1 2010-10-29 Oracle, Inc.
4 2.6.16.60-0.69.1 (SLES 10),
5 2.6.32.19-0.2.1 (SLES11),
6 2.6.18-194.17.1.el5 (RHEL 5)
7 2.6.18-194.17.1.0.1.el5 (OEL 5)
8 * Client support for unpatched kernels:
9 (see http://wiki.lustre.org/index.php?title=Patchless_Client)
10 2.6.16 - 2.6.32 vanilla (kernel.org)
11 * Recommended e2fsprogs version: 1.41.10-sun2
12 * The async journal commit feature (bug 19128) and the cancel
13 lock before replay feature (bug 16774) are disabled by default.
15 Severity : enhancement
17 Description: Update to RHEL5.5 kernel 2.6.19-194.17.1.el5.
18 Update to OEL5.5 kernel 2.6.19-194.17.1.0.1.el5.
20 Severity : enhancement
22 Description: Update to SLES10 SP3 kernel 2.6.16.60-0.69.1.
25 Frequency : only with SLES10
27 Description: Use OFED "KMP" provided by Novell
28 Details : SLES10 SP3 ships with OFED in a separate "KMP" package.
29 Lustre is now built against this package. That means you need to
30 install the ofed-kmp package from Novell for the patchless client
31 and from our download site for the server. Note that the ofed-kmp
32 that Novell ships may not exactly match the kernel version but
33 should still be compatible.
35 Severity : enhancement
37 Description: Update SLES11 SP1 kernel to 2.6.32.19-0.2.1.
41 Description: Enabling quotas fails with non-consecutive OST numbering.
45 Description: Fix kernel warning due to lookup_on_len() called without i_mutex
50 Description: Account direct i/o inflight rpcs separately from non-direct i/o so
51 that direct i/o, which is limited by max_rpcs_in_flight, should not
52 block non-direct i/o, which is not limited by max_rpcs_in_flight.
56 Description: Fix per-NID reporting on outstanding writes
60 Description: Reduce stack pressure by uninlining some mds and ptlrpc functions.
64 Description: Remove LASSERT in lprocfs_rd_conn_uuid() since conn == NULL is a
69 Description: fix obdo leak issue in ll_setattr_raw()
73 Description: limit MMP interval
75 Severity : enhancement
77 Description: add several lfs ost enhancements
81 Description: Too many default ACLs break directory access on new directories
85 Description: Lustre inode size is not coherent across nodes.
86 Details : Update lvbo from disk when AST fails with EINVAL. Lvbo is updated
87 on EINVAL error in ldlm_handle_ast_error(). The updates in
88 filter_intent_policy() and ldlm_cb_interpret() have been removed as
93 Description: Oops at __percpu_counter_add+0x1b
94 Details : Use bdi_init()/bdi_destroy() to proper initialize backing_dev_info
99 Description: add mount option to generate 32bit ino, this can be used for 32bit
100 application compatibility.
104 Description: keep reference count for "lli_sai" to prevent it to be released
105 when "statahead_enter()"
109 Description: allow quotacheck over OSTs with sparse indices
113 Description: Objects not getting deleted for files which have been removed
114 Details : ll_have_md_lock() should differentiate between CR and CW OPEN
119 Description: pin object's inode in memory to avoid certain timeouts
123 Description: fix LBUG when obdfilter-survey is interrupted.
125 -------------------------------------------------------------------------------
127 2010-07-31 Oracle, Inc.
129 * Support for kernels:
130 2.6.16.60-0.42.8 (SLES 10),
131 2.6.27.39-0.3.1 (SLES11),
132 2.6.18-194.3.1.el5 (RHEL 5)
133 2.6.18-194.3.1.0.1.el5 (OEL 5)
134 * Client support for unpatched kernels:
135 (see http://wiki.lustre.org/index.php?title=Patchless_Client)
136 2.6.16 - 2.6.30 vanilla (kernel.org)
137 * Recommended e2fsprogs version: 1.41.10-sun2
138 * The async journal commit feature (bug 19128) and the cancel
139 lock before replay feature (bug 16774) are disabled by default.
141 Severity : enhancement
143 Description: Update RHEL5.5 kernel to 2.6.18-194.8.1.el5 and OEL5.5 kernel to
144 2.6.18-194.8.1.0.1.el5.
146 Severity : enhancement
148 Description: using inkernel OFED stack for rhel5 & oel5.
150 Severity : enhancement
152 Description: Add "lfs_migrate" script from manual into lustre/scripts and RPMs
153 Details : lfs_migrate does a "poor man's" migration of files from their
154 current OST layout to a new OST layout as chosen by the MDS.
158 Description: mds_orphan_add_link()) error linking orphan to PENDING
159 Details : quota limits might disallow linking orphans to PENDING
160 when unlinking a file - temporary raise threads' privileges
161 when processing unlinks.
163 Severity : enhancement
165 Description: add conf-param -d option to remove permanent settings.
166 Details : Add the ability to remove permanent lctl conf_param settings.
167 (Previously conf_param settings could only be changed, not
168 removed.) This also provides a method to change failover
169 nid locations. Improve lctl man page.
171 Severity : enhancement
173 Description: add list_param to b1_8 and add "-R" option to list params
176 Severity : enhancement
178 Description: lfs quota output is not very convenient for awk/sed-parsing
179 Details : Some positions in lfs quota output table could be empty or
180 non-empty which made it hard to parse it with scripts, now a dash
181 is put instead of space where there is not supposed to be any data.
183 Severity : enhancement
185 Description: fix obdfilter-survey script to work properly with remote oss-s
187 Severity : enhancement
189 Description: add new OBDFILTER_SURVEY test suite
191 Severity : enhancement
193 Description: add new multiple mount protection (MMP) test suite
195 Severity : enhancement
197 Description: add support for async journal commit in echo client
199 Severity : enhancement
201 Description: allow userland programs to include <lustre/lustre_idl.h>
202 from stardard include directories
204 Severity : enhancement
206 Description: The prune-icache-use-trylock is no longer needed now that
207 the patch from bug 20008 is landed.
211 Description: The shrink grant feature is still active on the client although the
212 connect flag is not set.
216 Description: Don't leak grant space if the write failed with quota exceeded.
220 Description: Don't consume grant space twice on recoverable resent.
224 Description: a race condition could lead to SIGBUS being sent to an
225 application using mmap-ped files from Lustre
226 Details : truncate_complete_page implementation for the patchless
227 client could arbitrarily unset PG_Uptodate flag for a
228 page being kicked from the page cache, an uptodate check
229 right after a readpage call in filemap_fault could fail
230 because of that as though the page read had been unsuccessful.
234 Description: dlm lock slab shrinking is not efficient
235 Details : The dlm_locks slab can grow significantly and consumes a lot of
236 memory on the server. Set a hardlimit to grant_plan.
240 Description: Lustre does not do 1MB IOs to HW RAID
241 Details : Bump MAX_PHYS/HW_SEGMENTS and SG_ALL to 256 in the RHEL5 kernel.
242 This is what we do already for SLES kernels.
246 Description: bump maximum number of phys/hw segments in the SLES11 kernel
247 until s/g chaining works properly.
252 Description: LSI Fusion MPT driver hacks to improve performance
253 Details : Set CONFIG_FUSION_MAX_SGE to 256 for RHEL5
255 Severity : enhancement
257 Description: increase default md stripe_cache_size to 16k
260 Bugzilla : 15587/21439
261 Description: don't handle security.capability xattr
262 Details : CONFIG_SECURITY_FILE_CAPABILITIES is enabled by default on SLES11.
263 This results in additional getxattr calls, causing VBR test
264 failures as well as a preformance drop when writing.
268 Description: obdfilter-survey is no longer working
269 Details : revert patch from bug 20355 to resolve an issue with lctl --threads
270 not working correctly with $(PTHREAD_LIBS) being linked to lctl.
274 Description: ll_shrink_cache does not handle __GFP_FS properly
278 Description: lfs getstripe shows wrong info for directories
279 Details : Set correct LOVEA default values for filesystem-wide.
283 Description: FSX checksum false positves due to mmap IO
284 Details : Use OBD_FL_MMAP flag for IOs on a memory mapped file. Do not print
285 checksum errors, if the flag is set on a request.
289 Description: file operations after eviction have successful return values
290 Details : use vfs ->flush callback to return any pending async errors
295 Description: mdsrate fails to write after 1.3+M files opened
296 Details : decrease memory usage on clients by recycling dentries and
301 Description: obdfilter-survey gives unreasonably high numbers
302 Details : Wait for all threads to complete when running test_brw.
306 Description: do not set lustre read_only device when server umount and keep
307 client records for recoverable ones
311 Description: move sync_on_lock_cancel tunable to the obdfilter layer
312 Details : move the tunable to trigger a journal flush on lock cancel from
313 the ost layer to the obdfilter layer. This tunable is useful
314 when using the async journal commit feature.
318 Description: exp->exp_nid_stats == NULL in filter_tally()
319 Details : fix race with per-nid stats by delaying procfs cleanup until
324 Description: extent lock cancellation on client can keep the cpu busy for too
329 Description: Do not fail OST activation when a llog is not found, just
330 issue an error message.
334 Description: Don't enable extents by default for MDT.
338 Description: Protect bitfield access to ptlrpc_request's rq_flags, since
339 the AT code can access it concurrently while sending early
344 Description: Disable lockless truncate by default since it is sometimes flawed
345 and causes the write_disjoint test to fail.
349 Description: OSSs which don't have the patch from bug 20278 can trigger an
352 Severity : enhancement
354 Description: don't print message to the console when we have not managed to
359 Description: The MDS fails to synchronize OSTs which registered with the MGS
360 after the MDT. The problem is that OBD_NOTIFY_CREATE events are
361 raised too early and thus discarded by the MDT stack.
362 The fix consists of issuing OBD_NOTIFY_CREATE event in the lov
367 Description: Fix race when the ping evictor and a service thread execute
368 target_recovery_check_and_stop() concurrently.
372 Description: quota broadcast can trigger a LBUG on the MDT if there are
377 Description: Resetting the lov_objid values to last_id reported by the OST
378 during orphan recovery is incorrect and can cause the same
379 objects to be allocated twice.
381 Severity : enhancement
383 Description: "weak-modules" support
384 Details : Implement "weak-modules" support which enables kernel modules
385 to be used with any kernel that implements the same kABI. In
386 order to achieve this modules are now installed in
387 /lib/modules/$(uname -r)/updates/kernel on all distributions.
389 Severity : enhancement
391 Description: add writeconf as mount option
393 Severity : enhancement
395 Description: produce debuginfo packages for SLES.
397 Severity : enhancement
399 Description: add failover nidlist to the import proc file.
401 Severity : enhancement
403 Description: fix LUSTRE_SEQ_MAX_WIDTH for interoperability between 1.8
404 clients and 2.0 servers.
406 Severity : enhancement
408 Description: lfs find -s does not work correctly because of a bug in
413 Description: ll_read_ahead_page() must validate the dlm lock before using
418 Description: Prevent failover nids from registering with MGS first.
422 Description: fix lock inversion in ll_setattr_raw().
426 Description: object allocation is not balanced across OSTs.
427 Details : osc_precreate() should return 0, if there are enough objects left.
429 -------------------------------------------------------------------------------
431 2010-04-30 Oracle, Inc.
433 * Support for kernels:
434 2.6.16.60-0.42.8 (SLES 10),
435 2.6.27.39-0.3.1 (SLES11),
436 2.6.18-164.11.1.el5 (RHEL 5)
437 2.6.18-164.11.1.0.1.el5 (OEL 5)
438 * Client support for unpatched kernels:
439 (see http://wiki.lustre.org/index.php?title=Patchless_Client)
440 2.6.16 - 2.6.30 vanilla (kernel.org)
441 * Recommended e2fsprogs version: 1.41.10-sun2
442 * The async journal commit feature (bug 19128) and the cancel
443 lock before replay feature (bug 16774) are disabled by default.
447 Description: open-unlinked directories trigger MDS LBUG
448 Details : Fix regression introduced by the patch from bug 19640.
449 ext3_inc_count() can reset nlink to 1 when the directory
450 is indexed and inode->i_nlink == 2. Work around the problem
451 by incrementing nlink by 2 instead of 1.
455 Description: Reconnects are not throttled
456 Details : Don't wake up pinger on reconnect failures and rely on regular
457 pings to trigger the next reconnection. Please note that the
458 pinger already uses a smaller interval if the import is
462 Frequency : only with NFS export
464 Description: Console flooded with error message from ll_inode_from_lock()
466 Details : in mds_open, initialize the child_res_id before enqueuing
467 the OPEN lock for the child inode, then to avoid senting
468 wrong ldlm_res_id to client.
472 Description: allow multiple instances of the same nid in NID hash
473 Details : Case of multiple separate clients from the same NID (as
474 with liblustre) is legitimate and so we should allow
475 multiple instances of the same NID in nid hash.
479 Description: the readahead code can sleep on a semaphore while holding a
481 Details : in ras_update, "lov_get_info" could be called during increasing
482 readahead windows, which tries to get the mutex lock "lov_lock"
483 while holding the spin_lock "ras_lock", then causes system
488 Description: ASSERTION(cli->cl_avail_grant >= 0) failed
489 Details : fix assertion failure in the grant code.
493 Description: Use CNETERR (which is rate limited) in specific places in
494 the portal's LNET driver to avoid flooding the console.
498 Description: include last created object in precreate slow case
502 Description: don't do rep-ack if not created anything
503 Details : mds_open currently always put a lock into a rep-ack regardless
504 if something was created or not. This is pointless and only
505 creates needless contention. In fact the entire idea was to do
506 this for real creates as a recovery protection.
510 Description: Spurious error messages from smp_processor_id() on preemptible
512 Details : Disable a preemption by grabbing the lock in fs_trace_get_tcd()
513 first. The function fs_trace_get_tcd() was moved up.
517 Description: interval_erase() fix
518 Details : interval_erase() calls update_maxhigh() properly when child
523 Description: "lfs df" does not print stats for all mountpoints
524 Details : Print all mounted lustre filesystems with "lfs df"
528 Description: lfs setstripe -p not longer work with a relative pathname
529 Details : Use realpath() to provide absolute pathname.
533 Description: fix for truncated reply buffer
534 Details : reply buffer could be referred by reply_in_callback after released
539 Description: lustre.lov error when backing up symlinks with extended attributes
540 Details : Improved logic in ll_listxattr()
544 Description: properly handle null value for setattr -n lustre.lov
545 Details : Running "setfattr -n trusted.lov ." causes a NULL dereference
546 in ll_setxattr() due to no checking if "value" is NULL.
547 This command now resets to the default striping when executed
552 Description: stack overflow on lock cancellation due to fsync call
553 Details : sync_on_lock_cancel is needed for recovery when async journal
554 is enabled, but we actually just need to make sure that
555 metadata blocks have hit the journal, so doing a fs sync
556 should be enough and should consume less stack (just create an
557 empty handle and commmit it).
561 Description: using current->journal_info to store per-thread data leads
562 to problem under memory pressure
563 Details : disable the per-thread data (current->journal_info) containing
564 the lock info during I/O to work around the issue for short tem
568 Description: control DCACHE_LUSTRE_INVALID flag with MDS_INODELOCK_LOOKUP lock
569 Details : DCACHE_LUSTRE_INVALID is controlled by MDS_INODELOCK_LOOKUP
570 lock which is corresponding to "IT_LOOKUP", do not skip invalidate
575 Description: Cannot send after transport shutdown
576 Details : Clear imp_vbr_failed flag upon eviction
580 Description: soft lock in request set code during recovery
581 Details : during recovery, uses req->rq_set itself to replay the request
582 instead of ptlrpcd_recovery_pc
586 Description: Use CFS_ALLOC_IO instead of _STD in llap_from_page_with_lockh
587 Details : During an ll_readahead under ll_readpage, we have seen the the
588 OBD_SLAB_ALLOC hang under ldlm_pools_shrink when trying to lock
589 a page that is already locked by the readahead code.
594 Description: stop waitting for next replay transno if shutdown
595 Details : if the system is shutting down, wake up service thread blocked
596 to wait for next replay transno during recovery, then all the
597 references held by queued requests can be dropped and device
602 Description: lov_merge_lvb()) ASSERTION(spin_is_locked(&lsm->lsm_lock)) failed
603 Details : Protect lli->lli_smd pointer updates with lli->lli_lock.
607 Description: per-nid stats should not access lustre-hash internal structures
612 Description: mount.lustre fails to pass some options to mount()
617 Description: ext4 extent allocation is slower than in ext3
618 Details : Increase the default value of MB_DEFAULT_ORDER2_REQS to 8,
619 enlarge ext4 preallocation table for 2048 4K blocks extents
624 Description: incorrect triggering of synchronous IO
625 Details : The OSC can mistakenly fall back to synchronous IO when the
626 max_dirty_mb limit is reached and no write requests have yet
627 been issued. This can occur when the dirty pages are spread
628 over many files all of which are below the optimal request size.
632 Description: Optimize quota_ctl operations by sending requests in parallel
633 Details : Send MDS->OST quota_ctl requests in parallel, do not resend.
634 Compiled from two attachments in the ticket.
638 Description: ordering issue between transaction start & i_mutex
639 Details : start the transaction earlier in llog_lvfs_destroy to get
640 transaction start and inode mutex lock nested properly.
644 Description: lru resize SLV can get stuck
645 Details : calculate SLV with a greater precision to not lose small
646 changes due to interger math truncation; round up SLV only
647 if the amount of granted locks less than the limit to not
648 get stuck with this SLV
652 Description: avoid divide-by-zero in lprocfs_rd_import()
656 Description: lfs quota failed when OSTs are down
657 Details : really return approximate block/inode usage when OSTs are down
661 Description: llobdstat fix and enhancement
662 Details : add a counter to set a limit to how many samples will be returned
663 fix a wildcard in the path to limit to obdfilter stats only
667 Description: recovery-small 51 hang - MDT can not stop, Mount still busy with
669 Details : abort lock enqueue processing sooner when we umount is in progress.
671 Severity : enhancement
673 Description: Update RHEL5.4 kernel to 2.6.18-164.11.1.el5 and OEL5.4 kernel to
674 2.6.18-164.11.1.0.1.el5.
676 Severity : enhancement
678 Description: error message improvements
679 Details : Use INFO/WARN instead of WARN/ERROR for the slow messages.
680 Simplify MDT/OST service start message.
681 Suppress "changing the import ..." warning.
684 Bugzilla : 21961/17914
685 Description: ignore trailing -mdc when determining index number
688 Frequency : only with SLES11
690 Description: fix for a race condition in linux quotas implementation
691 Details : dq_flags(struct dquot) access is not properly locked which could
692 lead to certain inconsistencies when accessing it using non-atomic
693 bit operations like __set_bit in do_set_dqblk.
697 Description: return any pending async errors in close(2) using flush callback
699 -------------------------------------------------------------------------------
701 2010-01-29 Sun Microsystems, Inc.
703 * Support for kernels:
704 2.6.16.60-0.42.8 (SLES 10),
705 2.6.27.39-0.3.1 (SLES11),
706 2.6.18-164.11.1.el5 (RHEL 5)
707 2.6.18-164.6.1.0.1.el5 (OEL 5)
708 * Client support for unpatched kernels:
709 (see http://wiki.lustre.org/index.php?title=Patchless_Client)
710 2.6.16 - 2.6.30 vanilla (kernel.org)
711 * Recommended e2fsprogs version: 1.41.6.sun1
712 * The async journal commit feature (bug 19128) and the cancel
713 lock before replay feature (bug 16774) are disabled by default.
715 Severity : enhancement
717 Description: Update RHEL5.4 kernel to 2.6.18-164.11.1.el5 and
718 OEL5.4 kernel to 2.6.18-164.11.1.0.1.el5.
720 Severity : enhancement
721 Bugzilla : 21511/19848
722 Description: Update SLES11 kernel to 2.6.27.39-0.3.1.
724 Severity : enhancement
726 Description: Update supported SLES10 kernel to 2.6.16.60-0.42.8.
728 Severity : enhancement
730 Description: Update kernel to RHEL5.4 2.6.18-164.6.1.el5 and
731 OEL5 2.6.18-164.6.1.0.1.el5(Both in-kernel OFED enabled).
733 Severity : enhancement
735 Description: Build kernels (RHEL5, OEL5 and SLES10/11) using the vendor's own
738 Severity : enhancement
740 Description: Vanilla kernel 2.6.30 patchless client support.
745 Description: bad entry in directory xxx: inode out of bounds
746 Details : fix locking issue in the rename path which could race with any
747 other operations updating the same directory.
749 Severity : enhancement
751 Description: Make watchdog timer messages to be more clear and descriptive.
755 Description: cp -p command does not preserve the dates and timestamp
756 Details : mtime could be spoiled by a write callback
760 Description: Clear imp_force_reconnect correctly in ptlrpc_connect_interpret()
764 Description: Allow non-root access for "lfs check".
765 Details : Added a check in obd_class_ioctl() for OBD_IOC_PING_TARGET.
767 Severity : enhancement
769 Description: quotacheck performance/scaling issues
770 Details : reduce quotacheck time on empty filesystem by skipping uninit
773 Severity : enhancement
775 Description: Enhancement for lfs(1) command to use numeric uid/gid.
777 Severity : enhancement
779 Description: Adjust locks' extents on their first enqueue, so that at the time
780 they get granted, there is no need for another pass through the
781 queues since they are already shaped into the proper forms.
785 Description: Fix mds_shrink_intent_reply()/mds_intent_policy() to pass correct
786 arguments and prevent LBUG() in lustre_shrink_reply_v2().
790 Description: Change tunefs.lustre and mkfs.lustre --mountfsoptions so that
791 exactly the specified mount options are used. Leaving off
792 any "mandatory" mount options is an error. Leaving off any
793 default mount options causes a warning, but is allowed.
794 Change errors=remount-ro from mandatory to default. Sanitize
795 the mount string before storing it. Update man pages accordingly.
799 Description: mds_getattr() should return 0, even if mds_fid2entry() fails with
800 -ENOENT. Also fix in ptlrpc_expire_one_request() to print signed
803 Severity : enhancement
805 Description: Remove set_info(KEY_UNLINKED) from MDS/OSC
807 Severity : enhancement
809 Description: Clients can replay thousands of unused locks during recovery
810 Details : Don't replay unused locks (only read locks for now) during
811 recovery. This feature is disabled by default and can be
812 enabled by running the following command on the clients:
813 lctl get_param ldlm.cancel_unused_locks_before_replay
817 Description: can't stat file in some situation.
818 Details : improve initialize osc date when target is added to mds and
819 ability to resend too big getattr request is client isn't have info
824 Description: Prevent inconsistences between linux and lustre mount structures.
825 Details : Wait indefinitely in server_wait_finished() until mnt_count drops.
826 Make the sleep interruptible.
828 Severity : enhancement
830 Description: Communicate OST degraded/readonly state via statfs to MDS
831 Details : Flags in the statfs returned from OSTs indicate whether the
832 OST is in a degraded RAID state, or if the filesystem has
833 turned read-only after a filesystem error is detected.
838 Description: don't panic if EPROTO was hit when reading symlink
839 Details : correctly handling request reference in error cases.
844 Description: open sometimes returns ENOENT instead of EACCES
845 Details : checking permission should be part of open part of mds_open, not
846 lookup part. so server should be set DISP_OPEN_OPEN disposition
847 before starting permission check.
848 Also not need revalidate dentry if client already have LOOKUP lock.
852 Description: enable client interface failover
853 Frequency : on servers with multiple network interfaces
854 Details : When a child reconnects from another NID, properly update export
855 nid hash position and ldlm reverse import.
857 Severity : enhancement
859 Description: implemented direct I/O with arbitrary (nonaligned) memory
860 addresses and file offsets.
862 Severity : enhancement
864 Description: added more recovery timeout options.
866 Severity : enhancement
868 Description: added llapi_file_open, llapi_file_create, llapi_file_get_stripe
873 Description: Avoid deadlock for local client writes
874 Frequency : only on systems with clients writing to an OST on the same node
875 Details : Use new OBD_BRW_MEMALLOC flag to notify OST about writes in the
876 memory freeing context. This allows OST threads to set the
877 PF_MEMALLOC flag on task structures in order to allocate memory
878 from reserved pools and complete IO.
879 Use GFP_HIGHUSER for OST allocations for non-local client writes,
880 so that the OST threads generate memory pressure and allow
881 inactive pages to be reclaimed.
886 Descriptoin: lock ordering violation between &cli->cl_sem and _lprocfs_lock
887 Details : move ldlm namespace creation in setup phase to avoid grab
888 _lprocfs_lock with cli_sem held.
892 Frequency : only during format of test systems
893 Description: Unable to run several mkfs.lustre on loop devices at the same time
894 Details : mkfs.lustre returns error 256 on the concurrent loop devices
895 formatting. The solution is to proper handle the error.
897 Severity : enhancement
899 Description: implement async create (obd_async_create) method for osc, to avoid
900 too long waiting new ost objects with holding ldlm lock.
904 Frequency : occasionally during network problems
905 Description: client not allowed to reconnect to OST because of active request
906 Details : abort bulk requests received by the OST once the client has timed
907 out since the client will resend the request anyway. The client
908 also now retries to reconnect to the same server if a connect
909 request failed with EBUSY or -EAGAIN.
912 Frequency : rare, if used wide striped file and one ost in down.
914 Descriptoin: don't return error if we created a subset of objects for file.
915 Details : lov_update_create_set() uses set->set_success as index for created
916 objects, so if some requests failed, they will have hole at end of
917 array and we can use qos_shrink_lsm for allocate correct lsm.
921 Description: Slow stale export processing during normal start up
922 Details : The global mgc lock prevents OST setup to be run in parallel.
923 Replace the global lock with a per-config_llog_data semaphore.
927 Description: Out or order replies might be lost on replay
928 Details : In ptlrpc_retain_replayable_request if we cannot find retained
929 request with tid smaller then one currently being added, add it
930 to the start, not end of the list.
934 Description: BUG: soft lockup - CPU#1 stuck for 10s! [ll_mdt_07:4523]
935 Details : add cond_resched() calls to avoid hogging the cpu for too long
936 in the hash code. Make also lustre_hash_for_each_empty() more
939 Severity : enhancement
941 Description: Performance improvements for debug messages with D_RPCTRACE,
942 D_LDLM, D_QUOTA options.
945 Frequency : only with NFS export
947 Description: (lov_merge.c:74:lov_merge_lvb())
948 ASSERTION(spin_is_locked(&lsm->lsm_lock)) failed (SR 71691004)
949 Details : Fix a race in the nfs export code by populating inode
950 info while the new inode is still locked
952 Severity : enhancement
954 Description: add a new file in procfs called force_lbug. Writting to this
955 file triggers a LBUG. Only for test purpose.
959 Description: OOM killer causes node hang
960 Details : really interrupt the sleep in osc_enter_cache on signals
964 Description: LustreError: 9153:0:(quota_context.c:622:dqacq_completion()) LBUG
965 Details : fix race during quota release on the slave.
967 Severity : enhancement
969 Description: smaller hash bucket sizes, cleanups
970 Details : increase hash table sizes and enabled rehashing for pools, quota,
971 uuid, nid & per-nid stats.
973 Severity : enhancement
975 Description: Add ldiskfs maxdirsize mount option
976 Details : add max_dir size mount option
980 Description: panic in ll_statahead_thread
981 Details : prevent parent thread to be killed before its child
984 Frequency : only with 16TB device
986 Description: unable to perform "mount -t lustre" of 16TB OST device
987 Details : Mounting 16TB LUNs failed due to three bugs in mkfs.lustre.
991 Description: ASSERTION(atomic_read(&imp->imp_inflight) == 0) failed
992 Details : unregistering should be zero if no RPC inflight.
996 Description: hyperion: Oops during metabench
997 Details : Correct the refcount of lov_request_set
999 Severity : enhancement
1001 Description: Add mptlinux and nxge drivers to Lustre builds
1003 Severity : enhancement
1005 Description: Fix watchdog timer message to be more clear
1006 Details : Make watchdog timer messages more clear and descriptive.
1010 Description: LNET soft lockups in socknal_cd thread
1011 Details : don't hog CPU for active-connecting if another connd is
1012 accepting connecting-requst from the same peer
1016 Description: recovery-small test_17 hang
1017 Details : Land several AT improvements & fixes.
1021 Description: MDS panic and hanging client processes
1022 Details : Replace exp_ops_stats with exp_nid_stats->nid_stats
1026 Description: OSS stuck in recovery.
1027 Details : fix race during recovery. class_unlink_export,
1028 class_set_export_delayed and target_queue_last_replay_reply
1029 may race while increasing/decreasing obd_recoverable_clients
1030 and obd_delayed_clients, causing recovery to wait forever.
1032 Severity : enhancement
1034 Description: add cascading_rw.c to lustre/tests
1038 Description: filter_last_id() NULL deref
1039 Details : lprocfs_filter_rd_last_id() should check for the fully
1040 setup obd device, before proceeding further.
1042 Severity : enhancement
1044 Description: Loadgen improvements
1045 Details : stacksize and locking fixes for loadgen
1049 Description: Quiet CERROR("dirty %d > system dirty_max %d\n"
1050 Details : The atomic_read() allowing the atomic_inc() are not covered
1051 by a lock. Thus they may safely race and trip this CERROR()
1052 unless we add in a small fudge factor (+1).
1054 Severity : enhancement
1056 Description: shrink_slab: nr=-9223362083340912175
1057 Details : fix spurious message from shrink_slab reporing negative nr
1061 Description: Quiet bogus previously committed transno error
1062 Details : suppress the "server went back in time" error message which
1063 is always printed even in the common case after a client eviction
1065 Severity : enhancement
1067 Description: Parallel statfs() calls result in client eviction
1068 Details : cache statfs data for 1s.
1072 Description: parallel-scale test_compilebench: @@@@@@ FAIL: compilebench
1074 Details : fix serveral issues in pinger code causing clients not to ping
1075 servers for too long, resulting in evictions.
1079 Description: e2fsck should warn when MMP update interval is extended
1080 Details : print mmp_check_interval and make it possible to abort mount
1081 operation in case it takes too long.
1085 Description: mdsrate-create-large.sh, BUG: soft lockup - CPU#0 stuck for 10s!
1086 Details : fix bug in the RHEL5's jbd2 callback patch.
1090 Description: drop number of active requests when queued for recovery
1091 Details : Now that we take a reference on the original request instead of
1092 making a copy of it for recovery. We need to drop the number of
1093 active requests or the queued requests will prevent all request
1094 processing when they exceed (srv->srv_threads_running - 1).
1096 Severity : enhancement
1098 Description: refuse to invalidate operational quota files when they are in use
1099 Details : an attempt to invalidate operational quota files on the quota
1100 master is not actually permitted by VFS (returning -EPERM), but we
1101 should not depend on that and should return the error earlier.
1105 Description: Applications stuck in jbd2_log_wait_commit during exit
1106 Details : fix deadlock between kjournald2 trying to acquire the page lock
1107 owned by an ost_io thread waiting for journal commit.
1109 -------------------------------------------------------------------------------
1111 2009-10-16 Sun Microsystems, Inc.
1113 * Support for kernels:
1114 2.6.16.60-0.42.4 (SLES 10),
1115 2.6.27.29-0.1 (SLES11, i686 & x84_64 only),
1116 2.6.18-128.7.1.el5 (RHEL 5),
1117 2.6.18-128.7.1.el5 (OEL 5),
1118 * Client support for unpatched kernels:
1119 (see http://wiki.lustre.org/index.php?title=Patchless_Client)
1120 2.6.16 - 2.6.27 vanilla (kernel.org)
1121 * Recommended e2fsprogs version: 1.41.6.sun1
1122 * ext4 support for RHEL5 is experimental and thus should not be
1125 Severity : enhancement
1127 Description: Add OEL5 support.
1129 Severity : enhancement
1131 Description: Update kernel to SLES11 2.6.27.29-0.1.
1135 Description: File checksum failures with OST read cache on
1136 Details : Disable page poisoning when the bulk transfer has to be aborted
1137 because the client got evicted.
1141 Description: Don't allow make backward step on assiging osc next id.
1142 Details : race between allocation next id and ll_sync thread can be cause
1143 of set wrong osc next id and can be kill valid ost objects.
1145 Severity : enhancement
1147 Description: Update kernel to RHEL5 2.6.18-128.7.1.el5.
1149 Severity : enhancement
1151 Description: Update kernel to SLES10 SP2 2.6.16.60-0.42.4.
1155 Description: Changes in raid5-large-io-rhel5.patch to calculate sectors properly
1159 Description: Increase the default BLK_DEF_MAX_SECTORS value for RHEL5 and SLES11
1163 Description: Error handling in osc_statfs_interpret() has been improved.
1164 Details : Check in osc_statfs_interpret() for EBADR.
1168 Description: Do not send statfs() requests to OSTs disabled by administrator.
1169 Details : Check in lov_prep_statfs_set() for non-NULL ltd_exp.
1173 Description: Do not update ctime for the deleted inode.
1174 Details : Check in mds_reint_unlink() before calling fsfilt_setattr().
1178 Description: Increase of the size of the LDLM resource hash.
1179 Details : Bump up RES_HASH_BITS=12.
1183 Description: correctly send lsm on open replay
1184 Details : MDS is trust to LSM size on replay open, but client can set wrong size
1189 Description: Deadlock between filter_destroy() and filter_commitrw_write().
1190 Details : filter_destroy() does not hold the DLM lock over the whole
1191 operation. If the DLM lock is dropped, filter_commitrw() can go
1192 through, causing the deadlock between page lock and i_mutex.
1193 The i_alloc_sem should also be hold in filter_destroy() while
1194 truncating the file.
1198 Description: truncate starts GFP_FS allocation under transaction causing deadlock
1199 Details : ldiskfs_truncate calls grab_cache_page which may start page
1200 allocation under an open transaction. This may lead to
1201 calling prune_icache with consequent lustre reentrance.
1204 Frequency : only when down/upgrading the MDS to 1.6/1.8 while 1.8 clients are
1205 still up and when the OST pool feature is used
1207 Description: interop testing got LBUG when run dd with OST pool
1208 :LustreError: 30032:0:(llite_lib.c:1913:ll_replace_lsm()) LBUG
1209 Details : down/upgrading the MDS to a version that doesn't/does support OST
1210 pool can cause clients to crash because the lsm has changed
1215 Description: missing tree_status on 1.8.1 RPM build
1216 Details : make rpms failed due because the tree_status file is missing.
1220 Description: continuing LustreError "mds adjust qunit failed!"
1221 Details : don't print message on the console when ->adjust_qunit fails.
1225 Description: don't increase ldlm timeout if previous client was evicted
1226 Details : if a client doesn't respond to a blocking callback within the
1227 adaptive ldlm enqueue timeout, don't adjust the adaptive estimate
1228 when the lock is next granted.
1232 Description: ost is being unmounted w/o all writes to last_rcvd landing on disk.
1233 affects recovery negatively.
1234 Details : make sure all exports have been properly destroyed by the zombie
1235 thread processed before stopping the target.
1239 Description: Performance degradation with O_DIRECT between 1.6 & 1.8.1 b190
1240 Details : disable write barrier for ext4/SLES11.
1244 Description: Kernel panic - not syncing: Out of memory and no killable
1245 processes... on OSS when iozone
1246 Details : fix memory leak in the journal checksum patch.
1250 Description: group quota "too many blocks" OSS crashes
1251 Details : we should keep the same uid/gid for lquota_chkquota() and
1252 lquota_pending_commit()
1256 Description: LustreError: 9153:0:(quota_context.c:622:dqacq_completion()) LBUG
1257 Details : don't LBUG on release quota error. Just a workaround until the
1258 problem is understood.
1260 ------------------------------------------------------------------------------
1261 2009-07-31 Sun Microsystems, Inc.
1263 * Support for kernels:
1264 2.6.16.60-0.39.3 (SLES 10),
1265 2.6.27.23-0.1 (SLES11, i686 & x84_64 only),
1266 2.6.18-128.1.14.el5 (RHEL 5),
1267 * Client support for unpatched kernels:
1268 (see http://wiki.lustre.org/index.php?title=Patchless_Client)
1269 2.6.16 - 2.6.27 vanilla (kernel.org)
1270 * Recommended e2fsprogs version: 1.41.6.sun1
1271 * File join has been disabled in this release, refer to Bugzilla 16929.
1272 * NFS export disabled when stack size < 8192. Since the NFSv4 export
1273 of Lustre filesystem with 4K stack may cause a stack overflow. For
1274 more information, please refer to bugzilla 17630.
1275 * ext4 support for RHEL5 is experimental and thus should not be
1278 Severity : enhancement
1280 Description: Update kernel to SLES10 SP2 2.6.16.60-0.39.3.
1283 Frequency : with 1.8 server and 1.6 clients
1285 Description: correctly shrink reply for avoid send too big message to client.
1286 Details : 1.8 mds is allocate to big buffer to LOV EA data and this produce
1287 some problems with sending this reply to 1.6 client.
1291 Description: Repeated atomic allocation failures.
1292 Details : Use GFP_HIGHUSER | __GFP_NOMEMALLOC flags for memory allocations
1293 to generate memory pressure and allow reclaiming of inactive pages.
1294 At the same time, do not allow to exhaust emergency pools.
1295 For local clients the use of GFP_NOFS will be introduced in 1.8.2
1297 Severity : enhancement
1298 Bugzilla : 19846, 18289
1299 Description: Update kernel to RHEL5 2.6.18-128.1.14.el5.
1301 Severity : enhancement
1302 Bugzilla : 19625, 16893, 18668, 19848
1303 Description: Add support for SLES11 2.6.27.23-0.1.
1305 Severity : enhancement
1307 Description: Update client support to vanila kernels up to 2.6.27.
1309 Severity : enhancement
1311 Description: Update kernel to SLES10 SP2 2.6.16.60-0.37.
1313 Severity : enhancement
1315 Description: Compile with -Werror by default for i686 and x86_64.
1319 Description: resolve race between obd_disconnect and class_disconnect_exports
1320 Details : if obd_disconnect will be called to already disconnected export he
1321 forget release one reference and osc module can't unloaded.
1323 Severity : enhancement
1325 Description: move AT tunable parameters for more consistent usage
1326 Details : add AT tunables under /proc/sys/lustre, add to conf_param parsing
1330 Descriptoin: correctly skip time estimate if in recovery
1331 Details : rq_send_state insn't bitmask so using bitwise ops is forbid.
1335 Descriptoin: OSS DeadLock
1336 Details : Use trylock to prevent deadlock when shrink icache.
1338 Severity : enhancement
1340 Description: Allow tuning service thread via /proc
1341 Details : For each service a new
1342 /proc/fs/lustre/{service}/*/thread_{min,max,started} entry is
1343 created that can be used to set min/max thread counts, and get the
1344 current number of running threads.
1346 Severity : enhancement
1348 Description: Add state history info file, enhance import info file
1349 Details : Track import connection state changes in a new osc/mdc proc file;
1350 add overview-type data to the osc/mdc import proc file.
1354 Description: Reduce small size read RPC
1355 Details : Set read-ahead limite for every file and only do read-ahead when
1356 available read-ahead pages are bigger than 1M to avoid small size
1361 Description: free_entry erroneously used groups_free instead of put_group_info
1363 Severity : enhancement
1365 Description: Make read-ahead stripe size aligned.
1367 Severity : enhancement
1369 Description: MDS create should not wait for statfs RPC while holding DLM lock.
1372 Frequency : rare, connect and disconnect target at same time
1374 Descriptoin: ASSERTION(atomic_read(&imp->imp_inflight) == 0
1375 Details : don't call obd_disconnect under lov_lock. this long time
1376 operation and can block ptlrpcd which answer to connect request.
1379 Frequency : start MDS on uncleanly shutdowned MDS device
1381 Descriptoin: ll_sync thread stay in waiting mds<>ost recovery finished
1382 Details : stay in waiting mds<>ost recovery finished produce random bugs
1383 due race between two ll_sync thread for one lov target. send
1384 ACTIVATE event only if connect realy finished and import have
1388 Frequency : start MDS on uncleanly shutdowned MDS device
1390 Descriptoin: aborting recovery hang on MDS
1391 Details : don't throttle destroy RPCs for the MDT.
1395 Description: Slow reads beyond 8Tb offsets.
1396 Details : Page index integer overflow in ll_read_ahead_page
1400 Description: Soft lockup on OSS after MDS failover
1401 Details : MSG_CONNECT_INITIAL is not set on the initial MDS->OST connect.
1402 As a conseqence, the patch from bug 18224 is not operational
1403 and the MDS export cannot be reused on the OSTs until it gets
1407 Frequency : rare, only if using MMP with Linux RAID
1409 Description: MMP doesn't work with Linux RAID
1410 Details : While using HA for Lustre servers with Linux RAID, it is possible
1411 that MMP will not detect multiple mounts. To make this work we
1412 need to unplug the device queue in RAID when the MMP block is being
1413 written. Also while reading the MMP block, we should read it from
1414 disk and not the cached one.
1417 Frequency : rare, during recovery
1419 Description: Assertion failure in ldlm_lock_put
1420 Details : Do not put cancelled locks into replay list, hold references on
1421 locks in replay list
1425 Description: 1.6.5 mdsrate performance is slower than 1.4.11/12 (MDS is not cpu bound!)
1426 Details : create_count always drops to the min value (=32) because grow_count
1427 is being changed before the precreate RPC completes.
1430 Frequency : Only in RHEL5 when mounting multiple ext3 filesystems
1433 Description: "kmem_cache_create: duplicate cache jbd_4k" error message
1434 Details : add proper locking for creation of jbd_4k slab cache
1438 Description: MMP check in ext3_remount() fails without displaying any error
1439 Details : When multiple mount protection fails during remount, proper error
1444 Description: Rare Client crash on resend if the file was deleted.
1445 Details : When file is opened, but open reply is lost and file is
1446 subsequently deleted before resend, resend processing logic
1447 breaks trying to open the file again, should not try to open.
1451 Description: add check for >8TB ldiskfs filesystems
1452 Details : ext3-based ldiskfs does not support greater than 8TB LUNs.
1453 Don't allow >8TB ldiskfs filesystems to be mounted without
1454 force_over_8tb mount option
1458 Description: Client locked up when running multiple instances of an app. on
1459 multiple mount points
1460 Details : ll_shrink_cache() can sleep while holding the ll_sb_lock.
1461 Convert ll_sb_lock to a read/write semaphore to fix the problem.
1465 Description: Cannot acces an NFS-mounted Lustre filesystem
1466 Details : An NFS client cannot access the Lustre filesystem NFS-mounted
1467 from a Lustre-client exporting the Lustre filesystem via NFS.
1471 Description: panic in ll_statahead_thread
1472 Details : grab dentry reference in parent process.
1475 -------------------------------------------------------------------------------
1477 tbd Sun Microsystems, Inc.
1479 * Support for kernels:
1480 2.6.16.60-0.31 (SLES 10),
1481 2.6.18-128.1.6.el5 (RHEL 5),
1482 2.6.22.14 vanilla (kernel.org)
1483 * Client support for unpatched kernels:
1484 (see http://wiki.lustre.org/index.php?title=Patchless_Client)
1485 2.6.16 - 2.6.22 vanilla (kernel.org)
1486 * Recommended e2fsprogs version: 1.40.11-sun1
1487 * File join has been disabled, refer to Bugzilla 16929.
1488 * A new Lustre ADIO driver is available for MPICH2-1.0.7.
1489 * NFS export disabled when stack size < 8192. Since the NFSv4 export of
1490 Lustre filesystem with 4K stack may cause a stack overflow. For more
1491 information, please refer to bugzilla 17630.
1493 Severity : enhancement
1495 Description: Update to RHEL5.3 kernel-2.6.18-128.1.6.el5.
1497 Severity : enhancement
1499 Description: Update OFED release to 1.4.1 RC4
1501 Severity : major, only with big OST
1503 Description: Very poor metadata performance on Infiniband lustre configuration
1504 Details : OST object precreation becomes very slow on big OSTs. This is due
1505 to the ialloc patch spending too much time scanning groups.
1508 Frequency : during recovery
1510 Description: don't mix llog inodes with normal.
1511 Details : allocate inodes for log in last inode group
1516 Description: fix lqs' reference which won't be put in some situations
1517 Details : This patch fixes:
1518 1. In quota_check_common(), this function will check quota
1519 for user and group, but only send one return via "pending".
1520 In most cases, the pendings should be same. But that is not
1522 2. If quotaoff runs between lquota_chkquota() and
1523 lquota_pending_commit(), the same thing will happen too.
1524 That is why it comes:
1525 - if (!ll_sb_any_quota_active(qctxt->lqc_sb))
1528 -------------------------------------------------------------------------------
1530 2008-12-31 Sun Microsystems, Inc.
1532 * Support for kernels:
1533 2.6.5-7.314 (SLES 9),
1534 2.6.9-67.0.22.EL (RHEL 4),
1535 2.6.16.60-0.31 (SLES 10),
1536 2.6.18-92.1.17.el5 (RHEL 5),
1537 2.6.22.14 vanilla (kernel.org)
1538 * Client support for unpatched kernels:
1539 (see http://wiki.lustre.org/index.php?title=Patchless_Client)
1540 2.6.16 - 2.6.22 vanilla (kernel.org)
1541 * Client support for unpatched kernels:
1542 we do not recommend using patchless RHEL4 clients with kernels
1543 prior to 2.6.9-55EL (RHEL4U5).
1544 * Recommended e2fsprogs version: 1.40.11-sun1
1545 * Note that reiserfs quotas are disabled on SLES 10 in this kernel.
1546 * RHEL 4 and RHEL 5/SLES 10 clients behaves differently on 'cd' to a
1547 removed cwd "./" (refer to Bugzilla 14399).
1548 * A new quota file format has been introduced in 1.6.5.
1549 The format conversion from prior releases is handled transparently,
1550 but releases older than 1.4.12/1.6.5 don't understand this new
1551 format. The automatic format conversion can be avoided by running
1552 the following command on the MDS:
1553 'tunefs.lustre --param="mdt.quota_type=ug1" $MDTDEV'.
1554 For more information, please refer to bugzilla 13904.
1555 * A new quota file format was introduced in 1.6.6/1.8.0 (kernels 2.6.16+).
1556 The format conversion from prior releases is handled transparently,
1557 but releases older than 1.6.6/1.8.0 don't understand this new
1558 format. The automatic format conversion can be avoided by running
1559 the following commands on the MDS and OSS servers (for
1560 pre 1.4.12-1.6.5 quota files):
1561 'tunefs.lustre --param="mdt.quota_type=ug1" $MDTDEV',
1562 'tunefs.lustre --param="ost.quota_type=ug1" $MDTDEV'
1563 or (for 1.4.12/1.6.5 quota files)
1564 'tunefs.lustre --param="mdt.quota_type=ug2" $MDTDEV',
1565 'tunefs.lustre --param="ost.quota_type=ug2" $MDTDEV'
1566 For more information, please refer to bugzilla 13904.
1567 * Output of lfs quota has been made less detailed by default,
1568 old (verbose) output can be obtained by using -v option.
1569 * File join has been disabled in this release, refer to Bugzilla 16929.
1570 * A new Lustre ADIO driver is available for MPICH2-1.0.7.
1571 * NFS export disabled when stack size < 8192. Since the NFSv4 export of
1572 Lustre filesystem with 4K stack may cause a stack overflow. For more
1573 information, please refer to bugzilla 17630.
1575 Severity : enhancement
1577 Description: Caching OSS
1578 Details : introduce data caching on the OSS. The OSS now relies on the linux
1579 kernel page cache to keep recently accessed data in memory.
1580 It is worth noting that all write requests are still flushed
1581 synchronously as in lustre 1.6.
1583 Severity : enhancement
1585 Description: Version based recovery
1586 Details : introduce finer grained recovery able to detect transaction
1587 dependencies and can deal with transaction gaps caused by clients
1588 failing at the same time as the server.
1590 Severity : enhancement
1592 Description: Enable adaptive timeouts by default
1593 Details : The Lustre timeout value in /proc/sys/lustre/timeout is now
1594 managed dynamically based on server load and should not need
1595 to be tuned manually based on cluster size. This allows Lustre
1596 to work under a wider variety of system sizes and loads, without
1597 unnecessarily causing lengthy recovery times.
1599 Severity : enhancement
1601 Description: Add OST Pools support
1602 Details : File striping can now be set to use an arbitrary pool of OSTs
1604 Severity : enhancement
1606 Description: add lazystatfs mount option to allow statfs(2) to skip down OSTs
1607 Details : allow skip disconnected ost for send statfs request and hide error
1611 Frequency : rare, on llog test 6
1613 Descriptoin: don't allow connect to already connected import
1614 Details : allowing connect to already connected import is hide connecting problem.
1617 Frequency : rare, on failed llog setup
1619 Descriptoin: don't leak obd reference on failed llog setup
1620 Details : for failed llog setup - mgc forget call class_destroy_import for
1621 client import, move destroy import to more generic place.
1626 Descriptoin: allow kill process which wait statahead result
1627 Details : for some reasons 'ls' can stick in waiting result from statahead,
1628 in this case need way for kill this process.
1631 Frequency : rare, at shutdown
1633 Descriptoin: panic at umount
1634 Details : llap_shrinker can be raced with killing super block from list and
1635 this produce panic with access to already freeded pointer
1640 Descriptoin: don't lose wakeup for imp_recovery_waitq
1641 Details : recover_import_no_retry or invalidate_import and import_close can
1642 both sleep on imp_recovery_waitq, but we was send only one wakeup
1648 Descriptoin: panic in mds_open
1649 Details : don't confuse mds_finish_transno() with PTR_ERR(-ENOENT)
1654 Descriptoin: stuck in cache_remove_extent() or panic with accessing to already
1656 Details : release lock refernce only after add page to pages list.
1659 Frequency : always with long access acl
1661 Descriptoin: mds can't pack reply with long acl.
1662 Details : mds don't control size of acl but they limited by reint/getattr
1667 Frequency : on remount
1669 Description: external journal device not working after the remount
1670 Details : clear dev_rdonly flag for external journal devices in
1676 Description: shutdown vs evict race
1677 Details : client_disconnect_export vs connect request race.
1678 if client will evicted at this time - we start invalidate
1679 thread without referece to import and import can be freed
1685 Description: shrink LOV EAs before replying
1686 Details : correctly adjust LOV EA buffer for reply.
1691 Description: don't skip ost target if they assigned to file
1692 Details : Drop slow OSCs if we can, but not for requested start idx.
1693 This means "if OSC is slow and it is not the requested
1694 start OST, then it can be skipped, otherwise skip it only
1695 if it is inactive/recovering/out-of-space.
1697 Severity : enhancement
1699 Description: Update to RHEL5 kernel-2.6.18-92.1.17.el5.
1702 Frequency : rare, need acl's on inode.
1704 Description: client can't handle ost additional correctly
1705 Details : if ost was added after client connected to mds client can have
1706 hit lnet_try_match_md ... to big messages to wide striped files.
1707 in this case need teach client to handle config events about add
1708 lov target and update client max ea size at that event.
1710 Severity : enhancement
1712 Description: Update to sles9 kernel-2.6.5-7.314.
1714 Severity : enhancement
1716 Description: Update to SLES10 SP2 kernel-2.6.16.60-0.31.
1719 Frequency : Create a symlink file with a very long name
1721 Description: ldlm_cancel_pack()) ASSERTION(max >= dlm->lock_count + count)
1722 Details : If there is no extra space in the request for early cancels,
1723 ldlm_req_handles_avail() returns 0 instead of a negative value.
1728 Description: mds is deadlocked
1729 Details : in rare cases, inode in catalog can have i_no less than have parent
1730 i_no, this produce wrong order for locking during open, and
1731 parallel unlink can be lock open. this need teach mds_open to grab
1732 locks in resource id order, not at parent -> child order.
1734 Severity : enhancement
1736 Description: Add /proc entry for import status
1737 Details : The mdc, osc, and mgc import directories now have
1738 an import directory that contains useful import data for debugging
1739 connection problems.
1741 Severity : enhancement
1743 Description: Re-disable certain /proc logging
1744 Details : Enable and disable client's offset_stats, extents_stats and
1745 extents_stats_per_process stats logging on the fly.
1748 Frequency : Only on FC kernels 2.6.22+
1750 Description: oops in statahead
1751 Details : Do not drop reference count for the dentry from VFS when lookup,
1752 VFS will do that by itself.
1754 Severity : enhancement
1756 Description: Generic /proc file permissions
1757 Details : Set /Proc file permissions in a more generic way to enable non-
1758 root users operate on some /proc files.
1762 Description: Hitting mdc_commit_close() ASSERTION
1763 Details : Properly handle request reference release in
1764 ll_release_openhandle().
1768 Frequency : only patchless client
1769 Description: add workaround for race between add/remove dentry from hash
1771 Severity : enhancement
1773 Description: Allow OST glimpses to return PW locks
1777 Description: LBUG when llog conf file is full
1778 Details : When llog bitmap is full, ENOSPC should be returned for plain
1783 Description: Prevent import from entering FULL state when server in recovery
1787 Description: service mount cannot take device name with ":"
1788 Details : Only when device name contains ":/" will mount treat it as
1794 Description: replace ptlrpcd with the statahead thread to interpret the async
1795 statahead RPC callback
1799 Frequency : on recovery
1800 Description: I/O failures after umount during fail back
1801 Details : if client reconnected to restarted server we need join to recovery
1802 instead of find server handler is changed and process self eviction
1803 with cancel all locks.
1805 Severity : enhancement
1807 Description: Update to RHEL5 kernel-2.6.18-92.1.10.el5.
1811 Description: Kernel BUG tries to release flock
1812 Details : Lustre does not destroy flock lock before last reference goes
1813 away. So always drop flock locks when client is evicted and
1814 perform unlock regardless of successfulness of speaking to MDS.
1816 Severity : enhancement
1818 Description: Update to SLES10 SP2 kernel-2.6.16.60-0.27.
1820 Severity : enhancement
1822 Description: Upcall on Lustre log has been dumped
1823 Details : Allow for a user mode script to be called once a Lustre log has
1824 been dumped. It passes the filename of the dumped log to the
1825 script, the location of the script can be specified via
1826 /proc/sys/lnet/debug_log_upcall.
1831 Description: avoid messages about idr_remove called for id that is not allocated
1832 Details : Move assigment s_dev for clustered nfs to end of initialization,
1833 for avoid problem with error handling.
1838 Description: avoid Already found the key in hash [CONN_UNUSED_HASH] messages
1839 Details : When connection is reused this not moved from CONN_UNUSED_HASH into
1840 CONN_USED_HASH and this prodice warning when put connection again
1846 Description: avoid ASSERTION(client_stat->nid_exp_ref_count == 0) failed
1847 Details : release reference to stats when client disconnected, not
1848 when export destroyed for avoid races when client destroyed
1849 after main ost export.
1853 Description: more cleanup in mds_lov
1854 Details : add workaround for get valid ost count for avoid warnings about
1855 drop too big messages, not init llog cat under semphore which
1856 can be blocked on reconnect and break normal replay, fix access
1859 Severity : enhancement
1861 Description: Export bytes_read/bytes_write count on OSC/OST.
1865 Description: Early reply size mismatch, MGC loses connection
1866 Details : Apply the MGS_CONNECT_SUPPORTED mask at reconnect time so
1867 the connect flags are properly negotiated.
1871 Description: Properly propagate oinfo flags from lov to osc for statfs
1872 Details : restore missing copy oi_flags to lov requests.
1876 Description: exports in /proc are broken
1877 Details : recreate /proc entries for clients when they reconnect.
1879 Severity : enhancement
1881 Description: Add man pages for llobdstat(8), llstat(8), plot-llstat(8),
1882 : l_getgroups(8), lst(8), routerstat(8)
1883 Details : included man pages for llobdstat(8), llstat(8),
1884 : plot-llstat(8), l_getgroups(8), lst(8), routerstat(8)
1886 Severity : enhancement
1888 Description: Implement lustre ll_show_options method.
1890 Severity : enhancement
1892 Description: Update to SLES9 kernel-2.6.5-7.312.
1894 Severity : enhancement
1896 Description: Update to RHEL4 kernel-2.6.9-67.0.22.EL.
1900 Description: exports in /proc are broken
1901 Details : recreate /proc entries for clients when they reconnect.
1905 Description: don't fail open with -ERANGE
1906 Details : if client connected until mds will be know about real ost count
1907 get LOV EA can be fail because mds not allocate enougth buffer
1912 Description: Resolve device initialization race
1913 Details : Prevent proc handler from accessing devices added to the
1914 obd_devs array but yet be intialized.
1916 Severity : enhancement
1918 Description: configure's --enable-quota should check the
1919 : kernel .config for CONFIG_QUOTA
1920 Details : configure is terminated if --enable-quota is passed but
1921 : no quota support is in kernel
1923 Severity : enhancement
1925 Description: Update to SLES10 SP2 kernel-2.6.16.60-0.23.
1927 Severity : enhancement
1929 Description: Update to RHEL5 kernel-2.6.18-92.1.6.el5.
1933 Frequency : rare, on PPC clients
1934 Description: don't swab ost objects in response about directory, because
1936 Details : bug similar bug 14856, but in different function.
1938 Severity : enhancement
1940 Description: lfs quota tool enhancement
1941 Details : added units specifiers support for setquota, default to
1942 current uid/gid for quota report, short quota stats by
1943 default, nonpositional parameters for setquota, added
1944 llapi_quotactl manual page.
1946 Severity : enhancement
1948 Description: *optional* service tags registration
1949 Details : if the "service tags" package is installed on a Lustre node
1950 When the filesystem is mounted, a local-node service tag will
1951 be created. See http://inventory.sun.com/ for more information
1952 about the Service Tags asset management system.
1956 Description: Client runs out of low memory
1957 Details : Consider only lowmem when counting initial number of llap pages
1960 Frequency : occasional
1962 Description: add refcount for osc callbacks, so avoid panic on shutdown
1964 Severity : enhancement
1966 Description: Update to RHEL4 kernel-2.6.9-67.0.20.
1969 Frequency : testing only
1971 Description: sanity test 65a fails if stripecount of -1 is set
1972 Details : handle -1 striping on filesystem in ll_dirstripe_verify
1975 Frequency : only in unusual configurations
1977 Description: Kernel panic with find ost index.
1978 Details : lov_obd have panic if some OST's have sparse indexes.
1980 Severity : enhancement
1982 Description: Update to RHEL5 kernel-2.6.18-53.1.21.el5.
1985 Frequency : rarely, if filesystem is mounted with -o flock
1987 Description: do not process already freed flock
1988 Details : flock can possibly be freed by another thread before it reaches
1989 to ldlm_flock_completion_ast.
1992 Frequency : rarely, if filesystem is mounted with -o flock
1994 Description: LBUG during stress test
1995 Details : Need properly lock accesses the flock deadlock detection list.
1998 Frequency : rarely, if binaries are being run from Lustre
2000 Description: oops in page fault handler
2001 Details : kernel page fault handler can return two special 'pages' in
2002 error case, don't try dereference NOPAGE_SIGBUS and NOPAGE_OMM.
2005 Frequency : rarely, during shutdown
2007 Description: timeout with invalidate import.
2008 Details : ptlrpcd_check call obd_zombie_impexp_cull and wait request which
2009 should be handled by ptlrpcd. This produce long age waiting and
2010 -ETIMEOUT ptlrpc_invalidate_import and as result LASSERT.
2016 Description: ASSERTION(CheckWriteback(page,cmd)) failed
2017 Details : badly clear PG_Writeback bit in ll_ap_completion can produce false
2021 Frequency : only with broken builds/installations
2023 Description: no LBUG if lquota.ko and fsfilt_ldiskfs.ko are different versions
2024 Details : just return an error to a user, put a console error message
2026 Severity : enhancement
2028 Description: Update to RHEL5 kernel-2.6.18-53.1.19.el5.
2030 Severity : enhancement
2032 Description: Update to RHEL4 kernel-2.6.9-67.0.15.
2034 Severity : enhancement
2036 Description: enable MGS and MDT services start separately
2037 Details : add a 'nomgs' option in mount.lustre to enable start a MDT with
2038 a co-located MGS without starting the MGS, which is a complement
2039 to 'nosvc' mount option.
2042 Frequency : always, on big-endian systems
2044 Description: cleanup in ptlrpc code, related to PPC platform
2045 Details : store magic in native order avoid panic's in recovery on PPC
2046 node and forbid from this error in future. Also fix posibily
2047 of twice swab data. Fix get lov striping to userpace.
2050 Frequency : rarely, if replay get lost on server
2052 Description: server incorrectly drop resent replays lead to recovery failure.
2053 Details : do not drop replay according to msg flags, instead we check the
2054 per-export recovery request queue for duplication of transno.
2057 Frequency : after recovery
2059 Description: precreate to many object's after del orphan.
2060 Details : del orphan st in oscc last_id == next_id and this triger growing
2061 count of precreated objects. Set flag LOW to skip increase count
2062 of precreated objects.
2065 Frequency : rare, on clear nid stats
2067 Description: ASSERTION(client_stat->nid_exp_ref_count == 0)
2068 Details : when clean nid stats sometimes try destroy live entry,
2069 and this produce panic in free.
2072 Frequency : occasionally since 1.6.4
2074 Description: Stack overflow during MDS log replay
2075 Details : ease stack pressure by using a thread dealing llog_process.
2078 Frequency : very rare
2080 Description: MDT cannot be unmounted, reporting "Mount still busy"
2081 Details : Mountpoint references were being leaked during open reply
2082 reconstruction after an MDS restart. Drop mountpoint reference
2083 in reconstruct_open() and free dentry reference also.
2088 Description: wait until IO finished before start new when do lock cancel.
2089 Details : VM protocol want old IO finished before start new, in this case
2090 need wait until PG_writeback is cleared until check dirty flag
2091 and call writepages in lock cancel callback.
2096 Description: mds_mfd_close() ASSERTION(rc == 0)
2097 Details : In mds_mfd_close(), we need protect inode's writecount change
2098 within its orphan write semaphore to prevent possible races.
2101 Frequency : rare, on shutdown ost
2103 Description: don't hit live lock with umount ost.
2104 Details : shrink_dcache_parent can be in long loop with destroy dentries,
2105 use shrink_dcache_sb instead.
2108 Frequency : only when echo_client is used
2110 Description: don't panic with use echo_client
2111 Details : echo client pass NULL as client nid pointer and this produce NULL
2112 pointer dereference.
2115 Frequency : Always on 32-bit PowerPC systems
2117 Description: fix build on PPC32
2118 Details : compile code with -m64 flag produce wrong object file for PPC32.
2123 Description: MDS LBUG: ASSERTION(!IS_ERR(dchild))
2124 Details : In reconstruct_* functions, LASSERTs on both the data supplied
2125 by a client, and the data on disk are dangerous and incorrect.
2126 Change them with client eviction.
2128 Severity : enhancement
2130 Description: skiplist implementation simplification
2131 Details : skiplists are used to group compatible locks on granted list
2132 that was implemented as tracking first and last lock of each lock
2133 group the patch changes that to using doubly linked lists
2137 Description: delete compatibility for 32bit qdata
2138 Details : as planned, when lustre is beyond b1_8, lquota won't support 32bit
2139 qunit. That means servers of b1_4 and servers of b1_8 can't be
2140 used together if users want to use quota.
2143 Frequency : only with administrator action
2145 Description: mount failure if config log has invalid conf_param setting
2146 Details : If administrator specified an incorrect configuration parameter
2147 with "lctl conf_param" this would cause an error during future
2148 client mounts. Instead, ignore the bad configuration parameter.
2151 Frequency : blocks per group < blocksize*8 and uninit_groups is enabled
2153 Description: ldiskfs error: XXX blocks in bitmap, YYY in gd
2154 Details : If blocks per group is less than blocksize*8, set rest of the
2158 Frequency : Application do stride read on lustre
2160 Description: The read performance will drop a lot if the application does
2162 Details : Because the stride_start_offset are missing in stride read-ahead,
2163 it will cause clients read a lot of unused pages in read-ahead,
2164 then the read-performance drops.
2168 Description: more ldlm soft lockups
2169 Details : In ldlm_resource_add_lock(), call to ldlm_resource_dump()
2170 starve other threads from the resource lock for a long time in
2171 case of long waiting queue, so change the debug level from
2172 D_OTHER to the less frequently used D_INFO.
2174 Severity : enhancement
2176 Description: add -gid, -group, -uid, -user options to lfs find
2178 Severity : enhancement
2180 Description: ll_recover_lost_found_objs - recover objects in lost+found
2181 Details : OST corruption and subsequent e2fsck can leave objects in the
2182 lost+found directory. Using the "ll_recover_lost_found_objs"
2183 tool, these objects can be retrieved and data can be salvaged
2184 by using the object ID saved in the fid EA on each object.
2189 Description: this bug _only_ happens when inode quota limitation is very low
2190 (less than 12), so that inode quota unit is 1 at initialization.
2191 Details : if remaining quota equates 1, it is a sign to demonstate that quota
2192 is effective now. So least quota qunit should be 2.
2196 Description: Hung threads in invalidate_inode_pages2_range
2197 Details : The direct IO path doesn't call check_rpcs to submit a new RPC once
2198 one is completed. As a result, some RPCs are stuck in the queue
2203 Description: Procfs and llog threads access destoryed import sometimes.
2204 Details : Sync the import destoryed process with procfs and llog threads by
2205 the import refcount and semaphore.
2209 Description: mds fails to respond, threads stuck in ldlm_completion_ast
2210 Details : Sort source/child resource pair after updating child resource.
2215 Description: kernel BUG at ldiskfs2_ext_new_extent_cb
2216 Details : If insertion of an extent fails, then discard the inode
2217 preallocation and free data blocks else it can lead to duplicate
2222 Description: don't always update ctime in ext3_xattr_set_handle()
2223 Details : Current xattr code updates inode ctime in ext3_xattr_set_handle()
2224 In some cases the ctime should not be updated, for example for
2225 2.0->1.8 compatibility it is necessary to delete an xattr and it
2226 should not update the ctime.
2230 Description: add quota statistics
2231 Details : 1. sort out quota proc entries and proc code.
2232 2. add quota statistics
2237 Description: quotas are not honored with O_DIRECT
2238 Details : all writes with the flag O_DIRECT will use grants which leads to
2239 this problem. Now using OBD_BRW_SYNC to guard this.
2243 Bugzilla : 15713/16362
2244 Description: Assertion in iopen_connect_dentry in 1.6.3
2245 Details : looking up an inode via iopen with the wrong generation number can
2246 populate the dcache with a disconneced dentry while the inode
2247 number is in the process of being reallocated. This causes an
2248 assertion failure in iopen since the inode's dentry list contains
2249 both a connected and disconnected dentry.
2253 Description: assertion failure in ldlm_handle2lock()
2254 Details : fix a race between class_handle_unhash() and class_handle2object()
2255 introduced in lustre 1.6.5 by bug 13622.
2257 Severity : enhancement
2259 Description: superblock lock contention with many SMP cores on one client
2260 Details : several client filesystem locks were highly contended on SMP
2261 NUMA systems with 8 or more cores. Per-CPU datastructures
2262 and more efficient locking implemented to reduce contention.
2267 Description: Kernel BUG: sd_iostats_bump: unexpected disk index
2268 Details : remove the limit of 256 scsi disks in the sd_iostat patch
2273 Description: oops in sd_iostats_seq_show()
2274 Details : unloading/reloading the scsi low level driver triggers a kernel
2275 bug when trying to access the sd iostat file.
2280 Description: Kernel panics during QLogic driver reload
2281 Details : REQ_BLOCK_PC requests are not handled properly in the sd iostat
2282 patch, causing memory corruption.
2287 Description: journal_dev option does not work in b1_6
2288 Details : pass mount option during pre-mount.
2290 Severity : enhancement
2292 Description: Add a FIEMAP(FIle Extent MAP) ioctl for ldiskfs
2293 Details : FIEMAP ioctl will allow an application to efficiently fetch the
2294 extent information of a file. It can be used to map logical blocks
2295 in a file to physical blocks in the block device.
2298 Frequency : only with adaptive timeout enabled
2300 Description: DEBUG_REQ() bad paging request
2301 Details : ptlrpc_at_recv_early_reply() should not modify req->rq_repmsg
2302 because it can be accessed by reply_in_callback() without the
2306 Frequency : only on Cray X2
2308 Description: X2 build failures
2309 Details : fix build failures on Cray X2.
2313 Description: xid & resent requests
2314 Details : Initialize RPC XID from clock at startup (randomly if clock is
2319 Description: quota recovery deadlock during mds failover
2320 Details : This patch includes att18982, att18236, att18237 in bz14840.
2322 1. fix osts hang when mds does failover with quotaon
2323 2. prevent watchdog storm when osts threads wait for the
2328 Description: kernel panic on racer
2329 Details : Do not access dchild->d_inode when IS_ERR(dchild) is true.
2331 Severity : enhancement
2333 Description: Add lustre_start utility to start or stop multiple Lustre servers
2338 Description: Lustre GPF in {:ptlrpc:ptlrpc_server_free_request+373}
2339 Details : In case of memory pressure, list_del() can be called twice on
2340 req->rq_history_list, causing a kernel oops.
2344 Description: (ptllnd_peer.c:557:kptllnd_peer_check_sends()) ASSERTION(!in_interrupt()) failed
2345 Details : fix stack overflow in the distributed lock manager by defering
2346 export eviction after a failed ast to the elt thread instead of
2347 handling it in the dlm interpret routine.
2349 Severity : enhancement
2351 Description: More exported tunables for mballoc
2352 Details : Add support for tunable preallocation window and new tunables for
2353 large/small requests
2357 Description: Detect corruption of block bitmap and checking for preallocations
2358 Details : Checks validity of on-disk block bitmap. Also it does better
2359 checking of number of applied preallocations. When corruption is
2360 found, it turns filesystem readonly to prevent further corruptions.
2364 Frequency : only for big-endian servers
2365 Description: Check for big-endian system while mounting fs with extents feature
2366 Details : Mounting a filesystem with extents feature will fail on big-endian
2367 systems since ext3-based ldiskfs is not supported on big-endian
2368 systems. Can be over-ridden with "bigendian_extents" mount option.
2372 Description: Excessive recovery window
2373 Details : With AT enabled, the recovery window can be excessively long (6000+
2374 seconds). To address this problem, we no longer use
2375 OBD_RECOVERY_FACTOR when extending the recovery window (the connect
2376 timeout no longer depends on the service time, it is set to
2377 INITIAL_CONNECT_TIMEOUT now) and clients report the old service
2378 time via pb_service_time.
2381 Descriptoin: Don't sync journal after every i/o
2382 Details : Implement write RPC replay to allow server replies for write RPCs
2383 before data is on disk. However, this feature is disabled by
2384 default since some issues leading to data corruptions have been
2385 found during recovery (e.g. bug 19128). This feature can be enabled
2386 by running the following command on the OSSs:
2387 lctl set_param obdfilter.*.sync_journal=0
2391 Description: Watchdog triggered on MDS failover
2392 Details : enable OBD_CONNECT_MDT flag when connecting from the MDS so that
2393 the OSTs know that the MDS "UUID" can be reused for the same export
2394 from a different NID, so we do not need to wait for the export to
2399 Descriptoin: Lustre detected file system corruption with inode out of bounds
2400 Details : don't update i_size on MDS_CLOSE for directories. This causes
2401 directory corruptions on the MDT.
2403 -------------------------------------------------------------------------------
2405 2008-05-26 Sun Microsystems, Inc.
2407 * Support for kernels:
2408 2.6.5-7.311 (SLES 9),
2409 2.6.9-67.0.7.EL (RHEL 4),
2410 2.6.16.54-0.2.5 (SLES 10),
2411 2.6.18-53.1.14.el5 (RHEL 5),
2412 2.6.22.14 vanilla (kernel.org)
2413 * Client support for unpatched kernels:
2414 (see http://wiki.lustre.org/index.php?title=Patchless_Client)
2415 2.6.16 - 2.6.22 vanilla (kernel.org)
2416 * Due to problems with nested symlinks and FMODE_EXEC (bug 12652),
2417 we do not recommend using patchless RHEL4 clients with kernels
2418 prior to 2.6.9-55EL (RHEL4U5).
2419 * Recommended e2fsprogs version: 1.40.7-sun1
2420 * Note that reiserfs quotas are disabled on SLES 10 in this kernel.
2421 * RHEL 4 and RHEL 5/SLES 10 clients behaves differently on 'cd' to a
2422 removed cwd "./" (refer to Bugzilla 14399).
2423 * A new quota file format has been introduced in 1.6.5.
2424 The format conversion from prior releases is handled transparently,
2425 but releases older than 1.4.12/1.6.5 will not understand this new
2426 format. The automatic format conversion can be avoided by running
2427 the following command on the MDS before upgrading:
2428 'tunefs.lustre --param="mdt.quota_type=ug1" $MDTDEV'.
2429 For more information, please refer to bugzilla 13904.
2433 Description: quota performance fix
2434 Details : quota data is written in journalled mode instead of ordered to
2435 increase performance
2439 Description: lfs support for human-readable quota grace time strings
2440 Details : lfs setquota -t and lfs quota -t represent quota grace times
2441 in "XXwXXdXXhXXmXXs" format instead of large values in seconds
2444 Frequency : always with o2ib 1.3 and sles10
2446 Description: fix build with SLES10 and o2ib v3.
2447 Details : sles10 uses diffrent name for Module.symver file but configure
2448 assume this file has same name on RHEL/SLES/vanila kernels.
2451 Frequency : very rare, if additional xattrs are used on kernels >= 2.6.12
2453 Description: MDS may lose file striping (and hence file data) in some cases
2454 Details : If there are additional extended attributes stored on the MDS,
2455 in particular ACLs, SELinux, or user attributes (if user_xattr
2456 is specified for the client mount options) then there is a risk
2457 of attribute loss. Additionally, the Lustre file striping
2458 needs to be larger than default (e.g. striped over all OSTs),
2459 and an additional attribute must be stored initially in the
2460 inode and then increase in size enough to be moved to the
2461 external attribute block (e.g. ACL growing in size) for file
2464 Severity : enhancement
2466 Description: add message levels for liblustreapi
2471 Description: MDT cannot be unmounted, reporting "Mount still busy"
2472 Details : Mountpoint references were being leaked during open reply
2473 reconstruction after an MDS restart. Drop mountpoint reference
2474 in reconstruct_open() and free dentry reference also.
2479 Description: fix for occasional failure case of -ENOSPC in recovery-small tests
2480 Details : Move the 'good_osts' check before the 'total_bavail' check. This
2481 will result in an -EAGAIN and in the exit call path we call
2482 alloc_rr() which will with increasing aggressiveness attempt to
2483 aquire precreated objects on the minimum number of required OSCs.
2487 Description: Use old size assignment to avoid deadlock
2488 Details : This reverts the changes in bugs 2369 and bug 14138 that introduced
2489 the scheduling while holding a spinlock. We do not need locking
2490 for size in ll_update_inode() because size is only updated from
2491 the MDS for directories or files without objects, so there is no
2492 other place to do the update, and concurrent access to such inodes
2493 are protected by the inode lock.
2497 Description: Use __u64 instead of int for valid bits
2501 Description: resolve "_IOWR redefined" build error on SLES10
2505 Description: dump the memory debugging after all modules are unloaded to
2506 suppress false negative in conf_sanity test 39
2510 Description: the recovery timer never expires
2511 Details : for new client connect request, the recovery timer should not be
2512 reset, otherwise recovery timer will never expired, if the old
2513 client never come. Only old client connect and first connection
2514 req should trigger recovery timer reset.
2518 Description: the min numbers of lproc stats are wrong
2519 Details : adding a new constant LC_MIN_INIT and use it for initialization
2523 Frequency : always with interactive lfs
2525 Description: Reinitialize optind to 0 so that interactive lfs works in all cases
2528 Frequency : with multiple concurrent readdir processes in same directory
2529 Bugzilla : 15406, 15169, 15175
2530 Description: misc fixes for directory readahead.
2531 Details : prevent previous statahead async RPC callback from processing the
2532 current "statahead_info", race condition between async RPC callback
2533 add dentry into dentry hash table and "ls" thread revalidate such
2534 dentry, statahead his/miss control for hidden items, and so on.
2536 Severity : enhancement
2538 Description: build kernel-ib packages for OFED 1.3 in our release cycle
2542 Description: incore types cleaning in quota code (with respect to 64-bit limits)
2543 Details : several u32 variables declarations are replaced with u64 declarations
2548 Description: fix SLES kernel versioning
2549 Details : the kernel version for our SLES 10 kernel did not include a "-"
2550 before the "smp" at the end. while this was not a problem in
2551 general, it did mean that software trying to use the kernel
2552 version to try to detect a vendor specific kernel would fail.
2553 this was most evident by the OFED build scripts.
2558 Description: Don't update lov_desc members until making sure they are valid
2559 Details : When updating lov_desc members via proc fs, need fix their
2560 validities before doing the real update.
2563 Frequency : very rare
2565 Description: don't put request into delay list while invalidate in flight.
2566 Details : ptlrpc_delay_request sometimes put in delay list while invalidate
2567 import in flight. this produce timeout for invalidate and sometimes
2568 can cause stale data.
2570 Severity : enhancement
2572 Description: Update kernel to SLES9 2.6.5-7.311.
2574 Severity : enhancement
2576 Description: Update kernel to RHEL4 2.6.9-67.0.7.
2581 Frequency : on PPC only
2582 Description: not convert ost objects for directory because it's not exist.
2583 Details : ll_dir_getstripe assume dirrectory has ost objects but this wrong.
2585 Severity : enhancement
2587 Description: Fix warnings with compile liblustre at sles10/rhel5 which have
2588 __u64 as usingied long long type.
2591 Frequency : rare, on shutdown
2593 Description: race process ast vs remove callback
2594 Details : removing callback before disconnect import open race with
2595 processing callback.
2597 Severity : enhancement
2599 Description: Update kernel to SLES9 2.6.5-7.311.
2601 Severity : enhancement
2603 Description: Files open for execute are not marked busy on SLES10
2604 Details : Add FMODE_EXEC to SLES10 SP1 server kernel series.
2606 Severity : enhancement
2608 Description: Add server support for vanilla-2.6.22.14.
2611 Frequency : occasional
2613 Description: Avoid lov_create() getting stuck in obd_statfs_rqset()
2614 Details : If an OST is down the MDS will hang indefinitely in
2615 obd_statfs_rqset() waiting for the statfs data. While for
2616 MDS QOS usage of statfs, it should not stuck in waiting.
2618 Severity : enhancement
2620 Description: Disable adaptive timeouts by default
2623 Frequency : on network error
2625 Description: panic with double free request if network error
2626 Details : mdc_finish_enqueue is finish request if any network error occuring,
2627 but it's true only for synchronus enqueue, for async enqueue
2628 (via ptlrpcd) this incorrect and ptlrpcd want finish request
2632 Frequency : rare, on recovery
2634 Description: read procfs can produce deadlock in some situation
2635 Details : Holding lprocfs lock with send rpc can produce block for destroy
2636 obd objects and this also block reconnect with -EALREADY.
2637 This isn't fix all lprocfs bugs - but make it rare.
2639 Severity : enhancement
2641 Description: Update kernel to RHEL5 2.6.18-53.1.14.el5.
2644 Frequency : frequent on X2 node
2646 Description: mdc_set_open_replay_data LBUG
2647 Details : Set replay data for requests that are eligible for replay.
2652 Description: lustre_mgs: operation 101 on unconnected MGS
2653 Details : When MGC is disconnected from MGS long enough, MGS will evict the
2654 MGC, and late on MGC cannot successfully connect to MGS and a lot
2655 of the error messages complaining that MGS is not connected.
2658 Frequency : rare, depends on device drivers and load
2660 Description: MDS or OSS nodes crash due to stack overflow
2661 Details : Code changes in 1.6.4 increased the stack usage of some functions.
2662 In some cases, in conjunction with device drivers that use a lot
2663 of stack, the MDS (or possibly OSS) service threads could overflow
2664 the stack. One change which was identified to consume additional
2665 stack has been reworked to avoid the extra stack usage.
2667 Severity : enhancement
2669 Description: Update to RHEL5 latest kernel-2.6.18-53.1.13.el5.
2671 Severity : enhancement
2673 Description: Update to SLES10 SP1 latest kernel-2.6.16.54-0.2.5.
2675 Severity : enhancement
2677 Description: Update to RHEL5 latest kernel-2.6.18-53.1.6.el5.
2679 Serverity : enhancement
2681 Description: Update RHEL4 kernel to 2.6.9-67.0.4.
2684 Frequency : rare on shutdown OST
2686 Description: Don't allow skipping OSTs if index has been specified.
2687 Details : Don't allow skipping OSTs if index has been specified, make locking
2688 in internal create lots better.
2693 Description: ASSERTION(!PageDirty(page)) failed
2694 Details : Wrong check could lead to an assertion failure under specific
2700 Description: LBUG in ptlrpc_check_set() bad phase ebc0de00
2701 Details : access to bitfield in structure is always rounded to long
2702 and this produce problem with not atomic change any bit.
2707 Description: Lustre make rpms failed.
2708 Details : Remove ldiskfs spec file to avoids rpmbuild be confused when
2709 builds Lustre rpms from tarball.
2711 Severity : enhancement
2713 Description: Update to SLES9 SP4 kernel-2.6.5-7.308.
2716 Frequency : rare on shutdown OST
2718 Description: If llog cancel was not send before clean_exports phase, this can
2719 produce deadlock in llog code.
2720 Details : If llog thread has last reference to obd and call class_import_put
2721 this produce deadlock because llog_cleanup_commit_master wait when
2722 last llog_commit_thread exited, but this never success because was
2723 called from llog_commit_thread.
2726 Frequency : only if OST index is skipped
2728 Description: NULL lov_tgts causing MDS oops
2729 Details : more safe checks for NULL lov_tgts for avoid oops.
2731 Severity : enhancement
2733 Description: Update to RHEL4 latest kernel-2.6.9-67.0.1.EL.
2735 Severity : enhancement
2737 Description: Update to RHEL5 latest kernel-2.6.18-53.1.4.el5.
2742 Description: make mgs_setparam() handle fsname containing dash
2743 Details : fsname containing a dash does not work with lctl conf_param
2745 Severity : enhancement
2747 Description: Update to RHEL4 Update-6 kernel-2.6.9-67.EL.
2750 Frequency : rare, in recovery and (or) destroy lovobjid file.
2752 Description: rewrite lov objid code.
2753 Details : Cleanup for lov objid code, remove scability problems and wrong
2754 locking. Fix sending last_id into ost.
2756 Severity : enhancement
2758 Description: Update to SLES10 SP1 latest kernel-2.6.16.54-0.2.3.
2760 Severity : enhancement
2762 Description: Update to RHEL5 Update-1 kernel 2.6.18-53.el5.
2763 Details : Use d_move_locked instead of __d_move.
2766 Frequency : rare, at shutdown
2768 Description: access already free / zero obd_namespace.
2769 Details : if client_disconnect_export was called without force flag set,
2770 and exist connect request in flight, this can produce access to
2771 NULL pointer (or already free pointer) when connect_interpret
2772 store ocd flags in obd_namespace.
2775 Frequency : only at startup
2777 Description: not alloc memory with spinlock held.
2778 Details : allocation memory with GFP_KERNEL can produce sleep deadlock,
2779 if any spinlock held.
2784 Description: lfs find does not continue on file error
2785 Details : Continue other files processing when a file/dir is absent.
2789 Description: Inconsistent usage of lustre_pack_reply()
2790 Details : Standardize the usage of lustre_pack_reply() such that it
2791 always generate a CERROR on failure.
2794 Frequency : very rare
2796 Description: Fix replay if there is an un-replied request and open
2797 Details : In some cases, older replay request will revert the
2798 mcd->mcd_last_xid on MDS which is used to record the client's
2799 latest sent request.
2801 Severity : enhancement
2803 Description: Update to RHEL5 kernel 2.6.18-8.1.15.el5.
2805 Severity : enhancement
2807 Description: Update to SLES10 SP1 kernel 2.6.16.53-0.16
2809 Severity : enhancement
2811 Description: Update to SLES9 kernel-2.6.5-7.287.3.
2813 Severity : enhancement
2815 Description: Update to RHEL4 kernel-2.6.9-55.0.12.EL.
2817 Severity : enhancement
2819 Description: Build SLES10 patchless client fails
2820 Details : The configure was broken by run ./configure with
2821 --with-linux-obj=.... argument for patchless client. When the
2822 configure use --with-linux-obj, the LINUXINCLUDE= -Iinclude
2823 can't search header adequately. Use absolute path such as
2824 -I($LINUX)/include instead.
2826 Severity : enhancement
2828 Description: Lustre Page Accounting
2829 Details : New macros for page alloc and free which enable accounting
2830 of page allocation of Lustre. Use percpu counters to store memory
2831 and page statistics.
2834 Frequency : only if debugging is disabled
2836 Description: LASSERT_{REQ,REP}SWAB macros are buggy
2837 Details : If SWAB_PARANOIA is disabled, the LASSERT_REQSWAB and
2838 LASSERT_REPSWAB macros become no-ops, which is incorrect. Drop
2839 these macros and replace them with their definitions instead.
2844 Description: interrupt oig_wait produce painc on resend.
2845 Details : brw_redo_request can be used for resend requests from ptlrpcd and
2846 private set, and this produce situation when rq_ptlrpcd_data not
2847 copyed to new allocated request and triggered LBUG on assert
2848 req->rq_ptlrpcd_data != NULL. But this member used only for wakeup
2849 ptlrpcd set if request is changed and can be safety changed to use
2852 Severity : enhancement
2854 Description: organize the server-side client stats on per-nid basis
2855 Details : Change the structure of stats under obdfilter and mds to
2863 The "uuid"s file would list the uuids of _active_ exports.
2864 And the clear entry is to clear all stats and stale nids.
2869 Description: Processes looping in ll_readdir() on Lustre clients finally causing
2870 a full node pseudo-hang
2871 Details : Concurrent access to the same directory from multiple clients with
2872 intensive file creation/removal can cause a client node to spin in
2873 ll_readdir(). i_version must be increased every time the lock is
2874 cancelled to ensure a revalidate is done.
2879 Description: touch file failed when fs is not full
2880 Details : OST in recovery should not be discarded by MDS in alloc_qos(),
2881 otherwise we can get ENOSP while fs is not full.
2884 Frequency : only for Cray XT3
2885 Bugzilla : 12829/13455
2886 Description: Changing primary group doesn't change the group lustre assigns to
2888 Details : When CRAY_XT3 is defined, the fsgid supplied by the client is
2889 overridden with the primary group provided by the group upcall,
2890 whereas the supplied fsgid can be trusted if it is in the list of
2891 supplementary groups returned by the group upcall.
2893 Severity : enhancement
2895 Description: Root Squash Functionality
2896 Details : Implementation of NFS-like root squash capability. Specifically,
2897 don't allow someone with root access on a client node to be able
2898 to manipulate files owned by root on a server node.
2900 Severity : enhancement
2902 Description: Slow trucate/writes to huge files at high offsets.
2903 Details : Directly associate cached pages to lock that protect those pages,
2904 this allows us to quickly find what pages to write and remove
2905 once lock callback is received.
2910 Description: Too many locks accumulating on client during NFS usage
2911 Details : mds_open improperly used accmode to find out access mode to a
2912 file. Also mdc_intent_lock logic to find out if we already have
2913 lock similar to just received was flawed since introduction of
2914 skiplists - locks are now added to the front of the granted
2919 Description: Hit ASSERTION(obd->obd_stopping == 1) failed in some setup failed
2921 Details : In obd setup failure handler, obd_stopping will not necessarily to
2922 be 1, and obd_set_up should also be checked to make sure whether
2923 obd is completely setup.
2925 Severity : enhancement
2927 Description: Allow masking D_WARNING, D_ERROR messages from console
2928 Details : Console messages can now be disabled via lnet.printk.
2933 Description: User code with malformed file open parameter crashes client node
2934 Details : Before packing join_file req, all the related reference should be
2935 checked carefully in case some malformed flags cause fake join_file
2941 Description: shrink/enlarge qunit size when needed; fix the problem of coarse
2942 grain of quota doing harm to quota's accuracy
2943 Details : qunit size will be changed when quota limitation is too low/high;
2944 record the pending quota write in order to get more accureate
2945 quota; delete the patch for bug12588, which is unnecessary when
2946 this patch is landed. This bug also contains fixes for bug 14526,
2947 14299, 14601 and 13794.
2951 Description: LDLM_ENQUEUE races with LDLM_CP_CALLBACK
2952 Details : ldlm_completion_ast() assumes that a lock is granted when the req
2953 mode is equal to the granted mode. However, it should also check
2954 that LDLM_FL_CP_REQD is not set.
2958 Description: Heavy nfs access might result in deadlocks
2959 Details : After ELC code landed, it is now improper to enqueue any mds
2960 locks under och_sem, because enqueue might want to decide to
2961 cancel open locks for same inode we are holding och_sem for.
2965 Description: 35% write performance drop with ldiskfs2 when quotas are on
2966 Details : Enable ext3 journalled quota by default to improve performance
2967 when quotas are turned on.
2971 Description: Client eviction while running blogbench
2972 Details : A lot of unlink operations with concurrent I/O can lead to a
2973 deadlock causing evictions. To address the problem, the number of
2974 oustanding OST_DESTROY requests is now throttled to
2975 max_rpcs_in_flight per OSC and LDLM_FL_DISCARD_DATA blocking
2976 callbacks are processed in priority.
2979 Frequency : RHEL4 only
2981 Description: mkfs is very slow on IA64/RHEL4
2982 Details : A performance regression has been discovered in the MPT Fusion
2983 driver between versions 3.02.73rh and 3.02.99.00rh. As a
2984 consequence, we have downgraded the MPT Fusion driver in the RHEL4
2985 kernel from 3.02.99.00 to 3.02.73 until this problem is fixed.
2988 Frequency : PPC/PPC64 only
2990 Description: conflicts between asm-ppc64/types.h and lustre_types.h
2991 Details : fix duplicated definitions between asm-ppc64/types.h and
2992 lustre_types.h on PPC.
2995 Frequency : PPC/PPC64 only
2997 Description: asm-ppc/segment.h does not exist
2998 Details : fix compile issue on PPC.
3002 Description: data checksumming impacts single node performance
3003 Details : add support for several checksum algorithms. Currently, CRC32 and
3004 Adler-32 are supported. The checksum type can be changed on the fly
3005 through /proc/fs/lustre/osc/*/checksum_type.
3009 Description: use adler32 for page checksums
3010 Details : when available, use the Adler-32 algorithm instead of CRC32 for
3015 Description: better handle error messages in extents code
3017 Severity : enhancement
3019 Description: SNMP support enhancement
3020 Details : Adding total number of sampled request for an MDS node in snmp
3023 Severity : enhancement
3025 Description: Optimize ldlm waiting list processing for PR extent locks
3026 Details : When processing waiting list for read extent lock and meeting read
3027 lock that is same or wider to it that is not contended, skip
3028 processing rest of the list and immediatelly return current
3029 status of conflictness, since we are guaranteed there are no
3030 conflicting locks in the rest of the list.
3034 Description: Time out and refuse to reconnect
3035 Details : When the failover node is the primary node, it is possible
3036 to have two identical connections in imp_conn_list. We must
3037 compare not conn's pointers but NIDs, otherwise we
3038 can defeat connection throttling.
3042 Description: Client not clear own cache if answer to reconnect is lost.
3043 Details : Client gets evicted from server. Now client also thinks it is
3044 disconnected (or gets ENOTCONN on its operation) and decides to
3045 reconnect. Server receives reconnect message, but cannot find
3046 export. New export is created that is fully valid (new cookie!),
3047 but reply is lost and not reported to client. Client reconnects
3048 again and gets back a just-created connection, but it is not new
3049 so client thinks it was not evicted and does not do recovery.
3053 Description: Detect stride IO mode in read-ahead
3054 Details : When a client does stride read, read-ahead should detect that and
3055 read-ahead pages according to the detected stride pattern.
3059 Description: build for x2 fails
3060 Details : fix compile issue on Cray systems.
3062 Severity : enhancement
3064 Description: implement readv/writev APIs(aio_read/aio_writes in newer kernels)
3065 Details : This greatly improves speed of NFS writes on 2.6 kernels.
3068 Frequency : only on PPC/SLES10
3070 Description: "BITS_PER_LONG is not 32 or 64" in linux/idr.h
3071 Details : On SLES10/PPC, fs.h includes idr.h which requires BITS_PER_LONG to
3072 be defined. Add a hack in mkfs_lustre.c to work around this compile
3077 Description: LASSERT on MDS when client holding flock lock dies
3078 Details : ldlm pool logic depends on number of granted locks equal to
3079 number of released locks which is not true for flock locks, so
3080 just exclude such locks from consideration.
3084 Description: MDS deadlock with many ll_sync_lov threads and I/O stalled
3085 Details : Use fsfilt_sync() for both the whole filesystem sync and
3086 individual file sync to eliminate dangerous inode locking
3087 with I_LOCK that can lead to a deadlock.
3091 Description: Update an obsolete wirecheck.c generator
3092 Details : Update wirecheck.c/wirehdr.c and regenerate wiretest.c
3096 Description: Client can panic on open sometimes
3097 Details : It is possible that we try to free already freed request in
3098 ll_file_open in some error cases when we send request from
3103 Description: performance in 1.6.3
3104 Details : Force q->max_phys_segments to MAX_PHYS_SEGMENTS on SLES10 to be
3105 sure that 1MB requests are not fragmented by the block layer.
3109 Description: LDLM soft lockups - improvement
3110 Details : It is be possible to send the lock handle along with each read
3111 or write request because the client is already doing a lock match
3112 itself so there isn't any reason the OST should have to re-do that
3118 Description: lfs quota fails with deactivated OSTS
3119 Details : With this patch, three improvements are included:
3120 1. detete the softlimit in mds and osts when use "lfs quota".
3121 2. display the inaccurate data in the output of "lfs quota".
3122 3. try to get quota info when "lfs quota" is executed.
3127 Description: Extent locks not granted with no conflicts sometimes.
3128 Details : When race occurs in glimpse handler and nothing is returned,
3129 we do not reprocess the queue after lock cancel, and that leads
3130 to a stall until next activity on a resource
3133 Frequency : failover with quotaon
3135 Description: during mds failovers with quota on, OSTs got into deadlock state
3136 and causing dumpstack.
3137 Details : for every quota slave, at any time, there is only one quota req
3138 is sent to quota master for every uid/gid. Before that quota req
3139 returns, all the thread relative to the same uid/gid will wait.
3140 So if the quota req is lost because mds failovers or any other
3141 reasons, this bug will be hit. Now, dqacq_interpret() will handle
3142 quota reqs who time out.
3144 Severity : enhancement
3147 Description: when quota slave checks if quota is enough, there is an unnecessary
3149 Details : place this wait on necessary place instead of always waiting.
3151 --------------------------------------------------------------------------------
3153 2007-12-07 Cluster File Systems, Inc. <info@clusterfs.com>
3155 * Support for kernels:
3156 2.6.5-7.286 (SLES 9),
3157 2.6.9-55.0.9.EL (RHEL 4),
3158 2.6.16.53-0.8 (SLES 10),
3159 2.6.18-8.1.14.el5 (RHEL 5),
3160 2.6.18.8 vanilla (kernel.org)
3161 * Client support for unpatched kernels:
3162 (see http://wiki.lustre.org/index.php?title=Patchless_Client)
3163 2.6.16 - 2.6.22 vanilla (kernel.org)
3164 * Due to recently discovered recovery problems, we do not recommend
3165 using patchless RHEL 4 clients with this or any earlier release.
3166 * Recommended e2fsprogs version: 1.40.2-cfs1
3167 * Note that reiserfs quotas are disabled on SLES 10 in this kernel.
3170 Frequency : occasional
3172 Description: MDS hang or stay in waiting lock
3173 Details : If client receive lock with CBPENDING flag ldlm need send lock
3174 cancel as separate rpc, to avoid situation when cancel request
3175 can't processed due all i/o threads stay in wait lock.
3178 Frequency : occasional
3180 Description: Do not fail import if osc_interpret_create gets -EAGAIN
3181 Details : If osc_interpret_create got -EAGAIN it immediately exits and
3182 wakeup oscc_waitq. After wakeup oscc_wait_for_objects call
3183 oscc_has_objects and see OSC has no objests and call
3184 oscc_internal_create to resend create request.
3186 Severity : enhancement
3188 Description: Update kernel patches for SLES10 2.6.16.53-0.8.
3189 Details : Update which_patch & target file for SLES10 latest kernel.
3191 Severity : enhancement
3193 Description: add --type and --size parameters to lfs find
3194 Details : Enhance lfs find by adding filetype and filesize parameters. Also
3195 multiple OBDs can now be specified for the --obd option.
3197 Severity : enhancement
3199 Description: eliminate client locks in face of contention
3200 Details : file contention detection and lockless i/o implementation
3201 for contended files.
3203 Severity : enhancement
3205 Description: Remove client patches from SLES 10 kernel.
3206 Details : This causes SLES 10 clients to behave as patchless clients
3207 even on a Lustre-patched (server) kernel.
3209 Severity : enhancement
3211 Description: use i_size_read and i_size_write in 2.6 port
3212 Details : replace inode->i_size access with i_size_read/write()
3215 Frequency : when removing large files
3217 Description: scheduling issue during removal of large Lustre files
3218 Details : Don't take the BKL in fsfilt_ext3_setattr() for 2.6 kernels.
3219 It causes scheduling issues when removing large files (17TB in the
3225 Description: 1.4.11 Can't handle directories with stripe set and extended ACLs
3226 Details : Impossible (EPROTO is returned) to access a directory that has a
3227 non-default striping and ACLs.
3230 Frequency : only on ppc
3232 Description: /proc/fs/lustre/devices broken on ppc
3233 Details : The patch as applied to 1.6.2 doesn't look correct for all arches.
3234 We should make sure the type of 'index' is loff_t and then cast
3235 explicitly as needed below. Do not assign an explicitly cast
3239 Frequency : only for rhel5
3241 Description: Kernel patches update for RHEL5 2.6.18-8.1.10.el5.
3242 Details : Modify the target file & which_kernel.
3245 Frequency : if the uninit_groups feature is enabled on ldiskfs
3247 Description: e2fsck reports "invalid unused inodes count"
3248 Details : If a new ldiskfs filesystem is created with the "uninit_groups"
3249 feature and only a single inode is created in a group then the
3250 "bg_unused_inodes" count is incorrectly updated. Creating a
3251 second inode in that group would update it correctly.
3256 Description: buffer overruns could theoretically occur
3257 Details : llapi_semantic_traverse() modifies the "path" argument by
3258 appending values to the end of the origin string, and a buffer
3259 overrun may occur. Adding buffer overrun check in liblustreapi.
3261 Severity : enhancement
3263 Description: Add jbd statistics patch for RHEL5 and 2.6.18-vanilla.
3266 Frequency : only if filesystem is inconsistent
3268 Description: handle "serious error: objid * already exists" more gracefully
3269 Details : If LAST_ID value on disk is smaller than the objects existing in
3270 the O/0/d* directories, it indicates disk corruption and causes an
3271 LBUG(). If the object is 0-length, then we should use the existing
3272 object. This will help to avoid a full fsck in most cases.
3274 Severity : enhancement
3276 Description: Kernel patches update for RHEL4 2.6.9-55.0.6.
3277 Details : Modify vm-tunables-rhel4.patch.
3279 Severity : enhancement
3281 Description: Kernel config for 2.6.18-vanilla.
3282 Details : Modify targets/2.6-vanilla.target.in.
3283 Add config file kernel-2.6.18-2.6-vanilla-i686.config.
3284 Add config file kernel-2.6.18-2.6-vanilla-i686-smp.config.
3285 Add config file kernel-2.6.18-2.6-vanilla-x86_64.config.
3286 Add config file kernel-2.6.18-2.6-vanilla-x86_64-smp.config.
3289 Frequency : occasional
3291 Description: improve handling recoverable errors
3292 Details : If request processed with error which can be recoverable on server
3293 request should be resend, otherwise page released from cache and
3299 Description: Kernel patches update for RHEL5 2.6.18-8.1.14.el5.
3300 Details : Modify target file & which_patch.
3301 A flaw was found in the IA32 system call emulation provided
3302 on AMD64 and Intel 64 platforms. An improperly validated 64-bit
3303 value could be stored in the %RAX register, which could trigger an
3304 out-of-bounds system call table access. An untrusted local user
3305 could exploit this flaw to run code in the kernel
3306 (ie a root privilege escalation). (CVE-2007-4573).
3310 Description: change order of libsysio includes
3311 Details : '#include sysio.h' should always come before '#include xtio.h'
3313 Severity : enhancement
3315 Description: adapt the lustre_config script to support the upgrade case
3316 Details : Add "-u" option for lustre_config script to support upgrading 1.4
3317 server targets to 1.6 in parallel.
3322 Description: To avoid grant space > avaible space when the disk is almost
3323 full. Without this patch you might see the error "grant XXXX >
3324 available" or some LBUG about grant, when the disk is almost
3326 Details : In filter_check_grant, for non_grant cache write, we should
3327 check the left space by if (*left > ungranted + bytes), instead
3328 of (*left > ungranted), because only we are sure the left space
3329 is enough for another "bytes", then the ungrant space should be
3330 increase. In client, we should update cl_avail_grant only there is
3331 OBD_MD_FLGRANT in the reply.
3336 Description: Update RHEL 4 kernel to fix local root privilege escalation.
3337 Details : Update to the latest RHEL 4 kernel to fix the vulnerability
3338 described in CVE-2007-4573. This problem could allow untrusted
3339 local users to gain root access.
3342 Frequency : when using O_DIRECT and quotas
3344 Description: Incorrect file ownership on O_DIRECT output files
3345 Details : block usage reported by 'lfs quota' does not take into account
3346 files that have been written with O_DIRECT.
3350 Description: (rw.c:1323:ll_read_ahead_pages()) ASSERTION(page_idx > ria->ria_stoff) failed
3351 Details : Once the unmatched stride IO mode is detected, shrink the stride-ahead
3352 window to 0. If it does hit cache miss, and read-pattern is still
3353 stride-io mode, does not reset the stride window, but also does not
3354 increase the stride window length in this case.
3356 --------------------------------------------------------------------------------
3358 2007-09-27 Cluster File Systems, Inc. <info@clusterfs.com>
3360 * Support for kernels:
3361 2.6.5-7.286 (SLES 9),
3362 2.6.9-55.0.2.EL (RHEL 4),
3363 2.6.16.46-0.14 (SLES 10),
3364 2.6.18-8.1.8.el5 (RHEL 5),
3365 2.6.18.8 vanilla (kernel.org)
3366 * Client support for unpatched kernels:
3367 (see http://wiki.lustre.org/index.php?title=Patchless_Client)
3368 2.6.16 - 2.6.21 vanilla (kernel.org)
3369 * Due to recently discovered recovery problems, we do not recommend
3370 using patchless RHEL 4 clients with this or any earlier release.
3371 * Recommended e2fsprogs version: 1.40.2-cfs1
3372 * Note that reiserfs quotas are disabled on SLES 10 in this kernel.
3376 Description: Fix errors in lfs documentation
3377 Details : Fixes man pages
3379 Severity : enhancement
3381 Description: Adaptive timeouts
3382 Details : RPC timeouts adapt to changing server load and network
3383 conditions to reduce resend attempts and improve recovery time.
3385 Severity : enhancement
3387 Description: llapi_file_create() does not allow some changes
3388 Details : add llapi_file_open() that allows specifying the file creation
3389 mode and open flags, and also returns an open file handle.
3391 Severity : enhancement
3393 Description: df doesn't work properly if diskfs blocksize != 4K
3394 Details : Choose biggest blocksize of OST's as the LOV's blocksize.
3397 Frequency : always on directories with default striping set
3399 Description: lfs find on -1 stripe looping in lsm_lmm_verify_common()
3400 Details : Avoid lov_verify_lmm_common() on directory with -1 stripe count.
3402 Severity : enhancement
3404 Description: merge and cleanup kernel patches.
3405 Details : Remove mnt_lustre_list in vfs_intent-2.6-rhel4.patch.
3407 Severity : enhancement
3409 Description: RedHat Update kernel for RHEL5
3410 Details : Kernel config file for RHEL5.
3412 Severity : enhancement
3414 Description: OSS needs mutliple precreate threads
3415 Details : Add ability to start more than one create thread per OSS.
3418 Frequency : only with quota on the root user
3420 Description: mds_obd_create error creating tmp object
3421 Details : When the user sets quota on root, llog will be affected and can't
3422 create files and write files.
3425 Frequency : Always on ia64 patchless client, and possibly others.
3427 Description: Add EXPORT_SYMBOL check for node_to_cpumask symbol.
3428 Details : This allows the patchless client to be loaded on architectures
3429 without this export.
3434 Description: disorder of journal start and llog_add cause deadlock.
3435 Details : in llog_origin_connect, journal start should happen before
3436 llog_add keep the same order as other functions to avoid
3439 Severity : enhancement
3441 Description: RedHat Update kernel for RHEL5
3442 Details : Modify the kernel config file more closer RHEL5.
3444 Severity : enhancement
3446 Description: Build failure against Centos5 (RHEL5)
3447 Details : Define PAGE_SIZE when it isn't present.
3450 Frequency : occasionally when using NFS
3452 Description: "ll_intent_file_open()) lock enqueue: err: -13" with nfs
3453 Details : with NFS, the anon dentry's parent was set to itself in
3454 d_alloc_anon(), so in MDS, we use rec->ur_fid1 to find the
3455 corresponding dentry other than use rec->ur_name.
3457 Severity : enhancement
3459 Description: enable data checksumming by default
3460 Details : enable checksum by default, allow --disable-checksum
3461 configure option and "-o nochecksum" mount option. Checksums
3462 can also be disabled at runtime via $LPROC/osc/*/checksum_pages.
3465 Frequency : Occasionally with failover
3467 Description: Client eviction due to failover config
3468 Details : after a connection loss, the lustre client should attempt to
3469 reconnect to the last active server first before trying the
3470 other potential connections.
3472 Severity : enhancement
3474 Description: client-side metadata stat-ahead during readdir(directory readahead)
3475 Details : perform client-side metadata stat-ahead when the client detects
3476 readdir and sequential stat of dir entries therein
3481 Description: when mds and osts use different quota unit(32bit and 64bit),
3482 quota will be released repeatly.
3483 Details : void sending multiple quota reqs to mds, which will keep the
3484 status between the reqs.
3487 Frequency : only with liblustre clients on XT3
3489 Description: evictions taking too long
3490 Details : allow llrd to evict clients directly on OSTs
3494 Description: osts not allocated evenly to files
3495 Details : change the condition to increase offset_idx
3497 Severity : enhancement
3499 Description: self-adjustable client's lru lists
3500 Details : use adaptive algorithm for managing client cached locks lru
3501 lists according to current server load, other client's work
3502 pattern, memory activities, etc. Both, server and client
3503 side namespaces provide number of proc tunables for controlling
3508 Description: rewrite ext2-derived code in llite/dir.c and obdclass/uuid.c
3509 Details : rewrite inherited code (uuid parsing code from ext2 utils and
3510 readdir code from ext3) from scratch preserving functionality.
3514 Description: Only those disconnect error should be returned by rq_status.
3515 Details : In open/enqueue processs, Some errors, which will cause client
3516 disconnected, should be returned by rq_status, while other
3517 errors should still be returned by intent, then mdc or llite will
3520 Severity : enhancement
3522 Description: Tune the kernel for good SCSI performance.
3523 Details : Set the value of /sys/block/{dev}/queue/max_sectors_kb
3524 to the value of /sys/block/{dev}/queue/max_hw_sectors_kb
3528 Frequency : Always for filesystems larger than 2TB on 32-bit systems.
3529 Bugzilla : 13547, 13627
3530 Description: Data corruption for OSTs that are formatted larger than 2TB
3532 Details : When generating the bio request for lustre file writes the
3533 sector number would overflow a temporary variable before being
3534 used for the IO. The data reads correctly from Lustre (which
3535 will overflow in a similar manner) but other file data or
3536 filesystem metadata may be corrupted in some cases.
3540 Description: "lfs find -obd UUID" prints directories
3541 Details : "lfs find -obd UUID" will return all directory names instead
3542 of just file names. It is incorrect because the directories
3543 do not reside on the OSTs.
3546 Bugzilla : 13596, 13614
3547 Description: MDS hang after unclean shutdown of lots of clients
3548 Details : Never resend AST requests.
3552 Frequency : Always, for kernels after 2.6.16
3553 Description: Fix warning idr_remove called for id=.. which is not allocated.
3554 Details : Last kernels save old s_dev before kill super and not allow
3555 to restore from callback - restore it before call kill_anon_super.
3559 Description: Fix error on 'ls .' at the top of the Lustre mount.
3560 Details : Don't revalidate dentry if it is a root dentry.
3562 --------------------------------------------------------------------------------
3564 2007-08-27 Cluster File Systems, Inc. <info@clusterfs.com>
3566 * Support for kernels:
3567 2.6.5-7.286 (SLES 9),
3568 2.6.9-55.0.2.EL (RHEL 4),
3569 2.6.16.46-0.14 (SLES 10),
3570 2.6.18-8.1.8.el5 (RHEL 5),
3571 2.6.18.8 vanilla (kernel.org)
3572 * Client support for unpatched kernels:
3573 (see http://wiki.lustre.org/index.php?title=Patchless_Client)
3574 2.6.16 - 2.6.22 vanilla (kernel.org)
3575 * Due to recently discovered recovery problems, we do not recommend
3576 using patchless RHEL 4 clients with this or any earlier release.
3577 * Recommended e2fsprogs version: 1.39.cfs8
3578 * Note that reiserfs quotas are disabled on SLES 10 in this kernel.
3583 Description: block reactivating mgc import until all deactivates complete
3584 Details : Fix race when failing back MDT/MGS to itself (testing)
3586 Severity : enhancement
3588 Description: lfs setstripe enhancement
3589 Details : Make lfs setstripe understand 'k', 'm' and 'g' for stripe size.
3592 Frequency : mds/oss recovery
3594 Description: llog ctxt is refrenced after it has been freed.
3595 Details : llog ctxt refcount was added to avoide the race between ctxt free
3596 and llog recovery process. Each llog user must hold ctxt refcount
3597 before it access the llog. And the llog ctxt can only be freed
3598 when its refcount is zero.
3600 Severity : enhancement
3602 Description: randomly memory allocation failure util
3603 Details : Make lustre randomly failed allocating memory for testing purpose.
3606 Frequency : only for SLES10
3608 Description: Update kernel patch for SLES10 SP1
3609 Details : Add patch blkdev_tunables-2.6-sles10.patch to 2.6-sles10.series.
3611 Severity : enhancement
3613 Description: omit set fsid for export NFS
3614 Details : fix set/restore device id for avoid EMFILE error and mark lustre fs
3615 as FS_REQUIRES_DEV for avoid problems with generate fsid.
3618 Frequency : after network failures
3620 Description: Add sync option to mount_lustre.c
3621 Details : Client loses data written to lustre after a network interruption.
3623 Severity : enhancement
3625 Description: Error message improvement.
3626 Details : Merging of two LCONSOLE_ERROR_MSG into one.
3628 Severity : enhancement
3630 Description: don't use GFP_* in generic Lustre code.
3631 Details : Use cfs_alloc_* functions and CFS_* flags for code portability.
3633 Severity : enhancement
3635 Description: obdclass is limited by single OBD_ALLOC(idarray)
3636 Details : replace OBD_ALLOC/OBD_FREE with OBD_VMALLOC/OBD_VFREE
3638 Severity : enhancement
3640 Description: updated patchess for new RHEL4 kernel
3641 Details : Fixed ext3-unlink-race.patch per Kalpak's comment.
3643 Severity : enhancement
3645 Description: warnings with build patchless client with vanila 2.6.19 and up
3646 Details : change old ctl_table style and replace ctl_table/ctl_table_header
3647 with cfs_sysctl_table_t/cfs_sysctl_table_header_t
3649 Severity : enhancement
3651 Description: O_DIRECT bypasses client statistics.
3652 Details : When running with O_DIRECT I/O, neither the client rpc_stats nor
3653 read_ahead_stats were updated. Copied stats section from
3654 osc_send_oap_rpc() into async_internal().
3657 Frequency : only for Cray XT3
3659 Description: peer credits not enough on many OST per OSS systems.
3660 Details : Use new lnet way to add credits as we need those for pings and ASTs
3663 Frequency : only with liblustre
3665 Description: Liblustre is not releasing flock locks on file close.
3666 Details : Release flock locks on file close.
3669 Frequency : only for RHEL4
3671 Description: Update kernel patches for kernel-2.6.9-55.0.2.EL
3672 Details : Remove inode-nr_unused-2.6.9-rhel4.patch from 2.6-rhel4.series
3673 Update target file and kernel config.
3677 Description: lustre support for RHEL5
3678 Details : Add support for RHEL5.
3683 Description: ASSERTION(export != NULL) failed in target_handle_connect
3684 Details : Assetion hit is result of rare race between disconnect and connet
3685 to same nid. target_handle_connect found old connect cockie and
3686 tried to reconnect, but can't find export for this cockie.
3691 Description: umount blocks forever on error
3692 Details : In result of wrong using obd_no_recov and obd_force flags client
3693 can hand if cancel or some other requests is lost.
3696 Frequency : Only for SLES
3698 Description: sanity_quota fail test_1
3699 Details : There are multiple occurences of $TSTUSR in SLES's /etc/group
3700 file, which makes TSTID[2] inunique.
3702 Severity : enhancement
3704 Description: Kernel patches for SLES9 2.6.5-7.286 kernel
3705 Details : Update target/ChangeLog/which_patch .
3707 Severity : enhancement
3709 Description: jbd statistics
3710 Details : Port older jbd statistics patch for sles10
3715 Description: lvbo_init failed for resource with missing objects.
3716 Details : Fix returning error if we do stat for file with missing/corrupted
3717 objects and i_size set to all sum of size of all avaible objects.
3718 if we to truncate/write to missing object - it is recreated.
3723 Description: Oops in read and write path when failing to allocate lock.
3724 Details : Check if lock allocation failed and return error back.
3727 Frequency : When flocks are used.
3729 Description: assertion failure in ldlm_cli_enquque_fini for non NULL lock.
3730 Details : Flock locks might destroy just granted lock if it could be merged
3731 with another existing flock, this is done in completion handler,
3732 so teach ldlm_cli_enquque_fini that this is a valid case for
3738 Description: reply_lock_interpret crash due to race with it and lock cancel.
3739 Details : Do not replay locks that are being cancelled. Do not reference
3740 locks by their address during replay, just by their handle.
3743 Frequency : only with deactivated OSTs
3745 Description: lstripe command fails for valid OST index
3746 Details : The stripe offset is compared to 'lov->desc.ld_tgt_count'
3747 instead of lov->desc.ld_active_tgt_count.
3749 Severity : enhancement
3751 Description: Build failure against Centos5 (RHEL5)
3752 Details : Use getpagesize() instead of PAGE_SIZE.
3754 --------------------------------------------------------------------------------
3756 2007-07-30 Cluster File Systems, Inc. <info@clusterfs.com>
3758 * Support for kernels:
3759 2.6.5-7.283 (SLES 9),
3760 2.6.9-55.EL (RHEL 4),
3761 2.6.16.46-0.14 (SLES 10),
3762 2.6.18.8 vanilla (kernel.org)
3763 * Client support for unpatched kernels:
3764 (see http://wiki.lustre.org/index.php?title=Patchless_Client)
3765 2.6.16 - 2.6.22 vanilla (kernel.org)
3766 * Due to recently discovered recovery problems, we do not recommend
3767 using patchless RHEL 4 clients with this or any earlier release.
3768 * Recommended e2fsprogs version: 1.39.cfs8
3769 * Note that reiserfs quotas are disabled on SLES 10 in this kernel.
3770 * Starting with this release, the ldiskfs backing filesystem required
3771 by Lustre is now in its own package, lustre-ldiskfs. This package
3772 should be installed. It is versioned separately from Lustre and
3773 may be released separately in future.
3778 Description: server LBUG when shutting down
3779 Details : Block umount forever until the mount refcount is zero rather
3780 than giving up after an arbitrary timeout.
3782 Severity : enhancement
3784 Description: add optional extra BUILD_VERSION info
3785 Details : add a new environment variable (namely LUSTRE_VERS) which allows
3786 to override the lustre version.
3789 Frequency : 2.6.18 servers only
3791 Description: ll_kern_mount() doesn't release the module reference
3792 Details : The ldiskfs module reference count never drops down to 0
3793 because ll_kern_mount() doesn't release the module reference.
3798 Description: server LBUG when using old ost_num_threads parameter
3799 Details : Accept the old ost_num_threads parameter but warn that it
3800 is deprecated, and fix an off-by-one error that caused an LBUG.
3805 Description: Transient SCSI error results in persistent IO issue
3806 Details : iobuf->dr_error is not reinitialized to 0 between two
3810 Frequency : sometimes when underlying device returns I/O errors
3812 Description: OSTs not going read-only during write failures
3813 Details : OSTs are not remounted read-only when the journal commit threads
3814 get I/O errors because fsfilt_ext3 calls journal_start/stop()
3815 instead of the ext3 wrappers.
3818 Frequency : only with 10000 clients or more
3820 Description: poor connect scaling with increasing client count
3821 Details : Don't run filter_grant_sanity_check for more than 100 exports
3822 to improve scaling for large numbers of clients.
3825 Frequency : SLES10 only
3827 Description: sanity-quota.sh quotacheck failed: rc = -22
3828 Details : Quotas cannot be enabled on SLES10.
3831 Frequency : liblustre clients only
3833 Description: getdirentries does not give error when run on compute nodes
3834 Details : getdirentries does not fail when the size specified as an argument
3835 is too small to contain at least one entry
3837 Severity : enhancement
3839 Description: Add LNET router traceability for debug purposes
3840 Details : If a checksum failure occurs with a router as part of the
3841 IO path, the NID of the last router that forwarded the bulk data
3842 is printed so it can be identified.
3847 Description: OST "spontaneously" evicts client; client has imp_pingable == 0
3848 Details : Due to a race condition, liblustre clients were occasionally
3849 evicted incorrectly.
3851 Severity : enhancement
3853 Description: lfs setstripe use optional parameters instead of postional
3856 Severity : enhancement
3858 Description: Nanosecond timestamp support for ldiskfs
3859 Details : The on-disk ldiskfs filesystem has added support for nanosecond
3860 resolution timestamps. There is not yet support for this at
3861 the Lustre filesystem level.
3864 Frequency : during server recovery
3866 Description: MDS failing to send precreate requests due to OSCC_FLAG_RECOVERING
3867 Details : request with rq_no_resend flag not awake l_wait_event if they get a
3871 Frequency : nfs export on patchless client
3873 Description: connectathon hang when test nfs export over patchless client
3874 Details : Disconnected dentry cannot be found with lookup, so we do not need
3875 to unhash it or make it invalid
3878 Description: fix llapi_lov_get_uuids() to allow many OSTs to be returned
3879 Details: : Change llapi_lov_get_uuids() to read the UUIDs from /proc instead
3880 of using an ioctl. This allows lfsck for > 160 OSTs to succeed.
3885 Description: open req refcounting wrong on reconnect
3886 Details : If reconnect happened between getting open reply from server and
3887 call to mdc_set_replay_data in ll_file_open, we will schedule
3888 replay for unreferenced request that we are about to free.
3889 Subsequent close will crash in variety of ways.
3890 Check that request is still eligible for replay in
3891 mdc_set_replay_data().
3896 Description: disable writes to filesystem when reading health_check file
3897 Details : the default for reading the health_check proc file has changed
3898 to NOT do a journal transaction and write to disk, because this
3899 can cause reads of the /proc file to hang and block HA state
3900 checking on a healthy but otherwise heavily loaded system. It
3901 is possible to return to the previous behaviour during configure
3902 with --enable-health-write.
3904 Severity : enhancement
3906 Description: 64-bit inode version
3907 Details: : Add a on-disk 64-bit inode version for ext3 to track changes made
3908 to the inode. This will be required for version-based recovery.
3913 Description: MDS fails to start if a duplicate client export is detected
3914 Details : in some rare cases it was possible for a client to connect to
3915 an MDS multiple times. Upon recovery the MDS would detect this
3916 and fail during startup. Handle this more gracefully.
3918 Severity : enhancement
3920 Description: Add -o localflock option to simulate old noflock behaviour.
3921 Details : This will achieve local-only flock/fcntl locks coherentness.
3926 Description: log_commit_thread vs filter_destroy race leads to crash
3927 Details : Take import reference before releasing llog record semaphore
3932 Description: Wrong request locking in request set processing
3933 Details : ptlrpc_check_set wrongly uses req->rq_lock for proctect add to
3934 imp_delayed_list, in this place should be used imp_lock.
3937 Frequency : when reconnecting
3939 Description: Grant leak when OSC reconnect to OST
3940 Details : When osc reconnect ost, OST(filter) should check whether it
3941 should grant more space to client by comparing fed_grant and
3942 cl_avail_grant, and return the granted space to client instead
3943 of "new granted" space, because client will call osc_init_grant
3944 to update the client grant space info.
3947 Frequency : when client reconnects to OST
3949 Description: Grant leak when OSC does a resend and replays bulk write
3950 Details : When osc reconnect to OST, OST(filter) should clear grant info of
3951 bulk write request, because the grant info while be sync between
3952 OSC and OST when reconnect, and we should ignore the grant info
3953 these of resend/replay write req.
3958 Description: Grant space more than avaiable space sometimes.
3959 Details : When then OST is about to be full, if two bulk writing from
3960 different clients came to OST. Accord the avaliable space of the
3961 OST, the first req should be permitted, and the second one
3962 should be denied by ENOSPC. But if the seconde arrived before
3963 the first one is commited. The OST might wrongly permit second
3964 writing, which will cause grant space > avaiable space.
3967 Frequency : when client is evicted
3969 Description: Grant might be wrongly erased when osc is evicted by OST
3970 Details : when the import is evicted by server, it will fork another
3971 thread ptlrpc_invalidate_import_thread to invalidate the
3972 import, where the grant will be set to 0. While the original
3973 thread will update the grant it got when connecting. So if
3974 the former happened latter, the grant will be wrongly errased
3975 because of this race.
3980 Description: Checking Stale with correct fid
3981 Details : ll_revalidate_it should uses de_inode instead of op_data.fid2
3982 to check whether it is stale, because sometimes, we want the
3983 enqueue happened anyway, and op_data.fid2 will not be initialized.
3985 Severity : enhancement
3987 Description: update patchless client
3988 Details : Add support for patchless client with 2.6.20, 2.6.21 and RHEL 5
3991 Frequency : only with 2.4 kernel
3993 Description: random memory corruption
3994 Details : size of struct ll_inode_info is to big for union inode.u and this
3995 can be cause of random memory corruption.
4000 Description: Memory leak in recovery
4001 Details : Lov_mds_md was not free in an error handler in mds_create_object.
4002 It should also check obd_fail before fsfilt_start, otherwise if
4003 fsfilt_start return -EROFS,(failover mds during mds recovery).
4004 then the req will return with repmsg->transno = 0 and rc = EROFS.
4005 and we met hit the assert LASSERT(req->rq_reqmsg->transno ==
4006 req->rq_repmsg->transno) in ptlrpc_replay_interpret. Fcc should
4007 be freed no matter whether fsfilt_commit success or not.
4010 Frequency : only with large numbers of cores on a single node
4012 Description: Prevents from taking the superblock lock in llap_from_page for
4014 Details : using LL_ORIGIN_REMOVEPAGE origin flag instead of LL_ORIGIN_UNKNOW
4015 for llap_from_page call in ll_removepage() prevents from taking
4016 the superblock lock for a soon died page.
4021 Description: Not check open intent error before release open handle
4022 Details : in some rare cases, the open intent error is not checked before
4023 release open handle, which may cause
4024 ASSERTION(open_req->rq_transno != 0), because it tries to release
4025 the failed open handle.
4030 Description: Set cat log bitmap only after create log success.
4031 Details : in some rare cases, the cat log bitmap is set too early. and it
4032 should be set only after create log success.
4036 Description: Accessing a block bevice can re-enable I/O when Lustre is
4037 tearing down a device.
4038 Details : dev_clear_rdonly(bdev) must be called in kill_bdev() instead of
4043 Description: service threads may hog cpus when there are a lot of requests
4044 Details : Insert cond_resched to give other threads a chance to use some CPU
4049 Description: the cat log was not initialized in recovery
4050 Details : When mds(mgs) do recovery, the tgt_count might be zero, so the
4051 unlink log on mds will not be initialized until mds post
4052 recovery. And also in mds post recovery, the unlink log will
4053 initialization will be done asynchronausly, so there will be race
4054 between add unlink log and unlink log initialization.
4058 Description: brw_stats were being printed incorrectly
4059 Details : brw_stats were being printed as log2 but all of them were not
4060 recorded as log2. Also remove some code duplication arising from
4061 filter_tally_{read,write}.
4065 Frequency : rare, only in recovery.
4066 Description: ASSERTION(req->rq_type != LI_POISON) failed
4067 Details : imp_lock should be held while iterating over imp_sending_list for
4068 prevent destroy request after get timeout in ptlrpc_queue_wait.
4072 Description: replay-single.sh test 52 fails
4073 Details : A lock's skiplist need to be cleanup when it being unlinked
4074 from its resource list.
4079 Description: Short directio read returns full requested size rather than
4081 Details : Direct I/O operations should return actual amount of bytes
4082 transferred rather than requested size.
4084 Severity : enhancement
4086 Description: metadata RPC reduction (e.g. for rm performance)
4087 Details : decrease the amount of synchronous RPC between clients and servers
4088 by canceling conflicing lock before the operation on the client
4089 and packing thier handles into the main operation RPC to server.
4091 Severity : enhancement
4093 Description: add #ifdef HAVE_KERNEL_CONFIG_H
4094 Details : kernels from 2.6.19 not need include linux/config.h, but add
4095 include linux/autoconf.h in commpiler command line.
4097 Severity : enhancement
4099 Description: patchless client support for 2.6.22 kernel
4100 Details : 2.6.22 has only one visble change, SLAB_CTOR_* constants is
4101 removed. In this case we need drop using os depended interface to
4102 kmem_cache and use cfs_mem_cache API.
4107 Description: fix mal-formatted messages
4108 Details : fix some mal-formatted DEBUG_REQ and LCONSOLE_ERROR_MSG messages
4112 Frequency : always in liblustre
4113 Description: wrong IS_ERR implementation in liblustre.h
4114 Details : fix IS_ERR implementation in liblustre.h for right detect errors.
4119 Description: Correct condition for output debug message.
4120 Details : inode i_nlink equal zero is not enough for output message about
4121 disk corruption, i_ctime and i_mode should be also checked.
4125 Frequency : always in patchless client
4126 Description: add configure check for truncate_complete_page
4127 Details : improve checks for exported symbols. This allow run check without
4128 sources, but with Module.symvers shipped with kernel distribution.
4129 add check for truncate_complete_page used by patchless client.
4133 Description: sanity.sh test_77h fails with "test_77h file compare failed"
4134 Details : test_77h uses a file which was messed by other test case.
4138 Description: Not Check whether lov_tgts is NULL in some lov functions
4139 Details : Checking whether lov_tgts is NULL in some functions.
4143 Description: replace obdo_alloc() with OBDO_ALLOC macro
4144 Details : nothing special is done in obdo_alloc() function, and for
4145 debugging purpose, it needs to be replaced with macros.
4149 Description: bad return value and errno from fcntl call
4150 Details : In liblustre API, errno should be a negative value if error
4155 Description: ptlrpc_check_set() LBUG
4156 Details : In case of positive reply from server and failed client bulk
4157 callback after bulk transfer shouldn't LBUG, but process this
4158 request as erroneous.
4160 Severity : enhancement
4162 Description: VFS operations stats tool.
4163 Details : Tool which collects stats by tracking value written in pid,
4164 ppid, gid and uses llstat to generate output to plot graph using
4166 Updated lustre/utils/Makefile.am
4167 Added lustre/utils/ltrack_stats.c
4169 Severity : enhancement
4171 Description: 2.6.18 server support (lustre 1.6.1)
4172 Details : Support for 2.6.18 kernels on the server side.
4177 Description: ASSERTION(imp->imp_conn_current) failed
4178 Details : an assertion failure is hit if a client node boots and attempts to
4179 mount a lustre filesystem faster than RECONNECT_INTERVAL seconds.
4182 Frequency : only for i686
4184 Description: 1.4.11 RC1 build fails for RHEL 4, i686
4185 Details : Fixed config variable for build.
4190 Description: Updated patchess for new RHEL4 kernel
4191 Details : Updated patch inode-nr_unused-2.6.9-rhel4.patch
4192 Updated patch jbd-stats-2.6.9.patch
4193 Updated patch qsnet-rhel4-2.6.patch
4194 Updated patch quota-deadlock-on-pagelock-core.patch
4195 Updated patch vfs_intent-2.6-rhel4.patch
4196 Updated patch vfs_races-2.6-rhel4.patch
4197 Updated series file 2.6-rhel4-titech.series
4198 Updated series file 2.6-rhel4.series
4199 Updated kernel config files
4204 Description: lquota slave complains LBUG when reconnecting with mds
4206 Details : quota slave depends qctxt->lqc_import to send its quota request.
4207 This pointer will be invalid if mds did failover or broke its
4208 connect to osts, which leads to LBUG.
4211 Frequency : when qunit size is too small(less than 20M)
4213 Description: write is stopped by improper -EDQUOT
4214 Details : If the master is busy and qunit size is small enough(let's say 1M),
4215 the slave can not get quota from master on time, which will lead
4216 slave to trigger a -EQUOTA to client.
4221 Description: Deadlock during metadata tests
4222 Details : in prune_dir_dentries(), shrink_dcache_parent() should not be
4223 called with the per-dentry lock held.
4226 Frequency : SLES9 only
4228 Description: Lustre patched kernel for SLES9 SP3 has NR_CPUS set to 8
4229 Details : set CONFIG_NR_CPUS to 128 instead of 8.
4231 Severity : enhancement
4233 Description: remove fs_prep_san_write operation and related patches
4234 Details : remove the ext3-san-jdike patches which are no longer useful.
4239 Description: LDISKFS-fs error (device sdc): ldiskfs_free_blocks
4240 Details : a disk corruption can cause the mballoc code to assert on a
4241 double free or other extent corruptions. Handle these with
4242 ext3_error() instead of assertions.
4245 Frequency : only with mballoc3 code and deep extent trees
4247 Description: ldiskfs_ext_search_right: bad header in inode: unexpected eh_depth
4248 Details : a wrong check of extent headers in ldiskfs_ext_search_right()
4249 can cause the filesystem to be remounted read-only.
4254 Description: lfsck built against 1.4.x cannot run against 1.6.0 lustre
4255 Details : the definition for OBD_IOC_GETNAME changed in 1.6.0. One of the
4256 few external users of this ioctl number is lfsck's call to
4257 llapi_lov_get_uuids() and this caused lfsck to fail at startup.
4258 Add the old ioctl number to the handler so both old and new
4263 Description: parallel lock callbacks
4264 Details : Instead of sending blocking and completion callbacks as separated
4265 requests, adding them to a set and sending in parallel.
4269 Description: Disable most debugging by default
4270 Details : To improve performance, disable most logging (for debug purposes)
4271 by default. VFSTRACE, RPCTRACE, and DLMTRACE are now off by
4272 default, and HA includes fewer messages.
4276 Frequency : only run on patchless client.
4277 Description: use do_facet on sanity.sh for test handling recoverables errors
4278 Details : use do_facet instead of direct use sysctl for set fail_loc on OST
4282 Description: hash tables for lists of nids, connections and uuids
4283 Details : Hash tables noticeably help when a lot of clients connect to a
4284 server, to faster identify duplicate connections or reconnects,
4285 also to faster find export to evict in manual eviction case.
4289 Description: Sometimes, when the server evict a client, and the client will
4290 not be evicted as soon as possible.
4291 Details : In enqueue req, the error was returned by intent, instead of
4292 rq_status which make ptlrpc layer not detect this error, and
4293 does not evict the client. So enqueue error should be returned
4297 Frequency : only at startup
4299 Description: Delay client connections to MDT until fist MDT->OST connect
4300 Details : If a client tried to create a new file before the MDT had
4301 connected to any OSTs, the create would return EIO. Now
4302 the client will simply block until the MDT connects to the
4303 first OST and the create can succeed.
4306 Frequency : only for SLES9
4308 Description: Routinely utilize latest Quadrics drivers in CFS releases
4309 Details : Update patch qsnet-suse-2.6.patch.
4312 Frequency : only for sles10
4314 Description: Update patches for SLES 10 SP1 kernel.
4315 Details : Update the patch vfs_intent-2.6-sles10.patch.
4320 Description: Routinely utilize latest Quadrics drivers in CFS releases
4321 Details : Update patch qsnet-rhel4-2.6.patch.
4324 Frequency : at statup only
4326 Description: mds_lov_synchronize race leads to various problems
4327 Details : simultaneous MDT->OST connections at startup can cause the
4328 sync to abort, leaving the OSC in a bad state.
4333 Description: Using wrong pointer in osc_brw_prep_request
4334 Details : Access to array[-1] can produce panic if kernel compiled with
4335 CONFIG_PAGE_ALLOC enabled
4337 Severity : enhancement
4339 Description: Async OSC create to avoid the blocking unnecessarily.
4340 Details : If a OST has no remain object, system will block on the creating
4341 when need to create a new object on this OST. Now, ways use
4342 pre-created objects when available, instead of blocking on an
4343 empty osc while others are not empty. If we must block, we block
4344 for the shortest possible period of time.
4348 Frequency : only in recovery
4349 Description: Mark OST as early accessible if his start SYNC.
4350 Details : osc_precreate return flag early accessible if oscc marked as
4351 OSCC_FLAG_SYNC_IN_PROGRESS.
4356 Description: Sometimes precreate code can triger create object on wrong ost
4357 Details : Wrong protected or not not restored variables aftre precreate loop
4358 can produce creation object on wrong ost.
4361 Frequency : oss recovery
4363 Description: llog_commit_thread cleanup should sync with llog_commit_thread
4365 Details : llog_commit_thread_count should be synced between llog_commit
4366 start and cleanup, so new llog_commit thread should not be started
4367 when llog_commit threads being stopped to avoid accessing some
4370 Severity : enhancement
4372 Description: Add printing inode info into message about error in writepage.
4374 --------------------------------------------------------------------------------
4376 2007-05-03 Cluster File Systems, Inc. <info@clusterfs.com>
4381 Frequency : on some architectures
4383 Description: 1.6 client sometimes fails to mount from a 1.4 MDT
4384 Details : Uninitialized flags sometimes cause configuration commands to
4388 Frequency : patchless clients only
4390 Description: missing __iget() symbol export
4391 Details : The __iget() symbol export is missing. To avoid the need for
4392 this on patchless clients the deathrow inode reaper is turned
4393 off, and we depend on the VM to clean up old inodes. This
4394 dependency was during via the fix for bug 12181.
4399 Description: sanity.sh fail: test_52b
4400 Details : The ll_inode_to_ext_flags() has a glitch which makes MDS return
4401 incorrect inode's flags to client.
4403 --------------------------------------------------------------------------------
4405 2007-04-19 Cluster File Systems, Inc. <info@clusterfs.com>
4407 * CONFIGURATION CHANGE. This version of Lustre WILL NOT
4408 INTEROPERATE with older versions automatically. In many cases a
4409 special upgrade step is needed. Please read the
4410 user documentation before upgrading any part of a 1.4.x system.
4411 * WARNING: Lustre configuration and startup changes are required with
4412 this release. See https://mail.clusterfs.com/wikis/lustre/MountConf
4414 * Support for kernels:
4415 2.4.21-47.0.1.EL (RHEL 3),
4416 2.6.5-7.283 (SLES 9),
4417 2.6.9-42.0.10.EL (RHEL 4),
4418 2.6.12.6 vanilla (kernel.org),
4419 2.6.16.27-0.9 (SLES10)
4420 * Client support for unpatched kernels:
4421 (see https://mail.clusterfs.com/wikis/lustre/PatchlessClient)
4422 2.6.16 - 2.6.19 vanilla (kernel.org),
4423 2.6.9-42.0.8EL (RHEL 4)
4424 * Recommended e2fsprogs version: 1.39.cfs6
4425 * Note that reiserfs quotas are disabled on SLES 10 in this kernel
4428 Severity : enhancement
4430 Description: Async OSC create to avoid the blocking unnecessarily.
4431 Details : If a OST has no remain object, system will block on the creating
4432 when need to create a new object on this OST. Now, ways use
4433 pre-created objects when available, instead of blocking on an
4434 empty osc while others are not empty. If we must block, we block
4435 for the shortest possible period of time.
4437 Severity : enhancement
4439 Description: MountConf
4440 Details : Lustre configuration is now managed via mkfs and mount
4441 commands instead of lmc and lconf. New obd types (MGS, MGC)
4442 are added for dynamic configuration management. See
4443 https://mail.clusterfs.com/wikis/lustre/MountConf for
4446 Severity : enhancement
4448 Description: dynamic OST addition
4449 Details : OSTs can now be added to a live filesystem
4451 Severity : enhancement
4453 Description: startup order invariance
4454 Details : MDTs and OSTs can be started in any order. Clients only
4455 require the MDT to complete startup.
4457 Severity : enhancement
4459 Description: parallel, asynchronous orphan cleanup
4460 Details : orphan cleanup is now performed in separate threads for each
4461 OST, allowing parallel non-blocking operation.
4463 Severity : enhancement
4465 Description: optimized stripe assignment
4466 Details : stripe assignments are now made based on ost space available,
4467 ost previous usage, and OSS previous usage, in order to try
4468 to optimize storage space and networking resources.
4470 Severity : enhancement
4472 Description: Permanently set tunables
4473 Details : All writable /proc/fs/lustre tunables can now be permanently
4474 set on a per-server basis, at mkfs time or on a live
4477 Severity : enhancement
4479 Description: Lustre message v2
4480 Details : Add lustre message format v2.
4482 Severity : enhancement
4484 Description: client OST exclusion list
4485 Details : Clients can be started with a list of OSTs that should be
4486 declared "inactive" for known non-responsive OSTs.
4490 Description: ENOENT returned for valid filehandle during dbench.
4491 Details : Check if a directory has children when invalidating dentries
4492 associated with an inode during lock cancellation. This fixes
4493 an incorrect ENOENT sometimes seen for valid filehandles during
4494 testing with dbench.
4497 Frequency : SFS test only (otherwise harmless)
4499 Description: SPEC SFS validation failure on NFS v2 over lustre.
4500 Details : Changes the blocksize for regular files to be 2x RPC size,
4501 and not depend on stripe size.
4503 Severity : enhancement
4505 Description: fine-grained SMP locking inside DLM
4506 Details : Improve DLM performance on SMP systems by removing the single
4507 per-namespace lock and replace it with per-resource locks.
4509 Severity : enhancement
4511 Description: don't hold multiple extent locks at one time
4512 Details : To avoid client eviction during large writes, locks are not
4513 held on multiple stripes at one time or for very large writes.
4514 Otherwise, clients can block waiting for a lock on a failed OST
4515 while holding locks on other OSTs and be evicted.
4517 Severity : enhancement
4519 Description: Multiple MD RPCs in flight.
4520 Details : Further unserialise some read-only MDT RPCs - learn about intents.
4521 To avoid overly-overloading MDT, introduce a limit on number of
4522 MDT RPCs in flight for a single client and add /proc controls
4523 to adjust this limit.
4525 Severity : enhancement
4527 Description: client read/write statistics
4528 Details : Add client read/write call usage stats for performance
4529 analysis of user processes.
4530 /proc/fs/lustre/llite/*/offset_stats shows non-sequential
4531 file access. extents_stats shows chunk size distribution.
4532 extents_stats_per_process show chunk size distribution per
4535 Severity : enhancement
4537 Description: per-client statistics on server
4538 Details : Add ldlm and operations statistics for each client in
4539 /proc/fs/lustre/mds|obdfilter/*/exports/
4541 Severity : enhancement
4543 Description: improved MDT statistics
4544 Details : Add detailed MDT operations statistics in
4545 /proc/fs/lustre/mds/*/stats
4547 Severity : enhancement
4549 Description: VFS operations stats
4550 Details : Add client VFS call stats, trackable by pid, ppid, or gid
4551 /proc/fs/lustre/llite/*/stats_track_[pid|ppid|gid]
4556 Description: Fix client-side osc byte counters
4557 Details : The osc read/write byte counters in
4558 /proc/fs/lustre/osc/*/stats are now working
4561 Frequency : always as root on SLES
4563 Description: Failure of copying files with lustre special EAs.
4564 Details : Client side always return success for setxattr call for lustre
4565 special xattr (currently only "trusted.lov").
4570 Description: Refcount LNET uuids
4571 Details : The global LNET uuid list grew linearly with every startup;
4572 refcount repeated list entries instead of always adding to
4575 Severity : enhancement
4577 Description: Dynamic service threads
4578 Details : Within a small range, start extra service threads
4579 automatically when the request queue builds up.
4582 Frequency : mixed-endian client/server environments
4584 Description: mixed-endian crashes
4585 Details : The new msg_v2 system had some failures in mixed-endian
4588 Severity : enhancement
4590 Description: Easy OST removal
4591 Details : OSTs can be permanently deactivated with e.g. 'lctl
4592 conf_param lustre-OST0001.osc.active=0'
4594 Severity : enhancement
4596 Description: MGS proc entries
4597 Details : Added basic proc entries for the MGS showing what filesystems
4600 Severity : enhancement
4602 Description: provide MGS failover
4603 Details : Added config lock reacquisition after MGS server failover.
4605 Severity : enhancement
4607 Description: add Linux 2.4 support
4608 Details : Added support for RHEL 2.4.21 kernel for 1.6 servers and clients
4612 Description: a large application tries to do I/O to the same resource and dies
4613 in the middle of it.
4614 Details : Check the req->rq_arrival time after the call to
4615 ost_brw_lock_get(), but before we do anything about
4616 processing it & sending the BULK transfer request. This
4617 should help move old stale pending locks off the queue as
4618 quickly as obd_timeout.
4621 Frequency : when an incorrect nid is specified during startup
4623 Description: ptlrpc connect to non-existant node causes kernel crash
4624 Details : LNET can't be re-entered from an event callback, which
4625 happened when we expire a message after the export has been
4626 cleaned up. Instead, hand the zombie cleanup off to another
4629 Severity : enhancement
4631 Description: plain/inodebits lock performance improvement
4632 Details : Grouping plain/inodebits in granted list by their request modes
4633 and bits policy, thus improving the performance of search through
4637 Frequency : only if OST filesystem is corrupted
4639 Description: client incorrectly hits assertion in ptlrpc_replay_req()
4640 Details : for a short time RPCs with bulk IO are in the replay list,
4641 but replay of bulk IOs is unimplemented. If the OST filesystem
4642 is corrupted due to disk cache incoherency and then replay is
4643 started it is possible to trip an assertion. Avoid putting
4644 committed RPCs into the replay list at all to avoid this issue.
4647 Frequency : liblustre (e.g. catamount) on a large cluster with >= 8 OSTs/OSS
4649 Description: System hang on startup
4650 Details : This bug allowed the liblustre (e.g. catamount) client to
4651 return to the app before handling all startup RPCs. This
4652 could leave the node unresponsive to lustre network traffic
4653 and manifested as a server ptllnd timeout.
4655 Severity : enhancement
4657 Description: Add "/proc/sys/lustre/debug_peer_on_timeout"
4658 Details : liblustre envirable: LIBLUSTRE_DEBUG_PEER_ON_TIMEOUT
4659 boolean to control whether to print peer debug info when a
4660 client's RPC times out.
4663 Frequency : only for kernels with patches from Lustre below 1.4.3
4665 Description: Remove old rdonly API
4666 Details : Remove old rdonly API which unused from at least lustre 1.4.3
4669 Frequency : only for devices with external journals
4671 Description: Set external device read-only also
4672 Details : During a commanded failover stop, we set the disk device
4673 read-only while the server shuts down. We now also set any
4674 external journal device read-only at the same time.
4677 Frequency : when upgrading from 1.4 while trying to change parameters
4679 Description: The wrong (new) MDC name was used when setting parameters for
4680 upgraded MDT's. Also allows changing of OSC (and MDC)
4681 parameters if --writeconf is specified at tunefs upgrade time.
4684 Frequency : when setting specific ost indicies
4686 Description: QOS code breaks on skipped indicies
4687 Details : Add checks for missing OST indicies in the QOS code, so OSTs
4688 created with --index need not be sequential.
4690 Severity : enhancement
4692 Description: Add uninit_groups feature to ldiskfs2 to speed up e2fsck
4693 Details : The uninit_groups feature works in conjunction with the kernel
4694 filesystem code (ldiskfs2 only) and e2fsprogs-1.39-cfs6 to speed
4695 up the pass1 processing of e2fsck. This is a read-only feature
4696 in ldiskfs2 only, so older kernels and current ldiskfs cannot
4697 mount filesystems that have had this feature enabled.
4699 Severity : enhancement
4701 Description: Improve multi-block allocation algorithm to avoid fragmentation
4702 Details : The mballoc3 code (ldiskfs2 only) adds new mechanisms to improve
4703 allocation locality and avoid filesystem fragmentation.
4705 ------------------------------------------------------------------------------
4707 2007-04-01 Cluster File Systems, Inc. <info@clusterfs.com>
4709 * Support for kernels:
4710 2.4.21-47.0.1.EL (RHEL 3)
4711 2.6.5-7.283 (SLES 9)
4712 2.6.9-42.0.10.EL (RHEL 4)
4713 2.6.12.6 vanilla (kernel.org)
4714 2.6.16.27-0.9 (SLES 10)
4715 * Recommended e2fsprogs version: 1.39.cfs5
4717 * Note that reiserfs quotas are disabled on SLES 10 in this kernel
4721 Frequency : occasional, depends on client load and configuration
4722 Bugzilla : 12181, 12203
4723 Description: data loss for recently-modified files
4725 Details : In some cases it is possible that recently written or created
4726 files may not be written to disk in a timely manner (this should
4727 normally be within 30s unless client IO load is very high).
4728 The problem appears as zero-length files or files that are a
4729 multiple of 1MB in size after a client crash or client eviction
4730 that are missing data at the end of the file.
4732 This problem is more likely to be hit on clients where files are
4733 repeatedly created and unlinked in the same directory, clients
4734 have a large amount of RAM, have many CPUs, the filesystem has
4735 many OSTs, the clients are rebooted frequently, and/or the files
4736 are not accessed by other nodes after being written.
4738 The presence of the problem can be detected by looking at
4739 /proc/sys/fs/inode-state. If the first number (nr_inodes) is
4740 smaller than the second (nr_unused) then dirty files will not
4741 be flushed automatically to disk. "sync; sleep 10" should be
4742 run several times on the node before unmounting it to update
4743 Lustre (this is also safe to run on nodes without this problem).
4745 There is also a related kernel bug in the RHEL4 4 2.6.9 kernel
4746 that can cause this same problem, so customers using that kernel
4747 also need to update the kernel in addition to Lustre. In order
4748 to properly fix this bug, the RHEL3 2.4.21 kernel is also updated.
4750 It is normal that files written just before a client crash (less
4751 than 30s) may not yet have been flushed to disk, even for local
4755 Frequency : frequent on thin XT3 nodes
4757 Description: UUID collision on thin XT3 Linux nodes
4758 Details : UUIDs on Compute Node Linux XT3 nodes were not generated
4759 randomly, since we relied on an insufficiently-seeded PRNG.
4764 Description: OSS hangs after "All ost request buffers busy"
4765 Details : A deadlock between quota and journal operations caused OSS
4766 hangs after printing "All ost request buffers busy."
4769 Frequency : always on liblustre builds
4771 Description: Cleanup compiler warnings on liblustre
4774 Frequency : always on liblustre builds on XT3
4776 Description: LC_CONFIG_CDEBUG don't run while build liblustre on XT3.
4780 Description: Addition of EXT3_FEATURE_RO_COMPAT_DIR_NLINKS flag for
4781 > 32000 subdirectories
4782 Details : Add EXT3_FEATURE_RO_COMPAT_DIR_NLINK flag to
4783 EXT3_FEATURE_RO_COMPAT_SUPP. This flag will be set whenever
4784 subdirectory count crosses 32000. This will aid e2fsck to
4785 correctly handle more than 32000 subdirectories.
4788 Frequency : liblustre (e.g. catamount) on a large cluster with >= 8 OSTs/OSS
4790 Description: System hang on startup
4791 Details : This bug allowed the liblustre (e.g. catamount) client to
4792 return to the app before handling all startup RPCs. This
4793 could leave the node unresponsive to lustre network traffic
4794 and manifested as a server ptllnd timeout.
4796 Severity : enhancement
4798 Description: Add "/proc/sys/lustre/debug_peer_on_timeout"
4799 (liblustre envirable: LIBLUSTRE_DEBUG_PEER_ON_TIMEOUT)
4800 boolean to control whether to print peer debug info when a
4801 client's RPC times out.
4806 Description: make O_SYNC working on 2.6 kernels
4807 Details : 2.6 kernels use different method for mark pages for write,
4808 so need add a code to lustre for O_SYNC work.
4813 Description: Failure to close file and release space on NFS
4814 Details : Put inode details into lock acquired in ll_intent_file_open.
4815 Use mdc_intent_lock in ll_intent_open to properly
4816 detect all kind of errors unhandled by mdc_enqueue.
4821 Description: proc file read during shutdown sometimes raced obd removal,
4823 Details : Add lock to prevent obd access after proc file removal.
4826 Frequency : Only for files larger than 4GB on 32-bit clients.
4828 Description: improperly doing page alignment of locks
4829 Details : Modify lustre core code to use CFS_PAGE_* defines instead of
4830 PAGE_*. Make CFS_PAGE_MASK a 64-bit mask.
4835 Description: RPCs being resent when they shouldn't be
4836 Details : Some RPCs that should not be resent are being resent. This
4837 can cause inconsistencies in the RPC state machine. Do not
4838 resend such requests.
4841 Frequency : rare, only with NFS export
4843 Description: Crash on NFS re-export node
4844 Details : under very unusual load conditions an assertion is hit in
4845 ll_intent_file_open()
4848 Frequency : only if OST filesystem is corrupted
4850 Description: client incorrectly hits assertion in ptlrpc_replay_req()
4851 Details : for a short time RPCs with bulk IO are in the replay list,
4852 but replay of bulk IOs is unimplemented. If the OST filesystem
4853 is corrupted due to disk cache incoherency and then replay is
4854 started it is possible to trip an assertion. Avoid putting
4855 committed RPCs into the replay list at all to avoid this issue.
4860 Description: large O_DIRECT requests fail under memory pressure/fragmentation
4861 Details : Large single O_DIRECT read and write calls can fail to allocate
4862 a sufficiently large buffer to process the request. In case of
4863 allocation failure the allocation is retried with a smaller
4864 buffer and broken into smaller requests.
4866 Severity : enhancement
4868 Description: Add -o localflock option to simulate old noflock behaviour.
4869 Details : This will achieve local-only flock/fcntl locks coherentness.
4874 Description: versioning check is incomplete
4875 Details : Checking the version difference of client vs. server, report
4876 error if the gap is too big.
4881 Description: add support PG_writeback bit
4882 Details : add support for PG_writeback bit for Lustre, for more carefull
4883 work with page cache in 2.6 kernel. This also fix some deadlocks
4884 and remove hack for work O_SYNC with 2.6 kernel.
4886 Severity : enhancement
4888 Description: Add uninit_groups feature to ldiskfs2 to speed up e2fsck
4889 Details : The uninit_groups feature works in conjunction with the kernel
4890 filesystem code (ldiskfs2 only) and e2fsprogs-1.39-cfs6 to speed
4891 up the pass1 processing of e2fsck. This is a read-only feature
4892 in ldiskfs2 only, so older kernels and current ldiskfs cannot
4893 mount filesystems that have had this feature enabled.
4895 Severity : enhancement
4897 Description: Improve multi-block allocation algorithm to avoid fragmentation
4898 Details : The mballoc3 code (ldiskfs2 only) adds new mechanisms to improve
4899 allocation locality and avoid filesystem fragmentation.
4901 ------------------------------------------------------------------------------
4903 2007-02-09 Cluster File Systems, Inc. <info@clusterfs.com>
4905 * Support for kernels:
4906 2.6.9-42.0.3.EL (RHEL 4)
4907 2.6.5-7.276 (SLES 9)
4908 2.4.21-47.0.1.EL (RHEL 3)
4909 2.6.12.6 vanilla (kernel.org)
4910 2.6.16.21-0.8 (SLES10)
4911 * Recommended e2fsprogs version: 1.39.cfs2-0
4913 * The backwards-compatible /proc/sys/portals symlink has been removed
4914 in this release. Before upgrading, please ensure that you change
4915 any configuration scripts or /etc/sysctl.conf files that access
4916 /proc/sys/portals/* or sysctl portals.* to use the corresponding
4917 entry in /proc/sys/lnet or sysctl lnet.*. This change can be made
4918 in advance of the upgrade on any system running Lustre 1.4.6 or
4919 newer, since /proc/sys/lnet was added in that version.
4920 * Note that reiserfs quotas are disabled on SLES 10 in this kernel
4924 Frequency : only when quota is used
4926 Description: avoid scanning export list for quota master
4927 Details : Change the algorithms to avoid scanning export list in order
4928 to improve the efficiency.
4931 Frequency : MDS failover only, very rarely
4933 Description: "went back in time" messages on mds failover
4934 Details : The greatest transno may be lost when the current operation
4935 finishes with an error (transno==0) and the client's last_rcvd
4936 record is over-written. Save the greatest transno in the
4937 mds_last_transno for this case.
4940 Frequency : always for specific kernels and striping counts
4942 Description: client may get "Matching packet too big" without ACL support
4943 Details : Clients compiled without CONFIG_FS_POSIX_ACL get an error message
4944 when trying to access files in certain configurations. The
4945 clients should in fact be denied when mounting because they do
4946 not understand ACLs.
4949 Frequency : Cray XT3 with more than 4000 clients and multiple jobs
4951 Description: many clients connecting with IO in progress causes connect timeouts
4952 Details : Avoid synchronous journal commits to avoid delays caused by many
4953 clients connecting/disconnecting when bulk IO is in progress.
4954 Queue liblustre connect requests on OST_REQUEST_PORTAL instead of
4955 OST_IO_PORTAL to avoid delays behind potentially many pending
4959 Frequency : occasionally with multiple writers to a single file
4961 Description: shared writes to file may result in wrong size reported by stat()
4962 Details : Allow growing of kms when extent lock is cancelled
4965 Frequency : always with random mmap IO to multi-striped file
4967 Description: mmap write might be lost if we are writing to a 'hole' in stripe
4968 Details : Only if the hole is at the end of OST object so that kms is too
4969 small. Fix is to increase kms accordingly in ll_nopage.
4972 Frequency : rare, only if OST filesystem is inconsistent with MDS filesystem
4974 Description: writes to a missing object would leak memory on the OST
4975 Details : If there is an inconsistency between the MDS and OST filesystems,
4976 such that the MDS references an object that doesn't exist, writes
4977 to that object will leak memory due to incorrect cleanup in the
4978 error handling path, eventually running out of memory on the OST.
4983 Description: Creating too long symlink causes lustre errors
4984 Details : Check symlink and name lengths before sending requests to MDS.
4987 Frequency : only if flock is enabled (not on by default)
4989 Description: posix locks not released on fd closure on 2.6.9+
4990 Details : We failed to add posix locks to list of inode locks on 2.6.9+
4991 kernels, this caused such locks not to be released on fd close and
4992 then assertions on fs unmount about still used locks.
4995 Frequency : MDS failover only, very rarely
4997 Description: clients may get ASSERTION(granted_lock != NULL)
4998 Details : When request was taking a long time, and a client was resending
4999 a getattr by name lock request. The were multiple lock requests
5000 with the same client lock handle and
5001 mds_getattr_name->fixup_handle_for_resent_request found one of the
5002 lock handles but later failed with ASSERTION(granted_lock != NULL).
5007 Description: handle->h_buffer_credits > 0, assertion failure
5008 Details : h_buffer_credits is zero after truncate, causing assertion
5009 failure. This patch extends the transaction or creates a new
5013 Frequency : NFS re-export or patchless client
5014 Bugzilla : 11179, 10796
5015 Description: Crash on NFS re-export node (__d_move)
5016 Details : We do not want to hash the dentry if we don't have a lock.
5017 But if this dentry is later used in d_move, we'd hit uninitialised
5018 list head d_hash, so we just do this to init d_hash field but
5019 leave dentry unhashed.
5022 Frequency : NFS re-export or patchless client
5024 Description: NFS exports has problem with symbolic link
5025 Details : lustre client didn't properly install dentry when re-exported
5026 to NFS or running patchless client.
5029 Frequency : NFS re-export or patchless client
5031 Description: Various nfs/patchless fixes.
5032 Details : fixes reuse disconected alias for lookup process - this fixes
5033 warning "find_exported_dentry: npd != pd",
5034 fix permission error with open files at nfs.
5035 fix apply umask when do revalidate.
5038 Frequency : occasional
5040 Description: Crash on NFS re-export node
5041 Details : calling clear_page() on the wrong pointer triggered oops in
5042 generic_mapping_read().
5045 Frequency : rarely, using O_DIRECT IO
5047 Description: unaligned directio crashes client with LASSERT
5048 Details : check for unaligned buffers before trying any requests.
5051 Frequency : rarely, using CFS RAID5 patches in non-standard kernel series
5053 Description: stale data returned from RAID cache
5054 Details : If only a small amount of IO is done to the RAID device before
5055 reading it again it is possible to get stale data from the RAID
5056 cache instead of reading it from disk.
5059 Frequency : always for sles10 kernel
5061 Description: sles10 support
5062 Details : ll_follow_link: compile fixes and using of nd_set_link
5063 under newer kernels.
5066 Frequency : depends on arch, kernel and compiler version, always on sles10
5069 Description: recursive or deep enough symlinks cause stack overflow
5070 Details : getting rid of large stack-allocated variable in
5074 Frequency : depends on hardware
5076 Description: lustre write performance loss in the SLES10 kernel
5077 Details : the performance loss is caused by using of write barriers in the
5078 ext3 code. The SLES10 kernel turns barrier support on by
5079 default. The fix is to undo that change for ldiskfs.
5081 ------------------------------------------------------------------------------
5083 2006-12-09 Cluster File Systems, Inc. <info@clusterfs.com>
5085 * Support for kernels:
5086 2.6.9-42.0.3EL (RHEL 4)
5087 2.6.5-7.276 (SLES 9)
5088 2.4.21-47.0.1.EL (RHEL 3)
5089 2.6.12.6 vanilla (kernel.org)
5093 Frequency : quota enabled and large files being deleted
5095 Description: releasing more than 4GB of quota at once hangs OST
5096 Details : If a user deletes more than 4GB of files on a single OST it
5097 will cause the OST to spin in an infinite loop. Release
5098 quota in < 4GB chunks, or use a 64-bit value for 1.4.7.1+.
5103 Description: statfs data retrieved from /proc may be stale or zero
5104 Details : When reading per-device statfs data from /proc, in the
5105 {kbytes,files}_{total,free,avail} files, it may appear
5106 as zero or be out of date.
5109 Frequency : systems with MD RAID1 external journal devices
5111 Description: lconf's call to blkid is confused by RAID1 journal devices
5112 Details : Use the "blkid -l" flag to locate the MD RAID device instead
5113 of returning all block devices that match the journal UUID.
5116 Frequency : always, for aggregate stripe size over 4GB
5118 Description: "lfs setstripe" fails assertion when setting 4GB+ stripe width
5119 Details : Using "lfs setstripe" to set stripe size * stripe count over 4GB
5120 will fail the kernel with "ASSERTION(lsm->lsm_xfersize != 0)"
5123 Frequency : always if "lfs find" used on a local file/directory
5125 Description: "lfs find" segfaults if used on a local file/directory
5126 Details : The case where a directory component was not specified wasn't
5127 handled correctly. Handle this properly.
5130 Frequency : always on ppc64
5132 Description: the write to an ext3 filesystem mounted with mballoc got stuck
5133 Details : ext3_mb_generate_buddy() uses find_next_bit() which does not
5134 perform endianness conversion.
5137 Frequency : rarely (truncate to non-zero file size after write under load)
5138 Bugzilla : 10730, 10687
5139 Description: Files padded with zeros to next 4K multiple
5140 Details : With filesystems mounted using the "extents" option (2.6 kernels)
5141 it is possible that files that are truncated to a non-zero size
5142 immediately after being written are filled with zero bytes beyond
5143 the truncated size. No file data is lost.
5145 Severity : enhancement
5147 Description: Allow recovery/failover for liblustre clients.
5148 Details : liblustre clients were unaware of failover configurations until
5151 Severity : enhancement
5153 Description: user file locks should fail when not mounting with flock option
5154 Details : Set up an error-returning stub in ll_file_operations.lock field
5155 to prevent incorrect behaviour when client is mounted without
5156 flock option. Also, set up properly f_op->flock field for
5160 Frequency : always on ia64
5162 Description: "lfs df" loops on printing out MDS statfs information
5163 Details : The obd_ioctl_data was not initialized and in some systems
5164 this caused a failure during the ioctl that did not return
5165 an error. Initialize the struct and return an error on failure.
5168 Frequency : SLES 9 only
5170 Description: Error of copying files with lustre special EAs as root
5171 Details : Client side always return success for setxattr call for lustre
5172 special xattr (currently only "trusted.lov").
5175 Frequency : rarely on clusters with both ia64+i386 clients
5177 Description: ia64+i686 clients doing shared IO on the same file may LBUG
5178 Details : In rare cases when both ia64+i686 (or other mixed-PAGE_SIZE)
5179 clients are doing concurrent writes to the same file it is
5180 possible that the ia64 clients may LASSERT because the OST
5181 extent locks are not PAGE_SIZE aligned. Ensure that grown
5182 locks are always aligned on the request boundary.
5185 Frequency : specific use, occasional
5187 Description: Overwriting in use executable truncates on-disk binary image
5188 Details : If one node attempts to overwrite an executable in use by
5189 another node, we now correctly return ETXTBSY instead of
5190 truncating the file.
5192 Severity : enhancement
5194 Description: Async OSC create to avoid the blocking unnecessarily.
5195 Details : If a OST has no remain object, system will block on the creating
5196 when need to create a new object on this OST. Now, ways use
5197 pre-created objects when available, instead of blocking on an
5198 empty osc while others are not empty. If we must block, we block
5199 for the shortest possible period of time.
5204 Description: chmod on Lustre root is propagated to other clients
5205 Details : Re-validate root's dentry in ll_lookup_it to avoid having it
5206 invalid by the follow_mount time.
5209 Frequency : liblustre clients only
5211 Description: Race in 'instant cancel' lock handling could lead to such locks
5212 never to be granted in case of SMP MDS
5213 Details : Do not destroy not yet granted but cbpending locks in
5217 Frequency : replay/resend of open
5219 Description: non null lock assetion failure in mds_intent_policy
5220 Details : Trying to replay/resend lockless open requests resulted in
5221 mds_open() returning 0 with no lock. Now it sets a flag if
5222 a lock is going to be returned.
5224 Severity : enhancement
5226 Description: Checksum enhancements
5227 Details : New checksum enhancements allow for resending RPCs that failed
5230 Severity : enhancement
5232 Description: Tunables on number of dirty pages in cacche
5233 Details : Allow to set limit on number of dirty pages cached.
5238 Description: client crash on unmount - lock still has references
5239 Details : In some error handling cases it was possible to leak a lock
5240 reference on a client while accessing a file. This was not
5241 harmful to the client during operation, but would cause the
5242 client to crash when the filesystem is unmounted.
5245 Frequency : specific case, rare
5247 Description: ETXTBSY on mds though file not in use by client
5248 Details : ETXTBSY is no longer incorrectly returned when attempting to
5249 chmod or chown a directory that the user previously tried to
5250 execute or a currently-executing binary.
5253 Frequency : extremely rare except on liblustre-based clients
5255 Description: Lustre space not freed when files are deleted
5256 Details : Clean up open-unlinked files after client eviction. Previously
5257 the unlink was skipped and the files remained as orphans.
5262 Description: OST failure "would be an LBUG" in waiting_locks_callback()
5263 Details : In some cases it was possible to send a blocking callback to
5264 a client doing a glimpse, even though that client didn't get
5265 a lock granted. When the glimpse lock is cancelled on the OST
5266 the freed lock is left on the waiting list and corrupted the list.
5269 Frequency : all core dumps
5271 Description: Broke core dumps to lustre
5272 Details : Negative dentry may be unhashed if parent does not have UPDATE
5273 lock, but some callers, e.g. do_coredump, expect dentry to be
5274 hashed after successful create, hash it in ll_create_it.
5276 ------------------------------------------------------------------------------
5278 2006-09-13 Cluster File Systems, Inc. <info@clusterfs.com>
5280 * Support for kernels:
5281 2.6.9-42.0.2.EL (RHEL 4)
5282 2.6.5-7.276 (SLES 9)
5283 2.4.21-40.EL (RHEL 3)
5284 2.6.12.6 vanilla (kernel.org)
5288 Frequency : always on RHEL 3
5290 Description: Number of open files grows over time
5291 Details : The number of open files grows over time, whether or not
5292 Lustre is started. This was due to a filp leak introduced
5293 by one of our kernel patches.
5295 ------------------------------------------------------------------------------
5297 2006-08-20 Cluster File Systems, Inc. <info@clusterfs.com>
5299 * Support for kernels:
5300 2.6.9-42.EL (RHEL 4)
5301 2.6.5-7.267 (SLES 9)
5302 2.4.21-40.EL (RHEL 3)
5303 2.6.12.6 vanilla (kernel.org)
5308 Bugzilla : 5719, 9635, 9792, 9684
5309 Description: OST (or MDS) trips assertions in (re)connection under heavy load
5310 Details : If a server is under heavy load and cannot reply to new
5311 connection requests before the client resends the (re)connect,
5312 the connection handling code can behave badly if two service
5313 threads are concurrently handing separate (re)connections from
5314 the same client. Add better locking to the connection handling
5315 code, and ensure that only a single connection will be processed
5316 for a given client UUID, even if the lock is dropped.
5318 Severity : enhancement
5320 Description: add TCP zero-copy support to kernel
5321 Details : Add support to the kernel TCP stack to allow zero-copy bulk
5322 sends if the hardware supports scatter-gather and checksumming.
5323 This allows socklnd to do client-write and server-read more
5324 efficiently and reduce CPU utilization from skbuf copying.
5327 Frequency : only if NFS exporting from client
5329 Description: NULL pointer deref in ll_iocontrol() if chattr mknod file
5330 Details : If setting attributes on a file created under NFS that had
5331 never been opened it would be possible to oops the client
5332 if the file had no objects.
5336 Bugzilla : 9326, 10402, 10897
5337 Description: client crash in ptlrpcd_wake() thread when sending async RPC
5338 Details : It is possible that ptlrpcd_wake() dereferences a freed async
5339 RPC. In rare cases the ptlrpcd thread alread processed the RPC
5340 before ptlrpcd_wake() was called and the request was freed.
5343 Frequency : always for liblustre
5345 Description: liblustre client does MDS+OSTs setattr RPC for each write
5346 Details : When doing a write from a liblustre client, the client
5347 incorrectly issued an RPC to the MDS and each OST the file was
5348 striped over in order to update the timestamps. When writing
5349 with small chunks and many clients this could overwhelm the MDS
5350 with RPCs. In all cases it would slow down the write because
5351 these RPCs are unnecessary.
5353 Severity : enhancement
5355 Description: allow number of MDS service threads to be changed at module load
5356 Details : It is now possible to change the number of MDS service threads
5357 running. Adding "options mds mds_num_threads={N}" to the MDS's
5358 /etc/modprobe.conf will set the number of threads for the next
5359 time Lustre is restarted (assuming the "mds" module is also
5360 reloaded at that time). The default number of threads will
5361 stay the same, 32 for most systems.
5366 Description: OST crash if filesystem is unformatted or corrupt
5367 Details : If an OST is started on a device that has never been formatted
5368 or if the filesystem is corrupt and cannot even mount then the
5369 error handling cleanup routines would dereference a NULL pointer.
5374 Description: NULL pointer deref in llap_from_page.
5375 Details : get_cache_page_nowait can return a page with NULL (or otherwise
5376 incorrect) mapping if the page was truncated/reclaimed while it was
5377 searched for. Check for this condition and skip such pages when
5378 doing readahead. Introduce extra check to llap_from_page() to
5379 verify page->mapping->host is non-NULL (so page is not anonymous).
5382 Frequency : Sometimes when using sys_sendfile
5384 Description: "page not covered by a lock" warnings from ll_readpage
5385 Details : sendfile called ll_readpage without right page locks present.
5386 Now we introduced ll_file_sendfile that does necessary locking
5387 around call to generic_file_sendfile() much like we do in
5391 Frequency : with certain MDS communication failures at client mount time
5393 Description: NULL pointer deref after failed client mount
5394 Details : a client connection request may delayed by the network layer
5395 and not be sent until after the PTLRPC layer has timed out the
5396 request. If the client fails the mount immediately it will try
5397 to clean up before the network times out the request. Add a
5398 reference from the request import to the obd device and delay
5399 the cleanup until the network drops the request.
5402 Frequency : occasionally during client (re)connect
5404 Description: assertion failure during client (re)connect
5405 Details : processing a client connection request may be delayed by the
5406 client or server longer than the client connect timeout. This
5407 causes the client to resend the connection request. If the
5408 original connection request is replied in this interval, the
5409 client may trip an assertion failure in ptlrpc_connect_interpret()
5410 which thought it would be the only running connect process.
5413 Frequency : only with obd_echo servers and clients that are rebooted
5415 Description: kernel BUG accessing uninitialized data structure
5416 Details : When running an obd_echo server it did not start the ping_evictor
5417 thread, and when a client was evicted an uninitialized data
5418 structure was accessed. Start the ping_evictor in the RPC
5419 service startup instead of the OBD startup.
5421 Severity : enhancement
5422 Bugzilla : 10193 (patchless)
5423 Description: Remove dependency on various unexported kernel interfaces.
5424 Details : No longer need reparent_to_init, exit_mm, exit_files,
5425 sock_getsockopt, filemap_populate, FMODE_EXEC, put_filp.
5428 Frequency : rare (only users of deprecated and unsupported LDAP config)
5430 Description: write_conf for zeroconf mount queried LDAP incorrectly for client
5431 Details : LDAP apparently contains 'lustreName' attributes instead of
5432 'name'. A simple remapping of the name is sufficient.
5435 Frequency : rare (only with non-default dump_on_timeout debug enabled)
5437 Description: waiting_locks_callback trips kernel BUG if client is evicted
5438 Details : Running with the dump_on_timeout debug flag turned on makes
5439 it possible that the waiting_locks_callback() can try to dump
5440 the Lustre kernel debug logs from an interrupt handler. Defer
5441 this log dumping to the expired_lock_main() thread.
5443 Severity : enhancement
5445 Description: Support NFS exporting on 2.6 kernels.
5446 Details : Implement non-rawops metadata methods for NFS server to use without
5447 changing NFS server code.
5450 Frequency : very rare (synthetic metadata workload only)
5452 Description: two racing renames might cause an MDS thread to deadlock
5453 Details : Running the "racer" program may cause one MDS thread to rename
5454 a file from being the source of a rename to being the target of
5455 a rename at exactly the same time that another thread is doing
5456 so, and the second thread has already enqueued these locks after
5457 doing a lookup of the target and is trying to relock them in
5458 order. Ensure that we don't try to re-lock the same resource.
5461 Frequency : only very large systems with liblustre clients
5463 Description: slow eviction of liblustre clients with the "evict_by_nid" RPC
5464 Details : Use asynchronous set_info RPCs to send the "evict_by_nid" to
5465 all OSTs in parallel. This allows the eviction of stale liblustre
5466 clients to proceed much faster than if they were done in series,
5467 and also offers similar improvements for other set_info RPCs.
5472 Description: excessive CPU usage during initial read phase on client
5473 Details : During the initial read phase on a client, it would agressively
5474 retry readahead on the file, consuming too much CPU and impacting
5475 performance (since 1.4.5.8). Improve the readahead algorithm
5476 to avoid this, and also improve some other common cases (read
5477 of small files in particular, where "small" is files smaller than
5478 /proc/fs/lustre/llite/*/max_read_ahead_whole_mb, 2MB by default).
5483 Description: MDS crash when receiving packet with unknown intent.
5484 Details : Do not LBUG in unknown intent case, just return -EFAULT
5486 Severity : enhancement
5487 Bugzilla : 9293, 9385
5488 Description: MDS RPCs are serialised on client. This is unnecessary for some.
5489 Details : Do not serialize getattr (non-intent version) and statfs.
5492 Frequency : occasional, when OST network is overloaded/intermittent
5494 Description: client evicted by OST after bulk IO timeout
5495 Details : If a client sends a bulk IO request (read or write) the OST
5496 may evict the client if it is unresposive to its data GET/PUT
5497 request. This is incorrect if the network is overloaded (takes
5498 too long to transfer the RPC data) or dropped the OST GET/PUT
5499 request. There is no need to evict the client at all, since
5500 the pinger and/or lock callbacks will handle this, and the
5501 client can restart the bulk request.
5504 Frequency : Always when mmapping file with no objects
5506 Description: client crashes when mmapping file with no objects
5507 Details : Check that we actually have objects in a file before doing any
5508 operations on objects in ll_vm_open, ll_vm_close and
5514 Description: Request leak when working with deleted CWD
5515 Details : Introduce advanced request refcount tracking for requests
5516 referenced from lustre intent.
5518 Severity : Enhancement
5520 Description: Cache open file handles on client.
5521 Details : MDS now will return special lock along with openhandle, if
5522 requested and client is allowed to hold openhandle, even if unused,
5523 until such a lock is revoked. Helps NFS a lot, since NFS is opening
5524 closing files for every read/write openration.
5526 Severity : Enhancement
5528 Description: Cache open negative dentries on client when possible.
5529 Details : Guard negative dentries with UPDATE lock on parent dir, drop
5530 negative dentries on lock revocation.
5535 Description: Remounting a client read-only wasn't possible with a zconf mount
5536 Details : It wasn't possible to remount a client read-only with llmount.
5538 Severity : enhancement
5539 Description: Include MPICH 1.2.6 Lustre ADIO interface patch
5540 Details : In lustre/contrib/ or /usr/share/lustre in RPM a patch for
5541 MPICH is included to add Lustre-specific ADIO interfaces.
5542 This is based closely on the UFS ADIO layer and only differs
5543 in file creation, in order to allow the OST striping to be set.
5544 This is user-contributed code and not supported by CFS.
5549 Description: extended inode attributes (immutable, append-only) work improperly
5550 when 2.4 and 2.6 kernels are used on client/server or vice versa
5551 Details : Introduce kernel-independent values for these flags.
5553 Severity : enhancement
5556 Description: Allow fractional MB tunings for lustre in /proc/ filesystem.
5557 Details : Many of the /proc/ tunables can only be tuned at a megabyte
5558 granularity. Now, Fractional MB granularity is be supported,
5559 this is very useful for low memory system.
5561 Severity : enhancement
5563 Description: Getattr by fid
5564 Details : Getting a file attributes by its fid, obtaining UPDATE|LOOKUP
5565 locks, avoids extra getattr rpc requests to MDS, allows '/' to
5566 have locks and avoids getattr rpc requests for it on every stat.
5569 Frequency : Always, for filesystems larger than 2TB
5571 Description: ldiskfs crash at mount for filesystem larger than 2TB with mballoc
5572 Details : Kenrel kmalloc limits allocations to 128kB and this prevents
5573 filesystems larger than 2TB to be mounted with mballoc enabled.
5576 Frequency : Always, for 32-bit kernel without CONFIG_LBD and filesystem > 2TB
5578 Description: filesystem corruption for non-standard kernels and very large OSTs
5579 Details : If a 32-bit kernel is compiled without CONFIG_LBD enabled and a
5580 filesystems larger than 2TB is mounted then the kernel will
5581 silently corrupt the start of the filesystem. CONFIG_LBD is
5582 enabled for all CFS-supported kernels, but the possibility of
5583 this happening with a modified kernel config exists.
5585 Severity : enhancement
5587 Description: add client O_DIRECT support for 2.6 kernels
5588 Details : It is now possible to do O_DIRECT reads and writes to files
5589 in the Lustre client mountpoint on 2.6 kernel clients.
5591 Severity : enhancement
5593 Description: parallel glimpse, setattr, statfs, punch, destroy requests
5594 Details : Sends glimpse, setattr, statfs, punch, destroy requests to OSTs in
5595 parallel, not waiting for response from every OST before sending
5596 a rpc to the next OST.
5601 Description: setattr vs write race when updating file timestamps
5602 Details : Client processes that update a file timestamp into the past
5603 right after writing to the file (e.g. tar) it is possible that
5604 the updated file modification time can be reset to the current
5605 time due to a race between processing the setattr and write RPC.
5607 Severity : enhancement
5609 Description: Bring 'lfs find' closer in line with regular Linux find.
5610 Details : lfs find util supports -atime, -mtime, -ctime, -maxdepth, -print,
5611 -print0 options and obtains all the needed info through the lustre
5614 Severity : enhancement
5616 Description: support up to 1024 configured devices on one node
5617 Details : change obd_dev array from statically allocated to dynamically
5618 allocated structs as they are first used to reduce memory usage
5623 Description: Flush dirty partially truncated pages during truncate
5624 Details : Immediatelly flush partially truncated pages in filter_setattr,
5625 this way we completely avoid having any pages in page cache on OST
5626 and can retire ugly workarounds during writes to flush such pages.
5631 Description: i_sem vs transaction deadlock in mds_obd_destroy during unlink.
5632 Details : protect inode from truncation within vfs_unlink() context
5633 just take a reference before calling vfs_unlink() and release it
5634 when parent's i_sem is free.
5637 Frequency : always, if extents are used on OSTs
5639 Description: index ei_leaf_hi (48-bit extension) is not zeroed in extent index
5640 Details : OSTs using the extents format would not zero the high 16 bits of
5641 the index physical block number. This is not a problem for any
5642 OST filesystems smaller than 16TB, and no kernels support ext3
5643 filesystems larger than 16TB yet. This is fixed in 1.4.7 (all
5644 new/modified files) and can be fixed for existing filesystems
5645 with e2fsprogs-1.39-cfs1.
5650 Description: import connection selection may be incorrect if timer wraps
5651 Details : Using a 32-bit jiffies timer with HZ=1000 may cause backup
5652 import connections to be ignored if the 32-bit jiffies counter
5653 wraps. Use a 64-bit jiffies counter.
5656 Frequency : during server recovery
5658 Description: crash after server is denying duplicate export
5659 Details : If clients are resending connect requests to the server, the
5660 server refuses to allow a client to connect multiple times.
5661 Fixed a bug in the handling of this case.
5664 Frequency : very large clusters immediately after boot
5666 Description: LNET request buffers exhausted under heavy short-term load
5667 Details : If a large number of client requests are generated on a service
5668 that has previously never seen so many requests it is possible
5669 that the request buffer growth cannot keep up with the spike in
5670 demand. Instead of dropping incoming requests, they are held in
5671 the LND until the RPC service can accept more requests.
5674 Frequency : Sometimes during replay
5676 Description: Assertion failure in ll_local_open after replay.
5677 Details : If replay happened on an open request reply before we were able
5678 to set replay handler, reply will become not swabbed tripping the
5679 assertion in ll_local_open. Now we set the handler right after
5680 recognising of open request
5683 Frequency : very rare
5685 Description: kernel reports "badness in vsnprintf"
5686 Details : Reading from the "recovery_status" /proc file in small chunks
5687 may cause a negative length in lprocfs_obd_rd_recovery_status()
5688 call to vsnprintf() (which is otherwise harmless). Exit early
5689 if there is no more space in the output buffer.
5691 Severity : enhancement
5693 Description: clear OBD RPC statistics by writing to them
5694 Details : It is now possible to clear the OBD RPC statistics by writing
5695 to the "stats" file.
5700 Description: Client mtime is not the same on different clients after utimes
5701 Details : In some cases, the client was using the utimes() syscall on
5702 a file cached on another node. The clients now validate the
5703 ctime from the MDS + OSTs to determine which one is right.
5708 Description: Inability to activate failout mode
5709 Details : lconf script incorrectly assumed that in python string's numeric
5710 value is used in comparisons.
5713 Frequency : always with multiple stripes per file
5715 Description: Inefficient object allocation for mutli-stripe files
5716 Details : When selecting which OSTs to stripe files over, for files with
5717 a stripe count that divides evenly into the number of OSTs,
5718 the MDS is always picking the same starting OST for each file.
5719 Return the OST selection heuristic to the original design.
5724 Description: mount failures may take full timeout to return an error
5725 Details : Under some heavy load conditions it is possible that a
5726 failed mount can wait for the full obd_timeout interval,
5727 possibly several minutes, before reporting an error.
5728 Instead return an error as soon as the status is known.
5730 ------------------------------------------------------------------------------
5732 2006-02-14 Cluster File Systems, Inc. <info@clusterfs.com>
5734 * WIRE PROTOCOL CHANGE. This version of Lustre networking WILL NOT
5735 INTEROPERATE with older versions automatically. Please read the
5736 user documentation before upgrading any part of a live system.
5737 * WARNING: Lustre networking configuration changes are required with
5738 this release. See https://bugzilla.clusterfs.com/show_bug.cgi?id=10052
5741 * Support for kernels:
5742 2.6.9-22.0.2.EL (RHEL 4)
5743 2.6.5-7.244 (SLES 9)
5744 2.6.12.6 vanilla (kernel.org)
5747 Severity : enhancement
5748 Bugzilla : 7981/8208
5749 Description: Introduced Lustre Networking (LNET)
5750 Details : LNET is new networking infrastructure for Lustre, it includes
5751 a reorganized network configuration mode (see the user
5752 documentation for full details) as well as support for routing
5753 between different network fabrics. Lustre Networking Devices
5754 (LNDs) for the supported network fabrics have also been
5755 created for this new infrastructure.
5757 Severity : enhancement
5758 Description: Introduced Access control lists
5759 Details : clients can set ACLs on files and directories in order to have
5760 more fine-grained permissions than the standard Unix UGO+RWX.
5761 The MDS must be started with the "-o acl" mount option.
5763 Severity : enhancement
5764 Description: Introduced filesystem quotas
5765 Details : Administrators may now establish per-user quotas on the
5768 Severity : enhancement
5770 Description: Configuration change for the XT3
5771 The PTLLND is now used to run Lustre over Portals on the XT3
5772 The configure option(s) --with-cray-portals are no longer used.
5773 Rather --with-portals=<path-to-portals-includes> is used to
5774 enable building on the XT3. In addition to enable XT3 specific
5775 features the option --enable-cray-xt3 must be used.
5780 Description: Running on many-way SMP OSTs can trigger oops in llcd_send()
5781 Details : A race between allocating a new llcd and re-getting the llcd_lock
5782 allowed another thread to grab newly-allocated llcd.
5784 Severity : enhancement
5786 Description: 2.6 OST async journal commit and locking fix to improve performance
5787 Details : The filter_direct_io()+filter_commitrw_write() journal commits for
5788 2.6 kernels are now async as they already were in 2.4 kernels so
5789 that they can commit concurrently with the network bulk transfer.
5790 For block-allocated files the filter allocation semaphore is held
5791 to avoid filesystem fragmentation during allocation. BKL lock
5792 removed for 2.6 xattr operations where it is no longer needed.
5797 Description: lconf incorrectly determined whether two IP networks could talk
5798 Details : In some more complicated routing and multiple-network
5799 configurations, lconf will avoid trying to make a network
5800 connection to a disjoint part of the IP space. It was doing the
5801 math incorrectly for one set of cases.
5806 Description: Fix for potential infinite loop processing records in an llog.
5807 Details : If an llog record is corrupted/zeroed, it is possible to loop
5808 forever in llog_process(). Validate the llog record length
5809 and skip the remainder of the block on error.
5812 Frequency : occasional (liblustre only)
5814 Description: liblustre could not open files whose last component is a symlink
5815 Details : sysio_path_walk() would incorrectly pass the open intent to
5816 intermediate path components.
5819 Frequency : rare (liblustre only with non-standard tuning)
5820 Bugzilla : 7201 (7350)
5821 Description: Tuning the MDC DLM LRU size to zero triggers client LASSERT
5822 Details : llu_lookup_finish_locks() tries to set lock data on a lock
5823 after it has been released, only do this for referenced locks
5825 Severity : enhancement
5827 Description: specifying an (invalid) directory default stripe_size of -1
5828 would reset the directory default striping
5829 Details : stripe_size -1 was used internally to signal directory stripe
5830 removal, now use "all default" to signal dir stripe removal
5831 as a directory striping of "all default" is not useful
5834 Frequency : common for large clusters running liblustre clients
5836 Description: doing an ls when liblustre clients are running is slow
5837 Details : sending a glimpse AST to a liblustre client waits for every AST
5838 to time out, as liblustre clients will not respond. Since they
5839 cannot cache data we refresh the OST lock LVB from disk instead.
5841 Severity : enhancement
5843 Description: doing an ls at the same time as file IO can be slow
5844 Details : enqueue and other "small" requests can be blocked behind many
5845 large IO requests. Create a new OST IO portal for non-IO
5846 requests so they can be processed faster.
5849 Frequency : rare (only HPUX clients mounting unsupported re-exported NFS vol)
5851 Description: an HPUX NFS client would get -EACCESS when ftruncate()ing a newly
5852 created file with mode 000
5853 Details : the Linux NFS server relies on an MDS_OPEN_OWNEROVERRIDE hack to
5854 allow an ftruncate() as a non-root user to a file with mode 000.
5855 Lustre now respects this flag to disable mode checks when
5856 truncating a file owned by the user
5859 Frequency : liblustre-only, when liblustre client dies unexpectedly or becomes
5862 Description: Revoking locks from clients that went dead or catatonic might take
5864 Details : New lock flags FL_CANCEL_ON_BLOCK used by liblustre makes
5865 cancellation of such locks instant on servers without waiting for
5866 any reply from clients. Clients drops these locks when cancel
5867 notification from server is received without replying.
5870 Frequency : liblustre-only, when liblustre client dies or becomes busy
5872 Description: Doing ls on Linux clients can take a long time with active
5874 Details : Liblustre client cannot handle ASTs in timely manner, so avoid
5875 granting such locks to it in the first place if possible. Locks
5876 are taken by proxy on the OST during the read or write and
5877 dropped immediately afterward. Add connect flags handling, do
5878 not grant locks to liblustre clients for glimpse ASTs.
5880 Severity : enhancement
5882 Description: Improve read-ahead algorithm to avoid excessive IO for random reads
5883 Details : Existing read-ahead algorithm is tuned for the case of streamlined
5884 sequential reads and behaves badly with applications doing random
5885 reads. Improve it by reading ahead at least read region, and
5886 avoiding excessive large RPC for small reads.
5888 Severity : enhancement
5890 Description: Creating more than 1000 files for a single job may cause a load
5891 imbalance on the OSTs if there are also a large number of OSTs.
5892 Details : qos_prep_create() uses an OST index reseed value that is an
5893 even multiple of the number of available OSTs so that if the
5894 reseed happens in the middle of the object allocation it will
5895 still utilize the OSTs as uniformly as possible.
5900 Description: OST or MDS may oops in ping_evictor_main()
5901 Details : ping_evictor_main() drops obd_dev_lock if deleting a stale export
5902 but doesn't restart at beginning of obd_exports_timed list
5905 Severity : enhancement
5907 Description: improve by-nid export eviction on the MDS and OST
5908 Details : allow multiple exports with the same NID to be evicted at one
5909 time without re-searching the exports list.
5912 Frequency : rare, only with supplementary groups enabled on SMP 2.6 kernels
5914 Description: MDS may oops in groups_free()
5915 Details : in rare race conditions a newly allocated group_info struct is
5916 freed again, and this can be NULL. The 2.4 compatibility code
5917 for groups_free() checked for a NULL pointer, but 2.6 did not.
5920 Frequency : common for liblustre clients doing little filesystem IO
5921 Bugzilla : 9352, 7313
5922 Description: server may evict liblustre clients accessing contended locks
5923 Details : if a client is granted a lock or receives a completion AST
5924 with a blocking AST already set it would not reply to the AST
5925 for LDLM_FL_CANCEL_ON_BLOCK locks. It now replies to such ASTs.
5928 Frequency : lfs setstripe, only systems with more than 160 OSTs
5930 Description: unable to set striping with a starting offset beyond OST 160
5931 Details : llapi_create_file() incorrectly limited the starting stripe
5932 index to the maximum single-file stripe count.
5935 Frequency : LDAP users only
5937 Description: lconf did not handle in-kernel recovery with LDAP properly
5938 Details : lconf/LustreDB get_refs() is searching the wrong namespace
5940 Severity : enhancement
5942 Description: bind OST threads to NUMA nodes to improve performance
5943 Details : all OST threads are uniformly bound to CPUs on a single NUMA
5944 node and do their allocations there to localize memory access
5946 Severity : enhancement
5948 Description: llmount can determine client NID directly from Myrinet (GM)
5949 Details : the client NID code from gmnalnid was moved directly into
5950 llmount, removing the need to use this or specifying the
5951 client NID explicitly when mounting GM clients with zeroconf
5954 Frequency : if client is started with down MDS
5956 Description: if client is started with down MDS mount hangs in ptlrpc_queue_wait
5957 Details : Having an LWI_INTR() wait event (interruptible, but no timeout)
5958 will wait indefinitely in ptlrpc_queue_wait->l_wait_event() after
5959 ptlrpc_import_delayed_req() because we didn't check if the
5960 request was interrupted, and we also didn't break out of the
5961 event loop if there was no timeout
5966 Description: data loss during non-page-aligned writes to a single file from
5967 both multiple nodes and multiple threads on one node at same time
5968 Details : updates to KMS and lsm weren't protected by common lock. Resulting
5969 inconsistency led to false short-reads, that were cached and later
5970 used by ->prepare_write() to fill in partially written page,
5971 leading to data loss.
5974 Frequency : always, if lconf --abort_recovery used
5976 Description: lconf --abort_recovery fails with 'Operation not supported'
5977 Details : lconf was attempting to abort recovery on the MDT device and not
5980 Severity : enhancement
5982 Description: remove cleanup logs
5983 Details : replace lconf-generated cleanup logs with lustre internal
5984 cleanup routines. Eliminates the need for client-cleanup and
5987 Severity : enhancement
5989 Description: add support for EAs (user and system) on lustre filesystems
5990 Details : it is now possible to store extended attributes in the Lustre
5991 client filesystem, and with the user_xattr mount option it
5992 is possible to allow users to store EAs on their files also
5994 Severity : enhancement
5996 Description: Add possibility (config option) to show minimal available OST free
5998 Details : When compiled with --enable-mindf configure option, statfs(2)
5999 (and so, df) will return least minimal free space available from
6000 all OSTs as amount of free space on FS, instead of summary of
6001 free spaces of all OSTs.
6003 Severity : enhancement
6005 Description: do not expand extent locks acquired on OST-side
6006 Details : Modify ldlm_extent_policy() to not expand local locks, acquired
6007 by server: they are not cached anyway.
6010 Frequency : when mmap is used/binaries executed from Lustre
6012 Description: Unmmap pages before throwing them away from read cache.
6013 Details : llap_shrink cache now attempts to unmap pages before discarding
6014 them (if unmapping failed - do not discard). SLES9 kernel has
6015 extra checks that trigger if this unmapping is not done first.
6020 Description: lconf didn't resolve symlinks before checking to see whether a
6021 given mountpoint was already in use
6024 Frequency : when migrating failover services
6025 Bugzilla : 6395, 9514
6026 Description: When migrating a subset of services from a node (e.g. failback
6027 from a failover service node) the remaining services would
6028 time out and evict clients.
6029 Details : lconf --force (implied by --failover) sets the global obd_timeout
6030 to 5 seconds in order to quickly disconnect, but this caused
6031 other RPCs to time out too quickly. Do not change the global
6032 obd_timeout for force cleanup, only set it for DISCONNECT RPCs.
6034 Severity : enhancement
6035 Frequency : if MDS is started with down OST
6036 Bugzilla : 9439,5706
6037 Description: Allow startup/shutdown of an MDS without depending on the
6038 availability of the OSTs.
6039 Details : Asynchronously call mds_lov_synchronize during MDS startup.
6040 Add appropriate locking and lov-osc refcounts for safe
6041 cleaning. Add osc abort_inflight calls in case the
6042 synchronize never started.
6045 Frequency : occasional (Cray XT3 only)
6047 Description: root not authorized to access files in CRAY_PORTALS environment
6048 Details : The client process capabilities were not honoured on the MDS in
6049 a CRAY_PORTALS/CRAY_XT3 environment. If the file had previously
6050 been accessed by an authorized user then root was able to access
6051 the file on the local client also. The root user capabilities
6052 are now allowed on the MDS, as this environment has secure UID.
6055 Frequency : occasional
6057 Description: ldiskfs "too long searching" message happens too often
6058 Details : A debugging message (otherwise harmless) prints too often on
6059 the OST console. This has been reduced to only happen when
6060 there are fragmentation problems on the filesystem.
6065 Description: Division by zero in statfs when all OSCs are inactive
6066 Details : lov_get_stripecnt() returns zero due to incorrect order of checks,
6067 lov_statfs divides by value returned by lov_get_stripecnt().
6071 Bugzilla : 9489, 3273
6072 Description: First write from each client to each OST was only 4kB in size,
6073 to initialize client writeback cache, which caused sub-optimal
6074 RPCs and poor layout on disk for the first writen file.
6075 Details : Clients now request an initial cache grant at (re)connect time
6076 and so that they can start streaming writes to the cache right
6077 away and always do full-sized RPCs if there is enough data.
6078 If the OST is rebooted the client also re-establishes its grant
6079 so that client cached writes will be honoured under the grant.
6084 Description: Slow ls (and stat(2) syscall) on files residing on IO-loaded OSTs
6085 Details : Now I/O RPCs go to different portal number and (presumably) fast
6086 lock requests (and glimses) and other RPCs get their own service
6087 threads pool that should be able to service those RPCs
6090 Severity : enhancement
6092 Description: Ability to exchange lustre version between client and servers and
6093 issue warnings at client side if client is too old. Also for
6094 liblustre clients there is ability to refuse connection of too old
6096 Details : New 'version' field is added to connect data structure that is
6097 filled with version info. That info is later checked by server and
6101 Frequency : rare, liblustre only.
6102 Bugzilla : 9296, 9581
6103 Description: Two simultaneous writes from liblustre at offset within same page
6104 might proceed at the same time overwriting eachother with stale
6106 Details : I/O lock withing llu_file_prwv was released too early, before data
6107 actually was hitting the wire. Extended lock-holding time until
6108 server acknowledges receiving data.
6111 Frequency : extremely rare. Never observed in practice.
6113 Description: avoid generating lustre_handle cookie of 0.
6114 Details : class_handle_hash() generates handle cookies by incrementing
6115 global counter, and can hit 0 occasionaly (this is unlikely, but
6116 not impossible, because initial value of cookie counter is
6117 selected randonly). Value of 0 is used as a sentinel meaning
6118 "unassigned handle" --- avoid it. Also coalesce two critical
6119 sections in this function into one.
6121 Severity : enhancement
6123 Description: allow liblustre clients to delegate truncate locking to OST
6124 Details : To avoid overhead of locking, liblustre client instructs OST to
6125 take extent lock in ost_punch() on client's behalf. New connection
6126 flag is added to handle backward compatibility.
6128 Severity : enhancement
6129 Bugzilla : 4928, 7341, 9758
6130 Description: allow number of OST service threads to be specified
6131 Details : a module parameter allows the number of OST service threads
6132 to be specified via "options ost ost_num_threads={N}" in the
6133 OSS's /etc/modules.conf or /etc/modprobe.conf.
6137 Bugzilla : 6146, 9635, 9895
6138 Description: servers crash with bad pointer in target_handle_connect()
6139 Details : In rare cases when a client is reconnecting it was possible that
6140 the connection request was the last reference for that export.
6141 We would temporarily drop the export reference and get a new
6142 one, but this may have been the last reference and the export
6143 was just destroyed. Get new reference before dropping old one.
6145 Severity : enhancement
6146 Frequency : if client is started with failover MDS
6148 Description: Allow multiple MDS hostnames in the mount command
6149 Details : Try to read the configuration from all specified MDS
6150 hostnames during a client mount in case the "primary"
6153 Severity : enhancement
6155 Description: Stop sending data to evicted clients as soon as possible.
6156 Details : Check if the client we are about to send or are sending data to
6157 was evicted already. (Check is done every second of waiting,
6158 for which l_wait_event interface was extended to allow checking
6159 of exit condition at specified intervals).
6162 Frequency : rare, normally only when NFS exporting is done from client
6164 Description: 'bad disk LOV MAGIC: 0x00000000' error when chown'ing files
6166 Details : Make mds_get_md() recognise empty md case and set lmm size to 0.
6169 Frequency : always, if srand() is called before liblustre initialization
6171 Description: Liblustre uses system PRNG disturbing its usage by user application
6172 Details : Introduce internal to lustre fast and high-quality PRNG for
6173 lustre usage and make liblustre and some other places in generic
6174 lustre code to use it.
6176 Severity : enhancement
6177 Bugzilla : 9477, 9557, 9870
6178 Description: Verify that the MDS configuration logs are updated when xml is
6179 Details : Check if the .xml configuration logs are newer than the config
6180 logs stored on the MDS and report an error if this is the case.
6181 Request --write-conf, or allow starting with --old_conf.
6183 Severity : enhancement
6185 Description: Handle symlinks in the path when checking if Lustre is mounted.
6186 Details : Resolve intermediate symlinks when checking if a client has
6187 mounted a filesystem to avoid duplicate client mounts.
6192 Description: lconf can hit an error exception but still return success.
6193 Details : The lconf command catches the Command error exception at the top
6194 level script context and will exit with the associated exit
6195 status, but doesn't ensure that this exit status is non-zero.
6200 Description: failure of ptlrpc thread startup can cause oops
6201 Details : Starting a ptlrpc service thread can fail if there are a large
6202 number of threads or the server memory is very fragmented.
6203 Handle this without oopsing.
6206 Frequency : always, only if liblustre and non-default acceptor port was used
6208 Description: liblustre cannot connect to servers with non-default acceptor port
6209 Details : tcpnal_set_default_params() was not called and was therefore
6210 ignoring the environment varaible TCPNAL_PORT, as well as other
6211 TCPNAL_ environment variables
6216 Description: two objects could be created on the same OST for a single file
6217 Details : If an OST is down, in some cases it was possible to create two
6218 objects on a single OST for a single file. No problems other
6219 than potential performance impact and spurious error messages.
6223 Bugzilla : 5681, 9562
6224 Description: Client may oops in ll_unhash_aliases
6225 Details : Client dcache may become inconsistent in race condition.
6226 In some cases "getcwd" can fail if the current directory is
6232 Description: Inode refcounting problems in NFS export code
6233 Details : link_raw functions used to call d_instantiate without obtaining
6234 extra inode reference first.
6238 Bugzilla : 9942, 9903
6239 Description: Referencing freed requests leading to crash, memleaks with NFS.
6240 Details : We used to require that call to ll_revalidate_it was always
6241 followed by ll_lookup_it. Also with revalidate_special() it is
6242 possible to call ll_revalidate_it() twice for the same dentry
6243 even if first occurence returned success. This fix changes semantic
6244 between DISP_ENQ_COMPLETE disposition flag to mean there is extra
6245 reference on a request referred from the intent.
6246 ll_intent_release() then releases such a request.
6249 Frequency : rare, normally benchmark loads only
6251 Description: unlinked inodes were kept in memory on the client
6252 Details : If a client is repeatedly creating and unlinking files it
6253 can accumulate a lot of stale inodes in the inode slab cache.
6254 If there is no other client load running this can cause the
6255 client node to run out of memory. Instead flush old inodes
6256 from client cache that have the same inode number as a new inode.
6259 Frequency : SLES9 2.6.5 kernel and long filenames only
6260 Bugzilla : 9969, 10379
6261 Description: utime reports stale NFS file handle
6262 Details : SLES9 uses out-of-dentry names in some cases, which confused
6263 the lustre dentry revalidation. Change it to always use the
6267 Frequency : rare, unless heavy write-truncate concurrency is continuous
6268 Bugzilla : 4180, 6984, 7171, 9963, 9331
6269 Description: OST becomes very slow and/or deadlocked during object unlink
6270 Details : filter_destroy() was holding onto the parent directory lock
6271 while truncating+unlinking objects. For very large objects this
6272 may block other threads for a long time and slow overall OST
6273 responsiveness. It may also be possible to get a lock ordering
6274 deadlock in this case, or run out of journal credits because of
6275 the combined truncate+unlink. Solution is to do object truncate
6276 first in one transaction without parent lock, and then do the
6277 final unlink in a new transaction with the parent lock. This
6278 reduces the lock hold time dramatically.
6281 Frequency : rare, 2.4 kernels only
6283 Description: MDS or OST cleanup may trip kernel BUG when dropping kernel lock
6284 Details : mds_cleanup() and filter_cleanup() need to drop the kernel lock
6285 before unmounting their filesystem in order to avoid deadlock.
6286 The kernel_locked() function in 2.4 kernels only checks whether
6287 the kernel lock is held, not whether it is this process that is
6288 holding it as 2.6 kernels do.
6293 Description: MDS or OST may oops/LBUG if a client is connecting multiple times
6294 Details : The client ptlrpc code may be trying to reconnect to a down
6295 server before a previous connection attempt has timed out.
6296 Increase the reconnect interval to be longer than the connection
6297 timeout interval to avoid sending duplicate connections to
6301 Frequency : echo_client brw_test command
6303 Description: fix echo_client to work with OST preallocated code
6304 Details : OST preallocation code (5137) didn't take echo_client IO path
6305 into account: echo_client calls filter methods outside of any
6306 OST thread and, hence, there is no per-thread preallocated
6307 pages and buffers to use. Solution: hijack pga pages for IO. As
6308 a byproduct, this avoids unnecessary data copying.
6312 Bugzilla : 3555, 5962, 6025, 6155, 6296, 9574
6313 Description: Client can oops in mdc_commit_close() after open replay
6314 Details : It was possible for the MDS to return an open request with no
6315 transaction number in mds_finish_transno() if the client was
6316 evicted, but without actually returning an error. Clients
6317 would later try to replay that open and may trip an assertion
6318 Simplify the client close codepath, and always return an error
6319 from the MDS in case the open is not successful.
6322 Frequency : rare, 2.6 OSTs only
6324 Description: OST may deadlock under high load on fragmented files
6325 Details : If there was a heavy load and highly-fragmented OST filesystems
6326 it was possible to have all the OST threads deadlock waiting on
6327 allocation of biovecs, because the biovecs were not released
6328 until the entire RPC IO was completed. Instead, release biovecs
6329 as soon as they are complete to ensure forward IO progress.
6331 Severity : enhancement
6333 Description: Support for specifying external journal device at mount
6334 Details : If an OST or MDS device is formatted with an external journal
6335 device, this device major/minor is stored in the ext3 superblock
6336 and may not be valid for failover. Allow detecting and
6337 specifying the external journal at mount time.
6342 Description: Mounting an MDS with pending unlinked files may cause oops
6343 Details : target_finish_recovery() calls mds_postrecov() which returned
6344 the number of orphans unlinked. mds_lov_connect->mds_postsetup()
6345 considers this an error and immediately begins cleaning up the
6346 lov, just after starting the mds_lov process
6348 Severity : enhancement
6350 Description: Implement 'lfs df' to report actual free space on per-OST basis
6351 Details : Add sub-command 'df' on 'lfs' to report the disk space usage of
6352 MDS/OSDs. Usage: lfs df [-i][-h]. Command Options: '-i' to report
6353 usage of objects; '-h' to report in human readable format.
6355 ------------------------------------------------------------------------------
6357 2005-08-26 Cluster File Systems, Inc. <info@clusterfs.com>
6364 Description: Mounting an ldiskfs file system with mballoc may crash OST node.
6365 Details : ldiskfs mballoc code may reference an uninitialized buddy struct
6366 at startup during orphan unlinking. Instead, skip buddy update
6367 before setup, as it will be regenerated after recovery is complete.
6372 Description: If an OST is inactive, its locks might reference stale inodes.
6373 Details : lov_change_cbdata() must iterate over all namespaces, even if
6374 they are inactive to clear inode references from the lock.
6376 Severity : enhancement
6377 Frequency : occasional, if non-standard max_dirty_mb used
6379 Description: Client will block write RPCs if not enough grant
6380 Details : If a client has max_dirty_mb smaller than max_rpcs_in_flight,
6381 then the client will block writes while waiting for another RPC
6382 to complete instead of consuming its dirty limit. With change
6383 we get improved performance when max_dirty_mb is small.
6385 Severity : enhancement
6386 Bugzilla : 3389, 6253
6387 Description: Add support for supplementary groups on the MDS.
6388 Details : The MDS has an upcall /proc/fs/lustre/mds/{mds}/group_upcall
6389 (set to /usr/sbin/l_getgroups if enabled) which will do MDS-side
6390 lookups for user supplementary groups into a cache.
6394 Description: O_CREAT|O_EXCL open flags in liblustre always return -EEXIST
6395 Details : Make libsysio to not enforce O_EXCL by clearing the flag,
6396 for liblustre O_EXCL is enforced by MDS.
6400 Description: readdir never returns NULL in liblustre.
6401 Details : Corrected llu_iop_getdirentries logic, to return offset of next
6402 dentry in struct dirent.
6406 Frequency : liblustre only, depends on application IO pattern
6407 Description: liblustre clients evicted if not contacting servers
6408 Details : Don't put liblustre clients into the ping_evictor list, so
6409 they will not be evicted by the pinger ever.
6411 Severity : enhancement
6413 Description: Add ability to evict clients by NID from MDS.
6414 Details : By echoing "nid:$NID" string into
6415 /proc/fs/lustre/mds/.../evict_client client with nid that equals to
6416 $NID would be instantly evicted from this MDS and from all active
6417 OSTs connected to it.
6421 Description: Do not query file size twice, somewhat slowing stat(2) calls.
6422 Details : lookup_it_finish() used to query file size from OSTs that was not
6427 Description: service threads change working directory to that of init
6428 Details : Starting lustre service threads may pin the working directory
6429 of the parent thread, making that filesystem busy. Threads
6430 now change to the working directory of init to avoid this.
6434 Frequency : during shutdown only
6435 Description: shutdown with a failed MDS or OST can cause unmount to hang
6436 Details : Don't resend DISCONNECT messages in ptlrpc_disconnect_import()
6441 Frequency : 2.6 only
6442 Description: chmod/chown may include an extra supplementary group
6443 Details : ll{,u}_mdc_pack_op_data() does not properly initialize the
6444 supplementary group and if none is specified this is used.
6447 Bugzilla : 5479 (6816)
6449 Description: Racing open + rm can assert client in mdc_set_open_replay_data()
6450 Details : If lookup is in progress on a file that is unlinked we might try
6451 to revalidate the inode and fail in revalidate after lookup is
6452 complete and ll_file_open() enqueues the open again but
6453 it_open_error() was not checking DISP_OPEN_OPEN errors correctly.
6456 Frequency : always, if lconf --abort_recovery used
6458 Description: lconf --abort_recovery fails with 'Operation not supported'
6459 Details : lconf was attempting to abort recovery on the MDT device and not
6462 ------------------------------------------------------------------------------
6464 2005-08-08 Cluster File Systems, Inc. <info@clusterfs.com>
6469 Frequency : rare (only unsupported configurations with a node running as an
6471 Bugzilla : 6514, 5137
6472 Description: Mounting a Lustre file system on a node running as an OST could
6474 Details : OSTs now preallocates memory needed to write out data at
6475 startup, instead of when needed, to avoid having to
6476 allocate memory in possibly low memory situations.
6477 Specifically, if the file system is mounted on on OST,
6478 memory pressure could force it to try to write out data,
6479 which it needed to allocate memory to do. Due to the low
6480 memory, it would be unable to do so and the node would
6481 become unresponsive.
6483 Severity : enhancement
6485 Description: Addition of lconf --service command line option
6486 Details : lconf now accepts a '--service <arg>' option, which is
6487 shorthand for 'lconf --group <arg> --select <arg>=<hostname>'
6489 Severity : enhancement
6491 Description: Failover mode is now the default for OSTs.
6492 Details : By default, OSTs will now run in failover mode. To return to
6493 the old behaviour, add '--failout' to the lmc line for OSTs.
6495 Severity : enhancement
6497 Description: Health checks are now provided for MDS and OSTs
6498 Details : Additional detailed health check information on MSD and OSTs
6499 is now provided through the procfs health_check value.
6502 Frequency : occasional, depends on IO load
6504 Description: Disk fragmentation on the OSTs could eventually cause slowdowns
6505 after numerous create/delete cycles
6506 Details : The ext3 inode allocation policy would not allocate new inodes
6507 very well on the OSTs because there are no new directories
6508 being created. Instead we look for groups with free space if
6509 the parent directories are nearly full.
6514 Description: Network or server problems during mount may cause partially
6515 mounted clients instead of returning an error.
6516 Details : The config llog parsing code may overwrite the error return
6517 code during mount error handling, returning success instead
6523 Description: MDS can fail to allocate large reply buffers
6524 Details : After long uptimes the MDS can fail to allocate large reply
6525 buffers (e.g. zconf client mount config records) due to memory
6526 fragmentation or consumption by the buffer cache. Preallocate
6527 some large reply buffers so that these replies can be sent even
6528 under memory pressure.
6532 Frequency : rare (liblustre)
6533 Description: fsx running with liblustre complained that using truncate() to
6534 extend the file doesn't work. This patch corrects that issue.
6535 Details : This is the liblustre equivalent of the fix for bug 6196. Fixes
6536 ATTR_SIZE and lsm use in llu_setattr_raw.
6540 Frequency : rare, only 2.6 kernels
6541 Description: Unusual file access patterns on the MDS may result in inode
6542 data being lost in very rare circumstances.
6543 Details : Bad interaction between the ea-in-inode patch and the "no-read"
6544 code in the 2.6 kernel caused the inode and/or EA data not to
6545 be read from disk, causing single-file corruption.
6549 Frequency : rare, only 2.6 filesystems using extents
6550 Description: Heavy concurrent write and delete load may cause data corruption.
6551 Details : It was possible under high-load situations to have an extent
6552 metadata block in the block device cache from a just-unlinked
6553 file overwrite a newly-allocated data block. We now unmap any
6554 metadata buffers that alias just-allocated data blocks.
6558 Frequency : filesystems with default stripe_count larger than 77
6559 Description: lconf+mke2fs fail when formatting filesystem with > 77 stripes
6560 Details : lconf specifies an inode size of 4096 bytes when the default
6561 stripe_count is larger than 77. This conflicts with the default
6562 inode density of 1 per 4096 bytes. Allocate smaller inodes in
6563 this case to avoid pinning too much memory for large EAs.
6565 ------------------------------------------------------------------------------
6567 2005-07-07 Cluster File Systems, Inc. <info@clusterfs.com>
6572 Frequency : rare (extremely heavy IO load with hundreds of clients)
6574 Description: Client is evicted, gets IO error writing to file
6575 Details : lock ordering changes for bug 5492 reintroduced bug 3267 and
6576 caused clients to be evicted for AST timeouts. The fixes in
6577 bug 5192 mean we no longer need to have such short AST timeouts
6578 so ldlm_timeout has been increased.
6581 Frequency : occasional during --force or --failover shutdown under load
6582 Bugzilla : 5949, 4834
6583 Description: Server oops/LBUG if stopped with --force or --failover under load
6584 Details : a collection of import/export refcount and cleanup ordering
6585 issues fixed for safer force cleanup
6588 Frequency : only filesystems larger than 120 OSTs
6589 Bugzilla : 5990, 6223
6590 Description: lfs getstripe would oops on a very large filesystem
6591 Details : lov_getconfig used kfree on vmalloc'd memory
6594 Frequency : only filesystems exporting via NFS to Solaris 10 clients
6595 Bugzilla : 6242, 6243
6596 Description: reading from files that had been truncated to a non-zero size
6597 but never opened returned no data
6598 Details : ll_file_read() reads zeros from no-object files to EOF
6603 Description: A bug in MDS/OSS recovery could cause the OSS to fail an assertion
6604 Details : There's little harm in aborting MDS/OSS recovery and letting it
6605 try again, so I removed the LASSERT and return an error instead.
6607 Severity : enhancement
6609 Description: New debugging infrastructure for tracking down data corruption
6610 Details : The I/O checksum code was replaced to: (a) control it at runtime,
6611 (b) cover more of the client-side code path, and (c) try to narrow
6612 down where problems occurred
6616 Bugzilla : 3819, 4364, 4397, 6313
6617 Description: Racing close and eviction MDS could cause assertion in mds_close
6618 Details : It was possible to get multiple mfd references during close and
6619 client eviction, leading to one thread referencing a freed mfd.
6621 Severity: : enhancement
6622 Bugzilla : 3262, 6359
6623 Description: Attempts to reconnect to servers are now more aggressive.
6624 Details : This builds on the enhanced upcall-less recovery that was added
6625 in 1.4.2. When trying to reconnect to servers, clients will
6626 now try each server in the failover group every 10 seconds. By
6627 default, clients would previously try one server every 25 seconds.
6632 Description: After recovery, certain operations trigger a failed
6633 assertion on a client.
6634 Details : Failing over an mds, using lconf -d --failover, while a
6635 client was doing a readdir() call would cause the client to
6636 LBUG after recovery completed and the readdir() was resent.
6638 Severity : enhancement
6640 Description: Default groups are now added by lconf
6641 Details : You can now run lconf --group <servicename> without having to
6642 manually add groups with lmc.
6645 Frequency : occasional
6647 Description: Nodes with an elan id of 0 trigger a failed assertion
6650 Frequency : always when accessing e.g. tty/console device nodes
6652 Description: tty and some other devices nodes cannot be used on lustre
6653 Details : file's private_data field is used by device data and lustre
6654 values in there got lost. New field was added to struct file to
6655 store fs-specific private data.
6658 Frequency : when exporting Lustre via NFS
6660 Description: NFSD failed occasionally when looking up a path component
6661 Details : NFSD is looking up ".." which was broken in ext3 directories
6662 that had grown large enough to become hashed.
6665 Frequency : Clusters with multiple interfaces not on the same subnet
6667 Description: Nodes will repeatedly try to reconnect to an interface which it
6668 cannot reach and report an error to the log.
6669 Details : Extra peer list entries will be created by lconf with some peers
6670 unreachable. lconf now validates the peer before adding it.
6673 Frequency : Only if a default stripe is set on the filesystem root.
6675 Description: Setting a default stripe on the filesystem root prevented the
6676 filesystem from being remounted.
6677 Details : The client was sending extra request flags in the root getattr
6678 request and did not allocate a reply buffer for the dir EA.
6681 Frequency : occasional, higher if lots of files are accessed by one client
6682 Bugzilla : 6159, 6097
6683 Description: Client trips assertion regarding lsm mismatch/magic
6684 Details : While revalidating inodes the VFS looks up inodes with ifind()
6685 and in rare cases can find an inode that is being freed.
6686 The ll_test_inode() code will free the lsm during ifind()
6687 when it finds an existing inode and then the VFS later attaches
6688 this free lsm to a new inode.
6692 Bugzilla : 6422, 7030
6693 Description: MDS deadlock between mkdir and client eviction
6694 Details : Creating a new file via mkdir or mknod (starting a transaction
6695 and getting the ns lock) can deadlock with client eviction
6696 (gets ns lock and trying to finish a synchronous transaction).
6699 Frequency : occasional
6700 Description: While starting a server, the fsfilt_ext3 module could not be
6702 Details : CFS's improved ext3 filesystem is named ldiskfs for 2.6
6703 kernels. Previously, lconf would still use the ext3 name
6704 when trying to load modules. Now, it will correctly use
6705 ext3 on 2.4 and ldiskfs on 2.6.
6707 Severity : enhancement
6708 Description: The default stripe count has been changed to 1
6709 Details : The interpretation of the default stripe count (0, to lfs
6710 or lmc) has been changed to mean striping across a single
6711 OST, rather than all available. For general usage we have
6712 found a stripe count of 1 or 2 works best.
6714 Severity : enhancement
6715 Description: Add support for compiling against Cray portals.
6716 Details : Conditional compiling for some areas that are different
6720 Frequency : occasional
6721 Bugzilla : 6409, 6834
6722 Description: Creating files with an explicit stripe count may lead to
6723 a failed assertion on the MDS
6724 Details : If some OSTs are full or unavailable, creating files may
6725 trigger a failed assertion on the MDS. Now, Lustre will
6726 try to use other servers or return an error to the
6730 Frequency : occasional
6732 Description: Multiple concurrent overlapping read+write on multiple SMP nodes
6733 caused lock timeout during readahead (since 1.4.2).
6734 Details : Processes doing readahead might match a lock that hasn't been
6735 granted yet if there are overlapping and conflicting lock
6736 requests. The readahead process waits on ungranted lock
6737 (original lock is CBPENDING), while OST waits for that process
6738 to cancel CBPENDING read lock and eventually evicts client.
6740 Severity : enhancement
6742 Description: Initial enabling of flock support for clients
6743 Details : Implements fcntl advisory locking and file status functions.
6744 This feature is provided as an optional mount flag (default
6745 off), and is NOT CURRENTLY SUPPORTED. Not all types of record
6746 locking are implemented yet, and those that are are not guaranteed
6747 to be completely correct in production environments.
6748 mount -t lustre -o [flock|noflock] ...
6751 Frequency : occasional
6753 Description: OSTs running 2.4 kernels but with extents enabled might trip an
6754 assertion in the ext3 JBD (journaling) layer.
6755 Details : The b_committed_data struct is protected by the big kernel lock
6756 in 2.4 kernels, serializing journal_commit_transaction() and
6757 ext3_get_block_handle->ext3_new_block->find_next_usable_block()
6758 access to this struct. In 2.6 kernels there is finer grained
6759 locking to improve SMP performance of the JBD layer.
6763 Description: Changes the "SCSI I/O Stats" kernel patch to default to "enabled"
6765 -----------------------------------------------------------------------------
6767 2005-05-05 Cluster File Systems, Inc. <info@clusterfs.com>
6769 NOTE: Lustre 1.4.2 uses an incompatible network protocol than previous
6770 versions of Lustre. Please update all servers and clients to
6771 version 1.4.2 or later at the same time. You must also run
6772 "lconf --write-conf {config}.xml" on the MDS while it is stopped
6773 to update the configuration logs.
6775 - fix for HPUX NFS client breakage when NFS exporting Lustre (5781)
6776 - mdc_enqueue does not need max_mds_easize request buffer on send (5707)
6777 - swab llog records of type '0' so we get proper header size/idx (5861)
6778 - send llog cancel req to DLM cancel portal instead of cb portal (5515)
6779 - fix rename of one directory over another leaking an inode (5953)
6780 - avoid SetPageDirty on 2.6 (5981)
6781 - don't re-add just-being-destroyed locks to the waiting list (5653)
6782 - when creating new directories, inherit the parent's custom
6783 striping settings if present parent (3048)
6784 - flush buffers from cache before direct IO in 2.6 obdfilter (4982)
6785 - don't hold i_size_sem in ll_nopage() and ll_ap_refresh_count (6077)
6786 - don't hold client locks on temporary worklist from l_lru (5666)
6787 - handle IO errors in 2.6 obdfilter bio completion routine (6046)
6788 - automatically evict dead clients (5921)
6789 - Update file size properly in create+truncate+fstat case (6196)
6790 - Do not unhash mountpoint dentries, do not allow removal of
6792 - Avoid lock ordering deadlock issue with write/truncate (6203,5654)
6793 - reserve enough journal credits in fsfilt_start_log for setattr (4554)
6794 - ldlm_enqueue freed-export error path would always LBUG (6149,6184)
6795 - don't reference lr_lvb_data until after we hold lr_lvb_sem (6170)
6796 - don't overwrite last_rcvd if there is a *_client_add() error (6086)
6797 - Correctly handle reads of files with no objects (6243)
6798 - lctl recover will also mark a device active if deactivate used (5933)
6800 - by default create 1 inode per 4kB space on MDS, per 16kB on OSTs
6801 - allow --write-conf on an MDS with different nettype than client (5619)
6802 - don't write config llogs to MDS for mounts not from that MDS (5617)
6803 - lconf should create multiple TCP connections from a client (5201)
6804 - init scripts are now turned off by default; run chkconfig --on
6805 lustre and chkconfig --on lustrefs to use them
6806 - upcalls are no longer needed for clients to recover to failover
6808 - add --abort-recovery option to lconf to abort recovery on device
6810 - add support for an arbitrary number of OSTs (3026)
6811 - Quota support protocol changes.
6812 - forward compatibility changes to wire structs (6007)
6813 - rmmod NALs that might be loaded because of /etc/modules.conf (6133)
6814 - support for mountfsoptions and clientoptions to the Lustre LDAP (5873)
6815 - improved "lustre status" script
6816 - initialize blocksize for non-regular files (6062)
6817 - added --disable-server and --disable-client configure options (5782)
6818 - introduce a lookup cache for lconf to avoid repeated DB scans (6204)
6819 - Vanilla 2.4.29 support
6820 - increase maximum number of obd devices to 520 (6242)
6821 - remove the tcp-zero-copy patch from the suse-2.4 series (5902)
6822 - Quadrics Elan drivers are now included for the RHEL 3 2.4.21 and
6823 SLES 9 2.6.5 kernels
6824 - limit stripes per file to 160 (the maximum EA size) (6093)
6826 2005-03-22 Cluster File Systems, Inc. <info@clusterfs.com>
6829 - don't LASSERT in ll_release on NULL lld with NFS export (4655, 5760)
6830 - hold NS lock when calling handle_ast_error->del_waiting_lock (5746)
6831 - fix setattr mtime regression from lovcleanup merge (4829, 5669)
6832 - workaround for 2.6 crash in ll_unhash_aliases (5687, 5210)
6833 - small ext3 extents cleanups and fixes (5733)
6834 - improved mballoc code, several small races and bugs fixed (5733, 5638)
6835 - kernel version 43 - fix remove_suid bugs in both 2.4 and 2.6 (5695)
6836 - avoid needless client->OST connect, fix handle mismatch (5317)
6837 - fix DLM error path that led to out-of-sync client, long delays (5779)
6838 - support common vfs-enforced mount options (nodev,nosuid,noexec) (5637)
6839 - fix several locking issues related to i_size (5492,5624,5654,5672)
6840 - don't move pending lock onto export if it is already evicted (5683)
6841 - fix kernel oops when creating .foo in unlinked directory (5548)
6842 - fix deadlock in obdfilter statistics vs. object create (5811)
6843 - use time_{before,after} to avoid timer jiffies wrap (5882)
6844 - shutdown --force/--failover stability (3607,3651,4797,5203,4834)
6845 - Do not leak request if server was not able to process it (5154)
6846 - If mds_open unable to find parent dir, make that negative lookup(5154)
6847 - don't create new directories with extent-mapping (5909, 5936)
6849 - fix lustre/lustrefs init scripts for SuSE (patch from Scali, 5702)
6850 - don't hold the pinger_sem in ptlrpc_pinger_sending_on_import
6851 - change obd_increase_kms to obd_adjust_kms (up or down) (5654)
6852 - lconf, lmc search both /usr/lib and /usr/lib64 for Python libs (5800)
6853 - support for RHEL4 kernel on i686 (5773)
6854 - provide error messages when incompatible logs are encountered (5898)
6856 2005-02-18 Cluster File Systems, Inc. <info@clusterfs.com>
6857 * version 1.4.0.10 (1.4.1 release candidate 1)
6859 - don't keep a lock reference when lock is not granted (4238)
6860 - unsafe list practices (rarely) led to infinite eviction loop (4908)
6861 - add per-fs limit of Lustre pages in page cache, avoid OOM (4699)
6862 - drop import inflight refcount on signal_completed_replay error (5255)
6863 - unlock page after async write error during send (3677)
6864 - handle missing objects in filter_preprw_read properly (5265)
6865 - no transno return for symlink open, don't save no-trasno open (3440)
6866 - don't try to complete elan receive that already failed (4012)
6867 - free RPC server reply state on error (5406)
6868 - clean up thread from ptlrpc_start_thread() on error (5160)
6869 - readahead could read extra page into cache that wasn't ejected (5388)
6870 - prevent races in class_attach/setup/cleanup/detach (5260)
6871 - don't dereference de->d_inode after l_dput of de (5458)
6872 - use "int" for stripe value returned from lock_to_stripe (5544)
6873 - mballoc allocation and error-checking fixes in 2.6 (5504)
6874 - block device patches to fix I/O request sizes in 2.6 (5482)
6875 - look up hostnames for IB nals (5602)
6876 - 2.6 changed lock ordering of 2 semaphores, caused deadlock (5654)
6877 - don't start multiple acceptors for the same port (5277)
6878 - fix incorrect LASSERT in mds_getattr_name (5635)
6879 - export a proc file for general "ping" checking (5628)
6880 - fix "lfs check" to not block when the MDS is down (5628)
6882 - service request history (4965)
6883 - put {ll,lov,osc}_async_page structs in a single slab (4699)
6884 - create an "evict_client" /proc entry on OSTs, like the MDS has
6885 - fix mount usage message, return errors per mount(8) (5168)
6886 - change grep [] to grep "[]" in tests so they work in more UMLs
6887 - fix ppc64/x86_64 spec to use %{_libdir} instead of /usr/lib (5389)
6888 - remove ancient LOV_MAGIC_V0 EA support (5047)
6889 - add "disk I/Os in flight" and "I/O req time" stats in obdfilter
6890 - align r/w RPCs to PTLRPC_MAX_BRW_SIZE boundary for performance (3451)
6891 - allow readahead allocations to fail when low on memory (5383)
6892 - mmap locking landed again, after considerable improvement (2828)
6893 - add get_hostaddr() to lustreDB.py for LDAP support (5459)
6895 2004-11-23 Cluster File Systems, Inc. <info@clusterfs.com>
6898 - send OST transaction number in read/write reply to free req (4966)
6899 - don't ASSERT in ptl_send_rpc() if we run out of memory (5119)
6900 - lock /proc/sys/portals/routes internal state, avoiding oops (4827)
6901 - the watchdog thread now runs as interruptible (5246)
6902 - flock/lockf fixes (but it's still disabled, pending 5135)
6903 - don't use EXT3 constants in llite code (5094)
6904 - memory shortage at startup could cause assertion (5176)
6906 - reorganization of lov code
6907 - single portals codebase
6909 - add extents/mballoc support (5025)
6910 - direct I/O reads in the obdfilter (4048)
6911 - kernel patches from LNXI for 2.6 (bluesmoke, perfctr, mtd, kexec)
6913 tbd Cluster File Systems, Inc. <info@clusterfs.com>
6916 - send OST transaction number in read/write reply to free req (4966)
6917 - don't ASSERT in ptl_send_rpc() if we run out of memory (5119)
6918 - lock /proc/sys/portals/routes internal state, avoiding oops (4827)
6919 - the watchdog thread now runs as interruptible (5246)
6920 - handle missing objects in filter_preprw_read properly (5265)
6921 - unsafe list practices (rarely) led to infinite eviction loop (4908)
6922 - drop import inflight refcount on signal_completed_replay error (5255)
6923 - unlock page after async write error during send (3677)
6924 - return original error code on reconstructed replies (3761)
6925 - no transno return for symlink open, don't save no-trasno open (3440)
6927 - add pid to ldlm debugging output (4922)
6928 - bump the watchdog timeouts -- we can't handle 30sec yet
6929 - extra debugging for orphan dentry/inode bug (5259)
6931 2004-11-16 Cluster File Systems, Inc. <info@clusterfs.com>
6934 - fix TCP_NODELAY bug, which caused extreme perf regression (5134)
6935 - allocate qswnal tx descriptors singly to avoid fragmentation (4504)
6936 - don't LBUG on obdo_alloc() failure, use OBD_SLAB_ALLOC() (4800)
6937 - fix NULL dereference in /proc/sys/portals/routes (4827)
6938 - allow failed mdc_close() operations to be interrupted (4561)
6939 - stop precreate on OST before MDS would time out on it (4778)
6940 - don't send partial-page writes before EOF from client (4410)
6941 - discard client grant for sub-page writes on large-page clients (4520)
6942 - don't free dentries not owned by NFS code, check generation (4806)
6943 - fix lsm leak if mds_create_objects() fails (4801)
6944 - limit debug_daemon file size, always print CERROR messages (4789)
6945 - use transno after validating reply (3892)
6946 - process timed out requests if import state changes (3754)
6947 - update mtime on OST during writes, return in glimpse (4829)
6948 - add mkfsoptions to LDAP (4679)
6949 - use ->max_readahead method instead of zapping global ra (5039)
6950 - don't interrupt __l_wait_event() during strace
6952 - add software watchdogs to catch hung threads quickly (4941)
6953 - make lustrefs init script start after nfs is mounted
6954 - fix CWARN/ERROR duplication (4930)
6955 - return async write errors to application if possible (2248)
6956 - add /proc/sys/portal/memused (bytes allocated by PORTALS_ALLOC)
6957 - print NAL number in %x format (4645)
6958 - update barely-supported suse-2.4.21-171 series (4842)
6959 - support for sles 9 %post scripts
6960 - support for building 2.6 kernel-source packages
6961 - support for sles km_* packages
6963 2004-10-07 Cluster File Systems, Inc. <info@clusterfs.com>
6966 - ignore -ENOENT errors in osc_destroy (3639)
6967 - notify osc create thread that OSC is being cleaned up (4600)
6968 - add nettype argument for llmount in #5d in conf-sanity.sh (3936)
6969 - reconstruct ost_handle() like mds_handle() (4657)
6970 - create a new thread to do import eviction to avoid deadlock (3969)
6971 - let lconf resolve symlinked-to devices (4629)
6972 - don't unlink "objects" from directory with default EA (4554)
6973 - hold socknal file ref over connect in case target is down (4394)
6974 - allow more than 32000 subdirectories in a single directory (3244)
6975 - fix blocks count for O_DIRECT writes (3751)
6976 - OST returns ENOSPC from object create when no space left (4539)
6977 - don't send truncate RPC if file size isn't changing (4410)
6978 - limit OSC precreate to 1/2 of value OST considers bogus (4778)
6979 - bind to privileged port in socknal and tcpnal (3689)
6981 - rate limit CERROR/CWARN console message to avoid overload (4519)
6982 - GETFILEINFO dir ioctl returns LOV EA + MDS stat in 1 call (3327)
6983 - basic mmap support (3918)
6984 - kernel patch series update from b1_4 (4711)
6986 2004-09-16 Cluster File Systems, Inc. <info@clusterfs.com>
6989 - avoid crash during MDS cleanup with OST shut down (2775)
6990 - fix loi_list_lock/oig_lock inversion on interrupted IO (4136)
6991 - don't use bad inodes on the MDS (3744)
6992 - dynamic object preallocation to improve recovery speed (4236)
6993 - don't hold spinlock over lock dumping or change debug flags (4401)
6994 - don't zero obd_dev when it is force cleaned (3651)
6995 - print grants to console if they go negative (4431)
6996 - "lctl deactivate" will stop automatic recovery attempts (3406)
6997 - look for existing locks in ldlm_handle_enqueue() (3764)
6998 - don't resolve lock handle twice in recovery avoiding race (4401)
6999 - revalidate should check working dir is a directory (4134)
7001 - don't always mark "slow" obdfilter messages as errors (4418)
7003 2004-08-24 Cluster File Systems, Inc. <info@clusterfs.com>
7006 - don't close LustreDB during write_conf until it is done (3860)
7007 - fix typo in lconf for_each_profile (3821)
7008 - allow dumping logs from multiple threads at one time (3820)
7009 - don't allow multiple threads in OSC recovery (3812)
7010 - fix debug_size parameters (3864)
7011 - fix mds_postrecov to initialize import for llog ctxt (3121)
7012 - replace config semaphore with spinlock (3306)
7013 - be sure to send a reply for a CANCEL rpc with bad export (3863)
7014 - don't allow enqueue to complete on a destroyed export (3822)
7015 - down write_lock before checking llog header bitmap (3825)
7016 - recover from lock replay timeout (3764)
7017 - up llog sem before sending rpc (3652)
7018 - reduce ns lock hold times when setting kms (3267)
7019 - change a dlm LBUG to LASSERTF, to maybe learn something (4228)
7020 - fix NULL deref and obd_dev leak on setup error (3312)
7021 - replace some LBUG about llog ops with error handling (3841)
7022 - don't match INVALID dentries from d_lookup and spin (3784)
7023 - hold dcache_lock while marking dentries INVALID and hashing (4255)
7024 - fix invalid assertion in ptlrpc_set_wait (3880)
7026 - add libwrap support for the TCP acceptor (3996)
7027 - add /proc/sys/portals/routes for non-root route listing (3994)
7028 - allow setting MDS UUID in .xml (2580)
7029 - print the stack of a process that LBUGs (4228)
7031 2004-07-14 Cluster File Systems, Inc. <info@clusterfs.com>
7034 - don't cleanup request in ll_file_open() on failed MDS open (3430)
7035 - make sure to unset replay flag from failed open requests (3440)
7036 - if default stripe count is 0, use OST count for inode size (3636)
7037 - update parent mtime/ctime on client for create/unlink (2611)
7038 - drop dentry ref in ext3_add_link from open_connect_dentry (3266)
7039 - free recovery state on server during a forced cleanup (3571)
7040 - unregister_reply for resent reqs (3063)
7041 - loop back devices mounting and status check on 2.6 (3563)
7042 - fix resource-creation race that can provoke i_size == 0 (3513)
7043 - don't try to use bad inodes returned from MDS/OST fs lookup (3688)
7044 - more debugging for page-accounting assertion (3746)
7045 - return -ENOENT instead of asserting if ost getattr+unlink race (3558)
7046 - avoid deadlock after precreation failure (3758)
7047 - fix race and lock order deadlock in orphan handling (3450, 3750)
7048 - add validity checks when grabbing inodes from l_ast_data (3599)
7050 - add /proc/.../recovery_status to obdfilter (3428)
7051 - lightweight CDEBUG infrastructure, debug daemon (3668)
7052 - change default OSC RPC parameters to be better on small clusters
7053 - turn off OST read cache for files smaller than 32MB
7054 - install man pages and include them in rpms (3100)
7055 - add new init script for (un)mounting lustre filesystems (2593)
7056 - run chkconfig in %post for init scripts (3701)
7057 - drop scimac NAL (unmaintained)
7059 2004-06-17 Cluster File Systems, Inc. <info@clusterfs.com>
7062 - clean kiobufs before and after use (3485)
7063 - strip trailing '/'s before comparing paths with /proc/mounts (3486)
7064 - remove assertions to work around "in-flight rpcs" recovery bug (3063)
7065 - change init script to fail more clearly if not run as root (1528)
7066 - allow clients to reconnect during replay (1742)
7067 - fix ns_lock/i_sem lock ordering deadlock for kms update (3477)
7068 - don't do DNS lookups on NIDs too small for IP addresses (3442)
7069 - re-awaken ptlrpcd if new requests arrive during check_set (3554)
7070 - fix cond_resched (3554)
7071 - only evict unfinished clients after recovery (3515)
7072 - allow bulk resend, prevent data loss (3570)
7073 - dynamic ptlrpc request buffer allocation (2102)
7074 - don't allow unlinking open directory if it isn't empty (2904)
7075 - set MDS/OST threads to umask 0 to not clobber client modes (3359)
7076 - remove extraneous obd dereference causing LASSERT failure (3334)
7077 - don't use get_cycles() when creating temp. files on the mds (3156)
7078 - hold i_sem when setting i_size in ll_extent_lock() (3564)
7079 - handle EEXIST for set-stripe, set proper directory name (3336)
7081 - servers can dump a log evicting a client - lustre.dump_on_timeout=1
7082 - fix ksocknal_fmb_callback() error messages (2918)
7084 2004-05-27 Cluster File Systems, Inc. <info@clusterfs.com>
7087 - don't copy lvb into (possibly NULL) reply on error (2983)
7088 - don't deref dentry after dput, don't free lvb on error (2922)
7089 - use the kms to determine writeback rpc length (2947)
7090 - increment oti_logcookies when osc is inactive (2948)
7091 - update client's i_blocks count via lvb messages (2543)
7092 - handle intent open/close of special files properly (1557)
7093 - mount MDS with errors=remount-ro, like obdfilter (2009)
7094 - initialize lock handle to avoid ASSERT on error cleanup (3057)
7095 - don't use cancelling-locks' kms values (2947)
7096 - use highest lock extent for kms, not last one (2925)
7097 - don't dereference ERR_PTR() dentry in error handling path (3107)
7098 - fix thread race in portals_debug_dumplog() (3122)
7099 - create lprocfs device entries at setup instead of at attach (1519)
7100 - common AST error handler, don't evict client on completion race (3145)
7101 - zero nameidata in detach_mnt in 2.6 (3118)
7102 - verify d_inode after revalidate_special is valid in 2.6 (3116)
7103 - use lustre_put_super() to handle zconf unmounts in 2.6 (3064)
7104 - initialize RPC timeout timer earlier for 2.6 (3219)
7105 - don't dereference NULL reply buffer if mdc_close was never sent (2410)
7106 - print nal/nid for unknown nid (3258)
7107 - additional checks for oscc recovery before doing precreate (3284)
7108 - fix ll_extent_lock() error return code for 64-bit systems (3043)
7109 - don't crash in mdc_close for bad permissions on open (3285)
7110 - zero i_rdev for non-device files (3147)
7111 - clear page->private before handing to FS, better assertion (3119)
7112 - tune the read pipeline (3236)
7113 - fix incorrect decref of invalidated dentry (2350)
7114 - provide read-ahead stats and refine rpc in flight stats (3328)
7115 - don't hold journal transaction open across create RPC (3313)
7116 - update atime on MDS at close time (3265)
7117 - close LDAP connection when recovering to avoid server load (3315)
7118 - update iopen-2.6 patch with fixes from 2399,2517,2904 (3301)
7119 - don't leak open file on MDS after open resend (3325)
7120 - serialize filter_precreate and filter_destroy_precreated (3329)
7121 - loop device shouldn't call sync_dev() for nul device (3092)
7122 - clear page cache after eviction (2766)
7123 - resynchronize MDS->OST in background (2824)
7124 - refuse to mount the same filesystem twice on same mountpoint (3394)
7125 - allow llmount to create routes for mounting behind routers (3320)
7126 - push lock cancellation to blocking thread for glimpse ASTs (3409)
7127 - don't call osc_set_data_with_check() for TEST_LOCK matches (3159)
7128 - fix rare problem with rename on htree directories (3417)
7130 - allow default OST striping configuration per directory (1414)
7131 - fix compilation for qswnal for 2.6 kernels (3125)
7132 - increase maximum number of MDS request buffers for large systems
7133 - change liblustreapi to be useful for external progs like lfsck (3098)
7134 - increase local configuration timeout for slow disks (3353)
7135 - allow configuring ldlm AST timeout - lustre.ldlm_timeout=<seconds>
7137 2004-03-22 Cluster File Systems, Inc. <info@clusterfs.com>
7140 - fixes for glimpse AST timeouts / incorrectly 0-sized files (2818)
7141 - don't overwrite extent policy data in reply if lock was blocked (2901)
7142 - drop filter export grants atomically with removal from device (2663)
7143 - del obd_self_export from work_list in class_disconnect_exports (2908)
7144 - don't LBUG if MDS recovery times out during orphan cleanup (2530)
7145 - swab reply message in mdc_close, other PPC fixes (2464)
7146 - fix destroying of named logs (2325)
7147 - overwrite old logs when running lconf --write_conf (2264)
7148 - bump LLOG_CHUNKSIZE to 8k to allow for larger clusters (2306)
7149 - fix race in target_handle_connect (2898)
7150 - mds_reint_create() should take same inode create lock (2926)
7151 - correct journal credits calculated for CANCEL_UNLINK_LOG (2931)
7152 - don't close files for self_export to avoid uninitialized obd (2936)
7153 - allow MDS with the same name as client node (2939)
7154 - hold dentry reference for closed log files for unlink (2325)
7155 - reserve space for all logs during transactions (2059)
7156 - don't evict page beyond end of stripe extent (2925)
7157 - don't oops on a deleted current working directory (2399)
7158 - handle hard links to targets without a parent properly (2517)
7159 - don't dereference NULL lock when racing during eviction (2867)
7160 - don't grow lock extents when lots of conflicting locks (2919)
7162 2004-03-04 Cluster File Systems, Inc. <info@clusterfs.com>
7165 - account for cache space usage on clients to avoid data loss (974)
7166 - lfsck support in lustre kernel code (2349)
7167 - reduce journal credits needed for BRW writes (2370)
7168 - orphan handling to avoid losing space on client/server crashes
7169 - ptlrpcd can be blocked, stopping ALL progress (2477)
7170 - use lock value blocks to assist in proper KMS, faster stat (1021)
7171 - takes i_sem instead of DLM locks internally on obdfilter (2720)
7172 - recovery for initial connections (2355)
7173 - fixes for mds_cleanup_orphans (1934)
7174 - abort_recovery crashes MDS in b_eq (mds_unlink_orphan) (2584)
7175 - block all file creations until orphan recovery completes (1901)
7176 - client remove rq_connection from request struct (2423)
7177 - conf-sanity test_5, proper cleanup in umount log not availale (2640)
7178 - recovery timer race (2670)
7179 - mdc_close recovey bug (2532)
7180 - ptlrpc cleanup bug (2710)
7181 - mds timeout on local locks (2588)
7182 - namespace lock held during RPCs (2431)
7183 - handle interrupted sync write properly (2503)
7184 - don't try to handle a message that hasn't been replied to (2699)
7185 - client assert failure during cleanup after abort recovery (2701)
7186 - leak mdc device after failed mount (2712)
7187 - ptlrpc_check_set allows timedout requests to complete (2714)
7188 - wait for inflight reqs when ptlrpcd finishes (2710)
7189 - make sure unregistered services are removed from the srv_list
7190 - reset bulk XID's when resending them (caught by 1138 test)
7191 - unregister_bulk after timeout
7192 - fix lconf error (2694)
7193 - handle write after unfinished setstripe, stripe-only getstripe (2388)
7194 - readahead locks pages, leaves pending causing memory pressure (2673)
7195 - increase OST request buffers to 4096 on large machines (2729)
7196 - fix up permission of existing directories in simple_mkdir (2661)
7197 - init deleted item, add assertions ptlrpc_abort_inflight() (2725)
7198 - don't assign transno to errored transactions (2742)
7199 - don't delete objects on OST if given a bogus objid from MDS (2751)
7200 - handle large client PAGE_SIZE readdir on small PAGE_SIZE MDS (2777)
7201 - if rq_no_resend, then timeout request after recovery (2432)
7202 - fix MDS llog_logid record size, 64-bit array alignment (2733)
7203 - don't call usermode_helper from ptlrpcd, DEFAULT upcall (2773)
7204 - put magic in mount.lustre data, check for bad/NULL mount data (2529)
7205 - MDS recovery shouldn't delete objects that it has given out (2730)
7206 - if enqueue arrives after completion, don't clobber LVB (2819)
7207 - don't unlock pages twice when trigger_group_io returns error (2814)
7208 - don't deref NULL rq_repmsg if ldlm_handle_enqueue failed (2822)
7209 - don't write pages to disk if there was an error (1450)
7210 - don't ping imports that have recovery disabled (2676)
7211 - take buffered bytes into account when balancing socknal conn (2817)
7212 - hold a DLM lock over readdir always, use truncate_inode_pages (2706)
7213 - reconnect unlink llog connection after MDS reconnects to OST (2816)
7214 - remove little-endian swabbing of llog records (1987)
7215 - set/limit i_blksize to LL_MAX_BLKSIZE on client (2884)
7216 - retry reposting request buffers if they fail (1191)
7217 - grow extent at grant time to avoid granting a revoked lock (2809)
7218 - lock revoke doesn't evict page if covered by a second lock (2765)
7219 - disable VM readahead to avoid reading outside lock extents (2805)
7221 - return LL_SUPER_MAGIC from statfs for the filesystem type (1972)
7222 - updated kernel patches for hp-2.4.20 kernel (2681)
7224 2004-02-07 Cluster File Systems, Inc. <info@clusterfs.com>
7227 - fix truncated write corruption (2366)
7228 - fix for failed assertion in iopen_connect_dentry (1792,2517)
7230 - don't flag the ptlrpcd thread with PF_MEMALLOC (2636)
7231 - ensure len(uuid) < 37 in lmc (1171)
7232 - fix ia64 OOPS in llog_test (2255)
7233 - zero end of page at obdfilter for partial page writes (2648)
7234 - don't leave stale dentries around after renames (bug 2428)
7235 - fix timeouts when evicting a client with a single lock held (2642)
7236 - set deadline for the initial HELLO message to drain (2634)
7237 - print out dotted-quad IP addresses in the socknal (2302)
7239 - additional debugging for MDS client eviction problem (2443)
7240 - fix mkfsoptions support for osts (2603, 2604)
7242 2004-01-27 Cluster File Systems, Inc. <info@clusterfs.com>
7245 - add series for the vanilla 2.6.0 kernel
7246 - add series for the vanilla 2.4.24 kernel
7247 - add series for a cray x86/64 UL kernel drop
7248 - fix xattr patches for the vanilla 2.4.19 series
7250 - generate true UUIDs in lmc (1171)
7251 - have portals stack dumping break in UML (2466)
7252 - avoid bad dchild deref; avoid inum lock w/o creation (2362)
7253 - allocate with _NOFS in ldlm to avoid deadlock (1933)
7254 - wake callback waiting threads on client eviction (2460)
7255 - Add --ptldebug and --subsystem to lmc (1719)
7256 - update assertion to allow safe interrupt allocation
7257 - set rq_no_resend for cancel requests (2432)
7258 - recalculate ptlrpcd timeout after resend (2494)
7259 - call vfs_rmdir when removing pending directories (2368)
7260 - fix renaming a file to itself (2429)
7261 - lmc creates a default one-stripe lov (2454)
7262 - expand procfs space to handle large clusters (2326)
7263 - increase UML stack to avoid overflow
7264 - update lconf's list of debug and subsystem masks
7265 - fix lfs find --obd (2510)
7266 - /proc tunable for disabling filter read caching (2591)
7267 - stop rpm packages from altering slapd.conf (2301)
7268 - disable nagle in the socknal under 0conf (2578)
7269 - choose mds inode size based on stripe count (2572)
7270 - fix kernel-source rpm problems (2516)
7272 - add --disable-doc to avoid pdf generation (2421)
7273 - update documentation, tests, type-os, comments
7274 - avoid format warnings on ia64
7275 - remove the TOE NAL
7276 - tiny code cleanups by removing unused fields
7278 2004-01-07 Cluster File Systems, Inc. <info@clusterfs.com>
7281 - fix obvious semaphore misuse in as-yet-unused setattr path (2348)
7282 - remove the most blatant lies from BUILDING file (2371)
7283 - change default debug level to reasonable production setting
7284 - reduce client side cache size to reduce cache flush time
7285 - reduce max RPCs in flight to avoid unnecessary file fragmentation
7286 - make TCP zerocopy and pinger support enabled by default (2476)
7287 - sync writes completed after process exits caused crashes (2319)
7288 - maintain correct mount count on the MDS (2356)
7289 - backout 1557, because 2316 wasn't really fixed
7290 - better file I/O statistics gathering in /proc
7291 - don't take unnecessary, deadlock-inducing bug in readpage (2383)
7292 - another kernel patch to fix zero-copy TCP function export
7293 - don't take duplicate lock when processing re-sent getattr (2420)
7294 - lctl uses obd_self_export instead of creating new conn (2353)
7295 - MDS/OST recovery case which requires object creation asserted (2425)
7296 - move lfs from /usr/sbin to /usr/bin in packages
7297 - fix race between mds_client_add and mds_client_free (2417)
7298 - use kmalloc instead of slabs in portals (2430)
7299 - don't create duplicate records when a failover MDS is present (2442)
7300 - remove unnecessary mount age check (2332)
7301 - don't remove directory inodes from locks prematurely (2451)
7302 - don't break if MDS service name is the same as hostname (2103)
7303 - fix races in client write RPC generation when cache full (2482)
7305 2003-12-13 Cluster File Systems, Inc. <info@clusterfs.com>
7308 - remove now-unused request->rq_obd (278)
7309 - if an allocation fails, print out how much memory we've used (1933)
7310 - use PORTAL_SLAB_ALLOC for structures, to get GFP_MEMALLOC (1933)
7311 - add the "configurable stack size" patch to most series files (1256)
7312 - ability to write large log records, for 100+ OST configs (2306)
7313 - fix NULL deref when filter_prep fails (2314)
7314 - fix operator precedence error in filter_sync
7315 - dynamic allocation of socknal TX descriptors (2315)
7316 - fix a missed case in the GFP_MEMALLOC patch, can cause deadlock (2310)
7317 - fix gcc 2.96 compilation problem in xattr kernel patch (2294)
7318 - ensure that CWARN messages in Portals always get to the syslog
7319 - __init/__exit are not for prototype decls (ldlm_init/exit)
7320 - x86-64 compile warning fixes
7321 - fix gateway LMC keyword conflict (2318)
7322 - fix MDS lock inversions in getattr/reint paths (1844)
7323 - fix a rare lock re-ordering bug, which caused deadlock (2322)
7324 - fix i_sem/journal inversion in fsfilt_ext3_write_record (2306)
7325 - DLM race condition prevented some lock evictions (2328)
7326 - ENOMEM detection and retry on socknal sends (2230)
7327 - use GFP_NOFS throughout Lustre, to combat ENOMEM (2230)
7328 - move osc_rpcd into ptlrpc, for use in MDC and others (2329)
7329 - protect MDS inode fsdata with stronger locking; fixes assertion (2313)
7330 - better error messages when a client is rejected during recovery (1505)
7331 - avoid cancelling locks which were never granted, after failure (2330)
7332 - fix i_sem/journal inversion in mds_client_add (2333)
7333 - fix truncate/getattr lock cycle deadlock (2334)
7334 - use rpcd to send close; allows resend after timeout, avoid leak (1897)
7335 - fix two rare exit paths which could leak an l_lock() ref (2321)
7336 - fencepost error in MDS/OST orphan recovery (2226)
7337 - make log record alignment 8 bytes (1988)
7338 - lstripe now fails when requested offset > ost_count (2237)
7339 - ensure that all kernel series have a complete list.h (1607)
7340 - fix crashes in special-file operations (2316)
7341 - lctl create/brw OID mismatch, caused by obsolete filter loop (2339)
7343 - allow configurable automake binary, for testing new versions
7344 - small update to the lfs documentation
7346 2003-12-03 Cluster File Systems, Inc. <info@clusterfs.com>
7348 * fix negative export reference count in fsfilt_sync (2312)
7350 2003-12-01 Cluster File Systems, Inc. <info@clusterfs.com>
7351 * release candidate 0.9.1
7353 - orphans are moved into the PENDING directory for possible recovery
7354 - replayed opens now open by fid for orphan/rename safety (1042)
7355 - last close of an orphan inode generates a transno (683)
7356 - chdir() and mount() now pin the directory entry (1020)
7357 - avoid CERROR in normal ll_setattr_raw() error case (1500)
7358 - discard very old requests without processing them (1502)
7359 - remove some common, well-understood CERRORs (1505)
7360 - require O_DIRECT I/O to be page-sized to workaround IA64 crash (1609)
7361 - clear "grant" flags in OST replies until OST grant code lands (1644)
7362 - fix read performance by not clobbering i_blksize on client (1598)
7363 - fix __ldlm_handle2lock oops by not dereferencing lock after PUT (1625)
7364 - make LRU size a /proc tunable, clears locks when reduced (707)
7365 - fix some lprocfs rot that prevented ptlbd from loading (1732)
7366 - server locks take references on exports now (1558)
7367 - build fixes for 2.4.20-rh trees (1663)
7368 - return an error from lov_create if all OSCs are inactive (1751)
7369 - fix import levels when a reconnect happens without a timeout (1597)
7370 - exit early from mds_open if we get a lookup error (1749)
7371 - partial page read at EOF wouldn't wait for disk before sending (1642)
7372 - avoid NULL deref in obdfilter when reading page past EOF (1592)
7373 - avoid LASSERT in ll_intent_lock if server failed very early (1090)
7374 - fix LBUG in ll_it_open_error with rc = -2 (1861)
7375 - write/truncate lock inversion (1639)
7376 - Don't auto-load obdclass, portals modules during cleanup (1495)
7377 - fix timestamps from jumping to "now" (1763)
7378 - extra journal assertions (1648)
7379 - add an extra multiunlink test (1771)
7380 - fix read_record/write_record API (1776)
7381 - fix leak of offset_extent, possible incorrect i_size later (1772)
7382 - fix lasserts in mis-matched transnos during open-unlink testing (1541)
7383 - Debugging for the kqswnal_get_idle_tx problems (1820)
7384 - Allow recovery to be attempted multiple times (1536)
7385 - Write out MDS last_rcvd file after it is first created (1600)
7386 - Fix tx_descriptor leak in failed transmit situations (1827)
7387 - ext3 journaling fixes for assertion failure after IO error (1871)
7388 - class_export_put() on freed export after completion AST error (1896)
7389 - Fix revalidate looping in VFS (1322)
7390 - Don't access a freed export during MDS_REINT timeout (1521)
7391 - Add open-unlink recovery support on the MDS (1673,1764)
7392 - Return an error if no MDS data was read from last_rcvd (1946)
7393 - Fix for lookup "." or ".." crash on error (1932,1931,1935)
7394 - Don't setup a disk device that doesn't match exported UUID (317)
7395 - Reduce bulk RPC timeout to avoid cascading client/OST failures (1845)
7396 - avoid committing NULL handle in force close
7397 - local.sh is now a one-stripe LOV configuration
7398 - POSIX utime.4 -EPERM on FIFO not owned by user (56)
7399 - fix ext3 htree duplicate directory entry corruption (1516)
7400 - POSIX creat.13, fstat.1, open.18, stat.3 new file atime/mtime (2020)
7401 - update to new LOV EA format (2097)
7402 - interoperability for different PAGE_SIZE/wordsize (686,1821,1343,2042)
7404 2003-06-15 Phil Schwan <phil@clusterfs.com>
7407 - imports and exports cleanup too early, need refcounts (349, 879, 1045)
7408 - per-import/export recovery handling (958, 931, 959)
7409 - multiple last-rcvd slots, for serving multiple FSes (949)
7410 - connections are again shared between multiple imp/exports (963, 964)
7411 - "umount -f" would hang if any requests needed to be sent (393, 978)
7412 - avoid pinning large req buffer by copying for queued messages (989)
7413 - add "uuid" to "lctl device" command to help upcalls (991)
7414 - "open" RPCs with transnos would confuse recovery counters (1037)
7415 - do proper endian conversion of all wire messages (288, 340, 891)
7416 - remove OST bulk get LBUGs, fix ost_brw_write cleanup (1126)
7417 - call waiting locks callback from LDLM recovery thread (1127, 1151)
7418 - fix ptlrpc_connection leak in target_handle_connect (1174)
7419 - fix import refcounting bug in OST and MDS cleanup (1134)
7420 - if an invalid-at-open-time OSC returned before close(), LBUG (1150)
7421 - fix very unlikely obd_types race condition (501)
7422 - remove osc_open hack for echo_client (1187)
7423 - we leaked exports/dlmimps for forcibly disconnected clients (1143)
7424 - a failure in read_inode2 leads to deadlock (1139)
7425 - cancel ack-locks as soon as transaction is committed (1072)
7426 - fix major leaks and crashes in the bulk I/O path (937, 1057)
7427 - make sure to commitrw after any preprw to avoid deadlock (1162)
7428 - failing to execute a file in a lustre FS would lock inode (1203)
7429 - small DEBUG_REQ fix to avoid dereferencing a NULL (1227)
7430 - don't ASSERT while cleaning up an incompletely-setup obd (1248)
7431 - obd_uuid2tgt would walk off the end of the list (1255)
7432 - on IA64 the osc would give portals incorrect bulk size (1258)
7433 - fix debug daemon ioctl interface; allows daemon on ia64 (1274)
7434 - fix lock inversion caused by new llite matching code (1282)
7435 - limit the number of dirty pages on a client to 10MB (1286)
7436 - timed out locks were not being corrected cancelled (1289)
7437 - fix O_DIRECT above 4GB on IA-32 (1292)
7438 * major user-visible changes
7439 - fail out/fail over policy now controlled by the upcall (993)
7441 - add OBD_PING to check server availability and failure (954)
7442 - lustre messages are now sent in sending host order (288, 340, 891)
7443 - add eadatalen to MDS getattr reply (340)
7444 - OST read replies may contain second buffer, with per-page status (593)
7446 2003-03-11 Phil Schwan <phil@clusterfs.com>
7449 - LDLM_DEBUG macro fix, for gcc 3.2 (850)
7450 - failed open()s could cause deadlock; fixed (867, 869)
7451 - stop cancelling OST locks when files are closed (481)
7452 - overlapping XID spaces caused network corruption (851, 853)
7453 - fix unsafe fsfilt counter arithmetic; change to atomic_t
7454 - setattr_raw added, to do single-RPC, server-side setattrs
7455 - lmc/lconf syntax change for OST UUIDs
7456 - fix crashy race condition between ptlrpc_free_req and osc_close
7457 - don't use request in mdc_enqueue if we hit a timeout (889)
7458 - don't set the inode i_size for regular files from the MDS (896)
7459 - handle out of order completion AST (842)
7460 - don't LBUG if a lock request times out after receiving AST (913)
7461 - avoid d_rehash race in ll_find_alias by rehashing inside dcache_lock
7462 - if a bad lock AST arrives, send an error instead of dropping entirely
7463 - return 0 from revalidate2 if ll_intent_lock returns -EINTR (912)
7464 - fix leak in bulk IO when only partially completed (899, 900, 926)
7465 - fix O_DIRECT for ia64 (55)
7466 - (almost) eliminate Lustre-kernel-thread effects on load average (722)
7467 - C-z after timeout could hang a process forever; fixed (977)
7469 - client-side I/O cache (678, 924, 929, 941, 970)
7471 - READPAGE and SETATTRs which don't take server-side locks get
7474 2003-02-11 Phil Schwan <phil@clusterfs.com>
7477 - Fix ldlm_lock_match on the MDS to avoid matching remote locks (592)
7478 - Fix fsfilt_extN_readpage() to read a full page of directory
7479 entries, or fake the remainder if PAGE_SIZE != blocksize (500)
7480 - Avoid extra mdc_getattr() in ll_intent_lock when possible (534, 604)
7481 - Fix imbalanced LOV object allocation and out-of-bound access (469)
7482 - Most intent operations were removed, in favour of a new RPC mode
7483 that does a single RPC to the server and bypasses most of the VFS
7484 - All LDLM resource ID arrays were removed in favour of ldlm_res_id
7485 - Aggressively cancel local locks on DLM servers
7486 - mds_reint_unlink sends EA to the client if it's the last nlink.
7487 client uses that EA to unlink OST objects.
7488 - mds_reint_{rename,unlink,link} were rewritten to take ordered locks
7489 - recursive symlinks were fixed (439)
7490 - fixed NULL deref in DEBUG_REQ
7491 - filter_update_lastobjid no longer calls sync, which annoyed extN
7492 - fixed multi-client small-writes to a single file problem (445)
7493 - fixed mtime updates during file writes (607)
7494 - fixed vector writes on obdfilter causing problems when ENOSPC (670)
7495 - fixed bug in obd_brw_read/write() (under guise of testing 367)
7496 - fixed Linux OST size reporting problem (444, 656)
7497 - OST now updates object mtime with writes or setattr (607, 619)
7498 - client verifies file size before zeroing page past EOF (445)
7499 - OST now writes last allocated objid to disk with allocation (108)
7500 - LOV on echo now works (409)
7502 - mds_reint_unlink sends a new buffer, with the EA included. this
7503 buffer is only valid if body->valid & OBD_MD_FLEASIZE, which is only
7504 set if a regular file was being unlinked, and it was the last link
7505 - use PtlGet from the target for bulk writes (315)
7506 - OST now updates object mtime with writes or setattr (607, 619)
7507 - LDLM now has a grant-time callback to revalidate locked items, if
7509 - Many MDS operations were reorganized to combat race conditions
7511 - Merge b_intel branch (updated lprocfs code) - now at /proc/fs/lustre
7512 - configure check to avoid gcc version 2.96 20000731-2.96-98) (606)
7514 2003-01-06 Andreas Dilger <adilger@clusterfs.com>
7517 - Fully reactivate OST imports after reconnection (512, others)
7518 - Make sure client sees our -ENOTCONN from mds_handle (513 - partial)
7519 - More graceful error handling for truncating on dead OST (515)
7520 - Don't error out unless we're actually accessing dead stripes (474)
7521 - Fix garbage sizes when stripes are missing (410)
7522 - LRU counters were broken, causing constant lock purge (433, 432)
7523 - garbage on read from stripes with failed OSTs (441)
7524 - mark OSCs as active before reconnecting during recovery (438)
7525 - lov_enqueue and lov_cancel need to handle inactive OSTs (403)
7526 - lfind did not preserve OST order in output (443)
7527 - symlinks cause hung clients, incorrect data (439)
7528 - stop dereferencing request after dropping refcount (457)
7529 - don't LASSERT(spin_is_locked) on non-SMP (455)
7530 - fixes for many rename() bugs
7531 - fstat didn't correctly synchronize attributes (399)
7532 - server must handle lock cancellation during blocking AST prep (487)
7533 - bulk descriptors were free()d too soon (511)
7534 - fix paths in lconf, which would load incorrect modules (451, 507)
7535 - fix confusing lconf 'host not found' error message (386)
7536 - fix lock order deadlock on OST (O/R i_sem before journal ops, 478)
7537 - fix race condition in mdc_blocking_ast() for inode access (526)
7538 - fix lov_unpackmd() unpacking wrong number of stripes (537)
7539 - fix lov_set_osc_active() marking wrong OSC inactive (440)
7540 - fix bad lstripe lov_unpackmd() assertion (fix layering too) (527)
7541 - fix multiple writes of stripe MD to MDS (358, maybe 519)
7542 - fix lstripe in several ways (kernel side) (527)
7543 - fix request leak in ldlm_cli_enqueue (262)
7544 - incorrect OSC was marked inactive after OST failure
7545 - call mds_fs_cleanup before unmounting filesystem (524)
7546 - fix races between taking ns_lock and ldlm_lock_change_resource
7547 - fix races updating LOV export open file list
7548 - fix lov_enqueue error path, avoid decref-ing bad lock handle
7549 - fix recovery NULL deref in ldlm_cli_cancel_unused
7550 - fix some DLM races by using new hash table for lock handles (419)
7551 - permit the client to specify desired inodes, at replay
7552 - duplicate requests when we queue them for replay reintegration
7553 - fix last_rcvd offset calculation
7554 - sync after each recovered transaction, so we always make progress
7555 - never, not always, ERESTART requests without transnos
7556 - store the lov_desc in the MDS, so we don't depend on getlovinfo to
7558 - skip replay if the MDS says that the client is already connected
7559 - don't check for a recovery-enabled export to match lctl's UUID
7560 - don't INC_USE_COUNT for phantom exports
7561 - don't crash when cleaning up phantom exports (567)
7562 - don't double-finish or set replay data for errored mdc_open requests
7563 - abort requests when they time out, so we don't get old replies
7564 - send/receive replies for AST messages again
7565 - if the client says that it doesn't have the lock, cancel it on the
7567 - if we timeout during I/O, don't try to cancel an in-use lock; instead
7568 mark it as destroyed, it will all work out when decref is called
7569 - fix module use counts (22, 581)
7571 - ASTs now expect a reply (server cancels lock on error reply)
7573 2002-12-02 Andreas Dilger <adilger@clusterfs.com>
7576 - fix many simultaneous client startup (392)
7577 - fix dentry->d_it clobbering
7578 - credentials weren't being shipped for readdir/getattr operations
7579 - remove invalid assertions triggered during some concurrent MD
7581 - proper Lustre versions added (336, 389)
7582 - fix memory leak for create error case (398)
7583 - fix LOV locking bug that would get cli/srv out of sync
7584 - fix echo client over LOV (409)
7585 - fix dbench 2, extN refcount problem (170, 258, 356, 418)
7586 - fix double-O_EXCL intent crash (424)
7587 - avoid sending multiple lock CANCELs (352)
7589 - MDS can do multi-client recovery (modulo bugs in new code)
7591 - many updates, edits, cleanups
7593 2002-11-18 Phil Schwan <phil@clusterfs.com>
7596 - fix null d_it dereference (346)
7597 - fix full OST/dbench hang (333)
7598 - fix permission problem with file removal (286)
7599 - fix removal of OSCs from LOV when they fail
7600 - fix NULL deref during bulk timeout (214)
7601 - fix problems related to multiple filesystems on one MDS (241)
7602 - fixed serious subtle metadata locking bugs
7603 - free locks on clients when inodes are removed due to memory
7605 - fix inode pointer in lock data (285)
7606 - partial support for multiple MDS on a single host (241)
7607 - data locks weren't cancelled at clear_inode time (290, 311)
7608 - intent locks could lead to unbounded lock growth (205)
7609 - added a maximum lock count, an LRU list, and a flusher
7610 - fix multiple rename (365)
7611 - properly abstracted the echo client
7612 - OSC locked 1 byte too many; fixed
7613 - rewrote brw callback code:
7614 - fixed recovery bugs related to LOVs (306)
7615 - fixed too-many-pages-in-one-write crash (191)
7616 - fixed (again) crash in sync_io_timeout (214)
7617 - probably fixed callback-related race (385)
7619 - Add capability to MDS protocol
7620 - LDLM cancellations and callbacks on different portals
7622 2002-10-28 Andreas Dilger <adilger@clusterfs.com>
7625 - limit client IOV size to PTL_MD_MAX_IOV (611336, 191)
7626 - defer open object destruction to close time (601981, 138)
7627 - open/close OST file handle in obdo (OBD_MD_FLHANDLE) (601981, 138)
7628 - move LDLM_ENQUEUE/CONVERT back to MDS portal (625069)
7629 - abstract ll_lookup2, fix ll_revalidate2 to use abstraction (256)
7630 - don't call obd_setattr in ll_file_release for destroyed objects
7631 * protocol change to lustre_msg: move |version| and add |flags|
7632 * protocol change to osc_punch: "start" in "o_size", "end" in "o_blocks"
7633 * lock replay: for LDLM_FL_REPLAY trust client to do right thing
7634 * added replay of create, unlink, link and rename operations during
7635 MDS failover; recovery should be much more robust now
7636 * remove failed OSCs from LOVs (only lov_create uses this so far)
7637 * the lustre-HOWTO was brought (more) up to date (582544)
7639 2002-10-23 Phil Schwan <phil@clusterfs.com>
7642 - in-use dentries weren't being reused properly (617851)
7643 - prevent multiple LDLM setup (599178)
7644 - fix LOV size calculations for truncate (617853)
7645 - fix client handling of MDS intent errors (POSIX)
7646 - fix permission bug in lovstripe.c test (624321)
7647 - fix MDS thread deadlock - move LDLM handler to DLM portal (625069)
7648 - truncate past end of file could corrupt data
7649 - proper cleanup after timeouts, crashes, etc (592524, 550815)
7650 - a race in recovery could return ETIMEDOUT to apps (623947)
7651 - building outside the source directory was fixed
7652 * the lustre-HOWTO was brought (more) up to date (582544)
7653 * major progress was made on recovery functionality
7655 2002-10-10 Phil Schwan <phil@clusterfs.com>
7658 - recovery deadlock fix
7659 - rm -rf causes LBUG fix (617817)
7660 - file open by multiple tasks fix (618962)
7661 - directory permissions bugs (602707 and 620007)
7662 - journal_stop fixed with locking (611313)
7663 - O_APPEND failures resolved (618273, perhaps 614459)
7664 - lconf PATH fix (619770)
7665 - IA64 build fix (621450)
7666 - RPC buffer sizes scale with amount of memory
7668 2002-10-01 Phil Schwan <phil@clusterfs.com>
7671 - locks would be cancelled without throwing away data pages,
7672 resulting in inconsistent data (605627)
7673 - inode attributes were not always being refreshed (605627, 612449)
7674 - lconf now continues to cleanup after lctl reports an error
7675 - MDS now enforces user permissions (602707)
7676 - lprocfs cleanup fixed, but not yet enabled (614157)
7677 - fixed infinite server hang, should a client not respond to an AST
7678 - avoid going into recovery if user calls readlink() with a buffer
7679 that's too small (613941)
7680 - AST RPCs no longer require replies (614867) -- this may be changed
7681 - don't crash server if client sends an IOV that's too big (611336)
7682 - fixed lock conversion deadlock (611892)
7683 - fixed the following of symlinks (614622)
7684 * recovery: the server can remove locks from a client that dies, other
7685 clients can make progress
7686 * more extN patch fixes
7687 * compile-time configurable ptlrpc buffer allocations
7689 - collaborative read cache document
7690 - Lustre Lite Performance CDR document-in-progress
7692 2002-09-20 Andreas Dilger <adilger@clusterfs.com>
7695 - fix typo in patch-2.4.18
7697 2002-09-20 Andreas Dilger <adilger@clusterfs.com>
7700 - clear ptlrpc request each time in handle_incoming_request()
7701 - unlink of files now destroys the object on the OST
7703 2002-09-19 Peter Braam <braam@clusterfs.com>
7705 * add hard link support
7706 * change obdfile creation method
7707 * kernel patch changed
7709 2002-09-19 Peter Braam <braam@clusterfs.com>
7712 - stack overflow bug in extN fixed
7714 2002-09-18 Andreas Dilger <adilger@clusterfs.com>
7716 * documentation updates
7717 - add man pages for config tools
7718 - update tests/README to describe testing with new config tools
7719 - finish metadata API descriptions
7720 * bug fixes and cleanups
7721 - statfs workaround for 16TB limit
7722 - LOV stripe allocation improved, can stripe on subset of OSTs
7723 - LOV file size/IO offset was wrong for files > 4GB in size
7724 - object EA data was being dropped, caused files to be unreadable
7725 - memory overflow with non-LOV OST caused memory corruption
7726 - fixed regression tests to work with new config tools, obdfilter
7727 - fixed bug when directory size became larger than 1 block
7728 - fixed bug (for single client case) when PWD was deleted
7729 - invalidate local directory pages when doing intent-based ops
7730 - avoid LDLM oops when lock callback contained bad data
7732 2002-09-09 Andreas Dilger <adilger@clusterfs.com>
7734 * documentation updates
7735 * bug fixes and cleanups
7736 - configuration tools
7739 - 64-bit compile warnings
7740 - 64-bit internal statfs data
7742 * test_brw on persistent OST devices
7744 * lprocfs (disabled)
7746 2002-09-04 Andreas Dilger <adilger@clusterfs.com>
7748 * documentation updates
7749 * bug fixes and cleanups
7750 * configuration tools
7752 2002-08-30 Peter J. Braam <braam@clusterfs.com>
7755 * many small fixes to 0_5_4
7756 * io/network handling
7757 * thinkos in MDS operations
7759 2002-08-24 Peter J. Braam <braam@clusterfs.com>
7762 * crucial basic fixes to 0.5.3
7763 * IOR, Iozone work over Elan
7766 2002-08-07 Phil Schwan <phil@clusterfs.com>
7767 * version 0_5_3, our first alpha
7768 * we use the new Portals iovs
7769 * documentation updates
7770 * bug fixes and cleanups
7771 * small changes in the DLM wire protocol
7773 2002-07-25 Peter J. Braam <braam@clusterfs.com>
7774 * version 0_5_1 with some initial stability,
7775 * locking on MD and file I/O.
7776 * documentation updates
7777 * several bug fixes since 0.5.0
7778 * small changes in wire protocol
7780 2002-07-18 Phil Schwan <phil@clusterfs.com>
7782 * delivered as Lustre Light Alpha
7783 * fixed a crash after handling invalid MDS requests
7784 * fixed directory pages for architectures with non-4k pages sizes
7786 2002-07-11 Andreas Dilger <adilger@clusterfs.com>
7787 * release version v0_4_4
7788 * Moves TCP acceptor to be on port 2432 (unused Coda port) instead
7790 * Fixes a number of interruption problems with OST operations.
7791 * Update documentation for portals header changes
7792 * Move all wire protocol structs/defines to lustre_idl.h
7793 * Fixes symlink length bug.
7794 * Add tcpdump to repository.
7796 2002-07-05 Andreas Dilger <adilger@clusterfs.com>
7797 * release version v0_4_3
7798 * Fixes statfs for inodes on extN.
7799 * Fixes bug in runtests which would delete /etc/hosts.
7800 * Use 64-bit object IDs wherever possible (not into VFS though)
7801 Remove ost_get_info, which is unused by lustre, and out of date.
7803 2002-07-03 Peter Braam <braam@clusterfs.com>
7804 * release version v0_4_2 Fixes a lookup error (type not passed)
7805 * move forward to head of Portals
7806 * move forward to latest Lustre kernel
7808 2002-06-25 Peter Braam <braam@clusterfs.com>
7809 * release version v0_4_1. Hopefully stable on single node use.