1 2010-xx-xx Oracle, Inc.
4 2.6.16.60-0.42.8 (SLES 10),
5 2.6.32.19-0.2.1 (SLES11),
6 2.6.18-194.8.1.el5 (RHEL 5)
7 2.6.18-194.8.1.0.1.el5 (OEL 5)
8 * Client support for unpatched kernels:
9 (see http://wiki.lustre.org/index.php?title=Patchless_Client)
10 2.6.16 - 2.6.32 vanilla (kernel.org)
11 * Recommended e2fsprogs version: 1.41.10-sun2
12 * The async journal commit feature (bug 19128) and the cancel
13 lock before replay feature (bug 16774) are disabled by default.
15 Severity : enhancement
17 Description: Update SLES11 SP1 kernel to 2.6.32.19-0.2.1.
21 Description: Enabling quotas fails with non-consecutive OST numbering.
25 Description: Fix kernel warning due to lookup_on_len() called without i_mutex
30 Description: Account direct i/o inflight rpcs separately from non-direct i/o so
31 that direct i/o, which is limited by max_rpcs_in_flight, should not
32 block non-direct i/o, which is not limited by max_rpcs_in_flight.
36 Description: Fix per-NID reporting on outstanding writes
40 Description: Reduce stack pressure by uninlining some mds and ptlrpc functions.
44 Description: Remove LASSERT in lprocfs_rd_conn_uuid() since conn == NULL is a
49 Description: fix obdo leak issue in ll_setattr_raw()
53 Description: limit MMP interval
55 Severity : enhancement
57 Description: add several lfs ost enhancements
61 Description: Too many default ACLs break directory access on new directories
65 Description: Lustre inode size is not coherent across nodes.
66 Details : Update lvbo from disk when AST fails with EINVAL. Lvbo is updated
67 on EINVAL error in ldlm_handle_ast_error(). The updates in
68 filter_intent_policy() and ldlm_cb_interpret() have been removed as
73 Description: Oops at __percpu_counter_add+0x1b
74 Details : Use bdi_init()/bdi_destroy() to proper initialize backing_dev_info
79 Description: add mount option to generate 32bit ino, this can be used for 32bit
80 application compatibility.
84 Description: keep reference count for "lli_sai" to prevent it to be released
85 when "statahead_enter()"
89 Description: allow quotacheck over OSTs with sparse indices
93 Description: Objects not getting deleted for files which have been removed
94 Details : ll_have_md_lock() should differentiate between CR and CW OPEN
99 Description: pin object's inode in memory to avoid certain timeouts
101 -------------------------------------------------------------------------------
103 2010-07-31 Oracle, Inc.
105 * Support for kernels:
106 2.6.16.60-0.42.8 (SLES 10),
107 2.6.27.39-0.3.1 (SLES11),
108 2.6.18-194.3.1.el5 (RHEL 5)
109 2.6.18-194.3.1.0.1.el5 (OEL 5)
110 * Client support for unpatched kernels:
111 (see http://wiki.lustre.org/index.php?title=Patchless_Client)
112 2.6.16 - 2.6.30 vanilla (kernel.org)
113 * Recommended e2fsprogs version: 1.41.10-sun2
114 * The async journal commit feature (bug 19128) and the cancel
115 lock before replay feature (bug 16774) are disabled by default.
117 Severity : enhancement
119 Description: Update RHEL5.5 kernel to 2.6.18-194.8.1.el5 and OEL5.5 kernel to
120 2.6.18-194.8.1.0.1.el5.
122 Severity : enhancement
124 Description: using inkernel OFED stack for rhel5 & oel5.
126 Severity : enhancement
128 Description: Add "lfs_migrate" script from manual into lustre/scripts and RPMs
129 Details : lfs_migrate does a "poor man's" migration of files from their
130 current OST layout to a new OST layout as chosen by the MDS.
134 Description: mds_orphan_add_link()) error linking orphan to PENDING
135 Details : quota limits might disallow linking orphans to PENDING
136 when unlinking a file - temporary raise threads' privileges
137 when processing unlinks.
139 Severity : enhancement
141 Description: add conf-param -d option to remove permanent settings.
142 Details : Add the ability to remove permanent lctl conf_param settings.
143 (Previously conf_param settings could only be changed, not
144 removed.) This also provides a method to change failover
145 nid locations. Improve lctl man page.
147 Severity : enhancement
149 Description: add list_param to b1_8 and add "-R" option to list params
152 Severity : enhancement
154 Description: lfs quota output is not very convenient for awk/sed-parsing
155 Details : Some positions in lfs quota output table could be empty or
156 non-empty which made it hard to parse it with scripts, now a dash
157 is put instead of space where there is not supposed to be any data.
159 Severity : enhancement
161 Description: fix obdfilter-survey script to work properly with remote oss-s
163 Severity : enhancement
165 Description: add new OBDFILTER_SURVEY test suite
167 Severity : enhancement
169 Description: add new multiple mount protection (MMP) test suite
171 Severity : enhancement
173 Description: add support for async journal commit in echo client
175 Severity : enhancement
177 Description: allow userland programs to include <lustre/lustre_idl.h>
178 from stardard include directories
180 Severity : enhancement
182 Description: The prune-icache-use-trylock is no longer needed now that
183 the patch from bug 20008 is landed.
187 Description: The shrink grant feature is still active on the client although the
188 connect flag is not set.
192 Description: Don't leak grant space if the write failed with quota exceeded.
196 Description: Don't consume grant space twice on recoverable resent.
200 Description: a race condition could lead to SIGBUS being sent to an
201 application using mmap-ped files from Lustre
202 Details : truncate_complete_page implementation for the patchless
203 client could arbitrarily unset PG_Uptodate flag for a
204 page being kicked from the page cache, an uptodate check
205 right after a readpage call in filemap_fault could fail
206 because of that as though the page read had been unsuccessful.
210 Description: dlm lock slab shrinking is not efficient
211 Details : The dlm_locks slab can grow significantly and consumes a lot of
212 memory on the server. Set a hardlimit to grant_plan.
216 Description: Lustre does not do 1MB IOs to HW RAID
217 Details : Bump MAX_PHYS/HW_SEGMENTS and SG_ALL to 256 in the RHEL5 kernel.
218 This is what we do already for SLES kernels.
222 Description: bump maximum number of phys/hw segments in the SLES11 kernel
223 until s/g chaining works properly.
228 Description: LSI Fusion MPT driver hacks to improve performance
229 Details : Set CONFIG_FUSION_MAX_SGE to 256 for RHEL5
231 Severity : enhancement
233 Description: increase default md stripe_cache_size to 16k
236 Bugzilla : 15587/21439
237 Description: don't handle security.capability xattr
238 Details : CONFIG_SECURITY_FILE_CAPABILITIES is enabled by default on SLES11.
239 This results in additional getxattr calls, causing VBR test
240 failures as well as a preformance drop when writing.
244 Description: obdfilter-survey is no longer working
245 Details : revert patch from bug 20355 to resolve an issue with lctl --threads
246 not working correctly with $(PTHREAD_LIBS) being linked to lctl.
250 Description: ll_shrink_cache does not handle __GFP_FS properly
254 Description: lfs getstripe shows wrong info for directories
255 Details : Set correct LOVEA default values for filesystem-wide.
259 Description: FSX checksum false positves due to mmap IO
260 Details : Use OBD_FL_MMAP flag for IOs on a memory mapped file. Do not print
261 checksum errors, if the flag is set on a request.
265 Description: file operations after eviction have successful return values
266 Details : use vfs ->flush callback to return any pending async errors
271 Description: mdsrate fails to write after 1.3+M files opened
272 Details : decrease memory usage on clients by recycling dentries and
277 Description: obdfilter-survey gives unreasonably high numbers
278 Details : Wait for all threads to complete when running test_brw.
282 Description: do not set lustre read_only device when server umount and keep
283 client records for recoverable ones
287 Description: move sync_on_lock_cancel tunable to the obdfilter layer
288 Details : move the tunable to trigger a journal flush on lock cancel from
289 the ost layer to the obdfilter layer. This tunable is useful
290 when using the async journal commit feature.
294 Description: exp->exp_nid_stats == NULL in filter_tally()
295 Details : fix race with per-nid stats by delaying procfs cleanup until
300 Description: extent lock cancellation on client can keep the cpu busy for too
305 Description: Do not fail OST activation when a llog is not found, just
306 issue an error message.
310 Description: Don't enable extents by default for MDT.
314 Description: Protect bitfield access to ptlrpc_request's rq_flags, since
315 the AT code can access it concurrently while sending early
320 Description: Disable lockless truncate by default since it is sometimes flawed
321 and causes the write_disjoint test to fail.
325 Description: OSSs which don't have the patch from bug 20278 can trigger an
328 Severity : enhancement
330 Description: don't print message to the console when we have not managed to
335 Description: The MDS fails to synchronize OSTs which registered with the MGS
336 after the MDT. The problem is that OBD_NOTIFY_CREATE events are
337 raised too early and thus discarded by the MDT stack.
338 The fix consists of issuing OBD_NOTIFY_CREATE event in the lov
343 Description: Fix race when the ping evictor and a service thread execute
344 target_recovery_check_and_stop() concurrently.
348 Description: quota broadcast can trigger a LBUG on the MDT if there are
353 Description: Resetting the lov_objid values to last_id reported by the OST
354 during orphan recovery is incorrect and can cause the same
355 objects to be allocated twice.
357 Severity : enhancement
359 Description: "weak-modules" support
360 Details : Implement "weak-modules" support which enables kernel modules
361 to be used with any kernel that implements the same kABI. In
362 order to achieve this modules are now installed in
363 /lib/modules/$(uname -r)/updates/kernel on all distributions.
365 Severity : enhancement
367 Description: add writeconf as mount option
369 Severity : enhancement
371 Description: produce debuginfo packages for SLES.
373 Severity : enhancement
375 Description: add failover nidlist to the import proc file.
377 Severity : enhancement
379 Description: fix LUSTRE_SEQ_MAX_WIDTH for interoperability between 1.8
380 clients and 2.0 servers.
382 Severity : enhancement
384 Description: lfs find -s does not work correctly because of a bug in
389 Description: ll_read_ahead_page() must validate the dlm lock before using
394 Description: Prevent failover nids from registering with MGS first.
398 Description: fix lock inversion in ll_setattr_raw().
402 Description: object allocation is not balanced across OSTs.
403 Details : osc_precreate() should return 0, if there are enough objects left.
405 -------------------------------------------------------------------------------
407 2010-04-30 Oracle, Inc.
409 * Support for kernels:
410 2.6.16.60-0.42.8 (SLES 10),
411 2.6.27.39-0.3.1 (SLES11),
412 2.6.18-164.11.1.el5 (RHEL 5)
413 2.6.18-164.11.1.0.1.el5 (OEL 5)
414 * Client support for unpatched kernels:
415 (see http://wiki.lustre.org/index.php?title=Patchless_Client)
416 2.6.16 - 2.6.30 vanilla (kernel.org)
417 * Recommended e2fsprogs version: 1.41.10-sun2
418 * The async journal commit feature (bug 19128) and the cancel
419 lock before replay feature (bug 16774) are disabled by default.
423 Description: open-unlinked directories trigger MDS LBUG
424 Details : Fix regression introduced by the patch from bug 19640.
425 ext3_inc_count() can reset nlink to 1 when the directory
426 is indexed and inode->i_nlink == 2. Work around the problem
427 by incrementing nlink by 2 instead of 1.
431 Description: Reconnects are not throttled
432 Details : Don't wake up pinger on reconnect failures and rely on regular
433 pings to trigger the next reconnection. Please note that the
434 pinger already uses a smaller interval if the import is
438 Frequency : only with NFS export
440 Description: Console flooded with error message from ll_inode_from_lock()
442 Details : in mds_open, initialize the child_res_id before enqueuing
443 the OPEN lock for the child inode, then to avoid senting
444 wrong ldlm_res_id to client.
448 Description: allow multiple instances of the same nid in NID hash
449 Details : Case of multiple separate clients from the same NID (as
450 with liblustre) is legitimate and so we should allow
451 multiple instances of the same NID in nid hash.
455 Description: the readahead code can sleep on a semaphore while holding a
457 Details : in ras_update, "lov_get_info" could be called during increasing
458 readahead windows, which tries to get the mutex lock "lov_lock"
459 while holding the spin_lock "ras_lock", then causes system
464 Description: ASSERTION(cli->cl_avail_grant >= 0) failed
465 Details : fix assertion failure in the grant code.
469 Description: Use CNETERR (which is rate limited) in specific places in
470 the portal's LNET driver to avoid flooding the console.
474 Description: include last created object in precreate slow case
478 Description: don't do rep-ack if not created anything
479 Details : mds_open currently always put a lock into a rep-ack regardless
480 if something was created or not. This is pointless and only
481 creates needless contention. In fact the entire idea was to do
482 this for real creates as a recovery protection.
486 Description: Spurious error messages from smp_processor_id() on preemptible
488 Details : Disable a preemption by grabbing the lock in fs_trace_get_tcd()
489 first. The function fs_trace_get_tcd() was moved up.
493 Description: interval_erase() fix
494 Details : interval_erase() calls update_maxhigh() properly when child
499 Description: "lfs df" does not print stats for all mountpoints
500 Details : Print all mounted lustre filesystems with "lfs df"
504 Description: lfs setstripe -p not longer work with a relative pathname
505 Details : Use realpath() to provide absolute pathname.
509 Description: fix for truncated reply buffer
510 Details : reply buffer could be referred by reply_in_callback after released
515 Description: lustre.lov error when backing up symlinks with extended attributes
516 Details : Improved logic in ll_listxattr()
520 Description: properly handle null value for setattr -n lustre.lov
521 Details : Running "setfattr -n trusted.lov ." causes a NULL dereference
522 in ll_setxattr() due to no checking if "value" is NULL.
523 This command now resets to the default striping when executed
528 Description: stack overflow on lock cancellation due to fsync call
529 Details : sync_on_lock_cancel is needed for recovery when async journal
530 is enabled, but we actually just need to make sure that
531 metadata blocks have hit the journal, so doing a fs sync
532 should be enough and should consume less stack (just create an
533 empty handle and commmit it).
537 Description: using current->journal_info to store per-thread data leads
538 to problem under memory pressure
539 Details : disable the per-thread data (current->journal_info) containing
540 the lock info during I/O to work around the issue for short tem
544 Description: control DCACHE_LUSTRE_INVALID flag with MDS_INODELOCK_LOOKUP lock
545 Details : DCACHE_LUSTRE_INVALID is controlled by MDS_INODELOCK_LOOKUP
546 lock which is corresponding to "IT_LOOKUP", do not skip invalidate
551 Description: Cannot send after transport shutdown
552 Details : Clear imp_vbr_failed flag upon eviction
556 Description: soft lock in request set code during recovery
557 Details : during recovery, uses req->rq_set itself to replay the request
558 instead of ptlrpcd_recovery_pc
562 Description: Use CFS_ALLOC_IO instead of _STD in llap_from_page_with_lockh
563 Details : During an ll_readahead under ll_readpage, we have seen the the
564 OBD_SLAB_ALLOC hang under ldlm_pools_shrink when trying to lock
565 a page that is already locked by the readahead code.
570 Description: stop waitting for next replay transno if shutdown
571 Details : if the system is shutting down, wake up service thread blocked
572 to wait for next replay transno during recovery, then all the
573 references held by queued requests can be dropped and device
578 Description: lov_merge_lvb()) ASSERTION(spin_is_locked(&lsm->lsm_lock)) failed
579 Details : Protect lli->lli_smd pointer updates with lli->lli_lock.
583 Description: per-nid stats should not access lustre-hash internal structures
588 Description: mount.lustre fails to pass some options to mount()
593 Description: ext4 extent allocation is slower than in ext3
594 Details : Increase the default value of MB_DEFAULT_ORDER2_REQS to 8,
595 enlarge ext4 preallocation table for 2048 4K blocks extents
600 Description: incorrect triggering of synchronous IO
601 Details : The OSC can mistakenly fall back to synchronous IO when the
602 max_dirty_mb limit is reached and no write requests have yet
603 been issued. This can occur when the dirty pages are spread
604 over many files all of which are below the optimal request size.
608 Description: Optimize quota_ctl operations by sending requests in parallel
609 Details : Send MDS->OST quota_ctl requests in parallel, do not resend.
610 Compiled from two attachments in the ticket.
614 Description: ordering issue between transaction start & i_mutex
615 Details : start the transaction earlier in llog_lvfs_destroy to get
616 transaction start and inode mutex lock nested properly.
620 Description: lru resize SLV can get stuck
621 Details : calculate SLV with a greater precision to not lose small
622 changes due to interger math truncation; round up SLV only
623 if the amount of granted locks less than the limit to not
624 get stuck with this SLV
628 Description: avoid divide-by-zero in lprocfs_rd_import()
632 Description: lfs quota failed when OSTs are down
633 Details : really return approximate block/inode usage when OSTs are down
637 Description: llobdstat fix and enhancement
638 Details : add a counter to set a limit to how many samples will be returned
639 fix a wildcard in the path to limit to obdfilter stats only
643 Description: recovery-small 51 hang - MDT can not stop, Mount still busy with
645 Details : abort lock enqueue processing sooner when we umount is in progress.
647 Severity : enhancement
649 Description: Update RHEL5.4 kernel to 2.6.18-164.11.1.el5 and OEL5.4 kernel to
650 2.6.18-164.11.1.0.1.el5.
652 Severity : enhancement
654 Description: error message improvements
655 Details : Use INFO/WARN instead of WARN/ERROR for the slow messages.
656 Simplify MDT/OST service start message.
657 Suppress "changing the import ..." warning.
660 Bugzilla : 21961/17914
661 Description: ignore trailing -mdc when determining index number
664 Frequency : only with SLES11
666 Description: fix for a race condition in linux quotas implementation
667 Details : dq_flags(struct dquot) access is not properly locked which could
668 lead to certain inconsistencies when accessing it using non-atomic
669 bit operations like __set_bit in do_set_dqblk.
673 Description: return any pending async errors in close(2) using flush callback
675 -------------------------------------------------------------------------------
677 2010-01-29 Sun Microsystems, Inc.
679 * Support for kernels:
680 2.6.16.60-0.42.8 (SLES 10),
681 2.6.27.39-0.3.1 (SLES11),
682 2.6.18-164.11.1.el5 (RHEL 5)
683 2.6.18-164.6.1.0.1.el5 (OEL 5)
684 * Client support for unpatched kernels:
685 (see http://wiki.lustre.org/index.php?title=Patchless_Client)
686 2.6.16 - 2.6.30 vanilla (kernel.org)
687 * Recommended e2fsprogs version: 1.41.6.sun1
688 * The async journal commit feature (bug 19128) and the cancel
689 lock before replay feature (bug 16774) are disabled by default.
691 Severity : enhancement
693 Description: Update RHEL5.4 kernel to 2.6.18-164.11.1.el5 and
694 OEL5.4 kernel to 2.6.18-164.11.1.0.1.el5.
696 Severity : enhancement
697 Bugzilla : 21511/19848
698 Description: Update SLES11 kernel to 2.6.27.39-0.3.1.
700 Severity : enhancement
702 Description: Update supported SLES10 kernel to 2.6.16.60-0.42.8.
704 Severity : enhancement
706 Description: Update kernel to RHEL5.4 2.6.18-164.6.1.el5 and
707 OEL5 2.6.18-164.6.1.0.1.el5(Both in-kernel OFED enabled).
709 Severity : enhancement
711 Description: Build kernels (RHEL5, OEL5 and SLES10/11) using the vendor's own
714 Severity : enhancement
716 Description: Vanilla kernel 2.6.30 patchless client support.
721 Description: bad entry in directory xxx: inode out of bounds
722 Details : fix locking issue in the rename path which could race with any
723 other operations updating the same directory.
725 Severity : enhancement
727 Description: Make watchdog timer messages to be more clear and descriptive.
731 Description: cp -p command does not preserve the dates and timestamp
732 Details : mtime could be spoiled by a write callback
736 Description: Clear imp_force_reconnect correctly in ptlrpc_connect_interpret()
740 Description: Allow non-root access for "lfs check".
741 Details : Added a check in obd_class_ioctl() for OBD_IOC_PING_TARGET.
743 Severity : enhancement
745 Description: quotacheck performance/scaling issues
746 Details : reduce quotacheck time on empty filesystem by skipping uninit
749 Severity : enhancement
751 Description: Enhancement for lfs(1) command to use numeric uid/gid.
753 Severity : enhancement
755 Description: Adjust locks' extents on their first enqueue, so that at the time
756 they get granted, there is no need for another pass through the
757 queues since they are already shaped into the proper forms.
761 Description: Fix mds_shrink_intent_reply()/mds_intent_policy() to pass correct
762 arguments and prevent LBUG() in lustre_shrink_reply_v2().
766 Description: Change tunefs.lustre and mkfs.lustre --mountfsoptions so that
767 exactly the specified mount options are used. Leaving off
768 any "mandatory" mount options is an error. Leaving off any
769 default mount options causes a warning, but is allowed.
770 Change errors=remount-ro from mandatory to default. Sanitize
771 the mount string before storing it. Update man pages accordingly.
775 Description: mds_getattr() should return 0, even if mds_fid2entry() fails with
776 -ENOENT. Also fix in ptlrpc_expire_one_request() to print signed
779 Severity : enhancement
781 Description: Remove set_info(KEY_UNLINKED) from MDS/OSC
783 Severity : enhancement
785 Description: Clients can replay thousands of unused locks during recovery
786 Details : Don't replay unused locks (only read locks for now) during
787 recovery. This feature is disabled by default and can be
788 enabled by running the following command on the clients:
789 lctl get_param ldlm.cancel_unused_locks_before_replay
793 Description: can't stat file in some situation.
794 Details : improve initialize osc date when target is added to mds and
795 ability to resend too big getattr request is client isn't have info
800 Description: Prevent inconsistences between linux and lustre mount structures.
801 Details : Wait indefinitely in server_wait_finished() until mnt_count drops.
802 Make the sleep interruptible.
804 Severity : enhancement
806 Description: Communicate OST degraded/readonly state via statfs to MDS
807 Details : Flags in the statfs returned from OSTs indicate whether the
808 OST is in a degraded RAID state, or if the filesystem has
809 turned read-only after a filesystem error is detected.
814 Description: don't panic if EPROTO was hit when reading symlink
815 Details : correctly handling request reference in error cases.
820 Description: open sometimes returns ENOENT instead of EACCES
821 Details : checking permission should be part of open part of mds_open, not
822 lookup part. so server should be set DISP_OPEN_OPEN disposition
823 before starting permission check.
824 Also not need revalidate dentry if client already have LOOKUP lock.
828 Description: enable client interface failover
829 Frequency : on servers with multiple network interfaces
830 Details : When a child reconnects from another NID, properly update export
831 nid hash position and ldlm reverse import.
833 Severity : enhancement
835 Description: implemented direct I/O with arbitrary (nonaligned) memory
836 addresses and file offsets.
838 Severity : enhancement
840 Description: added more recovery timeout options.
842 Severity : enhancement
844 Description: added llapi_file_open, llapi_file_create, llapi_file_get_stripe
849 Description: Avoid deadlock for local client writes
850 Frequency : only on systems with clients writing to an OST on the same node
851 Details : Use new OBD_BRW_MEMALLOC flag to notify OST about writes in the
852 memory freeing context. This allows OST threads to set the
853 PF_MEMALLOC flag on task structures in order to allocate memory
854 from reserved pools and complete IO.
855 Use GFP_HIGHUSER for OST allocations for non-local client writes,
856 so that the OST threads generate memory pressure and allow
857 inactive pages to be reclaimed.
862 Descriptoin: lock ordering violation between &cli->cl_sem and _lprocfs_lock
863 Details : move ldlm namespace creation in setup phase to avoid grab
864 _lprocfs_lock with cli_sem held.
868 Frequency : only during format of test systems
869 Description: Unable to run several mkfs.lustre on loop devices at the same time
870 Details : mkfs.lustre returns error 256 on the concurrent loop devices
871 formatting. The solution is to proper handle the error.
873 Severity : enhancement
875 Description: implement async create (obd_async_create) method for osc, to avoid
876 too long waiting new ost objects with holding ldlm lock.
880 Frequency : occasionally during network problems
881 Description: client not allowed to reconnect to OST because of active request
882 Details : abort bulk requests received by the OST once the client has timed
883 out since the client will resend the request anyway. The client
884 also now retries to reconnect to the same server if a connect
885 request failed with EBUSY or -EAGAIN.
888 Frequency : rare, if used wide striped file and one ost in down.
890 Descriptoin: don't return error if we created a subset of objects for file.
891 Details : lov_update_create_set() uses set->set_success as index for created
892 objects, so if some requests failed, they will have hole at end of
893 array and we can use qos_shrink_lsm for allocate correct lsm.
897 Description: Slow stale export processing during normal start up
898 Details : The global mgc lock prevents OST setup to be run in parallel.
899 Replace the global lock with a per-config_llog_data semaphore.
903 Description: Out or order replies might be lost on replay
904 Details : In ptlrpc_retain_replayable_request if we cannot find retained
905 request with tid smaller then one currently being added, add it
906 to the start, not end of the list.
910 Description: BUG: soft lockup - CPU#1 stuck for 10s! [ll_mdt_07:4523]
911 Details : add cond_resched() calls to avoid hogging the cpu for too long
912 in the hash code. Make also lustre_hash_for_each_empty() more
915 Severity : enhancement
917 Description: Performance improvements for debug messages with D_RPCTRACE,
918 D_LDLM, D_QUOTA options.
921 Frequency : only with NFS export
923 Description: (lov_merge.c:74:lov_merge_lvb())
924 ASSERTION(spin_is_locked(&lsm->lsm_lock)) failed (SR 71691004)
925 Details : Fix a race in the nfs export code by populating inode
926 info while the new inode is still locked
928 Severity : enhancement
930 Description: add a new file in procfs called force_lbug. Writting to this
931 file triggers a LBUG. Only for test purpose.
935 Description: OOM killer causes node hang
936 Details : really interrupt the sleep in osc_enter_cache on signals
940 Description: LustreError: 9153:0:(quota_context.c:622:dqacq_completion()) LBUG
941 Details : fix race during quota release on the slave.
943 Severity : enhancement
945 Description: smaller hash bucket sizes, cleanups
946 Details : increase hash table sizes and enabled rehashing for pools, quota,
947 uuid, nid & per-nid stats.
949 Severity : enhancement
951 Description: Add ldiskfs maxdirsize mount option
952 Details : add max_dir size mount option
956 Description: panic in ll_statahead_thread
957 Details : prevent parent thread to be killed before its child
960 Frequency : only with 16TB device
962 Description: unable to perform "mount -t lustre" of 16TB OST device
963 Details : Mounting 16TB LUNs failed due to three bugs in mkfs.lustre.
967 Description: ASSERTION(atomic_read(&imp->imp_inflight) == 0) failed
968 Details : unregistering should be zero if no RPC inflight.
972 Description: hyperion: Oops during metabench
973 Details : Correct the refcount of lov_request_set
975 Severity : enhancement
977 Description: Add mptlinux and nxge drivers to Lustre builds
979 Severity : enhancement
981 Description: Fix watchdog timer message to be more clear
982 Details : Make watchdog timer messages more clear and descriptive.
986 Description: LNET soft lockups in socknal_cd thread
987 Details : don't hog CPU for active-connecting if another connd is
988 accepting connecting-requst from the same peer
992 Description: recovery-small test_17 hang
993 Details : Land several AT improvements & fixes.
997 Description: MDS panic and hanging client processes
998 Details : Replace exp_ops_stats with exp_nid_stats->nid_stats
1002 Description: OSS stuck in recovery.
1003 Details : fix race during recovery. class_unlink_export,
1004 class_set_export_delayed and target_queue_last_replay_reply
1005 may race while increasing/decreasing obd_recoverable_clients
1006 and obd_delayed_clients, causing recovery to wait forever.
1008 Severity : enhancement
1010 Description: add cascading_rw.c to lustre/tests
1014 Description: filter_last_id() NULL deref
1015 Details : lprocfs_filter_rd_last_id() should check for the fully
1016 setup obd device, before proceeding further.
1018 Severity : enhancement
1020 Description: Loadgen improvements
1021 Details : stacksize and locking fixes for loadgen
1025 Description: Quiet CERROR("dirty %d > system dirty_max %d\n"
1026 Details : The atomic_read() allowing the atomic_inc() are not covered
1027 by a lock. Thus they may safely race and trip this CERROR()
1028 unless we add in a small fudge factor (+1).
1030 Severity : enhancement
1032 Description: shrink_slab: nr=-9223362083340912175
1033 Details : fix spurious message from shrink_slab reporing negative nr
1037 Description: Quiet bogus previously committed transno error
1038 Details : suppress the "server went back in time" error message which
1039 is always printed even in the common case after a client eviction
1041 Severity : enhancement
1043 Description: Parallel statfs() calls result in client eviction
1044 Details : cache statfs data for 1s.
1048 Description: parallel-scale test_compilebench: @@@@@@ FAIL: compilebench
1050 Details : fix serveral issues in pinger code causing clients not to ping
1051 servers for too long, resulting in evictions.
1055 Description: e2fsck should warn when MMP update interval is extended
1056 Details : print mmp_check_interval and make it possible to abort mount
1057 operation in case it takes too long.
1061 Description: mdsrate-create-large.sh, BUG: soft lockup - CPU#0 stuck for 10s!
1062 Details : fix bug in the RHEL5's jbd2 callback patch.
1066 Description: drop number of active requests when queued for recovery
1067 Details : Now that we take a reference on the original request instead of
1068 making a copy of it for recovery. We need to drop the number of
1069 active requests or the queued requests will prevent all request
1070 processing when they exceed (srv->srv_threads_running - 1).
1072 Severity : enhancement
1074 Description: refuse to invalidate operational quota files when they are in use
1075 Details : an attempt to invalidate operational quota files on the quota
1076 master is not actually permitted by VFS (returning -EPERM), but we
1077 should not depend on that and should return the error earlier.
1081 Description: Applications stuck in jbd2_log_wait_commit during exit
1082 Details : fix deadlock between kjournald2 trying to acquire the page lock
1083 owned by an ost_io thread waiting for journal commit.
1085 -------------------------------------------------------------------------------
1087 2009-10-16 Sun Microsystems, Inc.
1089 * Support for kernels:
1090 2.6.16.60-0.42.4 (SLES 10),
1091 2.6.27.29-0.1 (SLES11, i686 & x84_64 only),
1092 2.6.18-128.7.1.el5 (RHEL 5),
1093 2.6.18-128.7.1.el5 (OEL 5),
1094 * Client support for unpatched kernels:
1095 (see http://wiki.lustre.org/index.php?title=Patchless_Client)
1096 2.6.16 - 2.6.27 vanilla (kernel.org)
1097 * Recommended e2fsprogs version: 1.41.6.sun1
1098 * ext4 support for RHEL5 is experimental and thus should not be
1101 Severity : enhancement
1103 Description: Add OEL5 support.
1105 Severity : enhancement
1107 Description: Update kernel to SLES11 2.6.27.29-0.1.
1111 Description: File checksum failures with OST read cache on
1112 Details : Disable page poisoning when the bulk transfer has to be aborted
1113 because the client got evicted.
1117 Description: Don't allow make backward step on assiging osc next id.
1118 Details : race between allocation next id and ll_sync thread can be cause
1119 of set wrong osc next id and can be kill valid ost objects.
1121 Severity : enhancement
1123 Description: Update kernel to RHEL5 2.6.18-128.7.1.el5.
1125 Severity : enhancement
1127 Description: Update kernel to SLES10 SP2 2.6.16.60-0.42.4.
1131 Description: Changes in raid5-large-io-rhel5.patch to calculate sectors properly
1135 Description: Increase the default BLK_DEF_MAX_SECTORS value for RHEL5 and SLES11
1139 Description: Error handling in osc_statfs_interpret() has been improved.
1140 Details : Check in osc_statfs_interpret() for EBADR.
1144 Description: Do not send statfs() requests to OSTs disabled by administrator.
1145 Details : Check in lov_prep_statfs_set() for non-NULL ltd_exp.
1149 Description: Do not update ctime for the deleted inode.
1150 Details : Check in mds_reint_unlink() before calling fsfilt_setattr().
1154 Description: Increase of the size of the LDLM resource hash.
1155 Details : Bump up RES_HASH_BITS=12.
1159 Description: correctly send lsm on open replay
1160 Details : MDS is trust to LSM size on replay open, but client can set wrong size
1165 Description: Deadlock between filter_destroy() and filter_commitrw_write().
1166 Details : filter_destroy() does not hold the DLM lock over the whole
1167 operation. If the DLM lock is dropped, filter_commitrw() can go
1168 through, causing the deadlock between page lock and i_mutex.
1169 The i_alloc_sem should also be hold in filter_destroy() while
1170 truncating the file.
1174 Description: truncate starts GFP_FS allocation under transaction causing deadlock
1175 Details : ldiskfs_truncate calls grab_cache_page which may start page
1176 allocation under an open transaction. This may lead to
1177 calling prune_icache with consequent lustre reentrance.
1180 Frequency : only when down/upgrading the MDS to 1.6/1.8 while 1.8 clients are
1181 still up and when the OST pool feature is used
1183 Description: interop testing got LBUG when run dd with OST pool
1184 :LustreError: 30032:0:(llite_lib.c:1913:ll_replace_lsm()) LBUG
1185 Details : down/upgrading the MDS to a version that doesn't/does support OST
1186 pool can cause clients to crash because the lsm has changed
1191 Description: missing tree_status on 1.8.1 RPM build
1192 Details : make rpms failed due because the tree_status file is missing.
1196 Description: continuing LustreError "mds adjust qunit failed!"
1197 Details : don't print message on the console when ->adjust_qunit fails.
1201 Description: don't increase ldlm timeout if previous client was evicted
1202 Details : if a client doesn't respond to a blocking callback within the
1203 adaptive ldlm enqueue timeout, don't adjust the adaptive estimate
1204 when the lock is next granted.
1208 Description: ost is being unmounted w/o all writes to last_rcvd landing on disk.
1209 affects recovery negatively.
1210 Details : make sure all exports have been properly destroyed by the zombie
1211 thread processed before stopping the target.
1215 Description: Performance degradation with O_DIRECT between 1.6 & 1.8.1 b190
1216 Details : disable write barrier for ext4/SLES11.
1220 Description: Kernel panic - not syncing: Out of memory and no killable
1221 processes... on OSS when iozone
1222 Details : fix memory leak in the journal checksum patch.
1226 Description: group quota "too many blocks" OSS crashes
1227 Details : we should keep the same uid/gid for lquota_chkquota() and
1228 lquota_pending_commit()
1232 Description: LustreError: 9153:0:(quota_context.c:622:dqacq_completion()) LBUG
1233 Details : don't LBUG on release quota error. Just a workaround until the
1234 problem is understood.
1236 ------------------------------------------------------------------------------
1237 2009-07-31 Sun Microsystems, Inc.
1239 * Support for kernels:
1240 2.6.16.60-0.39.3 (SLES 10),
1241 2.6.27.23-0.1 (SLES11, i686 & x84_64 only),
1242 2.6.18-128.1.14.el5 (RHEL 5),
1243 * Client support for unpatched kernels:
1244 (see http://wiki.lustre.org/index.php?title=Patchless_Client)
1245 2.6.16 - 2.6.27 vanilla (kernel.org)
1246 * Recommended e2fsprogs version: 1.41.6.sun1
1247 * File join has been disabled in this release, refer to Bugzilla 16929.
1248 * NFS export disabled when stack size < 8192. Since the NFSv4 export
1249 of Lustre filesystem with 4K stack may cause a stack overflow. For
1250 more information, please refer to bugzilla 17630.
1251 * ext4 support for RHEL5 is experimental and thus should not be
1254 Severity : enhancement
1256 Description: Update kernel to SLES10 SP2 2.6.16.60-0.39.3.
1259 Frequency : with 1.8 server and 1.6 clients
1261 Description: correctly shrink reply for avoid send too big message to client.
1262 Details : 1.8 mds is allocate to big buffer to LOV EA data and this produce
1263 some problems with sending this reply to 1.6 client.
1267 Description: Repeated atomic allocation failures.
1268 Details : Use GFP_HIGHUSER | __GFP_NOMEMALLOC flags for memory allocations
1269 to generate memory pressure and allow reclaiming of inactive pages.
1270 At the same time, do not allow to exhaust emergency pools.
1271 For local clients the use of GFP_NOFS will be introduced in 1.8.2
1273 Severity : enhancement
1274 Bugzilla : 19846, 18289
1275 Description: Update kernel to RHEL5 2.6.18-128.1.14.el5.
1277 Severity : enhancement
1278 Bugzilla : 19625, 16893, 18668, 19848
1279 Description: Add support for SLES11 2.6.27.23-0.1.
1281 Severity : enhancement
1283 Description: Update client support to vanila kernels up to 2.6.27.
1285 Severity : enhancement
1287 Description: Update kernel to SLES10 SP2 2.6.16.60-0.37.
1289 Severity : enhancement
1291 Description: Compile with -Werror by default for i686 and x86_64.
1295 Description: resolve race between obd_disconnect and class_disconnect_exports
1296 Details : if obd_disconnect will be called to already disconnected export he
1297 forget release one reference and osc module can't unloaded.
1299 Severity : enhancement
1301 Description: move AT tunable parameters for more consistent usage
1302 Details : add AT tunables under /proc/sys/lustre, add to conf_param parsing
1306 Descriptoin: correctly skip time estimate if in recovery
1307 Details : rq_send_state insn't bitmask so using bitwise ops is forbid.
1311 Descriptoin: OSS DeadLock
1312 Details : Use trylock to prevent deadlock when shrink icache.
1314 Severity : enhancement
1316 Description: Allow tuning service thread via /proc
1317 Details : For each service a new
1318 /proc/fs/lustre/{service}/*/thread_{min,max,started} entry is
1319 created that can be used to set min/max thread counts, and get the
1320 current number of running threads.
1322 Severity : enhancement
1324 Description: Add state history info file, enhance import info file
1325 Details : Track import connection state changes in a new osc/mdc proc file;
1326 add overview-type data to the osc/mdc import proc file.
1330 Description: Reduce small size read RPC
1331 Details : Set read-ahead limite for every file and only do read-ahead when
1332 available read-ahead pages are bigger than 1M to avoid small size
1337 Description: free_entry erroneously used groups_free instead of put_group_info
1339 Severity : enhancement
1341 Description: Make read-ahead stripe size aligned.
1343 Severity : enhancement
1345 Description: MDS create should not wait for statfs RPC while holding DLM lock.
1348 Frequency : rare, connect and disconnect target at same time
1350 Descriptoin: ASSERTION(atomic_read(&imp->imp_inflight) == 0
1351 Details : don't call obd_disconnect under lov_lock. this long time
1352 operation and can block ptlrpcd which answer to connect request.
1355 Frequency : start MDS on uncleanly shutdowned MDS device
1357 Descriptoin: ll_sync thread stay in waiting mds<>ost recovery finished
1358 Details : stay in waiting mds<>ost recovery finished produce random bugs
1359 due race between two ll_sync thread for one lov target. send
1360 ACTIVATE event only if connect realy finished and import have
1364 Frequency : start MDS on uncleanly shutdowned MDS device
1366 Descriptoin: aborting recovery hang on MDS
1367 Details : don't throttle destroy RPCs for the MDT.
1371 Description: Slow reads beyond 8Tb offsets.
1372 Details : Page index integer overflow in ll_read_ahead_page
1376 Description: Soft lockup on OSS after MDS failover
1377 Details : MSG_CONNECT_INITIAL is not set on the initial MDS->OST connect.
1378 As a conseqence, the patch from bug 18224 is not operational
1379 and the MDS export cannot be reused on the OSTs until it gets
1383 Frequency : rare, only if using MMP with Linux RAID
1385 Description: MMP doesn't work with Linux RAID
1386 Details : While using HA for Lustre servers with Linux RAID, it is possible
1387 that MMP will not detect multiple mounts. To make this work we
1388 need to unplug the device queue in RAID when the MMP block is being
1389 written. Also while reading the MMP block, we should read it from
1390 disk and not the cached one.
1393 Frequency : rare, during recovery
1395 Description: Assertion failure in ldlm_lock_put
1396 Details : Do not put cancelled locks into replay list, hold references on
1397 locks in replay list
1401 Description: 1.6.5 mdsrate performance is slower than 1.4.11/12 (MDS is not cpu bound!)
1402 Details : create_count always drops to the min value (=32) because grow_count
1403 is being changed before the precreate RPC completes.
1406 Frequency : Only in RHEL5 when mounting multiple ext3 filesystems
1409 Description: "kmem_cache_create: duplicate cache jbd_4k" error message
1410 Details : add proper locking for creation of jbd_4k slab cache
1414 Description: MMP check in ext3_remount() fails without displaying any error
1415 Details : When multiple mount protection fails during remount, proper error
1420 Description: Rare Client crash on resend if the file was deleted.
1421 Details : When file is opened, but open reply is lost and file is
1422 subsequently deleted before resend, resend processing logic
1423 breaks trying to open the file again, should not try to open.
1427 Description: add check for >8TB ldiskfs filesystems
1428 Details : ext3-based ldiskfs does not support greater than 8TB LUNs.
1429 Don't allow >8TB ldiskfs filesystems to be mounted without
1430 force_over_8tb mount option
1434 Description: Client locked up when running multiple instances of an app. on
1435 multiple mount points
1436 Details : ll_shrink_cache() can sleep while holding the ll_sb_lock.
1437 Convert ll_sb_lock to a read/write semaphore to fix the problem.
1441 Description: Cannot acces an NFS-mounted Lustre filesystem
1442 Details : An NFS client cannot access the Lustre filesystem NFS-mounted
1443 from a Lustre-client exporting the Lustre filesystem via NFS.
1447 Description: panic in ll_statahead_thread
1448 Details : grab dentry reference in parent process.
1451 -------------------------------------------------------------------------------
1453 tbd Sun Microsystems, Inc.
1455 * Support for kernels:
1456 2.6.16.60-0.31 (SLES 10),
1457 2.6.18-128.1.6.el5 (RHEL 5),
1458 2.6.22.14 vanilla (kernel.org)
1459 * Client support for unpatched kernels:
1460 (see http://wiki.lustre.org/index.php?title=Patchless_Client)
1461 2.6.16 - 2.6.22 vanilla (kernel.org)
1462 * Recommended e2fsprogs version: 1.40.11-sun1
1463 * File join has been disabled, refer to Bugzilla 16929.
1464 * A new Lustre ADIO driver is available for MPICH2-1.0.7.
1465 * NFS export disabled when stack size < 8192. Since the NFSv4 export of
1466 Lustre filesystem with 4K stack may cause a stack overflow. For more
1467 information, please refer to bugzilla 17630.
1469 Severity : enhancement
1471 Description: Update to RHEL5.3 kernel-2.6.18-128.1.6.el5.
1473 Severity : enhancement
1475 Description: Update OFED release to 1.4.1 RC4
1477 Severity : major, only with big OST
1479 Description: Very poor metadata performance on Infiniband lustre configuration
1480 Details : OST object precreation becomes very slow on big OSTs. This is due
1481 to the ialloc patch spending too much time scanning groups.
1484 Frequency : during recovery
1486 Description: don't mix llog inodes with normal.
1487 Details : allocate inodes for log in last inode group
1492 Description: fix lqs' reference which won't be put in some situations
1493 Details : This patch fixes:
1494 1. In quota_check_common(), this function will check quota
1495 for user and group, but only send one return via "pending".
1496 In most cases, the pendings should be same. But that is not
1498 2. If quotaoff runs between lquota_chkquota() and
1499 lquota_pending_commit(), the same thing will happen too.
1500 That is why it comes:
1501 - if (!ll_sb_any_quota_active(qctxt->lqc_sb))
1504 -------------------------------------------------------------------------------
1506 2008-12-31 Sun Microsystems, Inc.
1508 * Support for kernels:
1509 2.6.5-7.314 (SLES 9),
1510 2.6.9-67.0.22.EL (RHEL 4),
1511 2.6.16.60-0.31 (SLES 10),
1512 2.6.18-92.1.17.el5 (RHEL 5),
1513 2.6.22.14 vanilla (kernel.org)
1514 * Client support for unpatched kernels:
1515 (see http://wiki.lustre.org/index.php?title=Patchless_Client)
1516 2.6.16 - 2.6.22 vanilla (kernel.org)
1517 * Client support for unpatched kernels:
1518 we do not recommend using patchless RHEL4 clients with kernels
1519 prior to 2.6.9-55EL (RHEL4U5).
1520 * Recommended e2fsprogs version: 1.40.11-sun1
1521 * Note that reiserfs quotas are disabled on SLES 10 in this kernel.
1522 * RHEL 4 and RHEL 5/SLES 10 clients behaves differently on 'cd' to a
1523 removed cwd "./" (refer to Bugzilla 14399).
1524 * A new quota file format has been introduced in 1.6.5.
1525 The format conversion from prior releases is handled transparently,
1526 but releases older than 1.4.12/1.6.5 don't understand this new
1527 format. The automatic format conversion can be avoided by running
1528 the following command on the MDS:
1529 'tunefs.lustre --param="mdt.quota_type=ug1" $MDTDEV'.
1530 For more information, please refer to bugzilla 13904.
1531 * A new quota file format was introduced in 1.6.6/1.8.0 (kernels 2.6.16+).
1532 The format conversion from prior releases is handled transparently,
1533 but releases older than 1.6.6/1.8.0 don't understand this new
1534 format. The automatic format conversion can be avoided by running
1535 the following commands on the MDS and OSS servers (for
1536 pre 1.4.12-1.6.5 quota files):
1537 'tunefs.lustre --param="mdt.quota_type=ug1" $MDTDEV',
1538 'tunefs.lustre --param="ost.quota_type=ug1" $MDTDEV'
1539 or (for 1.4.12/1.6.5 quota files)
1540 'tunefs.lustre --param="mdt.quota_type=ug2" $MDTDEV',
1541 'tunefs.lustre --param="ost.quota_type=ug2" $MDTDEV'
1542 For more information, please refer to bugzilla 13904.
1543 * Output of lfs quota has been made less detailed by default,
1544 old (verbose) output can be obtained by using -v option.
1545 * File join has been disabled in this release, refer to Bugzilla 16929.
1546 * A new Lustre ADIO driver is available for MPICH2-1.0.7.
1547 * NFS export disabled when stack size < 8192. Since the NFSv4 export of
1548 Lustre filesystem with 4K stack may cause a stack overflow. For more
1549 information, please refer to bugzilla 17630.
1551 Severity : enhancement
1553 Description: Caching OSS
1554 Details : introduce data caching on the OSS. The OSS now relies on the linux
1555 kernel page cache to keep recently accessed data in memory.
1556 It is worth noting that all write requests are still flushed
1557 synchronously as in lustre 1.6.
1559 Severity : enhancement
1561 Description: Version based recovery
1562 Details : introduce finer grained recovery able to detect transaction
1563 dependencies and can deal with transaction gaps caused by clients
1564 failing at the same time as the server.
1566 Severity : enhancement
1568 Description: Enable adaptive timeouts by default
1569 Details : The Lustre timeout value in /proc/sys/lustre/timeout is now
1570 managed dynamically based on server load and should not need
1571 to be tuned manually based on cluster size. This allows Lustre
1572 to work under a wider variety of system sizes and loads, without
1573 unnecessarily causing lengthy recovery times.
1575 Severity : enhancement
1577 Description: Add OST Pools support
1578 Details : File striping can now be set to use an arbitrary pool of OSTs
1580 Severity : enhancement
1582 Description: add lazystatfs mount option to allow statfs(2) to skip down OSTs
1583 Details : allow skip disconnected ost for send statfs request and hide error
1587 Frequency : rare, on llog test 6
1589 Descriptoin: don't allow connect to already connected import
1590 Details : allowing connect to already connected import is hide connecting problem.
1593 Frequency : rare, on failed llog setup
1595 Descriptoin: don't leak obd reference on failed llog setup
1596 Details : for failed llog setup - mgc forget call class_destroy_import for
1597 client import, move destroy import to more generic place.
1602 Descriptoin: allow kill process which wait statahead result
1603 Details : for some reasons 'ls' can stick in waiting result from statahead,
1604 in this case need way for kill this process.
1607 Frequency : rare, at shutdown
1609 Descriptoin: panic at umount
1610 Details : llap_shrinker can be raced with killing super block from list and
1611 this produce panic with access to already freeded pointer
1616 Descriptoin: don't lose wakeup for imp_recovery_waitq
1617 Details : recover_import_no_retry or invalidate_import and import_close can
1618 both sleep on imp_recovery_waitq, but we was send only one wakeup
1624 Descriptoin: panic in mds_open
1625 Details : don't confuse mds_finish_transno() with PTR_ERR(-ENOENT)
1630 Descriptoin: stuck in cache_remove_extent() or panic with accessing to already
1632 Details : release lock refernce only after add page to pages list.
1635 Frequency : always with long access acl
1637 Descriptoin: mds can't pack reply with long acl.
1638 Details : mds don't control size of acl but they limited by reint/getattr
1643 Frequency : on remount
1645 Description: external journal device not working after the remount
1646 Details : clear dev_rdonly flag for external journal devices in
1652 Description: shutdown vs evict race
1653 Details : client_disconnect_export vs connect request race.
1654 if client will evicted at this time - we start invalidate
1655 thread without referece to import and import can be freed
1661 Description: shrink LOV EAs before replying
1662 Details : correctly adjust LOV EA buffer for reply.
1667 Description: don't skip ost target if they assigned to file
1668 Details : Drop slow OSCs if we can, but not for requested start idx.
1669 This means "if OSC is slow and it is not the requested
1670 start OST, then it can be skipped, otherwise skip it only
1671 if it is inactive/recovering/out-of-space.
1673 Severity : enhancement
1675 Description: Update to RHEL5 kernel-2.6.18-92.1.17.el5.
1678 Frequency : rare, need acl's on inode.
1680 Description: client can't handle ost additional correctly
1681 Details : if ost was added after client connected to mds client can have
1682 hit lnet_try_match_md ... to big messages to wide striped files.
1683 in this case need teach client to handle config events about add
1684 lov target and update client max ea size at that event.
1686 Severity : enhancement
1688 Description: Update to sles9 kernel-2.6.5-7.314.
1690 Severity : enhancement
1692 Description: Update to SLES10 SP2 kernel-2.6.16.60-0.31.
1695 Frequency : Create a symlink file with a very long name
1697 Description: ldlm_cancel_pack()) ASSERTION(max >= dlm->lock_count + count)
1698 Details : If there is no extra space in the request for early cancels,
1699 ldlm_req_handles_avail() returns 0 instead of a negative value.
1704 Description: mds is deadlocked
1705 Details : in rare cases, inode in catalog can have i_no less than have parent
1706 i_no, this produce wrong order for locking during open, and
1707 parallel unlink can be lock open. this need teach mds_open to grab
1708 locks in resource id order, not at parent -> child order.
1710 Severity : enhancement
1712 Description: Add /proc entry for import status
1713 Details : The mdc, osc, and mgc import directories now have
1714 an import directory that contains useful import data for debugging
1715 connection problems.
1717 Severity : enhancement
1719 Description: Re-disable certain /proc logging
1720 Details : Enable and disable client's offset_stats, extents_stats and
1721 extents_stats_per_process stats logging on the fly.
1724 Frequency : Only on FC kernels 2.6.22+
1726 Description: oops in statahead
1727 Details : Do not drop reference count for the dentry from VFS when lookup,
1728 VFS will do that by itself.
1730 Severity : enhancement
1732 Description: Generic /proc file permissions
1733 Details : Set /Proc file permissions in a more generic way to enable non-
1734 root users operate on some /proc files.
1738 Description: Hitting mdc_commit_close() ASSERTION
1739 Details : Properly handle request reference release in
1740 ll_release_openhandle().
1744 Frequency : only patchless client
1745 Description: add workaround for race between add/remove dentry from hash
1747 Severity : enhancement
1749 Description: Allow OST glimpses to return PW locks
1753 Description: LBUG when llog conf file is full
1754 Details : When llog bitmap is full, ENOSPC should be returned for plain
1759 Description: Prevent import from entering FULL state when server in recovery
1763 Description: service mount cannot take device name with ":"
1764 Details : Only when device name contains ":/" will mount treat it as
1770 Description: replace ptlrpcd with the statahead thread to interpret the async
1771 statahead RPC callback
1775 Frequency : on recovery
1776 Description: I/O failures after umount during fail back
1777 Details : if client reconnected to restarted server we need join to recovery
1778 instead of find server handler is changed and process self eviction
1779 with cancel all locks.
1781 Severity : enhancement
1783 Description: Update to RHEL5 kernel-2.6.18-92.1.10.el5.
1787 Description: Kernel BUG tries to release flock
1788 Details : Lustre does not destroy flock lock before last reference goes
1789 away. So always drop flock locks when client is evicted and
1790 perform unlock regardless of successfulness of speaking to MDS.
1792 Severity : enhancement
1794 Description: Update to SLES10 SP2 kernel-2.6.16.60-0.27.
1796 Severity : enhancement
1798 Description: Upcall on Lustre log has been dumped
1799 Details : Allow for a user mode script to be called once a Lustre log has
1800 been dumped. It passes the filename of the dumped log to the
1801 script, the location of the script can be specified via
1802 /proc/sys/lnet/debug_log_upcall.
1807 Description: avoid messages about idr_remove called for id that is not allocated
1808 Details : Move assigment s_dev for clustered nfs to end of initialization,
1809 for avoid problem with error handling.
1814 Description: avoid Already found the key in hash [CONN_UNUSED_HASH] messages
1815 Details : When connection is reused this not moved from CONN_UNUSED_HASH into
1816 CONN_USED_HASH and this prodice warning when put connection again
1822 Description: avoid ASSERTION(client_stat->nid_exp_ref_count == 0) failed
1823 Details : release reference to stats when client disconnected, not
1824 when export destroyed for avoid races when client destroyed
1825 after main ost export.
1829 Description: more cleanup in mds_lov
1830 Details : add workaround for get valid ost count for avoid warnings about
1831 drop too big messages, not init llog cat under semphore which
1832 can be blocked on reconnect and break normal replay, fix access
1835 Severity : enhancement
1837 Description: Export bytes_read/bytes_write count on OSC/OST.
1841 Description: Early reply size mismatch, MGC loses connection
1842 Details : Apply the MGS_CONNECT_SUPPORTED mask at reconnect time so
1843 the connect flags are properly negotiated.
1847 Description: Properly propagate oinfo flags from lov to osc for statfs
1848 Details : restore missing copy oi_flags to lov requests.
1852 Description: exports in /proc are broken
1853 Details : recreate /proc entries for clients when they reconnect.
1855 Severity : enhancement
1857 Description: Add man pages for llobdstat(8), llstat(8), plot-llstat(8),
1858 : l_getgroups(8), lst(8), routerstat(8)
1859 Details : included man pages for llobdstat(8), llstat(8),
1860 : plot-llstat(8), l_getgroups(8), lst(8), routerstat(8)
1862 Severity : enhancement
1864 Description: Implement lustre ll_show_options method.
1866 Severity : enhancement
1868 Description: Update to SLES9 kernel-2.6.5-7.312.
1870 Severity : enhancement
1872 Description: Update to RHEL4 kernel-2.6.9-67.0.22.EL.
1876 Description: exports in /proc are broken
1877 Details : recreate /proc entries for clients when they reconnect.
1881 Description: don't fail open with -ERANGE
1882 Details : if client connected until mds will be know about real ost count
1883 get LOV EA can be fail because mds not allocate enougth buffer
1888 Description: Resolve device initialization race
1889 Details : Prevent proc handler from accessing devices added to the
1890 obd_devs array but yet be intialized.
1892 Severity : enhancement
1894 Description: configure's --enable-quota should check the
1895 : kernel .config for CONFIG_QUOTA
1896 Details : configure is terminated if --enable-quota is passed but
1897 : no quota support is in kernel
1899 Severity : enhancement
1901 Description: Update to SLES10 SP2 kernel-2.6.16.60-0.23.
1903 Severity : enhancement
1905 Description: Update to RHEL5 kernel-2.6.18-92.1.6.el5.
1909 Frequency : rare, on PPC clients
1910 Description: don't swab ost objects in response about directory, because
1912 Details : bug similar bug 14856, but in different function.
1914 Severity : enhancement
1916 Description: lfs quota tool enhancement
1917 Details : added units specifiers support for setquota, default to
1918 current uid/gid for quota report, short quota stats by
1919 default, nonpositional parameters for setquota, added
1920 llapi_quotactl manual page.
1922 Severity : enhancement
1924 Description: *optional* service tags registration
1925 Details : if the "service tags" package is installed on a Lustre node
1926 When the filesystem is mounted, a local-node service tag will
1927 be created. See http://inventory.sun.com/ for more information
1928 about the Service Tags asset management system.
1932 Description: Client runs out of low memory
1933 Details : Consider only lowmem when counting initial number of llap pages
1936 Frequency : occasional
1938 Description: add refcount for osc callbacks, so avoid panic on shutdown
1940 Severity : enhancement
1942 Description: Update to RHEL4 kernel-2.6.9-67.0.20.
1945 Frequency : testing only
1947 Description: sanity test 65a fails if stripecount of -1 is set
1948 Details : handle -1 striping on filesystem in ll_dirstripe_verify
1951 Frequency : only in unusual configurations
1953 Description: Kernel panic with find ost index.
1954 Details : lov_obd have panic if some OST's have sparse indexes.
1956 Severity : enhancement
1958 Description: Update to RHEL5 kernel-2.6.18-53.1.21.el5.
1961 Frequency : rarely, if filesystem is mounted with -o flock
1963 Description: do not process already freed flock
1964 Details : flock can possibly be freed by another thread before it reaches
1965 to ldlm_flock_completion_ast.
1968 Frequency : rarely, if filesystem is mounted with -o flock
1970 Description: LBUG during stress test
1971 Details : Need properly lock accesses the flock deadlock detection list.
1974 Frequency : rarely, if binaries are being run from Lustre
1976 Description: oops in page fault handler
1977 Details : kernel page fault handler can return two special 'pages' in
1978 error case, don't try dereference NOPAGE_SIGBUS and NOPAGE_OMM.
1981 Frequency : rarely, during shutdown
1983 Description: timeout with invalidate import.
1984 Details : ptlrpcd_check call obd_zombie_impexp_cull and wait request which
1985 should be handled by ptlrpcd. This produce long age waiting and
1986 -ETIMEOUT ptlrpc_invalidate_import and as result LASSERT.
1992 Description: ASSERTION(CheckWriteback(page,cmd)) failed
1993 Details : badly clear PG_Writeback bit in ll_ap_completion can produce false
1997 Frequency : only with broken builds/installations
1999 Description: no LBUG if lquota.ko and fsfilt_ldiskfs.ko are different versions
2000 Details : just return an error to a user, put a console error message
2002 Severity : enhancement
2004 Description: Update to RHEL5 kernel-2.6.18-53.1.19.el5.
2006 Severity : enhancement
2008 Description: Update to RHEL4 kernel-2.6.9-67.0.15.
2010 Severity : enhancement
2012 Description: enable MGS and MDT services start separately
2013 Details : add a 'nomgs' option in mount.lustre to enable start a MDT with
2014 a co-located MGS without starting the MGS, which is a complement
2015 to 'nosvc' mount option.
2018 Frequency : always, on big-endian systems
2020 Description: cleanup in ptlrpc code, related to PPC platform
2021 Details : store magic in native order avoid panic's in recovery on PPC
2022 node and forbid from this error in future. Also fix posibily
2023 of twice swab data. Fix get lov striping to userpace.
2026 Frequency : rarely, if replay get lost on server
2028 Description: server incorrectly drop resent replays lead to recovery failure.
2029 Details : do not drop replay according to msg flags, instead we check the
2030 per-export recovery request queue for duplication of transno.
2033 Frequency : after recovery
2035 Description: precreate to many object's after del orphan.
2036 Details : del orphan st in oscc last_id == next_id and this triger growing
2037 count of precreated objects. Set flag LOW to skip increase count
2038 of precreated objects.
2041 Frequency : rare, on clear nid stats
2043 Description: ASSERTION(client_stat->nid_exp_ref_count == 0)
2044 Details : when clean nid stats sometimes try destroy live entry,
2045 and this produce panic in free.
2048 Frequency : occasionally since 1.6.4
2050 Description: Stack overflow during MDS log replay
2051 Details : ease stack pressure by using a thread dealing llog_process.
2054 Frequency : very rare
2056 Description: MDT cannot be unmounted, reporting "Mount still busy"
2057 Details : Mountpoint references were being leaked during open reply
2058 reconstruction after an MDS restart. Drop mountpoint reference
2059 in reconstruct_open() and free dentry reference also.
2064 Description: wait until IO finished before start new when do lock cancel.
2065 Details : VM protocol want old IO finished before start new, in this case
2066 need wait until PG_writeback is cleared until check dirty flag
2067 and call writepages in lock cancel callback.
2072 Description: mds_mfd_close() ASSERTION(rc == 0)
2073 Details : In mds_mfd_close(), we need protect inode's writecount change
2074 within its orphan write semaphore to prevent possible races.
2077 Frequency : rare, on shutdown ost
2079 Description: don't hit live lock with umount ost.
2080 Details : shrink_dcache_parent can be in long loop with destroy dentries,
2081 use shrink_dcache_sb instead.
2084 Frequency : only when echo_client is used
2086 Description: don't panic with use echo_client
2087 Details : echo client pass NULL as client nid pointer and this produce NULL
2088 pointer dereference.
2091 Frequency : Always on 32-bit PowerPC systems
2093 Description: fix build on PPC32
2094 Details : compile code with -m64 flag produce wrong object file for PPC32.
2099 Description: MDS LBUG: ASSERTION(!IS_ERR(dchild))
2100 Details : In reconstruct_* functions, LASSERTs on both the data supplied
2101 by a client, and the data on disk are dangerous and incorrect.
2102 Change them with client eviction.
2104 Severity : enhancement
2106 Description: skiplist implementation simplification
2107 Details : skiplists are used to group compatible locks on granted list
2108 that was implemented as tracking first and last lock of each lock
2109 group the patch changes that to using doubly linked lists
2113 Description: delete compatibility for 32bit qdata
2114 Details : as planned, when lustre is beyond b1_8, lquota won't support 32bit
2115 qunit. That means servers of b1_4 and servers of b1_8 can't be
2116 used together if users want to use quota.
2119 Frequency : only with administrator action
2121 Description: mount failure if config log has invalid conf_param setting
2122 Details : If administrator specified an incorrect configuration parameter
2123 with "lctl conf_param" this would cause an error during future
2124 client mounts. Instead, ignore the bad configuration parameter.
2127 Frequency : blocks per group < blocksize*8 and uninit_groups is enabled
2129 Description: ldiskfs error: XXX blocks in bitmap, YYY in gd
2130 Details : If blocks per group is less than blocksize*8, set rest of the
2134 Frequency : Application do stride read on lustre
2136 Description: The read performance will drop a lot if the application does
2138 Details : Because the stride_start_offset are missing in stride read-ahead,
2139 it will cause clients read a lot of unused pages in read-ahead,
2140 then the read-performance drops.
2144 Description: more ldlm soft lockups
2145 Details : In ldlm_resource_add_lock(), call to ldlm_resource_dump()
2146 starve other threads from the resource lock for a long time in
2147 case of long waiting queue, so change the debug level from
2148 D_OTHER to the less frequently used D_INFO.
2150 Severity : enhancement
2152 Description: add -gid, -group, -uid, -user options to lfs find
2154 Severity : enhancement
2156 Description: ll_recover_lost_found_objs - recover objects in lost+found
2157 Details : OST corruption and subsequent e2fsck can leave objects in the
2158 lost+found directory. Using the "ll_recover_lost_found_objs"
2159 tool, these objects can be retrieved and data can be salvaged
2160 by using the object ID saved in the fid EA on each object.
2165 Description: this bug _only_ happens when inode quota limitation is very low
2166 (less than 12), so that inode quota unit is 1 at initialization.
2167 Details : if remaining quota equates 1, it is a sign to demonstate that quota
2168 is effective now. So least quota qunit should be 2.
2172 Description: Hung threads in invalidate_inode_pages2_range
2173 Details : The direct IO path doesn't call check_rpcs to submit a new RPC once
2174 one is completed. As a result, some RPCs are stuck in the queue
2179 Description: Procfs and llog threads access destoryed import sometimes.
2180 Details : Sync the import destoryed process with procfs and llog threads by
2181 the import refcount and semaphore.
2185 Description: mds fails to respond, threads stuck in ldlm_completion_ast
2186 Details : Sort source/child resource pair after updating child resource.
2191 Description: kernel BUG at ldiskfs2_ext_new_extent_cb
2192 Details : If insertion of an extent fails, then discard the inode
2193 preallocation and free data blocks else it can lead to duplicate
2198 Description: don't always update ctime in ext3_xattr_set_handle()
2199 Details : Current xattr code updates inode ctime in ext3_xattr_set_handle()
2200 In some cases the ctime should not be updated, for example for
2201 2.0->1.8 compatibility it is necessary to delete an xattr and it
2202 should not update the ctime.
2206 Description: add quota statistics
2207 Details : 1. sort out quota proc entries and proc code.
2208 2. add quota statistics
2213 Description: quotas are not honored with O_DIRECT
2214 Details : all writes with the flag O_DIRECT will use grants which leads to
2215 this problem. Now using OBD_BRW_SYNC to guard this.
2219 Bugzilla : 15713/16362
2220 Description: Assertion in iopen_connect_dentry in 1.6.3
2221 Details : looking up an inode via iopen with the wrong generation number can
2222 populate the dcache with a disconneced dentry while the inode
2223 number is in the process of being reallocated. This causes an
2224 assertion failure in iopen since the inode's dentry list contains
2225 both a connected and disconnected dentry.
2229 Description: assertion failure in ldlm_handle2lock()
2230 Details : fix a race between class_handle_unhash() and class_handle2object()
2231 introduced in lustre 1.6.5 by bug 13622.
2233 Severity : enhancement
2235 Description: superblock lock contention with many SMP cores on one client
2236 Details : several client filesystem locks were highly contended on SMP
2237 NUMA systems with 8 or more cores. Per-CPU datastructures
2238 and more efficient locking implemented to reduce contention.
2243 Description: Kernel BUG: sd_iostats_bump: unexpected disk index
2244 Details : remove the limit of 256 scsi disks in the sd_iostat patch
2249 Description: oops in sd_iostats_seq_show()
2250 Details : unloading/reloading the scsi low level driver triggers a kernel
2251 bug when trying to access the sd iostat file.
2256 Description: Kernel panics during QLogic driver reload
2257 Details : REQ_BLOCK_PC requests are not handled properly in the sd iostat
2258 patch, causing memory corruption.
2263 Description: journal_dev option does not work in b1_6
2264 Details : pass mount option during pre-mount.
2266 Severity : enhancement
2268 Description: Add a FIEMAP(FIle Extent MAP) ioctl for ldiskfs
2269 Details : FIEMAP ioctl will allow an application to efficiently fetch the
2270 extent information of a file. It can be used to map logical blocks
2271 in a file to physical blocks in the block device.
2274 Frequency : only with adaptive timeout enabled
2276 Description: DEBUG_REQ() bad paging request
2277 Details : ptlrpc_at_recv_early_reply() should not modify req->rq_repmsg
2278 because it can be accessed by reply_in_callback() without the
2282 Frequency : only on Cray X2
2284 Description: X2 build failures
2285 Details : fix build failures on Cray X2.
2289 Description: xid & resent requests
2290 Details : Initialize RPC XID from clock at startup (randomly if clock is
2295 Description: quota recovery deadlock during mds failover
2296 Details : This patch includes att18982, att18236, att18237 in bz14840.
2298 1. fix osts hang when mds does failover with quotaon
2299 2. prevent watchdog storm when osts threads wait for the
2304 Description: kernel panic on racer
2305 Details : Do not access dchild->d_inode when IS_ERR(dchild) is true.
2307 Severity : enhancement
2309 Description: Add lustre_start utility to start or stop multiple Lustre servers
2314 Description: Lustre GPF in {:ptlrpc:ptlrpc_server_free_request+373}
2315 Details : In case of memory pressure, list_del() can be called twice on
2316 req->rq_history_list, causing a kernel oops.
2320 Description: (ptllnd_peer.c:557:kptllnd_peer_check_sends()) ASSERTION(!in_interrupt()) failed
2321 Details : fix stack overflow in the distributed lock manager by defering
2322 export eviction after a failed ast to the elt thread instead of
2323 handling it in the dlm interpret routine.
2325 Severity : enhancement
2327 Description: More exported tunables for mballoc
2328 Details : Add support for tunable preallocation window and new tunables for
2329 large/small requests
2333 Description: Detect corruption of block bitmap and checking for preallocations
2334 Details : Checks validity of on-disk block bitmap. Also it does better
2335 checking of number of applied preallocations. When corruption is
2336 found, it turns filesystem readonly to prevent further corruptions.
2340 Frequency : only for big-endian servers
2341 Description: Check for big-endian system while mounting fs with extents feature
2342 Details : Mounting a filesystem with extents feature will fail on big-endian
2343 systems since ext3-based ldiskfs is not supported on big-endian
2344 systems. Can be over-ridden with "bigendian_extents" mount option.
2348 Description: Excessive recovery window
2349 Details : With AT enabled, the recovery window can be excessively long (6000+
2350 seconds). To address this problem, we no longer use
2351 OBD_RECOVERY_FACTOR when extending the recovery window (the connect
2352 timeout no longer depends on the service time, it is set to
2353 INITIAL_CONNECT_TIMEOUT now) and clients report the old service
2354 time via pb_service_time.
2357 Descriptoin: Don't sync journal after every i/o
2358 Details : Implement write RPC replay to allow server replies for write RPCs
2359 before data is on disk. However, this feature is disabled by
2360 default since some issues leading to data corruptions have been
2361 found during recovery (e.g. bug 19128). This feature can be enabled
2362 by running the following command on the OSSs:
2363 lctl set_param obdfilter.*.sync_journal=0
2367 Description: Watchdog triggered on MDS failover
2368 Details : enable OBD_CONNECT_MDT flag when connecting from the MDS so that
2369 the OSTs know that the MDS "UUID" can be reused for the same export
2370 from a different NID, so we do not need to wait for the export to
2375 Descriptoin: Lustre detected file system corruption with inode out of bounds
2376 Details : don't update i_size on MDS_CLOSE for directories. This causes
2377 directory corruptions on the MDT.
2379 -------------------------------------------------------------------------------
2381 2008-05-26 Sun Microsystems, Inc.
2383 * Support for kernels:
2384 2.6.5-7.311 (SLES 9),
2385 2.6.9-67.0.7.EL (RHEL 4),
2386 2.6.16.54-0.2.5 (SLES 10),
2387 2.6.18-53.1.14.el5 (RHEL 5),
2388 2.6.22.14 vanilla (kernel.org)
2389 * Client support for unpatched kernels:
2390 (see http://wiki.lustre.org/index.php?title=Patchless_Client)
2391 2.6.16 - 2.6.22 vanilla (kernel.org)
2392 * Due to problems with nested symlinks and FMODE_EXEC (bug 12652),
2393 we do not recommend using patchless RHEL4 clients with kernels
2394 prior to 2.6.9-55EL (RHEL4U5).
2395 * Recommended e2fsprogs version: 1.40.7-sun1
2396 * Note that reiserfs quotas are disabled on SLES 10 in this kernel.
2397 * RHEL 4 and RHEL 5/SLES 10 clients behaves differently on 'cd' to a
2398 removed cwd "./" (refer to Bugzilla 14399).
2399 * A new quota file format has been introduced in 1.6.5.
2400 The format conversion from prior releases is handled transparently,
2401 but releases older than 1.4.12/1.6.5 will not understand this new
2402 format. The automatic format conversion can be avoided by running
2403 the following command on the MDS before upgrading:
2404 'tunefs.lustre --param="mdt.quota_type=ug1" $MDTDEV'.
2405 For more information, please refer to bugzilla 13904.
2409 Description: quota performance fix
2410 Details : quota data is written in journalled mode instead of ordered to
2411 increase performance
2415 Description: lfs support for human-readable quota grace time strings
2416 Details : lfs setquota -t and lfs quota -t represent quota grace times
2417 in "XXwXXdXXhXXmXXs" format instead of large values in seconds
2420 Frequency : always with o2ib 1.3 and sles10
2422 Description: fix build with SLES10 and o2ib v3.
2423 Details : sles10 uses diffrent name for Module.symver file but configure
2424 assume this file has same name on RHEL/SLES/vanila kernels.
2427 Frequency : very rare, if additional xattrs are used on kernels >= 2.6.12
2429 Description: MDS may lose file striping (and hence file data) in some cases
2430 Details : If there are additional extended attributes stored on the MDS,
2431 in particular ACLs, SELinux, or user attributes (if user_xattr
2432 is specified for the client mount options) then there is a risk
2433 of attribute loss. Additionally, the Lustre file striping
2434 needs to be larger than default (e.g. striped over all OSTs),
2435 and an additional attribute must be stored initially in the
2436 inode and then increase in size enough to be moved to the
2437 external attribute block (e.g. ACL growing in size) for file
2440 Severity : enhancement
2442 Description: add message levels for liblustreapi
2447 Description: MDT cannot be unmounted, reporting "Mount still busy"
2448 Details : Mountpoint references were being leaked during open reply
2449 reconstruction after an MDS restart. Drop mountpoint reference
2450 in reconstruct_open() and free dentry reference also.
2455 Description: fix for occasional failure case of -ENOSPC in recovery-small tests
2456 Details : Move the 'good_osts' check before the 'total_bavail' check. This
2457 will result in an -EAGAIN and in the exit call path we call
2458 alloc_rr() which will with increasing aggressiveness attempt to
2459 aquire precreated objects on the minimum number of required OSCs.
2463 Description: Use old size assignment to avoid deadlock
2464 Details : This reverts the changes in bugs 2369 and bug 14138 that introduced
2465 the scheduling while holding a spinlock. We do not need locking
2466 for size in ll_update_inode() because size is only updated from
2467 the MDS for directories or files without objects, so there is no
2468 other place to do the update, and concurrent access to such inodes
2469 are protected by the inode lock.
2473 Description: Use __u64 instead of int for valid bits
2477 Description: resolve "_IOWR redefined" build error on SLES10
2481 Description: dump the memory debugging after all modules are unloaded to
2482 suppress false negative in conf_sanity test 39
2486 Description: the recovery timer never expires
2487 Details : for new client connect request, the recovery timer should not be
2488 reset, otherwise recovery timer will never expired, if the old
2489 client never come. Only old client connect and first connection
2490 req should trigger recovery timer reset.
2494 Description: the min numbers of lproc stats are wrong
2495 Details : adding a new constant LC_MIN_INIT and use it for initialization
2499 Frequency : always with interactive lfs
2501 Description: Reinitialize optind to 0 so that interactive lfs works in all cases
2504 Frequency : with multiple concurrent readdir processes in same directory
2505 Bugzilla : 15406, 15169, 15175
2506 Description: misc fixes for directory readahead.
2507 Details : prevent previous statahead async RPC callback from processing the
2508 current "statahead_info", race condition between async RPC callback
2509 add dentry into dentry hash table and "ls" thread revalidate such
2510 dentry, statahead his/miss control for hidden items, and so on.
2512 Severity : enhancement
2514 Description: build kernel-ib packages for OFED 1.3 in our release cycle
2518 Description: incore types cleaning in quota code (with respect to 64-bit limits)
2519 Details : several u32 variables declarations are replaced with u64 declarations
2524 Description: fix SLES kernel versioning
2525 Details : the kernel version for our SLES 10 kernel did not include a "-"
2526 before the "smp" at the end. while this was not a problem in
2527 general, it did mean that software trying to use the kernel
2528 version to try to detect a vendor specific kernel would fail.
2529 this was most evident by the OFED build scripts.
2534 Description: Don't update lov_desc members until making sure they are valid
2535 Details : When updating lov_desc members via proc fs, need fix their
2536 validities before doing the real update.
2539 Frequency : very rare
2541 Description: don't put request into delay list while invalidate in flight.
2542 Details : ptlrpc_delay_request sometimes put in delay list while invalidate
2543 import in flight. this produce timeout for invalidate and sometimes
2544 can cause stale data.
2546 Severity : enhancement
2548 Description: Update kernel to SLES9 2.6.5-7.311.
2550 Severity : enhancement
2552 Description: Update kernel to RHEL4 2.6.9-67.0.7.
2557 Frequency : on PPC only
2558 Description: not convert ost objects for directory because it's not exist.
2559 Details : ll_dir_getstripe assume dirrectory has ost objects but this wrong.
2561 Severity : enhancement
2563 Description: Fix warnings with compile liblustre at sles10/rhel5 which have
2564 __u64 as usingied long long type.
2567 Frequency : rare, on shutdown
2569 Description: race process ast vs remove callback
2570 Details : removing callback before disconnect import open race with
2571 processing callback.
2573 Severity : enhancement
2575 Description: Update kernel to SLES9 2.6.5-7.311.
2577 Severity : enhancement
2579 Description: Files open for execute are not marked busy on SLES10
2580 Details : Add FMODE_EXEC to SLES10 SP1 server kernel series.
2582 Severity : enhancement
2584 Description: Add server support for vanilla-2.6.22.14.
2587 Frequency : occasional
2589 Description: Avoid lov_create() getting stuck in obd_statfs_rqset()
2590 Details : If an OST is down the MDS will hang indefinitely in
2591 obd_statfs_rqset() waiting for the statfs data. While for
2592 MDS QOS usage of statfs, it should not stuck in waiting.
2594 Severity : enhancement
2596 Description: Disable adaptive timeouts by default
2599 Frequency : on network error
2601 Description: panic with double free request if network error
2602 Details : mdc_finish_enqueue is finish request if any network error occuring,
2603 but it's true only for synchronus enqueue, for async enqueue
2604 (via ptlrpcd) this incorrect and ptlrpcd want finish request
2608 Frequency : rare, on recovery
2610 Description: read procfs can produce deadlock in some situation
2611 Details : Holding lprocfs lock with send rpc can produce block for destroy
2612 obd objects and this also block reconnect with -EALREADY.
2613 This isn't fix all lprocfs bugs - but make it rare.
2615 Severity : enhancement
2617 Description: Update kernel to RHEL5 2.6.18-53.1.14.el5.
2620 Frequency : frequent on X2 node
2622 Description: mdc_set_open_replay_data LBUG
2623 Details : Set replay data for requests that are eligible for replay.
2628 Description: lustre_mgs: operation 101 on unconnected MGS
2629 Details : When MGC is disconnected from MGS long enough, MGS will evict the
2630 MGC, and late on MGC cannot successfully connect to MGS and a lot
2631 of the error messages complaining that MGS is not connected.
2634 Frequency : rare, depends on device drivers and load
2636 Description: MDS or OSS nodes crash due to stack overflow
2637 Details : Code changes in 1.6.4 increased the stack usage of some functions.
2638 In some cases, in conjunction with device drivers that use a lot
2639 of stack, the MDS (or possibly OSS) service threads could overflow
2640 the stack. One change which was identified to consume additional
2641 stack has been reworked to avoid the extra stack usage.
2643 Severity : enhancement
2645 Description: Update to RHEL5 latest kernel-2.6.18-53.1.13.el5.
2647 Severity : enhancement
2649 Description: Update to SLES10 SP1 latest kernel-2.6.16.54-0.2.5.
2651 Severity : enhancement
2653 Description: Update to RHEL5 latest kernel-2.6.18-53.1.6.el5.
2655 Serverity : enhancement
2657 Description: Update RHEL4 kernel to 2.6.9-67.0.4.
2660 Frequency : rare on shutdown OST
2662 Description: Don't allow skipping OSTs if index has been specified.
2663 Details : Don't allow skipping OSTs if index has been specified, make locking
2664 in internal create lots better.
2669 Description: ASSERTION(!PageDirty(page)) failed
2670 Details : Wrong check could lead to an assertion failure under specific
2676 Description: LBUG in ptlrpc_check_set() bad phase ebc0de00
2677 Details : access to bitfield in structure is always rounded to long
2678 and this produce problem with not atomic change any bit.
2683 Description: Lustre make rpms failed.
2684 Details : Remove ldiskfs spec file to avoids rpmbuild be confused when
2685 builds Lustre rpms from tarball.
2687 Severity : enhancement
2689 Description: Update to SLES9 SP4 kernel-2.6.5-7.308.
2692 Frequency : rare on shutdown OST
2694 Description: If llog cancel was not send before clean_exports phase, this can
2695 produce deadlock in llog code.
2696 Details : If llog thread has last reference to obd and call class_import_put
2697 this produce deadlock because llog_cleanup_commit_master wait when
2698 last llog_commit_thread exited, but this never success because was
2699 called from llog_commit_thread.
2702 Frequency : only if OST index is skipped
2704 Description: NULL lov_tgts causing MDS oops
2705 Details : more safe checks for NULL lov_tgts for avoid oops.
2707 Severity : enhancement
2709 Description: Update to RHEL4 latest kernel-2.6.9-67.0.1.EL.
2711 Severity : enhancement
2713 Description: Update to RHEL5 latest kernel-2.6.18-53.1.4.el5.
2718 Description: make mgs_setparam() handle fsname containing dash
2719 Details : fsname containing a dash does not work with lctl conf_param
2721 Severity : enhancement
2723 Description: Update to RHEL4 Update-6 kernel-2.6.9-67.EL.
2726 Frequency : rare, in recovery and (or) destroy lovobjid file.
2728 Description: rewrite lov objid code.
2729 Details : Cleanup for lov objid code, remove scability problems and wrong
2730 locking. Fix sending last_id into ost.
2732 Severity : enhancement
2734 Description: Update to SLES10 SP1 latest kernel-2.6.16.54-0.2.3.
2736 Severity : enhancement
2738 Description: Update to RHEL5 Update-1 kernel 2.6.18-53.el5.
2739 Details : Use d_move_locked instead of __d_move.
2742 Frequency : rare, at shutdown
2744 Description: access already free / zero obd_namespace.
2745 Details : if client_disconnect_export was called without force flag set,
2746 and exist connect request in flight, this can produce access to
2747 NULL pointer (or already free pointer) when connect_interpret
2748 store ocd flags in obd_namespace.
2751 Frequency : only at startup
2753 Description: not alloc memory with spinlock held.
2754 Details : allocation memory with GFP_KERNEL can produce sleep deadlock,
2755 if any spinlock held.
2760 Description: lfs find does not continue on file error
2761 Details : Continue other files processing when a file/dir is absent.
2765 Description: Inconsistent usage of lustre_pack_reply()
2766 Details : Standardize the usage of lustre_pack_reply() such that it
2767 always generate a CERROR on failure.
2770 Frequency : very rare
2772 Description: Fix replay if there is an un-replied request and open
2773 Details : In some cases, older replay request will revert the
2774 mcd->mcd_last_xid on MDS which is used to record the client's
2775 latest sent request.
2777 Severity : enhancement
2779 Description: Update to RHEL5 kernel 2.6.18-8.1.15.el5.
2781 Severity : enhancement
2783 Description: Update to SLES10 SP1 kernel 2.6.16.53-0.16
2785 Severity : enhancement
2787 Description: Update to SLES9 kernel-2.6.5-7.287.3.
2789 Severity : enhancement
2791 Description: Update to RHEL4 kernel-2.6.9-55.0.12.EL.
2793 Severity : enhancement
2795 Description: Build SLES10 patchless client fails
2796 Details : The configure was broken by run ./configure with
2797 --with-linux-obj=.... argument for patchless client. When the
2798 configure use --with-linux-obj, the LINUXINCLUDE= -Iinclude
2799 can't search header adequately. Use absolute path such as
2800 -I($LINUX)/include instead.
2802 Severity : enhancement
2804 Description: Lustre Page Accounting
2805 Details : New macros for page alloc and free which enable accounting
2806 of page allocation of Lustre. Use percpu counters to store memory
2807 and page statistics.
2810 Frequency : only if debugging is disabled
2812 Description: LASSERT_{REQ,REP}SWAB macros are buggy
2813 Details : If SWAB_PARANOIA is disabled, the LASSERT_REQSWAB and
2814 LASSERT_REPSWAB macros become no-ops, which is incorrect. Drop
2815 these macros and replace them with their definitions instead.
2820 Description: interrupt oig_wait produce painc on resend.
2821 Details : brw_redo_request can be used for resend requests from ptlrpcd and
2822 private set, and this produce situation when rq_ptlrpcd_data not
2823 copyed to new allocated request and triggered LBUG on assert
2824 req->rq_ptlrpcd_data != NULL. But this member used only for wakeup
2825 ptlrpcd set if request is changed and can be safety changed to use
2828 Severity : enhancement
2830 Description: organize the server-side client stats on per-nid basis
2831 Details : Change the structure of stats under obdfilter and mds to
2839 The "uuid"s file would list the uuids of _active_ exports.
2840 And the clear entry is to clear all stats and stale nids.
2845 Description: Processes looping in ll_readdir() on Lustre clients finally causing
2846 a full node pseudo-hang
2847 Details : Concurrent access to the same directory from multiple clients with
2848 intensive file creation/removal can cause a client node to spin in
2849 ll_readdir(). i_version must be increased every time the lock is
2850 cancelled to ensure a revalidate is done.
2855 Description: touch file failed when fs is not full
2856 Details : OST in recovery should not be discarded by MDS in alloc_qos(),
2857 otherwise we can get ENOSP while fs is not full.
2860 Frequency : only for Cray XT3
2861 Bugzilla : 12829/13455
2862 Description: Changing primary group doesn't change the group lustre assigns to
2864 Details : When CRAY_XT3 is defined, the fsgid supplied by the client is
2865 overridden with the primary group provided by the group upcall,
2866 whereas the supplied fsgid can be trusted if it is in the list of
2867 supplementary groups returned by the group upcall.
2869 Severity : enhancement
2871 Description: Root Squash Functionality
2872 Details : Implementation of NFS-like root squash capability. Specifically,
2873 don't allow someone with root access on a client node to be able
2874 to manipulate files owned by root on a server node.
2876 Severity : enhancement
2878 Description: Slow trucate/writes to huge files at high offsets.
2879 Details : Directly associate cached pages to lock that protect those pages,
2880 this allows us to quickly find what pages to write and remove
2881 once lock callback is received.
2886 Description: Too many locks accumulating on client during NFS usage
2887 Details : mds_open improperly used accmode to find out access mode to a
2888 file. Also mdc_intent_lock logic to find out if we already have
2889 lock similar to just received was flawed since introduction of
2890 skiplists - locks are now added to the front of the granted
2895 Description: Hit ASSERTION(obd->obd_stopping == 1) failed in some setup failed
2897 Details : In obd setup failure handler, obd_stopping will not necessarily to
2898 be 1, and obd_set_up should also be checked to make sure whether
2899 obd is completely setup.
2901 Severity : enhancement
2903 Description: Allow masking D_WARNING, D_ERROR messages from console
2904 Details : Console messages can now be disabled via lnet.printk.
2909 Description: User code with malformed file open parameter crashes client node
2910 Details : Before packing join_file req, all the related reference should be
2911 checked carefully in case some malformed flags cause fake join_file
2917 Description: shrink/enlarge qunit size when needed; fix the problem of coarse
2918 grain of quota doing harm to quota's accuracy
2919 Details : qunit size will be changed when quota limitation is too low/high;
2920 record the pending quota write in order to get more accureate
2921 quota; delete the patch for bug12588, which is unnecessary when
2922 this patch is landed. This bug also contains fixes for bug 14526,
2923 14299, 14601 and 13794.
2927 Description: LDLM_ENQUEUE races with LDLM_CP_CALLBACK
2928 Details : ldlm_completion_ast() assumes that a lock is granted when the req
2929 mode is equal to the granted mode. However, it should also check
2930 that LDLM_FL_CP_REQD is not set.
2934 Description: Heavy nfs access might result in deadlocks
2935 Details : After ELC code landed, it is now improper to enqueue any mds
2936 locks under och_sem, because enqueue might want to decide to
2937 cancel open locks for same inode we are holding och_sem for.
2941 Description: 35% write performance drop with ldiskfs2 when quotas are on
2942 Details : Enable ext3 journalled quota by default to improve performance
2943 when quotas are turned on.
2947 Description: Client eviction while running blogbench
2948 Details : A lot of unlink operations with concurrent I/O can lead to a
2949 deadlock causing evictions. To address the problem, the number of
2950 oustanding OST_DESTROY requests is now throttled to
2951 max_rpcs_in_flight per OSC and LDLM_FL_DISCARD_DATA blocking
2952 callbacks are processed in priority.
2955 Frequency : RHEL4 only
2957 Description: mkfs is very slow on IA64/RHEL4
2958 Details : A performance regression has been discovered in the MPT Fusion
2959 driver between versions 3.02.73rh and 3.02.99.00rh. As a
2960 consequence, we have downgraded the MPT Fusion driver in the RHEL4
2961 kernel from 3.02.99.00 to 3.02.73 until this problem is fixed.
2964 Frequency : PPC/PPC64 only
2966 Description: conflicts between asm-ppc64/types.h and lustre_types.h
2967 Details : fix duplicated definitions between asm-ppc64/types.h and
2968 lustre_types.h on PPC.
2971 Frequency : PPC/PPC64 only
2973 Description: asm-ppc/segment.h does not exist
2974 Details : fix compile issue on PPC.
2978 Description: data checksumming impacts single node performance
2979 Details : add support for several checksum algorithms. Currently, CRC32 and
2980 Adler-32 are supported. The checksum type can be changed on the fly
2981 through /proc/fs/lustre/osc/*/checksum_type.
2985 Description: use adler32 for page checksums
2986 Details : when available, use the Adler-32 algorithm instead of CRC32 for
2991 Description: better handle error messages in extents code
2993 Severity : enhancement
2995 Description: SNMP support enhancement
2996 Details : Adding total number of sampled request for an MDS node in snmp
2999 Severity : enhancement
3001 Description: Optimize ldlm waiting list processing for PR extent locks
3002 Details : When processing waiting list for read extent lock and meeting read
3003 lock that is same or wider to it that is not contended, skip
3004 processing rest of the list and immediatelly return current
3005 status of conflictness, since we are guaranteed there are no
3006 conflicting locks in the rest of the list.
3010 Description: Time out and refuse to reconnect
3011 Details : When the failover node is the primary node, it is possible
3012 to have two identical connections in imp_conn_list. We must
3013 compare not conn's pointers but NIDs, otherwise we
3014 can defeat connection throttling.
3018 Description: Client not clear own cache if answer to reconnect is lost.
3019 Details : Client gets evicted from server. Now client also thinks it is
3020 disconnected (or gets ENOTCONN on its operation) and decides to
3021 reconnect. Server receives reconnect message, but cannot find
3022 export. New export is created that is fully valid (new cookie!),
3023 but reply is lost and not reported to client. Client reconnects
3024 again and gets back a just-created connection, but it is not new
3025 so client thinks it was not evicted and does not do recovery.
3029 Description: Detect stride IO mode in read-ahead
3030 Details : When a client does stride read, read-ahead should detect that and
3031 read-ahead pages according to the detected stride pattern.
3035 Description: build for x2 fails
3036 Details : fix compile issue on Cray systems.
3038 Severity : enhancement
3040 Description: implement readv/writev APIs(aio_read/aio_writes in newer kernels)
3041 Details : This greatly improves speed of NFS writes on 2.6 kernels.
3044 Frequency : only on PPC/SLES10
3046 Description: "BITS_PER_LONG is not 32 or 64" in linux/idr.h
3047 Details : On SLES10/PPC, fs.h includes idr.h which requires BITS_PER_LONG to
3048 be defined. Add a hack in mkfs_lustre.c to work around this compile
3053 Description: LASSERT on MDS when client holding flock lock dies
3054 Details : ldlm pool logic depends on number of granted locks equal to
3055 number of released locks which is not true for flock locks, so
3056 just exclude such locks from consideration.
3060 Description: MDS deadlock with many ll_sync_lov threads and I/O stalled
3061 Details : Use fsfilt_sync() for both the whole filesystem sync and
3062 individual file sync to eliminate dangerous inode locking
3063 with I_LOCK that can lead to a deadlock.
3067 Description: Update an obsolete wirecheck.c generator
3068 Details : Update wirecheck.c/wirehdr.c and regenerate wiretest.c
3072 Description: Client can panic on open sometimes
3073 Details : It is possible that we try to free already freed request in
3074 ll_file_open in some error cases when we send request from
3079 Description: performance in 1.6.3
3080 Details : Force q->max_phys_segments to MAX_PHYS_SEGMENTS on SLES10 to be
3081 sure that 1MB requests are not fragmented by the block layer.
3085 Description: LDLM soft lockups - improvement
3086 Details : It is be possible to send the lock handle along with each read
3087 or write request because the client is already doing a lock match
3088 itself so there isn't any reason the OST should have to re-do that
3094 Description: lfs quota fails with deactivated OSTS
3095 Details : With this patch, three improvements are included:
3096 1. detete the softlimit in mds and osts when use "lfs quota".
3097 2. display the inaccurate data in the output of "lfs quota".
3098 3. try to get quota info when "lfs quota" is executed.
3103 Description: Extent locks not granted with no conflicts sometimes.
3104 Details : When race occurs in glimpse handler and nothing is returned,
3105 we do not reprocess the queue after lock cancel, and that leads
3106 to a stall until next activity on a resource
3109 Frequency : failover with quotaon
3111 Description: during mds failovers with quota on, OSTs got into deadlock state
3112 and causing dumpstack.
3113 Details : for every quota slave, at any time, there is only one quota req
3114 is sent to quota master for every uid/gid. Before that quota req
3115 returns, all the thread relative to the same uid/gid will wait.
3116 So if the quota req is lost because mds failovers or any other
3117 reasons, this bug will be hit. Now, dqacq_interpret() will handle
3118 quota reqs who time out.
3120 Severity : enhancement
3123 Description: when quota slave checks if quota is enough, there is an unnecessary
3125 Details : place this wait on necessary place instead of always waiting.
3127 --------------------------------------------------------------------------------
3129 2007-12-07 Cluster File Systems, Inc. <info@clusterfs.com>
3131 * Support for kernels:
3132 2.6.5-7.286 (SLES 9),
3133 2.6.9-55.0.9.EL (RHEL 4),
3134 2.6.16.53-0.8 (SLES 10),
3135 2.6.18-8.1.14.el5 (RHEL 5),
3136 2.6.18.8 vanilla (kernel.org)
3137 * Client support for unpatched kernels:
3138 (see http://wiki.lustre.org/index.php?title=Patchless_Client)
3139 2.6.16 - 2.6.22 vanilla (kernel.org)
3140 * Due to recently discovered recovery problems, we do not recommend
3141 using patchless RHEL 4 clients with this or any earlier release.
3142 * Recommended e2fsprogs version: 1.40.2-cfs1
3143 * Note that reiserfs quotas are disabled on SLES 10 in this kernel.
3146 Frequency : occasional
3148 Description: MDS hang or stay in waiting lock
3149 Details : If client receive lock with CBPENDING flag ldlm need send lock
3150 cancel as separate rpc, to avoid situation when cancel request
3151 can't processed due all i/o threads stay in wait lock.
3154 Frequency : occasional
3156 Description: Do not fail import if osc_interpret_create gets -EAGAIN
3157 Details : If osc_interpret_create got -EAGAIN it immediately exits and
3158 wakeup oscc_waitq. After wakeup oscc_wait_for_objects call
3159 oscc_has_objects and see OSC has no objests and call
3160 oscc_internal_create to resend create request.
3162 Severity : enhancement
3164 Description: Update kernel patches for SLES10 2.6.16.53-0.8.
3165 Details : Update which_patch & target file for SLES10 latest kernel.
3167 Severity : enhancement
3169 Description: add --type and --size parameters to lfs find
3170 Details : Enhance lfs find by adding filetype and filesize parameters. Also
3171 multiple OBDs can now be specified for the --obd option.
3173 Severity : enhancement
3175 Description: eliminate client locks in face of contention
3176 Details : file contention detection and lockless i/o implementation
3177 for contended files.
3179 Severity : enhancement
3181 Description: Remove client patches from SLES 10 kernel.
3182 Details : This causes SLES 10 clients to behave as patchless clients
3183 even on a Lustre-patched (server) kernel.
3185 Severity : enhancement
3187 Description: use i_size_read and i_size_write in 2.6 port
3188 Details : replace inode->i_size access with i_size_read/write()
3191 Frequency : when removing large files
3193 Description: scheduling issue during removal of large Lustre files
3194 Details : Don't take the BKL in fsfilt_ext3_setattr() for 2.6 kernels.
3195 It causes scheduling issues when removing large files (17TB in the
3201 Description: 1.4.11 Can't handle directories with stripe set and extended ACLs
3202 Details : Impossible (EPROTO is returned) to access a directory that has a
3203 non-default striping and ACLs.
3206 Frequency : only on ppc
3208 Description: /proc/fs/lustre/devices broken on ppc
3209 Details : The patch as applied to 1.6.2 doesn't look correct for all arches.
3210 We should make sure the type of 'index' is loff_t and then cast
3211 explicitly as needed below. Do not assign an explicitly cast
3215 Frequency : only for rhel5
3217 Description: Kernel patches update for RHEL5 2.6.18-8.1.10.el5.
3218 Details : Modify the target file & which_kernel.
3221 Frequency : if the uninit_groups feature is enabled on ldiskfs
3223 Description: e2fsck reports "invalid unused inodes count"
3224 Details : If a new ldiskfs filesystem is created with the "uninit_groups"
3225 feature and only a single inode is created in a group then the
3226 "bg_unused_inodes" count is incorrectly updated. Creating a
3227 second inode in that group would update it correctly.
3232 Description: buffer overruns could theoretically occur
3233 Details : llapi_semantic_traverse() modifies the "path" argument by
3234 appending values to the end of the origin string, and a buffer
3235 overrun may occur. Adding buffer overrun check in liblustreapi.
3237 Severity : enhancement
3239 Description: Add jbd statistics patch for RHEL5 and 2.6.18-vanilla.
3242 Frequency : only if filesystem is inconsistent
3244 Description: handle "serious error: objid * already exists" more gracefully
3245 Details : If LAST_ID value on disk is smaller than the objects existing in
3246 the O/0/d* directories, it indicates disk corruption and causes an
3247 LBUG(). If the object is 0-length, then we should use the existing
3248 object. This will help to avoid a full fsck in most cases.
3250 Severity : enhancement
3252 Description: Kernel patches update for RHEL4 2.6.9-55.0.6.
3253 Details : Modify vm-tunables-rhel4.patch.
3255 Severity : enhancement
3257 Description: Kernel config for 2.6.18-vanilla.
3258 Details : Modify targets/2.6-vanilla.target.in.
3259 Add config file kernel-2.6.18-2.6-vanilla-i686.config.
3260 Add config file kernel-2.6.18-2.6-vanilla-i686-smp.config.
3261 Add config file kernel-2.6.18-2.6-vanilla-x86_64.config.
3262 Add config file kernel-2.6.18-2.6-vanilla-x86_64-smp.config.
3265 Frequency : occasional
3267 Description: improve handling recoverable errors
3268 Details : If request processed with error which can be recoverable on server
3269 request should be resend, otherwise page released from cache and
3275 Description: Kernel patches update for RHEL5 2.6.18-8.1.14.el5.
3276 Details : Modify target file & which_patch.
3277 A flaw was found in the IA32 system call emulation provided
3278 on AMD64 and Intel 64 platforms. An improperly validated 64-bit
3279 value could be stored in the %RAX register, which could trigger an
3280 out-of-bounds system call table access. An untrusted local user
3281 could exploit this flaw to run code in the kernel
3282 (ie a root privilege escalation). (CVE-2007-4573).
3286 Description: change order of libsysio includes
3287 Details : '#include sysio.h' should always come before '#include xtio.h'
3289 Severity : enhancement
3291 Description: adapt the lustre_config script to support the upgrade case
3292 Details : Add "-u" option for lustre_config script to support upgrading 1.4
3293 server targets to 1.6 in parallel.
3298 Description: To avoid grant space > avaible space when the disk is almost
3299 full. Without this patch you might see the error "grant XXXX >
3300 available" or some LBUG about grant, when the disk is almost
3302 Details : In filter_check_grant, for non_grant cache write, we should
3303 check the left space by if (*left > ungranted + bytes), instead
3304 of (*left > ungranted), because only we are sure the left space
3305 is enough for another "bytes", then the ungrant space should be
3306 increase. In client, we should update cl_avail_grant only there is
3307 OBD_MD_FLGRANT in the reply.
3312 Description: Update RHEL 4 kernel to fix local root privilege escalation.
3313 Details : Update to the latest RHEL 4 kernel to fix the vulnerability
3314 described in CVE-2007-4573. This problem could allow untrusted
3315 local users to gain root access.
3318 Frequency : when using O_DIRECT and quotas
3320 Description: Incorrect file ownership on O_DIRECT output files
3321 Details : block usage reported by 'lfs quota' does not take into account
3322 files that have been written with O_DIRECT.
3326 Description: (rw.c:1323:ll_read_ahead_pages()) ASSERTION(page_idx > ria->ria_stoff) failed
3327 Details : Once the unmatched stride IO mode is detected, shrink the stride-ahead
3328 window to 0. If it does hit cache miss, and read-pattern is still
3329 stride-io mode, does not reset the stride window, but also does not
3330 increase the stride window length in this case.
3332 --------------------------------------------------------------------------------
3334 2007-09-27 Cluster File Systems, Inc. <info@clusterfs.com>
3336 * Support for kernels:
3337 2.6.5-7.286 (SLES 9),
3338 2.6.9-55.0.2.EL (RHEL 4),
3339 2.6.16.46-0.14 (SLES 10),
3340 2.6.18-8.1.8.el5 (RHEL 5),
3341 2.6.18.8 vanilla (kernel.org)
3342 * Client support for unpatched kernels:
3343 (see http://wiki.lustre.org/index.php?title=Patchless_Client)
3344 2.6.16 - 2.6.21 vanilla (kernel.org)
3345 * Due to recently discovered recovery problems, we do not recommend
3346 using patchless RHEL 4 clients with this or any earlier release.
3347 * Recommended e2fsprogs version: 1.40.2-cfs1
3348 * Note that reiserfs quotas are disabled on SLES 10 in this kernel.
3352 Description: Fix errors in lfs documentation
3353 Details : Fixes man pages
3355 Severity : enhancement
3357 Description: Adaptive timeouts
3358 Details : RPC timeouts adapt to changing server load and network
3359 conditions to reduce resend attempts and improve recovery time.
3361 Severity : enhancement
3363 Description: llapi_file_create() does not allow some changes
3364 Details : add llapi_file_open() that allows specifying the file creation
3365 mode and open flags, and also returns an open file handle.
3367 Severity : enhancement
3369 Description: df doesn't work properly if diskfs blocksize != 4K
3370 Details : Choose biggest blocksize of OST's as the LOV's blocksize.
3373 Frequency : always on directories with default striping set
3375 Description: lfs find on -1 stripe looping in lsm_lmm_verify_common()
3376 Details : Avoid lov_verify_lmm_common() on directory with -1 stripe count.
3378 Severity : enhancement
3380 Description: merge and cleanup kernel patches.
3381 Details : Remove mnt_lustre_list in vfs_intent-2.6-rhel4.patch.
3383 Severity : enhancement
3385 Description: RedHat Update kernel for RHEL5
3386 Details : Kernel config file for RHEL5.
3388 Severity : enhancement
3390 Description: OSS needs mutliple precreate threads
3391 Details : Add ability to start more than one create thread per OSS.
3394 Frequency : only with quota on the root user
3396 Description: mds_obd_create error creating tmp object
3397 Details : When the user sets quota on root, llog will be affected and can't
3398 create files and write files.
3401 Frequency : Always on ia64 patchless client, and possibly others.
3403 Description: Add EXPORT_SYMBOL check for node_to_cpumask symbol.
3404 Details : This allows the patchless client to be loaded on architectures
3405 without this export.
3410 Description: disorder of journal start and llog_add cause deadlock.
3411 Details : in llog_origin_connect, journal start should happen before
3412 llog_add keep the same order as other functions to avoid
3415 Severity : enhancement
3417 Description: RedHat Update kernel for RHEL5
3418 Details : Modify the kernel config file more closer RHEL5.
3420 Severity : enhancement
3422 Description: Build failure against Centos5 (RHEL5)
3423 Details : Define PAGE_SIZE when it isn't present.
3426 Frequency : occasionally when using NFS
3428 Description: "ll_intent_file_open()) lock enqueue: err: -13" with nfs
3429 Details : with NFS, the anon dentry's parent was set to itself in
3430 d_alloc_anon(), so in MDS, we use rec->ur_fid1 to find the
3431 corresponding dentry other than use rec->ur_name.
3433 Severity : enhancement
3435 Description: enable data checksumming by default
3436 Details : enable checksum by default, allow --disable-checksum
3437 configure option and "-o nochecksum" mount option. Checksums
3438 can also be disabled at runtime via $LPROC/osc/*/checksum_pages.
3441 Frequency : Occasionally with failover
3443 Description: Client eviction due to failover config
3444 Details : after a connection loss, the lustre client should attempt to
3445 reconnect to the last active server first before trying the
3446 other potential connections.
3448 Severity : enhancement
3450 Description: client-side metadata stat-ahead during readdir(directory readahead)
3451 Details : perform client-side metadata stat-ahead when the client detects
3452 readdir and sequential stat of dir entries therein
3457 Description: when mds and osts use different quota unit(32bit and 64bit),
3458 quota will be released repeatly.
3459 Details : void sending multiple quota reqs to mds, which will keep the
3460 status between the reqs.
3463 Frequency : only with liblustre clients on XT3
3465 Description: evictions taking too long
3466 Details : allow llrd to evict clients directly on OSTs
3470 Description: osts not allocated evenly to files
3471 Details : change the condition to increase offset_idx
3473 Severity : enhancement
3475 Description: self-adjustable client's lru lists
3476 Details : use adaptive algorithm for managing client cached locks lru
3477 lists according to current server load, other client's work
3478 pattern, memory activities, etc. Both, server and client
3479 side namespaces provide number of proc tunables for controlling
3484 Description: rewrite ext2-derived code in llite/dir.c and obdclass/uuid.c
3485 Details : rewrite inherited code (uuid parsing code from ext2 utils and
3486 readdir code from ext3) from scratch preserving functionality.
3490 Description: Only those disconnect error should be returned by rq_status.
3491 Details : In open/enqueue processs, Some errors, which will cause client
3492 disconnected, should be returned by rq_status, while other
3493 errors should still be returned by intent, then mdc or llite will
3496 Severity : enhancement
3498 Description: Tune the kernel for good SCSI performance.
3499 Details : Set the value of /sys/block/{dev}/queue/max_sectors_kb
3500 to the value of /sys/block/{dev}/queue/max_hw_sectors_kb
3504 Frequency : Always for filesystems larger than 2TB on 32-bit systems.
3505 Bugzilla : 13547, 13627
3506 Description: Data corruption for OSTs that are formatted larger than 2TB
3508 Details : When generating the bio request for lustre file writes the
3509 sector number would overflow a temporary variable before being
3510 used for the IO. The data reads correctly from Lustre (which
3511 will overflow in a similar manner) but other file data or
3512 filesystem metadata may be corrupted in some cases.
3516 Description: "lfs find -obd UUID" prints directories
3517 Details : "lfs find -obd UUID" will return all directory names instead
3518 of just file names. It is incorrect because the directories
3519 do not reside on the OSTs.
3522 Bugzilla : 13596, 13614
3523 Description: MDS hang after unclean shutdown of lots of clients
3524 Details : Never resend AST requests.
3528 Frequency : Always, for kernels after 2.6.16
3529 Description: Fix warning idr_remove called for id=.. which is not allocated.
3530 Details : Last kernels save old s_dev before kill super and not allow
3531 to restore from callback - restore it before call kill_anon_super.
3535 Description: Fix error on 'ls .' at the top of the Lustre mount.
3536 Details : Don't revalidate dentry if it is a root dentry.
3538 --------------------------------------------------------------------------------
3540 2007-08-27 Cluster File Systems, Inc. <info@clusterfs.com>
3542 * Support for kernels:
3543 2.6.5-7.286 (SLES 9),
3544 2.6.9-55.0.2.EL (RHEL 4),
3545 2.6.16.46-0.14 (SLES 10),
3546 2.6.18-8.1.8.el5 (RHEL 5),
3547 2.6.18.8 vanilla (kernel.org)
3548 * Client support for unpatched kernels:
3549 (see http://wiki.lustre.org/index.php?title=Patchless_Client)
3550 2.6.16 - 2.6.22 vanilla (kernel.org)
3551 * Due to recently discovered recovery problems, we do not recommend
3552 using patchless RHEL 4 clients with this or any earlier release.
3553 * Recommended e2fsprogs version: 1.39.cfs8
3554 * Note that reiserfs quotas are disabled on SLES 10 in this kernel.
3559 Description: block reactivating mgc import until all deactivates complete
3560 Details : Fix race when failing back MDT/MGS to itself (testing)
3562 Severity : enhancement
3564 Description: lfs setstripe enhancement
3565 Details : Make lfs setstripe understand 'k', 'm' and 'g' for stripe size.
3568 Frequency : mds/oss recovery
3570 Description: llog ctxt is refrenced after it has been freed.
3571 Details : llog ctxt refcount was added to avoide the race between ctxt free
3572 and llog recovery process. Each llog user must hold ctxt refcount
3573 before it access the llog. And the llog ctxt can only be freed
3574 when its refcount is zero.
3576 Severity : enhancement
3578 Description: randomly memory allocation failure util
3579 Details : Make lustre randomly failed allocating memory for testing purpose.
3582 Frequency : only for SLES10
3584 Description: Update kernel patch for SLES10 SP1
3585 Details : Add patch blkdev_tunables-2.6-sles10.patch to 2.6-sles10.series.
3587 Severity : enhancement
3589 Description: omit set fsid for export NFS
3590 Details : fix set/restore device id for avoid EMFILE error and mark lustre fs
3591 as FS_REQUIRES_DEV for avoid problems with generate fsid.
3594 Frequency : after network failures
3596 Description: Add sync option to mount_lustre.c
3597 Details : Client loses data written to lustre after a network interruption.
3599 Severity : enhancement
3601 Description: Error message improvement.
3602 Details : Merging of two LCONSOLE_ERROR_MSG into one.
3604 Severity : enhancement
3606 Description: don't use GFP_* in generic Lustre code.
3607 Details : Use cfs_alloc_* functions and CFS_* flags for code portability.
3609 Severity : enhancement
3611 Description: obdclass is limited by single OBD_ALLOC(idarray)
3612 Details : replace OBD_ALLOC/OBD_FREE with OBD_VMALLOC/OBD_VFREE
3614 Severity : enhancement
3616 Description: updated patchess for new RHEL4 kernel
3617 Details : Fixed ext3-unlink-race.patch per Kalpak's comment.
3619 Severity : enhancement
3621 Description: warnings with build patchless client with vanila 2.6.19 and up
3622 Details : change old ctl_table style and replace ctl_table/ctl_table_header
3623 with cfs_sysctl_table_t/cfs_sysctl_table_header_t
3625 Severity : enhancement
3627 Description: O_DIRECT bypasses client statistics.
3628 Details : When running with O_DIRECT I/O, neither the client rpc_stats nor
3629 read_ahead_stats were updated. Copied stats section from
3630 osc_send_oap_rpc() into async_internal().
3633 Frequency : only for Cray XT3
3635 Description: peer credits not enough on many OST per OSS systems.
3636 Details : Use new lnet way to add credits as we need those for pings and ASTs
3639 Frequency : only with liblustre
3641 Description: Liblustre is not releasing flock locks on file close.
3642 Details : Release flock locks on file close.
3645 Frequency : only for RHEL4
3647 Description: Update kernel patches for kernel-2.6.9-55.0.2.EL
3648 Details : Remove inode-nr_unused-2.6.9-rhel4.patch from 2.6-rhel4.series
3649 Update target file and kernel config.
3653 Description: lustre support for RHEL5
3654 Details : Add support for RHEL5.
3659 Description: ASSERTION(export != NULL) failed in target_handle_connect
3660 Details : Assetion hit is result of rare race between disconnect and connet
3661 to same nid. target_handle_connect found old connect cockie and
3662 tried to reconnect, but can't find export for this cockie.
3667 Description: umount blocks forever on error
3668 Details : In result of wrong using obd_no_recov and obd_force flags client
3669 can hand if cancel or some other requests is lost.
3672 Frequency : Only for SLES
3674 Description: sanity_quota fail test_1
3675 Details : There are multiple occurences of $TSTUSR in SLES's /etc/group
3676 file, which makes TSTID[2] inunique.
3678 Severity : enhancement
3680 Description: Kernel patches for SLES9 2.6.5-7.286 kernel
3681 Details : Update target/ChangeLog/which_patch .
3683 Severity : enhancement
3685 Description: jbd statistics
3686 Details : Port older jbd statistics patch for sles10
3691 Description: lvbo_init failed for resource with missing objects.
3692 Details : Fix returning error if we do stat for file with missing/corrupted
3693 objects and i_size set to all sum of size of all avaible objects.
3694 if we to truncate/write to missing object - it is recreated.
3699 Description: Oops in read and write path when failing to allocate lock.
3700 Details : Check if lock allocation failed and return error back.
3703 Frequency : When flocks are used.
3705 Description: assertion failure in ldlm_cli_enquque_fini for non NULL lock.
3706 Details : Flock locks might destroy just granted lock if it could be merged
3707 with another existing flock, this is done in completion handler,
3708 so teach ldlm_cli_enquque_fini that this is a valid case for
3714 Description: reply_lock_interpret crash due to race with it and lock cancel.
3715 Details : Do not replay locks that are being cancelled. Do not reference
3716 locks by their address during replay, just by their handle.
3719 Frequency : only with deactivated OSTs
3721 Description: lstripe command fails for valid OST index
3722 Details : The stripe offset is compared to 'lov->desc.ld_tgt_count'
3723 instead of lov->desc.ld_active_tgt_count.
3725 Severity : enhancement
3727 Description: Build failure against Centos5 (RHEL5)
3728 Details : Use getpagesize() instead of PAGE_SIZE.
3730 --------------------------------------------------------------------------------
3732 2007-07-30 Cluster File Systems, Inc. <info@clusterfs.com>
3734 * Support for kernels:
3735 2.6.5-7.283 (SLES 9),
3736 2.6.9-55.EL (RHEL 4),
3737 2.6.16.46-0.14 (SLES 10),
3738 2.6.18.8 vanilla (kernel.org)
3739 * Client support for unpatched kernels:
3740 (see http://wiki.lustre.org/index.php?title=Patchless_Client)
3741 2.6.16 - 2.6.22 vanilla (kernel.org)
3742 * Due to recently discovered recovery problems, we do not recommend
3743 using patchless RHEL 4 clients with this or any earlier release.
3744 * Recommended e2fsprogs version: 1.39.cfs8
3745 * Note that reiserfs quotas are disabled on SLES 10 in this kernel.
3746 * Starting with this release, the ldiskfs backing filesystem required
3747 by Lustre is now in its own package, lustre-ldiskfs. This package
3748 should be installed. It is versioned separately from Lustre and
3749 may be released separately in future.
3754 Description: server LBUG when shutting down
3755 Details : Block umount forever until the mount refcount is zero rather
3756 than giving up after an arbitrary timeout.
3758 Severity : enhancement
3760 Description: add optional extra BUILD_VERSION info
3761 Details : add a new environment variable (namely LUSTRE_VERS) which allows
3762 to override the lustre version.
3765 Frequency : 2.6.18 servers only
3767 Description: ll_kern_mount() doesn't release the module reference
3768 Details : The ldiskfs module reference count never drops down to 0
3769 because ll_kern_mount() doesn't release the module reference.
3774 Description: server LBUG when using old ost_num_threads parameter
3775 Details : Accept the old ost_num_threads parameter but warn that it
3776 is deprecated, and fix an off-by-one error that caused an LBUG.
3781 Description: Transient SCSI error results in persistent IO issue
3782 Details : iobuf->dr_error is not reinitialized to 0 between two
3786 Frequency : sometimes when underlying device returns I/O errors
3788 Description: OSTs not going read-only during write failures
3789 Details : OSTs are not remounted read-only when the journal commit threads
3790 get I/O errors because fsfilt_ext3 calls journal_start/stop()
3791 instead of the ext3 wrappers.
3794 Frequency : only with 10000 clients or more
3796 Description: poor connect scaling with increasing client count
3797 Details : Don't run filter_grant_sanity_check for more than 100 exports
3798 to improve scaling for large numbers of clients.
3801 Frequency : SLES10 only
3803 Description: sanity-quota.sh quotacheck failed: rc = -22
3804 Details : Quotas cannot be enabled on SLES10.
3807 Frequency : liblustre clients only
3809 Description: getdirentries does not give error when run on compute nodes
3810 Details : getdirentries does not fail when the size specified as an argument
3811 is too small to contain at least one entry
3813 Severity : enhancement
3815 Description: Add LNET router traceability for debug purposes
3816 Details : If a checksum failure occurs with a router as part of the
3817 IO path, the NID of the last router that forwarded the bulk data
3818 is printed so it can be identified.
3823 Description: OST "spontaneously" evicts client; client has imp_pingable == 0
3824 Details : Due to a race condition, liblustre clients were occasionally
3825 evicted incorrectly.
3827 Severity : enhancement
3829 Description: lfs setstripe use optional parameters instead of postional
3832 Severity : enhancement
3834 Description: Nanosecond timestamp support for ldiskfs
3835 Details : The on-disk ldiskfs filesystem has added support for nanosecond
3836 resolution timestamps. There is not yet support for this at
3837 the Lustre filesystem level.
3840 Frequency : during server recovery
3842 Description: MDS failing to send precreate requests due to OSCC_FLAG_RECOVERING
3843 Details : request with rq_no_resend flag not awake l_wait_event if they get a
3847 Frequency : nfs export on patchless client
3849 Description: connectathon hang when test nfs export over patchless client
3850 Details : Disconnected dentry cannot be found with lookup, so we do not need
3851 to unhash it or make it invalid
3854 Description: fix llapi_lov_get_uuids() to allow many OSTs to be returned
3855 Details: : Change llapi_lov_get_uuids() to read the UUIDs from /proc instead
3856 of using an ioctl. This allows lfsck for > 160 OSTs to succeed.
3861 Description: open req refcounting wrong on reconnect
3862 Details : If reconnect happened between getting open reply from server and
3863 call to mdc_set_replay_data in ll_file_open, we will schedule
3864 replay for unreferenced request that we are about to free.
3865 Subsequent close will crash in variety of ways.
3866 Check that request is still eligible for replay in
3867 mdc_set_replay_data().
3872 Description: disable writes to filesystem when reading health_check file
3873 Details : the default for reading the health_check proc file has changed
3874 to NOT do a journal transaction and write to disk, because this
3875 can cause reads of the /proc file to hang and block HA state
3876 checking on a healthy but otherwise heavily loaded system. It
3877 is possible to return to the previous behaviour during configure
3878 with --enable-health-write.
3880 Severity : enhancement
3882 Description: 64-bit inode version
3883 Details: : Add a on-disk 64-bit inode version for ext3 to track changes made
3884 to the inode. This will be required for version-based recovery.
3889 Description: MDS fails to start if a duplicate client export is detected
3890 Details : in some rare cases it was possible for a client to connect to
3891 an MDS multiple times. Upon recovery the MDS would detect this
3892 and fail during startup. Handle this more gracefully.
3894 Severity : enhancement
3896 Description: Add -o localflock option to simulate old noflock behaviour.
3897 Details : This will achieve local-only flock/fcntl locks coherentness.
3902 Description: log_commit_thread vs filter_destroy race leads to crash
3903 Details : Take import reference before releasing llog record semaphore
3908 Description: Wrong request locking in request set processing
3909 Details : ptlrpc_check_set wrongly uses req->rq_lock for proctect add to
3910 imp_delayed_list, in this place should be used imp_lock.
3913 Frequency : when reconnecting
3915 Description: Grant leak when OSC reconnect to OST
3916 Details : When osc reconnect ost, OST(filter) should check whether it
3917 should grant more space to client by comparing fed_grant and
3918 cl_avail_grant, and return the granted space to client instead
3919 of "new granted" space, because client will call osc_init_grant
3920 to update the client grant space info.
3923 Frequency : when client reconnects to OST
3925 Description: Grant leak when OSC does a resend and replays bulk write
3926 Details : When osc reconnect to OST, OST(filter) should clear grant info of
3927 bulk write request, because the grant info while be sync between
3928 OSC and OST when reconnect, and we should ignore the grant info
3929 these of resend/replay write req.
3934 Description: Grant space more than avaiable space sometimes.
3935 Details : When then OST is about to be full, if two bulk writing from
3936 different clients came to OST. Accord the avaliable space of the
3937 OST, the first req should be permitted, and the second one
3938 should be denied by ENOSPC. But if the seconde arrived before
3939 the first one is commited. The OST might wrongly permit second
3940 writing, which will cause grant space > avaiable space.
3943 Frequency : when client is evicted
3945 Description: Grant might be wrongly erased when osc is evicted by OST
3946 Details : when the import is evicted by server, it will fork another
3947 thread ptlrpc_invalidate_import_thread to invalidate the
3948 import, where the grant will be set to 0. While the original
3949 thread will update the grant it got when connecting. So if
3950 the former happened latter, the grant will be wrongly errased
3951 because of this race.
3956 Description: Checking Stale with correct fid
3957 Details : ll_revalidate_it should uses de_inode instead of op_data.fid2
3958 to check whether it is stale, because sometimes, we want the
3959 enqueue happened anyway, and op_data.fid2 will not be initialized.
3961 Severity : enhancement
3963 Description: update patchless client
3964 Details : Add support for patchless client with 2.6.20, 2.6.21 and RHEL 5
3967 Frequency : only with 2.4 kernel
3969 Description: random memory corruption
3970 Details : size of struct ll_inode_info is to big for union inode.u and this
3971 can be cause of random memory corruption.
3976 Description: Memory leak in recovery
3977 Details : Lov_mds_md was not free in an error handler in mds_create_object.
3978 It should also check obd_fail before fsfilt_start, otherwise if
3979 fsfilt_start return -EROFS,(failover mds during mds recovery).
3980 then the req will return with repmsg->transno = 0 and rc = EROFS.
3981 and we met hit the assert LASSERT(req->rq_reqmsg->transno ==
3982 req->rq_repmsg->transno) in ptlrpc_replay_interpret. Fcc should
3983 be freed no matter whether fsfilt_commit success or not.
3986 Frequency : only with large numbers of cores on a single node
3988 Description: Prevents from taking the superblock lock in llap_from_page for
3990 Details : using LL_ORIGIN_REMOVEPAGE origin flag instead of LL_ORIGIN_UNKNOW
3991 for llap_from_page call in ll_removepage() prevents from taking
3992 the superblock lock for a soon died page.
3997 Description: Not check open intent error before release open handle
3998 Details : in some rare cases, the open intent error is not checked before
3999 release open handle, which may cause
4000 ASSERTION(open_req->rq_transno != 0), because it tries to release
4001 the failed open handle.
4006 Description: Set cat log bitmap only after create log success.
4007 Details : in some rare cases, the cat log bitmap is set too early. and it
4008 should be set only after create log success.
4012 Description: Accessing a block bevice can re-enable I/O when Lustre is
4013 tearing down a device.
4014 Details : dev_clear_rdonly(bdev) must be called in kill_bdev() instead of
4019 Description: service threads may hog cpus when there are a lot of requests
4020 Details : Insert cond_resched to give other threads a chance to use some CPU
4025 Description: the cat log was not initialized in recovery
4026 Details : When mds(mgs) do recovery, the tgt_count might be zero, so the
4027 unlink log on mds will not be initialized until mds post
4028 recovery. And also in mds post recovery, the unlink log will
4029 initialization will be done asynchronausly, so there will be race
4030 between add unlink log and unlink log initialization.
4034 Description: brw_stats were being printed incorrectly
4035 Details : brw_stats were being printed as log2 but all of them were not
4036 recorded as log2. Also remove some code duplication arising from
4037 filter_tally_{read,write}.
4041 Frequency : rare, only in recovery.
4042 Description: ASSERTION(req->rq_type != LI_POISON) failed
4043 Details : imp_lock should be held while iterating over imp_sending_list for
4044 prevent destroy request after get timeout in ptlrpc_queue_wait.
4048 Description: replay-single.sh test 52 fails
4049 Details : A lock's skiplist need to be cleanup when it being unlinked
4050 from its resource list.
4055 Description: Short directio read returns full requested size rather than
4057 Details : Direct I/O operations should return actual amount of bytes
4058 transferred rather than requested size.
4060 Severity : enhancement
4062 Description: metadata RPC reduction (e.g. for rm performance)
4063 Details : decrease the amount of synchronous RPC between clients and servers
4064 by canceling conflicing lock before the operation on the client
4065 and packing thier handles into the main operation RPC to server.
4067 Severity : enhancement
4069 Description: add #ifdef HAVE_KERNEL_CONFIG_H
4070 Details : kernels from 2.6.19 not need include linux/config.h, but add
4071 include linux/autoconf.h in commpiler command line.
4073 Severity : enhancement
4075 Description: patchless client support for 2.6.22 kernel
4076 Details : 2.6.22 has only one visble change, SLAB_CTOR_* constants is
4077 removed. In this case we need drop using os depended interface to
4078 kmem_cache and use cfs_mem_cache API.
4083 Description: fix mal-formatted messages
4084 Details : fix some mal-formatted DEBUG_REQ and LCONSOLE_ERROR_MSG messages
4088 Frequency : always in liblustre
4089 Description: wrong IS_ERR implementation in liblustre.h
4090 Details : fix IS_ERR implementation in liblustre.h for right detect errors.
4095 Description: Correct condition for output debug message.
4096 Details : inode i_nlink equal zero is not enough for output message about
4097 disk corruption, i_ctime and i_mode should be also checked.
4101 Frequency : always in patchless client
4102 Description: add configure check for truncate_complete_page
4103 Details : improve checks for exported symbols. This allow run check without
4104 sources, but with Module.symvers shipped with kernel distribution.
4105 add check for truncate_complete_page used by patchless client.
4109 Description: sanity.sh test_77h fails with "test_77h file compare failed"
4110 Details : test_77h uses a file which was messed by other test case.
4114 Description: Not Check whether lov_tgts is NULL in some lov functions
4115 Details : Checking whether lov_tgts is NULL in some functions.
4119 Description: replace obdo_alloc() with OBDO_ALLOC macro
4120 Details : nothing special is done in obdo_alloc() function, and for
4121 debugging purpose, it needs to be replaced with macros.
4125 Description: bad return value and errno from fcntl call
4126 Details : In liblustre API, errno should be a negative value if error
4131 Description: ptlrpc_check_set() LBUG
4132 Details : In case of positive reply from server and failed client bulk
4133 callback after bulk transfer shouldn't LBUG, but process this
4134 request as erroneous.
4136 Severity : enhancement
4138 Description: VFS operations stats tool.
4139 Details : Tool which collects stats by tracking value written in pid,
4140 ppid, gid and uses llstat to generate output to plot graph using
4142 Updated lustre/utils/Makefile.am
4143 Added lustre/utils/ltrack_stats.c
4145 Severity : enhancement
4147 Description: 2.6.18 server support (lustre 1.6.1)
4148 Details : Support for 2.6.18 kernels on the server side.
4153 Description: ASSERTION(imp->imp_conn_current) failed
4154 Details : an assertion failure is hit if a client node boots and attempts to
4155 mount a lustre filesystem faster than RECONNECT_INTERVAL seconds.
4158 Frequency : only for i686
4160 Description: 1.4.11 RC1 build fails for RHEL 4, i686
4161 Details : Fixed config variable for build.
4166 Description: Updated patchess for new RHEL4 kernel
4167 Details : Updated patch inode-nr_unused-2.6.9-rhel4.patch
4168 Updated patch jbd-stats-2.6.9.patch
4169 Updated patch qsnet-rhel4-2.6.patch
4170 Updated patch quota-deadlock-on-pagelock-core.patch
4171 Updated patch vfs_intent-2.6-rhel4.patch
4172 Updated patch vfs_races-2.6-rhel4.patch
4173 Updated series file 2.6-rhel4-titech.series
4174 Updated series file 2.6-rhel4.series
4175 Updated kernel config files
4180 Description: lquota slave complains LBUG when reconnecting with mds
4182 Details : quota slave depends qctxt->lqc_import to send its quota request.
4183 This pointer will be invalid if mds did failover or broke its
4184 connect to osts, which leads to LBUG.
4187 Frequency : when qunit size is too small(less than 20M)
4189 Description: write is stopped by improper -EDQUOT
4190 Details : If the master is busy and qunit size is small enough(let's say 1M),
4191 the slave can not get quota from master on time, which will lead
4192 slave to trigger a -EQUOTA to client.
4197 Description: Deadlock during metadata tests
4198 Details : in prune_dir_dentries(), shrink_dcache_parent() should not be
4199 called with the per-dentry lock held.
4202 Frequency : SLES9 only
4204 Description: Lustre patched kernel for SLES9 SP3 has NR_CPUS set to 8
4205 Details : set CONFIG_NR_CPUS to 128 instead of 8.
4207 Severity : enhancement
4209 Description: remove fs_prep_san_write operation and related patches
4210 Details : remove the ext3-san-jdike patches which are no longer useful.
4215 Description: LDISKFS-fs error (device sdc): ldiskfs_free_blocks
4216 Details : a disk corruption can cause the mballoc code to assert on a
4217 double free or other extent corruptions. Handle these with
4218 ext3_error() instead of assertions.
4221 Frequency : only with mballoc3 code and deep extent trees
4223 Description: ldiskfs_ext_search_right: bad header in inode: unexpected eh_depth
4224 Details : a wrong check of extent headers in ldiskfs_ext_search_right()
4225 can cause the filesystem to be remounted read-only.
4230 Description: lfsck built against 1.4.x cannot run against 1.6.0 lustre
4231 Details : the definition for OBD_IOC_GETNAME changed in 1.6.0. One of the
4232 few external users of this ioctl number is lfsck's call to
4233 llapi_lov_get_uuids() and this caused lfsck to fail at startup.
4234 Add the old ioctl number to the handler so both old and new
4239 Description: parallel lock callbacks
4240 Details : Instead of sending blocking and completion callbacks as separated
4241 requests, adding them to a set and sending in parallel.
4245 Description: Disable most debugging by default
4246 Details : To improve performance, disable most logging (for debug purposes)
4247 by default. VFSTRACE, RPCTRACE, and DLMTRACE are now off by
4248 default, and HA includes fewer messages.
4252 Frequency : only run on patchless client.
4253 Description: use do_facet on sanity.sh for test handling recoverables errors
4254 Details : use do_facet instead of direct use sysctl for set fail_loc on OST
4258 Description: hash tables for lists of nids, connections and uuids
4259 Details : Hash tables noticeably help when a lot of clients connect to a
4260 server, to faster identify duplicate connections or reconnects,
4261 also to faster find export to evict in manual eviction case.
4265 Description: Sometimes, when the server evict a client, and the client will
4266 not be evicted as soon as possible.
4267 Details : In enqueue req, the error was returned by intent, instead of
4268 rq_status which make ptlrpc layer not detect this error, and
4269 does not evict the client. So enqueue error should be returned
4273 Frequency : only at startup
4275 Description: Delay client connections to MDT until fist MDT->OST connect
4276 Details : If a client tried to create a new file before the MDT had
4277 connected to any OSTs, the create would return EIO. Now
4278 the client will simply block until the MDT connects to the
4279 first OST and the create can succeed.
4282 Frequency : only for SLES9
4284 Description: Routinely utilize latest Quadrics drivers in CFS releases
4285 Details : Update patch qsnet-suse-2.6.patch.
4288 Frequency : only for sles10
4290 Description: Update patches for SLES 10 SP1 kernel.
4291 Details : Update the patch vfs_intent-2.6-sles10.patch.
4296 Description: Routinely utilize latest Quadrics drivers in CFS releases
4297 Details : Update patch qsnet-rhel4-2.6.patch.
4300 Frequency : at statup only
4302 Description: mds_lov_synchronize race leads to various problems
4303 Details : simultaneous MDT->OST connections at startup can cause the
4304 sync to abort, leaving the OSC in a bad state.
4309 Description: Using wrong pointer in osc_brw_prep_request
4310 Details : Access to array[-1] can produce panic if kernel compiled with
4311 CONFIG_PAGE_ALLOC enabled
4313 Severity : enhancement
4315 Description: Async OSC create to avoid the blocking unnecessarily.
4316 Details : If a OST has no remain object, system will block on the creating
4317 when need to create a new object on this OST. Now, ways use
4318 pre-created objects when available, instead of blocking on an
4319 empty osc while others are not empty. If we must block, we block
4320 for the shortest possible period of time.
4324 Frequency : only in recovery
4325 Description: Mark OST as early accessible if his start SYNC.
4326 Details : osc_precreate return flag early accessible if oscc marked as
4327 OSCC_FLAG_SYNC_IN_PROGRESS.
4332 Description: Sometimes precreate code can triger create object on wrong ost
4333 Details : Wrong protected or not not restored variables aftre precreate loop
4334 can produce creation object on wrong ost.
4337 Frequency : oss recovery
4339 Description: llog_commit_thread cleanup should sync with llog_commit_thread
4341 Details : llog_commit_thread_count should be synced between llog_commit
4342 start and cleanup, so new llog_commit thread should not be started
4343 when llog_commit threads being stopped to avoid accessing some
4346 Severity : enhancement
4348 Description: Add printing inode info into message about error in writepage.
4350 --------------------------------------------------------------------------------
4352 2007-05-03 Cluster File Systems, Inc. <info@clusterfs.com>
4357 Frequency : on some architectures
4359 Description: 1.6 client sometimes fails to mount from a 1.4 MDT
4360 Details : Uninitialized flags sometimes cause configuration commands to
4364 Frequency : patchless clients only
4366 Description: missing __iget() symbol export
4367 Details : The __iget() symbol export is missing. To avoid the need for
4368 this on patchless clients the deathrow inode reaper is turned
4369 off, and we depend on the VM to clean up old inodes. This
4370 dependency was during via the fix for bug 12181.
4375 Description: sanity.sh fail: test_52b
4376 Details : The ll_inode_to_ext_flags() has a glitch which makes MDS return
4377 incorrect inode's flags to client.
4379 --------------------------------------------------------------------------------
4381 2007-04-19 Cluster File Systems, Inc. <info@clusterfs.com>
4383 * CONFIGURATION CHANGE. This version of Lustre WILL NOT
4384 INTEROPERATE with older versions automatically. In many cases a
4385 special upgrade step is needed. Please read the
4386 user documentation before upgrading any part of a 1.4.x system.
4387 * WARNING: Lustre configuration and startup changes are required with
4388 this release. See https://mail.clusterfs.com/wikis/lustre/MountConf
4390 * Support for kernels:
4391 2.4.21-47.0.1.EL (RHEL 3),
4392 2.6.5-7.283 (SLES 9),
4393 2.6.9-42.0.10.EL (RHEL 4),
4394 2.6.12.6 vanilla (kernel.org),
4395 2.6.16.27-0.9 (SLES10)
4396 * Client support for unpatched kernels:
4397 (see https://mail.clusterfs.com/wikis/lustre/PatchlessClient)
4398 2.6.16 - 2.6.19 vanilla (kernel.org),
4399 2.6.9-42.0.8EL (RHEL 4)
4400 * Recommended e2fsprogs version: 1.39.cfs6
4401 * Note that reiserfs quotas are disabled on SLES 10 in this kernel
4404 Severity : enhancement
4406 Description: Async OSC create to avoid the blocking unnecessarily.
4407 Details : If a OST has no remain object, system will block on the creating
4408 when need to create a new object on this OST. Now, ways use
4409 pre-created objects when available, instead of blocking on an
4410 empty osc while others are not empty. If we must block, we block
4411 for the shortest possible period of time.
4413 Severity : enhancement
4415 Description: MountConf
4416 Details : Lustre configuration is now managed via mkfs and mount
4417 commands instead of lmc and lconf. New obd types (MGS, MGC)
4418 are added for dynamic configuration management. See
4419 https://mail.clusterfs.com/wikis/lustre/MountConf for
4422 Severity : enhancement
4424 Description: dynamic OST addition
4425 Details : OSTs can now be added to a live filesystem
4427 Severity : enhancement
4429 Description: startup order invariance
4430 Details : MDTs and OSTs can be started in any order. Clients only
4431 require the MDT to complete startup.
4433 Severity : enhancement
4435 Description: parallel, asynchronous orphan cleanup
4436 Details : orphan cleanup is now performed in separate threads for each
4437 OST, allowing parallel non-blocking operation.
4439 Severity : enhancement
4441 Description: optimized stripe assignment
4442 Details : stripe assignments are now made based on ost space available,
4443 ost previous usage, and OSS previous usage, in order to try
4444 to optimize storage space and networking resources.
4446 Severity : enhancement
4448 Description: Permanently set tunables
4449 Details : All writable /proc/fs/lustre tunables can now be permanently
4450 set on a per-server basis, at mkfs time or on a live
4453 Severity : enhancement
4455 Description: Lustre message v2
4456 Details : Add lustre message format v2.
4458 Severity : enhancement
4460 Description: client OST exclusion list
4461 Details : Clients can be started with a list of OSTs that should be
4462 declared "inactive" for known non-responsive OSTs.
4466 Description: ENOENT returned for valid filehandle during dbench.
4467 Details : Check if a directory has children when invalidating dentries
4468 associated with an inode during lock cancellation. This fixes
4469 an incorrect ENOENT sometimes seen for valid filehandles during
4470 testing with dbench.
4473 Frequency : SFS test only (otherwise harmless)
4475 Description: SPEC SFS validation failure on NFS v2 over lustre.
4476 Details : Changes the blocksize for regular files to be 2x RPC size,
4477 and not depend on stripe size.
4479 Severity : enhancement
4481 Description: fine-grained SMP locking inside DLM
4482 Details : Improve DLM performance on SMP systems by removing the single
4483 per-namespace lock and replace it with per-resource locks.
4485 Severity : enhancement
4487 Description: don't hold multiple extent locks at one time
4488 Details : To avoid client eviction during large writes, locks are not
4489 held on multiple stripes at one time or for very large writes.
4490 Otherwise, clients can block waiting for a lock on a failed OST
4491 while holding locks on other OSTs and be evicted.
4493 Severity : enhancement
4495 Description: Multiple MD RPCs in flight.
4496 Details : Further unserialise some read-only MDT RPCs - learn about intents.
4497 To avoid overly-overloading MDT, introduce a limit on number of
4498 MDT RPCs in flight for a single client and add /proc controls
4499 to adjust this limit.
4501 Severity : enhancement
4503 Description: client read/write statistics
4504 Details : Add client read/write call usage stats for performance
4505 analysis of user processes.
4506 /proc/fs/lustre/llite/*/offset_stats shows non-sequential
4507 file access. extents_stats shows chunk size distribution.
4508 extents_stats_per_process show chunk size distribution per
4511 Severity : enhancement
4513 Description: per-client statistics on server
4514 Details : Add ldlm and operations statistics for each client in
4515 /proc/fs/lustre/mds|obdfilter/*/exports/
4517 Severity : enhancement
4519 Description: improved MDT statistics
4520 Details : Add detailed MDT operations statistics in
4521 /proc/fs/lustre/mds/*/stats
4523 Severity : enhancement
4525 Description: VFS operations stats
4526 Details : Add client VFS call stats, trackable by pid, ppid, or gid
4527 /proc/fs/lustre/llite/*/stats_track_[pid|ppid|gid]
4532 Description: Fix client-side osc byte counters
4533 Details : The osc read/write byte counters in
4534 /proc/fs/lustre/osc/*/stats are now working
4537 Frequency : always as root on SLES
4539 Description: Failure of copying files with lustre special EAs.
4540 Details : Client side always return success for setxattr call for lustre
4541 special xattr (currently only "trusted.lov").
4546 Description: Refcount LNET uuids
4547 Details : The global LNET uuid list grew linearly with every startup;
4548 refcount repeated list entries instead of always adding to
4551 Severity : enhancement
4553 Description: Dynamic service threads
4554 Details : Within a small range, start extra service threads
4555 automatically when the request queue builds up.
4558 Frequency : mixed-endian client/server environments
4560 Description: mixed-endian crashes
4561 Details : The new msg_v2 system had some failures in mixed-endian
4564 Severity : enhancement
4566 Description: Easy OST removal
4567 Details : OSTs can be permanently deactivated with e.g. 'lctl
4568 conf_param lustre-OST0001.osc.active=0'
4570 Severity : enhancement
4572 Description: MGS proc entries
4573 Details : Added basic proc entries for the MGS showing what filesystems
4576 Severity : enhancement
4578 Description: provide MGS failover
4579 Details : Added config lock reacquisition after MGS server failover.
4581 Severity : enhancement
4583 Description: add Linux 2.4 support
4584 Details : Added support for RHEL 2.4.21 kernel for 1.6 servers and clients
4588 Description: a large application tries to do I/O to the same resource and dies
4589 in the middle of it.
4590 Details : Check the req->rq_arrival time after the call to
4591 ost_brw_lock_get(), but before we do anything about
4592 processing it & sending the BULK transfer request. This
4593 should help move old stale pending locks off the queue as
4594 quickly as obd_timeout.
4597 Frequency : when an incorrect nid is specified during startup
4599 Description: ptlrpc connect to non-existant node causes kernel crash
4600 Details : LNET can't be re-entered from an event callback, which
4601 happened when we expire a message after the export has been
4602 cleaned up. Instead, hand the zombie cleanup off to another
4605 Severity : enhancement
4607 Description: plain/inodebits lock performance improvement
4608 Details : Grouping plain/inodebits in granted list by their request modes
4609 and bits policy, thus improving the performance of search through
4613 Frequency : only if OST filesystem is corrupted
4615 Description: client incorrectly hits assertion in ptlrpc_replay_req()
4616 Details : for a short time RPCs with bulk IO are in the replay list,
4617 but replay of bulk IOs is unimplemented. If the OST filesystem
4618 is corrupted due to disk cache incoherency and then replay is
4619 started it is possible to trip an assertion. Avoid putting
4620 committed RPCs into the replay list at all to avoid this issue.
4623 Frequency : liblustre (e.g. catamount) on a large cluster with >= 8 OSTs/OSS
4625 Description: System hang on startup
4626 Details : This bug allowed the liblustre (e.g. catamount) client to
4627 return to the app before handling all startup RPCs. This
4628 could leave the node unresponsive to lustre network traffic
4629 and manifested as a server ptllnd timeout.
4631 Severity : enhancement
4633 Description: Add "/proc/sys/lustre/debug_peer_on_timeout"
4634 Details : liblustre envirable: LIBLUSTRE_DEBUG_PEER_ON_TIMEOUT
4635 boolean to control whether to print peer debug info when a
4636 client's RPC times out.
4639 Frequency : only for kernels with patches from Lustre below 1.4.3
4641 Description: Remove old rdonly API
4642 Details : Remove old rdonly API which unused from at least lustre 1.4.3
4645 Frequency : only for devices with external journals
4647 Description: Set external device read-only also
4648 Details : During a commanded failover stop, we set the disk device
4649 read-only while the server shuts down. We now also set any
4650 external journal device read-only at the same time.
4653 Frequency : when upgrading from 1.4 while trying to change parameters
4655 Description: The wrong (new) MDC name was used when setting parameters for
4656 upgraded MDT's. Also allows changing of OSC (and MDC)
4657 parameters if --writeconf is specified at tunefs upgrade time.
4660 Frequency : when setting specific ost indicies
4662 Description: QOS code breaks on skipped indicies
4663 Details : Add checks for missing OST indicies in the QOS code, so OSTs
4664 created with --index need not be sequential.
4666 Severity : enhancement
4668 Description: Add uninit_groups feature to ldiskfs2 to speed up e2fsck
4669 Details : The uninit_groups feature works in conjunction with the kernel
4670 filesystem code (ldiskfs2 only) and e2fsprogs-1.39-cfs6 to speed
4671 up the pass1 processing of e2fsck. This is a read-only feature
4672 in ldiskfs2 only, so older kernels and current ldiskfs cannot
4673 mount filesystems that have had this feature enabled.
4675 Severity : enhancement
4677 Description: Improve multi-block allocation algorithm to avoid fragmentation
4678 Details : The mballoc3 code (ldiskfs2 only) adds new mechanisms to improve
4679 allocation locality and avoid filesystem fragmentation.
4681 ------------------------------------------------------------------------------
4683 2007-04-01 Cluster File Systems, Inc. <info@clusterfs.com>
4685 * Support for kernels:
4686 2.4.21-47.0.1.EL (RHEL 3)
4687 2.6.5-7.283 (SLES 9)
4688 2.6.9-42.0.10.EL (RHEL 4)
4689 2.6.12.6 vanilla (kernel.org)
4690 2.6.16.27-0.9 (SLES 10)
4691 * Recommended e2fsprogs version: 1.39.cfs5
4693 * Note that reiserfs quotas are disabled on SLES 10 in this kernel
4697 Frequency : occasional, depends on client load and configuration
4698 Bugzilla : 12181, 12203
4699 Description: data loss for recently-modified files
4701 Details : In some cases it is possible that recently written or created
4702 files may not be written to disk in a timely manner (this should
4703 normally be within 30s unless client IO load is very high).
4704 The problem appears as zero-length files or files that are a
4705 multiple of 1MB in size after a client crash or client eviction
4706 that are missing data at the end of the file.
4708 This problem is more likely to be hit on clients where files are
4709 repeatedly created and unlinked in the same directory, clients
4710 have a large amount of RAM, have many CPUs, the filesystem has
4711 many OSTs, the clients are rebooted frequently, and/or the files
4712 are not accessed by other nodes after being written.
4714 The presence of the problem can be detected by looking at
4715 /proc/sys/fs/inode-state. If the first number (nr_inodes) is
4716 smaller than the second (nr_unused) then dirty files will not
4717 be flushed automatically to disk. "sync; sleep 10" should be
4718 run several times on the node before unmounting it to update
4719 Lustre (this is also safe to run on nodes without this problem).
4721 There is also a related kernel bug in the RHEL4 4 2.6.9 kernel
4722 that can cause this same problem, so customers using that kernel
4723 also need to update the kernel in addition to Lustre. In order
4724 to properly fix this bug, the RHEL3 2.4.21 kernel is also updated.
4726 It is normal that files written just before a client crash (less
4727 than 30s) may not yet have been flushed to disk, even for local
4731 Frequency : frequent on thin XT3 nodes
4733 Description: UUID collision on thin XT3 Linux nodes
4734 Details : UUIDs on Compute Node Linux XT3 nodes were not generated
4735 randomly, since we relied on an insufficiently-seeded PRNG.
4740 Description: OSS hangs after "All ost request buffers busy"
4741 Details : A deadlock between quota and journal operations caused OSS
4742 hangs after printing "All ost request buffers busy."
4745 Frequency : always on liblustre builds
4747 Description: Cleanup compiler warnings on liblustre
4750 Frequency : always on liblustre builds on XT3
4752 Description: LC_CONFIG_CDEBUG don't run while build liblustre on XT3.
4756 Description: Addition of EXT3_FEATURE_RO_COMPAT_DIR_NLINKS flag for
4757 > 32000 subdirectories
4758 Details : Add EXT3_FEATURE_RO_COMPAT_DIR_NLINK flag to
4759 EXT3_FEATURE_RO_COMPAT_SUPP. This flag will be set whenever
4760 subdirectory count crosses 32000. This will aid e2fsck to
4761 correctly handle more than 32000 subdirectories.
4764 Frequency : liblustre (e.g. catamount) on a large cluster with >= 8 OSTs/OSS
4766 Description: System hang on startup
4767 Details : This bug allowed the liblustre (e.g. catamount) client to
4768 return to the app before handling all startup RPCs. This
4769 could leave the node unresponsive to lustre network traffic
4770 and manifested as a server ptllnd timeout.
4772 Severity : enhancement
4774 Description: Add "/proc/sys/lustre/debug_peer_on_timeout"
4775 (liblustre envirable: LIBLUSTRE_DEBUG_PEER_ON_TIMEOUT)
4776 boolean to control whether to print peer debug info when a
4777 client's RPC times out.
4782 Description: make O_SYNC working on 2.6 kernels
4783 Details : 2.6 kernels use different method for mark pages for write,
4784 so need add a code to lustre for O_SYNC work.
4789 Description: Failure to close file and release space on NFS
4790 Details : Put inode details into lock acquired in ll_intent_file_open.
4791 Use mdc_intent_lock in ll_intent_open to properly
4792 detect all kind of errors unhandled by mdc_enqueue.
4797 Description: proc file read during shutdown sometimes raced obd removal,
4799 Details : Add lock to prevent obd access after proc file removal.
4802 Frequency : Only for files larger than 4GB on 32-bit clients.
4804 Description: improperly doing page alignment of locks
4805 Details : Modify lustre core code to use CFS_PAGE_* defines instead of
4806 PAGE_*. Make CFS_PAGE_MASK a 64-bit mask.
4811 Description: RPCs being resent when they shouldn't be
4812 Details : Some RPCs that should not be resent are being resent. This
4813 can cause inconsistencies in the RPC state machine. Do not
4814 resend such requests.
4817 Frequency : rare, only with NFS export
4819 Description: Crash on NFS re-export node
4820 Details : under very unusual load conditions an assertion is hit in
4821 ll_intent_file_open()
4824 Frequency : only if OST filesystem is corrupted
4826 Description: client incorrectly hits assertion in ptlrpc_replay_req()
4827 Details : for a short time RPCs with bulk IO are in the replay list,
4828 but replay of bulk IOs is unimplemented. If the OST filesystem
4829 is corrupted due to disk cache incoherency and then replay is
4830 started it is possible to trip an assertion. Avoid putting
4831 committed RPCs into the replay list at all to avoid this issue.
4836 Description: large O_DIRECT requests fail under memory pressure/fragmentation
4837 Details : Large single O_DIRECT read and write calls can fail to allocate
4838 a sufficiently large buffer to process the request. In case of
4839 allocation failure the allocation is retried with a smaller
4840 buffer and broken into smaller requests.
4842 Severity : enhancement
4844 Description: Add -o localflock option to simulate old noflock behaviour.
4845 Details : This will achieve local-only flock/fcntl locks coherentness.
4850 Description: versioning check is incomplete
4851 Details : Checking the version difference of client vs. server, report
4852 error if the gap is too big.
4857 Description: add support PG_writeback bit
4858 Details : add support for PG_writeback bit for Lustre, for more carefull
4859 work with page cache in 2.6 kernel. This also fix some deadlocks
4860 and remove hack for work O_SYNC with 2.6 kernel.
4862 Severity : enhancement
4864 Description: Add uninit_groups feature to ldiskfs2 to speed up e2fsck
4865 Details : The uninit_groups feature works in conjunction with the kernel
4866 filesystem code (ldiskfs2 only) and e2fsprogs-1.39-cfs6 to speed
4867 up the pass1 processing of e2fsck. This is a read-only feature
4868 in ldiskfs2 only, so older kernels and current ldiskfs cannot
4869 mount filesystems that have had this feature enabled.
4871 Severity : enhancement
4873 Description: Improve multi-block allocation algorithm to avoid fragmentation
4874 Details : The mballoc3 code (ldiskfs2 only) adds new mechanisms to improve
4875 allocation locality and avoid filesystem fragmentation.
4877 ------------------------------------------------------------------------------
4879 2007-02-09 Cluster File Systems, Inc. <info@clusterfs.com>
4881 * Support for kernels:
4882 2.6.9-42.0.3.EL (RHEL 4)
4883 2.6.5-7.276 (SLES 9)
4884 2.4.21-47.0.1.EL (RHEL 3)
4885 2.6.12.6 vanilla (kernel.org)
4886 2.6.16.21-0.8 (SLES10)
4887 * Recommended e2fsprogs version: 1.39.cfs2-0
4889 * The backwards-compatible /proc/sys/portals symlink has been removed
4890 in this release. Before upgrading, please ensure that you change
4891 any configuration scripts or /etc/sysctl.conf files that access
4892 /proc/sys/portals/* or sysctl portals.* to use the corresponding
4893 entry in /proc/sys/lnet or sysctl lnet.*. This change can be made
4894 in advance of the upgrade on any system running Lustre 1.4.6 or
4895 newer, since /proc/sys/lnet was added in that version.
4896 * Note that reiserfs quotas are disabled on SLES 10 in this kernel
4900 Frequency : only when quota is used
4902 Description: avoid scanning export list for quota master
4903 Details : Change the algorithms to avoid scanning export list in order
4904 to improve the efficiency.
4907 Frequency : MDS failover only, very rarely
4909 Description: "went back in time" messages on mds failover
4910 Details : The greatest transno may be lost when the current operation
4911 finishes with an error (transno==0) and the client's last_rcvd
4912 record is over-written. Save the greatest transno in the
4913 mds_last_transno for this case.
4916 Frequency : always for specific kernels and striping counts
4918 Description: client may get "Matching packet too big" without ACL support
4919 Details : Clients compiled without CONFIG_FS_POSIX_ACL get an error message
4920 when trying to access files in certain configurations. The
4921 clients should in fact be denied when mounting because they do
4922 not understand ACLs.
4925 Frequency : Cray XT3 with more than 4000 clients and multiple jobs
4927 Description: many clients connecting with IO in progress causes connect timeouts
4928 Details : Avoid synchronous journal commits to avoid delays caused by many
4929 clients connecting/disconnecting when bulk IO is in progress.
4930 Queue liblustre connect requests on OST_REQUEST_PORTAL instead of
4931 OST_IO_PORTAL to avoid delays behind potentially many pending
4935 Frequency : occasionally with multiple writers to a single file
4937 Description: shared writes to file may result in wrong size reported by stat()
4938 Details : Allow growing of kms when extent lock is cancelled
4941 Frequency : always with random mmap IO to multi-striped file
4943 Description: mmap write might be lost if we are writing to a 'hole' in stripe
4944 Details : Only if the hole is at the end of OST object so that kms is too
4945 small. Fix is to increase kms accordingly in ll_nopage.
4948 Frequency : rare, only if OST filesystem is inconsistent with MDS filesystem
4950 Description: writes to a missing object would leak memory on the OST
4951 Details : If there is an inconsistency between the MDS and OST filesystems,
4952 such that the MDS references an object that doesn't exist, writes
4953 to that object will leak memory due to incorrect cleanup in the
4954 error handling path, eventually running out of memory on the OST.
4959 Description: Creating too long symlink causes lustre errors
4960 Details : Check symlink and name lengths before sending requests to MDS.
4963 Frequency : only if flock is enabled (not on by default)
4965 Description: posix locks not released on fd closure on 2.6.9+
4966 Details : We failed to add posix locks to list of inode locks on 2.6.9+
4967 kernels, this caused such locks not to be released on fd close and
4968 then assertions on fs unmount about still used locks.
4971 Frequency : MDS failover only, very rarely
4973 Description: clients may get ASSERTION(granted_lock != NULL)
4974 Details : When request was taking a long time, and a client was resending
4975 a getattr by name lock request. The were multiple lock requests
4976 with the same client lock handle and
4977 mds_getattr_name->fixup_handle_for_resent_request found one of the
4978 lock handles but later failed with ASSERTION(granted_lock != NULL).
4983 Description: handle->h_buffer_credits > 0, assertion failure
4984 Details : h_buffer_credits is zero after truncate, causing assertion
4985 failure. This patch extends the transaction or creates a new
4989 Frequency : NFS re-export or patchless client
4990 Bugzilla : 11179, 10796
4991 Description: Crash on NFS re-export node (__d_move)
4992 Details : We do not want to hash the dentry if we don't have a lock.
4993 But if this dentry is later used in d_move, we'd hit uninitialised
4994 list head d_hash, so we just do this to init d_hash field but
4995 leave dentry unhashed.
4998 Frequency : NFS re-export or patchless client
5000 Description: NFS exports has problem with symbolic link
5001 Details : lustre client didn't properly install dentry when re-exported
5002 to NFS or running patchless client.
5005 Frequency : NFS re-export or patchless client
5007 Description: Various nfs/patchless fixes.
5008 Details : fixes reuse disconected alias for lookup process - this fixes
5009 warning "find_exported_dentry: npd != pd",
5010 fix permission error with open files at nfs.
5011 fix apply umask when do revalidate.
5014 Frequency : occasional
5016 Description: Crash on NFS re-export node
5017 Details : calling clear_page() on the wrong pointer triggered oops in
5018 generic_mapping_read().
5021 Frequency : rarely, using O_DIRECT IO
5023 Description: unaligned directio crashes client with LASSERT
5024 Details : check for unaligned buffers before trying any requests.
5027 Frequency : rarely, using CFS RAID5 patches in non-standard kernel series
5029 Description: stale data returned from RAID cache
5030 Details : If only a small amount of IO is done to the RAID device before
5031 reading it again it is possible to get stale data from the RAID
5032 cache instead of reading it from disk.
5035 Frequency : always for sles10 kernel
5037 Description: sles10 support
5038 Details : ll_follow_link: compile fixes and using of nd_set_link
5039 under newer kernels.
5042 Frequency : depends on arch, kernel and compiler version, always on sles10
5045 Description: recursive or deep enough symlinks cause stack overflow
5046 Details : getting rid of large stack-allocated variable in
5050 Frequency : depends on hardware
5052 Description: lustre write performance loss in the SLES10 kernel
5053 Details : the performance loss is caused by using of write barriers in the
5054 ext3 code. The SLES10 kernel turns barrier support on by
5055 default. The fix is to undo that change for ldiskfs.
5057 ------------------------------------------------------------------------------
5059 2006-12-09 Cluster File Systems, Inc. <info@clusterfs.com>
5061 * Support for kernels:
5062 2.6.9-42.0.3EL (RHEL 4)
5063 2.6.5-7.276 (SLES 9)
5064 2.4.21-47.0.1.EL (RHEL 3)
5065 2.6.12.6 vanilla (kernel.org)
5069 Frequency : quota enabled and large files being deleted
5071 Description: releasing more than 4GB of quota at once hangs OST
5072 Details : If a user deletes more than 4GB of files on a single OST it
5073 will cause the OST to spin in an infinite loop. Release
5074 quota in < 4GB chunks, or use a 64-bit value for 1.4.7.1+.
5079 Description: statfs data retrieved from /proc may be stale or zero
5080 Details : When reading per-device statfs data from /proc, in the
5081 {kbytes,files}_{total,free,avail} files, it may appear
5082 as zero or be out of date.
5085 Frequency : systems with MD RAID1 external journal devices
5087 Description: lconf's call to blkid is confused by RAID1 journal devices
5088 Details : Use the "blkid -l" flag to locate the MD RAID device instead
5089 of returning all block devices that match the journal UUID.
5092 Frequency : always, for aggregate stripe size over 4GB
5094 Description: "lfs setstripe" fails assertion when setting 4GB+ stripe width
5095 Details : Using "lfs setstripe" to set stripe size * stripe count over 4GB
5096 will fail the kernel with "ASSERTION(lsm->lsm_xfersize != 0)"
5099 Frequency : always if "lfs find" used on a local file/directory
5101 Description: "lfs find" segfaults if used on a local file/directory
5102 Details : The case where a directory component was not specified wasn't
5103 handled correctly. Handle this properly.
5106 Frequency : always on ppc64
5108 Description: the write to an ext3 filesystem mounted with mballoc got stuck
5109 Details : ext3_mb_generate_buddy() uses find_next_bit() which does not
5110 perform endianness conversion.
5113 Frequency : rarely (truncate to non-zero file size after write under load)
5114 Bugzilla : 10730, 10687
5115 Description: Files padded with zeros to next 4K multiple
5116 Details : With filesystems mounted using the "extents" option (2.6 kernels)
5117 it is possible that files that are truncated to a non-zero size
5118 immediately after being written are filled with zero bytes beyond
5119 the truncated size. No file data is lost.
5121 Severity : enhancement
5123 Description: Allow recovery/failover for liblustre clients.
5124 Details : liblustre clients were unaware of failover configurations until
5127 Severity : enhancement
5129 Description: user file locks should fail when not mounting with flock option
5130 Details : Set up an error-returning stub in ll_file_operations.lock field
5131 to prevent incorrect behaviour when client is mounted without
5132 flock option. Also, set up properly f_op->flock field for
5136 Frequency : always on ia64
5138 Description: "lfs df" loops on printing out MDS statfs information
5139 Details : The obd_ioctl_data was not initialized and in some systems
5140 this caused a failure during the ioctl that did not return
5141 an error. Initialize the struct and return an error on failure.
5144 Frequency : SLES 9 only
5146 Description: Error of copying files with lustre special EAs as root
5147 Details : Client side always return success for setxattr call for lustre
5148 special xattr (currently only "trusted.lov").
5151 Frequency : rarely on clusters with both ia64+i386 clients
5153 Description: ia64+i686 clients doing shared IO on the same file may LBUG
5154 Details : In rare cases when both ia64+i686 (or other mixed-PAGE_SIZE)
5155 clients are doing concurrent writes to the same file it is
5156 possible that the ia64 clients may LASSERT because the OST
5157 extent locks are not PAGE_SIZE aligned. Ensure that grown
5158 locks are always aligned on the request boundary.
5161 Frequency : specific use, occasional
5163 Description: Overwriting in use executable truncates on-disk binary image
5164 Details : If one node attempts to overwrite an executable in use by
5165 another node, we now correctly return ETXTBSY instead of
5166 truncating the file.
5168 Severity : enhancement
5170 Description: Async OSC create to avoid the blocking unnecessarily.
5171 Details : If a OST has no remain object, system will block on the creating
5172 when need to create a new object on this OST. Now, ways use
5173 pre-created objects when available, instead of blocking on an
5174 empty osc while others are not empty. If we must block, we block
5175 for the shortest possible period of time.
5180 Description: chmod on Lustre root is propagated to other clients
5181 Details : Re-validate root's dentry in ll_lookup_it to avoid having it
5182 invalid by the follow_mount time.
5185 Frequency : liblustre clients only
5187 Description: Race in 'instant cancel' lock handling could lead to such locks
5188 never to be granted in case of SMP MDS
5189 Details : Do not destroy not yet granted but cbpending locks in
5193 Frequency : replay/resend of open
5195 Description: non null lock assetion failure in mds_intent_policy
5196 Details : Trying to replay/resend lockless open requests resulted in
5197 mds_open() returning 0 with no lock. Now it sets a flag if
5198 a lock is going to be returned.
5200 Severity : enhancement
5202 Description: Checksum enhancements
5203 Details : New checksum enhancements allow for resending RPCs that failed
5206 Severity : enhancement
5208 Description: Tunables on number of dirty pages in cacche
5209 Details : Allow to set limit on number of dirty pages cached.
5214 Description: client crash on unmount - lock still has references
5215 Details : In some error handling cases it was possible to leak a lock
5216 reference on a client while accessing a file. This was not
5217 harmful to the client during operation, but would cause the
5218 client to crash when the filesystem is unmounted.
5221 Frequency : specific case, rare
5223 Description: ETXTBSY on mds though file not in use by client
5224 Details : ETXTBSY is no longer incorrectly returned when attempting to
5225 chmod or chown a directory that the user previously tried to
5226 execute or a currently-executing binary.
5229 Frequency : extremely rare except on liblustre-based clients
5231 Description: Lustre space not freed when files are deleted
5232 Details : Clean up open-unlinked files after client eviction. Previously
5233 the unlink was skipped and the files remained as orphans.
5238 Description: OST failure "would be an LBUG" in waiting_locks_callback()
5239 Details : In some cases it was possible to send a blocking callback to
5240 a client doing a glimpse, even though that client didn't get
5241 a lock granted. When the glimpse lock is cancelled on the OST
5242 the freed lock is left on the waiting list and corrupted the list.
5245 Frequency : all core dumps
5247 Description: Broke core dumps to lustre
5248 Details : Negative dentry may be unhashed if parent does not have UPDATE
5249 lock, but some callers, e.g. do_coredump, expect dentry to be
5250 hashed after successful create, hash it in ll_create_it.
5252 ------------------------------------------------------------------------------
5254 2006-09-13 Cluster File Systems, Inc. <info@clusterfs.com>
5256 * Support for kernels:
5257 2.6.9-42.0.2.EL (RHEL 4)
5258 2.6.5-7.276 (SLES 9)
5259 2.4.21-40.EL (RHEL 3)
5260 2.6.12.6 vanilla (kernel.org)
5264 Frequency : always on RHEL 3
5266 Description: Number of open files grows over time
5267 Details : The number of open files grows over time, whether or not
5268 Lustre is started. This was due to a filp leak introduced
5269 by one of our kernel patches.
5271 ------------------------------------------------------------------------------
5273 2006-08-20 Cluster File Systems, Inc. <info@clusterfs.com>
5275 * Support for kernels:
5276 2.6.9-42.EL (RHEL 4)
5277 2.6.5-7.267 (SLES 9)
5278 2.4.21-40.EL (RHEL 3)
5279 2.6.12.6 vanilla (kernel.org)
5284 Bugzilla : 5719, 9635, 9792, 9684
5285 Description: OST (or MDS) trips assertions in (re)connection under heavy load
5286 Details : If a server is under heavy load and cannot reply to new
5287 connection requests before the client resends the (re)connect,
5288 the connection handling code can behave badly if two service
5289 threads are concurrently handing separate (re)connections from
5290 the same client. Add better locking to the connection handling
5291 code, and ensure that only a single connection will be processed
5292 for a given client UUID, even if the lock is dropped.
5294 Severity : enhancement
5296 Description: add TCP zero-copy support to kernel
5297 Details : Add support to the kernel TCP stack to allow zero-copy bulk
5298 sends if the hardware supports scatter-gather and checksumming.
5299 This allows socklnd to do client-write and server-read more
5300 efficiently and reduce CPU utilization from skbuf copying.
5303 Frequency : only if NFS exporting from client
5305 Description: NULL pointer deref in ll_iocontrol() if chattr mknod file
5306 Details : If setting attributes on a file created under NFS that had
5307 never been opened it would be possible to oops the client
5308 if the file had no objects.
5312 Bugzilla : 9326, 10402, 10897
5313 Description: client crash in ptlrpcd_wake() thread when sending async RPC
5314 Details : It is possible that ptlrpcd_wake() dereferences a freed async
5315 RPC. In rare cases the ptlrpcd thread alread processed the RPC
5316 before ptlrpcd_wake() was called and the request was freed.
5319 Frequency : always for liblustre
5321 Description: liblustre client does MDS+OSTs setattr RPC for each write
5322 Details : When doing a write from a liblustre client, the client
5323 incorrectly issued an RPC to the MDS and each OST the file was
5324 striped over in order to update the timestamps. When writing
5325 with small chunks and many clients this could overwhelm the MDS
5326 with RPCs. In all cases it would slow down the write because
5327 these RPCs are unnecessary.
5329 Severity : enhancement
5331 Description: allow number of MDS service threads to be changed at module load
5332 Details : It is now possible to change the number of MDS service threads
5333 running. Adding "options mds mds_num_threads={N}" to the MDS's
5334 /etc/modprobe.conf will set the number of threads for the next
5335 time Lustre is restarted (assuming the "mds" module is also
5336 reloaded at that time). The default number of threads will
5337 stay the same, 32 for most systems.
5342 Description: OST crash if filesystem is unformatted or corrupt
5343 Details : If an OST is started on a device that has never been formatted
5344 or if the filesystem is corrupt and cannot even mount then the
5345 error handling cleanup routines would dereference a NULL pointer.
5350 Description: NULL pointer deref in llap_from_page.
5351 Details : get_cache_page_nowait can return a page with NULL (or otherwise
5352 incorrect) mapping if the page was truncated/reclaimed while it was
5353 searched for. Check for this condition and skip such pages when
5354 doing readahead. Introduce extra check to llap_from_page() to
5355 verify page->mapping->host is non-NULL (so page is not anonymous).
5358 Frequency : Sometimes when using sys_sendfile
5360 Description: "page not covered by a lock" warnings from ll_readpage
5361 Details : sendfile called ll_readpage without right page locks present.
5362 Now we introduced ll_file_sendfile that does necessary locking
5363 around call to generic_file_sendfile() much like we do in
5367 Frequency : with certain MDS communication failures at client mount time
5369 Description: NULL pointer deref after failed client mount
5370 Details : a client connection request may delayed by the network layer
5371 and not be sent until after the PTLRPC layer has timed out the
5372 request. If the client fails the mount immediately it will try
5373 to clean up before the network times out the request. Add a
5374 reference from the request import to the obd device and delay
5375 the cleanup until the network drops the request.
5378 Frequency : occasionally during client (re)connect
5380 Description: assertion failure during client (re)connect
5381 Details : processing a client connection request may be delayed by the
5382 client or server longer than the client connect timeout. This
5383 causes the client to resend the connection request. If the
5384 original connection request is replied in this interval, the
5385 client may trip an assertion failure in ptlrpc_connect_interpret()
5386 which thought it would be the only running connect process.
5389 Frequency : only with obd_echo servers and clients that are rebooted
5391 Description: kernel BUG accessing uninitialized data structure
5392 Details : When running an obd_echo server it did not start the ping_evictor
5393 thread, and when a client was evicted an uninitialized data
5394 structure was accessed. Start the ping_evictor in the RPC
5395 service startup instead of the OBD startup.
5397 Severity : enhancement
5398 Bugzilla : 10193 (patchless)
5399 Description: Remove dependency on various unexported kernel interfaces.
5400 Details : No longer need reparent_to_init, exit_mm, exit_files,
5401 sock_getsockopt, filemap_populate, FMODE_EXEC, put_filp.
5404 Frequency : rare (only users of deprecated and unsupported LDAP config)
5406 Description: write_conf for zeroconf mount queried LDAP incorrectly for client
5407 Details : LDAP apparently contains 'lustreName' attributes instead of
5408 'name'. A simple remapping of the name is sufficient.
5411 Frequency : rare (only with non-default dump_on_timeout debug enabled)
5413 Description: waiting_locks_callback trips kernel BUG if client is evicted
5414 Details : Running with the dump_on_timeout debug flag turned on makes
5415 it possible that the waiting_locks_callback() can try to dump
5416 the Lustre kernel debug logs from an interrupt handler. Defer
5417 this log dumping to the expired_lock_main() thread.
5419 Severity : enhancement
5421 Description: Support NFS exporting on 2.6 kernels.
5422 Details : Implement non-rawops metadata methods for NFS server to use without
5423 changing NFS server code.
5426 Frequency : very rare (synthetic metadata workload only)
5428 Description: two racing renames might cause an MDS thread to deadlock
5429 Details : Running the "racer" program may cause one MDS thread to rename
5430 a file from being the source of a rename to being the target of
5431 a rename at exactly the same time that another thread is doing
5432 so, and the second thread has already enqueued these locks after
5433 doing a lookup of the target and is trying to relock them in
5434 order. Ensure that we don't try to re-lock the same resource.
5437 Frequency : only very large systems with liblustre clients
5439 Description: slow eviction of liblustre clients with the "evict_by_nid" RPC
5440 Details : Use asynchronous set_info RPCs to send the "evict_by_nid" to
5441 all OSTs in parallel. This allows the eviction of stale liblustre
5442 clients to proceed much faster than if they were done in series,
5443 and also offers similar improvements for other set_info RPCs.
5448 Description: excessive CPU usage during initial read phase on client
5449 Details : During the initial read phase on a client, it would agressively
5450 retry readahead on the file, consuming too much CPU and impacting
5451 performance (since 1.4.5.8). Improve the readahead algorithm
5452 to avoid this, and also improve some other common cases (read
5453 of small files in particular, where "small" is files smaller than
5454 /proc/fs/lustre/llite/*/max_read_ahead_whole_mb, 2MB by default).
5459 Description: MDS crash when receiving packet with unknown intent.
5460 Details : Do not LBUG in unknown intent case, just return -EFAULT
5462 Severity : enhancement
5463 Bugzilla : 9293, 9385
5464 Description: MDS RPCs are serialised on client. This is unnecessary for some.
5465 Details : Do not serialize getattr (non-intent version) and statfs.
5468 Frequency : occasional, when OST network is overloaded/intermittent
5470 Description: client evicted by OST after bulk IO timeout
5471 Details : If a client sends a bulk IO request (read or write) the OST
5472 may evict the client if it is unresposive to its data GET/PUT
5473 request. This is incorrect if the network is overloaded (takes
5474 too long to transfer the RPC data) or dropped the OST GET/PUT
5475 request. There is no need to evict the client at all, since
5476 the pinger and/or lock callbacks will handle this, and the
5477 client can restart the bulk request.
5480 Frequency : Always when mmapping file with no objects
5482 Description: client crashes when mmapping file with no objects
5483 Details : Check that we actually have objects in a file before doing any
5484 operations on objects in ll_vm_open, ll_vm_close and
5490 Description: Request leak when working with deleted CWD
5491 Details : Introduce advanced request refcount tracking for requests
5492 referenced from lustre intent.
5494 Severity : Enhancement
5496 Description: Cache open file handles on client.
5497 Details : MDS now will return special lock along with openhandle, if
5498 requested and client is allowed to hold openhandle, even if unused,
5499 until such a lock is revoked. Helps NFS a lot, since NFS is opening
5500 closing files for every read/write openration.
5502 Severity : Enhancement
5504 Description: Cache open negative dentries on client when possible.
5505 Details : Guard negative dentries with UPDATE lock on parent dir, drop
5506 negative dentries on lock revocation.
5511 Description: Remounting a client read-only wasn't possible with a zconf mount
5512 Details : It wasn't possible to remount a client read-only with llmount.
5514 Severity : enhancement
5515 Description: Include MPICH 1.2.6 Lustre ADIO interface patch
5516 Details : In lustre/contrib/ or /usr/share/lustre in RPM a patch for
5517 MPICH is included to add Lustre-specific ADIO interfaces.
5518 This is based closely on the UFS ADIO layer and only differs
5519 in file creation, in order to allow the OST striping to be set.
5520 This is user-contributed code and not supported by CFS.
5525 Description: extended inode attributes (immutable, append-only) work improperly
5526 when 2.4 and 2.6 kernels are used on client/server or vice versa
5527 Details : Introduce kernel-independent values for these flags.
5529 Severity : enhancement
5532 Description: Allow fractional MB tunings for lustre in /proc/ filesystem.
5533 Details : Many of the /proc/ tunables can only be tuned at a megabyte
5534 granularity. Now, Fractional MB granularity is be supported,
5535 this is very useful for low memory system.
5537 Severity : enhancement
5539 Description: Getattr by fid
5540 Details : Getting a file attributes by its fid, obtaining UPDATE|LOOKUP
5541 locks, avoids extra getattr rpc requests to MDS, allows '/' to
5542 have locks and avoids getattr rpc requests for it on every stat.
5545 Frequency : Always, for filesystems larger than 2TB
5547 Description: ldiskfs crash at mount for filesystem larger than 2TB with mballoc
5548 Details : Kenrel kmalloc limits allocations to 128kB and this prevents
5549 filesystems larger than 2TB to be mounted with mballoc enabled.
5552 Frequency : Always, for 32-bit kernel without CONFIG_LBD and filesystem > 2TB
5554 Description: filesystem corruption for non-standard kernels and very large OSTs
5555 Details : If a 32-bit kernel is compiled without CONFIG_LBD enabled and a
5556 filesystems larger than 2TB is mounted then the kernel will
5557 silently corrupt the start of the filesystem. CONFIG_LBD is
5558 enabled for all CFS-supported kernels, but the possibility of
5559 this happening with a modified kernel config exists.
5561 Severity : enhancement
5563 Description: add client O_DIRECT support for 2.6 kernels
5564 Details : It is now possible to do O_DIRECT reads and writes to files
5565 in the Lustre client mountpoint on 2.6 kernel clients.
5567 Severity : enhancement
5569 Description: parallel glimpse, setattr, statfs, punch, destroy requests
5570 Details : Sends glimpse, setattr, statfs, punch, destroy requests to OSTs in
5571 parallel, not waiting for response from every OST before sending
5572 a rpc to the next OST.
5577 Description: setattr vs write race when updating file timestamps
5578 Details : Client processes that update a file timestamp into the past
5579 right after writing to the file (e.g. tar) it is possible that
5580 the updated file modification time can be reset to the current
5581 time due to a race between processing the setattr and write RPC.
5583 Severity : enhancement
5585 Description: Bring 'lfs find' closer in line with regular Linux find.
5586 Details : lfs find util supports -atime, -mtime, -ctime, -maxdepth, -print,
5587 -print0 options and obtains all the needed info through the lustre
5590 Severity : enhancement
5592 Description: support up to 1024 configured devices on one node
5593 Details : change obd_dev array from statically allocated to dynamically
5594 allocated structs as they are first used to reduce memory usage
5599 Description: Flush dirty partially truncated pages during truncate
5600 Details : Immediatelly flush partially truncated pages in filter_setattr,
5601 this way we completely avoid having any pages in page cache on OST
5602 and can retire ugly workarounds during writes to flush such pages.
5607 Description: i_sem vs transaction deadlock in mds_obd_destroy during unlink.
5608 Details : protect inode from truncation within vfs_unlink() context
5609 just take a reference before calling vfs_unlink() and release it
5610 when parent's i_sem is free.
5613 Frequency : always, if extents are used on OSTs
5615 Description: index ei_leaf_hi (48-bit extension) is not zeroed in extent index
5616 Details : OSTs using the extents format would not zero the high 16 bits of
5617 the index physical block number. This is not a problem for any
5618 OST filesystems smaller than 16TB, and no kernels support ext3
5619 filesystems larger than 16TB yet. This is fixed in 1.4.7 (all
5620 new/modified files) and can be fixed for existing filesystems
5621 with e2fsprogs-1.39-cfs1.
5626 Description: import connection selection may be incorrect if timer wraps
5627 Details : Using a 32-bit jiffies timer with HZ=1000 may cause backup
5628 import connections to be ignored if the 32-bit jiffies counter
5629 wraps. Use a 64-bit jiffies counter.
5632 Frequency : during server recovery
5634 Description: crash after server is denying duplicate export
5635 Details : If clients are resending connect requests to the server, the
5636 server refuses to allow a client to connect multiple times.
5637 Fixed a bug in the handling of this case.
5640 Frequency : very large clusters immediately after boot
5642 Description: LNET request buffers exhausted under heavy short-term load
5643 Details : If a large number of client requests are generated on a service
5644 that has previously never seen so many requests it is possible
5645 that the request buffer growth cannot keep up with the spike in
5646 demand. Instead of dropping incoming requests, they are held in
5647 the LND until the RPC service can accept more requests.
5650 Frequency : Sometimes during replay
5652 Description: Assertion failure in ll_local_open after replay.
5653 Details : If replay happened on an open request reply before we were able
5654 to set replay handler, reply will become not swabbed tripping the
5655 assertion in ll_local_open. Now we set the handler right after
5656 recognising of open request
5659 Frequency : very rare
5661 Description: kernel reports "badness in vsnprintf"
5662 Details : Reading from the "recovery_status" /proc file in small chunks
5663 may cause a negative length in lprocfs_obd_rd_recovery_status()
5664 call to vsnprintf() (which is otherwise harmless). Exit early
5665 if there is no more space in the output buffer.
5667 Severity : enhancement
5669 Description: clear OBD RPC statistics by writing to them
5670 Details : It is now possible to clear the OBD RPC statistics by writing
5671 to the "stats" file.
5676 Description: Client mtime is not the same on different clients after utimes
5677 Details : In some cases, the client was using the utimes() syscall on
5678 a file cached on another node. The clients now validate the
5679 ctime from the MDS + OSTs to determine which one is right.
5684 Description: Inability to activate failout mode
5685 Details : lconf script incorrectly assumed that in python string's numeric
5686 value is used in comparisons.
5689 Frequency : always with multiple stripes per file
5691 Description: Inefficient object allocation for mutli-stripe files
5692 Details : When selecting which OSTs to stripe files over, for files with
5693 a stripe count that divides evenly into the number of OSTs,
5694 the MDS is always picking the same starting OST for each file.
5695 Return the OST selection heuristic to the original design.
5700 Description: mount failures may take full timeout to return an error
5701 Details : Under some heavy load conditions it is possible that a
5702 failed mount can wait for the full obd_timeout interval,
5703 possibly several minutes, before reporting an error.
5704 Instead return an error as soon as the status is known.
5706 ------------------------------------------------------------------------------
5708 2006-02-14 Cluster File Systems, Inc. <info@clusterfs.com>
5710 * WIRE PROTOCOL CHANGE. This version of Lustre networking WILL NOT
5711 INTEROPERATE with older versions automatically. Please read the
5712 user documentation before upgrading any part of a live system.
5713 * WARNING: Lustre networking configuration changes are required with
5714 this release. See https://bugzilla.clusterfs.com/show_bug.cgi?id=10052
5717 * Support for kernels:
5718 2.6.9-22.0.2.EL (RHEL 4)
5719 2.6.5-7.244 (SLES 9)
5720 2.6.12.6 vanilla (kernel.org)
5723 Severity : enhancement
5724 Bugzilla : 7981/8208
5725 Description: Introduced Lustre Networking (LNET)
5726 Details : LNET is new networking infrastructure for Lustre, it includes
5727 a reorganized network configuration mode (see the user
5728 documentation for full details) as well as support for routing
5729 between different network fabrics. Lustre Networking Devices
5730 (LNDs) for the supported network fabrics have also been
5731 created for this new infrastructure.
5733 Severity : enhancement
5734 Description: Introduced Access control lists
5735 Details : clients can set ACLs on files and directories in order to have
5736 more fine-grained permissions than the standard Unix UGO+RWX.
5737 The MDS must be started with the "-o acl" mount option.
5739 Severity : enhancement
5740 Description: Introduced filesystem quotas
5741 Details : Administrators may now establish per-user quotas on the
5744 Severity : enhancement
5746 Description: Configuration change for the XT3
5747 The PTLLND is now used to run Lustre over Portals on the XT3
5748 The configure option(s) --with-cray-portals are no longer used.
5749 Rather --with-portals=<path-to-portals-includes> is used to
5750 enable building on the XT3. In addition to enable XT3 specific
5751 features the option --enable-cray-xt3 must be used.
5756 Description: Running on many-way SMP OSTs can trigger oops in llcd_send()
5757 Details : A race between allocating a new llcd and re-getting the llcd_lock
5758 allowed another thread to grab newly-allocated llcd.
5760 Severity : enhancement
5762 Description: 2.6 OST async journal commit and locking fix to improve performance
5763 Details : The filter_direct_io()+filter_commitrw_write() journal commits for
5764 2.6 kernels are now async as they already were in 2.4 kernels so
5765 that they can commit concurrently with the network bulk transfer.
5766 For block-allocated files the filter allocation semaphore is held
5767 to avoid filesystem fragmentation during allocation. BKL lock
5768 removed for 2.6 xattr operations where it is no longer needed.
5773 Description: lconf incorrectly determined whether two IP networks could talk
5774 Details : In some more complicated routing and multiple-network
5775 configurations, lconf will avoid trying to make a network
5776 connection to a disjoint part of the IP space. It was doing the
5777 math incorrectly for one set of cases.
5782 Description: Fix for potential infinite loop processing records in an llog.
5783 Details : If an llog record is corrupted/zeroed, it is possible to loop
5784 forever in llog_process(). Validate the llog record length
5785 and skip the remainder of the block on error.
5788 Frequency : occasional (liblustre only)
5790 Description: liblustre could not open files whose last component is a symlink
5791 Details : sysio_path_walk() would incorrectly pass the open intent to
5792 intermediate path components.
5795 Frequency : rare (liblustre only with non-standard tuning)
5796 Bugzilla : 7201 (7350)
5797 Description: Tuning the MDC DLM LRU size to zero triggers client LASSERT
5798 Details : llu_lookup_finish_locks() tries to set lock data on a lock
5799 after it has been released, only do this for referenced locks
5801 Severity : enhancement
5803 Description: specifying an (invalid) directory default stripe_size of -1
5804 would reset the directory default striping
5805 Details : stripe_size -1 was used internally to signal directory stripe
5806 removal, now use "all default" to signal dir stripe removal
5807 as a directory striping of "all default" is not useful
5810 Frequency : common for large clusters running liblustre clients
5812 Description: doing an ls when liblustre clients are running is slow
5813 Details : sending a glimpse AST to a liblustre client waits for every AST
5814 to time out, as liblustre clients will not respond. Since they
5815 cannot cache data we refresh the OST lock LVB from disk instead.
5817 Severity : enhancement
5819 Description: doing an ls at the same time as file IO can be slow
5820 Details : enqueue and other "small" requests can be blocked behind many
5821 large IO requests. Create a new OST IO portal for non-IO
5822 requests so they can be processed faster.
5825 Frequency : rare (only HPUX clients mounting unsupported re-exported NFS vol)
5827 Description: an HPUX NFS client would get -EACCESS when ftruncate()ing a newly
5828 created file with mode 000
5829 Details : the Linux NFS server relies on an MDS_OPEN_OWNEROVERRIDE hack to
5830 allow an ftruncate() as a non-root user to a file with mode 000.
5831 Lustre now respects this flag to disable mode checks when
5832 truncating a file owned by the user
5835 Frequency : liblustre-only, when liblustre client dies unexpectedly or becomes
5838 Description: Revoking locks from clients that went dead or catatonic might take
5840 Details : New lock flags FL_CANCEL_ON_BLOCK used by liblustre makes
5841 cancellation of such locks instant on servers without waiting for
5842 any reply from clients. Clients drops these locks when cancel
5843 notification from server is received without replying.
5846 Frequency : liblustre-only, when liblustre client dies or becomes busy
5848 Description: Doing ls on Linux clients can take a long time with active
5850 Details : Liblustre client cannot handle ASTs in timely manner, so avoid
5851 granting such locks to it in the first place if possible. Locks
5852 are taken by proxy on the OST during the read or write and
5853 dropped immediately afterward. Add connect flags handling, do
5854 not grant locks to liblustre clients for glimpse ASTs.
5856 Severity : enhancement
5858 Description: Improve read-ahead algorithm to avoid excessive IO for random reads
5859 Details : Existing read-ahead algorithm is tuned for the case of streamlined
5860 sequential reads and behaves badly with applications doing random
5861 reads. Improve it by reading ahead at least read region, and
5862 avoiding excessive large RPC for small reads.
5864 Severity : enhancement
5866 Description: Creating more than 1000 files for a single job may cause a load
5867 imbalance on the OSTs if there are also a large number of OSTs.
5868 Details : qos_prep_create() uses an OST index reseed value that is an
5869 even multiple of the number of available OSTs so that if the
5870 reseed happens in the middle of the object allocation it will
5871 still utilize the OSTs as uniformly as possible.
5876 Description: OST or MDS may oops in ping_evictor_main()
5877 Details : ping_evictor_main() drops obd_dev_lock if deleting a stale export
5878 but doesn't restart at beginning of obd_exports_timed list
5881 Severity : enhancement
5883 Description: improve by-nid export eviction on the MDS and OST
5884 Details : allow multiple exports with the same NID to be evicted at one
5885 time without re-searching the exports list.
5888 Frequency : rare, only with supplementary groups enabled on SMP 2.6 kernels
5890 Description: MDS may oops in groups_free()
5891 Details : in rare race conditions a newly allocated group_info struct is
5892 freed again, and this can be NULL. The 2.4 compatibility code
5893 for groups_free() checked for a NULL pointer, but 2.6 did not.
5896 Frequency : common for liblustre clients doing little filesystem IO
5897 Bugzilla : 9352, 7313
5898 Description: server may evict liblustre clients accessing contended locks
5899 Details : if a client is granted a lock or receives a completion AST
5900 with a blocking AST already set it would not reply to the AST
5901 for LDLM_FL_CANCEL_ON_BLOCK locks. It now replies to such ASTs.
5904 Frequency : lfs setstripe, only systems with more than 160 OSTs
5906 Description: unable to set striping with a starting offset beyond OST 160
5907 Details : llapi_create_file() incorrectly limited the starting stripe
5908 index to the maximum single-file stripe count.
5911 Frequency : LDAP users only
5913 Description: lconf did not handle in-kernel recovery with LDAP properly
5914 Details : lconf/LustreDB get_refs() is searching the wrong namespace
5916 Severity : enhancement
5918 Description: bind OST threads to NUMA nodes to improve performance
5919 Details : all OST threads are uniformly bound to CPUs on a single NUMA
5920 node and do their allocations there to localize memory access
5922 Severity : enhancement
5924 Description: llmount can determine client NID directly from Myrinet (GM)
5925 Details : the client NID code from gmnalnid was moved directly into
5926 llmount, removing the need to use this or specifying the
5927 client NID explicitly when mounting GM clients with zeroconf
5930 Frequency : if client is started with down MDS
5932 Description: if client is started with down MDS mount hangs in ptlrpc_queue_wait
5933 Details : Having an LWI_INTR() wait event (interruptible, but no timeout)
5934 will wait indefinitely in ptlrpc_queue_wait->l_wait_event() after
5935 ptlrpc_import_delayed_req() because we didn't check if the
5936 request was interrupted, and we also didn't break out of the
5937 event loop if there was no timeout
5942 Description: data loss during non-page-aligned writes to a single file from
5943 both multiple nodes and multiple threads on one node at same time
5944 Details : updates to KMS and lsm weren't protected by common lock. Resulting
5945 inconsistency led to false short-reads, that were cached and later
5946 used by ->prepare_write() to fill in partially written page,
5947 leading to data loss.
5950 Frequency : always, if lconf --abort_recovery used
5952 Description: lconf --abort_recovery fails with 'Operation not supported'
5953 Details : lconf was attempting to abort recovery on the MDT device and not
5956 Severity : enhancement
5958 Description: remove cleanup logs
5959 Details : replace lconf-generated cleanup logs with lustre internal
5960 cleanup routines. Eliminates the need for client-cleanup and
5963 Severity : enhancement
5965 Description: add support for EAs (user and system) on lustre filesystems
5966 Details : it is now possible to store extended attributes in the Lustre
5967 client filesystem, and with the user_xattr mount option it
5968 is possible to allow users to store EAs on their files also
5970 Severity : enhancement
5972 Description: Add possibility (config option) to show minimal available OST free
5974 Details : When compiled with --enable-mindf configure option, statfs(2)
5975 (and so, df) will return least minimal free space available from
5976 all OSTs as amount of free space on FS, instead of summary of
5977 free spaces of all OSTs.
5979 Severity : enhancement
5981 Description: do not expand extent locks acquired on OST-side
5982 Details : Modify ldlm_extent_policy() to not expand local locks, acquired
5983 by server: they are not cached anyway.
5986 Frequency : when mmap is used/binaries executed from Lustre
5988 Description: Unmmap pages before throwing them away from read cache.
5989 Details : llap_shrink cache now attempts to unmap pages before discarding
5990 them (if unmapping failed - do not discard). SLES9 kernel has
5991 extra checks that trigger if this unmapping is not done first.
5996 Description: lconf didn't resolve symlinks before checking to see whether a
5997 given mountpoint was already in use
6000 Frequency : when migrating failover services
6001 Bugzilla : 6395, 9514
6002 Description: When migrating a subset of services from a node (e.g. failback
6003 from a failover service node) the remaining services would
6004 time out and evict clients.
6005 Details : lconf --force (implied by --failover) sets the global obd_timeout
6006 to 5 seconds in order to quickly disconnect, but this caused
6007 other RPCs to time out too quickly. Do not change the global
6008 obd_timeout for force cleanup, only set it for DISCONNECT RPCs.
6010 Severity : enhancement
6011 Frequency : if MDS is started with down OST
6012 Bugzilla : 9439,5706
6013 Description: Allow startup/shutdown of an MDS without depending on the
6014 availability of the OSTs.
6015 Details : Asynchronously call mds_lov_synchronize during MDS startup.
6016 Add appropriate locking and lov-osc refcounts for safe
6017 cleaning. Add osc abort_inflight calls in case the
6018 synchronize never started.
6021 Frequency : occasional (Cray XT3 only)
6023 Description: root not authorized to access files in CRAY_PORTALS environment
6024 Details : The client process capabilities were not honoured on the MDS in
6025 a CRAY_PORTALS/CRAY_XT3 environment. If the file had previously
6026 been accessed by an authorized user then root was able to access
6027 the file on the local client also. The root user capabilities
6028 are now allowed on the MDS, as this environment has secure UID.
6031 Frequency : occasional
6033 Description: ldiskfs "too long searching" message happens too often
6034 Details : A debugging message (otherwise harmless) prints too often on
6035 the OST console. This has been reduced to only happen when
6036 there are fragmentation problems on the filesystem.
6041 Description: Division by zero in statfs when all OSCs are inactive
6042 Details : lov_get_stripecnt() returns zero due to incorrect order of checks,
6043 lov_statfs divides by value returned by lov_get_stripecnt().
6047 Bugzilla : 9489, 3273
6048 Description: First write from each client to each OST was only 4kB in size,
6049 to initialize client writeback cache, which caused sub-optimal
6050 RPCs and poor layout on disk for the first writen file.
6051 Details : Clients now request an initial cache grant at (re)connect time
6052 and so that they can start streaming writes to the cache right
6053 away and always do full-sized RPCs if there is enough data.
6054 If the OST is rebooted the client also re-establishes its grant
6055 so that client cached writes will be honoured under the grant.
6060 Description: Slow ls (and stat(2) syscall) on files residing on IO-loaded OSTs
6061 Details : Now I/O RPCs go to different portal number and (presumably) fast
6062 lock requests (and glimses) and other RPCs get their own service
6063 threads pool that should be able to service those RPCs
6066 Severity : enhancement
6068 Description: Ability to exchange lustre version between client and servers and
6069 issue warnings at client side if client is too old. Also for
6070 liblustre clients there is ability to refuse connection of too old
6072 Details : New 'version' field is added to connect data structure that is
6073 filled with version info. That info is later checked by server and
6077 Frequency : rare, liblustre only.
6078 Bugzilla : 9296, 9581
6079 Description: Two simultaneous writes from liblustre at offset within same page
6080 might proceed at the same time overwriting eachother with stale
6082 Details : I/O lock withing llu_file_prwv was released too early, before data
6083 actually was hitting the wire. Extended lock-holding time until
6084 server acknowledges receiving data.
6087 Frequency : extremely rare. Never observed in practice.
6089 Description: avoid generating lustre_handle cookie of 0.
6090 Details : class_handle_hash() generates handle cookies by incrementing
6091 global counter, and can hit 0 occasionaly (this is unlikely, but
6092 not impossible, because initial value of cookie counter is
6093 selected randonly). Value of 0 is used as a sentinel meaning
6094 "unassigned handle" --- avoid it. Also coalesce two critical
6095 sections in this function into one.
6097 Severity : enhancement
6099 Description: allow liblustre clients to delegate truncate locking to OST
6100 Details : To avoid overhead of locking, liblustre client instructs OST to
6101 take extent lock in ost_punch() on client's behalf. New connection
6102 flag is added to handle backward compatibility.
6104 Severity : enhancement
6105 Bugzilla : 4928, 7341, 9758
6106 Description: allow number of OST service threads to be specified
6107 Details : a module parameter allows the number of OST service threads
6108 to be specified via "options ost ost_num_threads={N}" in the
6109 OSS's /etc/modules.conf or /etc/modprobe.conf.
6113 Bugzilla : 6146, 9635, 9895
6114 Description: servers crash with bad pointer in target_handle_connect()
6115 Details : In rare cases when a client is reconnecting it was possible that
6116 the connection request was the last reference for that export.
6117 We would temporarily drop the export reference and get a new
6118 one, but this may have been the last reference and the export
6119 was just destroyed. Get new reference before dropping old one.
6121 Severity : enhancement
6122 Frequency : if client is started with failover MDS
6124 Description: Allow multiple MDS hostnames in the mount command
6125 Details : Try to read the configuration from all specified MDS
6126 hostnames during a client mount in case the "primary"
6129 Severity : enhancement
6131 Description: Stop sending data to evicted clients as soon as possible.
6132 Details : Check if the client we are about to send or are sending data to
6133 was evicted already. (Check is done every second of waiting,
6134 for which l_wait_event interface was extended to allow checking
6135 of exit condition at specified intervals).
6138 Frequency : rare, normally only when NFS exporting is done from client
6140 Description: 'bad disk LOV MAGIC: 0x00000000' error when chown'ing files
6142 Details : Make mds_get_md() recognise empty md case and set lmm size to 0.
6145 Frequency : always, if srand() is called before liblustre initialization
6147 Description: Liblustre uses system PRNG disturbing its usage by user application
6148 Details : Introduce internal to lustre fast and high-quality PRNG for
6149 lustre usage and make liblustre and some other places in generic
6150 lustre code to use it.
6152 Severity : enhancement
6153 Bugzilla : 9477, 9557, 9870
6154 Description: Verify that the MDS configuration logs are updated when xml is
6155 Details : Check if the .xml configuration logs are newer than the config
6156 logs stored on the MDS and report an error if this is the case.
6157 Request --write-conf, or allow starting with --old_conf.
6159 Severity : enhancement
6161 Description: Handle symlinks in the path when checking if Lustre is mounted.
6162 Details : Resolve intermediate symlinks when checking if a client has
6163 mounted a filesystem to avoid duplicate client mounts.
6168 Description: lconf can hit an error exception but still return success.
6169 Details : The lconf command catches the Command error exception at the top
6170 level script context and will exit with the associated exit
6171 status, but doesn't ensure that this exit status is non-zero.
6176 Description: failure of ptlrpc thread startup can cause oops
6177 Details : Starting a ptlrpc service thread can fail if there are a large
6178 number of threads or the server memory is very fragmented.
6179 Handle this without oopsing.
6182 Frequency : always, only if liblustre and non-default acceptor port was used
6184 Description: liblustre cannot connect to servers with non-default acceptor port
6185 Details : tcpnal_set_default_params() was not called and was therefore
6186 ignoring the environment varaible TCPNAL_PORT, as well as other
6187 TCPNAL_ environment variables
6192 Description: two objects could be created on the same OST for a single file
6193 Details : If an OST is down, in some cases it was possible to create two
6194 objects on a single OST for a single file. No problems other
6195 than potential performance impact and spurious error messages.
6199 Bugzilla : 5681, 9562
6200 Description: Client may oops in ll_unhash_aliases
6201 Details : Client dcache may become inconsistent in race condition.
6202 In some cases "getcwd" can fail if the current directory is
6208 Description: Inode refcounting problems in NFS export code
6209 Details : link_raw functions used to call d_instantiate without obtaining
6210 extra inode reference first.
6214 Bugzilla : 9942, 9903
6215 Description: Referencing freed requests leading to crash, memleaks with NFS.
6216 Details : We used to require that call to ll_revalidate_it was always
6217 followed by ll_lookup_it. Also with revalidate_special() it is
6218 possible to call ll_revalidate_it() twice for the same dentry
6219 even if first occurence returned success. This fix changes semantic
6220 between DISP_ENQ_COMPLETE disposition flag to mean there is extra
6221 reference on a request referred from the intent.
6222 ll_intent_release() then releases such a request.
6225 Frequency : rare, normally benchmark loads only
6227 Description: unlinked inodes were kept in memory on the client
6228 Details : If a client is repeatedly creating and unlinking files it
6229 can accumulate a lot of stale inodes in the inode slab cache.
6230 If there is no other client load running this can cause the
6231 client node to run out of memory. Instead flush old inodes
6232 from client cache that have the same inode number as a new inode.
6235 Frequency : SLES9 2.6.5 kernel and long filenames only
6236 Bugzilla : 9969, 10379
6237 Description: utime reports stale NFS file handle
6238 Details : SLES9 uses out-of-dentry names in some cases, which confused
6239 the lustre dentry revalidation. Change it to always use the
6243 Frequency : rare, unless heavy write-truncate concurrency is continuous
6244 Bugzilla : 4180, 6984, 7171, 9963, 9331
6245 Description: OST becomes very slow and/or deadlocked during object unlink
6246 Details : filter_destroy() was holding onto the parent directory lock
6247 while truncating+unlinking objects. For very large objects this
6248 may block other threads for a long time and slow overall OST
6249 responsiveness. It may also be possible to get a lock ordering
6250 deadlock in this case, or run out of journal credits because of
6251 the combined truncate+unlink. Solution is to do object truncate
6252 first in one transaction without parent lock, and then do the
6253 final unlink in a new transaction with the parent lock. This
6254 reduces the lock hold time dramatically.
6257 Frequency : rare, 2.4 kernels only
6259 Description: MDS or OST cleanup may trip kernel BUG when dropping kernel lock
6260 Details : mds_cleanup() and filter_cleanup() need to drop the kernel lock
6261 before unmounting their filesystem in order to avoid deadlock.
6262 The kernel_locked() function in 2.4 kernels only checks whether
6263 the kernel lock is held, not whether it is this process that is
6264 holding it as 2.6 kernels do.
6269 Description: MDS or OST may oops/LBUG if a client is connecting multiple times
6270 Details : The client ptlrpc code may be trying to reconnect to a down
6271 server before a previous connection attempt has timed out.
6272 Increase the reconnect interval to be longer than the connection
6273 timeout interval to avoid sending duplicate connections to
6277 Frequency : echo_client brw_test command
6279 Description: fix echo_client to work with OST preallocated code
6280 Details : OST preallocation code (5137) didn't take echo_client IO path
6281 into account: echo_client calls filter methods outside of any
6282 OST thread and, hence, there is no per-thread preallocated
6283 pages and buffers to use. Solution: hijack pga pages for IO. As
6284 a byproduct, this avoids unnecessary data copying.
6288 Bugzilla : 3555, 5962, 6025, 6155, 6296, 9574
6289 Description: Client can oops in mdc_commit_close() after open replay
6290 Details : It was possible for the MDS to return an open request with no
6291 transaction number in mds_finish_transno() if the client was
6292 evicted, but without actually returning an error. Clients
6293 would later try to replay that open and may trip an assertion
6294 Simplify the client close codepath, and always return an error
6295 from the MDS in case the open is not successful.
6298 Frequency : rare, 2.6 OSTs only
6300 Description: OST may deadlock under high load on fragmented files
6301 Details : If there was a heavy load and highly-fragmented OST filesystems
6302 it was possible to have all the OST threads deadlock waiting on
6303 allocation of biovecs, because the biovecs were not released
6304 until the entire RPC IO was completed. Instead, release biovecs
6305 as soon as they are complete to ensure forward IO progress.
6307 Severity : enhancement
6309 Description: Support for specifying external journal device at mount
6310 Details : If an OST or MDS device is formatted with an external journal
6311 device, this device major/minor is stored in the ext3 superblock
6312 and may not be valid for failover. Allow detecting and
6313 specifying the external journal at mount time.
6318 Description: Mounting an MDS with pending unlinked files may cause oops
6319 Details : target_finish_recovery() calls mds_postrecov() which returned
6320 the number of orphans unlinked. mds_lov_connect->mds_postsetup()
6321 considers this an error and immediately begins cleaning up the
6322 lov, just after starting the mds_lov process
6324 Severity : enhancement
6326 Description: Implement 'lfs df' to report actual free space on per-OST basis
6327 Details : Add sub-command 'df' on 'lfs' to report the disk space usage of
6328 MDS/OSDs. Usage: lfs df [-i][-h]. Command Options: '-i' to report
6329 usage of objects; '-h' to report in human readable format.
6331 ------------------------------------------------------------------------------
6333 2005-08-26 Cluster File Systems, Inc. <info@clusterfs.com>
6340 Description: Mounting an ldiskfs file system with mballoc may crash OST node.
6341 Details : ldiskfs mballoc code may reference an uninitialized buddy struct
6342 at startup during orphan unlinking. Instead, skip buddy update
6343 before setup, as it will be regenerated after recovery is complete.
6348 Description: If an OST is inactive, its locks might reference stale inodes.
6349 Details : lov_change_cbdata() must iterate over all namespaces, even if
6350 they are inactive to clear inode references from the lock.
6352 Severity : enhancement
6353 Frequency : occasional, if non-standard max_dirty_mb used
6355 Description: Client will block write RPCs if not enough grant
6356 Details : If a client has max_dirty_mb smaller than max_rpcs_in_flight,
6357 then the client will block writes while waiting for another RPC
6358 to complete instead of consuming its dirty limit. With change
6359 we get improved performance when max_dirty_mb is small.
6361 Severity : enhancement
6362 Bugzilla : 3389, 6253
6363 Description: Add support for supplementary groups on the MDS.
6364 Details : The MDS has an upcall /proc/fs/lustre/mds/{mds}/group_upcall
6365 (set to /usr/sbin/l_getgroups if enabled) which will do MDS-side
6366 lookups for user supplementary groups into a cache.
6370 Description: O_CREAT|O_EXCL open flags in liblustre always return -EEXIST
6371 Details : Make libsysio to not enforce O_EXCL by clearing the flag,
6372 for liblustre O_EXCL is enforced by MDS.
6376 Description: readdir never returns NULL in liblustre.
6377 Details : Corrected llu_iop_getdirentries logic, to return offset of next
6378 dentry in struct dirent.
6382 Frequency : liblustre only, depends on application IO pattern
6383 Description: liblustre clients evicted if not contacting servers
6384 Details : Don't put liblustre clients into the ping_evictor list, so
6385 they will not be evicted by the pinger ever.
6387 Severity : enhancement
6389 Description: Add ability to evict clients by NID from MDS.
6390 Details : By echoing "nid:$NID" string into
6391 /proc/fs/lustre/mds/.../evict_client client with nid that equals to
6392 $NID would be instantly evicted from this MDS and from all active
6393 OSTs connected to it.
6397 Description: Do not query file size twice, somewhat slowing stat(2) calls.
6398 Details : lookup_it_finish() used to query file size from OSTs that was not
6403 Description: service threads change working directory to that of init
6404 Details : Starting lustre service threads may pin the working directory
6405 of the parent thread, making that filesystem busy. Threads
6406 now change to the working directory of init to avoid this.
6410 Frequency : during shutdown only
6411 Description: shutdown with a failed MDS or OST can cause unmount to hang
6412 Details : Don't resend DISCONNECT messages in ptlrpc_disconnect_import()
6417 Frequency : 2.6 only
6418 Description: chmod/chown may include an extra supplementary group
6419 Details : ll{,u}_mdc_pack_op_data() does not properly initialize the
6420 supplementary group and if none is specified this is used.
6423 Bugzilla : 5479 (6816)
6425 Description: Racing open + rm can assert client in mdc_set_open_replay_data()
6426 Details : If lookup is in progress on a file that is unlinked we might try
6427 to revalidate the inode and fail in revalidate after lookup is
6428 complete and ll_file_open() enqueues the open again but
6429 it_open_error() was not checking DISP_OPEN_OPEN errors correctly.
6432 Frequency : always, if lconf --abort_recovery used
6434 Description: lconf --abort_recovery fails with 'Operation not supported'
6435 Details : lconf was attempting to abort recovery on the MDT device and not
6438 ------------------------------------------------------------------------------
6440 2005-08-08 Cluster File Systems, Inc. <info@clusterfs.com>
6445 Frequency : rare (only unsupported configurations with a node running as an
6447 Bugzilla : 6514, 5137
6448 Description: Mounting a Lustre file system on a node running as an OST could
6450 Details : OSTs now preallocates memory needed to write out data at
6451 startup, instead of when needed, to avoid having to
6452 allocate memory in possibly low memory situations.
6453 Specifically, if the file system is mounted on on OST,
6454 memory pressure could force it to try to write out data,
6455 which it needed to allocate memory to do. Due to the low
6456 memory, it would be unable to do so and the node would
6457 become unresponsive.
6459 Severity : enhancement
6461 Description: Addition of lconf --service command line option
6462 Details : lconf now accepts a '--service <arg>' option, which is
6463 shorthand for 'lconf --group <arg> --select <arg>=<hostname>'
6465 Severity : enhancement
6467 Description: Failover mode is now the default for OSTs.
6468 Details : By default, OSTs will now run in failover mode. To return to
6469 the old behaviour, add '--failout' to the lmc line for OSTs.
6471 Severity : enhancement
6473 Description: Health checks are now provided for MDS and OSTs
6474 Details : Additional detailed health check information on MSD and OSTs
6475 is now provided through the procfs health_check value.
6478 Frequency : occasional, depends on IO load
6480 Description: Disk fragmentation on the OSTs could eventually cause slowdowns
6481 after numerous create/delete cycles
6482 Details : The ext3 inode allocation policy would not allocate new inodes
6483 very well on the OSTs because there are no new directories
6484 being created. Instead we look for groups with free space if
6485 the parent directories are nearly full.
6490 Description: Network or server problems during mount may cause partially
6491 mounted clients instead of returning an error.
6492 Details : The config llog parsing code may overwrite the error return
6493 code during mount error handling, returning success instead
6499 Description: MDS can fail to allocate large reply buffers
6500 Details : After long uptimes the MDS can fail to allocate large reply
6501 buffers (e.g. zconf client mount config records) due to memory
6502 fragmentation or consumption by the buffer cache. Preallocate
6503 some large reply buffers so that these replies can be sent even
6504 under memory pressure.
6508 Frequency : rare (liblustre)
6509 Description: fsx running with liblustre complained that using truncate() to
6510 extend the file doesn't work. This patch corrects that issue.
6511 Details : This is the liblustre equivalent of the fix for bug 6196. Fixes
6512 ATTR_SIZE and lsm use in llu_setattr_raw.
6516 Frequency : rare, only 2.6 kernels
6517 Description: Unusual file access patterns on the MDS may result in inode
6518 data being lost in very rare circumstances.
6519 Details : Bad interaction between the ea-in-inode patch and the "no-read"
6520 code in the 2.6 kernel caused the inode and/or EA data not to
6521 be read from disk, causing single-file corruption.
6525 Frequency : rare, only 2.6 filesystems using extents
6526 Description: Heavy concurrent write and delete load may cause data corruption.
6527 Details : It was possible under high-load situations to have an extent
6528 metadata block in the block device cache from a just-unlinked
6529 file overwrite a newly-allocated data block. We now unmap any
6530 metadata buffers that alias just-allocated data blocks.
6534 Frequency : filesystems with default stripe_count larger than 77
6535 Description: lconf+mke2fs fail when formatting filesystem with > 77 stripes
6536 Details : lconf specifies an inode size of 4096 bytes when the default
6537 stripe_count is larger than 77. This conflicts with the default
6538 inode density of 1 per 4096 bytes. Allocate smaller inodes in
6539 this case to avoid pinning too much memory for large EAs.
6541 ------------------------------------------------------------------------------
6543 2005-07-07 Cluster File Systems, Inc. <info@clusterfs.com>
6548 Frequency : rare (extremely heavy IO load with hundreds of clients)
6550 Description: Client is evicted, gets IO error writing to file
6551 Details : lock ordering changes for bug 5492 reintroduced bug 3267 and
6552 caused clients to be evicted for AST timeouts. The fixes in
6553 bug 5192 mean we no longer need to have such short AST timeouts
6554 so ldlm_timeout has been increased.
6557 Frequency : occasional during --force or --failover shutdown under load
6558 Bugzilla : 5949, 4834
6559 Description: Server oops/LBUG if stopped with --force or --failover under load
6560 Details : a collection of import/export refcount and cleanup ordering
6561 issues fixed for safer force cleanup
6564 Frequency : only filesystems larger than 120 OSTs
6565 Bugzilla : 5990, 6223
6566 Description: lfs getstripe would oops on a very large filesystem
6567 Details : lov_getconfig used kfree on vmalloc'd memory
6570 Frequency : only filesystems exporting via NFS to Solaris 10 clients
6571 Bugzilla : 6242, 6243
6572 Description: reading from files that had been truncated to a non-zero size
6573 but never opened returned no data
6574 Details : ll_file_read() reads zeros from no-object files to EOF
6579 Description: A bug in MDS/OSS recovery could cause the OSS to fail an assertion
6580 Details : There's little harm in aborting MDS/OSS recovery and letting it
6581 try again, so I removed the LASSERT and return an error instead.
6583 Severity : enhancement
6585 Description: New debugging infrastructure for tracking down data corruption
6586 Details : The I/O checksum code was replaced to: (a) control it at runtime,
6587 (b) cover more of the client-side code path, and (c) try to narrow
6588 down where problems occurred
6592 Bugzilla : 3819, 4364, 4397, 6313
6593 Description: Racing close and eviction MDS could cause assertion in mds_close
6594 Details : It was possible to get multiple mfd references during close and
6595 client eviction, leading to one thread referencing a freed mfd.
6597 Severity: : enhancement
6598 Bugzilla : 3262, 6359
6599 Description: Attempts to reconnect to servers are now more aggressive.
6600 Details : This builds on the enhanced upcall-less recovery that was added
6601 in 1.4.2. When trying to reconnect to servers, clients will
6602 now try each server in the failover group every 10 seconds. By
6603 default, clients would previously try one server every 25 seconds.
6608 Description: After recovery, certain operations trigger a failed
6609 assertion on a client.
6610 Details : Failing over an mds, using lconf -d --failover, while a
6611 client was doing a readdir() call would cause the client to
6612 LBUG after recovery completed and the readdir() was resent.
6614 Severity : enhancement
6616 Description: Default groups are now added by lconf
6617 Details : You can now run lconf --group <servicename> without having to
6618 manually add groups with lmc.
6621 Frequency : occasional
6623 Description: Nodes with an elan id of 0 trigger a failed assertion
6626 Frequency : always when accessing e.g. tty/console device nodes
6628 Description: tty and some other devices nodes cannot be used on lustre
6629 Details : file's private_data field is used by device data and lustre
6630 values in there got lost. New field was added to struct file to
6631 store fs-specific private data.
6634 Frequency : when exporting Lustre via NFS
6636 Description: NFSD failed occasionally when looking up a path component
6637 Details : NFSD is looking up ".." which was broken in ext3 directories
6638 that had grown large enough to become hashed.
6641 Frequency : Clusters with multiple interfaces not on the same subnet
6643 Description: Nodes will repeatedly try to reconnect to an interface which it
6644 cannot reach and report an error to the log.
6645 Details : Extra peer list entries will be created by lconf with some peers
6646 unreachable. lconf now validates the peer before adding it.
6649 Frequency : Only if a default stripe is set on the filesystem root.
6651 Description: Setting a default stripe on the filesystem root prevented the
6652 filesystem from being remounted.
6653 Details : The client was sending extra request flags in the root getattr
6654 request and did not allocate a reply buffer for the dir EA.
6657 Frequency : occasional, higher if lots of files are accessed by one client
6658 Bugzilla : 6159, 6097
6659 Description: Client trips assertion regarding lsm mismatch/magic
6660 Details : While revalidating inodes the VFS looks up inodes with ifind()
6661 and in rare cases can find an inode that is being freed.
6662 The ll_test_inode() code will free the lsm during ifind()
6663 when it finds an existing inode and then the VFS later attaches
6664 this free lsm to a new inode.
6668 Bugzilla : 6422, 7030
6669 Description: MDS deadlock between mkdir and client eviction
6670 Details : Creating a new file via mkdir or mknod (starting a transaction
6671 and getting the ns lock) can deadlock with client eviction
6672 (gets ns lock and trying to finish a synchronous transaction).
6675 Frequency : occasional
6676 Description: While starting a server, the fsfilt_ext3 module could not be
6678 Details : CFS's improved ext3 filesystem is named ldiskfs for 2.6
6679 kernels. Previously, lconf would still use the ext3 name
6680 when trying to load modules. Now, it will correctly use
6681 ext3 on 2.4 and ldiskfs on 2.6.
6683 Severity : enhancement
6684 Description: The default stripe count has been changed to 1
6685 Details : The interpretation of the default stripe count (0, to lfs
6686 or lmc) has been changed to mean striping across a single
6687 OST, rather than all available. For general usage we have
6688 found a stripe count of 1 or 2 works best.
6690 Severity : enhancement
6691 Description: Add support for compiling against Cray portals.
6692 Details : Conditional compiling for some areas that are different
6696 Frequency : occasional
6697 Bugzilla : 6409, 6834
6698 Description: Creating files with an explicit stripe count may lead to
6699 a failed assertion on the MDS
6700 Details : If some OSTs are full or unavailable, creating files may
6701 trigger a failed assertion on the MDS. Now, Lustre will
6702 try to use other servers or return an error to the
6706 Frequency : occasional
6708 Description: Multiple concurrent overlapping read+write on multiple SMP nodes
6709 caused lock timeout during readahead (since 1.4.2).
6710 Details : Processes doing readahead might match a lock that hasn't been
6711 granted yet if there are overlapping and conflicting lock
6712 requests. The readahead process waits on ungranted lock
6713 (original lock is CBPENDING), while OST waits for that process
6714 to cancel CBPENDING read lock and eventually evicts client.
6716 Severity : enhancement
6718 Description: Initial enabling of flock support for clients
6719 Details : Implements fcntl advisory locking and file status functions.
6720 This feature is provided as an optional mount flag (default
6721 off), and is NOT CURRENTLY SUPPORTED. Not all types of record
6722 locking are implemented yet, and those that are are not guaranteed
6723 to be completely correct in production environments.
6724 mount -t lustre -o [flock|noflock] ...
6727 Frequency : occasional
6729 Description: OSTs running 2.4 kernels but with extents enabled might trip an
6730 assertion in the ext3 JBD (journaling) layer.
6731 Details : The b_committed_data struct is protected by the big kernel lock
6732 in 2.4 kernels, serializing journal_commit_transaction() and
6733 ext3_get_block_handle->ext3_new_block->find_next_usable_block()
6734 access to this struct. In 2.6 kernels there is finer grained
6735 locking to improve SMP performance of the JBD layer.
6739 Description: Changes the "SCSI I/O Stats" kernel patch to default to "enabled"
6741 -----------------------------------------------------------------------------
6743 2005-05-05 Cluster File Systems, Inc. <info@clusterfs.com>
6745 NOTE: Lustre 1.4.2 uses an incompatible network protocol than previous
6746 versions of Lustre. Please update all servers and clients to
6747 version 1.4.2 or later at the same time. You must also run
6748 "lconf --write-conf {config}.xml" on the MDS while it is stopped
6749 to update the configuration logs.
6751 - fix for HPUX NFS client breakage when NFS exporting Lustre (5781)
6752 - mdc_enqueue does not need max_mds_easize request buffer on send (5707)
6753 - swab llog records of type '0' so we get proper header size/idx (5861)
6754 - send llog cancel req to DLM cancel portal instead of cb portal (5515)
6755 - fix rename of one directory over another leaking an inode (5953)
6756 - avoid SetPageDirty on 2.6 (5981)
6757 - don't re-add just-being-destroyed locks to the waiting list (5653)
6758 - when creating new directories, inherit the parent's custom
6759 striping settings if present parent (3048)
6760 - flush buffers from cache before direct IO in 2.6 obdfilter (4982)
6761 - don't hold i_size_sem in ll_nopage() and ll_ap_refresh_count (6077)
6762 - don't hold client locks on temporary worklist from l_lru (5666)
6763 - handle IO errors in 2.6 obdfilter bio completion routine (6046)
6764 - automatically evict dead clients (5921)
6765 - Update file size properly in create+truncate+fstat case (6196)
6766 - Do not unhash mountpoint dentries, do not allow removal of
6768 - Avoid lock ordering deadlock issue with write/truncate (6203,5654)
6769 - reserve enough journal credits in fsfilt_start_log for setattr (4554)
6770 - ldlm_enqueue freed-export error path would always LBUG (6149,6184)
6771 - don't reference lr_lvb_data until after we hold lr_lvb_sem (6170)
6772 - don't overwrite last_rcvd if there is a *_client_add() error (6086)
6773 - Correctly handle reads of files with no objects (6243)
6774 - lctl recover will also mark a device active if deactivate used (5933)
6776 - by default create 1 inode per 4kB space on MDS, per 16kB on OSTs
6777 - allow --write-conf on an MDS with different nettype than client (5619)
6778 - don't write config llogs to MDS for mounts not from that MDS (5617)
6779 - lconf should create multiple TCP connections from a client (5201)
6780 - init scripts are now turned off by default; run chkconfig --on
6781 lustre and chkconfig --on lustrefs to use them
6782 - upcalls are no longer needed for clients to recover to failover
6784 - add --abort-recovery option to lconf to abort recovery on device
6786 - add support for an arbitrary number of OSTs (3026)
6787 - Quota support protocol changes.
6788 - forward compatibility changes to wire structs (6007)
6789 - rmmod NALs that might be loaded because of /etc/modules.conf (6133)
6790 - support for mountfsoptions and clientoptions to the Lustre LDAP (5873)
6791 - improved "lustre status" script
6792 - initialize blocksize for non-regular files (6062)
6793 - added --disable-server and --disable-client configure options (5782)
6794 - introduce a lookup cache for lconf to avoid repeated DB scans (6204)
6795 - Vanilla 2.4.29 support
6796 - increase maximum number of obd devices to 520 (6242)
6797 - remove the tcp-zero-copy patch from the suse-2.4 series (5902)
6798 - Quadrics Elan drivers are now included for the RHEL 3 2.4.21 and
6799 SLES 9 2.6.5 kernels
6800 - limit stripes per file to 160 (the maximum EA size) (6093)
6802 2005-03-22 Cluster File Systems, Inc. <info@clusterfs.com>
6805 - don't LASSERT in ll_release on NULL lld with NFS export (4655, 5760)
6806 - hold NS lock when calling handle_ast_error->del_waiting_lock (5746)
6807 - fix setattr mtime regression from lovcleanup merge (4829, 5669)
6808 - workaround for 2.6 crash in ll_unhash_aliases (5687, 5210)
6809 - small ext3 extents cleanups and fixes (5733)
6810 - improved mballoc code, several small races and bugs fixed (5733, 5638)
6811 - kernel version 43 - fix remove_suid bugs in both 2.4 and 2.6 (5695)
6812 - avoid needless client->OST connect, fix handle mismatch (5317)
6813 - fix DLM error path that led to out-of-sync client, long delays (5779)
6814 - support common vfs-enforced mount options (nodev,nosuid,noexec) (5637)
6815 - fix several locking issues related to i_size (5492,5624,5654,5672)
6816 - don't move pending lock onto export if it is already evicted (5683)
6817 - fix kernel oops when creating .foo in unlinked directory (5548)
6818 - fix deadlock in obdfilter statistics vs. object create (5811)
6819 - use time_{before,after} to avoid timer jiffies wrap (5882)
6820 - shutdown --force/--failover stability (3607,3651,4797,5203,4834)
6821 - Do not leak request if server was not able to process it (5154)
6822 - If mds_open unable to find parent dir, make that negative lookup(5154)
6823 - don't create new directories with extent-mapping (5909, 5936)
6825 - fix lustre/lustrefs init scripts for SuSE (patch from Scali, 5702)
6826 - don't hold the pinger_sem in ptlrpc_pinger_sending_on_import
6827 - change obd_increase_kms to obd_adjust_kms (up or down) (5654)
6828 - lconf, lmc search both /usr/lib and /usr/lib64 for Python libs (5800)
6829 - support for RHEL4 kernel on i686 (5773)
6830 - provide error messages when incompatible logs are encountered (5898)
6832 2005-02-18 Cluster File Systems, Inc. <info@clusterfs.com>
6833 * version 1.4.0.10 (1.4.1 release candidate 1)
6835 - don't keep a lock reference when lock is not granted (4238)
6836 - unsafe list practices (rarely) led to infinite eviction loop (4908)
6837 - add per-fs limit of Lustre pages in page cache, avoid OOM (4699)
6838 - drop import inflight refcount on signal_completed_replay error (5255)
6839 - unlock page after async write error during send (3677)
6840 - handle missing objects in filter_preprw_read properly (5265)
6841 - no transno return for symlink open, don't save no-trasno open (3440)
6842 - don't try to complete elan receive that already failed (4012)
6843 - free RPC server reply state on error (5406)
6844 - clean up thread from ptlrpc_start_thread() on error (5160)
6845 - readahead could read extra page into cache that wasn't ejected (5388)
6846 - prevent races in class_attach/setup/cleanup/detach (5260)
6847 - don't dereference de->d_inode after l_dput of de (5458)
6848 - use "int" for stripe value returned from lock_to_stripe (5544)
6849 - mballoc allocation and error-checking fixes in 2.6 (5504)
6850 - block device patches to fix I/O request sizes in 2.6 (5482)
6851 - look up hostnames for IB nals (5602)
6852 - 2.6 changed lock ordering of 2 semaphores, caused deadlock (5654)
6853 - don't start multiple acceptors for the same port (5277)
6854 - fix incorrect LASSERT in mds_getattr_name (5635)
6855 - export a proc file for general "ping" checking (5628)
6856 - fix "lfs check" to not block when the MDS is down (5628)
6858 - service request history (4965)
6859 - put {ll,lov,osc}_async_page structs in a single slab (4699)
6860 - create an "evict_client" /proc entry on OSTs, like the MDS has
6861 - fix mount usage message, return errors per mount(8) (5168)
6862 - change grep [] to grep "[]" in tests so they work in more UMLs
6863 - fix ppc64/x86_64 spec to use %{_libdir} instead of /usr/lib (5389)
6864 - remove ancient LOV_MAGIC_V0 EA support (5047)
6865 - add "disk I/Os in flight" and "I/O req time" stats in obdfilter
6866 - align r/w RPCs to PTLRPC_MAX_BRW_SIZE boundary for performance (3451)
6867 - allow readahead allocations to fail when low on memory (5383)
6868 - mmap locking landed again, after considerable improvement (2828)
6869 - add get_hostaddr() to lustreDB.py for LDAP support (5459)
6871 2004-11-23 Cluster File Systems, Inc. <info@clusterfs.com>
6874 - send OST transaction number in read/write reply to free req (4966)
6875 - don't ASSERT in ptl_send_rpc() if we run out of memory (5119)
6876 - lock /proc/sys/portals/routes internal state, avoiding oops (4827)
6877 - the watchdog thread now runs as interruptible (5246)
6878 - flock/lockf fixes (but it's still disabled, pending 5135)
6879 - don't use EXT3 constants in llite code (5094)
6880 - memory shortage at startup could cause assertion (5176)
6882 - reorganization of lov code
6883 - single portals codebase
6885 - add extents/mballoc support (5025)
6886 - direct I/O reads in the obdfilter (4048)
6887 - kernel patches from LNXI for 2.6 (bluesmoke, perfctr, mtd, kexec)
6889 tbd Cluster File Systems, Inc. <info@clusterfs.com>
6892 - send OST transaction number in read/write reply to free req (4966)
6893 - don't ASSERT in ptl_send_rpc() if we run out of memory (5119)
6894 - lock /proc/sys/portals/routes internal state, avoiding oops (4827)
6895 - the watchdog thread now runs as interruptible (5246)
6896 - handle missing objects in filter_preprw_read properly (5265)
6897 - unsafe list practices (rarely) led to infinite eviction loop (4908)
6898 - drop import inflight refcount on signal_completed_replay error (5255)
6899 - unlock page after async write error during send (3677)
6900 - return original error code on reconstructed replies (3761)
6901 - no transno return for symlink open, don't save no-trasno open (3440)
6903 - add pid to ldlm debugging output (4922)
6904 - bump the watchdog timeouts -- we can't handle 30sec yet
6905 - extra debugging for orphan dentry/inode bug (5259)
6907 2004-11-16 Cluster File Systems, Inc. <info@clusterfs.com>
6910 - fix TCP_NODELAY bug, which caused extreme perf regression (5134)
6911 - allocate qswnal tx descriptors singly to avoid fragmentation (4504)
6912 - don't LBUG on obdo_alloc() failure, use OBD_SLAB_ALLOC() (4800)
6913 - fix NULL dereference in /proc/sys/portals/routes (4827)
6914 - allow failed mdc_close() operations to be interrupted (4561)
6915 - stop precreate on OST before MDS would time out on it (4778)
6916 - don't send partial-page writes before EOF from client (4410)
6917 - discard client grant for sub-page writes on large-page clients (4520)
6918 - don't free dentries not owned by NFS code, check generation (4806)
6919 - fix lsm leak if mds_create_objects() fails (4801)
6920 - limit debug_daemon file size, always print CERROR messages (4789)
6921 - use transno after validating reply (3892)
6922 - process timed out requests if import state changes (3754)
6923 - update mtime on OST during writes, return in glimpse (4829)
6924 - add mkfsoptions to LDAP (4679)
6925 - use ->max_readahead method instead of zapping global ra (5039)
6926 - don't interrupt __l_wait_event() during strace
6928 - add software watchdogs to catch hung threads quickly (4941)
6929 - make lustrefs init script start after nfs is mounted
6930 - fix CWARN/ERROR duplication (4930)
6931 - return async write errors to application if possible (2248)
6932 - add /proc/sys/portal/memused (bytes allocated by PORTALS_ALLOC)
6933 - print NAL number in %x format (4645)
6934 - update barely-supported suse-2.4.21-171 series (4842)
6935 - support for sles 9 %post scripts
6936 - support for building 2.6 kernel-source packages
6937 - support for sles km_* packages
6939 2004-10-07 Cluster File Systems, Inc. <info@clusterfs.com>
6942 - ignore -ENOENT errors in osc_destroy (3639)
6943 - notify osc create thread that OSC is being cleaned up (4600)
6944 - add nettype argument for llmount in #5d in conf-sanity.sh (3936)
6945 - reconstruct ost_handle() like mds_handle() (4657)
6946 - create a new thread to do import eviction to avoid deadlock (3969)
6947 - let lconf resolve symlinked-to devices (4629)
6948 - don't unlink "objects" from directory with default EA (4554)
6949 - hold socknal file ref over connect in case target is down (4394)
6950 - allow more than 32000 subdirectories in a single directory (3244)
6951 - fix blocks count for O_DIRECT writes (3751)
6952 - OST returns ENOSPC from object create when no space left (4539)
6953 - don't send truncate RPC if file size isn't changing (4410)
6954 - limit OSC precreate to 1/2 of value OST considers bogus (4778)
6955 - bind to privileged port in socknal and tcpnal (3689)
6957 - rate limit CERROR/CWARN console message to avoid overload (4519)
6958 - GETFILEINFO dir ioctl returns LOV EA + MDS stat in 1 call (3327)
6959 - basic mmap support (3918)
6960 - kernel patch series update from b1_4 (4711)
6962 2004-09-16 Cluster File Systems, Inc. <info@clusterfs.com>
6965 - avoid crash during MDS cleanup with OST shut down (2775)
6966 - fix loi_list_lock/oig_lock inversion on interrupted IO (4136)
6967 - don't use bad inodes on the MDS (3744)
6968 - dynamic object preallocation to improve recovery speed (4236)
6969 - don't hold spinlock over lock dumping or change debug flags (4401)
6970 - don't zero obd_dev when it is force cleaned (3651)
6971 - print grants to console if they go negative (4431)
6972 - "lctl deactivate" will stop automatic recovery attempts (3406)
6973 - look for existing locks in ldlm_handle_enqueue() (3764)
6974 - don't resolve lock handle twice in recovery avoiding race (4401)
6975 - revalidate should check working dir is a directory (4134)
6977 - don't always mark "slow" obdfilter messages as errors (4418)
6979 2004-08-24 Cluster File Systems, Inc. <info@clusterfs.com>
6982 - don't close LustreDB during write_conf until it is done (3860)
6983 - fix typo in lconf for_each_profile (3821)
6984 - allow dumping logs from multiple threads at one time (3820)
6985 - don't allow multiple threads in OSC recovery (3812)
6986 - fix debug_size parameters (3864)
6987 - fix mds_postrecov to initialize import for llog ctxt (3121)
6988 - replace config semaphore with spinlock (3306)
6989 - be sure to send a reply for a CANCEL rpc with bad export (3863)
6990 - don't allow enqueue to complete on a destroyed export (3822)
6991 - down write_lock before checking llog header bitmap (3825)
6992 - recover from lock replay timeout (3764)
6993 - up llog sem before sending rpc (3652)
6994 - reduce ns lock hold times when setting kms (3267)
6995 - change a dlm LBUG to LASSERTF, to maybe learn something (4228)
6996 - fix NULL deref and obd_dev leak on setup error (3312)
6997 - replace some LBUG about llog ops with error handling (3841)
6998 - don't match INVALID dentries from d_lookup and spin (3784)
6999 - hold dcache_lock while marking dentries INVALID and hashing (4255)
7000 - fix invalid assertion in ptlrpc_set_wait (3880)
7002 - add libwrap support for the TCP acceptor (3996)
7003 - add /proc/sys/portals/routes for non-root route listing (3994)
7004 - allow setting MDS UUID in .xml (2580)
7005 - print the stack of a process that LBUGs (4228)
7007 2004-07-14 Cluster File Systems, Inc. <info@clusterfs.com>
7010 - don't cleanup request in ll_file_open() on failed MDS open (3430)
7011 - make sure to unset replay flag from failed open requests (3440)
7012 - if default stripe count is 0, use OST count for inode size (3636)
7013 - update parent mtime/ctime on client for create/unlink (2611)
7014 - drop dentry ref in ext3_add_link from open_connect_dentry (3266)
7015 - free recovery state on server during a forced cleanup (3571)
7016 - unregister_reply for resent reqs (3063)
7017 - loop back devices mounting and status check on 2.6 (3563)
7018 - fix resource-creation race that can provoke i_size == 0 (3513)
7019 - don't try to use bad inodes returned from MDS/OST fs lookup (3688)
7020 - more debugging for page-accounting assertion (3746)
7021 - return -ENOENT instead of asserting if ost getattr+unlink race (3558)
7022 - avoid deadlock after precreation failure (3758)
7023 - fix race and lock order deadlock in orphan handling (3450, 3750)
7024 - add validity checks when grabbing inodes from l_ast_data (3599)
7026 - add /proc/.../recovery_status to obdfilter (3428)
7027 - lightweight CDEBUG infrastructure, debug daemon (3668)
7028 - change default OSC RPC parameters to be better on small clusters
7029 - turn off OST read cache for files smaller than 32MB
7030 - install man pages and include them in rpms (3100)
7031 - add new init script for (un)mounting lustre filesystems (2593)
7032 - run chkconfig in %post for init scripts (3701)
7033 - drop scimac NAL (unmaintained)
7035 2004-06-17 Cluster File Systems, Inc. <info@clusterfs.com>
7038 - clean kiobufs before and after use (3485)
7039 - strip trailing '/'s before comparing paths with /proc/mounts (3486)
7040 - remove assertions to work around "in-flight rpcs" recovery bug (3063)
7041 - change init script to fail more clearly if not run as root (1528)
7042 - allow clients to reconnect during replay (1742)
7043 - fix ns_lock/i_sem lock ordering deadlock for kms update (3477)
7044 - don't do DNS lookups on NIDs too small for IP addresses (3442)
7045 - re-awaken ptlrpcd if new requests arrive during check_set (3554)
7046 - fix cond_resched (3554)
7047 - only evict unfinished clients after recovery (3515)
7048 - allow bulk resend, prevent data loss (3570)
7049 - dynamic ptlrpc request buffer allocation (2102)
7050 - don't allow unlinking open directory if it isn't empty (2904)
7051 - set MDS/OST threads to umask 0 to not clobber client modes (3359)
7052 - remove extraneous obd dereference causing LASSERT failure (3334)
7053 - don't use get_cycles() when creating temp. files on the mds (3156)
7054 - hold i_sem when setting i_size in ll_extent_lock() (3564)
7055 - handle EEXIST for set-stripe, set proper directory name (3336)
7057 - servers can dump a log evicting a client - lustre.dump_on_timeout=1
7058 - fix ksocknal_fmb_callback() error messages (2918)
7060 2004-05-27 Cluster File Systems, Inc. <info@clusterfs.com>
7063 - don't copy lvb into (possibly NULL) reply on error (2983)
7064 - don't deref dentry after dput, don't free lvb on error (2922)
7065 - use the kms to determine writeback rpc length (2947)
7066 - increment oti_logcookies when osc is inactive (2948)
7067 - update client's i_blocks count via lvb messages (2543)
7068 - handle intent open/close of special files properly (1557)
7069 - mount MDS with errors=remount-ro, like obdfilter (2009)
7070 - initialize lock handle to avoid ASSERT on error cleanup (3057)
7071 - don't use cancelling-locks' kms values (2947)
7072 - use highest lock extent for kms, not last one (2925)
7073 - don't dereference ERR_PTR() dentry in error handling path (3107)
7074 - fix thread race in portals_debug_dumplog() (3122)
7075 - create lprocfs device entries at setup instead of at attach (1519)
7076 - common AST error handler, don't evict client on completion race (3145)
7077 - zero nameidata in detach_mnt in 2.6 (3118)
7078 - verify d_inode after revalidate_special is valid in 2.6 (3116)
7079 - use lustre_put_super() to handle zconf unmounts in 2.6 (3064)
7080 - initialize RPC timeout timer earlier for 2.6 (3219)
7081 - don't dereference NULL reply buffer if mdc_close was never sent (2410)
7082 - print nal/nid for unknown nid (3258)
7083 - additional checks for oscc recovery before doing precreate (3284)
7084 - fix ll_extent_lock() error return code for 64-bit systems (3043)
7085 - don't crash in mdc_close for bad permissions on open (3285)
7086 - zero i_rdev for non-device files (3147)
7087 - clear page->private before handing to FS, better assertion (3119)
7088 - tune the read pipeline (3236)
7089 - fix incorrect decref of invalidated dentry (2350)
7090 - provide read-ahead stats and refine rpc in flight stats (3328)
7091 - don't hold journal transaction open across create RPC (3313)
7092 - update atime on MDS at close time (3265)
7093 - close LDAP connection when recovering to avoid server load (3315)
7094 - update iopen-2.6 patch with fixes from 2399,2517,2904 (3301)
7095 - don't leak open file on MDS after open resend (3325)
7096 - serialize filter_precreate and filter_destroy_precreated (3329)
7097 - loop device shouldn't call sync_dev() for nul device (3092)
7098 - clear page cache after eviction (2766)
7099 - resynchronize MDS->OST in background (2824)
7100 - refuse to mount the same filesystem twice on same mountpoint (3394)
7101 - allow llmount to create routes for mounting behind routers (3320)
7102 - push lock cancellation to blocking thread for glimpse ASTs (3409)
7103 - don't call osc_set_data_with_check() for TEST_LOCK matches (3159)
7104 - fix rare problem with rename on htree directories (3417)
7106 - allow default OST striping configuration per directory (1414)
7107 - fix compilation for qswnal for 2.6 kernels (3125)
7108 - increase maximum number of MDS request buffers for large systems
7109 - change liblustreapi to be useful for external progs like lfsck (3098)
7110 - increase local configuration timeout for slow disks (3353)
7111 - allow configuring ldlm AST timeout - lustre.ldlm_timeout=<seconds>
7113 2004-03-22 Cluster File Systems, Inc. <info@clusterfs.com>
7116 - fixes for glimpse AST timeouts / incorrectly 0-sized files (2818)
7117 - don't overwrite extent policy data in reply if lock was blocked (2901)
7118 - drop filter export grants atomically with removal from device (2663)
7119 - del obd_self_export from work_list in class_disconnect_exports (2908)
7120 - don't LBUG if MDS recovery times out during orphan cleanup (2530)
7121 - swab reply message in mdc_close, other PPC fixes (2464)
7122 - fix destroying of named logs (2325)
7123 - overwrite old logs when running lconf --write_conf (2264)
7124 - bump LLOG_CHUNKSIZE to 8k to allow for larger clusters (2306)
7125 - fix race in target_handle_connect (2898)
7126 - mds_reint_create() should take same inode create lock (2926)
7127 - correct journal credits calculated for CANCEL_UNLINK_LOG (2931)
7128 - don't close files for self_export to avoid uninitialized obd (2936)
7129 - allow MDS with the same name as client node (2939)
7130 - hold dentry reference for closed log files for unlink (2325)
7131 - reserve space for all logs during transactions (2059)
7132 - don't evict page beyond end of stripe extent (2925)
7133 - don't oops on a deleted current working directory (2399)
7134 - handle hard links to targets without a parent properly (2517)
7135 - don't dereference NULL lock when racing during eviction (2867)
7136 - don't grow lock extents when lots of conflicting locks (2919)
7138 2004-03-04 Cluster File Systems, Inc. <info@clusterfs.com>
7141 - account for cache space usage on clients to avoid data loss (974)
7142 - lfsck support in lustre kernel code (2349)
7143 - reduce journal credits needed for BRW writes (2370)
7144 - orphan handling to avoid losing space on client/server crashes
7145 - ptlrpcd can be blocked, stopping ALL progress (2477)
7146 - use lock value blocks to assist in proper KMS, faster stat (1021)
7147 - takes i_sem instead of DLM locks internally on obdfilter (2720)
7148 - recovery for initial connections (2355)
7149 - fixes for mds_cleanup_orphans (1934)
7150 - abort_recovery crashes MDS in b_eq (mds_unlink_orphan) (2584)
7151 - block all file creations until orphan recovery completes (1901)
7152 - client remove rq_connection from request struct (2423)
7153 - conf-sanity test_5, proper cleanup in umount log not availale (2640)
7154 - recovery timer race (2670)
7155 - mdc_close recovey bug (2532)
7156 - ptlrpc cleanup bug (2710)
7157 - mds timeout on local locks (2588)
7158 - namespace lock held during RPCs (2431)
7159 - handle interrupted sync write properly (2503)
7160 - don't try to handle a message that hasn't been replied to (2699)
7161 - client assert failure during cleanup after abort recovery (2701)
7162 - leak mdc device after failed mount (2712)
7163 - ptlrpc_check_set allows timedout requests to complete (2714)
7164 - wait for inflight reqs when ptlrpcd finishes (2710)
7165 - make sure unregistered services are removed from the srv_list
7166 - reset bulk XID's when resending them (caught by 1138 test)
7167 - unregister_bulk after timeout
7168 - fix lconf error (2694)
7169 - handle write after unfinished setstripe, stripe-only getstripe (2388)
7170 - readahead locks pages, leaves pending causing memory pressure (2673)
7171 - increase OST request buffers to 4096 on large machines (2729)
7172 - fix up permission of existing directories in simple_mkdir (2661)
7173 - init deleted item, add assertions ptlrpc_abort_inflight() (2725)
7174 - don't assign transno to errored transactions (2742)
7175 - don't delete objects on OST if given a bogus objid from MDS (2751)
7176 - handle large client PAGE_SIZE readdir on small PAGE_SIZE MDS (2777)
7177 - if rq_no_resend, then timeout request after recovery (2432)
7178 - fix MDS llog_logid record size, 64-bit array alignment (2733)
7179 - don't call usermode_helper from ptlrpcd, DEFAULT upcall (2773)
7180 - put magic in mount.lustre data, check for bad/NULL mount data (2529)
7181 - MDS recovery shouldn't delete objects that it has given out (2730)
7182 - if enqueue arrives after completion, don't clobber LVB (2819)
7183 - don't unlock pages twice when trigger_group_io returns error (2814)
7184 - don't deref NULL rq_repmsg if ldlm_handle_enqueue failed (2822)
7185 - don't write pages to disk if there was an error (1450)
7186 - don't ping imports that have recovery disabled (2676)
7187 - take buffered bytes into account when balancing socknal conn (2817)
7188 - hold a DLM lock over readdir always, use truncate_inode_pages (2706)
7189 - reconnect unlink llog connection after MDS reconnects to OST (2816)
7190 - remove little-endian swabbing of llog records (1987)
7191 - set/limit i_blksize to LL_MAX_BLKSIZE on client (2884)
7192 - retry reposting request buffers if they fail (1191)
7193 - grow extent at grant time to avoid granting a revoked lock (2809)
7194 - lock revoke doesn't evict page if covered by a second lock (2765)
7195 - disable VM readahead to avoid reading outside lock extents (2805)
7197 - return LL_SUPER_MAGIC from statfs for the filesystem type (1972)
7198 - updated kernel patches for hp-2.4.20 kernel (2681)
7200 2004-02-07 Cluster File Systems, Inc. <info@clusterfs.com>
7203 - fix truncated write corruption (2366)
7204 - fix for failed assertion in iopen_connect_dentry (1792,2517)
7206 - don't flag the ptlrpcd thread with PF_MEMALLOC (2636)
7207 - ensure len(uuid) < 37 in lmc (1171)
7208 - fix ia64 OOPS in llog_test (2255)
7209 - zero end of page at obdfilter for partial page writes (2648)
7210 - don't leave stale dentries around after renames (bug 2428)
7211 - fix timeouts when evicting a client with a single lock held (2642)
7212 - set deadline for the initial HELLO message to drain (2634)
7213 - print out dotted-quad IP addresses in the socknal (2302)
7215 - additional debugging for MDS client eviction problem (2443)
7216 - fix mkfsoptions support for osts (2603, 2604)
7218 2004-01-27 Cluster File Systems, Inc. <info@clusterfs.com>
7221 - add series for the vanilla 2.6.0 kernel
7222 - add series for the vanilla 2.4.24 kernel
7223 - add series for a cray x86/64 UL kernel drop
7224 - fix xattr patches for the vanilla 2.4.19 series
7226 - generate true UUIDs in lmc (1171)
7227 - have portals stack dumping break in UML (2466)
7228 - avoid bad dchild deref; avoid inum lock w/o creation (2362)
7229 - allocate with _NOFS in ldlm to avoid deadlock (1933)
7230 - wake callback waiting threads on client eviction (2460)
7231 - Add --ptldebug and --subsystem to lmc (1719)
7232 - update assertion to allow safe interrupt allocation
7233 - set rq_no_resend for cancel requests (2432)
7234 - recalculate ptlrpcd timeout after resend (2494)
7235 - call vfs_rmdir when removing pending directories (2368)
7236 - fix renaming a file to itself (2429)
7237 - lmc creates a default one-stripe lov (2454)
7238 - expand procfs space to handle large clusters (2326)
7239 - increase UML stack to avoid overflow
7240 - update lconf's list of debug and subsystem masks
7241 - fix lfs find --obd (2510)
7242 - /proc tunable for disabling filter read caching (2591)
7243 - stop rpm packages from altering slapd.conf (2301)
7244 - disable nagle in the socknal under 0conf (2578)
7245 - choose mds inode size based on stripe count (2572)
7246 - fix kernel-source rpm problems (2516)
7248 - add --disable-doc to avoid pdf generation (2421)
7249 - update documentation, tests, type-os, comments
7250 - avoid format warnings on ia64
7251 - remove the TOE NAL
7252 - tiny code cleanups by removing unused fields
7254 2004-01-07 Cluster File Systems, Inc. <info@clusterfs.com>
7257 - fix obvious semaphore misuse in as-yet-unused setattr path (2348)
7258 - remove the most blatant lies from BUILDING file (2371)
7259 - change default debug level to reasonable production setting
7260 - reduce client side cache size to reduce cache flush time
7261 - reduce max RPCs in flight to avoid unnecessary file fragmentation
7262 - make TCP zerocopy and pinger support enabled by default (2476)
7263 - sync writes completed after process exits caused crashes (2319)
7264 - maintain correct mount count on the MDS (2356)
7265 - backout 1557, because 2316 wasn't really fixed
7266 - better file I/O statistics gathering in /proc
7267 - don't take unnecessary, deadlock-inducing bug in readpage (2383)
7268 - another kernel patch to fix zero-copy TCP function export
7269 - don't take duplicate lock when processing re-sent getattr (2420)
7270 - lctl uses obd_self_export instead of creating new conn (2353)
7271 - MDS/OST recovery case which requires object creation asserted (2425)
7272 - move lfs from /usr/sbin to /usr/bin in packages
7273 - fix race between mds_client_add and mds_client_free (2417)
7274 - use kmalloc instead of slabs in portals (2430)
7275 - don't create duplicate records when a failover MDS is present (2442)
7276 - remove unnecessary mount age check (2332)
7277 - don't remove directory inodes from locks prematurely (2451)
7278 - don't break if MDS service name is the same as hostname (2103)
7279 - fix races in client write RPC generation when cache full (2482)
7281 2003-12-13 Cluster File Systems, Inc. <info@clusterfs.com>
7284 - remove now-unused request->rq_obd (278)
7285 - if an allocation fails, print out how much memory we've used (1933)
7286 - use PORTAL_SLAB_ALLOC for structures, to get GFP_MEMALLOC (1933)
7287 - add the "configurable stack size" patch to most series files (1256)
7288 - ability to write large log records, for 100+ OST configs (2306)
7289 - fix NULL deref when filter_prep fails (2314)
7290 - fix operator precedence error in filter_sync
7291 - dynamic allocation of socknal TX descriptors (2315)
7292 - fix a missed case in the GFP_MEMALLOC patch, can cause deadlock (2310)
7293 - fix gcc 2.96 compilation problem in xattr kernel patch (2294)
7294 - ensure that CWARN messages in Portals always get to the syslog
7295 - __init/__exit are not for prototype decls (ldlm_init/exit)
7296 - x86-64 compile warning fixes
7297 - fix gateway LMC keyword conflict (2318)
7298 - fix MDS lock inversions in getattr/reint paths (1844)
7299 - fix a rare lock re-ordering bug, which caused deadlock (2322)
7300 - fix i_sem/journal inversion in fsfilt_ext3_write_record (2306)
7301 - DLM race condition prevented some lock evictions (2328)
7302 - ENOMEM detection and retry on socknal sends (2230)
7303 - use GFP_NOFS throughout Lustre, to combat ENOMEM (2230)
7304 - move osc_rpcd into ptlrpc, for use in MDC and others (2329)
7305 - protect MDS inode fsdata with stronger locking; fixes assertion (2313)
7306 - better error messages when a client is rejected during recovery (1505)
7307 - avoid cancelling locks which were never granted, after failure (2330)
7308 - fix i_sem/journal inversion in mds_client_add (2333)
7309 - fix truncate/getattr lock cycle deadlock (2334)
7310 - use rpcd to send close; allows resend after timeout, avoid leak (1897)
7311 - fix two rare exit paths which could leak an l_lock() ref (2321)
7312 - fencepost error in MDS/OST orphan recovery (2226)
7313 - make log record alignment 8 bytes (1988)
7314 - lstripe now fails when requested offset > ost_count (2237)
7315 - ensure that all kernel series have a complete list.h (1607)
7316 - fix crashes in special-file operations (2316)
7317 - lctl create/brw OID mismatch, caused by obsolete filter loop (2339)
7319 - allow configurable automake binary, for testing new versions
7320 - small update to the lfs documentation
7322 2003-12-03 Cluster File Systems, Inc. <info@clusterfs.com>
7324 * fix negative export reference count in fsfilt_sync (2312)
7326 2003-12-01 Cluster File Systems, Inc. <info@clusterfs.com>
7327 * release candidate 0.9.1
7329 - orphans are moved into the PENDING directory for possible recovery
7330 - replayed opens now open by fid for orphan/rename safety (1042)
7331 - last close of an orphan inode generates a transno (683)
7332 - chdir() and mount() now pin the directory entry (1020)
7333 - avoid CERROR in normal ll_setattr_raw() error case (1500)
7334 - discard very old requests without processing them (1502)
7335 - remove some common, well-understood CERRORs (1505)
7336 - require O_DIRECT I/O to be page-sized to workaround IA64 crash (1609)
7337 - clear "grant" flags in OST replies until OST grant code lands (1644)
7338 - fix read performance by not clobbering i_blksize on client (1598)
7339 - fix __ldlm_handle2lock oops by not dereferencing lock after PUT (1625)
7340 - make LRU size a /proc tunable, clears locks when reduced (707)
7341 - fix some lprocfs rot that prevented ptlbd from loading (1732)
7342 - server locks take references on exports now (1558)
7343 - build fixes for 2.4.20-rh trees (1663)
7344 - return an error from lov_create if all OSCs are inactive (1751)
7345 - fix import levels when a reconnect happens without a timeout (1597)
7346 - exit early from mds_open if we get a lookup error (1749)
7347 - partial page read at EOF wouldn't wait for disk before sending (1642)
7348 - avoid NULL deref in obdfilter when reading page past EOF (1592)
7349 - avoid LASSERT in ll_intent_lock if server failed very early (1090)
7350 - fix LBUG in ll_it_open_error with rc = -2 (1861)
7351 - write/truncate lock inversion (1639)
7352 - Don't auto-load obdclass, portals modules during cleanup (1495)
7353 - fix timestamps from jumping to "now" (1763)
7354 - extra journal assertions (1648)
7355 - add an extra multiunlink test (1771)
7356 - fix read_record/write_record API (1776)
7357 - fix leak of offset_extent, possible incorrect i_size later (1772)
7358 - fix lasserts in mis-matched transnos during open-unlink testing (1541)
7359 - Debugging for the kqswnal_get_idle_tx problems (1820)
7360 - Allow recovery to be attempted multiple times (1536)
7361 - Write out MDS last_rcvd file after it is first created (1600)
7362 - Fix tx_descriptor leak in failed transmit situations (1827)
7363 - ext3 journaling fixes for assertion failure after IO error (1871)
7364 - class_export_put() on freed export after completion AST error (1896)
7365 - Fix revalidate looping in VFS (1322)
7366 - Don't access a freed export during MDS_REINT timeout (1521)
7367 - Add open-unlink recovery support on the MDS (1673,1764)
7368 - Return an error if no MDS data was read from last_rcvd (1946)
7369 - Fix for lookup "." or ".." crash on error (1932,1931,1935)
7370 - Don't setup a disk device that doesn't match exported UUID (317)
7371 - Reduce bulk RPC timeout to avoid cascading client/OST failures (1845)
7372 - avoid committing NULL handle in force close
7373 - local.sh is now a one-stripe LOV configuration
7374 - POSIX utime.4 -EPERM on FIFO not owned by user (56)
7375 - fix ext3 htree duplicate directory entry corruption (1516)
7376 - POSIX creat.13, fstat.1, open.18, stat.3 new file atime/mtime (2020)
7377 - update to new LOV EA format (2097)
7378 - interoperability for different PAGE_SIZE/wordsize (686,1821,1343,2042)
7380 2003-06-15 Phil Schwan <phil@clusterfs.com>
7383 - imports and exports cleanup too early, need refcounts (349, 879, 1045)
7384 - per-import/export recovery handling (958, 931, 959)
7385 - multiple last-rcvd slots, for serving multiple FSes (949)
7386 - connections are again shared between multiple imp/exports (963, 964)
7387 - "umount -f" would hang if any requests needed to be sent (393, 978)
7388 - avoid pinning large req buffer by copying for queued messages (989)
7389 - add "uuid" to "lctl device" command to help upcalls (991)
7390 - "open" RPCs with transnos would confuse recovery counters (1037)
7391 - do proper endian conversion of all wire messages (288, 340, 891)
7392 - remove OST bulk get LBUGs, fix ost_brw_write cleanup (1126)
7393 - call waiting locks callback from LDLM recovery thread (1127, 1151)
7394 - fix ptlrpc_connection leak in target_handle_connect (1174)
7395 - fix import refcounting bug in OST and MDS cleanup (1134)
7396 - if an invalid-at-open-time OSC returned before close(), LBUG (1150)
7397 - fix very unlikely obd_types race condition (501)
7398 - remove osc_open hack for echo_client (1187)
7399 - we leaked exports/dlmimps for forcibly disconnected clients (1143)
7400 - a failure in read_inode2 leads to deadlock (1139)
7401 - cancel ack-locks as soon as transaction is committed (1072)
7402 - fix major leaks and crashes in the bulk I/O path (937, 1057)
7403 - make sure to commitrw after any preprw to avoid deadlock (1162)
7404 - failing to execute a file in a lustre FS would lock inode (1203)
7405 - small DEBUG_REQ fix to avoid dereferencing a NULL (1227)
7406 - don't ASSERT while cleaning up an incompletely-setup obd (1248)
7407 - obd_uuid2tgt would walk off the end of the list (1255)
7408 - on IA64 the osc would give portals incorrect bulk size (1258)
7409 - fix debug daemon ioctl interface; allows daemon on ia64 (1274)
7410 - fix lock inversion caused by new llite matching code (1282)
7411 - limit the number of dirty pages on a client to 10MB (1286)
7412 - timed out locks were not being corrected cancelled (1289)
7413 - fix O_DIRECT above 4GB on IA-32 (1292)
7414 * major user-visible changes
7415 - fail out/fail over policy now controlled by the upcall (993)
7417 - add OBD_PING to check server availability and failure (954)
7418 - lustre messages are now sent in sending host order (288, 340, 891)
7419 - add eadatalen to MDS getattr reply (340)
7420 - OST read replies may contain second buffer, with per-page status (593)
7422 2003-03-11 Phil Schwan <phil@clusterfs.com>
7425 - LDLM_DEBUG macro fix, for gcc 3.2 (850)
7426 - failed open()s could cause deadlock; fixed (867, 869)
7427 - stop cancelling OST locks when files are closed (481)
7428 - overlapping XID spaces caused network corruption (851, 853)
7429 - fix unsafe fsfilt counter arithmetic; change to atomic_t
7430 - setattr_raw added, to do single-RPC, server-side setattrs
7431 - lmc/lconf syntax change for OST UUIDs
7432 - fix crashy race condition between ptlrpc_free_req and osc_close
7433 - don't use request in mdc_enqueue if we hit a timeout (889)
7434 - don't set the inode i_size for regular files from the MDS (896)
7435 - handle out of order completion AST (842)
7436 - don't LBUG if a lock request times out after receiving AST (913)
7437 - avoid d_rehash race in ll_find_alias by rehashing inside dcache_lock
7438 - if a bad lock AST arrives, send an error instead of dropping entirely
7439 - return 0 from revalidate2 if ll_intent_lock returns -EINTR (912)
7440 - fix leak in bulk IO when only partially completed (899, 900, 926)
7441 - fix O_DIRECT for ia64 (55)
7442 - (almost) eliminate Lustre-kernel-thread effects on load average (722)
7443 - C-z after timeout could hang a process forever; fixed (977)
7445 - client-side I/O cache (678, 924, 929, 941, 970)
7447 - READPAGE and SETATTRs which don't take server-side locks get
7450 2003-02-11 Phil Schwan <phil@clusterfs.com>
7453 - Fix ldlm_lock_match on the MDS to avoid matching remote locks (592)
7454 - Fix fsfilt_extN_readpage() to read a full page of directory
7455 entries, or fake the remainder if PAGE_SIZE != blocksize (500)
7456 - Avoid extra mdc_getattr() in ll_intent_lock when possible (534, 604)
7457 - Fix imbalanced LOV object allocation and out-of-bound access (469)
7458 - Most intent operations were removed, in favour of a new RPC mode
7459 that does a single RPC to the server and bypasses most of the VFS
7460 - All LDLM resource ID arrays were removed in favour of ldlm_res_id
7461 - Aggressively cancel local locks on DLM servers
7462 - mds_reint_unlink sends EA to the client if it's the last nlink.
7463 client uses that EA to unlink OST objects.
7464 - mds_reint_{rename,unlink,link} were rewritten to take ordered locks
7465 - recursive symlinks were fixed (439)
7466 - fixed NULL deref in DEBUG_REQ
7467 - filter_update_lastobjid no longer calls sync, which annoyed extN
7468 - fixed multi-client small-writes to a single file problem (445)
7469 - fixed mtime updates during file writes (607)
7470 - fixed vector writes on obdfilter causing problems when ENOSPC (670)
7471 - fixed bug in obd_brw_read/write() (under guise of testing 367)
7472 - fixed Linux OST size reporting problem (444, 656)
7473 - OST now updates object mtime with writes or setattr (607, 619)
7474 - client verifies file size before zeroing page past EOF (445)
7475 - OST now writes last allocated objid to disk with allocation (108)
7476 - LOV on echo now works (409)
7478 - mds_reint_unlink sends a new buffer, with the EA included. this
7479 buffer is only valid if body->valid & OBD_MD_FLEASIZE, which is only
7480 set if a regular file was being unlinked, and it was the last link
7481 - use PtlGet from the target for bulk writes (315)
7482 - OST now updates object mtime with writes or setattr (607, 619)
7483 - LDLM now has a grant-time callback to revalidate locked items, if
7485 - Many MDS operations were reorganized to combat race conditions
7487 - Merge b_intel branch (updated lprocfs code) - now at /proc/fs/lustre
7488 - configure check to avoid gcc version 2.96 20000731-2.96-98) (606)
7490 2003-01-06 Andreas Dilger <adilger@clusterfs.com>
7493 - Fully reactivate OST imports after reconnection (512, others)
7494 - Make sure client sees our -ENOTCONN from mds_handle (513 - partial)
7495 - More graceful error handling for truncating on dead OST (515)
7496 - Don't error out unless we're actually accessing dead stripes (474)
7497 - Fix garbage sizes when stripes are missing (410)
7498 - LRU counters were broken, causing constant lock purge (433, 432)
7499 - garbage on read from stripes with failed OSTs (441)
7500 - mark OSCs as active before reconnecting during recovery (438)
7501 - lov_enqueue and lov_cancel need to handle inactive OSTs (403)
7502 - lfind did not preserve OST order in output (443)
7503 - symlinks cause hung clients, incorrect data (439)
7504 - stop dereferencing request after dropping refcount (457)
7505 - don't LASSERT(spin_is_locked) on non-SMP (455)
7506 - fixes for many rename() bugs
7507 - fstat didn't correctly synchronize attributes (399)
7508 - server must handle lock cancellation during blocking AST prep (487)
7509 - bulk descriptors were free()d too soon (511)
7510 - fix paths in lconf, which would load incorrect modules (451, 507)
7511 - fix confusing lconf 'host not found' error message (386)
7512 - fix lock order deadlock on OST (O/R i_sem before journal ops, 478)
7513 - fix race condition in mdc_blocking_ast() for inode access (526)
7514 - fix lov_unpackmd() unpacking wrong number of stripes (537)
7515 - fix lov_set_osc_active() marking wrong OSC inactive (440)
7516 - fix bad lstripe lov_unpackmd() assertion (fix layering too) (527)
7517 - fix multiple writes of stripe MD to MDS (358, maybe 519)
7518 - fix lstripe in several ways (kernel side) (527)
7519 - fix request leak in ldlm_cli_enqueue (262)
7520 - incorrect OSC was marked inactive after OST failure
7521 - call mds_fs_cleanup before unmounting filesystem (524)
7522 - fix races between taking ns_lock and ldlm_lock_change_resource
7523 - fix races updating LOV export open file list
7524 - fix lov_enqueue error path, avoid decref-ing bad lock handle
7525 - fix recovery NULL deref in ldlm_cli_cancel_unused
7526 - fix some DLM races by using new hash table for lock handles (419)
7527 - permit the client to specify desired inodes, at replay
7528 - duplicate requests when we queue them for replay reintegration
7529 - fix last_rcvd offset calculation
7530 - sync after each recovered transaction, so we always make progress
7531 - never, not always, ERESTART requests without transnos
7532 - store the lov_desc in the MDS, so we don't depend on getlovinfo to
7534 - skip replay if the MDS says that the client is already connected
7535 - don't check for a recovery-enabled export to match lctl's UUID
7536 - don't INC_USE_COUNT for phantom exports
7537 - don't crash when cleaning up phantom exports (567)
7538 - don't double-finish or set replay data for errored mdc_open requests
7539 - abort requests when they time out, so we don't get old replies
7540 - send/receive replies for AST messages again
7541 - if the client says that it doesn't have the lock, cancel it on the
7543 - if we timeout during I/O, don't try to cancel an in-use lock; instead
7544 mark it as destroyed, it will all work out when decref is called
7545 - fix module use counts (22, 581)
7547 - ASTs now expect a reply (server cancels lock on error reply)
7549 2002-12-02 Andreas Dilger <adilger@clusterfs.com>
7552 - fix many simultaneous client startup (392)
7553 - fix dentry->d_it clobbering
7554 - credentials weren't being shipped for readdir/getattr operations
7555 - remove invalid assertions triggered during some concurrent MD
7557 - proper Lustre versions added (336, 389)
7558 - fix memory leak for create error case (398)
7559 - fix LOV locking bug that would get cli/srv out of sync
7560 - fix echo client over LOV (409)
7561 - fix dbench 2, extN refcount problem (170, 258, 356, 418)
7562 - fix double-O_EXCL intent crash (424)
7563 - avoid sending multiple lock CANCELs (352)
7565 - MDS can do multi-client recovery (modulo bugs in new code)
7567 - many updates, edits, cleanups
7569 2002-11-18 Phil Schwan <phil@clusterfs.com>
7572 - fix null d_it dereference (346)
7573 - fix full OST/dbench hang (333)
7574 - fix permission problem with file removal (286)
7575 - fix removal of OSCs from LOV when they fail
7576 - fix NULL deref during bulk timeout (214)
7577 - fix problems related to multiple filesystems on one MDS (241)
7578 - fixed serious subtle metadata locking bugs
7579 - free locks on clients when inodes are removed due to memory
7581 - fix inode pointer in lock data (285)
7582 - partial support for multiple MDS on a single host (241)
7583 - data locks weren't cancelled at clear_inode time (290, 311)
7584 - intent locks could lead to unbounded lock growth (205)
7585 - added a maximum lock count, an LRU list, and a flusher
7586 - fix multiple rename (365)
7587 - properly abstracted the echo client
7588 - OSC locked 1 byte too many; fixed
7589 - rewrote brw callback code:
7590 - fixed recovery bugs related to LOVs (306)
7591 - fixed too-many-pages-in-one-write crash (191)
7592 - fixed (again) crash in sync_io_timeout (214)
7593 - probably fixed callback-related race (385)
7595 - Add capability to MDS protocol
7596 - LDLM cancellations and callbacks on different portals
7598 2002-10-28 Andreas Dilger <adilger@clusterfs.com>
7601 - limit client IOV size to PTL_MD_MAX_IOV (611336, 191)
7602 - defer open object destruction to close time (601981, 138)
7603 - open/close OST file handle in obdo (OBD_MD_FLHANDLE) (601981, 138)
7604 - move LDLM_ENQUEUE/CONVERT back to MDS portal (625069)
7605 - abstract ll_lookup2, fix ll_revalidate2 to use abstraction (256)
7606 - don't call obd_setattr in ll_file_release for destroyed objects
7607 * protocol change to lustre_msg: move |version| and add |flags|
7608 * protocol change to osc_punch: "start" in "o_size", "end" in "o_blocks"
7609 * lock replay: for LDLM_FL_REPLAY trust client to do right thing
7610 * added replay of create, unlink, link and rename operations during
7611 MDS failover; recovery should be much more robust now
7612 * remove failed OSCs from LOVs (only lov_create uses this so far)
7613 * the lustre-HOWTO was brought (more) up to date (582544)
7615 2002-10-23 Phil Schwan <phil@clusterfs.com>
7618 - in-use dentries weren't being reused properly (617851)
7619 - prevent multiple LDLM setup (599178)
7620 - fix LOV size calculations for truncate (617853)
7621 - fix client handling of MDS intent errors (POSIX)
7622 - fix permission bug in lovstripe.c test (624321)
7623 - fix MDS thread deadlock - move LDLM handler to DLM portal (625069)
7624 - truncate past end of file could corrupt data
7625 - proper cleanup after timeouts, crashes, etc (592524, 550815)
7626 - a race in recovery could return ETIMEDOUT to apps (623947)
7627 - building outside the source directory was fixed
7628 * the lustre-HOWTO was brought (more) up to date (582544)
7629 * major progress was made on recovery functionality
7631 2002-10-10 Phil Schwan <phil@clusterfs.com>
7634 - recovery deadlock fix
7635 - rm -rf causes LBUG fix (617817)
7636 - file open by multiple tasks fix (618962)
7637 - directory permissions bugs (602707 and 620007)
7638 - journal_stop fixed with locking (611313)
7639 - O_APPEND failures resolved (618273, perhaps 614459)
7640 - lconf PATH fix (619770)
7641 - IA64 build fix (621450)
7642 - RPC buffer sizes scale with amount of memory
7644 2002-10-01 Phil Schwan <phil@clusterfs.com>
7647 - locks would be cancelled without throwing away data pages,
7648 resulting in inconsistent data (605627)
7649 - inode attributes were not always being refreshed (605627, 612449)
7650 - lconf now continues to cleanup after lctl reports an error
7651 - MDS now enforces user permissions (602707)
7652 - lprocfs cleanup fixed, but not yet enabled (614157)
7653 - fixed infinite server hang, should a client not respond to an AST
7654 - avoid going into recovery if user calls readlink() with a buffer
7655 that's too small (613941)
7656 - AST RPCs no longer require replies (614867) -- this may be changed
7657 - don't crash server if client sends an IOV that's too big (611336)
7658 - fixed lock conversion deadlock (611892)
7659 - fixed the following of symlinks (614622)
7660 * recovery: the server can remove locks from a client that dies, other
7661 clients can make progress
7662 * more extN patch fixes
7663 * compile-time configurable ptlrpc buffer allocations
7665 - collaborative read cache document
7666 - Lustre Lite Performance CDR document-in-progress
7668 2002-09-20 Andreas Dilger <adilger@clusterfs.com>
7671 - fix typo in patch-2.4.18
7673 2002-09-20 Andreas Dilger <adilger@clusterfs.com>
7676 - clear ptlrpc request each time in handle_incoming_request()
7677 - unlink of files now destroys the object on the OST
7679 2002-09-19 Peter Braam <braam@clusterfs.com>
7681 * add hard link support
7682 * change obdfile creation method
7683 * kernel patch changed
7685 2002-09-19 Peter Braam <braam@clusterfs.com>
7688 - stack overflow bug in extN fixed
7690 2002-09-18 Andreas Dilger <adilger@clusterfs.com>
7692 * documentation updates
7693 - add man pages for config tools
7694 - update tests/README to describe testing with new config tools
7695 - finish metadata API descriptions
7696 * bug fixes and cleanups
7697 - statfs workaround for 16TB limit
7698 - LOV stripe allocation improved, can stripe on subset of OSTs
7699 - LOV file size/IO offset was wrong for files > 4GB in size
7700 - object EA data was being dropped, caused files to be unreadable
7701 - memory overflow with non-LOV OST caused memory corruption
7702 - fixed regression tests to work with new config tools, obdfilter
7703 - fixed bug when directory size became larger than 1 block
7704 - fixed bug (for single client case) when PWD was deleted
7705 - invalidate local directory pages when doing intent-based ops
7706 - avoid LDLM oops when lock callback contained bad data
7708 2002-09-09 Andreas Dilger <adilger@clusterfs.com>
7710 * documentation updates
7711 * bug fixes and cleanups
7712 - configuration tools
7715 - 64-bit compile warnings
7716 - 64-bit internal statfs data
7718 * test_brw on persistent OST devices
7720 * lprocfs (disabled)
7722 2002-09-04 Andreas Dilger <adilger@clusterfs.com>
7724 * documentation updates
7725 * bug fixes and cleanups
7726 * configuration tools
7728 2002-08-30 Peter J. Braam <braam@clusterfs.com>
7731 * many small fixes to 0_5_4
7732 * io/network handling
7733 * thinkos in MDS operations
7735 2002-08-24 Peter J. Braam <braam@clusterfs.com>
7738 * crucial basic fixes to 0.5.3
7739 * IOR, Iozone work over Elan
7742 2002-08-07 Phil Schwan <phil@clusterfs.com>
7743 * version 0_5_3, our first alpha
7744 * we use the new Portals iovs
7745 * documentation updates
7746 * bug fixes and cleanups
7747 * small changes in the DLM wire protocol
7749 2002-07-25 Peter J. Braam <braam@clusterfs.com>
7750 * version 0_5_1 with some initial stability,
7751 * locking on MD and file I/O.
7752 * documentation updates
7753 * several bug fixes since 0.5.0
7754 * small changes in wire protocol
7756 2002-07-18 Phil Schwan <phil@clusterfs.com>
7758 * delivered as Lustre Light Alpha
7759 * fixed a crash after handling invalid MDS requests
7760 * fixed directory pages for architectures with non-4k pages sizes
7762 2002-07-11 Andreas Dilger <adilger@clusterfs.com>
7763 * release version v0_4_4
7764 * Moves TCP acceptor to be on port 2432 (unused Coda port) instead
7766 * Fixes a number of interruption problems with OST operations.
7767 * Update documentation for portals header changes
7768 * Move all wire protocol structs/defines to lustre_idl.h
7769 * Fixes symlink length bug.
7770 * Add tcpdump to repository.
7772 2002-07-05 Andreas Dilger <adilger@clusterfs.com>
7773 * release version v0_4_3
7774 * Fixes statfs for inodes on extN.
7775 * Fixes bug in runtests which would delete /etc/hosts.
7776 * Use 64-bit object IDs wherever possible (not into VFS though)
7777 Remove ost_get_info, which is unused by lustre, and out of date.
7779 2002-07-03 Peter Braam <braam@clusterfs.com>
7780 * release version v0_4_2 Fixes a lookup error (type not passed)
7781 * move forward to head of Portals
7782 * move forward to latest Lustre kernel
7784 2002-06-25 Peter Braam <braam@clusterfs.com>
7785 * release version v0_4_1. Hopefully stable on single node use.