1 2010-xx-xx Oracle, Inc.
4 2.6.16.60-0.42.8 (SLES 10),
5 2.6.27.39-0.3.1 (SLES11),
6 2.6.18-194.3.1.el5 (RHEL 5)
7 2.6.18-194.3.1.0.1.el5 (OEL 5)
8 * Client support for unpatched kernels:
9 (see http://wiki.lustre.org/index.php?title=Patchless_Client)
10 2.6.16 - 2.6.30 vanilla (kernel.org)
11 * Recommended e2fsprogs version: 1.41.10-sun2
12 * The async journal commit feature (bug 19128) and the cancel
13 lock before replay feature (bug 16774) are disabled by default.
16 -------------------------------------------------------------------------------
18 2010-07-31 Oracle, Inc.
20 * Support for kernels:
21 2.6.16.60-0.42.8 (SLES 10),
22 2.6.27.39-0.3.1 (SLES11),
23 2.6.18-194.3.1.el5 (RHEL 5)
24 2.6.18-194.3.1.0.1.el5 (OEL 5)
25 * Client support for unpatched kernels:
26 (see http://wiki.lustre.org/index.php?title=Patchless_Client)
27 2.6.16 - 2.6.30 vanilla (kernel.org)
28 * Recommended e2fsprogs version: 1.41.10-sun2
29 * The async journal commit feature (bug 19128) and the cancel
30 lock before replay feature (bug 16774) are disabled by default.
32 Severity : enhancement
34 Description: Update RHEL5.5 kernel to 2.6.18-194.3.1.el5 and OEL5.5 kernel to
35 2.6.18-194.3.1.0.1.el5.
37 Severity : enhancement
39 Description: using inkernel OFED stack for rhel5 & oel5.
41 Severity : enhancement
43 Description: Add "lfs_migrate" script from manual into lustre/scripts and RPMs
44 Details : lfs_migrate does a "poor man's" migration of files from their
45 current OST layout to a new OST layout as chosen by the MDS.
49 Description: mds_orphan_add_link()) error linking orphan to PENDING
50 Details : quota limits might disallow linking orphans to PENDING
51 when unlinking a file - temporary raise threads' privileges
52 when processing unlinks.
54 Severity : enhancement
56 Description: add conf-param -d option to remove permanent settings.
57 Details : Add the ability to remove permanent lctl conf_param settings.
58 (Previously conf_param settings could only be changed, not
59 removed.) This also provides a method to change failover
60 nid locations. Improve lctl man page.
62 Severity : enhancement
64 Description: add list_param to b1_8 and add "-R" option to list params
67 Severity : enhancement
69 Description: lfs quota output is not very convenient for awk/sed-parsing
70 Details : Some positions in lfs quota output table could be empty or
71 non-empty which made it hard to parse it with scripts, now a dash
72 is put instead of space where there is not supposed to be any data.
74 Severity : enhancement
76 Description: fix obdfilter-survey script to work properly with remote oss-s
78 Severity : enhancement
80 Description: add new OBDFILTER_SURVEY test suite
82 Severity : enhancement
84 Description: add new multiple mount protection (MMP) test suite
86 Severity : enhancement
88 Description: add support for async journal commit in echo client
90 Severity : enhancement
92 Description: allow userland programs to include <lustre/lustre_idl.h>
93 from stardard include directories
95 Severity : enhancement
97 Description: The prune-icache-use-trylock is no longer needed now that
98 the patch from bug 20008 is landed.
102 Description: The shrink grant feature is still active on the client although the
103 connect flag is not set.
107 Description: Don't leak grant space if the write failed with quota exceeded.
111 Description: Don't consume grant space twice on recoverable resent.
115 Description: a race condition could lead to SIGBUS being sent to an
116 application using mmap-ped files from Lustre
117 Details : truncate_complete_page implementation for the patchless
118 client could arbitrarily unset PG_Uptodate flag for a
119 page being kicked from the page cache, an uptodate check
120 right after a readpage call in filemap_fault could fail
121 because of that as though the page read had been unsuccessful.
125 Description: dlm lock slab shrinking is not efficient
126 Details : The dlm_locks slab can grow significantly and consumes a lot of
127 memory on the server. Set a hardlimit to grant_plan.
131 Description: Lustre does not do 1MB IOs to HW RAID
132 Details : Bump MAX_PHYS/HW_SEGMENTS and SG_ALL to 256 in the RHEL5 kernel.
133 This is what we do already for SLES kernels.
137 Description: bump maximum number of phys/hw segments in the SLES11 kernel
138 until s/g chaining works properly.
143 Description: LSI Fusion MPT driver hacks to improve performance
144 Details : Set CONFIG_FUSION_MAX_SGE to 256 for RHEL5
146 Severity : enhancement
148 Description: increase default md stripe_cache_size to 16k
151 Bugzilla : 15587/21439
152 Description: don't handle security.capability xattr
153 Details : CONFIG_SECURITY_FILE_CAPABILITIES is enabled by default on SLES11.
154 This results in additional getxattr calls, causing VBR test
155 failures as well as a preformance drop when writing.
159 Description: obdfilter-survey is no longer working
160 Details : revert patch from bug 20355 to resolve an issue with lctl --threads
161 not working correctly with $(PTHREAD_LIBS) being linked to lctl.
165 Description: ll_shrink_cache does not handle __GFP_FS properly
169 Description: lfs getstripe shows wrong info for directories
170 Details : Set correct LOVEA default values for filesystem-wide.
174 Description: FSX checksum false positves due to mmap IO
175 Details : Use OBD_FL_MMAP flag for IOs on a memory mapped file. Do not print
176 checksum errors, if the flag is set on a request.
180 Description: file operations after eviction have successful return values
181 Details : use vfs ->flush callback to return any pending async errors
186 Description: mdsrate fails to write after 1.3+M files opened
187 Details : decrease memory usage on clients by recycling dentries and
192 Description: obdfilter-survey gives unreasonably high numbers
193 Details : Wait for all threads to complete when running test_brw.
197 Description: do not set lustre read_only device when server umount and keep
198 client records for recoverable ones
202 Description: move sync_on_lock_cancel tunable to the obdfilter layer
203 Details : move the tunable to trigger a journal flush on lock cancel from
204 the ost layer to the obdfilter layer. This tunable is useful
205 when using the async journal commit feature.
209 Description: exp->exp_nid_stats == NULL in filter_tally()
210 Details : fix race with per-nid stats by delaying procfs cleanup until
215 Description: extent lock cancellation on client can keep the cpu busy for too
220 Description: Do not fail OST activation when a llog is not found, just
221 issue an error message.
225 Description: Don't enable extents by default for MDT.
229 Description: Protect bitfield access to ptlrpc_request's rq_flags, since
230 the AT code can access it concurrently while sending early
235 Description: Disable lockless truncate by default since it is sometimes flawed
236 and causes the write_disjoint test to fail.
240 Description: OSSs which don't have the patch from bug 20278 can trigger an
243 Severity : enhancement
245 Description: don't print message to the console when we have not managed to
250 Description: The MDS fails to synchronize OSTs which registered with the MGS
251 after the MDT. The problem is that OBD_NOTIFY_CREATE events are
252 raised too early and thus discarded by the MDT stack.
253 The fix consists of issuing OBD_NOTIFY_CREATE event in the lov
258 Description: Fix race when the ping evictor and a service thread execute
259 target_recovery_check_and_stop() concurrently.
263 Description: quota broadcast can trigger a LBUG on the MDT if there are
268 Description: Resetting the lov_objid values to last_id reported by the OST
269 during orphan recovery is incorrect and can cause the same
270 objects to be allocated twice.
272 Severity : enhancement
274 Description: "weak-modules" support
275 Details : Implement "weak-modules" support which enables kernel modules
276 to be used with any kernel that implements the same kABI. In
277 order to achieve this modules are now installed in
278 /lib/modules/$(uname -r)/updates/kernel on all distributions.
280 Severity : enhancement
282 Description: add writeconf as mount option
284 Severity : enhancement
286 Description: produce debuginfo packages for SLES.
288 Severity : enhancement
290 Description: add failover nidlist to the import proc file.
292 Severity : enhancement
294 Description: fix LUSTRE_SEQ_MAX_WIDTH for interoperability between 1.8
295 clients and 2.0 servers.
297 Severity : enhancement
299 Description: lfs find -s does not work correctly because of a bug in
304 Description: ll_read_ahead_page() must validate the dlm lock before using
309 Description: Prevent failover nids from registering with MGS first.
313 Description: fix lock inversion in ll_setattr_raw().
317 Description: object allocation is not balanced across OSTs.
318 Details : osc_precreate() should return 0, if there are enough objects left.
320 -------------------------------------------------------------------------------
322 2010-04-30 Oracle, Inc.
324 * Support for kernels:
325 2.6.16.60-0.42.8 (SLES 10),
326 2.6.27.39-0.3.1 (SLES11),
327 2.6.18-164.11.1.el5 (RHEL 5)
328 2.6.18-164.11.1.0.1.el5 (OEL 5)
329 * Client support for unpatched kernels:
330 (see http://wiki.lustre.org/index.php?title=Patchless_Client)
331 2.6.16 - 2.6.30 vanilla (kernel.org)
332 * Recommended e2fsprogs version: 1.41.10-sun2
333 * The async journal commit feature (bug 19128) and the cancel
334 lock before replay feature (bug 16774) are disabled by default.
338 Description: open-unlinked directories trigger MDS LBUG
339 Details : Fix regression introduced by the patch from bug 19640.
340 ext3_inc_count() can reset nlink to 1 when the directory
341 is indexed and inode->i_nlink == 2. Work around the problem
342 by incrementing nlink by 2 instead of 1.
346 Description: Reconnects are not throttled
347 Details : Don't wake up pinger on reconnect failures and rely on regular
348 pings to trigger the next reconnection. Please note that the
349 pinger already uses a smaller interval if the import is
353 Frequency : only with NFS export
355 Description: Console flooded with error message from ll_inode_from_lock()
357 Details : in mds_open, initialize the child_res_id before enqueuing
358 the OPEN lock for the child inode, then to avoid senting
359 wrong ldlm_res_id to client.
363 Description: allow multiple instances of the same nid in NID hash
364 Details : Case of multiple separate clients from the same NID (as
365 with liblustre) is legitimate and so we should allow
366 multiple instances of the same NID in nid hash.
370 Description: the readahead code can sleep on a semaphore while holding a
372 Details : in ras_update, "lov_get_info" could be called during increasing
373 readahead windows, which tries to get the mutex lock "lov_lock"
374 while holding the spin_lock "ras_lock", then causes system
379 Description: ASSERTION(cli->cl_avail_grant >= 0) failed
380 Details : fix assertion failure in the grant code.
384 Description: Use CNETERR (which is rate limited) in specific places in
385 the portal's LNET driver to avoid flooding the console.
389 Description: include last created object in precreate slow case
393 Description: don't do rep-ack if not created anything
394 Details : mds_open currently always put a lock into a rep-ack regardless
395 if something was created or not. This is pointless and only
396 creates needless contention. In fact the entire idea was to do
397 this for real creates as a recovery protection.
401 Description: Spurious error messages from smp_processor_id() on preemptible
403 Details : Disable a preemption by grabbing the lock in fs_trace_get_tcd()
404 first. The function fs_trace_get_tcd() was moved up.
408 Description: interval_erase() fix
409 Details : interval_erase() calls update_maxhigh() properly when child
414 Description: "lfs df" does not print stats for all mountpoints
415 Details : Print all mounted lustre filesystems with "lfs df"
419 Description: lfs setstripe -p not longer work with a relative pathname
420 Details : Use realpath() to provide absolute pathname.
424 Description: fix for truncated reply buffer
425 Details : reply buffer could be referred by reply_in_callback after released
430 Description: lustre.lov error when backing up symlinks with extended attributes
431 Details : Improved logic in ll_listxattr()
435 Description: properly handle null value for setattr -n lustre.lov
436 Details : Running "setfattr -n trusted.lov ." causes a NULL dereference
437 in ll_setxattr() due to no checking if "value" is NULL.
438 This command now resets to the default striping when executed
443 Description: stack overflow on lock cancellation due to fsync call
444 Details : sync_on_lock_cancel is needed for recovery when async journal
445 is enabled, but we actually just need to make sure that
446 metadata blocks have hit the journal, so doing a fs sync
447 should be enough and should consume less stack (just create an
448 empty handle and commmit it).
452 Description: using current->journal_info to store per-thread data leads
453 to problem under memory pressure
454 Details : disable the per-thread data (current->journal_info) containing
455 the lock info during I/O to work around the issue for short tem
459 Description: control DCACHE_LUSTRE_INVALID flag with MDS_INODELOCK_LOOKUP lock
460 Details : DCACHE_LUSTRE_INVALID is controlled by MDS_INODELOCK_LOOKUP
461 lock which is corresponding to "IT_LOOKUP", do not skip invalidate
466 Description: Cannot send after transport shutdown
467 Details : Clear imp_vbr_failed flag upon eviction
471 Description: soft lock in request set code during recovery
472 Details : during recovery, uses req->rq_set itself to replay the request
473 instead of ptlrpcd_recovery_pc
477 Description: Use CFS_ALLOC_IO instead of _STD in llap_from_page_with_lockh
478 Details : During an ll_readahead under ll_readpage, we have seen the the
479 OBD_SLAB_ALLOC hang under ldlm_pools_shrink when trying to lock
480 a page that is already locked by the readahead code.
485 Description: stop waitting for next replay transno if shutdown
486 Details : if the system is shutting down, wake up service thread blocked
487 to wait for next replay transno during recovery, then all the
488 references held by queued requests can be dropped and device
493 Description: lov_merge_lvb()) ASSERTION(spin_is_locked(&lsm->lsm_lock)) failed
494 Details : Protect lli->lli_smd pointer updates with lli->lli_lock.
498 Description: per-nid stats should not access lustre-hash internal structures
503 Description: mount.lustre fails to pass some options to mount()
508 Description: ext4 extent allocation is slower than in ext3
509 Details : Increase the default value of MB_DEFAULT_ORDER2_REQS to 8,
510 enlarge ext4 preallocation table for 2048 4K blocks extents
515 Description: incorrect triggering of synchronous IO
516 Details : The OSC can mistakenly fall back to synchronous IO when the
517 max_dirty_mb limit is reached and no write requests have yet
518 been issued. This can occur when the dirty pages are spread
519 over many files all of which are below the optimal request size.
523 Description: Optimize quota_ctl operations by sending requests in parallel
524 Details : Send MDS->OST quota_ctl requests in parallel, do not resend.
525 Compiled from two attachments in the ticket.
529 Description: ordering issue between transaction start & i_mutex
530 Details : start the transaction earlier in llog_lvfs_destroy to get
531 transaction start and inode mutex lock nested properly.
535 Description: lru resize SLV can get stuck
536 Details : calculate SLV with a greater precision to not lose small
537 changes due to interger math truncation; round up SLV only
538 if the amount of granted locks less than the limit to not
539 get stuck with this SLV
543 Description: avoid divide-by-zero in lprocfs_rd_import()
547 Description: lfs quota failed when OSTs are down
548 Details : really return approximate block/inode usage when OSTs are down
552 Description: llobdstat fix and enhancement
553 Details : add a counter to set a limit to how many samples will be returned
554 fix a wildcard in the path to limit to obdfilter stats only
558 Description: recovery-small 51 hang - MDT can not stop, Mount still busy with
560 Details : abort lock enqueue processing sooner when we umount is in progress.
562 Severity : enhancement
564 Description: Update RHEL5.4 kernel to 2.6.18-164.11.1.el5 and OEL5.4 kernel to
565 2.6.18-164.11.1.0.1.el5.
567 Severity : enhancement
569 Description: error message improvements
570 Details : Use INFO/WARN instead of WARN/ERROR for the slow messages.
571 Simplify MDT/OST service start message.
572 Suppress "changing the import ..." warning.
575 Bugzilla : 21961/17914
576 Description: ignore trailing -mdc when determining index number
579 Frequency : only with SLES11
581 Description: fix for a race condition in linux quotas implementation
582 Details : dq_flags(struct dquot) access is not properly locked which could
583 lead to certain inconsistencies when accessing it using non-atomic
584 bit operations like __set_bit in do_set_dqblk.
588 Description: return any pending async errors in close(2) using flush callback
590 -------------------------------------------------------------------------------
592 2010-01-29 Sun Microsystems, Inc.
594 * Support for kernels:
595 2.6.16.60-0.42.8 (SLES 10),
596 2.6.27.39-0.3.1 (SLES11),
597 2.6.18-164.11.1.el5 (RHEL 5)
598 2.6.18-164.6.1.0.1.el5 (OEL 5)
599 * Client support for unpatched kernels:
600 (see http://wiki.lustre.org/index.php?title=Patchless_Client)
601 2.6.16 - 2.6.30 vanilla (kernel.org)
602 * Recommended e2fsprogs version: 1.41.6.sun1
603 * The async journal commit feature (bug 19128) and the cancel
604 lock before replay feature (bug 16774) are disabled by default.
606 Severity : enhancement
608 Description: Update RHEL5.4 kernel to 2.6.18-164.11.1.el5 and
609 OEL5.4 kernel to 2.6.18-164.11.1.0.1.el5.
611 Severity : enhancement
612 Bugzilla : 21511/19848
613 Description: Update SLES11 kernel to 2.6.27.39-0.3.1.
615 Severity : enhancement
617 Description: Update supported SLES10 kernel to 2.6.16.60-0.42.8.
619 Severity : enhancement
621 Description: Update kernel to RHEL5.4 2.6.18-164.6.1.el5 and
622 OEL5 2.6.18-164.6.1.0.1.el5(Both in-kernel OFED enabled).
624 Severity : enhancement
626 Description: Build kernels (RHEL5, OEL5 and SLES10/11) using the vendor's own
629 Severity : enhancement
631 Description: Vanilla kernel 2.6.30 patchless client support.
636 Description: bad entry in directory xxx: inode out of bounds
637 Details : fix locking issue in the rename path which could race with any
638 other operations updating the same directory.
640 Severity : enhancement
642 Description: Make watchdog timer messages to be more clear and descriptive.
646 Description: cp -p command does not preserve the dates and timestamp
647 Details : mtime could be spoiled by a write callback
651 Description: Clear imp_force_reconnect correctly in ptlrpc_connect_interpret()
655 Description: Allow non-root access for "lfs check".
656 Details : Added a check in obd_class_ioctl() for OBD_IOC_PING_TARGET.
658 Severity : enhancement
660 Description: quotacheck performance/scaling issues
661 Details : reduce quotacheck time on empty filesystem by skipping uninit
664 Severity : enhancement
666 Description: Enhancement for lfs(1) command to use numeric uid/gid.
668 Severity : enhancement
670 Description: Adjust locks' extents on their first enqueue, so that at the time
671 they get granted, there is no need for another pass through the
672 queues since they are already shaped into the proper forms.
676 Description: Fix mds_shrink_intent_reply()/mds_intent_policy() to pass correct
677 arguments and prevent LBUG() in lustre_shrink_reply_v2().
681 Description: Change tunefs.lustre and mkfs.lustre --mountfsoptions so that
682 exactly the specified mount options are used. Leaving off
683 any "mandatory" mount options is an error. Leaving off any
684 default mount options causes a warning, but is allowed.
685 Change errors=remount-ro from mandatory to default. Sanitize
686 the mount string before storing it. Update man pages accordingly.
690 Description: mds_getattr() should return 0, even if mds_fid2entry() fails with
691 -ENOENT. Also fix in ptlrpc_expire_one_request() to print signed
694 Severity : enhancement
696 Description: Remove set_info(KEY_UNLINKED) from MDS/OSC
698 Severity : enhancement
700 Description: Clients can replay thousands of unused locks during recovery
701 Details : Don't replay unused locks (only read locks for now) during
702 recovery. This feature is disabled by default and can be
703 enabled by running the following command on the clients:
704 lctl get_param ldlm.cancel_unused_locks_before_replay
708 Description: can't stat file in some situation.
709 Details : improve initialize osc date when target is added to mds and
710 ability to resend too big getattr request is client isn't have info
715 Description: Prevent inconsistences between linux and lustre mount structures.
716 Details : Wait indefinitely in server_wait_finished() until mnt_count drops.
717 Make the sleep interruptible.
719 Severity : enhancement
721 Description: Communicate OST degraded/readonly state via statfs to MDS
722 Details : Flags in the statfs returned from OSTs indicate whether the
723 OST is in a degraded RAID state, or if the filesystem has
724 turned read-only after a filesystem error is detected.
729 Description: don't panic if EPROTO was hit when reading symlink
730 Details : correctly handling request reference in error cases.
735 Description: open sometimes returns ENOENT instead of EACCES
736 Details : checking permission should be part of open part of mds_open, not
737 lookup part. so server should be set DISP_OPEN_OPEN disposition
738 before starting permission check.
739 Also not need revalidate dentry if client already have LOOKUP lock.
743 Description: enable client interface failover
744 Frequency : on servers with multiple network interfaces
745 Details : When a child reconnects from another NID, properly update export
746 nid hash position and ldlm reverse import.
748 Severity : enhancement
750 Description: implemented direct I/O with arbitrary (nonaligned) memory
751 addresses and file offsets.
753 Severity : enhancement
755 Description: added more recovery timeout options.
757 Severity : enhancement
759 Description: added llapi_file_open, llapi_file_create, llapi_file_get_stripe
764 Description: Avoid deadlock for local client writes
765 Frequency : only on systems with clients writing to an OST on the same node
766 Details : Use new OBD_BRW_MEMALLOC flag to notify OST about writes in the
767 memory freeing context. This allows OST threads to set the
768 PF_MEMALLOC flag on task structures in order to allocate memory
769 from reserved pools and complete IO.
770 Use GFP_HIGHUSER for OST allocations for non-local client writes,
771 so that the OST threads generate memory pressure and allow
772 inactive pages to be reclaimed.
777 Descriptoin: lock ordering violation between &cli->cl_sem and _lprocfs_lock
778 Details : move ldlm namespace creation in setup phase to avoid grab
779 _lprocfs_lock with cli_sem held.
783 Frequency : only during format of test systems
784 Description: Unable to run several mkfs.lustre on loop devices at the same time
785 Details : mkfs.lustre returns error 256 on the concurrent loop devices
786 formatting. The solution is to proper handle the error.
788 Severity : enhancement
790 Description: implement async create (obd_async_create) method for osc, to avoid
791 too long waiting new ost objects with holding ldlm lock.
795 Frequency : occasionally during network problems
796 Description: client not allowed to reconnect to OST because of active request
797 Details : abort bulk requests received by the OST once the client has timed
798 out since the client will resend the request anyway. The client
799 also now retries to reconnect to the same server if a connect
800 request failed with EBUSY or -EAGAIN.
803 Frequency : rare, if used wide striped file and one ost in down.
805 Descriptoin: don't return error if we created a subset of objects for file.
806 Details : lov_update_create_set() uses set->set_success as index for created
807 objects, so if some requests failed, they will have hole at end of
808 array and we can use qos_shrink_lsm for allocate correct lsm.
812 Description: Slow stale export processing during normal start up
813 Details : The global mgc lock prevents OST setup to be run in parallel.
814 Replace the global lock with a per-config_llog_data semaphore.
818 Description: Out or order replies might be lost on replay
819 Details : In ptlrpc_retain_replayable_request if we cannot find retained
820 request with tid smaller then one currently being added, add it
821 to the start, not end of the list.
825 Description: BUG: soft lockup - CPU#1 stuck for 10s! [ll_mdt_07:4523]
826 Details : add cond_resched() calls to avoid hogging the cpu for too long
827 in the hash code. Make also lustre_hash_for_each_empty() more
830 Severity : enhancement
832 Description: Performance improvements for debug messages with D_RPCTRACE,
833 D_LDLM, D_QUOTA options.
836 Frequency : only with NFS export
838 Description: (lov_merge.c:74:lov_merge_lvb())
839 ASSERTION(spin_is_locked(&lsm->lsm_lock)) failed (SR 71691004)
840 Details : Fix a race in the nfs export code by populating inode
841 info while the new inode is still locked
843 Severity : enhancement
845 Description: add a new file in procfs called force_lbug. Writting to this
846 file triggers a LBUG. Only for test purpose.
850 Description: OOM killer causes node hang
851 Details : really interrupt the sleep in osc_enter_cache on signals
855 Description: LustreError: 9153:0:(quota_context.c:622:dqacq_completion()) LBUG
856 Details : fix race during quota release on the slave.
858 Severity : enhancement
860 Description: smaller hash bucket sizes, cleanups
861 Details : increase hash table sizes and enabled rehashing for pools, quota,
862 uuid, nid & per-nid stats.
864 Severity : enhancement
866 Description: Add ldiskfs maxdirsize mount option
867 Details : add max_dir size mount option
871 Description: panic in ll_statahead_thread
872 Details : prevent parent thread to be killed before its child
875 Frequency : only with 16TB device
877 Description: unable to perform "mount -t lustre" of 16TB OST device
878 Details : Mounting 16TB LUNs failed due to three bugs in mkfs.lustre.
882 Description: ASSERTION(atomic_read(&imp->imp_inflight) == 0) failed
883 Details : unregistering should be zero if no RPC inflight.
887 Description: hyperion: Oops during metabench
888 Details : Correct the refcount of lov_request_set
890 Severity : enhancement
892 Description: Add mptlinux and nxge drivers to Lustre builds
894 Severity : enhancement
896 Description: Fix watchdog timer message to be more clear
897 Details : Make watchdog timer messages more clear and descriptive.
901 Description: LNET soft lockups in socknal_cd thread
902 Details : don't hog CPU for active-connecting if another connd is
903 accepting connecting-requst from the same peer
907 Description: recovery-small test_17 hang
908 Details : Land several AT improvements & fixes.
912 Description: MDS panic and hanging client processes
913 Details : Replace exp_ops_stats with exp_nid_stats->nid_stats
917 Description: OSS stuck in recovery.
918 Details : fix race during recovery. class_unlink_export,
919 class_set_export_delayed and target_queue_last_replay_reply
920 may race while increasing/decreasing obd_recoverable_clients
921 and obd_delayed_clients, causing recovery to wait forever.
923 Severity : enhancement
925 Description: add cascading_rw.c to lustre/tests
929 Description: filter_last_id() NULL deref
930 Details : lprocfs_filter_rd_last_id() should check for the fully
931 setup obd device, before proceeding further.
933 Severity : enhancement
935 Description: Loadgen improvements
936 Details : stacksize and locking fixes for loadgen
940 Description: Quiet CERROR("dirty %d > system dirty_max %d\n"
941 Details : The atomic_read() allowing the atomic_inc() are not covered
942 by a lock. Thus they may safely race and trip this CERROR()
943 unless we add in a small fudge factor (+1).
945 Severity : enhancement
947 Description: shrink_slab: nr=-9223362083340912175
948 Details : fix spurious message from shrink_slab reporing negative nr
952 Description: Quiet bogus previously committed transno error
953 Details : suppress the "server went back in time" error message which
954 is always printed even in the common case after a client eviction
956 Severity : enhancement
958 Description: Parallel statfs() calls result in client eviction
959 Details : cache statfs data for 1s.
963 Description: parallel-scale test_compilebench: @@@@@@ FAIL: compilebench
965 Details : fix serveral issues in pinger code causing clients not to ping
966 servers for too long, resulting in evictions.
970 Description: e2fsck should warn when MMP update interval is extended
971 Details : print mmp_check_interval and make it possible to abort mount
972 operation in case it takes too long.
976 Description: mdsrate-create-large.sh, BUG: soft lockup - CPU#0 stuck for 10s!
977 Details : fix bug in the RHEL5's jbd2 callback patch.
981 Description: drop number of active requests when queued for recovery
982 Details : Now that we take a reference on the original request instead of
983 making a copy of it for recovery. We need to drop the number of
984 active requests or the queued requests will prevent all request
985 processing when they exceed (srv->srv_threads_running - 1).
987 Severity : enhancement
989 Description: refuse to invalidate operational quota files when they are in use
990 Details : an attempt to invalidate operational quota files on the quota
991 master is not actually permitted by VFS (returning -EPERM), but we
992 should not depend on that and should return the error earlier.
996 Description: Applications stuck in jbd2_log_wait_commit during exit
997 Details : fix deadlock between kjournald2 trying to acquire the page lock
998 owned by an ost_io thread waiting for journal commit.
1000 -------------------------------------------------------------------------------
1002 2009-10-16 Sun Microsystems, Inc.
1004 * Support for kernels:
1005 2.6.16.60-0.42.4 (SLES 10),
1006 2.6.27.29-0.1 (SLES11, i686 & x84_64 only),
1007 2.6.18-128.7.1.el5 (RHEL 5),
1008 2.6.18-128.7.1.el5 (OEL 5),
1009 * Client support for unpatched kernels:
1010 (see http://wiki.lustre.org/index.php?title=Patchless_Client)
1011 2.6.16 - 2.6.27 vanilla (kernel.org)
1012 * Recommended e2fsprogs version: 1.41.6.sun1
1013 * ext4 support for RHEL5 is experimental and thus should not be
1016 Severity : enhancement
1018 Description: Add OEL5 support.
1020 Severity : enhancement
1022 Description: Update kernel to SLES11 2.6.27.29-0.1.
1026 Description: File checksum failures with OST read cache on
1027 Details : Disable page poisoning when the bulk transfer has to be aborted
1028 because the client got evicted.
1032 Description: Don't allow make backward step on assiging osc next id.
1033 Details : race between allocation next id and ll_sync thread can be cause
1034 of set wrong osc next id and can be kill valid ost objects.
1036 Severity : enhancement
1038 Description: Update kernel to RHEL5 2.6.18-128.7.1.el5.
1040 Severity : enhancement
1042 Description: Update kernel to SLES10 SP2 2.6.16.60-0.42.4.
1046 Description: Changes in raid5-large-io-rhel5.patch to calculate sectors properly
1050 Description: Increase the default BLK_DEF_MAX_SECTORS value for RHEL5 and SLES11
1054 Description: Error handling in osc_statfs_interpret() has been improved.
1055 Details : Check in osc_statfs_interpret() for EBADR.
1059 Description: Do not send statfs() requests to OSTs disabled by administrator.
1060 Details : Check in lov_prep_statfs_set() for non-NULL ltd_exp.
1064 Description: Do not update ctime for the deleted inode.
1065 Details : Check in mds_reint_unlink() before calling fsfilt_setattr().
1069 Description: Increase of the size of the LDLM resource hash.
1070 Details : Bump up RES_HASH_BITS=12.
1074 Description: correctly send lsm on open replay
1075 Details : MDS is trust to LSM size on replay open, but client can set wrong size
1080 Description: Deadlock between filter_destroy() and filter_commitrw_write().
1081 Details : filter_destroy() does not hold the DLM lock over the whole
1082 operation. If the DLM lock is dropped, filter_commitrw() can go
1083 through, causing the deadlock between page lock and i_mutex.
1084 The i_alloc_sem should also be hold in filter_destroy() while
1085 truncating the file.
1089 Description: truncate starts GFP_FS allocation under transaction causing deadlock
1090 Details : ldiskfs_truncate calls grab_cache_page which may start page
1091 allocation under an open transaction. This may lead to
1092 calling prune_icache with consequent lustre reentrance.
1095 Frequency : only when down/upgrading the MDS to 1.6/1.8 while 1.8 clients are
1096 still up and when the OST pool feature is used
1098 Description: interop testing got LBUG when run dd with OST pool
1099 :LustreError: 30032:0:(llite_lib.c:1913:ll_replace_lsm()) LBUG
1100 Details : down/upgrading the MDS to a version that doesn't/does support OST
1101 pool can cause clients to crash because the lsm has changed
1106 Description: missing tree_status on 1.8.1 RPM build
1107 Details : make rpms failed due because the tree_status file is missing.
1111 Description: continuing LustreError "mds adjust qunit failed!"
1112 Details : don't print message on the console when ->adjust_qunit fails.
1116 Description: don't increase ldlm timeout if previous client was evicted
1117 Details : if a client doesn't respond to a blocking callback within the
1118 adaptive ldlm enqueue timeout, don't adjust the adaptive estimate
1119 when the lock is next granted.
1123 Description: ost is being unmounted w/o all writes to last_rcvd landing on disk.
1124 affects recovery negatively.
1125 Details : make sure all exports have been properly destroyed by the zombie
1126 thread processed before stopping the target.
1130 Description: Performance degradation with O_DIRECT between 1.6 & 1.8.1 b190
1131 Details : disable write barrier for ext4/SLES11.
1135 Description: Kernel panic - not syncing: Out of memory and no killable
1136 processes... on OSS when iozone
1137 Details : fix memory leak in the journal checksum patch.
1141 Description: group quota "too many blocks" OSS crashes
1142 Details : we should keep the same uid/gid for lquota_chkquota() and
1143 lquota_pending_commit()
1147 Description: LustreError: 9153:0:(quota_context.c:622:dqacq_completion()) LBUG
1148 Details : don't LBUG on release quota error. Just a workaround until the
1149 problem is understood.
1151 ------------------------------------------------------------------------------
1152 2009-07-31 Sun Microsystems, Inc.
1154 * Support for kernels:
1155 2.6.16.60-0.39.3 (SLES 10),
1156 2.6.27.23-0.1 (SLES11, i686 & x84_64 only),
1157 2.6.18-128.1.14.el5 (RHEL 5),
1158 * Client support for unpatched kernels:
1159 (see http://wiki.lustre.org/index.php?title=Patchless_Client)
1160 2.6.16 - 2.6.27 vanilla (kernel.org)
1161 * Recommended e2fsprogs version: 1.41.6.sun1
1162 * File join has been disabled in this release, refer to Bugzilla 16929.
1163 * NFS export disabled when stack size < 8192. Since the NFSv4 export
1164 of Lustre filesystem with 4K stack may cause a stack overflow. For
1165 more information, please refer to bugzilla 17630.
1166 * ext4 support for RHEL5 is experimental and thus should not be
1169 Severity : enhancement
1171 Description: Update kernel to SLES10 SP2 2.6.16.60-0.39.3.
1174 Frequency : with 1.8 server and 1.6 clients
1176 Description: correctly shrink reply for avoid send too big message to client.
1177 Details : 1.8 mds is allocate to big buffer to LOV EA data and this produce
1178 some problems with sending this reply to 1.6 client.
1182 Description: Repeated atomic allocation failures.
1183 Details : Use GFP_HIGHUSER | __GFP_NOMEMALLOC flags for memory allocations
1184 to generate memory pressure and allow reclaiming of inactive pages.
1185 At the same time, do not allow to exhaust emergency pools.
1186 For local clients the use of GFP_NOFS will be introduced in 1.8.2
1188 Severity : enhancement
1189 Bugzilla : 19846, 18289
1190 Description: Update kernel to RHEL5 2.6.18-128.1.14.el5.
1192 Severity : enhancement
1193 Bugzilla : 19625, 16893, 18668, 19848
1194 Description: Add support for SLES11 2.6.27.23-0.1.
1196 Severity : enhancement
1198 Description: Update client support to vanila kernels up to 2.6.27.
1200 Severity : enhancement
1202 Description: Update kernel to SLES10 SP2 2.6.16.60-0.37.
1204 Severity : enhancement
1206 Description: Compile with -Werror by default for i686 and x86_64.
1210 Description: resolve race between obd_disconnect and class_disconnect_exports
1211 Details : if obd_disconnect will be called to already disconnected export he
1212 forget release one reference and osc module can't unloaded.
1214 Severity : enhancement
1216 Description: move AT tunable parameters for more consistent usage
1217 Details : add AT tunables under /proc/sys/lustre, add to conf_param parsing
1221 Descriptoin: correctly skip time estimate if in recovery
1222 Details : rq_send_state insn't bitmask so using bitwise ops is forbid.
1226 Descriptoin: OSS DeadLock
1227 Details : Use trylock to prevent deadlock when shrink icache.
1229 Severity : enhancement
1231 Description: Allow tuning service thread via /proc
1232 Details : For each service a new
1233 /proc/fs/lustre/{service}/*/thread_{min,max,started} entry is
1234 created that can be used to set min/max thread counts, and get the
1235 current number of running threads.
1237 Severity : enhancement
1239 Description: Add state history info file, enhance import info file
1240 Details : Track import connection state changes in a new osc/mdc proc file;
1241 add overview-type data to the osc/mdc import proc file.
1245 Description: Reduce small size read RPC
1246 Details : Set read-ahead limite for every file and only do read-ahead when
1247 available read-ahead pages are bigger than 1M to avoid small size
1252 Description: free_entry erroneously used groups_free instead of put_group_info
1254 Severity : enhancement
1256 Description: Make read-ahead stripe size aligned.
1258 Severity : enhancement
1260 Description: MDS create should not wait for statfs RPC while holding DLM lock.
1263 Frequency : rare, connect and disconnect target at same time
1265 Descriptoin: ASSERTION(atomic_read(&imp->imp_inflight) == 0
1266 Details : don't call obd_disconnect under lov_lock. this long time
1267 operation and can block ptlrpcd which answer to connect request.
1270 Frequency : start MDS on uncleanly shutdowned MDS device
1272 Descriptoin: ll_sync thread stay in waiting mds<>ost recovery finished
1273 Details : stay in waiting mds<>ost recovery finished produce random bugs
1274 due race between two ll_sync thread for one lov target. send
1275 ACTIVATE event only if connect realy finished and import have
1279 Frequency : start MDS on uncleanly shutdowned MDS device
1281 Descriptoin: aborting recovery hang on MDS
1282 Details : don't throttle destroy RPCs for the MDT.
1286 Description: Slow reads beyond 8Tb offsets.
1287 Details : Page index integer overflow in ll_read_ahead_page
1291 Description: Soft lockup on OSS after MDS failover
1292 Details : MSG_CONNECT_INITIAL is not set on the initial MDS->OST connect.
1293 As a conseqence, the patch from bug 18224 is not operational
1294 and the MDS export cannot be reused on the OSTs until it gets
1298 Frequency : rare, only if using MMP with Linux RAID
1300 Description: MMP doesn't work with Linux RAID
1301 Details : While using HA for Lustre servers with Linux RAID, it is possible
1302 that MMP will not detect multiple mounts. To make this work we
1303 need to unplug the device queue in RAID when the MMP block is being
1304 written. Also while reading the MMP block, we should read it from
1305 disk and not the cached one.
1308 Frequency : rare, during recovery
1310 Description: Assertion failure in ldlm_lock_put
1311 Details : Do not put cancelled locks into replay list, hold references on
1312 locks in replay list
1316 Description: 1.6.5 mdsrate performance is slower than 1.4.11/12 (MDS is not cpu bound!)
1317 Details : create_count always drops to the min value (=32) because grow_count
1318 is being changed before the precreate RPC completes.
1321 Frequency : Only in RHEL5 when mounting multiple ext3 filesystems
1324 Description: "kmem_cache_create: duplicate cache jbd_4k" error message
1325 Details : add proper locking for creation of jbd_4k slab cache
1329 Description: MMP check in ext3_remount() fails without displaying any error
1330 Details : When multiple mount protection fails during remount, proper error
1335 Description: Rare Client crash on resend if the file was deleted.
1336 Details : When file is opened, but open reply is lost and file is
1337 subsequently deleted before resend, resend processing logic
1338 breaks trying to open the file again, should not try to open.
1342 Description: add check for >8TB ldiskfs filesystems
1343 Details : ext3-based ldiskfs does not support greater than 8TB LUNs.
1344 Don't allow >8TB ldiskfs filesystems to be mounted without
1345 force_over_8tb mount option
1349 Description: Client locked up when running multiple instances of an app. on
1350 multiple mount points
1351 Details : ll_shrink_cache() can sleep while holding the ll_sb_lock.
1352 Convert ll_sb_lock to a read/write semaphore to fix the problem.
1356 Description: Cannot acces an NFS-mounted Lustre filesystem
1357 Details : An NFS client cannot access the Lustre filesystem NFS-mounted
1358 from a Lustre-client exporting the Lustre filesystem via NFS.
1362 Description: panic in ll_statahead_thread
1363 Details : grab dentry reference in parent process.
1366 -------------------------------------------------------------------------------
1368 tbd Sun Microsystems, Inc.
1370 * Support for kernels:
1371 2.6.16.60-0.31 (SLES 10),
1372 2.6.18-128.1.6.el5 (RHEL 5),
1373 2.6.22.14 vanilla (kernel.org)
1374 * Client support for unpatched kernels:
1375 (see http://wiki.lustre.org/index.php?title=Patchless_Client)
1376 2.6.16 - 2.6.22 vanilla (kernel.org)
1377 * Recommended e2fsprogs version: 1.40.11-sun1
1378 * File join has been disabled, refer to Bugzilla 16929.
1379 * A new Lustre ADIO driver is available for MPICH2-1.0.7.
1380 * NFS export disabled when stack size < 8192. Since the NFSv4 export of
1381 Lustre filesystem with 4K stack may cause a stack overflow. For more
1382 information, please refer to bugzilla 17630.
1384 Severity : enhancement
1386 Description: Update to RHEL5.3 kernel-2.6.18-128.1.6.el5.
1388 Severity : enhancement
1390 Description: Update OFED release to 1.4.1 RC4
1392 Severity : major, only with big OST
1394 Description: Very poor metadata performance on Infiniband lustre configuration
1395 Details : OST object precreation becomes very slow on big OSTs. This is due
1396 to the ialloc patch spending too much time scanning groups.
1399 Frequency : during recovery
1401 Description: don't mix llog inodes with normal.
1402 Details : allocate inodes for log in last inode group
1407 Description: fix lqs' reference which won't be put in some situations
1408 Details : This patch fixes:
1409 1. In quota_check_common(), this function will check quota
1410 for user and group, but only send one return via "pending".
1411 In most cases, the pendings should be same. But that is not
1413 2. If quotaoff runs between lquota_chkquota() and
1414 lquota_pending_commit(), the same thing will happen too.
1415 That is why it comes:
1416 - if (!ll_sb_any_quota_active(qctxt->lqc_sb))
1419 -------------------------------------------------------------------------------
1421 2008-12-31 Sun Microsystems, Inc.
1423 * Support for kernels:
1424 2.6.5-7.314 (SLES 9),
1425 2.6.9-67.0.22.EL (RHEL 4),
1426 2.6.16.60-0.31 (SLES 10),
1427 2.6.18-92.1.17.el5 (RHEL 5),
1428 2.6.22.14 vanilla (kernel.org)
1429 * Client support for unpatched kernels:
1430 (see http://wiki.lustre.org/index.php?title=Patchless_Client)
1431 2.6.16 - 2.6.22 vanilla (kernel.org)
1432 * Client support for unpatched kernels:
1433 we do not recommend using patchless RHEL4 clients with kernels
1434 prior to 2.6.9-55EL (RHEL4U5).
1435 * Recommended e2fsprogs version: 1.40.11-sun1
1436 * Note that reiserfs quotas are disabled on SLES 10 in this kernel.
1437 * RHEL 4 and RHEL 5/SLES 10 clients behaves differently on 'cd' to a
1438 removed cwd "./" (refer to Bugzilla 14399).
1439 * A new quota file format has been introduced in 1.6.5.
1440 The format conversion from prior releases is handled transparently,
1441 but releases older than 1.4.12/1.6.5 don't understand this new
1442 format. The automatic format conversion can be avoided by running
1443 the following command on the MDS:
1444 'tunefs.lustre --param="mdt.quota_type=ug1" $MDTDEV'.
1445 For more information, please refer to bugzilla 13904.
1446 * A new quota file format was introduced in 1.6.6/1.8.0 (kernels 2.6.16+).
1447 The format conversion from prior releases is handled transparently,
1448 but releases older than 1.6.6/1.8.0 don't understand this new
1449 format. The automatic format conversion can be avoided by running
1450 the following commands on the MDS and OSS servers (for
1451 pre 1.4.12-1.6.5 quota files):
1452 'tunefs.lustre --param="mdt.quota_type=ug1" $MDTDEV',
1453 'tunefs.lustre --param="ost.quota_type=ug1" $MDTDEV'
1454 or (for 1.4.12/1.6.5 quota files)
1455 'tunefs.lustre --param="mdt.quota_type=ug2" $MDTDEV',
1456 'tunefs.lustre --param="ost.quota_type=ug2" $MDTDEV'
1457 For more information, please refer to bugzilla 13904.
1458 * Output of lfs quota has been made less detailed by default,
1459 old (verbose) output can be obtained by using -v option.
1460 * File join has been disabled in this release, refer to Bugzilla 16929.
1461 * A new Lustre ADIO driver is available for MPICH2-1.0.7.
1462 * NFS export disabled when stack size < 8192. Since the NFSv4 export of
1463 Lustre filesystem with 4K stack may cause a stack overflow. For more
1464 information, please refer to bugzilla 17630.
1466 Severity : enhancement
1468 Description: Caching OSS
1469 Details : introduce data caching on the OSS. The OSS now relies on the linux
1470 kernel page cache to keep recently accessed data in memory.
1471 It is worth noting that all write requests are still flushed
1472 synchronously as in lustre 1.6.
1474 Severity : enhancement
1476 Description: Version based recovery
1477 Details : introduce finer grained recovery able to detect transaction
1478 dependencies and can deal with transaction gaps caused by clients
1479 failing at the same time as the server.
1481 Severity : enhancement
1483 Description: Enable adaptive timeouts by default
1484 Details : The Lustre timeout value in /proc/sys/lustre/timeout is now
1485 managed dynamically based on server load and should not need
1486 to be tuned manually based on cluster size. This allows Lustre
1487 to work under a wider variety of system sizes and loads, without
1488 unnecessarily causing lengthy recovery times.
1490 Severity : enhancement
1492 Description: Add OST Pools support
1493 Details : File striping can now be set to use an arbitrary pool of OSTs
1495 Severity : enhancement
1497 Description: add lazystatfs mount option to allow statfs(2) to skip down OSTs
1498 Details : allow skip disconnected ost for send statfs request and hide error
1502 Frequency : rare, on llog test 6
1504 Descriptoin: don't allow connect to already connected import
1505 Details : allowing connect to already connected import is hide connecting problem.
1508 Frequency : rare, on failed llog setup
1510 Descriptoin: don't leak obd reference on failed llog setup
1511 Details : for failed llog setup - mgc forget call class_destroy_import for
1512 client import, move destroy import to more generic place.
1517 Descriptoin: allow kill process which wait statahead result
1518 Details : for some reasons 'ls' can stick in waiting result from statahead,
1519 in this case need way for kill this process.
1522 Frequency : rare, at shutdown
1524 Descriptoin: panic at umount
1525 Details : llap_shrinker can be raced with killing super block from list and
1526 this produce panic with access to already freeded pointer
1531 Descriptoin: don't lose wakeup for imp_recovery_waitq
1532 Details : recover_import_no_retry or invalidate_import and import_close can
1533 both sleep on imp_recovery_waitq, but we was send only one wakeup
1539 Descriptoin: panic in mds_open
1540 Details : don't confuse mds_finish_transno() with PTR_ERR(-ENOENT)
1545 Descriptoin: stuck in cache_remove_extent() or panic with accessing to already
1547 Details : release lock refernce only after add page to pages list.
1550 Frequency : always with long access acl
1552 Descriptoin: mds can't pack reply with long acl.
1553 Details : mds don't control size of acl but they limited by reint/getattr
1558 Frequency : on remount
1560 Description: external journal device not working after the remount
1561 Details : clear dev_rdonly flag for external journal devices in
1567 Description: shutdown vs evict race
1568 Details : client_disconnect_export vs connect request race.
1569 if client will evicted at this time - we start invalidate
1570 thread without referece to import and import can be freed
1576 Description: shrink LOV EAs before replying
1577 Details : correctly adjust LOV EA buffer for reply.
1582 Description: don't skip ost target if they assigned to file
1583 Details : Drop slow OSCs if we can, but not for requested start idx.
1584 This means "if OSC is slow and it is not the requested
1585 start OST, then it can be skipped, otherwise skip it only
1586 if it is inactive/recovering/out-of-space.
1588 Severity : enhancement
1590 Description: Update to RHEL5 kernel-2.6.18-92.1.17.el5.
1593 Frequency : rare, need acl's on inode.
1595 Description: client can't handle ost additional correctly
1596 Details : if ost was added after client connected to mds client can have
1597 hit lnet_try_match_md ... to big messages to wide striped files.
1598 in this case need teach client to handle config events about add
1599 lov target and update client max ea size at that event.
1601 Severity : enhancement
1603 Description: Update to sles9 kernel-2.6.5-7.314.
1605 Severity : enhancement
1607 Description: Update to SLES10 SP2 kernel-2.6.16.60-0.31.
1610 Frequency : Create a symlink file with a very long name
1612 Description: ldlm_cancel_pack()) ASSERTION(max >= dlm->lock_count + count)
1613 Details : If there is no extra space in the request for early cancels,
1614 ldlm_req_handles_avail() returns 0 instead of a negative value.
1619 Description: mds is deadlocked
1620 Details : in rare cases, inode in catalog can have i_no less than have parent
1621 i_no, this produce wrong order for locking during open, and
1622 parallel unlink can be lock open. this need teach mds_open to grab
1623 locks in resource id order, not at parent -> child order.
1625 Severity : enhancement
1627 Description: Add /proc entry for import status
1628 Details : The mdc, osc, and mgc import directories now have
1629 an import directory that contains useful import data for debugging
1630 connection problems.
1632 Severity : enhancement
1634 Description: Re-disable certain /proc logging
1635 Details : Enable and disable client's offset_stats, extents_stats and
1636 extents_stats_per_process stats logging on the fly.
1639 Frequency : Only on FC kernels 2.6.22+
1641 Description: oops in statahead
1642 Details : Do not drop reference count for the dentry from VFS when lookup,
1643 VFS will do that by itself.
1645 Severity : enhancement
1647 Description: Generic /proc file permissions
1648 Details : Set /Proc file permissions in a more generic way to enable non-
1649 root users operate on some /proc files.
1653 Description: Hitting mdc_commit_close() ASSERTION
1654 Details : Properly handle request reference release in
1655 ll_release_openhandle().
1659 Frequency : only patchless client
1660 Description: add workaround for race between add/remove dentry from hash
1662 Severity : enhancement
1664 Description: Allow OST glimpses to return PW locks
1668 Description: LBUG when llog conf file is full
1669 Details : When llog bitmap is full, ENOSPC should be returned for plain
1674 Description: Prevent import from entering FULL state when server in recovery
1678 Description: service mount cannot take device name with ":"
1679 Details : Only when device name contains ":/" will mount treat it as
1685 Description: replace ptlrpcd with the statahead thread to interpret the async
1686 statahead RPC callback
1690 Frequency : on recovery
1691 Description: I/O failures after umount during fail back
1692 Details : if client reconnected to restarted server we need join to recovery
1693 instead of find server handler is changed and process self eviction
1694 with cancel all locks.
1696 Severity : enhancement
1698 Description: Update to RHEL5 kernel-2.6.18-92.1.10.el5.
1702 Description: Kernel BUG tries to release flock
1703 Details : Lustre does not destroy flock lock before last reference goes
1704 away. So always drop flock locks when client is evicted and
1705 perform unlock regardless of successfulness of speaking to MDS.
1707 Severity : enhancement
1709 Description: Update to SLES10 SP2 kernel-2.6.16.60-0.27.
1711 Severity : enhancement
1713 Description: Upcall on Lustre log has been dumped
1714 Details : Allow for a user mode script to be called once a Lustre log has
1715 been dumped. It passes the filename of the dumped log to the
1716 script, the location of the script can be specified via
1717 /proc/sys/lnet/debug_log_upcall.
1722 Description: avoid messages about idr_remove called for id that is not allocated
1723 Details : Move assigment s_dev for clustered nfs to end of initialization,
1724 for avoid problem with error handling.
1729 Description: avoid Already found the key in hash [CONN_UNUSED_HASH] messages
1730 Details : When connection is reused this not moved from CONN_UNUSED_HASH into
1731 CONN_USED_HASH and this prodice warning when put connection again
1737 Description: avoid ASSERTION(client_stat->nid_exp_ref_count == 0) failed
1738 Details : release reference to stats when client disconnected, not
1739 when export destroyed for avoid races when client destroyed
1740 after main ost export.
1744 Description: more cleanup in mds_lov
1745 Details : add workaround for get valid ost count for avoid warnings about
1746 drop too big messages, not init llog cat under semphore which
1747 can be blocked on reconnect and break normal replay, fix access
1750 Severity : enhancement
1752 Description: Export bytes_read/bytes_write count on OSC/OST.
1756 Description: Early reply size mismatch, MGC loses connection
1757 Details : Apply the MGS_CONNECT_SUPPORTED mask at reconnect time so
1758 the connect flags are properly negotiated.
1762 Description: Properly propagate oinfo flags from lov to osc for statfs
1763 Details : restore missing copy oi_flags to lov requests.
1767 Description: exports in /proc are broken
1768 Details : recreate /proc entries for clients when they reconnect.
1770 Severity : enhancement
1772 Description: Add man pages for llobdstat(8), llstat(8), plot-llstat(8),
1773 : l_getgroups(8), lst(8), routerstat(8)
1774 Details : included man pages for llobdstat(8), llstat(8),
1775 : plot-llstat(8), l_getgroups(8), lst(8), routerstat(8)
1777 Severity : enhancement
1779 Description: Implement lustre ll_show_options method.
1781 Severity : enhancement
1783 Description: Update to SLES9 kernel-2.6.5-7.312.
1785 Severity : enhancement
1787 Description: Update to RHEL4 kernel-2.6.9-67.0.22.EL.
1791 Description: exports in /proc are broken
1792 Details : recreate /proc entries for clients when they reconnect.
1796 Description: don't fail open with -ERANGE
1797 Details : if client connected until mds will be know about real ost count
1798 get LOV EA can be fail because mds not allocate enougth buffer
1803 Description: Resolve device initialization race
1804 Details : Prevent proc handler from accessing devices added to the
1805 obd_devs array but yet be intialized.
1807 Severity : enhancement
1809 Description: configure's --enable-quota should check the
1810 : kernel .config for CONFIG_QUOTA
1811 Details : configure is terminated if --enable-quota is passed but
1812 : no quota support is in kernel
1814 Severity : enhancement
1816 Description: Update to SLES10 SP2 kernel-2.6.16.60-0.23.
1818 Severity : enhancement
1820 Description: Update to RHEL5 kernel-2.6.18-92.1.6.el5.
1824 Frequency : rare, on PPC clients
1825 Description: don't swab ost objects in response about directory, because
1827 Details : bug similar bug 14856, but in different function.
1829 Severity : enhancement
1831 Description: lfs quota tool enhancement
1832 Details : added units specifiers support for setquota, default to
1833 current uid/gid for quota report, short quota stats by
1834 default, nonpositional parameters for setquota, added
1835 llapi_quotactl manual page.
1837 Severity : enhancement
1839 Description: *optional* service tags registration
1840 Details : if the "service tags" package is installed on a Lustre node
1841 When the filesystem is mounted, a local-node service tag will
1842 be created. See http://inventory.sun.com/ for more information
1843 about the Service Tags asset management system.
1847 Description: Client runs out of low memory
1848 Details : Consider only lowmem when counting initial number of llap pages
1851 Frequency : occasional
1853 Description: add refcount for osc callbacks, so avoid panic on shutdown
1855 Severity : enhancement
1857 Description: Update to RHEL4 kernel-2.6.9-67.0.20.
1860 Frequency : testing only
1862 Description: sanity test 65a fails if stripecount of -1 is set
1863 Details : handle -1 striping on filesystem in ll_dirstripe_verify
1866 Frequency : only in unusual configurations
1868 Description: Kernel panic with find ost index.
1869 Details : lov_obd have panic if some OST's have sparse indexes.
1871 Severity : enhancement
1873 Description: Update to RHEL5 kernel-2.6.18-53.1.21.el5.
1876 Frequency : rarely, if filesystem is mounted with -o flock
1878 Description: do not process already freed flock
1879 Details : flock can possibly be freed by another thread before it reaches
1880 to ldlm_flock_completion_ast.
1883 Frequency : rarely, if filesystem is mounted with -o flock
1885 Description: LBUG during stress test
1886 Details : Need properly lock accesses the flock deadlock detection list.
1889 Frequency : rarely, if binaries are being run from Lustre
1891 Description: oops in page fault handler
1892 Details : kernel page fault handler can return two special 'pages' in
1893 error case, don't try dereference NOPAGE_SIGBUS and NOPAGE_OMM.
1896 Frequency : rarely, during shutdown
1898 Description: timeout with invalidate import.
1899 Details : ptlrpcd_check call obd_zombie_impexp_cull and wait request which
1900 should be handled by ptlrpcd. This produce long age waiting and
1901 -ETIMEOUT ptlrpc_invalidate_import and as result LASSERT.
1907 Description: ASSERTION(CheckWriteback(page,cmd)) failed
1908 Details : badly clear PG_Writeback bit in ll_ap_completion can produce false
1912 Frequency : only with broken builds/installations
1914 Description: no LBUG if lquota.ko and fsfilt_ldiskfs.ko are different versions
1915 Details : just return an error to a user, put a console error message
1917 Severity : enhancement
1919 Description: Update to RHEL5 kernel-2.6.18-53.1.19.el5.
1921 Severity : enhancement
1923 Description: Update to RHEL4 kernel-2.6.9-67.0.15.
1925 Severity : enhancement
1927 Description: enable MGS and MDT services start separately
1928 Details : add a 'nomgs' option in mount.lustre to enable start a MDT with
1929 a co-located MGS without starting the MGS, which is a complement
1930 to 'nosvc' mount option.
1933 Frequency : always, on big-endian systems
1935 Description: cleanup in ptlrpc code, related to PPC platform
1936 Details : store magic in native order avoid panic's in recovery on PPC
1937 node and forbid from this error in future. Also fix posibily
1938 of twice swab data. Fix get lov striping to userpace.
1941 Frequency : rarely, if replay get lost on server
1943 Description: server incorrectly drop resent replays lead to recovery failure.
1944 Details : do not drop replay according to msg flags, instead we check the
1945 per-export recovery request queue for duplication of transno.
1948 Frequency : after recovery
1950 Description: precreate to many object's after del orphan.
1951 Details : del orphan st in oscc last_id == next_id and this triger growing
1952 count of precreated objects. Set flag LOW to skip increase count
1953 of precreated objects.
1956 Frequency : rare, on clear nid stats
1958 Description: ASSERTION(client_stat->nid_exp_ref_count == 0)
1959 Details : when clean nid stats sometimes try destroy live entry,
1960 and this produce panic in free.
1963 Frequency : occasionally since 1.6.4
1965 Description: Stack overflow during MDS log replay
1966 Details : ease stack pressure by using a thread dealing llog_process.
1969 Frequency : very rare
1971 Description: MDT cannot be unmounted, reporting "Mount still busy"
1972 Details : Mountpoint references were being leaked during open reply
1973 reconstruction after an MDS restart. Drop mountpoint reference
1974 in reconstruct_open() and free dentry reference also.
1979 Description: wait until IO finished before start new when do lock cancel.
1980 Details : VM protocol want old IO finished before start new, in this case
1981 need wait until PG_writeback is cleared until check dirty flag
1982 and call writepages in lock cancel callback.
1987 Description: mds_mfd_close() ASSERTION(rc == 0)
1988 Details : In mds_mfd_close(), we need protect inode's writecount change
1989 within its orphan write semaphore to prevent possible races.
1992 Frequency : rare, on shutdown ost
1994 Description: don't hit live lock with umount ost.
1995 Details : shrink_dcache_parent can be in long loop with destroy dentries,
1996 use shrink_dcache_sb instead.
1999 Frequency : only when echo_client is used
2001 Description: don't panic with use echo_client
2002 Details : echo client pass NULL as client nid pointer and this produce NULL
2003 pointer dereference.
2006 Frequency : Always on 32-bit PowerPC systems
2008 Description: fix build on PPC32
2009 Details : compile code with -m64 flag produce wrong object file for PPC32.
2014 Description: MDS LBUG: ASSERTION(!IS_ERR(dchild))
2015 Details : In reconstruct_* functions, LASSERTs on both the data supplied
2016 by a client, and the data on disk are dangerous and incorrect.
2017 Change them with client eviction.
2019 Severity : enhancement
2021 Description: skiplist implementation simplification
2022 Details : skiplists are used to group compatible locks on granted list
2023 that was implemented as tracking first and last lock of each lock
2024 group the patch changes that to using doubly linked lists
2028 Description: delete compatibility for 32bit qdata
2029 Details : as planned, when lustre is beyond b1_8, lquota won't support 32bit
2030 qunit. That means servers of b1_4 and servers of b1_8 can't be
2031 used together if users want to use quota.
2034 Frequency : only with administrator action
2036 Description: mount failure if config log has invalid conf_param setting
2037 Details : If administrator specified an incorrect configuration parameter
2038 with "lctl conf_param" this would cause an error during future
2039 client mounts. Instead, ignore the bad configuration parameter.
2042 Frequency : blocks per group < blocksize*8 and uninit_groups is enabled
2044 Description: ldiskfs error: XXX blocks in bitmap, YYY in gd
2045 Details : If blocks per group is less than blocksize*8, set rest of the
2049 Frequency : Application do stride read on lustre
2051 Description: The read performance will drop a lot if the application does
2053 Details : Because the stride_start_offset are missing in stride read-ahead,
2054 it will cause clients read a lot of unused pages in read-ahead,
2055 then the read-performance drops.
2059 Description: more ldlm soft lockups
2060 Details : In ldlm_resource_add_lock(), call to ldlm_resource_dump()
2061 starve other threads from the resource lock for a long time in
2062 case of long waiting queue, so change the debug level from
2063 D_OTHER to the less frequently used D_INFO.
2065 Severity : enhancement
2067 Description: add -gid, -group, -uid, -user options to lfs find
2069 Severity : enhancement
2071 Description: ll_recover_lost_found_objs - recover objects in lost+found
2072 Details : OST corruption and subsequent e2fsck can leave objects in the
2073 lost+found directory. Using the "ll_recover_lost_found_objs"
2074 tool, these objects can be retrieved and data can be salvaged
2075 by using the object ID saved in the fid EA on each object.
2080 Description: this bug _only_ happens when inode quota limitation is very low
2081 (less than 12), so that inode quota unit is 1 at initialization.
2082 Details : if remaining quota equates 1, it is a sign to demonstate that quota
2083 is effective now. So least quota qunit should be 2.
2087 Description: Hung threads in invalidate_inode_pages2_range
2088 Details : The direct IO path doesn't call check_rpcs to submit a new RPC once
2089 one is completed. As a result, some RPCs are stuck in the queue
2094 Description: Procfs and llog threads access destoryed import sometimes.
2095 Details : Sync the import destoryed process with procfs and llog threads by
2096 the import refcount and semaphore.
2100 Description: mds fails to respond, threads stuck in ldlm_completion_ast
2101 Details : Sort source/child resource pair after updating child resource.
2106 Description: kernel BUG at ldiskfs2_ext_new_extent_cb
2107 Details : If insertion of an extent fails, then discard the inode
2108 preallocation and free data blocks else it can lead to duplicate
2113 Description: don't always update ctime in ext3_xattr_set_handle()
2114 Details : Current xattr code updates inode ctime in ext3_xattr_set_handle()
2115 In some cases the ctime should not be updated, for example for
2116 2.0->1.8 compatibility it is necessary to delete an xattr and it
2117 should not update the ctime.
2121 Description: add quota statistics
2122 Details : 1. sort out quota proc entries and proc code.
2123 2. add quota statistics
2128 Description: quotas are not honored with O_DIRECT
2129 Details : all writes with the flag O_DIRECT will use grants which leads to
2130 this problem. Now using OBD_BRW_SYNC to guard this.
2134 Bugzilla : 15713/16362
2135 Description: Assertion in iopen_connect_dentry in 1.6.3
2136 Details : looking up an inode via iopen with the wrong generation number can
2137 populate the dcache with a disconneced dentry while the inode
2138 number is in the process of being reallocated. This causes an
2139 assertion failure in iopen since the inode's dentry list contains
2140 both a connected and disconnected dentry.
2144 Description: assertion failure in ldlm_handle2lock()
2145 Details : fix a race between class_handle_unhash() and class_handle2object()
2146 introduced in lustre 1.6.5 by bug 13622.
2148 Severity : enhancement
2150 Description: superblock lock contention with many SMP cores on one client
2151 Details : several client filesystem locks were highly contended on SMP
2152 NUMA systems with 8 or more cores. Per-CPU datastructures
2153 and more efficient locking implemented to reduce contention.
2158 Description: Kernel BUG: sd_iostats_bump: unexpected disk index
2159 Details : remove the limit of 256 scsi disks in the sd_iostat patch
2164 Description: oops in sd_iostats_seq_show()
2165 Details : unloading/reloading the scsi low level driver triggers a kernel
2166 bug when trying to access the sd iostat file.
2171 Description: Kernel panics during QLogic driver reload
2172 Details : REQ_BLOCK_PC requests are not handled properly in the sd iostat
2173 patch, causing memory corruption.
2178 Description: journal_dev option does not work in b1_6
2179 Details : pass mount option during pre-mount.
2181 Severity : enhancement
2183 Description: Add a FIEMAP(FIle Extent MAP) ioctl for ldiskfs
2184 Details : FIEMAP ioctl will allow an application to efficiently fetch the
2185 extent information of a file. It can be used to map logical blocks
2186 in a file to physical blocks in the block device.
2189 Frequency : only with adaptive timeout enabled
2191 Description: DEBUG_REQ() bad paging request
2192 Details : ptlrpc_at_recv_early_reply() should not modify req->rq_repmsg
2193 because it can be accessed by reply_in_callback() without the
2197 Frequency : only on Cray X2
2199 Description: X2 build failures
2200 Details : fix build failures on Cray X2.
2204 Description: xid & resent requests
2205 Details : Initialize RPC XID from clock at startup (randomly if clock is
2210 Description: quota recovery deadlock during mds failover
2211 Details : This patch includes att18982, att18236, att18237 in bz14840.
2213 1. fix osts hang when mds does failover with quotaon
2214 2. prevent watchdog storm when osts threads wait for the
2219 Description: kernel panic on racer
2220 Details : Do not access dchild->d_inode when IS_ERR(dchild) is true.
2222 Severity : enhancement
2224 Description: Add lustre_start utility to start or stop multiple Lustre servers
2229 Description: Lustre GPF in {:ptlrpc:ptlrpc_server_free_request+373}
2230 Details : In case of memory pressure, list_del() can be called twice on
2231 req->rq_history_list, causing a kernel oops.
2235 Description: (ptllnd_peer.c:557:kptllnd_peer_check_sends()) ASSERTION(!in_interrupt()) failed
2236 Details : fix stack overflow in the distributed lock manager by defering
2237 export eviction after a failed ast to the elt thread instead of
2238 handling it in the dlm interpret routine.
2240 Severity : enhancement
2242 Description: More exported tunables for mballoc
2243 Details : Add support for tunable preallocation window and new tunables for
2244 large/small requests
2248 Description: Detect corruption of block bitmap and checking for preallocations
2249 Details : Checks validity of on-disk block bitmap. Also it does better
2250 checking of number of applied preallocations. When corruption is
2251 found, it turns filesystem readonly to prevent further corruptions.
2255 Frequency : only for big-endian servers
2256 Description: Check for big-endian system while mounting fs with extents feature
2257 Details : Mounting a filesystem with extents feature will fail on big-endian
2258 systems since ext3-based ldiskfs is not supported on big-endian
2259 systems. Can be over-ridden with "bigendian_extents" mount option.
2263 Description: Excessive recovery window
2264 Details : With AT enabled, the recovery window can be excessively long (6000+
2265 seconds). To address this problem, we no longer use
2266 OBD_RECOVERY_FACTOR when extending the recovery window (the connect
2267 timeout no longer depends on the service time, it is set to
2268 INITIAL_CONNECT_TIMEOUT now) and clients report the old service
2269 time via pb_service_time.
2272 Descriptoin: Don't sync journal after every i/o
2273 Details : Implement write RPC replay to allow server replies for write RPCs
2274 before data is on disk. However, this feature is disabled by
2275 default since some issues leading to data corruptions have been
2276 found during recovery (e.g. bug 19128). This feature can be enabled
2277 by running the following command on the OSSs:
2278 lctl set_param obdfilter.*.sync_journal=0
2282 Description: Watchdog triggered on MDS failover
2283 Details : enable OBD_CONNECT_MDT flag when connecting from the MDS so that
2284 the OSTs know that the MDS "UUID" can be reused for the same export
2285 from a different NID, so we do not need to wait for the export to
2290 Descriptoin: Lustre detected file system corruption with inode out of bounds
2291 Details : don't update i_size on MDS_CLOSE for directories. This causes
2292 directory corruptions on the MDT.
2294 -------------------------------------------------------------------------------
2296 2008-05-26 Sun Microsystems, Inc.
2298 * Support for kernels:
2299 2.6.5-7.311 (SLES 9),
2300 2.6.9-67.0.7.EL (RHEL 4),
2301 2.6.16.54-0.2.5 (SLES 10),
2302 2.6.18-53.1.14.el5 (RHEL 5),
2303 2.6.22.14 vanilla (kernel.org)
2304 * Client support for unpatched kernels:
2305 (see http://wiki.lustre.org/index.php?title=Patchless_Client)
2306 2.6.16 - 2.6.22 vanilla (kernel.org)
2307 * Due to problems with nested symlinks and FMODE_EXEC (bug 12652),
2308 we do not recommend using patchless RHEL4 clients with kernels
2309 prior to 2.6.9-55EL (RHEL4U5).
2310 * Recommended e2fsprogs version: 1.40.7-sun1
2311 * Note that reiserfs quotas are disabled on SLES 10 in this kernel.
2312 * RHEL 4 and RHEL 5/SLES 10 clients behaves differently on 'cd' to a
2313 removed cwd "./" (refer to Bugzilla 14399).
2314 * A new quota file format has been introduced in 1.6.5.
2315 The format conversion from prior releases is handled transparently,
2316 but releases older than 1.4.12/1.6.5 will not understand this new
2317 format. The automatic format conversion can be avoided by running
2318 the following command on the MDS before upgrading:
2319 'tunefs.lustre --param="mdt.quota_type=ug1" $MDTDEV'.
2320 For more information, please refer to bugzilla 13904.
2324 Description: quota performance fix
2325 Details : quota data is written in journalled mode instead of ordered to
2326 increase performance
2330 Description: lfs support for human-readable quota grace time strings
2331 Details : lfs setquota -t and lfs quota -t represent quota grace times
2332 in "XXwXXdXXhXXmXXs" format instead of large values in seconds
2335 Frequency : always with o2ib 1.3 and sles10
2337 Description: fix build with SLES10 and o2ib v3.
2338 Details : sles10 uses diffrent name for Module.symver file but configure
2339 assume this file has same name on RHEL/SLES/vanila kernels.
2342 Frequency : very rare, if additional xattrs are used on kernels >= 2.6.12
2344 Description: MDS may lose file striping (and hence file data) in some cases
2345 Details : If there are additional extended attributes stored on the MDS,
2346 in particular ACLs, SELinux, or user attributes (if user_xattr
2347 is specified for the client mount options) then there is a risk
2348 of attribute loss. Additionally, the Lustre file striping
2349 needs to be larger than default (e.g. striped over all OSTs),
2350 and an additional attribute must be stored initially in the
2351 inode and then increase in size enough to be moved to the
2352 external attribute block (e.g. ACL growing in size) for file
2355 Severity : enhancement
2357 Description: add message levels for liblustreapi
2362 Description: MDT cannot be unmounted, reporting "Mount still busy"
2363 Details : Mountpoint references were being leaked during open reply
2364 reconstruction after an MDS restart. Drop mountpoint reference
2365 in reconstruct_open() and free dentry reference also.
2370 Description: fix for occasional failure case of -ENOSPC in recovery-small tests
2371 Details : Move the 'good_osts' check before the 'total_bavail' check. This
2372 will result in an -EAGAIN and in the exit call path we call
2373 alloc_rr() which will with increasing aggressiveness attempt to
2374 aquire precreated objects on the minimum number of required OSCs.
2378 Description: Use old size assignment to avoid deadlock
2379 Details : This reverts the changes in bugs 2369 and bug 14138 that introduced
2380 the scheduling while holding a spinlock. We do not need locking
2381 for size in ll_update_inode() because size is only updated from
2382 the MDS for directories or files without objects, so there is no
2383 other place to do the update, and concurrent access to such inodes
2384 are protected by the inode lock.
2388 Description: Use __u64 instead of int for valid bits
2392 Description: resolve "_IOWR redefined" build error on SLES10
2396 Description: dump the memory debugging after all modules are unloaded to
2397 suppress false negative in conf_sanity test 39
2401 Description: the recovery timer never expires
2402 Details : for new client connect request, the recovery timer should not be
2403 reset, otherwise recovery timer will never expired, if the old
2404 client never come. Only old client connect and first connection
2405 req should trigger recovery timer reset.
2409 Description: the min numbers of lproc stats are wrong
2410 Details : adding a new constant LC_MIN_INIT and use it for initialization
2414 Frequency : always with interactive lfs
2416 Description: Reinitialize optind to 0 so that interactive lfs works in all cases
2419 Frequency : with multiple concurrent readdir processes in same directory
2420 Bugzilla : 15406, 15169, 15175
2421 Description: misc fixes for directory readahead.
2422 Details : prevent previous statahead async RPC callback from processing the
2423 current "statahead_info", race condition between async RPC callback
2424 add dentry into dentry hash table and "ls" thread revalidate such
2425 dentry, statahead his/miss control for hidden items, and so on.
2427 Severity : enhancement
2429 Description: build kernel-ib packages for OFED 1.3 in our release cycle
2433 Description: incore types cleaning in quota code (with respect to 64-bit limits)
2434 Details : several u32 variables declarations are replaced with u64 declarations
2439 Description: fix SLES kernel versioning
2440 Details : the kernel version for our SLES 10 kernel did not include a "-"
2441 before the "smp" at the end. while this was not a problem in
2442 general, it did mean that software trying to use the kernel
2443 version to try to detect a vendor specific kernel would fail.
2444 this was most evident by the OFED build scripts.
2449 Description: Don't update lov_desc members until making sure they are valid
2450 Details : When updating lov_desc members via proc fs, need fix their
2451 validities before doing the real update.
2454 Frequency : very rare
2456 Description: don't put request into delay list while invalidate in flight.
2457 Details : ptlrpc_delay_request sometimes put in delay list while invalidate
2458 import in flight. this produce timeout for invalidate and sometimes
2459 can cause stale data.
2461 Severity : enhancement
2463 Description: Update kernel to SLES9 2.6.5-7.311.
2465 Severity : enhancement
2467 Description: Update kernel to RHEL4 2.6.9-67.0.7.
2472 Frequency : on PPC only
2473 Description: not convert ost objects for directory because it's not exist.
2474 Details : ll_dir_getstripe assume dirrectory has ost objects but this wrong.
2476 Severity : enhancement
2478 Description: Fix warnings with compile liblustre at sles10/rhel5 which have
2479 __u64 as usingied long long type.
2482 Frequency : rare, on shutdown
2484 Description: race process ast vs remove callback
2485 Details : removing callback before disconnect import open race with
2486 processing callback.
2488 Severity : enhancement
2490 Description: Update kernel to SLES9 2.6.5-7.311.
2492 Severity : enhancement
2494 Description: Files open for execute are not marked busy on SLES10
2495 Details : Add FMODE_EXEC to SLES10 SP1 server kernel series.
2497 Severity : enhancement
2499 Description: Add server support for vanilla-2.6.22.14.
2502 Frequency : occasional
2504 Description: Avoid lov_create() getting stuck in obd_statfs_rqset()
2505 Details : If an OST is down the MDS will hang indefinitely in
2506 obd_statfs_rqset() waiting for the statfs data. While for
2507 MDS QOS usage of statfs, it should not stuck in waiting.
2509 Severity : enhancement
2511 Description: Disable adaptive timeouts by default
2514 Frequency : on network error
2516 Description: panic with double free request if network error
2517 Details : mdc_finish_enqueue is finish request if any network error occuring,
2518 but it's true only for synchronus enqueue, for async enqueue
2519 (via ptlrpcd) this incorrect and ptlrpcd want finish request
2523 Frequency : rare, on recovery
2525 Description: read procfs can produce deadlock in some situation
2526 Details : Holding lprocfs lock with send rpc can produce block for destroy
2527 obd objects and this also block reconnect with -EALREADY.
2528 This isn't fix all lprocfs bugs - but make it rare.
2530 Severity : enhancement
2532 Description: Update kernel to RHEL5 2.6.18-53.1.14.el5.
2535 Frequency : frequent on X2 node
2537 Description: mdc_set_open_replay_data LBUG
2538 Details : Set replay data for requests that are eligible for replay.
2543 Description: lustre_mgs: operation 101 on unconnected MGS
2544 Details : When MGC is disconnected from MGS long enough, MGS will evict the
2545 MGC, and late on MGC cannot successfully connect to MGS and a lot
2546 of the error messages complaining that MGS is not connected.
2549 Frequency : rare, depends on device drivers and load
2551 Description: MDS or OSS nodes crash due to stack overflow
2552 Details : Code changes in 1.6.4 increased the stack usage of some functions.
2553 In some cases, in conjunction with device drivers that use a lot
2554 of stack, the MDS (or possibly OSS) service threads could overflow
2555 the stack. One change which was identified to consume additional
2556 stack has been reworked to avoid the extra stack usage.
2558 Severity : enhancement
2560 Description: Update to RHEL5 latest kernel-2.6.18-53.1.13.el5.
2562 Severity : enhancement
2564 Description: Update to SLES10 SP1 latest kernel-2.6.16.54-0.2.5.
2566 Severity : enhancement
2568 Description: Update to RHEL5 latest kernel-2.6.18-53.1.6.el5.
2570 Serverity : enhancement
2572 Description: Update RHEL4 kernel to 2.6.9-67.0.4.
2575 Frequency : rare on shutdown OST
2577 Description: Don't allow skipping OSTs if index has been specified.
2578 Details : Don't allow skipping OSTs if index has been specified, make locking
2579 in internal create lots better.
2584 Description: ASSERTION(!PageDirty(page)) failed
2585 Details : Wrong check could lead to an assertion failure under specific
2591 Description: LBUG in ptlrpc_check_set() bad phase ebc0de00
2592 Details : access to bitfield in structure is always rounded to long
2593 and this produce problem with not atomic change any bit.
2598 Description: Lustre make rpms failed.
2599 Details : Remove ldiskfs spec file to avoids rpmbuild be confused when
2600 builds Lustre rpms from tarball.
2602 Severity : enhancement
2604 Description: Update to SLES9 SP4 kernel-2.6.5-7.308.
2607 Frequency : rare on shutdown OST
2609 Description: If llog cancel was not send before clean_exports phase, this can
2610 produce deadlock in llog code.
2611 Details : If llog thread has last reference to obd and call class_import_put
2612 this produce deadlock because llog_cleanup_commit_master wait when
2613 last llog_commit_thread exited, but this never success because was
2614 called from llog_commit_thread.
2617 Frequency : only if OST index is skipped
2619 Description: NULL lov_tgts causing MDS oops
2620 Details : more safe checks for NULL lov_tgts for avoid oops.
2622 Severity : enhancement
2624 Description: Update to RHEL4 latest kernel-2.6.9-67.0.1.EL.
2626 Severity : enhancement
2628 Description: Update to RHEL5 latest kernel-2.6.18-53.1.4.el5.
2633 Description: make mgs_setparam() handle fsname containing dash
2634 Details : fsname containing a dash does not work with lctl conf_param
2636 Severity : enhancement
2638 Description: Update to RHEL4 Update-6 kernel-2.6.9-67.EL.
2641 Frequency : rare, in recovery and (or) destroy lovobjid file.
2643 Description: rewrite lov objid code.
2644 Details : Cleanup for lov objid code, remove scability problems and wrong
2645 locking. Fix sending last_id into ost.
2647 Severity : enhancement
2649 Description: Update to SLES10 SP1 latest kernel-2.6.16.54-0.2.3.
2651 Severity : enhancement
2653 Description: Update to RHEL5 Update-1 kernel 2.6.18-53.el5.
2654 Details : Use d_move_locked instead of __d_move.
2657 Frequency : rare, at shutdown
2659 Description: access already free / zero obd_namespace.
2660 Details : if client_disconnect_export was called without force flag set,
2661 and exist connect request in flight, this can produce access to
2662 NULL pointer (or already free pointer) when connect_interpret
2663 store ocd flags in obd_namespace.
2666 Frequency : only at startup
2668 Description: not alloc memory with spinlock held.
2669 Details : allocation memory with GFP_KERNEL can produce sleep deadlock,
2670 if any spinlock held.
2675 Description: lfs find does not continue on file error
2676 Details : Continue other files processing when a file/dir is absent.
2680 Description: Inconsistent usage of lustre_pack_reply()
2681 Details : Standardize the usage of lustre_pack_reply() such that it
2682 always generate a CERROR on failure.
2685 Frequency : very rare
2687 Description: Fix replay if there is an un-replied request and open
2688 Details : In some cases, older replay request will revert the
2689 mcd->mcd_last_xid on MDS which is used to record the client's
2690 latest sent request.
2692 Severity : enhancement
2694 Description: Update to RHEL5 kernel 2.6.18-8.1.15.el5.
2696 Severity : enhancement
2698 Description: Update to SLES10 SP1 kernel 2.6.16.53-0.16
2700 Severity : enhancement
2702 Description: Update to SLES9 kernel-2.6.5-7.287.3.
2704 Severity : enhancement
2706 Description: Update to RHEL4 kernel-2.6.9-55.0.12.EL.
2708 Severity : enhancement
2710 Description: Build SLES10 patchless client fails
2711 Details : The configure was broken by run ./configure with
2712 --with-linux-obj=.... argument for patchless client. When the
2713 configure use --with-linux-obj, the LINUXINCLUDE= -Iinclude
2714 can't search header adequately. Use absolute path such as
2715 -I($LINUX)/include instead.
2717 Severity : enhancement
2719 Description: Lustre Page Accounting
2720 Details : New macros for page alloc and free which enable accounting
2721 of page allocation of Lustre. Use percpu counters to store memory
2722 and page statistics.
2725 Frequency : only if debugging is disabled
2727 Description: LASSERT_{REQ,REP}SWAB macros are buggy
2728 Details : If SWAB_PARANOIA is disabled, the LASSERT_REQSWAB and
2729 LASSERT_REPSWAB macros become no-ops, which is incorrect. Drop
2730 these macros and replace them with their definitions instead.
2735 Description: interrupt oig_wait produce painc on resend.
2736 Details : brw_redo_request can be used for resend requests from ptlrpcd and
2737 private set, and this produce situation when rq_ptlrpcd_data not
2738 copyed to new allocated request and triggered LBUG on assert
2739 req->rq_ptlrpcd_data != NULL. But this member used only for wakeup
2740 ptlrpcd set if request is changed and can be safety changed to use
2743 Severity : enhancement
2745 Description: organize the server-side client stats on per-nid basis
2746 Details : Change the structure of stats under obdfilter and mds to
2754 The "uuid"s file would list the uuids of _active_ exports.
2755 And the clear entry is to clear all stats and stale nids.
2760 Description: Processes looping in ll_readdir() on Lustre clients finally causing
2761 a full node pseudo-hang
2762 Details : Concurrent access to the same directory from multiple clients with
2763 intensive file creation/removal can cause a client node to spin in
2764 ll_readdir(). i_version must be increased every time the lock is
2765 cancelled to ensure a revalidate is done.
2770 Description: touch file failed when fs is not full
2771 Details : OST in recovery should not be discarded by MDS in alloc_qos(),
2772 otherwise we can get ENOSP while fs is not full.
2775 Frequency : only for Cray XT3
2776 Bugzilla : 12829/13455
2777 Description: Changing primary group doesn't change the group lustre assigns to
2779 Details : When CRAY_XT3 is defined, the fsgid supplied by the client is
2780 overridden with the primary group provided by the group upcall,
2781 whereas the supplied fsgid can be trusted if it is in the list of
2782 supplementary groups returned by the group upcall.
2784 Severity : enhancement
2786 Description: Root Squash Functionality
2787 Details : Implementation of NFS-like root squash capability. Specifically,
2788 don't allow someone with root access on a client node to be able
2789 to manipulate files owned by root on a server node.
2791 Severity : enhancement
2793 Description: Slow trucate/writes to huge files at high offsets.
2794 Details : Directly associate cached pages to lock that protect those pages,
2795 this allows us to quickly find what pages to write and remove
2796 once lock callback is received.
2801 Description: Too many locks accumulating on client during NFS usage
2802 Details : mds_open improperly used accmode to find out access mode to a
2803 file. Also mdc_intent_lock logic to find out if we already have
2804 lock similar to just received was flawed since introduction of
2805 skiplists - locks are now added to the front of the granted
2810 Description: Hit ASSERTION(obd->obd_stopping == 1) failed in some setup failed
2812 Details : In obd setup failure handler, obd_stopping will not necessarily to
2813 be 1, and obd_set_up should also be checked to make sure whether
2814 obd is completely setup.
2816 Severity : enhancement
2818 Description: Allow masking D_WARNING, D_ERROR messages from console
2819 Details : Console messages can now be disabled via lnet.printk.
2824 Description: User code with malformed file open parameter crashes client node
2825 Details : Before packing join_file req, all the related reference should be
2826 checked carefully in case some malformed flags cause fake join_file
2832 Description: shrink/enlarge qunit size when needed; fix the problem of coarse
2833 grain of quota doing harm to quota's accuracy
2834 Details : qunit size will be changed when quota limitation is too low/high;
2835 record the pending quota write in order to get more accureate
2836 quota; delete the patch for bug12588, which is unnecessary when
2837 this patch is landed. This bug also contains fixes for bug 14526,
2838 14299, 14601 and 13794.
2842 Description: LDLM_ENQUEUE races with LDLM_CP_CALLBACK
2843 Details : ldlm_completion_ast() assumes that a lock is granted when the req
2844 mode is equal to the granted mode. However, it should also check
2845 that LDLM_FL_CP_REQD is not set.
2849 Description: Heavy nfs access might result in deadlocks
2850 Details : After ELC code landed, it is now improper to enqueue any mds
2851 locks under och_sem, because enqueue might want to decide to
2852 cancel open locks for same inode we are holding och_sem for.
2856 Description: 35% write performance drop with ldiskfs2 when quotas are on
2857 Details : Enable ext3 journalled quota by default to improve performance
2858 when quotas are turned on.
2862 Description: Client eviction while running blogbench
2863 Details : A lot of unlink operations with concurrent I/O can lead to a
2864 deadlock causing evictions. To address the problem, the number of
2865 oustanding OST_DESTROY requests is now throttled to
2866 max_rpcs_in_flight per OSC and LDLM_FL_DISCARD_DATA blocking
2867 callbacks are processed in priority.
2870 Frequency : RHEL4 only
2872 Description: mkfs is very slow on IA64/RHEL4
2873 Details : A performance regression has been discovered in the MPT Fusion
2874 driver between versions 3.02.73rh and 3.02.99.00rh. As a
2875 consequence, we have downgraded the MPT Fusion driver in the RHEL4
2876 kernel from 3.02.99.00 to 3.02.73 until this problem is fixed.
2879 Frequency : PPC/PPC64 only
2881 Description: conflicts between asm-ppc64/types.h and lustre_types.h
2882 Details : fix duplicated definitions between asm-ppc64/types.h and
2883 lustre_types.h on PPC.
2886 Frequency : PPC/PPC64 only
2888 Description: asm-ppc/segment.h does not exist
2889 Details : fix compile issue on PPC.
2893 Description: data checksumming impacts single node performance
2894 Details : add support for several checksum algorithms. Currently, CRC32 and
2895 Adler-32 are supported. The checksum type can be changed on the fly
2896 through /proc/fs/lustre/osc/*/checksum_type.
2900 Description: use adler32 for page checksums
2901 Details : when available, use the Adler-32 algorithm instead of CRC32 for
2906 Description: better handle error messages in extents code
2908 Severity : enhancement
2910 Description: SNMP support enhancement
2911 Details : Adding total number of sampled request for an MDS node in snmp
2914 Severity : enhancement
2916 Description: Optimize ldlm waiting list processing for PR extent locks
2917 Details : When processing waiting list for read extent lock and meeting read
2918 lock that is same or wider to it that is not contended, skip
2919 processing rest of the list and immediatelly return current
2920 status of conflictness, since we are guaranteed there are no
2921 conflicting locks in the rest of the list.
2925 Description: Time out and refuse to reconnect
2926 Details : When the failover node is the primary node, it is possible
2927 to have two identical connections in imp_conn_list. We must
2928 compare not conn's pointers but NIDs, otherwise we
2929 can defeat connection throttling.
2933 Description: Client not clear own cache if answer to reconnect is lost.
2934 Details : Client gets evicted from server. Now client also thinks it is
2935 disconnected (or gets ENOTCONN on its operation) and decides to
2936 reconnect. Server receives reconnect message, but cannot find
2937 export. New export is created that is fully valid (new cookie!),
2938 but reply is lost and not reported to client. Client reconnects
2939 again and gets back a just-created connection, but it is not new
2940 so client thinks it was not evicted and does not do recovery.
2944 Description: Detect stride IO mode in read-ahead
2945 Details : When a client does stride read, read-ahead should detect that and
2946 read-ahead pages according to the detected stride pattern.
2950 Description: build for x2 fails
2951 Details : fix compile issue on Cray systems.
2953 Severity : enhancement
2955 Description: implement readv/writev APIs(aio_read/aio_writes in newer kernels)
2956 Details : This greatly improves speed of NFS writes on 2.6 kernels.
2959 Frequency : only on PPC/SLES10
2961 Description: "BITS_PER_LONG is not 32 or 64" in linux/idr.h
2962 Details : On SLES10/PPC, fs.h includes idr.h which requires BITS_PER_LONG to
2963 be defined. Add a hack in mkfs_lustre.c to work around this compile
2968 Description: LASSERT on MDS when client holding flock lock dies
2969 Details : ldlm pool logic depends on number of granted locks equal to
2970 number of released locks which is not true for flock locks, so
2971 just exclude such locks from consideration.
2975 Description: MDS deadlock with many ll_sync_lov threads and I/O stalled
2976 Details : Use fsfilt_sync() for both the whole filesystem sync and
2977 individual file sync to eliminate dangerous inode locking
2978 with I_LOCK that can lead to a deadlock.
2982 Description: Update an obsolete wirecheck.c generator
2983 Details : Update wirecheck.c/wirehdr.c and regenerate wiretest.c
2987 Description: Client can panic on open sometimes
2988 Details : It is possible that we try to free already freed request in
2989 ll_file_open in some error cases when we send request from
2994 Description: performance in 1.6.3
2995 Details : Force q->max_phys_segments to MAX_PHYS_SEGMENTS on SLES10 to be
2996 sure that 1MB requests are not fragmented by the block layer.
3000 Description: LDLM soft lockups - improvement
3001 Details : It is be possible to send the lock handle along with each read
3002 or write request because the client is already doing a lock match
3003 itself so there isn't any reason the OST should have to re-do that
3009 Description: lfs quota fails with deactivated OSTS
3010 Details : With this patch, three improvements are included:
3011 1. detete the softlimit in mds and osts when use "lfs quota".
3012 2. display the inaccurate data in the output of "lfs quota".
3013 3. try to get quota info when "lfs quota" is executed.
3018 Description: Extent locks not granted with no conflicts sometimes.
3019 Details : When race occurs in glimpse handler and nothing is returned,
3020 we do not reprocess the queue after lock cancel, and that leads
3021 to a stall until next activity on a resource
3024 Frequency : failover with quotaon
3026 Description: during mds failovers with quota on, OSTs got into deadlock state
3027 and causing dumpstack.
3028 Details : for every quota slave, at any time, there is only one quota req
3029 is sent to quota master for every uid/gid. Before that quota req
3030 returns, all the thread relative to the same uid/gid will wait.
3031 So if the quota req is lost because mds failovers or any other
3032 reasons, this bug will be hit. Now, dqacq_interpret() will handle
3033 quota reqs who time out.
3035 Severity : enhancement
3038 Description: when quota slave checks if quota is enough, there is an unnecessary
3040 Details : place this wait on necessary place instead of always waiting.
3042 --------------------------------------------------------------------------------
3044 2007-12-07 Cluster File Systems, Inc. <info@clusterfs.com>
3046 * Support for kernels:
3047 2.6.5-7.286 (SLES 9),
3048 2.6.9-55.0.9.EL (RHEL 4),
3049 2.6.16.53-0.8 (SLES 10),
3050 2.6.18-8.1.14.el5 (RHEL 5),
3051 2.6.18.8 vanilla (kernel.org)
3052 * Client support for unpatched kernels:
3053 (see http://wiki.lustre.org/index.php?title=Patchless_Client)
3054 2.6.16 - 2.6.22 vanilla (kernel.org)
3055 * Due to recently discovered recovery problems, we do not recommend
3056 using patchless RHEL 4 clients with this or any earlier release.
3057 * Recommended e2fsprogs version: 1.40.2-cfs1
3058 * Note that reiserfs quotas are disabled on SLES 10 in this kernel.
3061 Frequency : occasional
3063 Description: MDS hang or stay in waiting lock
3064 Details : If client receive lock with CBPENDING flag ldlm need send lock
3065 cancel as separate rpc, to avoid situation when cancel request
3066 can't processed due all i/o threads stay in wait lock.
3069 Frequency : occasional
3071 Description: Do not fail import if osc_interpret_create gets -EAGAIN
3072 Details : If osc_interpret_create got -EAGAIN it immediately exits and
3073 wakeup oscc_waitq. After wakeup oscc_wait_for_objects call
3074 oscc_has_objects and see OSC has no objests and call
3075 oscc_internal_create to resend create request.
3077 Severity : enhancement
3079 Description: Update kernel patches for SLES10 2.6.16.53-0.8.
3080 Details : Update which_patch & target file for SLES10 latest kernel.
3082 Severity : enhancement
3084 Description: add --type and --size parameters to lfs find
3085 Details : Enhance lfs find by adding filetype and filesize parameters. Also
3086 multiple OBDs can now be specified for the --obd option.
3088 Severity : enhancement
3090 Description: eliminate client locks in face of contention
3091 Details : file contention detection and lockless i/o implementation
3092 for contended files.
3094 Severity : enhancement
3096 Description: Remove client patches from SLES 10 kernel.
3097 Details : This causes SLES 10 clients to behave as patchless clients
3098 even on a Lustre-patched (server) kernel.
3100 Severity : enhancement
3102 Description: use i_size_read and i_size_write in 2.6 port
3103 Details : replace inode->i_size access with i_size_read/write()
3106 Frequency : when removing large files
3108 Description: scheduling issue during removal of large Lustre files
3109 Details : Don't take the BKL in fsfilt_ext3_setattr() for 2.6 kernels.
3110 It causes scheduling issues when removing large files (17TB in the
3116 Description: 1.4.11 Can't handle directories with stripe set and extended ACLs
3117 Details : Impossible (EPROTO is returned) to access a directory that has a
3118 non-default striping and ACLs.
3121 Frequency : only on ppc
3123 Description: /proc/fs/lustre/devices broken on ppc
3124 Details : The patch as applied to 1.6.2 doesn't look correct for all arches.
3125 We should make sure the type of 'index' is loff_t and then cast
3126 explicitly as needed below. Do not assign an explicitly cast
3130 Frequency : only for rhel5
3132 Description: Kernel patches update for RHEL5 2.6.18-8.1.10.el5.
3133 Details : Modify the target file & which_kernel.
3136 Frequency : if the uninit_groups feature is enabled on ldiskfs
3138 Description: e2fsck reports "invalid unused inodes count"
3139 Details : If a new ldiskfs filesystem is created with the "uninit_groups"
3140 feature and only a single inode is created in a group then the
3141 "bg_unused_inodes" count is incorrectly updated. Creating a
3142 second inode in that group would update it correctly.
3147 Description: buffer overruns could theoretically occur
3148 Details : llapi_semantic_traverse() modifies the "path" argument by
3149 appending values to the end of the origin string, and a buffer
3150 overrun may occur. Adding buffer overrun check in liblustreapi.
3152 Severity : enhancement
3154 Description: Add jbd statistics patch for RHEL5 and 2.6.18-vanilla.
3157 Frequency : only if filesystem is inconsistent
3159 Description: handle "serious error: objid * already exists" more gracefully
3160 Details : If LAST_ID value on disk is smaller than the objects existing in
3161 the O/0/d* directories, it indicates disk corruption and causes an
3162 LBUG(). If the object is 0-length, then we should use the existing
3163 object. This will help to avoid a full fsck in most cases.
3165 Severity : enhancement
3167 Description: Kernel patches update for RHEL4 2.6.9-55.0.6.
3168 Details : Modify vm-tunables-rhel4.patch.
3170 Severity : enhancement
3172 Description: Kernel config for 2.6.18-vanilla.
3173 Details : Modify targets/2.6-vanilla.target.in.
3174 Add config file kernel-2.6.18-2.6-vanilla-i686.config.
3175 Add config file kernel-2.6.18-2.6-vanilla-i686-smp.config.
3176 Add config file kernel-2.6.18-2.6-vanilla-x86_64.config.
3177 Add config file kernel-2.6.18-2.6-vanilla-x86_64-smp.config.
3180 Frequency : occasional
3182 Description: improve handling recoverable errors
3183 Details : If request processed with error which can be recoverable on server
3184 request should be resend, otherwise page released from cache and
3190 Description: Kernel patches update for RHEL5 2.6.18-8.1.14.el5.
3191 Details : Modify target file & which_patch.
3192 A flaw was found in the IA32 system call emulation provided
3193 on AMD64 and Intel 64 platforms. An improperly validated 64-bit
3194 value could be stored in the %RAX register, which could trigger an
3195 out-of-bounds system call table access. An untrusted local user
3196 could exploit this flaw to run code in the kernel
3197 (ie a root privilege escalation). (CVE-2007-4573).
3201 Description: change order of libsysio includes
3202 Details : '#include sysio.h' should always come before '#include xtio.h'
3204 Severity : enhancement
3206 Description: adapt the lustre_config script to support the upgrade case
3207 Details : Add "-u" option for lustre_config script to support upgrading 1.4
3208 server targets to 1.6 in parallel.
3213 Description: To avoid grant space > avaible space when the disk is almost
3214 full. Without this patch you might see the error "grant XXXX >
3215 available" or some LBUG about grant, when the disk is almost
3217 Details : In filter_check_grant, for non_grant cache write, we should
3218 check the left space by if (*left > ungranted + bytes), instead
3219 of (*left > ungranted), because only we are sure the left space
3220 is enough for another "bytes", then the ungrant space should be
3221 increase. In client, we should update cl_avail_grant only there is
3222 OBD_MD_FLGRANT in the reply.
3227 Description: Update RHEL 4 kernel to fix local root privilege escalation.
3228 Details : Update to the latest RHEL 4 kernel to fix the vulnerability
3229 described in CVE-2007-4573. This problem could allow untrusted
3230 local users to gain root access.
3233 Frequency : when using O_DIRECT and quotas
3235 Description: Incorrect file ownership on O_DIRECT output files
3236 Details : block usage reported by 'lfs quota' does not take into account
3237 files that have been written with O_DIRECT.
3241 Description: (rw.c:1323:ll_read_ahead_pages()) ASSERTION(page_idx > ria->ria_stoff) failed
3242 Details : Once the unmatched stride IO mode is detected, shrink the stride-ahead
3243 window to 0. If it does hit cache miss, and read-pattern is still
3244 stride-io mode, does not reset the stride window, but also does not
3245 increase the stride window length in this case.
3247 --------------------------------------------------------------------------------
3249 2007-09-27 Cluster File Systems, Inc. <info@clusterfs.com>
3251 * Support for kernels:
3252 2.6.5-7.286 (SLES 9),
3253 2.6.9-55.0.2.EL (RHEL 4),
3254 2.6.16.46-0.14 (SLES 10),
3255 2.6.18-8.1.8.el5 (RHEL 5),
3256 2.6.18.8 vanilla (kernel.org)
3257 * Client support for unpatched kernels:
3258 (see http://wiki.lustre.org/index.php?title=Patchless_Client)
3259 2.6.16 - 2.6.21 vanilla (kernel.org)
3260 * Due to recently discovered recovery problems, we do not recommend
3261 using patchless RHEL 4 clients with this or any earlier release.
3262 * Recommended e2fsprogs version: 1.40.2-cfs1
3263 * Note that reiserfs quotas are disabled on SLES 10 in this kernel.
3267 Description: Fix errors in lfs documentation
3268 Details : Fixes man pages
3270 Severity : enhancement
3272 Description: Adaptive timeouts
3273 Details : RPC timeouts adapt to changing server load and network
3274 conditions to reduce resend attempts and improve recovery time.
3276 Severity : enhancement
3278 Description: llapi_file_create() does not allow some changes
3279 Details : add llapi_file_open() that allows specifying the file creation
3280 mode and open flags, and also returns an open file handle.
3282 Severity : enhancement
3284 Description: df doesn't work properly if diskfs blocksize != 4K
3285 Details : Choose biggest blocksize of OST's as the LOV's blocksize.
3288 Frequency : always on directories with default striping set
3290 Description: lfs find on -1 stripe looping in lsm_lmm_verify_common()
3291 Details : Avoid lov_verify_lmm_common() on directory with -1 stripe count.
3293 Severity : enhancement
3295 Description: merge and cleanup kernel patches.
3296 Details : Remove mnt_lustre_list in vfs_intent-2.6-rhel4.patch.
3298 Severity : enhancement
3300 Description: RedHat Update kernel for RHEL5
3301 Details : Kernel config file for RHEL5.
3303 Severity : enhancement
3305 Description: OSS needs mutliple precreate threads
3306 Details : Add ability to start more than one create thread per OSS.
3309 Frequency : only with quota on the root user
3311 Description: mds_obd_create error creating tmp object
3312 Details : When the user sets quota on root, llog will be affected and can't
3313 create files and write files.
3316 Frequency : Always on ia64 patchless client, and possibly others.
3318 Description: Add EXPORT_SYMBOL check for node_to_cpumask symbol.
3319 Details : This allows the patchless client to be loaded on architectures
3320 without this export.
3325 Description: disorder of journal start and llog_add cause deadlock.
3326 Details : in llog_origin_connect, journal start should happen before
3327 llog_add keep the same order as other functions to avoid
3330 Severity : enhancement
3332 Description: RedHat Update kernel for RHEL5
3333 Details : Modify the kernel config file more closer RHEL5.
3335 Severity : enhancement
3337 Description: Build failure against Centos5 (RHEL5)
3338 Details : Define PAGE_SIZE when it isn't present.
3341 Frequency : occasionally when using NFS
3343 Description: "ll_intent_file_open()) lock enqueue: err: -13" with nfs
3344 Details : with NFS, the anon dentry's parent was set to itself in
3345 d_alloc_anon(), so in MDS, we use rec->ur_fid1 to find the
3346 corresponding dentry other than use rec->ur_name.
3348 Severity : enhancement
3350 Description: enable data checksumming by default
3351 Details : enable checksum by default, allow --disable-checksum
3352 configure option and "-o nochecksum" mount option. Checksums
3353 can also be disabled at runtime via $LPROC/osc/*/checksum_pages.
3356 Frequency : Occasionally with failover
3358 Description: Client eviction due to failover config
3359 Details : after a connection loss, the lustre client should attempt to
3360 reconnect to the last active server first before trying the
3361 other potential connections.
3363 Severity : enhancement
3365 Description: client-side metadata stat-ahead during readdir(directory readahead)
3366 Details : perform client-side metadata stat-ahead when the client detects
3367 readdir and sequential stat of dir entries therein
3372 Description: when mds and osts use different quota unit(32bit and 64bit),
3373 quota will be released repeatly.
3374 Details : void sending multiple quota reqs to mds, which will keep the
3375 status between the reqs.
3378 Frequency : only with liblustre clients on XT3
3380 Description: evictions taking too long
3381 Details : allow llrd to evict clients directly on OSTs
3385 Description: osts not allocated evenly to files
3386 Details : change the condition to increase offset_idx
3388 Severity : enhancement
3390 Description: self-adjustable client's lru lists
3391 Details : use adaptive algorithm for managing client cached locks lru
3392 lists according to current server load, other client's work
3393 pattern, memory activities, etc. Both, server and client
3394 side namespaces provide number of proc tunables for controlling
3399 Description: rewrite ext2-derived code in llite/dir.c and obdclass/uuid.c
3400 Details : rewrite inherited code (uuid parsing code from ext2 utils and
3401 readdir code from ext3) from scratch preserving functionality.
3405 Description: Only those disconnect error should be returned by rq_status.
3406 Details : In open/enqueue processs, Some errors, which will cause client
3407 disconnected, should be returned by rq_status, while other
3408 errors should still be returned by intent, then mdc or llite will
3411 Severity : enhancement
3413 Description: Tune the kernel for good SCSI performance.
3414 Details : Set the value of /sys/block/{dev}/queue/max_sectors_kb
3415 to the value of /sys/block/{dev}/queue/max_hw_sectors_kb
3419 Frequency : Always for filesystems larger than 2TB on 32-bit systems.
3420 Bugzilla : 13547, 13627
3421 Description: Data corruption for OSTs that are formatted larger than 2TB
3423 Details : When generating the bio request for lustre file writes the
3424 sector number would overflow a temporary variable before being
3425 used for the IO. The data reads correctly from Lustre (which
3426 will overflow in a similar manner) but other file data or
3427 filesystem metadata may be corrupted in some cases.
3431 Description: "lfs find -obd UUID" prints directories
3432 Details : "lfs find -obd UUID" will return all directory names instead
3433 of just file names. It is incorrect because the directories
3434 do not reside on the OSTs.
3437 Bugzilla : 13596, 13614
3438 Description: MDS hang after unclean shutdown of lots of clients
3439 Details : Never resend AST requests.
3443 Frequency : Always, for kernels after 2.6.16
3444 Description: Fix warning idr_remove called for id=.. which is not allocated.
3445 Details : Last kernels save old s_dev before kill super and not allow
3446 to restore from callback - restore it before call kill_anon_super.
3450 Description: Fix error on 'ls .' at the top of the Lustre mount.
3451 Details : Don't revalidate dentry if it is a root dentry.
3453 --------------------------------------------------------------------------------
3455 2007-08-27 Cluster File Systems, Inc. <info@clusterfs.com>
3457 * Support for kernels:
3458 2.6.5-7.286 (SLES 9),
3459 2.6.9-55.0.2.EL (RHEL 4),
3460 2.6.16.46-0.14 (SLES 10),
3461 2.6.18-8.1.8.el5 (RHEL 5),
3462 2.6.18.8 vanilla (kernel.org)
3463 * Client support for unpatched kernels:
3464 (see http://wiki.lustre.org/index.php?title=Patchless_Client)
3465 2.6.16 - 2.6.22 vanilla (kernel.org)
3466 * Due to recently discovered recovery problems, we do not recommend
3467 using patchless RHEL 4 clients with this or any earlier release.
3468 * Recommended e2fsprogs version: 1.39.cfs8
3469 * Note that reiserfs quotas are disabled on SLES 10 in this kernel.
3474 Description: block reactivating mgc import until all deactivates complete
3475 Details : Fix race when failing back MDT/MGS to itself (testing)
3477 Severity : enhancement
3479 Description: lfs setstripe enhancement
3480 Details : Make lfs setstripe understand 'k', 'm' and 'g' for stripe size.
3483 Frequency : mds/oss recovery
3485 Description: llog ctxt is refrenced after it has been freed.
3486 Details : llog ctxt refcount was added to avoide the race between ctxt free
3487 and llog recovery process. Each llog user must hold ctxt refcount
3488 before it access the llog. And the llog ctxt can only be freed
3489 when its refcount is zero.
3491 Severity : enhancement
3493 Description: randomly memory allocation failure util
3494 Details : Make lustre randomly failed allocating memory for testing purpose.
3497 Frequency : only for SLES10
3499 Description: Update kernel patch for SLES10 SP1
3500 Details : Add patch blkdev_tunables-2.6-sles10.patch to 2.6-sles10.series.
3502 Severity : enhancement
3504 Description: omit set fsid for export NFS
3505 Details : fix set/restore device id for avoid EMFILE error and mark lustre fs
3506 as FS_REQUIRES_DEV for avoid problems with generate fsid.
3509 Frequency : after network failures
3511 Description: Add sync option to mount_lustre.c
3512 Details : Client loses data written to lustre after a network interruption.
3514 Severity : enhancement
3516 Description: Error message improvement.
3517 Details : Merging of two LCONSOLE_ERROR_MSG into one.
3519 Severity : enhancement
3521 Description: don't use GFP_* in generic Lustre code.
3522 Details : Use cfs_alloc_* functions and CFS_* flags for code portability.
3524 Severity : enhancement
3526 Description: obdclass is limited by single OBD_ALLOC(idarray)
3527 Details : replace OBD_ALLOC/OBD_FREE with OBD_VMALLOC/OBD_VFREE
3529 Severity : enhancement
3531 Description: updated patchess for new RHEL4 kernel
3532 Details : Fixed ext3-unlink-race.patch per Kalpak's comment.
3534 Severity : enhancement
3536 Description: warnings with build patchless client with vanila 2.6.19 and up
3537 Details : change old ctl_table style and replace ctl_table/ctl_table_header
3538 with cfs_sysctl_table_t/cfs_sysctl_table_header_t
3540 Severity : enhancement
3542 Description: O_DIRECT bypasses client statistics.
3543 Details : When running with O_DIRECT I/O, neither the client rpc_stats nor
3544 read_ahead_stats were updated. Copied stats section from
3545 osc_send_oap_rpc() into async_internal().
3548 Frequency : only for Cray XT3
3550 Description: peer credits not enough on many OST per OSS systems.
3551 Details : Use new lnet way to add credits as we need those for pings and ASTs
3554 Frequency : only with liblustre
3556 Description: Liblustre is not releasing flock locks on file close.
3557 Details : Release flock locks on file close.
3560 Frequency : only for RHEL4
3562 Description: Update kernel patches for kernel-2.6.9-55.0.2.EL
3563 Details : Remove inode-nr_unused-2.6.9-rhel4.patch from 2.6-rhel4.series
3564 Update target file and kernel config.
3568 Description: lustre support for RHEL5
3569 Details : Add support for RHEL5.
3574 Description: ASSERTION(export != NULL) failed in target_handle_connect
3575 Details : Assetion hit is result of rare race between disconnect and connet
3576 to same nid. target_handle_connect found old connect cockie and
3577 tried to reconnect, but can't find export for this cockie.
3582 Description: umount blocks forever on error
3583 Details : In result of wrong using obd_no_recov and obd_force flags client
3584 can hand if cancel or some other requests is lost.
3587 Frequency : Only for SLES
3589 Description: sanity_quota fail test_1
3590 Details : There are multiple occurences of $TSTUSR in SLES's /etc/group
3591 file, which makes TSTID[2] inunique.
3593 Severity : enhancement
3595 Description: Kernel patches for SLES9 2.6.5-7.286 kernel
3596 Details : Update target/ChangeLog/which_patch .
3598 Severity : enhancement
3600 Description: jbd statistics
3601 Details : Port older jbd statistics patch for sles10
3606 Description: lvbo_init failed for resource with missing objects.
3607 Details : Fix returning error if we do stat for file with missing/corrupted
3608 objects and i_size set to all sum of size of all avaible objects.
3609 if we to truncate/write to missing object - it is recreated.
3614 Description: Oops in read and write path when failing to allocate lock.
3615 Details : Check if lock allocation failed and return error back.
3618 Frequency : When flocks are used.
3620 Description: assertion failure in ldlm_cli_enquque_fini for non NULL lock.
3621 Details : Flock locks might destroy just granted lock if it could be merged
3622 with another existing flock, this is done in completion handler,
3623 so teach ldlm_cli_enquque_fini that this is a valid case for
3629 Description: reply_lock_interpret crash due to race with it and lock cancel.
3630 Details : Do not replay locks that are being cancelled. Do not reference
3631 locks by their address during replay, just by their handle.
3634 Frequency : only with deactivated OSTs
3636 Description: lstripe command fails for valid OST index
3637 Details : The stripe offset is compared to 'lov->desc.ld_tgt_count'
3638 instead of lov->desc.ld_active_tgt_count.
3640 Severity : enhancement
3642 Description: Build failure against Centos5 (RHEL5)
3643 Details : Use getpagesize() instead of PAGE_SIZE.
3645 --------------------------------------------------------------------------------
3647 2007-07-30 Cluster File Systems, Inc. <info@clusterfs.com>
3649 * Support for kernels:
3650 2.6.5-7.283 (SLES 9),
3651 2.6.9-55.EL (RHEL 4),
3652 2.6.16.46-0.14 (SLES 10),
3653 2.6.18.8 vanilla (kernel.org)
3654 * Client support for unpatched kernels:
3655 (see http://wiki.lustre.org/index.php?title=Patchless_Client)
3656 2.6.16 - 2.6.22 vanilla (kernel.org)
3657 * Due to recently discovered recovery problems, we do not recommend
3658 using patchless RHEL 4 clients with this or any earlier release.
3659 * Recommended e2fsprogs version: 1.39.cfs8
3660 * Note that reiserfs quotas are disabled on SLES 10 in this kernel.
3661 * Starting with this release, the ldiskfs backing filesystem required
3662 by Lustre is now in its own package, lustre-ldiskfs. This package
3663 should be installed. It is versioned separately from Lustre and
3664 may be released separately in future.
3669 Description: server LBUG when shutting down
3670 Details : Block umount forever until the mount refcount is zero rather
3671 than giving up after an arbitrary timeout.
3673 Severity : enhancement
3675 Description: add optional extra BUILD_VERSION info
3676 Details : add a new environment variable (namely LUSTRE_VERS) which allows
3677 to override the lustre version.
3680 Frequency : 2.6.18 servers only
3682 Description: ll_kern_mount() doesn't release the module reference
3683 Details : The ldiskfs module reference count never drops down to 0
3684 because ll_kern_mount() doesn't release the module reference.
3689 Description: server LBUG when using old ost_num_threads parameter
3690 Details : Accept the old ost_num_threads parameter but warn that it
3691 is deprecated, and fix an off-by-one error that caused an LBUG.
3696 Description: Transient SCSI error results in persistent IO issue
3697 Details : iobuf->dr_error is not reinitialized to 0 between two
3701 Frequency : sometimes when underlying device returns I/O errors
3703 Description: OSTs not going read-only during write failures
3704 Details : OSTs are not remounted read-only when the journal commit threads
3705 get I/O errors because fsfilt_ext3 calls journal_start/stop()
3706 instead of the ext3 wrappers.
3709 Frequency : only with 10000 clients or more
3711 Description: poor connect scaling with increasing client count
3712 Details : Don't run filter_grant_sanity_check for more than 100 exports
3713 to improve scaling for large numbers of clients.
3716 Frequency : SLES10 only
3718 Description: sanity-quota.sh quotacheck failed: rc = -22
3719 Details : Quotas cannot be enabled on SLES10.
3722 Frequency : liblustre clients only
3724 Description: getdirentries does not give error when run on compute nodes
3725 Details : getdirentries does not fail when the size specified as an argument
3726 is too small to contain at least one entry
3728 Severity : enhancement
3730 Description: Add LNET router traceability for debug purposes
3731 Details : If a checksum failure occurs with a router as part of the
3732 IO path, the NID of the last router that forwarded the bulk data
3733 is printed so it can be identified.
3738 Description: OST "spontaneously" evicts client; client has imp_pingable == 0
3739 Details : Due to a race condition, liblustre clients were occasionally
3740 evicted incorrectly.
3742 Severity : enhancement
3744 Description: lfs setstripe use optional parameters instead of postional
3747 Severity : enhancement
3749 Description: Nanosecond timestamp support for ldiskfs
3750 Details : The on-disk ldiskfs filesystem has added support for nanosecond
3751 resolution timestamps. There is not yet support for this at
3752 the Lustre filesystem level.
3755 Frequency : during server recovery
3757 Description: MDS failing to send precreate requests due to OSCC_FLAG_RECOVERING
3758 Details : request with rq_no_resend flag not awake l_wait_event if they get a
3762 Frequency : nfs export on patchless client
3764 Description: connectathon hang when test nfs export over patchless client
3765 Details : Disconnected dentry cannot be found with lookup, so we do not need
3766 to unhash it or make it invalid
3769 Description: fix llapi_lov_get_uuids() to allow many OSTs to be returned
3770 Details: : Change llapi_lov_get_uuids() to read the UUIDs from /proc instead
3771 of using an ioctl. This allows lfsck for > 160 OSTs to succeed.
3776 Description: open req refcounting wrong on reconnect
3777 Details : If reconnect happened between getting open reply from server and
3778 call to mdc_set_replay_data in ll_file_open, we will schedule
3779 replay for unreferenced request that we are about to free.
3780 Subsequent close will crash in variety of ways.
3781 Check that request is still eligible for replay in
3782 mdc_set_replay_data().
3787 Description: disable writes to filesystem when reading health_check file
3788 Details : the default for reading the health_check proc file has changed
3789 to NOT do a journal transaction and write to disk, because this
3790 can cause reads of the /proc file to hang and block HA state
3791 checking on a healthy but otherwise heavily loaded system. It
3792 is possible to return to the previous behaviour during configure
3793 with --enable-health-write.
3795 Severity : enhancement
3797 Description: 64-bit inode version
3798 Details: : Add a on-disk 64-bit inode version for ext3 to track changes made
3799 to the inode. This will be required for version-based recovery.
3804 Description: MDS fails to start if a duplicate client export is detected
3805 Details : in some rare cases it was possible for a client to connect to
3806 an MDS multiple times. Upon recovery the MDS would detect this
3807 and fail during startup. Handle this more gracefully.
3809 Severity : enhancement
3811 Description: Add -o localflock option to simulate old noflock behaviour.
3812 Details : This will achieve local-only flock/fcntl locks coherentness.
3817 Description: log_commit_thread vs filter_destroy race leads to crash
3818 Details : Take import reference before releasing llog record semaphore
3823 Description: Wrong request locking in request set processing
3824 Details : ptlrpc_check_set wrongly uses req->rq_lock for proctect add to
3825 imp_delayed_list, in this place should be used imp_lock.
3828 Frequency : when reconnecting
3830 Description: Grant leak when OSC reconnect to OST
3831 Details : When osc reconnect ost, OST(filter) should check whether it
3832 should grant more space to client by comparing fed_grant and
3833 cl_avail_grant, and return the granted space to client instead
3834 of "new granted" space, because client will call osc_init_grant
3835 to update the client grant space info.
3838 Frequency : when client reconnects to OST
3840 Description: Grant leak when OSC does a resend and replays bulk write
3841 Details : When osc reconnect to OST, OST(filter) should clear grant info of
3842 bulk write request, because the grant info while be sync between
3843 OSC and OST when reconnect, and we should ignore the grant info
3844 these of resend/replay write req.
3849 Description: Grant space more than avaiable space sometimes.
3850 Details : When then OST is about to be full, if two bulk writing from
3851 different clients came to OST. Accord the avaliable space of the
3852 OST, the first req should be permitted, and the second one
3853 should be denied by ENOSPC. But if the seconde arrived before
3854 the first one is commited. The OST might wrongly permit second
3855 writing, which will cause grant space > avaiable space.
3858 Frequency : when client is evicted
3860 Description: Grant might be wrongly erased when osc is evicted by OST
3861 Details : when the import is evicted by server, it will fork another
3862 thread ptlrpc_invalidate_import_thread to invalidate the
3863 import, where the grant will be set to 0. While the original
3864 thread will update the grant it got when connecting. So if
3865 the former happened latter, the grant will be wrongly errased
3866 because of this race.
3871 Description: Checking Stale with correct fid
3872 Details : ll_revalidate_it should uses de_inode instead of op_data.fid2
3873 to check whether it is stale, because sometimes, we want the
3874 enqueue happened anyway, and op_data.fid2 will not be initialized.
3876 Severity : enhancement
3878 Description: update patchless client
3879 Details : Add support for patchless client with 2.6.20, 2.6.21 and RHEL 5
3882 Frequency : only with 2.4 kernel
3884 Description: random memory corruption
3885 Details : size of struct ll_inode_info is to big for union inode.u and this
3886 can be cause of random memory corruption.
3891 Description: Memory leak in recovery
3892 Details : Lov_mds_md was not free in an error handler in mds_create_object.
3893 It should also check obd_fail before fsfilt_start, otherwise if
3894 fsfilt_start return -EROFS,(failover mds during mds recovery).
3895 then the req will return with repmsg->transno = 0 and rc = EROFS.
3896 and we met hit the assert LASSERT(req->rq_reqmsg->transno ==
3897 req->rq_repmsg->transno) in ptlrpc_replay_interpret. Fcc should
3898 be freed no matter whether fsfilt_commit success or not.
3901 Frequency : only with large numbers of cores on a single node
3903 Description: Prevents from taking the superblock lock in llap_from_page for
3905 Details : using LL_ORIGIN_REMOVEPAGE origin flag instead of LL_ORIGIN_UNKNOW
3906 for llap_from_page call in ll_removepage() prevents from taking
3907 the superblock lock for a soon died page.
3912 Description: Not check open intent error before release open handle
3913 Details : in some rare cases, the open intent error is not checked before
3914 release open handle, which may cause
3915 ASSERTION(open_req->rq_transno != 0), because it tries to release
3916 the failed open handle.
3921 Description: Set cat log bitmap only after create log success.
3922 Details : in some rare cases, the cat log bitmap is set too early. and it
3923 should be set only after create log success.
3927 Description: Accessing a block bevice can re-enable I/O when Lustre is
3928 tearing down a device.
3929 Details : dev_clear_rdonly(bdev) must be called in kill_bdev() instead of
3934 Description: service threads may hog cpus when there are a lot of requests
3935 Details : Insert cond_resched to give other threads a chance to use some CPU
3940 Description: the cat log was not initialized in recovery
3941 Details : When mds(mgs) do recovery, the tgt_count might be zero, so the
3942 unlink log on mds will not be initialized until mds post
3943 recovery. And also in mds post recovery, the unlink log will
3944 initialization will be done asynchronausly, so there will be race
3945 between add unlink log and unlink log initialization.
3949 Description: brw_stats were being printed incorrectly
3950 Details : brw_stats were being printed as log2 but all of them were not
3951 recorded as log2. Also remove some code duplication arising from
3952 filter_tally_{read,write}.
3956 Frequency : rare, only in recovery.
3957 Description: ASSERTION(req->rq_type != LI_POISON) failed
3958 Details : imp_lock should be held while iterating over imp_sending_list for
3959 prevent destroy request after get timeout in ptlrpc_queue_wait.
3963 Description: replay-single.sh test 52 fails
3964 Details : A lock's skiplist need to be cleanup when it being unlinked
3965 from its resource list.
3970 Description: Short directio read returns full requested size rather than
3972 Details : Direct I/O operations should return actual amount of bytes
3973 transferred rather than requested size.
3975 Severity : enhancement
3977 Description: metadata RPC reduction (e.g. for rm performance)
3978 Details : decrease the amount of synchronous RPC between clients and servers
3979 by canceling conflicing lock before the operation on the client
3980 and packing thier handles into the main operation RPC to server.
3982 Severity : enhancement
3984 Description: add #ifdef HAVE_KERNEL_CONFIG_H
3985 Details : kernels from 2.6.19 not need include linux/config.h, but add
3986 include linux/autoconf.h in commpiler command line.
3988 Severity : enhancement
3990 Description: patchless client support for 2.6.22 kernel
3991 Details : 2.6.22 has only one visble change, SLAB_CTOR_* constants is
3992 removed. In this case we need drop using os depended interface to
3993 kmem_cache and use cfs_mem_cache API.
3998 Description: fix mal-formatted messages
3999 Details : fix some mal-formatted DEBUG_REQ and LCONSOLE_ERROR_MSG messages
4003 Frequency : always in liblustre
4004 Description: wrong IS_ERR implementation in liblustre.h
4005 Details : fix IS_ERR implementation in liblustre.h for right detect errors.
4010 Description: Correct condition for output debug message.
4011 Details : inode i_nlink equal zero is not enough for output message about
4012 disk corruption, i_ctime and i_mode should be also checked.
4016 Frequency : always in patchless client
4017 Description: add configure check for truncate_complete_page
4018 Details : improve checks for exported symbols. This allow run check without
4019 sources, but with Module.symvers shipped with kernel distribution.
4020 add check for truncate_complete_page used by patchless client.
4024 Description: sanity.sh test_77h fails with "test_77h file compare failed"
4025 Details : test_77h uses a file which was messed by other test case.
4029 Description: Not Check whether lov_tgts is NULL in some lov functions
4030 Details : Checking whether lov_tgts is NULL in some functions.
4034 Description: replace obdo_alloc() with OBDO_ALLOC macro
4035 Details : nothing special is done in obdo_alloc() function, and for
4036 debugging purpose, it needs to be replaced with macros.
4040 Description: bad return value and errno from fcntl call
4041 Details : In liblustre API, errno should be a negative value if error
4046 Description: ptlrpc_check_set() LBUG
4047 Details : In case of positive reply from server and failed client bulk
4048 callback after bulk transfer shouldn't LBUG, but process this
4049 request as erroneous.
4051 Severity : enhancement
4053 Description: VFS operations stats tool.
4054 Details : Tool which collects stats by tracking value written in pid,
4055 ppid, gid and uses llstat to generate output to plot graph using
4057 Updated lustre/utils/Makefile.am
4058 Added lustre/utils/ltrack_stats.c
4060 Severity : enhancement
4062 Description: 2.6.18 server support (lustre 1.6.1)
4063 Details : Support for 2.6.18 kernels on the server side.
4068 Description: ASSERTION(imp->imp_conn_current) failed
4069 Details : an assertion failure is hit if a client node boots and attempts to
4070 mount a lustre filesystem faster than RECONNECT_INTERVAL seconds.
4073 Frequency : only for i686
4075 Description: 1.4.11 RC1 build fails for RHEL 4, i686
4076 Details : Fixed config variable for build.
4081 Description: Updated patchess for new RHEL4 kernel
4082 Details : Updated patch inode-nr_unused-2.6.9-rhel4.patch
4083 Updated patch jbd-stats-2.6.9.patch
4084 Updated patch qsnet-rhel4-2.6.patch
4085 Updated patch quota-deadlock-on-pagelock-core.patch
4086 Updated patch vfs_intent-2.6-rhel4.patch
4087 Updated patch vfs_races-2.6-rhel4.patch
4088 Updated series file 2.6-rhel4-titech.series
4089 Updated series file 2.6-rhel4.series
4090 Updated kernel config files
4095 Description: lquota slave complains LBUG when reconnecting with mds
4097 Details : quota slave depends qctxt->lqc_import to send its quota request.
4098 This pointer will be invalid if mds did failover or broke its
4099 connect to osts, which leads to LBUG.
4102 Frequency : when qunit size is too small(less than 20M)
4104 Description: write is stopped by improper -EDQUOT
4105 Details : If the master is busy and qunit size is small enough(let's say 1M),
4106 the slave can not get quota from master on time, which will lead
4107 slave to trigger a -EQUOTA to client.
4112 Description: Deadlock during metadata tests
4113 Details : in prune_dir_dentries(), shrink_dcache_parent() should not be
4114 called with the per-dentry lock held.
4117 Frequency : SLES9 only
4119 Description: Lustre patched kernel for SLES9 SP3 has NR_CPUS set to 8
4120 Details : set CONFIG_NR_CPUS to 128 instead of 8.
4122 Severity : enhancement
4124 Description: remove fs_prep_san_write operation and related patches
4125 Details : remove the ext3-san-jdike patches which are no longer useful.
4130 Description: LDISKFS-fs error (device sdc): ldiskfs_free_blocks
4131 Details : a disk corruption can cause the mballoc code to assert on a
4132 double free or other extent corruptions. Handle these with
4133 ext3_error() instead of assertions.
4136 Frequency : only with mballoc3 code and deep extent trees
4138 Description: ldiskfs_ext_search_right: bad header in inode: unexpected eh_depth
4139 Details : a wrong check of extent headers in ldiskfs_ext_search_right()
4140 can cause the filesystem to be remounted read-only.
4145 Description: lfsck built against 1.4.x cannot run against 1.6.0 lustre
4146 Details : the definition for OBD_IOC_GETNAME changed in 1.6.0. One of the
4147 few external users of this ioctl number is lfsck's call to
4148 llapi_lov_get_uuids() and this caused lfsck to fail at startup.
4149 Add the old ioctl number to the handler so both old and new
4154 Description: parallel lock callbacks
4155 Details : Instead of sending blocking and completion callbacks as separated
4156 requests, adding them to a set and sending in parallel.
4160 Description: Disable most debugging by default
4161 Details : To improve performance, disable most logging (for debug purposes)
4162 by default. VFSTRACE, RPCTRACE, and DLMTRACE are now off by
4163 default, and HA includes fewer messages.
4167 Frequency : only run on patchless client.
4168 Description: use do_facet on sanity.sh for test handling recoverables errors
4169 Details : use do_facet instead of direct use sysctl for set fail_loc on OST
4173 Description: hash tables for lists of nids, connections and uuids
4174 Details : Hash tables noticeably help when a lot of clients connect to a
4175 server, to faster identify duplicate connections or reconnects,
4176 also to faster find export to evict in manual eviction case.
4180 Description: Sometimes, when the server evict a client, and the client will
4181 not be evicted as soon as possible.
4182 Details : In enqueue req, the error was returned by intent, instead of
4183 rq_status which make ptlrpc layer not detect this error, and
4184 does not evict the client. So enqueue error should be returned
4188 Frequency : only at startup
4190 Description: Delay client connections to MDT until fist MDT->OST connect
4191 Details : If a client tried to create a new file before the MDT had
4192 connected to any OSTs, the create would return EIO. Now
4193 the client will simply block until the MDT connects to the
4194 first OST and the create can succeed.
4197 Frequency : only for SLES9
4199 Description: Routinely utilize latest Quadrics drivers in CFS releases
4200 Details : Update patch qsnet-suse-2.6.patch.
4203 Frequency : only for sles10
4205 Description: Update patches for SLES 10 SP1 kernel.
4206 Details : Update the patch vfs_intent-2.6-sles10.patch.
4211 Description: Routinely utilize latest Quadrics drivers in CFS releases
4212 Details : Update patch qsnet-rhel4-2.6.patch.
4215 Frequency : at statup only
4217 Description: mds_lov_synchronize race leads to various problems
4218 Details : simultaneous MDT->OST connections at startup can cause the
4219 sync to abort, leaving the OSC in a bad state.
4224 Description: Using wrong pointer in osc_brw_prep_request
4225 Details : Access to array[-1] can produce panic if kernel compiled with
4226 CONFIG_PAGE_ALLOC enabled
4228 Severity : enhancement
4230 Description: Async OSC create to avoid the blocking unnecessarily.
4231 Details : If a OST has no remain object, system will block on the creating
4232 when need to create a new object on this OST. Now, ways use
4233 pre-created objects when available, instead of blocking on an
4234 empty osc while others are not empty. If we must block, we block
4235 for the shortest possible period of time.
4239 Frequency : only in recovery
4240 Description: Mark OST as early accessible if his start SYNC.
4241 Details : osc_precreate return flag early accessible if oscc marked as
4242 OSCC_FLAG_SYNC_IN_PROGRESS.
4247 Description: Sometimes precreate code can triger create object on wrong ost
4248 Details : Wrong protected or not not restored variables aftre precreate loop
4249 can produce creation object on wrong ost.
4252 Frequency : oss recovery
4254 Description: llog_commit_thread cleanup should sync with llog_commit_thread
4256 Details : llog_commit_thread_count should be synced between llog_commit
4257 start and cleanup, so new llog_commit thread should not be started
4258 when llog_commit threads being stopped to avoid accessing some
4261 Severity : enhancement
4263 Description: Add printing inode info into message about error in writepage.
4265 --------------------------------------------------------------------------------
4267 2007-05-03 Cluster File Systems, Inc. <info@clusterfs.com>
4272 Frequency : on some architectures
4274 Description: 1.6 client sometimes fails to mount from a 1.4 MDT
4275 Details : Uninitialized flags sometimes cause configuration commands to
4279 Frequency : patchless clients only
4281 Description: missing __iget() symbol export
4282 Details : The __iget() symbol export is missing. To avoid the need for
4283 this on patchless clients the deathrow inode reaper is turned
4284 off, and we depend on the VM to clean up old inodes. This
4285 dependency was during via the fix for bug 12181.
4290 Description: sanity.sh fail: test_52b
4291 Details : The ll_inode_to_ext_flags() has a glitch which makes MDS return
4292 incorrect inode's flags to client.
4294 --------------------------------------------------------------------------------
4296 2007-04-19 Cluster File Systems, Inc. <info@clusterfs.com>
4298 * CONFIGURATION CHANGE. This version of Lustre WILL NOT
4299 INTEROPERATE with older versions automatically. In many cases a
4300 special upgrade step is needed. Please read the
4301 user documentation before upgrading any part of a 1.4.x system.
4302 * WARNING: Lustre configuration and startup changes are required with
4303 this release. See https://mail.clusterfs.com/wikis/lustre/MountConf
4305 * Support for kernels:
4306 2.4.21-47.0.1.EL (RHEL 3),
4307 2.6.5-7.283 (SLES 9),
4308 2.6.9-42.0.10.EL (RHEL 4),
4309 2.6.12.6 vanilla (kernel.org),
4310 2.6.16.27-0.9 (SLES10)
4311 * Client support for unpatched kernels:
4312 (see https://mail.clusterfs.com/wikis/lustre/PatchlessClient)
4313 2.6.16 - 2.6.19 vanilla (kernel.org),
4314 2.6.9-42.0.8EL (RHEL 4)
4315 * Recommended e2fsprogs version: 1.39.cfs6
4316 * Note that reiserfs quotas are disabled on SLES 10 in this kernel
4319 Severity : enhancement
4321 Description: Async OSC create to avoid the blocking unnecessarily.
4322 Details : If a OST has no remain object, system will block on the creating
4323 when need to create a new object on this OST. Now, ways use
4324 pre-created objects when available, instead of blocking on an
4325 empty osc while others are not empty. If we must block, we block
4326 for the shortest possible period of time.
4328 Severity : enhancement
4330 Description: MountConf
4331 Details : Lustre configuration is now managed via mkfs and mount
4332 commands instead of lmc and lconf. New obd types (MGS, MGC)
4333 are added for dynamic configuration management. See
4334 https://mail.clusterfs.com/wikis/lustre/MountConf for
4337 Severity : enhancement
4339 Description: dynamic OST addition
4340 Details : OSTs can now be added to a live filesystem
4342 Severity : enhancement
4344 Description: startup order invariance
4345 Details : MDTs and OSTs can be started in any order. Clients only
4346 require the MDT to complete startup.
4348 Severity : enhancement
4350 Description: parallel, asynchronous orphan cleanup
4351 Details : orphan cleanup is now performed in separate threads for each
4352 OST, allowing parallel non-blocking operation.
4354 Severity : enhancement
4356 Description: optimized stripe assignment
4357 Details : stripe assignments are now made based on ost space available,
4358 ost previous usage, and OSS previous usage, in order to try
4359 to optimize storage space and networking resources.
4361 Severity : enhancement
4363 Description: Permanently set tunables
4364 Details : All writable /proc/fs/lustre tunables can now be permanently
4365 set on a per-server basis, at mkfs time or on a live
4368 Severity : enhancement
4370 Description: Lustre message v2
4371 Details : Add lustre message format v2.
4373 Severity : enhancement
4375 Description: client OST exclusion list
4376 Details : Clients can be started with a list of OSTs that should be
4377 declared "inactive" for known non-responsive OSTs.
4381 Description: ENOENT returned for valid filehandle during dbench.
4382 Details : Check if a directory has children when invalidating dentries
4383 associated with an inode during lock cancellation. This fixes
4384 an incorrect ENOENT sometimes seen for valid filehandles during
4385 testing with dbench.
4388 Frequency : SFS test only (otherwise harmless)
4390 Description: SPEC SFS validation failure on NFS v2 over lustre.
4391 Details : Changes the blocksize for regular files to be 2x RPC size,
4392 and not depend on stripe size.
4394 Severity : enhancement
4396 Description: fine-grained SMP locking inside DLM
4397 Details : Improve DLM performance on SMP systems by removing the single
4398 per-namespace lock and replace it with per-resource locks.
4400 Severity : enhancement
4402 Description: don't hold multiple extent locks at one time
4403 Details : To avoid client eviction during large writes, locks are not
4404 held on multiple stripes at one time or for very large writes.
4405 Otherwise, clients can block waiting for a lock on a failed OST
4406 while holding locks on other OSTs and be evicted.
4408 Severity : enhancement
4410 Description: Multiple MD RPCs in flight.
4411 Details : Further unserialise some read-only MDT RPCs - learn about intents.
4412 To avoid overly-overloading MDT, introduce a limit on number of
4413 MDT RPCs in flight for a single client and add /proc controls
4414 to adjust this limit.
4416 Severity : enhancement
4418 Description: client read/write statistics
4419 Details : Add client read/write call usage stats for performance
4420 analysis of user processes.
4421 /proc/fs/lustre/llite/*/offset_stats shows non-sequential
4422 file access. extents_stats shows chunk size distribution.
4423 extents_stats_per_process show chunk size distribution per
4426 Severity : enhancement
4428 Description: per-client statistics on server
4429 Details : Add ldlm and operations statistics for each client in
4430 /proc/fs/lustre/mds|obdfilter/*/exports/
4432 Severity : enhancement
4434 Description: improved MDT statistics
4435 Details : Add detailed MDT operations statistics in
4436 /proc/fs/lustre/mds/*/stats
4438 Severity : enhancement
4440 Description: VFS operations stats
4441 Details : Add client VFS call stats, trackable by pid, ppid, or gid
4442 /proc/fs/lustre/llite/*/stats_track_[pid|ppid|gid]
4447 Description: Fix client-side osc byte counters
4448 Details : The osc read/write byte counters in
4449 /proc/fs/lustre/osc/*/stats are now working
4452 Frequency : always as root on SLES
4454 Description: Failure of copying files with lustre special EAs.
4455 Details : Client side always return success for setxattr call for lustre
4456 special xattr (currently only "trusted.lov").
4461 Description: Refcount LNET uuids
4462 Details : The global LNET uuid list grew linearly with every startup;
4463 refcount repeated list entries instead of always adding to
4466 Severity : enhancement
4468 Description: Dynamic service threads
4469 Details : Within a small range, start extra service threads
4470 automatically when the request queue builds up.
4473 Frequency : mixed-endian client/server environments
4475 Description: mixed-endian crashes
4476 Details : The new msg_v2 system had some failures in mixed-endian
4479 Severity : enhancement
4481 Description: Easy OST removal
4482 Details : OSTs can be permanently deactivated with e.g. 'lctl
4483 conf_param lustre-OST0001.osc.active=0'
4485 Severity : enhancement
4487 Description: MGS proc entries
4488 Details : Added basic proc entries for the MGS showing what filesystems
4491 Severity : enhancement
4493 Description: provide MGS failover
4494 Details : Added config lock reacquisition after MGS server failover.
4496 Severity : enhancement
4498 Description: add Linux 2.4 support
4499 Details : Added support for RHEL 2.4.21 kernel for 1.6 servers and clients
4503 Description: a large application tries to do I/O to the same resource and dies
4504 in the middle of it.
4505 Details : Check the req->rq_arrival time after the call to
4506 ost_brw_lock_get(), but before we do anything about
4507 processing it & sending the BULK transfer request. This
4508 should help move old stale pending locks off the queue as
4509 quickly as obd_timeout.
4512 Frequency : when an incorrect nid is specified during startup
4514 Description: ptlrpc connect to non-existant node causes kernel crash
4515 Details : LNET can't be re-entered from an event callback, which
4516 happened when we expire a message after the export has been
4517 cleaned up. Instead, hand the zombie cleanup off to another
4520 Severity : enhancement
4522 Description: plain/inodebits lock performance improvement
4523 Details : Grouping plain/inodebits in granted list by their request modes
4524 and bits policy, thus improving the performance of search through
4528 Frequency : only if OST filesystem is corrupted
4530 Description: client incorrectly hits assertion in ptlrpc_replay_req()
4531 Details : for a short time RPCs with bulk IO are in the replay list,
4532 but replay of bulk IOs is unimplemented. If the OST filesystem
4533 is corrupted due to disk cache incoherency and then replay is
4534 started it is possible to trip an assertion. Avoid putting
4535 committed RPCs into the replay list at all to avoid this issue.
4538 Frequency : liblustre (e.g. catamount) on a large cluster with >= 8 OSTs/OSS
4540 Description: System hang on startup
4541 Details : This bug allowed the liblustre (e.g. catamount) client to
4542 return to the app before handling all startup RPCs. This
4543 could leave the node unresponsive to lustre network traffic
4544 and manifested as a server ptllnd timeout.
4546 Severity : enhancement
4548 Description: Add "/proc/sys/lustre/debug_peer_on_timeout"
4549 Details : liblustre envirable: LIBLUSTRE_DEBUG_PEER_ON_TIMEOUT
4550 boolean to control whether to print peer debug info when a
4551 client's RPC times out.
4554 Frequency : only for kernels with patches from Lustre below 1.4.3
4556 Description: Remove old rdonly API
4557 Details : Remove old rdonly API which unused from at least lustre 1.4.3
4560 Frequency : only for devices with external journals
4562 Description: Set external device read-only also
4563 Details : During a commanded failover stop, we set the disk device
4564 read-only while the server shuts down. We now also set any
4565 external journal device read-only at the same time.
4568 Frequency : when upgrading from 1.4 while trying to change parameters
4570 Description: The wrong (new) MDC name was used when setting parameters for
4571 upgraded MDT's. Also allows changing of OSC (and MDC)
4572 parameters if --writeconf is specified at tunefs upgrade time.
4575 Frequency : when setting specific ost indicies
4577 Description: QOS code breaks on skipped indicies
4578 Details : Add checks for missing OST indicies in the QOS code, so OSTs
4579 created with --index need not be sequential.
4581 Severity : enhancement
4583 Description: Add uninit_groups feature to ldiskfs2 to speed up e2fsck
4584 Details : The uninit_groups feature works in conjunction with the kernel
4585 filesystem code (ldiskfs2 only) and e2fsprogs-1.39-cfs6 to speed
4586 up the pass1 processing of e2fsck. This is a read-only feature
4587 in ldiskfs2 only, so older kernels and current ldiskfs cannot
4588 mount filesystems that have had this feature enabled.
4590 Severity : enhancement
4592 Description: Improve multi-block allocation algorithm to avoid fragmentation
4593 Details : The mballoc3 code (ldiskfs2 only) adds new mechanisms to improve
4594 allocation locality and avoid filesystem fragmentation.
4596 ------------------------------------------------------------------------------
4598 2007-04-01 Cluster File Systems, Inc. <info@clusterfs.com>
4600 * Support for kernels:
4601 2.4.21-47.0.1.EL (RHEL 3)
4602 2.6.5-7.283 (SLES 9)
4603 2.6.9-42.0.10.EL (RHEL 4)
4604 2.6.12.6 vanilla (kernel.org)
4605 2.6.16.27-0.9 (SLES 10)
4606 * Recommended e2fsprogs version: 1.39.cfs5
4608 * Note that reiserfs quotas are disabled on SLES 10 in this kernel
4612 Frequency : occasional, depends on client load and configuration
4613 Bugzilla : 12181, 12203
4614 Description: data loss for recently-modified files
4616 Details : In some cases it is possible that recently written or created
4617 files may not be written to disk in a timely manner (this should
4618 normally be within 30s unless client IO load is very high).
4619 The problem appears as zero-length files or files that are a
4620 multiple of 1MB in size after a client crash or client eviction
4621 that are missing data at the end of the file.
4623 This problem is more likely to be hit on clients where files are
4624 repeatedly created and unlinked in the same directory, clients
4625 have a large amount of RAM, have many CPUs, the filesystem has
4626 many OSTs, the clients are rebooted frequently, and/or the files
4627 are not accessed by other nodes after being written.
4629 The presence of the problem can be detected by looking at
4630 /proc/sys/fs/inode-state. If the first number (nr_inodes) is
4631 smaller than the second (nr_unused) then dirty files will not
4632 be flushed automatically to disk. "sync; sleep 10" should be
4633 run several times on the node before unmounting it to update
4634 Lustre (this is also safe to run on nodes without this problem).
4636 There is also a related kernel bug in the RHEL4 4 2.6.9 kernel
4637 that can cause this same problem, so customers using that kernel
4638 also need to update the kernel in addition to Lustre. In order
4639 to properly fix this bug, the RHEL3 2.4.21 kernel is also updated.
4641 It is normal that files written just before a client crash (less
4642 than 30s) may not yet have been flushed to disk, even for local
4646 Frequency : frequent on thin XT3 nodes
4648 Description: UUID collision on thin XT3 Linux nodes
4649 Details : UUIDs on Compute Node Linux XT3 nodes were not generated
4650 randomly, since we relied on an insufficiently-seeded PRNG.
4655 Description: OSS hangs after "All ost request buffers busy"
4656 Details : A deadlock between quota and journal operations caused OSS
4657 hangs after printing "All ost request buffers busy."
4660 Frequency : always on liblustre builds
4662 Description: Cleanup compiler warnings on liblustre
4665 Frequency : always on liblustre builds on XT3
4667 Description: LC_CONFIG_CDEBUG don't run while build liblustre on XT3.
4671 Description: Addition of EXT3_FEATURE_RO_COMPAT_DIR_NLINKS flag for
4672 > 32000 subdirectories
4673 Details : Add EXT3_FEATURE_RO_COMPAT_DIR_NLINK flag to
4674 EXT3_FEATURE_RO_COMPAT_SUPP. This flag will be set whenever
4675 subdirectory count crosses 32000. This will aid e2fsck to
4676 correctly handle more than 32000 subdirectories.
4679 Frequency : liblustre (e.g. catamount) on a large cluster with >= 8 OSTs/OSS
4681 Description: System hang on startup
4682 Details : This bug allowed the liblustre (e.g. catamount) client to
4683 return to the app before handling all startup RPCs. This
4684 could leave the node unresponsive to lustre network traffic
4685 and manifested as a server ptllnd timeout.
4687 Severity : enhancement
4689 Description: Add "/proc/sys/lustre/debug_peer_on_timeout"
4690 (liblustre envirable: LIBLUSTRE_DEBUG_PEER_ON_TIMEOUT)
4691 boolean to control whether to print peer debug info when a
4692 client's RPC times out.
4697 Description: make O_SYNC working on 2.6 kernels
4698 Details : 2.6 kernels use different method for mark pages for write,
4699 so need add a code to lustre for O_SYNC work.
4704 Description: Failure to close file and release space on NFS
4705 Details : Put inode details into lock acquired in ll_intent_file_open.
4706 Use mdc_intent_lock in ll_intent_open to properly
4707 detect all kind of errors unhandled by mdc_enqueue.
4712 Description: proc file read during shutdown sometimes raced obd removal,
4714 Details : Add lock to prevent obd access after proc file removal.
4717 Frequency : Only for files larger than 4GB on 32-bit clients.
4719 Description: improperly doing page alignment of locks
4720 Details : Modify lustre core code to use CFS_PAGE_* defines instead of
4721 PAGE_*. Make CFS_PAGE_MASK a 64-bit mask.
4726 Description: RPCs being resent when they shouldn't be
4727 Details : Some RPCs that should not be resent are being resent. This
4728 can cause inconsistencies in the RPC state machine. Do not
4729 resend such requests.
4732 Frequency : rare, only with NFS export
4734 Description: Crash on NFS re-export node
4735 Details : under very unusual load conditions an assertion is hit in
4736 ll_intent_file_open()
4739 Frequency : only if OST filesystem is corrupted
4741 Description: client incorrectly hits assertion in ptlrpc_replay_req()
4742 Details : for a short time RPCs with bulk IO are in the replay list,
4743 but replay of bulk IOs is unimplemented. If the OST filesystem
4744 is corrupted due to disk cache incoherency and then replay is
4745 started it is possible to trip an assertion. Avoid putting
4746 committed RPCs into the replay list at all to avoid this issue.
4751 Description: large O_DIRECT requests fail under memory pressure/fragmentation
4752 Details : Large single O_DIRECT read and write calls can fail to allocate
4753 a sufficiently large buffer to process the request. In case of
4754 allocation failure the allocation is retried with a smaller
4755 buffer and broken into smaller requests.
4757 Severity : enhancement
4759 Description: Add -o localflock option to simulate old noflock behaviour.
4760 Details : This will achieve local-only flock/fcntl locks coherentness.
4765 Description: versioning check is incomplete
4766 Details : Checking the version difference of client vs. server, report
4767 error if the gap is too big.
4772 Description: add support PG_writeback bit
4773 Details : add support for PG_writeback bit for Lustre, for more carefull
4774 work with page cache in 2.6 kernel. This also fix some deadlocks
4775 and remove hack for work O_SYNC with 2.6 kernel.
4777 Severity : enhancement
4779 Description: Add uninit_groups feature to ldiskfs2 to speed up e2fsck
4780 Details : The uninit_groups feature works in conjunction with the kernel
4781 filesystem code (ldiskfs2 only) and e2fsprogs-1.39-cfs6 to speed
4782 up the pass1 processing of e2fsck. This is a read-only feature
4783 in ldiskfs2 only, so older kernels and current ldiskfs cannot
4784 mount filesystems that have had this feature enabled.
4786 Severity : enhancement
4788 Description: Improve multi-block allocation algorithm to avoid fragmentation
4789 Details : The mballoc3 code (ldiskfs2 only) adds new mechanisms to improve
4790 allocation locality and avoid filesystem fragmentation.
4792 ------------------------------------------------------------------------------
4794 2007-02-09 Cluster File Systems, Inc. <info@clusterfs.com>
4796 * Support for kernels:
4797 2.6.9-42.0.3.EL (RHEL 4)
4798 2.6.5-7.276 (SLES 9)
4799 2.4.21-47.0.1.EL (RHEL 3)
4800 2.6.12.6 vanilla (kernel.org)
4801 2.6.16.21-0.8 (SLES10)
4802 * Recommended e2fsprogs version: 1.39.cfs2-0
4804 * The backwards-compatible /proc/sys/portals symlink has been removed
4805 in this release. Before upgrading, please ensure that you change
4806 any configuration scripts or /etc/sysctl.conf files that access
4807 /proc/sys/portals/* or sysctl portals.* to use the corresponding
4808 entry in /proc/sys/lnet or sysctl lnet.*. This change can be made
4809 in advance of the upgrade on any system running Lustre 1.4.6 or
4810 newer, since /proc/sys/lnet was added in that version.
4811 * Note that reiserfs quotas are disabled on SLES 10 in this kernel
4815 Frequency : only when quota is used
4817 Description: avoid scanning export list for quota master
4818 Details : Change the algorithms to avoid scanning export list in order
4819 to improve the efficiency.
4822 Frequency : MDS failover only, very rarely
4824 Description: "went back in time" messages on mds failover
4825 Details : The greatest transno may be lost when the current operation
4826 finishes with an error (transno==0) and the client's last_rcvd
4827 record is over-written. Save the greatest transno in the
4828 mds_last_transno for this case.
4831 Frequency : always for specific kernels and striping counts
4833 Description: client may get "Matching packet too big" without ACL support
4834 Details : Clients compiled without CONFIG_FS_POSIX_ACL get an error message
4835 when trying to access files in certain configurations. The
4836 clients should in fact be denied when mounting because they do
4837 not understand ACLs.
4840 Frequency : Cray XT3 with more than 4000 clients and multiple jobs
4842 Description: many clients connecting with IO in progress causes connect timeouts
4843 Details : Avoid synchronous journal commits to avoid delays caused by many
4844 clients connecting/disconnecting when bulk IO is in progress.
4845 Queue liblustre connect requests on OST_REQUEST_PORTAL instead of
4846 OST_IO_PORTAL to avoid delays behind potentially many pending
4850 Frequency : occasionally with multiple writers to a single file
4852 Description: shared writes to file may result in wrong size reported by stat()
4853 Details : Allow growing of kms when extent lock is cancelled
4856 Frequency : always with random mmap IO to multi-striped file
4858 Description: mmap write might be lost if we are writing to a 'hole' in stripe
4859 Details : Only if the hole is at the end of OST object so that kms is too
4860 small. Fix is to increase kms accordingly in ll_nopage.
4863 Frequency : rare, only if OST filesystem is inconsistent with MDS filesystem
4865 Description: writes to a missing object would leak memory on the OST
4866 Details : If there is an inconsistency between the MDS and OST filesystems,
4867 such that the MDS references an object that doesn't exist, writes
4868 to that object will leak memory due to incorrect cleanup in the
4869 error handling path, eventually running out of memory on the OST.
4874 Description: Creating too long symlink causes lustre errors
4875 Details : Check symlink and name lengths before sending requests to MDS.
4878 Frequency : only if flock is enabled (not on by default)
4880 Description: posix locks not released on fd closure on 2.6.9+
4881 Details : We failed to add posix locks to list of inode locks on 2.6.9+
4882 kernels, this caused such locks not to be released on fd close and
4883 then assertions on fs unmount about still used locks.
4886 Frequency : MDS failover only, very rarely
4888 Description: clients may get ASSERTION(granted_lock != NULL)
4889 Details : When request was taking a long time, and a client was resending
4890 a getattr by name lock request. The were multiple lock requests
4891 with the same client lock handle and
4892 mds_getattr_name->fixup_handle_for_resent_request found one of the
4893 lock handles but later failed with ASSERTION(granted_lock != NULL).
4898 Description: handle->h_buffer_credits > 0, assertion failure
4899 Details : h_buffer_credits is zero after truncate, causing assertion
4900 failure. This patch extends the transaction or creates a new
4904 Frequency : NFS re-export or patchless client
4905 Bugzilla : 11179, 10796
4906 Description: Crash on NFS re-export node (__d_move)
4907 Details : We do not want to hash the dentry if we don't have a lock.
4908 But if this dentry is later used in d_move, we'd hit uninitialised
4909 list head d_hash, so we just do this to init d_hash field but
4910 leave dentry unhashed.
4913 Frequency : NFS re-export or patchless client
4915 Description: NFS exports has problem with symbolic link
4916 Details : lustre client didn't properly install dentry when re-exported
4917 to NFS or running patchless client.
4920 Frequency : NFS re-export or patchless client
4922 Description: Various nfs/patchless fixes.
4923 Details : fixes reuse disconected alias for lookup process - this fixes
4924 warning "find_exported_dentry: npd != pd",
4925 fix permission error with open files at nfs.
4926 fix apply umask when do revalidate.
4929 Frequency : occasional
4931 Description: Crash on NFS re-export node
4932 Details : calling clear_page() on the wrong pointer triggered oops in
4933 generic_mapping_read().
4936 Frequency : rarely, using O_DIRECT IO
4938 Description: unaligned directio crashes client with LASSERT
4939 Details : check for unaligned buffers before trying any requests.
4942 Frequency : rarely, using CFS RAID5 patches in non-standard kernel series
4944 Description: stale data returned from RAID cache
4945 Details : If only a small amount of IO is done to the RAID device before
4946 reading it again it is possible to get stale data from the RAID
4947 cache instead of reading it from disk.
4950 Frequency : always for sles10 kernel
4952 Description: sles10 support
4953 Details : ll_follow_link: compile fixes and using of nd_set_link
4954 under newer kernels.
4957 Frequency : depends on arch, kernel and compiler version, always on sles10
4960 Description: recursive or deep enough symlinks cause stack overflow
4961 Details : getting rid of large stack-allocated variable in
4965 Frequency : depends on hardware
4967 Description: lustre write performance loss in the SLES10 kernel
4968 Details : the performance loss is caused by using of write barriers in the
4969 ext3 code. The SLES10 kernel turns barrier support on by
4970 default. The fix is to undo that change for ldiskfs.
4972 ------------------------------------------------------------------------------
4974 2006-12-09 Cluster File Systems, Inc. <info@clusterfs.com>
4976 * Support for kernels:
4977 2.6.9-42.0.3EL (RHEL 4)
4978 2.6.5-7.276 (SLES 9)
4979 2.4.21-47.0.1.EL (RHEL 3)
4980 2.6.12.6 vanilla (kernel.org)
4984 Frequency : quota enabled and large files being deleted
4986 Description: releasing more than 4GB of quota at once hangs OST
4987 Details : If a user deletes more than 4GB of files on a single OST it
4988 will cause the OST to spin in an infinite loop. Release
4989 quota in < 4GB chunks, or use a 64-bit value for 1.4.7.1+.
4994 Description: statfs data retrieved from /proc may be stale or zero
4995 Details : When reading per-device statfs data from /proc, in the
4996 {kbytes,files}_{total,free,avail} files, it may appear
4997 as zero or be out of date.
5000 Frequency : systems with MD RAID1 external journal devices
5002 Description: lconf's call to blkid is confused by RAID1 journal devices
5003 Details : Use the "blkid -l" flag to locate the MD RAID device instead
5004 of returning all block devices that match the journal UUID.
5007 Frequency : always, for aggregate stripe size over 4GB
5009 Description: "lfs setstripe" fails assertion when setting 4GB+ stripe width
5010 Details : Using "lfs setstripe" to set stripe size * stripe count over 4GB
5011 will fail the kernel with "ASSERTION(lsm->lsm_xfersize != 0)"
5014 Frequency : always if "lfs find" used on a local file/directory
5016 Description: "lfs find" segfaults if used on a local file/directory
5017 Details : The case where a directory component was not specified wasn't
5018 handled correctly. Handle this properly.
5021 Frequency : always on ppc64
5023 Description: the write to an ext3 filesystem mounted with mballoc got stuck
5024 Details : ext3_mb_generate_buddy() uses find_next_bit() which does not
5025 perform endianness conversion.
5028 Frequency : rarely (truncate to non-zero file size after write under load)
5029 Bugzilla : 10730, 10687
5030 Description: Files padded with zeros to next 4K multiple
5031 Details : With filesystems mounted using the "extents" option (2.6 kernels)
5032 it is possible that files that are truncated to a non-zero size
5033 immediately after being written are filled with zero bytes beyond
5034 the truncated size. No file data is lost.
5036 Severity : enhancement
5038 Description: Allow recovery/failover for liblustre clients.
5039 Details : liblustre clients were unaware of failover configurations until
5042 Severity : enhancement
5044 Description: user file locks should fail when not mounting with flock option
5045 Details : Set up an error-returning stub in ll_file_operations.lock field
5046 to prevent incorrect behaviour when client is mounted without
5047 flock option. Also, set up properly f_op->flock field for
5051 Frequency : always on ia64
5053 Description: "lfs df" loops on printing out MDS statfs information
5054 Details : The obd_ioctl_data was not initialized and in some systems
5055 this caused a failure during the ioctl that did not return
5056 an error. Initialize the struct and return an error on failure.
5059 Frequency : SLES 9 only
5061 Description: Error of copying files with lustre special EAs as root
5062 Details : Client side always return success for setxattr call for lustre
5063 special xattr (currently only "trusted.lov").
5066 Frequency : rarely on clusters with both ia64+i386 clients
5068 Description: ia64+i686 clients doing shared IO on the same file may LBUG
5069 Details : In rare cases when both ia64+i686 (or other mixed-PAGE_SIZE)
5070 clients are doing concurrent writes to the same file it is
5071 possible that the ia64 clients may LASSERT because the OST
5072 extent locks are not PAGE_SIZE aligned. Ensure that grown
5073 locks are always aligned on the request boundary.
5076 Frequency : specific use, occasional
5078 Description: Overwriting in use executable truncates on-disk binary image
5079 Details : If one node attempts to overwrite an executable in use by
5080 another node, we now correctly return ETXTBSY instead of
5081 truncating the file.
5083 Severity : enhancement
5085 Description: Async OSC create to avoid the blocking unnecessarily.
5086 Details : If a OST has no remain object, system will block on the creating
5087 when need to create a new object on this OST. Now, ways use
5088 pre-created objects when available, instead of blocking on an
5089 empty osc while others are not empty. If we must block, we block
5090 for the shortest possible period of time.
5095 Description: chmod on Lustre root is propagated to other clients
5096 Details : Re-validate root's dentry in ll_lookup_it to avoid having it
5097 invalid by the follow_mount time.
5100 Frequency : liblustre clients only
5102 Description: Race in 'instant cancel' lock handling could lead to such locks
5103 never to be granted in case of SMP MDS
5104 Details : Do not destroy not yet granted but cbpending locks in
5108 Frequency : replay/resend of open
5110 Description: non null lock assetion failure in mds_intent_policy
5111 Details : Trying to replay/resend lockless open requests resulted in
5112 mds_open() returning 0 with no lock. Now it sets a flag if
5113 a lock is going to be returned.
5115 Severity : enhancement
5117 Description: Checksum enhancements
5118 Details : New checksum enhancements allow for resending RPCs that failed
5121 Severity : enhancement
5123 Description: Tunables on number of dirty pages in cacche
5124 Details : Allow to set limit on number of dirty pages cached.
5129 Description: client crash on unmount - lock still has references
5130 Details : In some error handling cases it was possible to leak a lock
5131 reference on a client while accessing a file. This was not
5132 harmful to the client during operation, but would cause the
5133 client to crash when the filesystem is unmounted.
5136 Frequency : specific case, rare
5138 Description: ETXTBSY on mds though file not in use by client
5139 Details : ETXTBSY is no longer incorrectly returned when attempting to
5140 chmod or chown a directory that the user previously tried to
5141 execute or a currently-executing binary.
5144 Frequency : extremely rare except on liblustre-based clients
5146 Description: Lustre space not freed when files are deleted
5147 Details : Clean up open-unlinked files after client eviction. Previously
5148 the unlink was skipped and the files remained as orphans.
5153 Description: OST failure "would be an LBUG" in waiting_locks_callback()
5154 Details : In some cases it was possible to send a blocking callback to
5155 a client doing a glimpse, even though that client didn't get
5156 a lock granted. When the glimpse lock is cancelled on the OST
5157 the freed lock is left on the waiting list and corrupted the list.
5160 Frequency : all core dumps
5162 Description: Broke core dumps to lustre
5163 Details : Negative dentry may be unhashed if parent does not have UPDATE
5164 lock, but some callers, e.g. do_coredump, expect dentry to be
5165 hashed after successful create, hash it in ll_create_it.
5167 ------------------------------------------------------------------------------
5169 2006-09-13 Cluster File Systems, Inc. <info@clusterfs.com>
5171 * Support for kernels:
5172 2.6.9-42.0.2.EL (RHEL 4)
5173 2.6.5-7.276 (SLES 9)
5174 2.4.21-40.EL (RHEL 3)
5175 2.6.12.6 vanilla (kernel.org)
5179 Frequency : always on RHEL 3
5181 Description: Number of open files grows over time
5182 Details : The number of open files grows over time, whether or not
5183 Lustre is started. This was due to a filp leak introduced
5184 by one of our kernel patches.
5186 ------------------------------------------------------------------------------
5188 2006-08-20 Cluster File Systems, Inc. <info@clusterfs.com>
5190 * Support for kernels:
5191 2.6.9-42.EL (RHEL 4)
5192 2.6.5-7.267 (SLES 9)
5193 2.4.21-40.EL (RHEL 3)
5194 2.6.12.6 vanilla (kernel.org)
5199 Bugzilla : 5719, 9635, 9792, 9684
5200 Description: OST (or MDS) trips assertions in (re)connection under heavy load
5201 Details : If a server is under heavy load and cannot reply to new
5202 connection requests before the client resends the (re)connect,
5203 the connection handling code can behave badly if two service
5204 threads are concurrently handing separate (re)connections from
5205 the same client. Add better locking to the connection handling
5206 code, and ensure that only a single connection will be processed
5207 for a given client UUID, even if the lock is dropped.
5209 Severity : enhancement
5211 Description: add TCP zero-copy support to kernel
5212 Details : Add support to the kernel TCP stack to allow zero-copy bulk
5213 sends if the hardware supports scatter-gather and checksumming.
5214 This allows socklnd to do client-write and server-read more
5215 efficiently and reduce CPU utilization from skbuf copying.
5218 Frequency : only if NFS exporting from client
5220 Description: NULL pointer deref in ll_iocontrol() if chattr mknod file
5221 Details : If setting attributes on a file created under NFS that had
5222 never been opened it would be possible to oops the client
5223 if the file had no objects.
5227 Bugzilla : 9326, 10402, 10897
5228 Description: client crash in ptlrpcd_wake() thread when sending async RPC
5229 Details : It is possible that ptlrpcd_wake() dereferences a freed async
5230 RPC. In rare cases the ptlrpcd thread alread processed the RPC
5231 before ptlrpcd_wake() was called and the request was freed.
5234 Frequency : always for liblustre
5236 Description: liblustre client does MDS+OSTs setattr RPC for each write
5237 Details : When doing a write from a liblustre client, the client
5238 incorrectly issued an RPC to the MDS and each OST the file was
5239 striped over in order to update the timestamps. When writing
5240 with small chunks and many clients this could overwhelm the MDS
5241 with RPCs. In all cases it would slow down the write because
5242 these RPCs are unnecessary.
5244 Severity : enhancement
5246 Description: allow number of MDS service threads to be changed at module load
5247 Details : It is now possible to change the number of MDS service threads
5248 running. Adding "options mds mds_num_threads={N}" to the MDS's
5249 /etc/modprobe.conf will set the number of threads for the next
5250 time Lustre is restarted (assuming the "mds" module is also
5251 reloaded at that time). The default number of threads will
5252 stay the same, 32 for most systems.
5257 Description: OST crash if filesystem is unformatted or corrupt
5258 Details : If an OST is started on a device that has never been formatted
5259 or if the filesystem is corrupt and cannot even mount then the
5260 error handling cleanup routines would dereference a NULL pointer.
5265 Description: NULL pointer deref in llap_from_page.
5266 Details : get_cache_page_nowait can return a page with NULL (or otherwise
5267 incorrect) mapping if the page was truncated/reclaimed while it was
5268 searched for. Check for this condition and skip such pages when
5269 doing readahead. Introduce extra check to llap_from_page() to
5270 verify page->mapping->host is non-NULL (so page is not anonymous).
5273 Frequency : Sometimes when using sys_sendfile
5275 Description: "page not covered by a lock" warnings from ll_readpage
5276 Details : sendfile called ll_readpage without right page locks present.
5277 Now we introduced ll_file_sendfile that does necessary locking
5278 around call to generic_file_sendfile() much like we do in
5282 Frequency : with certain MDS communication failures at client mount time
5284 Description: NULL pointer deref after failed client mount
5285 Details : a client connection request may delayed by the network layer
5286 and not be sent until after the PTLRPC layer has timed out the
5287 request. If the client fails the mount immediately it will try
5288 to clean up before the network times out the request. Add a
5289 reference from the request import to the obd device and delay
5290 the cleanup until the network drops the request.
5293 Frequency : occasionally during client (re)connect
5295 Description: assertion failure during client (re)connect
5296 Details : processing a client connection request may be delayed by the
5297 client or server longer than the client connect timeout. This
5298 causes the client to resend the connection request. If the
5299 original connection request is replied in this interval, the
5300 client may trip an assertion failure in ptlrpc_connect_interpret()
5301 which thought it would be the only running connect process.
5304 Frequency : only with obd_echo servers and clients that are rebooted
5306 Description: kernel BUG accessing uninitialized data structure
5307 Details : When running an obd_echo server it did not start the ping_evictor
5308 thread, and when a client was evicted an uninitialized data
5309 structure was accessed. Start the ping_evictor in the RPC
5310 service startup instead of the OBD startup.
5312 Severity : enhancement
5313 Bugzilla : 10193 (patchless)
5314 Description: Remove dependency on various unexported kernel interfaces.
5315 Details : No longer need reparent_to_init, exit_mm, exit_files,
5316 sock_getsockopt, filemap_populate, FMODE_EXEC, put_filp.
5319 Frequency : rare (only users of deprecated and unsupported LDAP config)
5321 Description: write_conf for zeroconf mount queried LDAP incorrectly for client
5322 Details : LDAP apparently contains 'lustreName' attributes instead of
5323 'name'. A simple remapping of the name is sufficient.
5326 Frequency : rare (only with non-default dump_on_timeout debug enabled)
5328 Description: waiting_locks_callback trips kernel BUG if client is evicted
5329 Details : Running with the dump_on_timeout debug flag turned on makes
5330 it possible that the waiting_locks_callback() can try to dump
5331 the Lustre kernel debug logs from an interrupt handler. Defer
5332 this log dumping to the expired_lock_main() thread.
5334 Severity : enhancement
5336 Description: Support NFS exporting on 2.6 kernels.
5337 Details : Implement non-rawops metadata methods for NFS server to use without
5338 changing NFS server code.
5341 Frequency : very rare (synthetic metadata workload only)
5343 Description: two racing renames might cause an MDS thread to deadlock
5344 Details : Running the "racer" program may cause one MDS thread to rename
5345 a file from being the source of a rename to being the target of
5346 a rename at exactly the same time that another thread is doing
5347 so, and the second thread has already enqueued these locks after
5348 doing a lookup of the target and is trying to relock them in
5349 order. Ensure that we don't try to re-lock the same resource.
5352 Frequency : only very large systems with liblustre clients
5354 Description: slow eviction of liblustre clients with the "evict_by_nid" RPC
5355 Details : Use asynchronous set_info RPCs to send the "evict_by_nid" to
5356 all OSTs in parallel. This allows the eviction of stale liblustre
5357 clients to proceed much faster than if they were done in series,
5358 and also offers similar improvements for other set_info RPCs.
5363 Description: excessive CPU usage during initial read phase on client
5364 Details : During the initial read phase on a client, it would agressively
5365 retry readahead on the file, consuming too much CPU and impacting
5366 performance (since 1.4.5.8). Improve the readahead algorithm
5367 to avoid this, and also improve some other common cases (read
5368 of small files in particular, where "small" is files smaller than
5369 /proc/fs/lustre/llite/*/max_read_ahead_whole_mb, 2MB by default).
5374 Description: MDS crash when receiving packet with unknown intent.
5375 Details : Do not LBUG in unknown intent case, just return -EFAULT
5377 Severity : enhancement
5378 Bugzilla : 9293, 9385
5379 Description: MDS RPCs are serialised on client. This is unnecessary for some.
5380 Details : Do not serialize getattr (non-intent version) and statfs.
5383 Frequency : occasional, when OST network is overloaded/intermittent
5385 Description: client evicted by OST after bulk IO timeout
5386 Details : If a client sends a bulk IO request (read or write) the OST
5387 may evict the client if it is unresposive to its data GET/PUT
5388 request. This is incorrect if the network is overloaded (takes
5389 too long to transfer the RPC data) or dropped the OST GET/PUT
5390 request. There is no need to evict the client at all, since
5391 the pinger and/or lock callbacks will handle this, and the
5392 client can restart the bulk request.
5395 Frequency : Always when mmapping file with no objects
5397 Description: client crashes when mmapping file with no objects
5398 Details : Check that we actually have objects in a file before doing any
5399 operations on objects in ll_vm_open, ll_vm_close and
5405 Description: Request leak when working with deleted CWD
5406 Details : Introduce advanced request refcount tracking for requests
5407 referenced from lustre intent.
5409 Severity : Enhancement
5411 Description: Cache open file handles on client.
5412 Details : MDS now will return special lock along with openhandle, if
5413 requested and client is allowed to hold openhandle, even if unused,
5414 until such a lock is revoked. Helps NFS a lot, since NFS is opening
5415 closing files for every read/write openration.
5417 Severity : Enhancement
5419 Description: Cache open negative dentries on client when possible.
5420 Details : Guard negative dentries with UPDATE lock on parent dir, drop
5421 negative dentries on lock revocation.
5426 Description: Remounting a client read-only wasn't possible with a zconf mount
5427 Details : It wasn't possible to remount a client read-only with llmount.
5429 Severity : enhancement
5430 Description: Include MPICH 1.2.6 Lustre ADIO interface patch
5431 Details : In lustre/contrib/ or /usr/share/lustre in RPM a patch for
5432 MPICH is included to add Lustre-specific ADIO interfaces.
5433 This is based closely on the UFS ADIO layer and only differs
5434 in file creation, in order to allow the OST striping to be set.
5435 This is user-contributed code and not supported by CFS.
5440 Description: extended inode attributes (immutable, append-only) work improperly
5441 when 2.4 and 2.6 kernels are used on client/server or vice versa
5442 Details : Introduce kernel-independent values for these flags.
5444 Severity : enhancement
5447 Description: Allow fractional MB tunings for lustre in /proc/ filesystem.
5448 Details : Many of the /proc/ tunables can only be tuned at a megabyte
5449 granularity. Now, Fractional MB granularity is be supported,
5450 this is very useful for low memory system.
5452 Severity : enhancement
5454 Description: Getattr by fid
5455 Details : Getting a file attributes by its fid, obtaining UPDATE|LOOKUP
5456 locks, avoids extra getattr rpc requests to MDS, allows '/' to
5457 have locks and avoids getattr rpc requests for it on every stat.
5460 Frequency : Always, for filesystems larger than 2TB
5462 Description: ldiskfs crash at mount for filesystem larger than 2TB with mballoc
5463 Details : Kenrel kmalloc limits allocations to 128kB and this prevents
5464 filesystems larger than 2TB to be mounted with mballoc enabled.
5467 Frequency : Always, for 32-bit kernel without CONFIG_LBD and filesystem > 2TB
5469 Description: filesystem corruption for non-standard kernels and very large OSTs
5470 Details : If a 32-bit kernel is compiled without CONFIG_LBD enabled and a
5471 filesystems larger than 2TB is mounted then the kernel will
5472 silently corrupt the start of the filesystem. CONFIG_LBD is
5473 enabled for all CFS-supported kernels, but the possibility of
5474 this happening with a modified kernel config exists.
5476 Severity : enhancement
5478 Description: add client O_DIRECT support for 2.6 kernels
5479 Details : It is now possible to do O_DIRECT reads and writes to files
5480 in the Lustre client mountpoint on 2.6 kernel clients.
5482 Severity : enhancement
5484 Description: parallel glimpse, setattr, statfs, punch, destroy requests
5485 Details : Sends glimpse, setattr, statfs, punch, destroy requests to OSTs in
5486 parallel, not waiting for response from every OST before sending
5487 a rpc to the next OST.
5492 Description: setattr vs write race when updating file timestamps
5493 Details : Client processes that update a file timestamp into the past
5494 right after writing to the file (e.g. tar) it is possible that
5495 the updated file modification time can be reset to the current
5496 time due to a race between processing the setattr and write RPC.
5498 Severity : enhancement
5500 Description: Bring 'lfs find' closer in line with regular Linux find.
5501 Details : lfs find util supports -atime, -mtime, -ctime, -maxdepth, -print,
5502 -print0 options and obtains all the needed info through the lustre
5505 Severity : enhancement
5507 Description: support up to 1024 configured devices on one node
5508 Details : change obd_dev array from statically allocated to dynamically
5509 allocated structs as they are first used to reduce memory usage
5514 Description: Flush dirty partially truncated pages during truncate
5515 Details : Immediatelly flush partially truncated pages in filter_setattr,
5516 this way we completely avoid having any pages in page cache on OST
5517 and can retire ugly workarounds during writes to flush such pages.
5522 Description: i_sem vs transaction deadlock in mds_obd_destroy during unlink.
5523 Details : protect inode from truncation within vfs_unlink() context
5524 just take a reference before calling vfs_unlink() and release it
5525 when parent's i_sem is free.
5528 Frequency : always, if extents are used on OSTs
5530 Description: index ei_leaf_hi (48-bit extension) is not zeroed in extent index
5531 Details : OSTs using the extents format would not zero the high 16 bits of
5532 the index physical block number. This is not a problem for any
5533 OST filesystems smaller than 16TB, and no kernels support ext3
5534 filesystems larger than 16TB yet. This is fixed in 1.4.7 (all
5535 new/modified files) and can be fixed for existing filesystems
5536 with e2fsprogs-1.39-cfs1.
5541 Description: import connection selection may be incorrect if timer wraps
5542 Details : Using a 32-bit jiffies timer with HZ=1000 may cause backup
5543 import connections to be ignored if the 32-bit jiffies counter
5544 wraps. Use a 64-bit jiffies counter.
5547 Frequency : during server recovery
5549 Description: crash after server is denying duplicate export
5550 Details : If clients are resending connect requests to the server, the
5551 server refuses to allow a client to connect multiple times.
5552 Fixed a bug in the handling of this case.
5555 Frequency : very large clusters immediately after boot
5557 Description: LNET request buffers exhausted under heavy short-term load
5558 Details : If a large number of client requests are generated on a service
5559 that has previously never seen so many requests it is possible
5560 that the request buffer growth cannot keep up with the spike in
5561 demand. Instead of dropping incoming requests, they are held in
5562 the LND until the RPC service can accept more requests.
5565 Frequency : Sometimes during replay
5567 Description: Assertion failure in ll_local_open after replay.
5568 Details : If replay happened on an open request reply before we were able
5569 to set replay handler, reply will become not swabbed tripping the
5570 assertion in ll_local_open. Now we set the handler right after
5571 recognising of open request
5574 Frequency : very rare
5576 Description: kernel reports "badness in vsnprintf"
5577 Details : Reading from the "recovery_status" /proc file in small chunks
5578 may cause a negative length in lprocfs_obd_rd_recovery_status()
5579 call to vsnprintf() (which is otherwise harmless). Exit early
5580 if there is no more space in the output buffer.
5582 Severity : enhancement
5584 Description: clear OBD RPC statistics by writing to them
5585 Details : It is now possible to clear the OBD RPC statistics by writing
5586 to the "stats" file.
5591 Description: Client mtime is not the same on different clients after utimes
5592 Details : In some cases, the client was using the utimes() syscall on
5593 a file cached on another node. The clients now validate the
5594 ctime from the MDS + OSTs to determine which one is right.
5599 Description: Inability to activate failout mode
5600 Details : lconf script incorrectly assumed that in python string's numeric
5601 value is used in comparisons.
5604 Frequency : always with multiple stripes per file
5606 Description: Inefficient object allocation for mutli-stripe files
5607 Details : When selecting which OSTs to stripe files over, for files with
5608 a stripe count that divides evenly into the number of OSTs,
5609 the MDS is always picking the same starting OST for each file.
5610 Return the OST selection heuristic to the original design.
5615 Description: mount failures may take full timeout to return an error
5616 Details : Under some heavy load conditions it is possible that a
5617 failed mount can wait for the full obd_timeout interval,
5618 possibly several minutes, before reporting an error.
5619 Instead return an error as soon as the status is known.
5621 ------------------------------------------------------------------------------
5623 2006-02-14 Cluster File Systems, Inc. <info@clusterfs.com>
5625 * WIRE PROTOCOL CHANGE. This version of Lustre networking WILL NOT
5626 INTEROPERATE with older versions automatically. Please read the
5627 user documentation before upgrading any part of a live system.
5628 * WARNING: Lustre networking configuration changes are required with
5629 this release. See https://bugzilla.clusterfs.com/show_bug.cgi?id=10052
5632 * Support for kernels:
5633 2.6.9-22.0.2.EL (RHEL 4)
5634 2.6.5-7.244 (SLES 9)
5635 2.6.12.6 vanilla (kernel.org)
5638 Severity : enhancement
5639 Bugzilla : 7981/8208
5640 Description: Introduced Lustre Networking (LNET)
5641 Details : LNET is new networking infrastructure for Lustre, it includes
5642 a reorganized network configuration mode (see the user
5643 documentation for full details) as well as support for routing
5644 between different network fabrics. Lustre Networking Devices
5645 (LNDs) for the supported network fabrics have also been
5646 created for this new infrastructure.
5648 Severity : enhancement
5649 Description: Introduced Access control lists
5650 Details : clients can set ACLs on files and directories in order to have
5651 more fine-grained permissions than the standard Unix UGO+RWX.
5652 The MDS must be started with the "-o acl" mount option.
5654 Severity : enhancement
5655 Description: Introduced filesystem quotas
5656 Details : Administrators may now establish per-user quotas on the
5659 Severity : enhancement
5661 Description: Configuration change for the XT3
5662 The PTLLND is now used to run Lustre over Portals on the XT3
5663 The configure option(s) --with-cray-portals are no longer used.
5664 Rather --with-portals=<path-to-portals-includes> is used to
5665 enable building on the XT3. In addition to enable XT3 specific
5666 features the option --enable-cray-xt3 must be used.
5671 Description: Running on many-way SMP OSTs can trigger oops in llcd_send()
5672 Details : A race between allocating a new llcd and re-getting the llcd_lock
5673 allowed another thread to grab newly-allocated llcd.
5675 Severity : enhancement
5677 Description: 2.6 OST async journal commit and locking fix to improve performance
5678 Details : The filter_direct_io()+filter_commitrw_write() journal commits for
5679 2.6 kernels are now async as they already were in 2.4 kernels so
5680 that they can commit concurrently with the network bulk transfer.
5681 For block-allocated files the filter allocation semaphore is held
5682 to avoid filesystem fragmentation during allocation. BKL lock
5683 removed for 2.6 xattr operations where it is no longer needed.
5688 Description: lconf incorrectly determined whether two IP networks could talk
5689 Details : In some more complicated routing and multiple-network
5690 configurations, lconf will avoid trying to make a network
5691 connection to a disjoint part of the IP space. It was doing the
5692 math incorrectly for one set of cases.
5697 Description: Fix for potential infinite loop processing records in an llog.
5698 Details : If an llog record is corrupted/zeroed, it is possible to loop
5699 forever in llog_process(). Validate the llog record length
5700 and skip the remainder of the block on error.
5703 Frequency : occasional (liblustre only)
5705 Description: liblustre could not open files whose last component is a symlink
5706 Details : sysio_path_walk() would incorrectly pass the open intent to
5707 intermediate path components.
5710 Frequency : rare (liblustre only with non-standard tuning)
5711 Bugzilla : 7201 (7350)
5712 Description: Tuning the MDC DLM LRU size to zero triggers client LASSERT
5713 Details : llu_lookup_finish_locks() tries to set lock data on a lock
5714 after it has been released, only do this for referenced locks
5716 Severity : enhancement
5718 Description: specifying an (invalid) directory default stripe_size of -1
5719 would reset the directory default striping
5720 Details : stripe_size -1 was used internally to signal directory stripe
5721 removal, now use "all default" to signal dir stripe removal
5722 as a directory striping of "all default" is not useful
5725 Frequency : common for large clusters running liblustre clients
5727 Description: doing an ls when liblustre clients are running is slow
5728 Details : sending a glimpse AST to a liblustre client waits for every AST
5729 to time out, as liblustre clients will not respond. Since they
5730 cannot cache data we refresh the OST lock LVB from disk instead.
5732 Severity : enhancement
5734 Description: doing an ls at the same time as file IO can be slow
5735 Details : enqueue and other "small" requests can be blocked behind many
5736 large IO requests. Create a new OST IO portal for non-IO
5737 requests so they can be processed faster.
5740 Frequency : rare (only HPUX clients mounting unsupported re-exported NFS vol)
5742 Description: an HPUX NFS client would get -EACCESS when ftruncate()ing a newly
5743 created file with mode 000
5744 Details : the Linux NFS server relies on an MDS_OPEN_OWNEROVERRIDE hack to
5745 allow an ftruncate() as a non-root user to a file with mode 000.
5746 Lustre now respects this flag to disable mode checks when
5747 truncating a file owned by the user
5750 Frequency : liblustre-only, when liblustre client dies unexpectedly or becomes
5753 Description: Revoking locks from clients that went dead or catatonic might take
5755 Details : New lock flags FL_CANCEL_ON_BLOCK used by liblustre makes
5756 cancellation of such locks instant on servers without waiting for
5757 any reply from clients. Clients drops these locks when cancel
5758 notification from server is received without replying.
5761 Frequency : liblustre-only, when liblustre client dies or becomes busy
5763 Description: Doing ls on Linux clients can take a long time with active
5765 Details : Liblustre client cannot handle ASTs in timely manner, so avoid
5766 granting such locks to it in the first place if possible. Locks
5767 are taken by proxy on the OST during the read or write and
5768 dropped immediately afterward. Add connect flags handling, do
5769 not grant locks to liblustre clients for glimpse ASTs.
5771 Severity : enhancement
5773 Description: Improve read-ahead algorithm to avoid excessive IO for random reads
5774 Details : Existing read-ahead algorithm is tuned for the case of streamlined
5775 sequential reads and behaves badly with applications doing random
5776 reads. Improve it by reading ahead at least read region, and
5777 avoiding excessive large RPC for small reads.
5779 Severity : enhancement
5781 Description: Creating more than 1000 files for a single job may cause a load
5782 imbalance on the OSTs if there are also a large number of OSTs.
5783 Details : qos_prep_create() uses an OST index reseed value that is an
5784 even multiple of the number of available OSTs so that if the
5785 reseed happens in the middle of the object allocation it will
5786 still utilize the OSTs as uniformly as possible.
5791 Description: OST or MDS may oops in ping_evictor_main()
5792 Details : ping_evictor_main() drops obd_dev_lock if deleting a stale export
5793 but doesn't restart at beginning of obd_exports_timed list
5796 Severity : enhancement
5798 Description: improve by-nid export eviction on the MDS and OST
5799 Details : allow multiple exports with the same NID to be evicted at one
5800 time without re-searching the exports list.
5803 Frequency : rare, only with supplementary groups enabled on SMP 2.6 kernels
5805 Description: MDS may oops in groups_free()
5806 Details : in rare race conditions a newly allocated group_info struct is
5807 freed again, and this can be NULL. The 2.4 compatibility code
5808 for groups_free() checked for a NULL pointer, but 2.6 did not.
5811 Frequency : common for liblustre clients doing little filesystem IO
5812 Bugzilla : 9352, 7313
5813 Description: server may evict liblustre clients accessing contended locks
5814 Details : if a client is granted a lock or receives a completion AST
5815 with a blocking AST already set it would not reply to the AST
5816 for LDLM_FL_CANCEL_ON_BLOCK locks. It now replies to such ASTs.
5819 Frequency : lfs setstripe, only systems with more than 160 OSTs
5821 Description: unable to set striping with a starting offset beyond OST 160
5822 Details : llapi_create_file() incorrectly limited the starting stripe
5823 index to the maximum single-file stripe count.
5826 Frequency : LDAP users only
5828 Description: lconf did not handle in-kernel recovery with LDAP properly
5829 Details : lconf/LustreDB get_refs() is searching the wrong namespace
5831 Severity : enhancement
5833 Description: bind OST threads to NUMA nodes to improve performance
5834 Details : all OST threads are uniformly bound to CPUs on a single NUMA
5835 node and do their allocations there to localize memory access
5837 Severity : enhancement
5839 Description: llmount can determine client NID directly from Myrinet (GM)
5840 Details : the client NID code from gmnalnid was moved directly into
5841 llmount, removing the need to use this or specifying the
5842 client NID explicitly when mounting GM clients with zeroconf
5845 Frequency : if client is started with down MDS
5847 Description: if client is started with down MDS mount hangs in ptlrpc_queue_wait
5848 Details : Having an LWI_INTR() wait event (interruptible, but no timeout)
5849 will wait indefinitely in ptlrpc_queue_wait->l_wait_event() after
5850 ptlrpc_import_delayed_req() because we didn't check if the
5851 request was interrupted, and we also didn't break out of the
5852 event loop if there was no timeout
5857 Description: data loss during non-page-aligned writes to a single file from
5858 both multiple nodes and multiple threads on one node at same time
5859 Details : updates to KMS and lsm weren't protected by common lock. Resulting
5860 inconsistency led to false short-reads, that were cached and later
5861 used by ->prepare_write() to fill in partially written page,
5862 leading to data loss.
5865 Frequency : always, if lconf --abort_recovery used
5867 Description: lconf --abort_recovery fails with 'Operation not supported'
5868 Details : lconf was attempting to abort recovery on the MDT device and not
5871 Severity : enhancement
5873 Description: remove cleanup logs
5874 Details : replace lconf-generated cleanup logs with lustre internal
5875 cleanup routines. Eliminates the need for client-cleanup and
5878 Severity : enhancement
5880 Description: add support for EAs (user and system) on lustre filesystems
5881 Details : it is now possible to store extended attributes in the Lustre
5882 client filesystem, and with the user_xattr mount option it
5883 is possible to allow users to store EAs on their files also
5885 Severity : enhancement
5887 Description: Add possibility (config option) to show minimal available OST free
5889 Details : When compiled with --enable-mindf configure option, statfs(2)
5890 (and so, df) will return least minimal free space available from
5891 all OSTs as amount of free space on FS, instead of summary of
5892 free spaces of all OSTs.
5894 Severity : enhancement
5896 Description: do not expand extent locks acquired on OST-side
5897 Details : Modify ldlm_extent_policy() to not expand local locks, acquired
5898 by server: they are not cached anyway.
5901 Frequency : when mmap is used/binaries executed from Lustre
5903 Description: Unmmap pages before throwing them away from read cache.
5904 Details : llap_shrink cache now attempts to unmap pages before discarding
5905 them (if unmapping failed - do not discard). SLES9 kernel has
5906 extra checks that trigger if this unmapping is not done first.
5911 Description: lconf didn't resolve symlinks before checking to see whether a
5912 given mountpoint was already in use
5915 Frequency : when migrating failover services
5916 Bugzilla : 6395, 9514
5917 Description: When migrating a subset of services from a node (e.g. failback
5918 from a failover service node) the remaining services would
5919 time out and evict clients.
5920 Details : lconf --force (implied by --failover) sets the global obd_timeout
5921 to 5 seconds in order to quickly disconnect, but this caused
5922 other RPCs to time out too quickly. Do not change the global
5923 obd_timeout for force cleanup, only set it for DISCONNECT RPCs.
5925 Severity : enhancement
5926 Frequency : if MDS is started with down OST
5927 Bugzilla : 9439,5706
5928 Description: Allow startup/shutdown of an MDS without depending on the
5929 availability of the OSTs.
5930 Details : Asynchronously call mds_lov_synchronize during MDS startup.
5931 Add appropriate locking and lov-osc refcounts for safe
5932 cleaning. Add osc abort_inflight calls in case the
5933 synchronize never started.
5936 Frequency : occasional (Cray XT3 only)
5938 Description: root not authorized to access files in CRAY_PORTALS environment
5939 Details : The client process capabilities were not honoured on the MDS in
5940 a CRAY_PORTALS/CRAY_XT3 environment. If the file had previously
5941 been accessed by an authorized user then root was able to access
5942 the file on the local client also. The root user capabilities
5943 are now allowed on the MDS, as this environment has secure UID.
5946 Frequency : occasional
5948 Description: ldiskfs "too long searching" message happens too often
5949 Details : A debugging message (otherwise harmless) prints too often on
5950 the OST console. This has been reduced to only happen when
5951 there are fragmentation problems on the filesystem.
5956 Description: Division by zero in statfs when all OSCs are inactive
5957 Details : lov_get_stripecnt() returns zero due to incorrect order of checks,
5958 lov_statfs divides by value returned by lov_get_stripecnt().
5962 Bugzilla : 9489, 3273
5963 Description: First write from each client to each OST was only 4kB in size,
5964 to initialize client writeback cache, which caused sub-optimal
5965 RPCs and poor layout on disk for the first writen file.
5966 Details : Clients now request an initial cache grant at (re)connect time
5967 and so that they can start streaming writes to the cache right
5968 away and always do full-sized RPCs if there is enough data.
5969 If the OST is rebooted the client also re-establishes its grant
5970 so that client cached writes will be honoured under the grant.
5975 Description: Slow ls (and stat(2) syscall) on files residing on IO-loaded OSTs
5976 Details : Now I/O RPCs go to different portal number and (presumably) fast
5977 lock requests (and glimses) and other RPCs get their own service
5978 threads pool that should be able to service those RPCs
5981 Severity : enhancement
5983 Description: Ability to exchange lustre version between client and servers and
5984 issue warnings at client side if client is too old. Also for
5985 liblustre clients there is ability to refuse connection of too old
5987 Details : New 'version' field is added to connect data structure that is
5988 filled with version info. That info is later checked by server and
5992 Frequency : rare, liblustre only.
5993 Bugzilla : 9296, 9581
5994 Description: Two simultaneous writes from liblustre at offset within same page
5995 might proceed at the same time overwriting eachother with stale
5997 Details : I/O lock withing llu_file_prwv was released too early, before data
5998 actually was hitting the wire. Extended lock-holding time until
5999 server acknowledges receiving data.
6002 Frequency : extremely rare. Never observed in practice.
6004 Description: avoid generating lustre_handle cookie of 0.
6005 Details : class_handle_hash() generates handle cookies by incrementing
6006 global counter, and can hit 0 occasionaly (this is unlikely, but
6007 not impossible, because initial value of cookie counter is
6008 selected randonly). Value of 0 is used as a sentinel meaning
6009 "unassigned handle" --- avoid it. Also coalesce two critical
6010 sections in this function into one.
6012 Severity : enhancement
6014 Description: allow liblustre clients to delegate truncate locking to OST
6015 Details : To avoid overhead of locking, liblustre client instructs OST to
6016 take extent lock in ost_punch() on client's behalf. New connection
6017 flag is added to handle backward compatibility.
6019 Severity : enhancement
6020 Bugzilla : 4928, 7341, 9758
6021 Description: allow number of OST service threads to be specified
6022 Details : a module parameter allows the number of OST service threads
6023 to be specified via "options ost ost_num_threads={N}" in the
6024 OSS's /etc/modules.conf or /etc/modprobe.conf.
6028 Bugzilla : 6146, 9635, 9895
6029 Description: servers crash with bad pointer in target_handle_connect()
6030 Details : In rare cases when a client is reconnecting it was possible that
6031 the connection request was the last reference for that export.
6032 We would temporarily drop the export reference and get a new
6033 one, but this may have been the last reference and the export
6034 was just destroyed. Get new reference before dropping old one.
6036 Severity : enhancement
6037 Frequency : if client is started with failover MDS
6039 Description: Allow multiple MDS hostnames in the mount command
6040 Details : Try to read the configuration from all specified MDS
6041 hostnames during a client mount in case the "primary"
6044 Severity : enhancement
6046 Description: Stop sending data to evicted clients as soon as possible.
6047 Details : Check if the client we are about to send or are sending data to
6048 was evicted already. (Check is done every second of waiting,
6049 for which l_wait_event interface was extended to allow checking
6050 of exit condition at specified intervals).
6053 Frequency : rare, normally only when NFS exporting is done from client
6055 Description: 'bad disk LOV MAGIC: 0x00000000' error when chown'ing files
6057 Details : Make mds_get_md() recognise empty md case and set lmm size to 0.
6060 Frequency : always, if srand() is called before liblustre initialization
6062 Description: Liblustre uses system PRNG disturbing its usage by user application
6063 Details : Introduce internal to lustre fast and high-quality PRNG for
6064 lustre usage and make liblustre and some other places in generic
6065 lustre code to use it.
6067 Severity : enhancement
6068 Bugzilla : 9477, 9557, 9870
6069 Description: Verify that the MDS configuration logs are updated when xml is
6070 Details : Check if the .xml configuration logs are newer than the config
6071 logs stored on the MDS and report an error if this is the case.
6072 Request --write-conf, or allow starting with --old_conf.
6074 Severity : enhancement
6076 Description: Handle symlinks in the path when checking if Lustre is mounted.
6077 Details : Resolve intermediate symlinks when checking if a client has
6078 mounted a filesystem to avoid duplicate client mounts.
6083 Description: lconf can hit an error exception but still return success.
6084 Details : The lconf command catches the Command error exception at the top
6085 level script context and will exit with the associated exit
6086 status, but doesn't ensure that this exit status is non-zero.
6091 Description: failure of ptlrpc thread startup can cause oops
6092 Details : Starting a ptlrpc service thread can fail if there are a large
6093 number of threads or the server memory is very fragmented.
6094 Handle this without oopsing.
6097 Frequency : always, only if liblustre and non-default acceptor port was used
6099 Description: liblustre cannot connect to servers with non-default acceptor port
6100 Details : tcpnal_set_default_params() was not called and was therefore
6101 ignoring the environment varaible TCPNAL_PORT, as well as other
6102 TCPNAL_ environment variables
6107 Description: two objects could be created on the same OST for a single file
6108 Details : If an OST is down, in some cases it was possible to create two
6109 objects on a single OST for a single file. No problems other
6110 than potential performance impact and spurious error messages.
6114 Bugzilla : 5681, 9562
6115 Description: Client may oops in ll_unhash_aliases
6116 Details : Client dcache may become inconsistent in race condition.
6117 In some cases "getcwd" can fail if the current directory is
6123 Description: Inode refcounting problems in NFS export code
6124 Details : link_raw functions used to call d_instantiate without obtaining
6125 extra inode reference first.
6129 Bugzilla : 9942, 9903
6130 Description: Referencing freed requests leading to crash, memleaks with NFS.
6131 Details : We used to require that call to ll_revalidate_it was always
6132 followed by ll_lookup_it. Also with revalidate_special() it is
6133 possible to call ll_revalidate_it() twice for the same dentry
6134 even if first occurence returned success. This fix changes semantic
6135 between DISP_ENQ_COMPLETE disposition flag to mean there is extra
6136 reference on a request referred from the intent.
6137 ll_intent_release() then releases such a request.
6140 Frequency : rare, normally benchmark loads only
6142 Description: unlinked inodes were kept in memory on the client
6143 Details : If a client is repeatedly creating and unlinking files it
6144 can accumulate a lot of stale inodes in the inode slab cache.
6145 If there is no other client load running this can cause the
6146 client node to run out of memory. Instead flush old inodes
6147 from client cache that have the same inode number as a new inode.
6150 Frequency : SLES9 2.6.5 kernel and long filenames only
6151 Bugzilla : 9969, 10379
6152 Description: utime reports stale NFS file handle
6153 Details : SLES9 uses out-of-dentry names in some cases, which confused
6154 the lustre dentry revalidation. Change it to always use the
6158 Frequency : rare, unless heavy write-truncate concurrency is continuous
6159 Bugzilla : 4180, 6984, 7171, 9963, 9331
6160 Description: OST becomes very slow and/or deadlocked during object unlink
6161 Details : filter_destroy() was holding onto the parent directory lock
6162 while truncating+unlinking objects. For very large objects this
6163 may block other threads for a long time and slow overall OST
6164 responsiveness. It may also be possible to get a lock ordering
6165 deadlock in this case, or run out of journal credits because of
6166 the combined truncate+unlink. Solution is to do object truncate
6167 first in one transaction without parent lock, and then do the
6168 final unlink in a new transaction with the parent lock. This
6169 reduces the lock hold time dramatically.
6172 Frequency : rare, 2.4 kernels only
6174 Description: MDS or OST cleanup may trip kernel BUG when dropping kernel lock
6175 Details : mds_cleanup() and filter_cleanup() need to drop the kernel lock
6176 before unmounting their filesystem in order to avoid deadlock.
6177 The kernel_locked() function in 2.4 kernels only checks whether
6178 the kernel lock is held, not whether it is this process that is
6179 holding it as 2.6 kernels do.
6184 Description: MDS or OST may oops/LBUG if a client is connecting multiple times
6185 Details : The client ptlrpc code may be trying to reconnect to a down
6186 server before a previous connection attempt has timed out.
6187 Increase the reconnect interval to be longer than the connection
6188 timeout interval to avoid sending duplicate connections to
6192 Frequency : echo_client brw_test command
6194 Description: fix echo_client to work with OST preallocated code
6195 Details : OST preallocation code (5137) didn't take echo_client IO path
6196 into account: echo_client calls filter methods outside of any
6197 OST thread and, hence, there is no per-thread preallocated
6198 pages and buffers to use. Solution: hijack pga pages for IO. As
6199 a byproduct, this avoids unnecessary data copying.
6203 Bugzilla : 3555, 5962, 6025, 6155, 6296, 9574
6204 Description: Client can oops in mdc_commit_close() after open replay
6205 Details : It was possible for the MDS to return an open request with no
6206 transaction number in mds_finish_transno() if the client was
6207 evicted, but without actually returning an error. Clients
6208 would later try to replay that open and may trip an assertion
6209 Simplify the client close codepath, and always return an error
6210 from the MDS in case the open is not successful.
6213 Frequency : rare, 2.6 OSTs only
6215 Description: OST may deadlock under high load on fragmented files
6216 Details : If there was a heavy load and highly-fragmented OST filesystems
6217 it was possible to have all the OST threads deadlock waiting on
6218 allocation of biovecs, because the biovecs were not released
6219 until the entire RPC IO was completed. Instead, release biovecs
6220 as soon as they are complete to ensure forward IO progress.
6222 Severity : enhancement
6224 Description: Support for specifying external journal device at mount
6225 Details : If an OST or MDS device is formatted with an external journal
6226 device, this device major/minor is stored in the ext3 superblock
6227 and may not be valid for failover. Allow detecting and
6228 specifying the external journal at mount time.
6233 Description: Mounting an MDS with pending unlinked files may cause oops
6234 Details : target_finish_recovery() calls mds_postrecov() which returned
6235 the number of orphans unlinked. mds_lov_connect->mds_postsetup()
6236 considers this an error and immediately begins cleaning up the
6237 lov, just after starting the mds_lov process
6239 Severity : enhancement
6241 Description: Implement 'lfs df' to report actual free space on per-OST basis
6242 Details : Add sub-command 'df' on 'lfs' to report the disk space usage of
6243 MDS/OSDs. Usage: lfs df [-i][-h]. Command Options: '-i' to report
6244 usage of objects; '-h' to report in human readable format.
6246 ------------------------------------------------------------------------------
6248 2005-08-26 Cluster File Systems, Inc. <info@clusterfs.com>
6255 Description: Mounting an ldiskfs file system with mballoc may crash OST node.
6256 Details : ldiskfs mballoc code may reference an uninitialized buddy struct
6257 at startup during orphan unlinking. Instead, skip buddy update
6258 before setup, as it will be regenerated after recovery is complete.
6263 Description: If an OST is inactive, its locks might reference stale inodes.
6264 Details : lov_change_cbdata() must iterate over all namespaces, even if
6265 they are inactive to clear inode references from the lock.
6267 Severity : enhancement
6268 Frequency : occasional, if non-standard max_dirty_mb used
6270 Description: Client will block write RPCs if not enough grant
6271 Details : If a client has max_dirty_mb smaller than max_rpcs_in_flight,
6272 then the client will block writes while waiting for another RPC
6273 to complete instead of consuming its dirty limit. With change
6274 we get improved performance when max_dirty_mb is small.
6276 Severity : enhancement
6277 Bugzilla : 3389, 6253
6278 Description: Add support for supplementary groups on the MDS.
6279 Details : The MDS has an upcall /proc/fs/lustre/mds/{mds}/group_upcall
6280 (set to /usr/sbin/l_getgroups if enabled) which will do MDS-side
6281 lookups for user supplementary groups into a cache.
6285 Description: O_CREAT|O_EXCL open flags in liblustre always return -EEXIST
6286 Details : Make libsysio to not enforce O_EXCL by clearing the flag,
6287 for liblustre O_EXCL is enforced by MDS.
6291 Description: readdir never returns NULL in liblustre.
6292 Details : Corrected llu_iop_getdirentries logic, to return offset of next
6293 dentry in struct dirent.
6297 Frequency : liblustre only, depends on application IO pattern
6298 Description: liblustre clients evicted if not contacting servers
6299 Details : Don't put liblustre clients into the ping_evictor list, so
6300 they will not be evicted by the pinger ever.
6302 Severity : enhancement
6304 Description: Add ability to evict clients by NID from MDS.
6305 Details : By echoing "nid:$NID" string into
6306 /proc/fs/lustre/mds/.../evict_client client with nid that equals to
6307 $NID would be instantly evicted from this MDS and from all active
6308 OSTs connected to it.
6312 Description: Do not query file size twice, somewhat slowing stat(2) calls.
6313 Details : lookup_it_finish() used to query file size from OSTs that was not
6318 Description: service threads change working directory to that of init
6319 Details : Starting lustre service threads may pin the working directory
6320 of the parent thread, making that filesystem busy. Threads
6321 now change to the working directory of init to avoid this.
6325 Frequency : during shutdown only
6326 Description: shutdown with a failed MDS or OST can cause unmount to hang
6327 Details : Don't resend DISCONNECT messages in ptlrpc_disconnect_import()
6332 Frequency : 2.6 only
6333 Description: chmod/chown may include an extra supplementary group
6334 Details : ll{,u}_mdc_pack_op_data() does not properly initialize the
6335 supplementary group and if none is specified this is used.
6338 Bugzilla : 5479 (6816)
6340 Description: Racing open + rm can assert client in mdc_set_open_replay_data()
6341 Details : If lookup is in progress on a file that is unlinked we might try
6342 to revalidate the inode and fail in revalidate after lookup is
6343 complete and ll_file_open() enqueues the open again but
6344 it_open_error() was not checking DISP_OPEN_OPEN errors correctly.
6347 Frequency : always, if lconf --abort_recovery used
6349 Description: lconf --abort_recovery fails with 'Operation not supported'
6350 Details : lconf was attempting to abort recovery on the MDT device and not
6353 ------------------------------------------------------------------------------
6355 2005-08-08 Cluster File Systems, Inc. <info@clusterfs.com>
6360 Frequency : rare (only unsupported configurations with a node running as an
6362 Bugzilla : 6514, 5137
6363 Description: Mounting a Lustre file system on a node running as an OST could
6365 Details : OSTs now preallocates memory needed to write out data at
6366 startup, instead of when needed, to avoid having to
6367 allocate memory in possibly low memory situations.
6368 Specifically, if the file system is mounted on on OST,
6369 memory pressure could force it to try to write out data,
6370 which it needed to allocate memory to do. Due to the low
6371 memory, it would be unable to do so and the node would
6372 become unresponsive.
6374 Severity : enhancement
6376 Description: Addition of lconf --service command line option
6377 Details : lconf now accepts a '--service <arg>' option, which is
6378 shorthand for 'lconf --group <arg> --select <arg>=<hostname>'
6380 Severity : enhancement
6382 Description: Failover mode is now the default for OSTs.
6383 Details : By default, OSTs will now run in failover mode. To return to
6384 the old behaviour, add '--failout' to the lmc line for OSTs.
6386 Severity : enhancement
6388 Description: Health checks are now provided for MDS and OSTs
6389 Details : Additional detailed health check information on MSD and OSTs
6390 is now provided through the procfs health_check value.
6393 Frequency : occasional, depends on IO load
6395 Description: Disk fragmentation on the OSTs could eventually cause slowdowns
6396 after numerous create/delete cycles
6397 Details : The ext3 inode allocation policy would not allocate new inodes
6398 very well on the OSTs because there are no new directories
6399 being created. Instead we look for groups with free space if
6400 the parent directories are nearly full.
6405 Description: Network or server problems during mount may cause partially
6406 mounted clients instead of returning an error.
6407 Details : The config llog parsing code may overwrite the error return
6408 code during mount error handling, returning success instead
6414 Description: MDS can fail to allocate large reply buffers
6415 Details : After long uptimes the MDS can fail to allocate large reply
6416 buffers (e.g. zconf client mount config records) due to memory
6417 fragmentation or consumption by the buffer cache. Preallocate
6418 some large reply buffers so that these replies can be sent even
6419 under memory pressure.
6423 Frequency : rare (liblustre)
6424 Description: fsx running with liblustre complained that using truncate() to
6425 extend the file doesn't work. This patch corrects that issue.
6426 Details : This is the liblustre equivalent of the fix for bug 6196. Fixes
6427 ATTR_SIZE and lsm use in llu_setattr_raw.
6431 Frequency : rare, only 2.6 kernels
6432 Description: Unusual file access patterns on the MDS may result in inode
6433 data being lost in very rare circumstances.
6434 Details : Bad interaction between the ea-in-inode patch and the "no-read"
6435 code in the 2.6 kernel caused the inode and/or EA data not to
6436 be read from disk, causing single-file corruption.
6440 Frequency : rare, only 2.6 filesystems using extents
6441 Description: Heavy concurrent write and delete load may cause data corruption.
6442 Details : It was possible under high-load situations to have an extent
6443 metadata block in the block device cache from a just-unlinked
6444 file overwrite a newly-allocated data block. We now unmap any
6445 metadata buffers that alias just-allocated data blocks.
6449 Frequency : filesystems with default stripe_count larger than 77
6450 Description: lconf+mke2fs fail when formatting filesystem with > 77 stripes
6451 Details : lconf specifies an inode size of 4096 bytes when the default
6452 stripe_count is larger than 77. This conflicts with the default
6453 inode density of 1 per 4096 bytes. Allocate smaller inodes in
6454 this case to avoid pinning too much memory for large EAs.
6456 ------------------------------------------------------------------------------
6458 2005-07-07 Cluster File Systems, Inc. <info@clusterfs.com>
6463 Frequency : rare (extremely heavy IO load with hundreds of clients)
6465 Description: Client is evicted, gets IO error writing to file
6466 Details : lock ordering changes for bug 5492 reintroduced bug 3267 and
6467 caused clients to be evicted for AST timeouts. The fixes in
6468 bug 5192 mean we no longer need to have such short AST timeouts
6469 so ldlm_timeout has been increased.
6472 Frequency : occasional during --force or --failover shutdown under load
6473 Bugzilla : 5949, 4834
6474 Description: Server oops/LBUG if stopped with --force or --failover under load
6475 Details : a collection of import/export refcount and cleanup ordering
6476 issues fixed for safer force cleanup
6479 Frequency : only filesystems larger than 120 OSTs
6480 Bugzilla : 5990, 6223
6481 Description: lfs getstripe would oops on a very large filesystem
6482 Details : lov_getconfig used kfree on vmalloc'd memory
6485 Frequency : only filesystems exporting via NFS to Solaris 10 clients
6486 Bugzilla : 6242, 6243
6487 Description: reading from files that had been truncated to a non-zero size
6488 but never opened returned no data
6489 Details : ll_file_read() reads zeros from no-object files to EOF
6494 Description: A bug in MDS/OSS recovery could cause the OSS to fail an assertion
6495 Details : There's little harm in aborting MDS/OSS recovery and letting it
6496 try again, so I removed the LASSERT and return an error instead.
6498 Severity : enhancement
6500 Description: New debugging infrastructure for tracking down data corruption
6501 Details : The I/O checksum code was replaced to: (a) control it at runtime,
6502 (b) cover more of the client-side code path, and (c) try to narrow
6503 down where problems occurred
6507 Bugzilla : 3819, 4364, 4397, 6313
6508 Description: Racing close and eviction MDS could cause assertion in mds_close
6509 Details : It was possible to get multiple mfd references during close and
6510 client eviction, leading to one thread referencing a freed mfd.
6512 Severity: : enhancement
6513 Bugzilla : 3262, 6359
6514 Description: Attempts to reconnect to servers are now more aggressive.
6515 Details : This builds on the enhanced upcall-less recovery that was added
6516 in 1.4.2. When trying to reconnect to servers, clients will
6517 now try each server in the failover group every 10 seconds. By
6518 default, clients would previously try one server every 25 seconds.
6523 Description: After recovery, certain operations trigger a failed
6524 assertion on a client.
6525 Details : Failing over an mds, using lconf -d --failover, while a
6526 client was doing a readdir() call would cause the client to
6527 LBUG after recovery completed and the readdir() was resent.
6529 Severity : enhancement
6531 Description: Default groups are now added by lconf
6532 Details : You can now run lconf --group <servicename> without having to
6533 manually add groups with lmc.
6536 Frequency : occasional
6538 Description: Nodes with an elan id of 0 trigger a failed assertion
6541 Frequency : always when accessing e.g. tty/console device nodes
6543 Description: tty and some other devices nodes cannot be used on lustre
6544 Details : file's private_data field is used by device data and lustre
6545 values in there got lost. New field was added to struct file to
6546 store fs-specific private data.
6549 Frequency : when exporting Lustre via NFS
6551 Description: NFSD failed occasionally when looking up a path component
6552 Details : NFSD is looking up ".." which was broken in ext3 directories
6553 that had grown large enough to become hashed.
6556 Frequency : Clusters with multiple interfaces not on the same subnet
6558 Description: Nodes will repeatedly try to reconnect to an interface which it
6559 cannot reach and report an error to the log.
6560 Details : Extra peer list entries will be created by lconf with some peers
6561 unreachable. lconf now validates the peer before adding it.
6564 Frequency : Only if a default stripe is set on the filesystem root.
6566 Description: Setting a default stripe on the filesystem root prevented the
6567 filesystem from being remounted.
6568 Details : The client was sending extra request flags in the root getattr
6569 request and did not allocate a reply buffer for the dir EA.
6572 Frequency : occasional, higher if lots of files are accessed by one client
6573 Bugzilla : 6159, 6097
6574 Description: Client trips assertion regarding lsm mismatch/magic
6575 Details : While revalidating inodes the VFS looks up inodes with ifind()
6576 and in rare cases can find an inode that is being freed.
6577 The ll_test_inode() code will free the lsm during ifind()
6578 when it finds an existing inode and then the VFS later attaches
6579 this free lsm to a new inode.
6583 Bugzilla : 6422, 7030
6584 Description: MDS deadlock between mkdir and client eviction
6585 Details : Creating a new file via mkdir or mknod (starting a transaction
6586 and getting the ns lock) can deadlock with client eviction
6587 (gets ns lock and trying to finish a synchronous transaction).
6590 Frequency : occasional
6591 Description: While starting a server, the fsfilt_ext3 module could not be
6593 Details : CFS's improved ext3 filesystem is named ldiskfs for 2.6
6594 kernels. Previously, lconf would still use the ext3 name
6595 when trying to load modules. Now, it will correctly use
6596 ext3 on 2.4 and ldiskfs on 2.6.
6598 Severity : enhancement
6599 Description: The default stripe count has been changed to 1
6600 Details : The interpretation of the default stripe count (0, to lfs
6601 or lmc) has been changed to mean striping across a single
6602 OST, rather than all available. For general usage we have
6603 found a stripe count of 1 or 2 works best.
6605 Severity : enhancement
6606 Description: Add support for compiling against Cray portals.
6607 Details : Conditional compiling for some areas that are different
6611 Frequency : occasional
6612 Bugzilla : 6409, 6834
6613 Description: Creating files with an explicit stripe count may lead to
6614 a failed assertion on the MDS
6615 Details : If some OSTs are full or unavailable, creating files may
6616 trigger a failed assertion on the MDS. Now, Lustre will
6617 try to use other servers or return an error to the
6621 Frequency : occasional
6623 Description: Multiple concurrent overlapping read+write on multiple SMP nodes
6624 caused lock timeout during readahead (since 1.4.2).
6625 Details : Processes doing readahead might match a lock that hasn't been
6626 granted yet if there are overlapping and conflicting lock
6627 requests. The readahead process waits on ungranted lock
6628 (original lock is CBPENDING), while OST waits for that process
6629 to cancel CBPENDING read lock and eventually evicts client.
6631 Severity : enhancement
6633 Description: Initial enabling of flock support for clients
6634 Details : Implements fcntl advisory locking and file status functions.
6635 This feature is provided as an optional mount flag (default
6636 off), and is NOT CURRENTLY SUPPORTED. Not all types of record
6637 locking are implemented yet, and those that are are not guaranteed
6638 to be completely correct in production environments.
6639 mount -t lustre -o [flock|noflock] ...
6642 Frequency : occasional
6644 Description: OSTs running 2.4 kernels but with extents enabled might trip an
6645 assertion in the ext3 JBD (journaling) layer.
6646 Details : The b_committed_data struct is protected by the big kernel lock
6647 in 2.4 kernels, serializing journal_commit_transaction() and
6648 ext3_get_block_handle->ext3_new_block->find_next_usable_block()
6649 access to this struct. In 2.6 kernels there is finer grained
6650 locking to improve SMP performance of the JBD layer.
6654 Description: Changes the "SCSI I/O Stats" kernel patch to default to "enabled"
6656 -----------------------------------------------------------------------------
6658 2005-05-05 Cluster File Systems, Inc. <info@clusterfs.com>
6660 NOTE: Lustre 1.4.2 uses an incompatible network protocol than previous
6661 versions of Lustre. Please update all servers and clients to
6662 version 1.4.2 or later at the same time. You must also run
6663 "lconf --write-conf {config}.xml" on the MDS while it is stopped
6664 to update the configuration logs.
6666 - fix for HPUX NFS client breakage when NFS exporting Lustre (5781)
6667 - mdc_enqueue does not need max_mds_easize request buffer on send (5707)
6668 - swab llog records of type '0' so we get proper header size/idx (5861)
6669 - send llog cancel req to DLM cancel portal instead of cb portal (5515)
6670 - fix rename of one directory over another leaking an inode (5953)
6671 - avoid SetPageDirty on 2.6 (5981)
6672 - don't re-add just-being-destroyed locks to the waiting list (5653)
6673 - when creating new directories, inherit the parent's custom
6674 striping settings if present parent (3048)
6675 - flush buffers from cache before direct IO in 2.6 obdfilter (4982)
6676 - don't hold i_size_sem in ll_nopage() and ll_ap_refresh_count (6077)
6677 - don't hold client locks on temporary worklist from l_lru (5666)
6678 - handle IO errors in 2.6 obdfilter bio completion routine (6046)
6679 - automatically evict dead clients (5921)
6680 - Update file size properly in create+truncate+fstat case (6196)
6681 - Do not unhash mountpoint dentries, do not allow removal of
6683 - Avoid lock ordering deadlock issue with write/truncate (6203,5654)
6684 - reserve enough journal credits in fsfilt_start_log for setattr (4554)
6685 - ldlm_enqueue freed-export error path would always LBUG (6149,6184)
6686 - don't reference lr_lvb_data until after we hold lr_lvb_sem (6170)
6687 - don't overwrite last_rcvd if there is a *_client_add() error (6086)
6688 - Correctly handle reads of files with no objects (6243)
6689 - lctl recover will also mark a device active if deactivate used (5933)
6691 - by default create 1 inode per 4kB space on MDS, per 16kB on OSTs
6692 - allow --write-conf on an MDS with different nettype than client (5619)
6693 - don't write config llogs to MDS for mounts not from that MDS (5617)
6694 - lconf should create multiple TCP connections from a client (5201)
6695 - init scripts are now turned off by default; run chkconfig --on
6696 lustre and chkconfig --on lustrefs to use them
6697 - upcalls are no longer needed for clients to recover to failover
6699 - add --abort-recovery option to lconf to abort recovery on device
6701 - add support for an arbitrary number of OSTs (3026)
6702 - Quota support protocol changes.
6703 - forward compatibility changes to wire structs (6007)
6704 - rmmod NALs that might be loaded because of /etc/modules.conf (6133)
6705 - support for mountfsoptions and clientoptions to the Lustre LDAP (5873)
6706 - improved "lustre status" script
6707 - initialize blocksize for non-regular files (6062)
6708 - added --disable-server and --disable-client configure options (5782)
6709 - introduce a lookup cache for lconf to avoid repeated DB scans (6204)
6710 - Vanilla 2.4.29 support
6711 - increase maximum number of obd devices to 520 (6242)
6712 - remove the tcp-zero-copy patch from the suse-2.4 series (5902)
6713 - Quadrics Elan drivers are now included for the RHEL 3 2.4.21 and
6714 SLES 9 2.6.5 kernels
6715 - limit stripes per file to 160 (the maximum EA size) (6093)
6717 2005-03-22 Cluster File Systems, Inc. <info@clusterfs.com>
6720 - don't LASSERT in ll_release on NULL lld with NFS export (4655, 5760)
6721 - hold NS lock when calling handle_ast_error->del_waiting_lock (5746)
6722 - fix setattr mtime regression from lovcleanup merge (4829, 5669)
6723 - workaround for 2.6 crash in ll_unhash_aliases (5687, 5210)
6724 - small ext3 extents cleanups and fixes (5733)
6725 - improved mballoc code, several small races and bugs fixed (5733, 5638)
6726 - kernel version 43 - fix remove_suid bugs in both 2.4 and 2.6 (5695)
6727 - avoid needless client->OST connect, fix handle mismatch (5317)
6728 - fix DLM error path that led to out-of-sync client, long delays (5779)
6729 - support common vfs-enforced mount options (nodev,nosuid,noexec) (5637)
6730 - fix several locking issues related to i_size (5492,5624,5654,5672)
6731 - don't move pending lock onto export if it is already evicted (5683)
6732 - fix kernel oops when creating .foo in unlinked directory (5548)
6733 - fix deadlock in obdfilter statistics vs. object create (5811)
6734 - use time_{before,after} to avoid timer jiffies wrap (5882)
6735 - shutdown --force/--failover stability (3607,3651,4797,5203,4834)
6736 - Do not leak request if server was not able to process it (5154)
6737 - If mds_open unable to find parent dir, make that negative lookup(5154)
6738 - don't create new directories with extent-mapping (5909, 5936)
6740 - fix lustre/lustrefs init scripts for SuSE (patch from Scali, 5702)
6741 - don't hold the pinger_sem in ptlrpc_pinger_sending_on_import
6742 - change obd_increase_kms to obd_adjust_kms (up or down) (5654)
6743 - lconf, lmc search both /usr/lib and /usr/lib64 for Python libs (5800)
6744 - support for RHEL4 kernel on i686 (5773)
6745 - provide error messages when incompatible logs are encountered (5898)
6747 2005-02-18 Cluster File Systems, Inc. <info@clusterfs.com>
6748 * version 1.4.0.10 (1.4.1 release candidate 1)
6750 - don't keep a lock reference when lock is not granted (4238)
6751 - unsafe list practices (rarely) led to infinite eviction loop (4908)
6752 - add per-fs limit of Lustre pages in page cache, avoid OOM (4699)
6753 - drop import inflight refcount on signal_completed_replay error (5255)
6754 - unlock page after async write error during send (3677)
6755 - handle missing objects in filter_preprw_read properly (5265)
6756 - no transno return for symlink open, don't save no-trasno open (3440)
6757 - don't try to complete elan receive that already failed (4012)
6758 - free RPC server reply state on error (5406)
6759 - clean up thread from ptlrpc_start_thread() on error (5160)
6760 - readahead could read extra page into cache that wasn't ejected (5388)
6761 - prevent races in class_attach/setup/cleanup/detach (5260)
6762 - don't dereference de->d_inode after l_dput of de (5458)
6763 - use "int" for stripe value returned from lock_to_stripe (5544)
6764 - mballoc allocation and error-checking fixes in 2.6 (5504)
6765 - block device patches to fix I/O request sizes in 2.6 (5482)
6766 - look up hostnames for IB nals (5602)
6767 - 2.6 changed lock ordering of 2 semaphores, caused deadlock (5654)
6768 - don't start multiple acceptors for the same port (5277)
6769 - fix incorrect LASSERT in mds_getattr_name (5635)
6770 - export a proc file for general "ping" checking (5628)
6771 - fix "lfs check" to not block when the MDS is down (5628)
6773 - service request history (4965)
6774 - put {ll,lov,osc}_async_page structs in a single slab (4699)
6775 - create an "evict_client" /proc entry on OSTs, like the MDS has
6776 - fix mount usage message, return errors per mount(8) (5168)
6777 - change grep [] to grep "[]" in tests so they work in more UMLs
6778 - fix ppc64/x86_64 spec to use %{_libdir} instead of /usr/lib (5389)
6779 - remove ancient LOV_MAGIC_V0 EA support (5047)
6780 - add "disk I/Os in flight" and "I/O req time" stats in obdfilter
6781 - align r/w RPCs to PTLRPC_MAX_BRW_SIZE boundary for performance (3451)
6782 - allow readahead allocations to fail when low on memory (5383)
6783 - mmap locking landed again, after considerable improvement (2828)
6784 - add get_hostaddr() to lustreDB.py for LDAP support (5459)
6786 2004-11-23 Cluster File Systems, Inc. <info@clusterfs.com>
6789 - send OST transaction number in read/write reply to free req (4966)
6790 - don't ASSERT in ptl_send_rpc() if we run out of memory (5119)
6791 - lock /proc/sys/portals/routes internal state, avoiding oops (4827)
6792 - the watchdog thread now runs as interruptible (5246)
6793 - flock/lockf fixes (but it's still disabled, pending 5135)
6794 - don't use EXT3 constants in llite code (5094)
6795 - memory shortage at startup could cause assertion (5176)
6797 - reorganization of lov code
6798 - single portals codebase
6800 - add extents/mballoc support (5025)
6801 - direct I/O reads in the obdfilter (4048)
6802 - kernel patches from LNXI for 2.6 (bluesmoke, perfctr, mtd, kexec)
6804 tbd Cluster File Systems, Inc. <info@clusterfs.com>
6807 - send OST transaction number in read/write reply to free req (4966)
6808 - don't ASSERT in ptl_send_rpc() if we run out of memory (5119)
6809 - lock /proc/sys/portals/routes internal state, avoiding oops (4827)
6810 - the watchdog thread now runs as interruptible (5246)
6811 - handle missing objects in filter_preprw_read properly (5265)
6812 - unsafe list practices (rarely) led to infinite eviction loop (4908)
6813 - drop import inflight refcount on signal_completed_replay error (5255)
6814 - unlock page after async write error during send (3677)
6815 - return original error code on reconstructed replies (3761)
6816 - no transno return for symlink open, don't save no-trasno open (3440)
6818 - add pid to ldlm debugging output (4922)
6819 - bump the watchdog timeouts -- we can't handle 30sec yet
6820 - extra debugging for orphan dentry/inode bug (5259)
6822 2004-11-16 Cluster File Systems, Inc. <info@clusterfs.com>
6825 - fix TCP_NODELAY bug, which caused extreme perf regression (5134)
6826 - allocate qswnal tx descriptors singly to avoid fragmentation (4504)
6827 - don't LBUG on obdo_alloc() failure, use OBD_SLAB_ALLOC() (4800)
6828 - fix NULL dereference in /proc/sys/portals/routes (4827)
6829 - allow failed mdc_close() operations to be interrupted (4561)
6830 - stop precreate on OST before MDS would time out on it (4778)
6831 - don't send partial-page writes before EOF from client (4410)
6832 - discard client grant for sub-page writes on large-page clients (4520)
6833 - don't free dentries not owned by NFS code, check generation (4806)
6834 - fix lsm leak if mds_create_objects() fails (4801)
6835 - limit debug_daemon file size, always print CERROR messages (4789)
6836 - use transno after validating reply (3892)
6837 - process timed out requests if import state changes (3754)
6838 - update mtime on OST during writes, return in glimpse (4829)
6839 - add mkfsoptions to LDAP (4679)
6840 - use ->max_readahead method instead of zapping global ra (5039)
6841 - don't interrupt __l_wait_event() during strace
6843 - add software watchdogs to catch hung threads quickly (4941)
6844 - make lustrefs init script start after nfs is mounted
6845 - fix CWARN/ERROR duplication (4930)
6846 - return async write errors to application if possible (2248)
6847 - add /proc/sys/portal/memused (bytes allocated by PORTALS_ALLOC)
6848 - print NAL number in %x format (4645)
6849 - update barely-supported suse-2.4.21-171 series (4842)
6850 - support for sles 9 %post scripts
6851 - support for building 2.6 kernel-source packages
6852 - support for sles km_* packages
6854 2004-10-07 Cluster File Systems, Inc. <info@clusterfs.com>
6857 - ignore -ENOENT errors in osc_destroy (3639)
6858 - notify osc create thread that OSC is being cleaned up (4600)
6859 - add nettype argument for llmount in #5d in conf-sanity.sh (3936)
6860 - reconstruct ost_handle() like mds_handle() (4657)
6861 - create a new thread to do import eviction to avoid deadlock (3969)
6862 - let lconf resolve symlinked-to devices (4629)
6863 - don't unlink "objects" from directory with default EA (4554)
6864 - hold socknal file ref over connect in case target is down (4394)
6865 - allow more than 32000 subdirectories in a single directory (3244)
6866 - fix blocks count for O_DIRECT writes (3751)
6867 - OST returns ENOSPC from object create when no space left (4539)
6868 - don't send truncate RPC if file size isn't changing (4410)
6869 - limit OSC precreate to 1/2 of value OST considers bogus (4778)
6870 - bind to privileged port in socknal and tcpnal (3689)
6872 - rate limit CERROR/CWARN console message to avoid overload (4519)
6873 - GETFILEINFO dir ioctl returns LOV EA + MDS stat in 1 call (3327)
6874 - basic mmap support (3918)
6875 - kernel patch series update from b1_4 (4711)
6877 2004-09-16 Cluster File Systems, Inc. <info@clusterfs.com>
6880 - avoid crash during MDS cleanup with OST shut down (2775)
6881 - fix loi_list_lock/oig_lock inversion on interrupted IO (4136)
6882 - don't use bad inodes on the MDS (3744)
6883 - dynamic object preallocation to improve recovery speed (4236)
6884 - don't hold spinlock over lock dumping or change debug flags (4401)
6885 - don't zero obd_dev when it is force cleaned (3651)
6886 - print grants to console if they go negative (4431)
6887 - "lctl deactivate" will stop automatic recovery attempts (3406)
6888 - look for existing locks in ldlm_handle_enqueue() (3764)
6889 - don't resolve lock handle twice in recovery avoiding race (4401)
6890 - revalidate should check working dir is a directory (4134)
6892 - don't always mark "slow" obdfilter messages as errors (4418)
6894 2004-08-24 Cluster File Systems, Inc. <info@clusterfs.com>
6897 - don't close LustreDB during write_conf until it is done (3860)
6898 - fix typo in lconf for_each_profile (3821)
6899 - allow dumping logs from multiple threads at one time (3820)
6900 - don't allow multiple threads in OSC recovery (3812)
6901 - fix debug_size parameters (3864)
6902 - fix mds_postrecov to initialize import for llog ctxt (3121)
6903 - replace config semaphore with spinlock (3306)
6904 - be sure to send a reply for a CANCEL rpc with bad export (3863)
6905 - don't allow enqueue to complete on a destroyed export (3822)
6906 - down write_lock before checking llog header bitmap (3825)
6907 - recover from lock replay timeout (3764)
6908 - up llog sem before sending rpc (3652)
6909 - reduce ns lock hold times when setting kms (3267)
6910 - change a dlm LBUG to LASSERTF, to maybe learn something (4228)
6911 - fix NULL deref and obd_dev leak on setup error (3312)
6912 - replace some LBUG about llog ops with error handling (3841)
6913 - don't match INVALID dentries from d_lookup and spin (3784)
6914 - hold dcache_lock while marking dentries INVALID and hashing (4255)
6915 - fix invalid assertion in ptlrpc_set_wait (3880)
6917 - add libwrap support for the TCP acceptor (3996)
6918 - add /proc/sys/portals/routes for non-root route listing (3994)
6919 - allow setting MDS UUID in .xml (2580)
6920 - print the stack of a process that LBUGs (4228)
6922 2004-07-14 Cluster File Systems, Inc. <info@clusterfs.com>
6925 - don't cleanup request in ll_file_open() on failed MDS open (3430)
6926 - make sure to unset replay flag from failed open requests (3440)
6927 - if default stripe count is 0, use OST count for inode size (3636)
6928 - update parent mtime/ctime on client for create/unlink (2611)
6929 - drop dentry ref in ext3_add_link from open_connect_dentry (3266)
6930 - free recovery state on server during a forced cleanup (3571)
6931 - unregister_reply for resent reqs (3063)
6932 - loop back devices mounting and status check on 2.6 (3563)
6933 - fix resource-creation race that can provoke i_size == 0 (3513)
6934 - don't try to use bad inodes returned from MDS/OST fs lookup (3688)
6935 - more debugging for page-accounting assertion (3746)
6936 - return -ENOENT instead of asserting if ost getattr+unlink race (3558)
6937 - avoid deadlock after precreation failure (3758)
6938 - fix race and lock order deadlock in orphan handling (3450, 3750)
6939 - add validity checks when grabbing inodes from l_ast_data (3599)
6941 - add /proc/.../recovery_status to obdfilter (3428)
6942 - lightweight CDEBUG infrastructure, debug daemon (3668)
6943 - change default OSC RPC parameters to be better on small clusters
6944 - turn off OST read cache for files smaller than 32MB
6945 - install man pages and include them in rpms (3100)
6946 - add new init script for (un)mounting lustre filesystems (2593)
6947 - run chkconfig in %post for init scripts (3701)
6948 - drop scimac NAL (unmaintained)
6950 2004-06-17 Cluster File Systems, Inc. <info@clusterfs.com>
6953 - clean kiobufs before and after use (3485)
6954 - strip trailing '/'s before comparing paths with /proc/mounts (3486)
6955 - remove assertions to work around "in-flight rpcs" recovery bug (3063)
6956 - change init script to fail more clearly if not run as root (1528)
6957 - allow clients to reconnect during replay (1742)
6958 - fix ns_lock/i_sem lock ordering deadlock for kms update (3477)
6959 - don't do DNS lookups on NIDs too small for IP addresses (3442)
6960 - re-awaken ptlrpcd if new requests arrive during check_set (3554)
6961 - fix cond_resched (3554)
6962 - only evict unfinished clients after recovery (3515)
6963 - allow bulk resend, prevent data loss (3570)
6964 - dynamic ptlrpc request buffer allocation (2102)
6965 - don't allow unlinking open directory if it isn't empty (2904)
6966 - set MDS/OST threads to umask 0 to not clobber client modes (3359)
6967 - remove extraneous obd dereference causing LASSERT failure (3334)
6968 - don't use get_cycles() when creating temp. files on the mds (3156)
6969 - hold i_sem when setting i_size in ll_extent_lock() (3564)
6970 - handle EEXIST for set-stripe, set proper directory name (3336)
6972 - servers can dump a log evicting a client - lustre.dump_on_timeout=1
6973 - fix ksocknal_fmb_callback() error messages (2918)
6975 2004-05-27 Cluster File Systems, Inc. <info@clusterfs.com>
6978 - don't copy lvb into (possibly NULL) reply on error (2983)
6979 - don't deref dentry after dput, don't free lvb on error (2922)
6980 - use the kms to determine writeback rpc length (2947)
6981 - increment oti_logcookies when osc is inactive (2948)
6982 - update client's i_blocks count via lvb messages (2543)
6983 - handle intent open/close of special files properly (1557)
6984 - mount MDS with errors=remount-ro, like obdfilter (2009)
6985 - initialize lock handle to avoid ASSERT on error cleanup (3057)
6986 - don't use cancelling-locks' kms values (2947)
6987 - use highest lock extent for kms, not last one (2925)
6988 - don't dereference ERR_PTR() dentry in error handling path (3107)
6989 - fix thread race in portals_debug_dumplog() (3122)
6990 - create lprocfs device entries at setup instead of at attach (1519)
6991 - common AST error handler, don't evict client on completion race (3145)
6992 - zero nameidata in detach_mnt in 2.6 (3118)
6993 - verify d_inode after revalidate_special is valid in 2.6 (3116)
6994 - use lustre_put_super() to handle zconf unmounts in 2.6 (3064)
6995 - initialize RPC timeout timer earlier for 2.6 (3219)
6996 - don't dereference NULL reply buffer if mdc_close was never sent (2410)
6997 - print nal/nid for unknown nid (3258)
6998 - additional checks for oscc recovery before doing precreate (3284)
6999 - fix ll_extent_lock() error return code for 64-bit systems (3043)
7000 - don't crash in mdc_close for bad permissions on open (3285)
7001 - zero i_rdev for non-device files (3147)
7002 - clear page->private before handing to FS, better assertion (3119)
7003 - tune the read pipeline (3236)
7004 - fix incorrect decref of invalidated dentry (2350)
7005 - provide read-ahead stats and refine rpc in flight stats (3328)
7006 - don't hold journal transaction open across create RPC (3313)
7007 - update atime on MDS at close time (3265)
7008 - close LDAP connection when recovering to avoid server load (3315)
7009 - update iopen-2.6 patch with fixes from 2399,2517,2904 (3301)
7010 - don't leak open file on MDS after open resend (3325)
7011 - serialize filter_precreate and filter_destroy_precreated (3329)
7012 - loop device shouldn't call sync_dev() for nul device (3092)
7013 - clear page cache after eviction (2766)
7014 - resynchronize MDS->OST in background (2824)
7015 - refuse to mount the same filesystem twice on same mountpoint (3394)
7016 - allow llmount to create routes for mounting behind routers (3320)
7017 - push lock cancellation to blocking thread for glimpse ASTs (3409)
7018 - don't call osc_set_data_with_check() for TEST_LOCK matches (3159)
7019 - fix rare problem with rename on htree directories (3417)
7021 - allow default OST striping configuration per directory (1414)
7022 - fix compilation for qswnal for 2.6 kernels (3125)
7023 - increase maximum number of MDS request buffers for large systems
7024 - change liblustreapi to be useful for external progs like lfsck (3098)
7025 - increase local configuration timeout for slow disks (3353)
7026 - allow configuring ldlm AST timeout - lustre.ldlm_timeout=<seconds>
7028 2004-03-22 Cluster File Systems, Inc. <info@clusterfs.com>
7031 - fixes for glimpse AST timeouts / incorrectly 0-sized files (2818)
7032 - don't overwrite extent policy data in reply if lock was blocked (2901)
7033 - drop filter export grants atomically with removal from device (2663)
7034 - del obd_self_export from work_list in class_disconnect_exports (2908)
7035 - don't LBUG if MDS recovery times out during orphan cleanup (2530)
7036 - swab reply message in mdc_close, other PPC fixes (2464)
7037 - fix destroying of named logs (2325)
7038 - overwrite old logs when running lconf --write_conf (2264)
7039 - bump LLOG_CHUNKSIZE to 8k to allow for larger clusters (2306)
7040 - fix race in target_handle_connect (2898)
7041 - mds_reint_create() should take same inode create lock (2926)
7042 - correct journal credits calculated for CANCEL_UNLINK_LOG (2931)
7043 - don't close files for self_export to avoid uninitialized obd (2936)
7044 - allow MDS with the same name as client node (2939)
7045 - hold dentry reference for closed log files for unlink (2325)
7046 - reserve space for all logs during transactions (2059)
7047 - don't evict page beyond end of stripe extent (2925)
7048 - don't oops on a deleted current working directory (2399)
7049 - handle hard links to targets without a parent properly (2517)
7050 - don't dereference NULL lock when racing during eviction (2867)
7051 - don't grow lock extents when lots of conflicting locks (2919)
7053 2004-03-04 Cluster File Systems, Inc. <info@clusterfs.com>
7056 - account for cache space usage on clients to avoid data loss (974)
7057 - lfsck support in lustre kernel code (2349)
7058 - reduce journal credits needed for BRW writes (2370)
7059 - orphan handling to avoid losing space on client/server crashes
7060 - ptlrpcd can be blocked, stopping ALL progress (2477)
7061 - use lock value blocks to assist in proper KMS, faster stat (1021)
7062 - takes i_sem instead of DLM locks internally on obdfilter (2720)
7063 - recovery for initial connections (2355)
7064 - fixes for mds_cleanup_orphans (1934)
7065 - abort_recovery crashes MDS in b_eq (mds_unlink_orphan) (2584)
7066 - block all file creations until orphan recovery completes (1901)
7067 - client remove rq_connection from request struct (2423)
7068 - conf-sanity test_5, proper cleanup in umount log not availale (2640)
7069 - recovery timer race (2670)
7070 - mdc_close recovey bug (2532)
7071 - ptlrpc cleanup bug (2710)
7072 - mds timeout on local locks (2588)
7073 - namespace lock held during RPCs (2431)
7074 - handle interrupted sync write properly (2503)
7075 - don't try to handle a message that hasn't been replied to (2699)
7076 - client assert failure during cleanup after abort recovery (2701)
7077 - leak mdc device after failed mount (2712)
7078 - ptlrpc_check_set allows timedout requests to complete (2714)
7079 - wait for inflight reqs when ptlrpcd finishes (2710)
7080 - make sure unregistered services are removed from the srv_list
7081 - reset bulk XID's when resending them (caught by 1138 test)
7082 - unregister_bulk after timeout
7083 - fix lconf error (2694)
7084 - handle write after unfinished setstripe, stripe-only getstripe (2388)
7085 - readahead locks pages, leaves pending causing memory pressure (2673)
7086 - increase OST request buffers to 4096 on large machines (2729)
7087 - fix up permission of existing directories in simple_mkdir (2661)
7088 - init deleted item, add assertions ptlrpc_abort_inflight() (2725)
7089 - don't assign transno to errored transactions (2742)
7090 - don't delete objects on OST if given a bogus objid from MDS (2751)
7091 - handle large client PAGE_SIZE readdir on small PAGE_SIZE MDS (2777)
7092 - if rq_no_resend, then timeout request after recovery (2432)
7093 - fix MDS llog_logid record size, 64-bit array alignment (2733)
7094 - don't call usermode_helper from ptlrpcd, DEFAULT upcall (2773)
7095 - put magic in mount.lustre data, check for bad/NULL mount data (2529)
7096 - MDS recovery shouldn't delete objects that it has given out (2730)
7097 - if enqueue arrives after completion, don't clobber LVB (2819)
7098 - don't unlock pages twice when trigger_group_io returns error (2814)
7099 - don't deref NULL rq_repmsg if ldlm_handle_enqueue failed (2822)
7100 - don't write pages to disk if there was an error (1450)
7101 - don't ping imports that have recovery disabled (2676)
7102 - take buffered bytes into account when balancing socknal conn (2817)
7103 - hold a DLM lock over readdir always, use truncate_inode_pages (2706)
7104 - reconnect unlink llog connection after MDS reconnects to OST (2816)
7105 - remove little-endian swabbing of llog records (1987)
7106 - set/limit i_blksize to LL_MAX_BLKSIZE on client (2884)
7107 - retry reposting request buffers if they fail (1191)
7108 - grow extent at grant time to avoid granting a revoked lock (2809)
7109 - lock revoke doesn't evict page if covered by a second lock (2765)
7110 - disable VM readahead to avoid reading outside lock extents (2805)
7112 - return LL_SUPER_MAGIC from statfs for the filesystem type (1972)
7113 - updated kernel patches for hp-2.4.20 kernel (2681)
7115 2004-02-07 Cluster File Systems, Inc. <info@clusterfs.com>
7118 - fix truncated write corruption (2366)
7119 - fix for failed assertion in iopen_connect_dentry (1792,2517)
7121 - don't flag the ptlrpcd thread with PF_MEMALLOC (2636)
7122 - ensure len(uuid) < 37 in lmc (1171)
7123 - fix ia64 OOPS in llog_test (2255)
7124 - zero end of page at obdfilter for partial page writes (2648)
7125 - don't leave stale dentries around after renames (bug 2428)
7126 - fix timeouts when evicting a client with a single lock held (2642)
7127 - set deadline for the initial HELLO message to drain (2634)
7128 - print out dotted-quad IP addresses in the socknal (2302)
7130 - additional debugging for MDS client eviction problem (2443)
7131 - fix mkfsoptions support for osts (2603, 2604)
7133 2004-01-27 Cluster File Systems, Inc. <info@clusterfs.com>
7136 - add series for the vanilla 2.6.0 kernel
7137 - add series for the vanilla 2.4.24 kernel
7138 - add series for a cray x86/64 UL kernel drop
7139 - fix xattr patches for the vanilla 2.4.19 series
7141 - generate true UUIDs in lmc (1171)
7142 - have portals stack dumping break in UML (2466)
7143 - avoid bad dchild deref; avoid inum lock w/o creation (2362)
7144 - allocate with _NOFS in ldlm to avoid deadlock (1933)
7145 - wake callback waiting threads on client eviction (2460)
7146 - Add --ptldebug and --subsystem to lmc (1719)
7147 - update assertion to allow safe interrupt allocation
7148 - set rq_no_resend for cancel requests (2432)
7149 - recalculate ptlrpcd timeout after resend (2494)
7150 - call vfs_rmdir when removing pending directories (2368)
7151 - fix renaming a file to itself (2429)
7152 - lmc creates a default one-stripe lov (2454)
7153 - expand procfs space to handle large clusters (2326)
7154 - increase UML stack to avoid overflow
7155 - update lconf's list of debug and subsystem masks
7156 - fix lfs find --obd (2510)
7157 - /proc tunable for disabling filter read caching (2591)
7158 - stop rpm packages from altering slapd.conf (2301)
7159 - disable nagle in the socknal under 0conf (2578)
7160 - choose mds inode size based on stripe count (2572)
7161 - fix kernel-source rpm problems (2516)
7163 - add --disable-doc to avoid pdf generation (2421)
7164 - update documentation, tests, type-os, comments
7165 - avoid format warnings on ia64
7166 - remove the TOE NAL
7167 - tiny code cleanups by removing unused fields
7169 2004-01-07 Cluster File Systems, Inc. <info@clusterfs.com>
7172 - fix obvious semaphore misuse in as-yet-unused setattr path (2348)
7173 - remove the most blatant lies from BUILDING file (2371)
7174 - change default debug level to reasonable production setting
7175 - reduce client side cache size to reduce cache flush time
7176 - reduce max RPCs in flight to avoid unnecessary file fragmentation
7177 - make TCP zerocopy and pinger support enabled by default (2476)
7178 - sync writes completed after process exits caused crashes (2319)
7179 - maintain correct mount count on the MDS (2356)
7180 - backout 1557, because 2316 wasn't really fixed
7181 - better file I/O statistics gathering in /proc
7182 - don't take unnecessary, deadlock-inducing bug in readpage (2383)
7183 - another kernel patch to fix zero-copy TCP function export
7184 - don't take duplicate lock when processing re-sent getattr (2420)
7185 - lctl uses obd_self_export instead of creating new conn (2353)
7186 - MDS/OST recovery case which requires object creation asserted (2425)
7187 - move lfs from /usr/sbin to /usr/bin in packages
7188 - fix race between mds_client_add and mds_client_free (2417)
7189 - use kmalloc instead of slabs in portals (2430)
7190 - don't create duplicate records when a failover MDS is present (2442)
7191 - remove unnecessary mount age check (2332)
7192 - don't remove directory inodes from locks prematurely (2451)
7193 - don't break if MDS service name is the same as hostname (2103)
7194 - fix races in client write RPC generation when cache full (2482)
7196 2003-12-13 Cluster File Systems, Inc. <info@clusterfs.com>
7199 - remove now-unused request->rq_obd (278)
7200 - if an allocation fails, print out how much memory we've used (1933)
7201 - use PORTAL_SLAB_ALLOC for structures, to get GFP_MEMALLOC (1933)
7202 - add the "configurable stack size" patch to most series files (1256)
7203 - ability to write large log records, for 100+ OST configs (2306)
7204 - fix NULL deref when filter_prep fails (2314)
7205 - fix operator precedence error in filter_sync
7206 - dynamic allocation of socknal TX descriptors (2315)
7207 - fix a missed case in the GFP_MEMALLOC patch, can cause deadlock (2310)
7208 - fix gcc 2.96 compilation problem in xattr kernel patch (2294)
7209 - ensure that CWARN messages in Portals always get to the syslog
7210 - __init/__exit are not for prototype decls (ldlm_init/exit)
7211 - x86-64 compile warning fixes
7212 - fix gateway LMC keyword conflict (2318)
7213 - fix MDS lock inversions in getattr/reint paths (1844)
7214 - fix a rare lock re-ordering bug, which caused deadlock (2322)
7215 - fix i_sem/journal inversion in fsfilt_ext3_write_record (2306)
7216 - DLM race condition prevented some lock evictions (2328)
7217 - ENOMEM detection and retry on socknal sends (2230)
7218 - use GFP_NOFS throughout Lustre, to combat ENOMEM (2230)
7219 - move osc_rpcd into ptlrpc, for use in MDC and others (2329)
7220 - protect MDS inode fsdata with stronger locking; fixes assertion (2313)
7221 - better error messages when a client is rejected during recovery (1505)
7222 - avoid cancelling locks which were never granted, after failure (2330)
7223 - fix i_sem/journal inversion in mds_client_add (2333)
7224 - fix truncate/getattr lock cycle deadlock (2334)
7225 - use rpcd to send close; allows resend after timeout, avoid leak (1897)
7226 - fix two rare exit paths which could leak an l_lock() ref (2321)
7227 - fencepost error in MDS/OST orphan recovery (2226)
7228 - make log record alignment 8 bytes (1988)
7229 - lstripe now fails when requested offset > ost_count (2237)
7230 - ensure that all kernel series have a complete list.h (1607)
7231 - fix crashes in special-file operations (2316)
7232 - lctl create/brw OID mismatch, caused by obsolete filter loop (2339)
7234 - allow configurable automake binary, for testing new versions
7235 - small update to the lfs documentation
7237 2003-12-03 Cluster File Systems, Inc. <info@clusterfs.com>
7239 * fix negative export reference count in fsfilt_sync (2312)
7241 2003-12-01 Cluster File Systems, Inc. <info@clusterfs.com>
7242 * release candidate 0.9.1
7244 - orphans are moved into the PENDING directory for possible recovery
7245 - replayed opens now open by fid for orphan/rename safety (1042)
7246 - last close of an orphan inode generates a transno (683)
7247 - chdir() and mount() now pin the directory entry (1020)
7248 - avoid CERROR in normal ll_setattr_raw() error case (1500)
7249 - discard very old requests without processing them (1502)
7250 - remove some common, well-understood CERRORs (1505)
7251 - require O_DIRECT I/O to be page-sized to workaround IA64 crash (1609)
7252 - clear "grant" flags in OST replies until OST grant code lands (1644)
7253 - fix read performance by not clobbering i_blksize on client (1598)
7254 - fix __ldlm_handle2lock oops by not dereferencing lock after PUT (1625)
7255 - make LRU size a /proc tunable, clears locks when reduced (707)
7256 - fix some lprocfs rot that prevented ptlbd from loading (1732)
7257 - server locks take references on exports now (1558)
7258 - build fixes for 2.4.20-rh trees (1663)
7259 - return an error from lov_create if all OSCs are inactive (1751)
7260 - fix import levels when a reconnect happens without a timeout (1597)
7261 - exit early from mds_open if we get a lookup error (1749)
7262 - partial page read at EOF wouldn't wait for disk before sending (1642)
7263 - avoid NULL deref in obdfilter when reading page past EOF (1592)
7264 - avoid LASSERT in ll_intent_lock if server failed very early (1090)
7265 - fix LBUG in ll_it_open_error with rc = -2 (1861)
7266 - write/truncate lock inversion (1639)
7267 - Don't auto-load obdclass, portals modules during cleanup (1495)
7268 - fix timestamps from jumping to "now" (1763)
7269 - extra journal assertions (1648)
7270 - add an extra multiunlink test (1771)
7271 - fix read_record/write_record API (1776)
7272 - fix leak of offset_extent, possible incorrect i_size later (1772)
7273 - fix lasserts in mis-matched transnos during open-unlink testing (1541)
7274 - Debugging for the kqswnal_get_idle_tx problems (1820)
7275 - Allow recovery to be attempted multiple times (1536)
7276 - Write out MDS last_rcvd file after it is first created (1600)
7277 - Fix tx_descriptor leak in failed transmit situations (1827)
7278 - ext3 journaling fixes for assertion failure after IO error (1871)
7279 - class_export_put() on freed export after completion AST error (1896)
7280 - Fix revalidate looping in VFS (1322)
7281 - Don't access a freed export during MDS_REINT timeout (1521)
7282 - Add open-unlink recovery support on the MDS (1673,1764)
7283 - Return an error if no MDS data was read from last_rcvd (1946)
7284 - Fix for lookup "." or ".." crash on error (1932,1931,1935)
7285 - Don't setup a disk device that doesn't match exported UUID (317)
7286 - Reduce bulk RPC timeout to avoid cascading client/OST failures (1845)
7287 - avoid committing NULL handle in force close
7288 - local.sh is now a one-stripe LOV configuration
7289 - POSIX utime.4 -EPERM on FIFO not owned by user (56)
7290 - fix ext3 htree duplicate directory entry corruption (1516)
7291 - POSIX creat.13, fstat.1, open.18, stat.3 new file atime/mtime (2020)
7292 - update to new LOV EA format (2097)
7293 - interoperability for different PAGE_SIZE/wordsize (686,1821,1343,2042)
7295 2003-06-15 Phil Schwan <phil@clusterfs.com>
7298 - imports and exports cleanup too early, need refcounts (349, 879, 1045)
7299 - per-import/export recovery handling (958, 931, 959)
7300 - multiple last-rcvd slots, for serving multiple FSes (949)
7301 - connections are again shared between multiple imp/exports (963, 964)
7302 - "umount -f" would hang if any requests needed to be sent (393, 978)
7303 - avoid pinning large req buffer by copying for queued messages (989)
7304 - add "uuid" to "lctl device" command to help upcalls (991)
7305 - "open" RPCs with transnos would confuse recovery counters (1037)
7306 - do proper endian conversion of all wire messages (288, 340, 891)
7307 - remove OST bulk get LBUGs, fix ost_brw_write cleanup (1126)
7308 - call waiting locks callback from LDLM recovery thread (1127, 1151)
7309 - fix ptlrpc_connection leak in target_handle_connect (1174)
7310 - fix import refcounting bug in OST and MDS cleanup (1134)
7311 - if an invalid-at-open-time OSC returned before close(), LBUG (1150)
7312 - fix very unlikely obd_types race condition (501)
7313 - remove osc_open hack for echo_client (1187)
7314 - we leaked exports/dlmimps for forcibly disconnected clients (1143)
7315 - a failure in read_inode2 leads to deadlock (1139)
7316 - cancel ack-locks as soon as transaction is committed (1072)
7317 - fix major leaks and crashes in the bulk I/O path (937, 1057)
7318 - make sure to commitrw after any preprw to avoid deadlock (1162)
7319 - failing to execute a file in a lustre FS would lock inode (1203)
7320 - small DEBUG_REQ fix to avoid dereferencing a NULL (1227)
7321 - don't ASSERT while cleaning up an incompletely-setup obd (1248)
7322 - obd_uuid2tgt would walk off the end of the list (1255)
7323 - on IA64 the osc would give portals incorrect bulk size (1258)
7324 - fix debug daemon ioctl interface; allows daemon on ia64 (1274)
7325 - fix lock inversion caused by new llite matching code (1282)
7326 - limit the number of dirty pages on a client to 10MB (1286)
7327 - timed out locks were not being corrected cancelled (1289)
7328 - fix O_DIRECT above 4GB on IA-32 (1292)
7329 * major user-visible changes
7330 - fail out/fail over policy now controlled by the upcall (993)
7332 - add OBD_PING to check server availability and failure (954)
7333 - lustre messages are now sent in sending host order (288, 340, 891)
7334 - add eadatalen to MDS getattr reply (340)
7335 - OST read replies may contain second buffer, with per-page status (593)
7337 2003-03-11 Phil Schwan <phil@clusterfs.com>
7340 - LDLM_DEBUG macro fix, for gcc 3.2 (850)
7341 - failed open()s could cause deadlock; fixed (867, 869)
7342 - stop cancelling OST locks when files are closed (481)
7343 - overlapping XID spaces caused network corruption (851, 853)
7344 - fix unsafe fsfilt counter arithmetic; change to atomic_t
7345 - setattr_raw added, to do single-RPC, server-side setattrs
7346 - lmc/lconf syntax change for OST UUIDs
7347 - fix crashy race condition between ptlrpc_free_req and osc_close
7348 - don't use request in mdc_enqueue if we hit a timeout (889)
7349 - don't set the inode i_size for regular files from the MDS (896)
7350 - handle out of order completion AST (842)
7351 - don't LBUG if a lock request times out after receiving AST (913)
7352 - avoid d_rehash race in ll_find_alias by rehashing inside dcache_lock
7353 - if a bad lock AST arrives, send an error instead of dropping entirely
7354 - return 0 from revalidate2 if ll_intent_lock returns -EINTR (912)
7355 - fix leak in bulk IO when only partially completed (899, 900, 926)
7356 - fix O_DIRECT for ia64 (55)
7357 - (almost) eliminate Lustre-kernel-thread effects on load average (722)
7358 - C-z after timeout could hang a process forever; fixed (977)
7360 - client-side I/O cache (678, 924, 929, 941, 970)
7362 - READPAGE and SETATTRs which don't take server-side locks get
7365 2003-02-11 Phil Schwan <phil@clusterfs.com>
7368 - Fix ldlm_lock_match on the MDS to avoid matching remote locks (592)
7369 - Fix fsfilt_extN_readpage() to read a full page of directory
7370 entries, or fake the remainder if PAGE_SIZE != blocksize (500)
7371 - Avoid extra mdc_getattr() in ll_intent_lock when possible (534, 604)
7372 - Fix imbalanced LOV object allocation and out-of-bound access (469)
7373 - Most intent operations were removed, in favour of a new RPC mode
7374 that does a single RPC to the server and bypasses most of the VFS
7375 - All LDLM resource ID arrays were removed in favour of ldlm_res_id
7376 - Aggressively cancel local locks on DLM servers
7377 - mds_reint_unlink sends EA to the client if it's the last nlink.
7378 client uses that EA to unlink OST objects.
7379 - mds_reint_{rename,unlink,link} were rewritten to take ordered locks
7380 - recursive symlinks were fixed (439)
7381 - fixed NULL deref in DEBUG_REQ
7382 - filter_update_lastobjid no longer calls sync, which annoyed extN
7383 - fixed multi-client small-writes to a single file problem (445)
7384 - fixed mtime updates during file writes (607)
7385 - fixed vector writes on obdfilter causing problems when ENOSPC (670)
7386 - fixed bug in obd_brw_read/write() (under guise of testing 367)
7387 - fixed Linux OST size reporting problem (444, 656)
7388 - OST now updates object mtime with writes or setattr (607, 619)
7389 - client verifies file size before zeroing page past EOF (445)
7390 - OST now writes last allocated objid to disk with allocation (108)
7391 - LOV on echo now works (409)
7393 - mds_reint_unlink sends a new buffer, with the EA included. this
7394 buffer is only valid if body->valid & OBD_MD_FLEASIZE, which is only
7395 set if a regular file was being unlinked, and it was the last link
7396 - use PtlGet from the target for bulk writes (315)
7397 - OST now updates object mtime with writes or setattr (607, 619)
7398 - LDLM now has a grant-time callback to revalidate locked items, if
7400 - Many MDS operations were reorganized to combat race conditions
7402 - Merge b_intel branch (updated lprocfs code) - now at /proc/fs/lustre
7403 - configure check to avoid gcc version 2.96 20000731-2.96-98) (606)
7405 2003-01-06 Andreas Dilger <adilger@clusterfs.com>
7408 - Fully reactivate OST imports after reconnection (512, others)
7409 - Make sure client sees our -ENOTCONN from mds_handle (513 - partial)
7410 - More graceful error handling for truncating on dead OST (515)
7411 - Don't error out unless we're actually accessing dead stripes (474)
7412 - Fix garbage sizes when stripes are missing (410)
7413 - LRU counters were broken, causing constant lock purge (433, 432)
7414 - garbage on read from stripes with failed OSTs (441)
7415 - mark OSCs as active before reconnecting during recovery (438)
7416 - lov_enqueue and lov_cancel need to handle inactive OSTs (403)
7417 - lfind did not preserve OST order in output (443)
7418 - symlinks cause hung clients, incorrect data (439)
7419 - stop dereferencing request after dropping refcount (457)
7420 - don't LASSERT(spin_is_locked) on non-SMP (455)
7421 - fixes for many rename() bugs
7422 - fstat didn't correctly synchronize attributes (399)
7423 - server must handle lock cancellation during blocking AST prep (487)
7424 - bulk descriptors were free()d too soon (511)
7425 - fix paths in lconf, which would load incorrect modules (451, 507)
7426 - fix confusing lconf 'host not found' error message (386)
7427 - fix lock order deadlock on OST (O/R i_sem before journal ops, 478)
7428 - fix race condition in mdc_blocking_ast() for inode access (526)
7429 - fix lov_unpackmd() unpacking wrong number of stripes (537)
7430 - fix lov_set_osc_active() marking wrong OSC inactive (440)
7431 - fix bad lstripe lov_unpackmd() assertion (fix layering too) (527)
7432 - fix multiple writes of stripe MD to MDS (358, maybe 519)
7433 - fix lstripe in several ways (kernel side) (527)
7434 - fix request leak in ldlm_cli_enqueue (262)
7435 - incorrect OSC was marked inactive after OST failure
7436 - call mds_fs_cleanup before unmounting filesystem (524)
7437 - fix races between taking ns_lock and ldlm_lock_change_resource
7438 - fix races updating LOV export open file list
7439 - fix lov_enqueue error path, avoid decref-ing bad lock handle
7440 - fix recovery NULL deref in ldlm_cli_cancel_unused
7441 - fix some DLM races by using new hash table for lock handles (419)
7442 - permit the client to specify desired inodes, at replay
7443 - duplicate requests when we queue them for replay reintegration
7444 - fix last_rcvd offset calculation
7445 - sync after each recovered transaction, so we always make progress
7446 - never, not always, ERESTART requests without transnos
7447 - store the lov_desc in the MDS, so we don't depend on getlovinfo to
7449 - skip replay if the MDS says that the client is already connected
7450 - don't check for a recovery-enabled export to match lctl's UUID
7451 - don't INC_USE_COUNT for phantom exports
7452 - don't crash when cleaning up phantom exports (567)
7453 - don't double-finish or set replay data for errored mdc_open requests
7454 - abort requests when they time out, so we don't get old replies
7455 - send/receive replies for AST messages again
7456 - if the client says that it doesn't have the lock, cancel it on the
7458 - if we timeout during I/O, don't try to cancel an in-use lock; instead
7459 mark it as destroyed, it will all work out when decref is called
7460 - fix module use counts (22, 581)
7462 - ASTs now expect a reply (server cancels lock on error reply)
7464 2002-12-02 Andreas Dilger <adilger@clusterfs.com>
7467 - fix many simultaneous client startup (392)
7468 - fix dentry->d_it clobbering
7469 - credentials weren't being shipped for readdir/getattr operations
7470 - remove invalid assertions triggered during some concurrent MD
7472 - proper Lustre versions added (336, 389)
7473 - fix memory leak for create error case (398)
7474 - fix LOV locking bug that would get cli/srv out of sync
7475 - fix echo client over LOV (409)
7476 - fix dbench 2, extN refcount problem (170, 258, 356, 418)
7477 - fix double-O_EXCL intent crash (424)
7478 - avoid sending multiple lock CANCELs (352)
7480 - MDS can do multi-client recovery (modulo bugs in new code)
7482 - many updates, edits, cleanups
7484 2002-11-18 Phil Schwan <phil@clusterfs.com>
7487 - fix null d_it dereference (346)
7488 - fix full OST/dbench hang (333)
7489 - fix permission problem with file removal (286)
7490 - fix removal of OSCs from LOV when they fail
7491 - fix NULL deref during bulk timeout (214)
7492 - fix problems related to multiple filesystems on one MDS (241)
7493 - fixed serious subtle metadata locking bugs
7494 - free locks on clients when inodes are removed due to memory
7496 - fix inode pointer in lock data (285)
7497 - partial support for multiple MDS on a single host (241)
7498 - data locks weren't cancelled at clear_inode time (290, 311)
7499 - intent locks could lead to unbounded lock growth (205)
7500 - added a maximum lock count, an LRU list, and a flusher
7501 - fix multiple rename (365)
7502 - properly abstracted the echo client
7503 - OSC locked 1 byte too many; fixed
7504 - rewrote brw callback code:
7505 - fixed recovery bugs related to LOVs (306)
7506 - fixed too-many-pages-in-one-write crash (191)
7507 - fixed (again) crash in sync_io_timeout (214)
7508 - probably fixed callback-related race (385)
7510 - Add capability to MDS protocol
7511 - LDLM cancellations and callbacks on different portals
7513 2002-10-28 Andreas Dilger <adilger@clusterfs.com>
7516 - limit client IOV size to PTL_MD_MAX_IOV (611336, 191)
7517 - defer open object destruction to close time (601981, 138)
7518 - open/close OST file handle in obdo (OBD_MD_FLHANDLE) (601981, 138)
7519 - move LDLM_ENQUEUE/CONVERT back to MDS portal (625069)
7520 - abstract ll_lookup2, fix ll_revalidate2 to use abstraction (256)
7521 - don't call obd_setattr in ll_file_release for destroyed objects
7522 * protocol change to lustre_msg: move |version| and add |flags|
7523 * protocol change to osc_punch: "start" in "o_size", "end" in "o_blocks"
7524 * lock replay: for LDLM_FL_REPLAY trust client to do right thing
7525 * added replay of create, unlink, link and rename operations during
7526 MDS failover; recovery should be much more robust now
7527 * remove failed OSCs from LOVs (only lov_create uses this so far)
7528 * the lustre-HOWTO was brought (more) up to date (582544)
7530 2002-10-23 Phil Schwan <phil@clusterfs.com>
7533 - in-use dentries weren't being reused properly (617851)
7534 - prevent multiple LDLM setup (599178)
7535 - fix LOV size calculations for truncate (617853)
7536 - fix client handling of MDS intent errors (POSIX)
7537 - fix permission bug in lovstripe.c test (624321)
7538 - fix MDS thread deadlock - move LDLM handler to DLM portal (625069)
7539 - truncate past end of file could corrupt data
7540 - proper cleanup after timeouts, crashes, etc (592524, 550815)
7541 - a race in recovery could return ETIMEDOUT to apps (623947)
7542 - building outside the source directory was fixed
7543 * the lustre-HOWTO was brought (more) up to date (582544)
7544 * major progress was made on recovery functionality
7546 2002-10-10 Phil Schwan <phil@clusterfs.com>
7549 - recovery deadlock fix
7550 - rm -rf causes LBUG fix (617817)
7551 - file open by multiple tasks fix (618962)
7552 - directory permissions bugs (602707 and 620007)
7553 - journal_stop fixed with locking (611313)
7554 - O_APPEND failures resolved (618273, perhaps 614459)
7555 - lconf PATH fix (619770)
7556 - IA64 build fix (621450)
7557 - RPC buffer sizes scale with amount of memory
7559 2002-10-01 Phil Schwan <phil@clusterfs.com>
7562 - locks would be cancelled without throwing away data pages,
7563 resulting in inconsistent data (605627)
7564 - inode attributes were not always being refreshed (605627, 612449)
7565 - lconf now continues to cleanup after lctl reports an error
7566 - MDS now enforces user permissions (602707)
7567 - lprocfs cleanup fixed, but not yet enabled (614157)
7568 - fixed infinite server hang, should a client not respond to an AST
7569 - avoid going into recovery if user calls readlink() with a buffer
7570 that's too small (613941)
7571 - AST RPCs no longer require replies (614867) -- this may be changed
7572 - don't crash server if client sends an IOV that's too big (611336)
7573 - fixed lock conversion deadlock (611892)
7574 - fixed the following of symlinks (614622)
7575 * recovery: the server can remove locks from a client that dies, other
7576 clients can make progress
7577 * more extN patch fixes
7578 * compile-time configurable ptlrpc buffer allocations
7580 - collaborative read cache document
7581 - Lustre Lite Performance CDR document-in-progress
7583 2002-09-20 Andreas Dilger <adilger@clusterfs.com>
7586 - fix typo in patch-2.4.18
7588 2002-09-20 Andreas Dilger <adilger@clusterfs.com>
7591 - clear ptlrpc request each time in handle_incoming_request()
7592 - unlink of files now destroys the object on the OST
7594 2002-09-19 Peter Braam <braam@clusterfs.com>
7596 * add hard link support
7597 * change obdfile creation method
7598 * kernel patch changed
7600 2002-09-19 Peter Braam <braam@clusterfs.com>
7603 - stack overflow bug in extN fixed
7605 2002-09-18 Andreas Dilger <adilger@clusterfs.com>
7607 * documentation updates
7608 - add man pages for config tools
7609 - update tests/README to describe testing with new config tools
7610 - finish metadata API descriptions
7611 * bug fixes and cleanups
7612 - statfs workaround for 16TB limit
7613 - LOV stripe allocation improved, can stripe on subset of OSTs
7614 - LOV file size/IO offset was wrong for files > 4GB in size
7615 - object EA data was being dropped, caused files to be unreadable
7616 - memory overflow with non-LOV OST caused memory corruption
7617 - fixed regression tests to work with new config tools, obdfilter
7618 - fixed bug when directory size became larger than 1 block
7619 - fixed bug (for single client case) when PWD was deleted
7620 - invalidate local directory pages when doing intent-based ops
7621 - avoid LDLM oops when lock callback contained bad data
7623 2002-09-09 Andreas Dilger <adilger@clusterfs.com>
7625 * documentation updates
7626 * bug fixes and cleanups
7627 - configuration tools
7630 - 64-bit compile warnings
7631 - 64-bit internal statfs data
7633 * test_brw on persistent OST devices
7635 * lprocfs (disabled)
7637 2002-09-04 Andreas Dilger <adilger@clusterfs.com>
7639 * documentation updates
7640 * bug fixes and cleanups
7641 * configuration tools
7643 2002-08-30 Peter J. Braam <braam@clusterfs.com>
7646 * many small fixes to 0_5_4
7647 * io/network handling
7648 * thinkos in MDS operations
7650 2002-08-24 Peter J. Braam <braam@clusterfs.com>
7653 * crucial basic fixes to 0.5.3
7654 * IOR, Iozone work over Elan
7657 2002-08-07 Phil Schwan <phil@clusterfs.com>
7658 * version 0_5_3, our first alpha
7659 * we use the new Portals iovs
7660 * documentation updates
7661 * bug fixes and cleanups
7662 * small changes in the DLM wire protocol
7664 2002-07-25 Peter J. Braam <braam@clusterfs.com>
7665 * version 0_5_1 with some initial stability,
7666 * locking on MD and file I/O.
7667 * documentation updates
7668 * several bug fixes since 0.5.0
7669 * small changes in wire protocol
7671 2002-07-18 Phil Schwan <phil@clusterfs.com>
7673 * delivered as Lustre Light Alpha
7674 * fixed a crash after handling invalid MDS requests
7675 * fixed directory pages for architectures with non-4k pages sizes
7677 2002-07-11 Andreas Dilger <adilger@clusterfs.com>
7678 * release version v0_4_4
7679 * Moves TCP acceptor to be on port 2432 (unused Coda port) instead
7681 * Fixes a number of interruption problems with OST operations.
7682 * Update documentation for portals header changes
7683 * Move all wire protocol structs/defines to lustre_idl.h
7684 * Fixes symlink length bug.
7685 * Add tcpdump to repository.
7687 2002-07-05 Andreas Dilger <adilger@clusterfs.com>
7688 * release version v0_4_3
7689 * Fixes statfs for inodes on extN.
7690 * Fixes bug in runtests which would delete /etc/hosts.
7691 * Use 64-bit object IDs wherever possible (not into VFS though)
7692 Remove ost_get_info, which is unused by lustre, and out of date.
7694 2002-07-03 Peter Braam <braam@clusterfs.com>
7695 * release version v0_4_2 Fixes a lookup error (type not passed)
7696 * move forward to head of Portals
7697 * move forward to latest Lustre kernel
7699 2002-06-25 Peter Braam <braam@clusterfs.com>
7700 * release version v0_4_1. Hopefully stable on single node use.