the left space by if (*left > ungranted + bytes), instead
1745 of (*left > ungranted), because only we are sure the left space
1746 is enough for another "bytes", then the ungrant space should be
1747 increase. In client, we should update cl_avail_grant only there is
1748 OBD_MD_FLGRANT in the reply.
1753 Description: Update RHEL 4 kernel to fix local root privilege escalation.
1754 Details : Update to the latest RHEL 4 kernel to fix the vulnerability
1755 described in CVE-2007-4573. This problem could allow untrusted
1756 local users to gain root access.
1759 Frequency : when using O_DIRECT and quotas
1761 Description: Incorrect file ownership on O_DIRECT output files
1762 Details : block usage reported by 'lfs quota' does not take into account
1763 files that have been written with O_DIRECT.
1767 Description: (rw.c:1323:ll_read_ahead_pages()) ASSERTION(page_idx > ria->ria_stoff) failed
1768 Details : Once the unmatched stride IO mode is detected, shrink the stride-ahead
1769 window to 0. If it does hit cache miss, and read-pattern is still
1770 stride-io mode, does not reset the stride window, but also does not
1771 increase the stride window length in this case.
1773 --------------------------------------------------------------------------------
1775 2007-09-27 Cluster File Systems, Inc. <info@clusterfs.com>
1777 * Support for kernels:
1778 2.6.5-7.286 (SLES 9),
1779 2.6.9-55.0.2.EL (RHEL 4),
1780 2.6.16.46-0.14 (SLES 10),
1781 2.6.18-8.1.8.el5 (RHEL 5),
1782 2.6.18.8 vanilla (kernel.org)
1783 * Client support for unpatched kernels:
1784 (see http://wiki.lustre.org/index.php?title=Patchless_Client)
1785 2.6.16 - 2.6.21 vanilla (kernel.org)
1786 * Due to recently discovered recovery problems, we do not recommend
1787 using patchless RHEL 4 clients with this or any earlier release.
1788 * Recommended e2fsprogs version: 1.40.2-cfs1
1789 * Note that reiserfs quotas are disabled on SLES 10 in this kernel.
1793 Description: Fix errors in lfs documentation
1794 Details : Fixes man pages
1796 Severity : enhancement
1798 Description: Adaptive timeouts
1799 Details : RPC timeouts adapt to changing server load and network
1800 conditions to reduce resend attempts and improve recovery time.
1802 Severity : enhancement
1804 Description: llapi_file_create() does not allow some changes
1805 Details : add llapi_file_open() that allows specifying the file creation
1806 mode and open flags, and also returns an open file handle.
1808 Severity : enhancement
1810 Description: df doesn't work properly if diskfs blocksize != 4K
1811 Details : Choose biggest blocksize of OST's as the LOV's blocksize.
1814 Frequency : always on directories with default striping set
1816 Description: lfs find on -1 stripe looping in lsm_lmm_verify_common()
1817 Details : Avoid lov_verify_lmm_common() on directory with -1 stripe count.
1819 Severity : enhancement
1821 Description: merge and cleanup kernel patches.
1822 Details : Remove mnt_lustre_list in vfs_intent-2.6-rhel4.patch.
1824 Severity : enhancement
1826 Description: RedHat Update kernel for RHEL5
1827 Details : Kernel config file for RHEL5.
1829 Severity : enhancement
1831 Description: OSS needs mutliple precreate threads
1832 Details : Add ability to start more than one create thread per OSS.
1835 Frequency : only with quota on the root user
1837 Description: mds_obd_create error creating tmp object
1838 Details : When the user sets quota on root, llog will be affected and can't
1839 create files and write files.
1842 Frequency : Always on ia64 patchless client, and possibly others.
1844 Description: Add EXPORT_SYMBOL check for node_to_cpumask symbol.
1845 Details : This allows the patchless client to be loaded on architectures
1846 without this export.
1851 Description: disorder of journal start and llog_add cause deadlock.
1852 Details : in llog_origin_connect, journal start should happen before
1853 llog_add keep the same order as other functions to avoid
1856 Severity : enhancement
1858 Description: RedHat Update kernel for RHEL5
1859 Details : Modify the kernel config file more closer RHEL5.
1861 Severity : enhancement
1863 Description: Build failure against Centos5 (RHEL5)
1864 Details : Define PAGE_SIZE when it isn't present.
1867 Frequency : occasionally when using NFS
1869 Description: "ll_intent_file_open()) lock enqueue: err: -13" with nfs
1870 Details : with NFS, the anon dentry's parent was set to itself in
1871 d_alloc_anon(), so in MDS, we use rec->ur_fid1 to find the
1872 corresponding dentry other than use rec->ur_name.
1874 Severity : enhancement
1876 Description: enable data checksumming by default
1877 Details : enable checksum by default, allow --disable-checksum
1878 configure option and "-o nochecksum" mount option. Checksums
1879 can also be disabled at runtime via $LPROC/osc/*/checksum_pages.
1882 Frequency : Occasionally with failover
1884 Description: Client eviction due to failover config
1885 Details : after a connection loss, the lustre client should attempt to
1886 reconnect to the last active server first before trying the
1887 other potential connections.
1889 Severity : enhancement
1891 Description: client-side metadata stat-ahead during readdir(directory readahead)
1892 Details : perform client-side metadata stat-ahead when the client detects
1893 readdir and sequential stat of dir entries therein
1898 Description: when mds and osts use different quota unit(32bit and 64bit),
1899 quota will be released repeatly.
1900 Details : void sending multiple quota reqs to mds, which will keep the
1901 status between the reqs.
1904 Frequency : only with liblustre clients on XT3
1906 Description: evictions taking too long
1907 Details : allow llrd to evict clients directly on OSTs
1911 Description: osts not allocated evenly to files
1912 Details : change the condition to increase offset_idx
1914 Severity : enhancement
1916 Description: self-adjustable client's lru lists
1917 Details : use adaptive algorithm for managing client cached locks lru
1918 lists according to current server load, other client's work
1919 pattern, memory activities, etc. Both, server and client
1920 side namespaces provide number of proc tunables for controlling
1925 Description: rewrite ext2-derived code in llite/dir.c and obdclass/uuid.c
1926 Details : rewrite inherited code (uuid parsing code from ext2 utils and
1927 readdir code from ext3) from scratch preserving functionality.
1931 Description: Only those disconnect error should be returned by rq_status.
1932 Details : In open/enqueue processs, Some errors, which will cause client
1933 disconnected, should be returned by rq_status, while other
1934 errors should still be returned by intent, then mdc or llite will
1937 Severity : enhancement
1939 Description: Tune the kernel for good SCSI performance.
1940 Details : Set the value of /sys/block/{dev}/queue/max_sectors_kb
1941 to the value of /sys/block/{dev}/queue/max_hw_sectors_kb
1945 Frequency : Always for filesystems larger than 2TB on 32-bit systems.
1946 Bugzilla : 13547, 13627
1947 Description: Data corruption for OSTs that are formatted larger than 2TB
1949 Details : When generating the bio request for lustre file writes the
1950 sector number would overflow a temporary variable before being
1951 used for the IO. The data reads correctly from Lustre (which
1952 will overflow in a similar manner) but other file data or
1953 filesystem metadata may be corrupted in some cases.
1957 Description: "lfs find -obd UUID" prints directories
1958 Details : "lfs find -obd UUID" will return all directory names instead
1959 of just file names. It is incorrect because the directories
1960 do not reside on the OSTs.
1963 Bugzilla : 13596, 13614
1964 Description: MDS hang after unclean shutdown of lots of clients
1965 Details : Never resend AST requests.
1969 Frequency : Always, for kernels after 2.6.16
1970 Description: Fix warning idr_remove called for id=.. which is not allocated.
1971 Details : Last kernels save old s_dev before kill super and not allow
1972 to restore from callback - restore it before call kill_anon_super.
1976 Description: Fix error on 'ls .' at the top of the Lustre mount.
1977 Details : Don't revalidate dentry if it is a root dentry.
1979 --------------------------------------------------------------------------------
1981 2007-08-27 Cluster File Systems, Inc. <info@clusterfs.com>
1983 * Support for kernels:
1984 2.6.5-7.286 (SLES 9),
1985 2.6.9-55.0.2.EL (RHEL 4),
1986 2.6.16.46-0.14 (SLES 10),
1987 2.6.18-8.1.8.el5 (RHEL 5),
1988 2.6.18.8 vanilla (kernel.org)
1989 * Client support for unpatched kernels:
1990 (see http://wiki.lustre.org/index.php?title=Patchless_Client)
1991 2.6.16 - 2.6.22 vanilla (kernel.org)
1992 * Due to recently discovered recovery problems, we do not recommend
1993 using patchless RHEL 4 clients with this or any earlier release.
1994 * Recommended e2fsprogs version: 1.39.cfs8
1995 * Note that reiserfs quotas are disabled on SLES 10 in this kernel.
2000 Description: block reactivating mgc import until all deactivates complete
2001 Details : Fix race when failing back MDT/MGS to itself (testing)
2003 Severity : enhancement
2005 Description: lfs setstripe enhancement
2006 Details : Make lfs setstripe understand 'k', 'm' and 'g' for stripe size.
2009 Frequency : mds/oss recovery
2011 Description: llog ctxt is refrenced after it has been freed.
2012 Details : llog ctxt refcount was added to avoide the race between ctxt free
2013 and llog recovery process. Each llog user must hold ctxt refcount
2014 before it access the llog. And the llog ctxt can only be freed
2015 when its refcount is zero.
2017 Severity : enhancement
2019 Description: randomly memory allocation failure util
2020 Details : Make lustre randomly failed allocating memory for testing purpose.
2023 Frequency : only for SLES10
2025 Description: Update kernel patch for SLES10 SP1
2026 Details : Add patch blkdev_tunables-2.6-sles10.patch to 2.6-sles10.series.
2028 Severity : enhancement
2030 Description: omit set fsid for export NFS
2031 Details : fix set/restore device id for avoid EMFILE error and mark lustre fs
2032 as FS_REQUIRES_DEV for avoid problems with generate fsid.
2035 Frequency : after network failures
2037 Description: Add sync option to mount_lustre.c
2038 Details : Client loses data written to lustre after a network interruption.
2040 Severity : enhancement
2042 Description: Error message improvement.
2043 Details : Merging of two LCONSOLE_ERROR_MSG into one.
2045 Severity : enhancement
2047 Description: don't use GFP_* in generic Lustre code.
2048 Details : Use cfs_alloc_* functions and CFS_* flags for code portability.
2050 Severity : enhancement
2052 Description: obdclass is limited by single OBD_ALLOC(idarray)
2053 Details : replace OBD_ALLOC/OBD_FREE with OBD_VMALLOC/OBD_VFREE
2055 Severity : enhancement
2057 Description: updated patchess for new RHEL4 kernel
2058 Details : Fixed ext3-unlink-race.patch per Kalpak's comment.
2060 Severity : enhancement
2062 Description: warnings with build patchless client with vanila 2.6.19 and up
2063 Details : change old ctl_table style and replace ctl_table/ctl_table_header
2064 with cfs_sysctl_table_t/cfs_sysctl_table_header_t
2066 Severity : enhancement
2068 Description: O_DIRECT bypasses client statistics.
2069 Details : When running with O_DIRECT I/O, neither the client rpc_stats nor
2070 read_ahead_stats were updated. Copied stats section from
2071 osc_send_oap_rpc() into async_internal().
2074 Frequency : only for Cray XT3
2076 Description: peer credits not enough on many OST per OSS systems.
2077 Details : Use new lnet way to add credits as we need those for pings and ASTs
2080 Frequency : only with liblustre
2082 Description: Liblustre is not releasing flock locks on file close.
2083 Details : Release flock locks on file close.
2086 Frequency : only for RHEL4
2088 Description: Update kernel patches for kernel-2.6.9-55.0.2.EL
2089 Details : Remove inode-nr_unused-2.6.9-rhel4.patch from 2.6-rhel4.series
2090 Update target file and kernel config.
2094 Description: lustre support for RHEL5
2095 Details : Add support for RHEL5.
2100 Description: ASSERTION(export != NULL) failed in target_handle_connect
2101 Details : Assetion hit is result of rare race between disconnect and connet
2102 to same nid. target_handle_connect found old connect cockie and
2103 tried to reconnect, but can't find export for this cockie.
2108 Description: umount blocks forever on error
2109 Details : In result of wrong using obd_no_recov and obd_force flags client
2110 can hand if cancel or some other requests is lost.
2113 Frequency : Only for SLES
2115 Description: sanity_quota fail test_1
2116 Details : There are multiple occurences of $TSTUSR in SLES's /etc/group
2117 file, which makes TSTID[2] inunique.
2119 Severity : enhancement
2121 Description: Kernel patches for SLES9 2.6.5-7.286 kernel
2122 Details : Update target/ChangeLog/which_patch .
2124 Severity : enhancement
2126 Description: jbd statistics
2127 Details : Port older jbd statistics patch for sles10
2132 Description: lvbo_init failed for resource with missing objects.
2133 Details : Fix returning error if we do stat for file with missing/corrupted
2134 objects and i_size set to all sum of size of all avaible objects.
2135 if we to truncate/write to missing object - it is recreated.
2140 Description: Oops in read and write path when failing to allocate lock.
2141 Details : Check if lock allocation failed and return error back.
2144 Frequency : When flocks are used.
2146 Description: assertion failure in ldlm_cli_enquque_fini for non NULL lock.
2147 Details : Flock locks might destroy just granted lock if it could be merged
2148 with another existing flock, this is done in completion handler,
2149 so teach ldlm_cli_enquque_fini that this is a valid case for
2155 Description: reply_lock_interpret crash due to race with it and lock cancel.
2156 Details : Do not replay locks that are being cancelled. Do not reference
2157 locks by their address during replay, just by their handle.
2160 Frequency : only with deactivated OSTs
2162 Description: lstripe command fails for valid OST index
2163 Details : The stripe offset is compared to 'lov->desc.ld_tgt_count'
2164 instead of lov->desc.ld_active_tgt_count.
2166 Severity : enhancement
2168 Description: Build failure against Centos5 (RHEL5)
2169 Details : Use getpagesize() instead of PAGE_SIZE.
2171 --------------------------------------------------------------------------------
2173 2007-07-30 Cluster File Systems, Inc. <info@clusterfs.com>
2175 * Support for kernels:
2176 2.6.5-7.283 (SLES 9),
2177 2.6.9-55.EL (RHEL 4),
2178 2.6.16.46-0.14 (SLES 10),
2179 2.6.18.8 vanilla (kernel.org)
2180 * Client support for unpatched kernels:
2181 (see http://wiki.lustre.org/index.php?title=Patchless_Client)
2182 2.6.16 - 2.6.22 vanilla (kernel.org)
2183 * Due to recently discovered recovery problems, we do not recommend
2184 using patchless RHEL 4 clients with this or any earlier release.
2185 * Recommended e2fsprogs version: 1.39.cfs8
2186 * Note that reiserfs quotas are disabled on SLES 10 in this kernel.
2187 * Starting with this release, the ldiskfs backing filesystem required
2188 by Lustre is now in its own package, lustre-ldiskfs. This package
2189 should be installed. It is versioned separately from Lustre and
2190 may be released separately in future.
2195 Description: server LBUG when shutting down
2196 Details : Block umount forever until the mount refcount is zero rather
2197 than giving up after an arbitrary timeout.
2199 Severity : enhancement
2201 Description: add optional extra BUILD_VERSION info
2202 Details : add a new environment variable (namely LUSTRE_VERS) which allows
2203 to override the lustre version.
2206 Frequency : 2.6.18 servers only
2208 Description: ll_kern_mount() doesn't release the module reference
2209 Details : The ldiskfs module reference count never drops down to 0
2210 because ll_kern_mount() doesn't release the module reference.
2215 Description: server LBUG when using old ost_num_threads parameter
2216 Details : Accept the old ost_num_threads parameter but warn that it
2217 is deprecated, and fix an off-by-one error that caused an LBUG.
2222 Description: Transient SCSI error results in persistent IO issue
2223 Details : iobuf->dr_error is not reinitialized to 0 between two
2227 Frequency : sometimes when underlying device returns I/O errors
2229 Description: OSTs not going read-only during write failures
2230 Details : OSTs are not remounted read-only when the journal commit threads
2231 get I/O errors because fsfilt_ext3 calls journal_start/stop()
2232 instead of the ext3 wrappers.
2235 Frequency : only with 10000 clients or more
2237 Description: poor connect scaling with increasing client count
2238 Details : Don't run filter_grant_sanity_check for more than 100 exports
2239 to improve scaling for large numbers of clients.
2242 Frequency : SLES10 only
2244 Description: sanity-quota.sh quotacheck failed: rc = -22
2245 Details : Quotas cannot be enabled on SLES10.
2248 Frequency : liblustre clients only
2250 Description: getdirentries does not give error when run on compute nodes
2251 Details : getdirentries does not fail when the size specified as an argument
2252 is too small to contain at least one entry
2254 Severity : enhancement
2256 Description: Add LNET router traceability for debug purposes
2257 Details : If a checksum failure occurs with a router as part of the
2258 IO path, the NID of the last router that forwarded the bulk data
2259 is printed so it can be identified.
2264 Description: OST "spontaneously" evicts client; client has imp_pingable == 0
2265 Details : Due to a race condition, liblustre clients were occasionally
2266 evicted incorrectly.
2268 Severity : enhancement
2270 Description: lfs setstripe use optional parameters instead of postional
2273 Severity : enhancement
2275 Description: Nanosecond timestamp support for ldiskfs
2276 Details : The on-disk ldiskfs filesystem has added support for nanosecond
2277 resolution timestamps. There is not yet support for this at
2278 the Lustre filesystem level.
2281 Frequency : during server recovery
2283 Description: MDS failing to send precreate requests due to OSCC_FLAG_RECOVERING
2284 Details : request with rq_no_resend flag not awake l_wait_event if they get a
2288 Frequency : nfs export on patchless client
2290 Description: connectathon hang when test nfs export over patchless client
2291 Details : Disconnected dentry cannot be found with lookup, so we do not need
2292 to unhash it or make it invalid
2295 Description: fix llapi_lov_get_uuids() to allow many OSTs to be returned
2296 Details: : Change llapi_lov_get_uuids() to read the UUIDs from /proc instead
2297 of using an ioctl. This allows lfsck for > 160 OSTs to succeed.
2302 Description: open req refcounting wrong on reconnect
2303 Details : If reconnect happened between getting open reply from server and
2304 call to mdc_set_replay_data in ll_file_open, we will schedule
2305 replay for unreferenced request that we are about to free.
2306 Subsequent close will crash in variety of ways.
2307 Check that request is still eligible for replay in
2308 mdc_set_replay_data().
2313 Description: disable writes to filesystem when reading health_check file
2314 Details : the default for reading the health_check proc file has changed
2315 to NOT do a journal transaction and write to disk, because this
2316 can cause reads of the /proc file to hang and block HA state
2317 checking on a healthy but otherwise heavily loaded system. It
2318 is possible to return to the previous behaviour during configure
2319 with --enable-health-write.
2321 Severity : enhancement
2323 Description: 64-bit inode version
2324 Details: : Add a on-disk 64-bit inode version for ext3 to track changes made
2325 to the inode. This will be required for version-based recovery.
2330 Description: MDS fails to start if a duplicate client export is detected
2331 Details : in some rare cases it was possible for a client to connect to
2332 an MDS multiple times. Upon recovery the MDS would detect this
2333 and fail during startup. Handle this more gracefully.
2335 Severity : enhancement
2337 Description: Add -o localflock option to simulate old noflock behaviour.
2338 Details : This will achieve local-only flock/fcntl locks coherentness.
2343 Description: log_commit_thread vs filter_destroy race leads to crash
2344 Details : Take import reference before releasing llog record semaphore
2349 Description: Wrong request locking in request set processing
2350 Details : ptlrpc_check_set wrongly uses req->rq_lock for proctect add to
2351 imp_delayed_list, in this place should be used imp_lock.
2354 Frequency : when reconnecting
2356 Description: Grant leak when OSC reconnect to OST
2357 Details : When osc reconnect ost, OST(filter) should check whether it
2358 should grant more space to client by comparing fed_grant and
2359 cl_avail_grant, and return the granted space to client instead
2360 of "new granted" space, because client will call osc_init_grant
2361 to update the client grant space info.
2364 Frequency : when client reconnects to OST
2366 Description: Grant leak when OSC does a resend and replays bulk write
2367 Details : When osc reconnect to OST, OST(filter) should clear grant info of
2368 bulk write request, because the grant info while be sync between
2369 OSC and OST when reconnect, and we should ignore the grant info
2370 these of resend/replay write req.
2375 Description: Grant space more than avaiable space sometimes.
2376 Details : When then OST is about to be full, if two bulk writing from
2377 different clients came to OST. Accord the avaliable space of the
2378 OST, the first req should be permitted, and the second one
2379 should be denied by ENOSPC. But if the seconde arrived before
2380 the first one is commited. The OST might wrongly permit second
2381 writing, which will cause grant space > avaiable space.
2384 Frequency : when client is evicted
2386 Description: Grant might be wrongly erased when osc is evicted by OST
2387 Details : when the import is evicted by server, it will fork another
2388 thread ptlrpc_invalidate_import_thread to invalidate the
2389 import, where the grant will be set to 0. While the original
2390 thread will update the grant it got when connecting. So if
2391 the former happened latter, the grant will be wrongly errased
2392 because of this race.
2397 Description: Checking Stale with correct fid
2398 Details : ll_revalidate_it should uses de_inode instead of op_data.fid2
2399 to check whether it is stale, because sometimes, we want the
2400 enqueue happened anyway, and op_data.fid2 will not be initialized.
2402 Severity : enhancement
2404 Description: update patchless client
2405 Details : Add support for patchless client with 2.6.20, 2.6.21 and RHEL 5
2408 Frequency : only with 2.4 kernel
2410 Description: random memory corruption
2411 Details : size of struct ll_inode_info is to big for union inode.u and this
2412 can be cause of random memory corruption.
2417 Description: Memory leak in recovery
2418 Details : Lov_mds_md was not free in an error handler in mds_create_object.
2419 It should also check obd_fail before fsfilt_start, otherwise if
2420 fsfilt_start return -EROFS,(failover mds during mds recovery).
2421 then the req will return with repmsg->transno = 0 and rc = EROFS.
2422 and we met hit the assert LASSERT(req->rq_reqmsg->transno ==
2423 req->rq_repmsg->transno) in ptlrpc_replay_interpret. Fcc should
2424 be freed no matter whether fsfilt_commit success or not.
2427 Frequency : only with large numbers of cores on a single node
2429 Description: Prevents from taking the superblock lock in llap_from_page for
2431 Details : using LL_ORIGIN_REMOVEPAGE origin flag instead of LL_ORIGIN_UNKNOW
2432 for llap_from_page call in ll_removepage() prevents from taking
2433 the superblock lock for a soon died page.
2438 Description: Not check open intent error before release open handle
2439 Details : in some rare cases, the open intent error is not checked before
2440 release open handle, which may cause
2441 ASSERTION(open_req->rq_transno != 0), because it tries to release
2442 the failed open handle.
2447 Description: Set cat log bitmap only after create log success.
2448 Details : in some rare cases, the cat log bitmap is set too early. and it
2449 should be set only after create log success.
2453 Description: Accessing a block bevice can re-enable I/O when Lustre is
2454 tearing down a device.
2455 Details : dev_clear_rdonly(bdev) must be called in kill_bdev() instead of
2460 Description: service threads may hog cpus when there are a lot of requests
2461 Details : Insert cond_resched to give other threads a chance to use some CPU
2466 Description: the cat log was not initialized in recovery
2467 Details : When mds(mgs) do recovery, the tgt_count might be zero, so the
2468 unlink log on mds will not be initialized until mds post
2469 recovery. And also in mds post recovery, the unlink log will
2470 initialization will be done asynchronausly, so there will be race
2471 between add unlink log and unlink log initialization.
2475 Description: brw_stats were being printed incorrectly
2476 Details : brw_stats were being printed as log2 but all of them were not
2477 recorded as log2. Also remove some code duplication arising from
2478 filter_tally_{read,write}.
2482 Frequency : rare, only in recovery.
2483 Description: ASSERTION(req->rq_type != LI_POISON) failed
2484 Details : imp_lock should be held while iterating over imp_sending_list for
2485 prevent destroy request after get timeout in ptlrpc_queue_wait.
2489 Description: replay-single.sh test 52 fails
2490 Details : A lock's skiplist need to be cleanup when it being unlinked
2491 from its resource list.
2496 Description: Short directio read returns full requested size rather than
2498 Details : Direct I/O operations should return actual amount of bytes
2499 transferred rather than requested size.
2501 Severity : enhancement
2503 Description: metadata RPC reduction (e.g. for rm performance)
2504 Details : decrease the amount of synchronous RPC between clients and servers
2505 by canceling conflicing lock before the operation on the client
2506 and packing thier handles into the main operation RPC to server.
2508 Severity : enhancement
2510 Description: add #ifdef HAVE_KERNEL_CONFIG_H
2511 Details : kernels from 2.6.19 not need include linux/config.h, but add
2512 include linux/autoconf.h in commpiler command line.
2514 Severity : enhancement
2516 Description: patchless client support for 2.6.22 kernel
2517 Details : 2.6.22 has only one visble change, SLAB_CTOR_* constants is
2518 removed. In this case we need drop using os depended interface to
2519 kmem_cache and use cfs_mem_cache API.
2524 Description: fix mal-formatted messages
2525 Details : fix some mal-formatted DEBUG_REQ and LCONSOLE_ERROR_MSG messages
2529 Frequency : always in liblustre
2530 Description: wrong IS_ERR implementation in liblustre.h
2531 Details : fix IS_ERR implementation in liblustre.h for right detect errors.
2536 Description: Correct condition for output debug message.
2537 Details : inode i_nlink equal zero is not enough for output message about
2538 disk corruption, i_ctime and i_mode should be also checked.
2542 Frequency : always in patchless client
2543 Description: add configure check for truncate_complete_page
2544 Details : improve checks for exported symbols. This allow run check without
2545 sources, but with Module.symvers shipped with kernel distribution.
2546 add check for truncate_complete_page used by patchless client.
2550 Description: sanity.sh test_77h fails with "test_77h file compare failed"
2551 Details : test_77h uses a file which was messed by other test case.
2555 Description: Not Check whether lov_tgts is NULL in some lov functions
2556 Details : Checking whether lov_tgts is NULL in some functions.
2560 Description: replace obdo_alloc() with OBDO_ALLOC macro
2561 Details : nothing special is done in obdo_alloc() function, and for
2562 debugging purpose, it needs to be replaced with macros.
2566 Description: bad return value and errno from fcntl call
2567 Details : In liblustre API, errno should be a negative value if error
2572 Description: ptlrpc_check_set() LBUG
2573 Details : In case of positive reply from server and failed client bulk
2574 callback after bulk transfer shouldn't LBUG, but process this
2575 request as erroneous.
2577 Severity : enhancement
2579 Description: VFS operations stats tool.
2580 Details : Tool which collects stats by tracking value written in pid,
2581 ppid, gid and uses llstat to generate output to plot graph using
2583 Updated lustre/utils/Makefile.am
2584 Added lustre/utils/ltrack_stats.c
2586 Severity : enhancement
2588 Description: 2.6.18 server support (lustre 1.6.1)
2589 Details : Support for 2.6.18 kernels on the server side.
2594 Description: ASSERTION(imp->imp_conn_current) failed
2595 Details : an assertion failure is hit if a client node boots and attempts to
2596 mount a lustre filesystem faster than RECONNECT_INTERVAL seconds.
2599 Frequency : only for i686
2601 Description: 1.4.11 RC1 build fails for RHEL 4, i686
2602 Details : Fixed config variable for build.
2607 Description: Updated patchess for new RHEL4 kernel
2608 Details : Updated patch inode-nr_unused-2.6.9-rhel4.patch
2609 Updated patch jbd-stats-2.6.9.patch
2610 Updated patch qsnet-rhel4-2.6.patch
2611 Updated patch quota-deadlock-on-pagelock-core.patch
2612 Updated patch vfs_intent-2.6-rhel4.patch
2613 Updated patch vfs_races-2.6-rhel4.patch
2614 Updated series file 2.6-rhel4-titech.series
2615 Updated series file 2.6-rhel4.series
2616 Updated kernel config files
2621 Description: lquota slave complains LBUG when reconnecting with mds
2623 Details : quota slave depends qctxt->lqc_import to send its quota request.
2624 This pointer will be invalid if mds did failover or broke its
2625 connect to osts, which leads to LBUG.
2628 Frequency : when qunit size is too small(less than 20M)
2630 Description: write is stopped by improper -EDQUOT
2631 Details : If the master is busy and qunit size is small enough(let's say 1M),
2632 the slave can not get quota from master on time, which will lead
2633 slave to trigger a -EQUOTA to client.
2638 Description: Deadlock during metadata tests
2639 Details : in prune_dir_dentries(), shrink_dcache_parent() should not be
2640 called with the per-dentry lock held.
2643 Frequency : SLES9 only
2645 Description: Lustre patched kernel for SLES9 SP3 has NR_CPUS set to 8
2646 Details : set CONFIG_NR_CPUS to 128 instead of 8.
2648 Severity : enhancement
2650 Description: remove fs_prep_san_write operation and related patches
2651 Details : remove the ext3-san-jdike patches which are no longer useful.
2656 Description: LDISKFS-fs error (device sdc): ldiskfs_free_blocks
2657 Details : a disk corruption can cause the mballoc code to assert on a
2658 double free or other extent corruptions. Handle these with
2659 ext3_error() instead of assertions.
2662 Frequency : only with mballoc3 code and deep extent trees
2664 Description: ldiskfs_ext_search_right: bad header in inode: unexpected eh_depth
2665 Details : a wrong check of extent headers in ldiskfs_ext_search_right()
2666 can cause the filesystem to be remounted read-only.
2671 Description: lfsck built against 1.4.x cannot run against 1.6.0 lustre
2672 Details : the definition for OBD_IOC_GETNAME changed in 1.6.0. One of the
2673 few external users of this ioctl number is lfsck's call to
2674 llapi_lov_get_uuids() and this caused lfsck to fail at startup.
2675 Add the old ioctl number to the handler so both old and new
2680 Description: parallel lock callbacks
2681 Details : Instead of sending blocking and completion callbacks as separated
2682 requests, adding them to a set and sending in parallel.
2686 Description: Disable most debugging by default
2687 Details : To improve performance, disable most logging (for debug purposes)
2688 by default. VFSTRACE, RPCTRACE, and DLMTRACE are now off by
2689 default, and HA includes fewer messages.
2693 Frequency : only run on patchless client.
2694 Description: use do_facet on sanity.sh for test handling recoverables errors
2695 Details : use do_facet instead of direct use sysctl for set fail_loc on OST
2699 Description: hash tables for lists of nids, connections and uuids
2700 Details : Hash tables noticeably help when a lot of clients connect to a
2701 server, to faster identify duplicate connections or reconnects,
2702 also to faster find export to evict in manual eviction case.
2706 Description: Sometimes, when the server evict a client, and the client will
2707 not be evicted as soon as possible.
2708 Details : In enqueue req, the error was returned by intent, instead of
2709 rq_status which make ptlrpc layer not detect this error, and
2710 does not evict the client. So enqueue error should be returned
2714 Frequency : only at startup
2716 Description: Delay client connections to MDT until fist MDT->OST connect
2717 Details : If a client tried to create a new file before the MDT had
2718 connected to any OSTs, the create would return EIO. Now
2719 the client will simply block until the MDT connects to the
2720 first OST and the create can succeed.
2723 Frequency : only for SLES9
2725 Description: Routinely utilize latest Quadrics drivers in CFS releases
2726 Details : Update patch qsnet-suse-2.6.patch.
2729 Frequency : only for sles10
2731 Description: Update patches for SLES 10 SP1 kernel.
2732 Details : Update the patch vfs_intent-2.6-sles10.patch.
2737 Description: Routinely utilize latest Quadrics drivers in CFS releases
2738 Details : Update patch qsnet-rhel4-2.6.patch.
2741 Frequency : at statup only
2743 Description: mds_lov_synchronize race leads to various problems
2744 Details : simultaneous MDT->OST connections at startup can cause the
2745 sync to abort, leaving the OSC in a bad state.
2750 Description: Using wrong pointer in osc_brw_prep_request
2751 Details : Access to array[-1] can produce panic if kernel compiled with
2752 CONFIG_PAGE_ALLOC enabled
2754 Severity : enhancement
2756 Description: Async OSC create to avoid the blocking unnecessarily.
2757 Details : If a OST has no remain object, system will block on the creating
2758 when need to create a new object on this OST. Now, ways use
2759 pre-created objects when available, instead of blocking on an
2760 empty osc while others are not empty. If we must block, we block
2761 for the shortest possible period of time.
2765 Frequency : only in recovery
2766 Description: Mark OST as early accessible if his start SYNC.
2767 Details : osc_precreate return flag early accessible if oscc marked as
2768 OSCC_FLAG_SYNC_IN_PROGRESS.
2773 Description: Sometimes precreate code can triger create object on wrong ost
2774 Details : Wrong protected or not not restored variables aftre precreate loop
2775 can produce creation object on wrong ost.
2778 Frequency : oss recovery
2780 Description: llog_commit_thread cleanup should sync with llog_commit_thread
2782 Details : llog_commit_thread_count should be synced between llog_commit
2783 start and cleanup, so new llog_commit thread should not be started
2784 when llog_commit threads being stopped to avoid accessing some
2787 Severity : enhancement
2789 Description: Add printing inode info into message about error in writepage.
2791 --------------------------------------------------------------------------------
2793 2007-05-03 Cluster File Systems, Inc. <info@clusterfs.com>
2798 Frequency : on some architectures
2800 Description: 1.6 client sometimes fails to mount from a 1.4 MDT
2801 Details : Uninitialized flags sometimes cause configuration commands to
2805 Frequency : patchless clients only
2807 Description: missing __iget() symbol export
2808 Details : The __iget() symbol export is missing. To avoid the need for
2809 this on patchless clients the deathrow inode reaper is turned
2810 off, and we depend on the VM to clean up old inodes. This
2811 dependency was during via the fix for bug 12181.
2816 Description: sanity.sh fail: test_52b
2817 Details : The ll_inode_to_ext_flags() has a glitch which makes MDS return
2818 incorrect inode's flags to client.
2820 --------------------------------------------------------------------------------
2822 2007-04-19 Cluster File Systems, Inc. <info@clusterfs.com>
2824 * CONFIGURATION CHANGE. This version of Lustre WILL NOT
2825 INTEROPERATE with older versions automatically. In many cases a
2826 special upgrade step is needed. Please read the
2827 user documentation before upgrading any part of a 1.4.x system.
2828 * WARNING: Lustre configuration and startup changes are required with
2829 this release. See https://mail.clusterfs.com/wikis/lustre/MountConf
2831 * Support for kernels:
2832 2.4.21-47.0.1.EL (RHEL 3),
2833 2.6.5-7.283 (SLES 9),
2834 2.6.9-42.0.10.EL (RHEL 4),
2835 2.6.12.6 vanilla (kernel.org),
2836 2.6.16.27-0.9 (SLES10)
2837 * Client support for unpatched kernels:
2838 (see https://mail.clusterfs.com/wikis/lustre/PatchlessClient)
2839 2.6.16 - 2.6.19 vanilla (kernel.org),
2840 2.6.9-42.0.8EL (RHEL 4)
2841 * Recommended e2fsprogs version: 1.39.cfs6
2842 * Note that reiserfs quotas are disabled on SLES 10 in this kernel
2845 Severity : enhancement
2847 Description: Async OSC create to avoid the blocking unnecessarily.
2848 Details : If a OST has no remain object, system will block on the creating
2849 when need to create a new object on this OST. Now, ways use
2850 pre-created objects when available, instead of blocking on an
2851 empty osc while others are not empty. If we must block, we block
2852 for the shortest possible period of time.
2854 Severity : enhancement
2856 Description: MountConf
2857 Details : Lustre configuration is now managed via mkfs and mount
2858 commands instead of lmc and lconf. New obd types (MGS, MGC)
2859 are added for dynamic configuration management. See
2860 https://mail.clusterfs.com/wikis/lustre/MountConf for
2863 Severity : enhancement
2865 Description: dynamic OST addition
2866 Details : OSTs can now be added to a live filesystem
2868 Severity : enhancement
2870 Description: startup order invariance
2871 Details : MDTs and OSTs can be started in any order. Clients only
2872 require the MDT to complete startup.
2874 Severity : enhancement
2876 Description: parallel, asynchronous orphan cleanup
2877 Details : orphan cleanup is now performed in separate threads for each
2878 OST, allowing parallel non-blocking operation.
2880 Severity : enhancement
2882 Description: optimized stripe assignment
2883 Details : stripe assignments are now made based on ost space available,
2884 ost previous usage, and OSS previous usage, in order to try
2885 to optimize storage space and networking resources.
2887 Severity : enhancement
2889 Description: Permanently set tunables
2890 Details : All writable /proc/fs/lustre tunables can now be permanently
2891 set on a per-server basis, at mkfs time or on a live
2894 Severity : enhancement
2896 Description: Lustre message v2
2897 Details : Add lustre message format v2.
2899 Severity : enhancement
2901 Description: client OST exclusion list
2902 Details : Clients can be started with a list of OSTs that should be
2903 declared "inactive" for known non-responsive OSTs.
2907 Description: ENOENT returned for valid filehandle during dbench.
2908 Details : Check if a directory has children when invalidating dentries
2909 associated with an inode during lock cancellation. This fixes
2910 an incorrect ENOENT sometimes seen for valid filehandles during
2911 testing with dbench.
2914 Frequency : SFS test only (otherwise harmless)
2916 Description: SPEC SFS validation failure on NFS v2 over lustre.
2917 Details : Changes the blocksize for regular files to be 2x RPC size,
2918 and not depend on stripe size.
2920 Severity : enhancement
2922 Description: fine-grained SMP locking inside DLM
2923 Details : Improve DLM performance on SMP systems by removing the single
2924 per-namespace lock and replace it with per-resource locks.
2926 Severity : enhancement
2928 Description: don't hold multiple extent locks at one time
2929 Details : To avoid client eviction during large writes, locks are not
2930 held on multiple stripes at one time or for very large writes.
2931 Otherwise, clients can block waiting for a lock on a failed OST
2932 while holding locks on other OSTs and be evicted.
2934 Severity : enhancement
2936 Description: Multiple MD RPCs in flight.
2937 Details : Further unserialise some read-only MDT RPCs - learn about intents.
2938 To avoid overly-overloading MDT, introduce a limit on number of
2939 MDT RPCs in flight for a single client and add /proc controls
2940 to adjust this limit.
2942 Severity : enhancement
2944 Description: client read/write statistics
2945 Details : Add client read/write call usage stats for performance
2946 analysis of user processes.
2947 /proc/fs/lustre/llite/*/offset_stats shows non-sequential
2948 file access. extents_stats shows chunk size distribution.
2949 extents_stats_per_process show chunk size distribution per
2952 Severity : enhancement
2954 Description: per-client statistics on server
2955 Details : Add ldlm and operations statistics for each client in
2956 /proc/fs/lustre/mds|obdfilter/*/exports/
2958 Severity : enhancement
2960 Description: improved MDT statistics
2961 Details : Add detailed MDT operations statistics in
2962 /proc/fs/lustre/mds/*/stats
2964 Severity : enhancement
2966 Description: VFS operations stats
2967 Details : Add client VFS call stats, trackable by pid, ppid, or gid
2968 /proc/fs/lustre/llite/*/stats_track_[pid|ppid|gid]
2973 Description: Fix client-side osc byte counters
2974 Details : The osc read/write byte counters in
2975 /proc/fs/lustre/osc/*/stats are now working
2978 Frequency : always as root on SLES
2980 Description: Failure of copying files with lustre special EAs.
2981 Details : Client side always return success for setxattr call for lustre
2982 special xattr (currently only "trusted.lov").
2987 Description: Refcount LNET uuids
2988 Details : The global LNET uuid list grew linearly with every startup;
2989 refcount repeated list entries instead of always adding to
2992 Severity : enhancement
2994 Description: Dynamic service threads
2995 Details : Within a small range, start extra service threads
2996 automatically when the request queue builds up.
2999 Frequency : mixed-endian client/server environments
3001 Description: mixed-endian crashes
3002 Details : The new msg_v2 system had some failures in mixed-endian
3005 Severity : enhancement
3007 Description: Easy OST removal
3008 Details : OSTs can be permanently deactivated with e.g. 'lctl
3009 conf_param lustre-OST0001.osc.active=0'
3011 Severity : enhancement
3013 Description: MGS proc entries
3014 Details : Added basic proc entries for the MGS showing what filesystems
3017 Severity : enhancement
3019 Description: provide MGS failover
3020 Details : Added config lock reacquisition after MGS server failover.
3022 Severity : enhancement
3024 Description: add Linux 2.4 support
3025 Details : Added support for RHEL 2.4.21 kernel for 1.6 servers and clients
3029 Description: a large application tries to do I/O to the same resource and dies
3030 in the middle of it.
3031 Details : Check the req->rq_arrival time after the call to
3032 ost_brw_lock_get(), but before we do anything about
3033 processing it & sending the BULK transfer request. This
3034 should help move old stale pending locks off the queue as
3035 quickly as obd_timeout.
3038 Frequency : when an incorrect nid is specified during startup
3040 Description: ptlrpc connect to non-existant node causes kernel crash
3041 Details : LNET can't be re-entered from an event callback, which
3042 happened when we expire a message after the export has been
3043 cleaned up. Instead, hand the zombie cleanup off to another
3046 Severity : enhancement
3048 Description: plain/inodebits lock performance improvement
3049 Details : Grouping plain/inodebits in granted list by their request modes
3050 and bits policy, thus improving the performance of search through
3054 Frequency : only if OST filesystem is corrupted
3056 Description: client incorrectly hits assertion in ptlrpc_replay_req()
3057 Details : for a short time RPCs with bulk IO are in the replay list,
3058 but replay of bulk IOs is unimplemented. If the OST filesystem
3059 is corrupted due to disk cache incoherency and then replay is
3060 started it is possible to trip an assertion. Avoid putting
3061 committed RPCs into the replay list at all to avoid this issue.
3064 Frequency : liblustre (e.g. catamount) on a large cluster with >= 8 OSTs/OSS
3066 Description: System hang on startup
3067 Details : This bug allowed the liblustre (e.g. catamount) client to
3068 return to the app before handling all startup RPCs. This
3069 could leave the node unresponsive to lustre network traffic
3070 and manifested as a server ptllnd timeout.
3072 Severity : enhancement
3074 Description: Add "/proc/sys/lustre/debug_peer_on_timeout"
3075 Details : liblustre envirable: LIBLUSTRE_DEBUG_PEER_ON_TIMEOUT
3076 boolean to control whether to print peer debug info when a
3077 client's RPC times out.
3080 Frequency : only for kernels with patches from Lustre below 1.4.3
3082 Description: Remove old rdonly API
3083 Details : Remove old rdonly API which unused from at least lustre 1.4.3
3086 Frequency : only for devices with external journals
3088 Description: Set external device read-only also
3089 Details : During a commanded failover stop, we set the disk device
3090 read-only while the server shuts down. We now also set any
3091 external journal device read-only at the same time.
3094 Frequency : when upgrading from 1.4 while trying to change parameters
3096 Description: The wrong (new) MDC name was used when setting parameters for
3097 upgraded MDT's. Also allows changing of OSC (and MDC)
3098 parameters if --writeconf is specified at tunefs upgrade time.
3101 Frequency : when setting specific ost indicies
3103 Description: QOS code breaks on skipped indicies
3104 Details : Add checks for missing OST indicies in the QOS code, so OSTs
3105 created with --index need not be sequential.
3107 Severity : enhancement
3109 Description: Add uninit_groups feature to ldiskfs2 to speed up e2fsck
3110 Details : The uninit_groups feature works in conjunction with the kernel
3111 filesystem code (ldiskfs2 only) and e2fsprogs-1.39-cfs6 to speed
3112 up the pass1 processing of e2fsck. This is a read-only feature
3113 in ldiskfs2 only, so older kernels and current ldiskfs cannot
3114 mount filesystems that have had this feature enabled.
3116 Severity : enhancement
3118 Description: Improve multi-block allocation algorithm to avoid fragmentation
3119 Details : The mballoc3 code (ldiskfs2 only) adds new mechanisms to improve
3120 allocation locality and avoid filesystem fragmentation.
3122 ------------------------------------------------------------------------------
3124 2007-04-01 Cluster File Systems, Inc. <info@clusterfs.com>
3126 * Support for kernels:
3127 2.4.21-47.0.1.EL (RHEL 3)
3128 2.6.5-7.283 (SLES 9)
3129 2.6.9-42.0.10.EL (RHEL 4)
3130 2.6.12.6 vanilla (kernel.org)
3131 2.6.16.27-0.9 (SLES 10)
3132 * Recommended e2fsprogs version: 1.39.cfs5
3134 * Note that reiserfs quotas are disabled on SLES 10 in this kernel
3138 Frequency : occasional, depends on client load and configuration
3139 Bugzilla : 12181, 12203
3140 Description: data loss for recently-modified files
3142 Details : In some cases it is possible that recently written or created
3143 files may not be written to disk in a timely manner (this should
3144 normally be within 30s unless client IO load is very high).
3145 The problem appears as zero-length files or files that are a
3146 multiple of 1MB in size after a client crash or client eviction
3147 that are missing data at the end of the file.
3149 This problem is more likely to be hit on clients where files are
3150 repeatedly created and unlinked in the same directory, clients
3151 have a large amount of RAM, have many CPUs, the filesystem has
3152 many OSTs, the clients are rebooted frequently, and/or the files
3153 are not accessed by other nodes after being written.
3155 The presence of the problem can be detected by looking at
3156 /proc/sys/fs/inode-state. If the first number (nr_inodes) is
3157 smaller than the second (nr_unused) then dirty files will not
3158 be flushed automatically to disk. "sync; sleep 10" should be
3159 run several times on the node before unmounting it to update
3160 Lustre (this is also safe to run on nodes without this problem).
3162 There is also a related kernel bug in the RHEL4 4 2.6.9 kernel
3163 that can cause this same problem, so customers using that kernel
3164 also need to update the kernel in addition to Lustre. In order
3165 to properly fix this bug, the RHEL3 2.4.21 kernel is also updated.
3167 It is normal that files written just before a client crash (less
3168 than 30s) may not yet have been flushed to disk, even for local
3172 Frequency : frequent on thin XT3 nodes
3174 Description: UUID collision on thin XT3 Linux nodes
3175 Details : UUIDs on Compute Node Linux XT3 nodes were not generated
3176 randomly, since we relied on an insufficiently-seeded PRNG.
3181 Description: OSS hangs after "All ost request buffers busy"
3182 Details : A deadlock between quota and journal operations caused OSS
3183 hangs after printing "All ost request buffers busy."
3186 Frequency : always on liblustre builds
3188 Description: Cleanup compiler warnings on liblustre
3191 Frequency : always on liblustre builds on XT3
3193 Description: LC_CONFIG_CDEBUG don't run while build liblustre on XT3.
3197 Description: Addition of EXT3_FEATURE_RO_COMPAT_DIR_NLINKS flag for
3198 > 32000 subdirectories
3199 Details : Add EXT3_FEATURE_RO_COMPAT_DIR_NLINK flag to
3200 EXT3_FEATURE_RO_COMPAT_SUPP. This flag will be set whenever
3201 subdirectory count crosses 32000. This will aid e2fsck to
3202 correctly handle more than 32000 subdirectories.
3205 Frequency : liblustre (e.g. catamount) on a large cluster with >= 8 OSTs/OSS
3207 Description: System hang on startup
3208 Details : This bug allowed the liblustre (e.g. catamount) client to
3209 return to the app before handling all startup RPCs. This
3210 could leave the node unresponsive to lustre network traffic
3211 and manifested as a server ptllnd timeout.
3213 Severity : enhancement
3215 Description: Add "/proc/sys/lustre/debug_peer_on_timeout"
3216 (liblustre envirable: LIBLUSTRE_DEBUG_PEER_ON_TIMEOUT)
3217 boolean to control whether to print peer debug info when a
3218 client's RPC times out.
3223 Description: make O_SYNC working on 2.6 kernels
3224 Details : 2.6 kernels use different method for mark pages for write,
3225 so need add a code to lustre for O_SYNC work.
3230 Description: Failure to close file and release space on NFS
3231 Details : Put inode details into lock acquired in ll_intent_file_open.
3232 Use mdc_intent_lock in ll_intent_open to properly
3233 detect all kind of errors unhandled by mdc_enqueue.
3238 Description: proc file read during shutdown sometimes raced obd removal,
3240 Details : Add lock to prevent obd access after proc file removal.
3243 Frequency : Only for files larger than 4GB on 32-bit clients.
3245 Description: improperly doing page alignment of locks
3246 Details : Modify lustre core code to use CFS_PAGE_* defines instead of
3247 PAGE_*. Make CFS_PAGE_MASK a 64-bit mask.
3252 Description: RPCs being resent when they shouldn't be
3253 Details : Some RPCs that should not be resent are being resent. This
3254 can cause inconsistencies in the RPC state machine. Do not
3255 resend such requests.
3258 Frequency : rare, only with NFS export
3260 Description: Crash on NFS re-export node
3261 Details : under very unusual load conditions an assertion is hit in
3262 ll_intent_file_open()
3265 Frequency : only if OST filesystem is corrupted
3267 Description: client incorrectly hits assertion in ptlrpc_replay_req()
3268 Details : for a short time RPCs with bulk IO are in the replay list,
3269 but replay of bulk IOs is unimplemented. If the OST filesystem
3270 is corrupted due to disk cache incoherency and then replay is
3271 started it is possible to trip an assertion. Avoid putting
3272 committed RPCs into the replay list at all to avoid this issue.
3277 Description: large O_DIRECT requests fail under memory pressure/fragmentation
3278 Details : Large single O_DIRECT read and write calls can fail to allocate
3279 a sufficiently large buffer to process the request. In case of
3280 allocation failure the allocation is retried with a smaller
3281 buffer and broken into smaller requests.
3283 Severity : enhancement
3285 Description: Add -o localflock option to simulate old noflock behaviour.
3286 Details : This will achieve local-only flock/fcntl locks coherentness.
3291 Description: versioning check is incomplete
3292 Details : Checking the version difference of client vs. server, report
3293 error if the gap is too big.
3298 Description: add support PG_writeback bit
3299 Details : add support for PG_writeback bit for Lustre, for more carefull
3300 work with page cache in 2.6 kernel. This also fix some deadlocks
3301 and remove hack for work O_SYNC with 2.6 kernel.
3303 Severity : enhancement
3305 Description: Add uninit_groups feature to ldiskfs2 to speed up e2fsck
3306 Details : The uninit_groups feature works in conjunction with the kernel
3307 filesystem code (ldiskfs2 only) and e2fsprogs-1.39-cfs6 to speed
3308 up the pass1 processing of e2fsck. This is a read-only feature
3309 in ldiskfs2 only, so older kernels and current ldiskfs cannot
3310 mount filesystems that have had this feature enabled.
3312 Severity : enhancement
3314 Description: Improve multi-block allocation algorithm to avoid fragmentation
3315 Details : The mballoc3 code (ldiskfs2 only) adds new mechanisms to improve
3316 allocation locality and avoid filesystem fragmentation.
3318 ------------------------------------------------------------------------------
3320 2007-02-09 Cluster File Systems, Inc. <info@clusterfs.com>
3322 * Support for kernels:
3323 2.6.9-42.0.3.EL (RHEL 4)
3324 2.6.5-7.276 (SLES 9)
3325 2.4.21-47.0.1.EL (RHEL 3)
3326 2.6.12.6 vanilla (kernel.org)
3327 2.6.16.21-0.8 (SLES10)
3328 * Recommended e2fsprogs version: 1.39.cfs2-0
3330 * The backwards-compatible /proc/sys/portals symlink has been removed
3331 in this release. Before upgrading, please ensure that you change
3332 any configuration scripts or /etc/sysctl.conf files that access
3333 /proc/sys/portals/* or sysctl portals.* to use the corresponding
3334 entry in /proc/sys/lnet or sysctl lnet.*. This change can be made
3335 in advance of the upgrade on any system running Lustre 1.4.6 or
3336 newer, since /proc/sys/lnet was added in that version.
3337 * Note that reiserfs quotas are disabled on SLES 10 in this kernel
3341 Frequency : only when quota is used
3343 Description: avoid scanning export list for quota master
3344 Details : Change the algorithms to avoid scanning export list in order
3345 to improve the efficiency.
3348 Frequency : MDS failover only, very rarely
3350 Description: "went back in time" messages on mds failover
3351 Details : The greatest transno may be lost when the current operation
3352 finishes with an error (transno==0) and the client's last_rcvd
3353 record is over-written. Save the greatest transno in the
3354 mds_last_transno for this case.
3357 Frequency : always for specific kernels and striping counts
3359 Description: client may get "Matching packet too big" without ACL support
3360 Details : Clients compiled without CONFIG_FS_POSIX_ACL get an error message
3361 when trying to access files in certain configurations. The
3362 clients should in fact be denied when mounting because they do
3363 not understand ACLs.
3366 Frequency : Cray XT3 with more than 4000 clients and multiple jobs
3368 Description: many clients connecting with IO in progress causes connect timeouts
3369 Details : Avoid synchronous journal commits to avoid delays caused by many
3370 clients connecting/disconnecting when bulk IO is in progress.
3371 Queue liblustre connect requests on OST_REQUEST_PORTAL instead of
3372 OST_IO_PORTAL to avoid delays behind potentially many pending
3376 Frequency : occasionally with multiple writers to a single file
3378 Description: shared writes to file may result in wrong size reported by stat()
3379 Details : Allow growing of kms when extent lock is cancelled
3382 Frequency : always with random mmap IO to multi-striped file
3384 Description: mmap write might be lost if we are writing to a 'hole' in stripe
3385 Details : Only if the hole is at the end of OST object so that kms is too
3386 small. Fix is to increase kms accordingly in ll_nopage.
3389 Frequency : rare, only if OST filesystem is inconsistent with MDS filesystem
3391 Description: writes to a missing object would leak memory on the OST
3392 Details : If there is an inconsistency between the MDS and OST filesystems,
3393 such that the MDS references an object that doesn't exist, writes
3394 to that object will leak memory due to incorrect cleanup in the
3395 error handling path, eventually running out of memory on the OST.
3400 Description: Creating too long symlink causes lustre errors
3401 Details : Check symlink and name lengths before sending requests to MDS.
3404 Frequency : only if flock is enabled (not on by default)
3406 Description: posix locks not released on fd closure on 2.6.9+
3407 Details : We failed to add posix locks to list of inode locks on 2.6.9+
3408 kernels, this caused such locks not to be released on fd close and
3409 then assertions on fs unmount about still used locks.
3412 Frequency : MDS failover only, very rarely
3414 Description: clients may get ASSERTION(granted_lock != NULL)
3415 Details : When request was taking a long time, and a client was resending
3416 a getattr by name lock request. The were multiple lock requests
3417 with the same client lock handle and
3418 mds_getattr_name->fixup_handle_for_resent_request found one of the
3419 lock handles but later failed with ASSERTION(granted_lock != NULL).
3424 Description: handle->h_buffer_credits > 0, assertion failure
3425 Details : h_buffer_credits is zero after truncate, causing assertion
3426 failure. This patch extends the transaction or creates a new
3430 Frequency : NFS re-export or patchless client
3431 Bugzilla : 11179, 10796
3432 Description: Crash on NFS re-export node (__d_move)
3433 Details : We do not want to hash the dentry if we don't have a lock.
3434 But if this dentry is later used in d_move, we'd hit uninitialised
3435 list head d_hash, so we just do this to init d_hash field but
3436 leave dentry unhashed.
3439 Frequency : NFS re-export or patchless client
3441 Description: NFS exports has problem with symbolic link
3442 Details : lustre client didn't properly install dentry when re-exported
3443 to NFS or running patchless client.
3446 Frequency : NFS re-export or patchless client
3448 Description: Various nfs/patchless fixes.
3449 Details : fixes reuse disconected alias for lookup process - this fixes
3450 warning "find_exported_dentry: npd != pd",
3451 fix permission error with open files at nfs.
3452 fix apply umask when do revalidate.
3455 Frequency : occasional
3457 Description: Crash on NFS re-export node
3458 Details : calling clear_page() on the wrong pointer triggered oops in
3459 generic_mapping_read().
3462 Frequency : rarely, using O_DIRECT IO
3464 Description: unaligned directio crashes client with LASSERT
3465 Details : check for unaligned buffers before trying any requests.
3468 Frequency : rarely, using CFS RAID5 patches in non-standard kernel series
3470 Description: stale data returned from RAID cache
3471 Details : If only a small amount of IO is done to the RAID device before
3472 reading it again it is possible to get stale data from the RAID
3473 cache instead of reading it from disk.
3476 Frequency : always for sles10 kernel
3478 Description: sles10 support
3479 Details : ll_follow_link: compile fixes and using of nd_set_link
3480 under newer kernels.
3483 Frequency : depends on arch, kernel and compiler version, always on sles10
3486 Description: recursive or deep enough symlinks cause stack overflow
3487 Details : getting rid of large stack-allocated variable in
3491 Frequency : depends on hardware
3493 Description: lustre write performance loss in the SLES10 kernel
3494 Details : the performance loss is caused by using of write barriers in the
3495 ext3 code. The SLES10 kernel turns barrier support on by
3496 default. The fix is to undo that change for ldiskfs.
3498 ------------------------------------------------------------------------------
3500 2006-12-09 Cluster File Systems, Inc. <info@clusterfs.com>
3502 * Support for kernels:
3503 2.6.9-42.0.3EL (RHEL 4)
3504 2.6.5-7.276 (SLES 9)
3505 2.4.21-47.0.1.EL (RHEL 3)
3506 2.6.12.6 vanilla (kernel.org)
3510 Frequency : quota enabled and large files being deleted
3512 Description: releasing more than 4GB of quota at once hangs OST
3513 Details : If a user deletes more than 4GB of files on a single OST it
3514 will cause the OST to spin in an infinite loop. Release
3515 quota in < 4GB chunks, or use a 64-bit value for 1.4.7.1+.
3520 Description: statfs data retrieved from /proc may be stale or zero
3521 Details : When reading per-device statfs data from /proc, in the
3522 {kbytes,files}_{total,free,avail} files, it may appear
3523 as zero or be out of date.
3526 Frequency : systems with MD RAID1 external journal devices
3528 Description: lconf's call to blkid is confused by RAID1 journal devices
3529 Details : Use the "blkid -l" flag to locate the MD RAID device instead
3530 of returning all block devices that match the journal UUID.
3533 Frequency : always, for aggregate stripe size over 4GB
3535 Description: "lfs setstripe" fails assertion when setting 4GB+ stripe width
3536 Details : Using "lfs setstripe" to set stripe size * stripe count over 4GB
3537 will fail the kernel with "ASSERTION(lsm->lsm_xfersize != 0)"
3540 Frequency : always if "lfs find" used on a local file/directory
3542 Description: "lfs find" segfaults if used on a local file/directory
3543 Details : The case where a directory component was not specified wasn't
3544 handled correctly. Handle this properly.
3547 Frequency : always on ppc64
3549 Description: the write to an ext3 filesystem mounted with mballoc got stuck
3550 Details : ext3_mb_generate_buddy() uses find_next_bit() which does not
3551 perform endianness conversion.
3554 Frequency : rarely (truncate to non-zero file size after write under load)
3555 Bugzilla : 10730, 10687
3556 Description: Files padded with zeros to next 4K multiple
3557 Details : With filesystems mounted using the "extents" option (2.6 kernels)
3558 it is possible that files that are truncated to a non-zero size
3559 immediately after being written are filled with zero bytes beyond
3560 the truncated size. No file data is lost.
3562 Severity : enhancement
3564 Description: Allow recovery/failover for liblustre clients.
3565 Details : liblustre clients were unaware of failover configurations until
3568 Severity : enhancement
3570 Description: user file locks should fail when not mounting with flock option
3571 Details : Set up an error-returning stub in ll_file_operations.lock field
3572 to prevent incorrect behaviour when client is mounted without
3573 flock option. Also, set up properly f_op->flock field for
3577 Frequency : always on ia64
3579 Description: "lfs df" loops on printing out MDS statfs information
3580 Details : The obd_ioctl_data was not initialized and in some systems
3581 this caused a failure during the ioctl that did not return
3582 an error. Initialize the struct and return an error on failure.
3585 Frequency : SLES 9 only
3587 Description: Error of copying files with lustre special EAs as root
3588 Details : Client side always return success for setxattr call for lustre
3589 special xattr (currently only "trusted.lov").
3592 Frequency : rarely on clusters with both ia64+i386 clients
3594 Description: ia64+i686 clients doing shared IO on the same file may LBUG
3595 Details : In rare cases when both ia64+i686 (or other mixed-PAGE_SIZE)
3596 clients are doing concurrent writes to the same file it is
3597 possible that the ia64 clients may LASSERT because the OST
3598 extent locks are not PAGE_SIZE aligned. Ensure that grown
3599 locks are always aligned on the request boundary.
3602 Frequency : specific use, occasional
3604 Description: Overwriting in use executable truncates on-disk binary image
3605 Details : If one node attempts to overwrite an executable in use by
3606 another node, we now correctly return ETXTBSY instead of
3607 truncating the file.
3609 Severity : enhancement
3611 Description: Async OSC create to avoid the blocking unnecessarily.
3612 Details : If a OST has no remain object, system will block on the creating
3613 when need to create a new object on this OST. Now, ways use
3614 pre-created objects when available, instead of blocking on an
3615 empty osc while others are not empty. If we must block, we block
3616 for the shortest possible period of time.
3621 Description: chmod on Lustre root is propagated to other clients
3622 Details : Re-validate root's dentry in ll_lookup_it to avoid having it
3623 invalid by the follow_mount time.
3626 Frequency : liblustre clients only
3628 Description: Race in 'instant cancel' lock handling could lead to such locks
3629 never to be granted in case of SMP MDS
3630 Details : Do not destroy not yet granted but cbpending locks in
3634 Frequency : replay/resend of open
3636 Description: non null lock assetion failure in mds_intent_policy
3637 Details : Trying to replay/resend lockless open requests resulted in
3638 mds_open() returning 0 with no lock. Now it sets a flag if
3639 a lock is going to be returned.
3641 Severity : enhancement
3643 Description: Checksum enhancements
3644 Details : New checksum enhancements allow for resending RPCs that failed
3647 Severity : enhancement
3649 Description: Tunables on number of dirty pages in cacche
3650 Details : Allow to set limit on number of dirty pages cached.
3655 Description: client crash on unmount - lock still has references
3656 Details : In some error handling cases it was possible to leak a lock
3657 reference on a client while accessing a file. This was not
3658 harmful to the client during operation, but would cause the
3659 client to crash when the filesystem is unmounted.
3662 Frequency : specific case, rare
3664 Description: ETXTBSY on mds though file not in use by client
3665 Details : ETXTBSY is no longer incorrectly returned when attempting to
3666 chmod or chown a directory that the user previously tried to
3667 execute or a currently-executing binary.
3670 Frequency : extremely rare except on liblustre-based clients
3672 Description: Lustre space not freed when files are deleted
3673 Details : Clean up open-unlinked files after client eviction. Previously
3674 the unlink was skipped and the files remained as orphans.
3679 Description: OST failure "would be an LBUG" in waiting_locks_callback()
3680 Details : In some cases it was possible to send a blocking callback to
3681 a client doing a glimpse, even though that client didn't get
3682 a lock granted. When the glimpse lock is cancelled on the OST
3683 the freed lock is left on the waiting list and corrupted the list.
3686 Frequency : all core dumps
3688 Description: Broke core dumps to lustre
3689 Details : Negative dentry may be unhashed if parent does not have UPDATE
3690 lock, but some callers, e.g. do_coredump, expect dentry to be
3691 hashed after successful create, hash it in ll_create_it.
3693 ------------------------------------------------------------------------------
3695 2006-09-13 Cluster File Systems, Inc. <info@clusterfs.com>
3697 * Support for kernels:
3698 2.6.9-42.0.2.EL (RHEL 4)
3699 2.6.5-7.276 (SLES 9)
3700 2.4.21-40.EL (RHEL 3)
3701 2.6.12.6 vanilla (kernel.org)
3705 Frequency : always on RHEL 3
3707 Description: Number of open files grows over time
3708 Details : The number of open files grows over time, whether or not
3709 Lustre is started. This was due to a filp leak introduced
3710 by one of our kernel patches.
3712 ------------------------------------------------------------------------------
3714 2006-08-20 Cluster File Systems, Inc. <info@clusterfs.com>
3716 * Support for kernels:
3717 2.6.9-42.EL (RHEL 4)
3718 2.6.5-7.267 (SLES 9)
3719 2.4.21-40.EL (RHEL 3)
3720 2.6.12.6 vanilla (kernel.org)
3725 Bugzilla : 5719, 9635, 9792, 9684
3726 Description: OST (or MDS) trips assertions in (re)connection under heavy load
3727 Details : If a server is under heavy load and cannot reply to new
3728 connection requests before the client resends the (re)connect,
3729 the connection handling code can behave badly if two service
3730 threads are concurrently handing separate (re)connections from
3731 the same client. Add better locking to the connection handling
3732 code, and ensure that only a single connection will be processed
3733 for a given client UUID, even if the lock is dropped.
3735 Severity : enhancement
3737 Description: add TCP zero-copy support to kernel
3738 Details : Add support to the kernel TCP stack to allow zero-copy bulk
3739 sends if the hardware supports scatter-gather and checksumming.
3740 This allows socklnd to do client-write and server-read more
3741 efficiently and reduce CPU utilization from skbuf copying.
3744 Frequency : only if NFS exporting from client
3746 Description: NULL pointer deref in ll_iocontrol() if chattr mknod file
3747 Details : If setting attributes on a file created under NFS that had
3748 never been opened it would be possible to oops the client
3749 if the file had no objects.
3753 Bugzilla : 9326, 10402, 10897
3754 Description: client crash in ptlrpcd_wake() thread when sending async RPC
3755 Details : It is possible that ptlrpcd_wake() dereferences a freed async
3756 RPC. In rare cases the ptlrpcd thread alread processed the RPC
3757 before ptlrpcd_wake() was called and the request was freed.
3760 Frequency : always for liblustre
3762 Description: liblustre client does MDS+OSTs setattr RPC for each write
3763 Details : When doing a write from a liblustre client, the client
3764 incorrectly issued an RPC to the MDS and each OST the file was
3765 striped over in order to update the timestamps. When writing
3766 with small chunks and many clients this could overwhelm the MDS
3767 with RPCs. In all cases it would slow down the write because
3768 these RPCs are unnecessary.
3770 Severity : enhancement
3772 Description: allow number of MDS service threads to be changed at module load
3773 Details : It is now possible to change the number of MDS service threads
3774 running. Adding "options mds mds_num_threads={N}" to the MDS's
3775 /etc/modprobe.conf will set the number of threads for the next
3776 time Lustre is restarted (assuming the "mds" module is also
3777 reloaded at that time). The default number of threads will
3778 stay the same, 32 for most systems.
3783 Description: OST crash if filesystem is unformatted or corrupt
3784 Details : If an OST is started on a device that has never been formatted
3785 or if the filesystem is corrupt and cannot even mount then the
3786 error handling cleanup routines would dereference a NULL pointer.
3791 Description: NULL pointer deref in llap_from_page.
3792 Details : get_cache_page_nowait can return a page with NULL (or otherwise
3793 incorrect) mapping if the page was truncated/reclaimed while it was
3794 searched for. Check for this condition and skip such pages when
3795 doing readahead. Introduce extra check to llap_from_page() to
3796 verify page->mapping->host is non-NULL (so page is not anonymous).
3799 Frequency : Sometimes when using sys_sendfile
3801 Description: "page not covered by a lock" warnings from ll_readpage
3802 Details : sendfile called ll_readpage without right page locks present.
3803 Now we introduced ll_file_sendfile that does necessary locking
3804 around call to generic_file_sendfile() much like we do in
3808 Frequency : with certain MDS communication failures at client mount time
3810 Description: NULL pointer deref after failed client mount
3811 Details : a client connection request may delayed by the network layer
3812 and not be sent until after the PTLRPC layer has timed out the
3813 request. If the client fails the mount immediately it will try
3814 to clean up before the network times out the request. Add a
3815 reference from the request import to the obd device and delay
3816 the cleanup until the network drops the request.
3819 Frequency : occasionally during client (re)connect
3821 Description: assertion failure during client (re)connect
3822 Details : processing a client connection request may be delayed by the
3823 client or server longer than the client connect timeout. This
3824 causes the client to resend the connection request. If the
3825 original connection request is replied in this interval, the
3826 client may trip an assertion failure in ptlrpc_connect_interpret()
3827 which thought it would be the only running connect process.
3830 Frequency : only with obd_echo servers and clients that are rebooted
3832 Description: kernel BUG accessing uninitialized data structure
3833 Details : When running an obd_echo server it did not start the ping_evictor
3834 thread, and when a client was evicted an uninitialized data
3835 structure was accessed. Start the ping_evictor in the RPC
3836 service startup instead of the OBD startup.
3838 Severity : enhancement
3839 Bugzilla : 10193 (patchless)
3840 Description: Remove dependency on various unexported kernel interfaces.
3841 Details : No longer need reparent_to_init, exit_mm, exit_files,
3842 sock_getsockopt, filemap_populate, FMODE_EXEC, put_filp.
3845 Frequency : rare (only users of deprecated and unsupported LDAP config)
3847 Description: write_conf for zeroconf mount queried LDAP incorrectly for client
3848 Details : LDAP apparently contains 'lustreName' attributes instead of
3849 'name'. A simple remapping of the name is sufficient.
3852 Frequency : rare (only with non-default dump_on_timeout debug enabled)
3854 Description: waiting_locks_callback trips kernel BUG if client is evicted
3855 Details : Running with the dump_on_timeout debug flag turned on makes
3856 it possible that the waiting_locks_callback() can try to dump
3857 the Lustre kernel debug logs from an interrupt handler. Defer
3858 this log dumping to the expired_lock_main() thread.
3860 Severity : enhancement
3862 Description: Support NFS exporting on 2.6 kernels.
3863 Details : Implement non-rawops metadata methods for NFS server to use without
3864 changing NFS server code.
3867 Frequency : very rare (synthetic metadata workload only)
3869 Description: two racing renames might cause an MDS thread to deadlock
3870 Details : Running the "racer" program may cause one MDS thread to rename
3871 a file from being the source of a rename to being the target of
3872 a rename at exactly the same time that another thread is doing
3873 so, and the second thread has already enqueued these locks after
3874 doing a lookup of the target and is trying to relock them in
3875 order. Ensure that we don't try to re-lock the same resource.
3878 Frequency : only very large systems with liblustre clients
3880 Description: slow eviction of liblustre clients with the "evict_by_nid" RPC
3881 Details : Use asynchronous set_info RPCs to send the "evict_by_nid" to
3882 all OSTs in parallel. This allows the eviction of stale liblustre
3883 clients to proceed much faster than if they were done in series,
3884 and also offers similar improvements for other set_info RPCs.
3889 Description: excessive CPU usage during initial read phase on client
3890 Details : During the initial read phase on a client, it would agressively
3891 retry readahead on the file, consuming too much CPU and impacting
3892 performance (since 1.4.5.8). Improve the readahead algorithm
3893 to avoid this, and also improve some other common cases (read
3894 of small files in particular, where "small" is files smaller than
3895 /proc/fs/lustre/llite/*/max_read_ahead_whole_mb, 2MB by default).
3900 Description: MDS crash when receiving packet with unknown intent.
3901 Details : Do not LBUG in unknown intent case, just return -EFAULT
3903 Severity : enhancement
3904 Bugzilla : 9293, 9385
3905 Description: MDS RPCs are serialised on client. This is unnecessary for some.
3906 Details : Do not serialize getattr (non-intent version) and statfs.
3909 Frequency : occasional, when OST network is overloaded/intermittent
3911 Description: client evicted by OST after bulk IO timeout
3912 Details : If a client sends a bulk IO request (read or write) the OST
3913 may evict the client if it is unresposive to its data GET/PUT
3914 request. This is incorrect if the network is overloaded (takes
3915 too long to transfer the RPC data) or dropped the OST GET/PUT
3916 request. There is no need to evict the client at all, since
3917 the pinger and/or lock callbacks will handle this, and the
3918 client can restart the bulk request.
3921 Frequency : Always when mmapping file with no objects
3923 Description: client crashes when mmapping file with no objects
3924 Details : Check that we actually have objects in a file before doing any
3925 operations on objects in ll_vm_open, ll_vm_close and
3931 Description: Request leak when working with deleted CWD
3932 Details : Introduce advanced request refcount tracking for requests
3933 referenced from lustre intent.
3935 Severity : Enhancement
3937 Description: Cache open file handles on client.
3938 Details : MDS now will return special lock along with openhandle, if
3939 requested and client is allowed to hold openhandle, even if unused,
3940 until such a lock is revoked. Helps NFS a lot, since NFS is opening
3941 closing files for every read/write openration.
3943 Severity : Enhancement
3945 Description: Cache open negative dentries on client when possible.
3946 Details : Guard negative dentries with UPDATE lock on parent dir, drop
3947 negative dentries on lock revocation.
3952 Description: Remounting a client read-only wasn't possible with a zconf mount
3953 Details : It wasn't possible to remount a client read-only with llmount.
3955 Severity : enhancement
3956 Description: Include MPICH 1.2.6 Lustre ADIO interface patch
3957 Details : In lustre/contrib/ or /usr/share/lustre in RPM a patch for
3958 MPICH is included to add Lustre-specific ADIO interfaces.
3959 This is based closely on the UFS ADIO layer and only differs
3960 in file creation, in order to allow the OST striping to be set.
3961 This is user-contributed code and not supported by CFS.
3966 Description: extended inode attributes (immutable, append-only) work improperly
3967 when 2.4 and 2.6 kernels are used on client/server or vice versa
3968 Details : Introduce kernel-independent values for these flags.
3970 Severity : enhancement
3973 Description: Allow fractional MB tunings for lustre in /proc/ filesystem.
3974 Details : Many of the /proc/ tunables can only be tuned at a megabyte
3975 granularity. Now, Fractional MB granularity is be supported,
3976 this is very useful for low memory system.
3978 Severity : enhancement
3980 Description: Getattr by fid
3981 Details : Getting a file attributes by its fid, obtaining UPDATE|LOOKUP
3982 locks, avoids extra getattr rpc requests to MDS, allows '/' to
3983 have locks and avoids getattr rpc requests for it on every stat.
3986 Frequency : Always, for filesystems larger than 2TB
3988 Description: ldiskfs crash at mount for filesystem larger than 2TB with mballoc
3989 Details : Kenrel kmalloc limits allocations to 128kB and this prevents
3990 filesystems larger than 2TB to be mounted with mballoc enabled.
3993 Frequency : Always, for 32-bit kernel without CONFIG_LBD and filesystem > 2TB
3995 Description: filesystem corruption for non-standard kernels and very large OSTs
3996 Details : If a 32-bit kernel is compiled without CONFIG_LBD enabled and a
3997 filesystems larger than 2TB is mounted then the kernel will
3998 silently corrupt the start of the filesystem. CONFIG_LBD is
3999 enabled for all CFS-supported kernels, but the possibility of
4000 this happening with a modified kernel config exists.
4002 Severity : enhancement
4004 Description: add client O_DIRECT support for 2.6 kernels
4005 Details : It is now possible to do O_DIRECT reads and writes to files
4006 in the Lustre client mountpoint on 2.6 kernel clients.
4008 Severity : enhancement
4010 Description: parallel glimpse, setattr, statfs, punch, destroy requests
4011 Details : Sends glimpse, setattr, statfs, punch, destroy requests to OSTs in
4012 parallel, not waiting for response from every OST before sending
4013 a rpc to the next OST.
4018 Description: setattr vs write race when updating file timestamps
4019 Details : Client processes that update a file timestamp into the past
4020 right after writing to the file (e.g. tar) it is possible that
4021 the updated file modification time can be reset to the current
4022 time due to a race between processing the setattr and write RPC.
4024 Severity : enhancement
4026 Description: Bring 'lfs find' closer in line with regular Linux find.
4027 Details : lfs find util supports -atime, -mtime, -ctime, -maxdepth, -print,
4028 -print0 options and obtains all the needed info through the lustre
4031 Severity : enhancement
4033 Description: support up to 1024 configured devices on one node
4034 Details : change obd_dev array from statically allocated to dynamically
4035 allocated structs as they are first used to reduce memory usage
4040 Description: Flush dirty partially truncated pages during truncate
4041 Details : Immediatelly flush partially truncated pages in filter_setattr,
4042 this way we completely avoid having any pages in page cache on OST
4043 and can retire ugly workarounds during writes to flush such pages.
4048 Description: i_sem vs transaction deadlock in mds_obd_destroy during unlink.
4049 Details : protect inode from truncation within vfs_unlink() context
4050 just take a reference before calling vfs_unlink() and release it
4051 when parent's i_sem is free.
4054 Frequency : always, if extents are used on OSTs
4056 Description: index ei_leaf_hi (48-bit extension) is not zeroed in extent index
4057 Details : OSTs using the extents format would not zero the high 16 bits of
4058 the index physical block number. This is not a problem for any
4059 OST filesystems smaller than 16TB, and no kernels support ext3
4060 filesystems larger than 16TB yet. This is fixed in 1.4.7 (all
4061 new/modified files) and can be fixed for existing filesystems
4062 with e2fsprogs-1.39-cfs1.
4067 Description: import connection selection may be incorrect if timer wraps
4068 Details : Using a 32-bit jiffies timer with HZ=1000 may cause backup
4069 import connections to be ignored if the 32-bit jiffies counter
4070 wraps. Use a 64-bit jiffies counter.
4073 Frequency : during server recovery
4075 Description: crash after server is denying duplicate export
4076 Details : If clients are resending connect requests to the server, the
4077 server refuses to allow a client to connect multiple times.
4078 Fixed a bug in the handling of this case.
4081 Frequency : very large clusters immediately after boot
4083 Description: LNET request buffers exhausted under heavy short-term load
4084 Details : If a large number of client requests are generated on a service
4085 that has previously never seen so many requests it is possible
4086 that the request buffer growth cannot keep up with the spike in
4087 demand. Instead of dropping incoming requests, they are held in
4088 the LND until the RPC service can accept more requests.
4091 Frequency : Sometimes during replay
4093 Description: Assertion failure in ll_local_open after replay.
4094 Details : If replay happened on an open request reply before we were able
4095 to set replay handler, reply will become not swabbed tripping the
4096 assertion in ll_local_open. Now we set the handler right after
4097 recognising of open request
4100 Frequency : very rare
4102 Description: kernel reports "badness in vsnprintf"
4103 Details : Reading from the "recovery_status" /proc file in small chunks
4104 may cause a negative length in lprocfs_obd_rd_recovery_status()
4105 call to vsnprintf() (which is otherwise harmless). Exit early
4106 if there is no more space in the output buffer.
4108 Severity : enhancement
4110 Description: clear OBD RPC statistics by writing to them
4111 Details : It is now possible to clear the OBD RPC statistics by writing
4112 to the "stats" file.
4117 Description: Client mtime is not the same on different clients after utimes
4118 Details : In some cases, the client was using the utimes() syscall on
4119 a file cached on another node. The clients now validate the
4120 ctime from the MDS + OSTs to determine which one is right.
4125 Description: Inability to activate failout mode
4126 Details : lconf script incorrectly assumed that in python string's numeric
4127 value is used in comparisons.
4130 Frequency : always with multiple stripes per file
4132 Description: Inefficient object allocation for mutli-stripe files
4133 Details : When selecting which OSTs to stripe files over, for files with
4134 a stripe count that divides evenly into the number of OSTs,
4135 the MDS is always picking the same starting OST for each file.
4136 Return the OST selection heuristic to the original design.
4141 Description: mount failures may take full timeout to return an error
4142 Details : Under some heavy load conditions it is possible that a
4143 failed mount can wait for the full obd_timeout interval,
4144 possibly several minutes, before reporting an error.
4145 Instead return an error as soon as the status is known.
4147 ------------------------------------------------------------------------------
4149 2006-02-14 Cluster File Systems, Inc. <info@clusterfs.com>
4151 * WIRE PROTOCOL CHANGE. This version of Lustre networking WILL NOT
4152 INTEROPERATE with older versions automatically. Please read the
4153 user documentation before upgrading any part of a live system.
4154 * WARNING: Lustre networking configuration changes are required with
4155 this release. See https://bugzilla.clusterfs.com/show_bug.cgi?id=10052
4158 * Support for kernels:
4159 2.6.9-22.0.2.EL (RHEL 4)
4160 2.6.5-7.244 (SLES 9)
4161 2.6.12.6 vanilla (kernel.org)
4164 Severity : enhancement
4165 Bugzilla : 7981/8208
4166 Description: Introduced Lustre Networking (LNET)
4167 Details : LNET is new networking infrastructure for Lustre, it includes
4168 a reorganized network configuration mode (see the user
4169 documentation for full details) as well as support for routing
4170 between different network fabrics. Lustre Networking Devices
4171 (LNDs) for the supported network fabrics have also been
4172 created for this new infrastructure.
4174 Severity : enhancement
4175 Description: Introduced Access control lists
4176 Details : clients can set ACLs on files and directories in order to have
4177 more fine-grained permissions than the standard Unix UGO+RWX.
4178 The MDS must be started with the "-o acl" mount option.
4180 Severity : enhancement
4181 Description: Introduced filesystem quotas
4182 Details : Administrators may now establish per-user quotas on the
4185 Severity : enhancement
4187 Description: Configuration change for the XT3
4188 The PTLLND is now used to run Lustre over Portals on the XT3
4189 The configure option(s) --with-cray-portals are no longer used.
4190 Rather --with-portals=<path-to-portals-includes> is used to
4191 enable building on the XT3. In addition to enable XT3 specific
4192 features the option --enable-cray-xt3 must be used.
4197 Description: Running on many-way SMP OSTs can trigger oops in llcd_send()
4198 Details : A race between allocating a new llcd and re-getting the llcd_lock
4199 allowed another thread to grab newly-allocated llcd.
4201 Severity : enhancement
4203 Description: 2.6 OST async journal commit and locking fix to improve performance
4204 Details : The filter_direct_io()+filter_commitrw_write() journal commits for
4205 2.6 kernels are now async as they already were in 2.4 kernels so
4206 that they can commit concurrently with the network bulk transfer.
4207 For block-allocated files the filter allocation semaphore is held
4208 to avoid filesystem fragmentation during allocation. BKL lock
4209 removed for 2.6 xattr operations where it is no longer needed.
4214 Description: lconf incorrectly determined whether two IP networks could talk
4215 Details : In some more complicated routing and multiple-network
4216 configurations, lconf will avoid trying to make a network
4217 connection to a disjoint part of the IP space. It was doing the
4218 math incorrectly for one set of cases.
4223 Description: Fix for potential infinite loop processing records in an llog.
4224 Details : If an llog record is corrupted/zeroed, it is possible to loop
4225 forever in llog_process(). Validate the llog record length
4226 and skip the remainder of the block on error.
4229 Frequency : occasional (liblustre only)
4231 Description: liblustre could not open files whose last component is a symlink
4232 Details : sysio_path_walk() would incorrectly pass the open intent to
4233 intermediate path components.
4236 Frequency : rare (liblustre only with non-standard tuning)
4237 Bugzilla : 7201 (7350)
4238 Description: Tuning the MDC DLM LRU size to zero triggers client LASSERT
4239 Details : llu_lookup_finish_locks() tries to set lock data on a lock
4240 after it has been released, only do this for referenced locks
4242 Severity : enhancement
4244 Description: specifying an (invalid) directory default stripe_size of -1
4245 would reset the directory default striping
4246 Details : stripe_size -1 was used internally to signal directory stripe
4247 removal, now use "all default" to signal dir stripe removal
4248 as a directory striping of "all default" is not useful
4251 Frequency : common for large clusters running liblustre clients
4253 Description: doing an ls when liblustre clients are running is slow
4254 Details : sending a glimpse AST to a liblustre client waits for every AST
4255 to time out, as liblustre clients will not respond. Since they
4256 cannot cache data we refresh the OST lock LVB from disk instead.
4258 Severity : enhancement
4260 Description: doing an ls at the same time as file IO can be slow
4261 Details : enqueue and other "small" requests can be blocked behind many
4262 large IO requests. Create a new OST IO portal for non-IO
4263 requests so they can be processed faster.
4266 Frequency : rare (only HPUX clients mounting unsupported re-exported NFS vol)
4268 Description: an HPUX NFS client would get -EACCESS when ftruncate()ing a newly
4269 created file with mode 000
4270 Details : the Linux NFS server relies on an MDS_OPEN_OWNEROVERRIDE hack to
4271 allow an ftruncate() as a non-root user to a file with mode 000.
4272 Lustre now respects this flag to disable mode checks when
4273 truncating a file owned by the user
4276 Frequency : liblustre-only, when liblustre client dies unexpectedly or becomes
4279 Description: Revoking locks from clients that went dead or catatonic might take
4281 Details : New lock flags FL_CANCEL_ON_BLOCK used by liblustre makes
4282 cancellation of such locks instant on servers without waiting for
4283 any reply from clients. Clients drops these locks when cancel
4284 notification from server is received without replying.
4287 Frequency : liblustre-only, when liblustre client dies or becomes busy
4289 Description: Doing ls on Linux clients can take a long time with active
4291 Details : Liblustre client cannot handle ASTs in timely manner, so avoid
4292 granting such locks to it in the first place if possible. Locks
4293 are taken by proxy on the OST during the read or write and
4294 dropped immediately afterward. Add connect flags handling, do
4295 not grant locks to liblustre clients for glimpse ASTs.
4297 Severity : enhancement
4299 Description: Improve read-ahead algorithm to avoid excessive IO for random reads
4300 Details : Existing read-ahead algorithm is tuned for the case of streamlined
4301 sequential reads and behaves badly with applications doing random
4302 reads. Improve it by reading ahead at least read region, and
4303 avoiding excessive large RPC for small reads.
4305 Severity : enhancement
4307 Description: Creating more than 1000 files for a single job may cause a load
4308 imbalance on the OSTs if there are also a large number of OSTs.
4309 Details : qos_prep_create() uses an OST index reseed value that is an
4310 even multiple of the number of available OSTs so that if the
4311 reseed happens in the middle of the object allocation it will
4312 still utilize the OSTs as uniformly as possible.
4317 Description: OST or MDS may oops in ping_evictor_main()
4318 Details : ping_evictor_main() drops obd_dev_lock if deleting a stale export
4319 but doesn't restart at beginning of obd_exports_timed list
4322 Severity : enhancement
4324 Description: improve by-nid export eviction on the MDS and OST
4325 Details : allow multiple exports with the same NID to be evicted at one
4326 time without re-searching the exports list.
4329 Frequency : rare, only with supplementary groups enabled on SMP 2.6 kernels
4331 Description: MDS may oops in groups_free()
4332 Details : in rare race conditions a newly allocated group_info struct is
4333 freed again, and this can be NULL. The 2.4 compatibility code
4334 for groups_free() checked for a NULL pointer, but 2.6 did not.
4337 Frequency : common for liblustre clients doing little filesystem IO
4338 Bugzilla : 9352, 7313
4339 Description: server may evict liblustre clients accessing contended locks
4340 Details : if a client is granted a lock or receives a completion AST
4341 with a blocking AST already set it would not reply to the AST
4342 for LDLM_FL_CANCEL_ON_BLOCK locks. It now replies to such ASTs.
4345 Frequency : lfs setstripe, only systems with more than 160 OSTs
4347 Description: unable to set striping with a starting offset beyond OST 160
4348 Details : llapi_create_file() incorrectly limited the starting stripe
4349 index to the maximum single-file stripe count.
4352 Frequency : LDAP users only
4354 Description: lconf did not handle in-kernel recovery with LDAP properly
4355 Details : lconf/LustreDB get_refs() is searching the wrong namespace
4357 Severity : enhancement
4359 Description: bind OST threads to NUMA nodes to improve performance
4360 Details : all OST threads are uniformly bound to CPUs on a single NUMA
4361 node and do their allocations there to localize memory access
4363 Severity : enhancement
4365 Description: llmount can determine client NID directly from Myrinet (GM)
4366 Details : the client NID code from gmnalnid was moved directly into
4367 llmount, removing the need to use this or specifying the
4368 client NID explicitly when mounting GM clients with zeroconf
4371 Frequency : if client is started with down MDS
4373 Description: if client is started with down MDS mount hangs in ptlrpc_queue_wait
4374 Details : Having an LWI_INTR() wait event (interruptible, but no timeout)
4375 will wait indefinitely in ptlrpc_queue_wait->l_wait_event() after
4376 ptlrpc_import_delayed_req() because we didn't check if the
4377 request was interrupted, and we also didn't break out of the
4378 event loop if there was no timeout
4383 Description: data loss during non-page-aligned writes to a single file from
4384 both multiple nodes and multiple threads on one node at same time
4385 Details : updates to KMS and lsm weren't protected by common lock. Resulting
4386 inconsistency led to false short-reads, that were cached and later
4387 used by ->prepare_write() to fill in partially written page,
4388 leading to data loss.
4391 Frequency : always, if lconf --abort_recovery used
4393 Description: lconf --abort_recovery fails with 'Operation not supported'
4394 Details : lconf was attempting to abort recovery on the MDT device and not
4397 Severity : enhancement
4399 Description: remove cleanup logs
4400 Details : replace lconf-generated cleanup logs with lustre internal
4401 cleanup routines. Eliminates the need for client-cleanup and
4404 Severity : enhancement
4406 Description: add support for EAs (user and system) on lustre filesystems
4407 Details : it is now possible to store extended attributes in the Lustre
4408 client filesystem, and with the user_xattr mount option it
4409 is possible to allow users to store EAs on their files also
4411 Severity : enhancement
4413 Description: Add possibility (config option) to show minimal available OST free
4415 Details : When compiled with --enable-mindf configure option, statfs(2)
4416 (and so, df) will return least minimal free space available from
4417 all OSTs as amount of free space on FS, instead of summary of
4418 free spaces of all OSTs.
4420 Severity : enhancement
4422 Description: do not expand extent locks acquired on OST-side
4423 Details : Modify ldlm_extent_policy() to not expand local locks, acquired
4424 by server: they are not cached anyway.
4427 Frequency : when mmap is used/binaries executed from Lustre
4429 Description: Unmmap pages before throwing them away from read cache.
4430 Details : llap_shrink cache now attempts to unmap pages before discarding
4431 them (if unmapping failed - do not discard). SLES9 kernel has
4432 extra checks that trigger if this unmapping is not done first.
4437 Description: lconf didn't resolve symlinks before checking to see whether a
4438 given mountpoint was already in use
4441 Frequency : when migrating failover services
4442 Bugzilla : 6395, 9514
4443 Description: When migrating a subset of services from a node (e.g. failback
4444 from a failover service node) the remaining services would
4445 time out and evict clients.
4446 Details : lconf --force (implied by --failover) sets the global obd_timeout
4447 to 5 seconds in order to quickly disconnect, but this caused
4448 other RPCs to time out too quickly. Do not change the global
4449 obd_timeout for force cleanup, only set it for DISCONNECT RPCs.
4451 Severity : enhancement
4452 Frequency : if MDS is started with down OST
4453 Bugzilla : 9439,5706
4454 Description: Allow startup/shutdown of an MDS without depending on the
4455 availability of the OSTs.
4456 Details : Asynchronously call mds_lov_synchronize during MDS startup.
4457 Add appropriate locking and lov-osc refcounts for safe
4458 cleaning. Add osc abort_inflight calls in case the
4459 synchronize never started.
4462 Frequency : occasional (Cray XT3 only)
4464 Description: root not authorized to access files in CRAY_PORTALS environment
4465 Details : The client process capabilities were not honoured on the MDS in
4466 a CRAY_PORTALS/CRAY_XT3 environment. If the file had previously
4467 been accessed by an authorized user then root was able to access
4468 the file on the local client also. The root user capabilities
4469 are now allowed on the MDS, as this environment has secure UID.
4472 Frequency : occasional
4474 Description: ldiskfs "too long searching" message happens too often
4475 Details : A debugging message (otherwise harmless) prints too often on
4476 the OST console. This has been reduced to only happen when
4477 there are fragmentation problems on the filesystem.
4482 Description: Division by zero in statfs when all OSCs are inactive
4483 Details : lov_get_stripecnt() returns zero due to incorrect order of checks,
4484 lov_statfs divides by value returned by lov_get_stripecnt().
4488 Bugzilla : 9489, 3273
4489 Description: First write from each client to each OST was only 4kB in size,
4490 to initialize client writeback cache, which caused sub-optimal
4491 RPCs and poor layout on disk for the first writen file.
4492 Details : Clients now request an initial cache grant at (re)connect time
4493 and so that they can start streaming writes to the cache right
4494 away and always do full-sized RPCs if there is enough data.
4495 If the OST is rebooted the client also re-establishes its grant
4496 so that client cached writes will be honoured under the grant.
4501 Description: Slow ls (and stat(2) syscall) on files residing on IO-loaded OSTs
4502 Details : Now I/O RPCs go to different portal number and (presumably) fast
4503 lock requests (and glimses) and other RPCs get their own service
4504 threads pool that should be able to service those RPCs
4507 Severity : enhancement
4509 Description: Ability to exchange lustre version between client and servers and
4510 issue warnings at client side if client is too old. Also for
4511 liblustre clients there is ability to refuse connection of too old
4513 Details : New 'version' field is added to connect data structure that is
4514 filled with version info. That info is later checked by server and
4518 Frequency : rare, liblustre only.
4519 Bugzilla : 9296, 9581
4520 Description: Two simultaneous writes from liblustre at offset within same page
4521 might proceed at the same time overwriting eachother with stale
4523 Details : I/O lock withing llu_file_prwv was released too early, before data
4524 actually was hitting the wire. Extended lock-holding time until
4525 server acknowledges receiving data.
4528 Frequency : extremely rare. Never observed in practice.
4530 Description: avoid generating lustre_handle cookie of 0.
4531 Details : class_handle_hash() generates handle cookies by incrementing
4532 global counter, and can hit 0 occasionaly (this is unlikely, but
4533 not impossible, because initial value of cookie counter is
4534 selected randonly). Value of 0 is used as a sentinel meaning
4535 "unassigned handle" --- avoid it. Also coalesce two critical
4536 sections in this function into one.
4538 Severity : enhancement
4540 Description: allow liblustre clients to delegate truncate locking to OST
4541 Details : To avoid overhead of locking, liblustre client instructs OST to
4542 take extent lock in ost_punch() on client's behalf. New connection
4543 flag is added to handle backward compatibility.
4545 Severity : enhancement
4546 Bugzilla : 4928, 7341, 9758
4547 Description: allow number of OST service threads to be specified
4548 Details : a module parameter allows the number of OST service threads
4549 to be specified via "options ost ost_num_threads={N}" in the
4550 OSS's /etc/modules.conf or /etc/modprobe.conf.
4554 Bugzilla : 6146, 9635, 9895
4555 Description: servers crash with bad pointer in target_handle_connect()
4556 Details : In rare cases when a client is reconnecting it was possible that
4557 the connection request was the last reference for that export.
4558 We would temporarily drop the export reference and get a new
4559 one, but this may have been the last reference and the export
4560 was just destroyed. Get new reference before dropping old one.
4562 Severity : enhancement
4563 Frequency : if client is started with failover MDS
4565 Description: Allow multiple MDS hostnames in the mount command
4566 Details : Try to read the configuration from all specified MDS
4567 hostnames during a client mount in case the "primary"
4570 Severity : enhancement
4572 Description: Stop sending data to evicted clients as soon as possible.
4573 Details : Check if the client we are about to send or are sending data to
4574 was evicted already. (Check is done every second of waiting,
4575 for which l_wait_event interface was extended to allow checking
4576 of exit condition at specified intervals).
4579 Frequency : rare, normally only when NFS exporting is done from client
4581 Description: 'bad disk LOV MAGIC: 0x00000000' error when chown'ing files
4583 Details : Make mds_get_md() recognise empty md case and set lmm size to 0.
4586 Frequency : always, if srand() is called before liblustre initialization
4588 Description: Liblustre uses system PRNG disturbing its usage by user application
4589 Details : Introduce internal to lustre fast and high-quality PRNG for
4590 lustre usage and make liblustre and some other places in generic
4591 lustre code to use it.
4593 Severity : enhancement
4594 Bugzilla : 9477, 9557, 9870
4595 Description: Verify that the MDS configuration logs are updated when xml is
4596 Details : Check if the .xml configuration logs are newer than the config
4597 logs stored on the MDS and report an error if this is the case.
4598 Request --write-conf, or allow starting with --old_conf.
4600 Severity : enhancement
4602 Description: Handle symlinks in the path when checking if Lustre is mounted.
4603 Details : Resolve intermediate symlinks when checking if a client has
4604 mounted a filesystem to avoid duplicate client mounts.
4609 Description: lconf can hit an error exception but still return success.
4610 Details : The lconf command catches the Command error exception at the top
4611 level script context and will exit with the associated exit
4612 status, but doesn't ensure that this exit status is non-zero.
4617 Description: failure of ptlrpc thread startup can cause oops
4618 Details : Starting a ptlrpc service thread can fail if there are a large
4619 number of threads or the server memory is very fragmented.
4620 Handle this without oopsing.
4623 Frequency : always, only if liblustre and non-default acceptor port was used
4625 Description: liblustre cannot connect to servers with non-default acceptor port
4626 Details : tcpnal_set_default_params() was not called and was therefore
4627 ignoring the environment varaible TCPNAL_PORT, as well as other
4628 TCPNAL_ environment variables
4633 Description: two objects could be created on the same OST for a single file
4634 Details : If an OST is down, in some cases it was possible to create two
4635 objects on a single OST for a single file. No problems other
4636 than potential performance impact and spurious error messages.
4640 Bugzilla : 5681, 9562
4641 Description: Client may oops in ll_unhash_aliases
4642 Details : Client dcache may become inconsistent in race condition.
4643 In some cases "getcwd" can fail if the current directory is
4649 Description: Inode refcounting problems in NFS export code
4650 Details : link_raw functions used to call d_instantiate without obtaining
4651 extra inode reference first.
4655 Bugzilla : 9942, 9903
4656 Description: Referencing freed requests leading to crash, memleaks with NFS.
4657 Details : We used to require that call to ll_revalidate_it was always
4658 followed by ll_lookup_it. Also with revalidate_special() it is
4659 possible to call ll_revalidate_it() twice for the same dentry
4660 even if first occurence returned success. This fix changes semantic
4661 between DISP_ENQ_COMPLETE disposition flag to mean there is extra
4662 reference on a request referred from the intent.
4663 ll_intent_release() then releases such a request.
4666 Frequency : rare, normally benchmark loads only
4668 Description: unlinked inodes were kept in memory on the client
4669 Details : If a client is repeatedly creating and unlinking files it
4670 can accumulate a lot of stale inodes in the inode slab cache.
4671 If there is no other client load running this can cause the
4672 client node to run out of memory. Instead flush old inodes
4673 from client cache that have the same inode number as a new inode.
4676 Frequency : SLES9 2.6.5 kernel and long filenames only
4677 Bugzilla : 9969, 10379
4678 Description: utime reports stale NFS file handle
4679 Details : SLES9 uses out-of-dentry names in some cases, which confused
4680 the lustre dentry revalidation. Change it to always use the
4684 Frequency : rare, unless heavy write-truncate concurrency is continuous
4685 Bugzilla : 4180, 6984, 7171, 9963, 9331
4686 Description: OST becomes very slow and/or deadlocked during object unlink
4687 Details : filter_destroy() was holding onto the parent directory lock
4688 while truncating+unlinking objects. For very large objects this
4689 may block other threads for a long time and slow overall OST
4690 responsiveness. It may also be possible to get a lock ordering
4691 deadlock in this case, or run out of journal credits because of
4692 the combined truncate+unlink. Solution is to do object truncate
4693 first in one transaction without parent lock, and then do the
4694 final unlink in a new transaction with the parent lock. This
4695 reduces the lock hold time dramatically.
4698 Frequency : rare, 2.4 kernels only
4700 Description: MDS or OST cleanup may trip kernel BUG when dropping kernel lock
4701 Details : mds_cleanup() and filter_cleanup() need to drop the kernel lock
4702 before unmounting their filesystem in order to avoid deadlock.
4703 The kernel_locked() function in 2.4 kernels only checks whether
4704 the kernel lock is held, not whether it is this process that is
4705 holding it as 2.6 kernels do.
4710 Description: MDS or OST may oops/LBUG if a client is connecting multiple times
4711 Details : The client ptlrpc code may be trying to reconnect to a down
4712 server before a previous connection attempt has timed out.
4713 Increase the reconnect interval to be longer than the connection
4714 timeout interval to avoid sending duplicate connections to
4718 Frequency : echo_client brw_test command
4720 Description: fix echo_client to work with OST preallocated code
4721 Details : OST preallocation code (5137) didn't take echo_client IO path
4722 into account: echo_client calls filter methods outside of any
4723 OST thread and, hence, there is no per-thread preallocated
4724 pages and buffers to use. Solution: hijack pga pages for IO. As
4725 a byproduct, this avoids unnecessary data copying.
4729 Bugzilla : 3555, 5962, 6025, 6155, 6296, 9574
4730 Description: Client can oops in mdc_commit_close() after open replay
4731 Details : It was possible for the MDS to return an open request with no
4732 transaction number in mds_finish_transno() if the client was
4733 evicted, but without actually returning an error. Clients
4734 would later try to replay that open and may trip an assertion
4735 Simplify the client close codepath, and always return an error
4736 from the MDS in case the open is not successful.
4739 Frequency : rare, 2.6 OSTs only
4741 Description: OST may deadlock under high load on fragmented files
4742 Details : If there was a heavy load and highly-fragmented OST filesystems
4743 it was possible to have all the OST threads deadlock waiting on
4744 allocation of biovecs, because the biovecs were not released
4745 until the entire RPC IO was completed. Instead, release biovecs
4746 as soon as they are complete to ensure forward IO progress.
4748 Severity : enhancement
4750 Description: Support for specifying external journal device at mount
4751 Details : If an OST or MDS device is formatted with an external journal
4752 device, this device major/minor is stored in the ext3 superblock
4753 and may not be valid for failover. Allow detecting and
4754 specifying the external journal at mount time.
4759 Description: Mounting an MDS with pending unlinked files may cause oops
4760 Details : target_finish_recovery() calls mds_postrecov() which returned
4761 the number of orphans unlinked. mds_lov_connect->mds_postsetup()
4762 considers this an error and immediately begins cleaning up the
4763 lov, just after starting the mds_lov process
4765 Severity : enhancement
4767 Description: Implement 'lfs df' to report actual free space on per-OST basis
4768 Details : Add sub-command 'df' on 'lfs' to report the disk space usage of
4769 MDS/OSDs. Usage: lfs df [-i][-h]. Command Options: '-i' to report
4770 usage of objects; '-h' to report in human readable format.
4772 ------------------------------------------------------------------------------
4774 2005-08-26 Cluster File Systems, Inc. <info@clusterfs.com>
4781 Description: Mounting an ldiskfs file system with mballoc may crash OST node.
4782 Details : ldiskfs mballoc code may reference an uninitialized buddy struct
4783 at startup during orphan unlinking. Instead, skip buddy update
4784 before setup, as it will be regenerated after recovery is complete.
4789 Description: If an OST is inactive, its locks might reference stale inodes.
4790 Details : lov_change_cbdata() must iterate over all namespaces, even if
4791 they are inactive to clear inode references from the lock.
4793 Severity : enhancement
4794 Frequency : occasional, if non-standard max_dirty_mb used
4796 Description: Client will block write RPCs if not enough grant
4797 Details : If a client has max_dirty_mb smaller than max_rpcs_in_flight,
4798 then the client will block writes while waiting for another RPC
4799 to complete instead of consuming its dirty limit. With change
4800 we get improved performance when max_dirty_mb is small.
4802 Severity : enhancement
4803 Bugzilla : 3389, 6253
4804 Description: Add support for supplementary groups on the MDS.
4805 Details : The MDS has an upcall /proc/fs/lustre/mds/{mds}/group_upcall
4806 (set to /usr/sbin/l_getgroups if enabled) which will do MDS-side
4807 lookups for user supplementary groups into a cache.
4811 Description: O_CREAT|O_EXCL open flags in liblustre always return -EEXIST
4812 Details : Make libsysio to not enforce O_EXCL by clearing the flag,
4813 for liblustre O_EXCL is enforced by MDS.
4817 Description: readdir never returns NULL in liblustre.
4818 Details : Corrected llu_iop_getdirentries logic, to return offset of next
4819 dentry in struct dirent.
4823 Frequency : liblustre only, depends on application IO pattern
4824 Description: liblustre clients evicted if not contacting servers
4825 Details : Don't put liblustre clients into the ping_evictor list, so
4826 they will not be evicted by the pinger ever.
4828 Severity : enhancement
4830 Description: Add ability to evict clients by NID from MDS.
4831 Details : By echoing "nid:$NID" string into
4832 /proc/fs/lustre/mds/.../evict_client client with nid that equals to
4833 $NID would be instantly evicted from this MDS and from all active
4834 OSTs connected to it.
4838 Description: Do not query file size twice, somewhat slowing stat(2) calls.
4839 Details : lookup_it_finish() used to query file size from OSTs that was not
4844 Description: service threads change working directory to that of init
4845 Details : Starting lustre service threads may pin the working directory
4846 of the parent thread, making that filesystem busy. Threads
4847 now change to the working directory of init to avoid this.
4851 Frequency : during shutdown only
4852 Description: shutdown with a failed MDS or OST can cause unmount to hang
4853 Details : Don't resend DISCONNECT messages in ptlrpc_disconnect_import()
4858 Frequency : 2.6 only
4859 Description: chmod/chown may include an extra supplementary group
4860 Details : ll{,u}_mdc_pack_op_data() does not properly initialize the
4861 supplementary group and if none is specified this is used.
4864 Bugzilla : 5479 (6816)
4866 Description: Racing open + rm can assert client in mdc_set_open_replay_data()
4867 Details : If lookup is in progress on a file that is unlinked we might try
4868 to revalidate the inode and fail in revalidate after lookup is
4869 complete and ll_file_open() enqueues the open again but
4870 it_open_error() was not checking DISP_OPEN_OPEN errors correctly.
4873 Frequency : always, if lconf --abort_recovery used
4875 Description: lconf --abort_recovery fails with 'Operation not supported'
4876 Details : lconf was attempting to abort recovery on the MDT device and not
4879 ------------------------------------------------------------------------------
4881 2005-08-08 Cluster File Systems, Inc. <info@clusterfs.com>
4886 Frequency : rare (only unsupported configurations with a node running as an
4888 Bugzilla : 6514, 5137
4889 Description: Mounting a Lustre file system on a node running as an OST could
4891 Details : OSTs now preallocates memory needed to write out data at
4892 startup, instead of when needed, to avoid having to
4893 allocate memory in possibly low memory situations.
4894 Specifically, if the file system is mounted on on OST,
4895 memory pressure could force it to try to write out data,
4896 which it needed to allocate memory to do. Due to the low
4897 memory, it would be unable to do so and the node would
4898 become unresponsive.
4900 Severity : enhancement
4902 Description: Addition of lconf --service command line option
4903 Details : lconf now accepts a '--service <arg>' option, which is
4904 shorthand for 'lconf --group <arg> --select <arg>=<hostname>'
4906 Severity : enhancement
4908 Description: Failover mode is now the default for OSTs.
4909 Details : By default, OSTs will now run in failover mode. To return to
4910 the old behaviour, add '--failout' to the lmc line for OSTs.
4912 Severity : enhancement
4914 Description: Health checks are now provided for MDS and OSTs
4915 Details : Additional detailed health check information on MSD and OSTs
4916 is now provided through the procfs health_check value.
4919 Frequency : occasional, depends on IO load
4921 Description: Disk fragmentation on the OSTs could eventually cause slowdowns
4922 after numerous create/delete cycles
4923 Details : The ext3 inode allocation policy would not allocate new inodes
4924 very well on the OSTs because there are no new directories
4925 being created. Instead we look for groups with free space if
4926 the parent directories are nearly full.
4931 Description: Network or server problems during mount may cause partially
4932 mounted clients instead of returning an error.
4933 Details : The config llog parsing code may overwrite the error return
4934 code during mount error handling, returning success instead
4940 Description: MDS can fail to allocate large reply buffers
4941 Details : After long uptimes the MDS can fail to allocate large reply
4942 buffers (e.g. zconf client mount config records) due to memory
4943 fragmentation or consumption by the buffer cache. Preallocate
4944 some large reply buffers so that these replies can be sent even
4945 under memory pressure.
4949 Frequency : rare (liblustre)
4950 Description: fsx running with liblustre complained that using truncate() to
4951 extend the file doesn't work. This patch corrects that issue.
4952 Details : This is the liblustre equivalent of the fix for bug 6196. Fixes
4953 ATTR_SIZE and lsm use in llu_setattr_raw.
4957 Frequency : rare, only 2.6 kernels
4958 Description: Unusual file access patterns on the MDS may result in inode
4959 data being lost in very rare circumstances.
4960 Details : Bad interaction between the ea-in-inode patch and the "no-read"
4961 code in the 2.6 kernel caused the inode and/or EA data not to
4962 be read from disk, causing single-file corruption.
4966 Frequency : rare, only 2.6 filesystems using extents
4967 Description: Heavy concurrent write and delete load may cause data corruption.
4968 Details : It was possible under high-load situations to have an extent
4969 metadata block in the block device cache from a just-unlinked
4970 file overwrite a newly-allocated data block. We now unmap any
4971 metadata buffers that alias just-allocated data blocks.
4975 Frequency : filesystems with default stripe_count larger than 77
4976 Description: lconf+mke2fs fail when formatting filesystem with > 77 stripes
4977 Details : lconf specifies an inode size of 4096 bytes when the default
4978 stripe_count is larger than 77. This conflicts with the default
4979 inode density of 1 per 4096 bytes. Allocate smaller inodes in
4980 this case to avoid pinning too much memory for large EAs.
4982 ------------------------------------------------------------------------------
4984 2005-07-07 Cluster File Systems, Inc. <info@clusterfs.com>
4989 Frequency : rare (extremely heavy IO load with hundreds of clients)
4991 Description: Client is evicted, gets IO error writing to file
4992 Details : lock ordering changes for bug 5492 reintroduced bug 3267 and
4993 caused clients to be evicted for AST timeouts. The fixes in
4994 bug 5192 mean we no longer need to have such short AST timeouts
4995 so ldlm_timeout has been increased.
4998 Frequency : occasional during --force or --failover shutdown under load
4999 Bugzilla : 5949, 4834
5000 Description: Server oops/LBUG if stopped with --force or --failover under load
5001 Details : a collection of import/export refcount and cleanup ordering
5002 issues fixed for safer force cleanup
5005 Frequency : only filesystems larger than 120 OSTs
5006 Bugzilla : 5990, 6223
5007 Description: lfs getstripe would oops on a very large filesystem
5008 Details : lov_getconfig used kfree on vmalloc'd memory
5011 Frequency : only filesystems exporting via NFS to Solaris 10 clients
5012 Bugzilla : 6242, 6243
5013 Description: reading from files that had been truncated to a non-zero size
5014 but never opened returned no data
5015 Details : ll_file_read() reads zeros from no-object files to EOF
5020 Description: A bug in MDS/OSS recovery could cause the OSS to fail an assertion
5021 Details : There's little harm in aborting MDS/OSS recovery and letting it
5022 try again, so I removed the LASSERT and return an error instead.
5024 Severity : enhancement
5026 Description: New debugging infrastructure for tracking down data corruption
5027 Details : The I/O checksum code was replaced to: (a) control it at runtime,
5028 (b) cover more of the client-side code path, and (c) try to narrow
5029 down where problems occurred
5033 Bugzilla : 3819, 4364, 4397, 6313
5034 Description: Racing close and eviction MDS could cause assertion in mds_close
5035 Details : It was possible to get multiple mfd references during close and
5036 client eviction, leading to one thread referencing a freed mfd.
5038 Severity: : enhancement
5039 Bugzilla : 3262, 6359
5040 Description: Attempts to reconnect to servers are now more aggressive.
5041 Details : This builds on the enhanced upcall-less recovery that was added
5042 in 1.4.2. When trying to reconnect to servers, clients will
5043 now try each server in the failover group every 10 seconds. By
5044 default, clients would previously try one server every 25 seconds.
5049 Description: After recovery, certain operations trigger a failed
5050 assertion on a client.
5051 Details : Failing over an mds, using lconf -d --failover, while a
5052 client was doing a readdir() call would cause the client to
5053 LBUG after recovery completed and the readdir() was resent.
5055 Severity : enhancement
5057 Description: Default groups are now added by lconf
5058 Details : You can now run lconf --group <servicename> without having to
5059 manually add groups with lmc.
5062 Frequency : occasional
5064 Description: Nodes with an elan id of 0 trigger a failed assertion
5067 Frequency : always when accessing e.g. tty/console device nodes
5069 Description: tty and some other devices nodes cannot be used on lustre
5070 Details : file's private_data field is used by device data and lustre
5071 values in there got lost. New field was added to struct file to
5072 store fs-specific private data.
5075 Frequency : when exporting Lustre via NFS
5077 Description: NFSD failed occasionally when looking up a path component
5078 Details : NFSD is looking up ".." which was broken in ext3 directories
5079 that had grown large enough to become hashed.
5082 Frequency : Clusters with multiple interfaces not on the same subnet
5084 Description: Nodes will repeatedly try to reconnect to an interface which it
5085 cannot reach and report an error to the log.
5086 Details : Extra peer list entries will be created by lconf with some peers
5087 unreachable. lconf now validates the peer before adding it.
5090 Frequency : Only if a default stripe is set on the filesystem root.
5092 Description: Setting a default stripe on the filesystem root prevented the
5093 filesystem from being remounted.
5094 Details : The client was sending extra request flags in the root getattr
5095 request and did not allocate a reply buffer for the dir EA.
5098 Frequency : occasional, higher if lots of files are accessed by one client
5099 Bugzilla : 6159, 6097
5100 Description: Client trips assertion regarding lsm mismatch/magic
5101 Details : While revalidating inodes the VFS looks up inodes with ifind()
5102 and in rare cases can find an inode that is being freed.
5103 The ll_test_inode() code will free the lsm during ifind()
5104 when it finds an existing inode and then the VFS later attaches
5105 this free lsm to a new inode.
5109 Bugzilla : 6422, 7030
5110 Description: MDS deadlock between mkdir and client eviction
5111 Details : Creating a new file via mkdir or mknod (starting a transaction
5112 and getting the ns lock) can deadlock with client eviction
5113 (gets ns lock and trying to finish a synchronous transaction).
5116 Frequency : occasional
5117 Description: While starting a server, the fsfilt_ext3 module could not be
5119 Details : CFS's improved ext3 filesystem is named ldiskfs for 2.6
5120 kernels. Previously, lconf would still use the ext3 name
5121 when trying to load modules. Now, it will correctly use
5122 ext3 on 2.4 and ldiskfs on 2.6.
5124 Severity : enhancement
5125 Description: The default stripe count has been changed to 1
5126 Details : The interpretation of the default stripe count (0, to lfs
5127 or lmc) has been changed to mean striping across a single
5128 OST, rather than all available. For general usage we have
5129 found a stripe count of 1 or 2 works best.
5131 Severity : enhancement
5132 Description: Add support for compiling against Cray portals.
5133 Details : Conditional compiling for some areas that are different
5137 Frequency : occasional
5138 Bugzilla : 6409, 6834
5139 Description: Creating files with an explicit stripe count may lead to
5140 a failed assertion on the MDS
5141 Details : If some OSTs are full or unavailable, creating files may
5142 trigger a failed assertion on the MDS. Now, Lustre will
5143 try to use other servers or return an error to the
5147 Frequency : occasional
5149 Description: Multiple concurrent overlapping read+write on multiple SMP nodes
5150 caused lock timeout during readahead (since 1.4.2).
5151 Details : Processes doing readahead might match a lock that hasn't been
5152 granted yet if there are overlapping and conflicting lock
5153 requests. The readahead process waits on ungranted lock
5154 (original lock is CBPENDING), while OST waits for that process
5155 to cancel CBPENDING read lock and eventually evicts client.
5157 Severity : enhancement
5159 Description: Initial enabling of flock support for clients
5160 Details : Implements fcntl advisory locking and file status functions.
5161 This feature is provided as an optional mount flag (default
5162 off), and is NOT CURRENTLY SUPPORTED. Not all types of record
5163 locking are implemented yet, and those that are are not guaranteed
5164 to be completely correct in production environments.
5165 mount -t lustre -o [flock|noflock] ...
5168 Frequency : occasional
5170 Description: OSTs running 2.4 kernels but with extents enabled might trip an
5171 assertion in the ext3 JBD (journaling) layer.
5172 Details : The b_committed_data struct is protected by the big kernel lock
5173 in 2.4 kernels, serializing journal_commit_transaction() and
5174 ext3_get_block_handle->ext3_new_block->find_next_usable_block()
5175 access to this struct. In 2.6 kernels there is finer grained
5176 locking to improve SMP performance of the JBD layer.
5180 Description: Changes the "SCSI I/O Stats" kernel patch to default to "enabled"
5182 -----------------------------------------------------------------------------
5184 2005-05-05 Cluster File Systems, Inc. <info@clusterfs.com>
5186 NOTE: Lustre 1.4.2 uses an incompatible network protocol than previous
5187 versions of Lustre. Please update all servers and clients to
5188 version 1.4.2 or later at the same time. You must also run
5189 "lconf --write-conf {config}.xml" on the MDS while it is stopped
5190 to update the configuration logs.
5192 - fix for HPUX NFS client breakage when NFS exporting Lustre (5781)
5193 - mdc_enqueue does not need max_mds_easize request buffer on send (5707)
5194 - swab llog records of type '0' so we get proper header size/idx (5861)
5195 - send llog cancel req to DLM cancel portal instead of cb portal (5515)
5196 - fix rename of one directory over another leaking an inode (5953)
5197 - avoid SetPageDirty on 2.6 (5981)
5198 - don't re-add just-being-destroyed locks to the waiting list (5653)
5199 - when creating new directories, inherit the parent's custom
5200 striping settings if present parent (3048)
5201 - flush buffers from cache before direct IO in 2.6 obdfilter (4982)
5202 - don't hold i_size_sem in ll_nopage() and ll_ap_refresh_count (6077)
5203 - don't hold client locks on temporary worklist from l_lru (5666)
5204 - handle IO errors in 2.6 obdfilter bio completion routine (6046)
5205 - automatically evict dead clients (5921)
5206 - Update file size properly in create+truncate+fstat case (6196)
5207 - Do not unhash mountpoint dentries, do not allow removal of
5209 - Avoid lock ordering deadlock issue with write/truncate (6203,5654)
5210 - reserve enough journal credits in fsfilt_start_log for setattr (4554)
5211 - ldlm_enqueue freed-export error path would always LBUG (6149,6184)
5212 - don't reference lr_lvb_data until after we hold lr_lvb_sem (6170)
5213 - don't overwrite last_rcvd if there is a *_client_add() error (6086)
5214 - Correctly handle reads of files with no objects (6243)
5215 - lctl recover will also mark a device active if deactivate used (5933)
5217 - by default create 1 inode per 4kB space on MDS, per 16kB on OSTs
5218 - allow --write-conf on an MDS with different nettype than client (5619)
5219 - don't write config llogs to MDS for mounts not from that MDS (5617)
5220 - lconf should create multiple TCP connections from a client (5201)
5221 - init scripts are now turned off by default; run chkconfig --on
5222 lustre and chkconfig --on lustrefs to use them
5223 - upcalls are no longer needed for clients to recover to failover
5225 - add --abort-recovery option to lconf to abort recovery on device
5227 - add support for an arbitrary number of OSTs (3026)
5228 - Quota support protocol changes.
5229 - forward compatibility changes to wire structs (6007)
5230 - rmmod NALs that might be loaded because of /etc/modules.conf (6133)
5231 - support for mountfsoptions and clientoptions to the Lustre LDAP (5873)
5232 - improved "lustre status" script
5233 - initialize blocksize for non-regular files (6062)
5234 - added --disable-server and --disable-client configure options (5782)
5235 - introduce a lookup cache for lconf to avoid repeated DB scans (6204)
5236 - Vanilla 2.4.29 support
5237 - increase maximum number of obd devices to 520 (6242)
5238 - remove the tcp-zero-copy patch from the suse-2.4 series (5902)
5239 - Quadrics Elan drivers are now included for the RHEL 3 2.4.21 and
5240 SLES 9 2.6.5 kernels
5241 - limit stripes per file to 160 (the maximum EA size) (6093)
5243 2005-03-22 Cluster File Systems, Inc. <info@clusterfs.com>
5246 - don't LASSERT in ll_release on NULL lld with NFS export (4655, 5760)
5247 - hold NS lock when calling handle_ast_error->del_waiting_lock (5746)
5248 - fix setattr mtime regression from lovcleanup merge (4829, 5669)
5249 - workaround for 2.6 crash in ll_unhash_aliases (5687, 5210)
5250 - small ext3 extents cleanups and fixes (5733)
5251 - improved mballoc code, several small races and bugs fixed (5733, 5638)
5252 - kernel version 43 - fix remove_suid bugs in both 2.4 and 2.6 (5695)
5253 - avoid needless client->OST connect, fix handle mismatch (5317)
5254 - fix DLM error path that led to out-of-sync client, long delays (5779)
5255 - support common vfs-enforced mount options (nodev,nosuid,noexec) (5637)
5256 - fix several locking issues related to i_size (5492,5624,5654,5672)
5257 - don't move pending lock onto export if it is already evicted (5683)
5258 - fix kernel oops when creating .foo in unlinked directory (5548)
5259 - fix deadlock in obdfilter statistics vs. object create (5811)
5260 - use time_{before,after} to avoid timer jiffies wrap (5882)
5261 - shutdown --force/--failover stability (3607,3651,4797,5203,4834)
5262 - Do not leak request if server was not able to process it (5154)
5263 - If mds_open unable to find parent dir, make that negative lookup(5154)
5264 - don't create new directories with extent-mapping (5909, 5936)
5266 - fix lustre/lustrefs init scripts for SuSE (patch from Scali, 5702)
5267 - don't hold the pinger_sem in ptlrpc_pinger_sending_on_import
5268 - change obd_increase_kms to obd_adjust_kms (up or down) (5654)
5269 - lconf, lmc search both /usr/lib and /usr/lib64 for Python libs (5800)
5270 - support for RHEL4 kernel on i686 (5773)
5271 - provide error messages when incompatible logs are encountered (5898)
5273 2005-02-18 Cluster File Systems, Inc. <info@clusterfs.com>
5274 * version 1.4.0.10 (1.4.1 release candidate 1)
5276 - don't keep a lock reference when lock is not granted (4238)
5277 - unsafe list practices (rarely) led to infinite eviction loop (4908)
5278 - add per-fs limit of Lustre pages in page cache, avoid OOM (4699)
5279 - drop import inflight refcount on signal_completed_replay error (5255)
5280 - unlock page after async write error during send (3677)
5281 - handle missing objects in filter_preprw_read properly (5265)
5282 - no transno return for symlink open, don't save no-trasno open (3440)
5283 - don't try to complete elan receive that already failed (4012)
5284 - free RPC server reply state on error (5406)
5285 - clean up thread from ptlrpc_start_thread() on error (5160)
5286 - readahead could read extra page into cache that wasn't ejected (5388)
5287 - prevent races in class_attach/setup/cleanup/detach (5260)
5288 - don't dereference de->d_inode after l_dput of de (5458)
5289 - use "int" for stripe value returned from lock_to_stripe (5544)
5290 - mballoc allocation and error-checking fixes in 2.6 (5504)
5291 - block device patches to fix I/O request sizes in 2.6 (5482)
5292 - look up hostnames for IB nals (5602)
5293 - 2.6 changed lock ordering of 2 semaphores, caused deadlock (5654)
5294 - don't start multiple acceptors for the same port (5277)
5295 - fix incorrect LASSERT in mds_getattr_name (5635)
5296 - export a proc file for general "ping" checking (5628)
5297 - fix "lfs check" to not block when the MDS is down (5628)
5299 - service request history (4965)
5300 - put {ll,lov,osc}_async_page structs in a single slab (4699)
5301 - create an "evict_client" /proc entry on OSTs, like the MDS has
5302 - fix mount usage message, return errors per mount(8) (5168)
5303 - change grep [] to grep "[]" in tests so they work in more UMLs
5304 - fix ppc64/x86_64 spec to use %{_libdir} instead of /usr/lib (5389)
5305 - remove ancient LOV_MAGIC_V0 EA support (5047)
5306 - add "disk I/Os in flight" and "I/O req time" stats in obdfilter
5307 - align r/w RPCs to PTLRPC_MAX_BRW_SIZE boundary for performance (3451)
5308 - allow readahead allocations to fail when low on memory (5383)
5309 - mmap locking landed again, after considerable improvement (2828)
5310 - add get_hostaddr() to lustreDB.py for LDAP support (5459)
5312 2004-11-23 Cluster File Systems, Inc. <info@clusterfs.com>
5315 - send OST transaction number in read/write reply to free req (4966)
5316 - don't ASSERT in ptl_send_rpc() if we run out of memory (5119)
5317 - lock /proc/sys/portals/routes internal state, avoiding oops (4827)
5318 - the watchdog thread now runs as interruptible (5246)
5319 - flock/lockf fixes (but it's still disabled, pending 5135)
5320 - don't use EXT3 constants in llite code (5094)
5321 - memory shortage at startup could cause assertion (5176)
5323 - reorganization of lov code
5324 - single portals codebase
5326 - add extents/mballoc support (5025)
5327 - direct I/O reads in the obdfilter (4048)
5328 - kernel patches from LNXI for 2.6 (bluesmoke, perfctr, mtd, kexec)
5330 tbd Cluster File Systems, Inc. <info@clusterfs.com>
5333 - send OST transaction number in read/write reply to free req (4966)
5334 - don't ASSERT in ptl_send_rpc() if we run out of memory (5119)
5335 - lock /proc/sys/portals/routes internal state, avoiding oops (4827)
5336 - the watchdog thread now runs as interruptible (5246)
5337 - handle missing objects in filter_preprw_read properly (5265)
5338 - unsafe list practices (rarely) led to infinite eviction loop (4908)
5339 - drop import inflight refcount on signal_completed_replay error (5255)
5340 - unlock page after async write error during send (3677)
5341 - return original error code on reconstructed replies (3761)
5342 - no transno return for symlink open, don't save no-trasno open (3440)
5344 - add pid to ldlm debugging output (4922)
5345 - bump the watchdog timeouts -- we can't handle 30sec yet
5346 - extra debugging for orphan dentry/inode bug (5259)
5348 2004-11-16 Cluster File Systems, Inc. <info@clusterfs.com>
5351 - fix TCP_NODELAY bug, which caused extreme perf regression (5134)
5352 - allocate qswnal tx descriptors singly to avoid fragmentation (4504)
5353 - don't LBUG on obdo_alloc() failure, use OBD_SLAB_ALLOC() (4800)
5354 - fix NULL dereference in /proc/sys/portals/routes (4827)
5355 - allow failed mdc_close() operations to be interrupted (4561)
5356 - stop precreate on OST before MDS would time out on it (4778)
5357 - don't send partial-page writes before EOF from client (4410)
5358 - discard client grant for sub-page writes on large-page clients (4520)
5359 - don't free dentries not owned by NFS code, check generation (4806)
5360 - fix lsm leak if mds_create_objects() fails (4801)
5361 - limit debug_daemon file size, always print CERROR messages (4789)
5362 - use transno after validating reply (3892)
5363 - process timed out requests if import state changes (3754)
5364 - update mtime on OST during writes, return in glimpse (4829)
5365 - add mkfsoptions to LDAP (4679)
5366 - use ->max_readahead method instead of zapping global ra (5039)
5367 - don't interrupt __l_wait_event() during strace
5369 - add software watchdogs to catch hung threads quickly (4941)
5370 - make lustrefs init script start after nfs is mounted
5371 - fix CWARN/ERROR duplication (4930)
5372 - return async write errors to application if possible (2248)
5373 - add /proc/sys/portal/memused (bytes allocated by PORTALS_ALLOC)
5374 - print NAL number in %x format (4645)
5375 - update barely-supported suse-2.4.21-171 series (4842)
5376 - support for sles 9 %post scripts
5377 - support for building 2.6 kernel-source packages
5378 - support for sles km_* packages
5380 2004-10-07 Cluster File Systems, Inc. <info@clusterfs.com>
5383 - ignore -ENOENT errors in osc_destroy (3639)
5384 - notify osc create thread that OSC is being cleaned up (4600)
5385 - add nettype argument for llmount in #5d in conf-sanity.sh (3936)
5386 - reconstruct ost_handle() like mds_handle() (4657)
5387 - create a new thread to do import eviction to avoid deadlock (3969)
5388 - let lconf resolve symlinked-to devices (4629)
5389 - don't unlink "objects" from directory with default EA (4554)
5390 - hold socknal file ref over connect in case target is down (4394)
5391 - allow more than 32000 subdirectories in a single directory (3244)
5392 - fix blocks count for O_DIRECT writes (3751)
5393 - OST returns ENOSPC from object create when no space left (4539)
5394 - don't send truncate RPC if file size isn't changing (4410)
5395 - limit OSC precreate to 1/2 of value OST considers bogus (4778)
5396 - bind to privileged port in socknal and tcpnal (3689)
5398 - rate limit CERROR/CWARN console message to avoid overload (4519)
5399 - GETFILEINFO dir ioctl returns LOV EA + MDS stat in 1 call (3327)
5400 - basic mmap support (3918)
5401 - kernel patch series update from b1_4 (4711)
5403 2004-09-16 Cluster File Systems, Inc. <info@clusterfs.com>
5406 - avoid crash during MDS cleanup with OST shut down (2775)
5407 - fix loi_list_lock/oig_lock inversion on interrupted IO (4136)
5408 - don't use bad inodes on the MDS (3744)
5409 - dynamic object preallocation to improve recovery speed (4236)
5410 - don't hold spinlock over lock dumping or change debug flags (4401)
5411 - don't zero obd_dev when it is force cleaned (3651)
5412 - print grants to console if they go negative (4431)
5413 - "lctl deactivate" will stop automatic recovery attempts (3406)
5414 - look for existing locks in ldlm_handle_enqueue() (3764)
5415 - don't resolve lock handle twice in recovery avoiding race (4401)
5416 - revalidate should check working dir is a directory (4134)
5418 - don't always mark "slow" obdfilter messages as errors (4418)
5420 2004-08-24 Cluster File Systems, Inc. <info@clusterfs.com>
5423 - don't close LustreDB during write_conf until it is done (3860)
5424 - fix typo in lconf for_each_profile (3821)
5425 - allow dumping logs from multiple threads at one time (3820)
5426 - don't allow multiple threads in OSC recovery (3812)
5427 - fix debug_size parameters (3864)
5428 - fix mds_postrecov to initialize import for llog ctxt (3121)
5429 - replace config semaphore with spinlock (3306)
5430 - be sure to send a reply for a CANCEL rpc with bad export (3863)
5431 - don't allow enqueue to complete on a destroyed export (3822)
5432 - down write_lock before checking llog header bitmap (3825)
5433 - recover from lock replay timeout (3764)
5434 - up llog sem before sending rpc (3652)
5435 - reduce ns lock hold times when setting kms (3267)
5436 - change a dlm LBUG to LASSERTF, to maybe learn something (4228)
5437 - fix NULL deref and obd_dev leak on setup error (3312)
5438 - replace some LBUG about llog ops with error handling (3841)
5439 - don't match INVALID dentries from d_lookup and spin (3784)
5440 - hold dcache_lock while marking dentries INVALID and hashing (4255)
5441 - fix invalid assertion in ptlrpc_set_wait (3880)
5443 - add libwrap support for the TCP acceptor (3996)
5444 - add /proc/sys/portals/routes for non-root route listing (3994)
5445 - allow setting MDS UUID in .xml (2580)
5446 - print the stack of a process that LBUGs (4228)
5448 2004-07-14 Cluster File Systems, Inc. <info@clusterfs.com>
5451 - don't cleanup request in ll_file_open() on failed MDS open (3430)
5452 - make sure to unset replay flag from failed open requests (3440)
5453 - if default stripe count is 0, use OST count for inode size (3636)
5454 - update parent mtime/ctime on client for create/unlink (2611)
5455 - drop dentry ref in ext3_add_link from open_connect_dentry (3266)
5456 - free recovery state on server during a forced cleanup (3571)
5457 - unregister_reply for resent reqs (3063)
5458 - loop back devices mounting and status check on 2.6 (3563)
5459 - fix resource-creation race that can provoke i_size == 0 (3513)
5460 - don't try to use bad inodes returned from MDS/OST fs lookup (3688)
5461 - more debugging for page-accounting assertion (3746)
5462 - return -ENOENT instead of asserting if ost getattr+unlink race (3558)
5463 - avoid deadlock after precreation failure (3758)
5464 - fix race and lock order deadlock in orphan handling (3450, 3750)
5465 - add validity checks when grabbing inodes from l_ast_data (3599)
5467 - add /proc/.../recovery_status to obdfilter (3428)
5468 - lightweight CDEBUG infrastructure, debug daemon (3668)
5469 - change default OSC RPC parameters to be better on small clusters
5470 - turn off OST read cache for files smaller than 32MB
5471 - install man pages and include them in rpms (3100)
5472 - add new init script for (un)mounting lustre filesystems (2593)
5473 - run chkconfig in %post for init scripts (3701)
5474 - drop scimac NAL (unmaintained)
5476 2004-06-17 Cluster File Systems, Inc. <info@clusterfs.com>
5479 - clean kiobufs before and after use (3485)
5480 - strip trailing '/'s before comparing paths with /proc/mounts (3486)
5481 - remove assertions to work around "in-flight rpcs" recovery bug (3063)
5482 - change init script to fail more clearly if not run as root (1528)
5483 - allow clients to reconnect during replay (1742)
5484 - fix ns_lock/i_sem lock ordering deadlock for kms update (3477)
5485 - don't do DNS lookups on NIDs too small for IP addresses (3442)
5486 - re-awaken ptlrpcd if new requests arrive during check_set (3554)
5487 - fix cond_resched (3554)
5488 - only evict unfinished clients after recovery (3515)
5489 - allow bulk resend, prevent data loss (3570)
5490 - dynamic ptlrpc request buffer allocation (2102)
5491 - don't allow unlinking open directory if it isn't empty (2904)
5492 - set MDS/OST threads to umask 0 to not clobber client modes (3359)
5493 - remove extraneous obd dereference causing LASSERT failure (3334)
5494 - don't use get_cycles() when creating temp. files on the mds (3156)
5495 - hold i_sem when setting i_size in ll_extent_lock() (3564)
5496 - handle EEXIST for set-stripe, set proper directory name (3336)
5498 - servers can dump a log evicting a client - lustre.dump_on_timeout=1
5499 - fix ksocknal_fmb_callback() error messages (2918)
5501 2004-05-27 Cluster File Systems, Inc. <info@clusterfs.com>
5504 - don't copy lvb into (possibly NULL) reply on error (2983)
5505 - don't deref dentry after dput, don't free lvb on error (2922)
5506 - use the kms to determine writeback rpc length (2947)
5507 - increment oti_logcookies when osc is inactive (2948)
5508 - update client's i_blocks count via lvb messages (2543)
5509 - handle intent open/close of special files properly (1557)
5510 - mount MDS with errors=remount-ro, like obdfilter (2009)
5511 - initialize lock handle to avoid ASSERT on error cleanup (3057)
5512 - don't use cancelling-locks' kms values (2947)
5513 - use highest lock extent for kms, not last one (2925)
5514 - don't dereference ERR_PTR() dentry in error handling path (3107)
5515 - fix thread race in portals_debug_dumplog() (3122)
5516 - create lprocfs device entries at setup instead of at attach (1519)
5517 - common AST error handler, don't evict client on completion race (3145)
5518 - zero nameidata in detach_mnt in 2.6 (3118)
5519 - verify d_inode after revalidate_special is valid in 2.6 (3116)
5520 - use lustre_put_super() to handle zconf unmounts in 2.6 (3064)
5521 - initialize RPC timeout timer earlier for 2.6 (3219)
5522 - don't dereference NULL reply buffer if mdc_close was never sent (2410)
5523 - print nal/nid for unknown nid (3258)
5524 - additional checks for oscc recovery before doing precreate (3284)
5525 - fix ll_extent_lock() error return code for 64-bit systems (3043)
5526 - don't crash in mdc_close for bad permissions on open (3285)
5527 - zero i_rdev for non-device files (3147)
5528 - clear page->private before handing to FS, better assertion (3119)
5529 - tune the read pipeline (3236)
5530 - fix incorrect decref of invalidated dentry (2350)
5531 - provide read-ahead stats and refine rpc in flight stats (3328)
5532 - don't hold journal transaction open across create RPC (3313)
5533 - update atime on MDS at close time (3265)
5534 - close LDAP connection when recovering to avoid server load (3315)
5535 - update iopen-2.6 patch with fixes from 2399,2517,2904 (3301)
5536 - don't leak open file on MDS after open resend (3325)
5537 - serialize filter_precreate and filter_destroy_precreated (3329)
5538 - loop device shouldn't call sync_dev() for nul device (3092)
5539 - clear page cache after eviction (2766)
5540 - resynchronize MDS->OST in background (2824)
5541 - refuse to mount the same filesystem twice on same mountpoint (3394)
5542 - allow llmount to create routes for mounting behind routers (3320)
5543 - push lock cancellation to blocking thread for glimpse ASTs (3409)
5544 - don't call osc_set_data_with_check() for TEST_LOCK matches (3159)
5545 - fix rare problem with rename on htree directories (3417)
5547 - allow default OST striping configuration per directory (1414)
5548 - fix compilation for qswnal for 2.6 kernels (3125)
5549 - increase maximum number of MDS request buffers for large systems
5550 - change liblustreapi to be useful for external progs like lfsck (3098)
5551 - increase local configuration timeout for slow disks (3353)
5552 - allow configuring ldlm AST timeout - lustre.ldlm_timeout=<seconds>
5554 2004-03-22 Cluster File Systems, Inc. <info@clusterfs.com>
5557 - fixes for glimpse AST timeouts / incorrectly 0-sized files (2818)
5558 - don't overwrite extent policy data in reply if lock was blocked (2901)
5559 - drop filter export grants atomically with removal from device (2663)
5560 - del obd_self_export from work_list in class_disconnect_exports (2908)
5561 - don't LBUG if MDS recovery times out during orphan cleanup (2530)
5562 - swab reply message in mdc_close, other PPC fixes (2464)
5563 - fix destroying of named logs (2325)
5564 - overwrite old logs when running lconf --write_conf (2264)
5565 - bump LLOG_CHUNKSIZE to 8k to allow for larger clusters (2306)
5566 - fix race in target_handle_connect (2898)
5567 - mds_reint_create() should take same inode create lock (2926)
5568 - correct journal credits calculated for CANCEL_UNLINK_LOG (2931)
5569 - don't close files for self_export to avoid uninitialized obd (2936)
5570 - allow MDS with the same name as client node (2939)
5571 - hold dentry reference for closed log files for unlink (2325)
5572 - reserve space for all logs during transactions (2059)
5573 - don't evict page beyond end of stripe extent (2925)
5574 - don't oops on a deleted current working directory (2399)
5575 - handle hard links to targets without a parent properly (2517)
5576 - don't dereference NULL lock when racing during eviction (2867)
5577 - don't grow lock extents when lots of conflicting locks (2919)
5579 2004-03-04 Cluster File Systems, Inc. <info@clusterfs.com>
5582 - account for cache space usage on clients to avoid data loss (974)
5583 - lfsck support in lustre kernel code (2349)
5584 - reduce journal credits needed for BRW writes (2370)
5585 - orphan handling to avoid losing space on client/server crashes
5586 - ptlrpcd can be blocked, stopping ALL progress (2477)
5587 - use lock value blocks to assist in proper KMS, faster stat (1021)
5588 - takes i_sem instead of DLM locks internally on obdfilter (2720)
5589 - recovery for initial connections (2355)
5590 - fixes for mds_cleanup_orphans (1934)
5591 - abort_recovery crashes MDS in b_eq (mds_unlink_orphan) (2584)
5592 - block all file creations until orphan recovery completes (1901)
5593 - client remove rq_connection from request struct (2423)
5594 - conf-sanity test_5, proper cleanup in umount log not availale (2640)
5595 - recovery timer race (2670)
5596 - mdc_close recovey bug (2532)
5597 - ptlrpc cleanup bug (2710)
5598 - mds timeout on local locks (2588)
5599 - namespace lock held during RPCs (2431)
5600 - handle interrupted sync write properly (2503)
5601 - don't try to handle a message that hasn't been replied to (2699)
5602 - client assert failure during cleanup after abort recovery (2701)
5603 - leak mdc device after failed mount (2712)
5604 - ptlrpc_check_set allows timedout requests to complete (2714)
5605 - wait for inflight reqs when ptlrpcd finishes (2710)
5606 - make sure unregistered services are removed from the srv_list
5607 - reset bulk XID's when resending them (caught by 1138 test)
5608 - unregister_bulk after timeout
5609 - fix lconf error (2694)
5610 - handle write after unfinished setstripe, stripe-only getstripe (2388)
5611 - readahead locks pages, leaves pending causing memory pressure (2673)
5612 - increase OST request buffers to 4096 on large machines (2729)
5613 - fix up permission of existing directories in simple_mkdir (2661)
5614 - init deleted item, add assertions ptlrpc_abort_inflight() (2725)
5615 - don't assign transno to errored transactions (2742)
5616 - don't delete objects on OST if given a bogus objid from MDS (2751)
5617 - handle large client PAGE_SIZE readdir on small PAGE_SIZE MDS (2777)
5618 - if rq_no_resend, then timeout request after recovery (2432)
5619 - fix MDS llog_logid record size, 64-bit array alignment (2733)
5620 - don't call usermode_helper from ptlrpcd, DEFAULT upcall (2773)
5621 - put magic in mount.lustre data, check for bad/NULL mount data (2529)
5622 - MDS recovery shouldn't delete objects that it has given out (2730)
5623 - if enqueue arrives after completion, don't clobber LVB (2819)
5624 - don't unlock pages twice when trigger_group_io returns error (2814)
5625 - don't deref NULL rq_repmsg if ldlm_handle_enqueue failed (2822)
5626 - don't write pages to disk if there was an error (1450)
5627 - don't ping imports that have recovery disabled (2676)
5628 - take buffered bytes into account when balancing socknal conn (2817)
5629 - hold a DLM lock over readdir always, use truncate_inode_pages (2706)
5630 - reconnect unlink llog connection after MDS reconnects to OST (2816)
5631 - remove little-endian swabbing of llog records (1987)
5632 - set/limit i_blksize to LL_MAX_BLKSIZE on client (2884)
5633 - retry reposting request buffers if they fail (1191)
5634 - grow extent at grant time to avoid granting a revoked lock (2809)
5635 - lock revoke doesn't evict page if covered by a second lock (2765)
5636 - disable VM readahead to avoid reading outside lock extents (2805)
5638 - return LL_SUPER_MAGIC from statfs for the filesystem type (1972)
5639 - updated kernel patches for hp-2.4.20 kernel (2681)
5641 2004-02-07 Cluster File Systems, Inc. <info@clusterfs.com>
5644 - fix truncated write corruption (2366)
5645 - fix for failed assertion in iopen_connect_dentry (1792,2517)
5647 - don't flag the ptlrpcd thread with PF_MEMALLOC (2636)
5648 - ensure len(uuid) < 37 in lmc (1171)
5649 - fix ia64 OOPS in llog_test (2255)
5650 - zero end of page at obdfilter for partial page writes (2648)
5651 - don't leave stale dentries around after renames (bug 2428)
5652 - fix timeouts when evicting a client with a single lock held (2642)
5653 - set deadline for the initial HELLO message to drain (2634)
5654 - print out dotted-quad IP addresses in the socknal (2302)
5656 - additional debugging for MDS client eviction problem (2443)
5657 - fix mkfsoptions support for osts (2603, 2604)
5659 2004-01-27 Cluster File Systems, Inc. <info@clusterfs.com>
5662 - add series for the vanilla 2.6.0 kernel
5663 - add series for the vanilla 2.4.24 kernel
5664 - add series for a cray x86/64 UL kernel drop
5665 - fix xattr patches for the vanilla 2.4.19 series
5667 - generate true UUIDs in lmc (1171)
5668 - have portals stack dumping break in UML (2466)
5669 - avoid bad dchild deref; avoid inum lock w/o creation (2362)
5670 - allocate with _NOFS in ldlm to avoid deadlock (1933)
5671 - wake callback waiting threads on client eviction (2460)
5672 - Add --ptldebug and --subsystem to lmc (1719)
5673 - update assertion to allow safe interrupt allocation
5674 - set rq_no_resend for cancel requests (2432)
5675 - recalculate ptlrpcd timeout after resend (2494)
5676 - call vfs_rmdir when removing pending directories (2368)
5677 - fix renaming a file to itself (2429)
5678 - lmc creates a default one-stripe lov (2454)
5679 - expand procfs space to handle large clusters (2326)
5680 - increase UML stack to avoid overflow
5681 - update lconf's list of debug and subsystem masks
5682 - fix lfs find --obd (2510)
5683 - /proc tunable for disabling filter read caching (2591)
5684 - stop rpm packages from altering slapd.conf (2301)
5685 - disable nagle in the socknal under 0conf (2578)
5686 - choose mds inode size based on stripe count (2572)
5687 - fix kernel-source rpm problems (2516)
5689 - add --disable-doc to avoid pdf generation (2421)
5690 - update documentation, tests, type-os, comments
5691 - avoid format warnings on ia64
5692 - remove the TOE NAL
5693 - tiny code cleanups by removing unused fields
5695 2004-01-07 Cluster File Systems, Inc. <info@clusterfs.com>
5698 - fix obvious semaphore misuse in as-yet-unused setattr path (2348)
5699 - remove the most blatant lies from BUILDING file (2371)
5700 - change default debug level to reasonable production setting
5701 - reduce client side cache size to reduce cache flush time
5702 - reduce max RPCs in flight to avoid unnecessary file fragmentation
5703 - make TCP zerocopy and pinger support enabled by default (2476)
5704 - sync writes completed after process exits caused crashes (2319)
5705 - maintain correct mount count on the MDS (2356)
5706 - backout 1557, because 2316 wasn't really fixed
5707 - better file I/O statistics gathering in /proc
5708 - don't take unnecessary, deadlock-inducing bug in readpage (2383)
5709 - another kernel patch to fix zero-copy TCP function export
5710 - don't take duplicate lock when processing re-sent getattr (2420)
5711 - lctl uses obd_self_export instead of creating new conn (2353)
5712 - MDS/OST recovery case which requires object creation asserted (2425)
5713 - move lfs from /usr/sbin to /usr/bin in packages
5714 - fix race between mds_client_add and mds_client_free (2417)
5715 - use kmalloc instead of slabs in portals (2430)
5716 - don't create duplicate records when a failover MDS is present (2442)
5717 - remove unnecessary mount age check (2332)
5718 - don't remove directory inodes from locks prematurely (2451)
5719 - don't break if MDS service name is the same as hostname (2103)
5720 - fix races in client write RPC generation when cache full (2482)
5722 2003-12-13 Cluster File Systems, Inc. <info@clusterfs.com>
5725 - remove now-unused request->rq_obd (278)
5726 - if an allocation fails, print out how much memory we've used (1933)
5727 - use PORTAL_SLAB_ALLOC for structures, to get GFP_MEMALLOC (1933)
5728 - add the "configurable stack size" patch to most series files (1256)
5729 - ability to write large log records, for 100+ OST configs (2306)
5730 - fix NULL deref when filter_prep fails (2314)
5731 - fix operator precedence error in filter_sync
5732 - dynamic allocation of socknal TX descriptors (2315)
5733 - fix a missed case in the GFP_MEMALLOC patch, can cause deadlock (2310)
5734 - fix gcc 2.96 compilation problem in xattr kernel patch (2294)
5735 - ensure that CWARN messages in Portals always get to the syslog
5736 - __init/__exit are not for prototype decls (ldlm_init/exit)
5737 - x86-64 compile warning fixes
5738 - fix gateway LMC keyword conflict (2318)
5739 - fix MDS lock inversions in getattr/reint paths (1844)
5740 - fix a rare lock re-ordering bug, which caused deadlock (2322)
5741 - fix i_sem/journal inversion in fsfilt_ext3_write_record (2306)
5742 - DLM race condition prevented some lock evictions (2328)
5743 - ENOMEM detection and retry on socknal sends (2230)
5744 - use GFP_NOFS throughout Lustre, to combat ENOMEM (2230)
5745 - move osc_rpcd into ptlrpc, for use in MDC and others (2329)
5746 - protect MDS inode fsdata with stronger locking; fixes assertion (2313)
5747 - better error messages when a client is rejected during recovery (1505)
5748 - avoid cancelling locks which were never granted, after failure (2330)
5749 - fix i_sem/journal inversion in mds_client_add (2333)
5750 - fix truncate/getattr lock cycle deadlock (2334)
5751 - use rpcd to send close; allows resend after timeout, avoid leak (1897)
5752 - fix two rare exit paths which could leak an l_lock() ref (2321)
5753 - fencepost error in MDS/OST orphan recovery (2226)
5754 - make log record alignment 8 bytes (1988)
5755 - lstripe now fails when requested offset > ost_count (2237)
5756 - ensure that all kernel series have a complete list.h (1607)
5757 - fix crashes in special-file operations (2316)
5758 - lctl create/brw OID mismatch, caused by obsolete filter loop (2339)
5760 - allow configurable automake binary, for testing new versions
5761 - small update to the lfs documentation
5763 2003-12-03 Cluster File Systems, Inc. <info@clusterfs.com>
5765 * fix negative export reference count in fsfilt_sync (2312)
5767 2003-12-01 Cluster File Systems, Inc. <info@clusterfs.com>
5768 * release candidate 0.9.1
5770 - orphans are moved into the PENDING directory for possible recovery
5771 - replayed opens now open by fid for orphan/rename safety (1042)
5772 - last close of an orphan inode generates a transno (683)
5773 - chdir() and mount() now pin the directory entry (1020)
5774 - avoid CERROR in normal ll_setattr_raw() error case (1500)
5775 - discard very old requests without processing them (1502)
5776 - remove some common, well-understood CERRORs (1505)
5777 - require O_DIRECT I/O to be page-sized to workaround IA64 crash (1609)
5778 - clear "grant" flags in OST replies until OST grant code lands (1644)
5779 - fix read performance by not clobbering i_blksize on client (1598)
5780 - fix __ldlm_handle2lock oops by not dereferencing lock after PUT (1625)
5781 - make LRU size a /proc tunable, clears locks when reduced (707)
5782 - fix some lprocfs rot that prevented ptlbd from loading (1732)
5783 - server locks take references on exports now (1558)
5784 - build fixes for 2.4.20-rh trees (1663)
5785 - return an error from lov_create if all OSCs are inactive (1751)
5786 - fix import levels when a reconnect happens without a timeout (1597)
5787 - exit early from mds_open if we get a lookup error (1749)
5788 - partial page read at EOF wouldn't wait for disk before sending (1642)
5789 - avoid NULL deref in obdfilter when reading page past EOF (1592)
5790 - avoid LASSERT in ll_intent_lock if server failed very early (1090)
5791 - fix LBUG in ll_it_open_error with rc = -2 (1861)
5792 - write/truncate lock inversion (1639)
5793 - Don't auto-load obdclass, portals modules during cleanup (1495)
5794 - fix timestamps from jumping to "now" (1763)
5795 - extra journal assertions (1648)
5796 - add an extra multiunlink test (1771)
5797 - fix read_record/write_record API (1776)
5798 - fix leak of offset_extent, possible incorrect i_size later (1772)
5799 - fix lasserts in mis-matched transnos during open-unlink testing (1541)
5800 - Debugging for the kqswnal_get_idle_tx problems (1820)
5801 - Allow recovery to be attempted multiple times (1536)
5802 - Write out MDS last_rcvd file after it is first created (1600)
5803 - Fix tx_descriptor leak in failed transmit situations (1827)
5804 - ext3 journaling fixes for assertion failure after IO error (1871)
5805 - class_export_put() on freed export after completion AST error (1896)
5806 - Fix revalidate looping in VFS (1322)
5807 - Don't access a freed export during MDS_REINT timeout (1521)
5808 - Add open-unlink recovery support on the MDS (1673,1764)
5809 - Return an error if no MDS data was read from last_rcvd (1946)
5810 - Fix for lookup "." or ".." crash on error (1932,1931,1935)
5811 - Don't setup a disk device that doesn't match exported UUID (317)
5812 - Reduce bulk RPC timeout to avoid cascading client/OST failures (1845)
5813 - avoid committing NULL handle in force close
5814 - local.sh is now a one-stripe LOV configuration
5815 - POSIX utime.4 -EPERM on FIFO not owned by user (56)
5816 - fix ext3 htree duplicate directory entry corruption (1516)
5817 - POSIX creat.13, fstat.1, open.18, stat.3 new file atime/mtime (2020)
5818 - update to new LOV EA format (2097)
5819 - interoperability for different PAGE_SIZE/wordsize (686,1821,1343,2042)
5821 2003-06-15 Phil Schwan <phil@clusterfs.com>
5824 - imports and exports cleanup too early, need refcounts (349, 879, 1045)
5825 - per-import/export recovery handling (958, 931, 959)
5826 - multiple last-rcvd slots, for serving multiple FSes (949)
5827 - connections are again shared between multiple imp/exports (963, 964)
5828 - "umount -f" would hang if any requests needed to be sent (393, 978)
5829 - avoid pinning large req buffer by copying for queued messages (989)
5830 - add "uuid" to "lctl device" command to help upcalls (991)
5831 - "open" RPCs with transnos would confuse recovery counters (1037)
5832 - do proper endian conversion of all wire messages (288, 340, 891)
5833 - remove OST bulk get LBUGs, fix ost_brw_write cleanup (1126)
5834 - call waiting locks callback from LDLM recovery thread (1127, 1151)
5835 - fix ptlrpc_connection leak in target_handle_connect (1174)
5836 - fix import refcounting bug in OST and MDS cleanup (1134)
5837 - if an invalid-at-open-time OSC returned before close(), LBUG (1150)
5838 - fix very unlikely obd_types race condition (501)
5839 - remove osc_open hack for echo_client (1187)
5840 - we leaked exports/dlmimps for forcibly disconnected clients (1143)
5841 - a failure in read_inode2 leads to deadlock (1139)
5842 - cancel ack-locks as soon as transaction is committed (1072)
5843 - fix major leaks and crashes in the bulk I/O path (937, 1057)
5844 - make sure to commitrw after any preprw to avoid deadlock (1162)
5845 - failing to execute a file in a lustre FS would lock inode (1203)
5846 - small DEBUG_REQ fix to avoid dereferencing a NULL (1227)
5847 - don't ASSERT while cleaning up an incompletely-setup obd (1248)
5848 - obd_uuid2tgt would walk off the end of the list (1255)
5849 - on IA64 the osc would give portals incorrect bulk size (1258)
5850 - fix debug daemon ioctl interface; allows daemon on ia64 (1274)
5851 - fix lock inversion caused by new llite matching code (1282)
5852 - limit the number of dirty pages on a client to 10MB (1286)
5853 - timed out locks were not being corrected cancelled (1289)
5854 - fix O_DIRECT above 4GB on IA-32 (1292)
5855 * major user-visible changes
5856 - fail out/fail over policy now controlled by the upcall (993)
5858 - add OBD_PING to check server availability and failure (954)
5859 - lustre messages are now sent in sending host order (288, 340, 891)
5860 - add eadatalen to MDS getattr reply (340)
5861 - OST read replies may contain second buffer, with per-page status (593)
5863 2003-03-11 Phil Schwan <phil@clusterfs.com>
5866 - LDLM_DEBUG macro fix, for gcc 3.2 (850)
5867 - failed open()s could cause deadlock; fixed (867, 869)
5868 - stop cancelling OST locks when files are closed (481)
5869 - overlapping XID spaces caused network corruption (851, 853)
5870 - fix unsafe fsfilt counter arithmetic; change to atomic_t
5871 - setattr_raw added, to do single-RPC, server-side setattrs
5872 - lmc/lconf syntax change for OST UUIDs
5873 - fix crashy race condition between ptlrpc_free_req and osc_close
5874 - don't use request in mdc_enqueue if we hit a timeout (889)
5875 - don't set the inode i_size for regular files from the MDS (896)
5876 - handle out of order completion AST (842)
5877 - don't LBUG if a lock request times out after receiving AST (913)
5878 - avoid d_rehash race in ll_find_alias by rehashing inside dcache_lock
5879 - if a bad lock AST arrives, send an error instead of dropping entirely
5880 - return 0 from revalidate2 if ll_intent_lock returns -EINTR (912)
5881 - fix leak in bulk IO when only partially completed (899, 900, 926)
5882 - fix O_DIRECT for ia64 (55)
5883 - (almost) eliminate Lustre-kernel-thread effects on load average (722)
5884 - C-z after timeout could hang a process forever; fixed (977)
5886 - client-side I/O cache (678, 924, 929, 941, 970)
5888 - READPAGE and SETATTRs which don't take server-side locks get
5891 2003-02-11 Phil Schwan <phil@clusterfs.com>
5894 - Fix ldlm_lock_match on the MDS to avoid matching remote locks (592)
5895 - Fix fsfilt_extN_readpage() to read a full page of directory
5896 entries, or fake the remainder if PAGE_SIZE != blocksize (500)
5897 - Avoid extra mdc_getattr() in ll_intent_lock when possible (534, 604)
5898 - Fix imbalanced LOV object allocation and out-of-bound access (469)
5899 - Most intent operations were removed, in favour of a new RPC mode
5900 that does a single RPC to the server and bypasses most of the VFS
5901 - All LDLM resource ID arrays were removed in favour of ldlm_res_id
5902 - Aggressively cancel local locks on DLM servers
5903 - mds_reint_unlink sends EA to the client if it's the last nlink.
5904 client uses that EA to unlink OST objects.
5905 - mds_reint_{rename,unlink,link} were rewritten to take ordered locks
5906 - recursive symlinks were fixed (439)
5907 - fixed NULL deref in DEBUG_REQ
5908 - filter_update_lastobjid no longer calls sync, which annoyed extN
5909 - fixed multi-client small-writes to a single file problem (445)
5910 - fixed mtime updates during file writes (607)
5911 - fixed vector writes on obdfilter causing problems when ENOSPC (670)
5912 - fixed bug in obd_brw_read/write() (under guise of testing 367)
5913 - fixed Linux OST size reporting problem (444, 656)
5914 - OST now updates object mtime with writes or setattr (607, 619)
5915 - client verifies file size before zeroing page past EOF (445)
5916 - OST now writes last allocated objid to disk with allocation (108)
5917 - LOV on echo now works (409)
5919 - mds_reint_unlink sends a new buffer, with the EA included. this
5920 buffer is only valid if body->valid & OBD_MD_FLEASIZE, which is only
5921 set if a regular file was being unlinked, and it was the last link
5922 - use PtlGet from the target for bulk writes (315)
5923 - OST now updates object mtime with writes or setattr (607, 619)
5924 - LDLM now has a grant-time callback to revalidate locked items, if
5926 - Many MDS operations were reorganized to combat race conditions
5928 - Merge b_intel branch (updated lprocfs code) - now at /proc/fs/lustre
5929 - configure check to avoid gcc version 2.96 20000731-2.96-98) (606)
5931 2003-01-06 Andreas Dilger <adilger@clusterfs.com>
5934 - Fully reactivate OST imports after reconnection (512, others)
5935 - Make sure client sees our -ENOTCONN from mds_handle (513 - partial)
5936 - More graceful error handling for truncating on dead OST (515)
5937 - Don't error out unless we're actually accessing dead stripes (474)
5938 - Fix garbage sizes when stripes are missing (410)
5939 - LRU counters were broken, causing constant lock purge (433, 432)
5940 - garbage on read from stripes with failed OSTs (441)
5941 - mark OSCs as active before reconnecting during recovery (438)
5942 - lov_enqueue and lov_cancel need to handle inactive OSTs (403)
5943 - lfind did not preserve OST order in output (443)
5944 - symlinks cause hung clients, incorrect data (439)
5945 - stop dereferencing request after dropping refcount (457)
5946 - don't LASSERT(spin_is_locked) on non-SMP (455)
5947 - fixes for many rename() bugs
5948 - fstat didn't correctly synchronize attributes (399)
5949 - server must handle lock cancellation during blocking AST prep (487)
5950 - bulk descriptors were free()d too soon (511)
5951 - fix paths in lconf, which would load incorrect modules (451, 507)
5952 - fix confusing lconf 'host not found' error message (386)
5953 - fix lock order deadlock on OST (O/R i_sem before journal ops, 478)
5954 - fix race condition in mdc_blocking_ast() for inode access (526)
5955 - fix lov_unpackmd() unpacking wrong number of stripes (537)
5956 - fix lov_set_osc_active() marking wrong OSC inactive (440)
5957 - fix bad lstripe lov_unpackmd() assertion (fix layering too) (527)
5958 - fix multiple writes of stripe MD to MDS (358, maybe 519)
5959 - fix lstripe in several ways (kernel side) (527)
5960 - fix request leak in ldlm_cli_enqueue (262)
5961 - incorrect OSC was marked inactive after OST failure
5962 - call mds_fs_cleanup before unmounting filesystem (524)
5963 - fix races between taking ns_lock and ldlm_lock_change_resource
5964 - fix races updating LOV export open file list
5965 - fix lov_enqueue error path, avoid decref-ing bad lock handle
5966 - fix recovery NULL deref in ldlm_cli_cancel_unused
5967 - fix some DLM races by using new hash table for lock handles (419)
5968 - permit the client to specify desired inodes, at replay
5969 - duplicate requests when we queue them for replay reintegration
5970 - fix last_rcvd offset calculation
5971 - sync after each recovered transaction, so we always make progress
5972 - never, not always, ERESTART requests without transnos
5973 - store the lov_desc in the MDS, so we don't depend on getlovinfo to
5975 - skip replay if the MDS says that the client is already connected
5976 - don't check for a recovery-enabled export to match lctl's UUID
5977 - don't INC_USE_COUNT for phantom exports
5978 - don't crash when cleaning up phantom exports (567)
5979 - don't double-finish or set replay data for errored mdc_open requests
5980 - abort requests when they time out, so we don't get old replies
5981 - send/receive replies for AST messages again
5982 - if the client says that it doesn't have the lock, cancel it on the
5984 - if we timeout during I/O, don't try to cancel an in-use lock; instead
5985 mark it as destroyed, it will all work out when decref is called
5986 - fix module use counts (22, 581)
5988 - ASTs now expect a reply (server cancels lock on error reply)
5990 2002-12-02 Andreas Dilger <adilger@clusterfs.com>
5993 - fix many simultaneous client startup (392)
5994 - fix dentry->d_it clobbering
5995 - credentials weren't being shipped for readdir/getattr operations
5996 - remove invalid assertions triggered during some concurrent MD
5998 - proper Lustre versions added (336, 389)
5999 - fix memory leak for create error case (398)
6000 - fix LOV locking bug that would get cli/srv out of sync
6001 - fix echo client over LOV (409)
6002 - fix dbench 2, extN refcount problem (170, 258, 356, 418)
6003 - fix double-O_EXCL intent crash (424)
6004 - avoid sending multiple lock CANCELs (352)
6006 - MDS can do multi-client recovery (modulo bugs in new code)
6008 - many updates, edits, cleanups
6010 2002-11-18 Phil Schwan <phil@clusterfs.com>
6013 - fix null d_it dereference (346)
6014 - fix full OST/dbench hang (333)
6015 - fix permission problem with file removal (286)
6016 - fix removal of OSCs from LOV when they fail
6017 - fix NULL deref during bulk timeout (214)
6018 - fix problems related to multiple filesystems on one MDS (241)
6019 - fixed serious subtle metadata locking bugs
6020 - free locks on clients when inodes are removed due to memory
6022 - fix inode pointer in lock data (285)
6023 - partial support for multiple MDS on a single host (241)
6024 - data locks weren't cancelled at clear_inode time (290, 311)
6025 - intent locks could lead to unbounded lock growth (205)
6026 - added a maximum lock count, an LRU list, and a flusher
6027 - fix multiple rename (365)
6028 - properly abstracted the echo client
6029 - OSC locked 1 byte too many; fixed
6030 - rewrote brw callback code:
6031 - fixed recovery bugs related to LOVs (306)
6032 - fixed too-many-pages-in-one-write crash (191)
6033 - fixed (again) crash in sync_io_timeout (214)
6034 - probably fixed callback-related race (385)
6036 - Add capability to MDS protocol
6037 - LDLM cancellations and callbacks on different portals
6039 2002-10-28 Andreas Dilger <adilger@clusterfs.com>
6042 - limit client IOV size to PTL_MD_MAX_IOV (611336, 191)
6043 - defer open object destruction to close time (601981, 138)
6044 - open/close OST file handle in obdo (OBD_MD_FLHANDLE) (601981, 138)
6045 - move LDLM_ENQUEUE/CONVERT back to MDS portal (625069)
6046 - abstract ll_lookup2, fix ll_revalidate2 to use abstraction (256)
6047 - don't call obd_setattr in ll_file_release for destroyed objects
6048 * protocol change to lustre_msg: move |version| and add |flags|
6049 * protocol change to osc_punch: "start" in "o_size", "end" in "o_blocks"
6050 * lock replay: for LDLM_FL_REPLAY trust client to do right thing
6051 * added replay of create, unlink, link and rename operations during
6052 MDS failover; recovery should be much more robust now
6053 * remove failed OSCs from LOVs (only lov_create uses this so far)
6054 * the lustre-HOWTO was brought (more) up to date (582544)
6056 2002-10-23 Phil Schwan <phil@clusterfs.com>
6059 - in-use dentries weren't being reused properly (617851)
6060 - prevent multiple LDLM setup (599178)
6061 - fix LOV size calculations for truncate (617853)
6062 - fix client handling of MDS intent errors (POSIX)
6063 - fix permission bug in lovstripe.c test (624321)
6064 - fix MDS thread deadlock - move LDLM handler to DLM portal (625069)
6065 - truncate past end of file could corrupt data
6066 - proper cleanup after timeouts, crashes, etc (592524, 550815)
6067 - a race in recovery could return ETIMEDOUT to apps (623947)
6068 - building outside the source directory was fixed
6069 * the lustre-HOWTO was brought (more) up to date (582544)
6070 * major progress was made on recovery functionality
6072 2002-10-10 Phil Schwan <phil@clusterfs.com>
6075 - recovery deadlock fix
6076 - rm -rf causes LBUG fix (617817)
6077 - file open by multiple tasks fix (618962)
6078 - directory permissions bugs (602707 and 620007)
6079 - journal_stop fixed with locking (611313)
6080 - O_APPEND failures resolved (618273, perhaps 614459)
6081 - lconf PATH fix (619770)
6082 - IA64 build fix (621450)
6083 - RPC buffer sizes scale with amount of memory
6085 2002-10-01 Phil Schwan <phil@clusterfs.com>
6088 - locks would be cancelled without throwing away data pages,
6089 resulting in inconsistent data (605627)
6090 - inode attributes were not always being refreshed (605627, 612449)
6091 - lconf now continues to cleanup after lctl reports an error
6092 - MDS now enforces user permissions (602707)
6093 - lprocfs cleanup fixed, but not yet enabled (614157)
6094 - fixed infinite server hang, should a client not respond to an AST
6095 - avoid going into recovery if user calls readlink() with a buffer
6096 that's too small (613941)
6097 - AST RPCs no longer require replies (614867) -- this may be changed
6098 - don't crash server if client sends an IOV that's too big (611336)
6099 - fixed lock conversion deadlock (611892)
6100 - fixed the following of symlinks (614622)
6101 * recovery: the server can remove locks from a client that dies, other
6102 clients can make progress
6103 * more extN patch fixes
6104 * compile-time configurable ptlrpc buffer allocations
6106 - collaborative read cache document
6107 - Lustre Lite Performance CDR document-in-progress
6109 2002-09-20 Andreas Dilger <adilger@clusterfs.com>
6112 - fix typo in patch-2.4.18
6114 2002-09-20 Andreas Dilger <adilger@clusterfs.com>
6117 - clear ptlrpc request each time in handle_incoming_request()
6118 - unlink of files now destroys the object on the OST
6120 2002-09-19 Peter Braam <braam@clusterfs.com>
6122 * add hard link support
6123 * change obdfile creation method
6124 * kernel patch changed
6126 2002-09-19 Peter Braam <braam@clusterfs.com>
6129 - stack overflow bug in extN fixed
6131 2002-09-18 Andreas Dilger <adilger@clusterfs.com>
6133 * documentation updates
6134 - add man pages for config tools
6135 - update tests/README to describe testing with new config tools
6136 - finish metadata API descriptions
6137 * bug fixes and cleanups
6138 - statfs workaround for 16TB limit
6139 - LOV stripe allocation improved, can stripe on subset of OSTs
6140 - LOV file size/IO offset was wrong for files > 4GB in size
6141 - object EA data was being dropped, caused files to be unreadable
6142 - memory overflow with non-LOV OST caused memory corruption
6143 - fixed regression tests to work with new config tools, obdfilter
6144 - fixed bug when directory size became larger than 1 block
6145 - fixed bug (for single client case) when PWD was deleted
6146 - invalidate local directory pages when doing intent-based ops
6147 - avoid LDLM oops when lock callback contained bad data
6149 2002-09-09 Andreas Dilger <adilger@clusterfs.com>
6151 * documentation updates
6152 * bug fixes and cleanups
6153 - configuration tools
6156 - 64-bit compile warnings
6157 - 64-bit internal statfs data
6159 * test_brw on persistent OST devices
6161 * lprocfs (disabled)
6163 2002-09-04 Andreas Dilger <adilger@clusterfs.com>
6165 * documentation updates
6166 * bug fixes and cleanups
6167 * configuration tools
6169 2002-08-30 Peter J. Braam <braam@clusterfs.com>
6172 * many small fixes to 0_5_4
6173 * io/network handling
6174 * thinkos in MDS operations
6176 2002-08-24 Peter J. Braam <braam@clusterfs.com>
6179 * crucial basic fixes to 0.5.3
6180 * IOR, Iozone work over Elan
6183 2002-08-07 Phil Schwan <phil@clusterfs.com>
6184 * version 0_5_3, our first alpha
6185 * we use the new Portals iovs
6186 * documentation updates
6187 * bug fixes and cleanups
6188 * small changes in the DLM wire protocol
6190 2002-07-25 Peter J. Braam <braam@clusterfs.com>
6191 * version 0_5_1 with some initial stability,
6192 * locking on MD and file I/O.
6193 * documentation updates
6194 * several bug fixes since 0.5.0
6195 * small changes in wire protocol
6197 2002-07-18 Phil Schwan <phil@clusterfs.com>
6199 * delivered as Lustre Light Alpha
6200 * fixed a crash after handling invalid MDS requests
6201 * fixed directory pages for architectures with non-4k pages sizes
6203 2002-07-11 Andreas Dilger <adilger@clusterfs.com>
6204 * release version v0_4_4
6205 * Moves TCP acceptor to be on port 2432 (unused Coda port) instead
6207 * Fixes a number of interruption problems with OST operations.
6208 * Update documentation for portals header changes
6209 * Move all wire protocol structs/defines to lustre_idl.h
6210 * Fixes symlink length bug.
6211 * Add tcpdump to repository.
6213 2002-07-05 Andreas Dilger <adilger@clusterfs.com>
6214 * release version v0_4_3
6215 * Fixes statfs for inodes on extN.
6216 * Fixes bug in runtests which would delete /etc/hosts.
6217 * Use 64-bit object IDs wherever possible (not into VFS though)
6218 Remove ost_get_info, which is unused by lustre, and out of date.
6220 2002-07-03 Peter Braam <braam@clusterfs.com>
6221 * release version v0_4_2 Fixes a lookup error (type not passed)
6222 * move forward to head of Portals
6223 * move forward to latest Lustre kernel
6225 2002-06-25 Peter Braam <braam@clusterfs.com>
6226 * release version v0_4_1. Hopefully stable on single node use.
git://git.whamcloud.com / fs / lustre-release.git / blob