1 2010-xx-xx Oracle, Inc.
4 2.6.16.60-0.42.8 (SLES 10),
5 2.6.32.13-0.5.1 (SLES11),
6 2.6.18-194.8.1.el5 (RHEL 5)
7 2.6.18-194.8.1.0.1.el5 (OEL 5)
8 * Client support for unpatched kernels:
9 (see http://wiki.lustre.org/index.php?title=Patchless_Client)
10 2.6.16 - 2.6.32 vanilla (kernel.org)
11 * Recommended e2fsprogs version: 1.41.10-sun2
12 * The async journal commit feature (bug 19128) and the cancel
13 lock before replay feature (bug 16774) are disabled by default.
15 Severity : enhancement
17 Description: Update SLES11 SP1 kernel to 2.6.32.13-0.5.1.
21 Description: Enabling quotas fails with non-consecutive OST numbering.
23 -------------------------------------------------------------------------------
25 2010-07-31 Oracle, Inc.
27 * Support for kernels:
28 2.6.16.60-0.42.8 (SLES 10),
29 2.6.27.39-0.3.1 (SLES11),
30 2.6.18-194.3.1.el5 (RHEL 5)
31 2.6.18-194.3.1.0.1.el5 (OEL 5)
32 * Client support for unpatched kernels:
33 (see http://wiki.lustre.org/index.php?title=Patchless_Client)
34 2.6.16 - 2.6.30 vanilla (kernel.org)
35 * Recommended e2fsprogs version: 1.41.10-sun2
36 * The async journal commit feature (bug 19128) and the cancel
37 lock before replay feature (bug 16774) are disabled by default.
39 Severity : enhancement
41 Description: Update RHEL5.5 kernel to 2.6.18-194.8.1.el5 and OEL5.5 kernel to
42 2.6.18-194.8.1.0.1.el5.
44 Severity : enhancement
46 Description: using inkernel OFED stack for rhel5 & oel5.
48 Severity : enhancement
50 Description: Add "lfs_migrate" script from manual into lustre/scripts and RPMs
51 Details : lfs_migrate does a "poor man's" migration of files from their
52 current OST layout to a new OST layout as chosen by the MDS.
56 Description: mds_orphan_add_link()) error linking orphan to PENDING
57 Details : quota limits might disallow linking orphans to PENDING
58 when unlinking a file - temporary raise threads' privileges
59 when processing unlinks.
61 Severity : enhancement
63 Description: add conf-param -d option to remove permanent settings.
64 Details : Add the ability to remove permanent lctl conf_param settings.
65 (Previously conf_param settings could only be changed, not
66 removed.) This also provides a method to change failover
67 nid locations. Improve lctl man page.
69 Severity : enhancement
71 Description: add list_param to b1_8 and add "-R" option to list params
74 Severity : enhancement
76 Description: lfs quota output is not very convenient for awk/sed-parsing
77 Details : Some positions in lfs quota output table could be empty or
78 non-empty which made it hard to parse it with scripts, now a dash
79 is put instead of space where there is not supposed to be any data.
81 Severity : enhancement
83 Description: fix obdfilter-survey script to work properly with remote oss-s
85 Severity : enhancement
87 Description: add new OBDFILTER_SURVEY test suite
89 Severity : enhancement
91 Description: add new multiple mount protection (MMP) test suite
93 Severity : enhancement
95 Description: add support for async journal commit in echo client
97 Severity : enhancement
99 Description: allow userland programs to include <lustre/lustre_idl.h>
100 from stardard include directories
102 Severity : enhancement
104 Description: The prune-icache-use-trylock is no longer needed now that
105 the patch from bug 20008 is landed.
109 Description: The shrink grant feature is still active on the client although the
110 connect flag is not set.
114 Description: Don't leak grant space if the write failed with quota exceeded.
118 Description: Don't consume grant space twice on recoverable resent.
122 Description: a race condition could lead to SIGBUS being sent to an
123 application using mmap-ped files from Lustre
124 Details : truncate_complete_page implementation for the patchless
125 client could arbitrarily unset PG_Uptodate flag for a
126 page being kicked from the page cache, an uptodate check
127 right after a readpage call in filemap_fault could fail
128 because of that as though the page read had been unsuccessful.
132 Description: dlm lock slab shrinking is not efficient
133 Details : The dlm_locks slab can grow significantly and consumes a lot of
134 memory on the server. Set a hardlimit to grant_plan.
138 Description: Lustre does not do 1MB IOs to HW RAID
139 Details : Bump MAX_PHYS/HW_SEGMENTS and SG_ALL to 256 in the RHEL5 kernel.
140 This is what we do already for SLES kernels.
144 Description: bump maximum number of phys/hw segments in the SLES11 kernel
145 until s/g chaining works properly.
150 Description: LSI Fusion MPT driver hacks to improve performance
151 Details : Set CONFIG_FUSION_MAX_SGE to 256 for RHEL5
153 Severity : enhancement
155 Description: increase default md stripe_cache_size to 16k
158 Bugzilla : 15587/21439
159 Description: don't handle security.capability xattr
160 Details : CONFIG_SECURITY_FILE_CAPABILITIES is enabled by default on SLES11.
161 This results in additional getxattr calls, causing VBR test
162 failures as well as a preformance drop when writing.
166 Description: obdfilter-survey is no longer working
167 Details : revert patch from bug 20355 to resolve an issue with lctl --threads
168 not working correctly with $(PTHREAD_LIBS) being linked to lctl.
172 Description: ll_shrink_cache does not handle __GFP_FS properly
176 Description: lfs getstripe shows wrong info for directories
177 Details : Set correct LOVEA default values for filesystem-wide.
181 Description: FSX checksum false positves due to mmap IO
182 Details : Use OBD_FL_MMAP flag for IOs on a memory mapped file. Do not print
183 checksum errors, if the flag is set on a request.
187 Description: file operations after eviction have successful return values
188 Details : use vfs ->flush callback to return any pending async errors
193 Description: mdsrate fails to write after 1.3+M files opened
194 Details : decrease memory usage on clients by recycling dentries and
199 Description: obdfilter-survey gives unreasonably high numbers
200 Details : Wait for all threads to complete when running test_brw.
204 Description: do not set lustre read_only device when server umount and keep
205 client records for recoverable ones
209 Description: move sync_on_lock_cancel tunable to the obdfilter layer
210 Details : move the tunable to trigger a journal flush on lock cancel from
211 the ost layer to the obdfilter layer. This tunable is useful
212 when using the async journal commit feature.
216 Description: exp->exp_nid_stats == NULL in filter_tally()
217 Details : fix race with per-nid stats by delaying procfs cleanup until
222 Description: extent lock cancellation on client can keep the cpu busy for too
227 Description: Do not fail OST activation when a llog is not found, just
228 issue an error message.
232 Description: Don't enable extents by default for MDT.
236 Description: Protect bitfield access to ptlrpc_request's rq_flags, since
237 the AT code can access it concurrently while sending early
242 Description: Disable lockless truncate by default since it is sometimes flawed
243 and causes the write_disjoint test to fail.
247 Description: OSSs which don't have the patch from bug 20278 can trigger an
250 Severity : enhancement
252 Description: don't print message to the console when we have not managed to
257 Description: The MDS fails to synchronize OSTs which registered with the MGS
258 after the MDT. The problem is that OBD_NOTIFY_CREATE events are
259 raised too early and thus discarded by the MDT stack.
260 The fix consists of issuing OBD_NOTIFY_CREATE event in the lov
265 Description: Fix race when the ping evictor and a service thread execute
266 target_recovery_check_and_stop() concurrently.
270 Description: quota broadcast can trigger a LBUG on the MDT if there are
275 Description: Resetting the lov_objid values to last_id reported by the OST
276 during orphan recovery is incorrect and can cause the same
277 objects to be allocated twice.
279 Severity : enhancement
281 Description: "weak-modules" support
282 Details : Implement "weak-modules" support which enables kernel modules
283 to be used with any kernel that implements the same kABI. In
284 order to achieve this modules are now installed in
285 /lib/modules/$(uname -r)/updates/kernel on all distributions.
287 Severity : enhancement
289 Description: add writeconf as mount option
291 Severity : enhancement
293 Description: produce debuginfo packages for SLES.
295 Severity : enhancement
297 Description: add failover nidlist to the import proc file.
299 Severity : enhancement
301 Description: fix LUSTRE_SEQ_MAX_WIDTH for interoperability between 1.8
302 clients and 2.0 servers.
304 Severity : enhancement
306 Description: lfs find -s does not work correctly because of a bug in
311 Description: ll_read_ahead_page() must validate the dlm lock before using
316 Description: Prevent failover nids from registering with MGS first.
320 Description: fix lock inversion in ll_setattr_raw().
324 Description: object allocation is not balanced across OSTs.
325 Details : osc_precreate() should return 0, if there are enough objects left.
327 -------------------------------------------------------------------------------
329 2010-04-30 Oracle, Inc.
331 * Support for kernels:
332 2.6.16.60-0.42.8 (SLES 10),
333 2.6.27.39-0.3.1 (SLES11),
334 2.6.18-164.11.1.el5 (RHEL 5)
335 2.6.18-164.11.1.0.1.el5 (OEL 5)
336 * Client support for unpatched kernels:
337 (see http://wiki.lustre.org/index.php?title=Patchless_Client)
338 2.6.16 - 2.6.30 vanilla (kernel.org)
339 * Recommended e2fsprogs version: 1.41.10-sun2
340 * The async journal commit feature (bug 19128) and the cancel
341 lock before replay feature (bug 16774) are disabled by default.
345 Description: open-unlinked directories trigger MDS LBUG
346 Details : Fix regression introduced by the patch from bug 19640.
347 ext3_inc_count() can reset nlink to 1 when the directory
348 is indexed and inode->i_nlink == 2. Work around the problem
349 by incrementing nlink by 2 instead of 1.
353 Description: Reconnects are not throttled
354 Details : Don't wake up pinger on reconnect failures and rely on regular
355 pings to trigger the next reconnection. Please note that the
356 pinger already uses a smaller interval if the import is
360 Frequency : only with NFS export
362 Description: Console flooded with error message from ll_inode_from_lock()
364 Details : in mds_open, initialize the child_res_id before enqueuing
365 the OPEN lock for the child inode, then to avoid senting
366 wrong ldlm_res_id to client.
370 Description: allow multiple instances of the same nid in NID hash
371 Details : Case of multiple separate clients from the same NID (as
372 with liblustre) is legitimate and so we should allow
373 multiple instances of the same NID in nid hash.
377 Description: the readahead code can sleep on a semaphore while holding a
379 Details : in ras_update, "lov_get_info" could be called during increasing
380 readahead windows, which tries to get the mutex lock "lov_lock"
381 while holding the spin_lock "ras_lock", then causes system
386 Description: ASSERTION(cli->cl_avail_grant >= 0) failed
387 Details : fix assertion failure in the grant code.
391 Description: Use CNETERR (which is rate limited) in specific places in
392 the portal's LNET driver to avoid flooding the console.
396 Description: include last created object in precreate slow case
400 Description: don't do rep-ack if not created anything
401 Details : mds_open currently always put a lock into a rep-ack regardless
402 if something was created or not. This is pointless and only
403 creates needless contention. In fact the entire idea was to do
404 this for real creates as a recovery protection.
408 Description: Spurious error messages from smp_processor_id() on preemptible
410 Details : Disable a preemption by grabbing the lock in fs_trace_get_tcd()
411 first. The function fs_trace_get_tcd() was moved up.
415 Description: interval_erase() fix
416 Details : interval_erase() calls update_maxhigh() properly when child
421 Description: "lfs df" does not print stats for all mountpoints
422 Details : Print all mounted lustre filesystems with "lfs df"
426 Description: lfs setstripe -p not longer work with a relative pathname
427 Details : Use realpath() to provide absolute pathname.
431 Description: fix for truncated reply buffer
432 Details : reply buffer could be referred by reply_in_callback after released
437 Description: lustre.lov error when backing up symlinks with extended attributes
438 Details : Improved logic in ll_listxattr()
442 Description: properly handle null value for setattr -n lustre.lov
443 Details : Running "setfattr -n trusted.lov ." causes a NULL dereference
444 in ll_setxattr() due to no checking if "value" is NULL.
445 This command now resets to the default striping when executed
450 Description: stack overflow on lock cancellation due to fsync call
451 Details : sync_on_lock_cancel is needed for recovery when async journal
452 is enabled, but we actually just need to make sure that
453 metadata blocks have hit the journal, so doing a fs sync
454 should be enough and should consume less stack (just create an
455 empty handle and commmit it).
459 Description: using current->journal_info to store per-thread data leads
460 to problem under memory pressure
461 Details : disable the per-thread data (current->journal_info) containing
462 the lock info during I/O to work around the issue for short tem
466 Description: control DCACHE_LUSTRE_INVALID flag with MDS_INODELOCK_LOOKUP lock
467 Details : DCACHE_LUSTRE_INVALID is controlled by MDS_INODELOCK_LOOKUP
468 lock which is corresponding to "IT_LOOKUP", do not skip invalidate
473 Description: Cannot send after transport shutdown
474 Details : Clear imp_vbr_failed flag upon eviction
478 Description: soft lock in request set code during recovery
479 Details : during recovery, uses req->rq_set itself to replay the request
480 instead of ptlrpcd_recovery_pc
484 Description: Use CFS_ALLOC_IO instead of _STD in llap_from_page_with_lockh
485 Details : During an ll_readahead under ll_readpage, we have seen the the
486 OBD_SLAB_ALLOC hang under ldlm_pools_shrink when trying to lock
487 a page that is already locked by the readahead code.
492 Description: stop waitting for next replay transno if shutdown
493 Details : if the system is shutting down, wake up service thread blocked
494 to wait for next replay transno during recovery, then all the
495 references held by queued requests can be dropped and device
500 Description: lov_merge_lvb()) ASSERTION(spin_is_locked(&lsm->lsm_lock)) failed
501 Details : Protect lli->lli_smd pointer updates with lli->lli_lock.
505 Description: per-nid stats should not access lustre-hash internal structures
510 Description: mount.lustre fails to pass some options to mount()
515 Description: ext4 extent allocation is slower than in ext3
516 Details : Increase the default value of MB_DEFAULT_ORDER2_REQS to 8,
517 enlarge ext4 preallocation table for 2048 4K blocks extents
522 Description: incorrect triggering of synchronous IO
523 Details : The OSC can mistakenly fall back to synchronous IO when the
524 max_dirty_mb limit is reached and no write requests have yet
525 been issued. This can occur when the dirty pages are spread
526 over many files all of which are below the optimal request size.
530 Description: Optimize quota_ctl operations by sending requests in parallel
531 Details : Send MDS->OST quota_ctl requests in parallel, do not resend.
532 Compiled from two attachments in the ticket.
536 Description: ordering issue between transaction start & i_mutex
537 Details : start the transaction earlier in llog_lvfs_destroy to get
538 transaction start and inode mutex lock nested properly.
542 Description: lru resize SLV can get stuck
543 Details : calculate SLV with a greater precision to not lose small
544 changes due to interger math truncation; round up SLV only
545 if the amount of granted locks less than the limit to not
546 get stuck with this SLV
550 Description: avoid divide-by-zero in lprocfs_rd_import()
554 Description: lfs quota failed when OSTs are down
555 Details : really return approximate block/inode usage when OSTs are down
559 Description: llobdstat fix and enhancement
560 Details : add a counter to set a limit to how many samples will be returned
561 fix a wildcard in the path to limit to obdfilter stats only
565 Description: recovery-small 51 hang - MDT can not stop, Mount still busy with
567 Details : abort lock enqueue processing sooner when we umount is in progress.
569 Severity : enhancement
571 Description: Update RHEL5.4 kernel to 2.6.18-164.11.1.el5 and OEL5.4 kernel to
572 2.6.18-164.11.1.0.1.el5.
574 Severity : enhancement
576 Description: error message improvements
577 Details : Use INFO/WARN instead of WARN/ERROR for the slow messages.
578 Simplify MDT/OST service start message.
579 Suppress "changing the import ..." warning.
582 Bugzilla : 21961/17914
583 Description: ignore trailing -mdc when determining index number
586 Frequency : only with SLES11
588 Description: fix for a race condition in linux quotas implementation
589 Details : dq_flags(struct dquot) access is not properly locked which could
590 lead to certain inconsistencies when accessing it using non-atomic
591 bit operations like __set_bit in do_set_dqblk.
595 Description: return any pending async errors in close(2) using flush callback
597 -------------------------------------------------------------------------------
599 2010-01-29 Sun Microsystems, Inc.
601 * Support for kernels:
602 2.6.16.60-0.42.8 (SLES 10),
603 2.6.27.39-0.3.1 (SLES11),
604 2.6.18-164.11.1.el5 (RHEL 5)
605 2.6.18-164.6.1.0.1.el5 (OEL 5)
606 * Client support for unpatched kernels:
607 (see http://wiki.lustre.org/index.php?title=Patchless_Client)
608 2.6.16 - 2.6.30 vanilla (kernel.org)
609 * Recommended e2fsprogs version: 1.41.6.sun1
610 * The async journal commit feature (bug 19128) and the cancel
611 lock before replay feature (bug 16774) are disabled by default.
613 Severity : enhancement
615 Description: Update RHEL5.4 kernel to 2.6.18-164.11.1.el5 and
616 OEL5.4 kernel to 2.6.18-164.11.1.0.1.el5.
618 Severity : enhancement
619 Bugzilla : 21511/19848
620 Description: Update SLES11 kernel to 2.6.27.39-0.3.1.
622 Severity : enhancement
624 Description: Update supported SLES10 kernel to 2.6.16.60-0.42.8.
626 Severity : enhancement
628 Description: Update kernel to RHEL5.4 2.6.18-164.6.1.el5 and
629 OEL5 2.6.18-164.6.1.0.1.el5(Both in-kernel OFED enabled).
631 Severity : enhancement
633 Description: Build kernels (RHEL5, OEL5 and SLES10/11) using the vendor's own
636 Severity : enhancement
638 Description: Vanilla kernel 2.6.30 patchless client support.
643 Description: bad entry in directory xxx: inode out of bounds
644 Details : fix locking issue in the rename path which could race with any
645 other operations updating the same directory.
647 Severity : enhancement
649 Description: Make watchdog timer messages to be more clear and descriptive.
653 Description: cp -p command does not preserve the dates and timestamp
654 Details : mtime could be spoiled by a write callback
658 Description: Clear imp_force_reconnect correctly in ptlrpc_connect_interpret()
662 Description: Allow non-root access for "lfs check".
663 Details : Added a check in obd_class_ioctl() for OBD_IOC_PING_TARGET.
665 Severity : enhancement
667 Description: quotacheck performance/scaling issues
668 Details : reduce quotacheck time on empty filesystem by skipping uninit
671 Severity : enhancement
673 Description: Enhancement for lfs(1) command to use numeric uid/gid.
675 Severity : enhancement
677 Description: Adjust locks' extents on their first enqueue, so that at the time
678 they get granted, there is no need for another pass through the
679 queues since they are already shaped into the proper forms.
683 Description: Fix mds_shrink_intent_reply()/mds_intent_policy() to pass correct
684 arguments and prevent LBUG() in lustre_shrink_reply_v2().
688 Description: Change tunefs.lustre and mkfs.lustre --mountfsoptions so that
689 exactly the specified mount options are used. Leaving off
690 any "mandatory" mount options is an error. Leaving off any
691 default mount options causes a warning, but is allowed.
692 Change errors=remount-ro from mandatory to default. Sanitize
693 the mount string before storing it. Update man pages accordingly.
697 Description: mds_getattr() should return 0, even if mds_fid2entry() fails with
698 -ENOENT. Also fix in ptlrpc_expire_one_request() to print signed
701 Severity : enhancement
703 Description: Remove set_info(KEY_UNLINKED) from MDS/OSC
705 Severity : enhancement
707 Description: Clients can replay thousands of unused locks during recovery
708 Details : Don't replay unused locks (only read locks for now) during
709 recovery. This feature is disabled by default and can be
710 enabled by running the following command on the clients:
711 lctl get_param ldlm.cancel_unused_locks_before_replay
715 Description: can't stat file in some situation.
716 Details : improve initialize osc date when target is added to mds and
717 ability to resend too big getattr request is client isn't have info
722 Description: Prevent inconsistences between linux and lustre mount structures.
723 Details : Wait indefinitely in server_wait_finished() until mnt_count drops.
724 Make the sleep interruptible.
726 Severity : enhancement
728 Description: Communicate OST degraded/readonly state via statfs to MDS
729 Details : Flags in the statfs returned from OSTs indicate whether the
730 OST is in a degraded RAID state, or if the filesystem has
731 turned read-only after a filesystem error is detected.
736 Description: don't panic if EPROTO was hit when reading symlink
737 Details : correctly handling request reference in error cases.
742 Description: open sometimes returns ENOENT instead of EACCES
743 Details : checking permission should be part of open part of mds_open, not
744 lookup part. so server should be set DISP_OPEN_OPEN disposition
745 before starting permission check.
746 Also not need revalidate dentry if client already have LOOKUP lock.
750 Description: enable client interface failover
751 Frequency : on servers with multiple network interfaces
752 Details : When a child reconnects from another NID, properly update export
753 nid hash position and ldlm reverse import.
755 Severity : enhancement
757 Description: implemented direct I/O with arbitrary (nonaligned) memory
758 addresses and file offsets.
760 Severity : enhancement
762 Description: added more recovery timeout options.
764 Severity : enhancement
766 Description: added llapi_file_open, llapi_file_create, llapi_file_get_stripe
771 Description: Avoid deadlock for local client writes
772 Frequency : only on systems with clients writing to an OST on the same node
773 Details : Use new OBD_BRW_MEMALLOC flag to notify OST about writes in the
774 memory freeing context. This allows OST threads to set the
775 PF_MEMALLOC flag on task structures in order to allocate memory
776 from reserved pools and complete IO.
777 Use GFP_HIGHUSER for OST allocations for non-local client writes,
778 so that the OST threads generate memory pressure and allow
779 inactive pages to be reclaimed.
784 Descriptoin: lock ordering violation between &cli->cl_sem and _lprocfs_lock
785 Details : move ldlm namespace creation in setup phase to avoid grab
786 _lprocfs_lock with cli_sem held.
790 Frequency : only during format of test systems
791 Description: Unable to run several mkfs.lustre on loop devices at the same time
792 Details : mkfs.lustre returns error 256 on the concurrent loop devices
793 formatting. The solution is to proper handle the error.
795 Severity : enhancement
797 Description: implement async create (obd_async_create) method for osc, to avoid
798 too long waiting new ost objects with holding ldlm lock.
802 Frequency : occasionally during network problems
803 Description: client not allowed to reconnect to OST because of active request
804 Details : abort bulk requests received by the OST once the client has timed
805 out since the client will resend the request anyway. The client
806 also now retries to reconnect to the same server if a connect
807 request failed with EBUSY or -EAGAIN.
810 Frequency : rare, if used wide striped file and one ost in down.
812 Descriptoin: don't return error if we created a subset of objects for file.
813 Details : lov_update_create_set() uses set->set_success as index for created
814 objects, so if some requests failed, they will have hole at end of
815 array and we can use qos_shrink_lsm for allocate correct lsm.
819 Description: Slow stale export processing during normal start up
820 Details : The global mgc lock prevents OST setup to be run in parallel.
821 Replace the global lock with a per-config_llog_data semaphore.
825 Description: Out or order replies might be lost on replay
826 Details : In ptlrpc_retain_replayable_request if we cannot find retained
827 request with tid smaller then one currently being added, add it
828 to the start, not end of the list.
832 Description: BUG: soft lockup - CPU#1 stuck for 10s! [ll_mdt_07:4523]
833 Details : add cond_resched() calls to avoid hogging the cpu for too long
834 in the hash code. Make also lustre_hash_for_each_empty() more
837 Severity : enhancement
839 Description: Performance improvements for debug messages with D_RPCTRACE,
840 D_LDLM, D_QUOTA options.
843 Frequency : only with NFS export
845 Description: (lov_merge.c:74:lov_merge_lvb())
846 ASSERTION(spin_is_locked(&lsm->lsm_lock)) failed (SR 71691004)
847 Details : Fix a race in the nfs export code by populating inode
848 info while the new inode is still locked
850 Severity : enhancement
852 Description: add a new file in procfs called force_lbug. Writting to this
853 file triggers a LBUG. Only for test purpose.
857 Description: OOM killer causes node hang
858 Details : really interrupt the sleep in osc_enter_cache on signals
862 Description: LustreError: 9153:0:(quota_context.c:622:dqacq_completion()) LBUG
863 Details : fix race during quota release on the slave.
865 Severity : enhancement
867 Description: smaller hash bucket sizes, cleanups
868 Details : increase hash table sizes and enabled rehashing for pools, quota,
869 uuid, nid & per-nid stats.
871 Severity : enhancement
873 Description: Add ldiskfs maxdirsize mount option
874 Details : add max_dir size mount option
878 Description: panic in ll_statahead_thread
879 Details : prevent parent thread to be killed before its child
882 Frequency : only with 16TB device
884 Description: unable to perform "mount -t lustre" of 16TB OST device
885 Details : Mounting 16TB LUNs failed due to three bugs in mkfs.lustre.
889 Description: ASSERTION(atomic_read(&imp->imp_inflight) == 0) failed
890 Details : unregistering should be zero if no RPC inflight.
894 Description: hyperion: Oops during metabench
895 Details : Correct the refcount of lov_request_set
897 Severity : enhancement
899 Description: Add mptlinux and nxge drivers to Lustre builds
901 Severity : enhancement
903 Description: Fix watchdog timer message to be more clear
904 Details : Make watchdog timer messages more clear and descriptive.
908 Description: LNET soft lockups in socknal_cd thread
909 Details : don't hog CPU for active-connecting if another connd is
910 accepting connecting-requst from the same peer
914 Description: recovery-small test_17 hang
915 Details : Land several AT improvements & fixes.
919 Description: MDS panic and hanging client processes
920 Details : Replace exp_ops_stats with exp_nid_stats->nid_stats
924 Description: OSS stuck in recovery.
925 Details : fix race during recovery. class_unlink_export,
926 class_set_export_delayed and target_queue_last_replay_reply
927 may race while increasing/decreasing obd_recoverable_clients
928 and obd_delayed_clients, causing recovery to wait forever.
930 Severity : enhancement
932 Description: add cascading_rw.c to lustre/tests
936 Description: filter_last_id() NULL deref
937 Details : lprocfs_filter_rd_last_id() should check for the fully
938 setup obd device, before proceeding further.
940 Severity : enhancement
942 Description: Loadgen improvements
943 Details : stacksize and locking fixes for loadgen
947 Description: Quiet CERROR("dirty %d > system dirty_max %d\n"
948 Details : The atomic_read() allowing the atomic_inc() are not covered
949 by a lock. Thus they may safely race and trip this CERROR()
950 unless we add in a small fudge factor (+1).
952 Severity : enhancement
954 Description: shrink_slab: nr=-9223362083340912175
955 Details : fix spurious message from shrink_slab reporing negative nr
959 Description: Quiet bogus previously committed transno error
960 Details : suppress the "server went back in time" error message which
961 is always printed even in the common case after a client eviction
963 Severity : enhancement
965 Description: Parallel statfs() calls result in client eviction
966 Details : cache statfs data for 1s.
970 Description: parallel-scale test_compilebench: @@@@@@ FAIL: compilebench
972 Details : fix serveral issues in pinger code causing clients not to ping
973 servers for too long, resulting in evictions.
977 Description: e2fsck should warn when MMP update interval is extended
978 Details : print mmp_check_interval and make it possible to abort mount
979 operation in case it takes too long.
983 Description: mdsrate-create-large.sh, BUG: soft lockup - CPU#0 stuck for 10s!
984 Details : fix bug in the RHEL5's jbd2 callback patch.
988 Description: drop number of active requests when queued for recovery
989 Details : Now that we take a reference on the original request instead of
990 making a copy of it for recovery. We need to drop the number of
991 active requests or the queued requests will prevent all request
992 processing when they exceed (srv->srv_threads_running - 1).
994 Severity : enhancement
996 Description: refuse to invalidate operational quota files when they are in use
997 Details : an attempt to invalidate operational quota files on the quota
998 master is not actually permitted by VFS (returning -EPERM), but we
999 should not depend on that and should return the error earlier.
1003 Description: Applications stuck in jbd2_log_wait_commit during exit
1004 Details : fix deadlock between kjournald2 trying to acquire the page lock
1005 owned by an ost_io thread waiting for journal commit.
1007 -------------------------------------------------------------------------------
1009 2009-10-16 Sun Microsystems, Inc.
1011 * Support for kernels:
1012 2.6.16.60-0.42.4 (SLES 10),
1013 2.6.27.29-0.1 (SLES11, i686 & x84_64 only),
1014 2.6.18-128.7.1.el5 (RHEL 5),
1015 2.6.18-128.7.1.el5 (OEL 5),
1016 * Client support for unpatched kernels:
1017 (see http://wiki.lustre.org/index.php?title=Patchless_Client)
1018 2.6.16 - 2.6.27 vanilla (kernel.org)
1019 * Recommended e2fsprogs version: 1.41.6.sun1
1020 * ext4 support for RHEL5 is experimental and thus should not be
1023 Severity : enhancement
1025 Description: Add OEL5 support.
1027 Severity : enhancement
1029 Description: Update kernel to SLES11 2.6.27.29-0.1.
1033 Description: File checksum failures with OST read cache on
1034 Details : Disable page poisoning when the bulk transfer has to be aborted
1035 because the client got evicted.
1039 Description: Don't allow make backward step on assiging osc next id.
1040 Details : race between allocation next id and ll_sync thread can be cause
1041 of set wrong osc next id and can be kill valid ost objects.
1043 Severity : enhancement
1045 Description: Update kernel to RHEL5 2.6.18-128.7.1.el5.
1047 Severity : enhancement
1049 Description: Update kernel to SLES10 SP2 2.6.16.60-0.42.4.
1053 Description: Changes in raid5-large-io-rhel5.patch to calculate sectors properly
1057 Description: Increase the default BLK_DEF_MAX_SECTORS value for RHEL5 and SLES11
1061 Description: Error handling in osc_statfs_interpret() has been improved.
1062 Details : Check in osc_statfs_interpret() for EBADR.
1066 Description: Do not send statfs() requests to OSTs disabled by administrator.
1067 Details : Check in lov_prep_statfs_set() for non-NULL ltd_exp.
1071 Description: Do not update ctime for the deleted inode.
1072 Details : Check in mds_reint_unlink() before calling fsfilt_setattr().
1076 Description: Increase of the size of the LDLM resource hash.
1077 Details : Bump up RES_HASH_BITS=12.
1081 Description: correctly send lsm on open replay
1082 Details : MDS is trust to LSM size on replay open, but client can set wrong size
1087 Description: Deadlock between filter_destroy() and filter_commitrw_write().
1088 Details : filter_destroy() does not hold the DLM lock over the whole
1089 operation. If the DLM lock is dropped, filter_commitrw() can go
1090 through, causing the deadlock between page lock and i_mutex.
1091 The i_alloc_sem should also be hold in filter_destroy() while
1092 truncating the file.
1096 Description: truncate starts GFP_FS allocation under transaction causing deadlock
1097 Details : ldiskfs_truncate calls grab_cache_page which may start page
1098 allocation under an open transaction. This may lead to
1099 calling prune_icache with consequent lustre reentrance.
1102 Frequency : only when down/upgrading the MDS to 1.6/1.8 while 1.8 clients are
1103 still up and when the OST pool feature is used
1105 Description: interop testing got LBUG when run dd with OST pool
1106 :LustreError: 30032:0:(llite_lib.c:1913:ll_replace_lsm()) LBUG
1107 Details : down/upgrading the MDS to a version that doesn't/does support OST
1108 pool can cause clients to crash because the lsm has changed
1113 Description: missing tree_status on 1.8.1 RPM build
1114 Details : make rpms failed due because the tree_status file is missing.
1118 Description: continuing LustreError "mds adjust qunit failed!"
1119 Details : don't print message on the console when ->adjust_qunit fails.
1123 Description: don't increase ldlm timeout if previous client was evicted
1124 Details : if a client doesn't respond to a blocking callback within the
1125 adaptive ldlm enqueue timeout, don't adjust the adaptive estimate
1126 when the lock is next granted.
1130 Description: ost is being unmounted w/o all writes to last_rcvd landing on disk.
1131 affects recovery negatively.
1132 Details : make sure all exports have been properly destroyed by the zombie
1133 thread processed before stopping the target.
1137 Description: Performance degradation with O_DIRECT between 1.6 & 1.8.1 b190
1138 Details : disable write barrier for ext4/SLES11.
1142 Description: Kernel panic - not syncing: Out of memory and no killable
1143 processes... on OSS when iozone
1144 Details : fix memory leak in the journal checksum patch.
1148 Description: group quota "too many blocks" OSS crashes
1149 Details : we should keep the same uid/gid for lquota_chkquota() and
1150 lquota_pending_commit()
1154 Description: LustreError: 9153:0:(quota_context.c:622:dqacq_completion()) LBUG
1155 Details : don't LBUG on release quota error. Just a workaround until the
1156 problem is understood.
1158 ------------------------------------------------------------------------------
1159 2009-07-31 Sun Microsystems, Inc.
1161 * Support for kernels:
1162 2.6.16.60-0.39.3 (SLES 10),
1163 2.6.27.23-0.1 (SLES11, i686 & x84_64 only),
1164 2.6.18-128.1.14.el5 (RHEL 5),
1165 * Client support for unpatched kernels:
1166 (see http://wiki.lustre.org/index.php?title=Patchless_Client)
1167 2.6.16 - 2.6.27 vanilla (kernel.org)
1168 * Recommended e2fsprogs version: 1.41.6.sun1
1169 * File join has been disabled in this release, refer to Bugzilla 16929.
1170 * NFS export disabled when stack size < 8192. Since the NFSv4 export
1171 of Lustre filesystem with 4K stack may cause a stack overflow. For
1172 more information, please refer to bugzilla 17630.
1173 * ext4 support for RHEL5 is experimental and thus should not be
1176 Severity : enhancement
1178 Description: Update kernel to SLES10 SP2 2.6.16.60-0.39.3.
1181 Frequency : with 1.8 server and 1.6 clients
1183 Description: correctly shrink reply for avoid send too big message to client.
1184 Details : 1.8 mds is allocate to big buffer to LOV EA data and this produce
1185 some problems with sending this reply to 1.6 client.
1189 Description: Repeated atomic allocation failures.
1190 Details : Use GFP_HIGHUSER | __GFP_NOMEMALLOC flags for memory allocations
1191 to generate memory pressure and allow reclaiming of inactive pages.
1192 At the same time, do not allow to exhaust emergency pools.
1193 For local clients the use of GFP_NOFS will be introduced in 1.8.2
1195 Severity : enhancement
1196 Bugzilla : 19846, 18289
1197 Description: Update kernel to RHEL5 2.6.18-128.1.14.el5.
1199 Severity : enhancement
1200 Bugzilla : 19625, 16893, 18668, 19848
1201 Description: Add support for SLES11 2.6.27.23-0.1.
1203 Severity : enhancement
1205 Description: Update client support to vanila kernels up to 2.6.27.
1207 Severity : enhancement
1209 Description: Update kernel to SLES10 SP2 2.6.16.60-0.37.
1211 Severity : enhancement
1213 Description: Compile with -Werror by default for i686 and x86_64.
1217 Description: resolve race between obd_disconnect and class_disconnect_exports
1218 Details : if obd_disconnect will be called to already disconnected export he
1219 forget release one reference and osc module can't unloaded.
1221 Severity : enhancement
1223 Description: move AT tunable parameters for more consistent usage
1224 Details : add AT tunables under /proc/sys/lustre, add to conf_param parsing
1228 Descriptoin: correctly skip time estimate if in recovery
1229 Details : rq_send_state insn't bitmask so using bitwise ops is forbid.
1233 Descriptoin: OSS DeadLock
1234 Details : Use trylock to prevent deadlock when shrink icache.
1236 Severity : enhancement
1238 Description: Allow tuning service thread via /proc
1239 Details : For each service a new
1240 /proc/fs/lustre/{service}/*/thread_{min,max,started} entry is
1241 created that can be used to set min/max thread counts, and get the
1242 current number of running threads.
1244 Severity : enhancement
1246 Description: Add state history info file, enhance import info file
1247 Details : Track import connection state changes in a new osc/mdc proc file;
1248 add overview-type data to the osc/mdc import proc file.
1252 Description: Reduce small size read RPC
1253 Details : Set read-ahead limite for every file and only do read-ahead when
1254 available read-ahead pages are bigger than 1M to avoid small size
1259 Description: free_entry erroneously used groups_free instead of put_group_info
1261 Severity : enhancement
1263 Description: Make read-ahead stripe size aligned.
1265 Severity : enhancement
1267 Description: MDS create should not wait for statfs RPC while holding DLM lock.
1270 Frequency : rare, connect and disconnect target at same time
1272 Descriptoin: ASSERTION(atomic_read(&imp->imp_inflight) == 0
1273 Details : don't call obd_disconnect under lov_lock. this long time
1274 operation and can block ptlrpcd which answer to connect request.
1277 Frequency : start MDS on uncleanly shutdowned MDS device
1279 Descriptoin: ll_sync thread stay in waiting mds<>ost recovery finished
1280 Details : stay in waiting mds<>ost recovery finished produce random bugs
1281 due race between two ll_sync thread for one lov target. send
1282 ACTIVATE event only if connect realy finished and import have
1286 Frequency : start MDS on uncleanly shutdowned MDS device
1288 Descriptoin: aborting recovery hang on MDS
1289 Details : don't throttle destroy RPCs for the MDT.
1293 Description: Slow reads beyond 8Tb offsets.
1294 Details : Page index integer overflow in ll_read_ahead_page
1298 Description: Soft lockup on OSS after MDS failover
1299 Details : MSG_CONNECT_INITIAL is not set on the initial MDS->OST connect.
1300 As a conseqence, the patch from bug 18224 is not operational
1301 and the MDS export cannot be reused on the OSTs until it gets
1305 Frequency : rare, only if using MMP with Linux RAID
1307 Description: MMP doesn't work with Linux RAID
1308 Details : While using HA for Lustre servers with Linux RAID, it is possible
1309 that MMP will not detect multiple mounts. To make this work we
1310 need to unplug the device queue in RAID when the MMP block is being
1311 written. Also while reading the MMP block, we should read it from
1312 disk and not the cached one.
1315 Frequency : rare, during recovery
1317 Description: Assertion failure in ldlm_lock_put
1318 Details : Do not put cancelled locks into replay list, hold references on
1319 locks in replay list
1323 Description: 1.6.5 mdsrate performance is slower than 1.4.11/12 (MDS is not cpu bound!)
1324 Details : create_count always drops to the min value (=32) because grow_count
1325 is being changed before the precreate RPC completes.
1328 Frequency : Only in RHEL5 when mounting multiple ext3 filesystems
1331 Description: "kmem_cache_create: duplicate cache jbd_4k" error message
1332 Details : add proper locking for creation of jbd_4k slab cache
1336 Description: MMP check in ext3_remount() fails without displaying any error
1337 Details : When multiple mount protection fails during remount, proper error
1342 Description: Rare Client crash on resend if the file was deleted.
1343 Details : When file is opened, but open reply is lost and file is
1344 subsequently deleted before resend, resend processing logic
1345 breaks trying to open the file again, should not try to open.
1349 Description: add check for >8TB ldiskfs filesystems
1350 Details : ext3-based ldiskfs does not support greater than 8TB LUNs.
1351 Don't allow >8TB ldiskfs filesystems to be mounted without
1352 force_over_8tb mount option
1356 Description: Client locked up when running multiple instances of an app. on
1357 multiple mount points
1358 Details : ll_shrink_cache() can sleep while holding the ll_sb_lock.
1359 Convert ll_sb_lock to a read/write semaphore to fix the problem.
1363 Description: Cannot acces an NFS-mounted Lustre filesystem
1364 Details : An NFS client cannot access the Lustre filesystem NFS-mounted
1365 from a Lustre-client exporting the Lustre filesystem via NFS.
1369 Description: panic in ll_statahead_thread
1370 Details : grab dentry reference in parent process.
1373 -------------------------------------------------------------------------------
1375 tbd Sun Microsystems, Inc.
1377 * Support for kernels:
1378 2.6.16.60-0.31 (SLES 10),
1379 2.6.18-128.1.6.el5 (RHEL 5),
1380 2.6.22.14 vanilla (kernel.org)
1381 * Client support for unpatched kernels:
1382 (see http://wiki.lustre.org/index.php?title=Patchless_Client)
1383 2.6.16 - 2.6.22 vanilla (kernel.org)
1384 * Recommended e2fsprogs version: 1.40.11-sun1
1385 * File join has been disabled, refer to Bugzilla 16929.
1386 * A new Lustre ADIO driver is available for MPICH2-1.0.7.
1387 * NFS export disabled when stack size < 8192. Since the NFSv4 export of
1388 Lustre filesystem with 4K stack may cause a stack overflow. For more
1389 information, please refer to bugzilla 17630.
1391 Severity : enhancement
1393 Description: Update to RHEL5.3 kernel-2.6.18-128.1.6.el5.
1395 Severity : enhancement
1397 Description: Update OFED release to 1.4.1 RC4
1399 Severity : major, only with big OST
1401 Description: Very poor metadata performance on Infiniband lustre configuration
1402 Details : OST object precreation becomes very slow on big OSTs. This is due
1403 to the ialloc patch spending too much time scanning groups.
1406 Frequency : during recovery
1408 Description: don't mix llog inodes with normal.
1409 Details : allocate inodes for log in last inode group
1414 Description: fix lqs' reference which won't be put in some situations
1415 Details : This patch fixes:
1416 1. In quota_check_common(), this function will check quota
1417 for user and group, but only send one return via "pending".
1418 In most cases, the pendings should be same. But that is not
1420 2. If quotaoff runs between lquota_chkquota() and
1421 lquota_pending_commit(), the same thing will happen too.
1422 That is why it comes:
1423 - if (!ll_sb_any_quota_active(qctxt->lqc_sb))
1426 -------------------------------------------------------------------------------
1428 2008-12-31 Sun Microsystems, Inc.
1430 * Support for kernels:
1431 2.6.5-7.314 (SLES 9),
1432 2.6.9-67.0.22.EL (RHEL 4),
1433 2.6.16.60-0.31 (SLES 10),
1434 2.6.18-92.1.17.el5 (RHEL 5),
1435 2.6.22.14 vanilla (kernel.org)
1436 * Client support for unpatched kernels:
1437 (see http://wiki.lustre.org/index.php?title=Patchless_Client)
1438 2.6.16 - 2.6.22 vanilla (kernel.org)
1439 * Client support for unpatched kernels:
1440 we do not recommend using patchless RHEL4 clients with kernels
1441 prior to 2.6.9-55EL (RHEL4U5).
1442 * Recommended e2fsprogs version: 1.40.11-sun1
1443 * Note that reiserfs quotas are disabled on SLES 10 in this kernel.
1444 * RHEL 4 and RHEL 5/SLES 10 clients behaves differently on 'cd' to a
1445 removed cwd "./" (refer to Bugzilla 14399).
1446 * A new quota file format has been introduced in 1.6.5.
1447 The format conversion from prior releases is handled transparently,
1448 but releases older than 1.4.12/1.6.5 don't understand this new
1449 format. The automatic format conversion can be avoided by running
1450 the following command on the MDS:
1451 'tunefs.lustre --param="mdt.quota_type=ug1" $MDTDEV'.
1452 For more information, please refer to bugzilla 13904.
1453 * A new quota file format was introduced in 1.6.6/1.8.0 (kernels 2.6.16+).
1454 The format conversion from prior releases is handled transparently,
1455 but releases older than 1.6.6/1.8.0 don't understand this new
1456 format. The automatic format conversion can be avoided by running
1457 the following commands on the MDS and OSS servers (for
1458 pre 1.4.12-1.6.5 quota files):
1459 'tunefs.lustre --param="mdt.quota_type=ug1" $MDTDEV',
1460 'tunefs.lustre --param="ost.quota_type=ug1" $MDTDEV'
1461 or (for 1.4.12/1.6.5 quota files)
1462 'tunefs.lustre --param="mdt.quota_type=ug2" $MDTDEV',
1463 'tunefs.lustre --param="ost.quota_type=ug2" $MDTDEV'
1464 For more information, please refer to bugzilla 13904.
1465 * Output of lfs quota has been made less detailed by default,
1466 old (verbose) output can be obtained by using -v option.
1467 * File join has been disabled in this release, refer to Bugzilla 16929.
1468 * A new Lustre ADIO driver is available for MPICH2-1.0.7.
1469 * NFS export disabled when stack size < 8192. Since the NFSv4 export of
1470 Lustre filesystem with 4K stack may cause a stack overflow. For more
1471 information, please refer to bugzilla 17630.
1473 Severity : enhancement
1475 Description: Caching OSS
1476 Details : introduce data caching on the OSS. The OSS now relies on the linux
1477 kernel page cache to keep recently accessed data in memory.
1478 It is worth noting that all write requests are still flushed
1479 synchronously as in lustre 1.6.
1481 Severity : enhancement
1483 Description: Version based recovery
1484 Details : introduce finer grained recovery able to detect transaction
1485 dependencies and can deal with transaction gaps caused by clients
1486 failing at the same time as the server.
1488 Severity : enhancement
1490 Description: Enable adaptive timeouts by default
1491 Details : The Lustre timeout value in /proc/sys/lustre/timeout is now
1492 managed dynamically based on server load and should not need
1493 to be tuned manually based on cluster size. This allows Lustre
1494 to work under a wider variety of system sizes and loads, without
1495 unnecessarily causing lengthy recovery times.
1497 Severity : enhancement
1499 Description: Add OST Pools support
1500 Details : File striping can now be set to use an arbitrary pool of OSTs
1502 Severity : enhancement
1504 Description: add lazystatfs mount option to allow statfs(2) to skip down OSTs
1505 Details : allow skip disconnected ost for send statfs request and hide error
1509 Frequency : rare, on llog test 6
1511 Descriptoin: don't allow connect to already connected import
1512 Details : allowing connect to already connected import is hide connecting problem.
1515 Frequency : rare, on failed llog setup
1517 Descriptoin: don't leak obd reference on failed llog setup
1518 Details : for failed llog setup - mgc forget call class_destroy_import for
1519 client import, move destroy import to more generic place.
1524 Descriptoin: allow kill process which wait statahead result
1525 Details : for some reasons 'ls' can stick in waiting result from statahead,
1526 in this case need way for kill this process.
1529 Frequency : rare, at shutdown
1531 Descriptoin: panic at umount
1532 Details : llap_shrinker can be raced with killing super block from list and
1533 this produce panic with access to already freeded pointer
1538 Descriptoin: don't lose wakeup for imp_recovery_waitq
1539 Details : recover_import_no_retry or invalidate_import and import_close can
1540 both sleep on imp_recovery_waitq, but we was send only one wakeup
1546 Descriptoin: panic in mds_open
1547 Details : don't confuse mds_finish_transno() with PTR_ERR(-ENOENT)
1552 Descriptoin: stuck in cache_remove_extent() or panic with accessing to already
1554 Details : release lock refernce only after add page to pages list.
1557 Frequency : always with long access acl
1559 Descriptoin: mds can't pack reply with long acl.
1560 Details : mds don't control size of acl but they limited by reint/getattr
1565 Frequency : on remount
1567 Description: external journal device not working after the remount
1568 Details : clear dev_rdonly flag for external journal devices in
1574 Description: shutdown vs evict race
1575 Details : client_disconnect_export vs connect request race.
1576 if client will evicted at this time - we start invalidate
1577 thread without referece to import and import can be freed
1583 Description: shrink LOV EAs before replying
1584 Details : correctly adjust LOV EA buffer for reply.
1589 Description: don't skip ost target if they assigned to file
1590 Details : Drop slow OSCs if we can, but not for requested start idx.
1591 This means "if OSC is slow and it is not the requested
1592 start OST, then it can be skipped, otherwise skip it only
1593 if it is inactive/recovering/out-of-space.
1595 Severity : enhancement
1597 Description: Update to RHEL5 kernel-2.6.18-92.1.17.el5.
1600 Frequency : rare, need acl's on inode.
1602 Description: client can't handle ost additional correctly
1603 Details : if ost was added after client connected to mds client can have
1604 hit lnet_try_match_md ... to big messages to wide striped files.
1605 in this case need teach client to handle config events about add
1606 lov target and update client max ea size at that event.
1608 Severity : enhancement
1610 Description: Update to sles9 kernel-2.6.5-7.314.
1612 Severity : enhancement
1614 Description: Update to SLES10 SP2 kernel-2.6.16.60-0.31.
1617 Frequency : Create a symlink file with a very long name
1619 Description: ldlm_cancel_pack()) ASSERTION(max >= dlm->lock_count + count)
1620 Details : If there is no extra space in the request for early cancels,
1621 ldlm_req_handles_avail() returns 0 instead of a negative value.
1626 Description: mds is deadlocked
1627 Details : in rare cases, inode in catalog can have i_no less than have parent
1628 i_no, this produce wrong order for locking during open, and
1629 parallel unlink can be lock open. this need teach mds_open to grab
1630 locks in resource id order, not at parent -> child order.
1632 Severity : enhancement
1634 Description: Add /proc entry for import status
1635 Details : The mdc, osc, and mgc import directories now have
1636 an import directory that contains useful import data for debugging
1637 connection problems.
1639 Severity : enhancement
1641 Description: Re-disable certain /proc logging
1642 Details : Enable and disable client's offset_stats, extents_stats and
1643 extents_stats_per_process stats logging on the fly.
1646 Frequency : Only on FC kernels 2.6.22+
1648 Description: oops in statahead
1649 Details : Do not drop reference count for the dentry from VFS when lookup,
1650 VFS will do that by itself.
1652 Severity : enhancement
1654 Description: Generic /proc file permissions
1655 Details : Set /Proc file permissions in a more generic way to enable non-
1656 root users operate on some /proc files.
1660 Description: Hitting mdc_commit_close() ASSERTION
1661 Details : Properly handle request reference release in
1662 ll_release_openhandle().
1666 Frequency : only patchless client
1667 Description: add workaround for race between add/remove dentry from hash
1669 Severity : enhancement
1671 Description: Allow OST glimpses to return PW locks
1675 Description: LBUG when llog conf file is full
1676 Details : When llog bitmap is full, ENOSPC should be returned for plain
1681 Description: Prevent import from entering FULL state when server in recovery
1685 Description: service mount cannot take device name with ":"
1686 Details : Only when device name contains ":/" will mount treat it as
1692 Description: replace ptlrpcd with the statahead thread to interpret the async
1693 statahead RPC callback
1697 Frequency : on recovery
1698 Description: I/O failures after umount during fail back
1699 Details : if client reconnected to restarted server we need join to recovery
1700 instead of find server handler is changed and process self eviction
1701 with cancel all locks.
1703 Severity : enhancement
1705 Description: Update to RHEL5 kernel-2.6.18-92.1.10.el5.
1709 Description: Kernel BUG tries to release flock
1710 Details : Lustre does not destroy flock lock before last reference goes
1711 away. So always drop flock locks when client is evicted and
1712 perform unlock regardless of successfulness of speaking to MDS.
1714 Severity : enhancement
1716 Description: Update to SLES10 SP2 kernel-2.6.16.60-0.27.
1718 Severity : enhancement
1720 Description: Upcall on Lustre log has been dumped
1721 Details : Allow for a user mode script to be called once a Lustre log has
1722 been dumped. It passes the filename of the dumped log to the
1723 script, the location of the script can be specified via
1724 /proc/sys/lnet/debug_log_upcall.
1729 Description: avoid messages about idr_remove called for id that is not allocated
1730 Details : Move assigment s_dev for clustered nfs to end of initialization,
1731 for avoid problem with error handling.
1736 Description: avoid Already found the key in hash [CONN_UNUSED_HASH] messages
1737 Details : When connection is reused this not moved from CONN_UNUSED_HASH into
1738 CONN_USED_HASH and this prodice warning when put connection again
1744 Description: avoid ASSERTION(client_stat->nid_exp_ref_count == 0) failed
1745 Details : release reference to stats when client disconnected, not
1746 when export destroyed for avoid races when client destroyed
1747 after main ost export.
1751 Description: more cleanup in mds_lov
1752 Details : add workaround for get valid ost count for avoid warnings about
1753 drop too big messages, not init llog cat under semphore which
1754 can be blocked on reconnect and break normal replay, fix access
1757 Severity : enhancement
1759 Description: Export bytes_read/bytes_write count on OSC/OST.
1763 Description: Early reply size mismatch, MGC loses connection
1764 Details : Apply the MGS_CONNECT_SUPPORTED mask at reconnect time so
1765 the connect flags are properly negotiated.
1769 Description: Properly propagate oinfo flags from lov to osc for statfs
1770 Details : restore missing copy oi_flags to lov requests.
1774 Description: exports in /proc are broken
1775 Details : recreate /proc entries for clients when they reconnect.
1777 Severity : enhancement
1779 Description: Add man pages for llobdstat(8), llstat(8), plot-llstat(8),
1780 : l_getgroups(8), lst(8), routerstat(8)
1781 Details : included man pages for llobdstat(8), llstat(8),
1782 : plot-llstat(8), l_getgroups(8), lst(8), routerstat(8)
1784 Severity : enhancement
1786 Description: Implement lustre ll_show_options method.
1788 Severity : enhancement
1790 Description: Update to SLES9 kernel-2.6.5-7.312.
1792 Severity : enhancement
1794 Description: Update to RHEL4 kernel-2.6.9-67.0.22.EL.
1798 Description: exports in /proc are broken
1799 Details : recreate /proc entries for clients when they reconnect.
1803 Description: don't fail open with -ERANGE
1804 Details : if client connected until mds will be know about real ost count
1805 get LOV EA can be fail because mds not allocate enougth buffer
1810 Description: Resolve device initialization race
1811 Details : Prevent proc handler from accessing devices added to the
1812 obd_devs array but yet be intialized.
1814 Severity : enhancement
1816 Description: configure's --enable-quota should check the
1817 : kernel .config for CONFIG_QUOTA
1818 Details : configure is terminated if --enable-quota is passed but
1819 : no quota support is in kernel
1821 Severity : enhancement
1823 Description: Update to SLES10 SP2 kernel-2.6.16.60-0.23.
1825 Severity : enhancement
1827 Description: Update to RHEL5 kernel-2.6.18-92.1.6.el5.
1831 Frequency : rare, on PPC clients
1832 Description: don't swab ost objects in response about directory, because
1834 Details : bug similar bug 14856, but in different function.
1836 Severity : enhancement
1838 Description: lfs quota tool enhancement
1839 Details : added units specifiers support for setquota, default to
1840 current uid/gid for quota report, short quota stats by
1841 default, nonpositional parameters for setquota, added
1842 llapi_quotactl manual page.
1844 Severity : enhancement
1846 Description: *optional* service tags registration
1847 Details : if the "service tags" package is installed on a Lustre node
1848 When the filesystem is mounted, a local-node service tag will
1849 be created. See http://inventory.sun.com/ for more information
1850 about the Service Tags asset management system.
1854 Description: Client runs out of low memory
1855 Details : Consider only lowmem when counting initial number of llap pages
1858 Frequency : occasional
1860 Description: add refcount for osc callbacks, so avoid panic on shutdown
1862 Severity : enhancement
1864 Description: Update to RHEL4 kernel-2.6.9-67.0.20.
1867 Frequency : testing only
1869 Description: sanity test 65a fails if stripecount of -1 is set
1870 Details : handle -1 striping on filesystem in ll_dirstripe_verify
1873 Frequency : only in unusual configurations
1875 Description: Kernel panic with find ost index.
1876 Details : lov_obd have panic if some OST's have sparse indexes.
1878 Severity : enhancement
1880 Description: Update to RHEL5 kernel-2.6.18-53.1.21.el5.
1883 Frequency : rarely, if filesystem is mounted with -o flock
1885 Description: do not process already freed flock
1886 Details : flock can possibly be freed by another thread before it reaches
1887 to ldlm_flock_completion_ast.
1890 Frequency : rarely, if filesystem is mounted with -o flock
1892 Description: LBUG during stress test
1893 Details : Need properly lock accesses the flock deadlock detection list.
1896 Frequency : rarely, if binaries are being run from Lustre
1898 Description: oops in page fault handler
1899 Details : kernel page fault handler can return two special 'pages' in
1900 error case, don't try dereference NOPAGE_SIGBUS and NOPAGE_OMM.
1903 Frequency : rarely, during shutdown
1905 Description: timeout with invalidate import.
1906 Details : ptlrpcd_check call obd_zombie_impexp_cull and wait request which
1907 should be handled by ptlrpcd. This produce long age waiting and
1908 -ETIMEOUT ptlrpc_invalidate_import and as result LASSERT.
1914 Description: ASSERTION(CheckWriteback(page,cmd)) failed
1915 Details : badly clear PG_Writeback bit in ll_ap_completion can produce false
1919 Frequency : only with broken builds/installations
1921 Description: no LBUG if lquota.ko and fsfilt_ldiskfs.ko are different versions
1922 Details : just return an error to a user, put a console error message
1924 Severity : enhancement
1926 Description: Update to RHEL5 kernel-2.6.18-53.1.19.el5.
1928 Severity : enhancement
1930 Description: Update to RHEL4 kernel-2.6.9-67.0.15.
1932 Severity : enhancement
1934 Description: enable MGS and MDT services start separately
1935 Details : add a 'nomgs' option in mount.lustre to enable start a MDT with
1936 a co-located MGS without starting the MGS, which is a complement
1937 to 'nosvc' mount option.
1940 Frequency : always, on big-endian systems
1942 Description: cleanup in ptlrpc code, related to PPC platform
1943 Details : store magic in native order avoid panic's in recovery on PPC
1944 node and forbid from this error in future. Also fix posibily
1945 of twice swab data. Fix get lov striping to userpace.
1948 Frequency : rarely, if replay get lost on server
1950 Description: server incorrectly drop resent replays lead to recovery failure.
1951 Details : do not drop replay according to msg flags, instead we check the
1952 per-export recovery request queue for duplication of transno.
1955 Frequency : after recovery
1957 Description: precreate to many object's after del orphan.
1958 Details : del orphan st in oscc last_id == next_id and this triger growing
1959 count of precreated objects. Set flag LOW to skip increase count
1960 of precreated objects.
1963 Frequency : rare, on clear nid stats
1965 Description: ASSERTION(client_stat->nid_exp_ref_count == 0)
1966 Details : when clean nid stats sometimes try destroy live entry,
1967 and this produce panic in free.
1970 Frequency : occasionally since 1.6.4
1972 Description: Stack overflow during MDS log replay
1973 Details : ease stack pressure by using a thread dealing llog_process.
1976 Frequency : very rare
1978 Description: MDT cannot be unmounted, reporting "Mount still busy"
1979 Details : Mountpoint references were being leaked during open reply
1980 reconstruction after an MDS restart. Drop mountpoint reference
1981 in reconstruct_open() and free dentry reference also.
1986 Description: wait until IO finished before start new when do lock cancel.
1987 Details : VM protocol want old IO finished before start new, in this case
1988 need wait until PG_writeback is cleared until check dirty flag
1989 and call writepages in lock cancel callback.
1994 Description: mds_mfd_close() ASSERTION(rc == 0)
1995 Details : In mds_mfd_close(), we need protect inode's writecount change
1996 within its orphan write semaphore to prevent possible races.
1999 Frequency : rare, on shutdown ost
2001 Description: don't hit live lock with umount ost.
2002 Details : shrink_dcache_parent can be in long loop with destroy dentries,
2003 use shrink_dcache_sb instead.
2006 Frequency : only when echo_client is used
2008 Description: don't panic with use echo_client
2009 Details : echo client pass NULL as client nid pointer and this produce NULL
2010 pointer dereference.
2013 Frequency : Always on 32-bit PowerPC systems
2015 Description: fix build on PPC32
2016 Details : compile code with -m64 flag produce wrong object file for PPC32.
2021 Description: MDS LBUG: ASSERTION(!IS_ERR(dchild))
2022 Details : In reconstruct_* functions, LASSERTs on both the data supplied
2023 by a client, and the data on disk are dangerous and incorrect.
2024 Change them with client eviction.
2026 Severity : enhancement
2028 Description: skiplist implementation simplification
2029 Details : skiplists are used to group compatible locks on granted list
2030 that was implemented as tracking first and last lock of each lock
2031 group the patch changes that to using doubly linked lists
2035 Description: delete compatibility for 32bit qdata
2036 Details : as planned, when lustre is beyond b1_8, lquota won't support 32bit
2037 qunit. That means servers of b1_4 and servers of b1_8 can't be
2038 used together if users want to use quota.
2041 Frequency : only with administrator action
2043 Description: mount failure if config log has invalid conf_param setting
2044 Details : If administrator specified an incorrect configuration parameter
2045 with "lctl conf_param" this would cause an error during future
2046 client mounts. Instead, ignore the bad configuration parameter.
2049 Frequency : blocks per group < blocksize*8 and uninit_groups is enabled
2051 Description: ldiskfs error: XXX blocks in bitmap, YYY in gd
2052 Details : If blocks per group is less than blocksize*8, set rest of the
2056 Frequency : Application do stride read on lustre
2058 Description: The read performance will drop a lot if the application does
2060 Details : Because the stride_start_offset are missing in stride read-ahead,
2061 it will cause clients read a lot of unused pages in read-ahead,
2062 then the read-performance drops.
2066 Description: more ldlm soft lockups
2067 Details : In ldlm_resource_add_lock(), call to ldlm_resource_dump()
2068 starve other threads from the resource lock for a long time in
2069 case of long waiting queue, so change the debug level from
2070 D_OTHER to the less frequently used D_INFO.
2072 Severity : enhancement
2074 Description: add -gid, -group, -uid, -user options to lfs find
2076 Severity : enhancement
2078 Description: ll_recover_lost_found_objs - recover objects in lost+found
2079 Details : OST corruption and subsequent e2fsck can leave objects in the
2080 lost+found directory. Using the "ll_recover_lost_found_objs"
2081 tool, these objects can be retrieved and data can be salvaged
2082 by using the object ID saved in the fid EA on each object.
2087 Description: this bug _only_ happens when inode quota limitation is very low
2088 (less than 12), so that inode quota unit is 1 at initialization.
2089 Details : if remaining quota equates 1, it is a sign to demonstate that quota
2090 is effective now. So least quota qunit should be 2.
2094 Description: Hung threads in invalidate_inode_pages2_range
2095 Details : The direct IO path doesn't call check_rpcs to submit a new RPC once
2096 one is completed. As a result, some RPCs are stuck in the queue
2101 Description: Procfs and llog threads access destoryed import sometimes.
2102 Details : Sync the import destoryed process with procfs and llog threads by
2103 the import refcount and semaphore.
2107 Description: mds fails to respond, threads stuck in ldlm_completion_ast
2108 Details : Sort source/child resource pair after updating child resource.
2113 Description: kernel BUG at ldiskfs2_ext_new_extent_cb
2114 Details : If insertion of an extent fails, then discard the inode
2115 preallocation and free data blocks else it can lead to duplicate
2120 Description: don't always update ctime in ext3_xattr_set_handle()
2121 Details : Current xattr code updates inode ctime in ext3_xattr_set_handle()
2122 In some cases the ctime should not be updated, for example for
2123 2.0->1.8 compatibility it is necessary to delete an xattr and it
2124 should not update the ctime.
2128 Description: add quota statistics
2129 Details : 1. sort out quota proc entries and proc code.
2130 2. add quota statistics
2135 Description: quotas are not honored with O_DIRECT
2136 Details : all writes with the flag O_DIRECT will use grants which leads to
2137 this problem. Now using OBD_BRW_SYNC to guard this.
2141 Bugzilla : 15713/16362
2142 Description: Assertion in iopen_connect_dentry in 1.6.3
2143 Details : looking up an inode via iopen with the wrong generation number can
2144 populate the dcache with a disconneced dentry while the inode
2145 number is in the process of being reallocated. This causes an
2146 assertion failure in iopen since the inode's dentry list contains
2147 both a connected and disconnected dentry.
2151 Description: assertion failure in ldlm_handle2lock()
2152 Details : fix a race between class_handle_unhash() and class_handle2object()
2153 introduced in lustre 1.6.5 by bug 13622.
2155 Severity : enhancement
2157 Description: superblock lock contention with many SMP cores on one client
2158 Details : several client filesystem locks were highly contended on SMP
2159 NUMA systems with 8 or more cores. Per-CPU datastructures
2160 and more efficient locking implemented to reduce contention.
2165 Description: Kernel BUG: sd_iostats_bump: unexpected disk index
2166 Details : remove the limit of 256 scsi disks in the sd_iostat patch
2171 Description: oops in sd_iostats_seq_show()
2172 Details : unloading/reloading the scsi low level driver triggers a kernel
2173 bug when trying to access the sd iostat file.
2178 Description: Kernel panics during QLogic driver reload
2179 Details : REQ_BLOCK_PC requests are not handled properly in the sd iostat
2180 patch, causing memory corruption.
2185 Description: journal_dev option does not work in b1_6
2186 Details : pass mount option during pre-mount.
2188 Severity : enhancement
2190 Description: Add a FIEMAP(FIle Extent MAP) ioctl for ldiskfs
2191 Details : FIEMAP ioctl will allow an application to efficiently fetch the
2192 extent information of a file. It can be used to map logical blocks
2193 in a file to physical blocks in the block device.
2196 Frequency : only with adaptive timeout enabled
2198 Description: DEBUG_REQ() bad paging request
2199 Details : ptlrpc_at_recv_early_reply() should not modify req->rq_repmsg
2200 because it can be accessed by reply_in_callback() without the
2204 Frequency : only on Cray X2
2206 Description: X2 build failures
2207 Details : fix build failures on Cray X2.
2211 Description: xid & resent requests
2212 Details : Initialize RPC XID from clock at startup (randomly if clock is
2217 Description: quota recovery deadlock during mds failover
2218 Details : This patch includes att18982, att18236, att18237 in bz14840.
2220 1. fix osts hang when mds does failover with quotaon
2221 2. prevent watchdog storm when osts threads wait for the
2226 Description: kernel panic on racer
2227 Details : Do not access dchild->d_inode when IS_ERR(dchild) is true.
2229 Severity : enhancement
2231 Description: Add lustre_start utility to start or stop multiple Lustre servers
2236 Description: Lustre GPF in {:ptlrpc:ptlrpc_server_free_request+373}
2237 Details : In case of memory pressure, list_del() can be called twice on
2238 req->rq_history_list, causing a kernel oops.
2242 Description: (ptllnd_peer.c:557:kptllnd_peer_check_sends()) ASSERTION(!in_interrupt()) failed
2243 Details : fix stack overflow in the distributed lock manager by defering
2244 export eviction after a failed ast to the elt thread instead of
2245 handling it in the dlm interpret routine.
2247 Severity : enhancement
2249 Description: More exported tunables for mballoc
2250 Details : Add support for tunable preallocation window and new tunables for
2251 large/small requests
2255 Description: Detect corruption of block bitmap and checking for preallocations
2256 Details : Checks validity of on-disk block bitmap. Also it does better
2257 checking of number of applied preallocations. When corruption is
2258 found, it turns filesystem readonly to prevent further corruptions.
2262 Frequency : only for big-endian servers
2263 Description: Check for big-endian system while mounting fs with extents feature
2264 Details : Mounting a filesystem with extents feature will fail on big-endian
2265 systems since ext3-based ldiskfs is not supported on big-endian
2266 systems. Can be over-ridden with "bigendian_extents" mount option.
2270 Description: Excessive recovery window
2271 Details : With AT enabled, the recovery window can be excessively long (6000+
2272 seconds). To address this problem, we no longer use
2273 OBD_RECOVERY_FACTOR when extending the recovery window (the connect
2274 timeout no longer depends on the service time, it is set to
2275 INITIAL_CONNECT_TIMEOUT now) and clients report the old service
2276 time via pb_service_time.
2279 Descriptoin: Don't sync journal after every i/o
2280 Details : Implement write RPC replay to allow server replies for write RPCs
2281 before data is on disk. However, this feature is disabled by
2282 default since some issues leading to data corruptions have been
2283 found during recovery (e.g. bug 19128). This feature can be enabled
2284 by running the following command on the OSSs:
2285 lctl set_param obdfilter.*.sync_journal=0
2289 Description: Watchdog triggered on MDS failover
2290 Details : enable OBD_CONNECT_MDT flag when connecting from the MDS so that
2291 the OSTs know that the MDS "UUID" can be reused for the same export
2292 from a different NID, so we do not need to wait for the export to
2297 Descriptoin: Lustre detected file system corruption with inode out of bounds
2298 Details : don't update i_size on MDS_CLOSE for directories. This causes
2299 directory corruptions on the MDT.
2301 -------------------------------------------------------------------------------
2303 2008-05-26 Sun Microsystems, Inc.
2305 * Support for kernels:
2306 2.6.5-7.311 (SLES 9),
2307 2.6.9-67.0.7.EL (RHEL 4),
2308 2.6.16.54-0.2.5 (SLES 10),
2309 2.6.18-53.1.14.el5 (RHEL 5),
2310 2.6.22.14 vanilla (kernel.org)
2311 * Client support for unpatched kernels:
2312 (see http://wiki.lustre.org/index.php?title=Patchless_Client)
2313 2.6.16 - 2.6.22 vanilla (kernel.org)
2314 * Due to problems with nested symlinks and FMODE_EXEC (bug 12652),
2315 we do not recommend using patchless RHEL4 clients with kernels
2316 prior to 2.6.9-55EL (RHEL4U5).
2317 * Recommended e2fsprogs version: 1.40.7-sun1
2318 * Note that reiserfs quotas are disabled on SLES 10 in this kernel.
2319 * RHEL 4 and RHEL 5/SLES 10 clients behaves differently on 'cd' to a
2320 removed cwd "./" (refer to Bugzilla 14399).
2321 * A new quota file format has been introduced in 1.6.5.
2322 The format conversion from prior releases is handled transparently,
2323 but releases older than 1.4.12/1.6.5 will not understand this new
2324 format. The automatic format conversion can be avoided by running
2325 the following command on the MDS before upgrading:
2326 'tunefs.lustre --param="mdt.quota_type=ug1" $MDTDEV'.
2327 For more information, please refer to bugzilla 13904.
2331 Description: quota performance fix
2332 Details : quota data is written in journalled mode instead of ordered to
2333 increase performance
2337 Description: lfs support for human-readable quota grace time strings
2338 Details : lfs setquota -t and lfs quota -t represent quota grace times
2339 in "XXwXXdXXhXXmXXs" format instead of large values in seconds
2342 Frequency : always with o2ib 1.3 and sles10
2344 Description: fix build with SLES10 and o2ib v3.
2345 Details : sles10 uses diffrent name for Module.symver file but configure
2346 assume this file has same name on RHEL/SLES/vanila kernels.
2349 Frequency : very rare, if additional xattrs are used on kernels >= 2.6.12
2351 Description: MDS may lose file striping (and hence file data) in some cases
2352 Details : If there are additional extended attributes stored on the MDS,
2353 in particular ACLs, SELinux, or user attributes (if user_xattr
2354 is specified for the client mount options) then there is a risk
2355 of attribute loss. Additionally, the Lustre file striping
2356 needs to be larger than default (e.g. striped over all OSTs),
2357 and an additional attribute must be stored initially in the
2358 inode and then increase in size enough to be moved to the
2359 external attribute block (e.g. ACL growing in size) for file
2362 Severity : enhancement
2364 Description: add message levels for liblustreapi
2369 Description: MDT cannot be unmounted, reporting "Mount still busy"
2370 Details : Mountpoint references were being leaked during open reply
2371 reconstruction after an MDS restart. Drop mountpoint reference
2372 in reconstruct_open() and free dentry reference also.
2377 Description: fix for occasional failure case of -ENOSPC in recovery-small tests
2378 Details : Move the 'good_osts' check before the 'total_bavail' check. This
2379 will result in an -EAGAIN and in the exit call path we call
2380 alloc_rr() which will with increasing aggressiveness attempt to
2381 aquire precreated objects on the minimum number of required OSCs.
2385 Description: Use old size assignment to avoid deadlock
2386 Details : This reverts the changes in bugs 2369 and bug 14138 that introduced
2387 the scheduling while holding a spinlock. We do not need locking
2388 for size in ll_update_inode() because size is only updated from
2389 the MDS for directories or files without objects, so there is no
2390 other place to do the update, and concurrent access to such inodes
2391 are protected by the inode lock.
2395 Description: Use __u64 instead of int for valid bits
2399 Description: resolve "_IOWR redefined" build error on SLES10
2403 Description: dump the memory debugging after all modules are unloaded to
2404 suppress false negative in conf_sanity test 39
2408 Description: the recovery timer never expires
2409 Details : for new client connect request, the recovery timer should not be
2410 reset, otherwise recovery timer will never expired, if the old
2411 client never come. Only old client connect and first connection
2412 req should trigger recovery timer reset.
2416 Description: the min numbers of lproc stats are wrong
2417 Details : adding a new constant LC_MIN_INIT and use it for initialization
2421 Frequency : always with interactive lfs
2423 Description: Reinitialize optind to 0 so that interactive lfs works in all cases
2426 Frequency : with multiple concurrent readdir processes in same directory
2427 Bugzilla : 15406, 15169, 15175
2428 Description: misc fixes for directory readahead.
2429 Details : prevent previous statahead async RPC callback from processing the
2430 current "statahead_info", race condition between async RPC callback
2431 add dentry into dentry hash table and "ls" thread revalidate such
2432 dentry, statahead his/miss control for hidden items, and so on.
2434 Severity : enhancement
2436 Description: build kernel-ib packages for OFED 1.3 in our release cycle
2440 Description: incore types cleaning in quota code (with respect to 64-bit limits)
2441 Details : several u32 variables declarations are replaced with u64 declarations
2446 Description: fix SLES kernel versioning
2447 Details : the kernel version for our SLES 10 kernel did not include a "-"
2448 before the "smp" at the end. while this was not a problem in
2449 general, it did mean that software trying to use the kernel
2450 version to try to detect a vendor specific kernel would fail.
2451 this was most evident by the OFED build scripts.
2456 Description: Don't update lov_desc members until making sure they are valid
2457 Details : When updating lov_desc members via proc fs, need fix their
2458 validities before doing the real update.
2461 Frequency : very rare
2463 Description: don't put request into delay list while invalidate in flight.
2464 Details : ptlrpc_delay_request sometimes put in delay list while invalidate
2465 import in flight. this produce timeout for invalidate and sometimes
2466 can cause stale data.
2468 Severity : enhancement
2470 Description: Update kernel to SLES9 2.6.5-7.311.
2472 Severity : enhancement
2474 Description: Update kernel to RHEL4 2.6.9-67.0.7.
2479 Frequency : on PPC only
2480 Description: not convert ost objects for directory because it's not exist.
2481 Details : ll_dir_getstripe assume dirrectory has ost objects but this wrong.
2483 Severity : enhancement
2485 Description: Fix warnings with compile liblustre at sles10/rhel5 which have
2486 __u64 as usingied long long type.
2489 Frequency : rare, on shutdown
2491 Description: race process ast vs remove callback
2492 Details : removing callback before disconnect import open race with
2493 processing callback.
2495 Severity : enhancement
2497 Description: Update kernel to SLES9 2.6.5-7.311.
2499 Severity : enhancement
2501 Description: Files open for execute are not marked busy on SLES10
2502 Details : Add FMODE_EXEC to SLES10 SP1 server kernel series.
2504 Severity : enhancement
2506 Description: Add server support for vanilla-2.6.22.14.
2509 Frequency : occasional
2511 Description: Avoid lov_create() getting stuck in obd_statfs_rqset()
2512 Details : If an OST is down the MDS will hang indefinitely in
2513 obd_statfs_rqset() waiting for the statfs data. While for
2514 MDS QOS usage of statfs, it should not stuck in waiting.
2516 Severity : enhancement
2518 Description: Disable adaptive timeouts by default
2521 Frequency : on network error
2523 Description: panic with double free request if network error
2524 Details : mdc_finish_enqueue is finish request if any network error occuring,
2525 but it's true only for synchronus enqueue, for async enqueue
2526 (via ptlrpcd) this incorrect and ptlrpcd want finish request
2530 Frequency : rare, on recovery
2532 Description: read procfs can produce deadlock in some situation
2533 Details : Holding lprocfs lock with send rpc can produce block for destroy
2534 obd objects and this also block reconnect with -EALREADY.
2535 This isn't fix all lprocfs bugs - but make it rare.
2537 Severity : enhancement
2539 Description: Update kernel to RHEL5 2.6.18-53.1.14.el5.
2542 Frequency : frequent on X2 node
2544 Description: mdc_set_open_replay_data LBUG
2545 Details : Set replay data for requests that are eligible for replay.
2550 Description: lustre_mgs: operation 101 on unconnected MGS
2551 Details : When MGC is disconnected from MGS long enough, MGS will evict the
2552 MGC, and late on MGC cannot successfully connect to MGS and a lot
2553 of the error messages complaining that MGS is not connected.
2556 Frequency : rare, depends on device drivers and load
2558 Description: MDS or OSS nodes crash due to stack overflow
2559 Details : Code changes in 1.6.4 increased the stack usage of some functions.
2560 In some cases, in conjunction with device drivers that use a lot
2561 of stack, the MDS (or possibly OSS) service threads could overflow
2562 the stack. One change which was identified to consume additional
2563 stack has been reworked to avoid the extra stack usage.
2565 Severity : enhancement
2567 Description: Update to RHEL5 latest kernel-2.6.18-53.1.13.el5.
2569 Severity : enhancement
2571 Description: Update to SLES10 SP1 latest kernel-2.6.16.54-0.2.5.
2573 Severity : enhancement
2575 Description: Update to RHEL5 latest kernel-2.6.18-53.1.6.el5.
2577 Serverity : enhancement
2579 Description: Update RHEL4 kernel to 2.6.9-67.0.4.
2582 Frequency : rare on shutdown OST
2584 Description: Don't allow skipping OSTs if index has been specified.
2585 Details : Don't allow skipping OSTs if index has been specified, make locking
2586 in internal create lots better.
2591 Description: ASSERTION(!PageDirty(page)) failed
2592 Details : Wrong check could lead to an assertion failure under specific
2598 Description: LBUG in ptlrpc_check_set() bad phase ebc0de00
2599 Details : access to bitfield in structure is always rounded to long
2600 and this produce problem with not atomic change any bit.
2605 Description: Lustre make rpms failed.
2606 Details : Remove ldiskfs spec file to avoids rpmbuild be confused when
2607 builds Lustre rpms from tarball.
2609 Severity : enhancement
2611 Description: Update to SLES9 SP4 kernel-2.6.5-7.308.
2614 Frequency : rare on shutdown OST
2616 Description: If llog cancel was not send before clean_exports phase, this can
2617 produce deadlock in llog code.
2618 Details : If llog thread has last reference to obd and call class_import_put
2619 this produce deadlock because llog_cleanup_commit_master wait when
2620 last llog_commit_thread exited, but this never success because was
2621 called from llog_commit_thread.
2624 Frequency : only if OST index is skipped
2626 Description: NULL lov_tgts causing MDS oops
2627 Details : more safe checks for NULL lov_tgts for avoid oops.
2629 Severity : enhancement
2631 Description: Update to RHEL4 latest kernel-2.6.9-67.0.1.EL.
2633 Severity : enhancement
2635 Description: Update to RHEL5 latest kernel-2.6.18-53.1.4.el5.
2640 Description: make mgs_setparam() handle fsname containing dash
2641 Details : fsname containing a dash does not work with lctl conf_param
2643 Severity : enhancement
2645 Description: Update to RHEL4 Update-6 kernel-2.6.9-67.EL.
2648 Frequency : rare, in recovery and (or) destroy lovobjid file.
2650 Description: rewrite lov objid code.
2651 Details : Cleanup for lov objid code, remove scability problems and wrong
2652 locking. Fix sending last_id into ost.
2654 Severity : enhancement
2656 Description: Update to SLES10 SP1 latest kernel-2.6.16.54-0.2.3.
2658 Severity : enhancement
2660 Description: Update to RHEL5 Update-1 kernel 2.6.18-53.el5.
2661 Details : Use d_move_locked instead of __d_move.
2664 Frequency : rare, at shutdown
2666 Description: access already free / zero obd_namespace.
2667 Details : if client_disconnect_export was called without force flag set,
2668 and exist connect request in flight, this can produce access to
2669 NULL pointer (or already free pointer) when connect_interpret
2670 store ocd flags in obd_namespace.
2673 Frequency : only at startup
2675 Description: not alloc memory with spinlock held.
2676 Details : allocation memory with GFP_KERNEL can produce sleep deadlock,
2677 if any spinlock held.
2682 Description: lfs find does not continue on file error
2683 Details : Continue other files processing when a file/dir is absent.
2687 Description: Inconsistent usage of lustre_pack_reply()
2688 Details : Standardize the usage of lustre_pack_reply() such that it
2689 always generate a CERROR on failure.
2692 Frequency : very rare
2694 Description: Fix replay if there is an un-replied request and open
2695 Details : In some cases, older replay request will revert the
2696 mcd->mcd_last_xid on MDS which is used to record the client's
2697 latest sent request.
2699 Severity : enhancement
2701 Description: Update to RHEL5 kernel 2.6.18-8.1.15.el5.
2703 Severity : enhancement
2705 Description: Update to SLES10 SP1 kernel 2.6.16.53-0.16
2707 Severity : enhancement
2709 Description: Update to SLES9 kernel-2.6.5-7.287.3.
2711 Severity : enhancement
2713 Description: Update to RHEL4 kernel-2.6.9-55.0.12.EL.
2715 Severity : enhancement
2717 Description: Build SLES10 patchless client fails
2718 Details : The configure was broken by run ./configure with
2719 --with-linux-obj=.... argument for patchless client. When the
2720 configure use --with-linux-obj, the LINUXINCLUDE= -Iinclude
2721 can't search header adequately. Use absolute path such as
2722 -I($LINUX)/include instead.
2724 Severity : enhancement
2726 Description: Lustre Page Accounting
2727 Details : New macros for page alloc and free which enable accounting
2728 of page allocation of Lustre. Use percpu counters to store memory
2729 and page statistics.
2732 Frequency : only if debugging is disabled
2734 Description: LASSERT_{REQ,REP}SWAB macros are buggy
2735 Details : If SWAB_PARANOIA is disabled, the LASSERT_REQSWAB and
2736 LASSERT_REPSWAB macros become no-ops, which is incorrect. Drop
2737 these macros and replace them with their definitions instead.
2742 Description: interrupt oig_wait produce painc on resend.
2743 Details : brw_redo_request can be used for resend requests from ptlrpcd and
2744 private set, and this produce situation when rq_ptlrpcd_data not
2745 copyed to new allocated request and triggered LBUG on assert
2746 req->rq_ptlrpcd_data != NULL. But this member used only for wakeup
2747 ptlrpcd set if request is changed and can be safety changed to use
2750 Severity : enhancement
2752 Description: organize the server-side client stats on per-nid basis
2753 Details : Change the structure of stats under obdfilter and mds to
2761 The "uuid"s file would list the uuids of _active_ exports.
2762 And the clear entry is to clear all stats and stale nids.
2767 Description: Processes looping in ll_readdir() on Lustre clients finally causing
2768 a full node pseudo-hang
2769 Details : Concurrent access to the same directory from multiple clients with
2770 intensive file creation/removal can cause a client node to spin in
2771 ll_readdir(). i_version must be increased every time the lock is
2772 cancelled to ensure a revalidate is done.
2777 Description: touch file failed when fs is not full
2778 Details : OST in recovery should not be discarded by MDS in alloc_qos(),
2779 otherwise we can get ENOSP while fs is not full.
2782 Frequency : only for Cray XT3
2783 Bugzilla : 12829/13455
2784 Description: Changing primary group doesn't change the group lustre assigns to
2786 Details : When CRAY_XT3 is defined, the fsgid supplied by the client is
2787 overridden with the primary group provided by the group upcall,
2788 whereas the supplied fsgid can be trusted if it is in the list of
2789 supplementary groups returned by the group upcall.
2791 Severity : enhancement
2793 Description: Root Squash Functionality
2794 Details : Implementation of NFS-like root squash capability. Specifically,
2795 don't allow someone with root access on a client node to be able
2796 to manipulate files owned by root on a server node.
2798 Severity : enhancement
2800 Description: Slow trucate/writes to huge files at high offsets.
2801 Details : Directly associate cached pages to lock that protect those pages,
2802 this allows us to quickly find what pages to write and remove
2803 once lock callback is received.
2808 Description: Too many locks accumulating on client during NFS usage
2809 Details : mds_open improperly used accmode to find out access mode to a
2810 file. Also mdc_intent_lock logic to find out if we already have
2811 lock similar to just received was flawed since introduction of
2812 skiplists - locks are now added to the front of the granted
2817 Description: Hit ASSERTION(obd->obd_stopping == 1) failed in some setup failed
2819 Details : In obd setup failure handler, obd_stopping will not necessarily to
2820 be 1, and obd_set_up should also be checked to make sure whether
2821 obd is completely setup.
2823 Severity : enhancement
2825 Description: Allow masking D_WARNING, D_ERROR messages from console
2826 Details : Console messages can now be disabled via lnet.printk.
2831 Description: User code with malformed file open parameter crashes client node
2832 Details : Before packing join_file req, all the related reference should be
2833 checked carefully in case some malformed flags cause fake join_file
2839 Description: shrink/enlarge qunit size when needed; fix the problem of coarse
2840 grain of quota doing harm to quota's accuracy
2841 Details : qunit size will be changed when quota limitation is too low/high;
2842 record the pending quota write in order to get more accureate
2843 quota; delete the patch for bug12588, which is unnecessary when
2844 this patch is landed. This bug also contains fixes for bug 14526,
2845 14299, 14601 and 13794.
2849 Description: LDLM_ENQUEUE races with LDLM_CP_CALLBACK
2850 Details : ldlm_completion_ast() assumes that a lock is granted when the req
2851 mode is equal to the granted mode. However, it should also check
2852 that LDLM_FL_CP_REQD is not set.
2856 Description: Heavy nfs access might result in deadlocks
2857 Details : After ELC code landed, it is now improper to enqueue any mds
2858 locks under och_sem, because enqueue might want to decide to
2859 cancel open locks for same inode we are holding och_sem for.
2863 Description: 35% write performance drop with ldiskfs2 when quotas are on
2864 Details : Enable ext3 journalled quota by default to improve performance
2865 when quotas are turned on.
2869 Description: Client eviction while running blogbench
2870 Details : A lot of unlink operations with concurrent I/O can lead to a
2871 deadlock causing evictions. To address the problem, the number of
2872 oustanding OST_DESTROY requests is now throttled to
2873 max_rpcs_in_flight per OSC and LDLM_FL_DISCARD_DATA blocking
2874 callbacks are processed in priority.
2877 Frequency : RHEL4 only
2879 Description: mkfs is very slow on IA64/RHEL4
2880 Details : A performance regression has been discovered in the MPT Fusion
2881 driver between versions 3.02.73rh and 3.02.99.00rh. As a
2882 consequence, we have downgraded the MPT Fusion driver in the RHEL4
2883 kernel from 3.02.99.00 to 3.02.73 until this problem is fixed.
2886 Frequency : PPC/PPC64 only
2888 Description: conflicts between asm-ppc64/types.h and lustre_types.h
2889 Details : fix duplicated definitions between asm-ppc64/types.h and
2890 lustre_types.h on PPC.
2893 Frequency : PPC/PPC64 only
2895 Description: asm-ppc/segment.h does not exist
2896 Details : fix compile issue on PPC.
2900 Description: data checksumming impacts single node performance
2901 Details : add support for several checksum algorithms. Currently, CRC32 and
2902 Adler-32 are supported. The checksum type can be changed on the fly
2903 through /proc/fs/lustre/osc/*/checksum_type.
2907 Description: use adler32 for page checksums
2908 Details : when available, use the Adler-32 algorithm instead of CRC32 for
2913 Description: better handle error messages in extents code
2915 Severity : enhancement
2917 Description: SNMP support enhancement
2918 Details : Adding total number of sampled request for an MDS node in snmp
2921 Severity : enhancement
2923 Description: Optimize ldlm waiting list processing for PR extent locks
2924 Details : When processing waiting list for read extent lock and meeting read
2925 lock that is same or wider to it that is not contended, skip
2926 processing rest of the list and immediatelly return current
2927 status of conflictness, since we are guaranteed there are no
2928 conflicting locks in the rest of the list.
2932 Description: Time out and refuse to reconnect
2933 Details : When the failover node is the primary node, it is possible
2934 to have two identical connections in imp_conn_list. We must
2935 compare not conn's pointers but NIDs, otherwise we
2936 can defeat connection throttling.
2940 Description: Client not clear own cache if answer to reconnect is lost.
2941 Details : Client gets evicted from server. Now client also thinks it is
2942 disconnected (or gets ENOTCONN on its operation) and decides to
2943 reconnect. Server receives reconnect message, but cannot find
2944 export. New export is created that is fully valid (new cookie!),
2945 but reply is lost and not reported to client. Client reconnects
2946 again and gets back a just-created connection, but it is not new
2947 so client thinks it was not evicted and does not do recovery.
2951 Description: Detect stride IO mode in read-ahead
2952 Details : When a client does stride read, read-ahead should detect that and
2953 read-ahead pages according to the detected stride pattern.
2957 Description: build for x2 fails
2958 Details : fix compile issue on Cray systems.
2960 Severity : enhancement
2962 Description: implement readv/writev APIs(aio_read/aio_writes in newer kernels)
2963 Details : This greatly improves speed of NFS writes on 2.6 kernels.
2966 Frequency : only on PPC/SLES10
2968 Description: "BITS_PER_LONG is not 32 or 64" in linux/idr.h
2969 Details : On SLES10/PPC, fs.h includes idr.h which requires BITS_PER_LONG to
2970 be defined. Add a hack in mkfs_lustre.c to work around this compile
2975 Description: LASSERT on MDS when client holding flock lock dies
2976 Details : ldlm pool logic depends on number of granted locks equal to
2977 number of released locks which is not true for flock locks, so
2978 just exclude such locks from consideration.
2982 Description: MDS deadlock with many ll_sync_lov threads and I/O stalled
2983 Details : Use fsfilt_sync() for both the whole filesystem sync and
2984 individual file sync to eliminate dangerous inode locking
2985 with I_LOCK that can lead to a deadlock.
2989 Description: Update an obsolete wirecheck.c generator
2990 Details : Update wirecheck.c/wirehdr.c and regenerate wiretest.c
2994 Description: Client can panic on open sometimes
2995 Details : It is possible that we try to free already freed request in
2996 ll_file_open in some error cases when we send request from
3001 Description: performance in 1.6.3
3002 Details : Force q->max_phys_segments to MAX_PHYS_SEGMENTS on SLES10 to be
3003 sure that 1MB requests are not fragmented by the block layer.
3007 Description: LDLM soft lockups - improvement
3008 Details : It is be possible to send the lock handle along with each read
3009 or write request because the client is already doing a lock match
3010 itself so there isn't any reason the OST should have to re-do that
3016 Description: lfs quota fails with deactivated OSTS
3017 Details : With this patch, three improvements are included:
3018 1. detete the softlimit in mds and osts when use "lfs quota".
3019 2. display the inaccurate data in the output of "lfs quota".
3020 3. try to get quota info when "lfs quota" is executed.
3025 Description: Extent locks not granted with no conflicts sometimes.
3026 Details : When race occurs in glimpse handler and nothing is returned,
3027 we do not reprocess the queue after lock cancel, and that leads
3028 to a stall until next activity on a resource
3031 Frequency : failover with quotaon
3033 Description: during mds failovers with quota on, OSTs got into deadlock state
3034 and causing dumpstack.
3035 Details : for every quota slave, at any time, there is only one quota req
3036 is sent to quota master for every uid/gid. Before that quota req
3037 returns, all the thread relative to the same uid/gid will wait.
3038 So if the quota req is lost because mds failovers or any other
3039 reasons, this bug will be hit. Now, dqacq_interpret() will handle
3040 quota reqs who time out.
3042 Severity : enhancement
3045 Description: when quota slave checks if quota is enough, there is an unnecessary
3047 Details : place this wait on necessary place instead of always waiting.
3049 --------------------------------------------------------------------------------
3051 2007-12-07 Cluster File Systems, Inc. <info@clusterfs.com>
3053 * Support for kernels:
3054 2.6.5-7.286 (SLES 9),
3055 2.6.9-55.0.9.EL (RHEL 4),
3056 2.6.16.53-0.8 (SLES 10),
3057 2.6.18-8.1.14.el5 (RHEL 5),
3058 2.6.18.8 vanilla (kernel.org)
3059 * Client support for unpatched kernels:
3060 (see http://wiki.lustre.org/index.php?title=Patchless_Client)
3061 2.6.16 - 2.6.22 vanilla (kernel.org)
3062 * Due to recently discovered recovery problems, we do not recommend
3063 using patchless RHEL 4 clients with this or any earlier release.
3064 * Recommended e2fsprogs version: 1.40.2-cfs1
3065 * Note that reiserfs quotas are disabled on SLES 10 in this kernel.
3068 Frequency : occasional
3070 Description: MDS hang or stay in waiting lock
3071 Details : If client receive lock with CBPENDING flag ldlm need send lock
3072 cancel as separate rpc, to avoid situation when cancel request
3073 can't processed due all i/o threads stay in wait lock.
3076 Frequency : occasional
3078 Description: Do not fail import if osc_interpret_create gets -EAGAIN
3079 Details : If osc_interpret_create got -EAGAIN it immediately exits and
3080 wakeup oscc_waitq. After wakeup oscc_wait_for_objects call
3081 oscc_has_objects and see OSC has no objests and call
3082 oscc_internal_create to resend create request.
3084 Severity : enhancement
3086 Description: Update kernel patches for SLES10 2.6.16.53-0.8.
3087 Details : Update which_patch & target file for SLES10 latest kernel.
3089 Severity : enhancement
3091 Description: add --type and --size parameters to lfs find
3092 Details : Enhance lfs find by adding filetype and filesize parameters. Also
3093 multiple OBDs can now be specified for the --obd option.
3095 Severity : enhancement
3097 Description: eliminate client locks in face of contention
3098 Details : file contention detection and lockless i/o implementation
3099 for contended files.
3101 Severity : enhancement
3103 Description: Remove client patches from SLES 10 kernel.
3104 Details : This causes SLES 10 clients to behave as patchless clients
3105 even on a Lustre-patched (server) kernel.
3107 Severity : enhancement
3109 Description: use i_size_read and i_size_write in 2.6 port
3110 Details : replace inode->i_size access with i_size_read/write()
3113 Frequency : when removing large files
3115 Description: scheduling issue during removal of large Lustre files
3116 Details : Don't take the BKL in fsfilt_ext3_setattr() for 2.6 kernels.
3117 It causes scheduling issues when removing large files (17TB in the
3123 Description: 1.4.11 Can't handle directories with stripe set and extended ACLs
3124 Details : Impossible (EPROTO is returned) to access a directory that has a
3125 non-default striping and ACLs.
3128 Frequency : only on ppc
3130 Description: /proc/fs/lustre/devices broken on ppc
3131 Details : The patch as applied to 1.6.2 doesn't look correct for all arches.
3132 We should make sure the type of 'index' is loff_t and then cast
3133 explicitly as needed below. Do not assign an explicitly cast
3137 Frequency : only for rhel5
3139 Description: Kernel patches update for RHEL5 2.6.18-8.1.10.el5.
3140 Details : Modify the target file & which_kernel.
3143 Frequency : if the uninit_groups feature is enabled on ldiskfs
3145 Description: e2fsck reports "invalid unused inodes count"
3146 Details : If a new ldiskfs filesystem is created with the "uninit_groups"
3147 feature and only a single inode is created in a group then the
3148 "bg_unused_inodes" count is incorrectly updated. Creating a
3149 second inode in that group would update it correctly.
3154 Description: buffer overruns could theoretically occur
3155 Details : llapi_semantic_traverse() modifies the "path" argument by
3156 appending values to the end of the origin string, and a buffer
3157 overrun may occur. Adding buffer overrun check in liblustreapi.
3159 Severity : enhancement
3161 Description: Add jbd statistics patch for RHEL5 and 2.6.18-vanilla.
3164 Frequency : only if filesystem is inconsistent
3166 Description: handle "serious error: objid * already exists" more gracefully
3167 Details : If LAST_ID value on disk is smaller than the objects existing in
3168 the O/0/d* directories, it indicates disk corruption and causes an
3169 LBUG(). If the object is 0-length, then we should use the existing
3170 object. This will help to avoid a full fsck in most cases.
3172 Severity : enhancement
3174 Description: Kernel patches update for RHEL4 2.6.9-55.0.6.
3175 Details : Modify vm-tunables-rhel4.patch.
3177 Severity : enhancement
3179 Description: Kernel config for 2.6.18-vanilla.
3180 Details : Modify targets/2.6-vanilla.target.in.
3181 Add config file kernel-2.6.18-2.6-vanilla-i686.config.
3182 Add config file kernel-2.6.18-2.6-vanilla-i686-smp.config.
3183 Add config file kernel-2.6.18-2.6-vanilla-x86_64.config.
3184 Add config file kernel-2.6.18-2.6-vanilla-x86_64-smp.config.
3187 Frequency : occasional
3189 Description: improve handling recoverable errors
3190 Details : If request processed with error which can be recoverable on server
3191 request should be resend, otherwise page released from cache and
3197 Description: Kernel patches update for RHEL5 2.6.18-8.1.14.el5.
3198 Details : Modify target file & which_patch.
3199 A flaw was found in the IA32 system call emulation provided
3200 on AMD64 and Intel 64 platforms. An improperly validated 64-bit
3201 value could be stored in the %RAX register, which could trigger an
3202 out-of-bounds system call table access. An untrusted local user
3203 could exploit this flaw to run code in the kernel
3204 (ie a root privilege escalation). (CVE-2007-4573).
3208 Description: change order of libsysio includes
3209 Details : '#include sysio.h' should always come before '#include xtio.h'
3211 Severity : enhancement
3213 Description: adapt the lustre_config script to support the upgrade case
3214 Details : Add "-u" option for lustre_config script to support upgrading 1.4
3215 server targets to 1.6 in parallel.
3220 Description: To avoid grant space > avaible space when the disk is almost
3221 full. Without this patch you might see the error "grant XXXX >
3222 available" or some LBUG about grant, when the disk is almost
3224 Details : In filter_check_grant, for non_grant cache write, we should
3225 check the left space by if (*left > ungranted + bytes), instead
3226 of (*left > ungranted), because only we are sure the left space
3227 is enough for another "bytes", then the ungrant space should be
3228 increase. In client, we should update cl_avail_grant only there is
3229 OBD_MD_FLGRANT in the reply.
3234 Description: Update RHEL 4 kernel to fix local root privilege escalation.
3235 Details : Update to the latest RHEL 4 kernel to fix the vulnerability
3236 described in CVE-2007-4573. This problem could allow untrusted
3237 local users to gain root access.
3240 Frequency : when using O_DIRECT and quotas
3242 Description: Incorrect file ownership on O_DIRECT output files
3243 Details : block usage reported by 'lfs quota' does not take into account
3244 files that have been written with O_DIRECT.
3248 Description: (rw.c:1323:ll_read_ahead_pages()) ASSERTION(page_idx > ria->ria_stoff) failed
3249 Details : Once the unmatched stride IO mode is detected, shrink the stride-ahead
3250 window to 0. If it does hit cache miss, and read-pattern is still
3251 stride-io mode, does not reset the stride window, but also does not
3252 increase the stride window length in this case.
3254 --------------------------------------------------------------------------------
3256 2007-09-27 Cluster File Systems, Inc. <info@clusterfs.com>
3258 * Support for kernels:
3259 2.6.5-7.286 (SLES 9),
3260 2.6.9-55.0.2.EL (RHEL 4),
3261 2.6.16.46-0.14 (SLES 10),
3262 2.6.18-8.1.8.el5 (RHEL 5),
3263 2.6.18.8 vanilla (kernel.org)
3264 * Client support for unpatched kernels:
3265 (see http://wiki.lustre.org/index.php?title=Patchless_Client)
3266 2.6.16 - 2.6.21 vanilla (kernel.org)
3267 * Due to recently discovered recovery problems, we do not recommend
3268 using patchless RHEL 4 clients with this or any earlier release.
3269 * Recommended e2fsprogs version: 1.40.2-cfs1
3270 * Note that reiserfs quotas are disabled on SLES 10 in this kernel.
3274 Description: Fix errors in lfs documentation
3275 Details : Fixes man pages
3277 Severity : enhancement
3279 Description: Adaptive timeouts
3280 Details : RPC timeouts adapt to changing server load and network
3281 conditions to reduce resend attempts and improve recovery time.
3283 Severity : enhancement
3285 Description: llapi_file_create() does not allow some changes
3286 Details : add llapi_file_open() that allows specifying the file creation
3287 mode and open flags, and also returns an open file handle.
3289 Severity : enhancement
3291 Description: df doesn't work properly if diskfs blocksize != 4K
3292 Details : Choose biggest blocksize of OST's as the LOV's blocksize.
3295 Frequency : always on directories with default striping set
3297 Description: lfs find on -1 stripe looping in lsm_lmm_verify_common()
3298 Details : Avoid lov_verify_lmm_common() on directory with -1 stripe count.
3300 Severity : enhancement
3302 Description: merge and cleanup kernel patches.
3303 Details : Remove mnt_lustre_list in vfs_intent-2.6-rhel4.patch.
3305 Severity : enhancement
3307 Description: RedHat Update kernel for RHEL5
3308 Details : Kernel config file for RHEL5.
3310 Severity : enhancement
3312 Description: OSS needs mutliple precreate threads
3313 Details : Add ability to start more than one create thread per OSS.
3316 Frequency : only with quota on the root user
3318 Description: mds_obd_create error creating tmp object
3319 Details : When the user sets quota on root, llog will be affected and can't
3320 create files and write files.
3323 Frequency : Always on ia64 patchless client, and possibly others.
3325 Description: Add EXPORT_SYMBOL check for node_to_cpumask symbol.
3326 Details : This allows the patchless client to be loaded on architectures
3327 without this export.
3332 Description: disorder of journal start and llog_add cause deadlock.
3333 Details : in llog_origin_connect, journal start should happen before
3334 llog_add keep the same order as other functions to avoid
3337 Severity : enhancement
3339 Description: RedHat Update kernel for RHEL5
3340 Details : Modify the kernel config file more closer RHEL5.
3342 Severity : enhancement
3344 Description: Build failure against Centos5 (RHEL5)
3345 Details : Define PAGE_SIZE when it isn't present.
3348 Frequency : occasionally when using NFS
3350 Description: "ll_intent_file_open()) lock enqueue: err: -13" with nfs
3351 Details : with NFS, the anon dentry's parent was set to itself in
3352 d_alloc_anon(), so in MDS, we use rec->ur_fid1 to find the
3353 corresponding dentry other than use rec->ur_name.
3355 Severity : enhancement
3357 Description: enable data checksumming by default
3358 Details : enable checksum by default, allow --disable-checksum
3359 configure option and "-o nochecksum" mount option. Checksums
3360 can also be disabled at runtime via $LPROC/osc/*/checksum_pages.
3363 Frequency : Occasionally with failover
3365 Description: Client eviction due to failover config
3366 Details : after a connection loss, the lustre client should attempt to
3367 reconnect to the last active server first before trying the
3368 other potential connections.
3370 Severity : enhancement
3372 Description: client-side metadata stat-ahead during readdir(directory readahead)
3373 Details : perform client-side metadata stat-ahead when the client detects
3374 readdir and sequential stat of dir entries therein
3379 Description: when mds and osts use different quota unit(32bit and 64bit),
3380 quota will be released repeatly.
3381 Details : void sending multiple quota reqs to mds, which will keep the
3382 status between the reqs.
3385 Frequency : only with liblustre clients on XT3
3387 Description: evictions taking too long
3388 Details : allow llrd to evict clients directly on OSTs
3392 Description: osts not allocated evenly to files
3393 Details : change the condition to increase offset_idx
3395 Severity : enhancement
3397 Description: self-adjustable client's lru lists
3398 Details : use adaptive algorithm for managing client cached locks lru
3399 lists according to current server load, other client's work
3400 pattern, memory activities, etc. Both, server and client
3401 side namespaces provide number of proc tunables for controlling
3406 Description: rewrite ext2-derived code in llite/dir.c and obdclass/uuid.c
3407 Details : rewrite inherited code (uuid parsing code from ext2 utils and
3408 readdir code from ext3) from scratch preserving functionality.
3412 Description: Only those disconnect error should be returned by rq_status.
3413 Details : In open/enqueue processs, Some errors, which will cause client
3414 disconnected, should be returned by rq_status, while other
3415 errors should still be returned by intent, then mdc or llite will
3418 Severity : enhancement
3420 Description: Tune the kernel for good SCSI performance.
3421 Details : Set the value of /sys/block/{dev}/queue/max_sectors_kb
3422 to the value of /sys/block/{dev}/queue/max_hw_sectors_kb
3426 Frequency : Always for filesystems larger than 2TB on 32-bit systems.
3427 Bugzilla : 13547, 13627
3428 Description: Data corruption for OSTs that are formatted larger than 2TB
3430 Details : When generating the bio request for lustre file writes the
3431 sector number would overflow a temporary variable before being
3432 used for the IO. The data reads correctly from Lustre (which
3433 will overflow in a similar manner) but other file data or
3434 filesystem metadata may be corrupted in some cases.
3438 Description: "lfs find -obd UUID" prints directories
3439 Details : "lfs find -obd UUID" will return all directory names instead
3440 of just file names. It is incorrect because the directories
3441 do not reside on the OSTs.
3444 Bugzilla : 13596, 13614
3445 Description: MDS hang after unclean shutdown of lots of clients
3446 Details : Never resend AST requests.
3450 Frequency : Always, for kernels after 2.6.16
3451 Description: Fix warning idr_remove called for id=.. which is not allocated.
3452 Details : Last kernels save old s_dev before kill super and not allow
3453 to restore from callback - restore it before call kill_anon_super.
3457 Description: Fix error on 'ls .' at the top of the Lustre mount.
3458 Details : Don't revalidate dentry if it is a root dentry.
3460 --------------------------------------------------------------------------------
3462 2007-08-27 Cluster File Systems, Inc. <info@clusterfs.com>
3464 * Support for kernels:
3465 2.6.5-7.286 (SLES 9),
3466 2.6.9-55.0.2.EL (RHEL 4),
3467 2.6.16.46-0.14 (SLES 10),
3468 2.6.18-8.1.8.el5 (RHEL 5),
3469 2.6.18.8 vanilla (kernel.org)
3470 * Client support for unpatched kernels:
3471 (see http://wiki.lustre.org/index.php?title=Patchless_Client)
3472 2.6.16 - 2.6.22 vanilla (kernel.org)
3473 * Due to recently discovered recovery problems, we do not recommend
3474 using patchless RHEL 4 clients with this or any earlier release.
3475 * Recommended e2fsprogs version: 1.39.cfs8
3476 * Note that reiserfs quotas are disabled on SLES 10 in this kernel.
3481 Description: block reactivating mgc import until all deactivates complete
3482 Details : Fix race when failing back MDT/MGS to itself (testing)
3484 Severity : enhancement
3486 Description: lfs setstripe enhancement
3487 Details : Make lfs setstripe understand 'k', 'm' and 'g' for stripe size.
3490 Frequency : mds/oss recovery
3492 Description: llog ctxt is refrenced after it has been freed.
3493 Details : llog ctxt refcount was added to avoide the race between ctxt free
3494 and llog recovery process. Each llog user must hold ctxt refcount
3495 before it access the llog. And the llog ctxt can only be freed
3496 when its refcount is zero.
3498 Severity : enhancement
3500 Description: randomly memory allocation failure util
3501 Details : Make lustre randomly failed allocating memory for testing purpose.
3504 Frequency : only for SLES10
3506 Description: Update kernel patch for SLES10 SP1
3507 Details : Add patch blkdev_tunables-2.6-sles10.patch to 2.6-sles10.series.
3509 Severity : enhancement
3511 Description: omit set fsid for export NFS
3512 Details : fix set/restore device id for avoid EMFILE error and mark lustre fs
3513 as FS_REQUIRES_DEV for avoid problems with generate fsid.
3516 Frequency : after network failures
3518 Description: Add sync option to mount_lustre.c
3519 Details : Client loses data written to lustre after a network interruption.
3521 Severity : enhancement
3523 Description: Error message improvement.
3524 Details : Merging of two LCONSOLE_ERROR_MSG into one.
3526 Severity : enhancement
3528 Description: don't use GFP_* in generic Lustre code.
3529 Details : Use cfs_alloc_* functions and CFS_* flags for code portability.
3531 Severity : enhancement
3533 Description: obdclass is limited by single OBD_ALLOC(idarray)
3534 Details : replace OBD_ALLOC/OBD_FREE with OBD_VMALLOC/OBD_VFREE
3536 Severity : enhancement
3538 Description: updated patchess for new RHEL4 kernel
3539 Details : Fixed ext3-unlink-race.patch per Kalpak's comment.
3541 Severity : enhancement
3543 Description: warnings with build patchless client with vanila 2.6.19 and up
3544 Details : change old ctl_table style and replace ctl_table/ctl_table_header
3545 with cfs_sysctl_table_t/cfs_sysctl_table_header_t
3547 Severity : enhancement
3549 Description: O_DIRECT bypasses client statistics.
3550 Details : When running with O_DIRECT I/O, neither the client rpc_stats nor
3551 read_ahead_stats were updated. Copied stats section from
3552 osc_send_oap_rpc() into async_internal().
3555 Frequency : only for Cray XT3
3557 Description: peer credits not enough on many OST per OSS systems.
3558 Details : Use new lnet way to add credits as we need those for pings and ASTs
3561 Frequency : only with liblustre
3563 Description: Liblustre is not releasing flock locks on file close.
3564 Details : Release flock locks on file close.
3567 Frequency : only for RHEL4
3569 Description: Update kernel patches for kernel-2.6.9-55.0.2.EL
3570 Details : Remove inode-nr_unused-2.6.9-rhel4.patch from 2.6-rhel4.series
3571 Update target file and kernel config.
3575 Description: lustre support for RHEL5
3576 Details : Add support for RHEL5.
3581 Description: ASSERTION(export != NULL) failed in target_handle_connect
3582 Details : Assetion hit is result of rare race between disconnect and connet
3583 to same nid. target_handle_connect found old connect cockie and
3584 tried to reconnect, but can't find export for this cockie.
3589 Description: umount blocks forever on error
3590 Details : In result of wrong using obd_no_recov and obd_force flags client
3591 can hand if cancel or some other requests is lost.
3594 Frequency : Only for SLES
3596 Description: sanity_quota fail test_1
3597 Details : There are multiple occurences of $TSTUSR in SLES's /etc/group
3598 file, which makes TSTID[2] inunique.
3600 Severity : enhancement
3602 Description: Kernel patches for SLES9 2.6.5-7.286 kernel
3603 Details : Update target/ChangeLog/which_patch .
3605 Severity : enhancement
3607 Description: jbd statistics
3608 Details : Port older jbd statistics patch for sles10
3613 Description: lvbo_init failed for resource with missing objects.
3614 Details : Fix returning error if we do stat for file with missing/corrupted
3615 objects and i_size set to all sum of size of all avaible objects.
3616 if we to truncate/write to missing object - it is recreated.
3621 Description: Oops in read and write path when failing to allocate lock.
3622 Details : Check if lock allocation failed and return error back.
3625 Frequency : When flocks are used.
3627 Description: assertion failure in ldlm_cli_enquque_fini for non NULL lock.
3628 Details : Flock locks might destroy just granted lock if it could be merged
3629 with another existing flock, this is done in completion handler,
3630 so teach ldlm_cli_enquque_fini that this is a valid case for
3636 Description: reply_lock_interpret crash due to race with it and lock cancel.
3637 Details : Do not replay locks that are being cancelled. Do not reference
3638 locks by their address during replay, just by their handle.
3641 Frequency : only with deactivated OSTs
3643 Description: lstripe command fails for valid OST index
3644 Details : The stripe offset is compared to 'lov->desc.ld_tgt_count'
3645 instead of lov->desc.ld_active_tgt_count.
3647 Severity : enhancement
3649 Description: Build failure against Centos5 (RHEL5)
3650 Details : Use getpagesize() instead of PAGE_SIZE.
3652 --------------------------------------------------------------------------------
3654 2007-07-30 Cluster File Systems, Inc. <info@clusterfs.com>
3656 * Support for kernels:
3657 2.6.5-7.283 (SLES 9),
3658 2.6.9-55.EL (RHEL 4),
3659 2.6.16.46-0.14 (SLES 10),
3660 2.6.18.8 vanilla (kernel.org)
3661 * Client support for unpatched kernels:
3662 (see http://wiki.lustre.org/index.php?title=Patchless_Client)
3663 2.6.16 - 2.6.22 vanilla (kernel.org)
3664 * Due to recently discovered recovery problems, we do not recommend
3665 using patchless RHEL 4 clients with this or any earlier release.
3666 * Recommended e2fsprogs version: 1.39.cfs8
3667 * Note that reiserfs quotas are disabled on SLES 10 in this kernel.
3668 * Starting with this release, the ldiskfs backing filesystem required
3669 by Lustre is now in its own package, lustre-ldiskfs. This package
3670 should be installed. It is versioned separately from Lustre and
3671 may be released separately in future.
3676 Description: server LBUG when shutting down
3677 Details : Block umount forever until the mount refcount is zero rather
3678 than giving up after an arbitrary timeout.
3680 Severity : enhancement
3682 Description: add optional extra BUILD_VERSION info
3683 Details : add a new environment variable (namely LUSTRE_VERS) which allows
3684 to override the lustre version.
3687 Frequency : 2.6.18 servers only
3689 Description: ll_kern_mount() doesn't release the module reference
3690 Details : The ldiskfs module reference count never drops down to 0
3691 because ll_kern_mount() doesn't release the module reference.
3696 Description: server LBUG when using old ost_num_threads parameter
3697 Details : Accept the old ost_num_threads parameter but warn that it
3698 is deprecated, and fix an off-by-one error that caused an LBUG.
3703 Description: Transient SCSI error results in persistent IO issue
3704 Details : iobuf->dr_error is not reinitialized to 0 between two
3708 Frequency : sometimes when underlying device returns I/O errors
3710 Description: OSTs not going read-only during write failures
3711 Details : OSTs are not remounted read-only when the journal commit threads
3712 get I/O errors because fsfilt_ext3 calls journal_start/stop()
3713 instead of the ext3 wrappers.
3716 Frequency : only with 10000 clients or more
3718 Description: poor connect scaling with increasing client count
3719 Details : Don't run filter_grant_sanity_check for more than 100 exports
3720 to improve scaling for large numbers of clients.
3723 Frequency : SLES10 only
3725 Description: sanity-quota.sh quotacheck failed: rc = -22
3726 Details : Quotas cannot be enabled on SLES10.
3729 Frequency : liblustre clients only
3731 Description: getdirentries does not give error when run on compute nodes
3732 Details : getdirentries does not fail when the size specified as an argument
3733 is too small to contain at least one entry
3735 Severity : enhancement
3737 Description: Add LNET router traceability for debug purposes
3738 Details : If a checksum failure occurs with a router as part of the
3739 IO path, the NID of the last router that forwarded the bulk data
3740 is printed so it can be identified.
3745 Description: OST "spontaneously" evicts client; client has imp_pingable == 0
3746 Details : Due to a race condition, liblustre clients were occasionally
3747 evicted incorrectly.
3749 Severity : enhancement
3751 Description: lfs setstripe use optional parameters instead of postional
3754 Severity : enhancement
3756 Description: Nanosecond timestamp support for ldiskfs
3757 Details : The on-disk ldiskfs filesystem has added support for nanosecond
3758 resolution timestamps. There is not yet support for this at
3759 the Lustre filesystem level.
3762 Frequency : during server recovery
3764 Description: MDS failing to send precreate requests due to OSCC_FLAG_RECOVERING
3765 Details : request with rq_no_resend flag not awake l_wait_event if they get a
3769 Frequency : nfs export on patchless client
3771 Description: connectathon hang when test nfs export over patchless client
3772 Details : Disconnected dentry cannot be found with lookup, so we do not need
3773 to unhash it or make it invalid
3776 Description: fix llapi_lov_get_uuids() to allow many OSTs to be returned
3777 Details: : Change llapi_lov_get_uuids() to read the UUIDs from /proc instead
3778 of using an ioctl. This allows lfsck for > 160 OSTs to succeed.
3783 Description: open req refcounting wrong on reconnect
3784 Details : If reconnect happened between getting open reply from server and
3785 call to mdc_set_replay_data in ll_file_open, we will schedule
3786 replay for unreferenced request that we are about to free.
3787 Subsequent close will crash in variety of ways.
3788 Check that request is still eligible for replay in
3789 mdc_set_replay_data().
3794 Description: disable writes to filesystem when reading health_check file
3795 Details : the default for reading the health_check proc file has changed
3796 to NOT do a journal transaction and write to disk, because this
3797 can cause reads of the /proc file to hang and block HA state
3798 checking on a healthy but otherwise heavily loaded system. It
3799 is possible to return to the previous behaviour during configure
3800 with --enable-health-write.
3802 Severity : enhancement
3804 Description: 64-bit inode version
3805 Details: : Add a on-disk 64-bit inode version for ext3 to track changes made
3806 to the inode. This will be required for version-based recovery.
3811 Description: MDS fails to start if a duplicate client export is detected
3812 Details : in some rare cases it was possible for a client to connect to
3813 an MDS multiple times. Upon recovery the MDS would detect this
3814 and fail during startup. Handle this more gracefully.
3816 Severity : enhancement
3818 Description: Add -o localflock option to simulate old noflock behaviour.
3819 Details : This will achieve local-only flock/fcntl locks coherentness.
3824 Description: log_commit_thread vs filter_destroy race leads to crash
3825 Details : Take import reference before releasing llog record semaphore
3830 Description: Wrong request locking in request set processing
3831 Details : ptlrpc_check_set wrongly uses req->rq_lock for proctect add to
3832 imp_delayed_list, in this place should be used imp_lock.
3835 Frequency : when reconnecting
3837 Description: Grant leak when OSC reconnect to OST
3838 Details : When osc reconnect ost, OST(filter) should check whether it
3839 should grant more space to client by comparing fed_grant and
3840 cl_avail_grant, and return the granted space to client instead
3841 of "new granted" space, because client will call osc_init_grant
3842 to update the client grant space info.
3845 Frequency : when client reconnects to OST
3847 Description: Grant leak when OSC does a resend and replays bulk write
3848 Details : When osc reconnect to OST, OST(filter) should clear grant info of
3849 bulk write request, because the grant info while be sync between
3850 OSC and OST when reconnect, and we should ignore the grant info
3851 these of resend/replay write req.
3856 Description: Grant space more than avaiable space sometimes.
3857 Details : When then OST is about to be full, if two bulk writing from
3858 different clients came to OST. Accord the avaliable space of the
3859 OST, the first req should be permitted, and the second one
3860 should be denied by ENOSPC. But if the seconde arrived before
3861 the first one is commited. The OST might wrongly permit second
3862 writing, which will cause grant space > avaiable space.
3865 Frequency : when client is evicted
3867 Description: Grant might be wrongly erased when osc is evicted by OST
3868 Details : when the import is evicted by server, it will fork another
3869 thread ptlrpc_invalidate_import_thread to invalidate the
3870 import, where the grant will be set to 0. While the original
3871 thread will update the grant it got when connecting. So if
3872 the former happened latter, the grant will be wrongly errased
3873 because of this race.
3878 Description: Checking Stale with correct fid
3879 Details : ll_revalidate_it should uses de_inode instead of op_data.fid2
3880 to check whether it is stale, because sometimes, we want the
3881 enqueue happened anyway, and op_data.fid2 will not be initialized.
3883 Severity : enhancement
3885 Description: update patchless client
3886 Details : Add support for patchless client with 2.6.20, 2.6.21 and RHEL 5
3889 Frequency : only with 2.4 kernel
3891 Description: random memory corruption
3892 Details : size of struct ll_inode_info is to big for union inode.u and this
3893 can be cause of random memory corruption.
3898 Description: Memory leak in recovery
3899 Details : Lov_mds_md was not free in an error handler in mds_create_object.
3900 It should also check obd_fail before fsfilt_start, otherwise if
3901 fsfilt_start return -EROFS,(failover mds during mds recovery).
3902 then the req will return with repmsg->transno = 0 and rc = EROFS.
3903 and we met hit the assert LASSERT(req->rq_reqmsg->transno ==
3904 req->rq_repmsg->transno) in ptlrpc_replay_interpret. Fcc should
3905 be freed no matter whether fsfilt_commit success or not.
3908 Frequency : only with large numbers of cores on a single node
3910 Description: Prevents from taking the superblock lock in llap_from_page for
3912 Details : using LL_ORIGIN_REMOVEPAGE origin flag instead of LL_ORIGIN_UNKNOW
3913 for llap_from_page call in ll_removepage() prevents from taking
3914 the superblock lock for a soon died page.
3919 Description: Not check open intent error before release open handle
3920 Details : in some rare cases, the open intent error is not checked before
3921 release open handle, which may cause
3922 ASSERTION(open_req->rq_transno != 0), because it tries to release
3923 the failed open handle.
3928 Description: Set cat log bitmap only after create log success.
3929 Details : in some rare cases, the cat log bitmap is set too early. and it
3930 should be set only after create log success.
3934 Description: Accessing a block bevice can re-enable I/O when Lustre is
3935 tearing down a device.
3936 Details : dev_clear_rdonly(bdev) must be called in kill_bdev() instead of
3941 Description: service threads may hog cpus when there are a lot of requests
3942 Details : Insert cond_resched to give other threads a chance to use some CPU
3947 Description: the cat log was not initialized in recovery
3948 Details : When mds(mgs) do recovery, the tgt_count might be zero, so the
3949 unlink log on mds will not be initialized until mds post
3950 recovery. And also in mds post recovery, the unlink log will
3951 initialization will be done asynchronausly, so there will be race
3952 between add unlink log and unlink log initialization.
3956 Description: brw_stats were being printed incorrectly
3957 Details : brw_stats were being printed as log2 but all of them were not
3958 recorded as log2. Also remove some code duplication arising from
3959 filter_tally_{read,write}.
3963 Frequency : rare, only in recovery.
3964 Description: ASSERTION(req->rq_type != LI_POISON) failed
3965 Details : imp_lock should be held while iterating over imp_sending_list for
3966 prevent destroy request after get timeout in ptlrpc_queue_wait.
3970 Description: replay-single.sh test 52 fails
3971 Details : A lock's skiplist need to be cleanup when it being unlinked
3972 from its resource list.
3977 Description: Short directio read returns full requested size rather than
3979 Details : Direct I/O operations should return actual amount of bytes
3980 transferred rather than requested size.
3982 Severity : enhancement
3984 Description: metadata RPC reduction (e.g. for rm performance)
3985 Details : decrease the amount of synchronous RPC between clients and servers
3986 by canceling conflicing lock before the operation on the client
3987 and packing thier handles into the main operation RPC to server.
3989 Severity : enhancement
3991 Description: add #ifdef HAVE_KERNEL_CONFIG_H
3992 Details : kernels from 2.6.19 not need include linux/config.h, but add
3993 include linux/autoconf.h in commpiler command line.
3995 Severity : enhancement
3997 Description: patchless client support for 2.6.22 kernel
3998 Details : 2.6.22 has only one visble change, SLAB_CTOR_* constants is
3999 removed. In this case we need drop using os depended interface to
4000 kmem_cache and use cfs_mem_cache API.
4005 Description: fix mal-formatted messages
4006 Details : fix some mal-formatted DEBUG_REQ and LCONSOLE_ERROR_MSG messages
4010 Frequency : always in liblustre
4011 Description: wrong IS_ERR implementation in liblustre.h
4012 Details : fix IS_ERR implementation in liblustre.h for right detect errors.
4017 Description: Correct condition for output debug message.
4018 Details : inode i_nlink equal zero is not enough for output message about
4019 disk corruption, i_ctime and i_mode should be also checked.
4023 Frequency : always in patchless client
4024 Description: add configure check for truncate_complete_page
4025 Details : improve checks for exported symbols. This allow run check without
4026 sources, but with Module.symvers shipped with kernel distribution.
4027 add check for truncate_complete_page used by patchless client.
4031 Description: sanity.sh test_77h fails with "test_77h file compare failed"
4032 Details : test_77h uses a file which was messed by other test case.
4036 Description: Not Check whether lov_tgts is NULL in some lov functions
4037 Details : Checking whether lov_tgts is NULL in some functions.
4041 Description: replace obdo_alloc() with OBDO_ALLOC macro
4042 Details : nothing special is done in obdo_alloc() function, and for
4043 debugging purpose, it needs to be replaced with macros.
4047 Description: bad return value and errno from fcntl call
4048 Details : In liblustre API, errno should be a negative value if error
4053 Description: ptlrpc_check_set() LBUG
4054 Details : In case of positive reply from server and failed client bulk
4055 callback after bulk transfer shouldn't LBUG, but process this
4056 request as erroneous.
4058 Severity : enhancement
4060 Description: VFS operations stats tool.
4061 Details : Tool which collects stats by tracking value written in pid,
4062 ppid, gid and uses llstat to generate output to plot graph using
4064 Updated lustre/utils/Makefile.am
4065 Added lustre/utils/ltrack_stats.c
4067 Severity : enhancement
4069 Description: 2.6.18 server support (lustre 1.6.1)
4070 Details : Support for 2.6.18 kernels on the server side.
4075 Description: ASSERTION(imp->imp_conn_current) failed
4076 Details : an assertion failure is hit if a client node boots and attempts to
4077 mount a lustre filesystem faster than RECONNECT_INTERVAL seconds.
4080 Frequency : only for i686
4082 Description: 1.4.11 RC1 build fails for RHEL 4, i686
4083 Details : Fixed config variable for build.
4088 Description: Updated patchess for new RHEL4 kernel
4089 Details : Updated patch inode-nr_unused-2.6.9-rhel4.patch
4090 Updated patch jbd-stats-2.6.9.patch
4091 Updated patch qsnet-rhel4-2.6.patch
4092 Updated patch quota-deadlock-on-pagelock-core.patch
4093 Updated patch vfs_intent-2.6-rhel4.patch
4094 Updated patch vfs_races-2.6-rhel4.patch
4095 Updated series file 2.6-rhel4-titech.series
4096 Updated series file 2.6-rhel4.series
4097 Updated kernel config files
4102 Description: lquota slave complains LBUG when reconnecting with mds
4104 Details : quota slave depends qctxt->lqc_import to send its quota request.
4105 This pointer will be invalid if mds did failover or broke its
4106 connect to osts, which leads to LBUG.
4109 Frequency : when qunit size is too small(less than 20M)
4111 Description: write is stopped by improper -EDQUOT
4112 Details : If the master is busy and qunit size is small enough(let's say 1M),
4113 the slave can not get quota from master on time, which will lead
4114 slave to trigger a -EQUOTA to client.
4119 Description: Deadlock during metadata tests
4120 Details : in prune_dir_dentries(), shrink_dcache_parent() should not be
4121 called with the per-dentry lock held.
4124 Frequency : SLES9 only
4126 Description: Lustre patched kernel for SLES9 SP3 has NR_CPUS set to 8
4127 Details : set CONFIG_NR_CPUS to 128 instead of 8.
4129 Severity : enhancement
4131 Description: remove fs_prep_san_write operation and related patches
4132 Details : remove the ext3-san-jdike patches which are no longer useful.
4137 Description: LDISKFS-fs error (device sdc): ldiskfs_free_blocks
4138 Details : a disk corruption can cause the mballoc code to assert on a
4139 double free or other extent corruptions. Handle these with
4140 ext3_error() instead of assertions.
4143 Frequency : only with mballoc3 code and deep extent trees
4145 Description: ldiskfs_ext_search_right: bad header in inode: unexpected eh_depth
4146 Details : a wrong check of extent headers in ldiskfs_ext_search_right()
4147 can cause the filesystem to be remounted read-only.
4152 Description: lfsck built against 1.4.x cannot run against 1.6.0 lustre
4153 Details : the definition for OBD_IOC_GETNAME changed in 1.6.0. One of the
4154 few external users of this ioctl number is lfsck's call to
4155 llapi_lov_get_uuids() and this caused lfsck to fail at startup.
4156 Add the old ioctl number to the handler so both old and new
4161 Description: parallel lock callbacks
4162 Details : Instead of sending blocking and completion callbacks as separated
4163 requests, adding them to a set and sending in parallel.
4167 Description: Disable most debugging by default
4168 Details : To improve performance, disable most logging (for debug purposes)
4169 by default. VFSTRACE, RPCTRACE, and DLMTRACE are now off by
4170 default, and HA includes fewer messages.
4174 Frequency : only run on patchless client.
4175 Description: use do_facet on sanity.sh for test handling recoverables errors
4176 Details : use do_facet instead of direct use sysctl for set fail_loc on OST
4180 Description: hash tables for lists of nids, connections and uuids
4181 Details : Hash tables noticeably help when a lot of clients connect to a
4182 server, to faster identify duplicate connections or reconnects,
4183 also to faster find export to evict in manual eviction case.
4187 Description: Sometimes, when the server evict a client, and the client will
4188 not be evicted as soon as possible.
4189 Details : In enqueue req, the error was returned by intent, instead of
4190 rq_status which make ptlrpc layer not detect this error, and
4191 does not evict the client. So enqueue error should be returned
4195 Frequency : only at startup
4197 Description: Delay client connections to MDT until fist MDT->OST connect
4198 Details : If a client tried to create a new file before the MDT had
4199 connected to any OSTs, the create would return EIO. Now
4200 the client will simply block until the MDT connects to the
4201 first OST and the create can succeed.
4204 Frequency : only for SLES9
4206 Description: Routinely utilize latest Quadrics drivers in CFS releases
4207 Details : Update patch qsnet-suse-2.6.patch.
4210 Frequency : only for sles10
4212 Description: Update patches for SLES 10 SP1 kernel.
4213 Details : Update the patch vfs_intent-2.6-sles10.patch.
4218 Description: Routinely utilize latest Quadrics drivers in CFS releases
4219 Details : Update patch qsnet-rhel4-2.6.patch.
4222 Frequency : at statup only
4224 Description: mds_lov_synchronize race leads to various problems
4225 Details : simultaneous MDT->OST connections at startup can cause the
4226 sync to abort, leaving the OSC in a bad state.
4231 Description: Using wrong pointer in osc_brw_prep_request
4232 Details : Access to array[-1] can produce panic if kernel compiled with
4233 CONFIG_PAGE_ALLOC enabled
4235 Severity : enhancement
4237 Description: Async OSC create to avoid the blocking unnecessarily.
4238 Details : If a OST has no remain object, system will block on the creating
4239 when need to create a new object on this OST. Now, ways use
4240 pre-created objects when available, instead of blocking on an
4241 empty osc while others are not empty. If we must block, we block
4242 for the shortest possible period of time.
4246 Frequency : only in recovery
4247 Description: Mark OST as early accessible if his start SYNC.
4248 Details : osc_precreate return flag early accessible if oscc marked as
4249 OSCC_FLAG_SYNC_IN_PROGRESS.
4254 Description: Sometimes precreate code can triger create object on wrong ost
4255 Details : Wrong protected or not not restored variables aftre precreate loop
4256 can produce creation object on wrong ost.
4259 Frequency : oss recovery
4261 Description: llog_commit_thread cleanup should sync with llog_commit_thread
4263 Details : llog_commit_thread_count should be synced between llog_commit
4264 start and cleanup, so new llog_commit thread should not be started
4265 when llog_commit threads being stopped to avoid accessing some
4268 Severity : enhancement
4270 Description: Add printing inode info into message about error in writepage.
4272 --------------------------------------------------------------------------------
4274 2007-05-03 Cluster File Systems, Inc. <info@clusterfs.com>
4279 Frequency : on some architectures
4281 Description: 1.6 client sometimes fails to mount from a 1.4 MDT
4282 Details : Uninitialized flags sometimes cause configuration commands to
4286 Frequency : patchless clients only
4288 Description: missing __iget() symbol export
4289 Details : The __iget() symbol export is missing. To avoid the need for
4290 this on patchless clients the deathrow inode reaper is turned
4291 off, and we depend on the VM to clean up old inodes. This
4292 dependency was during via the fix for bug 12181.
4297 Description: sanity.sh fail: test_52b
4298 Details : The ll_inode_to_ext_flags() has a glitch which makes MDS return
4299 incorrect inode's flags to client.
4301 --------------------------------------------------------------------------------
4303 2007-04-19 Cluster File Systems, Inc. <info@clusterfs.com>
4305 * CONFIGURATION CHANGE. This version of Lustre WILL NOT
4306 INTEROPERATE with older versions automatically. In many cases a
4307 special upgrade step is needed. Please read the
4308 user documentation before upgrading any part of a 1.4.x system.
4309 * WARNING: Lustre configuration and startup changes are required with
4310 this release. See https://mail.clusterfs.com/wikis/lustre/MountConf
4312 * Support for kernels:
4313 2.4.21-47.0.1.EL (RHEL 3),
4314 2.6.5-7.283 (SLES 9),
4315 2.6.9-42.0.10.EL (RHEL 4),
4316 2.6.12.6 vanilla (kernel.org),
4317 2.6.16.27-0.9 (SLES10)
4318 * Client support for unpatched kernels:
4319 (see https://mail.clusterfs.com/wikis/lustre/PatchlessClient)
4320 2.6.16 - 2.6.19 vanilla (kernel.org),
4321 2.6.9-42.0.8EL (RHEL 4)
4322 * Recommended e2fsprogs version: 1.39.cfs6
4323 * Note that reiserfs quotas are disabled on SLES 10 in this kernel
4326 Severity : enhancement
4328 Description: Async OSC create to avoid the blocking unnecessarily.
4329 Details : If a OST has no remain object, system will block on the creating
4330 when need to create a new object on this OST. Now, ways use
4331 pre-created objects when available, instead of blocking on an
4332 empty osc while others are not empty. If we must block, we block
4333 for the shortest possible period of time.
4335 Severity : enhancement
4337 Description: MountConf
4338 Details : Lustre configuration is now managed via mkfs and mount
4339 commands instead of lmc and lconf. New obd types (MGS, MGC)
4340 are added for dynamic configuration management. See
4341 https://mail.clusterfs.com/wikis/lustre/MountConf for
4344 Severity : enhancement
4346 Description: dynamic OST addition
4347 Details : OSTs can now be added to a live filesystem
4349 Severity : enhancement
4351 Description: startup order invariance
4352 Details : MDTs and OSTs can be started in any order. Clients only
4353 require the MDT to complete startup.
4355 Severity : enhancement
4357 Description: parallel, asynchronous orphan cleanup
4358 Details : orphan cleanup is now performed in separate threads for each
4359 OST, allowing parallel non-blocking operation.
4361 Severity : enhancement
4363 Description: optimized stripe assignment
4364 Details : stripe assignments are now made based on ost space available,
4365 ost previous usage, and OSS previous usage, in order to try
4366 to optimize storage space and networking resources.
4368 Severity : enhancement
4370 Description: Permanently set tunables
4371 Details : All writable /proc/fs/lustre tunables can now be permanently
4372 set on a per-server basis, at mkfs time or on a live
4375 Severity : enhancement
4377 Description: Lustre message v2
4378 Details : Add lustre message format v2.
4380 Severity : enhancement
4382 Description: client OST exclusion list
4383 Details : Clients can be started with a list of OSTs that should be
4384 declared "inactive" for known non-responsive OSTs.
4388 Description: ENOENT returned for valid filehandle during dbench.
4389 Details : Check if a directory has children when invalidating dentries
4390 associated with an inode during lock cancellation. This fixes
4391 an incorrect ENOENT sometimes seen for valid filehandles during
4392 testing with dbench.
4395 Frequency : SFS test only (otherwise harmless)
4397 Description: SPEC SFS validation failure on NFS v2 over lustre.
4398 Details : Changes the blocksize for regular files to be 2x RPC size,
4399 and not depend on stripe size.
4401 Severity : enhancement
4403 Description: fine-grained SMP locking inside DLM
4404 Details : Improve DLM performance on SMP systems by removing the single
4405 per-namespace lock and replace it with per-resource locks.
4407 Severity : enhancement
4409 Description: don't hold multiple extent locks at one time
4410 Details : To avoid client eviction during large writes, locks are not
4411 held on multiple stripes at one time or for very large writes.
4412 Otherwise, clients can block waiting for a lock on a failed OST
4413 while holding locks on other OSTs and be evicted.
4415 Severity : enhancement
4417 Description: Multiple MD RPCs in flight.
4418 Details : Further unserialise some read-only MDT RPCs - learn about intents.
4419 To avoid overly-overloading MDT, introduce a limit on number of
4420 MDT RPCs in flight for a single client and add /proc controls
4421 to adjust this limit.
4423 Severity : enhancement
4425 Description: client read/write statistics
4426 Details : Add client read/write call usage stats for performance
4427 analysis of user processes.
4428 /proc/fs/lustre/llite/*/offset_stats shows non-sequential
4429 file access. extents_stats shows chunk size distribution.
4430 extents_stats_per_process show chunk size distribution per
4433 Severity : enhancement
4435 Description: per-client statistics on server
4436 Details : Add ldlm and operations statistics for each client in
4437 /proc/fs/lustre/mds|obdfilter/*/exports/
4439 Severity : enhancement
4441 Description: improved MDT statistics
4442 Details : Add detailed MDT operations statistics in
4443 /proc/fs/lustre/mds/*/stats
4445 Severity : enhancement
4447 Description: VFS operations stats
4448 Details : Add client VFS call stats, trackable by pid, ppid, or gid
4449 /proc/fs/lustre/llite/*/stats_track_[pid|ppid|gid]
4454 Description: Fix client-side osc byte counters
4455 Details : The osc read/write byte counters in
4456 /proc/fs/lustre/osc/*/stats are now working
4459 Frequency : always as root on SLES
4461 Description: Failure of copying files with lustre special EAs.
4462 Details : Client side always return success for setxattr call for lustre
4463 special xattr (currently only "trusted.lov").
4468 Description: Refcount LNET uuids
4469 Details : The global LNET uuid list grew linearly with every startup;
4470 refcount repeated list entries instead of always adding to
4473 Severity : enhancement
4475 Description: Dynamic service threads
4476 Details : Within a small range, start extra service threads
4477 automatically when the request queue builds up.
4480 Frequency : mixed-endian client/server environments
4482 Description: mixed-endian crashes
4483 Details : The new msg_v2 system had some failures in mixed-endian
4486 Severity : enhancement
4488 Description: Easy OST removal
4489 Details : OSTs can be permanently deactivated with e.g. 'lctl
4490 conf_param lustre-OST0001.osc.active=0'
4492 Severity : enhancement
4494 Description: MGS proc entries
4495 Details : Added basic proc entries for the MGS showing what filesystems
4498 Severity : enhancement
4500 Description: provide MGS failover
4501 Details : Added config lock reacquisition after MGS server failover.
4503 Severity : enhancement
4505 Description: add Linux 2.4 support
4506 Details : Added support for RHEL 2.4.21 kernel for 1.6 servers and clients
4510 Description: a large application tries to do I/O to the same resource and dies
4511 in the middle of it.
4512 Details : Check the req->rq_arrival time after the call to
4513 ost_brw_lock_get(), but before we do anything about
4514 processing it & sending the BULK transfer request. This
4515 should help move old stale pending locks off the queue as
4516 quickly as obd_timeout.
4519 Frequency : when an incorrect nid is specified during startup
4521 Description: ptlrpc connect to non-existant node causes kernel crash
4522 Details : LNET can't be re-entered from an event callback, which
4523 happened when we expire a message after the export has been
4524 cleaned up. Instead, hand the zombie cleanup off to another
4527 Severity : enhancement
4529 Description: plain/inodebits lock performance improvement
4530 Details : Grouping plain/inodebits in granted list by their request modes
4531 and bits policy, thus improving the performance of search through
4535 Frequency : only if OST filesystem is corrupted
4537 Description: client incorrectly hits assertion in ptlrpc_replay_req()
4538 Details : for a short time RPCs with bulk IO are in the replay list,
4539 but replay of bulk IOs is unimplemented. If the OST filesystem
4540 is corrupted due to disk cache incoherency and then replay is
4541 started it is possible to trip an assertion. Avoid putting
4542 committed RPCs into the replay list at all to avoid this issue.
4545 Frequency : liblustre (e.g. catamount) on a large cluster with >= 8 OSTs/OSS
4547 Description: System hang on startup
4548 Details : This bug allowed the liblustre (e.g. catamount) client to
4549 return to the app before handling all startup RPCs. This
4550 could leave the node unresponsive to lustre network traffic
4551 and manifested as a server ptllnd timeout.
4553 Severity : enhancement
4555 Description: Add "/proc/sys/lustre/debug_peer_on_timeout"
4556 Details : liblustre envirable: LIBLUSTRE_DEBUG_PEER_ON_TIMEOUT
4557 boolean to control whether to print peer debug info when a
4558 client's RPC times out.
4561 Frequency : only for kernels with patches from Lustre below 1.4.3
4563 Description: Remove old rdonly API
4564 Details : Remove old rdonly API which unused from at least lustre 1.4.3
4567 Frequency : only for devices with external journals
4569 Description: Set external device read-only also
4570 Details : During a commanded failover stop, we set the disk device
4571 read-only while the server shuts down. We now also set any
4572 external journal device read-only at the same time.
4575 Frequency : when upgrading from 1.4 while trying to change parameters
4577 Description: The wrong (new) MDC name was used when setting parameters for
4578 upgraded MDT's. Also allows changing of OSC (and MDC)
4579 parameters if --writeconf is specified at tunefs upgrade time.
4582 Frequency : when setting specific ost indicies
4584 Description: QOS code breaks on skipped indicies
4585 Details : Add checks for missing OST indicies in the QOS code, so OSTs
4586 created with --index need not be sequential.
4588 Severity : enhancement
4590 Description: Add uninit_groups feature to ldiskfs2 to speed up e2fsck
4591 Details : The uninit_groups feature works in conjunction with the kernel
4592 filesystem code (ldiskfs2 only) and e2fsprogs-1.39-cfs6 to speed
4593 up the pass1 processing of e2fsck. This is a read-only feature
4594 in ldiskfs2 only, so older kernels and current ldiskfs cannot
4595 mount filesystems that have had this feature enabled.
4597 Severity : enhancement
4599 Description: Improve multi-block allocation algorithm to avoid fragmentation
4600 Details : The mballoc3 code (ldiskfs2 only) adds new mechanisms to improve
4601 allocation locality and avoid filesystem fragmentation.
4603 ------------------------------------------------------------------------------
4605 2007-04-01 Cluster File Systems, Inc. <info@clusterfs.com>
4607 * Support for kernels:
4608 2.4.21-47.0.1.EL (RHEL 3)
4609 2.6.5-7.283 (SLES 9)
4610 2.6.9-42.0.10.EL (RHEL 4)
4611 2.6.12.6 vanilla (kernel.org)
4612 2.6.16.27-0.9 (SLES 10)
4613 * Recommended e2fsprogs version: 1.39.cfs5
4615 * Note that reiserfs quotas are disabled on SLES 10 in this kernel
4619 Frequency : occasional, depends on client load and configuration
4620 Bugzilla : 12181, 12203
4621 Description: data loss for recently-modified files
4623 Details : In some cases it is possible that recently written or created
4624 files may not be written to disk in a timely manner (this should
4625 normally be within 30s unless client IO load is very high).
4626 The problem appears as zero-length files or files that are a
4627 multiple of 1MB in size after a client crash or client eviction
4628 that are missing data at the end of the file.
4630 This problem is more likely to be hit on clients where files are
4631 repeatedly created and unlinked in the same directory, clients
4632 have a large amount of RAM, have many CPUs, the filesystem has
4633 many OSTs, the clients are rebooted frequently, and/or the files
4634 are not accessed by other nodes after being written.
4636 The presence of the problem can be detected by looking at
4637 /proc/sys/fs/inode-state. If the first number (nr_inodes) is
4638 smaller than the second (nr_unused) then dirty files will not
4639 be flushed automatically to disk. "sync; sleep 10" should be
4640 run several times on the node before unmounting it to update
4641 Lustre (this is also safe to run on nodes without this problem).
4643 There is also a related kernel bug in the RHEL4 4 2.6.9 kernel
4644 that can cause this same problem, so customers using that kernel
4645 also need to update the kernel in addition to Lustre. In order
4646 to properly fix this bug, the RHEL3 2.4.21 kernel is also updated.
4648 It is normal that files written just before a client crash (less
4649 than 30s) may not yet have been flushed to disk, even for local
4653 Frequency : frequent on thin XT3 nodes
4655 Description: UUID collision on thin XT3 Linux nodes
4656 Details : UUIDs on Compute Node Linux XT3 nodes were not generated
4657 randomly, since we relied on an insufficiently-seeded PRNG.
4662 Description: OSS hangs after "All ost request buffers busy"
4663 Details : A deadlock between quota and journal operations caused OSS
4664 hangs after printing "All ost request buffers busy."
4667 Frequency : always on liblustre builds
4669 Description: Cleanup compiler warnings on liblustre
4672 Frequency : always on liblustre builds on XT3
4674 Description: LC_CONFIG_CDEBUG don't run while build liblustre on XT3.
4678 Description: Addition of EXT3_FEATURE_RO_COMPAT_DIR_NLINKS flag for
4679 > 32000 subdirectories
4680 Details : Add EXT3_FEATURE_RO_COMPAT_DIR_NLINK flag to
4681 EXT3_FEATURE_RO_COMPAT_SUPP. This flag will be set whenever
4682 subdirectory count crosses 32000. This will aid e2fsck to
4683 correctly handle more than 32000 subdirectories.
4686 Frequency : liblustre (e.g. catamount) on a large cluster with >= 8 OSTs/OSS
4688 Description: System hang on startup
4689 Details : This bug allowed the liblustre (e.g. catamount) client to
4690 return to the app before handling all startup RPCs. This
4691 could leave the node unresponsive to lustre network traffic
4692 and manifested as a server ptllnd timeout.
4694 Severity : enhancement
4696 Description: Add "/proc/sys/lustre/debug_peer_on_timeout"
4697 (liblustre envirable: LIBLUSTRE_DEBUG_PEER_ON_TIMEOUT)
4698 boolean to control whether to print peer debug info when a
4699 client's RPC times out.
4704 Description: make O_SYNC working on 2.6 kernels
4705 Details : 2.6 kernels use different method for mark pages for write,
4706 so need add a code to lustre for O_SYNC work.
4711 Description: Failure to close file and release space on NFS
4712 Details : Put inode details into lock acquired in ll_intent_file_open.
4713 Use mdc_intent_lock in ll_intent_open to properly
4714 detect all kind of errors unhandled by mdc_enqueue.
4719 Description: proc file read during shutdown sometimes raced obd removal,
4721 Details : Add lock to prevent obd access after proc file removal.
4724 Frequency : Only for files larger than 4GB on 32-bit clients.
4726 Description: improperly doing page alignment of locks
4727 Details : Modify lustre core code to use CFS_PAGE_* defines instead of
4728 PAGE_*. Make CFS_PAGE_MASK a 64-bit mask.
4733 Description: RPCs being resent when they shouldn't be
4734 Details : Some RPCs that should not be resent are being resent. This
4735 can cause inconsistencies in the RPC state machine. Do not
4736 resend such requests.
4739 Frequency : rare, only with NFS export
4741 Description: Crash on NFS re-export node
4742 Details : under very unusual load conditions an assertion is hit in
4743 ll_intent_file_open()
4746 Frequency : only if OST filesystem is corrupted
4748 Description: client incorrectly hits assertion in ptlrpc_replay_req()
4749 Details : for a short time RPCs with bulk IO are in the replay list,
4750 but replay of bulk IOs is unimplemented. If the OST filesystem
4751 is corrupted due to disk cache incoherency and then replay is
4752 started it is possible to trip an assertion. Avoid putting
4753 committed RPCs into the replay list at all to avoid this issue.
4758 Description: large O_DIRECT requests fail under memory pressure/fragmentation
4759 Details : Large single O_DIRECT read and write calls can fail to allocate
4760 a sufficiently large buffer to process the request. In case of
4761 allocation failure the allocation is retried with a smaller
4762 buffer and broken into smaller requests.
4764 Severity : enhancement
4766 Description: Add -o localflock option to simulate old noflock behaviour.
4767 Details : This will achieve local-only flock/fcntl locks coherentness.
4772 Description: versioning check is incomplete
4773 Details : Checking the version difference of client vs. server, report
4774 error if the gap is too big.
4779 Description: add support PG_writeback bit
4780 Details : add support for PG_writeback bit for Lustre, for more carefull
4781 work with page cache in 2.6 kernel. This also fix some deadlocks
4782 and remove hack for work O_SYNC with 2.6 kernel.
4784 Severity : enhancement
4786 Description: Add uninit_groups feature to ldiskfs2 to speed up e2fsck
4787 Details : The uninit_groups feature works in conjunction with the kernel
4788 filesystem code (ldiskfs2 only) and e2fsprogs-1.39-cfs6 to speed
4789 up the pass1 processing of e2fsck. This is a read-only feature
4790 in ldiskfs2 only, so older kernels and current ldiskfs cannot
4791 mount filesystems that have had this feature enabled.
4793 Severity : enhancement
4795 Description: Improve multi-block allocation algorithm to avoid fragmentation
4796 Details : The mballoc3 code (ldiskfs2 only) adds new mechanisms to improve
4797 allocation locality and avoid filesystem fragmentation.
4799 ------------------------------------------------------------------------------
4801 2007-02-09 Cluster File Systems, Inc. <info@clusterfs.com>
4803 * Support for kernels:
4804 2.6.9-42.0.3.EL (RHEL 4)
4805 2.6.5-7.276 (SLES 9)
4806 2.4.21-47.0.1.EL (RHEL 3)
4807 2.6.12.6 vanilla (kernel.org)
4808 2.6.16.21-0.8 (SLES10)
4809 * Recommended e2fsprogs version: 1.39.cfs2-0
4811 * The backwards-compatible /proc/sys/portals symlink has been removed
4812 in this release. Before upgrading, please ensure that you change
4813 any configuration scripts or /etc/sysctl.conf files that access
4814 /proc/sys/portals/* or sysctl portals.* to use the corresponding
4815 entry in /proc/sys/lnet or sysctl lnet.*. This change can be made
4816 in advance of the upgrade on any system running Lustre 1.4.6 or
4817 newer, since /proc/sys/lnet was added in that version.
4818 * Note that reiserfs quotas are disabled on SLES 10 in this kernel
4822 Frequency : only when quota is used
4824 Description: avoid scanning export list for quota master
4825 Details : Change the algorithms to avoid scanning export list in order
4826 to improve the efficiency.
4829 Frequency : MDS failover only, very rarely
4831 Description: "went back in time" messages on mds failover
4832 Details : The greatest transno may be lost when the current operation
4833 finishes with an error (transno==0) and the client's last_rcvd
4834 record is over-written. Save the greatest transno in the
4835 mds_last_transno for this case.
4838 Frequency : always for specific kernels and striping counts
4840 Description: client may get "Matching packet too big" without ACL support
4841 Details : Clients compiled without CONFIG_FS_POSIX_ACL get an error message
4842 when trying to access files in certain configurations. The
4843 clients should in fact be denied when mounting because they do
4844 not understand ACLs.
4847 Frequency : Cray XT3 with more than 4000 clients and multiple jobs
4849 Description: many clients connecting with IO in progress causes connect timeouts
4850 Details : Avoid synchronous journal commits to avoid delays caused by many
4851 clients connecting/disconnecting when bulk IO is in progress.
4852 Queue liblustre connect requests on OST_REQUEST_PORTAL instead of
4853 OST_IO_PORTAL to avoid delays behind potentially many pending
4857 Frequency : occasionally with multiple writers to a single file
4859 Description: shared writes to file may result in wrong size reported by stat()
4860 Details : Allow growing of kms when extent lock is cancelled
4863 Frequency : always with random mmap IO to multi-striped file
4865 Description: mmap write might be lost if we are writing to a 'hole' in stripe
4866 Details : Only if the hole is at the end of OST object so that kms is too
4867 small. Fix is to increase kms accordingly in ll_nopage.
4870 Frequency : rare, only if OST filesystem is inconsistent with MDS filesystem
4872 Description: writes to a missing object would leak memory on the OST
4873 Details : If there is an inconsistency between the MDS and OST filesystems,
4874 such that the MDS references an object that doesn't exist, writes
4875 to that object will leak memory due to incorrect cleanup in the
4876 error handling path, eventually running out of memory on the OST.
4881 Description: Creating too long symlink causes lustre errors
4882 Details : Check symlink and name lengths before sending requests to MDS.
4885 Frequency : only if flock is enabled (not on by default)
4887 Description: posix locks not released on fd closure on 2.6.9+
4888 Details : We failed to add posix locks to list of inode locks on 2.6.9+
4889 kernels, this caused such locks not to be released on fd close and
4890 then assertions on fs unmount about still used locks.
4893 Frequency : MDS failover only, very rarely
4895 Description: clients may get ASSERTION(granted_lock != NULL)
4896 Details : When request was taking a long time, and a client was resending
4897 a getattr by name lock request. The were multiple lock requests
4898 with the same client lock handle and
4899 mds_getattr_name->fixup_handle_for_resent_request found one of the
4900 lock handles but later failed with ASSERTION(granted_lock != NULL).
4905 Description: handle->h_buffer_credits > 0, assertion failure
4906 Details : h_buffer_credits is zero after truncate, causing assertion
4907 failure. This patch extends the transaction or creates a new
4911 Frequency : NFS re-export or patchless client
4912 Bugzilla : 11179, 10796
4913 Description: Crash on NFS re-export node (__d_move)
4914 Details : We do not want to hash the dentry if we don't have a lock.
4915 But if this dentry is later used in d_move, we'd hit uninitialised
4916 list head d_hash, so we just do this to init d_hash field but
4917 leave dentry unhashed.
4920 Frequency : NFS re-export or patchless client
4922 Description: NFS exports has problem with symbolic link
4923 Details : lustre client didn't properly install dentry when re-exported
4924 to NFS or running patchless client.
4927 Frequency : NFS re-export or patchless client
4929 Description: Various nfs/patchless fixes.
4930 Details : fixes reuse disconected alias for lookup process - this fixes
4931 warning "find_exported_dentry: npd != pd",
4932 fix permission error with open files at nfs.
4933 fix apply umask when do revalidate.
4936 Frequency : occasional
4938 Description: Crash on NFS re-export node
4939 Details : calling clear_page() on the wrong pointer triggered oops in
4940 generic_mapping_read().
4943 Frequency : rarely, using O_DIRECT IO
4945 Description: unaligned directio crashes client with LASSERT
4946 Details : check for unaligned buffers before trying any requests.
4949 Frequency : rarely, using CFS RAID5 patches in non-standard kernel series
4951 Description: stale data returned from RAID cache
4952 Details : If only a small amount of IO is done to the RAID device before
4953 reading it again it is possible to get stale data from the RAID
4954 cache instead of reading it from disk.
4957 Frequency : always for sles10 kernel
4959 Description: sles10 support
4960 Details : ll_follow_link: compile fixes and using of nd_set_link
4961 under newer kernels.
4964 Frequency : depends on arch, kernel and compiler version, always on sles10
4967 Description: recursive or deep enough symlinks cause stack overflow
4968 Details : getting rid of large stack-allocated variable in
4972 Frequency : depends on hardware
4974 Description: lustre write performance loss in the SLES10 kernel
4975 Details : the performance loss is caused by using of write barriers in the
4976 ext3 code. The SLES10 kernel turns barrier support on by
4977 default. The fix is to undo that change for ldiskfs.
4979 ------------------------------------------------------------------------------
4981 2006-12-09 Cluster File Systems, Inc. <info@clusterfs.com>
4983 * Support for kernels:
4984 2.6.9-42.0.3EL (RHEL 4)
4985 2.6.5-7.276 (SLES 9)
4986 2.4.21-47.0.1.EL (RHEL 3)
4987 2.6.12.6 vanilla (kernel.org)
4991 Frequency : quota enabled and large files being deleted
4993 Description: releasing more than 4GB of quota at once hangs OST
4994 Details : If a user deletes more than 4GB of files on a single OST it
4995 will cause the OST to spin in an infinite loop. Release
4996 quota in < 4GB chunks, or use a 64-bit value for 1.4.7.1+.
5001 Description: statfs data retrieved from /proc may be stale or zero
5002 Details : When reading per-device statfs data from /proc, in the
5003 {kbytes,files}_{total,free,avail} files, it may appear
5004 as zero or be out of date.
5007 Frequency : systems with MD RAID1 external journal devices
5009 Description: lconf's call to blkid is confused by RAID1 journal devices
5010 Details : Use the "blkid -l" flag to locate the MD RAID device instead
5011 of returning all block devices that match the journal UUID.
5014 Frequency : always, for aggregate stripe size over 4GB
5016 Description: "lfs setstripe" fails assertion when setting 4GB+ stripe width
5017 Details : Using "lfs setstripe" to set stripe size * stripe count over 4GB
5018 will fail the kernel with "ASSERTION(lsm->lsm_xfersize != 0)"
5021 Frequency : always if "lfs find" used on a local file/directory
5023 Description: "lfs find" segfaults if used on a local file/directory
5024 Details : The case where a directory component was not specified wasn't
5025 handled correctly. Handle this properly.
5028 Frequency : always on ppc64
5030 Description: the write to an ext3 filesystem mounted with mballoc got stuck
5031 Details : ext3_mb_generate_buddy() uses find_next_bit() which does not
5032 perform endianness conversion.
5035 Frequency : rarely (truncate to non-zero file size after write under load)
5036 Bugzilla : 10730, 10687
5037 Description: Files padded with zeros to next 4K multiple
5038 Details : With filesystems mounted using the "extents" option (2.6 kernels)
5039 it is possible that files that are truncated to a non-zero size
5040 immediately after being written are filled with zero bytes beyond
5041 the truncated size. No file data is lost.
5043 Severity : enhancement
5045 Description: Allow recovery/failover for liblustre clients.
5046 Details : liblustre clients were unaware of failover configurations until
5049 Severity : enhancement
5051 Description: user file locks should fail when not mounting with flock option
5052 Details : Set up an error-returning stub in ll_file_operations.lock field
5053 to prevent incorrect behaviour when client is mounted without
5054 flock option. Also, set up properly f_op->flock field for
5058 Frequency : always on ia64
5060 Description: "lfs df" loops on printing out MDS statfs information
5061 Details : The obd_ioctl_data was not initialized and in some systems
5062 this caused a failure during the ioctl that did not return
5063 an error. Initialize the struct and return an error on failure.
5066 Frequency : SLES 9 only
5068 Description: Error of copying files with lustre special EAs as root
5069 Details : Client side always return success for setxattr call for lustre
5070 special xattr (currently only "trusted.lov").
5073 Frequency : rarely on clusters with both ia64+i386 clients
5075 Description: ia64+i686 clients doing shared IO on the same file may LBUG
5076 Details : In rare cases when both ia64+i686 (or other mixed-PAGE_SIZE)
5077 clients are doing concurrent writes to the same file it is
5078 possible that the ia64 clients may LASSERT because the OST
5079 extent locks are not PAGE_SIZE aligned. Ensure that grown
5080 locks are always aligned on the request boundary.
5083 Frequency : specific use, occasional
5085 Description: Overwriting in use executable truncates on-disk binary image
5086 Details : If one node attempts to overwrite an executable in use by
5087 another node, we now correctly return ETXTBSY instead of
5088 truncating the file.
5090 Severity : enhancement
5092 Description: Async OSC create to avoid the blocking unnecessarily.
5093 Details : If a OST has no remain object, system will block on the creating
5094 when need to create a new object on this OST. Now, ways use
5095 pre-created objects when available, instead of blocking on an
5096 empty osc while others are not empty. If we must block, we block
5097 for the shortest possible period of time.
5102 Description: chmod on Lustre root is propagated to other clients
5103 Details : Re-validate root's dentry in ll_lookup_it to avoid having it
5104 invalid by the follow_mount time.
5107 Frequency : liblustre clients only
5109 Description: Race in 'instant cancel' lock handling could lead to such locks
5110 never to be granted in case of SMP MDS
5111 Details : Do not destroy not yet granted but cbpending locks in
5115 Frequency : replay/resend of open
5117 Description: non null lock assetion failure in mds_intent_policy
5118 Details : Trying to replay/resend lockless open requests resulted in
5119 mds_open() returning 0 with no lock. Now it sets a flag if
5120 a lock is going to be returned.
5122 Severity : enhancement
5124 Description: Checksum enhancements
5125 Details : New checksum enhancements allow for resending RPCs that failed
5128 Severity : enhancement
5130 Description: Tunables on number of dirty pages in cacche
5131 Details : Allow to set limit on number of dirty pages cached.
5136 Description: client crash on unmount - lock still has references
5137 Details : In some error handling cases it was possible to leak a lock
5138 reference on a client while accessing a file. This was not
5139 harmful to the client during operation, but would cause the
5140 client to crash when the filesystem is unmounted.
5143 Frequency : specific case, rare
5145 Description: ETXTBSY on mds though file not in use by client
5146 Details : ETXTBSY is no longer incorrectly returned when attempting to
5147 chmod or chown a directory that the user previously tried to
5148 execute or a currently-executing binary.
5151 Frequency : extremely rare except on liblustre-based clients
5153 Description: Lustre space not freed when files are deleted
5154 Details : Clean up open-unlinked files after client eviction. Previously
5155 the unlink was skipped and the files remained as orphans.
5160 Description: OST failure "would be an LBUG" in waiting_locks_callback()
5161 Details : In some cases it was possible to send a blocking callback to
5162 a client doing a glimpse, even though that client didn't get
5163 a lock granted. When the glimpse lock is cancelled on the OST
5164 the freed lock is left on the waiting list and corrupted the list.
5167 Frequency : all core dumps
5169 Description: Broke core dumps to lustre
5170 Details : Negative dentry may be unhashed if parent does not have UPDATE
5171 lock, but some callers, e.g. do_coredump, expect dentry to be
5172 hashed after successful create, hash it in ll_create_it.
5174 ------------------------------------------------------------------------------
5176 2006-09-13 Cluster File Systems, Inc. <info@clusterfs.com>
5178 * Support for kernels:
5179 2.6.9-42.0.2.EL (RHEL 4)
5180 2.6.5-7.276 (SLES 9)
5181 2.4.21-40.EL (RHEL 3)
5182 2.6.12.6 vanilla (kernel.org)
5186 Frequency : always on RHEL 3
5188 Description: Number of open files grows over time
5189 Details : The number of open files grows over time, whether or not
5190 Lustre is started. This was due to a filp leak introduced
5191 by one of our kernel patches.
5193 ------------------------------------------------------------------------------
5195 2006-08-20 Cluster File Systems, Inc. <info@clusterfs.com>
5197 * Support for kernels:
5198 2.6.9-42.EL (RHEL 4)
5199 2.6.5-7.267 (SLES 9)
5200 2.4.21-40.EL (RHEL 3)
5201 2.6.12.6 vanilla (kernel.org)
5206 Bugzilla : 5719, 9635, 9792, 9684
5207 Description: OST (or MDS) trips assertions in (re)connection under heavy load
5208 Details : If a server is under heavy load and cannot reply to new
5209 connection requests before the client resends the (re)connect,
5210 the connection handling code can behave badly if two service
5211 threads are concurrently handing separate (re)connections from
5212 the same client. Add better locking to the connection handling
5213 code, and ensure that only a single connection will be processed
5214 for a given client UUID, even if the lock is dropped.
5216 Severity : enhancement
5218 Description: add TCP zero-copy support to kernel
5219 Details : Add support to the kernel TCP stack to allow zero-copy bulk
5220 sends if the hardware supports scatter-gather and checksumming.
5221 This allows socklnd to do client-write and server-read more
5222 efficiently and reduce CPU utilization from skbuf copying.
5225 Frequency : only if NFS exporting from client
5227 Description: NULL pointer deref in ll_iocontrol() if chattr mknod file
5228 Details : If setting attributes on a file created under NFS that had
5229 never been opened it would be possible to oops the client
5230 if the file had no objects.
5234 Bugzilla : 9326, 10402, 10897
5235 Description: client crash in ptlrpcd_wake() thread when sending async RPC
5236 Details : It is possible that ptlrpcd_wake() dereferences a freed async
5237 RPC. In rare cases the ptlrpcd thread alread processed the RPC
5238 before ptlrpcd_wake() was called and the request was freed.
5241 Frequency : always for liblustre
5243 Description: liblustre client does MDS+OSTs setattr RPC for each write
5244 Details : When doing a write from a liblustre client, the client
5245 incorrectly issued an RPC to the MDS and each OST the file was
5246 striped over in order to update the timestamps. When writing
5247 with small chunks and many clients this could overwhelm the MDS
5248 with RPCs. In all cases it would slow down the write because
5249 these RPCs are unnecessary.
5251 Severity : enhancement
5253 Description: allow number of MDS service threads to be changed at module load
5254 Details : It is now possible to change the number of MDS service threads
5255 running. Adding "options mds mds_num_threads={N}" to the MDS's
5256 /etc/modprobe.conf will set the number of threads for the next
5257 time Lustre is restarted (assuming the "mds" module is also
5258 reloaded at that time). The default number of threads will
5259 stay the same, 32 for most systems.
5264 Description: OST crash if filesystem is unformatted or corrupt
5265 Details : If an OST is started on a device that has never been formatted
5266 or if the filesystem is corrupt and cannot even mount then the
5267 error handling cleanup routines would dereference a NULL pointer.
5272 Description: NULL pointer deref in llap_from_page.
5273 Details : get_cache_page_nowait can return a page with NULL (or otherwise
5274 incorrect) mapping if the page was truncated/reclaimed while it was
5275 searched for. Check for this condition and skip such pages when
5276 doing readahead. Introduce extra check to llap_from_page() to
5277 verify page->mapping->host is non-NULL (so page is not anonymous).
5280 Frequency : Sometimes when using sys_sendfile
5282 Description: "page not covered by a lock" warnings from ll_readpage
5283 Details : sendfile called ll_readpage without right page locks present.
5284 Now we introduced ll_file_sendfile that does necessary locking
5285 around call to generic_file_sendfile() much like we do in
5289 Frequency : with certain MDS communication failures at client mount time
5291 Description: NULL pointer deref after failed client mount
5292 Details : a client connection request may delayed by the network layer
5293 and not be sent until after the PTLRPC layer has timed out the
5294 request. If the client fails the mount immediately it will try
5295 to clean up before the network times out the request. Add a
5296 reference from the request import to the obd device and delay
5297 the cleanup until the network drops the request.
5300 Frequency : occasionally during client (re)connect
5302 Description: assertion failure during client (re)connect
5303 Details : processing a client connection request may be delayed by the
5304 client or server longer than the client connect timeout. This
5305 causes the client to resend the connection request. If the
5306 original connection request is replied in this interval, the
5307 client may trip an assertion failure in ptlrpc_connect_interpret()
5308 which thought it would be the only running connect process.
5311 Frequency : only with obd_echo servers and clients that are rebooted
5313 Description: kernel BUG accessing uninitialized data structure
5314 Details : When running an obd_echo server it did not start the ping_evictor
5315 thread, and when a client was evicted an uninitialized data
5316 structure was accessed. Start the ping_evictor in the RPC
5317 service startup instead of the OBD startup.
5319 Severity : enhancement
5320 Bugzilla : 10193 (patchless)
5321 Description: Remove dependency on various unexported kernel interfaces.
5322 Details : No longer need reparent_to_init, exit_mm, exit_files,
5323 sock_getsockopt, filemap_populate, FMODE_EXEC, put_filp.
5326 Frequency : rare (only users of deprecated and unsupported LDAP config)
5328 Description: write_conf for zeroconf mount queried LDAP incorrectly for client
5329 Details : LDAP apparently contains 'lustreName' attributes instead of
5330 'name'. A simple remapping of the name is sufficient.
5333 Frequency : rare (only with non-default dump_on_timeout debug enabled)
5335 Description: waiting_locks_callback trips kernel BUG if client is evicted
5336 Details : Running with the dump_on_timeout debug flag turned on makes
5337 it possible that the waiting_locks_callback() can try to dump
5338 the Lustre kernel debug logs from an interrupt handler. Defer
5339 this log dumping to the expired_lock_main() thread.
5341 Severity : enhancement
5343 Description: Support NFS exporting on 2.6 kernels.
5344 Details : Implement non-rawops metadata methods for NFS server to use without
5345 changing NFS server code.
5348 Frequency : very rare (synthetic metadata workload only)
5350 Description: two racing renames might cause an MDS thread to deadlock
5351 Details : Running the "racer" program may cause one MDS thread to rename
5352 a file from being the source of a rename to being the target of
5353 a rename at exactly the same time that another thread is doing
5354 so, and the second thread has already enqueued these locks after
5355 doing a lookup of the target and is trying to relock them in
5356 order. Ensure that we don't try to re-lock the same resource.
5359 Frequency : only very large systems with liblustre clients
5361 Description: slow eviction of liblustre clients with the "evict_by_nid" RPC
5362 Details : Use asynchronous set_info RPCs to send the "evict_by_nid" to
5363 all OSTs in parallel. This allows the eviction of stale liblustre
5364 clients to proceed much faster than if they were done in series,
5365 and also offers similar improvements for other set_info RPCs.
5370 Description: excessive CPU usage during initial read phase on client
5371 Details : During the initial read phase on a client, it would agressively
5372 retry readahead on the file, consuming too much CPU and impacting
5373 performance (since 1.4.5.8). Improve the readahead algorithm
5374 to avoid this, and also improve some other common cases (read
5375 of small files in particular, where "small" is files smaller than
5376 /proc/fs/lustre/llite/*/max_read_ahead_whole_mb, 2MB by default).
5381 Description: MDS crash when receiving packet with unknown intent.
5382 Details : Do not LBUG in unknown intent case, just return -EFAULT
5384 Severity : enhancement
5385 Bugzilla : 9293, 9385
5386 Description: MDS RPCs are serialised on client. This is unnecessary for some.
5387 Details : Do not serialize getattr (non-intent version) and statfs.
5390 Frequency : occasional, when OST network is overloaded/intermittent
5392 Description: client evicted by OST after bulk IO timeout
5393 Details : If a client sends a bulk IO request (read or write) the OST
5394 may evict the client if it is unresposive to its data GET/PUT
5395 request. This is incorrect if the network is overloaded (takes
5396 too long to transfer the RPC data) or dropped the OST GET/PUT
5397 request. There is no need to evict the client at all, since
5398 the pinger and/or lock callbacks will handle this, and the
5399 client can restart the bulk request.
5402 Frequency : Always when mmapping file with no objects
5404 Description: client crashes when mmapping file with no objects
5405 Details : Check that we actually have objects in a file before doing any
5406 operations on objects in ll_vm_open, ll_vm_close and
5412 Description: Request leak when working with deleted CWD
5413 Details : Introduce advanced request refcount tracking for requests
5414 referenced from lustre intent.
5416 Severity : Enhancement
5418 Description: Cache open file handles on client.
5419 Details : MDS now will return special lock along with openhandle, if
5420 requested and client is allowed to hold openhandle, even if unused,
5421 until such a lock is revoked. Helps NFS a lot, since NFS is opening
5422 closing files for every read/write openration.
5424 Severity : Enhancement
5426 Description: Cache open negative dentries on client when possible.
5427 Details : Guard negative dentries with UPDATE lock on parent dir, drop
5428 negative dentries on lock revocation.
5433 Description: Remounting a client read-only wasn't possible with a zconf mount
5434 Details : It wasn't possible to remount a client read-only with llmount.
5436 Severity : enhancement
5437 Description: Include MPICH 1.2.6 Lustre ADIO interface patch
5438 Details : In lustre/contrib/ or /usr/share/lustre in RPM a patch for
5439 MPICH is included to add Lustre-specific ADIO interfaces.
5440 This is based closely on the UFS ADIO layer and only differs
5441 in file creation, in order to allow the OST striping to be set.
5442 This is user-contributed code and not supported by CFS.
5447 Description: extended inode attributes (immutable, append-only) work improperly
5448 when 2.4 and 2.6 kernels are used on client/server or vice versa
5449 Details : Introduce kernel-independent values for these flags.
5451 Severity : enhancement
5454 Description: Allow fractional MB tunings for lustre in /proc/ filesystem.
5455 Details : Many of the /proc/ tunables can only be tuned at a megabyte
5456 granularity. Now, Fractional MB granularity is be supported,
5457 this is very useful for low memory system.
5459 Severity : enhancement
5461 Description: Getattr by fid
5462 Details : Getting a file attributes by its fid, obtaining UPDATE|LOOKUP
5463 locks, avoids extra getattr rpc requests to MDS, allows '/' to
5464 have locks and avoids getattr rpc requests for it on every stat.
5467 Frequency : Always, for filesystems larger than 2TB
5469 Description: ldiskfs crash at mount for filesystem larger than 2TB with mballoc
5470 Details : Kenrel kmalloc limits allocations to 128kB and this prevents
5471 filesystems larger than 2TB to be mounted with mballoc enabled.
5474 Frequency : Always, for 32-bit kernel without CONFIG_LBD and filesystem > 2TB
5476 Description: filesystem corruption for non-standard kernels and very large OSTs
5477 Details : If a 32-bit kernel is compiled without CONFIG_LBD enabled and a
5478 filesystems larger than 2TB is mounted then the kernel will
5479 silently corrupt the start of the filesystem. CONFIG_LBD is
5480 enabled for all CFS-supported kernels, but the possibility of
5481 this happening with a modified kernel config exists.
5483 Severity : enhancement
5485 Description: add client O_DIRECT support for 2.6 kernels
5486 Details : It is now possible to do O_DIRECT reads and writes to files
5487 in the Lustre client mountpoint on 2.6 kernel clients.
5489 Severity : enhancement
5491 Description: parallel glimpse, setattr, statfs, punch, destroy requests
5492 Details : Sends glimpse, setattr, statfs, punch, destroy requests to OSTs in
5493 parallel, not waiting for response from every OST before sending
5494 a rpc to the next OST.
5499 Description: setattr vs write race when updating file timestamps
5500 Details : Client processes that update a file timestamp into the past
5501 right after writing to the file (e.g. tar) it is possible that
5502 the updated file modification time can be reset to the current
5503 time due to a race between processing the setattr and write RPC.
5505 Severity : enhancement
5507 Description: Bring 'lfs find' closer in line with regular Linux find.
5508 Details : lfs find util supports -atime, -mtime, -ctime, -maxdepth, -print,
5509 -print0 options and obtains all the needed info through the lustre
5512 Severity : enhancement
5514 Description: support up to 1024 configured devices on one node
5515 Details : change obd_dev array from statically allocated to dynamically
5516 allocated structs as they are first used to reduce memory usage
5521 Description: Flush dirty partially truncated pages during truncate
5522 Details : Immediatelly flush partially truncated pages in filter_setattr,
5523 this way we completely avoid having any pages in page cache on OST
5524 and can retire ugly workarounds during writes to flush such pages.
5529 Description: i_sem vs transaction deadlock in mds_obd_destroy during unlink.
5530 Details : protect inode from truncation within vfs_unlink() context
5531 just take a reference before calling vfs_unlink() and release it
5532 when parent's i_sem is free.
5535 Frequency : always, if extents are used on OSTs
5537 Description: index ei_leaf_hi (48-bit extension) is not zeroed in extent index
5538 Details : OSTs using the extents format would not zero the high 16 bits of
5539 the index physical block number. This is not a problem for any
5540 OST filesystems smaller than 16TB, and no kernels support ext3
5541 filesystems larger than 16TB yet. This is fixed in 1.4.7 (all
5542 new/modified files) and can be fixed for existing filesystems
5543 with e2fsprogs-1.39-cfs1.
5548 Description: import connection selection may be incorrect if timer wraps
5549 Details : Using a 32-bit jiffies timer with HZ=1000 may cause backup
5550 import connections to be ignored if the 32-bit jiffies counter
5551 wraps. Use a 64-bit jiffies counter.
5554 Frequency : during server recovery
5556 Description: crash after server is denying duplicate export
5557 Details : If clients are resending connect requests to the server, the
5558 server refuses to allow a client to connect multiple times.
5559 Fixed a bug in the handling of this case.
5562 Frequency : very large clusters immediately after boot
5564 Description: LNET request buffers exhausted under heavy short-term load
5565 Details : If a large number of client requests are generated on a service
5566 that has previously never seen so many requests it is possible
5567 that the request buffer growth cannot keep up with the spike in
5568 demand. Instead of dropping incoming requests, they are held in
5569 the LND until the RPC service can accept more requests.
5572 Frequency : Sometimes during replay
5574 Description: Assertion failure in ll_local_open after replay.
5575 Details : If replay happened on an open request reply before we were able
5576 to set replay handler, reply will become not swabbed tripping the
5577 assertion in ll_local_open. Now we set the handler right after
5578 recognising of open request
5581 Frequency : very rare
5583 Description: kernel reports "badness in vsnprintf"
5584 Details : Reading from the "recovery_status" /proc file in small chunks
5585 may cause a negative length in lprocfs_obd_rd_recovery_status()
5586 call to vsnprintf() (which is otherwise harmless). Exit early
5587 if there is no more space in the output buffer.
5589 Severity : enhancement
5591 Description: clear OBD RPC statistics by writing to them
5592 Details : It is now possible to clear the OBD RPC statistics by writing
5593 to the "stats" file.
5598 Description: Client mtime is not the same on different clients after utimes
5599 Details : In some cases, the client was using the utimes() syscall on
5600 a file cached on another node. The clients now validate the
5601 ctime from the MDS + OSTs to determine which one is right.
5606 Description: Inability to activate failout mode
5607 Details : lconf script incorrectly assumed that in python string's numeric
5608 value is used in comparisons.
5611 Frequency : always with multiple stripes per file
5613 Description: Inefficient object allocation for mutli-stripe files
5614 Details : When selecting which OSTs to stripe files over, for files with
5615 a stripe count that divides evenly into the number of OSTs,
5616 the MDS is always picking the same starting OST for each file.
5617 Return the OST selection heuristic to the original design.
5622 Description: mount failures may take full timeout to return an error
5623 Details : Under some heavy load conditions it is possible that a
5624 failed mount can wait for the full obd_timeout interval,
5625 possibly several minutes, before reporting an error.
5626 Instead return an error as soon as the status is known.
5628 ------------------------------------------------------------------------------
5630 2006-02-14 Cluster File Systems, Inc. <info@clusterfs.com>
5632 * WIRE PROTOCOL CHANGE. This version of Lustre networking WILL NOT
5633 INTEROPERATE with older versions automatically. Please read the
5634 user documentation before upgrading any part of a live system.
5635 * WARNING: Lustre networking configuration changes are required with
5636 this release. See https://bugzilla.clusterfs.com/show_bug.cgi?id=10052
5639 * Support for kernels:
5640 2.6.9-22.0.2.EL (RHEL 4)
5641 2.6.5-7.244 (SLES 9)
5642 2.6.12.6 vanilla (kernel.org)
5645 Severity : enhancement
5646 Bugzilla : 7981/8208
5647 Description: Introduced Lustre Networking (LNET)
5648 Details : LNET is new networking infrastructure for Lustre, it includes
5649 a reorganized network configuration mode (see the user
5650 documentation for full details) as well as support for routing
5651 between different network fabrics. Lustre Networking Devices
5652 (LNDs) for the supported network fabrics have also been
5653 created for this new infrastructure.
5655 Severity : enhancement
5656 Description: Introduced Access control lists
5657 Details : clients can set ACLs on files and directories in order to have
5658 more fine-grained permissions than the standard Unix UGO+RWX.
5659 The MDS must be started with the "-o acl" mount option.
5661 Severity : enhancement
5662 Description: Introduced filesystem quotas
5663 Details : Administrators may now establish per-user quotas on the
5666 Severity : enhancement
5668 Description: Configuration change for the XT3
5669 The PTLLND is now used to run Lustre over Portals on the XT3
5670 The configure option(s) --with-cray-portals are no longer used.
5671 Rather --with-portals=<path-to-portals-includes> is used to
5672 enable building on the XT3. In addition to enable XT3 specific
5673 features the option --enable-cray-xt3 must be used.
5678 Description: Running on many-way SMP OSTs can trigger oops in llcd_send()
5679 Details : A race between allocating a new llcd and re-getting the llcd_lock
5680 allowed another thread to grab newly-allocated llcd.
5682 Severity : enhancement
5684 Description: 2.6 OST async journal commit and locking fix to improve performance
5685 Details : The filter_direct_io()+filter_commitrw_write() journal commits for
5686 2.6 kernels are now async as they already were in 2.4 kernels so
5687 that they can commit concurrently with the network bulk transfer.
5688 For block-allocated files the filter allocation semaphore is held
5689 to avoid filesystem fragmentation during allocation. BKL lock
5690 removed for 2.6 xattr operations where it is no longer needed.
5695 Description: lconf incorrectly determined whether two IP networks could talk
5696 Details : In some more complicated routing and multiple-network
5697 configurations, lconf will avoid trying to make a network
5698 connection to a disjoint part of the IP space. It was doing the
5699 math incorrectly for one set of cases.
5704 Description: Fix for potential infinite loop processing records in an llog.
5705 Details : If an llog record is corrupted/zeroed, it is possible to loop
5706 forever in llog_process(). Validate the llog record length
5707 and skip the remainder of the block on error.
5710 Frequency : occasional (liblustre only)
5712 Description: liblustre could not open files whose last component is a symlink
5713 Details : sysio_path_walk() would incorrectly pass the open intent to
5714 intermediate path components.
5717 Frequency : rare (liblustre only with non-standard tuning)
5718 Bugzilla : 7201 (7350)
5719 Description: Tuning the MDC DLM LRU size to zero triggers client LASSERT
5720 Details : llu_lookup_finish_locks() tries to set lock data on a lock
5721 after it has been released, only do this for referenced locks
5723 Severity : enhancement
5725 Description: specifying an (invalid) directory default stripe_size of -1
5726 would reset the directory default striping
5727 Details : stripe_size -1 was used internally to signal directory stripe
5728 removal, now use "all default" to signal dir stripe removal
5729 as a directory striping of "all default" is not useful
5732 Frequency : common for large clusters running liblustre clients
5734 Description: doing an ls when liblustre clients are running is slow
5735 Details : sending a glimpse AST to a liblustre client waits for every AST
5736 to time out, as liblustre clients will not respond. Since they
5737 cannot cache data we refresh the OST lock LVB from disk instead.
5739 Severity : enhancement
5741 Description: doing an ls at the same time as file IO can be slow
5742 Details : enqueue and other "small" requests can be blocked behind many
5743 large IO requests. Create a new OST IO portal for non-IO
5744 requests so they can be processed faster.
5747 Frequency : rare (only HPUX clients mounting unsupported re-exported NFS vol)
5749 Description: an HPUX NFS client would get -EACCESS when ftruncate()ing a newly
5750 created file with mode 000
5751 Details : the Linux NFS server relies on an MDS_OPEN_OWNEROVERRIDE hack to
5752 allow an ftruncate() as a non-root user to a file with mode 000.
5753 Lustre now respects this flag to disable mode checks when
5754 truncating a file owned by the user
5757 Frequency : liblustre-only, when liblustre client dies unexpectedly or becomes
5760 Description: Revoking locks from clients that went dead or catatonic might take
5762 Details : New lock flags FL_CANCEL_ON_BLOCK used by liblustre makes
5763 cancellation of such locks instant on servers without waiting for
5764 any reply from clients. Clients drops these locks when cancel
5765 notification from server is received without replying.
5768 Frequency : liblustre-only, when liblustre client dies or becomes busy
5770 Description: Doing ls on Linux clients can take a long time with active
5772 Details : Liblustre client cannot handle ASTs in timely manner, so avoid
5773 granting such locks to it in the first place if possible. Locks
5774 are taken by proxy on the OST during the read or write and
5775 dropped immediately afterward. Add connect flags handling, do
5776 not grant locks to liblustre clients for glimpse ASTs.
5778 Severity : enhancement
5780 Description: Improve read-ahead algorithm to avoid excessive IO for random reads
5781 Details : Existing read-ahead algorithm is tuned for the case of streamlined
5782 sequential reads and behaves badly with applications doing random
5783 reads. Improve it by reading ahead at least read region, and
5784 avoiding excessive large RPC for small reads.
5786 Severity : enhancement
5788 Description: Creating more than 1000 files for a single job may cause a load
5789 imbalance on the OSTs if there are also a large number of OSTs.
5790 Details : qos_prep_create() uses an OST index reseed value that is an
5791 even multiple of the number of available OSTs so that if the
5792 reseed happens in the middle of the object allocation it will
5793 still utilize the OSTs as uniformly as possible.
5798 Description: OST or MDS may oops in ping_evictor_main()
5799 Details : ping_evictor_main() drops obd_dev_lock if deleting a stale export
5800 but doesn't restart at beginning of obd_exports_timed list
5803 Severity : enhancement
5805 Description: improve by-nid export eviction on the MDS and OST
5806 Details : allow multiple exports with the same NID to be evicted at one
5807 time without re-searching the exports list.
5810 Frequency : rare, only with supplementary groups enabled on SMP 2.6 kernels
5812 Description: MDS may oops in groups_free()
5813 Details : in rare race conditions a newly allocated group_info struct is
5814 freed again, and this can be NULL. The 2.4 compatibility code
5815 for groups_free() checked for a NULL pointer, but 2.6 did not.
5818 Frequency : common for liblustre clients doing little filesystem IO
5819 Bugzilla : 9352, 7313
5820 Description: server may evict liblustre clients accessing contended locks
5821 Details : if a client is granted a lock or receives a completion AST
5822 with a blocking AST already set it would not reply to the AST
5823 for LDLM_FL_CANCEL_ON_BLOCK locks. It now replies to such ASTs.
5826 Frequency : lfs setstripe, only systems with more than 160 OSTs
5828 Description: unable to set striping with a starting offset beyond OST 160
5829 Details : llapi_create_file() incorrectly limited the starting stripe
5830 index to the maximum single-file stripe count.
5833 Frequency : LDAP users only
5835 Description: lconf did not handle in-kernel recovery with LDAP properly
5836 Details : lconf/LustreDB get_refs() is searching the wrong namespace
5838 Severity : enhancement
5840 Description: bind OST threads to NUMA nodes to improve performance
5841 Details : all OST threads are uniformly bound to CPUs on a single NUMA
5842 node and do their allocations there to localize memory access
5844 Severity : enhancement
5846 Description: llmount can determine client NID directly from Myrinet (GM)
5847 Details : the client NID code from gmnalnid was moved directly into
5848 llmount, removing the need to use this or specifying the
5849 client NID explicitly when mounting GM clients with zeroconf
5852 Frequency : if client is started with down MDS
5854 Description: if client is started with down MDS mount hangs in ptlrpc_queue_wait
5855 Details : Having an LWI_INTR() wait event (interruptible, but no timeout)
5856 will wait indefinitely in ptlrpc_queue_wait->l_wait_event() after
5857 ptlrpc_import_delayed_req() because we didn't check if the
5858 request was interrupted, and we also didn't break out of the
5859 event loop if there was no timeout
5864 Description: data loss during non-page-aligned writes to a single file from
5865 both multiple nodes and multiple threads on one node at same time
5866 Details : updates to KMS and lsm weren't protected by common lock. Resulting
5867 inconsistency led to false short-reads, that were cached and later
5868 used by ->prepare_write() to fill in partially written page,
5869 leading to data loss.
5872 Frequency : always, if lconf --abort_recovery used
5874 Description: lconf --abort_recovery fails with 'Operation not supported'
5875 Details : lconf was attempting to abort recovery on the MDT device and not
5878 Severity : enhancement
5880 Description: remove cleanup logs
5881 Details : replace lconf-generated cleanup logs with lustre internal
5882 cleanup routines. Eliminates the need for client-cleanup and
5885 Severity : enhancement
5887 Description: add support for EAs (user and system) on lustre filesystems
5888 Details : it is now possible to store extended attributes in the Lustre
5889 client filesystem, and with the user_xattr mount option it
5890 is possible to allow users to store EAs on their files also
5892 Severity : enhancement
5894 Description: Add possibility (config option) to show minimal available OST free
5896 Details : When compiled with --enable-mindf configure option, statfs(2)
5897 (and so, df) will return least minimal free space available from
5898 all OSTs as amount of free space on FS, instead of summary of
5899 free spaces of all OSTs.
5901 Severity : enhancement
5903 Description: do not expand extent locks acquired on OST-side
5904 Details : Modify ldlm_extent_policy() to not expand local locks, acquired
5905 by server: they are not cached anyway.
5908 Frequency : when mmap is used/binaries executed from Lustre
5910 Description: Unmmap pages before throwing them away from read cache.
5911 Details : llap_shrink cache now attempts to unmap pages before discarding
5912 them (if unmapping failed - do not discard). SLES9 kernel has
5913 extra checks that trigger if this unmapping is not done first.
5918 Description: lconf didn't resolve symlinks before checking to see whether a
5919 given mountpoint was already in use
5922 Frequency : when migrating failover services
5923 Bugzilla : 6395, 9514
5924 Description: When migrating a subset of services from a node (e.g. failback
5925 from a failover service node) the remaining services would
5926 time out and evict clients.
5927 Details : lconf --force (implied by --failover) sets the global obd_timeout
5928 to 5 seconds in order to quickly disconnect, but this caused
5929 other RPCs to time out too quickly. Do not change the global
5930 obd_timeout for force cleanup, only set it for DISCONNECT RPCs.
5932 Severity : enhancement
5933 Frequency : if MDS is started with down OST
5934 Bugzilla : 9439,5706
5935 Description: Allow startup/shutdown of an MDS without depending on the
5936 availability of the OSTs.
5937 Details : Asynchronously call mds_lov_synchronize during MDS startup.
5938 Add appropriate locking and lov-osc refcounts for safe
5939 cleaning. Add osc abort_inflight calls in case the
5940 synchronize never started.
5943 Frequency : occasional (Cray XT3 only)
5945 Description: root not authorized to access files in CRAY_PORTALS environment
5946 Details : The client process capabilities were not honoured on the MDS in
5947 a CRAY_PORTALS/CRAY_XT3 environment. If the file had previously
5948 been accessed by an authorized user then root was able to access
5949 the file on the local client also. The root user capabilities
5950 are now allowed on the MDS, as this environment has secure UID.
5953 Frequency : occasional
5955 Description: ldiskfs "too long searching" message happens too often
5956 Details : A debugging message (otherwise harmless) prints too often on
5957 the OST console. This has been reduced to only happen when
5958 there are fragmentation problems on the filesystem.
5963 Description: Division by zero in statfs when all OSCs are inactive
5964 Details : lov_get_stripecnt() returns zero due to incorrect order of checks,
5965 lov_statfs divides by value returned by lov_get_stripecnt().
5969 Bugzilla : 9489, 3273
5970 Description: First write from each client to each OST was only 4kB in size,
5971 to initialize client writeback cache, which caused sub-optimal
5972 RPCs and poor layout on disk for the first writen file.
5973 Details : Clients now request an initial cache grant at (re)connect time
5974 and so that they can start streaming writes to the cache right
5975 away and always do full-sized RPCs if there is enough data.
5976 If the OST is rebooted the client also re-establishes its grant
5977 so that client cached writes will be honoured under the grant.
5982 Description: Slow ls (and stat(2) syscall) on files residing on IO-loaded OSTs
5983 Details : Now I/O RPCs go to different portal number and (presumably) fast
5984 lock requests (and glimses) and other RPCs get their own service
5985 threads pool that should be able to service those RPCs
5988 Severity : enhancement
5990 Description: Ability to exchange lustre version between client and servers and
5991 issue warnings at client side if client is too old. Also for
5992 liblustre clients there is ability to refuse connection of too old
5994 Details : New 'version' field is added to connect data structure that is
5995 filled with version info. That info is later checked by server and
5999 Frequency : rare, liblustre only.
6000 Bugzilla : 9296, 9581
6001 Description: Two simultaneous writes from liblustre at offset within same page
6002 might proceed at the same time overwriting eachother with stale
6004 Details : I/O lock withing llu_file_prwv was released too early, before data
6005 actually was hitting the wire. Extended lock-holding time until
6006 server acknowledges receiving data.
6009 Frequency : extremely rare. Never observed in practice.
6011 Description: avoid generating lustre_handle cookie of 0.
6012 Details : class_handle_hash() generates handle cookies by incrementing
6013 global counter, and can hit 0 occasionaly (this is unlikely, but
6014 not impossible, because initial value of cookie counter is
6015 selected randonly). Value of 0 is used as a sentinel meaning
6016 "unassigned handle" --- avoid it. Also coalesce two critical
6017 sections in this function into one.
6019 Severity : enhancement
6021 Description: allow liblustre clients to delegate truncate locking to OST
6022 Details : To avoid overhead of locking, liblustre client instructs OST to
6023 take extent lock in ost_punch() on client's behalf. New connection
6024 flag is added to handle backward compatibility.
6026 Severity : enhancement
6027 Bugzilla : 4928, 7341, 9758
6028 Description: allow number of OST service threads to be specified
6029 Details : a module parameter allows the number of OST service threads
6030 to be specified via "options ost ost_num_threads={N}" in the
6031 OSS's /etc/modules.conf or /etc/modprobe.conf.
6035 Bugzilla : 6146, 9635, 9895
6036 Description: servers crash with bad pointer in target_handle_connect()
6037 Details : In rare cases when a client is reconnecting it was possible that
6038 the connection request was the last reference for that export.
6039 We would temporarily drop the export reference and get a new
6040 one, but this may have been the last reference and the export
6041 was just destroyed. Get new reference before dropping old one.
6043 Severity : enhancement
6044 Frequency : if client is started with failover MDS
6046 Description: Allow multiple MDS hostnames in the mount command
6047 Details : Try to read the configuration from all specified MDS
6048 hostnames during a client mount in case the "primary"
6051 Severity : enhancement
6053 Description: Stop sending data to evicted clients as soon as possible.
6054 Details : Check if the client we are about to send or are sending data to
6055 was evicted already. (Check is done every second of waiting,
6056 for which l_wait_event interface was extended to allow checking
6057 of exit condition at specified intervals).
6060 Frequency : rare, normally only when NFS exporting is done from client
6062 Description: 'bad disk LOV MAGIC: 0x00000000' error when chown'ing files
6064 Details : Make mds_get_md() recognise empty md case and set lmm size to 0.
6067 Frequency : always, if srand() is called before liblustre initialization
6069 Description: Liblustre uses system PRNG disturbing its usage by user application
6070 Details : Introduce internal to lustre fast and high-quality PRNG for
6071 lustre usage and make liblustre and some other places in generic
6072 lustre code to use it.
6074 Severity : enhancement
6075 Bugzilla : 9477, 9557, 9870
6076 Description: Verify that the MDS configuration logs are updated when xml is
6077 Details : Check if the .xml configuration logs are newer than the config
6078 logs stored on the MDS and report an error if this is the case.
6079 Request --write-conf, or allow starting with --old_conf.
6081 Severity : enhancement
6083 Description: Handle symlinks in the path when checking if Lustre is mounted.
6084 Details : Resolve intermediate symlinks when checking if a client has
6085 mounted a filesystem to avoid duplicate client mounts.
6090 Description: lconf can hit an error exception but still return success.
6091 Details : The lconf command catches the Command error exception at the top
6092 level script context and will exit with the associated exit
6093 status, but doesn't ensure that this exit status is non-zero.
6098 Description: failure of ptlrpc thread startup can cause oops
6099 Details : Starting a ptlrpc service thread can fail if there are a large
6100 number of threads or the server memory is very fragmented.
6101 Handle this without oopsing.
6104 Frequency : always, only if liblustre and non-default acceptor port was used
6106 Description: liblustre cannot connect to servers with non-default acceptor port
6107 Details : tcpnal_set_default_params() was not called and was therefore
6108 ignoring the environment varaible TCPNAL_PORT, as well as other
6109 TCPNAL_ environment variables
6114 Description: two objects could be created on the same OST for a single file
6115 Details : If an OST is down, in some cases it was possible to create two
6116 objects on a single OST for a single file. No problems other
6117 than potential performance impact and spurious error messages.
6121 Bugzilla : 5681, 9562
6122 Description: Client may oops in ll_unhash_aliases
6123 Details : Client dcache may become inconsistent in race condition.
6124 In some cases "getcwd" can fail if the current directory is
6130 Description: Inode refcounting problems in NFS export code
6131 Details : link_raw functions used to call d_instantiate without obtaining
6132 extra inode reference first.
6136 Bugzilla : 9942, 9903
6137 Description: Referencing freed requests leading to crash, memleaks with NFS.
6138 Details : We used to require that call to ll_revalidate_it was always
6139 followed by ll_lookup_it. Also with revalidate_special() it is
6140 possible to call ll_revalidate_it() twice for the same dentry
6141 even if first occurence returned success. This fix changes semantic
6142 between DISP_ENQ_COMPLETE disposition flag to mean there is extra
6143 reference on a request referred from the intent.
6144 ll_intent_release() then releases such a request.
6147 Frequency : rare, normally benchmark loads only
6149 Description: unlinked inodes were kept in memory on the client
6150 Details : If a client is repeatedly creating and unlinking files it
6151 can accumulate a lot of stale inodes in the inode slab cache.
6152 If there is no other client load running this can cause the
6153 client node to run out of memory. Instead flush old inodes
6154 from client cache that have the same inode number as a new inode.
6157 Frequency : SLES9 2.6.5 kernel and long filenames only
6158 Bugzilla : 9969, 10379
6159 Description: utime reports stale NFS file handle
6160 Details : SLES9 uses out-of-dentry names in some cases, which confused
6161 the lustre dentry revalidation. Change it to always use the
6165 Frequency : rare, unless heavy write-truncate concurrency is continuous
6166 Bugzilla : 4180, 6984, 7171, 9963, 9331
6167 Description: OST becomes very slow and/or deadlocked during object unlink
6168 Details : filter_destroy() was holding onto the parent directory lock
6169 while truncating+unlinking objects. For very large objects this
6170 may block other threads for a long time and slow overall OST
6171 responsiveness. It may also be possible to get a lock ordering
6172 deadlock in this case, or run out of journal credits because of
6173 the combined truncate+unlink. Solution is to do object truncate
6174 first in one transaction without parent lock, and then do the
6175 final unlink in a new transaction with the parent lock. This
6176 reduces the lock hold time dramatically.
6179 Frequency : rare, 2.4 kernels only
6181 Description: MDS or OST cleanup may trip kernel BUG when dropping kernel lock
6182 Details : mds_cleanup() and filter_cleanup() need to drop the kernel lock
6183 before unmounting their filesystem in order to avoid deadlock.
6184 The kernel_locked() function in 2.4 kernels only checks whether
6185 the kernel lock is held, not whether it is this process that is
6186 holding it as 2.6 kernels do.
6191 Description: MDS or OST may oops/LBUG if a client is connecting multiple times
6192 Details : The client ptlrpc code may be trying to reconnect to a down
6193 server before a previous connection attempt has timed out.
6194 Increase the reconnect interval to be longer than the connection
6195 timeout interval to avoid sending duplicate connections to
6199 Frequency : echo_client brw_test command
6201 Description: fix echo_client to work with OST preallocated code
6202 Details : OST preallocation code (5137) didn't take echo_client IO path
6203 into account: echo_client calls filter methods outside of any
6204 OST thread and, hence, there is no per-thread preallocated
6205 pages and buffers to use. Solution: hijack pga pages for IO. As
6206 a byproduct, this avoids unnecessary data copying.
6210 Bugzilla : 3555, 5962, 6025, 6155, 6296, 9574
6211 Description: Client can oops in mdc_commit_close() after open replay
6212 Details : It was possible for the MDS to return an open request with no
6213 transaction number in mds_finish_transno() if the client was
6214 evicted, but without actually returning an error. Clients
6215 would later try to replay that open and may trip an assertion
6216 Simplify the client close codepath, and always return an error
6217 from the MDS in case the open is not successful.
6220 Frequency : rare, 2.6 OSTs only
6222 Description: OST may deadlock under high load on fragmented files
6223 Details : If there was a heavy load and highly-fragmented OST filesystems
6224 it was possible to have all the OST threads deadlock waiting on
6225 allocation of biovecs, because the biovecs were not released
6226 until the entire RPC IO was completed. Instead, release biovecs
6227 as soon as they are complete to ensure forward IO progress.
6229 Severity : enhancement
6231 Description: Support for specifying external journal device at mount
6232 Details : If an OST or MDS device is formatted with an external journal
6233 device, this device major/minor is stored in the ext3 superblock
6234 and may not be valid for failover. Allow detecting and
6235 specifying the external journal at mount time.
6240 Description: Mounting an MDS with pending unlinked files may cause oops
6241 Details : target_finish_recovery() calls mds_postrecov() which returned
6242 the number of orphans unlinked. mds_lov_connect->mds_postsetup()
6243 considers this an error and immediately begins cleaning up the
6244 lov, just after starting the mds_lov process
6246 Severity : enhancement
6248 Description: Implement 'lfs df' to report actual free space on per-OST basis
6249 Details : Add sub-command 'df' on 'lfs' to report the disk space usage of
6250 MDS/OSDs. Usage: lfs df [-i][-h]. Command Options: '-i' to report
6251 usage of objects; '-h' to report in human readable format.
6253 ------------------------------------------------------------------------------
6255 2005-08-26 Cluster File Systems, Inc. <info@clusterfs.com>
6262 Description: Mounting an ldiskfs file system with mballoc may crash OST node.
6263 Details : ldiskfs mballoc code may reference an uninitialized buddy struct
6264 at startup during orphan unlinking. Instead, skip buddy update
6265 before setup, as it will be regenerated after recovery is complete.
6270 Description: If an OST is inactive, its locks might reference stale inodes.
6271 Details : lov_change_cbdata() must iterate over all namespaces, even if
6272 they are inactive to clear inode references from the lock.
6274 Severity : enhancement
6275 Frequency : occasional, if non-standard max_dirty_mb used
6277 Description: Client will block write RPCs if not enough grant
6278 Details : If a client has max_dirty_mb smaller than max_rpcs_in_flight,
6279 then the client will block writes while waiting for another RPC
6280 to complete instead of consuming its dirty limit. With change
6281 we get improved performance when max_dirty_mb is small.
6283 Severity : enhancement
6284 Bugzilla : 3389, 6253
6285 Description: Add support for supplementary groups on the MDS.
6286 Details : The MDS has an upcall /proc/fs/lustre/mds/{mds}/group_upcall
6287 (set to /usr/sbin/l_getgroups if enabled) which will do MDS-side
6288 lookups for user supplementary groups into a cache.
6292 Description: O_CREAT|O_EXCL open flags in liblustre always return -EEXIST
6293 Details : Make libsysio to not enforce O_EXCL by clearing the flag,
6294 for liblustre O_EXCL is enforced by MDS.
6298 Description: readdir never returns NULL in liblustre.
6299 Details : Corrected llu_iop_getdirentries logic, to return offset of next
6300 dentry in struct dirent.
6304 Frequency : liblustre only, depends on application IO pattern
6305 Description: liblustre clients evicted if not contacting servers
6306 Details : Don't put liblustre clients into the ping_evictor list, so
6307 they will not be evicted by the pinger ever.
6309 Severity : enhancement
6311 Description: Add ability to evict clients by NID from MDS.
6312 Details : By echoing "nid:$NID" string into
6313 /proc/fs/lustre/mds/.../evict_client client with nid that equals to
6314 $NID would be instantly evicted from this MDS and from all active
6315 OSTs connected to it.
6319 Description: Do not query file size twice, somewhat slowing stat(2) calls.
6320 Details : lookup_it_finish() used to query file size from OSTs that was not
6325 Description: service threads change working directory to that of init
6326 Details : Starting lustre service threads may pin the working directory
6327 of the parent thread, making that filesystem busy. Threads
6328 now change to the working directory of init to avoid this.
6332 Frequency : during shutdown only
6333 Description: shutdown with a failed MDS or OST can cause unmount to hang
6334 Details : Don't resend DISCONNECT messages in ptlrpc_disconnect_import()
6339 Frequency : 2.6 only
6340 Description: chmod/chown may include an extra supplementary group
6341 Details : ll{,u}_mdc_pack_op_data() does not properly initialize the
6342 supplementary group and if none is specified this is used.
6345 Bugzilla : 5479 (6816)
6347 Description: Racing open + rm can assert client in mdc_set_open_replay_data()
6348 Details : If lookup is in progress on a file that is unlinked we might try
6349 to revalidate the inode and fail in revalidate after lookup is
6350 complete and ll_file_open() enqueues the open again but
6351 it_open_error() was not checking DISP_OPEN_OPEN errors correctly.
6354 Frequency : always, if lconf --abort_recovery used
6356 Description: lconf --abort_recovery fails with 'Operation not supported'
6357 Details : lconf was attempting to abort recovery on the MDT device and not
6360 ------------------------------------------------------------------------------
6362 2005-08-08 Cluster File Systems, Inc. <info@clusterfs.com>
6367 Frequency : rare (only unsupported configurations with a node running as an
6369 Bugzilla : 6514, 5137
6370 Description: Mounting a Lustre file system on a node running as an OST could
6372 Details : OSTs now preallocates memory needed to write out data at
6373 startup, instead of when needed, to avoid having to
6374 allocate memory in possibly low memory situations.
6375 Specifically, if the file system is mounted on on OST,
6376 memory pressure could force it to try to write out data,
6377 which it needed to allocate memory to do. Due to the low
6378 memory, it would be unable to do so and the node would
6379 become unresponsive.
6381 Severity : enhancement
6383 Description: Addition of lconf --service command line option
6384 Details : lconf now accepts a '--service <arg>' option, which is
6385 shorthand for 'lconf --group <arg> --select <arg>=<hostname>'
6387 Severity : enhancement
6389 Description: Failover mode is now the default for OSTs.
6390 Details : By default, OSTs will now run in failover mode. To return to
6391 the old behaviour, add '--failout' to the lmc line for OSTs.
6393 Severity : enhancement
6395 Description: Health checks are now provided for MDS and OSTs
6396 Details : Additional detailed health check information on MSD and OSTs
6397 is now provided through the procfs health_check value.
6400 Frequency : occasional, depends on IO load
6402 Description: Disk fragmentation on the OSTs could eventually cause slowdowns
6403 after numerous create/delete cycles
6404 Details : The ext3 inode allocation policy would not allocate new inodes
6405 very well on the OSTs because there are no new directories
6406 being created. Instead we look for groups with free space if
6407 the parent directories are nearly full.
6412 Description: Network or server problems during mount may cause partially
6413 mounted clients instead of returning an error.
6414 Details : The config llog parsing code may overwrite the error return
6415 code during mount error handling, returning success instead
6421 Description: MDS can fail to allocate large reply buffers
6422 Details : After long uptimes the MDS can fail to allocate large reply
6423 buffers (e.g. zconf client mount config records) due to memory
6424 fragmentation or consumption by the buffer cache. Preallocate
6425 some large reply buffers so that these replies can be sent even
6426 under memory pressure.
6430 Frequency : rare (liblustre)
6431 Description: fsx running with liblustre complained that using truncate() to
6432 extend the file doesn't work. This patch corrects that issue.
6433 Details : This is the liblustre equivalent of the fix for bug 6196. Fixes
6434 ATTR_SIZE and lsm use in llu_setattr_raw.
6438 Frequency : rare, only 2.6 kernels
6439 Description: Unusual file access patterns on the MDS may result in inode
6440 data being lost in very rare circumstances.
6441 Details : Bad interaction between the ea-in-inode patch and the "no-read"
6442 code in the 2.6 kernel caused the inode and/or EA data not to
6443 be read from disk, causing single-file corruption.
6447 Frequency : rare, only 2.6 filesystems using extents
6448 Description: Heavy concurrent write and delete load may cause data corruption.
6449 Details : It was possible under high-load situations to have an extent
6450 metadata block in the block device cache from a just-unlinked
6451 file overwrite a newly-allocated data block. We now unmap any
6452 metadata buffers that alias just-allocated data blocks.
6456 Frequency : filesystems with default stripe_count larger than 77
6457 Description: lconf+mke2fs fail when formatting filesystem with > 77 stripes
6458 Details : lconf specifies an inode size of 4096 bytes when the default
6459 stripe_count is larger than 77. This conflicts with the default
6460 inode density of 1 per 4096 bytes. Allocate smaller inodes in
6461 this case to avoid pinning too much memory for large EAs.
6463 ------------------------------------------------------------------------------
6465 2005-07-07 Cluster File Systems, Inc. <info@clusterfs.com>
6470 Frequency : rare (extremely heavy IO load with hundreds of clients)
6472 Description: Client is evicted, gets IO error writing to file
6473 Details : lock ordering changes for bug 5492 reintroduced bug 3267 and
6474 caused clients to be evicted for AST timeouts. The fixes in
6475 bug 5192 mean we no longer need to have such short AST timeouts
6476 so ldlm_timeout has been increased.
6479 Frequency : occasional during --force or --failover shutdown under load
6480 Bugzilla : 5949, 4834
6481 Description: Server oops/LBUG if stopped with --force or --failover under load
6482 Details : a collection of import/export refcount and cleanup ordering
6483 issues fixed for safer force cleanup
6486 Frequency : only filesystems larger than 120 OSTs
6487 Bugzilla : 5990, 6223
6488 Description: lfs getstripe would oops on a very large filesystem
6489 Details : lov_getconfig used kfree on vmalloc'd memory
6492 Frequency : only filesystems exporting via NFS to Solaris 10 clients
6493 Bugzilla : 6242, 6243
6494 Description: reading from files that had been truncated to a non-zero size
6495 but never opened returned no data
6496 Details : ll_file_read() reads zeros from no-object files to EOF
6501 Description: A bug in MDS/OSS recovery could cause the OSS to fail an assertion
6502 Details : There's little harm in aborting MDS/OSS recovery and letting it
6503 try again, so I removed the LASSERT and return an error instead.
6505 Severity : enhancement
6507 Description: New debugging infrastructure for tracking down data corruption
6508 Details : The I/O checksum code was replaced to: (a) control it at runtime,
6509 (b) cover more of the client-side code path, and (c) try to narrow
6510 down where problems occurred
6514 Bugzilla : 3819, 4364, 4397, 6313
6515 Description: Racing close and eviction MDS could cause assertion in mds_close
6516 Details : It was possible to get multiple mfd references during close and
6517 client eviction, leading to one thread referencing a freed mfd.
6519 Severity: : enhancement
6520 Bugzilla : 3262, 6359
6521 Description: Attempts to reconnect to servers are now more aggressive.
6522 Details : This builds on the enhanced upcall-less recovery that was added
6523 in 1.4.2. When trying to reconnect to servers, clients will
6524 now try each server in the failover group every 10 seconds. By
6525 default, clients would previously try one server every 25 seconds.
6530 Description: After recovery, certain operations trigger a failed
6531 assertion on a client.
6532 Details : Failing over an mds, using lconf -d --failover, while a
6533 client was doing a readdir() call would cause the client to
6534 LBUG after recovery completed and the readdir() was resent.
6536 Severity : enhancement
6538 Description: Default groups are now added by lconf
6539 Details : You can now run lconf --group <servicename> without having to
6540 manually add groups with lmc.
6543 Frequency : occasional
6545 Description: Nodes with an elan id of 0 trigger a failed assertion
6548 Frequency : always when accessing e.g. tty/console device nodes
6550 Description: tty and some other devices nodes cannot be used on lustre
6551 Details : file's private_data field is used by device data and lustre
6552 values in there got lost. New field was added to struct file to
6553 store fs-specific private data.
6556 Frequency : when exporting Lustre via NFS
6558 Description: NFSD failed occasionally when looking up a path component
6559 Details : NFSD is looking up ".." which was broken in ext3 directories
6560 that had grown large enough to become hashed.
6563 Frequency : Clusters with multiple interfaces not on the same subnet
6565 Description: Nodes will repeatedly try to reconnect to an interface which it
6566 cannot reach and report an error to the log.
6567 Details : Extra peer list entries will be created by lconf with some peers
6568 unreachable. lconf now validates the peer before adding it.
6571 Frequency : Only if a default stripe is set on the filesystem root.
6573 Description: Setting a default stripe on the filesystem root prevented the
6574 filesystem from being remounted.
6575 Details : The client was sending extra request flags in the root getattr
6576 request and did not allocate a reply buffer for the dir EA.
6579 Frequency : occasional, higher if lots of files are accessed by one client
6580 Bugzilla : 6159, 6097
6581 Description: Client trips assertion regarding lsm mismatch/magic
6582 Details : While revalidating inodes the VFS looks up inodes with ifind()
6583 and in rare cases can find an inode that is being freed.
6584 The ll_test_inode() code will free the lsm during ifind()
6585 when it finds an existing inode and then the VFS later attaches
6586 this free lsm to a new inode.
6590 Bugzilla : 6422, 7030
6591 Description: MDS deadlock between mkdir and client eviction
6592 Details : Creating a new file via mkdir or mknod (starting a transaction
6593 and getting the ns lock) can deadlock with client eviction
6594 (gets ns lock and trying to finish a synchronous transaction).
6597 Frequency : occasional
6598 Description: While starting a server, the fsfilt_ext3 module could not be
6600 Details : CFS's improved ext3 filesystem is named ldiskfs for 2.6
6601 kernels. Previously, lconf would still use the ext3 name
6602 when trying to load modules. Now, it will correctly use
6603 ext3 on 2.4 and ldiskfs on 2.6.
6605 Severity : enhancement
6606 Description: The default stripe count has been changed to 1
6607 Details : The interpretation of the default stripe count (0, to lfs
6608 or lmc) has been changed to mean striping across a single
6609 OST, rather than all available. For general usage we have
6610 found a stripe count of 1 or 2 works best.
6612 Severity : enhancement
6613 Description: Add support for compiling against Cray portals.
6614 Details : Conditional compiling for some areas that are different
6618 Frequency : occasional
6619 Bugzilla : 6409, 6834
6620 Description: Creating files with an explicit stripe count may lead to
6621 a failed assertion on the MDS
6622 Details : If some OSTs are full or unavailable, creating files may
6623 trigger a failed assertion on the MDS. Now, Lustre will
6624 try to use other servers or return an error to the
6628 Frequency : occasional
6630 Description: Multiple concurrent overlapping read+write on multiple SMP nodes
6631 caused lock timeout during readahead (since 1.4.2).
6632 Details : Processes doing readahead might match a lock that hasn't been
6633 granted yet if there are overlapping and conflicting lock
6634 requests. The readahead process waits on ungranted lock
6635 (original lock is CBPENDING), while OST waits for that process
6636 to cancel CBPENDING read lock and eventually evicts client.
6638 Severity : enhancement
6640 Description: Initial enabling of flock support for clients
6641 Details : Implements fcntl advisory locking and file status functions.
6642 This feature is provided as an optional mount flag (default
6643 off), and is NOT CURRENTLY SUPPORTED. Not all types of record
6644 locking are implemented yet, and those that are are not guaranteed
6645 to be completely correct in production environments.
6646 mount -t lustre -o [flock|noflock] ...
6649 Frequency : occasional
6651 Description: OSTs running 2.4 kernels but with extents enabled might trip an
6652 assertion in the ext3 JBD (journaling) layer.
6653 Details : The b_committed_data struct is protected by the big kernel lock
6654 in 2.4 kernels, serializing journal_commit_transaction() and
6655 ext3_get_block_handle->ext3_new_block->find_next_usable_block()
6656 access to this struct. In 2.6 kernels there is finer grained
6657 locking to improve SMP performance of the JBD layer.
6661 Description: Changes the "SCSI I/O Stats" kernel patch to default to "enabled"
6663 -----------------------------------------------------------------------------
6665 2005-05-05 Cluster File Systems, Inc. <info@clusterfs.com>
6667 NOTE: Lustre 1.4.2 uses an incompatible network protocol than previous
6668 versions of Lustre. Please update all servers and clients to
6669 version 1.4.2 or later at the same time. You must also run
6670 "lconf --write-conf {config}.xml" on the MDS while it is stopped
6671 to update the configuration logs.
6673 - fix for HPUX NFS client breakage when NFS exporting Lustre (5781)
6674 - mdc_enqueue does not need max_mds_easize request buffer on send (5707)
6675 - swab llog records of type '0' so we get proper header size/idx (5861)
6676 - send llog cancel req to DLM cancel portal instead of cb portal (5515)
6677 - fix rename of one directory over another leaking an inode (5953)
6678 - avoid SetPageDirty on 2.6 (5981)
6679 - don't re-add just-being-destroyed locks to the waiting list (5653)
6680 - when creating new directories, inherit the parent's custom
6681 striping settings if present parent (3048)
6682 - flush buffers from cache before direct IO in 2.6 obdfilter (4982)
6683 - don't hold i_size_sem in ll_nopage() and ll_ap_refresh_count (6077)
6684 - don't hold client locks on temporary worklist from l_lru (5666)
6685 - handle IO errors in 2.6 obdfilter bio completion routine (6046)
6686 - automatically evict dead clients (5921)
6687 - Update file size properly in create+truncate+fstat case (6196)
6688 - Do not unhash mountpoint dentries, do not allow removal of
6690 - Avoid lock ordering deadlock issue with write/truncate (6203,5654)
6691 - reserve enough journal credits in fsfilt_start_log for setattr (4554)
6692 - ldlm_enqueue freed-export error path would always LBUG (6149,6184)
6693 - don't reference lr_lvb_data until after we hold lr_lvb_sem (6170)
6694 - don't overwrite last_rcvd if there is a *_client_add() error (6086)
6695 - Correctly handle reads of files with no objects (6243)
6696 - lctl recover will also mark a device active if deactivate used (5933)
6698 - by default create 1 inode per 4kB space on MDS, per 16kB on OSTs
6699 - allow --write-conf on an MDS with different nettype than client (5619)
6700 - don't write config llogs to MDS for mounts not from that MDS (5617)
6701 - lconf should create multiple TCP connections from a client (5201)
6702 - init scripts are now turned off by default; run chkconfig --on
6703 lustre and chkconfig --on lustrefs to use them
6704 - upcalls are no longer needed for clients to recover to failover
6706 - add --abort-recovery option to lconf to abort recovery on device
6708 - add support for an arbitrary number of OSTs (3026)
6709 - Quota support protocol changes.
6710 - forward compatibility changes to wire structs (6007)
6711 - rmmod NALs that might be loaded because of /etc/modules.conf (6133)
6712 - support for mountfsoptions and clientoptions to the Lustre LDAP (5873)
6713 - improved "lustre status" script
6714 - initialize blocksize for non-regular files (6062)
6715 - added --disable-server and --disable-client configure options (5782)
6716 - introduce a lookup cache for lconf to avoid repeated DB scans (6204)
6717 - Vanilla 2.4.29 support
6718 - increase maximum number of obd devices to 520 (6242)
6719 - remove the tcp-zero-copy patch from the suse-2.4 series (5902)
6720 - Quadrics Elan drivers are now included for the RHEL 3 2.4.21 and
6721 SLES 9 2.6.5 kernels
6722 - limit stripes per file to 160 (the maximum EA size) (6093)
6724 2005-03-22 Cluster File Systems, Inc. <info@clusterfs.com>
6727 - don't LASSERT in ll_release on NULL lld with NFS export (4655, 5760)
6728 - hold NS lock when calling handle_ast_error->del_waiting_lock (5746)
6729 - fix setattr mtime regression from lovcleanup merge (4829, 5669)
6730 - workaround for 2.6 crash in ll_unhash_aliases (5687, 5210)
6731 - small ext3 extents cleanups and fixes (5733)
6732 - improved mballoc code, several small races and bugs fixed (5733, 5638)
6733 - kernel version 43 - fix remove_suid bugs in both 2.4 and 2.6 (5695)
6734 - avoid needless client->OST connect, fix handle mismatch (5317)
6735 - fix DLM error path that led to out-of-sync client, long delays (5779)
6736 - support common vfs-enforced mount options (nodev,nosuid,noexec) (5637)
6737 - fix several locking issues related to i_size (5492,5624,5654,5672)
6738 - don't move pending lock onto export if it is already evicted (5683)
6739 - fix kernel oops when creating .foo in unlinked directory (5548)
6740 - fix deadlock in obdfilter statistics vs. object create (5811)
6741 - use time_{before,after} to avoid timer jiffies wrap (5882)
6742 - shutdown --force/--failover stability (3607,3651,4797,5203,4834)
6743 - Do not leak request if server was not able to process it (5154)
6744 - If mds_open unable to find parent dir, make that negative lookup(5154)
6745 - don't create new directories with extent-mapping (5909, 5936)
6747 - fix lustre/lustrefs init scripts for SuSE (patch from Scali, 5702)
6748 - don't hold the pinger_sem in ptlrpc_pinger_sending_on_import
6749 - change obd_increase_kms to obd_adjust_kms (up or down) (5654)
6750 - lconf, lmc search both /usr/lib and /usr/lib64 for Python libs (5800)
6751 - support for RHEL4 kernel on i686 (5773)
6752 - provide error messages when incompatible logs are encountered (5898)
6754 2005-02-18 Cluster File Systems, Inc. <info@clusterfs.com>
6755 * version 1.4.0.10 (1.4.1 release candidate 1)
6757 - don't keep a lock reference when lock is not granted (4238)
6758 - unsafe list practices (rarely) led to infinite eviction loop (4908)
6759 - add per-fs limit of Lustre pages in page cache, avoid OOM (4699)
6760 - drop import inflight refcount on signal_completed_replay error (5255)
6761 - unlock page after async write error during send (3677)
6762 - handle missing objects in filter_preprw_read properly (5265)
6763 - no transno return for symlink open, don't save no-trasno open (3440)
6764 - don't try to complete elan receive that already failed (4012)
6765 - free RPC server reply state on error (5406)
6766 - clean up thread from ptlrpc_start_thread() on error (5160)
6767 - readahead could read extra page into cache that wasn't ejected (5388)
6768 - prevent races in class_attach/setup/cleanup/detach (5260)
6769 - don't dereference de->d_inode after l_dput of de (5458)
6770 - use "int" for stripe value returned from lock_to_stripe (5544)
6771 - mballoc allocation and error-checking fixes in 2.6 (5504)
6772 - block device patches to fix I/O request sizes in 2.6 (5482)
6773 - look up hostnames for IB nals (5602)
6774 - 2.6 changed lock ordering of 2 semaphores, caused deadlock (5654)
6775 - don't start multiple acceptors for the same port (5277)
6776 - fix incorrect LASSERT in mds_getattr_name (5635)
6777 - export a proc file for general "ping" checking (5628)
6778 - fix "lfs check" to not block when the MDS is down (5628)
6780 - service request history (4965)
6781 - put {ll,lov,osc}_async_page structs in a single slab (4699)
6782 - create an "evict_client" /proc entry on OSTs, like the MDS has
6783 - fix mount usage message, return errors per mount(8) (5168)
6784 - change grep [] to grep "[]" in tests so they work in more UMLs
6785 - fix ppc64/x86_64 spec to use %{_libdir} instead of /usr/lib (5389)
6786 - remove ancient LOV_MAGIC_V0 EA support (5047)
6787 - add "disk I/Os in flight" and "I/O req time" stats in obdfilter
6788 - align r/w RPCs to PTLRPC_MAX_BRW_SIZE boundary for performance (3451)
6789 - allow readahead allocations to fail when low on memory (5383)
6790 - mmap locking landed again, after considerable improvement (2828)
6791 - add get_hostaddr() to lustreDB.py for LDAP support (5459)
6793 2004-11-23 Cluster File Systems, Inc. <info@clusterfs.com>
6796 - send OST transaction number in read/write reply to free req (4966)
6797 - don't ASSERT in ptl_send_rpc() if we run out of memory (5119)
6798 - lock /proc/sys/portals/routes internal state, avoiding oops (4827)
6799 - the watchdog thread now runs as interruptible (5246)
6800 - flock/lockf fixes (but it's still disabled, pending 5135)
6801 - don't use EXT3 constants in llite code (5094)
6802 - memory shortage at startup could cause assertion (5176)
6804 - reorganization of lov code
6805 - single portals codebase
6807 - add extents/mballoc support (5025)
6808 - direct I/O reads in the obdfilter (4048)
6809 - kernel patches from LNXI for 2.6 (bluesmoke, perfctr, mtd, kexec)
6811 tbd Cluster File Systems, Inc. <info@clusterfs.com>
6814 - send OST transaction number in read/write reply to free req (4966)
6815 - don't ASSERT in ptl_send_rpc() if we run out of memory (5119)
6816 - lock /proc/sys/portals/routes internal state, avoiding oops (4827)
6817 - the watchdog thread now runs as interruptible (5246)
6818 - handle missing objects in filter_preprw_read properly (5265)
6819 - unsafe list practices (rarely) led to infinite eviction loop (4908)
6820 - drop import inflight refcount on signal_completed_replay error (5255)
6821 - unlock page after async write error during send (3677)
6822 - return original error code on reconstructed replies (3761)
6823 - no transno return for symlink open, don't save no-trasno open (3440)
6825 - add pid to ldlm debugging output (4922)
6826 - bump the watchdog timeouts -- we can't handle 30sec yet
6827 - extra debugging for orphan dentry/inode bug (5259)
6829 2004-11-16 Cluster File Systems, Inc. <info@clusterfs.com>
6832 - fix TCP_NODELAY bug, which caused extreme perf regression (5134)
6833 - allocate qswnal tx descriptors singly to avoid fragmentation (4504)
6834 - don't LBUG on obdo_alloc() failure, use OBD_SLAB_ALLOC() (4800)
6835 - fix NULL dereference in /proc/sys/portals/routes (4827)
6836 - allow failed mdc_close() operations to be interrupted (4561)
6837 - stop precreate on OST before MDS would time out on it (4778)
6838 - don't send partial-page writes before EOF from client (4410)
6839 - discard client grant for sub-page writes on large-page clients (4520)
6840 - don't free dentries not owned by NFS code, check generation (4806)
6841 - fix lsm leak if mds_create_objects() fails (4801)
6842 - limit debug_daemon file size, always print CERROR messages (4789)
6843 - use transno after validating reply (3892)
6844 - process timed out requests if import state changes (3754)
6845 - update mtime on OST during writes, return in glimpse (4829)
6846 - add mkfsoptions to LDAP (4679)
6847 - use ->max_readahead method instead of zapping global ra (5039)
6848 - don't interrupt __l_wait_event() during strace
6850 - add software watchdogs to catch hung threads quickly (4941)
6851 - make lustrefs init script start after nfs is mounted
6852 - fix CWARN/ERROR duplication (4930)
6853 - return async write errors to application if possible (2248)
6854 - add /proc/sys/portal/memused (bytes allocated by PORTALS_ALLOC)
6855 - print NAL number in %x format (4645)
6856 - update barely-supported suse-2.4.21-171 series (4842)
6857 - support for sles 9 %post scripts
6858 - support for building 2.6 kernel-source packages
6859 - support for sles km_* packages
6861 2004-10-07 Cluster File Systems, Inc. <info@clusterfs.com>
6864 - ignore -ENOENT errors in osc_destroy (3639)
6865 - notify osc create thread that OSC is being cleaned up (4600)
6866 - add nettype argument for llmount in #5d in conf-sanity.sh (3936)
6867 - reconstruct ost_handle() like mds_handle() (4657)
6868 - create a new thread to do import eviction to avoid deadlock (3969)
6869 - let lconf resolve symlinked-to devices (4629)
6870 - don't unlink "objects" from directory with default EA (4554)
6871 - hold socknal file ref over connect in case target is down (4394)
6872 - allow more than 32000 subdirectories in a single directory (3244)
6873 - fix blocks count for O_DIRECT writes (3751)
6874 - OST returns ENOSPC from object create when no space left (4539)
6875 - don't send truncate RPC if file size isn't changing (4410)
6876 - limit OSC precreate to 1/2 of value OST considers bogus (4778)
6877 - bind to privileged port in socknal and tcpnal (3689)
6879 - rate limit CERROR/CWARN console message to avoid overload (4519)
6880 - GETFILEINFO dir ioctl returns LOV EA + MDS stat in 1 call (3327)
6881 - basic mmap support (3918)
6882 - kernel patch series update from b1_4 (4711)
6884 2004-09-16 Cluster File Systems, Inc. <info@clusterfs.com>
6887 - avoid crash during MDS cleanup with OST shut down (2775)
6888 - fix loi_list_lock/oig_lock inversion on interrupted IO (4136)
6889 - don't use bad inodes on the MDS (3744)
6890 - dynamic object preallocation to improve recovery speed (4236)
6891 - don't hold spinlock over lock dumping or change debug flags (4401)
6892 - don't zero obd_dev when it is force cleaned (3651)
6893 - print grants to console if they go negative (4431)
6894 - "lctl deactivate" will stop automatic recovery attempts (3406)
6895 - look for existing locks in ldlm_handle_enqueue() (3764)
6896 - don't resolve lock handle twice in recovery avoiding race (4401)
6897 - revalidate should check working dir is a directory (4134)
6899 - don't always mark "slow" obdfilter messages as errors (4418)
6901 2004-08-24 Cluster File Systems, Inc. <info@clusterfs.com>
6904 - don't close LustreDB during write_conf until it is done (3860)
6905 - fix typo in lconf for_each_profile (3821)
6906 - allow dumping logs from multiple threads at one time (3820)
6907 - don't allow multiple threads in OSC recovery (3812)
6908 - fix debug_size parameters (3864)
6909 - fix mds_postrecov to initialize import for llog ctxt (3121)
6910 - replace config semaphore with spinlock (3306)
6911 - be sure to send a reply for a CANCEL rpc with bad export (3863)
6912 - don't allow enqueue to complete on a destroyed export (3822)
6913 - down write_lock before checking llog header bitmap (3825)
6914 - recover from lock replay timeout (3764)
6915 - up llog sem before sending rpc (3652)
6916 - reduce ns lock hold times when setting kms (3267)
6917 - change a dlm LBUG to LASSERTF, to maybe learn something (4228)
6918 - fix NULL deref and obd_dev leak on setup error (3312)
6919 - replace some LBUG about llog ops with error handling (3841)
6920 - don't match INVALID dentries from d_lookup and spin (3784)
6921 - hold dcache_lock while marking dentries INVALID and hashing (4255)
6922 - fix invalid assertion in ptlrpc_set_wait (3880)
6924 - add libwrap support for the TCP acceptor (3996)
6925 - add /proc/sys/portals/routes for non-root route listing (3994)
6926 - allow setting MDS UUID in .xml (2580)
6927 - print the stack of a process that LBUGs (4228)
6929 2004-07-14 Cluster File Systems, Inc. <info@clusterfs.com>
6932 - don't cleanup request in ll_file_open() on failed MDS open (3430)
6933 - make sure to unset replay flag from failed open requests (3440)
6934 - if default stripe count is 0, use OST count for inode size (3636)
6935 - update parent mtime/ctime on client for create/unlink (2611)
6936 - drop dentry ref in ext3_add_link from open_connect_dentry (3266)
6937 - free recovery state on server during a forced cleanup (3571)
6938 - unregister_reply for resent reqs (3063)
6939 - loop back devices mounting and status check on 2.6 (3563)
6940 - fix resource-creation race that can provoke i_size == 0 (3513)
6941 - don't try to use bad inodes returned from MDS/OST fs lookup (3688)
6942 - more debugging for page-accounting assertion (3746)
6943 - return -ENOENT instead of asserting if ost getattr+unlink race (3558)
6944 - avoid deadlock after precreation failure (3758)
6945 - fix race and lock order deadlock in orphan handling (3450, 3750)
6946 - add validity checks when grabbing inodes from l_ast_data (3599)
6948 - add /proc/.../recovery_status to obdfilter (3428)
6949 - lightweight CDEBUG infrastructure, debug daemon (3668)
6950 - change default OSC RPC parameters to be better on small clusters
6951 - turn off OST read cache for files smaller than 32MB
6952 - install man pages and include them in rpms (3100)
6953 - add new init script for (un)mounting lustre filesystems (2593)
6954 - run chkconfig in %post for init scripts (3701)
6955 - drop scimac NAL (unmaintained)
6957 2004-06-17 Cluster File Systems, Inc. <info@clusterfs.com>
6960 - clean kiobufs before and after use (3485)
6961 - strip trailing '/'s before comparing paths with /proc/mounts (3486)
6962 - remove assertions to work around "in-flight rpcs" recovery bug (3063)
6963 - change init script to fail more clearly if not run as root (1528)
6964 - allow clients to reconnect during replay (1742)
6965 - fix ns_lock/i_sem lock ordering deadlock for kms update (3477)
6966 - don't do DNS lookups on NIDs too small for IP addresses (3442)
6967 - re-awaken ptlrpcd if new requests arrive during check_set (3554)
6968 - fix cond_resched (3554)
6969 - only evict unfinished clients after recovery (3515)
6970 - allow bulk resend, prevent data loss (3570)
6971 - dynamic ptlrpc request buffer allocation (2102)
6972 - don't allow unlinking open directory if it isn't empty (2904)
6973 - set MDS/OST threads to umask 0 to not clobber client modes (3359)
6974 - remove extraneous obd dereference causing LASSERT failure (3334)
6975 - don't use get_cycles() when creating temp. files on the mds (3156)
6976 - hold i_sem when setting i_size in ll_extent_lock() (3564)
6977 - handle EEXIST for set-stripe, set proper directory name (3336)
6979 - servers can dump a log evicting a client - lustre.dump_on_timeout=1
6980 - fix ksocknal_fmb_callback() error messages (2918)
6982 2004-05-27 Cluster File Systems, Inc. <info@clusterfs.com>
6985 - don't copy lvb into (possibly NULL) reply on error (2983)
6986 - don't deref dentry after dput, don't free lvb on error (2922)
6987 - use the kms to determine writeback rpc length (2947)
6988 - increment oti_logcookies when osc is inactive (2948)
6989 - update client's i_blocks count via lvb messages (2543)
6990 - handle intent open/close of special files properly (1557)
6991 - mount MDS with errors=remount-ro, like obdfilter (2009)
6992 - initialize lock handle to avoid ASSERT on error cleanup (3057)
6993 - don't use cancelling-locks' kms values (2947)
6994 - use highest lock extent for kms, not last one (2925)
6995 - don't dereference ERR_PTR() dentry in error handling path (3107)
6996 - fix thread race in portals_debug_dumplog() (3122)
6997 - create lprocfs device entries at setup instead of at attach (1519)
6998 - common AST error handler, don't evict client on completion race (3145)
6999 - zero nameidata in detach_mnt in 2.6 (3118)
7000 - verify d_inode after revalidate_special is valid in 2.6 (3116)
7001 - use lustre_put_super() to handle zconf unmounts in 2.6 (3064)
7002 - initialize RPC timeout timer earlier for 2.6 (3219)
7003 - don't dereference NULL reply buffer if mdc_close was never sent (2410)
7004 - print nal/nid for unknown nid (3258)
7005 - additional checks for oscc recovery before doing precreate (3284)
7006 - fix ll_extent_lock() error return code for 64-bit systems (3043)
7007 - don't crash in mdc_close for bad permissions on open (3285)
7008 - zero i_rdev for non-device files (3147)
7009 - clear page->private before handing to FS, better assertion (3119)
7010 - tune the read pipeline (3236)
7011 - fix incorrect decref of invalidated dentry (2350)
7012 - provide read-ahead stats and refine rpc in flight stats (3328)
7013 - don't hold journal transaction open across create RPC (3313)
7014 - update atime on MDS at close time (3265)
7015 - close LDAP connection when recovering to avoid server load (3315)
7016 - update iopen-2.6 patch with fixes from 2399,2517,2904 (3301)
7017 - don't leak open file on MDS after open resend (3325)
7018 - serialize filter_precreate and filter_destroy_precreated (3329)
7019 - loop device shouldn't call sync_dev() for nul device (3092)
7020 - clear page cache after eviction (2766)
7021 - resynchronize MDS->OST in background (2824)
7022 - refuse to mount the same filesystem twice on same mountpoint (3394)
7023 - allow llmount to create routes for mounting behind routers (3320)
7024 - push lock cancellation to blocking thread for glimpse ASTs (3409)
7025 - don't call osc_set_data_with_check() for TEST_LOCK matches (3159)
7026 - fix rare problem with rename on htree directories (3417)
7028 - allow default OST striping configuration per directory (1414)
7029 - fix compilation for qswnal for 2.6 kernels (3125)
7030 - increase maximum number of MDS request buffers for large systems
7031 - change liblustreapi to be useful for external progs like lfsck (3098)
7032 - increase local configuration timeout for slow disks (3353)
7033 - allow configuring ldlm AST timeout - lustre.ldlm_timeout=<seconds>
7035 2004-03-22 Cluster File Systems, Inc. <info@clusterfs.com>
7038 - fixes for glimpse AST timeouts / incorrectly 0-sized files (2818)
7039 - don't overwrite extent policy data in reply if lock was blocked (2901)
7040 - drop filter export grants atomically with removal from device (2663)
7041 - del obd_self_export from work_list in class_disconnect_exports (2908)
7042 - don't LBUG if MDS recovery times out during orphan cleanup (2530)
7043 - swab reply message in mdc_close, other PPC fixes (2464)
7044 - fix destroying of named logs (2325)
7045 - overwrite old logs when running lconf --write_conf (2264)
7046 - bump LLOG_CHUNKSIZE to 8k to allow for larger clusters (2306)
7047 - fix race in target_handle_connect (2898)
7048 - mds_reint_create() should take same inode create lock (2926)
7049 - correct journal credits calculated for CANCEL_UNLINK_LOG (2931)
7050 - don't close files for self_export to avoid uninitialized obd (2936)
7051 - allow MDS with the same name as client node (2939)
7052 - hold dentry reference for closed log files for unlink (2325)
7053 - reserve space for all logs during transactions (2059)
7054 - don't evict page beyond end of stripe extent (2925)
7055 - don't oops on a deleted current working directory (2399)
7056 - handle hard links to targets without a parent properly (2517)
7057 - don't dereference NULL lock when racing during eviction (2867)
7058 - don't grow lock extents when lots of conflicting locks (2919)
7060 2004-03-04 Cluster File Systems, Inc. <info@clusterfs.com>
7063 - account for cache space usage on clients to avoid data loss (974)
7064 - lfsck support in lustre kernel code (2349)
7065 - reduce journal credits needed for BRW writes (2370)
7066 - orphan handling to avoid losing space on client/server crashes
7067 - ptlrpcd can be blocked, stopping ALL progress (2477)
7068 - use lock value blocks to assist in proper KMS, faster stat (1021)
7069 - takes i_sem instead of DLM locks internally on obdfilter (2720)
7070 - recovery for initial connections (2355)
7071 - fixes for mds_cleanup_orphans (1934)
7072 - abort_recovery crashes MDS in b_eq (mds_unlink_orphan) (2584)
7073 - block all file creations until orphan recovery completes (1901)
7074 - client remove rq_connection from request struct (2423)
7075 - conf-sanity test_5, proper cleanup in umount log not availale (2640)
7076 - recovery timer race (2670)
7077 - mdc_close recovey bug (2532)
7078 - ptlrpc cleanup bug (2710)
7079 - mds timeout on local locks (2588)
7080 - namespace lock held during RPCs (2431)
7081 - handle interrupted sync write properly (2503)
7082 - don't try to handle a message that hasn't been replied to (2699)
7083 - client assert failure during cleanup after abort recovery (2701)
7084 - leak mdc device after failed mount (2712)
7085 - ptlrpc_check_set allows timedout requests to complete (2714)
7086 - wait for inflight reqs when ptlrpcd finishes (2710)
7087 - make sure unregistered services are removed from the srv_list
7088 - reset bulk XID's when resending them (caught by 1138 test)
7089 - unregister_bulk after timeout
7090 - fix lconf error (2694)
7091 - handle write after unfinished setstripe, stripe-only getstripe (2388)
7092 - readahead locks pages, leaves pending causing memory pressure (2673)
7093 - increase OST request buffers to 4096 on large machines (2729)
7094 - fix up permission of existing directories in simple_mkdir (2661)
7095 - init deleted item, add assertions ptlrpc_abort_inflight() (2725)
7096 - don't assign transno to errored transactions (2742)
7097 - don't delete objects on OST if given a bogus objid from MDS (2751)
7098 - handle large client PAGE_SIZE readdir on small PAGE_SIZE MDS (2777)
7099 - if rq_no_resend, then timeout request after recovery (2432)
7100 - fix MDS llog_logid record size, 64-bit array alignment (2733)
7101 - don't call usermode_helper from ptlrpcd, DEFAULT upcall (2773)
7102 - put magic in mount.lustre data, check for bad/NULL mount data (2529)
7103 - MDS recovery shouldn't delete objects that it has given out (2730)
7104 - if enqueue arrives after completion, don't clobber LVB (2819)
7105 - don't unlock pages twice when trigger_group_io returns error (2814)
7106 - don't deref NULL rq_repmsg if ldlm_handle_enqueue failed (2822)
7107 - don't write pages to disk if there was an error (1450)
7108 - don't ping imports that have recovery disabled (2676)
7109 - take buffered bytes into account when balancing socknal conn (2817)
7110 - hold a DLM lock over readdir always, use truncate_inode_pages (2706)
7111 - reconnect unlink llog connection after MDS reconnects to OST (2816)
7112 - remove little-endian swabbing of llog records (1987)
7113 - set/limit i_blksize to LL_MAX_BLKSIZE on client (2884)
7114 - retry reposting request buffers if they fail (1191)
7115 - grow extent at grant time to avoid granting a revoked lock (2809)
7116 - lock revoke doesn't evict page if covered by a second lock (2765)
7117 - disable VM readahead to avoid reading outside lock extents (2805)
7119 - return LL_SUPER_MAGIC from statfs for the filesystem type (1972)
7120 - updated kernel patches for hp-2.4.20 kernel (2681)
7122 2004-02-07 Cluster File Systems, Inc. <info@clusterfs.com>
7125 - fix truncated write corruption (2366)
7126 - fix for failed assertion in iopen_connect_dentry (1792,2517)
7128 - don't flag the ptlrpcd thread with PF_MEMALLOC (2636)
7129 - ensure len(uuid) < 37 in lmc (1171)
7130 - fix ia64 OOPS in llog_test (2255)
7131 - zero end of page at obdfilter for partial page writes (2648)
7132 - don't leave stale dentries around after renames (bug 2428)
7133 - fix timeouts when evicting a client with a single lock held (2642)
7134 - set deadline for the initial HELLO message to drain (2634)
7135 - print out dotted-quad IP addresses in the socknal (2302)
7137 - additional debugging for MDS client eviction problem (2443)
7138 - fix mkfsoptions support for osts (2603, 2604)
7140 2004-01-27 Cluster File Systems, Inc. <info@clusterfs.com>
7143 - add series for the vanilla 2.6.0 kernel
7144 - add series for the vanilla 2.4.24 kernel
7145 - add series for a cray x86/64 UL kernel drop
7146 - fix xattr patches for the vanilla 2.4.19 series
7148 - generate true UUIDs in lmc (1171)
7149 - have portals stack dumping break in UML (2466)
7150 - avoid bad dchild deref; avoid inum lock w/o creation (2362)
7151 - allocate with _NOFS in ldlm to avoid deadlock (1933)
7152 - wake callback waiting threads on client eviction (2460)
7153 - Add --ptldebug and --subsystem to lmc (1719)
7154 - update assertion to allow safe interrupt allocation
7155 - set rq_no_resend for cancel requests (2432)
7156 - recalculate ptlrpcd timeout after resend (2494)
7157 - call vfs_rmdir when removing pending directories (2368)
7158 - fix renaming a file to itself (2429)
7159 - lmc creates a default one-stripe lov (2454)
7160 - expand procfs space to handle large clusters (2326)
7161 - increase UML stack to avoid overflow
7162 - update lconf's list of debug and subsystem masks
7163 - fix lfs find --obd (2510)
7164 - /proc tunable for disabling filter read caching (2591)
7165 - stop rpm packages from altering slapd.conf (2301)
7166 - disable nagle in the socknal under 0conf (2578)
7167 - choose mds inode size based on stripe count (2572)
7168 - fix kernel-source rpm problems (2516)
7170 - add --disable-doc to avoid pdf generation (2421)
7171 - update documentation, tests, type-os, comments
7172 - avoid format warnings on ia64
7173 - remove the TOE NAL
7174 - tiny code cleanups by removing unused fields
7176 2004-01-07 Cluster File Systems, Inc. <info@clusterfs.com>
7179 - fix obvious semaphore misuse in as-yet-unused setattr path (2348)
7180 - remove the most blatant lies from BUILDING file (2371)
7181 - change default debug level to reasonable production setting
7182 - reduce client side cache size to reduce cache flush time
7183 - reduce max RPCs in flight to avoid unnecessary file fragmentation
7184 - make TCP zerocopy and pinger support enabled by default (2476)
7185 - sync writes completed after process exits caused crashes (2319)
7186 - maintain correct mount count on the MDS (2356)
7187 - backout 1557, because 2316 wasn't really fixed
7188 - better file I/O statistics gathering in /proc
7189 - don't take unnecessary, deadlock-inducing bug in readpage (2383)
7190 - another kernel patch to fix zero-copy TCP function export
7191 - don't take duplicate lock when processing re-sent getattr (2420)
7192 - lctl uses obd_self_export instead of creating new conn (2353)
7193 - MDS/OST recovery case which requires object creation asserted (2425)
7194 - move lfs from /usr/sbin to /usr/bin in packages
7195 - fix race between mds_client_add and mds_client_free (2417)
7196 - use kmalloc instead of slabs in portals (2430)
7197 - don't create duplicate records when a failover MDS is present (2442)
7198 - remove unnecessary mount age check (2332)
7199 - don't remove directory inodes from locks prematurely (2451)
7200 - don't break if MDS service name is the same as hostname (2103)
7201 - fix races in client write RPC generation when cache full (2482)
7203 2003-12-13 Cluster File Systems, Inc. <info@clusterfs.com>
7206 - remove now-unused request->rq_obd (278)
7207 - if an allocation fails, print out how much memory we've used (1933)
7208 - use PORTAL_SLAB_ALLOC for structures, to get GFP_MEMALLOC (1933)
7209 - add the "configurable stack size" patch to most series files (1256)
7210 - ability to write large log records, for 100+ OST configs (2306)
7211 - fix NULL deref when filter_prep fails (2314)
7212 - fix operator precedence error in filter_sync
7213 - dynamic allocation of socknal TX descriptors (2315)
7214 - fix a missed case in the GFP_MEMALLOC patch, can cause deadlock (2310)
7215 - fix gcc 2.96 compilation problem in xattr kernel patch (2294)
7216 - ensure that CWARN messages in Portals always get to the syslog
7217 - __init/__exit are not for prototype decls (ldlm_init/exit)
7218 - x86-64 compile warning fixes
7219 - fix gateway LMC keyword conflict (2318)
7220 - fix MDS lock inversions in getattr/reint paths (1844)
7221 - fix a rare lock re-ordering bug, which caused deadlock (2322)
7222 - fix i_sem/journal inversion in fsfilt_ext3_write_record (2306)
7223 - DLM race condition prevented some lock evictions (2328)
7224 - ENOMEM detection and retry on socknal sends (2230)
7225 - use GFP_NOFS throughout Lustre, to combat ENOMEM (2230)
7226 - move osc_rpcd into ptlrpc, for use in MDC and others (2329)
7227 - protect MDS inode fsdata with stronger locking; fixes assertion (2313)
7228 - better error messages when a client is rejected during recovery (1505)
7229 - avoid cancelling locks which were never granted, after failure (2330)
7230 - fix i_sem/journal inversion in mds_client_add (2333)
7231 - fix truncate/getattr lock cycle deadlock (2334)
7232 - use rpcd to send close; allows resend after timeout, avoid leak (1897)
7233 - fix two rare exit paths which could leak an l_lock() ref (2321)
7234 - fencepost error in MDS/OST orphan recovery (2226)
7235 - make log record alignment 8 bytes (1988)
7236 - lstripe now fails when requested offset > ost_count (2237)
7237 - ensure that all kernel series have a complete list.h (1607)
7238 - fix crashes in special-file operations (2316)
7239 - lctl create/brw OID mismatch, caused by obsolete filter loop (2339)
7241 - allow configurable automake binary, for testing new versions
7242 - small update to the lfs documentation
7244 2003-12-03 Cluster File Systems, Inc. <info@clusterfs.com>
7246 * fix negative export reference count in fsfilt_sync (2312)
7248 2003-12-01 Cluster File Systems, Inc. <info@clusterfs.com>
7249 * release candidate 0.9.1
7251 - orphans are moved into the PENDING directory for possible recovery
7252 - replayed opens now open by fid for orphan/rename safety (1042)
7253 - last close of an orphan inode generates a transno (683)
7254 - chdir() and mount() now pin the directory entry (1020)
7255 - avoid CERROR in normal ll_setattr_raw() error case (1500)
7256 - discard very old requests without processing them (1502)
7257 - remove some common, well-understood CERRORs (1505)
7258 - require O_DIRECT I/O to be page-sized to workaround IA64 crash (1609)
7259 - clear "grant" flags in OST replies until OST grant code lands (1644)
7260 - fix read performance by not clobbering i_blksize on client (1598)
7261 - fix __ldlm_handle2lock oops by not dereferencing lock after PUT (1625)
7262 - make LRU size a /proc tunable, clears locks when reduced (707)
7263 - fix some lprocfs rot that prevented ptlbd from loading (1732)
7264 - server locks take references on exports now (1558)
7265 - build fixes for 2.4.20-rh trees (1663)
7266 - return an error from lov_create if all OSCs are inactive (1751)
7267 - fix import levels when a reconnect happens without a timeout (1597)
7268 - exit early from mds_open if we get a lookup error (1749)
7269 - partial page read at EOF wouldn't wait for disk before sending (1642)
7270 - avoid NULL deref in obdfilter when reading page past EOF (1592)
7271 - avoid LASSERT in ll_intent_lock if server failed very early (1090)
7272 - fix LBUG in ll_it_open_error with rc = -2 (1861)
7273 - write/truncate lock inversion (1639)
7274 - Don't auto-load obdclass, portals modules during cleanup (1495)
7275 - fix timestamps from jumping to "now" (1763)
7276 - extra journal assertions (1648)
7277 - add an extra multiunlink test (1771)
7278 - fix read_record/write_record API (1776)
7279 - fix leak of offset_extent, possible incorrect i_size later (1772)
7280 - fix lasserts in mis-matched transnos during open-unlink testing (1541)
7281 - Debugging for the kqswnal_get_idle_tx problems (1820)
7282 - Allow recovery to be attempted multiple times (1536)
7283 - Write out MDS last_rcvd file after it is first created (1600)
7284 - Fix tx_descriptor leak in failed transmit situations (1827)
7285 - ext3 journaling fixes for assertion failure after IO error (1871)
7286 - class_export_put() on freed export after completion AST error (1896)
7287 - Fix revalidate looping in VFS (1322)
7288 - Don't access a freed export during MDS_REINT timeout (1521)
7289 - Add open-unlink recovery support on the MDS (1673,1764)
7290 - Return an error if no MDS data was read from last_rcvd (1946)
7291 - Fix for lookup "." or ".." crash on error (1932,1931,1935)
7292 - Don't setup a disk device that doesn't match exported UUID (317)
7293 - Reduce bulk RPC timeout to avoid cascading client/OST failures (1845)
7294 - avoid committing NULL handle in force close
7295 - local.sh is now a one-stripe LOV configuration
7296 - POSIX utime.4 -EPERM on FIFO not owned by user (56)
7297 - fix ext3 htree duplicate directory entry corruption (1516)
7298 - POSIX creat.13, fstat.1, open.18, stat.3 new file atime/mtime (2020)
7299 - update to new LOV EA format (2097)
7300 - interoperability for different PAGE_SIZE/wordsize (686,1821,1343,2042)
7302 2003-06-15 Phil Schwan <phil@clusterfs.com>
7305 - imports and exports cleanup too early, need refcounts (349, 879, 1045)
7306 - per-import/export recovery handling (958, 931, 959)
7307 - multiple last-rcvd slots, for serving multiple FSes (949)
7308 - connections are again shared between multiple imp/exports (963, 964)
7309 - "umount -f" would hang if any requests needed to be sent (393, 978)
7310 - avoid pinning large req buffer by copying for queued messages (989)
7311 - add "uuid" to "lctl device" command to help upcalls (991)
7312 - "open" RPCs with transnos would confuse recovery counters (1037)
7313 - do proper endian conversion of all wire messages (288, 340, 891)
7314 - remove OST bulk get LBUGs, fix ost_brw_write cleanup (1126)
7315 - call waiting locks callback from LDLM recovery thread (1127, 1151)
7316 - fix ptlrpc_connection leak in target_handle_connect (1174)
7317 - fix import refcounting bug in OST and MDS cleanup (1134)
7318 - if an invalid-at-open-time OSC returned before close(), LBUG (1150)
7319 - fix very unlikely obd_types race condition (501)
7320 - remove osc_open hack for echo_client (1187)
7321 - we leaked exports/dlmimps for forcibly disconnected clients (1143)
7322 - a failure in read_inode2 leads to deadlock (1139)
7323 - cancel ack-locks as soon as transaction is committed (1072)
7324 - fix major leaks and crashes in the bulk I/O path (937, 1057)
7325 - make sure to commitrw after any preprw to avoid deadlock (1162)
7326 - failing to execute a file in a lustre FS would lock inode (1203)
7327 - small DEBUG_REQ fix to avoid dereferencing a NULL (1227)
7328 - don't ASSERT while cleaning up an incompletely-setup obd (1248)
7329 - obd_uuid2tgt would walk off the end of the list (1255)
7330 - on IA64 the osc would give portals incorrect bulk size (1258)
7331 - fix debug daemon ioctl interface; allows daemon on ia64 (1274)
7332 - fix lock inversion caused by new llite matching code (1282)
7333 - limit the number of dirty pages on a client to 10MB (1286)
7334 - timed out locks were not being corrected cancelled (1289)
7335 - fix O_DIRECT above 4GB on IA-32 (1292)
7336 * major user-visible changes
7337 - fail out/fail over policy now controlled by the upcall (993)
7339 - add OBD_PING to check server availability and failure (954)
7340 - lustre messages are now sent in sending host order (288, 340, 891)
7341 - add eadatalen to MDS getattr reply (340)
7342 - OST read replies may contain second buffer, with per-page status (593)
7344 2003-03-11 Phil Schwan <phil@clusterfs.com>
7347 - LDLM_DEBUG macro fix, for gcc 3.2 (850)
7348 - failed open()s could cause deadlock; fixed (867, 869)
7349 - stop cancelling OST locks when files are closed (481)
7350 - overlapping XID spaces caused network corruption (851, 853)
7351 - fix unsafe fsfilt counter arithmetic; change to atomic_t
7352 - setattr_raw added, to do single-RPC, server-side setattrs
7353 - lmc/lconf syntax change for OST UUIDs
7354 - fix crashy race condition between ptlrpc_free_req and osc_close
7355 - don't use request in mdc_enqueue if we hit a timeout (889)
7356 - don't set the inode i_size for regular files from the MDS (896)
7357 - handle out of order completion AST (842)
7358 - don't LBUG if a lock request times out after receiving AST (913)
7359 - avoid d_rehash race in ll_find_alias by rehashing inside dcache_lock
7360 - if a bad lock AST arrives, send an error instead of dropping entirely
7361 - return 0 from revalidate2 if ll_intent_lock returns -EINTR (912)
7362 - fix leak in bulk IO when only partially completed (899, 900, 926)
7363 - fix O_DIRECT for ia64 (55)
7364 - (almost) eliminate Lustre-kernel-thread effects on load average (722)
7365 - C-z after timeout could hang a process forever; fixed (977)
7367 - client-side I/O cache (678, 924, 929, 941, 970)
7369 - READPAGE and SETATTRs which don't take server-side locks get
7372 2003-02-11 Phil Schwan <phil@clusterfs.com>
7375 - Fix ldlm_lock_match on the MDS to avoid matching remote locks (592)
7376 - Fix fsfilt_extN_readpage() to read a full page of directory
7377 entries, or fake the remainder if PAGE_SIZE != blocksize (500)
7378 - Avoid extra mdc_getattr() in ll_intent_lock when possible (534, 604)
7379 - Fix imbalanced LOV object allocation and out-of-bound access (469)
7380 - Most intent operations were removed, in favour of a new RPC mode
7381 that does a single RPC to the server and bypasses most of the VFS
7382 - All LDLM resource ID arrays were removed in favour of ldlm_res_id
7383 - Aggressively cancel local locks on DLM servers
7384 - mds_reint_unlink sends EA to the client if it's the last nlink.
7385 client uses that EA to unlink OST objects.
7386 - mds_reint_{rename,unlink,link} were rewritten to take ordered locks
7387 - recursive symlinks were fixed (439)
7388 - fixed NULL deref in DEBUG_REQ
7389 - filter_update_lastobjid no longer calls sync, which annoyed extN
7390 - fixed multi-client small-writes to a single file problem (445)
7391 - fixed mtime updates during file writes (607)
7392 - fixed vector writes on obdfilter causing problems when ENOSPC (670)
7393 - fixed bug in obd_brw_read/write() (under guise of testing 367)
7394 - fixed Linux OST size reporting problem (444, 656)
7395 - OST now updates object mtime with writes or setattr (607, 619)
7396 - client verifies file size before zeroing page past EOF (445)
7397 - OST now writes last allocated objid to disk with allocation (108)
7398 - LOV on echo now works (409)
7400 - mds_reint_unlink sends a new buffer, with the EA included. this
7401 buffer is only valid if body->valid & OBD_MD_FLEASIZE, which is only
7402 set if a regular file was being unlinked, and it was the last link
7403 - use PtlGet from the target for bulk writes (315)
7404 - OST now updates object mtime with writes or setattr (607, 619)
7405 - LDLM now has a grant-time callback to revalidate locked items, if
7407 - Many MDS operations were reorganized to combat race conditions
7409 - Merge b_intel branch (updated lprocfs code) - now at /proc/fs/lustre
7410 - configure check to avoid gcc version 2.96 20000731-2.96-98) (606)
7412 2003-01-06 Andreas Dilger <adilger@clusterfs.com>
7415 - Fully reactivate OST imports after reconnection (512, others)
7416 - Make sure client sees our -ENOTCONN from mds_handle (513 - partial)
7417 - More graceful error handling for truncating on dead OST (515)
7418 - Don't error out unless we're actually accessing dead stripes (474)
7419 - Fix garbage sizes when stripes are missing (410)
7420 - LRU counters were broken, causing constant lock purge (433, 432)
7421 - garbage on read from stripes with failed OSTs (441)
7422 - mark OSCs as active before reconnecting during recovery (438)
7423 - lov_enqueue and lov_cancel need to handle inactive OSTs (403)
7424 - lfind did not preserve OST order in output (443)
7425 - symlinks cause hung clients, incorrect data (439)
7426 - stop dereferencing request after dropping refcount (457)
7427 - don't LASSERT(spin_is_locked) on non-SMP (455)
7428 - fixes for many rename() bugs
7429 - fstat didn't correctly synchronize attributes (399)
7430 - server must handle lock cancellation during blocking AST prep (487)
7431 - bulk descriptors were free()d too soon (511)
7432 - fix paths in lconf, which would load incorrect modules (451, 507)
7433 - fix confusing lconf 'host not found' error message (386)
7434 - fix lock order deadlock on OST (O/R i_sem before journal ops, 478)
7435 - fix race condition in mdc_blocking_ast() for inode access (526)
7436 - fix lov_unpackmd() unpacking wrong number of stripes (537)
7437 - fix lov_set_osc_active() marking wrong OSC inactive (440)
7438 - fix bad lstripe lov_unpackmd() assertion (fix layering too) (527)
7439 - fix multiple writes of stripe MD to MDS (358, maybe 519)
7440 - fix lstripe in several ways (kernel side) (527)
7441 - fix request leak in ldlm_cli_enqueue (262)
7442 - incorrect OSC was marked inactive after OST failure
7443 - call mds_fs_cleanup before unmounting filesystem (524)
7444 - fix races between taking ns_lock and ldlm_lock_change_resource
7445 - fix races updating LOV export open file list
7446 - fix lov_enqueue error path, avoid decref-ing bad lock handle
7447 - fix recovery NULL deref in ldlm_cli_cancel_unused
7448 - fix some DLM races by using new hash table for lock handles (419)
7449 - permit the client to specify desired inodes, at replay
7450 - duplicate requests when we queue them for replay reintegration
7451 - fix last_rcvd offset calculation
7452 - sync after each recovered transaction, so we always make progress
7453 - never, not always, ERESTART requests without transnos
7454 - store the lov_desc in the MDS, so we don't depend on getlovinfo to
7456 - skip replay if the MDS says that the client is already connected
7457 - don't check for a recovery-enabled export to match lctl's UUID
7458 - don't INC_USE_COUNT for phantom exports
7459 - don't crash when cleaning up phantom exports (567)
7460 - don't double-finish or set replay data for errored mdc_open requests
7461 - abort requests when they time out, so we don't get old replies
7462 - send/receive replies for AST messages again
7463 - if the client says that it doesn't have the lock, cancel it on the
7465 - if we timeout during I/O, don't try to cancel an in-use lock; instead
7466 mark it as destroyed, it will all work out when decref is called
7467 - fix module use counts (22, 581)
7469 - ASTs now expect a reply (server cancels lock on error reply)
7471 2002-12-02 Andreas Dilger <adilger@clusterfs.com>
7474 - fix many simultaneous client startup (392)
7475 - fix dentry->d_it clobbering
7476 - credentials weren't being shipped for readdir/getattr operations
7477 - remove invalid assertions triggered during some concurrent MD
7479 - proper Lustre versions added (336, 389)
7480 - fix memory leak for create error case (398)
7481 - fix LOV locking bug that would get cli/srv out of sync
7482 - fix echo client over LOV (409)
7483 - fix dbench 2, extN refcount problem (170, 258, 356, 418)
7484 - fix double-O_EXCL intent crash (424)
7485 - avoid sending multiple lock CANCELs (352)
7487 - MDS can do multi-client recovery (modulo bugs in new code)
7489 - many updates, edits, cleanups
7491 2002-11-18 Phil Schwan <phil@clusterfs.com>
7494 - fix null d_it dereference (346)
7495 - fix full OST/dbench hang (333)
7496 - fix permission problem with file removal (286)
7497 - fix removal of OSCs from LOV when they fail
7498 - fix NULL deref during bulk timeout (214)
7499 - fix problems related to multiple filesystems on one MDS (241)
7500 - fixed serious subtle metadata locking bugs
7501 - free locks on clients when inodes are removed due to memory
7503 - fix inode pointer in lock data (285)
7504 - partial support for multiple MDS on a single host (241)
7505 - data locks weren't cancelled at clear_inode time (290, 311)
7506 - intent locks could lead to unbounded lock growth (205)
7507 - added a maximum lock count, an LRU list, and a flusher
7508 - fix multiple rename (365)
7509 - properly abstracted the echo client
7510 - OSC locked 1 byte too many; fixed
7511 - rewrote brw callback code:
7512 - fixed recovery bugs related to LOVs (306)
7513 - fixed too-many-pages-in-one-write crash (191)
7514 - fixed (again) crash in sync_io_timeout (214)
7515 - probably fixed callback-related race (385)
7517 - Add capability to MDS protocol
7518 - LDLM cancellations and callbacks on different portals
7520 2002-10-28 Andreas Dilger <adilger@clusterfs.com>
7523 - limit client IOV size to PTL_MD_MAX_IOV (611336, 191)
7524 - defer open object destruction to close time (601981, 138)
7525 - open/close OST file handle in obdo (OBD_MD_FLHANDLE) (601981, 138)
7526 - move LDLM_ENQUEUE/CONVERT back to MDS portal (625069)
7527 - abstract ll_lookup2, fix ll_revalidate2 to use abstraction (256)
7528 - don't call obd_setattr in ll_file_release for destroyed objects
7529 * protocol change to lustre_msg: move |version| and add |flags|
7530 * protocol change to osc_punch: "start" in "o_size", "end" in "o_blocks"
7531 * lock replay: for LDLM_FL_REPLAY trust client to do right thing
7532 * added replay of create, unlink, link and rename operations during
7533 MDS failover; recovery should be much more robust now
7534 * remove failed OSCs from LOVs (only lov_create uses this so far)
7535 * the lustre-HOWTO was brought (more) up to date (582544)
7537 2002-10-23 Phil Schwan <phil@clusterfs.com>
7540 - in-use dentries weren't being reused properly (617851)
7541 - prevent multiple LDLM setup (599178)
7542 - fix LOV size calculations for truncate (617853)
7543 - fix client handling of MDS intent errors (POSIX)
7544 - fix permission bug in lovstripe.c test (624321)
7545 - fix MDS thread deadlock - move LDLM handler to DLM portal (625069)
7546 - truncate past end of file could corrupt data
7547 - proper cleanup after timeouts, crashes, etc (592524, 550815)
7548 - a race in recovery could return ETIMEDOUT to apps (623947)
7549 - building outside the source directory was fixed
7550 * the lustre-HOWTO was brought (more) up to date (582544)
7551 * major progress was made on recovery functionality
7553 2002-10-10 Phil Schwan <phil@clusterfs.com>
7556 - recovery deadlock fix
7557 - rm -rf causes LBUG fix (617817)
7558 - file open by multiple tasks fix (618962)
7559 - directory permissions bugs (602707 and 620007)
7560 - journal_stop fixed with locking (611313)
7561 - O_APPEND failures resolved (618273, perhaps 614459)
7562 - lconf PATH fix (619770)
7563 - IA64 build fix (621450)
7564 - RPC buffer sizes scale with amount of memory
7566 2002-10-01 Phil Schwan <phil@clusterfs.com>
7569 - locks would be cancelled without throwing away data pages,
7570 resulting in inconsistent data (605627)
7571 - inode attributes were not always being refreshed (605627, 612449)
7572 - lconf now continues to cleanup after lctl reports an error
7573 - MDS now enforces user permissions (602707)
7574 - lprocfs cleanup fixed, but not yet enabled (614157)
7575 - fixed infinite server hang, should a client not respond to an AST
7576 - avoid going into recovery if user calls readlink() with a buffer
7577 that's too small (613941)
7578 - AST RPCs no longer require replies (614867) -- this may be changed
7579 - don't crash server if client sends an IOV that's too big (611336)
7580 - fixed lock conversion deadlock (611892)
7581 - fixed the following of symlinks (614622)
7582 * recovery: the server can remove locks from a client that dies, other
7583 clients can make progress
7584 * more extN patch fixes
7585 * compile-time configurable ptlrpc buffer allocations
7587 - collaborative read cache document
7588 - Lustre Lite Performance CDR document-in-progress
7590 2002-09-20 Andreas Dilger <adilger@clusterfs.com>
7593 - fix typo in patch-2.4.18
7595 2002-09-20 Andreas Dilger <adilger@clusterfs.com>
7598 - clear ptlrpc request each time in handle_incoming_request()
7599 - unlink of files now destroys the object on the OST
7601 2002-09-19 Peter Braam <braam@clusterfs.com>
7603 * add hard link support
7604 * change obdfile creation method
7605 * kernel patch changed
7607 2002-09-19 Peter Braam <braam@clusterfs.com>
7610 - stack overflow bug in extN fixed
7612 2002-09-18 Andreas Dilger <adilger@clusterfs.com>
7614 * documentation updates
7615 - add man pages for config tools
7616 - update tests/README to describe testing with new config tools
7617 - finish metadata API descriptions
7618 * bug fixes and cleanups
7619 - statfs workaround for 16TB limit
7620 - LOV stripe allocation improved, can stripe on subset of OSTs
7621 - LOV file size/IO offset was wrong for files > 4GB in size
7622 - object EA data was being dropped, caused files to be unreadable
7623 - memory overflow with non-LOV OST caused memory corruption
7624 - fixed regression tests to work with new config tools, obdfilter
7625 - fixed bug when directory size became larger than 1 block
7626 - fixed bug (for single client case) when PWD was deleted
7627 - invalidate local directory pages when doing intent-based ops
7628 - avoid LDLM oops when lock callback contained bad data
7630 2002-09-09 Andreas Dilger <adilger@clusterfs.com>
7632 * documentation updates
7633 * bug fixes and cleanups
7634 - configuration tools
7637 - 64-bit compile warnings
7638 - 64-bit internal statfs data
7640 * test_brw on persistent OST devices
7642 * lprocfs (disabled)
7644 2002-09-04 Andreas Dilger <adilger@clusterfs.com>
7646 * documentation updates
7647 * bug fixes and cleanups
7648 * configuration tools
7650 2002-08-30 Peter J. Braam <braam@clusterfs.com>
7653 * many small fixes to 0_5_4
7654 * io/network handling
7655 * thinkos in MDS operations
7657 2002-08-24 Peter J. Braam <braam@clusterfs.com>
7660 * crucial basic fixes to 0.5.3
7661 * IOR, Iozone work over Elan
7664 2002-08-07 Phil Schwan <phil@clusterfs.com>
7665 * version 0_5_3, our first alpha
7666 * we use the new Portals iovs
7667 * documentation updates
7668 * bug fixes and cleanups
7669 * small changes in the DLM wire protocol
7671 2002-07-25 Peter J. Braam <braam@clusterfs.com>
7672 * version 0_5_1 with some initial stability,
7673 * locking on MD and file I/O.
7674 * documentation updates
7675 * several bug fixes since 0.5.0
7676 * small changes in wire protocol
7678 2002-07-18 Phil Schwan <phil@clusterfs.com>
7680 * delivered as Lustre Light Alpha
7681 * fixed a crash after handling invalid MDS requests
7682 * fixed directory pages for architectures with non-4k pages sizes
7684 2002-07-11 Andreas Dilger <adilger@clusterfs.com>
7685 * release version v0_4_4
7686 * Moves TCP acceptor to be on port 2432 (unused Coda port) instead
7688 * Fixes a number of interruption problems with OST operations.
7689 * Update documentation for portals header changes
7690 * Move all wire protocol structs/defines to lustre_idl.h
7691 * Fixes symlink length bug.
7692 * Add tcpdump to repository.
7694 2002-07-05 Andreas Dilger <adilger@clusterfs.com>
7695 * release version v0_4_3
7696 * Fixes statfs for inodes on extN.
7697 * Fixes bug in runtests which would delete /etc/hosts.
7698 * Use 64-bit object IDs wherever possible (not into VFS though)
7699 Remove ost_get_info, which is unused by lustre, and out of date.
7701 2002-07-03 Peter Braam <braam@clusterfs.com>
7702 * release version v0_4_2 Fixes a lookup error (type not passed)
7703 * move forward to head of Portals
7704 * move forward to latest Lustre kernel
7706 2002-06-25 Peter Braam <braam@clusterfs.com>
7707 * release version v0_4_1. Hopefully stable on single node use.