1 xxxx-xx-xx Oracle, Inc.
4 2.6.16.60-0.42.8 (SLES 10),
5 2.6.27.39-0.3.1 (SLES11),
6 2.6.18-194.3.1.el5 (RHEL 5)
7 2.6.18-194.3.1.0.1.el5 (OEL 5)
8 * Client support for unpatched kernels:
9 (see http://wiki.lustre.org/index.php?title=Patchless_Client)
10 2.6.16 - 2.6.30 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.
17 Description: release request ref through ptlrpc_server_drop_request()
18 Details : make sure the request is protected by rq_refcount while
19 being processed in ptlrpc_server_handle_req_in().
23 Description: Handle unsent requests with rq_net_err in ptlrpc_check_set()
27 Description: apps stuck in ptlrpc_check_set() during direct I/O
28 Details : start async bulk unregistering at the same time at reply unlink
30 -------------------------------------------------------------------------------
32 2010-10-29 Oracle, Inc.
34 * Support for kernels:
35 2.6.16.60-0.69.1 (SLES 10),
36 2.6.32.19-0.2.1 (SLES11),
37 2.6.18-194.17.1.el5 (RHEL 5)
38 2.6.18-194.17.1.0.1.el5 (OEL 5)
39 * Client support for unpatched kernels:
40 (see http://wiki.lustre.org/index.php?title=Patchless_Client)
41 2.6.16 - 2.6.32 vanilla (kernel.org)
42 * Recommended e2fsprogs version: 1.41.10-sun2
43 * The async journal commit feature (bug 19128) and the cancel
44 lock before replay feature (bug 16774) are disabled by default.
48 Description: atime is not properly updated on an MDS
50 Severity : enhancement
52 Description: Update to RHEL5.5 kernel 2.6.19-194.17.1.el5.
53 Update to OEL5.5 kernel 2.6.19-194.17.1.0.1.el5.
55 Severity : enhancement
57 Description: Update to SLES10 SP3 kernel 2.6.16.60-0.69.1.
60 Frequency : only with SLES10
62 Description: Use OFED "KMP" provided by Novell
63 Details : SLES10 SP3 ships with OFED in a separate "KMP" package.
64 Lustre is now built against this package. That means you need to
65 install the ofed-kmp package from Novell for the patchless client
66 and from our download site for the server. Note that the ofed-kmp
67 that Novell ships may not exactly match the kernel version but
68 should still be compatible.
70 Severity : enhancement
72 Description: Update SLES11 SP1 kernel to 2.6.32.19-0.2.1.
76 Description: Enabling quotas fails with non-consecutive OST numbering.
80 Description: Fix kernel warning due to lookup_on_len() called without i_mutex
85 Description: Account direct i/o inflight rpcs separately from non-direct i/o so
86 that direct i/o, which is limited by max_rpcs_in_flight, should not
87 block non-direct i/o, which is not limited by max_rpcs_in_flight.
91 Description: Fix per-NID reporting on outstanding writes
95 Description: Reduce stack pressure by uninlining some mds and ptlrpc functions.
99 Description: Remove LASSERT in lprocfs_rd_conn_uuid() since conn == NULL is a
104 Description: fix obdo leak issue in ll_setattr_raw()
108 Description: limit MMP interval
110 Severity : enhancement
112 Description: add several lfs ost enhancements
116 Description: Too many default ACLs break directory access on new directories
120 Description: Lustre inode size is not coherent across nodes.
121 Details : Update lvbo from disk when AST fails with EINVAL. Lvbo is updated
122 on EINVAL error in ldlm_handle_ast_error(). The updates in
123 filter_intent_policy() and ldlm_cb_interpret() have been removed as
128 Description: Oops at __percpu_counter_add+0x1b
129 Details : Use bdi_init()/bdi_destroy() to proper initialize backing_dev_info
134 Description: add mount option to generate 32bit ino, this can be used for 32bit
135 application compatibility.
139 Description: keep reference count for "lli_sai" to prevent it to be released
140 when "statahead_enter()"
144 Description: allow quotacheck over OSTs with sparse indices
148 Description: Objects not getting deleted for files which have been removed
149 Details : ll_have_md_lock() should differentiate between CR and CW OPEN
154 Description: pin object's inode in memory to avoid certain timeouts
158 Description: fix LBUG when obdfilter-survey is interrupted.
160 -------------------------------------------------------------------------------
162 2010-07-31 Oracle, Inc.
164 * Support for kernels:
165 2.6.16.60-0.42.8 (SLES 10),
166 2.6.27.39-0.3.1 (SLES11),
167 2.6.18-194.3.1.el5 (RHEL 5)
168 2.6.18-194.3.1.0.1.el5 (OEL 5)
169 * Client support for unpatched kernels:
170 (see http://wiki.lustre.org/index.php?title=Patchless_Client)
171 2.6.16 - 2.6.30 vanilla (kernel.org)
172 * Recommended e2fsprogs version: 1.41.10-sun2
173 * The async journal commit feature (bug 19128) and the cancel
174 lock before replay feature (bug 16774) are disabled by default.
176 Severity : enhancement
178 Description: Update RHEL5.5 kernel to 2.6.18-194.8.1.el5 and OEL5.5 kernel to
179 2.6.18-194.8.1.0.1.el5.
181 Severity : enhancement
183 Description: using inkernel OFED stack for rhel5 & oel5.
185 Severity : enhancement
187 Description: Add "lfs_migrate" script from manual into lustre/scripts and RPMs
188 Details : lfs_migrate does a "poor man's" migration of files from their
189 current OST layout to a new OST layout as chosen by the MDS.
193 Description: mds_orphan_add_link()) error linking orphan to PENDING
194 Details : quota limits might disallow linking orphans to PENDING
195 when unlinking a file - temporary raise threads' privileges
196 when processing unlinks.
198 Severity : enhancement
200 Description: add conf-param -d option to remove permanent settings.
201 Details : Add the ability to remove permanent lctl conf_param settings.
202 (Previously conf_param settings could only be changed, not
203 removed.) This also provides a method to change failover
204 nid locations. Improve lctl man page.
206 Severity : enhancement
208 Description: add list_param to b1_8 and add "-R" option to list params
211 Severity : enhancement
213 Description: lfs quota output is not very convenient for awk/sed-parsing
214 Details : Some positions in lfs quota output table could be empty or
215 non-empty which made it hard to parse it with scripts, now a dash
216 is put instead of space where there is not supposed to be any data.
218 Severity : enhancement
220 Description: fix obdfilter-survey script to work properly with remote oss-s
222 Severity : enhancement
224 Description: add new OBDFILTER_SURVEY test suite
226 Severity : enhancement
228 Description: add new multiple mount protection (MMP) test suite
230 Severity : enhancement
232 Description: add support for async journal commit in echo client
234 Severity : enhancement
236 Description: allow userland programs to include <lustre/lustre_idl.h>
237 from stardard include directories
239 Severity : enhancement
241 Description: The prune-icache-use-trylock is no longer needed now that
242 the patch from bug 20008 is landed.
246 Description: The shrink grant feature is still active on the client although the
247 connect flag is not set.
251 Description: Don't leak grant space if the write failed with quota exceeded.
255 Description: Don't consume grant space twice on recoverable resent.
259 Description: a race condition could lead to SIGBUS being sent to an
260 application using mmap-ped files from Lustre
261 Details : truncate_complete_page implementation for the patchless
262 client could arbitrarily unset PG_Uptodate flag for a
263 page being kicked from the page cache, an uptodate check
264 right after a readpage call in filemap_fault could fail
265 because of that as though the page read had been unsuccessful.
269 Description: dlm lock slab shrinking is not efficient
270 Details : The dlm_locks slab can grow significantly and consumes a lot of
271 memory on the server. Set a hardlimit to grant_plan.
275 Description: Lustre does not do 1MB IOs to HW RAID
276 Details : Bump MAX_PHYS/HW_SEGMENTS and SG_ALL to 256 in the RHEL5 kernel.
277 This is what we do already for SLES kernels.
281 Description: bump maximum number of phys/hw segments in the SLES11 kernel
282 until s/g chaining works properly.
287 Description: LSI Fusion MPT driver hacks to improve performance
288 Details : Set CONFIG_FUSION_MAX_SGE to 256 for RHEL5
290 Severity : enhancement
292 Description: increase default md stripe_cache_size to 16k
295 Bugzilla : 15587/21439
296 Description: don't handle security.capability xattr
297 Details : CONFIG_SECURITY_FILE_CAPABILITIES is enabled by default on SLES11.
298 This results in additional getxattr calls, causing VBR test
299 failures as well as a preformance drop when writing.
303 Description: obdfilter-survey is no longer working
304 Details : revert patch from bug 20355 to resolve an issue with lctl --threads
305 not working correctly with $(PTHREAD_LIBS) being linked to lctl.
309 Description: ll_shrink_cache does not handle __GFP_FS properly
313 Description: lfs getstripe shows wrong info for directories
314 Details : Set correct LOVEA default values for filesystem-wide.
318 Description: FSX checksum false positves due to mmap IO
319 Details : Use OBD_FL_MMAP flag for IOs on a memory mapped file. Do not print
320 checksum errors, if the flag is set on a request.
324 Description: file operations after eviction have successful return values
325 Details : use vfs ->flush callback to return any pending async errors
330 Description: mdsrate fails to write after 1.3+M files opened
331 Details : decrease memory usage on clients by recycling dentries and
336 Description: obdfilter-survey gives unreasonably high numbers
337 Details : Wait for all threads to complete when running test_brw.
341 Description: do not set lustre read_only device when server umount and keep
342 client records for recoverable ones
346 Description: move sync_on_lock_cancel tunable to the obdfilter layer
347 Details : move the tunable to trigger a journal flush on lock cancel from
348 the ost layer to the obdfilter layer. This tunable is useful
349 when using the async journal commit feature.
353 Description: exp->exp_nid_stats == NULL in filter_tally()
354 Details : fix race with per-nid stats by delaying procfs cleanup until
359 Description: extent lock cancellation on client can keep the cpu busy for too
364 Description: Do not fail OST activation when a llog is not found, just
365 issue an error message.
369 Description: Don't enable extents by default for MDT.
373 Description: Protect bitfield access to ptlrpc_request's rq_flags, since
374 the AT code can access it concurrently while sending early
379 Description: Disable lockless truncate by default since it is sometimes flawed
380 and causes the write_disjoint test to fail.
384 Description: OSSs which don't have the patch from bug 20278 can trigger an
387 Severity : enhancement
389 Description: don't print message to the console when we have not managed to
394 Description: The MDS fails to synchronize OSTs which registered with the MGS
395 after the MDT. The problem is that OBD_NOTIFY_CREATE events are
396 raised too early and thus discarded by the MDT stack.
397 The fix consists of issuing OBD_NOTIFY_CREATE event in the lov
402 Description: Fix race when the ping evictor and a service thread execute
403 target_recovery_check_and_stop() concurrently.
407 Description: quota broadcast can trigger a LBUG on the MDT if there are
412 Description: Resetting the lov_objid values to last_id reported by the OST
413 during orphan recovery is incorrect and can cause the same
414 objects to be allocated twice.
416 Severity : enhancement
418 Description: "weak-modules" support
419 Details : Implement "weak-modules" support which enables kernel modules
420 to be used with any kernel that implements the same kABI. In
421 order to achieve this modules are now installed in
422 /lib/modules/$(uname -r)/updates/kernel on all distributions.
424 Severity : enhancement
426 Description: add writeconf as mount option
428 Severity : enhancement
430 Description: produce debuginfo packages for SLES.
432 Severity : enhancement
434 Description: add failover nidlist to the import proc file.
436 Severity : enhancement
438 Description: fix LUSTRE_SEQ_MAX_WIDTH for interoperability between 1.8
439 clients and 2.0 servers.
441 Severity : enhancement
443 Description: lfs find -s does not work correctly because of a bug in
448 Description: ll_read_ahead_page() must validate the dlm lock before using
453 Description: Prevent failover nids from registering with MGS first.
457 Description: fix lock inversion in ll_setattr_raw().
461 Description: object allocation is not balanced across OSTs.
462 Details : osc_precreate() should return 0, if there are enough objects left.
464 -------------------------------------------------------------------------------
466 2010-04-30 Oracle, Inc.
468 * Support for kernels:
469 2.6.16.60-0.42.8 (SLES 10),
470 2.6.27.39-0.3.1 (SLES11),
471 2.6.18-164.11.1.el5 (RHEL 5)
472 2.6.18-164.11.1.0.1.el5 (OEL 5)
473 * Client support for unpatched kernels:
474 (see http://wiki.lustre.org/index.php?title=Patchless_Client)
475 2.6.16 - 2.6.30 vanilla (kernel.org)
476 * Recommended e2fsprogs version: 1.41.10-sun2
477 * The async journal commit feature (bug 19128) and the cancel
478 lock before replay feature (bug 16774) are disabled by default.
482 Description: open-unlinked directories trigger MDS LBUG
483 Details : Fix regression introduced by the patch from bug 19640.
484 ext3_inc_count() can reset nlink to 1 when the directory
485 is indexed and inode->i_nlink == 2. Work around the problem
486 by incrementing nlink by 2 instead of 1.
490 Description: Reconnects are not throttled
491 Details : Don't wake up pinger on reconnect failures and rely on regular
492 pings to trigger the next reconnection. Please note that the
493 pinger already uses a smaller interval if the import is
497 Frequency : only with NFS export
499 Description: Console flooded with error message from ll_inode_from_lock()
501 Details : in mds_open, initialize the child_res_id before enqueuing
502 the OPEN lock for the child inode, then to avoid senting
503 wrong ldlm_res_id to client.
507 Description: allow multiple instances of the same nid in NID hash
508 Details : Case of multiple separate clients from the same NID (as
509 with liblustre) is legitimate and so we should allow
510 multiple instances of the same NID in nid hash.
514 Description: the readahead code can sleep on a semaphore while holding a
516 Details : in ras_update, "lov_get_info" could be called during increasing
517 readahead windows, which tries to get the mutex lock "lov_lock"
518 while holding the spin_lock "ras_lock", then causes system
523 Description: ASSERTION(cli->cl_avail_grant >= 0) failed
524 Details : fix assertion failure in the grant code.
528 Description: Use CNETERR (which is rate limited) in specific places in
529 the portal's LNET driver to avoid flooding the console.
533 Description: include last created object in precreate slow case
537 Description: don't do rep-ack if not created anything
538 Details : mds_open currently always put a lock into a rep-ack regardless
539 if something was created or not. This is pointless and only
540 creates needless contention. In fact the entire idea was to do
541 this for real creates as a recovery protection.
545 Description: Spurious error messages from smp_processor_id() on preemptible
547 Details : Disable a preemption by grabbing the lock in fs_trace_get_tcd()
548 first. The function fs_trace_get_tcd() was moved up.
552 Description: interval_erase() fix
553 Details : interval_erase() calls update_maxhigh() properly when child
558 Description: "lfs df" does not print stats for all mountpoints
559 Details : Print all mounted lustre filesystems with "lfs df"
563 Description: lfs setstripe -p not longer work with a relative pathname
564 Details : Use realpath() to provide absolute pathname.
568 Description: fix for truncated reply buffer
569 Details : reply buffer could be referred by reply_in_callback after released
574 Description: lustre.lov error when backing up symlinks with extended attributes
575 Details : Improved logic in ll_listxattr()
579 Description: properly handle null value for setattr -n lustre.lov
580 Details : Running "setfattr -n trusted.lov ." causes a NULL dereference
581 in ll_setxattr() due to no checking if "value" is NULL.
582 This command now resets to the default striping when executed
587 Description: stack overflow on lock cancellation due to fsync call
588 Details : sync_on_lock_cancel is needed for recovery when async journal
589 is enabled, but we actually just need to make sure that
590 metadata blocks have hit the journal, so doing a fs sync
591 should be enough and should consume less stack (just create an
592 empty handle and commmit it).
596 Description: using current->journal_info to store per-thread data leads
597 to problem under memory pressure
598 Details : disable the per-thread data (current->journal_info) containing
599 the lock info during I/O to work around the issue for short tem
603 Description: control DCACHE_LUSTRE_INVALID flag with MDS_INODELOCK_LOOKUP lock
604 Details : DCACHE_LUSTRE_INVALID is controlled by MDS_INODELOCK_LOOKUP
605 lock which is corresponding to "IT_LOOKUP", do not skip invalidate
610 Description: Cannot send after transport shutdown
611 Details : Clear imp_vbr_failed flag upon eviction
615 Description: soft lock in request set code during recovery
616 Details : during recovery, uses req->rq_set itself to replay the request
617 instead of ptlrpcd_recovery_pc
621 Description: Use CFS_ALLOC_IO instead of _STD in llap_from_page_with_lockh
622 Details : During an ll_readahead under ll_readpage, we have seen the the
623 OBD_SLAB_ALLOC hang under ldlm_pools_shrink when trying to lock
624 a page that is already locked by the readahead code.
629 Description: stop waitting for next replay transno if shutdown
630 Details : if the system is shutting down, wake up service thread blocked
631 to wait for next replay transno during recovery, then all the
632 references held by queued requests can be dropped and device
637 Description: lov_merge_lvb()) ASSERTION(spin_is_locked(&lsm->lsm_lock)) failed
638 Details : Protect lli->lli_smd pointer updates with lli->lli_lock.
642 Description: per-nid stats should not access lustre-hash internal structures
647 Description: mount.lustre fails to pass some options to mount()
652 Description: ext4 extent allocation is slower than in ext3
653 Details : Increase the default value of MB_DEFAULT_ORDER2_REQS to 8,
654 enlarge ext4 preallocation table for 2048 4K blocks extents
659 Description: incorrect triggering of synchronous IO
660 Details : The OSC can mistakenly fall back to synchronous IO when the
661 max_dirty_mb limit is reached and no write requests have yet
662 been issued. This can occur when the dirty pages are spread
663 over many files all of which are below the optimal request size.
667 Description: Optimize quota_ctl operations by sending requests in parallel
668 Details : Send MDS->OST quota_ctl requests in parallel, do not resend.
669 Compiled from two attachments in the ticket.
673 Description: ordering issue between transaction start & i_mutex
674 Details : start the transaction earlier in llog_lvfs_destroy to get
675 transaction start and inode mutex lock nested properly.
679 Description: lru resize SLV can get stuck
680 Details : calculate SLV with a greater precision to not lose small
681 changes due to interger math truncation; round up SLV only
682 if the amount of granted locks less than the limit to not
683 get stuck with this SLV
687 Description: avoid divide-by-zero in lprocfs_rd_import()
691 Description: lfs quota failed when OSTs are down
692 Details : really return approximate block/inode usage when OSTs are down
696 Description: llobdstat fix and enhancement
697 Details : add a counter to set a limit to how many samples will be returned
698 fix a wildcard in the path to limit to obdfilter stats only
702 Description: recovery-small 51 hang - MDT can not stop, Mount still busy with
704 Details : abort lock enqueue processing sooner when we umount is in progress.
706 Severity : enhancement
708 Description: Update RHEL5.4 kernel to 2.6.18-164.11.1.el5 and OEL5.4 kernel to
709 2.6.18-164.11.1.0.1.el5.
711 Severity : enhancement
713 Description: error message improvements
714 Details : Use INFO/WARN instead of WARN/ERROR for the slow messages.
715 Simplify MDT/OST service start message.
716 Suppress "changing the import ..." warning.
719 Bugzilla : 21961/17914
720 Description: ignore trailing -mdc when determining index number
723 Frequency : only with SLES11
725 Description: fix for a race condition in linux quotas implementation
726 Details : dq_flags(struct dquot) access is not properly locked which could
727 lead to certain inconsistencies when accessing it using non-atomic
728 bit operations like __set_bit in do_set_dqblk.
732 Description: return any pending async errors in close(2) using flush callback
734 -------------------------------------------------------------------------------
736 2010-01-29 Sun Microsystems, Inc.
738 * Support for kernels:
739 2.6.16.60-0.42.8 (SLES 10),
740 2.6.27.39-0.3.1 (SLES11),
741 2.6.18-164.11.1.el5 (RHEL 5)
742 2.6.18-164.6.1.0.1.el5 (OEL 5)
743 * Client support for unpatched kernels:
744 (see http://wiki.lustre.org/index.php?title=Patchless_Client)
745 2.6.16 - 2.6.30 vanilla (kernel.org)
746 * Recommended e2fsprogs version: 1.41.6.sun1
747 * The async journal commit feature (bug 19128) and the cancel
748 lock before replay feature (bug 16774) are disabled by default.
750 Severity : enhancement
752 Description: Update RHEL5.4 kernel to 2.6.18-164.11.1.el5 and
753 OEL5.4 kernel to 2.6.18-164.11.1.0.1.el5.
755 Severity : enhancement
756 Bugzilla : 21511/19848
757 Description: Update SLES11 kernel to 2.6.27.39-0.3.1.
759 Severity : enhancement
761 Description: Update supported SLES10 kernel to 2.6.16.60-0.42.8.
763 Severity : enhancement
765 Description: Update kernel to RHEL5.4 2.6.18-164.6.1.el5 and
766 OEL5 2.6.18-164.6.1.0.1.el5(Both in-kernel OFED enabled).
768 Severity : enhancement
770 Description: Build kernels (RHEL5, OEL5 and SLES10/11) using the vendor's own
773 Severity : enhancement
775 Description: Vanilla kernel 2.6.30 patchless client support.
780 Description: bad entry in directory xxx: inode out of bounds
781 Details : fix locking issue in the rename path which could race with any
782 other operations updating the same directory.
784 Severity : enhancement
786 Description: Make watchdog timer messages to be more clear and descriptive.
790 Description: cp -p command does not preserve the dates and timestamp
791 Details : mtime could be spoiled by a write callback
795 Description: Clear imp_force_reconnect correctly in ptlrpc_connect_interpret()
799 Description: Allow non-root access for "lfs check".
800 Details : Added a check in obd_class_ioctl() for OBD_IOC_PING_TARGET.
802 Severity : enhancement
804 Description: quotacheck performance/scaling issues
805 Details : reduce quotacheck time on empty filesystem by skipping uninit
808 Severity : enhancement
810 Description: Enhancement for lfs(1) command to use numeric uid/gid.
812 Severity : enhancement
814 Description: Adjust locks' extents on their first enqueue, so that at the time
815 they get granted, there is no need for another pass through the
816 queues since they are already shaped into the proper forms.
820 Description: Fix mds_shrink_intent_reply()/mds_intent_policy() to pass correct
821 arguments and prevent LBUG() in lustre_shrink_reply_v2().
825 Description: Change tunefs.lustre and mkfs.lustre --mountfsoptions so that
826 exactly the specified mount options are used. Leaving off
827 any "mandatory" mount options is an error. Leaving off any
828 default mount options causes a warning, but is allowed.
829 Change errors=remount-ro from mandatory to default. Sanitize
830 the mount string before storing it. Update man pages accordingly.
834 Description: mds_getattr() should return 0, even if mds_fid2entry() fails with
835 -ENOENT. Also fix in ptlrpc_expire_one_request() to print signed
838 Severity : enhancement
840 Description: Remove set_info(KEY_UNLINKED) from MDS/OSC
842 Severity : enhancement
844 Description: Clients can replay thousands of unused locks during recovery
845 Details : Don't replay unused locks (only read locks for now) during
846 recovery. This feature is disabled by default and can be
847 enabled by running the following command on the clients:
848 lctl get_param ldlm.cancel_unused_locks_before_replay
852 Description: can't stat file in some situation.
853 Details : improve initialize osc date when target is added to mds and
854 ability to resend too big getattr request is client isn't have info
859 Description: Prevent inconsistences between linux and lustre mount structures.
860 Details : Wait indefinitely in server_wait_finished() until mnt_count drops.
861 Make the sleep interruptible.
863 Severity : enhancement
865 Description: Communicate OST degraded/readonly state via statfs to MDS
866 Details : Flags in the statfs returned from OSTs indicate whether the
867 OST is in a degraded RAID state, or if the filesystem has
868 turned read-only after a filesystem error is detected.
873 Description: don't panic if EPROTO was hit when reading symlink
874 Details : correctly handling request reference in error cases.
879 Description: open sometimes returns ENOENT instead of EACCES
880 Details : checking permission should be part of open part of mds_open, not
881 lookup part. so server should be set DISP_OPEN_OPEN disposition
882 before starting permission check.
883 Also not need revalidate dentry if client already have LOOKUP lock.
887 Description: enable client interface failover
888 Frequency : on servers with multiple network interfaces
889 Details : When a child reconnects from another NID, properly update export
890 nid hash position and ldlm reverse import.
892 Severity : enhancement
894 Description: implemented direct I/O with arbitrary (nonaligned) memory
895 addresses and file offsets.
897 Severity : enhancement
899 Description: added more recovery timeout options.
901 Severity : enhancement
903 Description: added llapi_file_open, llapi_file_create, llapi_file_get_stripe
908 Description: Avoid deadlock for local client writes
909 Frequency : only on systems with clients writing to an OST on the same node
910 Details : Use new OBD_BRW_MEMALLOC flag to notify OST about writes in the
911 memory freeing context. This allows OST threads to set the
912 PF_MEMALLOC flag on task structures in order to allocate memory
913 from reserved pools and complete IO.
914 Use GFP_HIGHUSER for OST allocations for non-local client writes,
915 so that the OST threads generate memory pressure and allow
916 inactive pages to be reclaimed.
921 Descriptoin: lock ordering violation between &cli->cl_sem and _lprocfs_lock
922 Details : move ldlm namespace creation in setup phase to avoid grab
923 _lprocfs_lock with cli_sem held.
927 Frequency : only during format of test systems
928 Description: Unable to run several mkfs.lustre on loop devices at the same time
929 Details : mkfs.lustre returns error 256 on the concurrent loop devices
930 formatting. The solution is to proper handle the error.
932 Severity : enhancement
934 Description: implement async create (obd_async_create) method for osc, to avoid
935 too long waiting new ost objects with holding ldlm lock.
939 Frequency : occasionally during network problems
940 Description: client not allowed to reconnect to OST because of active request
941 Details : abort bulk requests received by the OST once the client has timed
942 out since the client will resend the request anyway. The client
943 also now retries to reconnect to the same server if a connect
944 request failed with EBUSY or -EAGAIN.
947 Frequency : rare, if used wide striped file and one ost in down.
949 Descriptoin: don't return error if we created a subset of objects for file.
950 Details : lov_update_create_set() uses set->set_success as index for created
951 objects, so if some requests failed, they will have hole at end of
952 array and we can use qos_shrink_lsm for allocate correct lsm.
956 Description: Slow stale export processing during normal start up
957 Details : The global mgc lock prevents OST setup to be run in parallel.
958 Replace the global lock with a per-config_llog_data semaphore.
962 Description: Out or order replies might be lost on replay
963 Details : In ptlrpc_retain_replayable_request if we cannot find retained
964 request with tid smaller then one currently being added, add it
965 to the start, not end of the list.
969 Description: BUG: soft lockup - CPU#1 stuck for 10s! [ll_mdt_07:4523]
970 Details : add cond_resched() calls to avoid hogging the cpu for too long
971 in the hash code. Make also lustre_hash_for_each_empty() more
974 Severity : enhancement
976 Description: Performance improvements for debug messages with D_RPCTRACE,
977 D_LDLM, D_QUOTA options.
980 Frequency : only with NFS export
982 Description: (lov_merge.c:74:lov_merge_lvb())
983 ASSERTION(spin_is_locked(&lsm->lsm_lock)) failed (SR 71691004)
984 Details : Fix a race in the nfs export code by populating inode
985 info while the new inode is still locked
987 Severity : enhancement
989 Description: add a new file in procfs called force_lbug. Writting to this
990 file triggers a LBUG. Only for test purpose.
994 Description: OOM killer causes node hang
995 Details : really interrupt the sleep in osc_enter_cache on signals
999 Description: LustreError: 9153:0:(quota_context.c:622:dqacq_completion()) LBUG
1000 Details : fix race during quota release on the slave.
1002 Severity : enhancement
1004 Description: smaller hash bucket sizes, cleanups
1005 Details : increase hash table sizes and enabled rehashing for pools, quota,
1006 uuid, nid & per-nid stats.
1008 Severity : enhancement
1010 Description: Add ldiskfs maxdirsize mount option
1011 Details : add max_dir size mount option
1015 Description: panic in ll_statahead_thread
1016 Details : prevent parent thread to be killed before its child
1019 Frequency : only with 16TB device
1021 Description: unable to perform "mount -t lustre" of 16TB OST device
1022 Details : Mounting 16TB LUNs failed due to three bugs in mkfs.lustre.
1026 Description: ASSERTION(atomic_read(&imp->imp_inflight) == 0) failed
1027 Details : unregistering should be zero if no RPC inflight.
1031 Description: hyperion: Oops during metabench
1032 Details : Correct the refcount of lov_request_set
1034 Severity : enhancement
1036 Description: Add mptlinux and nxge drivers to Lustre builds
1038 Severity : enhancement
1040 Description: Fix watchdog timer message to be more clear
1041 Details : Make watchdog timer messages more clear and descriptive.
1045 Description: LNET soft lockups in socknal_cd thread
1046 Details : don't hog CPU for active-connecting if another connd is
1047 accepting connecting-requst from the same peer
1051 Description: recovery-small test_17 hang
1052 Details : Land several AT improvements & fixes.
1056 Description: MDS panic and hanging client processes
1057 Details : Replace exp_ops_stats with exp_nid_stats->nid_stats
1061 Description: OSS stuck in recovery.
1062 Details : fix race during recovery. class_unlink_export,
1063 class_set_export_delayed and target_queue_last_replay_reply
1064 may race while increasing/decreasing obd_recoverable_clients
1065 and obd_delayed_clients, causing recovery to wait forever.
1067 Severity : enhancement
1069 Description: add cascading_rw.c to lustre/tests
1073 Description: filter_last_id() NULL deref
1074 Details : lprocfs_filter_rd_last_id() should check for the fully
1075 setup obd device, before proceeding further.
1077 Severity : enhancement
1079 Description: Loadgen improvements
1080 Details : stacksize and locking fixes for loadgen
1084 Description: Quiet CERROR("dirty %d > system dirty_max %d\n"
1085 Details : The atomic_read() allowing the atomic_inc() are not covered
1086 by a lock. Thus they may safely race and trip this CERROR()
1087 unless we add in a small fudge factor (+1).
1089 Severity : enhancement
1091 Description: shrink_slab: nr=-9223362083340912175
1092 Details : fix spurious message from shrink_slab reporing negative nr
1096 Description: Quiet bogus previously committed transno error
1097 Details : suppress the "server went back in time" error message which
1098 is always printed even in the common case after a client eviction
1100 Severity : enhancement
1102 Description: Parallel statfs() calls result in client eviction
1103 Details : cache statfs data for 1s.
1107 Description: parallel-scale test_compilebench: @@@@@@ FAIL: compilebench
1109 Details : fix serveral issues in pinger code causing clients not to ping
1110 servers for too long, resulting in evictions.
1114 Description: e2fsck should warn when MMP update interval is extended
1115 Details : print mmp_check_interval and make it possible to abort mount
1116 operation in case it takes too long.
1120 Description: mdsrate-create-large.sh, BUG: soft lockup - CPU#0 stuck for 10s!
1121 Details : fix bug in the RHEL5's jbd2 callback patch.
1125 Description: drop number of active requests when queued for recovery
1126 Details : Now that we take a reference on the original request instead of
1127 making a copy of it for recovery. We need to drop the number of
1128 active requests or the queued requests will prevent all request
1129 processing when they exceed (srv->srv_threads_running - 1).
1131 Severity : enhancement
1133 Description: refuse to invalidate operational quota files when they are in use
1134 Details : an attempt to invalidate operational quota files on the quota
1135 master is not actually permitted by VFS (returning -EPERM), but we
1136 should not depend on that and should return the error earlier.
1140 Description: Applications stuck in jbd2_log_wait_commit during exit
1141 Details : fix deadlock between kjournald2 trying to acquire the page lock
1142 owned by an ost_io thread waiting for journal commit.
1144 -------------------------------------------------------------------------------
1146 2009-10-16 Sun Microsystems, Inc.
1148 * Support for kernels:
1149 2.6.16.60-0.42.4 (SLES 10),
1150 2.6.27.29-0.1 (SLES11, i686 & x84_64 only),
1151 2.6.18-128.7.1.el5 (RHEL 5),
1152 2.6.18-128.7.1.el5 (OEL 5),
1153 * Client support for unpatched kernels:
1154 (see http://wiki.lustre.org/index.php?title=Patchless_Client)
1155 2.6.16 - 2.6.27 vanilla (kernel.org)
1156 * Recommended e2fsprogs version: 1.41.6.sun1
1157 * ext4 support for RHEL5 is experimental and thus should not be
1160 Severity : enhancement
1162 Description: Add OEL5 support.
1164 Severity : enhancement
1166 Description: Update kernel to SLES11 2.6.27.29-0.1.
1170 Description: File checksum failures with OST read cache on
1171 Details : Disable page poisoning when the bulk transfer has to be aborted
1172 because the client got evicted.
1176 Description: Don't allow make backward step on assiging osc next id.
1177 Details : race between allocation next id and ll_sync thread can be cause
1178 of set wrong osc next id and can be kill valid ost objects.
1180 Severity : enhancement
1182 Description: Update kernel to RHEL5 2.6.18-128.7.1.el5.
1184 Severity : enhancement
1186 Description: Update kernel to SLES10 SP2 2.6.16.60-0.42.4.
1190 Description: Changes in raid5-large-io-rhel5.patch to calculate sectors properly
1194 Description: Increase the default BLK_DEF_MAX_SECTORS value for RHEL5 and SLES11
1198 Description: Error handling in osc_statfs_interpret() has been improved.
1199 Details : Check in osc_statfs_interpret() for EBADR.
1203 Description: Do not send statfs() requests to OSTs disabled by administrator.
1204 Details : Check in lov_prep_statfs_set() for non-NULL ltd_exp.
1208 Description: Do not update ctime for the deleted inode.
1209 Details : Check in mds_reint_unlink() before calling fsfilt_setattr().
1213 Description: Increase of the size of the LDLM resource hash.
1214 Details : Bump up RES_HASH_BITS=12.
1218 Description: correctly send lsm on open replay
1219 Details : MDS is trust to LSM size on replay open, but client can set wrong size
1224 Description: Deadlock between filter_destroy() and filter_commitrw_write().
1225 Details : filter_destroy() does not hold the DLM lock over the whole
1226 operation. If the DLM lock is dropped, filter_commitrw() can go
1227 through, causing the deadlock between page lock and i_mutex.
1228 The i_alloc_sem should also be hold in filter_destroy() while
1229 truncating the file.
1233 Description: truncate starts GFP_FS allocation under transaction causing deadlock
1234 Details : ldiskfs_truncate calls grab_cache_page which may start page
1235 allocation under an open transaction. This may lead to
1236 calling prune_icache with consequent lustre reentrance.
1239 Frequency : only when down/upgrading the MDS to 1.6/1.8 while 1.8 clients are
1240 still up and when the OST pool feature is used
1242 Description: interop testing got LBUG when run dd with OST pool
1243 :LustreError: 30032:0:(llite_lib.c:1913:ll_replace_lsm()) LBUG
1244 Details : down/upgrading the MDS to a version that doesn't/does support OST
1245 pool can cause clients to crash because the lsm has changed
1250 Description: missing tree_status on 1.8.1 RPM build
1251 Details : make rpms failed due because the tree_status file is missing.
1255 Description: continuing LustreError "mds adjust qunit failed!"
1256 Details : don't print message on the console when ->adjust_qunit fails.
1260 Description: don't increase ldlm timeout if previous client was evicted
1261 Details : if a client doesn't respond to a blocking callback within the
1262 adaptive ldlm enqueue timeout, don't adjust the adaptive estimate
1263 when the lock is next granted.
1267 Description: ost is being unmounted w/o all writes to last_rcvd landing on disk.
1268 affects recovery negatively.
1269 Details : make sure all exports have been properly destroyed by the zombie
1270 thread processed before stopping the target.
1274 Description: Performance degradation with O_DIRECT between 1.6 & 1.8.1 b190
1275 Details : disable write barrier for ext4/SLES11.
1279 Description: Kernel panic - not syncing: Out of memory and no killable
1280 processes... on OSS when iozone
1281 Details : fix memory leak in the journal checksum patch.
1285 Description: group quota "too many blocks" OSS crashes
1286 Details : we should keep the same uid/gid for lquota_chkquota() and
1287 lquota_pending_commit()
1291 Description: LustreError: 9153:0:(quota_context.c:622:dqacq_completion()) LBUG
1292 Details : don't LBUG on release quota error. Just a workaround until the
1293 problem is understood.
1295 ------------------------------------------------------------------------------
1296 2009-07-31 Sun Microsystems, Inc.
1298 * Support for kernels:
1299 2.6.16.60-0.39.3 (SLES 10),
1300 2.6.27.23-0.1 (SLES11, i686 & x84_64 only),
1301 2.6.18-128.1.14.el5 (RHEL 5),
1302 * Client support for unpatched kernels:
1303 (see http://wiki.lustre.org/index.php?title=Patchless_Client)
1304 2.6.16 - 2.6.27 vanilla (kernel.org)
1305 * Recommended e2fsprogs version: 1.41.6.sun1
1306 * File join has been disabled in this release, refer to Bugzilla 16929.
1307 * NFS export disabled when stack size < 8192. Since the NFSv4 export
1308 of Lustre filesystem with 4K stack may cause a stack overflow. For
1309 more information, please refer to bugzilla 17630.
1310 * ext4 support for RHEL5 is experimental and thus should not be
1313 Severity : enhancement
1315 Description: Update kernel to SLES10 SP2 2.6.16.60-0.39.3.
1318 Frequency : with 1.8 server and 1.6 clients
1320 Description: correctly shrink reply for avoid send too big message to client.
1321 Details : 1.8 mds is allocate to big buffer to LOV EA data and this produce
1322 some problems with sending this reply to 1.6 client.
1326 Description: Repeated atomic allocation failures.
1327 Details : Use GFP_HIGHUSER | __GFP_NOMEMALLOC flags for memory allocations
1328 to generate memory pressure and allow reclaiming of inactive pages.
1329 At the same time, do not allow to exhaust emergency pools.
1330 For local clients the use of GFP_NOFS will be introduced in 1.8.2
1332 Severity : enhancement
1333 Bugzilla : 19846, 18289
1334 Description: Update kernel to RHEL5 2.6.18-128.1.14.el5.
1336 Severity : enhancement
1337 Bugzilla : 19625, 16893, 18668, 19848
1338 Description: Add support for SLES11 2.6.27.23-0.1.
1340 Severity : enhancement
1342 Description: Update client support to vanila kernels up to 2.6.27.
1344 Severity : enhancement
1346 Description: Update kernel to SLES10 SP2 2.6.16.60-0.37.
1348 Severity : enhancement
1350 Description: Compile with -Werror by default for i686 and x86_64.
1354 Description: resolve race between obd_disconnect and class_disconnect_exports
1355 Details : if obd_disconnect will be called to already disconnected export he
1356 forget release one reference and osc module can't unloaded.
1358 Severity : enhancement
1360 Description: move AT tunable parameters for more consistent usage
1361 Details : add AT tunables under /proc/sys/lustre, add to conf_param parsing
1365 Descriptoin: correctly skip time estimate if in recovery
1366 Details : rq_send_state insn't bitmask so using bitwise ops is forbid.
1370 Descriptoin: OSS DeadLock
1371 Details : Use trylock to prevent deadlock when shrink icache.
1373 Severity : enhancement
1375 Description: Allow tuning service thread via /proc
1376 Details : For each service a new
1377 /proc/fs/lustre/{service}/*/thread_{min,max,started} entry is
1378 created that can be used to set min/max thread counts, and get the
1379 current number of running threads.
1381 Severity : enhancement
1383 Description: Add state history info file, enhance import info file
1384 Details : Track import connection state changes in a new osc/mdc proc file;
1385 add overview-type data to the osc/mdc import proc file.
1389 Description: Reduce small size read RPC
1390 Details : Set read-ahead limite for every file and only do read-ahead when
1391 available read-ahead pages are bigger than 1M to avoid small size
1396 Description: free_entry erroneously used groups_free instead of put_group_info
1398 Severity : enhancement
1400 Description: Make read-ahead stripe size aligned.
1402 Severity : enhancement
1404 Description: MDS create should not wait for statfs RPC while holding DLM lock.
1407 Frequency : rare, connect and disconnect target at same time
1409 Descriptoin: ASSERTION(atomic_read(&imp->imp_inflight) == 0
1410 Details : don't call obd_disconnect under lov_lock. this long time
1411 operation and can block ptlrpcd which answer to connect request.
1414 Frequency : start MDS on uncleanly shutdowned MDS device
1416 Descriptoin: ll_sync thread stay in waiting mds<>ost recovery finished
1417 Details : stay in waiting mds<>ost recovery finished produce random bugs
1418 due race between two ll_sync thread for one lov target. send
1419 ACTIVATE event only if connect realy finished and import have
1423 Frequency : start MDS on uncleanly shutdowned MDS device
1425 Descriptoin: aborting recovery hang on MDS
1426 Details : don't throttle destroy RPCs for the MDT.
1430 Description: Slow reads beyond 8Tb offsets.
1431 Details : Page index integer overflow in ll_read_ahead_page
1435 Description: Soft lockup on OSS after MDS failover
1436 Details : MSG_CONNECT_INITIAL is not set on the initial MDS->OST connect.
1437 As a conseqence, the patch from bug 18224 is not operational
1438 and the MDS export cannot be reused on the OSTs until it gets
1442 Frequency : rare, only if using MMP with Linux RAID
1444 Description: MMP doesn't work with Linux RAID
1445 Details : While using HA for Lustre servers with Linux RAID, it is possible
1446 that MMP will not detect multiple mounts. To make this work we
1447 need to unplug the device queue in RAID when the MMP block is being
1448 written. Also while reading the MMP block, we should read it from
1449 disk and not the cached one.
1452 Frequency : rare, during recovery
1454 Description: Assertion failure in ldlm_lock_put
1455 Details : Do not put cancelled locks into replay list, hold references on
1456 locks in replay list
1460 Description: 1.6.5 mdsrate performance is slower than 1.4.11/12 (MDS is not cpu bound!)
1461 Details : create_count always drops to the min value (=32) because grow_count
1462 is being changed before the precreate RPC completes.
1465 Frequency : Only in RHEL5 when mounting multiple ext3 filesystems
1468 Description: "kmem_cache_create: duplicate cache jbd_4k" error message
1469 Details : add proper locking for creation of jbd_4k slab cache
1473 Description: MMP check in ext3_remount() fails without displaying any error
1474 Details : When multiple mount protection fails during remount, proper error
1479 Description: Rare Client crash on resend if the file was deleted.
1480 Details : When file is opened, but open reply is lost and file is
1481 subsequently deleted before resend, resend processing logic
1482 breaks trying to open the file again, should not try to open.
1486 Description: add check for >8TB ldiskfs filesystems
1487 Details : ext3-based ldiskfs does not support greater than 8TB LUNs.
1488 Don't allow >8TB ldiskfs filesystems to be mounted without
1489 force_over_8tb mount option
1493 Description: Client locked up when running multiple instances of an app. on
1494 multiple mount points
1495 Details : ll_shrink_cache() can sleep while holding the ll_sb_lock.
1496 Convert ll_sb_lock to a read/write semaphore to fix the problem.
1500 Description: Cannot acces an NFS-mounted Lustre filesystem
1501 Details : An NFS client cannot access the Lustre filesystem NFS-mounted
1502 from a Lustre-client exporting the Lustre filesystem via NFS.
1506 Description: panic in ll_statahead_thread
1507 Details : grab dentry reference in parent process.
1510 -------------------------------------------------------------------------------
1512 tbd Sun Microsystems, Inc.
1514 * Support for kernels:
1515 2.6.16.60-0.31 (SLES 10),
1516 2.6.18-128.1.6.el5 (RHEL 5),
1517 2.6.22.14 vanilla (kernel.org)
1518 * Client support for unpatched kernels:
1519 (see http://wiki.lustre.org/index.php?title=Patchless_Client)
1520 2.6.16 - 2.6.22 vanilla (kernel.org)
1521 * Recommended e2fsprogs version: 1.40.11-sun1
1522 * File join has been disabled, refer to Bugzilla 16929.
1523 * A new Lustre ADIO driver is available for MPICH2-1.0.7.
1524 * NFS export disabled when stack size < 8192. Since the NFSv4 export of
1525 Lustre filesystem with 4K stack may cause a stack overflow. For more
1526 information, please refer to bugzilla 17630.
1528 Severity : enhancement
1530 Description: Update to RHEL5.3 kernel-2.6.18-128.1.6.el5.
1532 Severity : enhancement
1534 Description: Update OFED release to 1.4.1 RC4
1536 Severity : major, only with big OST
1538 Description: Very poor metadata performance on Infiniband lustre configuration
1539 Details : OST object precreation becomes very slow on big OSTs. This is due
1540 to the ialloc patch spending too much time scanning groups.
1543 Frequency : during recovery
1545 Description: don't mix llog inodes with normal.
1546 Details : allocate inodes for log in last inode group
1551 Description: fix lqs' reference which won't be put in some situations
1552 Details : This patch fixes:
1553 1. In quota_check_common(), this function will check quota
1554 for user and group, but only send one return via "pending".
1555 In most cases, the pendings should be same. But that is not
1557 2. If quotaoff runs between lquota_chkquota() and
1558 lquota_pending_commit(), the same thing will happen too.
1559 That is why it comes:
1560 - if (!ll_sb_any_quota_active(qctxt->lqc_sb))
1563 -------------------------------------------------------------------------------
1565 2008-12-31 Sun Microsystems, Inc.
1567 * Support for kernels:
1568 2.6.5-7.314 (SLES 9),
1569 2.6.9-67.0.22.EL (RHEL 4),
1570 2.6.16.60-0.31 (SLES 10),
1571 2.6.18-92.1.17.el5 (RHEL 5),
1572 2.6.22.14 vanilla (kernel.org)
1573 * Client support for unpatched kernels:
1574 (see http://wiki.lustre.org/index.php?title=Patchless_Client)
1575 2.6.16 - 2.6.22 vanilla (kernel.org)
1576 * Client support for unpatched kernels:
1577 we do not recommend using patchless RHEL4 clients with kernels
1578 prior to 2.6.9-55EL (RHEL4U5).
1579 * Recommended e2fsprogs version: 1.40.11-sun1
1580 * Note that reiserfs quotas are disabled on SLES 10 in this kernel.
1581 * RHEL 4 and RHEL 5/SLES 10 clients behaves differently on 'cd' to a
1582 removed cwd "./" (refer to Bugzilla 14399).
1583 * A new quota file format has been introduced in 1.6.5.
1584 The format conversion from prior releases is handled transparently,
1585 but releases older than 1.4.12/1.6.5 don't understand this new
1586 format. The automatic format conversion can be avoided by running
1587 the following command on the MDS:
1588 'tunefs.lustre --param="mdt.quota_type=ug1" $MDTDEV'.
1589 For more information, please refer to bugzilla 13904.
1590 * A new quota file format was introduced in 1.6.6/1.8.0 (kernels 2.6.16+).
1591 The format conversion from prior releases is handled transparently,
1592 but releases older than 1.6.6/1.8.0 don't understand this new
1593 format. The automatic format conversion can be avoided by running
1594 the following commands on the MDS and OSS servers (for
1595 pre 1.4.12-1.6.5 quota files):
1596 'tunefs.lustre --param="mdt.quota_type=ug1" $MDTDEV',
1597 'tunefs.lustre --param="ost.quota_type=ug1" $MDTDEV'
1598 or (for 1.4.12/1.6.5 quota files)
1599 'tunefs.lustre --param="mdt.quota_type=ug2" $MDTDEV',
1600 'tunefs.lustre --param="ost.quota_type=ug2" $MDTDEV'
1601 For more information, please refer to bugzilla 13904.
1602 * Output of lfs quota has been made less detailed by default,
1603 old (verbose) output can be obtained by using -v option.
1604 * File join has been disabled in this release, refer to Bugzilla 16929.
1605 * A new Lustre ADIO driver is available for MPICH2-1.0.7.
1606 * NFS export disabled when stack size < 8192. Since the NFSv4 export of
1607 Lustre filesystem with 4K stack may cause a stack overflow. For more
1608 information, please refer to bugzilla 17630.
1610 Severity : enhancement
1612 Description: Caching OSS
1613 Details : introduce data caching on the OSS. The OSS now relies on the linux
1614 kernel page cache to keep recently accessed data in memory.
1615 It is worth noting that all write requests are still flushed
1616 synchronously as in lustre 1.6.
1618 Severity : enhancement
1620 Description: Version based recovery
1621 Details : introduce finer grained recovery able to detect transaction
1622 dependencies and can deal with transaction gaps caused by clients
1623 failing at the same time as the server.
1625 Severity : enhancement
1627 Description: Enable adaptive timeouts by default
1628 Details : The Lustre timeout value in /proc/sys/lustre/timeout is now
1629 managed dynamically based on server load and should not need
1630 to be tuned manually based on cluster size. This allows Lustre
1631 to work under a wider variety of system sizes and loads, without
1632 unnecessarily causing lengthy recovery times.
1634 Severity : enhancement
1636 Description: Add OST Pools support
1637 Details : File striping can now be set to use an arbitrary pool of OSTs
1639 Severity : enhancement
1641 Description: add lazystatfs mount option to allow statfs(2) to skip down OSTs
1642 Details : allow skip disconnected ost for send statfs request and hide error
1646 Frequency : rare, on llog test 6
1648 Descriptoin: don't allow connect to already connected import
1649 Details : allowing connect to already connected import is hide connecting problem.
1652 Frequency : rare, on failed llog setup
1654 Descriptoin: don't leak obd reference on failed llog setup
1655 Details : for failed llog setup - mgc forget call class_destroy_import for
1656 client import, move destroy import to more generic place.
1661 Descriptoin: allow kill process which wait statahead result
1662 Details : for some reasons 'ls' can stick in waiting result from statahead,
1663 in this case need way for kill this process.
1666 Frequency : rare, at shutdown
1668 Descriptoin: panic at umount
1669 Details : llap_shrinker can be raced with killing super block from list and
1670 this produce panic with access to already freeded pointer
1675 Descriptoin: don't lose wakeup for imp_recovery_waitq
1676 Details : recover_import_no_retry or invalidate_import and import_close can
1677 both sleep on imp_recovery_waitq, but we was send only one wakeup
1683 Descriptoin: panic in mds_open
1684 Details : don't confuse mds_finish_transno() with PTR_ERR(-ENOENT)
1689 Descriptoin: stuck in cache_remove_extent() or panic with accessing to already
1691 Details : release lock refernce only after add page to pages list.
1694 Frequency : always with long access acl
1696 Descriptoin: mds can't pack reply with long acl.
1697 Details : mds don't control size of acl but they limited by reint/getattr
1702 Frequency : on remount
1704 Description: external journal device not working after the remount
1705 Details : clear dev_rdonly flag for external journal devices in
1711 Description: shutdown vs evict race
1712 Details : client_disconnect_export vs connect request race.
1713 if client will evicted at this time - we start invalidate
1714 thread without referece to import and import can be freed
1720 Description: shrink LOV EAs before replying
1721 Details : correctly adjust LOV EA buffer for reply.
1726 Description: don't skip ost target if they assigned to file
1727 Details : Drop slow OSCs if we can, but not for requested start idx.
1728 This means "if OSC is slow and it is not the requested
1729 start OST, then it can be skipped, otherwise skip it only
1730 if it is inactive/recovering/out-of-space.
1732 Severity : enhancement
1734 Description: Update to RHEL5 kernel-2.6.18-92.1.17.el5.
1737 Frequency : rare, need acl's on inode.
1739 Description: client can't handle ost additional correctly
1740 Details : if ost was added after client connected to mds client can have
1741 hit lnet_try_match_md ... to big messages to wide striped files.
1742 in this case need teach client to handle config events about add
1743 lov target and update client max ea size at that event.
1745 Severity : enhancement
1747 Description: Update to sles9 kernel-2.6.5-7.314.
1749 Severity : enhancement
1751 Description: Update to SLES10 SP2 kernel-2.6.16.60-0.31.
1754 Frequency : Create a symlink file with a very long name
1756 Description: ldlm_cancel_pack()) ASSERTION(max >= dlm->lock_count + count)
1757 Details : If there is no extra space in the request for early cancels,
1758 ldlm_req_handles_avail() returns 0 instead of a negative value.
1763 Description: mds is deadlocked
1764 Details : in rare cases, inode in catalog can have i_no less than have parent
1765 i_no, this produce wrong order for locking during open, and
1766 parallel unlink can be lock open. this need teach mds_open to grab
1767 locks in resource id order, not at parent -> child order.
1769 Severity : enhancement
1771 Description: Add /proc entry for import status
1772 Details : The mdc, osc, and mgc import directories now have
1773 an import directory that contains useful import data for debugging
1774 connection problems.
1776 Severity : enhancement
1778 Description: Re-disable certain /proc logging
1779 Details : Enable and disable client's offset_stats, extents_stats and
1780 extents_stats_per_process stats logging on the fly.
1783 Frequency : Only on FC kernels 2.6.22+
1785 Description: oops in statahead
1786 Details : Do not drop reference count for the dentry from VFS when lookup,
1787 VFS will do that by itself.
1789 Severity : enhancement
1791 Description: Generic /proc file permissions
1792 Details : Set /Proc file permissions in a more generic way to enable non-
1793 root users operate on some /proc files.
1797 Description: Hitting mdc_commit_close() ASSERTION
1798 Details : Properly handle request reference release in
1799 ll_release_openhandle().
1803 Frequency : only patchless client
1804 Description: add workaround for race between add/remove dentry from hash
1806 Severity : enhancement
1808 Description: Allow OST glimpses to return PW locks
1812 Description: LBUG when llog conf file is full
1813 Details : When llog bitmap is full, ENOSPC should be returned for plain
1818 Description: Prevent import from entering FULL state when server in recovery
1822 Description: service mount cannot take device name with ":"
1823 Details : Only when device name contains ":/" will mount treat it as
1829 Description: replace ptlrpcd with the statahead thread to interpret the async
1830 statahead RPC callback
1834 Frequency : on recovery
1835 Description: I/O failures after umount during fail back
1836 Details : if client reconnected to restarted server we need join to recovery
1837 instead of find server handler is changed and process self eviction
1838 with cancel all locks.
1840 Severity : enhancement
1842 Description: Update to RHEL5 kernel-2.6.18-92.1.10.el5.
1846 Description: Kernel BUG tries to release flock
1847 Details : Lustre does not destroy flock lock before last reference goes
1848 away. So always drop flock locks when client is evicted and
1849 perform unlock regardless of successfulness of speaking to MDS.
1851 Severity : enhancement
1853 Description: Update to SLES10 SP2 kernel-2.6.16.60-0.27.
1855 Severity : enhancement
1857 Description: Upcall on Lustre log has been dumped
1858 Details : Allow for a user mode script to be called once a Lustre log has
1859 been dumped. It passes the filename of the dumped log to the
1860 script, the location of the script can be specified via
1861 /proc/sys/lnet/debug_log_upcall.
1866 Description: avoid messages about idr_remove called for id that is not allocated
1867 Details : Move assigment s_dev for clustered nfs to end of initialization,
1868 for avoid problem with error handling.
1873 Description: avoid Already found the key in hash [CONN_UNUSED_HASH] messages
1874 Details : When connection is reused this not moved from CONN_UNUSED_HASH into
1875 CONN_USED_HASH and this prodice warning when put connection again
1881 Description: avoid ASSERTION(client_stat->nid_exp_ref_count == 0) failed
1882 Details : release reference to stats when client disconnected, not
1883 when export destroyed for avoid races when client destroyed
1884 after main ost export.
1888 Description: more cleanup in mds_lov
1889 Details : add workaround for get valid ost count for avoid warnings about
1890 drop too big messages, not init llog cat under semphore which
1891 can be blocked on reconnect and break normal replay, fix access
1894 Severity : enhancement
1896 Description: Export bytes_read/bytes_write count on OSC/OST.
1900 Description: Early reply size mismatch, MGC loses connection
1901 Details : Apply the MGS_CONNECT_SUPPORTED mask at reconnect time so
1902 the connect flags are properly negotiated.
1906 Description: Properly propagate oinfo flags from lov to osc for statfs
1907 Details : restore missing copy oi_flags to lov requests.
1911 Description: exports in /proc are broken
1912 Details : recreate /proc entries for clients when they reconnect.
1914 Severity : enhancement
1916 Description: Add man pages for llobdstat(8), llstat(8), plot-llstat(8),
1917 : l_getgroups(8), lst(8), routerstat(8)
1918 Details : included man pages for llobdstat(8), llstat(8),
1919 : plot-llstat(8), l_getgroups(8), lst(8), routerstat(8)
1921 Severity : enhancement
1923 Description: Implement lustre ll_show_options method.
1925 Severity : enhancement
1927 Description: Update to SLES9 kernel-2.6.5-7.312.
1929 Severity : enhancement
1931 Description: Update to RHEL4 kernel-2.6.9-67.0.22.EL.
1935 Description: exports in /proc are broken
1936 Details : recreate /proc entries for clients when they reconnect.
1940 Description: don't fail open with -ERANGE
1941 Details : if client connected until mds will be know about real ost count
1942 get LOV EA can be fail because mds not allocate enougth buffer
1947 Description: Resolve device initialization race
1948 Details : Prevent proc handler from accessing devices added to the
1949 obd_devs array but yet be intialized.
1951 Severity : enhancement
1953 Description: configure's --enable-quota should check the
1954 : kernel .config for CONFIG_QUOTA
1955 Details : configure is terminated if --enable-quota is passed but
1956 : no quota support is in kernel
1958 Severity : enhancement
1960 Description: Update to SLES10 SP2 kernel-2.6.16.60-0.23.
1962 Severity : enhancement
1964 Description: Update to RHEL5 kernel-2.6.18-92.1.6.el5.
1968 Frequency : rare, on PPC clients
1969 Description: don't swab ost objects in response about directory, because
1971 Details : bug similar bug 14856, but in different function.
1973 Severity : enhancement
1975 Description: lfs quota tool enhancement
1976 Details : added units specifiers support for setquota, default to
1977 current uid/gid for quota report, short quota stats by
1978 default, nonpositional parameters for setquota, added
1979 llapi_quotactl manual page.
1981 Severity : enhancement
1983 Description: *optional* service tags registration
1984 Details : if the "service tags" package is installed on a Lustre node
1985 When the filesystem is mounted, a local-node service tag will
1986 be created. See http://inventory.sun.com/ for more information
1987 about the Service Tags asset management system.
1991 Description: Client runs out of low memory
1992 Details : Consider only lowmem when counting initial number of llap pages
1995 Frequency : occasional
1997 Description: add refcount for osc callbacks, so avoid panic on shutdown
1999 Severity : enhancement
2001 Description: Update to RHEL4 kernel-2.6.9-67.0.20.
2004 Frequency : testing only
2006 Description: sanity test 65a fails if stripecount of -1 is set
2007 Details : handle -1 striping on filesystem in ll_dirstripe_verify
2010 Frequency : only in unusual configurations
2012 Description: Kernel panic with find ost index.
2013 Details : lov_obd have panic if some OST's have sparse indexes.
2015 Severity : enhancement
2017 Description: Update to RHEL5 kernel-2.6.18-53.1.21.el5.
2020 Frequency : rarely, if filesystem is mounted with -o flock
2022 Description: do not process already freed flock
2023 Details : flock can possibly be freed by another thread before it reaches
2024 to ldlm_flock_completion_ast.
2027 Frequency : rarely, if filesystem is mounted with -o flock
2029 Description: LBUG during stress test
2030 Details : Need properly lock accesses the flock deadlock detection list.
2033 Frequency : rarely, if binaries are being run from Lustre
2035 Description: oops in page fault handler
2036 Details : kernel page fault handler can return two special 'pages' in
2037 error case, don't try dereference NOPAGE_SIGBUS and NOPAGE_OMM.
2040 Frequency : rarely, during shutdown
2042 Description: timeout with invalidate import.
2043 Details : ptlrpcd_check call obd_zombie_impexp_cull and wait request which
2044 should be handled by ptlrpcd. This produce long age waiting and
2045 -ETIMEOUT ptlrpc_invalidate_import and as result LASSERT.
2051 Description: ASSERTION(CheckWriteback(page,cmd)) failed
2052 Details : badly clear PG_Writeback bit in ll_ap_completion can produce false
2056 Frequency : only with broken builds/installations
2058 Description: no LBUG if lquota.ko and fsfilt_ldiskfs.ko are different versions
2059 Details : just return an error to a user, put a console error message
2061 Severity : enhancement
2063 Description: Update to RHEL5 kernel-2.6.18-53.1.19.el5.
2065 Severity : enhancement
2067 Description: Update to RHEL4 kernel-2.6.9-67.0.15.
2069 Severity : enhancement
2071 Description: enable MGS and MDT services start separately
2072 Details : add a 'nomgs' option in mount.lustre to enable start a MDT with
2073 a co-located MGS without starting the MGS, which is a complement
2074 to 'nosvc' mount option.
2077 Frequency : always, on big-endian systems
2079 Description: cleanup in ptlrpc code, related to PPC platform
2080 Details : store magic in native order avoid panic's in recovery on PPC
2081 node and forbid from this error in future. Also fix posibily
2082 of twice swab data. Fix get lov striping to userpace.
2085 Frequency : rarely, if replay get lost on server
2087 Description: server incorrectly drop resent replays lead to recovery failure.
2088 Details : do not drop replay according to msg flags, instead we check the
2089 per-export recovery request queue for duplication of transno.
2092 Frequency : after recovery
2094 Description: precreate to many object's after del orphan.
2095 Details : del orphan st in oscc last_id == next_id and this triger growing
2096 count of precreated objects. Set flag LOW to skip increase count
2097 of precreated objects.
2100 Frequency : rare, on clear nid stats
2102 Description: ASSERTION(client_stat->nid_exp_ref_count == 0)
2103 Details : when clean nid stats sometimes try destroy live entry,
2104 and this produce panic in free.
2107 Frequency : occasionally since 1.6.4
2109 Description: Stack overflow during MDS log replay
2110 Details : ease stack pressure by using a thread dealing llog_process.
2113 Frequency : very rare
2115 Description: MDT cannot be unmounted, reporting "Mount still busy"
2116 Details : Mountpoint references were being leaked during open reply
2117 reconstruction after an MDS restart. Drop mountpoint reference
2118 in reconstruct_open() and free dentry reference also.
2123 Description: wait until IO finished before start new when do lock cancel.
2124 Details : VM protocol want old IO finished before start new, in this case
2125 need wait until PG_writeback is cleared until check dirty flag
2126 and call writepages in lock cancel callback.
2131 Description: mds_mfd_close() ASSERTION(rc == 0)
2132 Details : In mds_mfd_close(), we need protect inode's writecount change
2133 within its orphan write semaphore to prevent possible races.
2136 Frequency : rare, on shutdown ost
2138 Description: don't hit live lock with umount ost.
2139 Details : shrink_dcache_parent can be in long loop with destroy dentries,
2140 use shrink_dcache_sb instead.
2143 Frequency : only when echo_client is used
2145 Description: don't panic with use echo_client
2146 Details : echo client pass NULL as client nid pointer and this produce NULL
2147 pointer dereference.
2150 Frequency : Always on 32-bit PowerPC systems
2152 Description: fix build on PPC32
2153 Details : compile code with -m64 flag produce wrong object file for PPC32.
2158 Description: MDS LBUG: ASSERTION(!IS_ERR(dchild))
2159 Details : In reconstruct_* functions, LASSERTs on both the data supplied
2160 by a client, and the data on disk are dangerous and incorrect.
2161 Change them with client eviction.
2163 Severity : enhancement
2165 Description: skiplist implementation simplification
2166 Details : skiplists are used to group compatible locks on granted list
2167 that was implemented as tracking first and last lock of each lock
2168 group the patch changes that to using doubly linked lists
2172 Description: delete compatibility for 32bit qdata
2173 Details : as planned, when lustre is beyond b1_8, lquota won't support 32bit
2174 qunit. That means servers of b1_4 and servers of b1_8 can't be
2175 used together if users want to use quota.
2178 Frequency : only with administrator action
2180 Description: mount failure if config log has invalid conf_param setting
2181 Details : If administrator specified an incorrect configuration parameter
2182 with "lctl conf_param" this would cause an error during future
2183 client mounts. Instead, ignore the bad configuration parameter.
2186 Frequency : blocks per group < blocksize*8 and uninit_groups is enabled
2188 Description: ldiskfs error: XXX blocks in bitmap, YYY in gd
2189 Details : If blocks per group is less than blocksize*8, set rest of the
2193 Frequency : Application do stride read on lustre
2195 Description: The read performance will drop a lot if the application does
2197 Details : Because the stride_start_offset are missing in stride read-ahead,
2198 it will cause clients read a lot of unused pages in read-ahead,
2199 then the read-performance drops.
2203 Description: more ldlm soft lockups
2204 Details : In ldlm_resource_add_lock(), call to ldlm_resource_dump()
2205 starve other threads from the resource lock for a long time in
2206 case of long waiting queue, so change the debug level from
2207 D_OTHER to the less frequently used D_INFO.
2209 Severity : enhancement
2211 Description: add -gid, -group, -uid, -user options to lfs find
2213 Severity : enhancement
2215 Description: ll_recover_lost_found_objs - recover objects in lost+found
2216 Details : OST corruption and subsequent e2fsck can leave objects in the
2217 lost+found directory. Using the "ll_recover_lost_found_objs"
2218 tool, these objects can be retrieved and data can be salvaged
2219 by using the object ID saved in the fid EA on each object.
2224 Description: this bug _only_ happens when inode quota limitation is very low
2225 (less than 12), so that inode quota unit is 1 at initialization.
2226 Details : if remaining quota equates 1, it is a sign to demonstate that quota
2227 is effective now. So least quota qunit should be 2.
2231 Description: Hung threads in invalidate_inode_pages2_range
2232 Details : The direct IO path doesn't call check_rpcs to submit a new RPC once
2233 one is completed. As a result, some RPCs are stuck in the queue
2238 Description: Procfs and llog threads access destoryed import sometimes.
2239 Details : Sync the import destoryed process with procfs and llog threads by
2240 the import refcount and semaphore.
2244 Description: mds fails to respond, threads stuck in ldlm_completion_ast
2245 Details : Sort source/child resource pair after updating child resource.
2250 Description: kernel BUG at ldiskfs2_ext_new_extent_cb
2251 Details : If insertion of an extent fails, then discard the inode
2252 preallocation and free data blocks else it can lead to duplicate
2257 Description: don't always update ctime in ext3_xattr_set_handle()
2258 Details : Current xattr code updates inode ctime in ext3_xattr_set_handle()
2259 In some cases the ctime should not be updated, for example for
2260 2.0->1.8 compatibility it is necessary to delete an xattr and it
2261 should not update the ctime.
2265 Description: add quota statistics
2266 Details : 1. sort out quota proc entries and proc code.
2267 2. add quota statistics
2272 Description: quotas are not honored with O_DIRECT
2273 Details : all writes with the flag O_DIRECT will use grants which leads to
2274 this problem. Now using OBD_BRW_SYNC to guard this.
2278 Bugzilla : 15713/16362
2279 Description: Assertion in iopen_connect_dentry in 1.6.3
2280 Details : looking up an inode via iopen with the wrong generation number can
2281 populate the dcache with a disconneced dentry while the inode
2282 number is in the process of being reallocated. This causes an
2283 assertion failure in iopen since the inode's dentry list contains
2284 both a connected and disconnected dentry.
2288 Description: assertion failure in ldlm_handle2lock()
2289 Details : fix a race between class_handle_unhash() and class_handle2object()
2290 introduced in lustre 1.6.5 by bug 13622.
2292 Severity : enhancement
2294 Description: superblock lock contention with many SMP cores on one client
2295 Details : several client filesystem locks were highly contended on SMP
2296 NUMA systems with 8 or more cores. Per-CPU datastructures
2297 and more efficient locking implemented to reduce contention.
2302 Description: Kernel BUG: sd_iostats_bump: unexpected disk index
2303 Details : remove the limit of 256 scsi disks in the sd_iostat patch
2308 Description: oops in sd_iostats_seq_show()
2309 Details : unloading/reloading the scsi low level driver triggers a kernel
2310 bug when trying to access the sd iostat file.
2315 Description: Kernel panics during QLogic driver reload
2316 Details : REQ_BLOCK_PC requests are not handled properly in the sd iostat
2317 patch, causing memory corruption.
2322 Description: journal_dev option does not work in b1_6
2323 Details : pass mount option during pre-mount.
2325 Severity : enhancement
2327 Description: Add a FIEMAP(FIle Extent MAP) ioctl for ldiskfs
2328 Details : FIEMAP ioctl will allow an application to efficiently fetch the
2329 extent information of a file. It can be used to map logical blocks
2330 in a file to physical blocks in the block device.
2333 Frequency : only with adaptive timeout enabled
2335 Description: DEBUG_REQ() bad paging request
2336 Details : ptlrpc_at_recv_early_reply() should not modify req->rq_repmsg
2337 because it can be accessed by reply_in_callback() without the
2341 Frequency : only on Cray X2
2343 Description: X2 build failures
2344 Details : fix build failures on Cray X2.
2348 Description: xid & resent requests
2349 Details : Initialize RPC XID from clock at startup (randomly if clock is
2354 Description: quota recovery deadlock during mds failover
2355 Details : This patch includes att18982, att18236, att18237 in bz14840.
2357 1. fix osts hang when mds does failover with quotaon
2358 2. prevent watchdog storm when osts threads wait for the
2363 Description: kernel panic on racer
2364 Details : Do not access dchild->d_inode when IS_ERR(dchild) is true.
2366 Severity : enhancement
2368 Description: Add lustre_start utility to start or stop multiple Lustre servers
2373 Description: Lustre GPF in {:ptlrpc:ptlrpc_server_free_request+373}
2374 Details : In case of memory pressure, list_del() can be called twice on
2375 req->rq_history_list, causing a kernel oops.
2379 Description: (ptllnd_peer.c:557:kptllnd_peer_check_sends()) ASSERTION(!in_interrupt()) failed
2380 Details : fix stack overflow in the distributed lock manager by defering
2381 export eviction after a failed ast to the elt thread instead of
2382 handling it in the dlm interpret routine.
2384 Severity : enhancement
2386 Description: More exported tunables for mballoc
2387 Details : Add support for tunable preallocation window and new tunables for
2388 large/small requests
2392 Description: Detect corruption of block bitmap and checking for preallocations
2393 Details : Checks validity of on-disk block bitmap. Also it does better
2394 checking of number of applied preallocations. When corruption is
2395 found, it turns filesystem readonly to prevent further corruptions.
2399 Frequency : only for big-endian servers
2400 Description: Check for big-endian system while mounting fs with extents feature
2401 Details : Mounting a filesystem with extents feature will fail on big-endian
2402 systems since ext3-based ldiskfs is not supported on big-endian
2403 systems. Can be over-ridden with "bigendian_extents" mount option.
2407 Description: Excessive recovery window
2408 Details : With AT enabled, the recovery window can be excessively long (6000+
2409 seconds). To address this problem, we no longer use
2410 OBD_RECOVERY_FACTOR when extending the recovery window (the connect
2411 timeout no longer depends on the service time, it is set to
2412 INITIAL_CONNECT_TIMEOUT now) and clients report the old service
2413 time via pb_service_time.
2416 Descriptoin: Don't sync journal after every i/o
2417 Details : Implement write RPC replay to allow server replies for write RPCs
2418 before data is on disk. However, this feature is disabled by
2419 default since some issues leading to data corruptions have been
2420 found during recovery (e.g. bug 19128). This feature can be enabled
2421 by running the following command on the OSSs:
2422 lctl set_param obdfilter.*.sync_journal=0
2426 Description: Watchdog triggered on MDS failover
2427 Details : enable OBD_CONNECT_MDT flag when connecting from the MDS so that
2428 the OSTs know that the MDS "UUID" can be reused for the same export
2429 from a different NID, so we do not need to wait for the export to
2434 Descriptoin: Lustre detected file system corruption with inode out of bounds
2435 Details : don't update i_size on MDS_CLOSE for directories. This causes
2436 directory corruptions on the MDT.
2438 -------------------------------------------------------------------------------
2440 2008-05-26 Sun Microsystems, Inc.
2442 * Support for kernels:
2443 2.6.5-7.311 (SLES 9),
2444 2.6.9-67.0.7.EL (RHEL 4),
2445 2.6.16.54-0.2.5 (SLES 10),
2446 2.6.18-53.1.14.el5 (RHEL 5),
2447 2.6.22.14 vanilla (kernel.org)
2448 * Client support for unpatched kernels:
2449 (see http://wiki.lustre.org/index.php?title=Patchless_Client)
2450 2.6.16 - 2.6.22 vanilla (kernel.org)
2451 * Due to problems with nested symlinks and FMODE_EXEC (bug 12652),
2452 we do not recommend using patchless RHEL4 clients with kernels
2453 prior to 2.6.9-55EL (RHEL4U5).
2454 * Recommended e2fsprogs version: 1.40.7-sun1
2455 * Note that reiserfs quotas are disabled on SLES 10 in this kernel.
2456 * RHEL 4 and RHEL 5/SLES 10 clients behaves differently on 'cd' to a
2457 removed cwd "./" (refer to Bugzilla 14399).
2458 * A new quota file format has been introduced in 1.6.5.
2459 The format conversion from prior releases is handled transparently,
2460 but releases older than 1.4.12/1.6.5 will not understand this new
2461 format. The automatic format conversion can be avoided by running
2462 the following command on the MDS before upgrading:
2463 'tunefs.lustre --param="mdt.quota_type=ug1" $MDTDEV'.
2464 For more information, please refer to bugzilla 13904.
2468 Description: quota performance fix
2469 Details : quota data is written in journalled mode instead of ordered to
2470 increase performance
2474 Description: lfs support for human-readable quota grace time strings
2475 Details : lfs setquota -t and lfs quota -t represent quota grace times
2476 in "XXwXXdXXhXXmXXs" format instead of large values in seconds
2479 Frequency : always with o2ib 1.3 and sles10
2481 Description: fix build with SLES10 and o2ib v3.
2482 Details : sles10 uses diffrent name for Module.symver file but configure
2483 assume this file has same name on RHEL/SLES/vanila kernels.
2486 Frequency : very rare, if additional xattrs are used on kernels >= 2.6.12
2488 Description: MDS may lose file striping (and hence file data) in some cases
2489 Details : If there are additional extended attributes stored on the MDS,
2490 in particular ACLs, SELinux, or user attributes (if user_xattr
2491 is specified for the client mount options) then there is a risk
2492 of attribute loss. Additionally, the Lustre file striping
2493 needs to be larger than default (e.g. striped over all OSTs),
2494 and an additional attribute must be stored initially in the
2495 inode and then increase in size enough to be moved to the
2496 external attribute block (e.g. ACL growing in size) for file
2499 Severity : enhancement
2501 Description: add message levels for liblustreapi
2506 Description: MDT cannot be unmounted, reporting "Mount still busy"
2507 Details : Mountpoint references were being leaked during open reply
2508 reconstruction after an MDS restart. Drop mountpoint reference
2509 in reconstruct_open() and free dentry reference also.
2514 Description: fix for occasional failure case of -ENOSPC in recovery-small tests
2515 Details : Move the 'good_osts' check before the 'total_bavail' check. This
2516 will result in an -EAGAIN and in the exit call path we call
2517 alloc_rr() which will with increasing aggressiveness attempt to
2518 aquire precreated objects on the minimum number of required OSCs.
2522 Description: Use old size assignment to avoid deadlock
2523 Details : This reverts the changes in bugs 2369 and bug 14138 that introduced
2524 the scheduling while holding a spinlock. We do not need locking
2525 for size in ll_update_inode() because size is only updated from
2526 the MDS for directories or files without objects, so there is no
2527 other place to do the update, and concurrent access to such inodes
2528 are protected by the inode lock.
2532 Description: Use __u64 instead of int for valid bits
2536 Description: resolve "_IOWR redefined" build error on SLES10
2540 Description: dump the memory debugging after all modules are unloaded to
2541 suppress false negative in conf_sanity test 39
2545 Description: the recovery timer never expires
2546 Details : for new client connect request, the recovery timer should not be
2547 reset, otherwise recovery timer will never expired, if the old
2548 client never come. Only old client connect and first connection
2549 req should trigger recovery timer reset.
2553 Description: the min numbers of lproc stats are wrong
2554 Details : adding a new constant LC_MIN_INIT and use it for initialization
2558 Frequency : always with interactive lfs
2560 Description: Reinitialize optind to 0 so that interactive lfs works in all cases
2563 Frequency : with multiple concurrent readdir processes in same directory
2564 Bugzilla : 15406, 15169, 15175
2565 Description: misc fixes for directory readahead.
2566 Details : prevent previous statahead async RPC callback from processing the
2567 current "statahead_info", race condition between async RPC callback
2568 add dentry into dentry hash table and "ls" thread revalidate such
2569 dentry, statahead his/miss control for hidden items, and so on.
2571 Severity : enhancement
2573 Description: build kernel-ib packages for OFED 1.3 in our release cycle
2577 Description: incore types cleaning in quota code (with respect to 64-bit limits)
2578 Details : several u32 variables declarations are replaced with u64 declarations
2583 Description: fix SLES kernel versioning
2584 Details : the kernel version for our SLES 10 kernel did not include a "-"
2585 before the "smp" at the end. while this was not a problem in
2586 general, it did mean that software trying to use the kernel
2587 version to try to detect a vendor specific kernel would fail.
2588 this was most evident by the OFED build scripts.
2593 Description: Don't update lov_desc members until making sure they are valid
2594 Details : When updating lov_desc members via proc fs, need fix their
2595 validities before doing the real update.
2598 Frequency : very rare
2600 Description: don't put request into delay list while invalidate in flight.
2601 Details : ptlrpc_delay_request sometimes put in delay list while invalidate
2602 import in flight. this produce timeout for invalidate and sometimes
2603 can cause stale data.
2605 Severity : enhancement
2607 Description: Update kernel to SLES9 2.6.5-7.311.
2609 Severity : enhancement
2611 Description: Update kernel to RHEL4 2.6.9-67.0.7.
2616 Frequency : on PPC only
2617 Description: not convert ost objects for directory because it's not exist.
2618 Details : ll_dir_getstripe assume dirrectory has ost objects but this wrong.
2620 Severity : enhancement
2622 Description: Fix warnings with compile liblustre at sles10/rhel5 which have
2623 __u64 as usingied long long type.
2626 Frequency : rare, on shutdown
2628 Description: race process ast vs remove callback
2629 Details : removing callback before disconnect import open race with
2630 processing callback.
2632 Severity : enhancement
2634 Description: Update kernel to SLES9 2.6.5-7.311.
2636 Severity : enhancement
2638 Description: Files open for execute are not marked busy on SLES10
2639 Details : Add FMODE_EXEC to SLES10 SP1 server kernel series.
2641 Severity : enhancement
2643 Description: Add server support for vanilla-2.6.22.14.
2646 Frequency : occasional
2648 Description: Avoid lov_create() getting stuck in obd_statfs_rqset()
2649 Details : If an OST is down the MDS will hang indefinitely in
2650 obd_statfs_rqset() waiting for the statfs data. While for
2651 MDS QOS usage of statfs, it should not stuck in waiting.
2653 Severity : enhancement
2655 Description: Disable adaptive timeouts by default
2658 Frequency : on network error
2660 Description: panic with double free request if network error
2661 Details : mdc_finish_enqueue is finish request if any network error occuring,
2662 but it's true only for synchronus enqueue, for async enqueue
2663 (via ptlrpcd) this incorrect and ptlrpcd want finish request
2667 Frequency : rare, on recovery
2669 Description: read procfs can produce deadlock in some situation
2670 Details : Holding lprocfs lock with send rpc can produce block for destroy
2671 obd objects and this also block reconnect with -EALREADY.
2672 This isn't fix all lprocfs bugs - but make it rare.
2674 Severity : enhancement
2676 Description: Update kernel to RHEL5 2.6.18-53.1.14.el5.
2679 Frequency : frequent on X2 node
2681 Description: mdc_set_open_replay_data LBUG
2682 Details : Set replay data for requests that are eligible for replay.
2687 Description: lustre_mgs: operation 101 on unconnected MGS
2688 Details : When MGC is disconnected from MGS long enough, MGS will evict the
2689 MGC, and late on MGC cannot successfully connect to MGS and a lot
2690 of the error messages complaining that MGS is not connected.
2693 Frequency : rare, depends on device drivers and load
2695 Description: MDS or OSS nodes crash due to stack overflow
2696 Details : Code changes in 1.6.4 increased the stack usage of some functions.
2697 In some cases, in conjunction with device drivers that use a lot
2698 of stack, the MDS (or possibly OSS) service threads could overflow
2699 the stack. One change which was identified to consume additional
2700 stack has been reworked to avoid the extra stack usage.
2702 Severity : enhancement
2704 Description: Update to RHEL5 latest kernel-2.6.18-53.1.13.el5.
2706 Severity : enhancement
2708 Description: Update to SLES10 SP1 latest kernel-2.6.16.54-0.2.5.
2710 Severity : enhancement
2712 Description: Update to RHEL5 latest kernel-2.6.18-53.1.6.el5.
2714 Serverity : enhancement
2716 Description: Update RHEL4 kernel to 2.6.9-67.0.4.
2719 Frequency : rare on shutdown OST
2721 Description: Don't allow skipping OSTs if index has been specified.
2722 Details : Don't allow skipping OSTs if index has been specified, make locking
2723 in internal create lots better.
2728 Description: ASSERTION(!PageDirty(page)) failed
2729 Details : Wrong check could lead to an assertion failure under specific
2735 Description: LBUG in ptlrpc_check_set() bad phase ebc0de00
2736 Details : access to bitfield in structure is always rounded to long
2737 and this produce problem with not atomic change any bit.
2742 Description: Lustre make rpms failed.
2743 Details : Remove ldiskfs spec file to avoids rpmbuild be confused when
2744 builds Lustre rpms from tarball.
2746 Severity : enhancement
2748 Description: Update to SLES9 SP4 kernel-2.6.5-7.308.
2751 Frequency : rare on shutdown OST
2753 Description: If llog cancel was not send before clean_exports phase, this can
2754 produce deadlock in llog code.
2755 Details : If llog thread has last reference to obd and call class_import_put
2756 this produce deadlock because llog_cleanup_commit_master wait when
2757 last llog_commit_thread exited, but this never success because was
2758 called from llog_commit_thread.
2761 Frequency : only if OST index is skipped
2763 Description: NULL lov_tgts causing MDS oops
2764 Details : more safe checks for NULL lov_tgts for avoid oops.
2766 Severity : enhancement
2768 Description: Update to RHEL4 latest kernel-2.6.9-67.0.1.EL.
2770 Severity : enhancement
2772 Description: Update to RHEL5 latest kernel-2.6.18-53.1.4.el5.
2777 Description: make mgs_setparam() handle fsname containing dash
2778 Details : fsname containing a dash does not work with lctl conf_param
2780 Severity : enhancement
2782 Description: Update to RHEL4 Update-6 kernel-2.6.9-67.EL.
2785 Frequency : rare, in recovery and (or) destroy lovobjid file.
2787 Description: rewrite lov objid code.
2788 Details : Cleanup for lov objid code, remove scability problems and wrong
2789 locking. Fix sending last_id into ost.
2791 Severity : enhancement
2793 Description: Update to SLES10 SP1 latest kernel-2.6.16.54-0.2.3.
2795 Severity : enhancement
2797 Description: Update to RHEL5 Update-1 kernel 2.6.18-53.el5.
2798 Details : Use d_move_locked instead of __d_move.
2801 Frequency : rare, at shutdown
2803 Description: access already free / zero obd_namespace.
2804 Details : if client_disconnect_export was called without force flag set,
2805 and exist connect request in flight, this can produce access to
2806 NULL pointer (or already free pointer) when connect_interpret
2807 store ocd flags in obd_namespace.
2810 Frequency : only at startup
2812 Description: not alloc memory with spinlock held.
2813 Details : allocation memory with GFP_KERNEL can produce sleep deadlock,
2814 if any spinlock held.
2819 Description: lfs find does not continue on file error
2820 Details : Continue other files processing when a file/dir is absent.
2824 Description: Inconsistent usage of lustre_pack_reply()
2825 Details : Standardize the usage of lustre_pack_reply() such that it
2826 always generate a CERROR on failure.
2829 Frequency : very rare
2831 Description: Fix replay if there is an un-replied request and open
2832 Details : In some cases, older replay request will revert the
2833 mcd->mcd_last_xid on MDS which is used to record the client's
2834 latest sent request.
2836 Severity : enhancement
2838 Description: Update to RHEL5 kernel 2.6.18-8.1.15.el5.
2840 Severity : enhancement
2842 Description: Update to SLES10 SP1 kernel 2.6.16.53-0.16
2844 Severity : enhancement
2846 Description: Update to SLES9 kernel-2.6.5-7.287.3.
2848 Severity : enhancement
2850 Description: Update to RHEL4 kernel-2.6.9-55.0.12.EL.
2852 Severity : enhancement
2854 Description: Build SLES10 patchless client fails
2855 Details : The configure was broken by run ./configure with
2856 --with-linux-obj=.... argument for patchless client. When the
2857 configure use --with-linux-obj, the LINUXINCLUDE= -Iinclude
2858 can't search header adequately. Use absolute path such as
2859 -I($LINUX)/include instead.
2861 Severity : enhancement
2863 Description: Lustre Page Accounting
2864 Details : New macros for page alloc and free which enable accounting
2865 of page allocation of Lustre. Use percpu counters to store memory
2866 and page statistics.
2869 Frequency : only if debugging is disabled
2871 Description: LASSERT_{REQ,REP}SWAB macros are buggy
2872 Details : If SWAB_PARANOIA is disabled, the LASSERT_REQSWAB and
2873 LASSERT_REPSWAB macros become no-ops, which is incorrect. Drop
2874 these macros and replace them with their definitions instead.
2879 Description: interrupt oig_wait produce painc on resend.
2880 Details : brw_redo_request can be used for resend requests from ptlrpcd and
2881 private set, and this produce situation when rq_ptlrpcd_data not
2882 copyed to new allocated request and triggered LBUG on assert
2883 req->rq_ptlrpcd_data != NULL. But this member used only for wakeup
2884 ptlrpcd set if request is changed and can be safety changed to use
2887 Severity : enhancement
2889 Description: organize the server-side client stats on per-nid basis
2890 Details : Change the structure of stats under obdfilter and mds to
2898 The "uuid"s file would list the uuids of _active_ exports.
2899 And the clear entry is to clear all stats and stale nids.
2904 Description: Processes looping in ll_readdir() on Lustre clients finally causing
2905 a full node pseudo-hang
2906 Details : Concurrent access to the same directory from multiple clients with
2907 intensive file creation/removal can cause a client node to spin in
2908 ll_readdir(). i_version must be increased every time the lock is
2909 cancelled to ensure a revalidate is done.
2914 Description: touch file failed when fs is not full
2915 Details : OST in recovery should not be discarded by MDS in alloc_qos(),
2916 otherwise we can get ENOSP while fs is not full.
2919 Frequency : only for Cray XT3
2920 Bugzilla : 12829/13455
2921 Description: Changing primary group doesn't change the group lustre assigns to
2923 Details : When CRAY_XT3 is defined, the fsgid supplied by the client is
2924 overridden with the primary group provided by the group upcall,
2925 whereas the supplied fsgid can be trusted if it is in the list of
2926 supplementary groups returned by the group upcall.
2928 Severity : enhancement
2930 Description: Root Squash Functionality
2931 Details : Implementation of NFS-like root squash capability. Specifically,
2932 don't allow someone with root access on a client node to be able
2933 to manipulate files owned by root on a server node.
2935 Severity : enhancement
2937 Description: Slow trucate/writes to huge files at high offsets.
2938 Details : Directly associate cached pages to lock that protect those pages,
2939 this allows us to quickly find what pages to write and remove
2940 once lock callback is received.
2945 Description: Too many locks accumulating on client during NFS usage
2946 Details : mds_open improperly used accmode to find out access mode to a
2947 file. Also mdc_intent_lock logic to find out if we already have
2948 lock similar to just received was flawed since introduction of
2949 skiplists - locks are now added to the front of the granted
2954 Description: Hit ASSERTION(obd->obd_stopping == 1) failed in some setup failed
2956 Details : In obd setup failure handler, obd_stopping will not necessarily to
2957 be 1, and obd_set_up should also be checked to make sure whether
2958 obd is completely setup.
2960 Severity : enhancement
2962 Description: Allow masking D_WARNING, D_ERROR messages from console
2963 Details : Console messages can now be disabled via lnet.printk.
2968 Description: User code with malformed file open parameter crashes client node
2969 Details : Before packing join_file req, all the related reference should be
2970 checked carefully in case some malformed flags cause fake join_file
2976 Description: shrink/enlarge qunit size when needed; fix the problem of coarse
2977 grain of quota doing harm to quota's accuracy
2978 Details : qunit size will be changed when quota limitation is too low/high;
2979 record the pending quota write in order to get more accureate
2980 quota; delete the patch for bug12588, which is unnecessary when
2981 this patch is landed. This bug also contains fixes for bug 14526,
2982 14299, 14601 and 13794.
2986 Description: LDLM_ENQUEUE races with LDLM_CP_CALLBACK
2987 Details : ldlm_completion_ast() assumes that a lock is granted when the req
2988 mode is equal to the granted mode. However, it should also check
2989 that LDLM_FL_CP_REQD is not set.
2993 Description: Heavy nfs access might result in deadlocks
2994 Details : After ELC code landed, it is now improper to enqueue any mds
2995 locks under och_sem, because enqueue might want to decide to
2996 cancel open locks for same inode we are holding och_sem for.
3000 Description: 35% write performance drop with ldiskfs2 when quotas are on
3001 Details : Enable ext3 journalled quota by default to improve performance
3002 when quotas are turned on.
3006 Description: Client eviction while running blogbench
3007 Details : A lot of unlink operations with concurrent I/O can lead to a
3008 deadlock causing evictions. To address the problem, the number of
3009 oustanding OST_DESTROY requests is now throttled to
3010 max_rpcs_in_flight per OSC and LDLM_FL_DISCARD_DATA blocking
3011 callbacks are processed in priority.
3014 Frequency : RHEL4 only
3016 Description: mkfs is very slow on IA64/RHEL4
3017 Details : A performance regression has been discovered in the MPT Fusion
3018 driver between versions 3.02.73rh and 3.02.99.00rh. As a
3019 consequence, we have downgraded the MPT Fusion driver in the RHEL4
3020 kernel from 3.02.99.00 to 3.02.73 until this problem is fixed.
3023 Frequency : PPC/PPC64 only
3025 Description: conflicts between asm-ppc64/types.h and lustre_types.h
3026 Details : fix duplicated definitions between asm-ppc64/types.h and
3027 lustre_types.h on PPC.
3030 Frequency : PPC/PPC64 only
3032 Description: asm-ppc/segment.h does not exist
3033 Details : fix compile issue on PPC.
3037 Description: data checksumming impacts single node performance
3038 Details : add support for several checksum algorithms. Currently, CRC32 and
3039 Adler-32 are supported. The checksum type can be changed on the fly
3040 through /proc/fs/lustre/osc/*/checksum_type.
3044 Description: use adler32 for page checksums
3045 Details : when available, use the Adler-32 algorithm instead of CRC32 for
3050 Description: better handle error messages in extents code
3052 Severity : enhancement
3054 Description: SNMP support enhancement
3055 Details : Adding total number of sampled request for an MDS node in snmp
3058 Severity : enhancement
3060 Description: Optimize ldlm waiting list processing for PR extent locks
3061 Details : When processing waiting list for read extent lock and meeting read
3062 lock that is same or wider to it that is not contended, skip
3063 processing rest of the list and immediatelly return current
3064 status of conflictness, since we are guaranteed there are no
3065 conflicting locks in the rest of the list.
3069 Description: Time out and refuse to reconnect
3070 Details : When the failover node is the primary node, it is possible
3071 to have two identical connections in imp_conn_list. We must
3072 compare not conn's pointers but NIDs, otherwise we
3073 can defeat connection throttling.
3077 Description: Client not clear own cache if answer to reconnect is lost.
3078 Details : Client gets evicted from server. Now client also thinks it is
3079 disconnected (or gets ENOTCONN on its operation) and decides to
3080 reconnect. Server receives reconnect message, but cannot find
3081 export. New export is created that is fully valid (new cookie!),
3082 but reply is lost and not reported to client. Client reconnects
3083 again and gets back a just-created connection, but it is not new
3084 so client thinks it was not evicted and does not do recovery.
3088 Description: Detect stride IO mode in read-ahead
3089 Details : When a client does stride read, read-ahead should detect that and
3090 read-ahead pages according to the detected stride pattern.
3094 Description: build for x2 fails
3095 Details : fix compile issue on Cray systems.
3097 Severity : enhancement
3099 Description: implement readv/writev APIs(aio_read/aio_writes in newer kernels)
3100 Details : This greatly improves speed of NFS writes on 2.6 kernels.
3103 Frequency : only on PPC/SLES10
3105 Description: "BITS_PER_LONG is not 32 or 64" in linux/idr.h
3106 Details : On SLES10/PPC, fs.h includes idr.h which requires BITS_PER_LONG to
3107 be defined. Add a hack in mkfs_lustre.c to work around this compile
3112 Description: LASSERT on MDS when client holding flock lock dies
3113 Details : ldlm pool logic depends on number of granted locks equal to
3114 number of released locks which is not true for flock locks, so
3115 just exclude such locks from consideration.
3119 Description: MDS deadlock with many ll_sync_lov threads and I/O stalled
3120 Details : Use fsfilt_sync() for both the whole filesystem sync and
3121 individual file sync to eliminate dangerous inode locking
3122 with I_LOCK that can lead to a deadlock.
3126 Description: Update an obsolete wirecheck.c generator
3127 Details : Update wirecheck.c/wirehdr.c and regenerate wiretest.c
3131 Description: Client can panic on open sometimes
3132 Details : It is possible that we try to free already freed request in
3133 ll_file_open in some error cases when we send request from
3138 Description: performance in 1.6.3
3139 Details : Force q->max_phys_segments to MAX_PHYS_SEGMENTS on SLES10 to be
3140 sure that 1MB requests are not fragmented by the block layer.
3144 Description: LDLM soft lockups - improvement
3145 Details : It is be possible to send the lock handle along with each read
3146 or write request because the client is already doing a lock match
3147 itself so there isn't any reason the OST should have to re-do that
3153 Description: lfs quota fails with deactivated OSTS
3154 Details : With this patch, three improvements are included:
3155 1. detete the softlimit in mds and osts when use "lfs quota".
3156 2. display the inaccurate data in the output of "lfs quota".
3157 3. try to get quota info when "lfs quota" is executed.
3162 Description: Extent locks not granted with no conflicts sometimes.
3163 Details : When race occurs in glimpse handler and nothing is returned,
3164 we do not reprocess the queue after lock cancel, and that leads
3165 to a stall until next activity on a resource
3168 Frequency : failover with quotaon
3170 Description: during mds failovers with quota on, OSTs got into deadlock state
3171 and causing dumpstack.
3172 Details : for every quota slave, at any time, there is only one quota req
3173 is sent to quota master for every uid/gid. Before that quota req
3174 returns, all the thread relative to the same uid/gid will wait.
3175 So if the quota req is lost because mds failovers or any other
3176 reasons, this bug will be hit. Now, dqacq_interpret() will handle
3177 quota reqs who time out.
3179 Severity : enhancement
3182 Description: when quota slave checks if quota is enough, there is an unnecessary
3184 Details : place this wait on necessary place instead of always waiting.
3186 --------------------------------------------------------------------------------
3188 2007-12-07 Cluster File Systems, Inc. <info@clusterfs.com>
3190 * Support for kernels:
3191 2.6.5-7.286 (SLES 9),
3192 2.6.9-55.0.9.EL (RHEL 4),
3193 2.6.16.53-0.8 (SLES 10),
3194 2.6.18-8.1.14.el5 (RHEL 5),
3195 2.6.18.8 vanilla (kernel.org)
3196 * Client support for unpatched kernels:
3197 (see http://wiki.lustre.org/index.php?title=Patchless_Client)
3198 2.6.16 - 2.6.22 vanilla (kernel.org)
3199 * Due to recently discovered recovery problems, we do not recommend
3200 using patchless RHEL 4 clients with this or any earlier release.
3201 * Recommended e2fsprogs version: 1.40.2-cfs1
3202 * Note that reiserfs quotas are disabled on SLES 10 in this kernel.
3205 Frequency : occasional
3207 Description: MDS hang or stay in waiting lock
3208 Details : If client receive lock with CBPENDING flag ldlm need send lock
3209 cancel as separate rpc, to avoid situation when cancel request
3210 can't processed due all i/o threads stay in wait lock.
3213 Frequency : occasional
3215 Description: Do not fail import if osc_interpret_create gets -EAGAIN
3216 Details : If osc_interpret_create got -EAGAIN it immediately exits and
3217 wakeup oscc_waitq. After wakeup oscc_wait_for_objects call
3218 oscc_has_objects and see OSC has no objests and call
3219 oscc_internal_create to resend create request.
3221 Severity : enhancement
3223 Description: Update kernel patches for SLES10 2.6.16.53-0.8.
3224 Details : Update which_patch & target file for SLES10 latest kernel.
3226 Severity : enhancement
3228 Description: add --type and --size parameters to lfs find
3229 Details : Enhance lfs find by adding filetype and filesize parameters. Also
3230 multiple OBDs can now be specified for the --obd option.
3232 Severity : enhancement
3234 Description: eliminate client locks in face of contention
3235 Details : file contention detection and lockless i/o implementation
3236 for contended files.
3238 Severity : enhancement
3240 Description: Remove client patches from SLES 10 kernel.
3241 Details : This causes SLES 10 clients to behave as patchless clients
3242 even on a Lustre-patched (server) kernel.
3244 Severity : enhancement
3246 Description: use i_size_read and i_size_write in 2.6 port
3247 Details : replace inode->i_size access with i_size_read/write()
3250 Frequency : when removing large files
3252 Description: scheduling issue during removal of large Lustre files
3253 Details : Don't take the BKL in fsfilt_ext3_setattr() for 2.6 kernels.
3254 It causes scheduling issues when removing large files (17TB in the
3260 Description: 1.4.11 Can't handle directories with stripe set and extended ACLs
3261 Details : Impossible (EPROTO is returned) to access a directory that has a
3262 non-default striping and ACLs.
3265 Frequency : only on ppc
3267 Description: /proc/fs/lustre/devices broken on ppc
3268 Details : The patch as applied to 1.6.2 doesn't look correct for all arches.
3269 We should make sure the type of 'index' is loff_t and then cast
3270 explicitly as needed below. Do not assign an explicitly cast
3274 Frequency : only for rhel5
3276 Description: Kernel patches update for RHEL5 2.6.18-8.1.10.el5.
3277 Details : Modify the target file & which_kernel.
3280 Frequency : if the uninit_groups feature is enabled on ldiskfs
3282 Description: e2fsck reports "invalid unused inodes count"
3283 Details : If a new ldiskfs filesystem is created with the "uninit_groups"
3284 feature and only a single inode is created in a group then the
3285 "bg_unused_inodes" count is incorrectly updated. Creating a
3286 second inode in that group would update it correctly.
3291 Description: buffer overruns could theoretically occur
3292 Details : llapi_semantic_traverse() modifies the "path" argument by
3293 appending values to the end of the origin string, and a buffer
3294 overrun may occur. Adding buffer overrun check in liblustreapi.
3296 Severity : enhancement
3298 Description: Add jbd statistics patch for RHEL5 and 2.6.18-vanilla.
3301 Frequency : only if filesystem is inconsistent
3303 Description: handle "serious error: objid * already exists" more gracefully
3304 Details : If LAST_ID value on disk is smaller than the objects existing in
3305 the O/0/d* directories, it indicates disk corruption and causes an
3306 LBUG(). If the object is 0-length, then we should use the existing
3307 object. This will help to avoid a full fsck in most cases.
3309 Severity : enhancement
3311 Description: Kernel patches update for RHEL4 2.6.9-55.0.6.
3312 Details : Modify vm-tunables-rhel4.patch.
3314 Severity : enhancement
3316 Description: Kernel config for 2.6.18-vanilla.
3317 Details : Modify targets/2.6-vanilla.target.in.
3318 Add config file kernel-2.6.18-2.6-vanilla-i686.config.
3319 Add config file kernel-2.6.18-2.6-vanilla-i686-smp.config.
3320 Add config file kernel-2.6.18-2.6-vanilla-x86_64.config.
3321 Add config file kernel-2.6.18-2.6-vanilla-x86_64-smp.config.
3324 Frequency : occasional
3326 Description: improve handling recoverable errors
3327 Details : If request processed with error which can be recoverable on server
3328 request should be resend, otherwise page released from cache and
3334 Description: Kernel patches update for RHEL5 2.6.18-8.1.14.el5.
3335 Details : Modify target file & which_patch.
3336 A flaw was found in the IA32 system call emulation provided
3337 on AMD64 and Intel 64 platforms. An improperly validated 64-bit
3338 value could be stored in the %RAX register, which could trigger an
3339 out-of-bounds system call table access. An untrusted local user
3340 could exploit this flaw to run code in the kernel
3341 (ie a root privilege escalation). (CVE-2007-4573).
3345 Description: change order of libsysio includes
3346 Details : '#include sysio.h' should always come before '#include xtio.h'
3348 Severity : enhancement
3350 Description: adapt the lustre_config script to support the upgrade case
3351 Details : Add "-u" option for lustre_config script to support upgrading 1.4
3352 server targets to 1.6 in parallel.
3357 Description: To avoid grant space > avaible space when the disk is almost
3358 full. Without this patch you might see the error "grant XXXX >
3359 available" or some LBUG about grant, when the disk is almost
3361 Details : In filter_check_grant, for non_grant cache write, we should
3362 check the left space by if (*left > ungranted + bytes), instead
3363 of (*left > ungranted), because only we are sure the left space
3364 is enough for another "bytes", then the ungrant space should be
3365 increase. In client, we should update cl_avail_grant only there is
3366 OBD_MD_FLGRANT in the reply.
3371 Description: Update RHEL 4 kernel to fix local root privilege escalation.
3372 Details : Update to the latest RHEL 4 kernel to fix the vulnerability
3373 described in CVE-2007-4573. This problem could allow untrusted
3374 local users to gain root access.
3377 Frequency : when using O_DIRECT and quotas
3379 Description: Incorrect file ownership on O_DIRECT output files
3380 Details : block usage reported by 'lfs quota' does not take into account
3381 files that have been written with O_DIRECT.
3385 Description: (rw.c:1323:ll_read_ahead_pages()) ASSERTION(page_idx > ria->ria_stoff) failed
3386 Details : Once the unmatched stride IO mode is detected, shrink the stride-ahead
3387 window to 0. If it does hit cache miss, and read-pattern is still
3388 stride-io mode, does not reset the stride window, but also does not
3389 increase the stride window length in this case.
3391 --------------------------------------------------------------------------------
3393 2007-09-27 Cluster File Systems, Inc. <info@clusterfs.com>
3395 * Support for kernels:
3396 2.6.5-7.286 (SLES 9),
3397 2.6.9-55.0.2.EL (RHEL 4),
3398 2.6.16.46-0.14 (SLES 10),
3399 2.6.18-8.1.8.el5 (RHEL 5),
3400 2.6.18.8 vanilla (kernel.org)
3401 * Client support for unpatched kernels:
3402 (see http://wiki.lustre.org/index.php?title=Patchless_Client)
3403 2.6.16 - 2.6.21 vanilla (kernel.org)
3404 * Due to recently discovered recovery problems, we do not recommend
3405 using patchless RHEL 4 clients with this or any earlier release.
3406 * Recommended e2fsprogs version: 1.40.2-cfs1
3407 * Note that reiserfs quotas are disabled on SLES 10 in this kernel.
3411 Description: Fix errors in lfs documentation
3412 Details : Fixes man pages
3414 Severity : enhancement
3416 Description: Adaptive timeouts
3417 Details : RPC timeouts adapt to changing server load and network
3418 conditions to reduce resend attempts and improve recovery time.
3420 Severity : enhancement
3422 Description: llapi_file_create() does not allow some changes
3423 Details : add llapi_file_open() that allows specifying the file creation
3424 mode and open flags, and also returns an open file handle.
3426 Severity : enhancement
3428 Description: df doesn't work properly if diskfs blocksize != 4K
3429 Details : Choose biggest blocksize of OST's as the LOV's blocksize.
3432 Frequency : always on directories with default striping set
3434 Description: lfs find on -1 stripe looping in lsm_lmm_verify_common()
3435 Details : Avoid lov_verify_lmm_common() on directory with -1 stripe count.
3437 Severity : enhancement
3439 Description: merge and cleanup kernel patches.
3440 Details : Remove mnt_lustre_list in vfs_intent-2.6-rhel4.patch.
3442 Severity : enhancement
3444 Description: RedHat Update kernel for RHEL5
3445 Details : Kernel config file for RHEL5.
3447 Severity : enhancement
3449 Description: OSS needs mutliple precreate threads
3450 Details : Add ability to start more than one create thread per OSS.
3453 Frequency : only with quota on the root user
3455 Description: mds_obd_create error creating tmp object
3456 Details : When the user sets quota on root, llog will be affected and can't
3457 create files and write files.
3460 Frequency : Always on ia64 patchless client, and possibly others.
3462 Description: Add EXPORT_SYMBOL check for node_to_cpumask symbol.
3463 Details : This allows the patchless client to be loaded on architectures
3464 without this export.
3469 Description: disorder of journal start and llog_add cause deadlock.
3470 Details : in llog_origin_connect, journal start should happen before
3471 llog_add keep the same order as other functions to avoid
3474 Severity : enhancement
3476 Description: RedHat Update kernel for RHEL5
3477 Details : Modify the kernel config file more closer RHEL5.
3479 Severity : enhancement
3481 Description: Build failure against Centos5 (RHEL5)
3482 Details : Define PAGE_SIZE when it isn't present.
3485 Frequency : occasionally when using NFS
3487 Description: "ll_intent_file_open()) lock enqueue: err: -13" with nfs
3488 Details : with NFS, the anon dentry's parent was set to itself in
3489 d_alloc_anon(), so in MDS, we use rec->ur_fid1 to find the
3490 corresponding dentry other than use rec->ur_name.
3492 Severity : enhancement
3494 Description: enable data checksumming by default
3495 Details : enable checksum by default, allow --disable-checksum
3496 configure option and "-o nochecksum" mount option. Checksums
3497 can also be disabled at runtime via $LPROC/osc/*/checksum_pages.
3500 Frequency : Occasionally with failover
3502 Description: Client eviction due to failover config
3503 Details : after a connection loss, the lustre client should attempt to
3504 reconnect to the last active server first before trying the
3505 other potential connections.
3507 Severity : enhancement
3509 Description: client-side metadata stat-ahead during readdir(directory readahead)
3510 Details : perform client-side metadata stat-ahead when the client detects
3511 readdir and sequential stat of dir entries therein
3516 Description: when mds and osts use different quota unit(32bit and 64bit),
3517 quota will be released repeatly.
3518 Details : void sending multiple quota reqs to mds, which will keep the
3519 status between the reqs.
3522 Frequency : only with liblustre clients on XT3
3524 Description: evictions taking too long
3525 Details : allow llrd to evict clients directly on OSTs
3529 Description: osts not allocated evenly to files
3530 Details : change the condition to increase offset_idx
3532 Severity : enhancement
3534 Description: self-adjustable client's lru lists
3535 Details : use adaptive algorithm for managing client cached locks lru
3536 lists according to current server load, other client's work
3537 pattern, memory activities, etc. Both, server and client
3538 side namespaces provide number of proc tunables for controlling
3543 Description: rewrite ext2-derived code in llite/dir.c and obdclass/uuid.c
3544 Details : rewrite inherited code (uuid parsing code from ext2 utils and
3545 readdir code from ext3) from scratch preserving functionality.
3549 Description: Only those disconnect error should be returned by rq_status.
3550 Details : In open/enqueue processs, Some errors, which will cause client
3551 disconnected, should be returned by rq_status, while other
3552 errors should still be returned by intent, then mdc or llite will
3555 Severity : enhancement
3557 Description: Tune the kernel for good SCSI performance.
3558 Details : Set the value of /sys/block/{dev}/queue/max_sectors_kb
3559 to the value of /sys/block/{dev}/queue/max_hw_sectors_kb
3563 Frequency : Always for filesystems larger than 2TB on 32-bit systems.
3564 Bugzilla : 13547, 13627
3565 Description: Data corruption for OSTs that are formatted larger than 2TB
3567 Details : When generating the bio request for lustre file writes the
3568 sector number would overflow a temporary variable before being
3569 used for the IO. The data reads correctly from Lustre (which
3570 will overflow in a similar manner) but other file data or
3571 filesystem metadata may be corrupted in some cases.
3575 Description: "lfs find -obd UUID" prints directories
3576 Details : "lfs find -obd UUID" will return all directory names instead
3577 of just file names. It is incorrect because the directories
3578 do not reside on the OSTs.
3581 Bugzilla : 13596, 13614
3582 Description: MDS hang after unclean shutdown of lots of clients
3583 Details : Never resend AST requests.
3587 Frequency : Always, for kernels after 2.6.16
3588 Description: Fix warning idr_remove called for id=.. which is not allocated.
3589 Details : Last kernels save old s_dev before kill super and not allow
3590 to restore from callback - restore it before call kill_anon_super.
3594 Description: Fix error on 'ls .' at the top of the Lustre mount.
3595 Details : Don't revalidate dentry if it is a root dentry.
3597 --------------------------------------------------------------------------------
3599 2007-08-27 Cluster File Systems, Inc. <info@clusterfs.com>
3601 * Support for kernels:
3602 2.6.5-7.286 (SLES 9),
3603 2.6.9-55.0.2.EL (RHEL 4),
3604 2.6.16.46-0.14 (SLES 10),
3605 2.6.18-8.1.8.el5 (RHEL 5),
3606 2.6.18.8 vanilla (kernel.org)
3607 * Client support for unpatched kernels:
3608 (see http://wiki.lustre.org/index.php?title=Patchless_Client)
3609 2.6.16 - 2.6.22 vanilla (kernel.org)
3610 * Due to recently discovered recovery problems, we do not recommend
3611 using patchless RHEL 4 clients with this or any earlier release.
3612 * Recommended e2fsprogs version: 1.39.cfs8
3613 * Note that reiserfs quotas are disabled on SLES 10 in this kernel.
3618 Description: block reactivating mgc import until all deactivates complete
3619 Details : Fix race when failing back MDT/MGS to itself (testing)
3621 Severity : enhancement
3623 Description: lfs setstripe enhancement
3624 Details : Make lfs setstripe understand 'k', 'm' and 'g' for stripe size.
3627 Frequency : mds/oss recovery
3629 Description: llog ctxt is refrenced after it has been freed.
3630 Details : llog ctxt refcount was added to avoide the race between ctxt free
3631 and llog recovery process. Each llog user must hold ctxt refcount
3632 before it access the llog. And the llog ctxt can only be freed
3633 when its refcount is zero.
3635 Severity : enhancement
3637 Description: randomly memory allocation failure util
3638 Details : Make lustre randomly failed allocating memory for testing purpose.
3641 Frequency : only for SLES10
3643 Description: Update kernel patch for SLES10 SP1
3644 Details : Add patch blkdev_tunables-2.6-sles10.patch to 2.6-sles10.series.
3646 Severity : enhancement
3648 Description: omit set fsid for export NFS
3649 Details : fix set/restore device id for avoid EMFILE error and mark lustre fs
3650 as FS_REQUIRES_DEV for avoid problems with generate fsid.
3653 Frequency : after network failures
3655 Description: Add sync option to mount_lustre.c
3656 Details : Client loses data written to lustre after a network interruption.
3658 Severity : enhancement
3660 Description: Error message improvement.
3661 Details : Merging of two LCONSOLE_ERROR_MSG into one.
3663 Severity : enhancement
3665 Description: don't use GFP_* in generic Lustre code.
3666 Details : Use cfs_alloc_* functions and CFS_* flags for code portability.
3668 Severity : enhancement
3670 Description: obdclass is limited by single OBD_ALLOC(idarray)
3671 Details : replace OBD_ALLOC/OBD_FREE with OBD_VMALLOC/OBD_VFREE
3673 Severity : enhancement
3675 Description: updated patchess for new RHEL4 kernel
3676 Details : Fixed ext3-unlink-race.patch per Kalpak's comment.
3678 Severity : enhancement
3680 Description: warnings with build patchless client with vanila 2.6.19 and up
3681 Details : change old ctl_table style and replace ctl_table/ctl_table_header
3682 with cfs_sysctl_table_t/cfs_sysctl_table_header_t
3684 Severity : enhancement
3686 Description: O_DIRECT bypasses client statistics.
3687 Details : When running with O_DIRECT I/O, neither the client rpc_stats nor
3688 read_ahead_stats were updated. Copied stats section from
3689 osc_send_oap_rpc() into async_internal().
3692 Frequency : only for Cray XT3
3694 Description: peer credits not enough on many OST per OSS systems.
3695 Details : Use new lnet way to add credits as we need those for pings and ASTs
3698 Frequency : only with liblustre
3700 Description: Liblustre is not releasing flock locks on file close.
3701 Details : Release flock locks on file close.
3704 Frequency : only for RHEL4
3706 Description: Update kernel patches for kernel-2.6.9-55.0.2.EL
3707 Details : Remove inode-nr_unused-2.6.9-rhel4.patch from 2.6-rhel4.series
3708 Update target file and kernel config.
3712 Description: lustre support for RHEL5
3713 Details : Add support for RHEL5.
3718 Description: ASSERTION(export != NULL) failed in target_handle_connect
3719 Details : Assetion hit is result of rare race between disconnect and connet
3720 to same nid. target_handle_connect found old connect cockie and
3721 tried to reconnect, but can't find export for this cockie.
3726 Description: umount blocks forever on error
3727 Details : In result of wrong using obd_no_recov and obd_force flags client
3728 can hand if cancel or some other requests is lost.
3731 Frequency : Only for SLES
3733 Description: sanity_quota fail test_1
3734 Details : There are multiple occurences of $TSTUSR in SLES's /etc/group
3735 file, which makes TSTID[2] inunique.
3737 Severity : enhancement
3739 Description: Kernel patches for SLES9 2.6.5-7.286 kernel
3740 Details : Update target/ChangeLog/which_patch .
3742 Severity : enhancement
3744 Description: jbd statistics
3745 Details : Port older jbd statistics patch for sles10
3750 Description: lvbo_init failed for resource with missing objects.
3751 Details : Fix returning error if we do stat for file with missing/corrupted
3752 objects and i_size set to all sum of size of all avaible objects.
3753 if we to truncate/write to missing object - it is recreated.
3758 Description: Oops in read and write path when failing to allocate lock.
3759 Details : Check if lock allocation failed and return error back.
3762 Frequency : When flocks are used.
3764 Description: assertion failure in ldlm_cli_enquque_fini for non NULL lock.
3765 Details : Flock locks might destroy just granted lock if it could be merged
3766 with another existing flock, this is done in completion handler,
3767 so teach ldlm_cli_enquque_fini that this is a valid case for
3773 Description: reply_lock_interpret crash due to race with it and lock cancel.
3774 Details : Do not replay locks that are being cancelled. Do not reference
3775 locks by their address during replay, just by their handle.
3778 Frequency : only with deactivated OSTs
3780 Description: lstripe command fails for valid OST index
3781 Details : The stripe offset is compared to 'lov->desc.ld_tgt_count'
3782 instead of lov->desc.ld_active_tgt_count.
3784 Severity : enhancement
3786 Description: Build failure against Centos5 (RHEL5)
3787 Details : Use getpagesize() instead of PAGE_SIZE.
3789 --------------------------------------------------------------------------------
3791 2007-07-30 Cluster File Systems, Inc. <info@clusterfs.com>
3793 * Support for kernels:
3794 2.6.5-7.283 (SLES 9),
3795 2.6.9-55.EL (RHEL 4),
3796 2.6.16.46-0.14 (SLES 10),
3797 2.6.18.8 vanilla (kernel.org)
3798 * Client support for unpatched kernels:
3799 (see http://wiki.lustre.org/index.php?title=Patchless_Client)
3800 2.6.16 - 2.6.22 vanilla (kernel.org)
3801 * Due to recently discovered recovery problems, we do not recommend
3802 using patchless RHEL 4 clients with this or any earlier release.
3803 * Recommended e2fsprogs version: 1.39.cfs8
3804 * Note that reiserfs quotas are disabled on SLES 10 in this kernel.
3805 * Starting with this release, the ldiskfs backing filesystem required
3806 by Lustre is now in its own package, lustre-ldiskfs. This package
3807 should be installed. It is versioned separately from Lustre and
3808 may be released separately in future.
3813 Description: server LBUG when shutting down
3814 Details : Block umount forever until the mount refcount is zero rather
3815 than giving up after an arbitrary timeout.
3817 Severity : enhancement
3819 Description: add optional extra BUILD_VERSION info
3820 Details : add a new environment variable (namely LUSTRE_VERS) which allows
3821 to override the lustre version.
3824 Frequency : 2.6.18 servers only
3826 Description: ll_kern_mount() doesn't release the module reference
3827 Details : The ldiskfs module reference count never drops down to 0
3828 because ll_kern_mount() doesn't release the module reference.
3833 Description: server LBUG when using old ost_num_threads parameter
3834 Details : Accept the old ost_num_threads parameter but warn that it
3835 is deprecated, and fix an off-by-one error that caused an LBUG.
3840 Description: Transient SCSI error results in persistent IO issue
3841 Details : iobuf->dr_error is not reinitialized to 0 between two
3845 Frequency : sometimes when underlying device returns I/O errors
3847 Description: OSTs not going read-only during write failures
3848 Details : OSTs are not remounted read-only when the journal commit threads
3849 get I/O errors because fsfilt_ext3 calls journal_start/stop()
3850 instead of the ext3 wrappers.
3853 Frequency : only with 10000 clients or more
3855 Description: poor connect scaling with increasing client count
3856 Details : Don't run filter_grant_sanity_check for more than 100 exports
3857 to improve scaling for large numbers of clients.
3860 Frequency : SLES10 only
3862 Description: sanity-quota.sh quotacheck failed: rc = -22
3863 Details : Quotas cannot be enabled on SLES10.
3866 Frequency : liblustre clients only
3868 Description: getdirentries does not give error when run on compute nodes
3869 Details : getdirentries does not fail when the size specified as an argument
3870 is too small to contain at least one entry
3872 Severity : enhancement
3874 Description: Add LNET router traceability for debug purposes
3875 Details : If a checksum failure occurs with a router as part of the
3876 IO path, the NID of the last router that forwarded the bulk data
3877 is printed so it can be identified.
3882 Description: OST "spontaneously" evicts client; client has imp_pingable == 0
3883 Details : Due to a race condition, liblustre clients were occasionally
3884 evicted incorrectly.
3886 Severity : enhancement
3888 Description: lfs setstripe use optional parameters instead of postional
3891 Severity : enhancement
3893 Description: Nanosecond timestamp support for ldiskfs
3894 Details : The on-disk ldiskfs filesystem has added support for nanosecond
3895 resolution timestamps. There is not yet support for this at
3896 the Lustre filesystem level.
3899 Frequency : during server recovery
3901 Description: MDS failing to send precreate requests due to OSCC_FLAG_RECOVERING
3902 Details : request with rq_no_resend flag not awake l_wait_event if they get a
3906 Frequency : nfs export on patchless client
3908 Description: connectathon hang when test nfs export over patchless client
3909 Details : Disconnected dentry cannot be found with lookup, so we do not need
3910 to unhash it or make it invalid
3913 Description: fix llapi_lov_get_uuids() to allow many OSTs to be returned
3914 Details: : Change llapi_lov_get_uuids() to read the UUIDs from /proc instead
3915 of using an ioctl. This allows lfsck for > 160 OSTs to succeed.
3920 Description: open req refcounting wrong on reconnect
3921 Details : If reconnect happened between getting open reply from server and
3922 call to mdc_set_replay_data in ll_file_open, we will schedule
3923 replay for unreferenced request that we are about to free.
3924 Subsequent close will crash in variety of ways.
3925 Check that request is still eligible for replay in
3926 mdc_set_replay_data().
3931 Description: disable writes to filesystem when reading health_check file
3932 Details : the default for reading the health_check proc file has changed
3933 to NOT do a journal transaction and write to disk, because this
3934 can cause reads of the /proc file to hang and block HA state
3935 checking on a healthy but otherwise heavily loaded system. It
3936 is possible to return to the previous behaviour during configure
3937 with --enable-health-write.
3939 Severity : enhancement
3941 Description: 64-bit inode version
3942 Details: : Add a on-disk 64-bit inode version for ext3 to track changes made
3943 to the inode. This will be required for version-based recovery.
3948 Description: MDS fails to start if a duplicate client export is detected
3949 Details : in some rare cases it was possible for a client to connect to
3950 an MDS multiple times. Upon recovery the MDS would detect this
3951 and fail during startup. Handle this more gracefully.
3953 Severity : enhancement
3955 Description: Add -o localflock option to simulate old noflock behaviour.
3956 Details : This will achieve local-only flock/fcntl locks coherentness.
3961 Description: log_commit_thread vs filter_destroy race leads to crash
3962 Details : Take import reference before releasing llog record semaphore
3967 Description: Wrong request locking in request set processing
3968 Details : ptlrpc_check_set wrongly uses req->rq_lock for proctect add to
3969 imp_delayed_list, in this place should be used imp_lock.
3972 Frequency : when reconnecting
3974 Description: Grant leak when OSC reconnect to OST
3975 Details : When osc reconnect ost, OST(filter) should check whether it
3976 should grant more space to client by comparing fed_grant and
3977 cl_avail_grant, and return the granted space to client instead
3978 of "new granted" space, because client will call osc_init_grant
3979 to update the client grant space info.
3982 Frequency : when client reconnects to OST
3984 Description: Grant leak when OSC does a resend and replays bulk write
3985 Details : When osc reconnect to OST, OST(filter) should clear grant info of
3986 bulk write request, because the grant info while be sync between
3987 OSC and OST when reconnect, and we should ignore the grant info
3988 these of resend/replay write req.
3993 Description: Grant space more than avaiable space sometimes.
3994 Details : When then OST is about to be full, if two bulk writing from
3995 different clients came to OST. Accord the avaliable space of the
3996 OST, the first req should be permitted, and the second one
3997 should be denied by ENOSPC. But if the seconde arrived before
3998 the first one is commited. The OST might wrongly permit second
3999 writing, which will cause grant space > avaiable space.
4002 Frequency : when client is evicted
4004 Description: Grant might be wrongly erased when osc is evicted by OST
4005 Details : when the import is evicted by server, it will fork another
4006 thread ptlrpc_invalidate_import_thread to invalidate the
4007 import, where the grant will be set to 0. While the original
4008 thread will update the grant it got when connecting. So if
4009 the former happened latter, the grant will be wrongly errased
4010 because of this race.
4015 Description: Checking Stale with correct fid
4016 Details : ll_revalidate_it should uses de_inode instead of op_data.fid2
4017 to check whether it is stale, because sometimes, we want the
4018 enqueue happened anyway, and op_data.fid2 will not be initialized.
4020 Severity : enhancement
4022 Description: update patchless client
4023 Details : Add support for patchless client with 2.6.20, 2.6.21 and RHEL 5
4026 Frequency : only with 2.4 kernel
4028 Description: random memory corruption
4029 Details : size of struct ll_inode_info is to big for union inode.u and this
4030 can be cause of random memory corruption.
4035 Description: Memory leak in recovery
4036 Details : Lov_mds_md was not free in an error handler in mds_create_object.
4037 It should also check obd_fail before fsfilt_start, otherwise if
4038 fsfilt_start return -EROFS,(failover mds during mds recovery).
4039 then the req will return with repmsg->transno = 0 and rc = EROFS.
4040 and we met hit the assert LASSERT(req->rq_reqmsg->transno ==
4041 req->rq_repmsg->transno) in ptlrpc_replay_interpret. Fcc should
4042 be freed no matter whether fsfilt_commit success or not.
4045 Frequency : only with large numbers of cores on a single node
4047 Description: Prevents from taking the superblock lock in llap_from_page for
4049 Details : using LL_ORIGIN_REMOVEPAGE origin flag instead of LL_ORIGIN_UNKNOW
4050 for llap_from_page call in ll_removepage() prevents from taking
4051 the superblock lock for a soon died page.
4056 Description: Not check open intent error before release open handle
4057 Details : in some rare cases, the open intent error is not checked before
4058 release open handle, which may cause
4059 ASSERTION(open_req->rq_transno != 0), because it tries to release
4060 the failed open handle.
4065 Description: Set cat log bitmap only after create log success.
4066 Details : in some rare cases, the cat log bitmap is set too early. and it
4067 should be set only after create log success.
4071 Description: Accessing a block bevice can re-enable I/O when Lustre is
4072 tearing down a device.
4073 Details : dev_clear_rdonly(bdev) must be called in kill_bdev() instead of
4078 Description: service threads may hog cpus when there are a lot of requests
4079 Details : Insert cond_resched to give other threads a chance to use some CPU
4084 Description: the cat log was not initialized in recovery
4085 Details : When mds(mgs) do recovery, the tgt_count might be zero, so the
4086 unlink log on mds will not be initialized until mds post
4087 recovery. And also in mds post recovery, the unlink log will
4088 initialization will be done asynchronausly, so there will be race
4089 between add unlink log and unlink log initialization.
4093 Description: brw_stats were being printed incorrectly
4094 Details : brw_stats were being printed as log2 but all of them were not
4095 recorded as log2. Also remove some code duplication arising from
4096 filter_tally_{read,write}.
4100 Frequency : rare, only in recovery.
4101 Description: ASSERTION(req->rq_type != LI_POISON) failed
4102 Details : imp_lock should be held while iterating over imp_sending_list for
4103 prevent destroy request after get timeout in ptlrpc_queue_wait.
4107 Description: replay-single.sh test 52 fails
4108 Details : A lock's skiplist need to be cleanup when it being unlinked
4109 from its resource list.
4114 Description: Short directio read returns full requested size rather than
4116 Details : Direct I/O operations should return actual amount of bytes
4117 transferred rather than requested size.
4119 Severity : enhancement
4121 Description: metadata RPC reduction (e.g. for rm performance)
4122 Details : decrease the amount of synchronous RPC between clients and servers
4123 by canceling conflicing lock before the operation on the client
4124 and packing thier handles into the main operation RPC to server.
4126 Severity : enhancement
4128 Description: add #ifdef HAVE_KERNEL_CONFIG_H
4129 Details : kernels from 2.6.19 not need include linux/config.h, but add
4130 include linux/autoconf.h in commpiler command line.
4132 Severity : enhancement
4134 Description: patchless client support for 2.6.22 kernel
4135 Details : 2.6.22 has only one visble change, SLAB_CTOR_* constants is
4136 removed. In this case we need drop using os depended interface to
4137 kmem_cache and use cfs_mem_cache API.
4142 Description: fix mal-formatted messages
4143 Details : fix some mal-formatted DEBUG_REQ and LCONSOLE_ERROR_MSG messages
4147 Frequency : always in liblustre
4148 Description: wrong IS_ERR implementation in liblustre.h
4149 Details : fix IS_ERR implementation in liblustre.h for right detect errors.
4154 Description: Correct condition for output debug message.
4155 Details : inode i_nlink equal zero is not enough for output message about
4156 disk corruption, i_ctime and i_mode should be also checked.
4160 Frequency : always in patchless client
4161 Description: add configure check for truncate_complete_page
4162 Details : improve checks for exported symbols. This allow run check without
4163 sources, but with Module.symvers shipped with kernel distribution.
4164 add check for truncate_complete_page used by patchless client.
4168 Description: sanity.sh test_77h fails with "test_77h file compare failed"
4169 Details : test_77h uses a file which was messed by other test case.
4173 Description: Not Check whether lov_tgts is NULL in some lov functions
4174 Details : Checking whether lov_tgts is NULL in some functions.
4178 Description: replace obdo_alloc() with OBDO_ALLOC macro
4179 Details : nothing special is done in obdo_alloc() function, and for
4180 debugging purpose, it needs to be replaced with macros.
4184 Description: bad return value and errno from fcntl call
4185 Details : In liblustre API, errno should be a negative value if error
4190 Description: ptlrpc_check_set() LBUG
4191 Details : In case of positive reply from server and failed client bulk
4192 callback after bulk transfer shouldn't LBUG, but process this
4193 request as erroneous.
4195 Severity : enhancement
4197 Description: VFS operations stats tool.
4198 Details : Tool which collects stats by tracking value written in pid,
4199 ppid, gid and uses llstat to generate output to plot graph using
4201 Updated lustre/utils/Makefile.am
4202 Added lustre/utils/ltrack_stats.c
4204 Severity : enhancement
4206 Description: 2.6.18 server support (lustre 1.6.1)
4207 Details : Support for 2.6.18 kernels on the server side.
4212 Description: ASSERTION(imp->imp_conn_current) failed
4213 Details : an assertion failure is hit if a client node boots and attempts to
4214 mount a lustre filesystem faster than RECONNECT_INTERVAL seconds.
4217 Frequency : only for i686
4219 Description: 1.4.11 RC1 build fails for RHEL 4, i686
4220 Details : Fixed config variable for build.
4225 Description: Updated patchess for new RHEL4 kernel
4226 Details : Updated patch inode-nr_unused-2.6.9-rhel4.patch
4227 Updated patch jbd-stats-2.6.9.patch
4228 Updated patch qsnet-rhel4-2.6.patch
4229 Updated patch quota-deadlock-on-pagelock-core.patch
4230 Updated patch vfs_intent-2.6-rhel4.patch
4231 Updated patch vfs_races-2.6-rhel4.patch
4232 Updated series file 2.6-rhel4-titech.series
4233 Updated series file 2.6-rhel4.series
4234 Updated kernel config files
4239 Description: lquota slave complains LBUG when reconnecting with mds
4241 Details : quota slave depends qctxt->lqc_import to send its quota request.
4242 This pointer will be invalid if mds did failover or broke its
4243 connect to osts, which leads to LBUG.
4246 Frequency : when qunit size is too small(less than 20M)
4248 Description: write is stopped by improper -EDQUOT
4249 Details : If the master is busy and qunit size is small enough(let's say 1M),
4250 the slave can not get quota from master on time, which will lead
4251 slave to trigger a -EQUOTA to client.
4256 Description: Deadlock during metadata tests
4257 Details : in prune_dir_dentries(), shrink_dcache_parent() should not be
4258 called with the per-dentry lock held.
4261 Frequency : SLES9 only
4263 Description: Lustre patched kernel for SLES9 SP3 has NR_CPUS set to 8
4264 Details : set CONFIG_NR_CPUS to 128 instead of 8.
4266 Severity : enhancement
4268 Description: remove fs_prep_san_write operation and related patches
4269 Details : remove the ext3-san-jdike patches which are no longer useful.
4274 Description: LDISKFS-fs error (device sdc): ldiskfs_free_blocks
4275 Details : a disk corruption can cause the mballoc code to assert on a
4276 double free or other extent corruptions. Handle these with
4277 ext3_error() instead of assertions.
4280 Frequency : only with mballoc3 code and deep extent trees
4282 Description: ldiskfs_ext_search_right: bad header in inode: unexpected eh_depth
4283 Details : a wrong check of extent headers in ldiskfs_ext_search_right()
4284 can cause the filesystem to be remounted read-only.
4289 Description: lfsck built against 1.4.x cannot run against 1.6.0 lustre
4290 Details : the definition for OBD_IOC_GETNAME changed in 1.6.0. One of the
4291 few external users of this ioctl number is lfsck's call to
4292 llapi_lov_get_uuids() and this caused lfsck to fail at startup.
4293 Add the old ioctl number to the handler so both old and new
4298 Description: parallel lock callbacks
4299 Details : Instead of sending blocking and completion callbacks as separated
4300 requests, adding them to a set and sending in parallel.
4304 Description: Disable most debugging by default
4305 Details : To improve performance, disable most logging (for debug purposes)
4306 by default. VFSTRACE, RPCTRACE, and DLMTRACE are now off by
4307 default, and HA includes fewer messages.
4311 Frequency : only run on patchless client.
4312 Description: use do_facet on sanity.sh for test handling recoverables errors
4313 Details : use do_facet instead of direct use sysctl for set fail_loc on OST
4317 Description: hash tables for lists of nids, connections and uuids
4318 Details : Hash tables noticeably help when a lot of clients connect to a
4319 server, to faster identify duplicate connections or reconnects,
4320 also to faster find export to evict in manual eviction case.
4324 Description: Sometimes, when the server evict a client, and the client will
4325 not be evicted as soon as possible.
4326 Details : In enqueue req, the error was returned by intent, instead of
4327 rq_status which make ptlrpc layer not detect this error, and
4328 does not evict the client. So enqueue error should be returned
4332 Frequency : only at startup
4334 Description: Delay client connections to MDT until fist MDT->OST connect
4335 Details : If a client tried to create a new file before the MDT had
4336 connected to any OSTs, the create would return EIO. Now
4337 the client will simply block until the MDT connects to the
4338 first OST and the create can succeed.
4341 Frequency : only for SLES9
4343 Description: Routinely utilize latest Quadrics drivers in CFS releases
4344 Details : Update patch qsnet-suse-2.6.patch.
4347 Frequency : only for sles10
4349 Description: Update patches for SLES 10 SP1 kernel.
4350 Details : Update the patch vfs_intent-2.6-sles10.patch.
4355 Description: Routinely utilize latest Quadrics drivers in CFS releases
4356 Details : Update patch qsnet-rhel4-2.6.patch.
4359 Frequency : at statup only
4361 Description: mds_lov_synchronize race leads to various problems
4362 Details : simultaneous MDT->OST connections at startup can cause the
4363 sync to abort, leaving the OSC in a bad state.
4368 Description: Using wrong pointer in osc_brw_prep_request
4369 Details : Access to array[-1] can produce panic if kernel compiled with
4370 CONFIG_PAGE_ALLOC enabled
4372 Severity : enhancement
4374 Description: Async OSC create to avoid the blocking unnecessarily.
4375 Details : If a OST has no remain object, system will block on the creating
4376 when need to create a new object on this OST. Now, ways use
4377 pre-created objects when available, instead of blocking on an
4378 empty osc while others are not empty. If we must block, we block
4379 for the shortest possible period of time.
4383 Frequency : only in recovery
4384 Description: Mark OST as early accessible if his start SYNC.
4385 Details : osc_precreate return flag early accessible if oscc marked as
4386 OSCC_FLAG_SYNC_IN_PROGRESS.
4391 Description: Sometimes precreate code can triger create object on wrong ost
4392 Details : Wrong protected or not not restored variables aftre precreate loop
4393 can produce creation object on wrong ost.
4396 Frequency : oss recovery
4398 Description: llog_commit_thread cleanup should sync with llog_commit_thread
4400 Details : llog_commit_thread_count should be synced between llog_commit
4401 start and cleanup, so new llog_commit thread should not be started
4402 when llog_commit threads being stopped to avoid accessing some
4405 Severity : enhancement
4407 Description: Add printing inode info into message about error in writepage.
4409 --------------------------------------------------------------------------------
4411 2007-05-03 Cluster File Systems, Inc. <info@clusterfs.com>
4416 Frequency : on some architectures
4418 Description: 1.6 client sometimes fails to mount from a 1.4 MDT
4419 Details : Uninitialized flags sometimes cause configuration commands to
4423 Frequency : patchless clients only
4425 Description: missing __iget() symbol export
4426 Details : The __iget() symbol export is missing. To avoid the need for
4427 this on patchless clients the deathrow inode reaper is turned
4428 off, and we depend on the VM to clean up old inodes. This
4429 dependency was during via the fix for bug 12181.
4434 Description: sanity.sh fail: test_52b
4435 Details : The ll_inode_to_ext_flags() has a glitch which makes MDS return
4436 incorrect inode's flags to client.
4438 --------------------------------------------------------------------------------
4440 2007-04-19 Cluster File Systems, Inc. <info@clusterfs.com>
4442 * CONFIGURATION CHANGE. This version of Lustre WILL NOT
4443 INTEROPERATE with older versions automatically. In many cases a
4444 special upgrade step is needed. Please read the
4445 user documentation before upgrading any part of a 1.4.x system.
4446 * WARNING: Lustre configuration and startup changes are required with
4447 this release. See https://mail.clusterfs.com/wikis/lustre/MountConf
4449 * Support for kernels:
4450 2.4.21-47.0.1.EL (RHEL 3),
4451 2.6.5-7.283 (SLES 9),
4452 2.6.9-42.0.10.EL (RHEL 4),
4453 2.6.12.6 vanilla (kernel.org),
4454 2.6.16.27-0.9 (SLES10)
4455 * Client support for unpatched kernels:
4456 (see https://mail.clusterfs.com/wikis/lustre/PatchlessClient)
4457 2.6.16 - 2.6.19 vanilla (kernel.org),
4458 2.6.9-42.0.8EL (RHEL 4)
4459 * Recommended e2fsprogs version: 1.39.cfs6
4460 * Note that reiserfs quotas are disabled on SLES 10 in this kernel
4463 Severity : enhancement
4465 Description: Async OSC create to avoid the blocking unnecessarily.
4466 Details : If a OST has no remain object, system will block on the creating
4467 when need to create a new object on this OST. Now, ways use
4468 pre-created objects when available, instead of blocking on an
4469 empty osc while others are not empty. If we must block, we block
4470 for the shortest possible period of time.
4472 Severity : enhancement
4474 Description: MountConf
4475 Details : Lustre configuration is now managed via mkfs and mount
4476 commands instead of lmc and lconf. New obd types (MGS, MGC)
4477 are added for dynamic configuration management. See
4478 https://mail.clusterfs.com/wikis/lustre/MountConf for
4481 Severity : enhancement
4483 Description: dynamic OST addition
4484 Details : OSTs can now be added to a live filesystem
4486 Severity : enhancement
4488 Description: startup order invariance
4489 Details : MDTs and OSTs can be started in any order. Clients only
4490 require the MDT to complete startup.
4492 Severity : enhancement
4494 Description: parallel, asynchronous orphan cleanup
4495 Details : orphan cleanup is now performed in separate threads for each
4496 OST, allowing parallel non-blocking operation.
4498 Severity : enhancement
4500 Description: optimized stripe assignment
4501 Details : stripe assignments are now made based on ost space available,
4502 ost previous usage, and OSS previous usage, in order to try
4503 to optimize storage space and networking resources.
4505 Severity : enhancement
4507 Description: Permanently set tunables
4508 Details : All writable /proc/fs/lustre tunables can now be permanently
4509 set on a per-server basis, at mkfs time or on a live
4512 Severity : enhancement
4514 Description: Lustre message v2
4515 Details : Add lustre message format v2.
4517 Severity : enhancement
4519 Description: client OST exclusion list
4520 Details : Clients can be started with a list of OSTs that should be
4521 declared "inactive" for known non-responsive OSTs.
4525 Description: ENOENT returned for valid filehandle during dbench.
4526 Details : Check if a directory has children when invalidating dentries
4527 associated with an inode during lock cancellation. This fixes
4528 an incorrect ENOENT sometimes seen for valid filehandles during
4529 testing with dbench.
4532 Frequency : SFS test only (otherwise harmless)
4534 Description: SPEC SFS validation failure on NFS v2 over lustre.
4535 Details : Changes the blocksize for regular files to be 2x RPC size,
4536 and not depend on stripe size.
4538 Severity : enhancement
4540 Description: fine-grained SMP locking inside DLM
4541 Details : Improve DLM performance on SMP systems by removing the single
4542 per-namespace lock and replace it with per-resource locks.
4544 Severity : enhancement
4546 Description: don't hold multiple extent locks at one time
4547 Details : To avoid client eviction during large writes, locks are not
4548 held on multiple stripes at one time or for very large writes.
4549 Otherwise, clients can block waiting for a lock on a failed OST
4550 while holding locks on other OSTs and be evicted.
4552 Severity : enhancement
4554 Description: Multiple MD RPCs in flight.
4555 Details : Further unserialise some read-only MDT RPCs - learn about intents.
4556 To avoid overly-overloading MDT, introduce a limit on number of
4557 MDT RPCs in flight for a single client and add /proc controls
4558 to adjust this limit.
4560 Severity : enhancement
4562 Description: client read/write statistics
4563 Details : Add client read/write call usage stats for performance
4564 analysis of user processes.
4565 /proc/fs/lustre/llite/*/offset_stats shows non-sequential
4566 file access. extents_stats shows chunk size distribution.
4567 extents_stats_per_process show chunk size distribution per
4570 Severity : enhancement
4572 Description: per-client statistics on server
4573 Details : Add ldlm and operations statistics for each client in
4574 /proc/fs/lustre/mds|obdfilter/*/exports/
4576 Severity : enhancement
4578 Description: improved MDT statistics
4579 Details : Add detailed MDT operations statistics in
4580 /proc/fs/lustre/mds/*/stats
4582 Severity : enhancement
4584 Description: VFS operations stats
4585 Details : Add client VFS call stats, trackable by pid, ppid, or gid
4586 /proc/fs/lustre/llite/*/stats_track_[pid|ppid|gid]
4591 Description: Fix client-side osc byte counters
4592 Details : The osc read/write byte counters in
4593 /proc/fs/lustre/osc/*/stats are now working
4596 Frequency : always as root on SLES
4598 Description: Failure of copying files with lustre special EAs.
4599 Details : Client side always return success for setxattr call for lustre
4600 special xattr (currently only "trusted.lov").
4605 Description: Refcount LNET uuids
4606 Details : The global LNET uuid list grew linearly with every startup;
4607 refcount repeated list entries instead of always adding to
4610 Severity : enhancement
4612 Description: Dynamic service threads
4613 Details : Within a small range, start extra service threads
4614 automatically when the request queue builds up.
4617 Frequency : mixed-endian client/server environments
4619 Description: mixed-endian crashes
4620 Details : The new msg_v2 system had some failures in mixed-endian
4623 Severity : enhancement
4625 Description: Easy OST removal
4626 Details : OSTs can be permanently deactivated with e.g. 'lctl
4627 conf_param lustre-OST0001.osc.active=0'
4629 Severity : enhancement
4631 Description: MGS proc entries
4632 Details : Added basic proc entries for the MGS showing what filesystems
4635 Severity : enhancement
4637 Description: provide MGS failover
4638 Details : Added config lock reacquisition after MGS server failover.
4640 Severity : enhancement
4642 Description: add Linux 2.4 support
4643 Details : Added support for RHEL 2.4.21 kernel for 1.6 servers and clients
4647 Description: a large application tries to do I/O to the same resource and dies
4648 in the middle of it.
4649 Details : Check the req->rq_arrival time after the call to
4650 ost_brw_lock_get(), but before we do anything about
4651 processing it & sending the BULK transfer request. This
4652 should help move old stale pending locks off the queue as
4653 quickly as obd_timeout.
4656 Frequency : when an incorrect nid is specified during startup
4658 Description: ptlrpc connect to non-existant node causes kernel crash
4659 Details : LNET can't be re-entered from an event callback, which
4660 happened when we expire a message after the export has been
4661 cleaned up. Instead, hand the zombie cleanup off to another
4664 Severity : enhancement
4666 Description: plain/inodebits lock performance improvement
4667 Details : Grouping plain/inodebits in granted list by their request modes
4668 and bits policy, thus improving the performance of search through
4672 Frequency : only if OST filesystem is corrupted
4674 Description: client incorrectly hits assertion in ptlrpc_replay_req()
4675 Details : for a short time RPCs with bulk IO are in the replay list,
4676 but replay of bulk IOs is unimplemented. If the OST filesystem
4677 is corrupted due to disk cache incoherency and then replay is
4678 started it is possible to trip an assertion. Avoid putting
4679 committed RPCs into the replay list at all to avoid this issue.
4682 Frequency : liblustre (e.g. catamount) on a large cluster with >= 8 OSTs/OSS
4684 Description: System hang on startup
4685 Details : This bug allowed the liblustre (e.g. catamount) client to
4686 return to the app before handling all startup RPCs. This
4687 could leave the node unresponsive to lustre network traffic
4688 and manifested as a server ptllnd timeout.
4690 Severity : enhancement
4692 Description: Add "/proc/sys/lustre/debug_peer_on_timeout"
4693 Details : liblustre envirable: LIBLUSTRE_DEBUG_PEER_ON_TIMEOUT
4694 boolean to control whether to print peer debug info when a
4695 client's RPC times out.
4698 Frequency : only for kernels with patches from Lustre below 1.4.3
4700 Description: Remove old rdonly API
4701 Details : Remove old rdonly API which unused from at least lustre 1.4.3
4704 Frequency : only for devices with external journals
4706 Description: Set external device read-only also
4707 Details : During a commanded failover stop, we set the disk device
4708 read-only while the server shuts down. We now also set any
4709 external journal device read-only at the same time.
4712 Frequency : when upgrading from 1.4 while trying to change parameters
4714 Description: The wrong (new) MDC name was used when setting parameters for
4715 upgraded MDT's. Also allows changing of OSC (and MDC)
4716 parameters if --writeconf is specified at tunefs upgrade time.
4719 Frequency : when setting specific ost indicies
4721 Description: QOS code breaks on skipped indicies
4722 Details : Add checks for missing OST indicies in the QOS code, so OSTs
4723 created with --index need not be sequential.
4725 Severity : enhancement
4727 Description: Add uninit_groups feature to ldiskfs2 to speed up e2fsck
4728 Details : The uninit_groups feature works in conjunction with the kernel
4729 filesystem code (ldiskfs2 only) and e2fsprogs-1.39-cfs6 to speed
4730 up the pass1 processing of e2fsck. This is a read-only feature
4731 in ldiskfs2 only, so older kernels and current ldiskfs cannot
4732 mount filesystems that have had this feature enabled.
4734 Severity : enhancement
4736 Description: Improve multi-block allocation algorithm to avoid fragmentation
4737 Details : The mballoc3 code (ldiskfs2 only) adds new mechanisms to improve
4738 allocation locality and avoid filesystem fragmentation.
4740 ------------------------------------------------------------------------------
4742 2007-04-01 Cluster File Systems, Inc. <info@clusterfs.com>
4744 * Support for kernels:
4745 2.4.21-47.0.1.EL (RHEL 3)
4746 2.6.5-7.283 (SLES 9)
4747 2.6.9-42.0.10.EL (RHEL 4)
4748 2.6.12.6 vanilla (kernel.org)
4749 2.6.16.27-0.9 (SLES 10)
4750 * Recommended e2fsprogs version: 1.39.cfs5
4752 * Note that reiserfs quotas are disabled on SLES 10 in this kernel
4756 Frequency : occasional, depends on client load and configuration
4757 Bugzilla : 12181, 12203
4758 Description: data loss for recently-modified files
4760 Details : In some cases it is possible that recently written or created
4761 files may not be written to disk in a timely manner (this should
4762 normally be within 30s unless client IO load is very high).
4763 The problem appears as zero-length files or files that are a
4764 multiple of 1MB in size after a client crash or client eviction
4765 that are missing data at the end of the file.
4767 This problem is more likely to be hit on clients where files are
4768 repeatedly created and unlinked in the same directory, clients
4769 have a large amount of RAM, have many CPUs, the filesystem has
4770 many OSTs, the clients are rebooted frequently, and/or the files
4771 are not accessed by other nodes after being written.
4773 The presence of the problem can be detected by looking at
4774 /proc/sys/fs/inode-state. If the first number (nr_inodes) is
4775 smaller than the second (nr_unused) then dirty files will not
4776 be flushed automatically to disk. "sync; sleep 10" should be
4777 run several times on the node before unmounting it to update
4778 Lustre (this is also safe to run on nodes without this problem).
4780 There is also a related kernel bug in the RHEL4 4 2.6.9 kernel
4781 that can cause this same problem, so customers using that kernel
4782 also need to update the kernel in addition to Lustre. In order
4783 to properly fix this bug, the RHEL3 2.4.21 kernel is also updated.
4785 It is normal that files written just before a client crash (less
4786 than 30s) may not yet have been flushed to disk, even for local
4790 Frequency : frequent on thin XT3 nodes
4792 Description: UUID collision on thin XT3 Linux nodes
4793 Details : UUIDs on Compute Node Linux XT3 nodes were not generated
4794 randomly, since we relied on an insufficiently-seeded PRNG.
4799 Description: OSS hangs after "All ost request buffers busy"
4800 Details : A deadlock between quota and journal operations caused OSS
4801 hangs after printing "All ost request buffers busy."
4804 Frequency : always on liblustre builds
4806 Description: Cleanup compiler warnings on liblustre
4809 Frequency : always on liblustre builds on XT3
4811 Description: LC_CONFIG_CDEBUG don't run while build liblustre on XT3.
4815 Description: Addition of EXT3_FEATURE_RO_COMPAT_DIR_NLINKS flag for
4816 > 32000 subdirectories
4817 Details : Add EXT3_FEATURE_RO_COMPAT_DIR_NLINK flag to
4818 EXT3_FEATURE_RO_COMPAT_SUPP. This flag will be set whenever
4819 subdirectory count crosses 32000. This will aid e2fsck to
4820 correctly handle more than 32000 subdirectories.
4823 Frequency : liblustre (e.g. catamount) on a large cluster with >= 8 OSTs/OSS
4825 Description: System hang on startup
4826 Details : This bug allowed the liblustre (e.g. catamount) client to
4827 return to the app before handling all startup RPCs. This
4828 could leave the node unresponsive to lustre network traffic
4829 and manifested as a server ptllnd timeout.
4831 Severity : enhancement
4833 Description: Add "/proc/sys/lustre/debug_peer_on_timeout"
4834 (liblustre envirable: LIBLUSTRE_DEBUG_PEER_ON_TIMEOUT)
4835 boolean to control whether to print peer debug info when a
4836 client's RPC times out.
4841 Description: make O_SYNC working on 2.6 kernels
4842 Details : 2.6 kernels use different method for mark pages for write,
4843 so need add a code to lustre for O_SYNC work.
4848 Description: Failure to close file and release space on NFS
4849 Details : Put inode details into lock acquired in ll_intent_file_open.
4850 Use mdc_intent_lock in ll_intent_open to properly
4851 detect all kind of errors unhandled by mdc_enqueue.
4856 Description: proc file read during shutdown sometimes raced obd removal,
4858 Details : Add lock to prevent obd access after proc file removal.
4861 Frequency : Only for files larger than 4GB on 32-bit clients.
4863 Description: improperly doing page alignment of locks
4864 Details : Modify lustre core code to use CFS_PAGE_* defines instead of
4865 PAGE_*. Make CFS_PAGE_MASK a 64-bit mask.
4870 Description: RPCs being resent when they shouldn't be
4871 Details : Some RPCs that should not be resent are being resent. This
4872 can cause inconsistencies in the RPC state machine. Do not
4873 resend such requests.
4876 Frequency : rare, only with NFS export
4878 Description: Crash on NFS re-export node
4879 Details : under very unusual load conditions an assertion is hit in
4880 ll_intent_file_open()
4883 Frequency : only if OST filesystem is corrupted
4885 Description: client incorrectly hits assertion in ptlrpc_replay_req()
4886 Details : for a short time RPCs with bulk IO are in the replay list,
4887 but replay of bulk IOs is unimplemented. If the OST filesystem
4888 is corrupted due to disk cache incoherency and then replay is
4889 started it is possible to trip an assertion. Avoid putting
4890 committed RPCs into the replay list at all to avoid this issue.
4895 Description: large O_DIRECT requests fail under memory pressure/fragmentation
4896 Details : Large single O_DIRECT read and write calls can fail to allocate
4897 a sufficiently large buffer to process the request. In case of
4898 allocation failure the allocation is retried with a smaller
4899 buffer and broken into smaller requests.
4901 Severity : enhancement
4903 Description: Add -o localflock option to simulate old noflock behaviour.
4904 Details : This will achieve local-only flock/fcntl locks coherentness.
4909 Description: versioning check is incomplete
4910 Details : Checking the version difference of client vs. server, report
4911 error if the gap is too big.
4916 Description: add support PG_writeback bit
4917 Details : add support for PG_writeback bit for Lustre, for more carefull
4918 work with page cache in 2.6 kernel. This also fix some deadlocks
4919 and remove hack for work O_SYNC with 2.6 kernel.
4921 Severity : enhancement
4923 Description: Add uninit_groups feature to ldiskfs2 to speed up e2fsck
4924 Details : The uninit_groups feature works in conjunction with the kernel
4925 filesystem code (ldiskfs2 only) and e2fsprogs-1.39-cfs6 to speed
4926 up the pass1 processing of e2fsck. This is a read-only feature
4927 in ldiskfs2 only, so older kernels and current ldiskfs cannot
4928 mount filesystems that have had this feature enabled.
4930 Severity : enhancement
4932 Description: Improve multi-block allocation algorithm to avoid fragmentation
4933 Details : The mballoc3 code (ldiskfs2 only) adds new mechanisms to improve
4934 allocation locality and avoid filesystem fragmentation.
4936 ------------------------------------------------------------------------------
4938 2007-02-09 Cluster File Systems, Inc. <info@clusterfs.com>
4940 * Support for kernels:
4941 2.6.9-42.0.3.EL (RHEL 4)
4942 2.6.5-7.276 (SLES 9)
4943 2.4.21-47.0.1.EL (RHEL 3)
4944 2.6.12.6 vanilla (kernel.org)
4945 2.6.16.21-0.8 (SLES10)
4946 * Recommended e2fsprogs version: 1.39.cfs2-0
4948 * The backwards-compatible /proc/sys/portals symlink has been removed
4949 in this release. Before upgrading, please ensure that you change
4950 any configuration scripts or /etc/sysctl.conf files that access
4951 /proc/sys/portals/* or sysctl portals.* to use the corresponding
4952 entry in /proc/sys/lnet or sysctl lnet.*. This change can be made
4953 in advance of the upgrade on any system running Lustre 1.4.6 or
4954 newer, since /proc/sys/lnet was added in that version.
4955 * Note that reiserfs quotas are disabled on SLES 10 in this kernel
4959 Frequency : only when quota is used
4961 Description: avoid scanning export list for quota master
4962 Details : Change the algorithms to avoid scanning export list in order
4963 to improve the efficiency.
4966 Frequency : MDS failover only, very rarely
4968 Description: "went back in time" messages on mds failover
4969 Details : The greatest transno may be lost when the current operation
4970 finishes with an error (transno==0) and the client's last_rcvd
4971 record is over-written. Save the greatest transno in the
4972 mds_last_transno for this case.
4975 Frequency : always for specific kernels and striping counts
4977 Description: client may get "Matching packet too big" without ACL support
4978 Details : Clients compiled without CONFIG_FS_POSIX_ACL get an error message
4979 when trying to access files in certain configurations. The
4980 clients should in fact be denied when mounting because they do
4981 not understand ACLs.
4984 Frequency : Cray XT3 with more than 4000 clients and multiple jobs
4986 Description: many clients connecting with IO in progress causes connect timeouts
4987 Details : Avoid synchronous journal commits to avoid delays caused by many
4988 clients connecting/disconnecting when bulk IO is in progress.
4989 Queue liblustre connect requests on OST_REQUEST_PORTAL instead of
4990 OST_IO_PORTAL to avoid delays behind potentially many pending
4994 Frequency : occasionally with multiple writers to a single file
4996 Description: shared writes to file may result in wrong size reported by stat()
4997 Details : Allow growing of kms when extent lock is cancelled
5000 Frequency : always with random mmap IO to multi-striped file
5002 Description: mmap write might be lost if we are writing to a 'hole' in stripe
5003 Details : Only if the hole is at the end of OST object so that kms is too
5004 small. Fix is to increase kms accordingly in ll_nopage.
5007 Frequency : rare, only if OST filesystem is inconsistent with MDS filesystem
5009 Description: writes to a missing object would leak memory on the OST
5010 Details : If there is an inconsistency between the MDS and OST filesystems,
5011 such that the MDS references an object that doesn't exist, writes
5012 to that object will leak memory due to incorrect cleanup in the
5013 error handling path, eventually running out of memory on the OST.
5018 Description: Creating too long symlink causes lustre errors
5019 Details : Check symlink and name lengths before sending requests to MDS.
5022 Frequency : only if flock is enabled (not on by default)
5024 Description: posix locks not released on fd closure on 2.6.9+
5025 Details : We failed to add posix locks to list of inode locks on 2.6.9+
5026 kernels, this caused such locks not to be released on fd close and
5027 then assertions on fs unmount about still used locks.
5030 Frequency : MDS failover only, very rarely
5032 Description: clients may get ASSERTION(granted_lock != NULL)
5033 Details : When request was taking a long time, and a client was resending
5034 a getattr by name lock request. The were multiple lock requests
5035 with the same client lock handle and
5036 mds_getattr_name->fixup_handle_for_resent_request found one of the
5037 lock handles but later failed with ASSERTION(granted_lock != NULL).
5042 Description: handle->h_buffer_credits > 0, assertion failure
5043 Details : h_buffer_credits is zero after truncate, causing assertion
5044 failure. This patch extends the transaction or creates a new
5048 Frequency : NFS re-export or patchless client
5049 Bugzilla : 11179, 10796
5050 Description: Crash on NFS re-export node (__d_move)
5051 Details : We do not want to hash the dentry if we don't have a lock.
5052 But if this dentry is later used in d_move, we'd hit uninitialised
5053 list head d_hash, so we just do this to init d_hash field but
5054 leave dentry unhashed.
5057 Frequency : NFS re-export or patchless client
5059 Description: NFS exports has problem with symbolic link
5060 Details : lustre client didn't properly install dentry when re-exported
5061 to NFS or running patchless client.
5064 Frequency : NFS re-export or patchless client
5066 Description: Various nfs/patchless fixes.
5067 Details : fixes reuse disconected alias for lookup process - this fixes
5068 warning "find_exported_dentry: npd != pd",
5069 fix permission error with open files at nfs.
5070 fix apply umask when do revalidate.
5073 Frequency : occasional
5075 Description: Crash on NFS re-export node
5076 Details : calling clear_page() on the wrong pointer triggered oops in
5077 generic_mapping_read().
5080 Frequency : rarely, using O_DIRECT IO
5082 Description: unaligned directio crashes client with LASSERT
5083 Details : check for unaligned buffers before trying any requests.
5086 Frequency : rarely, using CFS RAID5 patches in non-standard kernel series
5088 Description: stale data returned from RAID cache
5089 Details : If only a small amount of IO is done to the RAID device before
5090 reading it again it is possible to get stale data from the RAID
5091 cache instead of reading it from disk.
5094 Frequency : always for sles10 kernel
5096 Description: sles10 support
5097 Details : ll_follow_link: compile fixes and using of nd_set_link
5098 under newer kernels.
5101 Frequency : depends on arch, kernel and compiler version, always on sles10
5104 Description: recursive or deep enough symlinks cause stack overflow
5105 Details : getting rid of large stack-allocated variable in
5109 Frequency : depends on hardware
5111 Description: lustre write performance loss in the SLES10 kernel
5112 Details : the performance loss is caused by using of write barriers in the
5113 ext3 code. The SLES10 kernel turns barrier support on by
5114 default. The fix is to undo that change for ldiskfs.
5116 ------------------------------------------------------------------------------
5118 2006-12-09 Cluster File Systems, Inc. <info@clusterfs.com>
5120 * Support for kernels:
5121 2.6.9-42.0.3EL (RHEL 4)
5122 2.6.5-7.276 (SLES 9)
5123 2.4.21-47.0.1.EL (RHEL 3)
5124 2.6.12.6 vanilla (kernel.org)
5128 Frequency : quota enabled and large files being deleted
5130 Description: releasing more than 4GB of quota at once hangs OST
5131 Details : If a user deletes more than 4GB of files on a single OST it
5132 will cause the OST to spin in an infinite loop. Release
5133 quota in < 4GB chunks, or use a 64-bit value for 1.4.7.1+.
5138 Description: statfs data retrieved from /proc may be stale or zero
5139 Details : When reading per-device statfs data from /proc, in the
5140 {kbytes,files}_{total,free,avail} files, it may appear
5141 as zero or be out of date.
5144 Frequency : systems with MD RAID1 external journal devices
5146 Description: lconf's call to blkid is confused by RAID1 journal devices
5147 Details : Use the "blkid -l" flag to locate the MD RAID device instead
5148 of returning all block devices that match the journal UUID.
5151 Frequency : always, for aggregate stripe size over 4GB
5153 Description: "lfs setstripe" fails assertion when setting 4GB+ stripe width
5154 Details : Using "lfs setstripe" to set stripe size * stripe count over 4GB
5155 will fail the kernel with "ASSERTION(lsm->lsm_xfersize != 0)"
5158 Frequency : always if "lfs find" used on a local file/directory
5160 Description: "lfs find" segfaults if used on a local file/directory
5161 Details : The case where a directory component was not specified wasn't
5162 handled correctly. Handle this properly.
5165 Frequency : always on ppc64
5167 Description: the write to an ext3 filesystem mounted with mballoc got stuck
5168 Details : ext3_mb_generate_buddy() uses find_next_bit() which does not
5169 perform endianness conversion.
5172 Frequency : rarely (truncate to non-zero file size after write under load)
5173 Bugzilla : 10730, 10687
5174 Description: Files padded with zeros to next 4K multiple
5175 Details : With filesystems mounted using the "extents" option (2.6 kernels)
5176 it is possible that files that are truncated to a non-zero size
5177 immediately after being written are filled with zero bytes beyond
5178 the truncated size. No file data is lost.
5180 Severity : enhancement
5182 Description: Allow recovery/failover for liblustre clients.
5183 Details : liblustre clients were unaware of failover configurations until
5186 Severity : enhancement
5188 Description: user file locks should fail when not mounting with flock option
5189 Details : Set up an error-returning stub in ll_file_operations.lock field
5190 to prevent incorrect behaviour when client is mounted without
5191 flock option. Also, set up properly f_op->flock field for
5195 Frequency : always on ia64
5197 Description: "lfs df" loops on printing out MDS statfs information
5198 Details : The obd_ioctl_data was not initialized and in some systems
5199 this caused a failure during the ioctl that did not return
5200 an error. Initialize the struct and return an error on failure.
5203 Frequency : SLES 9 only
5205 Description: Error of copying files with lustre special EAs as root
5206 Details : Client side always return success for setxattr call for lustre
5207 special xattr (currently only "trusted.lov").
5210 Frequency : rarely on clusters with both ia64+i386 clients
5212 Description: ia64+i686 clients doing shared IO on the same file may LBUG
5213 Details : In rare cases when both ia64+i686 (or other mixed-PAGE_SIZE)
5214 clients are doing concurrent writes to the same file it is
5215 possible that the ia64 clients may LASSERT because the OST
5216 extent locks are not PAGE_SIZE aligned. Ensure that grown
5217 locks are always aligned on the request boundary.
5220 Frequency : specific use, occasional
5222 Description: Overwriting in use executable truncates on-disk binary image
5223 Details : If one node attempts to overwrite an executable in use by
5224 another node, we now correctly return ETXTBSY instead of
5225 truncating the file.
5227 Severity : enhancement
5229 Description: Async OSC create to avoid the blocking unnecessarily.
5230 Details : If a OST has no remain object, system will block on the creating
5231 when need to create a new object on this OST. Now, ways use
5232 pre-created objects when available, instead of blocking on an
5233 empty osc while others are not empty. If we must block, we block
5234 for the shortest possible period of time.
5239 Description: chmod on Lustre root is propagated to other clients
5240 Details : Re-validate root's dentry in ll_lookup_it to avoid having it
5241 invalid by the follow_mount time.
5244 Frequency : liblustre clients only
5246 Description: Race in 'instant cancel' lock handling could lead to such locks
5247 never to be granted in case of SMP MDS
5248 Details : Do not destroy not yet granted but cbpending locks in
5252 Frequency : replay/resend of open
5254 Description: non null lock assetion failure in mds_intent_policy
5255 Details : Trying to replay/resend lockless open requests resulted in
5256 mds_open() returning 0 with no lock. Now it sets a flag if
5257 a lock is going to be returned.
5259 Severity : enhancement
5261 Description: Checksum enhancements
5262 Details : New checksum enhancements allow for resending RPCs that failed
5265 Severity : enhancement
5267 Description: Tunables on number of dirty pages in cacche
5268 Details : Allow to set limit on number of dirty pages cached.
5273 Description: client crash on unmount - lock still has references
5274 Details : In some error handling cases it was possible to leak a lock
5275 reference on a client while accessing a file. This was not
5276 harmful to the client during operation, but would cause the
5277 client to crash when the filesystem is unmounted.
5280 Frequency : specific case, rare
5282 Description: ETXTBSY on mds though file not in use by client
5283 Details : ETXTBSY is no longer incorrectly returned when attempting to
5284 chmod or chown a directory that the user previously tried to
5285 execute or a currently-executing binary.
5288 Frequency : extremely rare except on liblustre-based clients
5290 Description: Lustre space not freed when files are deleted
5291 Details : Clean up open-unlinked files after client eviction. Previously
5292 the unlink was skipped and the files remained as orphans.
5297 Description: OST failure "would be an LBUG" in waiting_locks_callback()
5298 Details : In some cases it was possible to send a blocking callback to
5299 a client doing a glimpse, even though that client didn't get
5300 a lock granted. When the glimpse lock is cancelled on the OST
5301 the freed lock is left on the waiting list and corrupted the list.
5304 Frequency : all core dumps
5306 Description: Broke core dumps to lustre
5307 Details : Negative dentry may be unhashed if parent does not have UPDATE
5308 lock, but some callers, e.g. do_coredump, expect dentry to be
5309 hashed after successful create, hash it in ll_create_it.
5311 ------------------------------------------------------------------------------
5313 2006-09-13 Cluster File Systems, Inc. <info@clusterfs.com>
5315 * Support for kernels:
5316 2.6.9-42.0.2.EL (RHEL 4)
5317 2.6.5-7.276 (SLES 9)
5318 2.4.21-40.EL (RHEL 3)
5319 2.6.12.6 vanilla (kernel.org)
5323 Frequency : always on RHEL 3
5325 Description: Number of open files grows over time
5326 Details : The number of open files grows over time, whether or not
5327 Lustre is started. This was due to a filp leak introduced
5328 by one of our kernel patches.
5330 ------------------------------------------------------------------------------
5332 2006-08-20 Cluster File Systems, Inc. <info@clusterfs.com>
5334 * Support for kernels:
5335 2.6.9-42.EL (RHEL 4)
5336 2.6.5-7.267 (SLES 9)
5337 2.4.21-40.EL (RHEL 3)
5338 2.6.12.6 vanilla (kernel.org)
5343 Bugzilla : 5719, 9635, 9792, 9684
5344 Description: OST (or MDS) trips assertions in (re)connection under heavy load
5345 Details : If a server is under heavy load and cannot reply to new
5346 connection requests before the client resends the (re)connect,
5347 the connection handling code can behave badly if two service
5348 threads are concurrently handing separate (re)connections from
5349 the same client. Add better locking to the connection handling
5350 code, and ensure that only a single connection will be processed
5351 for a given client UUID, even if the lock is dropped.
5353 Severity : enhancement
5355 Description: add TCP zero-copy support to kernel
5356 Details : Add support to the kernel TCP stack to allow zero-copy bulk
5357 sends if the hardware supports scatter-gather and checksumming.
5358 This allows socklnd to do client-write and server-read more
5359 efficiently and reduce CPU utilization from skbuf copying.
5362 Frequency : only if NFS exporting from client
5364 Description: NULL pointer deref in ll_iocontrol() if chattr mknod file
5365 Details : If setting attributes on a file created under NFS that had
5366 never been opened it would be possible to oops the client
5367 if the file had no objects.
5371 Bugzilla : 9326, 10402, 10897
5372 Description: client crash in ptlrpcd_wake() thread when sending async RPC
5373 Details : It is possible that ptlrpcd_wake() dereferences a freed async
5374 RPC. In rare cases the ptlrpcd thread alread processed the RPC
5375 before ptlrpcd_wake() was called and the request was freed.
5378 Frequency : always for liblustre
5380 Description: liblustre client does MDS+OSTs setattr RPC for each write
5381 Details : When doing a write from a liblustre client, the client
5382 incorrectly issued an RPC to the MDS and each OST the file was
5383 striped over in order to update the timestamps. When writing
5384 with small chunks and many clients this could overwhelm the MDS
5385 with RPCs. In all cases it would slow down the write because
5386 these RPCs are unnecessary.
5388 Severity : enhancement
5390 Description: allow number of MDS service threads to be changed at module load
5391 Details : It is now possible to change the number of MDS service threads
5392 running. Adding "options mds mds_num_threads={N}" to the MDS's
5393 /etc/modprobe.conf will set the number of threads for the next
5394 time Lustre is restarted (assuming the "mds" module is also
5395 reloaded at that time). The default number of threads will
5396 stay the same, 32 for most systems.
5401 Description: OST crash if filesystem is unformatted or corrupt
5402 Details : If an OST is started on a device that has never been formatted
5403 or if the filesystem is corrupt and cannot even mount then the
5404 error handling cleanup routines would dereference a NULL pointer.
5409 Description: NULL pointer deref in llap_from_page.
5410 Details : get_cache_page_nowait can return a page with NULL (or otherwise
5411 incorrect) mapping if the page was truncated/reclaimed while it was
5412 searched for. Check for this condition and skip such pages when
5413 doing readahead. Introduce extra check to llap_from_page() to
5414 verify page->mapping->host is non-NULL (so page is not anonymous).
5417 Frequency : Sometimes when using sys_sendfile
5419 Description: "page not covered by a lock" warnings from ll_readpage
5420 Details : sendfile called ll_readpage without right page locks present.
5421 Now we introduced ll_file_sendfile that does necessary locking
5422 around call to generic_file_sendfile() much like we do in
5426 Frequency : with certain MDS communication failures at client mount time
5428 Description: NULL pointer deref after failed client mount
5429 Details : a client connection request may delayed by the network layer
5430 and not be sent until after the PTLRPC layer has timed out the
5431 request. If the client fails the mount immediately it will try
5432 to clean up before the network times out the request. Add a
5433 reference from the request import to the obd device and delay
5434 the cleanup until the network drops the request.
5437 Frequency : occasionally during client (re)connect
5439 Description: assertion failure during client (re)connect
5440 Details : processing a client connection request may be delayed by the
5441 client or server longer than the client connect timeout. This
5442 causes the client to resend the connection request. If the
5443 original connection request is replied in this interval, the
5444 client may trip an assertion failure in ptlrpc_connect_interpret()
5445 which thought it would be the only running connect process.
5448 Frequency : only with obd_echo servers and clients that are rebooted
5450 Description: kernel BUG accessing uninitialized data structure
5451 Details : When running an obd_echo server it did not start the ping_evictor
5452 thread, and when a client was evicted an uninitialized data
5453 structure was accessed. Start the ping_evictor in the RPC
5454 service startup instead of the OBD startup.
5456 Severity : enhancement
5457 Bugzilla : 10193 (patchless)
5458 Description: Remove dependency on various unexported kernel interfaces.
5459 Details : No longer need reparent_to_init, exit_mm, exit_files,
5460 sock_getsockopt, filemap_populate, FMODE_EXEC, put_filp.
5463 Frequency : rare (only users of deprecated and unsupported LDAP config)
5465 Description: write_conf for zeroconf mount queried LDAP incorrectly for client
5466 Details : LDAP apparently contains 'lustreName' attributes instead of
5467 'name'. A simple remapping of the name is sufficient.
5470 Frequency : rare (only with non-default dump_on_timeout debug enabled)
5472 Description: waiting_locks_callback trips kernel BUG if client is evicted
5473 Details : Running with the dump_on_timeout debug flag turned on makes
5474 it possible that the waiting_locks_callback() can try to dump
5475 the Lustre kernel debug logs from an interrupt handler. Defer
5476 this log dumping to the expired_lock_main() thread.
5478 Severity : enhancement
5480 Description: Support NFS exporting on 2.6 kernels.
5481 Details : Implement non-rawops metadata methods for NFS server to use without
5482 changing NFS server code.
5485 Frequency : very rare (synthetic metadata workload only)
5487 Description: two racing renames might cause an MDS thread to deadlock
5488 Details : Running the "racer" program may cause one MDS thread to rename
5489 a file from being the source of a rename to being the target of
5490 a rename at exactly the same time that another thread is doing
5491 so, and the second thread has already enqueued these locks after
5492 doing a lookup of the target and is trying to relock them in
5493 order. Ensure that we don't try to re-lock the same resource.
5496 Frequency : only very large systems with liblustre clients
5498 Description: slow eviction of liblustre clients with the "evict_by_nid" RPC
5499 Details : Use asynchronous set_info RPCs to send the "evict_by_nid" to
5500 all OSTs in parallel. This allows the eviction of stale liblustre
5501 clients to proceed much faster than if they were done in series,
5502 and also offers similar improvements for other set_info RPCs.
5507 Description: excessive CPU usage during initial read phase on client
5508 Details : During the initial read phase on a client, it would agressively
5509 retry readahead on the file, consuming too much CPU and impacting
5510 performance (since 1.4.5.8). Improve the readahead algorithm
5511 to avoid this, and also improve some other common cases (read
5512 of small files in particular, where "small" is files smaller than
5513 /proc/fs/lustre/llite/*/max_read_ahead_whole_mb, 2MB by default).
5518 Description: MDS crash when receiving packet with unknown intent.
5519 Details : Do not LBUG in unknown intent case, just return -EFAULT
5521 Severity : enhancement
5522 Bugzilla : 9293, 9385
5523 Description: MDS RPCs are serialised on client. This is unnecessary for some.
5524 Details : Do not serialize getattr (non-intent version) and statfs.
5527 Frequency : occasional, when OST network is overloaded/intermittent
5529 Description: client evicted by OST after bulk IO timeout
5530 Details : If a client sends a bulk IO request (read or write) the OST
5531 may evict the client if it is unresposive to its data GET/PUT
5532 request. This is incorrect if the network is overloaded (takes
5533 too long to transfer the RPC data) or dropped the OST GET/PUT
5534 request. There is no need to evict the client at all, since
5535 the pinger and/or lock callbacks will handle this, and the
5536 client can restart the bulk request.
5539 Frequency : Always when mmapping file with no objects
5541 Description: client crashes when mmapping file with no objects
5542 Details : Check that we actually have objects in a file before doing any
5543 operations on objects in ll_vm_open, ll_vm_close and
5549 Description: Request leak when working with deleted CWD
5550 Details : Introduce advanced request refcount tracking for requests
5551 referenced from lustre intent.
5553 Severity : Enhancement
5555 Description: Cache open file handles on client.
5556 Details : MDS now will return special lock along with openhandle, if
5557 requested and client is allowed to hold openhandle, even if unused,
5558 until such a lock is revoked. Helps NFS a lot, since NFS is opening
5559 closing files for every read/write openration.
5561 Severity : Enhancement
5563 Description: Cache open negative dentries on client when possible.
5564 Details : Guard negative dentries with UPDATE lock on parent dir, drop
5565 negative dentries on lock revocation.
5570 Description: Remounting a client read-only wasn't possible with a zconf mount
5571 Details : It wasn't possible to remount a client read-only with llmount.
5573 Severity : enhancement
5574 Description: Include MPICH 1.2.6 Lustre ADIO interface patch
5575 Details : In lustre/contrib/ or /usr/share/lustre in RPM a patch for
5576 MPICH is included to add Lustre-specific ADIO interfaces.
5577 This is based closely on the UFS ADIO layer and only differs
5578 in file creation, in order to allow the OST striping to be set.
5579 This is user-contributed code and not supported by CFS.
5584 Description: extended inode attributes (immutable, append-only) work improperly
5585 when 2.4 and 2.6 kernels are used on client/server or vice versa
5586 Details : Introduce kernel-independent values for these flags.
5588 Severity : enhancement
5591 Description: Allow fractional MB tunings for lustre in /proc/ filesystem.
5592 Details : Many of the /proc/ tunables can only be tuned at a megabyte
5593 granularity. Now, Fractional MB granularity is be supported,
5594 this is very useful for low memory system.
5596 Severity : enhancement
5598 Description: Getattr by fid
5599 Details : Getting a file attributes by its fid, obtaining UPDATE|LOOKUP
5600 locks, avoids extra getattr rpc requests to MDS, allows '/' to
5601 have locks and avoids getattr rpc requests for it on every stat.
5604 Frequency : Always, for filesystems larger than 2TB
5606 Description: ldiskfs crash at mount for filesystem larger than 2TB with mballoc
5607 Details : Kenrel kmalloc limits allocations to 128kB and this prevents
5608 filesystems larger than 2TB to be mounted with mballoc enabled.
5611 Frequency : Always, for 32-bit kernel without CONFIG_LBD and filesystem > 2TB
5613 Description: filesystem corruption for non-standard kernels and very large OSTs
5614 Details : If a 32-bit kernel is compiled without CONFIG_LBD enabled and a
5615 filesystems larger than 2TB is mounted then the kernel will
5616 silently corrupt the start of the filesystem. CONFIG_LBD is
5617 enabled for all CFS-supported kernels, but the possibility of
5618 this happening with a modified kernel config exists.
5620 Severity : enhancement
5622 Description: add client O_DIRECT support for 2.6 kernels
5623 Details : It is now possible to do O_DIRECT reads and writes to files
5624 in the Lustre client mountpoint on 2.6 kernel clients.
5626 Severity : enhancement
5628 Description: parallel glimpse, setattr, statfs, punch, destroy requests
5629 Details : Sends glimpse, setattr, statfs, punch, destroy requests to OSTs in
5630 parallel, not waiting for response from every OST before sending
5631 a rpc to the next OST.
5636 Description: setattr vs write race when updating file timestamps
5637 Details : Client processes that update a file timestamp into the past
5638 right after writing to the file (e.g. tar) it is possible that
5639 the updated file modification time can be reset to the current
5640 time due to a race between processing the setattr and write RPC.
5642 Severity : enhancement
5644 Description: Bring 'lfs find' closer in line with regular Linux find.
5645 Details : lfs find util supports -atime, -mtime, -ctime, -maxdepth, -print,
5646 -print0 options and obtains all the needed info through the lustre
5649 Severity : enhancement
5651 Description: support up to 1024 configured devices on one node
5652 Details : change obd_dev array from statically allocated to dynamically
5653 allocated structs as they are first used to reduce memory usage
5658 Description: Flush dirty partially truncated pages during truncate
5659 Details : Immediatelly flush partially truncated pages in filter_setattr,
5660 this way we completely avoid having any pages in page cache on OST
5661 and can retire ugly workarounds during writes to flush such pages.
5666 Description: i_sem vs transaction deadlock in mds_obd_destroy during unlink.
5667 Details : protect inode from truncation within vfs_unlink() context
5668 just take a reference before calling vfs_unlink() and release it
5669 when parent's i_sem is free.
5672 Frequency : always, if extents are used on OSTs
5674 Description: index ei_leaf_hi (48-bit extension) is not zeroed in extent index
5675 Details : OSTs using the extents format would not zero the high 16 bits of
5676 the index physical block number. This is not a problem for any
5677 OST filesystems smaller than 16TB, and no kernels support ext3
5678 filesystems larger than 16TB yet. This is fixed in 1.4.7 (all
5679 new/modified files) and can be fixed for existing filesystems
5680 with e2fsprogs-1.39-cfs1.
5685 Description: import connection selection may be incorrect if timer wraps
5686 Details : Using a 32-bit jiffies timer with HZ=1000 may cause backup
5687 import connections to be ignored if the 32-bit jiffies counter
5688 wraps. Use a 64-bit jiffies counter.
5691 Frequency : during server recovery
5693 Description: crash after server is denying duplicate export
5694 Details : If clients are resending connect requests to the server, the
5695 server refuses to allow a client to connect multiple times.
5696 Fixed a bug in the handling of this case.
5699 Frequency : very large clusters immediately after boot
5701 Description: LNET request buffers exhausted under heavy short-term load
5702 Details : If a large number of client requests are generated on a service
5703 that has previously never seen so many requests it is possible
5704 that the request buffer growth cannot keep up with the spike in
5705 demand. Instead of dropping incoming requests, they are held in
5706 the LND until the RPC service can accept more requests.
5709 Frequency : Sometimes during replay
5711 Description: Assertion failure in ll_local_open after replay.
5712 Details : If replay happened on an open request reply before we were able
5713 to set replay handler, reply will become not swabbed tripping the
5714 assertion in ll_local_open. Now we set the handler right after
5715 recognising of open request
5718 Frequency : very rare
5720 Description: kernel reports "badness in vsnprintf"
5721 Details : Reading from the "recovery_status" /proc file in small chunks
5722 may cause a negative length in lprocfs_obd_rd_recovery_status()
5723 call to vsnprintf() (which is otherwise harmless). Exit early
5724 if there is no more space in the output buffer.
5726 Severity : enhancement
5728 Description: clear OBD RPC statistics by writing to them
5729 Details : It is now possible to clear the OBD RPC statistics by writing
5730 to the "stats" file.
5735 Description: Client mtime is not the same on different clients after utimes
5736 Details : In some cases, the client was using the utimes() syscall on
5737 a file cached on another node. The clients now validate the
5738 ctime from the MDS + OSTs to determine which one is right.
5743 Description: Inability to activate failout mode
5744 Details : lconf script incorrectly assumed that in python string's numeric
5745 value is used in comparisons.
5748 Frequency : always with multiple stripes per file
5750 Description: Inefficient object allocation for mutli-stripe files
5751 Details : When selecting which OSTs to stripe files over, for files with
5752 a stripe count that divides evenly into the number of OSTs,
5753 the MDS is always picking the same starting OST for each file.
5754 Return the OST selection heuristic to the original design.
5759 Description: mount failures may take full timeout to return an error
5760 Details : Under some heavy load conditions it is possible that a
5761 failed mount can wait for the full obd_timeout interval,
5762 possibly several minutes, before reporting an error.
5763 Instead return an error as soon as the status is known.
5765 ------------------------------------------------------------------------------
5767 2006-02-14 Cluster File Systems, Inc. <info@clusterfs.com>
5769 * WIRE PROTOCOL CHANGE. This version of Lustre networking WILL NOT
5770 INTEROPERATE with older versions automatically. Please read the
5771 user documentation before upgrading any part of a live system.
5772 * WARNING: Lustre networking configuration changes are required with
5773 this release. See https://bugzilla.clusterfs.com/show_bug.cgi?id=10052
5776 * Support for kernels:
5777 2.6.9-22.0.2.EL (RHEL 4)
5778 2.6.5-7.244 (SLES 9)
5779 2.6.12.6 vanilla (kernel.org)
5782 Severity : enhancement
5783 Bugzilla : 7981/8208
5784 Description: Introduced Lustre Networking (LNET)
5785 Details : LNET is new networking infrastructure for Lustre, it includes
5786 a reorganized network configuration mode (see the user
5787 documentation for full details) as well as support for routing
5788 between different network fabrics. Lustre Networking Devices
5789 (LNDs) for the supported network fabrics have also been
5790 created for this new infrastructure.
5792 Severity : enhancement
5793 Description: Introduced Access control lists
5794 Details : clients can set ACLs on files and directories in order to have
5795 more fine-grained permissions than the standard Unix UGO+RWX.
5796 The MDS must be started with the "-o acl" mount option.
5798 Severity : enhancement
5799 Description: Introduced filesystem quotas
5800 Details : Administrators may now establish per-user quotas on the
5803 Severity : enhancement
5805 Description: Configuration change for the XT3
5806 The PTLLND is now used to run Lustre over Portals on the XT3
5807 The configure option(s) --with-cray-portals are no longer used.
5808 Rather --with-portals=<path-to-portals-includes> is used to
5809 enable building on the XT3. In addition to enable XT3 specific
5810 features the option --enable-cray-xt3 must be used.
5815 Description: Running on many-way SMP OSTs can trigger oops in llcd_send()
5816 Details : A race between allocating a new llcd and re-getting the llcd_lock
5817 allowed another thread to grab newly-allocated llcd.
5819 Severity : enhancement
5821 Description: 2.6 OST async journal commit and locking fix to improve performance
5822 Details : The filter_direct_io()+filter_commitrw_write() journal commits for
5823 2.6 kernels are now async as they already were in 2.4 kernels so
5824 that they can commit concurrently with the network bulk transfer.
5825 For block-allocated files the filter allocation semaphore is held
5826 to avoid filesystem fragmentation during allocation. BKL lock
5827 removed for 2.6 xattr operations where it is no longer needed.
5832 Description: lconf incorrectly determined whether two IP networks could talk
5833 Details : In some more complicated routing and multiple-network
5834 configurations, lconf will avoid trying to make a network
5835 connection to a disjoint part of the IP space. It was doing the
5836 math incorrectly for one set of cases.
5841 Description: Fix for potential infinite loop processing records in an llog.
5842 Details : If an llog record is corrupted/zeroed, it is possible to loop
5843 forever in llog_process(). Validate the llog record length
5844 and skip the remainder of the block on error.
5847 Frequency : occasional (liblustre only)
5849 Description: liblustre could not open files whose last component is a symlink
5850 Details : sysio_path_walk() would incorrectly pass the open intent to
5851 intermediate path components.
5854 Frequency : rare (liblustre only with non-standard tuning)
5855 Bugzilla : 7201 (7350)
5856 Description: Tuning the MDC DLM LRU size to zero triggers client LASSERT
5857 Details : llu_lookup_finish_locks() tries to set lock data on a lock
5858 after it has been released, only do this for referenced locks
5860 Severity : enhancement
5862 Description: specifying an (invalid) directory default stripe_size of -1
5863 would reset the directory default striping
5864 Details : stripe_size -1 was used internally to signal directory stripe
5865 removal, now use "all default" to signal dir stripe removal
5866 as a directory striping of "all default" is not useful
5869 Frequency : common for large clusters running liblustre clients
5871 Description: doing an ls when liblustre clients are running is slow
5872 Details : sending a glimpse AST to a liblustre client waits for every AST
5873 to time out, as liblustre clients will not respond. Since they
5874 cannot cache data we refresh the OST lock LVB from disk instead.
5876 Severity : enhancement
5878 Description: doing an ls at the same time as file IO can be slow
5879 Details : enqueue and other "small" requests can be blocked behind many
5880 large IO requests. Create a new OST IO portal for non-IO
5881 requests so they can be processed faster.
5884 Frequency : rare (only HPUX clients mounting unsupported re-exported NFS vol)
5886 Description: an HPUX NFS client would get -EACCESS when ftruncate()ing a newly
5887 created file with mode 000
5888 Details : the Linux NFS server relies on an MDS_OPEN_OWNEROVERRIDE hack to
5889 allow an ftruncate() as a non-root user to a file with mode 000.
5890 Lustre now respects this flag to disable mode checks when
5891 truncating a file owned by the user
5894 Frequency : liblustre-only, when liblustre client dies unexpectedly or becomes
5897 Description: Revoking locks from clients that went dead or catatonic might take
5899 Details : New lock flags FL_CANCEL_ON_BLOCK used by liblustre makes
5900 cancellation of such locks instant on servers without waiting for
5901 any reply from clients. Clients drops these locks when cancel
5902 notification from server is received without replying.
5905 Frequency : liblustre-only, when liblustre client dies or becomes busy
5907 Description: Doing ls on Linux clients can take a long time with active
5909 Details : Liblustre client cannot handle ASTs in timely manner, so avoid
5910 granting such locks to it in the first place if possible. Locks
5911 are taken by proxy on the OST during the read or write and
5912 dropped immediately afterward. Add connect flags handling, do
5913 not grant locks to liblustre clients for glimpse ASTs.
5915 Severity : enhancement
5917 Description: Improve read-ahead algorithm to avoid excessive IO for random reads
5918 Details : Existing read-ahead algorithm is tuned for the case of streamlined
5919 sequential reads and behaves badly with applications doing random
5920 reads. Improve it by reading ahead at least read region, and
5921 avoiding excessive large RPC for small reads.
5923 Severity : enhancement
5925 Description: Creating more than 1000 files for a single job may cause a load
5926 imbalance on the OSTs if there are also a large number of OSTs.
5927 Details : qos_prep_create() uses an OST index reseed value that is an
5928 even multiple of the number of available OSTs so that if the
5929 reseed happens in the middle of the object allocation it will
5930 still utilize the OSTs as uniformly as possible.
5935 Description: OST or MDS may oops in ping_evictor_main()
5936 Details : ping_evictor_main() drops obd_dev_lock if deleting a stale export
5937 but doesn't restart at beginning of obd_exports_timed list
5940 Severity : enhancement
5942 Description: improve by-nid export eviction on the MDS and OST
5943 Details : allow multiple exports with the same NID to be evicted at one
5944 time without re-searching the exports list.
5947 Frequency : rare, only with supplementary groups enabled on SMP 2.6 kernels
5949 Description: MDS may oops in groups_free()
5950 Details : in rare race conditions a newly allocated group_info struct is
5951 freed again, and this can be NULL. The 2.4 compatibility code
5952 for groups_free() checked for a NULL pointer, but 2.6 did not.
5955 Frequency : common for liblustre clients doing little filesystem IO
5956 Bugzilla : 9352, 7313
5957 Description: server may evict liblustre clients accessing contended locks
5958 Details : if a client is granted a lock or receives a completion AST
5959 with a blocking AST already set it would not reply to the AST
5960 for LDLM_FL_CANCEL_ON_BLOCK locks. It now replies to such ASTs.
5963 Frequency : lfs setstripe, only systems with more than 160 OSTs
5965 Description: unable to set striping with a starting offset beyond OST 160
5966 Details : llapi_create_file() incorrectly limited the starting stripe
5967 index to the maximum single-file stripe count.
5970 Frequency : LDAP users only
5972 Description: lconf did not handle in-kernel recovery with LDAP properly
5973 Details : lconf/LustreDB get_refs() is searching the wrong namespace
5975 Severity : enhancement
5977 Description: bind OST threads to NUMA nodes to improve performance
5978 Details : all OST threads are uniformly bound to CPUs on a single NUMA
5979 node and do their allocations there to localize memory access
5981 Severity : enhancement
5983 Description: llmount can determine client NID directly from Myrinet (GM)
5984 Details : the client NID code from gmnalnid was moved directly into
5985 llmount, removing the need to use this or specifying the
5986 client NID explicitly when mounting GM clients with zeroconf
5989 Frequency : if client is started with down MDS
5991 Description: if client is started with down MDS mount hangs in ptlrpc_queue_wait
5992 Details : Having an LWI_INTR() wait event (interruptible, but no timeout)
5993 will wait indefinitely in ptlrpc_queue_wait->l_wait_event() after
5994 ptlrpc_import_delayed_req() because we didn't check if the
5995 request was interrupted, and we also didn't break out of the
5996 event loop if there was no timeout
6001 Description: data loss during non-page-aligned writes to a single file from
6002 both multiple nodes and multiple threads on one node at same time
6003 Details : updates to KMS and lsm weren't protected by common lock. Resulting
6004 inconsistency led to false short-reads, that were cached and later
6005 used by ->prepare_write() to fill in partially written page,
6006 leading to data loss.
6009 Frequency : always, if lconf --abort_recovery used
6011 Description: lconf --abort_recovery fails with 'Operation not supported'
6012 Details : lconf was attempting to abort recovery on the MDT device and not
6015 Severity : enhancement
6017 Description: remove cleanup logs
6018 Details : replace lconf-generated cleanup logs with lustre internal
6019 cleanup routines. Eliminates the need for client-cleanup and
6022 Severity : enhancement
6024 Description: add support for EAs (user and system) on lustre filesystems
6025 Details : it is now possible to store extended attributes in the Lustre
6026 client filesystem, and with the user_xattr mount option it
6027 is possible to allow users to store EAs on their files also
6029 Severity : enhancement
6031 Description: Add possibility (config option) to show minimal available OST free
6033 Details : When compiled with --enable-mindf configure option, statfs(2)
6034 (and so, df) will return least minimal free space available from
6035 all OSTs as amount of free space on FS, instead of summary of
6036 free spaces of all OSTs.
6038 Severity : enhancement
6040 Description: do not expand extent locks acquired on OST-side
6041 Details : Modify ldlm_extent_policy() to not expand local locks, acquired
6042 by server: they are not cached anyway.
6045 Frequency : when mmap is used/binaries executed from Lustre
6047 Description: Unmmap pages before throwing them away from read cache.
6048 Details : llap_shrink cache now attempts to unmap pages before discarding
6049 them (if unmapping failed - do not discard). SLES9 kernel has
6050 extra checks that trigger if this unmapping is not done first.
6055 Description: lconf didn't resolve symlinks before checking to see whether a
6056 given mountpoint was already in use
6059 Frequency : when migrating failover services
6060 Bugzilla : 6395, 9514
6061 Description: When migrating a subset of services from a node (e.g. failback
6062 from a failover service node) the remaining services would
6063 time out and evict clients.
6064 Details : lconf --force (implied by --failover) sets the global obd_timeout
6065 to 5 seconds in order to quickly disconnect, but this caused
6066 other RPCs to time out too quickly. Do not change the global
6067 obd_timeout for force cleanup, only set it for DISCONNECT RPCs.
6069 Severity : enhancement
6070 Frequency : if MDS is started with down OST
6071 Bugzilla : 9439,5706
6072 Description: Allow startup/shutdown of an MDS without depending on the
6073 availability of the OSTs.
6074 Details : Asynchronously call mds_lov_synchronize during MDS startup.
6075 Add appropriate locking and lov-osc refcounts for safe
6076 cleaning. Add osc abort_inflight calls in case the
6077 synchronize never started.
6080 Frequency : occasional (Cray XT3 only)
6082 Description: root not authorized to access files in CRAY_PORTALS environment
6083 Details : The client process capabilities were not honoured on the MDS in
6084 a CRAY_PORTALS/CRAY_XT3 environment. If the file had previously
6085 been accessed by an authorized user then root was able to access
6086 the file on the local client also. The root user capabilities
6087 are now allowed on the MDS, as this environment has secure UID.
6090 Frequency : occasional
6092 Description: ldiskfs "too long searching" message happens too often
6093 Details : A debugging message (otherwise harmless) prints too often on
6094 the OST console. This has been reduced to only happen when
6095 there are fragmentation problems on the filesystem.
6100 Description: Division by zero in statfs when all OSCs are inactive
6101 Details : lov_get_stripecnt() returns zero due to incorrect order of checks,
6102 lov_statfs divides by value returned by lov_get_stripecnt().
6106 Bugzilla : 9489, 3273
6107 Description: First write from each client to each OST was only 4kB in size,
6108 to initialize client writeback cache, which caused sub-optimal
6109 RPCs and poor layout on disk for the first writen file.
6110 Details : Clients now request an initial cache grant at (re)connect time
6111 and so that they can start streaming writes to the cache right
6112 away and always do full-sized RPCs if there is enough data.
6113 If the OST is rebooted the client also re-establishes its grant
6114 so that client cached writes will be honoured under the grant.
6119 Description: Slow ls (and stat(2) syscall) on files residing on IO-loaded OSTs
6120 Details : Now I/O RPCs go to different portal number and (presumably) fast
6121 lock requests (and glimses) and other RPCs get their own service
6122 threads pool that should be able to service those RPCs
6125 Severity : enhancement
6127 Description: Ability to exchange lustre version between client and servers and
6128 issue warnings at client side if client is too old. Also for
6129 liblustre clients there is ability to refuse connection of too old
6131 Details : New 'version' field is added to connect data structure that is
6132 filled with version info. That info is later checked by server and
6136 Frequency : rare, liblustre only.
6137 Bugzilla : 9296, 9581
6138 Description: Two simultaneous writes from liblustre at offset within same page
6139 might proceed at the same time overwriting eachother with stale
6141 Details : I/O lock withing llu_file_prwv was released too early, before data
6142 actually was hitting the wire. Extended lock-holding time until
6143 server acknowledges receiving data.
6146 Frequency : extremely rare. Never observed in practice.
6148 Description: avoid generating lustre_handle cookie of 0.
6149 Details : class_handle_hash() generates handle cookies by incrementing
6150 global counter, and can hit 0 occasionaly (this is unlikely, but
6151 not impossible, because initial value of cookie counter is
6152 selected randonly). Value of 0 is used as a sentinel meaning
6153 "unassigned handle" --- avoid it. Also coalesce two critical
6154 sections in this function into one.
6156 Severity : enhancement
6158 Description: allow liblustre clients to delegate truncate locking to OST
6159 Details : To avoid overhead of locking, liblustre client instructs OST to
6160 take extent lock in ost_punch() on client's behalf. New connection
6161 flag is added to handle backward compatibility.
6163 Severity : enhancement
6164 Bugzilla : 4928, 7341, 9758
6165 Description: allow number of OST service threads to be specified
6166 Details : a module parameter allows the number of OST service threads
6167 to be specified via "options ost ost_num_threads={N}" in the
6168 OSS's /etc/modules.conf or /etc/modprobe.conf.
6172 Bugzilla : 6146, 9635, 9895
6173 Description: servers crash with bad pointer in target_handle_connect()
6174 Details : In rare cases when a client is reconnecting it was possible that
6175 the connection request was the last reference for that export.
6176 We would temporarily drop the export reference and get a new
6177 one, but this may have been the last reference and the export
6178 was just destroyed. Get new reference before dropping old one.
6180 Severity : enhancement
6181 Frequency : if client is started with failover MDS
6183 Description: Allow multiple MDS hostnames in the mount command
6184 Details : Try to read the configuration from all specified MDS
6185 hostnames during a client mount in case the "primary"
6188 Severity : enhancement
6190 Description: Stop sending data to evicted clients as soon as possible.
6191 Details : Check if the client we are about to send or are sending data to
6192 was evicted already. (Check is done every second of waiting,
6193 for which l_wait_event interface was extended to allow checking
6194 of exit condition at specified intervals).
6197 Frequency : rare, normally only when NFS exporting is done from client
6199 Description: 'bad disk LOV MAGIC: 0x00000000' error when chown'ing files
6201 Details : Make mds_get_md() recognise empty md case and set lmm size to 0.
6204 Frequency : always, if srand() is called before liblustre initialization
6206 Description: Liblustre uses system PRNG disturbing its usage by user application
6207 Details : Introduce internal to lustre fast and high-quality PRNG for
6208 lustre usage and make liblustre and some other places in generic
6209 lustre code to use it.
6211 Severity : enhancement
6212 Bugzilla : 9477, 9557, 9870
6213 Description: Verify that the MDS configuration logs are updated when xml is
6214 Details : Check if the .xml configuration logs are newer than the config
6215 logs stored on the MDS and report an error if this is the case.
6216 Request --write-conf, or allow starting with --old_conf.
6218 Severity : enhancement
6220 Description: Handle symlinks in the path when checking if Lustre is mounted.
6221 Details : Resolve intermediate symlinks when checking if a client has
6222 mounted a filesystem to avoid duplicate client mounts.
6227 Description: lconf can hit an error exception but still return success.
6228 Details : The lconf command catches the Command error exception at the top
6229 level script context and will exit with the associated exit
6230 status, but doesn't ensure that this exit status is non-zero.
6235 Description: failure of ptlrpc thread startup can cause oops
6236 Details : Starting a ptlrpc service thread can fail if there are a large
6237 number of threads or the server memory is very fragmented.
6238 Handle this without oopsing.
6241 Frequency : always, only if liblustre and non-default acceptor port was used
6243 Description: liblustre cannot connect to servers with non-default acceptor port
6244 Details : tcpnal_set_default_params() was not called and was therefore
6245 ignoring the environment varaible TCPNAL_PORT, as well as other
6246 TCPNAL_ environment variables
6251 Description: two objects could be created on the same OST for a single file
6252 Details : If an OST is down, in some cases it was possible to create two
6253 objects on a single OST for a single file. No problems other
6254 than potential performance impact and spurious error messages.
6258 Bugzilla : 5681, 9562
6259 Description: Client may oops in ll_unhash_aliases
6260 Details : Client dcache may become inconsistent in race condition.
6261 In some cases "getcwd" can fail if the current directory is
6267 Description: Inode refcounting problems in NFS export code
6268 Details : link_raw functions used to call d_instantiate without obtaining
6269 extra inode reference first.
6273 Bugzilla : 9942, 9903
6274 Description: Referencing freed requests leading to crash, memleaks with NFS.
6275 Details : We used to require that call to ll_revalidate_it was always
6276 followed by ll_lookup_it. Also with revalidate_special() it is
6277 possible to call ll_revalidate_it() twice for the same dentry
6278 even if first occurence returned success. This fix changes semantic
6279 between DISP_ENQ_COMPLETE disposition flag to mean there is extra
6280 reference on a request referred from the intent.
6281 ll_intent_release() then releases such a request.
6284 Frequency : rare, normally benchmark loads only
6286 Description: unlinked inodes were kept in memory on the client
6287 Details : If a client is repeatedly creating and unlinking files it
6288 can accumulate a lot of stale inodes in the inode slab cache.
6289 If there is no other client load running this can cause the
6290 client node to run out of memory. Instead flush old inodes
6291 from client cache that have the same inode number as a new inode.
6294 Frequency : SLES9 2.6.5 kernel and long filenames only
6295 Bugzilla : 9969, 10379
6296 Description: utime reports stale NFS file handle
6297 Details : SLES9 uses out-of-dentry names in some cases, which confused
6298 the lustre dentry revalidation. Change it to always use the
6302 Frequency : rare, unless heavy write-truncate concurrency is continuous
6303 Bugzilla : 4180, 6984, 7171, 9963, 9331
6304 Description: OST becomes very slow and/or deadlocked during object unlink
6305 Details : filter_destroy() was holding onto the parent directory lock
6306 while truncating+unlinking objects. For very large objects this
6307 may block other threads for a long time and slow overall OST
6308 responsiveness. It may also be possible to get a lock ordering
6309 deadlock in this case, or run out of journal credits because of
6310 the combined truncate+unlink. Solution is to do object truncate
6311 first in one transaction without parent lock, and then do the
6312 final unlink in a new transaction with the parent lock. This
6313 reduces the lock hold time dramatically.
6316 Frequency : rare, 2.4 kernels only
6318 Description: MDS or OST cleanup may trip kernel BUG when dropping kernel lock
6319 Details : mds_cleanup() and filter_cleanup() need to drop the kernel lock
6320 before unmounting their filesystem in order to avoid deadlock.
6321 The kernel_locked() function in 2.4 kernels only checks whether
6322 the kernel lock is held, not whether it is this process that is
6323 holding it as 2.6 kernels do.
6328 Description: MDS or OST may oops/LBUG if a client is connecting multiple times
6329 Details : The client ptlrpc code may be trying to reconnect to a down
6330 server before a previous connection attempt has timed out.
6331 Increase the reconnect interval to be longer than the connection
6332 timeout interval to avoid sending duplicate connections to
6336 Frequency : echo_client brw_test command
6338 Description: fix echo_client to work with OST preallocated code
6339 Details : OST preallocation code (5137) didn't take echo_client IO path
6340 into account: echo_client calls filter methods outside of any
6341 OST thread and, hence, there is no per-thread preallocated
6342 pages and buffers to use. Solution: hijack pga pages for IO. As
6343 a byproduct, this avoids unnecessary data copying.
6347 Bugzilla : 3555, 5962, 6025, 6155, 6296, 9574
6348 Description: Client can oops in mdc_commit_close() after open replay
6349 Details : It was possible for the MDS to return an open request with no
6350 transaction number in mds_finish_transno() if the client was
6351 evicted, but without actually returning an error. Clients
6352 would later try to replay that open and may trip an assertion
6353 Simplify the client close codepath, and always return an error
6354 from the MDS in case the open is not successful.
6357 Frequency : rare, 2.6 OSTs only
6359 Description: OST may deadlock under high load on fragmented files
6360 Details : If there was a heavy load and highly-fragmented OST filesystems
6361 it was possible to have all the OST threads deadlock waiting on
6362 allocation of biovecs, because the biovecs were not released
6363 until the entire RPC IO was completed. Instead, release biovecs
6364 as soon as they are complete to ensure forward IO progress.
6366 Severity : enhancement
6368 Description: Support for specifying external journal device at mount
6369 Details : If an OST or MDS device is formatted with an external journal
6370 device, this device major/minor is stored in the ext3 superblock
6371 and may not be valid for failover. Allow detecting and
6372 specifying the external journal at mount time.
6377 Description: Mounting an MDS with pending unlinked files may cause oops
6378 Details : target_finish_recovery() calls mds_postrecov() which returned
6379 the number of orphans unlinked. mds_lov_connect->mds_postsetup()
6380 considers this an error and immediately begins cleaning up the
6381 lov, just after starting the mds_lov process
6383 Severity : enhancement
6385 Description: Implement 'lfs df' to report actual free space on per-OST basis
6386 Details : Add sub-command 'df' on 'lfs' to report the disk space usage of
6387 MDS/OSDs. Usage: lfs df [-i][-h]. Command Options: '-i' to report
6388 usage of objects; '-h' to report in human readable format.
6390 ------------------------------------------------------------------------------
6392 2005-08-26 Cluster File Systems, Inc. <info@clusterfs.com>
6399 Description: Mounting an ldiskfs file system with mballoc may crash OST node.
6400 Details : ldiskfs mballoc code may reference an uninitialized buddy struct
6401 at startup during orphan unlinking. Instead, skip buddy update
6402 before setup, as it will be regenerated after recovery is complete.
6407 Description: If an OST is inactive, its locks might reference stale inodes.
6408 Details : lov_change_cbdata() must iterate over all namespaces, even if
6409 they are inactive to clear inode references from the lock.
6411 Severity : enhancement
6412 Frequency : occasional, if non-standard max_dirty_mb used
6414 Description: Client will block write RPCs if not enough grant
6415 Details : If a client has max_dirty_mb smaller than max_rpcs_in_flight,
6416 then the client will block writes while waiting for another RPC
6417 to complete instead of consuming its dirty limit. With change
6418 we get improved performance when max_dirty_mb is small.
6420 Severity : enhancement
6421 Bugzilla : 3389, 6253
6422 Description: Add support for supplementary groups on the MDS.
6423 Details : The MDS has an upcall /proc/fs/lustre/mds/{mds}/group_upcall
6424 (set to /usr/sbin/l_getgroups if enabled) which will do MDS-side
6425 lookups for user supplementary groups into a cache.
6429 Description: O_CREAT|O_EXCL open flags in liblustre always return -EEXIST
6430 Details : Make libsysio to not enforce O_EXCL by clearing the flag,
6431 for liblustre O_EXCL is enforced by MDS.
6435 Description: readdir never returns NULL in liblustre.
6436 Details : Corrected llu_iop_getdirentries logic, to return offset of next
6437 dentry in struct dirent.
6441 Frequency : liblustre only, depends on application IO pattern
6442 Description: liblustre clients evicted if not contacting servers
6443 Details : Don't put liblustre clients into the ping_evictor list, so
6444 they will not be evicted by the pinger ever.
6446 Severity : enhancement
6448 Description: Add ability to evict clients by NID from MDS.
6449 Details : By echoing "nid:$NID" string into
6450 /proc/fs/lustre/mds/.../evict_client client with nid that equals to
6451 $NID would be instantly evicted from this MDS and from all active
6452 OSTs connected to it.
6456 Description: Do not query file size twice, somewhat slowing stat(2) calls.
6457 Details : lookup_it_finish() used to query file size from OSTs that was not
6462 Description: service threads change working directory to that of init
6463 Details : Starting lustre service threads may pin the working directory
6464 of the parent thread, making that filesystem busy. Threads
6465 now change to the working directory of init to avoid this.
6469 Frequency : during shutdown only
6470 Description: shutdown with a failed MDS or OST can cause unmount to hang
6471 Details : Don't resend DISCONNECT messages in ptlrpc_disconnect_import()
6476 Frequency : 2.6 only
6477 Description: chmod/chown may include an extra supplementary group
6478 Details : ll{,u}_mdc_pack_op_data() does not properly initialize the
6479 supplementary group and if none is specified this is used.
6482 Bugzilla : 5479 (6816)
6484 Description: Racing open + rm can assert client in mdc_set_open_replay_data()
6485 Details : If lookup is in progress on a file that is unlinked we might try
6486 to revalidate the inode and fail in revalidate after lookup is
6487 complete and ll_file_open() enqueues the open again but
6488 it_open_error() was not checking DISP_OPEN_OPEN errors correctly.
6491 Frequency : always, if lconf --abort_recovery used
6493 Description: lconf --abort_recovery fails with 'Operation not supported'
6494 Details : lconf was attempting to abort recovery on the MDT device and not
6497 ------------------------------------------------------------------------------
6499 2005-08-08 Cluster File Systems, Inc. <info@clusterfs.com>
6504 Frequency : rare (only unsupported configurations with a node running as an
6506 Bugzilla : 6514, 5137
6507 Description: Mounting a Lustre file system on a node running as an OST could
6509 Details : OSTs now preallocates memory needed to write out data at
6510 startup, instead of when needed, to avoid having to
6511 allocate memory in possibly low memory situations.
6512 Specifically, if the file system is mounted on on OST,
6513 memory pressure could force it to try to write out data,
6514 which it needed to allocate memory to do. Due to the low
6515 memory, it would be unable to do so and the node would
6516 become unresponsive.
6518 Severity : enhancement
6520 Description: Addition of lconf --service command line option
6521 Details : lconf now accepts a '--service <arg>' option, which is
6522 shorthand for 'lconf --group <arg> --select <arg>=<hostname>'
6524 Severity : enhancement
6526 Description: Failover mode is now the default for OSTs.
6527 Details : By default, OSTs will now run in failover mode. To return to
6528 the old behaviour, add '--failout' to the lmc line for OSTs.
6530 Severity : enhancement
6532 Description: Health checks are now provided for MDS and OSTs
6533 Details : Additional detailed health check information on MSD and OSTs
6534 is now provided through the procfs health_check value.
6537 Frequency : occasional, depends on IO load
6539 Description: Disk fragmentation on the OSTs could eventually cause slowdowns
6540 after numerous create/delete cycles
6541 Details : The ext3 inode allocation policy would not allocate new inodes
6542 very well on the OSTs because there are no new directories
6543 being created. Instead we look for groups with free space if
6544 the parent directories are nearly full.
6549 Description: Network or server problems during mount may cause partially
6550 mounted clients instead of returning an error.
6551 Details : The config llog parsing code may overwrite the error return
6552 code during mount error handling, returning success instead
6558 Description: MDS can fail to allocate large reply buffers
6559 Details : After long uptimes the MDS can fail to allocate large reply
6560 buffers (e.g. zconf client mount config records) due to memory
6561 fragmentation or consumption by the buffer cache. Preallocate
6562 some large reply buffers so that these replies can be sent even
6563 under memory pressure.
6567 Frequency : rare (liblustre)
6568 Description: fsx running with liblustre complained that using truncate() to
6569 extend the file doesn't work. This patch corrects that issue.
6570 Details : This is the liblustre equivalent of the fix for bug 6196. Fixes
6571 ATTR_SIZE and lsm use in llu_setattr_raw.
6575 Frequency : rare, only 2.6 kernels
6576 Description: Unusual file access patterns on the MDS may result in inode
6577 data being lost in very rare circumstances.
6578 Details : Bad interaction between the ea-in-inode patch and the "no-read"
6579 code in the 2.6 kernel caused the inode and/or EA data not to
6580 be read from disk, causing single-file corruption.
6584 Frequency : rare, only 2.6 filesystems using extents
6585 Description: Heavy concurrent write and delete load may cause data corruption.
6586 Details : It was possible under high-load situations to have an extent
6587 metadata block in the block device cache from a just-unlinked
6588 file overwrite a newly-allocated data block. We now unmap any
6589 metadata buffers that alias just-allocated data blocks.
6593 Frequency : filesystems with default stripe_count larger than 77
6594 Description: lconf+mke2fs fail when formatting filesystem with > 77 stripes
6595 Details : lconf specifies an inode size of 4096 bytes when the default
6596 stripe_count is larger than 77. This conflicts with the default
6597 inode density of 1 per 4096 bytes. Allocate smaller inodes in
6598 this case to avoid pinning too much memory for large EAs.
6600 ------------------------------------------------------------------------------
6602 2005-07-07 Cluster File Systems, Inc. <info@clusterfs.com>
6607 Frequency : rare (extremely heavy IO load with hundreds of clients)
6609 Description: Client is evicted, gets IO error writing to file
6610 Details : lock ordering changes for bug 5492 reintroduced bug 3267 and
6611 caused clients to be evicted for AST timeouts. The fixes in
6612 bug 5192 mean we no longer need to have such short AST timeouts
6613 so ldlm_timeout has been increased.
6616 Frequency : occasional during --force or --failover shutdown under load
6617 Bugzilla : 5949, 4834
6618 Description: Server oops/LBUG if stopped with --force or --failover under load
6619 Details : a collection of import/export refcount and cleanup ordering
6620 issues fixed for safer force cleanup
6623 Frequency : only filesystems larger than 120 OSTs
6624 Bugzilla : 5990, 6223
6625 Description: lfs getstripe would oops on a very large filesystem
6626 Details : lov_getconfig used kfree on vmalloc'd memory
6629 Frequency : only filesystems exporting via NFS to Solaris 10 clients
6630 Bugzilla : 6242, 6243
6631 Description: reading from files that had been truncated to a non-zero size
6632 but never opened returned no data
6633 Details : ll_file_read() reads zeros from no-object files to EOF
6638 Description: A bug in MDS/OSS recovery could cause the OSS to fail an assertion
6639 Details : There's little harm in aborting MDS/OSS recovery and letting it
6640 try again, so I removed the LASSERT and return an error instead.
6642 Severity : enhancement
6644 Description: New debugging infrastructure for tracking down data corruption
6645 Details : The I/O checksum code was replaced to: (a) control it at runtime,
6646 (b) cover more of the client-side code path, and (c) try to narrow
6647 down where problems occurred
6651 Bugzilla : 3819, 4364, 4397, 6313
6652 Description: Racing close and eviction MDS could cause assertion in mds_close
6653 Details : It was possible to get multiple mfd references during close and
6654 client eviction, leading to one thread referencing a freed mfd.
6656 Severity: : enhancement
6657 Bugzilla : 3262, 6359
6658 Description: Attempts to reconnect to servers are now more aggressive.
6659 Details : This builds on the enhanced upcall-less recovery that was added
6660 in 1.4.2. When trying to reconnect to servers, clients will
6661 now try each server in the failover group every 10 seconds. By
6662 default, clients would previously try one server every 25 seconds.
6667 Description: After recovery, certain operations trigger a failed
6668 assertion on a client.
6669 Details : Failing over an mds, using lconf -d --failover, while a
6670 client was doing a readdir() call would cause the client to
6671 LBUG after recovery completed and the readdir() was resent.
6673 Severity : enhancement
6675 Description: Default groups are now added by lconf
6676 Details : You can now run lconf --group <servicename> without having to
6677 manually add groups with lmc.
6680 Frequency : occasional
6682 Description: Nodes with an elan id of 0 trigger a failed assertion
6685 Frequency : always when accessing e.g. tty/console device nodes
6687 Description: tty and some other devices nodes cannot be used on lustre
6688 Details : file's private_data field is used by device data and lustre
6689 values in there got lost. New field was added to struct file to
6690 store fs-specific private data.
6693 Frequency : when exporting Lustre via NFS
6695 Description: NFSD failed occasionally when looking up a path component
6696 Details : NFSD is looking up ".." which was broken in ext3 directories
6697 that had grown large enough to become hashed.
6700 Frequency : Clusters with multiple interfaces not on the same subnet
6702 Description: Nodes will repeatedly try to reconnect to an interface which it
6703 cannot reach and report an error to the log.
6704 Details : Extra peer list entries will be created by lconf with some peers
6705 unreachable. lconf now validates the peer before adding it.
6708 Frequency : Only if a default stripe is set on the filesystem root.
6710 Description: Setting a default stripe on the filesystem root prevented the
6711 filesystem from being remounted.
6712 Details : The client was sending extra request flags in the root getattr
6713 request and did not allocate a reply buffer for the dir EA.
6716 Frequency : occasional, higher if lots of files are accessed by one client
6717 Bugzilla : 6159, 6097
6718 Description: Client trips assertion regarding lsm mismatch/magic
6719 Details : While revalidating inodes the VFS looks up inodes with ifind()
6720 and in rare cases can find an inode that is being freed.
6721 The ll_test_inode() code will free the lsm during ifind()
6722 when it finds an existing inode and then the VFS later attaches
6723 this free lsm to a new inode.
6727 Bugzilla : 6422, 7030
6728 Description: MDS deadlock between mkdir and client eviction
6729 Details : Creating a new file via mkdir or mknod (starting a transaction
6730 and getting the ns lock) can deadlock with client eviction
6731 (gets ns lock and trying to finish a synchronous transaction).
6734 Frequency : occasional
6735 Description: While starting a server, the fsfilt_ext3 module could not be
6737 Details : CFS's improved ext3 filesystem is named ldiskfs for 2.6
6738 kernels. Previously, lconf would still use the ext3 name
6739 when trying to load modules. Now, it will correctly use
6740 ext3 on 2.4 and ldiskfs on 2.6.
6742 Severity : enhancement
6743 Description: The default stripe count has been changed to 1
6744 Details : The interpretation of the default stripe count (0, to lfs
6745 or lmc) has been changed to mean striping across a single
6746 OST, rather than all available. For general usage we have
6747 found a stripe count of 1 or 2 works best.
6749 Severity : enhancement
6750 Description: Add support for compiling against Cray portals.
6751 Details : Conditional compiling for some areas that are different
6755 Frequency : occasional
6756 Bugzilla : 6409, 6834
6757 Description: Creating files with an explicit stripe count may lead to
6758 a failed assertion on the MDS
6759 Details : If some OSTs are full or unavailable, creating files may
6760 trigger a failed assertion on the MDS. Now, Lustre will
6761 try to use other servers or return an error to the
6765 Frequency : occasional
6767 Description: Multiple concurrent overlapping read+write on multiple SMP nodes
6768 caused lock timeout during readahead (since 1.4.2).
6769 Details : Processes doing readahead might match a lock that hasn't been
6770 granted yet if there are overlapping and conflicting lock
6771 requests. The readahead process waits on ungranted lock
6772 (original lock is CBPENDING), while OST waits for that process
6773 to cancel CBPENDING read lock and eventually evicts client.
6775 Severity : enhancement
6777 Description: Initial enabling of flock support for clients
6778 Details : Implements fcntl advisory locking and file status functions.
6779 This feature is provided as an optional mount flag (default
6780 off), and is NOT CURRENTLY SUPPORTED. Not all types of record
6781 locking are implemented yet, and those that are are not guaranteed
6782 to be completely correct in production environments.
6783 mount -t lustre -o [flock|noflock] ...
6786 Frequency : occasional
6788 Description: OSTs running 2.4 kernels but with extents enabled might trip an
6789 assertion in the ext3 JBD (journaling) layer.
6790 Details : The b_committed_data struct is protected by the big kernel lock
6791 in 2.4 kernels, serializing journal_commit_transaction() and
6792 ext3_get_block_handle->ext3_new_block->find_next_usable_block()
6793 access to this struct. In 2.6 kernels there is finer grained
6794 locking to improve SMP performance of the JBD layer.
6798 Description: Changes the "SCSI I/O Stats" kernel patch to default to "enabled"
6800 -----------------------------------------------------------------------------
6802 2005-05-05 Cluster File Systems, Inc. <info@clusterfs.com>
6804 NOTE: Lustre 1.4.2 uses an incompatible network protocol than previous
6805 versions of Lustre. Please update all servers and clients to
6806 version 1.4.2 or later at the same time. You must also run
6807 "lconf --write-conf {config}.xml" on the MDS while it is stopped
6808 to update the configuration logs.
6810 - fix for HPUX NFS client breakage when NFS exporting Lustre (5781)
6811 - mdc_enqueue does not need max_mds_easize request buffer on send (5707)
6812 - swab llog records of type '0' so we get proper header size/idx (5861)
6813 - send llog cancel req to DLM cancel portal instead of cb portal (5515)
6814 - fix rename of one directory over another leaking an inode (5953)
6815 - avoid SetPageDirty on 2.6 (5981)
6816 - don't re-add just-being-destroyed locks to the waiting list (5653)
6817 - when creating new directories, inherit the parent's custom
6818 striping settings if present parent (3048)
6819 - flush buffers from cache before direct IO in 2.6 obdfilter (4982)
6820 - don't hold i_size_sem in ll_nopage() and ll_ap_refresh_count (6077)
6821 - don't hold client locks on temporary worklist from l_lru (5666)
6822 - handle IO errors in 2.6 obdfilter bio completion routine (6046)
6823 - automatically evict dead clients (5921)
6824 - Update file size properly in create+truncate+fstat case (6196)
6825 - Do not unhash mountpoint dentries, do not allow removal of
6827 - Avoid lock ordering deadlock issue with write/truncate (6203,5654)
6828 - reserve enough journal credits in fsfilt_start_log for setattr (4554)
6829 - ldlm_enqueue freed-export error path would always LBUG (6149,6184)
6830 - don't reference lr_lvb_data until after we hold lr_lvb_sem (6170)
6831 - don't overwrite last_rcvd if there is a *_client_add() error (6086)
6832 - Correctly handle reads of files with no objects (6243)
6833 - lctl recover will also mark a device active if deactivate used (5933)
6835 - by default create 1 inode per 4kB space on MDS, per 16kB on OSTs
6836 - allow --write-conf on an MDS with different nettype than client (5619)
6837 - don't write config llogs to MDS for mounts not from that MDS (5617)
6838 - lconf should create multiple TCP connections from a client (5201)
6839 - init scripts are now turned off by default; run chkconfig --on
6840 lustre and chkconfig --on lustrefs to use them
6841 - upcalls are no longer needed for clients to recover to failover
6843 - add --abort-recovery option to lconf to abort recovery on device
6845 - add support for an arbitrary number of OSTs (3026)
6846 - Quota support protocol changes.
6847 - forward compatibility changes to wire structs (6007)
6848 - rmmod NALs that might be loaded because of /etc/modules.conf (6133)
6849 - support for mountfsoptions and clientoptions to the Lustre LDAP (5873)
6850 - improved "lustre status" script
6851 - initialize blocksize for non-regular files (6062)
6852 - added --disable-server and --disable-client configure options (5782)
6853 - introduce a lookup cache for lconf to avoid repeated DB scans (6204)
6854 - Vanilla 2.4.29 support
6855 - increase maximum number of obd devices to 520 (6242)
6856 - remove the tcp-zero-copy patch from the suse-2.4 series (5902)
6857 - Quadrics Elan drivers are now included for the RHEL 3 2.4.21 and
6858 SLES 9 2.6.5 kernels
6859 - limit stripes per file to 160 (the maximum EA size) (6093)
6861 2005-03-22 Cluster File Systems, Inc. <info@clusterfs.com>
6864 - don't LASSERT in ll_release on NULL lld with NFS export (4655, 5760)
6865 - hold NS lock when calling handle_ast_error->del_waiting_lock (5746)
6866 - fix setattr mtime regression from lovcleanup merge (4829, 5669)
6867 - workaround for 2.6 crash in ll_unhash_aliases (5687, 5210)
6868 - small ext3 extents cleanups and fixes (5733)
6869 - improved mballoc code, several small races and bugs fixed (5733, 5638)
6870 - kernel version 43 - fix remove_suid bugs in both 2.4 and 2.6 (5695)
6871 - avoid needless client->OST connect, fix handle mismatch (5317)
6872 - fix DLM error path that led to out-of-sync client, long delays (5779)
6873 - support common vfs-enforced mount options (nodev,nosuid,noexec) (5637)
6874 - fix several locking issues related to i_size (5492,5624,5654,5672)
6875 - don't move pending lock onto export if it is already evicted (5683)
6876 - fix kernel oops when creating .foo in unlinked directory (5548)
6877 - fix deadlock in obdfilter statistics vs. object create (5811)
6878 - use time_{before,after} to avoid timer jiffies wrap (5882)
6879 - shutdown --force/--failover stability (3607,3651,4797,5203,4834)
6880 - Do not leak request if server was not able to process it (5154)
6881 - If mds_open unable to find parent dir, make that negative lookup(5154)
6882 - don't create new directories with extent-mapping (5909, 5936)
6884 - fix lustre/lustrefs init scripts for SuSE (patch from Scali, 5702)
6885 - don't hold the pinger_sem in ptlrpc_pinger_sending_on_import
6886 - change obd_increase_kms to obd_adjust_kms (up or down) (5654)
6887 - lconf, lmc search both /usr/lib and /usr/lib64 for Python libs (5800)
6888 - support for RHEL4 kernel on i686 (5773)
6889 - provide error messages when incompatible logs are encountered (5898)
6891 2005-02-18 Cluster File Systems, Inc. <info@clusterfs.com>
6892 * version 1.4.0.10 (1.4.1 release candidate 1)
6894 - don't keep a lock reference when lock is not granted (4238)
6895 - unsafe list practices (rarely) led to infinite eviction loop (4908)
6896 - add per-fs limit of Lustre pages in page cache, avoid OOM (4699)
6897 - drop import inflight refcount on signal_completed_replay error (5255)
6898 - unlock page after async write error during send (3677)
6899 - handle missing objects in filter_preprw_read properly (5265)
6900 - no transno return for symlink open, don't save no-trasno open (3440)
6901 - don't try to complete elan receive that already failed (4012)
6902 - free RPC server reply state on error (5406)
6903 - clean up thread from ptlrpc_start_thread() on error (5160)
6904 - readahead could read extra page into cache that wasn't ejected (5388)
6905 - prevent races in class_attach/setup/cleanup/detach (5260)
6906 - don't dereference de->d_inode after l_dput of de (5458)
6907 - use "int" for stripe value returned from lock_to_stripe (5544)
6908 - mballoc allocation and error-checking fixes in 2.6 (5504)
6909 - block device patches to fix I/O request sizes in 2.6 (5482)
6910 - look up hostnames for IB nals (5602)
6911 - 2.6 changed lock ordering of 2 semaphores, caused deadlock (5654)
6912 - don't start multiple acceptors for the same port (5277)
6913 - fix incorrect LASSERT in mds_getattr_name (5635)
6914 - export a proc file for general "ping" checking (5628)
6915 - fix "lfs check" to not block when the MDS is down (5628)
6917 - service request history (4965)
6918 - put {ll,lov,osc}_async_page structs in a single slab (4699)
6919 - create an "evict_client" /proc entry on OSTs, like the MDS has
6920 - fix mount usage message, return errors per mount(8) (5168)
6921 - change grep [] to grep "[]" in tests so they work in more UMLs
6922 - fix ppc64/x86_64 spec to use %{_libdir} instead of /usr/lib (5389)
6923 - remove ancient LOV_MAGIC_V0 EA support (5047)
6924 - add "disk I/Os in flight" and "I/O req time" stats in obdfilter
6925 - align r/w RPCs to PTLRPC_MAX_BRW_SIZE boundary for performance (3451)
6926 - allow readahead allocations to fail when low on memory (5383)
6927 - mmap locking landed again, after considerable improvement (2828)
6928 - add get_hostaddr() to lustreDB.py for LDAP support (5459)
6930 2004-11-23 Cluster File Systems, Inc. <info@clusterfs.com>
6933 - send OST transaction number in read/write reply to free req (4966)
6934 - don't ASSERT in ptl_send_rpc() if we run out of memory (5119)
6935 - lock /proc/sys/portals/routes internal state, avoiding oops (4827)
6936 - the watchdog thread now runs as interruptible (5246)
6937 - flock/lockf fixes (but it's still disabled, pending 5135)
6938 - don't use EXT3 constants in llite code (5094)
6939 - memory shortage at startup could cause assertion (5176)
6941 - reorganization of lov code
6942 - single portals codebase
6944 - add extents/mballoc support (5025)
6945 - direct I/O reads in the obdfilter (4048)
6946 - kernel patches from LNXI for 2.6 (bluesmoke, perfctr, mtd, kexec)
6948 tbd Cluster File Systems, Inc. <info@clusterfs.com>
6951 - send OST transaction number in read/write reply to free req (4966)
6952 - don't ASSERT in ptl_send_rpc() if we run out of memory (5119)
6953 - lock /proc/sys/portals/routes internal state, avoiding oops (4827)
6954 - the watchdog thread now runs as interruptible (5246)
6955 - handle missing objects in filter_preprw_read properly (5265)
6956 - unsafe list practices (rarely) led to infinite eviction loop (4908)
6957 - drop import inflight refcount on signal_completed_replay error (5255)
6958 - unlock page after async write error during send (3677)
6959 - return original error code on reconstructed replies (3761)
6960 - no transno return for symlink open, don't save no-trasno open (3440)
6962 - add pid to ldlm debugging output (4922)
6963 - bump the watchdog timeouts -- we can't handle 30sec yet
6964 - extra debugging for orphan dentry/inode bug (5259)
6966 2004-11-16 Cluster File Systems, Inc. <info@clusterfs.com>
6969 - fix TCP_NODELAY bug, which caused extreme perf regression (5134)
6970 - allocate qswnal tx descriptors singly to avoid fragmentation (4504)
6971 - don't LBUG on obdo_alloc() failure, use OBD_SLAB_ALLOC() (4800)
6972 - fix NULL dereference in /proc/sys/portals/routes (4827)
6973 - allow failed mdc_close() operations to be interrupted (4561)
6974 - stop precreate on OST before MDS would time out on it (4778)
6975 - don't send partial-page writes before EOF from client (4410)
6976 - discard client grant for sub-page writes on large-page clients (4520)
6977 - don't free dentries not owned by NFS code, check generation (4806)
6978 - fix lsm leak if mds_create_objects() fails (4801)
6979 - limit debug_daemon file size, always print CERROR messages (4789)
6980 - use transno after validating reply (3892)
6981 - process timed out requests if import state changes (3754)
6982 - update mtime on OST during writes, return in glimpse (4829)
6983 - add mkfsoptions to LDAP (4679)
6984 - use ->max_readahead method instead of zapping global ra (5039)
6985 - don't interrupt __l_wait_event() during strace
6987 - add software watchdogs to catch hung threads quickly (4941)
6988 - make lustrefs init script start after nfs is mounted
6989 - fix CWARN/ERROR duplication (4930)
6990 - return async write errors to application if possible (2248)
6991 - add /proc/sys/portal/memused (bytes allocated by PORTALS_ALLOC)
6992 - print NAL number in %x format (4645)
6993 - update barely-supported suse-2.4.21-171 series (4842)
6994 - support for sles 9 %post scripts
6995 - support for building 2.6 kernel-source packages
6996 - support for sles km_* packages
6998 2004-10-07 Cluster File Systems, Inc. <info@clusterfs.com>
7001 - ignore -ENOENT errors in osc_destroy (3639)
7002 - notify osc create thread that OSC is being cleaned up (4600)
7003 - add nettype argument for llmount in #5d in conf-sanity.sh (3936)
7004 - reconstruct ost_handle() like mds_handle() (4657)
7005 - create a new thread to do import eviction to avoid deadlock (3969)
7006 - let lconf resolve symlinked-to devices (4629)
7007 - don't unlink "objects" from directory with default EA (4554)
7008 - hold socknal file ref over connect in case target is down (4394)
7009 - allow more than 32000 subdirectories in a single directory (3244)
7010 - fix blocks count for O_DIRECT writes (3751)
7011 - OST returns ENOSPC from object create when no space left (4539)
7012 - don't send truncate RPC if file size isn't changing (4410)
7013 - limit OSC precreate to 1/2 of value OST considers bogus (4778)
7014 - bind to privileged port in socknal and tcpnal (3689)
7016 - rate limit CERROR/CWARN console message to avoid overload (4519)
7017 - GETFILEINFO dir ioctl returns LOV EA + MDS stat in 1 call (3327)
7018 - basic mmap support (3918)
7019 - kernel patch series update from b1_4 (4711)
7021 2004-09-16 Cluster File Systems, Inc. <info@clusterfs.com>
7024 - avoid crash during MDS cleanup with OST shut down (2775)
7025 - fix loi_list_lock/oig_lock inversion on interrupted IO (4136)
7026 - don't use bad inodes on the MDS (3744)
7027 - dynamic object preallocation to improve recovery speed (4236)
7028 - don't hold spinlock over lock dumping or change debug flags (4401)
7029 - don't zero obd_dev when it is force cleaned (3651)
7030 - print grants to console if they go negative (4431)
7031 - "lctl deactivate" will stop automatic recovery attempts (3406)
7032 - look for existing locks in ldlm_handle_enqueue() (3764)
7033 - don't resolve lock handle twice in recovery avoiding race (4401)
7034 - revalidate should check working dir is a directory (4134)
7036 - don't always mark "slow" obdfilter messages as errors (4418)
7038 2004-08-24 Cluster File Systems, Inc. <info@clusterfs.com>
7041 - don't close LustreDB during write_conf until it is done (3860)
7042 - fix typo in lconf for_each_profile (3821)
7043 - allow dumping logs from multiple threads at one time (3820)
7044 - don't allow multiple threads in OSC recovery (3812)
7045 - fix debug_size parameters (3864)
7046 - fix mds_postrecov to initialize import for llog ctxt (3121)
7047 - replace config semaphore with spinlock (3306)
7048 - be sure to send a reply for a CANCEL rpc with bad export (3863)
7049 - don't allow enqueue to complete on a destroyed export (3822)
7050 - down write_lock before checking llog header bitmap (3825)
7051 - recover from lock replay timeout (3764)
7052 - up llog sem before sending rpc (3652)
7053 - reduce ns lock hold times when setting kms (3267)
7054 - change a dlm LBUG to LASSERTF, to maybe learn something (4228)
7055 - fix NULL deref and obd_dev leak on setup error (3312)
7056 - replace some LBUG about llog ops with error handling (3841)
7057 - don't match INVALID dentries from d_lookup and spin (3784)
7058 - hold dcache_lock while marking dentries INVALID and hashing (4255)
7059 - fix invalid assertion in ptlrpc_set_wait (3880)
7061 - add libwrap support for the TCP acceptor (3996)
7062 - add /proc/sys/portals/routes for non-root route listing (3994)
7063 - allow setting MDS UUID in .xml (2580)
7064 - print the stack of a process that LBUGs (4228)
7066 2004-07-14 Cluster File Systems, Inc. <info@clusterfs.com>
7069 - don't cleanup request in ll_file_open() on failed MDS open (3430)
7070 - make sure to unset replay flag from failed open requests (3440)
7071 - if default stripe count is 0, use OST count for inode size (3636)
7072 - update parent mtime/ctime on client for create/unlink (2611)
7073 - drop dentry ref in ext3_add_link from open_connect_dentry (3266)
7074 - free recovery state on server during a forced cleanup (3571)
7075 - unregister_reply for resent reqs (3063)
7076 - loop back devices mounting and status check on 2.6 (3563)
7077 - fix resource-creation race that can provoke i_size == 0 (3513)
7078 - don't try to use bad inodes returned from MDS/OST fs lookup (3688)
7079 - more debugging for page-accounting assertion (3746)
7080 - return -ENOENT instead of asserting if ost getattr+unlink race (3558)
7081 - avoid deadlock after precreation failure (3758)
7082 - fix race and lock order deadlock in orphan handling (3450, 3750)
7083 - add validity checks when grabbing inodes from l_ast_data (3599)
7085 - add /proc/.../recovery_status to obdfilter (3428)
7086 - lightweight CDEBUG infrastructure, debug daemon (3668)
7087 - change default OSC RPC parameters to be better on small clusters
7088 - turn off OST read cache for files smaller than 32MB
7089 - install man pages and include them in rpms (3100)
7090 - add new init script for (un)mounting lustre filesystems (2593)
7091 - run chkconfig in %post for init scripts (3701)
7092 - drop scimac NAL (unmaintained)
7094 2004-06-17 Cluster File Systems, Inc. <info@clusterfs.com>
7097 - clean kiobufs before and after use (3485)
7098 - strip trailing '/'s before comparing paths with /proc/mounts (3486)
7099 - remove assertions to work around "in-flight rpcs" recovery bug (3063)
7100 - change init script to fail more clearly if not run as root (1528)
7101 - allow clients to reconnect during replay (1742)
7102 - fix ns_lock/i_sem lock ordering deadlock for kms update (3477)
7103 - don't do DNS lookups on NIDs too small for IP addresses (3442)
7104 - re-awaken ptlrpcd if new requests arrive during check_set (3554)
7105 - fix cond_resched (3554)
7106 - only evict unfinished clients after recovery (3515)
7107 - allow bulk resend, prevent data loss (3570)
7108 - dynamic ptlrpc request buffer allocation (2102)
7109 - don't allow unlinking open directory if it isn't empty (2904)
7110 - set MDS/OST threads to umask 0 to not clobber client modes (3359)
7111 - remove extraneous obd dereference causing LASSERT failure (3334)
7112 - don't use get_cycles() when creating temp. files on the mds (3156)
7113 - hold i_sem when setting i_size in ll_extent_lock() (3564)
7114 - handle EEXIST for set-stripe, set proper directory name (3336)
7116 - servers can dump a log evicting a client - lustre.dump_on_timeout=1
7117 - fix ksocknal_fmb_callback() error messages (2918)
7119 2004-05-27 Cluster File Systems, Inc. <info@clusterfs.com>
7122 - don't copy lvb into (possibly NULL) reply on error (2983)
7123 - don't deref dentry after dput, don't free lvb on error (2922)
7124 - use the kms to determine writeback rpc length (2947)
7125 - increment oti_logcookies when osc is inactive (2948)
7126 - update client's i_blocks count via lvb messages (2543)
7127 - handle intent open/close of special files properly (1557)
7128 - mount MDS with errors=remount-ro, like obdfilter (2009)
7129 - initialize lock handle to avoid ASSERT on error cleanup (3057)
7130 - don't use cancelling-locks' kms values (2947)
7131 - use highest lock extent for kms, not last one (2925)
7132 - don't dereference ERR_PTR() dentry in error handling path (3107)
7133 - fix thread race in portals_debug_dumplog() (3122)
7134 - create lprocfs device entries at setup instead of at attach (1519)
7135 - common AST error handler, don't evict client on completion race (3145)
7136 - zero nameidata in detach_mnt in 2.6 (3118)
7137 - verify d_inode after revalidate_special is valid in 2.6 (3116)
7138 - use lustre_put_super() to handle zconf unmounts in 2.6 (3064)
7139 - initialize RPC timeout timer earlier for 2.6 (3219)
7140 - don't dereference NULL reply buffer if mdc_close was never sent (2410)
7141 - print nal/nid for unknown nid (3258)
7142 - additional checks for oscc recovery before doing precreate (3284)
7143 - fix ll_extent_lock() error return code for 64-bit systems (3043)
7144 - don't crash in mdc_close for bad permissions on open (3285)
7145 - zero i_rdev for non-device files (3147)
7146 - clear page->private before handing to FS, better assertion (3119)
7147 - tune the read pipeline (3236)
7148 - fix incorrect decref of invalidated dentry (2350)
7149 - provide read-ahead stats and refine rpc in flight stats (3328)
7150 - don't hold journal transaction open across create RPC (3313)
7151 - update atime on MDS at close time (3265)
7152 - close LDAP connection when recovering to avoid server load (3315)
7153 - update iopen-2.6 patch with fixes from 2399,2517,2904 (3301)
7154 - don't leak open file on MDS after open resend (3325)
7155 - serialize filter_precreate and filter_destroy_precreated (3329)
7156 - loop device shouldn't call sync_dev() for nul device (3092)
7157 - clear page cache after eviction (2766)
7158 - resynchronize MDS->OST in background (2824)
7159 - refuse to mount the same filesystem twice on same mountpoint (3394)
7160 - allow llmount to create routes for mounting behind routers (3320)
7161 - push lock cancellation to blocking thread for glimpse ASTs (3409)
7162 - don't call osc_set_data_with_check() for TEST_LOCK matches (3159)
7163 - fix rare problem with rename on htree directories (3417)
7165 - allow default OST striping configuration per directory (1414)
7166 - fix compilation for qswnal for 2.6 kernels (3125)
7167 - increase maximum number of MDS request buffers for large systems
7168 - change liblustreapi to be useful for external progs like lfsck (3098)
7169 - increase local configuration timeout for slow disks (3353)
7170 - allow configuring ldlm AST timeout - lustre.ldlm_timeout=<seconds>
7172 2004-03-22 Cluster File Systems, Inc. <info@clusterfs.com>
7175 - fixes for glimpse AST timeouts / incorrectly 0-sized files (2818)
7176 - don't overwrite extent policy data in reply if lock was blocked (2901)
7177 - drop filter export grants atomically with removal from device (2663)
7178 - del obd_self_export from work_list in class_disconnect_exports (2908)
7179 - don't LBUG if MDS recovery times out during orphan cleanup (2530)
7180 - swab reply message in mdc_close, other PPC fixes (2464)
7181 - fix destroying of named logs (2325)
7182 - overwrite old logs when running lconf --write_conf (2264)
7183 - bump LLOG_CHUNKSIZE to 8k to allow for larger clusters (2306)
7184 - fix race in target_handle_connect (2898)
7185 - mds_reint_create() should take same inode create lock (2926)
7186 - correct journal credits calculated for CANCEL_UNLINK_LOG (2931)
7187 - don't close files for self_export to avoid uninitialized obd (2936)
7188 - allow MDS with the same name as client node (2939)
7189 - hold dentry reference for closed log files for unlink (2325)
7190 - reserve space for all logs during transactions (2059)
7191 - don't evict page beyond end of stripe extent (2925)
7192 - don't oops on a deleted current working directory (2399)
7193 - handle hard links to targets without a parent properly (2517)
7194 - don't dereference NULL lock when racing during eviction (2867)
7195 - don't grow lock extents when lots of conflicting locks (2919)
7197 2004-03-04 Cluster File Systems, Inc. <info@clusterfs.com>
7200 - account for cache space usage on clients to avoid data loss (974)
7201 - lfsck support in lustre kernel code (2349)
7202 - reduce journal credits needed for BRW writes (2370)
7203 - orphan handling to avoid losing space on client/server crashes
7204 - ptlrpcd can be blocked, stopping ALL progress (2477)
7205 - use lock value blocks to assist in proper KMS, faster stat (1021)
7206 - takes i_sem instead of DLM locks internally on obdfilter (2720)
7207 - recovery for initial connections (2355)
7208 - fixes for mds_cleanup_orphans (1934)
7209 - abort_recovery crashes MDS in b_eq (mds_unlink_orphan) (2584)
7210 - block all file creations until orphan recovery completes (1901)
7211 - client remove rq_connection from request struct (2423)
7212 - conf-sanity test_5, proper cleanup in umount log not availale (2640)
7213 - recovery timer race (2670)
7214 - mdc_close recovey bug (2532)
7215 - ptlrpc cleanup bug (2710)
7216 - mds timeout on local locks (2588)
7217 - namespace lock held during RPCs (2431)
7218 - handle interrupted sync write properly (2503)
7219 - don't try to handle a message that hasn't been replied to (2699)
7220 - client assert failure during cleanup after abort recovery (2701)
7221 - leak mdc device after failed mount (2712)
7222 - ptlrpc_check_set allows timedout requests to complete (2714)
7223 - wait for inflight reqs when ptlrpcd finishes (2710)
7224 - make sure unregistered services are removed from the srv_list
7225 - reset bulk XID's when resending them (caught by 1138 test)
7226 - unregister_bulk after timeout
7227 - fix lconf error (2694)
7228 - handle write after unfinished setstripe, stripe-only getstripe (2388)
7229 - readahead locks pages, leaves pending causing memory pressure (2673)
7230 - increase OST request buffers to 4096 on large machines (2729)
7231 - fix up permission of existing directories in simple_mkdir (2661)
7232 - init deleted item, add assertions ptlrpc_abort_inflight() (2725)
7233 - don't assign transno to errored transactions (2742)
7234 - don't delete objects on OST if given a bogus objid from MDS (2751)
7235 - handle large client PAGE_SIZE readdir on small PAGE_SIZE MDS (2777)
7236 - if rq_no_resend, then timeout request after recovery (2432)
7237 - fix MDS llog_logid record size, 64-bit array alignment (2733)
7238 - don't call usermode_helper from ptlrpcd, DEFAULT upcall (2773)
7239 - put magic in mount.lustre data, check for bad/NULL mount data (2529)
7240 - MDS recovery shouldn't delete objects that it has given out (2730)
7241 - if enqueue arrives after completion, don't clobber LVB (2819)
7242 - don't unlock pages twice when trigger_group_io returns error (2814)
7243 - don't deref NULL rq_repmsg if ldlm_handle_enqueue failed (2822)
7244 - don't write pages to disk if there was an error (1450)
7245 - don't ping imports that have recovery disabled (2676)
7246 - take buffered bytes into account when balancing socknal conn (2817)
7247 - hold a DLM lock over readdir always, use truncate_inode_pages (2706)
7248 - reconnect unlink llog connection after MDS reconnects to OST (2816)
7249 - remove little-endian swabbing of llog records (1987)
7250 - set/limit i_blksize to LL_MAX_BLKSIZE on client (2884)
7251 - retry reposting request buffers if they fail (1191)
7252 - grow extent at grant time to avoid granting a revoked lock (2809)
7253 - lock revoke doesn't evict page if covered by a second lock (2765)
7254 - disable VM readahead to avoid reading outside lock extents (2805)
7256 - return LL_SUPER_MAGIC from statfs for the filesystem type (1972)
7257 - updated kernel patches for hp-2.4.20 kernel (2681)
7259 2004-02-07 Cluster File Systems, Inc. <info@clusterfs.com>
7262 - fix truncated write corruption (2366)
7263 - fix for failed assertion in iopen_connect_dentry (1792,2517)
7265 - don't flag the ptlrpcd thread with PF_MEMALLOC (2636)
7266 - ensure len(uuid) < 37 in lmc (1171)
7267 - fix ia64 OOPS in llog_test (2255)
7268 - zero end of page at obdfilter for partial page writes (2648)
7269 - don't leave stale dentries around after renames (bug 2428)
7270 - fix timeouts when evicting a client with a single lock held (2642)
7271 - set deadline for the initial HELLO message to drain (2634)
7272 - print out dotted-quad IP addresses in the socknal (2302)
7274 - additional debugging for MDS client eviction problem (2443)
7275 - fix mkfsoptions support for osts (2603, 2604)
7277 2004-01-27 Cluster File Systems, Inc. <info@clusterfs.com>
7280 - add series for the vanilla 2.6.0 kernel
7281 - add series for the vanilla 2.4.24 kernel
7282 - add series for a cray x86/64 UL kernel drop
7283 - fix xattr patches for the vanilla 2.4.19 series
7285 - generate true UUIDs in lmc (1171)
7286 - have portals stack dumping break in UML (2466)
7287 - avoid bad dchild deref; avoid inum lock w/o creation (2362)
7288 - allocate with _NOFS in ldlm to avoid deadlock (1933)
7289 - wake callback waiting threads on client eviction (2460)
7290 - Add --ptldebug and --subsystem to lmc (1719)
7291 - update assertion to allow safe interrupt allocation
7292 - set rq_no_resend for cancel requests (2432)
7293 - recalculate ptlrpcd timeout after resend (2494)
7294 - call vfs_rmdir when removing pending directories (2368)
7295 - fix renaming a file to itself (2429)
7296 - lmc creates a default one-stripe lov (2454)
7297 - expand procfs space to handle large clusters (2326)
7298 - increase UML stack to avoid overflow
7299 - update lconf's list of debug and subsystem masks
7300 - fix lfs find --obd (2510)
7301 - /proc tunable for disabling filter read caching (2591)
7302 - stop rpm packages from altering slapd.conf (2301)
7303 - disable nagle in the socknal under 0conf (2578)
7304 - choose mds inode size based on stripe count (2572)
7305 - fix kernel-source rpm problems (2516)
7307 - add --disable-doc to avoid pdf generation (2421)
7308 - update documentation, tests, type-os, comments
7309 - avoid format warnings on ia64
7310 - remove the TOE NAL
7311 - tiny code cleanups by removing unused fields
7313 2004-01-07 Cluster File Systems, Inc. <info@clusterfs.com>
7316 - fix obvious semaphore misuse in as-yet-unused setattr path (2348)
7317 - remove the most blatant lies from BUILDING file (2371)
7318 - change default debug level to reasonable production setting
7319 - reduce client side cache size to reduce cache flush time
7320 - reduce max RPCs in flight to avoid unnecessary file fragmentation
7321 - make TCP zerocopy and pinger support enabled by default (2476)
7322 - sync writes completed after process exits caused crashes (2319)
7323 - maintain correct mount count on the MDS (2356)
7324 - backout 1557, because 2316 wasn't really fixed
7325 - better file I/O statistics gathering in /proc
7326 - don't take unnecessary, deadlock-inducing bug in readpage (2383)
7327 - another kernel patch to fix zero-copy TCP function export
7328 - don't take duplicate lock when processing re-sent getattr (2420)
7329 - lctl uses obd_self_export instead of creating new conn (2353)
7330 - MDS/OST recovery case which requires object creation asserted (2425)
7331 - move lfs from /usr/sbin to /usr/bin in packages
7332 - fix race between mds_client_add and mds_client_free (2417)
7333 - use kmalloc instead of slabs in portals (2430)
7334 - don't create duplicate records when a failover MDS is present (2442)
7335 - remove unnecessary mount age check (2332)
7336 - don't remove directory inodes from locks prematurely (2451)
7337 - don't break if MDS service name is the same as hostname (2103)
7338 - fix races in client write RPC generation when cache full (2482)
7340 2003-12-13 Cluster File Systems, Inc. <info@clusterfs.com>
7343 - remove now-unused request->rq_obd (278)
7344 - if an allocation fails, print out how much memory we've used (1933)
7345 - use PORTAL_SLAB_ALLOC for structures, to get GFP_MEMALLOC (1933)
7346 - add the "configurable stack size" patch to most series files (1256)
7347 - ability to write large log records, for 100+ OST configs (2306)
7348 - fix NULL deref when filter_prep fails (2314)
7349 - fix operator precedence error in filter_sync
7350 - dynamic allocation of socknal TX descriptors (2315)
7351 - fix a missed case in the GFP_MEMALLOC patch, can cause deadlock (2310)
7352 - fix gcc 2.96 compilation problem in xattr kernel patch (2294)
7353 - ensure that CWARN messages in Portals always get to the syslog
7354 - __init/__exit are not for prototype decls (ldlm_init/exit)
7355 - x86-64 compile warning fixes
7356 - fix gateway LMC keyword conflict (2318)
7357 - fix MDS lock inversions in getattr/reint paths (1844)
7358 - fix a rare lock re-ordering bug, which caused deadlock (2322)
7359 - fix i_sem/journal inversion in fsfilt_ext3_write_record (2306)
7360 - DLM race condition prevented some lock evictions (2328)
7361 - ENOMEM detection and retry on socknal sends (2230)
7362 - use GFP_NOFS throughout Lustre, to combat ENOMEM (2230)
7363 - move osc_rpcd into ptlrpc, for use in MDC and others (2329)
7364 - protect MDS inode fsdata with stronger locking; fixes assertion (2313)
7365 - better error messages when a client is rejected during recovery (1505)
7366 - avoid cancelling locks which were never granted, after failure (2330)
7367 - fix i_sem/journal inversion in mds_client_add (2333)
7368 - fix truncate/getattr lock cycle deadlock (2334)
7369 - use rpcd to send close; allows resend after timeout, avoid leak (1897)
7370 - fix two rare exit paths which could leak an l_lock() ref (2321)
7371 - fencepost error in MDS/OST orphan recovery (2226)
7372 - make log record alignment 8 bytes (1988)
7373 - lstripe now fails when requested offset > ost_count (2237)
7374 - ensure that all kernel series have a complete list.h (1607)
7375 - fix crashes in special-file operations (2316)
7376 - lctl create/brw OID mismatch, caused by obsolete filter loop (2339)
7378 - allow configurable automake binary, for testing new versions
7379 - small update to the lfs documentation
7381 2003-12-03 Cluster File Systems, Inc. <info@clusterfs.com>
7383 * fix negative export reference count in fsfilt_sync (2312)
7385 2003-12-01 Cluster File Systems, Inc. <info@clusterfs.com>
7386 * release candidate 0.9.1
7388 - orphans are moved into the PENDING directory for possible recovery
7389 - replayed opens now open by fid for orphan/rename safety (1042)
7390 - last close of an orphan inode generates a transno (683)
7391 - chdir() and mount() now pin the directory entry (1020)
7392 - avoid CERROR in normal ll_setattr_raw() error case (1500)
7393 - discard very old requests without processing them (1502)
7394 - remove some common, well-understood CERRORs (1505)
7395 - require O_DIRECT I/O to be page-sized to workaround IA64 crash (1609)
7396 - clear "grant" flags in OST replies until OST grant code lands (1644)
7397 - fix read performance by not clobbering i_blksize on client (1598)
7398 - fix __ldlm_handle2lock oops by not dereferencing lock after PUT (1625)
7399 - make LRU size a /proc tunable, clears locks when reduced (707)
7400 - fix some lprocfs rot that prevented ptlbd from loading (1732)
7401 - server locks take references on exports now (1558)
7402 - build fixes for 2.4.20-rh trees (1663)
7403 - return an error from lov_create if all OSCs are inactive (1751)
7404 - fix import levels when a reconnect happens without a timeout (1597)
7405 - exit early from mds_open if we get a lookup error (1749)
7406 - partial page read at EOF wouldn't wait for disk before sending (1642)
7407 - avoid NULL deref in obdfilter when reading page past EOF (1592)
7408 - avoid LASSERT in ll_intent_lock if server failed very early (1090)
7409 - fix LBUG in ll_it_open_error with rc = -2 (1861)
7410 - write/truncate lock inversion (1639)
7411 - Don't auto-load obdclass, portals modules during cleanup (1495)
7412 - fix timestamps from jumping to "now" (1763)
7413 - extra journal assertions (1648)
7414 - add an extra multiunlink test (1771)
7415 - fix read_record/write_record API (1776)
7416 - fix leak of offset_extent, possible incorrect i_size later (1772)
7417 - fix lasserts in mis-matched transnos during open-unlink testing (1541)
7418 - Debugging for the kqswnal_get_idle_tx problems (1820)
7419 - Allow recovery to be attempted multiple times (1536)
7420 - Write out MDS last_rcvd file after it is first created (1600)
7421 - Fix tx_descriptor leak in failed transmit situations (1827)
7422 - ext3 journaling fixes for assertion failure after IO error (1871)
7423 - class_export_put() on freed export after completion AST error (1896)
7424 - Fix revalidate looping in VFS (1322)
7425 - Don't access a freed export during MDS_REINT timeout (1521)
7426 - Add open-unlink recovery support on the MDS (1673,1764)
7427 - Return an error if no MDS data was read from last_rcvd (1946)
7428 - Fix for lookup "." or ".." crash on error (1932,1931,1935)
7429 - Don't setup a disk device that doesn't match exported UUID (317)
7430 - Reduce bulk RPC timeout to avoid cascading client/OST failures (1845)
7431 - avoid committing NULL handle in force close
7432 - local.sh is now a one-stripe LOV configuration
7433 - POSIX utime.4 -EPERM on FIFO not owned by user (56)
7434 - fix ext3 htree duplicate directory entry corruption (1516)
7435 - POSIX creat.13, fstat.1, open.18, stat.3 new file atime/mtime (2020)
7436 - update to new LOV EA format (2097)
7437 - interoperability for different PAGE_SIZE/wordsize (686,1821,1343,2042)
7439 2003-06-15 Phil Schwan <phil@clusterfs.com>
7442 - imports and exports cleanup too early, need refcounts (349, 879, 1045)
7443 - per-import/export recovery handling (958, 931, 959)
7444 - multiple last-rcvd slots, for serving multiple FSes (949)
7445 - connections are again shared between multiple imp/exports (963, 964)
7446 - "umount -f" would hang if any requests needed to be sent (393, 978)
7447 - avoid pinning large req buffer by copying for queued messages (989)
7448 - add "uuid" to "lctl device" command to help upcalls (991)
7449 - "open" RPCs with transnos would confuse recovery counters (1037)
7450 - do proper endian conversion of all wire messages (288, 340, 891)
7451 - remove OST bulk get LBUGs, fix ost_brw_write cleanup (1126)
7452 - call waiting locks callback from LDLM recovery thread (1127, 1151)
7453 - fix ptlrpc_connection leak in target_handle_connect (1174)
7454 - fix import refcounting bug in OST and MDS cleanup (1134)
7455 - if an invalid-at-open-time OSC returned before close(), LBUG (1150)
7456 - fix very unlikely obd_types race condition (501)
7457 - remove osc_open hack for echo_client (1187)
7458 - we leaked exports/dlmimps for forcibly disconnected clients (1143)
7459 - a failure in read_inode2 leads to deadlock (1139)
7460 - cancel ack-locks as soon as transaction is committed (1072)
7461 - fix major leaks and crashes in the bulk I/O path (937, 1057)
7462 - make sure to commitrw after any preprw to avoid deadlock (1162)
7463 - failing to execute a file in a lustre FS would lock inode (1203)
7464 - small DEBUG_REQ fix to avoid dereferencing a NULL (1227)
7465 - don't ASSERT while cleaning up an incompletely-setup obd (1248)
7466 - obd_uuid2tgt would walk off the end of the list (1255)
7467 - on IA64 the osc would give portals incorrect bulk size (1258)
7468 - fix debug daemon ioctl interface; allows daemon on ia64 (1274)
7469 - fix lock inversion caused by new llite matching code (1282)
7470 - limit the number of dirty pages on a client to 10MB (1286)
7471 - timed out locks were not being corrected cancelled (1289)
7472 - fix O_DIRECT above 4GB on IA-32 (1292)
7473 * major user-visible changes
7474 - fail out/fail over policy now controlled by the upcall (993)
7476 - add OBD_PING to check server availability and failure (954)
7477 - lustre messages are now sent in sending host order (288, 340, 891)
7478 - add eadatalen to MDS getattr reply (340)
7479 - OST read replies may contain second buffer, with per-page status (593)
7481 2003-03-11 Phil Schwan <phil@clusterfs.com>
7484 - LDLM_DEBUG macro fix, for gcc 3.2 (850)
7485 - failed open()s could cause deadlock; fixed (867, 869)
7486 - stop cancelling OST locks when files are closed (481)
7487 - overlapping XID spaces caused network corruption (851, 853)
7488 - fix unsafe fsfilt counter arithmetic; change to atomic_t
7489 - setattr_raw added, to do single-RPC, server-side setattrs
7490 - lmc/lconf syntax change for OST UUIDs
7491 - fix crashy race condition between ptlrpc_free_req and osc_close
7492 - don't use request in mdc_enqueue if we hit a timeout (889)
7493 - don't set the inode i_size for regular files from the MDS (896)
7494 - handle out of order completion AST (842)
7495 - don't LBUG if a lock request times out after receiving AST (913)
7496 - avoid d_rehash race in ll_find_alias by rehashing inside dcache_lock
7497 - if a bad lock AST arrives, send an error instead of dropping entirely
7498 - return 0 from revalidate2 if ll_intent_lock returns -EINTR (912)
7499 - fix leak in bulk IO when only partially completed (899, 900, 926)
7500 - fix O_DIRECT for ia64 (55)
7501 - (almost) eliminate Lustre-kernel-thread effects on load average (722)
7502 - C-z after timeout could hang a process forever; fixed (977)
7504 - client-side I/O cache (678, 924, 929, 941, 970)
7506 - READPAGE and SETATTRs which don't take server-side locks get
7509 2003-02-11 Phil Schwan <phil@clusterfs.com>
7512 - Fix ldlm_lock_match on the MDS to avoid matching remote locks (592)
7513 - Fix fsfilt_extN_readpage() to read a full page of directory
7514 entries, or fake the remainder if PAGE_SIZE != blocksize (500)
7515 - Avoid extra mdc_getattr() in ll_intent_lock when possible (534, 604)
7516 - Fix imbalanced LOV object allocation and out-of-bound access (469)
7517 - Most intent operations were removed, in favour of a new RPC mode
7518 that does a single RPC to the server and bypasses most of the VFS
7519 - All LDLM resource ID arrays were removed in favour of ldlm_res_id
7520 - Aggressively cancel local locks on DLM servers
7521 - mds_reint_unlink sends EA to the client if it's the last nlink.
7522 client uses that EA to unlink OST objects.
7523 - mds_reint_{rename,unlink,link} were rewritten to take ordered locks
7524 - recursive symlinks were fixed (439)
7525 - fixed NULL deref in DEBUG_REQ
7526 - filter_update_lastobjid no longer calls sync, which annoyed extN
7527 - fixed multi-client small-writes to a single file problem (445)
7528 - fixed mtime updates during file writes (607)
7529 - fixed vector writes on obdfilter causing problems when ENOSPC (670)
7530 - fixed bug in obd_brw_read/write() (under guise of testing 367)
7531 - fixed Linux OST size reporting problem (444, 656)
7532 - OST now updates object mtime with writes or setattr (607, 619)
7533 - client verifies file size before zeroing page past EOF (445)
7534 - OST now writes last allocated objid to disk with allocation (108)
7535 - LOV on echo now works (409)
7537 - mds_reint_unlink sends a new buffer, with the EA included. this
7538 buffer is only valid if body->valid & OBD_MD_FLEASIZE, which is only
7539 set if a regular file was being unlinked, and it was the last link
7540 - use PtlGet from the target for bulk writes (315)
7541 - OST now updates object mtime with writes or setattr (607, 619)
7542 - LDLM now has a grant-time callback to revalidate locked items, if
7544 - Many MDS operations were reorganized to combat race conditions
7546 - Merge b_intel branch (updated lprocfs code) - now at /proc/fs/lustre
7547 - configure check to avoid gcc version 2.96 20000731-2.96-98) (606)
7549 2003-01-06 Andreas Dilger <adilger@clusterfs.com>
7552 - Fully reactivate OST imports after reconnection (512, others)
7553 - Make sure client sees our -ENOTCONN from mds_handle (513 - partial)
7554 - More graceful error handling for truncating on dead OST (515)
7555 - Don't error out unless we're actually accessing dead stripes (474)
7556 - Fix garbage sizes when stripes are missing (410)
7557 - LRU counters were broken, causing constant lock purge (433, 432)
7558 - garbage on read from stripes with failed OSTs (441)
7559 - mark OSCs as active before reconnecting during recovery (438)
7560 - lov_enqueue and lov_cancel need to handle inactive OSTs (403)
7561 - lfind did not preserve OST order in output (443)
7562 - symlinks cause hung clients, incorrect data (439)
7563 - stop dereferencing request after dropping refcount (457)
7564 - don't LASSERT(spin_is_locked) on non-SMP (455)
7565 - fixes for many rename() bugs
7566 - fstat didn't correctly synchronize attributes (399)
7567 - server must handle lock cancellation during blocking AST prep (487)
7568 - bulk descriptors were free()d too soon (511)
7569 - fix paths in lconf, which would load incorrect modules (451, 507)
7570 - fix confusing lconf 'host not found' error message (386)
7571 - fix lock order deadlock on OST (O/R i_sem before journal ops, 478)
7572 - fix race condition in mdc_blocking_ast() for inode access (526)
7573 - fix lov_unpackmd() unpacking wrong number of stripes (537)
7574 - fix lov_set_osc_active() marking wrong OSC inactive (440)
7575 - fix bad lstripe lov_unpackmd() assertion (fix layering too) (527)
7576 - fix multiple writes of stripe MD to MDS (358, maybe 519)
7577 - fix lstripe in several ways (kernel side) (527)
7578 - fix request leak in ldlm_cli_enqueue (262)
7579 - incorrect OSC was marked inactive after OST failure
7580 - call mds_fs_cleanup before unmounting filesystem (524)
7581 - fix races between taking ns_lock and ldlm_lock_change_resource
7582 - fix races updating LOV export open file list
7583 - fix lov_enqueue error path, avoid decref-ing bad lock handle
7584 - fix recovery NULL deref in ldlm_cli_cancel_unused
7585 - fix some DLM races by using new hash table for lock handles (419)
7586 - permit the client to specify desired inodes, at replay
7587 - duplicate requests when we queue them for replay reintegration
7588 - fix last_rcvd offset calculation
7589 - sync after each recovered transaction, so we always make progress
7590 - never, not always, ERESTART requests without transnos
7591 - store the lov_desc in the MDS, so we don't depend on getlovinfo to
7593 - skip replay if the MDS says that the client is already connected
7594 - don't check for a recovery-enabled export to match lctl's UUID
7595 - don't INC_USE_COUNT for phantom exports
7596 - don't crash when cleaning up phantom exports (567)
7597 - don't double-finish or set replay data for errored mdc_open requests
7598 - abort requests when they time out, so we don't get old replies
7599 - send/receive replies for AST messages again
7600 - if the client says that it doesn't have the lock, cancel it on the
7602 - if we timeout during I/O, don't try to cancel an in-use lock; instead
7603 mark it as destroyed, it will all work out when decref is called
7604 - fix module use counts (22, 581)
7606 - ASTs now expect a reply (server cancels lock on error reply)
7608 2002-12-02 Andreas Dilger <adilger@clusterfs.com>
7611 - fix many simultaneous client startup (392)
7612 - fix dentry->d_it clobbering
7613 - credentials weren't being shipped for readdir/getattr operations
7614 - remove invalid assertions triggered during some concurrent MD
7616 - proper Lustre versions added (336, 389)
7617 - fix memory leak for create error case (398)
7618 - fix LOV locking bug that would get cli/srv out of sync
7619 - fix echo client over LOV (409)
7620 - fix dbench 2, extN refcount problem (170, 258, 356, 418)
7621 - fix double-O_EXCL intent crash (424)
7622 - avoid sending multiple lock CANCELs (352)
7624 - MDS can do multi-client recovery (modulo bugs in new code)
7626 - many updates, edits, cleanups
7628 2002-11-18 Phil Schwan <phil@clusterfs.com>
7631 - fix null d_it dereference (346)
7632 - fix full OST/dbench hang (333)
7633 - fix permission problem with file removal (286)
7634 - fix removal of OSCs from LOV when they fail
7635 - fix NULL deref during bulk timeout (214)
7636 - fix problems related to multiple filesystems on one MDS (241)
7637 - fixed serious subtle metadata locking bugs
7638 - free locks on clients when inodes are removed due to memory
7640 - fix inode pointer in lock data (285)
7641 - partial support for multiple MDS on a single host (241)
7642 - data locks weren't cancelled at clear_inode time (290, 311)
7643 - intent locks could lead to unbounded lock growth (205)
7644 - added a maximum lock count, an LRU list, and a flusher
7645 - fix multiple rename (365)
7646 - properly abstracted the echo client
7647 - OSC locked 1 byte too many; fixed
7648 - rewrote brw callback code:
7649 - fixed recovery bugs related to LOVs (306)
7650 - fixed too-many-pages-in-one-write crash (191)
7651 - fixed (again) crash in sync_io_timeout (214)
7652 - probably fixed callback-related race (385)
7654 - Add capability to MDS protocol
7655 - LDLM cancellations and callbacks on different portals
7657 2002-10-28 Andreas Dilger <adilger@clusterfs.com>
7660 - limit client IOV size to PTL_MD_MAX_IOV (611336, 191)
7661 - defer open object destruction to close time (601981, 138)
7662 - open/close OST file handle in obdo (OBD_MD_FLHANDLE) (601981, 138)
7663 - move LDLM_ENQUEUE/CONVERT back to MDS portal (625069)
7664 - abstract ll_lookup2, fix ll_revalidate2 to use abstraction (256)
7665 - don't call obd_setattr in ll_file_release for destroyed objects
7666 * protocol change to lustre_msg: move |version| and add |flags|
7667 * protocol change to osc_punch: "start" in "o_size", "end" in "o_blocks"
7668 * lock replay: for LDLM_FL_REPLAY trust client to do right thing
7669 * added replay of create, unlink, link and rename operations during
7670 MDS failover; recovery should be much more robust now
7671 * remove failed OSCs from LOVs (only lov_create uses this so far)
7672 * the lustre-HOWTO was brought (more) up to date (582544)
7674 2002-10-23 Phil Schwan <phil@clusterfs.com>
7677 - in-use dentries weren't being reused properly (617851)
7678 - prevent multiple LDLM setup (599178)
7679 - fix LOV size calculations for truncate (617853)
7680 - fix client handling of MDS intent errors (POSIX)
7681 - fix permission bug in lovstripe.c test (624321)
7682 - fix MDS thread deadlock - move LDLM handler to DLM portal (625069)
7683 - truncate past end of file could corrupt data
7684 - proper cleanup after timeouts, crashes, etc (592524, 550815)
7685 - a race in recovery could return ETIMEDOUT to apps (623947)
7686 - building outside the source directory was fixed
7687 * the lustre-HOWTO was brought (more) up to date (582544)
7688 * major progress was made on recovery functionality
7690 2002-10-10 Phil Schwan <phil@clusterfs.com>
7693 - recovery deadlock fix
7694 - rm -rf causes LBUG fix (617817)
7695 - file open by multiple tasks fix (618962)
7696 - directory permissions bugs (602707 and 620007)
7697 - journal_stop fixed with locking (611313)
7698 - O_APPEND failures resolved (618273, perhaps 614459)
7699 - lconf PATH fix (619770)
7700 - IA64 build fix (621450)
7701 - RPC buffer sizes scale with amount of memory
7703 2002-10-01 Phil Schwan <phil@clusterfs.com>
7706 - locks would be cancelled without throwing away data pages,
7707 resulting in inconsistent data (605627)
7708 - inode attributes were not always being refreshed (605627, 612449)
7709 - lconf now continues to cleanup after lctl reports an error
7710 - MDS now enforces user permissions (602707)
7711 - lprocfs cleanup fixed, but not yet enabled (614157)
7712 - fixed infinite server hang, should a client not respond to an AST
7713 - avoid going into recovery if user calls readlink() with a buffer
7714 that's too small (613941)
7715 - AST RPCs no longer require replies (614867) -- this may be changed
7716 - don't crash server if client sends an IOV that's too big (611336)
7717 - fixed lock conversion deadlock (611892)
7718 - fixed the following of symlinks (614622)
7719 * recovery: the server can remove locks from a client that dies, other
7720 clients can make progress
7721 * more extN patch fixes
7722 * compile-time configurable ptlrpc buffer allocations
7724 - collaborative read cache document
7725 - Lustre Lite Performance CDR document-in-progress
7727 2002-09-20 Andreas Dilger <adilger@clusterfs.com>
7730 - fix typo in patch-2.4.18
7732 2002-09-20 Andreas Dilger <adilger@clusterfs.com>
7735 - clear ptlrpc request each time in handle_incoming_request()
7736 - unlink of files now destroys the object on the OST
7738 2002-09-19 Peter Braam <braam@clusterfs.com>
7740 * add hard link support
7741 * change obdfile creation method
7742 * kernel patch changed
7744 2002-09-19 Peter Braam <braam@clusterfs.com>
7747 - stack overflow bug in extN fixed
7749 2002-09-18 Andreas Dilger <adilger@clusterfs.com>
7751 * documentation updates
7752 - add man pages for config tools
7753 - update tests/README to describe testing with new config tools
7754 - finish metadata API descriptions
7755 * bug fixes and cleanups
7756 - statfs workaround for 16TB limit
7757 - LOV stripe allocation improved, can stripe on subset of OSTs
7758 - LOV file size/IO offset was wrong for files > 4GB in size
7759 - object EA data was being dropped, caused files to be unreadable
7760 - memory overflow with non-LOV OST caused memory corruption
7761 - fixed regression tests to work with new config tools, obdfilter
7762 - fixed bug when directory size became larger than 1 block
7763 - fixed bug (for single client case) when PWD was deleted
7764 - invalidate local directory pages when doing intent-based ops
7765 - avoid LDLM oops when lock callback contained bad data
7767 2002-09-09 Andreas Dilger <adilger@clusterfs.com>
7769 * documentation updates
7770 * bug fixes and cleanups
7771 - configuration tools
7774 - 64-bit compile warnings
7775 - 64-bit internal statfs data
7777 * test_brw on persistent OST devices
7779 * lprocfs (disabled)
7781 2002-09-04 Andreas Dilger <adilger@clusterfs.com>
7783 * documentation updates
7784 * bug fixes and cleanups
7785 * configuration tools
7787 2002-08-30 Peter J. Braam <braam@clusterfs.com>
7790 * many small fixes to 0_5_4
7791 * io/network handling
7792 * thinkos in MDS operations
7794 2002-08-24 Peter J. Braam <braam@clusterfs.com>
7797 * crucial basic fixes to 0.5.3
7798 * IOR, Iozone work over Elan
7801 2002-08-07 Phil Schwan <phil@clusterfs.com>
7802 * version 0_5_3, our first alpha
7803 * we use the new Portals iovs
7804 * documentation updates
7805 * bug fixes and cleanups
7806 * small changes in the DLM wire protocol
7808 2002-07-25 Peter J. Braam <braam@clusterfs.com>
7809 * version 0_5_1 with some initial stability,
7810 * locking on MD and file I/O.
7811 * documentation updates
7812 * several bug fixes since 0.5.0
7813 * small changes in wire protocol
7815 2002-07-18 Phil Schwan <phil@clusterfs.com>
7817 * delivered as Lustre Light Alpha
7818 * fixed a crash after handling invalid MDS requests
7819 * fixed directory pages for architectures with non-4k pages sizes
7821 2002-07-11 Andreas Dilger <adilger@clusterfs.com>
7822 * release version v0_4_4
7823 * Moves TCP acceptor to be on port 2432 (unused Coda port) instead
7825 * Fixes a number of interruption problems with OST operations.
7826 * Update documentation for portals header changes
7827 * Move all wire protocol structs/defines to lustre_idl.h
7828 * Fixes symlink length bug.
7829 * Add tcpdump to repository.
7831 2002-07-05 Andreas Dilger <adilger@clusterfs.com>
7832 * release version v0_4_3
7833 * Fixes statfs for inodes on extN.
7834 * Fixes bug in runtests which would delete /etc/hosts.
7835 * Use 64-bit object IDs wherever possible (not into VFS though)
7836 Remove ost_get_info, which is unused by lustre, and out of date.
7838 2002-07-03 Peter Braam <braam@clusterfs.com>
7839 * release version v0_4_2 Fixes a lookup error (type not passed)
7840 * move forward to head of Portals
7841 * move forward to latest Lustre kernel
7843 2002-06-25 Peter Braam <braam@clusterfs.com>
7844 * release version v0_4_1. Hopefully stable on single node use.